[PATCH] avr32 architecture
Haavard Skinnemoen [Tue, 26 Sep 2006 06:32:13 +0000 (23:32 -0700)]
This adds support for the Atmel AVR32 architecture as well as the AT32AP7000
CPU and the AT32STK1000 development board.

AVR32 is a new high-performance 32-bit RISC microprocessor core, designed for
cost-sensitive embedded applications, with particular emphasis on low power
consumption and high code density.  The AVR32 architecture is not binary
compatible with earlier 8-bit AVR architectures.

The AVR32 architecture, including the instruction set, is described by the
AVR32 Architecture Manual, available from

http://www.atmel.com/dyn/resources/prod_documents/doc32000.pdf

The Atmel AT32AP7000 is the first CPU implementing the AVR32 architecture.  It
features a 7-stage pipeline, 16KB instruction and data caches and a full
Memory Management Unit.  It also comes with a large set of integrated
peripherals, many of which are shared with the AT91 ARM-based controllers from
Atmel.

Full data sheet is available from

http://www.atmel.com/dyn/resources/prod_documents/doc32003.pdf

while the CPU core implementation including caches and MMU is documented by
the AVR32 AP Technical Reference, available from

http://www.atmel.com/dyn/resources/prod_documents/doc32001.pdf

Information about the AT32STK1000 development board can be found at

http://www.atmel.com/dyn/products/tools_card.asp?tool_id=3918

including a BSP CD image with an earlier version of this patch, development
tools (binaries and source/patches) and a root filesystem image suitable for
booting from SD card.

Alternatively, there's a preliminary "getting started" guide available at
http://avr32linux.org/twiki/bin/view/Main/GettingStarted which provides links
to the sources and patches you will need in order to set up a cross-compiling
environment for avr32-linux.

This patch, as well as the other patches included with the BSP and the
toolchain patches, is actively supported by Atmel Corporation.

[dmccr@us.ibm.com: Fix more pxx_page macro locations]
[bunk@stusta.de: fix `make defconfig']
Signed-off-by: Haavard Skinnemoen <hskinnemoen@atmel.com>
Signed-off-by: Adrian Bunk <bunk@stusta.de>
Signed-off-by: Dave McCracken <dmccr@us.ibm.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

176 files changed:
MAINTAINERS
arch/avr32/Kconfig [new file with mode: 0644]
arch/avr32/Kconfig.debug [new file with mode: 0644]
arch/avr32/Makefile [new file with mode: 0644]
arch/avr32/boards/atstk1000/Makefile [new file with mode: 0644]
arch/avr32/boards/atstk1000/atstk1002.c [new file with mode: 0644]
arch/avr32/boards/atstk1000/setup.c [new file with mode: 0644]
arch/avr32/boards/atstk1000/spi.c [new file with mode: 0644]
arch/avr32/boot/images/Makefile [new file with mode: 0644]
arch/avr32/boot/u-boot/Makefile [new file with mode: 0644]
arch/avr32/boot/u-boot/empty.S [new file with mode: 0644]
arch/avr32/boot/u-boot/head.S [new file with mode: 0644]
arch/avr32/configs/atstk1002_defconfig [new file with mode: 0644]
arch/avr32/kernel/Makefile [new file with mode: 0644]
arch/avr32/kernel/asm-offsets.c [new file with mode: 0644]
arch/avr32/kernel/avr32_ksyms.c [new file with mode: 0644]
arch/avr32/kernel/cpu.c [new file with mode: 0644]
arch/avr32/kernel/entry-avr32b.S [new file with mode: 0644]
arch/avr32/kernel/head.S [new file with mode: 0644]
arch/avr32/kernel/init_task.c [new file with mode: 0644]
arch/avr32/kernel/irq.c [new file with mode: 0644]
arch/avr32/kernel/kprobes.c [new file with mode: 0644]
arch/avr32/kernel/module.c [new file with mode: 0644]
arch/avr32/kernel/process.c [new file with mode: 0644]
arch/avr32/kernel/ptrace.c [new file with mode: 0644]
arch/avr32/kernel/semaphore.c [new file with mode: 0644]
arch/avr32/kernel/setup.c [new file with mode: 0644]
arch/avr32/kernel/signal.c [new file with mode: 0644]
arch/avr32/kernel/switch_to.S [new file with mode: 0644]
arch/avr32/kernel/sys_avr32.c [new file with mode: 0644]
arch/avr32/kernel/syscall-stubs.S [new file with mode: 0644]
arch/avr32/kernel/syscall_table.S [new file with mode: 0644]
arch/avr32/kernel/time.c [new file with mode: 0644]
arch/avr32/kernel/traps.c [new file with mode: 0644]
arch/avr32/kernel/vmlinux.lds.c [new file with mode: 0644]
arch/avr32/lib/Makefile [new file with mode: 0644]
arch/avr32/lib/__avr32_asr64.S [new file with mode: 0644]
arch/avr32/lib/__avr32_lsl64.S [new file with mode: 0644]
arch/avr32/lib/__avr32_lsr64.S [new file with mode: 0644]
arch/avr32/lib/clear_user.S [new file with mode: 0644]
arch/avr32/lib/copy_user.S [new file with mode: 0644]
arch/avr32/lib/csum_partial.S [new file with mode: 0644]
arch/avr32/lib/csum_partial_copy_generic.S [new file with mode: 0644]
arch/avr32/lib/delay.c [new file with mode: 0644]
arch/avr32/lib/findbit.S [new file with mode: 0644]
arch/avr32/lib/io-readsl.S [new file with mode: 0644]
arch/avr32/lib/io-readsw.S [new file with mode: 0644]
arch/avr32/lib/io-writesl.S [new file with mode: 0644]
arch/avr32/lib/io-writesw.S [new file with mode: 0644]
arch/avr32/lib/libgcc.h [new file with mode: 0644]
arch/avr32/lib/longlong.h [new file with mode: 0644]
arch/avr32/lib/memcpy.S [new file with mode: 0644]
arch/avr32/lib/memset.S [new file with mode: 0644]
arch/avr32/lib/strncpy_from_user.S [new file with mode: 0644]
arch/avr32/lib/strnlen_user.S [new file with mode: 0644]
arch/avr32/mach-at32ap/Makefile [new file with mode: 0644]
arch/avr32/mach-at32ap/at32ap.c [new file with mode: 0644]
arch/avr32/mach-at32ap/at32ap7000.c [new file with mode: 0644]
arch/avr32/mach-at32ap/clock.c [new file with mode: 0644]
arch/avr32/mach-at32ap/clock.h [new file with mode: 0644]
arch/avr32/mach-at32ap/extint.c [new file with mode: 0644]
arch/avr32/mach-at32ap/intc.c [new file with mode: 0644]
arch/avr32/mach-at32ap/intc.h [new file with mode: 0644]
arch/avr32/mach-at32ap/pio.c [new file with mode: 0644]
arch/avr32/mach-at32ap/pio.h [new file with mode: 0644]
arch/avr32/mach-at32ap/sm.c [new file with mode: 0644]
arch/avr32/mach-at32ap/sm.h [new file with mode: 0644]
arch/avr32/mm/Makefile [new file with mode: 0644]
arch/avr32/mm/cache.c [new file with mode: 0644]
arch/avr32/mm/clear_page.S [new file with mode: 0644]
arch/avr32/mm/copy_page.S [new file with mode: 0644]
arch/avr32/mm/dma-coherent.c [new file with mode: 0644]
arch/avr32/mm/fault.c [new file with mode: 0644]
arch/avr32/mm/init.c [new file with mode: 0644]
arch/avr32/mm/ioremap.c [new file with mode: 0644]
arch/avr32/mm/tlb.c [new file with mode: 0644]
include/asm-avr32/Kbuild [new file with mode: 0644]
include/asm-avr32/a.out.h [new file with mode: 0644]
include/asm-avr32/addrspace.h [new file with mode: 0644]
include/asm-avr32/arch-at32ap/at91rm9200_pdc.h [new file with mode: 0644]
include/asm-avr32/arch-at32ap/at91rm9200_usart.h [new file with mode: 0644]
include/asm-avr32/arch-at32ap/board.h [new file with mode: 0644]
include/asm-avr32/arch-at32ap/init.h [new file with mode: 0644]
include/asm-avr32/arch-at32ap/portmux.h [new file with mode: 0644]
include/asm-avr32/arch-at32ap/sm.h [new file with mode: 0644]
include/asm-avr32/asm.h [new file with mode: 0644]
include/asm-avr32/atomic.h [new file with mode: 0644]
include/asm-avr32/auxvec.h [new file with mode: 0644]
include/asm-avr32/bitops.h [new file with mode: 0644]
include/asm-avr32/bug.h [new file with mode: 0644]
include/asm-avr32/bugs.h [new file with mode: 0644]
include/asm-avr32/byteorder.h [new file with mode: 0644]
include/asm-avr32/cache.h [new file with mode: 0644]
include/asm-avr32/cachectl.h [new file with mode: 0644]
include/asm-avr32/cacheflush.h [new file with mode: 0644]
include/asm-avr32/checksum.h [new file with mode: 0644]
include/asm-avr32/cputime.h [new file with mode: 0644]
include/asm-avr32/current.h [new file with mode: 0644]
include/asm-avr32/delay.h [new file with mode: 0644]
include/asm-avr32/div64.h [new file with mode: 0644]
include/asm-avr32/dma-mapping.h [new file with mode: 0644]
include/asm-avr32/dma.h [new file with mode: 0644]
include/asm-avr32/elf.h [new file with mode: 0644]
include/asm-avr32/emergency-restart.h [new file with mode: 0644]
include/asm-avr32/errno.h [new file with mode: 0644]
include/asm-avr32/fcntl.h [new file with mode: 0644]
include/asm-avr32/futex.h [new file with mode: 0644]
include/asm-avr32/hardirq.h [new file with mode: 0644]
include/asm-avr32/hw_irq.h [new file with mode: 0644]
include/asm-avr32/intc.h [new file with mode: 0644]
include/asm-avr32/io.h [new file with mode: 0644]
include/asm-avr32/ioctl.h [new file with mode: 0644]
include/asm-avr32/ioctls.h [new file with mode: 0644]
include/asm-avr32/ipcbuf.h [new file with mode: 0644]
include/asm-avr32/irq.h [new file with mode: 0644]
include/asm-avr32/irqflags.h [new file with mode: 0644]
include/asm-avr32/kdebug.h [new file with mode: 0644]
include/asm-avr32/kmap_types.h [new file with mode: 0644]
include/asm-avr32/kprobes.h [new file with mode: 0644]
include/asm-avr32/linkage.h [new file with mode: 0644]
include/asm-avr32/local.h [new file with mode: 0644]
include/asm-avr32/mach/serial_at91.h [new file with mode: 0644]
include/asm-avr32/mman.h [new file with mode: 0644]
include/asm-avr32/mmu.h [new file with mode: 0644]
include/asm-avr32/mmu_context.h [new file with mode: 0644]
include/asm-avr32/module.h [new file with mode: 0644]
include/asm-avr32/msgbuf.h [new file with mode: 0644]
include/asm-avr32/mutex.h [new file with mode: 0644]
include/asm-avr32/namei.h [new file with mode: 0644]
include/asm-avr32/numnodes.h [new file with mode: 0644]
include/asm-avr32/ocd.h [new file with mode: 0644]
include/asm-avr32/page.h [new file with mode: 0644]
include/asm-avr32/param.h [new file with mode: 0644]
include/asm-avr32/pci.h [new file with mode: 0644]
include/asm-avr32/percpu.h [new file with mode: 0644]
include/asm-avr32/pgalloc.h [new file with mode: 0644]
include/asm-avr32/pgtable-2level.h [new file with mode: 0644]
include/asm-avr32/pgtable.h [new file with mode: 0644]
include/asm-avr32/poll.h [new file with mode: 0644]
include/asm-avr32/posix_types.h [new file with mode: 0644]
include/asm-avr32/processor.h [new file with mode: 0644]
include/asm-avr32/ptrace.h [new file with mode: 0644]
include/asm-avr32/resource.h [new file with mode: 0644]
include/asm-avr32/scatterlist.h [new file with mode: 0644]
include/asm-avr32/sections.h [new file with mode: 0644]
include/asm-avr32/semaphore.h [new file with mode: 0644]
include/asm-avr32/sembuf.h [new file with mode: 0644]
include/asm-avr32/setup.h [new file with mode: 0644]
include/asm-avr32/shmbuf.h [new file with mode: 0644]
include/asm-avr32/shmparam.h [new file with mode: 0644]
include/asm-avr32/sigcontext.h [new file with mode: 0644]
include/asm-avr32/siginfo.h [new file with mode: 0644]
include/asm-avr32/signal.h [new file with mode: 0644]
include/asm-avr32/socket.h [new file with mode: 0644]
include/asm-avr32/sockios.h [new file with mode: 0644]
include/asm-avr32/stat.h [new file with mode: 0644]
include/asm-avr32/statfs.h [new file with mode: 0644]
include/asm-avr32/string.h [new file with mode: 0644]
include/asm-avr32/sysreg.h [new file with mode: 0644]
include/asm-avr32/system.h [new file with mode: 0644]
include/asm-avr32/termbits.h [new file with mode: 0644]
include/asm-avr32/termios.h [new file with mode: 0644]
include/asm-avr32/thread_info.h [new file with mode: 0644]
include/asm-avr32/timex.h [new file with mode: 0644]
include/asm-avr32/tlb.h [new file with mode: 0644]
include/asm-avr32/tlbflush.h [new file with mode: 0644]
include/asm-avr32/topology.h [new file with mode: 0644]
include/asm-avr32/traps.h [new file with mode: 0644]
include/asm-avr32/types.h [new file with mode: 0644]
include/asm-avr32/uaccess.h [new file with mode: 0644]
include/asm-avr32/ucontext.h [new file with mode: 0644]
include/asm-avr32/unaligned.h [new file with mode: 0644]
include/asm-avr32/unistd.h [new file with mode: 0644]
include/asm-avr32/user.h [new file with mode: 0644]
include/linux/elf-em.h
lib/Kconfig.debug

index 23348c0..767a434 100644 (file)
@@ -443,6 +443,23 @@ W: http://people.redhat.com/sgrubb/audit/
 T:     git kernel.org:/pub/scm/linux/kernel/git/dwmw2/audit-2.6.git
 S:     Maintained
 
+AVR32 ARCHITECTURE
+P:     Atmel AVR32 Support Team
+M:     avr32@atmel.com
+P:     Haavard Skinnemoen
+M:     hskinnemoen@atmel.com
+W:     http://www.atmel.com/products/AVR32/
+W:     http://avr32linux.org/
+W:     http://avrfreaks.net/
+S:     Supported
+
+AVR32/AT32AP MACHINE SUPPORT
+P:     Atmel AVR32 Support Team
+M:     avr32@atmel.com
+P:     Haavard Skinnemoen
+M:     hskinnemoen@atmel.com
+S:     Supported
+
 AX.25 NETWORK LAYER
 P:     Ralf Baechle
 M:     ralf@linux-mips.org
diff --git a/arch/avr32/Kconfig b/arch/avr32/Kconfig
new file mode 100644 (file)
index 0000000..5f1694e
--- /dev/null
@@ -0,0 +1,196 @@
+#
+# For a description of the syntax of this configuration file,
+# see Documentation/kbuild/kconfig-language.txt.
+#
+
+mainmenu "Linux Kernel Configuration"
+
+config AVR32
+       bool
+       default y
+       # With EMBEDDED=n, we get lots of stuff automatically selected
+       # that we usually don't need on AVR32.
+       select EMBEDDED
+       help
+         AVR32 is a high-performance 32-bit RISC microprocessor core,
+         designed for cost-sensitive embedded applications, with particular
+         emphasis on low power consumption and high code density.
+
+         There is an AVR32 Linux project with a web page at
+         http://avr32linux.org/.
+
+config UID16
+       bool
+
+config GENERIC_HARDIRQS
+       bool
+       default y
+
+config HARDIRQS_SW_RESEND
+       bool
+       default y
+
+config GENERIC_IRQ_PROBE
+       bool
+       default y
+
+config RWSEM_GENERIC_SPINLOCK
+       bool
+       default y
+
+config GENERIC_TIME
+       bool
+       default y
+
+config RWSEM_XCHGADD_ALGORITHM
+       bool
+
+config GENERIC_BUST_SPINLOCK
+       bool
+
+config GENERIC_HWEIGHT
+       bool
+       default y
+
+config GENERIC_CALIBRATE_DELAY
+       bool
+       default y
+
+source "init/Kconfig"
+
+menu "System Type and features"
+
+config SUBARCH_AVR32B
+       bool
+config MMU
+       bool
+config PERFORMANCE_COUNTERS
+       bool
+
+config PLATFORM_AT32AP
+       bool
+       select SUBARCH_AVR32B
+       select MMU
+       select PERFORMANCE_COUNTERS
+
+choice
+       prompt "AVR32 CPU type"
+       default CPU_AT32AP7000
+
+config CPU_AT32AP7000
+       bool "AT32AP7000"
+       select PLATFORM_AT32AP
+endchoice
+
+#
+# CPU Daughterboards for ATSTK1000
+config BOARD_ATSTK1002
+       bool
+
+choice
+       prompt "AVR32 board type"
+       default BOARD_ATSTK1000
+
+config BOARD_ATSTK1000
+       bool "ATSTK1000 evaluation board"
+       select BOARD_ATSTK1002 if CPU_AT32AP7000
+endchoice
+
+choice
+       prompt "Boot loader type"
+       default LOADER_U_BOOT
+
+config LOADER_U_BOOT
+       bool "U-Boot (or similar) bootloader"
+endchoice
+
+config LOAD_ADDRESS
+       hex
+       default 0x10000000 if LOADER_U_BOOT=y && CPU_AT32AP7000=y
+
+config ENTRY_ADDRESS
+       hex
+       default 0x90000000 if LOADER_U_BOOT=y && CPU_AT32AP7000=y
+
+config PHYS_OFFSET
+       hex
+       default 0x10000000 if CPU_AT32AP7000=y
+
+source "kernel/Kconfig.preempt"
+
+config HAVE_ARCH_BOOTMEM_NODE
+       bool
+       default n
+
+config ARCH_HAVE_MEMORY_PRESENT
+       bool
+       default n
+
+config NEED_NODE_MEMMAP_SIZE
+       bool
+       default n
+
+config ARCH_FLATMEM_ENABLE
+       bool
+       default y
+
+config ARCH_DISCONTIGMEM_ENABLE
+       bool
+       default n
+
+config ARCH_SPARSEMEM_ENABLE
+       bool
+       default n
+
+source "mm/Kconfig"
+
+config OWNERSHIP_TRACE
+       bool "Ownership trace support"
+       default y
+       help
+         Say Y to generate an Ownership Trace message on every context switch,
+         enabling Nexus-compliant debuggers to keep track of the PID of the
+         currently executing task.
+
+# FPU emulation goes here
+
+source "kernel/Kconfig.hz"
+
+config CMDLINE
+       string "Default kernel command line"
+       default ""
+       help
+         If you don't have a boot loader capable of passing a command line string
+         to the kernel, you may specify one here. As a minimum, you should specify
+         the memory size and the root device (e.g., mem=8M, root=/dev/nfs).
+
+endmenu
+
+menu "Bus options"
+
+config PCI
+       bool
+
+source "drivers/pci/Kconfig"
+
+source "drivers/pcmcia/Kconfig"
+
+endmenu
+
+menu "Executable file formats"
+source "fs/Kconfig.binfmt"
+endmenu
+
+source "net/Kconfig"
+
+source "drivers/Kconfig"
+
+source "fs/Kconfig"
+
+source "arch/avr32/Kconfig.debug"
+
+source "security/Kconfig"
+
+source "crypto/Kconfig"
+
+source "lib/Kconfig"
diff --git a/arch/avr32/Kconfig.debug b/arch/avr32/Kconfig.debug
new file mode 100644 (file)
index 0000000..64ace00
--- /dev/null
@@ -0,0 +1,19 @@
+menu "Kernel hacking"
+
+config TRACE_IRQFLAGS_SUPPORT
+       bool
+       default y
+
+source "lib/Kconfig.debug"
+
+config KPROBES
+       bool "Kprobes"
+       depends on DEBUG_KERNEL
+       help
+         Kprobes allows you to trap at almost any kernel address and
+          execute a callback function.  register_kprobe() establishes
+          a probepoint and specifies the callback.  Kprobes is useful
+          for kernel debugging, non-intrusive instrumentation and testing.
+          If in doubt, say "N".
+
+endmenu
diff --git a/arch/avr32/Makefile b/arch/avr32/Makefile
new file mode 100644 (file)
index 0000000..cefc95a
--- /dev/null
@@ -0,0 +1,84 @@
+#
+# This file is subject to the terms and conditions of the GNU General Public
+# License.  See the file "COPYING" in the main directory of this archive
+# for more details.
+#
+# Copyright (C) 2004-2006 Atmel Corporation.
+
+# Default target when executing plain make
+.PHONY: all
+all: uImage vmlinux.elf linux.lst
+
+KBUILD_DEFCONFIG       := atstk1002_defconfig
+
+CFLAGS         += -pipe -fno-builtin -mno-pic
+AFLAGS         += -mrelax -mno-pic
+CFLAGS_MODULE  += -mno-relax
+LDFLAGS_vmlinux        += --relax
+
+cpuflags-$(CONFIG_CPU_AP7000)  += -mcpu=ap7000
+
+CFLAGS         += $(cpuflags-y)
+AFLAGS         += $(cpuflags-y)
+
+CHECKFLAGS     += -D__avr32__
+
+LIBGCC         := $(shell $(CC) $(CFLAGS) -print-libgcc-file-name)
+
+head-$(CONFIG_LOADER_U_BOOT)           += arch/avr32/boot/u-boot/head.o
+head-y                                 += arch/avr32/kernel/head.o
+core-$(CONFIG_PLATFORM_AT32AP)         += arch/avr32/mach-at32ap/
+core-$(CONFIG_BOARD_ATSTK1000)         += arch/avr32/boards/atstk1000/
+core-$(CONFIG_LOADER_U_BOOT)           += arch/avr32/boot/u-boot/
+core-y                                 += arch/avr32/kernel/
+core-y                                 += arch/avr32/mm/
+libs-y                                 += arch/avr32/lib/ #$(LIBGCC)
+
+archincdir-$(CONFIG_PLATFORM_AT32AP)   := arch-at32ap
+
+include/asm-avr32/.arch: $(wildcard include/config/platform/*.h) include/config/auto.conf
+       @echo '  SYMLINK include/asm-avr32/arch -> include/asm-avr32/$(archincdir-y)'
+ifneq ($(KBUILD_SRC),)
+       $(Q)mkdir -p include/asm-avr32
+       $(Q)ln -fsn $(srctree)/include/asm-avr32/$(archincdir-y) include/asm-avr32/arch
+else
+       $(Q)ln -fsn $(archincdir-y) include/asm-avr32/arch
+endif
+       @touch $@
+
+archprepare: include/asm-avr32/.arch
+
+BOOT_TARGETS := vmlinux.elf vmlinux.bin uImage uImage.srec
+
+.PHONY: $(BOOT_TARGETS) install
+
+boot := arch/$(ARCH)/boot/images
+
+             KBUILD_IMAGE := $(boot)/uImage
+vmlinux.elf: KBUILD_IMAGE := $(boot)/vmlinux.elf
+vmlinux.cso: KBUILD_IMAGE := $(boot)/vmlinux.cso
+uImage.srec: KBUILD_IMAGE := $(boot)/uImage.srec
+uImage:      KBUILD_IMAGE := $(boot)/uImage
+
+quiet_cmd_listing = LST     $@
+      cmd_listing = avr32-linux-objdump $(OBJDUMPFLAGS) -lS $< > $@
+quiet_cmd_disasm  = DIS     $@
+      cmd_disasm  = avr32-linux-objdump $(OBJDUMPFLAGS) -d $< > $@
+
+vmlinux.elf vmlinux.bin uImage.srec uImage vmlinux.cso: vmlinux
+       $(Q)$(MAKE) $(build)=$(boot) $(boot)/$@
+
+install: vmlinux
+       $(Q)$(MAKE) $(build)=$(boot) BOOTIMAGE=$(KBUILD_IMAGE) $@
+
+linux.s: vmlinux
+       $(call if_changed,disasm)
+
+linux.lst: vmlinux
+       $(call if_changed,listing)
+
+define archhelp
+  @echo '* vmlinux.elf         - ELF image with load address 0'
+  @echo '  vmlinux.cso         - PathFinder CSO image'
+  @echo '  uImage              - Create a bootable image for U-Boot'
+endef
diff --git a/arch/avr32/boards/atstk1000/Makefile b/arch/avr32/boards/atstk1000/Makefile
new file mode 100644 (file)
index 0000000..c3e36ec
--- /dev/null
@@ -0,0 +1,2 @@
+obj-y                          += setup.o spi.o
+obj-$(CONFIG_BOARD_ATSTK1002)  += atstk1002.o
diff --git a/arch/avr32/boards/atstk1000/atstk1002.c b/arch/avr32/boards/atstk1000/atstk1002.c
new file mode 100644 (file)
index 0000000..49164e9
--- /dev/null
@@ -0,0 +1,37 @@
+/*
+ * ATSTK1002 daughterboard-specific init code
+ *
+ * Copyright (C) 2005-2006 Atmel Corporation
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ */
+#include <linux/init.h>
+
+#include <asm/arch/board.h>
+
+struct eth_platform_data __initdata eth0_data = {
+       .valid          = 1,
+       .mii_phy_addr   = 0x10,
+       .is_rmii        = 0,
+       .hw_addr        = { 0x6a, 0x87, 0x71, 0x14, 0xcd, 0xcb },
+};
+
+extern struct lcdc_platform_data atstk1000_fb0_data;
+
+static int __init atstk1002_init(void)
+{
+       at32_add_system_devices();
+
+       at32_add_device_usart(1);       /* /dev/ttyS0 */
+       at32_add_device_usart(2);       /* /dev/ttyS1 */
+       at32_add_device_usart(3);       /* /dev/ttyS2 */
+
+       at32_add_device_eth(0, &eth0_data);
+       at32_add_device_spi(0);
+       at32_add_device_lcdc(0, &atstk1000_fb0_data);
+
+       return 0;
+}
+postcore_initcall(atstk1002_init);
diff --git a/arch/avr32/boards/atstk1000/setup.c b/arch/avr32/boards/atstk1000/setup.c
new file mode 100644 (file)
index 0000000..191ab85
--- /dev/null
@@ -0,0 +1,59 @@
+/*
+ * ATSTK1000 board-specific setup code.
+ *
+ * Copyright (C) 2005-2006 Atmel Corporation
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ */
+#include <linux/bootmem.h>
+#include <linux/init.h>
+#include <linux/types.h>
+#include <linux/linkage.h>
+
+#include <asm/setup.h>
+
+#include <asm/arch/board.h>
+
+/* Initialized by bootloader-specific startup code. */
+struct tag *bootloader_tags __initdata;
+
+struct lcdc_platform_data __initdata atstk1000_fb0_data;
+
+asmlinkage void __init board_early_init(void)
+{
+       extern void sdram_init(void);
+
+#ifdef CONFIG_LOADER_STANDALONE
+       sdram_init();
+#endif
+}
+
+void __init board_setup_fbmem(unsigned long fbmem_start,
+                             unsigned long fbmem_size)
+{
+       if (!fbmem_size)
+               return;
+
+       if (!fbmem_start) {
+               void *fbmem;
+
+               fbmem = alloc_bootmem_low_pages(fbmem_size);
+               fbmem_start = __pa(fbmem);
+       } else {
+               pg_data_t *pgdat;
+
+               for_each_online_pgdat(pgdat) {
+                       if (fbmem_start >= pgdat->bdata->node_boot_start
+                           && fbmem_start <= pgdat->bdata->node_low_pfn)
+                               reserve_bootmem_node(pgdat, fbmem_start,
+                                                    fbmem_size);
+               }
+       }
+
+       printk("%luKiB framebuffer memory at address 0x%08lx\n",
+              fbmem_size >> 10, fbmem_start);
+       atstk1000_fb0_data.fbmem_start = fbmem_start;
+       atstk1000_fb0_data.fbmem_size = fbmem_size;
+}
diff --git a/arch/avr32/boards/atstk1000/spi.c b/arch/avr32/boards/atstk1000/spi.c
new file mode 100644 (file)
index 0000000..567726c
--- /dev/null
@@ -0,0 +1,27 @@
+/*
+ * ATSTK1000 SPI devices
+ *
+ * Copyright (C) 2005 Atmel Norway
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ */
+#include <linux/device.h>
+#include <linux/spi/spi.h>
+
+static struct spi_board_info spi_board_info[] __initdata = {
+       {
+               .modalias       = "ltv350qv",
+               .max_speed_hz   = 16000000,
+               .bus_num        = 0,
+               .chip_select    = 1,
+       },
+};
+
+static int board_init_spi(void)
+{
+       spi_register_board_info(spi_board_info, ARRAY_SIZE(spi_board_info));
+       return 0;
+}
+arch_initcall(board_init_spi);
diff --git a/arch/avr32/boot/images/Makefile b/arch/avr32/boot/images/Makefile
new file mode 100644 (file)
index 0000000..ccd74ee
--- /dev/null
@@ -0,0 +1,62 @@
+#
+# Copyright (C) 2004-2006 Atmel Corporation
+#
+# This file is subject to the terms and conditions of the GNU General Public
+# License.  See the file "COPYING" in the main directory of this archive
+# for more details.
+#
+
+MKIMAGE                := $(srctree)/scripts/mkuboot.sh
+
+extra-y                := vmlinux.bin vmlinux.gz
+
+OBJCOPYFLAGS_vmlinux.bin := -O binary
+$(obj)/vmlinux.bin: vmlinux FORCE
+       $(call if_changed,objcopy)
+
+$(obj)/vmlinux.gz: $(obj)/vmlinux.bin FORCE
+       $(call if_changed,gzip)
+
+quiet_cmd_uimage = UIMAGE $@
+      cmd_uimage = $(CONFIG_SHELL) $(MKIMAGE) -A avr32 -O linux -T kernel      \
+               -C gzip -a $(CONFIG_LOAD_ADDRESS) -e $(CONFIG_ENTRY_ADDRESS)    \
+               -n 'Linux-$(KERNELRELEASE)' -d $< $@
+
+targets += uImage uImage.srec
+$(obj)/uImage: $(obj)/vmlinux.gz
+       $(call if_changed,uimage)
+       @echo '  Image $@ is ready'
+
+OBJCOPYFLAGS_uImage.srec := -I binary -O srec
+$(obj)/uImage.srec: $(obj)/uImage
+       $(call if_changed,objcopy)
+
+OBJCOPYFLAGS_vmlinux.elf := --change-section-lma .text-0x80000000 \
+                           --change-section-lma __ex_table-0x80000000 \
+                           --change-section-lma .rodata-0x80000000 \
+                           --change-section-lma .data-0x80000000 \
+                           --change-section-lma .init-0x80000000 \
+                           --change-section-lma .bss-0x80000000 \
+                           --change-section-lma .initrd-0x80000000 \
+                           --change-section-lma __param-0x80000000 \
+                           --change-section-lma __ksymtab-0x80000000 \
+                           --change-section-lma __ksymtab_gpl-0x80000000 \
+                           --change-section-lma __kcrctab-0x80000000 \
+                           --change-section-lma __kcrctab_gpl-0x80000000 \
+                           --change-section-lma __ksymtab_strings-0x80000000 \
+                           --change-section-lma .got-0x80000000 \
+                           --set-start 0xa0000000
+$(obj)/vmlinux.elf: vmlinux FORCE
+       $(call if_changed,objcopy)
+
+quiet_cmd_sfdwarf = SFDWARF $@
+      cmd_sfdwarf = sfdwarf $< TO $@ GNUAVR IW $(SFDWARF_FLAGS) > $(obj)/sfdwarf.log
+
+$(obj)/vmlinux.cso: $(obj)/vmlinux.elf FORCE
+       $(call if_changed,sfdwarf)
+
+install: $(BOOTIMAGE)
+       sh $(srctree)/install-kernel.sh $<
+
+# Generated files to be removed upon make clean
+clean-files    := vmlinux* uImage uImage.srec
diff --git a/arch/avr32/boot/u-boot/Makefile b/arch/avr32/boot/u-boot/Makefile
new file mode 100644 (file)
index 0000000..125ddc9
--- /dev/null
@@ -0,0 +1,3 @@
+extra-y                := head.o
+
+obj-y          := empty.o
diff --git a/arch/avr32/boot/u-boot/empty.S b/arch/avr32/boot/u-boot/empty.S
new file mode 100644 (file)
index 0000000..8ac91a5
--- /dev/null
@@ -0,0 +1 @@
+/* Empty file */
diff --git a/arch/avr32/boot/u-boot/head.S b/arch/avr32/boot/u-boot/head.S
new file mode 100644 (file)
index 0000000..4488fa2
--- /dev/null
@@ -0,0 +1,60 @@
+/*
+ * Startup code for use with the u-boot bootloader.
+ *
+ * Copyright (C) 2004-2006 Atmel Corporation
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ */
+#include <asm/setup.h>
+
+       /*
+        * The kernel is loaded where we want it to be and all caches
+        * have just been flushed. We get two parameters from u-boot:
+        *
+        * r12 contains a magic number (ATAG_MAGIC)
+        * r11 points to a tag table providing information about
+        *     the system.
+        */
+       .section .init.text,"ax"
+       .global _start
+_start:
+       /* Check if the boot loader actually provided a tag table */
+       lddpc   r0, magic_number
+       cp.w    r12, r0
+       brne    no_tag_table
+
+       /* Initialize .bss */
+       lddpc   r2, bss_start_addr
+       lddpc   r3, end_addr
+       mov     r0, 0
+       mov     r1, 0
+1:      st.d    r2++, r0
+       cp      r2, r3
+       brlo    1b
+
+       /*
+        * Save the tag table address for later use. This must be done
+        * _after_ .bss has been initialized...
+        */
+       lddpc   r0, tag_table_addr
+       st.w    r0[0], r11
+
+       /* Jump to loader-independent setup code */
+       rjmp    kernel_entry
+
+       .align  2
+magic_number:
+       .long   ATAG_MAGIC
+tag_table_addr:
+       .long   bootloader_tags
+bss_start_addr:
+       .long   __bss_start
+end_addr:
+       .long   _end
+
+no_tag_table:
+       sub     r12, pc, (. - 2f)
+       bral    panic
+2:     .asciz  "Boot loader didn't provide correct magic number\n"
diff --git a/arch/avr32/configs/atstk1002_defconfig b/arch/avr32/configs/atstk1002_defconfig
new file mode 100644 (file)
index 0000000..1d22255
--- /dev/null
@@ -0,0 +1,754 @@
+#
+# Automatically generated make config: don't edit
+# Linux kernel version: 2.6.18-rc1
+# Tue Jul 11 12:41:36 2006
+#
+CONFIG_AVR32=y
+CONFIG_GENERIC_HARDIRQS=y
+CONFIG_HARDIRQS_SW_RESEND=y
+CONFIG_GENERIC_IRQ_PROBE=y
+CONFIG_RWSEM_GENERIC_SPINLOCK=y
+CONFIG_GENERIC_HWEIGHT=y
+CONFIG_GENERIC_CALIBRATE_DELAY=y
+CONFIG_DEFCONFIG_LIST="/lib/modules/$UNAME_RELEASE/.config"
+
+#
+# Code maturity level options
+#
+CONFIG_EXPERIMENTAL=y
+CONFIG_BROKEN_ON_SMP=y
+CONFIG_INIT_ENV_ARG_LIMIT=32
+
+#
+# General setup
+#
+CONFIG_LOCALVERSION=""
+# CONFIG_LOCALVERSION_AUTO is not set
+CONFIG_SWAP=y
+# CONFIG_SYSVIPC is not set
+# CONFIG_POSIX_MQUEUE is not set
+# CONFIG_BSD_PROCESS_ACCT is not set
+CONFIG_SYSCTL=y
+# CONFIG_AUDIT is not set
+# CONFIG_IKCONFIG is not set
+# CONFIG_RELAY is not set
+CONFIG_INITRAMFS_SOURCE=""
+CONFIG_CC_OPTIMIZE_FOR_SIZE=y
+CONFIG_EMBEDDED=y
+CONFIG_KALLSYMS=y
+# CONFIG_KALLSYMS_ALL is not set
+# CONFIG_KALLSYMS_EXTRA_PASS is not set
+CONFIG_HOTPLUG=y
+CONFIG_PRINTK=y
+CONFIG_BUG=y
+CONFIG_ELF_CORE=y
+# CONFIG_BASE_FULL is not set
+# CONFIG_FUTEX is not set
+# CONFIG_EPOLL is not set
+CONFIG_SHMEM=y
+# CONFIG_SLAB is not set
+# CONFIG_VM_EVENT_COUNTERS is not set
+# CONFIG_TINY_SHMEM is not set
+CONFIG_BASE_SMALL=1
+CONFIG_SLOB=y
+
+#
+# Loadable module support
+#
+CONFIG_MODULES=y
+CONFIG_MODULE_UNLOAD=y
+# CONFIG_MODULE_FORCE_UNLOAD is not set
+# CONFIG_MODVERSIONS is not set
+# CONFIG_MODULE_SRCVERSION_ALL is not set
+# CONFIG_KMOD is not set
+
+#
+# Block layer
+#
+# CONFIG_BLK_DEV_IO_TRACE is not set
+
+#
+# IO Schedulers
+#
+CONFIG_IOSCHED_NOOP=y
+# CONFIG_IOSCHED_AS is not set
+# CONFIG_IOSCHED_DEADLINE is not set
+# CONFIG_IOSCHED_CFQ is not set
+# CONFIG_DEFAULT_AS is not set
+# CONFIG_DEFAULT_DEADLINE is not set
+# CONFIG_DEFAULT_CFQ is not set
+CONFIG_DEFAULT_NOOP=y
+CONFIG_DEFAULT_IOSCHED="noop"
+
+#
+# System Type and features
+#
+CONFIG_SUBARCH_AVR32B=y
+CONFIG_MMU=y
+CONFIG_PERFORMANCE_COUNTERS=y
+CONFIG_PLATFORM_AT32AP=y
+CONFIG_CPU_AT32AP7000=y
+CONFIG_BOARD_ATSTK1002=y
+CONFIG_BOARD_ATSTK1000=y
+CONFIG_LOADER_U_BOOT=y
+CONFIG_LOAD_ADDRESS=0x10000000
+CONFIG_ENTRY_ADDRESS=0x90000000
+CONFIG_PHYS_OFFSET=0x10000000
+CONFIG_PREEMPT_NONE=y
+# CONFIG_PREEMPT_VOLUNTARY is not set
+# CONFIG_PREEMPT is not set
+# CONFIG_HAVE_ARCH_BOOTMEM_NODE is not set
+# CONFIG_ARCH_HAVE_MEMORY_PRESENT is not set
+# CONFIG_NEED_NODE_MEMMAP_SIZE is not set
+CONFIG_ARCH_FLATMEM_ENABLE=y
+# CONFIG_ARCH_DISCONTIGMEM_ENABLE is not set
+# CONFIG_ARCH_SPARSEMEM_ENABLE is not set
+CONFIG_SELECT_MEMORY_MODEL=y
+CONFIG_FLATMEM_MANUAL=y
+# CONFIG_DISCONTIGMEM_MANUAL is not set
+# CONFIG_SPARSEMEM_MANUAL is not set
+CONFIG_FLATMEM=y
+CONFIG_FLAT_NODE_MEM_MAP=y
+# CONFIG_SPARSEMEM_STATIC is not set
+CONFIG_SPLIT_PTLOCK_CPUS=4
+# CONFIG_RESOURCES_64BIT is not set
+# CONFIG_OWNERSHIP_TRACE is not set
+# CONFIG_HZ_100 is not set
+CONFIG_HZ_250=y
+# CONFIG_HZ_1000 is not set
+CONFIG_HZ=250
+CONFIG_CMDLINE=""
+
+#
+# Bus options
+#
+
+#
+# PCCARD (PCMCIA/CardBus) support
+#
+# CONFIG_PCCARD is not set
+
+#
+# Executable file formats
+#
+CONFIG_BINFMT_ELF=y
+# CONFIG_BINFMT_MISC is not set
+
+#
+# Networking
+#
+CONFIG_NET=y
+
+#
+# Networking options
+#
+# CONFIG_NETDEBUG is not set
+CONFIG_PACKET=y
+CONFIG_PACKET_MMAP=y
+CONFIG_UNIX=y
+# CONFIG_NET_KEY is not set
+CONFIG_INET=y
+# CONFIG_IP_MULTICAST is not set
+# CONFIG_IP_ADVANCED_ROUTER is not set
+CONFIG_IP_FIB_HASH=y
+CONFIG_IP_PNP=y
+CONFIG_IP_PNP_DHCP=y
+# CONFIG_IP_PNP_BOOTP is not set
+# CONFIG_IP_PNP_RARP is not set
+# CONFIG_NET_IPIP is not set
+# CONFIG_NET_IPGRE is not set
+# CONFIG_ARPD is not set
+# CONFIG_SYN_COOKIES is not set
+# CONFIG_INET_AH is not set
+# CONFIG_INET_ESP is not set
+# CONFIG_INET_IPCOMP is not set
+# CONFIG_INET_XFRM_TUNNEL is not set
+# CONFIG_INET_TUNNEL is not set
+# CONFIG_INET_XFRM_MODE_TRANSPORT is not set
+# CONFIG_INET_XFRM_MODE_TUNNEL is not set
+CONFIG_INET_DIAG=y
+CONFIG_INET_TCP_DIAG=y
+# CONFIG_TCP_CONG_ADVANCED is not set
+CONFIG_TCP_CONG_BIC=y
+# CONFIG_IPV6 is not set
+# CONFIG_INET6_XFRM_TUNNEL is not set
+# CONFIG_INET6_TUNNEL is not set
+# CONFIG_NETWORK_SECMARK is not set
+# CONFIG_NETFILTER is not set
+
+#
+# DCCP Configuration (EXPERIMENTAL)
+#
+# CONFIG_IP_DCCP is not set
+
+#
+# SCTP Configuration (EXPERIMENTAL)
+#
+# CONFIG_IP_SCTP is not set
+
+#
+# TIPC Configuration (EXPERIMENTAL)
+#
+# CONFIG_TIPC is not set
+# CONFIG_ATM is not set
+# CONFIG_BRIDGE is not set
+# CONFIG_VLAN_8021Q is not set
+# CONFIG_DECNET is not set
+# CONFIG_LLC2 is not set
+# CONFIG_IPX is not set
+# CONFIG_ATALK is not set
+# CONFIG_X25 is not set
+# CONFIG_LAPB is not set
+# CONFIG_NET_DIVERT is not set
+# CONFIG_ECONET is not set
+# CONFIG_WAN_ROUTER is not set
+
+#
+# QoS and/or fair queueing
+#
+# CONFIG_NET_SCHED is not set
+
+#
+# Network testing
+#
+# CONFIG_NET_PKTGEN is not set
+# CONFIG_NET_TCPPROBE is not set
+# CONFIG_HAMRADIO is not set
+# CONFIG_IRDA is not set
+# CONFIG_BT is not set
+# CONFIG_IEEE80211 is not set
+
+#
+# Device Drivers
+#
+
+#
+# Generic Driver Options
+#
+CONFIG_STANDALONE=y
+# CONFIG_PREVENT_FIRMWARE_BUILD is not set
+# CONFIG_FW_LOADER is not set
+# CONFIG_DEBUG_DRIVER is not set
+# CONFIG_SYS_HYPERVISOR is not set
+
+#
+# Connector - unified userspace <-> kernelspace linker
+#
+# CONFIG_CONNECTOR is not set
+
+#
+# Memory Technology Devices (MTD)
+#
+# CONFIG_MTD is not set
+
+#
+# Parallel port support
+#
+# CONFIG_PARPORT is not set
+
+#
+# Plug and Play support
+#
+
+#
+# Block devices
+#
+# CONFIG_BLK_DEV_COW_COMMON is not set
+CONFIG_BLK_DEV_LOOP=m
+# CONFIG_BLK_DEV_CRYPTOLOOP is not set
+CONFIG_BLK_DEV_NBD=m
+CONFIG_BLK_DEV_RAM=m
+CONFIG_BLK_DEV_RAM_COUNT=16
+CONFIG_BLK_DEV_RAM_SIZE=4096
+CONFIG_BLK_DEV_INITRD=y
+# CONFIG_CDROM_PKTCDVD is not set
+# CONFIG_ATA_OVER_ETH is not set
+
+#
+# ATA/ATAPI/MFM/RLL support
+#
+# CONFIG_IDE is not set
+
+#
+# SCSI device support
+#
+# CONFIG_RAID_ATTRS is not set
+# CONFIG_SCSI is not set
+
+#
+# Multi-device support (RAID and LVM)
+#
+# CONFIG_MD is not set
+
+#
+# Fusion MPT device support
+#
+# CONFIG_FUSION is not set
+
+#
+# IEEE 1394 (FireWire) support
+#
+
+#
+# I2O device support
+#
+
+#
+# Network device support
+#
+CONFIG_NETDEVICES=y
+CONFIG_DUMMY=y
+# CONFIG_BONDING is not set
+# CONFIG_EQUALIZER is not set
+CONFIG_TUN=m
+
+#
+# PHY device support
+#
+# CONFIG_PHYLIB is not set
+
+#
+# Ethernet (10 or 100Mbit)
+#
+CONFIG_NET_ETHERNET=y
+CONFIG_MII=y
+CONFIG_MACB=y
+
+#
+# Ethernet (1000 Mbit)
+#
+
+#
+# Ethernet (10000 Mbit)
+#
+
+#
+# Token Ring devices
+#
+
+#
+# Wireless LAN (non-hamradio)
+#
+# CONFIG_NET_RADIO is not set
+
+#
+# Wan interfaces
+#
+# CONFIG_WAN is not set
+CONFIG_PPP=m
+# CONFIG_PPP_MULTILINK is not set
+# CONFIG_PPP_FILTER is not set
+CONFIG_PPP_ASYNC=m
+# CONFIG_PPP_SYNC_TTY is not set
+CONFIG_PPP_DEFLATE=m
+# CONFIG_PPP_BSDCOMP is not set
+# CONFIG_PPP_MPPE is not set
+# CONFIG_PPPOE is not set
+# CONFIG_SLIP is not set
+# CONFIG_SHAPER is not set
+# CONFIG_NETCONSOLE is not set
+# CONFIG_NETPOLL is not set
+# CONFIG_NET_POLL_CONTROLLER is not set
+
+#
+# ISDN subsystem
+#
+# CONFIG_ISDN is not set
+
+#
+# Telephony Support
+#
+# CONFIG_PHONE is not set
+
+#
+# Input device support
+#
+# CONFIG_INPUT is not set
+
+#
+# Hardware I/O ports
+#
+# CONFIG_SERIO is not set
+# CONFIG_GAMEPORT is not set
+
+#
+# Character devices
+#
+# CONFIG_VT is not set
+# CONFIG_SERIAL_NONSTANDARD is not set
+
+#
+# Serial drivers
+#
+# CONFIG_SERIAL_8250 is not set
+
+#
+# Non-8250 serial port support
+#
+CONFIG_SERIAL_AT91=y
+CONFIG_SERIAL_AT91_CONSOLE=y
+# CONFIG_SERIAL_AT91_TTYAT is not set
+CONFIG_SERIAL_CORE=y
+CONFIG_SERIAL_CORE_CONSOLE=y
+CONFIG_UNIX98_PTYS=y
+# CONFIG_LEGACY_PTYS is not set
+
+#
+# IPMI
+#
+# CONFIG_IPMI_HANDLER is not set
+
+#
+# Watchdog Cards
+#
+# CONFIG_WATCHDOG is not set
+# CONFIG_HW_RANDOM is not set
+# CONFIG_RTC is not set
+# CONFIG_GEN_RTC is not set
+# CONFIG_DTLK is not set
+# CONFIG_R3964 is not set
+
+#
+# Ftape, the floppy tape device driver
+#
+# CONFIG_RAW_DRIVER is not set
+
+#
+# TPM devices
+#
+# CONFIG_TCG_TPM is not set
+# CONFIG_TELCLOCK is not set
+
+#
+# I2C support
+#
+# CONFIG_I2C is not set
+
+#
+# SPI support
+#
+CONFIG_SPI=y
+# CONFIG_SPI_DEBUG is not set
+CONFIG_SPI_MASTER=y
+
+#
+# SPI Master Controller Drivers
+#
+CONFIG_SPI_ATMEL=m
+# CONFIG_SPI_BITBANG is not set
+
+#
+# SPI Protocol Masters
+#
+
+#
+# Dallas's 1-wire bus
+#
+
+#
+# Hardware Monitoring support
+#
+# CONFIG_HWMON is not set
+# CONFIG_HWMON_VID is not set
+
+#
+# Misc devices
+#
+
+#
+# Multimedia devices
+#
+# CONFIG_VIDEO_DEV is not set
+CONFIG_VIDEO_V4L2=y
+
+#
+# Digital Video Broadcasting Devices
+#
+# CONFIG_DVB is not set
+
+#
+# Graphics support
+#
+# CONFIG_FIRMWARE_EDID is not set
+CONFIG_FB=m
+CONFIG_FB_CFB_FILLRECT=m
+CONFIG_FB_CFB_COPYAREA=m
+CONFIG_FB_CFB_IMAGEBLIT=m
+# CONFIG_FB_MACMODES is not set
+# CONFIG_FB_BACKLIGHT is not set
+# CONFIG_FB_MODE_HELPERS is not set
+# CONFIG_FB_TILEBLITTING is not set
+CONFIG_FB_SIDSA=m
+CONFIG_FB_SIDSA_DEFAULT_BPP=24
+# CONFIG_FB_S1D13XXX is not set
+# CONFIG_FB_VIRTUAL is not set
+
+#
+# Logo configuration
+#
+# CONFIG_LOGO is not set
+CONFIG_BACKLIGHT_LCD_SUPPORT=y
+# CONFIG_BACKLIGHT_CLASS_DEVICE is not set
+CONFIG_LCD_CLASS_DEVICE=m
+CONFIG_LCD_DEVICE=y
+CONFIG_LCD_LTV350QV=m
+
+#
+# Sound
+#
+# CONFIG_SOUND is not set
+
+#
+# USB support
+#
+# CONFIG_USB_ARCH_HAS_HCD is not set
+# CONFIG_USB_ARCH_HAS_OHCI is not set
+# CONFIG_USB_ARCH_HAS_EHCI is not set
+
+#
+# NOTE: USB_STORAGE enables SCSI, and 'SCSI disk support'
+#
+
+#
+# USB Gadget Support
+#
+# CONFIG_USB_GADGET is not set
+
+#
+# MMC/SD Card support
+#
+# CONFIG_MMC is not set
+
+#
+# LED devices
+#
+# CONFIG_NEW_LEDS is not set
+
+#
+# LED drivers
+#
+
+#
+# LED Triggers
+#
+
+#
+# InfiniBand support
+#
+
+#
+# EDAC - error detection and reporting (RAS) (EXPERIMENTAL)
+#
+
+#
+# Real Time Clock
+#
+# CONFIG_RTC_CLASS is not set
+
+#
+# DMA Engine support
+#
+# CONFIG_DMA_ENGINE is not set
+
+#
+# DMA Clients
+#
+
+#
+# DMA Devices
+#
+
+#
+# File systems
+#
+CONFIG_EXT2_FS=y
+# CONFIG_EXT2_FS_XATTR is not set
+# CONFIG_EXT2_FS_XIP is not set
+# CONFIG_EXT3_FS is not set
+# CONFIG_REISERFS_FS is not set
+# CONFIG_JFS_FS is not set
+# CONFIG_FS_POSIX_ACL is not set
+# CONFIG_XFS_FS is not set
+# CONFIG_OCFS2_FS is not set
+CONFIG_MINIX_FS=m
+CONFIG_ROMFS_FS=m
+# CONFIG_INOTIFY is not set
+# CONFIG_QUOTA is not set
+# CONFIG_DNOTIFY is not set
+# CONFIG_AUTOFS_FS is not set
+# CONFIG_AUTOFS4_FS is not set
+# CONFIG_FUSE_FS is not set
+
+#
+# CD-ROM/DVD Filesystems
+#
+# CONFIG_ISO9660_FS is not set
+# CONFIG_UDF_FS is not set
+
+#
+# DOS/FAT/NT Filesystems
+#
+CONFIG_FAT_FS=m
+CONFIG_MSDOS_FS=m
+CONFIG_VFAT_FS=m
+CONFIG_FAT_DEFAULT_CODEPAGE=437
+CONFIG_FAT_DEFAULT_IOCHARSET="iso8859-1"
+# CONFIG_NTFS_FS is not set
+
+#
+# Pseudo filesystems
+#
+CONFIG_PROC_FS=y
+CONFIG_PROC_KCORE=y
+CONFIG_SYSFS=y
+CONFIG_TMPFS=y
+# CONFIG_HUGETLB_PAGE is not set
+CONFIG_RAMFS=y
+CONFIG_CONFIGFS_FS=m
+
+#
+# Miscellaneous filesystems
+#
+# CONFIG_ADFS_FS is not set
+# CONFIG_AFFS_FS is not set
+# CONFIG_HFS_FS is not set
+# CONFIG_HFSPLUS_FS is not set
+# CONFIG_BEFS_FS is not set
+# CONFIG_BFS_FS is not set
+# CONFIG_EFS_FS is not set
+# CONFIG_CRAMFS is not set
+# CONFIG_VXFS_FS is not set
+# CONFIG_HPFS_FS is not set
+# CONFIG_QNX4FS_FS is not set
+# CONFIG_SYSV_FS is not set
+# CONFIG_UFS_FS is not set
+
+#
+# Network File Systems
+#
+CONFIG_NFS_FS=y
+CONFIG_NFS_V3=y
+# CONFIG_NFS_V3_ACL is not set
+# CONFIG_NFS_V4 is not set
+# CONFIG_NFS_DIRECTIO is not set
+# CONFIG_NFSD is not set
+CONFIG_ROOT_NFS=y
+CONFIG_LOCKD=y
+CONFIG_LOCKD_V4=y
+CONFIG_NFS_COMMON=y
+CONFIG_SUNRPC=y
+# CONFIG_RPCSEC_GSS_KRB5 is not set
+# CONFIG_RPCSEC_GSS_SPKM3 is not set
+# CONFIG_SMB_FS is not set
+CONFIG_CIFS=m
+# CONFIG_CIFS_STATS is not set
+# CONFIG_CIFS_WEAK_PW_HASH is not set
+# CONFIG_CIFS_XATTR is not set
+# CONFIG_CIFS_DEBUG2 is not set
+# CONFIG_CIFS_EXPERIMENTAL is not set
+# CONFIG_NCP_FS is not set
+# CONFIG_CODA_FS is not set
+# CONFIG_AFS_FS is not set
+# CONFIG_9P_FS is not set
+
+#
+# Partition Types
+#
+# CONFIG_PARTITION_ADVANCED is not set
+CONFIG_MSDOS_PARTITION=y
+
+#
+# Native Language Support
+#
+CONFIG_NLS=m
+CONFIG_NLS_DEFAULT="iso8859-1"
+CONFIG_NLS_CODEPAGE_437=m
+# CONFIG_NLS_CODEPAGE_737 is not set
+# CONFIG_NLS_CODEPAGE_775 is not set
+CONFIG_NLS_CODEPAGE_850=m
+# CONFIG_NLS_CODEPAGE_852 is not set
+# CONFIG_NLS_CODEPAGE_855 is not set
+# CONFIG_NLS_CODEPAGE_857 is not set
+# CONFIG_NLS_CODEPAGE_860 is not set
+# CONFIG_NLS_CODEPAGE_861 is not set
+# CONFIG_NLS_CODEPAGE_862 is not set
+# CONFIG_NLS_CODEPAGE_863 is not set
+# CONFIG_NLS_CODEPAGE_864 is not set
+# CONFIG_NLS_CODEPAGE_865 is not set
+# CONFIG_NLS_CODEPAGE_866 is not set
+# CONFIG_NLS_CODEPAGE_869 is not set
+# CONFIG_NLS_CODEPAGE_936 is not set
+# CONFIG_NLS_CODEPAGE_950 is not set
+# CONFIG_NLS_CODEPAGE_932 is not set
+# CONFIG_NLS_CODEPAGE_949 is not set
+# CONFIG_NLS_CODEPAGE_874 is not set
+# CONFIG_NLS_ISO8859_8 is not set
+# CONFIG_NLS_CODEPAGE_1250 is not set
+# CONFIG_NLS_CODEPAGE_1251 is not set
+# CONFIG_NLS_ASCII is not set
+CONFIG_NLS_ISO8859_1=m
+# CONFIG_NLS_ISO8859_2 is not set
+# CONFIG_NLS_ISO8859_3 is not set
+# CONFIG_NLS_ISO8859_4 is not set
+# CONFIG_NLS_ISO8859_5 is not set
+# CONFIG_NLS_ISO8859_6 is not set
+# CONFIG_NLS_ISO8859_7 is not set
+# CONFIG_NLS_ISO8859_9 is not set
+# CONFIG_NLS_ISO8859_13 is not set
+# CONFIG_NLS_ISO8859_14 is not set
+# CONFIG_NLS_ISO8859_15 is not set
+# CONFIG_NLS_KOI8_R is not set
+# CONFIG_NLS_KOI8_U is not set
+CONFIG_NLS_UTF8=m
+
+#
+# Kernel hacking
+#
+CONFIG_TRACE_IRQFLAGS_SUPPORT=y
+CONFIG_PRINTK_TIME=y
+CONFIG_MAGIC_SYSRQ=y
+# CONFIG_UNUSED_SYMBOLS is not set
+CONFIG_DEBUG_KERNEL=y
+CONFIG_LOG_BUF_SHIFT=14
+CONFIG_DETECT_SOFTLOCKUP=y
+# CONFIG_SCHEDSTATS is not set
+# CONFIG_DEBUG_SPINLOCK is not set
+# CONFIG_DEBUG_MUTEXES is not set
+# CONFIG_DEBUG_RWSEMS is not set
+# CONFIG_DEBUG_SPINLOCK_SLEEP is not set
+# CONFIG_DEBUG_LOCKING_API_SELFTESTS is not set
+# CONFIG_DEBUG_KOBJECT is not set
+CONFIG_DEBUG_BUGVERBOSE=y
+# CONFIG_DEBUG_INFO is not set
+CONFIG_DEBUG_FS=y
+# CONFIG_DEBUG_VM is not set
+CONFIG_FRAME_POINTER=y
+# CONFIG_UNWIND_INFO is not set
+CONFIG_FORCED_INLINING=y
+# CONFIG_RCU_TORTURE_TEST is not set
+CONFIG_KPROBES=y
+
+#
+# Security options
+#
+# CONFIG_KEYS is not set
+# CONFIG_SECURITY is not set
+
+#
+# Cryptographic options
+#
+# CONFIG_CRYPTO is not set
+
+#
+# Hardware crypto devices
+#
+
+#
+# Library routines
+#
+CONFIG_CRC_CCITT=m
+# CONFIG_CRC16 is not set
+CONFIG_CRC32=m
+# CONFIG_LIBCRC32C is not set
+CONFIG_ZLIB_INFLATE=m
+CONFIG_ZLIB_DEFLATE=m
diff --git a/arch/avr32/kernel/Makefile b/arch/avr32/kernel/Makefile
new file mode 100644 (file)
index 0000000..90e5aff
--- /dev/null
@@ -0,0 +1,18 @@
+#
+# Makefile for the Linux/AVR32 kernel.
+#
+
+extra-y                                := head.o vmlinux.lds
+
+obj-$(CONFIG_SUBARCH_AVR32B)   += entry-avr32b.o
+obj-y                          += syscall_table.o syscall-stubs.o irq.o
+obj-y                          += setup.o traps.o semaphore.o ptrace.o
+obj-y                          += signal.o sys_avr32.o process.o time.o
+obj-y                          += init_task.o switch_to.o cpu.o
+obj-$(CONFIG_MODULES)          += module.o avr32_ksyms.o
+obj-$(CONFIG_KPROBES)          += kprobes.o
+
+USE_STANDARD_AS_RULE           := true
+
+%.lds: %.lds.c FORCE
+       $(call if_changed_dep,cpp_lds_S)
diff --git a/arch/avr32/kernel/asm-offsets.c b/arch/avr32/kernel/asm-offsets.c
new file mode 100644 (file)
index 0000000..97d8658
--- /dev/null
@@ -0,0 +1,25 @@
+/*
+ * Generate definitions needed by assembly language modules.
+ * This code generates raw asm output which is post-processed
+ * to extract and format the required data.
+ */
+
+#include <linux/thread_info.h>
+
+#define DEFINE(sym, val) \
+        asm volatile("\n->" #sym " %0 " #val : : "i" (val))
+
+#define BLANK() asm volatile("\n->" : : )
+
+#define OFFSET(sym, str, mem) \
+        DEFINE(sym, offsetof(struct str, mem));
+
+void foo(void)
+{
+       OFFSET(TI_task, thread_info, task);
+       OFFSET(TI_exec_domain, thread_info, exec_domain);
+       OFFSET(TI_flags, thread_info, flags);
+       OFFSET(TI_cpu, thread_info, cpu);
+       OFFSET(TI_preempt_count, thread_info, preempt_count);
+       OFFSET(TI_restart_block, thread_info, restart_block);
+}
diff --git a/arch/avr32/kernel/avr32_ksyms.c b/arch/avr32/kernel/avr32_ksyms.c
new file mode 100644 (file)
index 0000000..04f767a
--- /dev/null
@@ -0,0 +1,55 @@
+/*
+ * Export AVR32-specific functions for loadable modules.
+ *
+ * Copyright (C) 2004-2006 Atmel Corporation
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ */
+#include <linux/module.h>
+
+#include <asm/checksum.h>
+#include <asm/uaccess.h>
+#include <asm/delay.h>
+
+/*
+ * GCC functions
+ */
+extern unsigned long long __avr32_lsl64(unsigned long long u, unsigned long b);
+extern unsigned long long __avr32_lsr64(unsigned long long u, unsigned long b);
+extern unsigned long long __avr32_asr64(unsigned long long u, unsigned long b);
+EXPORT_SYMBOL(__avr32_lsl64);
+EXPORT_SYMBOL(__avr32_lsr64);
+EXPORT_SYMBOL(__avr32_asr64);
+
+/*
+ * String functions
+ */
+EXPORT_SYMBOL(memset);
+EXPORT_SYMBOL(memcpy);
+
+/*
+ * Userspace access stuff.
+ */
+EXPORT_SYMBOL(copy_from_user);
+EXPORT_SYMBOL(copy_to_user);
+EXPORT_SYMBOL(__copy_user);
+EXPORT_SYMBOL(strncpy_from_user);
+EXPORT_SYMBOL(__strncpy_from_user);
+EXPORT_SYMBOL(clear_user);
+EXPORT_SYMBOL(__clear_user);
+EXPORT_SYMBOL(csum_partial);
+EXPORT_SYMBOL(csum_partial_copy_generic);
+
+/* Delay loops (lib/delay.S) */
+EXPORT_SYMBOL(__ndelay);
+EXPORT_SYMBOL(__udelay);
+EXPORT_SYMBOL(__const_udelay);
+
+/* Bit operations (lib/findbit.S) */
+EXPORT_SYMBOL(find_first_zero_bit);
+EXPORT_SYMBOL(find_next_zero_bit);
+EXPORT_SYMBOL(find_first_bit);
+EXPORT_SYMBOL(find_next_bit);
+EXPORT_SYMBOL(generic_find_next_zero_le_bit);
diff --git a/arch/avr32/kernel/cpu.c b/arch/avr32/kernel/cpu.c
new file mode 100644 (file)
index 0000000..342452b
--- /dev/null
@@ -0,0 +1,327 @@
+/*
+ * Copyright (C) 2005-2006 Atmel Corporation
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ */
+#include <linux/init.h>
+#include <linux/sysdev.h>
+#include <linux/seq_file.h>
+#include <linux/cpu.h>
+#include <linux/percpu.h>
+#include <linux/param.h>
+#include <linux/errno.h>
+
+#include <asm/setup.h>
+#include <asm/sysreg.h>
+
+static DEFINE_PER_CPU(struct cpu, cpu_devices);
+
+#ifdef CONFIG_PERFORMANCE_COUNTERS
+
+/*
+ * XXX: If/when a SMP-capable implementation of AVR32 will ever be
+ * made, we must make sure that the code executes on the correct CPU.
+ */
+static ssize_t show_pc0event(struct sys_device *dev, char *buf)
+{
+       unsigned long pccr;
+
+       pccr = sysreg_read(PCCR);
+       return sprintf(buf, "0x%lx\n", (pccr >> 12) & 0x3f);
+}
+static ssize_t store_pc0event(struct sys_device *dev, const char *buf,
+                             size_t count)
+{
+       unsigned long val;
+       char *endp;
+
+       val = simple_strtoul(buf, &endp, 0);
+       if (endp == buf || val > 0x3f)
+               return -EINVAL;
+       val = (val << 12) | (sysreg_read(PCCR) & 0xfffc0fff);
+       sysreg_write(PCCR, val);
+       return count;
+}
+static ssize_t show_pc0count(struct sys_device *dev, char *buf)
+{
+       unsigned long pcnt0;
+
+       pcnt0 = sysreg_read(PCNT0);
+       return sprintf(buf, "%lu\n", pcnt0);
+}
+static ssize_t store_pc0count(struct sys_device *dev, const char *buf,
+                             size_t count)
+{
+       unsigned long val;
+       char *endp;
+
+       val = simple_strtoul(buf, &endp, 0);
+       if (endp == buf)
+               return -EINVAL;
+       sysreg_write(PCNT0, val);
+
+       return count;
+}
+
+static ssize_t show_pc1event(struct sys_device *dev, char *buf)
+{
+       unsigned long pccr;
+
+       pccr = sysreg_read(PCCR);
+       return sprintf(buf, "0x%lx\n", (pccr >> 18) & 0x3f);
+}
+static ssize_t store_pc1event(struct sys_device *dev, const char *buf,
+                             size_t count)
+{
+       unsigned long val;
+       char *endp;
+
+       val = simple_strtoul(buf, &endp, 0);
+       if (endp == buf || val > 0x3f)
+               return -EINVAL;
+       val = (val << 18) | (sysreg_read(PCCR) & 0xff03ffff);
+       sysreg_write(PCCR, val);
+       return count;
+}
+static ssize_t show_pc1count(struct sys_device *dev, char *buf)
+{
+       unsigned long pcnt1;
+
+       pcnt1 = sysreg_read(PCNT1);
+       return sprintf(buf, "%lu\n", pcnt1);
+}
+static ssize_t store_pc1count(struct sys_device *dev, const char *buf,
+                             size_t count)
+{
+       unsigned long val;
+       char *endp;
+
+       val = simple_strtoul(buf, &endp, 0);
+       if (endp == buf)
+               return -EINVAL;
+       sysreg_write(PCNT1, val);
+
+       return count;
+}
+
+static ssize_t show_pccycles(struct sys_device *dev, char *buf)
+{
+       unsigned long pccnt;
+
+       pccnt = sysreg_read(PCCNT);
+       return sprintf(buf, "%lu\n", pccnt);
+}
+static ssize_t store_pccycles(struct sys_device *dev, const char *buf,
+                             size_t count)
+{
+       unsigned long val;
+       char *endp;
+
+       val = simple_strtoul(buf, &endp, 0);
+       if (endp == buf)
+               return -EINVAL;
+       sysreg_write(PCCNT, val);
+
+       return count;
+}
+
+static ssize_t show_pcenable(struct sys_device *dev, char *buf)
+{
+       unsigned long pccr;
+
+       pccr = sysreg_read(PCCR);
+       return sprintf(buf, "%c\n", (pccr & 1)?'1':'0');
+}
+static ssize_t store_pcenable(struct sys_device *dev, const char *buf,
+                             size_t count)
+{
+       unsigned long pccr, val;
+       char *endp;
+
+       val = simple_strtoul(buf, &endp, 0);
+       if (endp == buf)
+               return -EINVAL;
+       if (val)
+               val = 1;
+
+       pccr = sysreg_read(PCCR);
+       pccr = (pccr & ~1UL) | val;
+       sysreg_write(PCCR, pccr);
+
+       return count;
+}
+
+static SYSDEV_ATTR(pc0event, 0600, show_pc0event, store_pc0event);
+static SYSDEV_ATTR(pc0count, 0600, show_pc0count, store_pc0count);
+static SYSDEV_ATTR(pc1event, 0600, show_pc1event, store_pc1event);
+static SYSDEV_ATTR(pc1count, 0600, show_pc1count, store_pc1count);
+static SYSDEV_ATTR(pccycles, 0600, show_pccycles, store_pccycles);
+static SYSDEV_ATTR(pcenable, 0600, show_pcenable, store_pcenable);
+
+#endif /* CONFIG_PERFORMANCE_COUNTERS */
+
+static int __init topology_init(void)
+{
+       int cpu;
+
+       for_each_possible_cpu(cpu) {
+               struct cpu *c = &per_cpu(cpu_devices, cpu);
+
+               register_cpu(c, cpu);
+
+#ifdef CONFIG_PERFORMANCE_COUNTERS
+               sysdev_create_file(&c->sysdev, &attr_pc0event);
+               sysdev_create_file(&c->sysdev, &attr_pc0count);
+               sysdev_create_file(&c->sysdev, &attr_pc1event);
+               sysdev_create_file(&c->sysdev, &attr_pc1count);
+               sysdev_create_file(&c->sysdev, &attr_pccycles);
+               sysdev_create_file(&c->sysdev, &attr_pcenable);
+#endif
+       }
+
+       return 0;
+}
+
+subsys_initcall(topology_init);
+
+static const char *cpu_names[] = {
+       "Morgan",
+       "AP7000",
+};
+#define NR_CPU_NAMES ARRAY_SIZE(cpu_names)
+
+static const char *arch_names[] = {
+       "AVR32A",
+       "AVR32B",
+};
+#define NR_ARCH_NAMES ARRAY_SIZE(arch_names)
+
+static const char *mmu_types[] = {
+       "No MMU",
+       "ITLB and DTLB",
+       "Shared TLB",
+       "MPU"
+};
+
+void __init setup_processor(void)
+{
+       unsigned long config0, config1;
+       unsigned cpu_id, cpu_rev, arch_id, arch_rev, mmu_type;
+       unsigned tmp;
+
+       config0 = sysreg_read(CONFIG0); /* 0x0000013e; */
+       config1 = sysreg_read(CONFIG1); /* 0x01f689a2; */
+       cpu_id = config0 >> 24;
+       cpu_rev = (config0 >> 16) & 0xff;
+       arch_id = (config0 >> 13) & 0x07;
+       arch_rev = (config0 >> 10) & 0x07;
+       mmu_type = (config0 >> 7) & 0x03;
+
+       boot_cpu_data.arch_type = arch_id;
+       boot_cpu_data.cpu_type = cpu_id;
+       boot_cpu_data.arch_revision = arch_rev;
+       boot_cpu_data.cpu_revision = cpu_rev;
+       boot_cpu_data.tlb_config = mmu_type;
+
+       tmp = (config1 >> 13) & 0x07;
+       if (tmp) {
+               boot_cpu_data.icache.ways = 1 << ((config1 >> 10) & 0x07);
+               boot_cpu_data.icache.sets = 1 << ((config1 >> 16) & 0x0f);
+               boot_cpu_data.icache.linesz = 1 << (tmp + 1);
+       }
+       tmp = (config1 >> 3) & 0x07;
+       if (tmp) {
+               boot_cpu_data.dcache.ways = 1 << (config1 & 0x07);
+               boot_cpu_data.dcache.sets = 1 << ((config1 >> 6) & 0x0f);
+               boot_cpu_data.dcache.linesz = 1 << (tmp + 1);
+       }
+
+       if ((cpu_id >= NR_CPU_NAMES) || (arch_id >= NR_ARCH_NAMES)) {
+               printk ("Unknown CPU configuration (ID %02x, arch %02x), "
+                       "continuing anyway...\n",
+                       cpu_id, arch_id);
+               return;
+       }
+
+       printk ("CPU: %s [%02x] revision %d (%s revision %d)\n",
+               cpu_names[cpu_id], cpu_id, cpu_rev,
+               arch_names[arch_id], arch_rev);
+       printk ("CPU: MMU configuration: %s\n", mmu_types[mmu_type]);
+       printk ("CPU: features:");
+       if (config0 & (1 << 6))
+               printk(" fpu");
+       if (config0 & (1 << 5))
+               printk(" java");
+       if (config0 & (1 << 4))
+               printk(" perfctr");
+       if (config0 & (1 << 3))
+               printk(" ocd");
+       printk("\n");
+}
+
+#ifdef CONFIG_PROC_FS
+static int c_show(struct seq_file *m, void *v)
+{
+       unsigned int icache_size, dcache_size;
+       unsigned int cpu = smp_processor_id();
+
+       icache_size = boot_cpu_data.icache.ways *
+               boot_cpu_data.icache.sets *
+               boot_cpu_data.icache.linesz;
+       dcache_size = boot_cpu_data.dcache.ways *
+               boot_cpu_data.dcache.sets *
+               boot_cpu_data.dcache.linesz;
+
+       seq_printf(m, "processor\t: %d\n", cpu);
+
+       if (boot_cpu_data.arch_type < NR_ARCH_NAMES)
+               seq_printf(m, "cpu family\t: %s revision %d\n",
+                          arch_names[boot_cpu_data.arch_type],
+                          boot_cpu_data.arch_revision);
+       if (boot_cpu_data.cpu_type < NR_CPU_NAMES)
+               seq_printf(m, "cpu type\t: %s revision %d\n",
+                          cpu_names[boot_cpu_data.cpu_type],
+                          boot_cpu_data.cpu_revision);
+
+       seq_printf(m, "i-cache\t\t: %dK (%u ways x %u sets x %u)\n",
+                  icache_size >> 10,
+                  boot_cpu_data.icache.ways,
+                  boot_cpu_data.icache.sets,
+                  boot_cpu_data.icache.linesz);
+       seq_printf(m, "d-cache\t\t: %dK (%u ways x %u sets x %u)\n",
+                  dcache_size >> 10,
+                  boot_cpu_data.dcache.ways,
+                  boot_cpu_data.dcache.sets,
+                  boot_cpu_data.dcache.linesz);
+       seq_printf(m, "bogomips\t: %lu.%02lu\n",
+                  boot_cpu_data.loops_per_jiffy / (500000/HZ),
+                  (boot_cpu_data.loops_per_jiffy / (5000/HZ)) % 100);
+
+       return 0;
+}
+
+static void *c_start(struct seq_file *m, loff_t *pos)
+{
+       return *pos < 1 ? (void *)1 : NULL;
+}
+
+static void *c_next(struct seq_file *m, void *v, loff_t *pos)
+{
+       ++*pos;
+       return NULL;
+}
+
+static void c_stop(struct seq_file *m, void *v)
+{
+
+}
+
+struct seq_operations cpuinfo_op = {
+       .start  = c_start,
+       .next   = c_next,
+       .stop   = c_stop,
+       .show   = c_show
+};
+#endif /* CONFIG_PROC_FS */
diff --git a/arch/avr32/kernel/entry-avr32b.S b/arch/avr32/kernel/entry-avr32b.S
new file mode 100644 (file)
index 0000000..eeb6679
--- /dev/null
@@ -0,0 +1,678 @@
+/*
+ * Copyright (C) 2004-2006 Atmel Corporation
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ */
+
+/*
+ * This file contains the low-level entry-points into the kernel, that is,
+ * exception handlers, debug trap handlers, interrupt handlers and the
+ * system call handler.
+ */
+#include <linux/errno.h>
+
+#include <asm/asm.h>
+#include <asm/hardirq.h>
+#include <asm/irq.h>
+#include <asm/ocd.h>
+#include <asm/page.h>
+#include <asm/pgtable.h>
+#include <asm/ptrace.h>
+#include <asm/sysreg.h>
+#include <asm/thread_info.h>
+#include <asm/unistd.h>
+
+#ifdef CONFIG_PREEMPT
+# define preempt_stop          mask_interrupts
+#else
+# define preempt_stop
+# define fault_resume_kernel   fault_restore_all
+#endif
+
+#define __MASK(x)      ((1 << (x)) - 1)
+#define IRQ_MASK       ((__MASK(SOFTIRQ_BITS) << SOFTIRQ_SHIFT) | \
+                        (__MASK(HARDIRQ_BITS) << HARDIRQ_SHIFT))
+
+       .section .ex.text,"ax",@progbits
+       .align  2
+exception_vectors:
+       bral    handle_critical
+       .align  2
+       bral    handle_critical
+       .align  2
+       bral    do_bus_error_write
+       .align  2
+       bral    do_bus_error_read
+       .align  2
+       bral    do_nmi_ll
+       .align  2
+       bral    handle_address_fault
+       .align  2
+       bral    handle_protection_fault
+       .align  2
+       bral    handle_debug
+       .align  2
+       bral    do_illegal_opcode_ll
+       .align  2
+       bral    do_illegal_opcode_ll
+       .align  2
+       bral    do_illegal_opcode_ll
+       .align  2
+       bral    do_fpe_ll
+       .align  2
+       bral    do_illegal_opcode_ll
+       .align  2
+       bral    handle_address_fault
+       .align  2
+       bral    handle_address_fault
+       .align  2
+       bral    handle_protection_fault
+       .align  2
+       bral    handle_protection_fault
+       .align  2
+       bral    do_dtlb_modified
+
+       /*
+        * r0 : PGD/PT/PTE
+        * r1 : Offending address
+        * r2 : Scratch register
+        * r3 : Cause (5, 12 or 13)
+        */
+#define        tlbmiss_save    pushm   r0-r3
+#define tlbmiss_restore        popm    r0-r3
+
+       .section .tlbx.ex.text,"ax",@progbits
+       .global itlb_miss
+itlb_miss:
+       tlbmiss_save
+       rjmp    tlb_miss_common
+
+       .section .tlbr.ex.text,"ax",@progbits
+dtlb_miss_read:
+       tlbmiss_save
+       rjmp    tlb_miss_common
+
+       .section .tlbw.ex.text,"ax",@progbits
+dtlb_miss_write:
+       tlbmiss_save
+
+       .global tlb_miss_common
+tlb_miss_common:
+       mfsr    r0, SYSREG_PTBR
+       mfsr    r1, SYSREG_TLBEAR
+
+       /* Is it the vmalloc space? */
+       bld     r1, 31
+       brcs    handle_vmalloc_miss
+
+       /* First level lookup */
+pgtbl_lookup:
+       lsr     r2, r1, PGDIR_SHIFT
+       ld.w    r0, r0[r2 << 2]
+       bld     r0, _PAGE_BIT_PRESENT
+       brcc    page_table_not_present
+
+       /* TODO: Check access rights on page table if necessary */
+
+       /* Translate to virtual address in P1. */
+       andl    r0, 0xf000
+       sbr     r0, 31
+
+       /* Second level lookup */
+       lsl     r1, (32 - PGDIR_SHIFT)
+       lsr     r1, (32 - PGDIR_SHIFT) + PAGE_SHIFT
+       add     r2, r0, r1 << 2
+       ld.w    r1, r2[0]
+       bld     r1, _PAGE_BIT_PRESENT
+       brcc    page_not_present
+
+       /* Mark the page as accessed */
+       sbr     r1, _PAGE_BIT_ACCESSED
+       st.w    r2[0], r1
+
+       /* Drop software flags */
+       andl    r1, _PAGE_FLAGS_HARDWARE_MASK & 0xffff
+       mtsr    SYSREG_TLBELO, r1
+
+       /* Figure out which entry we want to replace */
+       mfsr    r0, SYSREG_TLBARLO
+       clz     r2, r0
+       brcc    1f
+       mov     r1, -1                  /* All entries have been accessed, */
+       mtsr    SYSREG_TLBARLO, r1      /* so reset TLBAR */
+       mov     r2, 0                   /* and start at 0 */
+1:     mfsr    r1, SYSREG_MMUCR
+       lsl     r2, 14
+       andl    r1, 0x3fff, COH
+       or      r1, r2
+       mtsr    SYSREG_MMUCR, r1
+
+       tlbw
+
+       tlbmiss_restore
+       rete
+
+handle_vmalloc_miss:
+       /* Simply do the lookup in init's page table */
+       mov     r0, lo(swapper_pg_dir)
+       orh     r0, hi(swapper_pg_dir)
+       rjmp    pgtbl_lookup
+
+
+       /* ---                    System Call                    --- */
+
+       .section .scall.text,"ax",@progbits
+system_call:
+       pushm   r12             /* r12_orig */
+       stmts   --sp, r0-lr
+       zero_fp
+       mfsr    r0, SYSREG_RAR_SUP
+       mfsr    r1, SYSREG_RSR_SUP
+       stm     --sp, r0-r1
+
+       /* check for syscall tracing */
+       get_thread_info r0
+       ld.w    r1, r0[TI_flags]
+       bld     r1, TIF_SYSCALL_TRACE
+       brcs    syscall_trace_enter
+
+syscall_trace_cont:
+       cp.w    r8, NR_syscalls
+       brhs    syscall_badsys
+
+       lddpc   lr, syscall_table_addr
+       ld.w    lr, lr[r8 << 2]
+       mov     r8, r5          /* 5th argument (6th is pushed by stub) */
+       icall   lr
+
+       .global syscall_return
+syscall_return:
+       get_thread_info r0
+       mask_interrupts         /* make sure we don't miss an interrupt
+                                  setting need_resched or sigpending
+                                  between sampling and the rets */
+
+       /* Store the return value so that the correct value is loaded below */
+       stdsp   sp[REG_R12], r12
+
+       ld.w    r1, r0[TI_flags]
+       andl    r1, _TIF_ALLWORK_MASK, COH
+       brne    syscall_exit_work
+
+syscall_exit_cont:
+       popm    r8-r9
+       mtsr    SYSREG_RAR_SUP, r8
+       mtsr    SYSREG_RSR_SUP, r9
+       ldmts   sp++, r0-lr
+       sub     sp, -4          /* r12_orig */
+       rets
+
+       .align  2
+syscall_table_addr:
+       .long   sys_call_table
+
+syscall_badsys:
+       mov     r12, -ENOSYS
+       rjmp    syscall_return
+
+       .global ret_from_fork
+ret_from_fork:
+       rcall   schedule_tail
+
+       /* check for syscall tracing */
+       get_thread_info r0
+       ld.w    r1, r0[TI_flags]
+       andl    r1, _TIF_ALLWORK_MASK, COH
+       brne    syscall_exit_work
+       rjmp    syscall_exit_cont
+
+syscall_trace_enter:
+       pushm   r8-r12
+       rcall   syscall_trace
+       popm    r8-r12
+       rjmp    syscall_trace_cont
+
+syscall_exit_work:
+       bld     r1, TIF_SYSCALL_TRACE
+       brcc    1f
+       unmask_interrupts
+       rcall   syscall_trace
+       mask_interrupts
+       ld.w    r1, r0[TI_flags]
+
+1:     bld     r1, TIF_NEED_RESCHED
+       brcc    2f
+       unmask_interrupts
+       rcall   schedule
+       mask_interrupts
+       ld.w    r1, r0[TI_flags]
+       rjmp    1b
+
+2:     mov     r2, _TIF_SIGPENDING | _TIF_RESTORE_SIGMASK
+       tst     r1, r2
+       breq    3f
+       unmask_interrupts
+       mov     r12, sp
+       mov     r11, r0
+       rcall   do_notify_resume
+       mask_interrupts
+       ld.w    r1, r0[TI_flags]
+       rjmp    1b
+
+3:     bld     r1, TIF_BREAKPOINT
+       brcc    syscall_exit_cont
+       mfsr    r3, SYSREG_TLBEHI
+       lddsp   r2, sp[REG_PC]
+       andl    r3, 0xff, COH
+       lsl     r3, 1
+       sbr     r3, 30
+       sbr     r3, 0
+       mtdr    DBGREG_BWA2A, r2
+       mtdr    DBGREG_BWC2A, r3
+       rjmp    syscall_exit_cont
+
+
+       /* The slow path of the TLB miss handler */
+page_table_not_present:
+page_not_present:
+       tlbmiss_restore
+       sub     sp, 4
+       stmts   --sp, r0-lr
+       rcall   save_full_context_ex
+       mfsr    r12, SYSREG_ECR
+       mov     r11, sp
+       rcall   do_page_fault
+       rjmp    ret_from_exception
+
+       /* This function expects to find offending PC in SYSREG_RAR_EX */
+save_full_context_ex:
+       mfsr    r8, SYSREG_RSR_EX
+       mov     r12, r8
+       andh    r8, (MODE_MASK >> 16), COH
+       mfsr    r11, SYSREG_RAR_EX
+       brne    2f
+
+1:     pushm   r11, r12        /* PC and SR */
+       unmask_exceptions
+       ret     r12
+
+2:     sub     r10, sp, -(FRAME_SIZE_FULL - REG_LR)
+       stdsp   sp[4], r10      /* replace saved SP */
+       rjmp    1b
+
+       /* Low-level exception handlers */
+handle_critical:
+       pushm   r12
+       pushm   r0-r12
+       rcall   save_full_context_ex
+       mfsr    r12, SYSREG_ECR
+       mov     r11, sp
+       rcall   do_critical_exception
+
+       /* We should never get here... */
+bad_return:
+       sub     r12, pc, (. - 1f)
+       bral    panic
+       .align  2
+1:     .asciz  "Return from critical exception!"
+
+       .align  1
+do_bus_error_write:
+       sub     sp, 4
+       stmts   --sp, r0-lr
+       rcall   save_full_context_ex
+       mov     r11, 1
+       rjmp    1f
+
+do_bus_error_read:
+       sub     sp, 4
+       stmts   --sp, r0-lr
+       rcall   save_full_context_ex
+       mov     r11, 0
+1:     mfsr    r12, SYSREG_BEAR
+       mov     r10, sp
+       rcall   do_bus_error
+       rjmp    ret_from_exception
+
+       .align  1
+do_nmi_ll:
+       sub     sp, 4
+       stmts   --sp, r0-lr
+       /* FIXME: Make sure RAR_NMI and RSR_NMI are pushed instead of *_EX */
+       rcall   save_full_context_ex
+       mfsr    r12, SYSREG_ECR
+       mov     r11, sp
+       rcall   do_nmi
+       rjmp    bad_return
+
+handle_address_fault:
+       sub     sp, 4
+       stmts   --sp, r0-lr
+       rcall   save_full_context_ex
+       mfsr    r12, SYSREG_ECR
+       mov     r11, sp
+       rcall   do_address_exception
+       rjmp    ret_from_exception
+
+handle_protection_fault:
+       sub     sp, 4
+       stmts   --sp, r0-lr
+       rcall   save_full_context_ex
+       mfsr    r12, SYSREG_ECR
+       mov     r11, sp
+       rcall   do_page_fault
+       rjmp    ret_from_exception
+
+       .align  1
+do_illegal_opcode_ll:
+       sub     sp, 4
+       stmts   --sp, r0-lr
+       rcall   save_full_context_ex
+       mfsr    r12, SYSREG_ECR
+       mov     r11, sp
+       rcall   do_illegal_opcode
+       rjmp    ret_from_exception
+
+do_dtlb_modified:
+       pushm   r0-r3
+       mfsr    r1, SYSREG_TLBEAR
+       mfsr    r0, SYSREG_PTBR
+       lsr     r2, r1, PGDIR_SHIFT
+       ld.w    r0, r0[r2 << 2]
+       lsl     r1, (32 - PGDIR_SHIFT)
+       lsr     r1, (32 - PGDIR_SHIFT) + PAGE_SHIFT
+
+       /* Translate to virtual address in P1 */
+       andl    r0, 0xf000
+       sbr     r0, 31
+       add     r2, r0, r1 << 2
+       ld.w    r3, r2[0]
+       sbr     r3, _PAGE_BIT_DIRTY
+       mov     r0, r3
+       st.w    r2[0], r3
+
+       /* The page table is up-to-date. Update the TLB entry as well */
+       andl    r0, lo(_PAGE_FLAGS_HARDWARE_MASK)
+       mtsr    SYSREG_TLBELO, r0
+
+       /* MMUCR[DRP] is updated automatically, so let's go... */
+       tlbw
+
+       popm    r0-r3
+       rete
+
+do_fpe_ll:
+       sub     sp, 4
+       stmts   --sp, r0-lr
+       rcall   save_full_context_ex
+       unmask_interrupts
+       mov     r12, 26
+       mov     r11, sp
+       rcall   do_fpe
+       rjmp    ret_from_exception
+
+ret_from_exception:
+       mask_interrupts
+       lddsp   r4, sp[REG_SR]
+       andh    r4, (MODE_MASK >> 16), COH
+       brne    fault_resume_kernel
+
+       get_thread_info r0
+       ld.w    r1, r0[TI_flags]
+       andl    r1, _TIF_WORK_MASK, COH
+       brne    fault_exit_work
+
+fault_resume_user:
+       popm    r8-r9
+       mask_exceptions
+       mtsr    SYSREG_RAR_EX, r8
+       mtsr    SYSREG_RSR_EX, r9
+       ldmts   sp++, r0-lr
+       sub     sp, -4
+       rete
+
+fault_resume_kernel:
+#ifdef CONFIG_PREEMPT
+       get_thread_info r0
+       ld.w    r2, r0[TI_preempt_count]
+       cp.w    r2, 0
+       brne    1f
+       ld.w    r1, r0[TI_flags]
+       bld     r1, TIF_NEED_RESCHED
+       brcc    1f
+       lddsp   r4, sp[REG_SR]
+       bld     r4, SYSREG_GM_OFFSET
+       brcs    1f
+       rcall   preempt_schedule_irq
+1:
+#endif
+
+       popm    r8-r9
+       mask_exceptions
+       mfsr    r1, SYSREG_SR
+       mtsr    SYSREG_RAR_EX, r8
+       mtsr    SYSREG_RSR_EX, r9
+       popm    lr
+       sub     sp, -4          /* ignore SP */
+       popm    r0-r12
+       sub     sp, -4          /* ignore r12_orig */
+       rete
+
+irq_exit_work:
+       /* Switch to exception mode so that we can share the same code. */
+       mfsr    r8, SYSREG_SR
+       cbr     r8, SYSREG_M0_OFFSET
+       orh     r8, hi(SYSREG_BIT(M1) | SYSREG_BIT(M2))
+       mtsr    SYSREG_SR, r8
+       sub     pc, -2
+       get_thread_info r0
+       ld.w    r1, r0[TI_flags]
+
+fault_exit_work:
+       bld     r1, TIF_NEED_RESCHED
+       brcc    1f
+       unmask_interrupts
+       rcall   schedule
+       mask_interrupts
+       ld.w    r1, r0[TI_flags]
+       rjmp    fault_exit_work
+
+1:     mov     r2, _TIF_SIGPENDING | _TIF_RESTORE_SIGMASK
+       tst     r1, r2
+       breq    2f
+       unmask_interrupts
+       mov     r12, sp
+       mov     r11, r0
+       rcall   do_notify_resume
+       mask_interrupts
+       ld.w    r1, r0[TI_flags]
+       rjmp    fault_exit_work
+
+2:     bld     r1, TIF_BREAKPOINT
+       brcc    fault_resume_user
+       mfsr    r3, SYSREG_TLBEHI
+       lddsp   r2, sp[REG_PC]
+       andl    r3, 0xff, COH
+       lsl     r3, 1
+       sbr     r3, 30
+       sbr     r3, 0
+       mtdr    DBGREG_BWA2A, r2
+       mtdr    DBGREG_BWC2A, r3
+       rjmp    fault_resume_user
+
+       /* If we get a debug trap from privileged context we end up here */
+handle_debug_priv:
+       /* Fix up LR and SP in regs. r11 contains the mode we came from */
+       mfsr    r8, SYSREG_SR
+       mov     r9, r8
+       andh    r8, hi(~MODE_MASK)
+       or      r8, r11
+       mtsr    SYSREG_SR, r8
+       sub     pc, -2
+       stdsp   sp[REG_LR], lr
+       mtsr    SYSREG_SR, r9
+       sub     pc, -2
+       sub     r10, sp, -FRAME_SIZE_FULL
+       stdsp   sp[REG_SP], r10
+       mov     r12, sp
+       rcall   do_debug_priv
+
+       /* Now, put everything back */
+       ssrf    SR_EM_BIT
+       popm    r10, r11
+       mtsr    SYSREG_RAR_DBG, r10
+       mtsr    SYSREG_RSR_DBG, r11
+       mfsr    r8, SYSREG_SR
+       mov     r9, r8
+       andh    r8, hi(~MODE_MASK)
+       andh    r11, hi(MODE_MASK)
+       or      r8, r11
+       mtsr    SYSREG_SR, r8
+       sub     pc, -2
+       popm    lr
+       mtsr    SYSREG_SR, r9
+       sub     pc, -2
+       sub     sp, -4          /* skip SP */
+       popm    r0-r12
+       sub     sp, -4
+       retd
+
+       /*
+        * At this point, everything is masked, that is, interrupts,
+        * exceptions and debugging traps. We might get called from
+        * interrupt or exception context in some rare cases, but this
+        * will be taken care of by do_debug(), so we're not going to
+        * do a 100% correct context save here.
+        */
+handle_debug:
+       sub     sp, 4           /* r12_orig */
+       stmts   --sp, r0-lr
+       mfsr    r10, SYSREG_RAR_DBG
+       mfsr    r11, SYSREG_RSR_DBG
+       unmask_exceptions
+       pushm   r10,r11
+       andh    r11, (MODE_MASK >> 16), COH
+       brne    handle_debug_priv
+
+       mov     r12, sp
+       rcall   do_debug
+
+       lddsp   r10, sp[REG_SR]
+       andh    r10, (MODE_MASK >> 16), COH
+       breq    debug_resume_user
+
+debug_restore_all:
+       popm    r10,r11
+       mask_exceptions
+       mtsr    SYSREG_RSR_DBG, r11
+       mtsr    SYSREG_RAR_DBG, r10
+       ldmts   sp++, r0-lr
+       sub     sp, -4
+       retd
+
+debug_resume_user:
+       get_thread_info r0
+       mask_interrupts
+
+       ld.w    r1, r0[TI_flags]
+       andl    r1, _TIF_DBGWORK_MASK, COH
+       breq    debug_restore_all
+
+1:     bld     r1, TIF_NEED_RESCHED
+       brcc    2f
+       unmask_interrupts
+       rcall   schedule
+       mask_interrupts
+       ld.w    r1, r0[TI_flags]
+       rjmp    1b
+
+2:     mov     r2, _TIF_SIGPENDING | _TIF_RESTORE_SIGMASK
+       tst     r1, r2
+       breq    3f
+       unmask_interrupts
+       mov     r12, sp
+       mov     r11, r0
+       rcall   do_notify_resume
+       mask_interrupts
+       ld.w    r1, r0[TI_flags]
+       rjmp    1b
+
+3:     bld     r1, TIF_SINGLE_STEP
+       brcc    debug_restore_all
+       mfdr    r2, DBGREG_DC
+       sbr     r2, DC_SS_BIT
+       mtdr    DBGREG_DC, r2
+       rjmp    debug_restore_all
+
+       .set    rsr_int0,       SYSREG_RSR_INT0
+       .set    rsr_int1,       SYSREG_RSR_INT1
+       .set    rsr_int2,       SYSREG_RSR_INT2
+       .set    rsr_int3,       SYSREG_RSR_INT3
+       .set    rar_int0,       SYSREG_RAR_INT0
+       .set    rar_int1,       SYSREG_RAR_INT1
+       .set    rar_int2,       SYSREG_RAR_INT2
+       .set    rar_int3,       SYSREG_RAR_INT3
+
+       .macro  IRQ_LEVEL level
+       .type   irq_level\level, @function
+irq_level\level:
+       sub     sp, 4           /* r12_orig */
+       stmts   --sp,r0-lr
+       mfsr    r8, rar_int\level
+       mfsr    r9, rsr_int\level
+       pushm   r8-r9
+
+       mov     r11, sp
+       mov     r12, \level
+
+       rcall   do_IRQ
+
+       lddsp   r4, sp[REG_SR]
+       andh    r4, (MODE_MASK >> 16), COH
+#ifdef CONFIG_PREEMPT
+       brne    2f
+#else
+       brne    1f
+#endif
+
+       get_thread_info r0
+       ld.w    r1, r0[TI_flags]
+       andl    r1, _TIF_WORK_MASK, COH
+       brne    irq_exit_work
+
+1:     popm    r8-r9
+       mtsr    rar_int\level, r8
+       mtsr    rsr_int\level, r9
+       ldmts   sp++,r0-lr
+       sub     sp, -4          /* ignore r12_orig */
+       rete
+
+#ifdef CONFIG_PREEMPT
+2:
+       get_thread_info r0
+       ld.w    r2, r0[TI_preempt_count]
+       cp.w    r2, 0
+       brne    1b
+       ld.w    r1, r0[TI_flags]
+       bld     r1, TIF_NEED_RESCHED
+       brcc    1b
+       lddsp   r4, sp[REG_SR]
+       bld     r4, SYSREG_GM_OFFSET
+       brcs    1b
+       rcall   preempt_schedule_irq
+       rjmp    1b
+#endif
+       .endm
+
+       .section .irq.text,"ax",@progbits
+
+       .global irq_level0
+       .global irq_level1
+       .global irq_level2
+       .global irq_level3
+       IRQ_LEVEL 0
+       IRQ_LEVEL 1
+       IRQ_LEVEL 2
+       IRQ_LEVEL 3
diff --git a/arch/avr32/kernel/head.S b/arch/avr32/kernel/head.S
new file mode 100644 (file)
index 0000000..773b7ad
--- /dev/null
@@ -0,0 +1,45 @@
+/*
+ * Non-board-specific low-level startup code
+ *
+ * Copyright (C) 2004-2006 Atmel Corporation
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ */
+#include <linux/linkage.h>
+
+#include <asm/page.h>
+#include <asm/thread_info.h>
+#include <asm/sysreg.h>
+
+       .section .init.text,"ax"
+       .global kernel_entry
+kernel_entry:
+       /* Initialize status register */
+       lddpc   r0, init_sr
+       mtsr    SYSREG_SR, r0
+
+       /* Set initial stack pointer */
+       lddpc   sp, stack_addr
+       sub     sp, -THREAD_SIZE
+
+#ifdef CONFIG_FRAME_POINTER
+       /* Mark last stack frame */
+       mov     lr, 0
+       mov     r7, 0
+#endif
+
+       /* Set up the PIO, SDRAM controller, early printk, etc. */
+       rcall   board_early_init
+
+       /* Start the show */
+       lddpc   pc, kernel_start_addr
+
+       .align  2
+init_sr:
+       .long   0x007f0000      /* Supervisor mode, everything masked */
+stack_addr:
+       .long   init_thread_union
+kernel_start_addr:
+       .long   start_kernel
diff --git a/arch/avr32/kernel/init_task.c b/arch/avr32/kernel/init_task.c
new file mode 100644 (file)
index 0000000..effcacf
--- /dev/null
@@ -0,0 +1,38 @@
+/*
+ * Copyright (C) 2004-2006 Atmel Corporation
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ */
+#include <linux/module.h>
+#include <linux/fs.h>
+#include <linux/sched.h>
+#include <linux/init_task.h>
+#include <linux/mqueue.h>
+
+#include <asm/pgtable.h>
+
+static struct fs_struct init_fs = INIT_FS;
+static struct files_struct init_files = INIT_FILES;
+static struct signal_struct init_signals = INIT_SIGNALS(init_signals);
+static struct sighand_struct init_sighand = INIT_SIGHAND(init_sighand);
+struct mm_struct init_mm = INIT_MM(init_mm);
+
+EXPORT_SYMBOL(init_mm);
+
+/*
+ * Initial thread structure. Must be aligned on an 8192-byte boundary.
+ */
+union thread_union init_thread_union
+       __attribute__((__section__(".data.init_task"))) =
+               { INIT_THREAD_INFO(init_task) };
+
+/*
+ * Initial task structure.
+ *
+ * All other task structs will be allocated on slabs in fork.c
+ */
+struct task_struct init_task = INIT_TASK(init_task);
+
+EXPORT_SYMBOL(init_task);
diff --git a/arch/avr32/kernel/irq.c b/arch/avr32/kernel/irq.c
new file mode 100644 (file)
index 0000000..856f354
--- /dev/null
@@ -0,0 +1,71 @@
+/*
+ * Copyright (C) 2004-2006 Atmel Corporation
+ *
+ * Based on arch/i386/kernel/irq.c
+ *   Copyright (C) 1992, 1998 Linus Torvalds, Ingo Molnar
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ *
+ * This file contains the code used by various IRQ handling routines:
+ * asking for different IRQ's should be done through these routines
+ * instead of just grabbing them. Thus setups with different IRQ numbers
+ * shouldn't result in any weird surprises, and installing new handlers
+ * should be easier.
+ *
+ * IRQ's are in fact implemented a bit like signal handlers for the kernel.
+ * Naturally it's not a 1:1 relation, but there are similarities.
+ */
+
+#include <linux/interrupt.h>
+#include <linux/irq.h>
+#include <linux/kernel_stat.h>
+#include <linux/proc_fs.h>
+#include <linux/seq_file.h>
+#include <linux/sysdev.h>
+
+/*
+ * 'what should we do if we get a hw irq event on an illegal vector'.
+ * each architecture has to answer this themselves.
+ */
+void ack_bad_irq(unsigned int irq)
+{
+       printk("unexpected IRQ %u\n", irq);
+}
+
+#ifdef CONFIG_PROC_FS
+int show_interrupts(struct seq_file *p, void *v)
+{
+       int i = *(loff_t *)v, cpu;
+       struct irqaction *action;
+       unsigned long flags;
+
+       if (i == 0) {
+               seq_puts(p, "           ");
+               for_each_online_cpu(cpu)
+                       seq_printf(p, "CPU%d       ", cpu);
+               seq_putc(p, '\n');
+       }
+
+       if (i < NR_IRQS) {
+               spin_lock_irqsave(&irq_desc[i].lock, flags);
+               action = irq_desc[i].action;
+               if (!action)
+                       goto unlock;
+
+               seq_printf(p, "%3d: ", i);
+               for_each_online_cpu(cpu)
+                       seq_printf(p, "%10u ", kstat_cpu(cpu).irqs[i]);
+               seq_printf(p, "  %s", action->name);
+               for (action = action->next; action; action = action->next)
+                       seq_printf(p, ", %s", action->name);
+
+               seq_putc(p, '\n');
+       unlock:
+               spin_unlock_irqrestore(&irq_desc[i].lock, flags);
+       }
+
+       return 0;
+}
+#endif
diff --git a/arch/avr32/kernel/kprobes.c b/arch/avr32/kernel/kprobes.c
new file mode 100644 (file)
index 0000000..6caf9e8
--- /dev/null
@@ -0,0 +1,270 @@
+/*
+ *  Kernel Probes (KProbes)
+ *
+ * Copyright (C) 2005-2006 Atmel Corporation
+ *
+ * Based on arch/ppc64/kernel/kprobes.c
+ *  Copyright (C) IBM Corporation, 2002, 2004
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ */
+
+#include <linux/kprobes.h>
+#include <linux/ptrace.h>
+
+#include <asm/cacheflush.h>
+#include <asm/kdebug.h>
+#include <asm/ocd.h>
+
+DEFINE_PER_CPU(struct kprobe *, current_kprobe);
+static unsigned long kprobe_status;
+static struct pt_regs jprobe_saved_regs;
+
+int __kprobes arch_prepare_kprobe(struct kprobe *p)
+{
+       int ret = 0;
+
+       if ((unsigned long)p->addr & 0x01) {
+               printk("Attempt to register kprobe at an unaligned address\n");
+               ret = -EINVAL;
+       }
+
+       /* XXX: Might be a good idea to check if p->addr is a valid
+        * kernel address as well... */
+
+       if (!ret) {
+               pr_debug("copy kprobe at %p\n", p->addr);
+               memcpy(p->ainsn.insn, p->addr, MAX_INSN_SIZE * sizeof(kprobe_opcode_t));
+               p->opcode = *p->addr;
+       }
+
+       return ret;
+}
+
+void __kprobes arch_arm_kprobe(struct kprobe *p)
+{
+       pr_debug("arming kprobe at %p\n", p->addr);
+       *p->addr = BREAKPOINT_INSTRUCTION;
+       flush_icache_range((unsigned long)p->addr,
+                          (unsigned long)p->addr + sizeof(kprobe_opcode_t));
+}
+
+void __kprobes arch_disarm_kprobe(struct kprobe *p)
+{
+       pr_debug("disarming kprobe at %p\n", p->addr);
+       *p->addr = p->opcode;
+       flush_icache_range((unsigned long)p->addr,
+                          (unsigned long)p->addr + sizeof(kprobe_opcode_t));
+}
+
+static void __kprobes prepare_singlestep(struct kprobe *p, struct pt_regs *regs)
+{
+       unsigned long dc;
+
+       pr_debug("preparing to singlestep over %p (PC=%08lx)\n",
+                p->addr, regs->pc);
+
+       BUG_ON(!(sysreg_read(SR) & SYSREG_BIT(SR_D)));
+
+       dc = __mfdr(DBGREG_DC);
+       dc |= DC_SS;
+       __mtdr(DBGREG_DC, dc);
+
+       /*
+        * We must run the instruction from its original location
+        * since it may actually reference PC.
+        *
+        * TODO: Do the instruction replacement directly in icache.
+        */
+       *p->addr = p->opcode;
+       flush_icache_range((unsigned long)p->addr,
+                          (unsigned long)p->addr + sizeof(kprobe_opcode_t));
+}
+
+static void __kprobes resume_execution(struct kprobe *p, struct pt_regs *regs)
+{
+       unsigned long dc;
+
+       pr_debug("resuming execution at PC=%08lx\n", regs->pc);
+
+       dc = __mfdr(DBGREG_DC);
+       dc &= ~DC_SS;
+       __mtdr(DBGREG_DC, dc);
+
+       *p->addr = BREAKPOINT_INSTRUCTION;
+       flush_icache_range((unsigned long)p->addr,
+                          (unsigned long)p->addr + sizeof(kprobe_opcode_t));
+}
+
+static void __kprobes set_current_kprobe(struct kprobe *p)
+{
+       __get_cpu_var(current_kprobe) = p;
+}
+
+static int __kprobes kprobe_handler(struct pt_regs *regs)
+{
+       struct kprobe *p;
+       void *addr = (void *)regs->pc;
+       int ret = 0;
+
+       pr_debug("kprobe_handler: kprobe_running=%d\n",
+                kprobe_running());
+
+       /*
+        * We don't want to be preempted for the entire
+        * duration of kprobe processing
+        */
+       preempt_disable();
+
+       /* Check that we're not recursing */
+       if (kprobe_running()) {
+               p = get_kprobe(addr);
+               if (p) {
+                       if (kprobe_status == KPROBE_HIT_SS) {
+                               printk("FIXME: kprobe hit while single-stepping!\n");
+                               goto no_kprobe;
+                       }
+
+                       printk("FIXME: kprobe hit while handling another kprobe\n");
+                       goto no_kprobe;
+               } else {
+                       p = kprobe_running();
+                       if (p->break_handler && p->break_handler(p, regs))
+                               goto ss_probe;
+               }
+               /* If it's not ours, can't be delete race, (we hold lock). */
+               goto no_kprobe;
+       }
+
+       p = get_kprobe(addr);
+       if (!p)
+               goto no_kprobe;
+
+       kprobe_status = KPROBE_HIT_ACTIVE;
+       set_current_kprobe(p);
+       if (p->pre_handler && p->pre_handler(p, regs))
+               /* handler has already set things up, so skip ss setup */
+               return 1;
+
+ss_probe:
+       prepare_singlestep(p, regs);
+       kprobe_status = KPROBE_HIT_SS;
+       return 1;
+
+no_kprobe:
+       return ret;
+}
+
+static int __kprobes post_kprobe_handler(struct pt_regs *regs)
+{
+       struct kprobe *cur = kprobe_running();
+
+       pr_debug("post_kprobe_handler, cur=%p\n", cur);
+
+       if (!cur)
+               return 0;
+
+       if (cur->post_handler) {
+               kprobe_status = KPROBE_HIT_SSDONE;
+               cur->post_handler(cur, regs, 0);
+       }
+
+       resume_execution(cur, regs);
+       reset_current_kprobe();
+       preempt_enable_no_resched();
+
+       return 1;
+}
+
+static int __kprobes kprobe_fault_handler(struct pt_regs *regs, int trapnr)
+{
+       struct kprobe *cur = kprobe_running();
+
+       pr_debug("kprobe_fault_handler: trapnr=%d\n", trapnr);
+
+       if (cur->fault_handler && cur->fault_handler(cur, regs, trapnr))
+               return 1;
+
+       if (kprobe_status & KPROBE_HIT_SS) {
+               resume_execution(cur, regs);
+               preempt_enable_no_resched();
+       }
+       return 0;
+}
+
+/*
+ * Wrapper routine to for handling exceptions.
+ */
+int __kprobes kprobe_exceptions_notify(struct notifier_block *self,
+                                      unsigned long val, void *data)
+{
+       struct die_args *args = (struct die_args *)data;
+       int ret = NOTIFY_DONE;
+
+       pr_debug("kprobe_exceptions_notify: val=%lu, data=%p\n",
+                val, data);
+
+       switch (val) {
+       case DIE_BREAKPOINT:
+               if (kprobe_handler(args->regs))
+                       ret = NOTIFY_STOP;
+               break;
+       case DIE_SSTEP:
+               if (post_kprobe_handler(args->regs))
+                       ret = NOTIFY_STOP;
+               break;
+       case DIE_FAULT:
+               if (kprobe_running()
+                   && kprobe_fault_handler(args->regs, args->trapnr))
+                       ret = NOTIFY_STOP;
+               break;
+       default:
+               break;
+       }
+
+       return ret;
+}
+
+int __kprobes setjmp_pre_handler(struct kprobe *p, struct pt_regs *regs)
+{
+       struct jprobe *jp = container_of(p, struct jprobe, kp);
+
+       memcpy(&jprobe_saved_regs, regs, sizeof(struct pt_regs));
+
+       /*
+        * TODO: We should probably save some of the stack here as
+        * well, since gcc may pass arguments on the stack for certain
+        * functions (lots of arguments, large aggregates, varargs)
+        */
+
+       /* setup return addr to the jprobe handler routine */
+       regs->pc = (unsigned long)jp->entry;
+       return 1;
+}
+
+void __kprobes jprobe_return(void)
+{
+       asm volatile("breakpoint" ::: "memory");
+}
+
+int __kprobes longjmp_break_handler(struct kprobe *p, struct pt_regs *regs)
+{
+       /*
+        * FIXME - we should ideally be validating that we got here 'cos
+        * of the "trap" in jprobe_return() above, before restoring the
+        * saved regs...
+        */
+       memcpy(regs, &jprobe_saved_regs, sizeof(struct pt_regs));
+       return 1;
+}
+
+int __init arch_init_kprobes(void)
+{
+       printk("KPROBES: Enabling monitor mode (MM|DBE)...\n");
+       __mtdr(DBGREG_DC, DC_MM | DC_DBE);
+
+       /* TODO: Register kretprobe trampoline */
+       return 0;
+}
diff --git a/arch/avr32/kernel/module.c b/arch/avr32/kernel/module.c
new file mode 100644 (file)
index 0000000..dfc32f2
--- /dev/null
@@ -0,0 +1,324 @@
+/*
+ * AVR32-specific kernel module loader
+ *
+ * Copyright (C) 2005-2006 Atmel Corporation
+ *
+ * GOT initialization parts are based on the s390 version
+ *   Copyright (C) 2002, 2003 IBM Deutschland Entwicklung GmbH,
+ *                            IBM Corporation
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ */
+
+#include <linux/moduleloader.h>
+#include <linux/module.h>
+#include <linux/kernel.h>
+#include <linux/elf.h>
+#include <linux/vmalloc.h>
+
+void *module_alloc(unsigned long size)
+{
+       if (size == 0)
+               return NULL;
+       return vmalloc(size);
+}
+
+void module_free(struct module *mod, void *module_region)
+{
+       vfree(mod->arch.syminfo);
+       mod->arch.syminfo = NULL;
+
+       vfree(module_region);
+       /* FIXME: if module_region == mod->init_region, trim exception
+        * table entries. */
+}
+
+static inline int check_rela(Elf32_Rela *rela, struct module *module,
+                            char *strings, Elf32_Sym *symbols)
+{
+       struct mod_arch_syminfo *info;
+
+       info = module->arch.syminfo + ELF32_R_SYM(rela->r_info);
+       switch (ELF32_R_TYPE(rela->r_info)) {
+       case R_AVR32_GOT32:
+       case R_AVR32_GOT16:
+       case R_AVR32_GOT8:
+       case R_AVR32_GOT21S:
+       case R_AVR32_GOT18SW:   /* mcall */
+       case R_AVR32_GOT16S:    /* ld.w */
+               if (rela->r_addend != 0) {
+                       printk(KERN_ERR
+                              "GOT relocation against %s at offset %u with addend\n",
+                              strings + symbols[ELF32_R_SYM(rela->r_info)].st_name,
+                              rela->r_offset);
+                       return -ENOEXEC;
+               }
+               if (info->got_offset == -1UL) {
+                       info->got_offset = module->arch.got_size;
+                       module->arch.got_size += sizeof(void *);
+               }
+               pr_debug("GOT[%3lu] %s\n", info->got_offset,
+                        strings + symbols[ELF32_R_SYM(rela->r_info)].st_name);
+               break;
+       }
+
+       return 0;
+}
+
+int module_frob_arch_sections(Elf_Ehdr *hdr, Elf_Shdr *sechdrs,
+                             char *secstrings, struct module *module)
+{
+       Elf32_Shdr *symtab;
+       Elf32_Sym *symbols;
+       Elf32_Rela *rela;
+       char *strings;
+       int nrela, i, j;
+       int ret;
+
+       /* Find the symbol table */
+       symtab = NULL;
+       for (i = 0; i < hdr->e_shnum; i++)
+               switch (sechdrs[i].sh_type) {
+               case SHT_SYMTAB:
+                       symtab = &sechdrs[i];
+                       break;
+               }
+       if (!symtab) {
+               printk(KERN_ERR "module %s: no symbol table\n", module->name);
+               return -ENOEXEC;
+       }
+
+       /* Allocate room for one syminfo structure per symbol. */
+       module->arch.nsyms = symtab->sh_size / sizeof(Elf_Sym);
+       module->arch.syminfo = vmalloc(module->arch.nsyms
+                                  * sizeof(struct mod_arch_syminfo));
+       if (!module->arch.syminfo)
+               return -ENOMEM;
+
+       symbols = (void *)hdr + symtab->sh_offset;
+       strings = (void *)hdr + sechdrs[symtab->sh_link].sh_offset;
+       for (i = 0; i < module->arch.nsyms; i++) {
+               if (symbols[i].st_shndx == SHN_UNDEF &&
+                   strcmp(strings + symbols[i].st_name,
+                          "_GLOBAL_OFFSET_TABLE_") == 0)
+                       /* "Define" it as absolute. */
+                       symbols[i].st_shndx = SHN_ABS;
+               module->arch.syminfo[i].got_offset = -1UL;
+               module->arch.syminfo[i].got_initialized = 0;
+       }
+
+       /* Allocate GOT entries for symbols that need it. */
+       module->arch.got_size = 0;
+       for (i = 0; i < hdr->e_shnum; i++) {
+               if (sechdrs[i].sh_type != SHT_RELA)
+                       continue;
+               nrela = sechdrs[i].sh_size / sizeof(Elf32_Rela);
+               rela = (void *)hdr + sechdrs[i].sh_offset;
+               for (j = 0; j < nrela; j++) {
+                       ret = check_rela(rela + j, module,
+                                        strings, symbols);
+                       if (ret)
+                               goto out_free_syminfo;
+               }
+       }
+
+       /*
+        * Increase core size to make room for GOT and set start
+        * offset for GOT.
+        */
+       module->core_size = ALIGN(module->core_size, 4);
+       module->arch.got_offset = module->core_size;
+       module->core_size += module->arch.got_size;
+
+       return 0;
+
+out_free_syminfo:
+       vfree(module->arch.syminfo);
+       module->arch.syminfo = NULL;
+
+       return ret;
+}
+
+static inline int reloc_overflow(struct module *module, const char *reloc_name,
+                                Elf32_Addr relocation)
+{
+       printk(KERN_ERR "module %s: Value %lx does not fit relocation %s\n",
+              module->name, (unsigned long)relocation, reloc_name);
+       return -ENOEXEC;
+}
+
+#define get_u16(loc)           (*((uint16_t *)loc))
+#define put_u16(loc, val)      (*((uint16_t *)loc) = (val))
+
+int apply_relocate_add(Elf32_Shdr *sechdrs, const char *strtab,
+                      unsigned int symindex, unsigned int relindex,
+                      struct module *module)
+{
+       Elf32_Shdr *symsec = sechdrs + symindex;
+       Elf32_Shdr *relsec = sechdrs + relindex;
+       Elf32_Shdr *dstsec = sechdrs + relsec->sh_info;
+       Elf32_Rela *rel = (void *)relsec->sh_addr;
+       unsigned int i;
+       int ret = 0;
+
+       for (i = 0; i < relsec->sh_size / sizeof(Elf32_Rela); i++, rel++) {
+               struct mod_arch_syminfo *info;
+               Elf32_Sym *sym;
+               Elf32_Addr relocation;
+               uint32_t *location;
+               uint32_t value;
+
+               location = (void *)dstsec->sh_addr + rel->r_offset;
+               sym = (Elf32_Sym *)symsec->sh_addr + ELF32_R_SYM(rel->r_info);
+               relocation = sym->st_value + rel->r_addend;
+
+               info = module->arch.syminfo + ELF32_R_SYM(rel->r_info);
+
+               /* Initialize GOT entry if necessary */
+               switch (ELF32_R_TYPE(rel->r_info)) {
+               case R_AVR32_GOT32:
+               case R_AVR32_GOT16:
+               case R_AVR32_GOT8:
+               case R_AVR32_GOT21S:
+               case R_AVR32_GOT18SW:
+               case R_AVR32_GOT16S:
+                       if (!info->got_initialized) {
+                               Elf32_Addr *gotent;
+
+                               gotent = (module->module_core
+                                         + module->arch.got_offset
+                                         + info->got_offset);
+                               *gotent = relocation;
+                               info->got_initialized = 1;
+                       }
+
+                       relocation = info->got_offset;
+                       break;
+               }
+
+               switch (ELF32_R_TYPE(rel->r_info)) {
+               case R_AVR32_32:
+               case R_AVR32_32_CPENT:
+                       *location = relocation;
+                       break;
+               case R_AVR32_22H_PCREL:
+                       relocation -= (Elf32_Addr)location;
+                       if ((relocation & 0xffe00001) != 0
+                           && (relocation & 0xffc00001) != 0xffc00000)
+                               return reloc_overflow(module,
+                                                     "R_AVR32_22H_PCREL",
+                                                     relocation);
+                       relocation >>= 1;
+
+                       value = *location;
+                       value = ((value & 0xe1ef0000)
+                                | (relocation & 0xffff)
+                                | ((relocation & 0x10000) << 4)
+                                | ((relocation & 0x1e0000) << 8));
+                       *location = value;
+                       break;
+               case R_AVR32_11H_PCREL:
+                       relocation -= (Elf32_Addr)location;
+                       if ((relocation & 0xfffffc01) != 0
+                           && (relocation & 0xfffff801) != 0xfffff800)
+                               return reloc_overflow(module,
+                                                     "R_AVR32_11H_PCREL",
+                                                     relocation);
+                       value = get_u16(location);
+                       value = ((value & 0xf00c)
+                                | ((relocation & 0x1fe) << 3)
+                                | ((relocation & 0x600) >> 9));
+                       put_u16(location, value);
+                       break;
+               case R_AVR32_9H_PCREL:
+                       relocation -= (Elf32_Addr)location;
+                       if ((relocation & 0xffffff01) != 0
+                           && (relocation & 0xfffffe01) != 0xfffffe00)
+                               return reloc_overflow(module,
+                                                     "R_AVR32_9H_PCREL",
+                                                     relocation);
+                       value = get_u16(location);
+                       value = ((value & 0xf00f)
+                                | ((relocation & 0x1fe) << 3));
+                       put_u16(location, value);
+                       break;
+               case R_AVR32_9UW_PCREL:
+                       relocation -= ((Elf32_Addr)location) & 0xfffffffc;
+                       if ((relocation & 0xfffffc03) != 0)
+                               return reloc_overflow(module,
+                                                     "R_AVR32_9UW_PCREL",
+                                                     relocation);
+                       value = get_u16(location);
+                       value = ((value & 0xf80f)
+                                | ((relocation & 0x1fc) << 2));
+                       put_u16(location, value);
+                       break;
+               case R_AVR32_GOTPC:
+                       /*
+                        * R6 = PC - (PC - GOT)
+                        *
+                        * At this point, relocation contains the
+                        * value of PC.  Just subtract the value of
+                        * GOT, and we're done.
+                        */
+                       pr_debug("GOTPC: PC=0x%lx, got_offset=0x%lx, core=0x%p\n",
+                                relocation, module->arch.got_offset,
+                                module->module_core);
+                       relocation -= ((unsigned long)module->module_core
+                                      + module->arch.got_offset);
+                       *location = relocation;
+                       break;
+               case R_AVR32_GOT18SW:
+                       if ((relocation & 0xfffe0003) != 0
+                           && (relocation & 0xfffc0003) != 0xffff0000)
+                               return reloc_overflow(module, "R_AVR32_GOT18SW",
+                                                    relocation);
+                       relocation >>= 2;
+                       /* fall through */
+               case R_AVR32_GOT16S:
+                       if ((relocation & 0xffff8000) != 0
+                           && (relocation & 0xffff0000) != 0xffff0000)
+                               return reloc_overflow(module, "R_AVR32_GOT16S",
+                                                     relocation);
+                       pr_debug("GOT reloc @ 0x%lx -> %lu\n",
+                                rel->r_offset, relocation);
+                       value = *location;
+                       value = ((value & 0xffff0000)
+                                | (relocation & 0xffff));
+                       *location = value;
+                       break;
+
+               default:
+                       printk(KERN_ERR "module %s: Unknown relocation: %u\n",
+                              module->name, ELF32_R_TYPE(rel->r_info));
+                       return -ENOEXEC;
+               }
+       }
+
+       return ret;
+}
+
+int apply_relocate(Elf32_Shdr *sechdrs, const char *strtab,
+                  unsigned int symindex, unsigned int relindex,
+                  struct module *module)
+{
+       printk(KERN_ERR "module %s: REL relocations are not supported\n",
+               module->name);
+       return -ENOEXEC;
+}
+
+int module_finalize(const Elf_Ehdr *hdr, const Elf_Shdr *sechdrs,
+                   struct module *module)
+{
+       vfree(module->arch.syminfo);
+       module->arch.syminfo = NULL;
+
+       return 0;
+}
+
+void module_arch_cleanup(struct module *module)
+{
+
+}
diff --git a/arch/avr32/kernel/process.c b/arch/avr32/kernel/process.c
new file mode 100644 (file)
index 0000000..317dc50
--- /dev/null
@@ -0,0 +1,276 @@
+/*
+ * Copyright (C) 2004-2006 Atmel Corporation
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ */
+#include <linux/sched.h>
+#include <linux/module.h>
+#include <linux/kallsyms.h>
+#include <linux/fs.h>
+#include <linux/ptrace.h>
+#include <linux/reboot.h>
+#include <linux/unistd.h>
+
+#include <asm/sysreg.h>
+#include <asm/ocd.h>
+
+void (*pm_power_off)(void) = NULL;
+EXPORT_SYMBOL(pm_power_off);
+
+/*
+ * This file handles the architecture-dependent parts of process handling..
+ */
+
+void cpu_idle(void)
+{
+       /* endless idle loop with no priority at all */
+       while (1) {
+               /* TODO: Enter sleep mode */
+               while (!need_resched())
+                       cpu_relax();
+               preempt_enable_no_resched();
+               schedule();
+               preempt_disable();
+       }
+}
+
+void machine_halt(void)
+{
+}
+
+void machine_power_off(void)
+{
+}
+
+void machine_restart(char *cmd)
+{
+       __mtdr(DBGREG_DC, DC_DBE);
+       __mtdr(DBGREG_DC, DC_RES);
+       while (1) ;
+}
+
+/*
+ * PC is actually discarded when returning from a system call -- the
+ * return address must be stored in LR. This function will make sure
+ * LR points to do_exit before starting the thread.
+ *
+ * Also, when returning from fork(), r12 is 0, so we must copy the
+ * argument as well.
+ *
+ *  r0 : The argument to the main thread function
+ *  r1 : The address of do_exit
+ *  r2 : The address of the main thread function
+ */
+asmlinkage extern void kernel_thread_helper(void);
+__asm__("      .type   kernel_thread_helper, @function\n"
+       "kernel_thread_helper:\n"
+       "       mov     r12, r0\n"
+       "       mov     lr, r2\n"
+       "       mov     pc, r1\n"
+       "       .size   kernel_thread_helper, . - kernel_thread_helper");
+
+int kernel_thread(int (*fn)(void *), void *arg, unsigned long flags)
+{
+       struct pt_regs regs;
+
+       memset(&regs, 0, sizeof(regs));
+
+       regs.r0 = (unsigned long)arg;
+       regs.r1 = (unsigned long)fn;
+       regs.r2 = (unsigned long)do_exit;
+       regs.lr = (unsigned long)kernel_thread_helper;
+       regs.pc = (unsigned long)kernel_thread_helper;
+       regs.sr = MODE_SUPERVISOR;
+
+       return do_fork(flags | CLONE_VM | CLONE_UNTRACED,
+                      0, &regs, 0, NULL, NULL);
+}
+EXPORT_SYMBOL(kernel_thread);
+
+/*
+ * Free current thread data structures etc
+ */
+void exit_thread(void)
+{
+       /* nothing to do */
+}
+
+void flush_thread(void)
+{
+       /* nothing to do */
+}
+
+void release_thread(struct task_struct *dead_task)
+{
+       /* do nothing */
+}
+
+static const char *cpu_modes[] = {
+       "Application", "Supervisor", "Interrupt level 0", "Interrupt level 1",
+       "Interrupt level 2", "Interrupt level 3", "Exception", "NMI"
+};
+
+void show_regs(struct pt_regs *regs)
+{
+       unsigned long sp = regs->sp;
+       unsigned long lr = regs->lr;
+       unsigned long mode = (regs->sr & MODE_MASK) >> MODE_SHIFT;
+
+       if (!user_mode(regs))
+               sp = (unsigned long)regs + FRAME_SIZE_FULL;
+
+       print_symbol("PC is at %s\n", instruction_pointer(regs));
+       print_symbol("LR is at %s\n", lr);
+       printk("pc : [<%08lx>]    lr : [<%08lx>]    %s\n"
+              "sp : %08lx  r12: %08lx  r11: %08lx\n",
+              instruction_pointer(regs),
+              lr, print_tainted(), sp, regs->r12, regs->r11);
+       printk("r10: %08lx  r9 : %08lx  r8 : %08lx\n",
+              regs->r10, regs->r9, regs->r8);
+       printk("r7 : %08lx  r6 : %08lx  r5 : %08lx  r4 : %08lx\n",
+              regs->r7, regs->r6, regs->r5, regs->r4);
+       printk("r3 : %08lx  r2 : %08lx  r1 : %08lx  r0 : %08lx\n",
+              regs->r3, regs->r2, regs->r1, regs->r0);
+       printk("Flags: %c%c%c%c%c\n",
+              regs->sr & SR_Q ? 'Q' : 'q',
+              regs->sr & SR_V ? 'V' : 'v',
+              regs->sr & SR_N ? 'N' : 'n',
+              regs->sr & SR_Z ? 'Z' : 'z',
+              regs->sr & SR_C ? 'C' : 'c');
+       printk("Mode bits: %c%c%c%c%c%c%c%c%c\n",
+              regs->sr & SR_H ? 'H' : 'h',
+              regs->sr & SR_R ? 'R' : 'r',
+              regs->sr & SR_J ? 'J' : 'j',
+              regs->sr & SR_EM ? 'E' : 'e',
+              regs->sr & SR_I3M ? '3' : '.',
+              regs->sr & SR_I2M ? '2' : '.',
+              regs->sr & SR_I1M ? '1' : '.',
+              regs->sr & SR_I0M ? '0' : '.',
+              regs->sr & SR_GM ? 'G' : 'g');
+       printk("CPU Mode: %s\n", cpu_modes[mode]);
+
+       show_trace(NULL, (unsigned long *)sp, regs);
+}
+EXPORT_SYMBOL(show_regs);
+
+/* Fill in the fpu structure for a core dump. This is easy -- we don't have any */
+int dump_fpu(struct pt_regs *regs, elf_fpregset_t *fpu)
+{
+       /* Not valid */
+       return 0;
+}
+
+asmlinkage void ret_from_fork(void);
+
+int copy_thread(int nr, unsigned long clone_flags, unsigned long usp,
+               unsigned long unused,
+               struct task_struct *p, struct pt_regs *regs)
+{
+       struct pt_regs *childregs;
+
+       childregs = ((struct pt_regs *)(THREAD_SIZE + (unsigned long)p->thread_info)) - 1;
+       *childregs = *regs;
+
+       if (user_mode(regs))
+               childregs->sp = usp;
+       else
+               childregs->sp = (unsigned long)p->thread_info + THREAD_SIZE;
+
+       childregs->r12 = 0; /* Set return value for child */
+
+       p->thread.cpu_context.sr = MODE_SUPERVISOR | SR_GM;
+       p->thread.cpu_context.ksp = (unsigned long)childregs;
+       p->thread.cpu_context.pc = (unsigned long)ret_from_fork;
+
+       return 0;
+}
+
+/* r12-r8 are dummy parameters to force the compiler to use the stack */
+asmlinkage int sys_fork(struct pt_regs *regs)
+{
+       return do_fork(SIGCHLD, regs->sp, regs, 0, NULL, NULL);
+}
+
+asmlinkage int sys_clone(unsigned long clone_flags, unsigned long newsp,
+                        unsigned long parent_tidptr,
+                        unsigned long child_tidptr, struct pt_regs *regs)
+{
+       if (!newsp)
+               newsp = regs->sp;
+       return do_fork(clone_flags, newsp, regs, 0,
+                      (int __user *)parent_tidptr,
+                      (int __user *)child_tidptr);
+}
+
+asmlinkage int sys_vfork(struct pt_regs *regs)
+{
+       return do_fork(CLONE_VFORK | CLONE_VM | SIGCHLD, regs->sp, regs,
+                      0, NULL, NULL);
+}
+
+asmlinkage int sys_execve(char __user *ufilename, char __user *__user *uargv,
+                         char __user *__user *uenvp, struct pt_regs *regs)
+{
+       int error;
+       char *filename;
+
+       filename = getname(ufilename);
+       error = PTR_ERR(filename);
+       if (IS_ERR(filename))
+               goto out;
+
+       error = do_execve(filename, uargv, uenvp, regs);
+       if (error == 0)
+               current->ptrace &= ~PT_DTRACE;
+       putname(filename);
+
+out:
+       return error;
+}
+
+
+/*
+ * This function is supposed to answer the question "who called
+ * schedule()?"
+ */
+unsigned long get_wchan(struct task_struct *p)
+{
+       unsigned long pc;
+       unsigned long stack_page;
+
+       if (!p || p == current || p->state == TASK_RUNNING)
+               return 0;
+
+       stack_page = (unsigned long)p->thread_info;
+       BUG_ON(!stack_page);
+
+       /*
+        * The stored value of PC is either the address right after
+        * the call to __switch_to() or ret_from_fork.
+        */
+       pc = thread_saved_pc(p);
+       if (in_sched_functions(pc)) {
+#ifdef CONFIG_FRAME_POINTER
+               unsigned long fp = p->thread.cpu_context.r7;
+               BUG_ON(fp < stack_page || fp > (THREAD_SIZE + stack_page));
+               pc = *(unsigned long *)fp;
+#else
+               /*
+                * We depend on the frame size of schedule here, which
+                * is actually quite ugly. It might be possible to
+                * determine the frame size automatically at build
+                * time by doing this:
+                *   - compile sched.c
+                *   - disassemble the resulting sched.o
+                *   - look for 'sub sp,??' shortly after '<schedule>:'
+                */
+               unsigned long sp = p->thread.cpu_context.ksp + 16;
+               BUG_ON(sp < stack_page || sp > (THREAD_SIZE + stack_page));
+               pc = *(unsigned long *)sp;
+#endif
+       }
+
+       return pc;
+}
diff --git a/arch/avr32/kernel/ptrace.c b/arch/avr32/kernel/ptrace.c
new file mode 100644 (file)
index 0000000..3c89e59
--- /dev/null
@@ -0,0 +1,371 @@
+/*
+ * Copyright (C) 2004-2006 Atmel Corporation
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ */
+#undef DEBUG
+#include <linux/kernel.h>
+#include <linux/sched.h>
+#include <linux/mm.h>
+#include <linux/smp_lock.h>
+#include <linux/ptrace.h>
+#include <linux/errno.h>
+#include <linux/user.h>
+#include <linux/security.h>
+#include <linux/unistd.h>
+#include <linux/notifier.h>
+
+#include <asm/traps.h>
+#include <asm/uaccess.h>
+#include <asm/ocd.h>
+#include <asm/mmu_context.h>
+#include <asm/kdebug.h>
+
+static struct pt_regs *get_user_regs(struct task_struct *tsk)
+{
+       return (struct pt_regs *)((unsigned long) tsk->thread_info +
+                                 THREAD_SIZE - sizeof(struct pt_regs));
+}
+
+static void ptrace_single_step(struct task_struct *tsk)
+{
+       pr_debug("ptrace_single_step: pid=%u, SR=0x%08lx\n",
+                tsk->pid, tsk->thread.cpu_context.sr);
+       if (!(tsk->thread.cpu_context.sr & SR_D)) {
+               /*
+                * Set a breakpoint at the current pc to force the
+                * process into debug mode.  The syscall/exception
+                * exit code will set a breakpoint at the return
+                * address when this flag is set.
+                */
+               pr_debug("ptrace_single_step: Setting TIF_BREAKPOINT\n");
+               set_tsk_thread_flag(tsk, TIF_BREAKPOINT);
+       }
+
+       /* The monitor code will do the actual step for us */
+       set_tsk_thread_flag(tsk, TIF_SINGLE_STEP);
+}
+
+/*
+ * Called by kernel/ptrace.c when detaching
+ *
+ * Make sure any single step bits, etc. are not set
+ */
+void ptrace_disable(struct task_struct *child)
+{
+       clear_tsk_thread_flag(child, TIF_SINGLE_STEP);
+}
+
+/*
+ * Handle hitting a breakpoint
+ */
+static void ptrace_break(struct task_struct *tsk, struct pt_regs *regs)
+{
+       siginfo_t info;
+
+       info.si_signo = SIGTRAP;
+       info.si_errno = 0;
+       info.si_code  = TRAP_BRKPT;
+       info.si_addr  = (void __user *)instruction_pointer(regs);
+
+       pr_debug("ptrace_break: Sending SIGTRAP to PID %u (pc = 0x%p)\n",
+                tsk->pid, info.si_addr);
+       force_sig_info(SIGTRAP, &info, tsk);
+}
+
+/*
+ * Read the word at offset "offset" into the task's "struct user". We
+ * actually access the pt_regs struct stored on the kernel stack.
+ */
+static int ptrace_read_user(struct task_struct *tsk, unsigned long offset,
+                           unsigned long __user *data)
+{
+       unsigned long *regs;
+       unsigned long value;
+
+       pr_debug("ptrace_read_user(%p, %#lx, %p)\n",
+                tsk, offset, data);
+
+       if (offset & 3 || offset >= sizeof(struct user)) {
+               printk("ptrace_read_user: invalid offset 0x%08lx\n", offset);
+               return -EIO;
+       }
+
+       regs = (unsigned long *)get_user_regs(tsk);
+
+       value = 0;
+       if (offset < sizeof(struct pt_regs))
+               value = regs[offset / sizeof(regs[0])];
+
+       return put_user(value, data);
+}
+
+/*
+ * Write the word "value" to offset "offset" into the task's "struct
+ * user". We actually access the pt_regs struct stored on the kernel
+ * stack.
+ */
+static int ptrace_write_user(struct task_struct *tsk, unsigned long offset,
+                            unsigned long value)
+{
+       unsigned long *regs;
+
+       if (offset & 3 || offset >= sizeof(struct user)) {
+               printk("ptrace_write_user: invalid offset 0x%08lx\n", offset);
+               return -EIO;
+       }
+
+       if (offset >= sizeof(struct pt_regs))
+               return 0;
+
+       regs = (unsigned long *)get_user_regs(tsk);
+       regs[offset / sizeof(regs[0])] = value;
+
+       return 0;
+}
+
+static int ptrace_getregs(struct task_struct *tsk, void __user *uregs)
+{
+       struct pt_regs *regs = get_user_regs(tsk);
+
+       return copy_to_user(uregs, regs, sizeof(*regs)) ? -EFAULT : 0;
+}
+
+static int ptrace_setregs(struct task_struct *tsk, const void __user *uregs)
+{
+       struct pt_regs newregs;
+       int ret;
+
+       ret = -EFAULT;
+       if (copy_from_user(&newregs, uregs, sizeof(newregs)) == 0) {
+               struct pt_regs *regs = get_user_regs(tsk);
+
+               ret = -EINVAL;
+               if (valid_user_regs(&newregs)) {
+                       *regs = newregs;
+                       ret = 0;
+               }
+       }
+
+       return ret;
+}
+
+long arch_ptrace(struct task_struct *child, long request, long addr, long data)
+{
+       unsigned long tmp;
+       int ret;
+
+       pr_debug("arch_ptrace(%ld, %ld, %#lx, %#lx)\n",
+                request, child->pid, addr, data);
+
+       pr_debug("ptrace: Enabling monitor mode...\n");
+       __mtdr(DBGREG_DC, __mfdr(DBGREG_DC) | DC_MM | DC_DBE);
+
+       switch (request) {
+       /* Read the word at location addr in the child process */
+       case PTRACE_PEEKTEXT:
+       case PTRACE_PEEKDATA:
+               ret = access_process_vm(child, addr, &tmp, sizeof(tmp), 0);
+               if (ret == sizeof(tmp))
+                       ret = put_user(tmp, (unsigned long __user *)data);
+               else
+                       ret = -EIO;
+               break;
+
+       case PTRACE_PEEKUSR:
+               ret = ptrace_read_user(child, addr,
+                                      (unsigned long __user *)data);
+               break;
+
+       /* Write the word in data at location addr */
+       case PTRACE_POKETEXT:
+       case PTRACE_POKEDATA:
+               ret = access_process_vm(child, addr, &data, sizeof(data), 1);
+               if (ret == sizeof(data))
+                       ret = 0;
+               else
+                       ret = -EIO;
+               break;
+
+       case PTRACE_POKEUSR:
+               ret = ptrace_write_user(child, addr, data);
+               break;
+
+       /* continue and stop at next (return from) syscall */
+       case PTRACE_SYSCALL:
+       /* restart after signal */
+       case PTRACE_CONT:
+               ret = -EIO;
+               if (!valid_signal(data))
+                       break;
+               if (request == PTRACE_SYSCALL)
+                       set_tsk_thread_flag(child, TIF_SYSCALL_TRACE);
+               else
+                       clear_tsk_thread_flag(child, TIF_SYSCALL_TRACE);
+               child->exit_code = data;
+               /* XXX: Are we sure no breakpoints are active here? */
+               wake_up_process(child);
+               ret = 0;
+               break;
+
+       /*
+        * Make the child exit. Best I can do is send it a
+        * SIGKILL. Perhaps it should be put in the status that it
+        * wants to exit.
+        */
+       case PTRACE_KILL:
+               ret = 0;
+               if (child->exit_state == EXIT_ZOMBIE)
+                       break;
+               child->exit_code = SIGKILL;
+               wake_up_process(child);
+               break;
+
+       /*
+        * execute single instruction.
+        */
+       case PTRACE_SINGLESTEP:
+               ret = -EIO;
+               if (!valid_signal(data))
+                       break;
+               clear_tsk_thread_flag(child, TIF_SYSCALL_TRACE);
+               ptrace_single_step(child);
+               child->exit_code = data;
+               wake_up_process(child);
+               ret = 0;
+               break;
+
+       /* Detach a process that was attached */
+       case PTRACE_DETACH:
+               ret = ptrace_detach(child, data);
+               break;
+
+       case PTRACE_GETREGS:
+               ret = ptrace_getregs(child, (void __user *)data);
+               break;
+
+       case PTRACE_SETREGS:
+               ret = ptrace_setregs(child, (const void __user *)data);
+               break;
+
+       default:
+               ret = ptrace_request(child, request, addr, data);
+               break;
+       }
+
+       pr_debug("sys_ptrace returning %d (DC = 0x%08lx)\n", ret, __mfdr(DBGREG_DC));
+       return ret;
+}
+
+asmlinkage void syscall_trace(void)
+{
+       pr_debug("syscall_trace called\n");
+       if (!test_thread_flag(TIF_SYSCALL_TRACE))
+               return;
+       if (!(current->ptrace & PT_PTRACED))
+               return;
+
+       pr_debug("syscall_trace: notifying parent\n");
+       /* The 0x80 provides a way for the tracing parent to
+        * distinguish between a syscall stop and SIGTRAP delivery */
+       ptrace_notify(SIGTRAP | ((current->ptrace & PT_TRACESYSGOOD)
+                                ? 0x80 : 0));
+
+       /*
+        * this isn't the same as continuing with a signal, but it
+        * will do for normal use.  strace only continues with a
+        * signal if the stopping signal is not SIGTRAP.  -brl
+        */
+       if (current->exit_code) {
+               pr_debug("syscall_trace: sending signal %d to PID %u\n",
+                        current->exit_code, current->pid);
+               send_sig(current->exit_code, current, 1);
+               current->exit_code = 0;
+       }
+}
+
+asmlinkage void do_debug_priv(struct pt_regs *regs)
+{
+       unsigned long dc, ds;
+       unsigned long die_val;
+
+       ds = __mfdr(DBGREG_DS);
+
+       pr_debug("do_debug_priv: pc = %08lx, ds = %08lx\n", regs->pc, ds);
+
+       if (ds & DS_SSS)
+               die_val = DIE_SSTEP;
+       else
+               die_val = DIE_BREAKPOINT;
+
+       if (notify_die(die_val, regs, 0, SIGTRAP) == NOTIFY_STOP)
+               return;
+
+       if (likely(ds & DS_SSS)) {
+               extern void itlb_miss(void);
+               extern void tlb_miss_common(void);
+               struct thread_info *ti;
+
+               dc = __mfdr(DBGREG_DC);
+               dc &= ~DC_SS;
+               __mtdr(DBGREG_DC, dc);
+
+               ti = current_thread_info();
+               ti->flags |= _TIF_BREAKPOINT;
+
+               /* The TLB miss handlers don't check thread flags */
+               if ((regs->pc >= (unsigned long)&itlb_miss)
+                   && (regs->pc <= (unsigned long)&tlb_miss_common)) {
+                       __mtdr(DBGREG_BWA2A, sysreg_read(RAR_EX));
+                       __mtdr(DBGREG_BWC2A, 0x40000001 | (get_asid() << 1));
+               }
+
+               /*
+                * If we're running in supervisor mode, the breakpoint
+                * will take us where we want directly, no need to
+                * single step.
+                */
+               if ((regs->sr & MODE_MASK) != MODE_SUPERVISOR)
+                       ti->flags |= TIF_SINGLE_STEP;
+       } else {
+               panic("Unable to handle debug trap at pc = %08lx\n",
+                     regs->pc);
+       }
+}
+
+/*
+ * Handle breakpoints, single steps and other debuggy things. To keep
+ * things simple initially, we run with interrupts and exceptions
+ * disabled all the time.
+ */
+asmlinkage void do_debug(struct pt_regs *regs)
+{
+       unsigned long dc, ds;
+
+       ds = __mfdr(DBGREG_DS);
+       pr_debug("do_debug: pc = %08lx, ds = %08lx\n", regs->pc, ds);
+
+       if (test_thread_flag(TIF_BREAKPOINT)) {
+               pr_debug("TIF_BREAKPOINT set\n");
+               /* We're taking care of it */
+               clear_thread_flag(TIF_BREAKPOINT);
+               __mtdr(DBGREG_BWC2A, 0);
+       }
+
+       if (test_thread_flag(TIF_SINGLE_STEP)) {
+               pr_debug("TIF_SINGLE_STEP set, ds = 0x%08lx\n", ds);
+               if (ds & DS_SSS) {
+                       dc = __mfdr(DBGREG_DC);
+                       dc &= ~DC_SS;
+                       __mtdr(DBGREG_DC, dc);
+
+                       clear_thread_flag(TIF_SINGLE_STEP);
+                       ptrace_break(current, regs);
+               }
+       } else {
+               /* regular breakpoint */
+               ptrace_break(current, regs);
+       }
+}
diff --git a/arch/avr32/kernel/semaphore.c b/arch/avr32/kernel/semaphore.c
new file mode 100644 (file)
index 0000000..1e2705a
--- /dev/null
@@ -0,0 +1,148 @@
+/*
+ * AVR32 sempahore implementation.
+ *
+ * Copyright (C) 2004-2006 Atmel Corporation
+ *
+ * Based on linux/arch/i386/kernel/semaphore.c
+ *  Copyright (C) 1999 Linus Torvalds
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ */
+
+#include <linux/sched.h>
+#include <linux/errno.h>
+#include <linux/module.h>
+
+#include <asm/semaphore.h>
+#include <asm/atomic.h>
+
+/*
+ * Semaphores are implemented using a two-way counter:
+ * The "count" variable is decremented for each process
+ * that tries to acquire the semaphore, while the "sleeping"
+ * variable is a count of such acquires.
+ *
+ * Notably, the inline "up()" and "down()" functions can
+ * efficiently test if they need to do any extra work (up
+ * needs to do something only if count was negative before
+ * the increment operation.
+ *
+ * "sleeping" and the contention routine ordering is protected
+ * by the spinlock in the semaphore's waitqueue head.
+ *
+ * Note that these functions are only called when there is
+ * contention on the lock, and as such all this is the
+ * "non-critical" part of the whole semaphore business. The
+ * critical part is the inline stuff in <asm/semaphore.h>
+ * where we want to avoid any extra jumps and calls.
+ */
+
+/*
+ * Logic:
+ *  - only on a boundary condition do we need to care. When we go
+ *    from a negative count to a non-negative, we wake people up.
+ *  - when we go from a non-negative count to a negative do we
+ *    (a) synchronize with the "sleeper" count and (b) make sure
+ *    that we're on the wakeup list before we synchronize so that
+ *    we cannot lose wakeup events.
+ */
+
+void __up(struct semaphore *sem)
+{
+       wake_up(&sem->wait);
+}
+EXPORT_SYMBOL(__up);
+
+void __sched __down(struct semaphore *sem)
+{
+       struct task_struct *tsk = current;
+        DECLARE_WAITQUEUE(wait, tsk);
+        unsigned long flags;
+
+        tsk->state = TASK_UNINTERRUPTIBLE;
+        spin_lock_irqsave(&sem->wait.lock, flags);
+        add_wait_queue_exclusive_locked(&sem->wait, &wait);
+
+        sem->sleepers++;
+        for (;;) {
+                int sleepers = sem->sleepers;
+
+                /*
+                 * Add "everybody else" into it. They aren't
+                 * playing, because we own the spinlock in
+                 * the wait_queue_head.
+                 */
+                if (atomic_add_return(sleepers - 1, &sem->count) >= 0) {
+                        sem->sleepers = 0;
+                        break;
+                }
+                sem->sleepers = 1;      /* us - see -1 above */
+                spin_unlock_irqrestore(&sem->wait.lock, flags);
+
+                schedule();
+
+                spin_lock_irqsave(&sem->wait.lock, flags);
+                tsk->state = TASK_UNINTERRUPTIBLE;
+        }
+        remove_wait_queue_locked(&sem->wait, &wait);
+        wake_up_locked(&sem->wait);
+        spin_unlock_irqrestore(&sem->wait.lock, flags);
+        tsk->state = TASK_RUNNING;
+}
+EXPORT_SYMBOL(__down);
+
+int __sched __down_interruptible(struct semaphore *sem)
+{
+       int retval = 0;
+       struct task_struct *tsk = current;
+        DECLARE_WAITQUEUE(wait, tsk);
+        unsigned long flags;
+
+        tsk->state = TASK_INTERRUPTIBLE;
+        spin_lock_irqsave(&sem->wait.lock, flags);
+        add_wait_queue_exclusive_locked(&sem->wait, &wait);
+
+        sem->sleepers++;
+        for (;;) {
+                int sleepers = sem->sleepers;
+
+               /*
+                * With signals pending, this turns into the trylock
+                * failure case - we won't be sleeping, and we can't
+                * get the lock as it has contention. Just correct the
+                * count and exit.
+                */
+               if (signal_pending(current)) {
+                       retval = -EINTR;
+                       sem->sleepers = 0;
+                       atomic_add(sleepers, &sem->count);
+                       break;
+               }
+
+                /*
+                 * Add "everybody else" into it. They aren't
+                 * playing, because we own the spinlock in
+                 * the wait_queue_head.
+                 */
+                if (atomic_add_return(sleepers - 1, &sem->count) >= 0) {
+                        sem->sleepers = 0;
+                        break;
+                }
+                sem->sleepers = 1;      /* us - see -1 above */
+                spin_unlock_irqrestore(&sem->wait.lock, flags);
+
+                schedule();
+
+                spin_lock_irqsave(&sem->wait.lock, flags);
+                tsk->state = TASK_INTERRUPTIBLE;
+        }
+        remove_wait_queue_locked(&sem->wait, &wait);
+        wake_up_locked(&sem->wait);
+        spin_unlock_irqrestore(&sem->wait.lock, flags);
+
+        tsk->state = TASK_RUNNING;
+       return retval;
+}
+EXPORT_SYMBOL(__down_interruptible);
diff --git a/arch/avr32/kernel/setup.c b/arch/avr32/kernel/setup.c
new file mode 100644 (file)
index 0000000..5d68f3c
--- /dev/null
@@ -0,0 +1,335 @@
+/*
+ * Copyright (C) 2004-2006 Atmel Corporation
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ */
+
+#include <linux/clk.h>
+#include <linux/init.h>
+#include <linux/sched.h>
+#include <linux/console.h>
+#include <linux/ioport.h>
+#include <linux/bootmem.h>
+#include <linux/fs.h>
+#include <linux/module.h>
+#include <linux/root_dev.h>
+#include <linux/cpu.h>
+
+#include <asm/sections.h>
+#include <asm/processor.h>
+#include <asm/pgtable.h>
+#include <asm/setup.h>
+#include <asm/sysreg.h>
+
+#include <asm/arch/board.h>
+#include <asm/arch/init.h>
+
+extern int root_mountflags;
+
+/*
+ * Bootloader-provided information about physical memory
+ */
+struct tag_mem_range *mem_phys;
+struct tag_mem_range *mem_reserved;
+struct tag_mem_range *mem_ramdisk;
+
+/*
+ * Initialize loops_per_jiffy as 5000000 (500MIPS).
+ * Better make it too large than too small...
+ */
+struct avr32_cpuinfo boot_cpu_data = {
+       .loops_per_jiffy = 5000000
+};
+EXPORT_SYMBOL(boot_cpu_data);
+
+static char command_line[COMMAND_LINE_SIZE];
+
+/*
+ * Should be more than enough, but if you have a _really_ complex
+ * setup, you might need to increase the size of this...
+ */
+static struct tag_mem_range __initdata mem_range_cache[32];
+static unsigned mem_range_next_free;
+
+/*
+ * Standard memory resources
+ */
+static struct resource mem_res[] = {
+       {
+               .name   = "Kernel code",
+               .start  = 0,
+               .end    = 0,
+               .flags  = IORESOURCE_MEM
+       },
+       {
+               .name   = "Kernel data",
+               .start  = 0,
+               .end    = 0,
+               .flags  = IORESOURCE_MEM,
+       },
+};
+
+#define kernel_code    mem_res[0]
+#define kernel_data    mem_res[1]
+
+/*
+ * Early framebuffer allocation. Works as follows:
+ *   - If fbmem_size is zero, nothing will be allocated or reserved.
+ *   - If fbmem_start is zero when setup_bootmem() is called,
+ *     fbmem_size bytes will be allocated from the bootmem allocator.
+ *   - If fbmem_start is nonzero, an area of size fbmem_size will be
+ *     reserved at the physical address fbmem_start if necessary. If
+ *     the area isn't in a memory region known to the kernel, it will
+ *     be left alone.
+ *
+ * Board-specific code may use these variables to set up platform data
+ * for the framebuffer driver if fbmem_size is nonzero.
+ */
+static unsigned long __initdata fbmem_start;
+static unsigned long __initdata fbmem_size;
+
+/*
+ * "fbmem=xxx[kKmM]" allocates the specified amount of boot memory for
+ * use as framebuffer.
+ *
+ * "fbmem=xxx[kKmM]@yyy[kKmM]" defines a memory region of size xxx and
+ * starting at yyy to be reserved for use as framebuffer.
+ *
+ * The kernel won't verify that the memory region starting at yyy
+ * actually contains usable RAM.
+ */
+static int __init early_parse_fbmem(char *p)
+{
+       fbmem_size = memparse(p, &p);
+       if (*p == '@')
+               fbmem_start = memparse(p, &p);
+       return 0;
+}
+early_param("fbmem", early_parse_fbmem);
+
+static inline void __init resource_init(void)
+{
+       struct tag_mem_range *region;
+
+       kernel_code.start = __pa(init_mm.start_code);
+       kernel_code.end = __pa(init_mm.end_code - 1);
+       kernel_data.start = __pa(init_mm.end_code);
+       kernel_data.end = __pa(init_mm.brk - 1);
+
+       for (region = mem_phys; region; region = region->next) {
+               struct resource *res;
+               unsigned long phys_start, phys_end;
+
+               if (region->size == 0)
+                       continue;
+
+               phys_start = region->addr;
+               phys_end = phys_start + region->size - 1;
+
+               res = alloc_bootmem_low(sizeof(*res));
+               res->name = "System RAM";
+               res->start = phys_start;
+               res->end = phys_end;
+               res->flags = IORESOURCE_MEM | IORESOURCE_BUSY;
+
+               request_resource (&iomem_resource, res);
+
+               if (kernel_code.start >= res->start &&
+                   kernel_code.end <= res->end)
+                       request_resource (res, &kernel_code);
+               if (kernel_data.start >= res->start &&
+                   kernel_data.end <= res->end)
+                       request_resource (res, &kernel_data);
+       }
+}
+
+static int __init parse_tag_core(struct tag *tag)
+{
+       if (tag->hdr.size > 2) {
+               if ((tag->u.core.flags & 1) == 0)
+                       root_mountflags &= ~MS_RDONLY;
+               ROOT_DEV = new_decode_dev(tag->u.core.rootdev);
+       }
+       return 0;
+}
+__tagtable(ATAG_CORE, parse_tag_core);
+
+static int __init parse_tag_mem_range(struct tag *tag,
+                                     struct tag_mem_range **root)
+{
+       struct tag_mem_range *cur, **pprev;
+       struct tag_mem_range *new;
+
+       /*
+        * Ignore zero-sized entries. If we're running standalone, the
+        * SDRAM code may emit such entries if something goes
+        * wrong...
+        */
+       if (tag->u.mem_range.size == 0)
+               return 0;
+
+       /*
+        * Copy the data so the bootmem init code doesn't need to care
+        * about it.
+        */
+       if (mem_range_next_free >=
+           (sizeof(mem_range_cache) / sizeof(mem_range_cache[0])))
+               panic("Physical memory map too complex!\n");
+
+       new = &mem_range_cache[mem_range_next_free++];
+       *new = tag->u.mem_range;
+
+       pprev = root;
+       cur = *root;
+       while (cur) {
+               pprev = &cur->next;
+               cur = cur->next;
+       }
+
+       *pprev = new;
+       new->next = NULL;
+
+       return 0;
+}
+
+static int __init parse_tag_mem(struct tag *tag)
+{
+       return parse_tag_mem_range(tag, &mem_phys);
+}
+__tagtable(ATAG_MEM, parse_tag_mem);
+
+static int __init parse_tag_cmdline(struct tag *tag)
+{
+       strlcpy(saved_command_line, tag->u.cmdline.cmdline, COMMAND_LINE_SIZE);
+       return 0;
+}
+__tagtable(ATAG_CMDLINE, parse_tag_cmdline);
+
+static int __init parse_tag_rdimg(struct tag *tag)
+{
+       return parse_tag_mem_range(tag, &mem_ramdisk);
+}
+__tagtable(ATAG_RDIMG, parse_tag_rdimg);
+
+static int __init parse_tag_clock(struct tag *tag)
+{
+       /*
+        * We'll figure out the clocks by peeking at the system
+        * manager regs directly.
+        */
+       return 0;
+}
+__tagtable(ATAG_CLOCK, parse_tag_clock);
+
+static int __init parse_tag_rsvd_mem(struct tag *tag)
+{
+       return parse_tag_mem_range(tag, &mem_reserved);
+}
+__tagtable(ATAG_RSVD_MEM, parse_tag_rsvd_mem);
+
+static int __init parse_tag_ethernet(struct tag *tag)
+{
+#if 0
+       const struct platform_device *pdev;
+
+       /*
+        * We really need a bus type that supports "classes"...this
+        * will do for now (until we must handle other kinds of
+        * ethernet controllers)
+        */
+       pdev = platform_get_device("macb", tag->u.ethernet.mac_index);
+       if (pdev && pdev->dev.platform_data) {
+               struct eth_platform_data *data = pdev->dev.platform_data;
+
+               data->valid = 1;
+               data->mii_phy_addr = tag->u.ethernet.mii_phy_addr;
+               memcpy(data->hw_addr, tag->u.ethernet.hw_address,
+                      sizeof(data->hw_addr));
+       }
+#endif
+       return 0;
+}
+__tagtable(ATAG_ETHERNET, parse_tag_ethernet);
+
+/*
+ * Scan the tag table for this tag, and call its parse function. The
+ * tag table is built by the linker from all the __tagtable
+ * declarations.
+ */
+static int __init parse_tag(struct tag *tag)
+{
+       extern struct tagtable __tagtable_begin, __tagtable_end;
+       struct tagtable *t;
+
+       for (t = &__tagtable_begin; t < &__tagtable_end; t++)
+               if (tag->hdr.tag == t->tag) {
+                       t->parse(tag);
+                       break;
+               }
+
+       return t < &__tagtable_end;
+}
+
+/*
+ * Parse all tags in the list we got from the boot loader
+ */
+static void __init parse_tags(struct tag *t)
+{
+       for (; t->hdr.tag != ATAG_NONE; t = tag_next(t))
+               if (!parse_tag(t))
+                       printk(KERN_WARNING
+                              "Ignoring unrecognised tag 0x%08x\n",
+                              t->hdr.tag);
+}
+
+void __init setup_arch (char **cmdline_p)
+{
+       struct clk *cpu_clk;
+
+       parse_tags(bootloader_tags);
+
+       setup_processor();
+       setup_platform();
+
+       cpu_clk = clk_get(NULL, "cpu");
+       if (IS_ERR(cpu_clk)) {
+               printk(KERN_WARNING "Warning: Unable to get CPU clock\n");
+       } else {
+               unsigned long cpu_hz = clk_get_rate(cpu_clk);
+
+               /*
+                * Well, duh, but it's probably a good idea to
+                * increment the use count.
+                */
+               clk_enable(cpu_clk);
+
+               boot_cpu_data.clk = cpu_clk;
+               boot_cpu_data.loops_per_jiffy = cpu_hz * 4;
+               printk("CPU: Running at %lu.%03lu MHz\n",
+                      ((cpu_hz + 500) / 1000) / 1000,
+                      ((cpu_hz + 500) / 1000) % 1000);
+       }
+
+       init_mm.start_code = (unsigned long) &_text;
+       init_mm.end_code = (unsigned long) &_etext;
+       init_mm.end_data = (unsigned long) &_edata;
+       init_mm.brk = (unsigned long) &_end;
+
+       strlcpy(command_line, saved_command_line, COMMAND_LINE_SIZE);
+       *cmdline_p = command_line;
+       parse_early_param();
+
+       setup_bootmem();
+
+       board_setup_fbmem(fbmem_start, fbmem_size);
+
+#ifdef CONFIG_VT
+       conswitchp = &dummy_con;
+#endif
+
+       paging_init();
+
+       resource_init();
+}
diff --git a/arch/avr32/kernel/signal.c b/arch/avr32/kernel/signal.c
new file mode 100644 (file)
index 0000000..3309665
--- /dev/null
@@ -0,0 +1,328 @@
+/*
+ * Copyright (C) 2004-2006 Atmel Corporation
+ *
+ * Based on linux/arch/sh/kernel/signal.c
+ *  Copyright (C) 1999, 2000  Niibe Yutaka & Kaz Kojima
+ *  Copyright (C) 1991, 1992  Linus Torvalds
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ */
+
+#include <linux/sched.h>
+#include <linux/mm.h>
+#include <linux/errno.h>
+#include <linux/ptrace.h>
+#include <linux/unistd.h>
+#include <linux/suspend.h>
+
+#include <asm/uaccess.h>
+#include <asm/ucontext.h>
+
+#define _BLOCKABLE (~(sigmask(SIGKILL) | sigmask(SIGSTOP)))
+
+asmlinkage int sys_sigaltstack(const stack_t __user *uss, stack_t __user *uoss,
+                              struct pt_regs *regs)
+{
+       return do_sigaltstack(uss, uoss, regs->sp);
+}
+
+struct rt_sigframe
+{
+       struct siginfo info;
+       struct ucontext uc;
+       unsigned long retcode;
+};
+
+static int
+restore_sigcontext(struct pt_regs *regs, struct sigcontext __user *sc)
+{
+       int err = 0;
+
+#define COPY(x)                err |= __get_user(regs->x, &sc->x)
+       COPY(sr);
+       COPY(pc);
+       COPY(lr);
+       COPY(sp);
+       COPY(r12);
+       COPY(r11);
+       COPY(r10);
+       COPY(r9);
+       COPY(r8);
+       COPY(r7);
+       COPY(r6);
+       COPY(r5);
+       COPY(r4);
+       COPY(r3);
+       COPY(r2);
+       COPY(r1);
+       COPY(r0);
+#undef COPY
+
+       /*
+        * Don't allow anyone to pretend they're running in supervisor
+        * mode or something...
+        */
+       err |= !valid_user_regs(regs);
+
+       return err;
+}
+
+
+asmlinkage int sys_rt_sigreturn(struct pt_regs *regs)
+{
+       struct rt_sigframe __user *frame;
+       sigset_t set;
+
+       frame = (struct rt_sigframe __user *)regs->sp;
+       pr_debug("SIG return: frame = %p\n", frame);
+
+       if (!access_ok(VERIFY_READ, frame, sizeof(*frame)))
+               goto badframe;
+
+       if (__copy_from_user(&set, &frame->uc.uc_sigmask, sizeof(set)))
+               goto badframe;
+
+       sigdelsetmask(&set, ~_BLOCKABLE);
+       spin_lock_irq(&current->sighand->siglock);
+       current->blocked = set;
+       recalc_sigpending();
+       spin_unlock_irq(&current->sighand->siglock);
+
+       if (restore_sigcontext(regs, &frame->uc.uc_mcontext))
+               goto badframe;
+
+       pr_debug("Context restored: pc = %08lx, lr = %08lx, sp = %08lx\n",
+                regs->pc, regs->lr, regs->sp);
+
+       return regs->r12;
+
+badframe:
+       force_sig(SIGSEGV, current);
+       return 0;
+}
+
+static int
+setup_sigcontext(struct sigcontext __user *sc, struct pt_regs *regs)
+{
+       int err = 0;
+
+#define COPY(x)                err |= __put_user(regs->x, &sc->x)
+       COPY(sr);
+       COPY(pc);
+       COPY(lr);
+       COPY(sp);
+       COPY(r12);
+       COPY(r11);
+       COPY(r10);
+       COPY(r9);
+       COPY(r8);
+       COPY(r7);
+       COPY(r6);
+       COPY(r5);
+       COPY(r4);
+       COPY(r3);
+       COPY(r2);
+       COPY(r1);
+       COPY(r0);
+#undef COPY
+
+       return err;
+}
+
+static inline void __user *
+get_sigframe(struct k_sigaction *ka, struct pt_regs *regs, int framesize)
+{
+       unsigned long sp = regs->sp;
+
+       if ((ka->sa.sa_flags & SA_ONSTACK) && !sas_ss_flags(sp))
+               sp = current->sas_ss_sp + current->sas_ss_size;
+
+       return (void __user *)((sp - framesize) & ~3);
+}
+
+static int
+setup_rt_frame(int sig, struct k_sigaction *ka, siginfo_t *info,
+              sigset_t *set, struct pt_regs *regs)
+{
+       struct rt_sigframe __user *frame;
+       int err = 0;
+
+       frame = get_sigframe(ka, regs, sizeof(*frame));
+       err = -EFAULT;
+       if (!access_ok(VERIFY_WRITE, frame, sizeof (*frame)))
+               goto out;
+
+       /*
+        * Set up the return code:
+        *
+        *      mov     r8, __NR_rt_sigreturn
+        *      scall
+        *
+        * Note: This will blow up since we're using a non-executable
+        * stack. Better use SA_RESTORER.
+        */
+#if __NR_rt_sigreturn > 127
+# error __NR_rt_sigreturn must be < 127 to fit in a short mov
+#endif
+       err = __put_user(0x3008d733 | (__NR_rt_sigreturn << 20),
+                        &frame->retcode);
+
+       err |= copy_siginfo_to_user(&frame->info, info);
+
+       /* Set up the ucontext */
+       err |= __put_user(0, &frame->uc.uc_flags);
+       err |= __put_user(NULL, &frame->uc.uc_link);
+       err |= __put_user((void __user *)current->sas_ss_sp,
+                         &frame->uc.uc_stack.ss_sp);
+       err |= __put_user(sas_ss_flags(regs->sp),
+                         &frame->uc.uc_stack.ss_flags);
+       err |= __put_user(current->sas_ss_size,
+                         &frame->uc.uc_stack.ss_size);
+       err |= setup_sigcontext(&frame->uc.uc_mcontext, regs);
+       err |= __copy_to_user(&frame->uc.uc_sigmask, set, sizeof(*set));
+
+       if (err)
+               goto out;
+
+       regs->r12 = sig;
+       regs->r11 = (unsigned long) &frame->info;
+       regs->r10 = (unsigned long) &frame->uc;
+       regs->sp = (unsigned long) frame;
+       if (ka->sa.sa_flags & SA_RESTORER)
+               regs->lr = (unsigned long)ka->sa.sa_restorer;
+       else {
+               printk(KERN_NOTICE "[%s:%d] did not set SA_RESTORER\n",
+                      current->comm, current->pid);
+               regs->lr = (unsigned long) &frame->retcode;
+       }
+
+       pr_debug("SIG deliver [%s:%d]: sig=%d sp=0x%lx pc=0x%lx->0x%p lr=0x%lx\n",
+                current->comm, current->pid, sig, regs->sp,
+                regs->pc, ka->sa.sa_handler, regs->lr);
+
+       regs->pc = (unsigned long) ka->sa.sa_handler;
+
+out:
+       return err;
+}
+
+static inline void restart_syscall(struct pt_regs *regs)
+{
+       if (regs->r12 == -ERESTART_RESTARTBLOCK)
+               regs->r8 = __NR_restart_syscall;
+       else
+               regs->r12 = regs->r12_orig;
+       regs->pc -= 2;
+}
+
+static inline void
+handle_signal(unsigned long sig, struct k_sigaction *ka, siginfo_t *info,
+             sigset_t *oldset, struct pt_regs *regs, int syscall)
+{
+       int ret;
+
+       /*
+        * Set up the stack frame
+        */
+       ret = setup_rt_frame(sig, ka, info, oldset, regs);
+
+       /*
+        * Check that the resulting registers are sane
+        */
+       ret |= !valid_user_regs(regs);
+
+       /*
+        * Block the signal if we were unsuccessful.
+        */
+       if (ret != 0 || !(ka->sa.sa_flags & SA_NODEFER)) {
+               spin_lock_irq(&current->sighand->siglock);
+               sigorsets(&current->blocked, &current->blocked,
+                         &ka->sa.sa_mask);
+               sigaddset(&current->blocked, sig);
+               recalc_sigpending();
+               spin_unlock_irq(&current->sighand->siglock);
+       }
+
+       if (ret == 0)
+               return;
+
+       force_sigsegv(sig, current);
+}
+
+/*
+ * Note that 'init' is a special process: it doesn't get signals it
+ * doesn't want to handle. Thus you cannot kill init even with a
+ * SIGKILL even by mistake.
+ */
+int do_signal(struct pt_regs *regs, sigset_t *oldset, int syscall)
+{
+       siginfo_t info;
+       int signr;
+       struct k_sigaction ka;
+
+       /*
+        * We want the common case to go fast, which is why we may in
+        * certain cases get here from kernel mode. Just return
+        * without doing anything if so.
+        */
+       if (!user_mode(regs))
+               return 0;
+
+       if (try_to_freeze()) {
+               signr = 0;
+               if (!signal_pending(current))
+                       goto no_signal;
+       }
+
+       if (test_thread_flag(TIF_RESTORE_SIGMASK))
+               oldset = &current->saved_sigmask;
+       else if (!oldset)
+               oldset = &current->blocked;
+
+       signr = get_signal_to_deliver(&info, &ka, regs, NULL);
+no_signal:
+       if (syscall) {
+               switch (regs->r12) {
+               case -ERESTART_RESTARTBLOCK:
+               case -ERESTARTNOHAND:
+                       if (signr > 0) {
+                               regs->r12 = -EINTR;
+                               break;
+                       }
+                       /* fall through */
+               case -ERESTARTSYS:
+                       if (signr > 0 && !(ka.sa.sa_flags & SA_RESTART)) {
+                               regs->r12 = -EINTR;
+                               break;
+                       }
+                       /* fall through */
+               case -ERESTARTNOINTR:
+                       restart_syscall(regs);
+               }
+       }
+
+       if (signr == 0) {
+               /* No signal to deliver -- put the saved sigmask back */
+               if (test_thread_flag(TIF_RESTORE_SIGMASK)) {
+                       clear_thread_flag(TIF_RESTORE_SIGMASK);
+                       sigprocmask(SIG_SETMASK, &current->saved_sigmask, NULL);
+               }
+               return 0;
+       }
+
+       handle_signal(signr, &ka, &info, oldset, regs, syscall);
+       return 1;
+}
+
+asmlinkage void do_notify_resume(struct pt_regs *regs, struct thread_info *ti)
+{
+       int syscall = 0;
+
+       if ((sysreg_read(SR) & MODE_MASK) == MODE_SUPERVISOR)
+               syscall = 1;
+
+       if (ti->flags & (_TIF_SIGPENDING | _TIF_RESTORE_SIGMASK))
+               do_signal(regs, &current->blocked, syscall);
+}
diff --git a/arch/avr32/kernel/switch_to.S b/arch/avr32/kernel/switch_to.S
new file mode 100644 (file)
index 0000000..a48d046
--- /dev/null
@@ -0,0 +1,35 @@
+/*
+ * Copyright (C) 2004-2006 Atmel Corporation
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ */
+
+#include <asm/sysreg.h>
+
+       .text
+       .global __switch_to
+       .type   __switch_to, @function
+
+       /* Switch thread context from "prev" to "next", returning "last"
+        *   r12 :      prev
+        *   r11 :      &prev->thread + 1
+        *   r10 :      &next->thread
+        */
+__switch_to:
+       stm     --r11, r0,r1,r2,r3,r4,r5,r6,r7,sp,lr
+       mfsr    r9, SYSREG_SR
+       st.w    --r11, r9
+       ld.w    r8, r10++
+       /*
+        * schedule() may have been called from a mode with a different
+        * set of registers. Make sure we don't lose anything here.
+        */
+       pushm   r10,r12
+       mtsr    SYSREG_SR, r8
+       frs                     /* flush the return stack */
+       sub     pc, -2          /* flush the pipeline */
+       popm    r10,r12
+       ldm     r10++, r0,r1,r2,r3,r4,r5,r6,r7,sp,pc
+       .size   __switch_to, . - __switch_to
diff --git a/arch/avr32/kernel/sys_avr32.c b/arch/avr32/kernel/sys_avr32.c
new file mode 100644 (file)
index 0000000..6ec5693
--- /dev/null
@@ -0,0 +1,51 @@
+/*
+ * Copyright (C) 2004-2006 Atmel Corporation
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ */
+#include <linux/errno.h>
+#include <linux/fs.h>
+#include <linux/file.h>
+#include <linux/mm.h>
+#include <linux/unistd.h>
+
+#include <asm/mman.h>
+#include <asm/uaccess.h>
+
+asmlinkage int sys_pipe(unsigned long __user *filedes)
+{
+       int fd[2];
+       int error;
+
+       error = do_pipe(fd);
+       if (!error) {
+               if (copy_to_user(filedes, fd, sizeof(fd)))
+                       error = -EFAULT;
+       }
+       return error;
+}
+
+asmlinkage long sys_mmap2(unsigned long addr, unsigned long len,
+                         unsigned long prot, unsigned long flags,
+                         unsigned long fd, off_t offset)
+{
+       int error = -EBADF;
+       struct file *file = NULL;
+
+       flags &= ~(MAP_EXECUTABLE | MAP_DENYWRITE);
+       if (!(flags & MAP_ANONYMOUS)) {
+               file = fget(fd);
+               if (!file)
+                       return error;
+       }
+
+       down_write(&current->mm->mmap_sem);
+       error = do_mmap_pgoff(file, addr, len, prot, flags, offset);
+       up_write(&current->mm->mmap_sem);
+
+       if (file)
+               fput(file);
+       return error;
+}
diff --git a/arch/avr32/kernel/syscall-stubs.S b/arch/avr32/kernel/syscall-stubs.S
new file mode 100644 (file)
index 0000000..7589a9b
--- /dev/null
@@ -0,0 +1,102 @@
+/*
+ * Copyright (C) 2005-2006 Atmel Corporation
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ */
+
+/*
+ * Stubs for syscalls that require access to pt_regs or that take more
+ * than five parameters.
+ */
+
+#define ARG6   r3
+
+       .text
+       .global __sys_rt_sigsuspend
+       .type   __sys_rt_sigsuspend,@function
+__sys_rt_sigsuspend:
+       mov     r10, sp
+       rjmp    sys_rt_sigsuspend
+
+       .global __sys_sigaltstack
+       .type   __sys_sigaltstack,@function
+__sys_sigaltstack:
+       mov     r10, sp
+       rjmp    sys_sigaltstack
+
+       .global __sys_rt_sigreturn
+       .type   __sys_rt_sigreturn,@function
+__sys_rt_sigreturn:
+       mov     r12, sp
+       rjmp    sys_rt_sigreturn
+
+       .global __sys_fork
+       .type   __sys_fork,@function
+__sys_fork:
+       mov     r12, sp
+       rjmp    sys_fork
+
+       .global __sys_clone
+       .type   __sys_clone,@function
+__sys_clone:
+       mov     r8, sp
+       rjmp    sys_clone
+
+       .global __sys_vfork
+       .type   __sys_vfork,@function
+__sys_vfork:
+       mov     r12, sp
+       rjmp    sys_vfork
+
+       .global __sys_execve
+       .type   __sys_execve,@function
+__sys_execve:
+       mov     r9, sp
+       rjmp    sys_execve
+
+       .global __sys_mmap2
+       .type   __sys_mmap2,@function
+__sys_mmap2:
+       pushm   lr
+       st.w    --sp, ARG6
+       rcall   sys_mmap2
+       sub     sp, -4
+       popm    pc
+
+       .global __sys_sendto
+       .type   __sys_sendto,@function
+__sys_sendto:
+       pushm   lr
+       st.w    --sp, ARG6
+       rcall   sys_sendto
+       sub     sp, -4
+       popm    pc
+
+       .global __sys_recvfrom
+       .type   __sys_recvfrom,@function
+__sys_recvfrom:
+       pushm   lr
+       st.w    --sp, ARG6
+       rcall   sys_recvfrom
+       sub     sp, -4
+       popm    pc
+
+       .global __sys_pselect6
+       .type   __sys_pselect6,@function
+__sys_pselect6:
+       pushm   lr
+       st.w    --sp, ARG6
+       rcall   sys_pselect6
+       sub     sp, -4
+       popm    pc
+
+       .global __sys_splice
+       .type   __sys_splice,@function
+__sys_splice:
+       pushm   lr
+       st.w    --sp, ARG6
+       rcall   sys_splice
+       sub     sp, -4
+       popm    pc
diff --git a/arch/avr32/kernel/syscall_table.S b/arch/avr32/kernel/syscall_table.S
new file mode 100644 (file)
index 0000000..63b2069
--- /dev/null
@@ -0,0 +1,289 @@
+/*
+ * AVR32 system call table
+ *
+ * Copyright (C) 2004-2006 Atmel Corporation
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ */
+
+#if !defined(CONFIG_NFSD) && !defined(CONFIG_NFSD_MODULE)
+#define sys_nfsservctl sys_ni_syscall
+#endif
+
+#if !defined(CONFIG_SYSV_IPC)
+# define sys_ipc       sys_ni_syscall
+#endif
+
+       .section .rodata,"a",@progbits
+       .type   sys_call_table,@object
+       .global sys_call_table
+       .align  2
+sys_call_table:
+       .long   sys_restart_syscall
+       .long   sys_exit
+       .long   __sys_fork
+       .long   sys_read
+       .long   sys_write
+       .long   sys_open                /* 5 */
+       .long   sys_close
+       .long   sys_umask
+       .long   sys_creat
+       .long   sys_link
+       .long   sys_unlink              /* 10 */
+       .long   __sys_execve
+       .long   sys_chdir
+       .long   sys_time
+       .long   sys_mknod
+       .long   sys_chmod               /* 15 */
+       .long   sys_chown
+       .long   sys_lchown
+       .long   sys_lseek
+       .long   sys_llseek
+       .long   sys_getpid              /* 20 */
+       .long   sys_mount
+       .long   sys_umount
+       .long   sys_setuid
+       .long   sys_getuid
+       .long   sys_stime               /* 25 */
+       .long   sys_ptrace
+       .long   sys_alarm
+       .long   sys_pause
+       .long   sys_utime
+       .long   sys_newstat             /* 30 */
+       .long   sys_newfstat
+       .long   sys_newlstat
+       .long   sys_access
+       .long   sys_chroot
+       .long   sys_sync                /* 35 */
+       .long   sys_fsync
+       .long   sys_kill
+       .long   sys_rename
+       .long   sys_mkdir
+       .long   sys_rmdir               /* 40 */
+       .long   sys_dup
+       .long   sys_pipe
+       .long   sys_times
+       .long   __sys_clone
+       .long   sys_brk                 /* 45 */
+       .long   sys_setgid
+       .long   sys_getgid
+       .long   sys_getcwd
+       .long   sys_geteuid
+       .long   sys_getegid             /* 50 */
+       .long   sys_acct
+       .long   sys_setfsuid
+       .long   sys_setfsgid
+       .long   sys_ioctl
+       .long   sys_fcntl               /* 55 */
+       .long   sys_setpgid
+       .long   sys_mremap
+       .long   sys_setresuid
+       .long   sys_getresuid
+       .long   sys_setreuid            /* 60 */
+       .long   sys_setregid
+       .long   sys_ustat
+       .long   sys_dup2
+       .long   sys_getppid
+       .long   sys_getpgrp             /* 65 */
+       .long   sys_setsid
+       .long   sys_rt_sigaction
+       .long   __sys_rt_sigreturn
+       .long   sys_rt_sigprocmask
+       .long   sys_rt_sigpending       /* 70 */
+       .long   sys_rt_sigtimedwait
+       .long   sys_rt_sigqueueinfo
+       .long   __sys_rt_sigsuspend
+       .long   sys_sethostname
+       .long   sys_setrlimit           /* 75 */
+       .long   sys_getrlimit
+       .long   sys_getrusage
+       .long   sys_gettimeofday
+       .long   sys_settimeofday
+       .long   sys_getgroups           /* 80 */
+       .long   sys_setgroups
+       .long   sys_select
+       .long   sys_symlink
+       .long   sys_fchdir
+       .long   sys_readlink            /* 85 */
+       .long   sys_pread64
+       .long   sys_pwrite64
+       .long   sys_swapon
+       .long   sys_reboot
+       .long   __sys_mmap2             /* 90 */
+       .long   sys_munmap
+       .long   sys_truncate
+       .long   sys_ftruncate
+       .long   sys_fchmod
+       .long   sys_fchown              /* 95 */
+       .long   sys_getpriority
+       .long   sys_setpriority
+       .long   sys_wait4
+       .long   sys_statfs
+       .long   sys_fstatfs             /* 100 */
+       .long   sys_vhangup
+       .long   __sys_sigaltstack
+       .long   sys_syslog
+       .long   sys_setitimer
+       .long   sys_getitimer           /* 105 */
+       .long   sys_swapoff
+       .long   sys_sysinfo
+       .long   sys_ipc
+       .long   sys_sendfile
+       .long   sys_setdomainname       /* 110 */
+       .long   sys_newuname
+       .long   sys_adjtimex
+       .long   sys_mprotect
+       .long   __sys_vfork
+       .long   sys_init_module         /* 115 */
+       .long   sys_delete_module
+       .long   sys_quotactl
+       .long   sys_getpgid
+       .long   sys_bdflush
+       .long   sys_sysfs               /* 120 */
+       .long   sys_personality
+       .long   sys_ni_syscall          /* reserved for afs_syscall */
+       .long   sys_getdents
+       .long   sys_flock
+       .long   sys_msync               /* 125 */
+       .long   sys_readv
+       .long   sys_writev
+       .long   sys_getsid
+       .long   sys_fdatasync
+       .long   sys_sysctl              /* 130 */
+       .long   sys_mlock
+       .long   sys_munlock
+       .long   sys_mlockall
+       .long   sys_munlockall
+       .long   sys_sched_setparam              /* 135 */
+       .long   sys_sched_getparam
+       .long   sys_sched_setscheduler
+       .long   sys_sched_getscheduler
+       .long   sys_sched_yield
+       .long   sys_sched_get_priority_max      /* 140 */
+       .long   sys_sched_get_priority_min
+       .long   sys_sched_rr_get_interval
+       .long   sys_nanosleep
+       .long   sys_poll
+       .long   sys_nfsservctl          /* 145 */
+       .long   sys_setresgid
+       .long   sys_getresgid
+       .long   sys_prctl
+       .long   sys_socket
+       .long   sys_bind                /* 150 */
+       .long   sys_connect
+       .long   sys_listen
+       .long   sys_accept
+       .long   sys_getsockname
+       .long   sys_getpeername         /* 155 */
+       .long   sys_socketpair
+       .long   sys_send
+       .long   sys_recv
+       .long   __sys_sendto
+       .long   __sys_recvfrom          /* 160 */
+       .long   sys_shutdown
+       .long   sys_setsockopt
+       .long   sys_getsockopt
+       .long   sys_sendmsg
+       .long   sys_recvmsg             /* 165 */
+       .long   sys_truncate64
+       .long   sys_ftruncate64
+       .long   sys_stat64
+       .long   sys_lstat64
+       .long   sys_fstat64             /* 170 */
+       .long   sys_pivot_root
+       .long   sys_mincore
+       .long   sys_madvise
+       .long   sys_getdents64
+       .long   sys_fcntl64             /* 175 */
+       .long   sys_gettid
+       .long   sys_readahead
+       .long   sys_setxattr
+       .long   sys_lsetxattr
+       .long   sys_fsetxattr           /* 180 */
+       .long   sys_getxattr
+       .long   sys_lgetxattr
+       .long   sys_fgetxattr
+       .long   sys_listxattr
+       .long   sys_llistxattr          /* 185 */
+       .long   sys_flistxattr
+       .long   sys_removexattr
+       .long   sys_lremovexattr
+       .long   sys_fremovexattr
+       .long   sys_tkill               /* 190 */
+       .long   sys_sendfile64
+       .long   sys_futex
+       .long   sys_sched_setaffinity
+       .long   sys_sched_getaffinity
+       .long   sys_capget              /* 195 */
+       .long   sys_capset
+       .long   sys_io_setup
+       .long   sys_io_destroy
+       .long   sys_io_getevents
+       .long   sys_io_submit           /* 200 */
+       .long   sys_io_cancel
+       .long   sys_fadvise64
+       .long   sys_exit_group
+       .long   sys_lookup_dcookie
+       .long   sys_epoll_create        /* 205 */
+       .long   sys_epoll_ctl
+       .long   sys_epoll_wait
+       .long   sys_remap_file_pages
+       .long   sys_set_tid_address
+       .long   sys_timer_create        /* 210 */
+       .long   sys_timer_settime
+       .long   sys_timer_gettime
+       .long   sys_timer_getoverrun
+       .long   sys_timer_delete
+       .long   sys_clock_settime       /* 215 */
+       .long   sys_clock_gettime
+       .long   sys_clock_getres
+       .long   sys_clock_nanosleep
+       .long   sys_statfs64
+       .long   sys_fstatfs64           /* 220 */
+       .long   sys_tgkill
+       .long   sys_ni_syscall          /* reserved for TUX */
+       .long   sys_utimes
+       .long   sys_fadvise64_64
+       .long   sys_cacheflush          /* 225 */
+       .long   sys_ni_syscall          /* sys_vserver */
+       .long   sys_mq_open
+       .long   sys_mq_unlink
+       .long   sys_mq_timedsend
+       .long   sys_mq_timedreceive     /* 230 */
+       .long   sys_mq_notify
+       .long   sys_mq_getsetattr
+       .long   sys_kexec_load
+       .long   sys_waitid
+       .long   sys_add_key             /* 235 */
+       .long   sys_request_key
+       .long   sys_keyctl
+       .long   sys_ioprio_set
+       .long   sys_ioprio_get
+       .long   sys_inotify_init        /* 240 */
+       .long   sys_inotify_add_watch
+       .long   sys_inotify_rm_watch
+       .long   sys_openat
+       .long   sys_mkdirat
+       .long   sys_mknodat             /* 245 */
+       .long   sys_fchownat
+       .long   sys_futimesat
+       .long   sys_fstatat64
+       .long   sys_unlinkat
+       .long   sys_renameat            /* 250 */
+       .long   sys_linkat
+       .long   sys_symlinkat
+       .long   sys_readlinkat
+       .long   sys_fchmodat
+       .long   sys_faccessat           /* 255 */
+       .long   __sys_pselect6
+       .long   sys_ppoll
+       .long   sys_unshare
+       .long   sys_set_robust_list
+       .long   sys_get_robust_list     /* 260 */
+       .long   __sys_splice
+       .long   sys_sync_file_range
+       .long   sys_tee
+       .long   sys_vmsplice
+       .long   sys_ni_syscall          /* r8 is saturated at nr_syscalls */
diff --git a/arch/avr32/kernel/time.c b/arch/avr32/kernel/time.c
new file mode 100644 (file)
index 0000000..b0e6b58
--- /dev/null
@@ -0,0 +1,238 @@
+/*
+ * Copyright (C) 2004-2006 Atmel Corporation
+ *
+ * Based on MIPS implementation arch/mips/kernel/time.c
+ *   Copyright 2001 MontaVista Software Inc.
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ */
+
+#include <linux/clk.h>
+#include <linux/clocksource.h>
+#include <linux/time.h>
+#include <linux/module.h>
+#include <linux/interrupt.h>
+#include <linux/irq.h>
+#include <linux/kernel_stat.h>
+#include <linux/errno.h>
+#include <linux/init.h>
+#include <linux/profile.h>
+#include <linux/sysdev.h>
+
+#include <asm/div64.h>
+#include <asm/sysreg.h>
+#include <asm/io.h>
+#include <asm/sections.h>
+
+static cycle_t read_cycle_count(void)
+{
+       return (cycle_t)sysreg_read(COUNT);
+}
+
+static struct clocksource clocksource_avr32 = {
+       .name           = "avr32",
+       .rating         = 350,
+       .read           = read_cycle_count,
+       .mask           = CLOCKSOURCE_MASK(32),
+       .shift          = 16,
+       .is_continuous  = 1,
+};
+
+/*
+ * By default we provide the null RTC ops
+ */
+static unsigned long null_rtc_get_time(void)
+{
+       return mktime(2004, 1, 1, 0, 0, 0);
+}
+
+static int null_rtc_set_time(unsigned long sec)
+{
+       return 0;
+}
+
+static unsigned long (*rtc_get_time)(void) = null_rtc_get_time;
+static int (*rtc_set_time)(unsigned long) = null_rtc_set_time;
+
+/* how many counter cycles in a jiffy? */
+static unsigned long cycles_per_jiffy;
+
+/* cycle counter value at the previous timer interrupt */
+static unsigned int timerhi, timerlo;
+
+/* the count value for the next timer interrupt */
+static unsigned int expirelo;
+
+static void avr32_timer_ack(void)
+{
+       unsigned int count;
+
+       /* Ack this timer interrupt and set the next one */
+       expirelo += cycles_per_jiffy;
+       if (expirelo == 0) {
+               printk(KERN_DEBUG "expirelo == 0\n");
+               sysreg_write(COMPARE, expirelo + 1);
+       } else {
+               sysreg_write(COMPARE, expirelo);
+       }
+
+       /* Check to see if we have missed any timer interrupts */
+       count = sysreg_read(COUNT);
+       if ((count - expirelo) < 0x7fffffff) {
+               expirelo = count + cycles_per_jiffy;
+               sysreg_write(COMPARE, expirelo);
+       }
+}
+
+static unsigned int avr32_hpt_read(void)
+{
+       return sysreg_read(COUNT);
+}
+
+/*
+ * Taken from MIPS c0_hpt_timer_init().
+ *
+ * Why is it so complicated, and what is "count"?  My assumption is
+ * that `count' specifies the "reference cycle", i.e. the cycle since
+ * reset that should mean "zero". The reason COUNT is written twice is
+ * probably to make sure we don't get any timer interrupts while we
+ * are messing with the counter.
+ */
+static void avr32_hpt_init(unsigned int count)
+{
+       count = sysreg_read(COUNT) - count;
+       expirelo = (count / cycles_per_jiffy + 1) * cycles_per_jiffy;
+       sysreg_write(COUNT, expirelo - cycles_per_jiffy);
+       sysreg_write(COMPARE, expirelo);
+       sysreg_write(COUNT, count);
+}
+
+/*
+ * Scheduler clock - returns current time in nanosec units.
+ */
+unsigned long long sched_clock(void)
+{
+       /* There must be better ways...? */
+       return (unsigned long long)jiffies * (1000000000 / HZ);
+}
+
+/*
+ * local_timer_interrupt() does profiling and process accounting on a
+ * per-CPU basis.
+ *
+ * In UP mode, it is invoked from the (global) timer_interrupt.
+ */
+static void local_timer_interrupt(int irq, void *dev_id, struct pt_regs *regs)
+{
+       if (current->pid)
+               profile_tick(CPU_PROFILING, regs);
+       update_process_times(user_mode(regs));
+}
+
+static irqreturn_t
+timer_interrupt(int irq, void *dev_id, struct pt_regs *regs)
+{
+       unsigned int count;
+
+       /* ack timer interrupt and try to set next interrupt */
+       count = avr32_hpt_read();
+       avr32_timer_ack();
+
+       /* Update timerhi/timerlo for intra-jiffy calibration */
+       timerhi += count < timerlo;     /* Wrap around */
+       timerlo = count;
+
+       /*
+        * Call the generic timer interrupt handler
+        */
+       write_seqlock(&xtime_lock);
+       do_timer(regs);
+       write_sequnlock(&xtime_lock);
+
+       /*
+        * In UP mode, we call local_timer_interrupt() to do profiling
+        * and process accounting.
+        *
+        * SMP is not supported yet.
+        */
+       local_timer_interrupt(irq, dev_id, regs);
+
+       return IRQ_HANDLED;
+}
+
+static struct irqaction timer_irqaction = {
+       .handler        = timer_interrupt,
+       .flags          = IRQF_DISABLED,
+       .name           = "timer",
+};
+
+void __init time_init(void)
+{
+       unsigned long mult, shift, count_hz;
+       int ret;
+
+       xtime.tv_sec = rtc_get_time();
+       xtime.tv_nsec = 0;
+
+       set_normalized_timespec(&wall_to_monotonic,
+                               -xtime.tv_sec, -xtime.tv_nsec);
+
+       printk("Before time_init: count=%08lx, compare=%08lx\n",
+              (unsigned long)sysreg_read(COUNT),
+              (unsigned long)sysreg_read(COMPARE));
+
+       count_hz = clk_get_rate(boot_cpu_data.clk);
+       shift = clocksource_avr32.shift;
+       mult = clocksource_hz2mult(count_hz, shift);
+       clocksource_avr32.mult = mult;
+
+       printk("Cycle counter: mult=%lu, shift=%lu\n", mult, shift);
+
+       {
+               u64 tmp;
+
+               tmp = TICK_NSEC;
+               tmp <<= shift;
+               tmp += mult / 2;
+               do_div(tmp, mult);
+
+               cycles_per_jiffy = tmp;
+       }
+
+       /* This sets up the high precision timer for the first interrupt. */
+       avr32_hpt_init(avr32_hpt_read());
+
+       printk("After time_init: count=%08lx, compare=%08lx\n",
+              (unsigned long)sysreg_read(COUNT),
+              (unsigned long)sysreg_read(COMPARE));
+
+       ret = clocksource_register(&clocksource_avr32);
+       if (ret)
+               printk(KERN_ERR
+                      "timer: could not register clocksource: %d\n", ret);
+
+       ret = setup_irq(0, &timer_irqaction);
+       if (ret)
+               printk("timer: could not request IRQ 0: %d\n", ret);
+}
+
+static struct sysdev_class timer_class = {
+       set_kset_name("timer"),
+};
+
+static struct sys_device timer_device = {
+       .id     = 0,
+       .cls    = &timer_class,
+};
+
+static int __init init_timer_sysfs(void)
+{
+       int err = sysdev_class_register(&timer_class);
+       if (!err)
+               err = sysdev_register(&timer_device);
+       return err;
+}
+
+device_initcall(init_timer_sysfs);
diff --git a/arch/avr32/kernel/traps.c b/arch/avr32/kernel/traps.c
new file mode 100644 (file)
index 0000000..7e803f4
--- /dev/null
@@ -0,0 +1,425 @@
+/*
+ * Copyright (C) 2004-2006 Atmel Corporation
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ */
+#undef DEBUG
+#include <linux/sched.h>
+#include <linux/init.h>
+#include <linux/module.h>
+#include <linux/kallsyms.h>
+#include <linux/notifier.h>
+
+#include <asm/traps.h>
+#include <asm/sysreg.h>
+#include <asm/addrspace.h>
+#include <asm/ocd.h>
+#include <asm/mmu_context.h>
+#include <asm/uaccess.h>
+
+static void dump_mem(const char *str, unsigned long bottom, unsigned long top)
+{
+       unsigned long p;
+       int i;
+
+       printk("%s(0x%08lx to 0x%08lx)\n", str, bottom, top);
+
+       for (p = bottom & ~31; p < top; ) {
+               printk("%04lx: ", p & 0xffff);
+
+               for (i = 0; i < 8; i++, p += 4) {
+                       unsigned int val;
+
+                       if (p < bottom || p >= top)
+                               printk("         ");
+                       else {
+                               if (__get_user(val, (unsigned int __user *)p)) {
+                                       printk("\n");
+                                       goto out;
+                               }
+                               printk("%08x ", val);
+                       }
+               }
+               printk("\n");
+       }
+
+out:
+       return;
+}
+
+#ifdef CONFIG_FRAME_POINTER
+static inline void __show_trace(struct task_struct *tsk, unsigned long *sp,
+                               struct pt_regs *regs)
+{
+       unsigned long __user *fp;
+       unsigned long __user *last_fp = NULL;
+
+       if (regs) {
+               fp = (unsigned long __user *)regs->r7;
+       } else if (tsk == current) {
+               register unsigned long __user *real_fp __asm__("r7");
+               fp = real_fp;
+       } else {
+               fp = (unsigned long __user *)tsk->thread.cpu_context.r7;
+       }
+
+       /*
+        * Walk the stack until (a) we get an exception, (b) the frame
+        * pointer becomes zero, or (c) the frame pointer gets stuck
+        * at the same value.
+        */
+       while (fp && fp != last_fp) {
+               unsigned long lr, new_fp = 0;
+
+               last_fp = fp;
+               if (__get_user(lr, fp))
+                       break;
+               if (fp && __get_user(new_fp, fp + 1))
+                       break;
+               fp = (unsigned long __user *)new_fp;
+
+               printk(" [<%08lx>] ", lr);
+               print_symbol("%s\n", lr);
+       }
+       printk("\n");
+}
+#else
+static inline void __show_trace(struct task_struct *tsk, unsigned long *sp,
+                               struct pt_regs *regs)
+{
+       unsigned long addr;
+
+       while (!kstack_end(sp)) {
+               addr = *sp++;
+               if (kernel_text_address(addr)) {
+                       printk(" [<%08lx>] ", addr);
+                       print_symbol("%s\n", addr);
+               }
+       }
+}
+#endif
+
+void show_trace(struct task_struct *tsk, unsigned long *sp,
+                      struct pt_regs *regs)
+{
+       if (regs &&
+           (((regs->sr & MODE_MASK) == MODE_EXCEPTION) ||
+            ((regs->sr & MODE_MASK) == MODE_USER)))
+               return;
+
+       printk ("Call trace:");
+#ifdef CONFIG_KALLSYMS
+       printk("\n");
+#endif
+
+       __show_trace(tsk, sp, regs);
+       printk("\n");
+}
+
+void show_stack(struct task_struct *tsk, unsigned long *sp)
+{
+       unsigned long stack;
+
+       if (!tsk)
+               tsk = current;
+       if (sp == 0) {
+               if (tsk == current) {
+                       register unsigned long *real_sp __asm__("sp");
+                       sp = real_sp;
+               } else {
+                       sp = (unsigned long *)tsk->thread.cpu_context.ksp;
+               }
+       }
+
+       stack = (unsigned long)sp;
+       dump_mem("Stack: ", stack,
+                THREAD_SIZE + (unsigned long)tsk->thread_info);
+       show_trace(tsk, sp, NULL);
+}
+
+void dump_stack(void)
+{
+       show_stack(NULL, NULL);
+}
+EXPORT_SYMBOL(dump_stack);
+
+ATOMIC_NOTIFIER_HEAD(avr32_die_chain);
+
+int register_die_notifier(struct notifier_block *nb)
+{
+       pr_debug("register_die_notifier: %p\n", nb);
+
+       return atomic_notifier_chain_register(&avr32_die_chain, nb);
+}
+EXPORT_SYMBOL(register_die_notifier);
+
+int unregister_die_notifier(struct notifier_block *nb)
+{
+       return atomic_notifier_chain_unregister(&avr32_die_chain, nb);
+}
+EXPORT_SYMBOL(unregister_die_notifier);
+
+static DEFINE_SPINLOCK(die_lock);
+
+void __die(const char *str, struct pt_regs *regs, unsigned long err,
+          const char *file, const char *func, unsigned long line)
+{
+       struct task_struct *tsk = current;
+       static int die_counter;
+
+       console_verbose();
+       spin_lock_irq(&die_lock);
+       bust_spinlocks(1);
+
+       printk(KERN_ALERT "%s", str);
+       if (file && func)
+               printk(" in %s:%s, line %ld", file, func, line);
+       printk("[#%d]:\n", ++die_counter);
+       print_modules();
+       show_regs(regs);
+       printk("Process %s (pid: %d, stack limit = 0x%p)\n",
+              tsk->comm, tsk->pid, tsk->thread_info + 1);
+
+       if (!user_mode(regs) || in_interrupt()) {
+               dump_mem("Stack: ", regs->sp,
+                        THREAD_SIZE + (unsigned long)tsk->thread_info);
+       }
+
+       bust_spinlocks(0);
+       spin_unlock_irq(&die_lock);
+       do_exit(SIGSEGV);
+}
+
+void __die_if_kernel(const char *str, struct pt_regs *regs, unsigned long err,
+                    const char *file, const char *func, unsigned long line)
+{
+       if (!user_mode(regs))
+               __die(str, regs, err, file, func, line);
+}
+
+asmlinkage void do_nmi(unsigned long ecr, struct pt_regs *regs)
+{
+#ifdef CONFIG_SUBARCH_AVR32B
+       /*
+        * The exception entry always saves RSR_EX. For NMI, this is
+        * wrong; it should be RSR_NMI
+        */
+       regs->sr = sysreg_read(RSR_NMI);
+#endif
+
+       printk("NMI taken!!!!\n");
+       die("NMI", regs, ecr);
+       BUG();
+}
+
+asmlinkage void do_critical_exception(unsigned long ecr, struct pt_regs *regs)
+{
+       printk("Unable to handle critical exception %lu at pc = %08lx!\n",
+              ecr, regs->pc);
+       die("Oops", regs, ecr);
+       BUG();
+}
+
+asmlinkage void do_address_exception(unsigned long ecr, struct pt_regs *regs)
+{
+       siginfo_t info;
+
+       die_if_kernel("Oops: Address exception in kernel mode", regs, ecr);
+
+#ifdef DEBUG
+       if (ecr == ECR_ADDR_ALIGN_X)
+               pr_debug("Instruction Address Exception at pc = %08lx\n",
+                        regs->pc);
+       else if (ecr == ECR_ADDR_ALIGN_R)
+               pr_debug("Data Address Exception (Read) at pc = %08lx\n",
+                        regs->pc);
+       else if (ecr == ECR_ADDR_ALIGN_W)
+               pr_debug("Data Address Exception (Write) at pc = %08lx\n",
+                        regs->pc);
+       else
+               BUG();
+
+       show_regs(regs);
+#endif
+
+       info.si_signo = SIGBUS;
+       info.si_errno = 0;
+       info.si_code = BUS_ADRALN;
+       info.si_addr = (void __user *)regs->pc;
+
+       force_sig_info(SIGBUS, &info, current);
+}
+
+/* This way of handling undefined instructions is stolen from ARM */
+static LIST_HEAD(undef_hook);
+static spinlock_t undef_lock = SPIN_LOCK_UNLOCKED;
+
+void register_undef_hook(struct undef_hook *hook)
+{
+       spin_lock_irq(&undef_lock);
+       list_add(&hook->node, &undef_hook);
+       spin_unlock_irq(&undef_lock);
+}
+
+void unregister_undef_hook(struct undef_hook *hook)
+{
+       spin_lock_irq(&undef_lock);
+       list_del(&hook->node);
+       spin_unlock_irq(&undef_lock);
+}
+
+static int do_cop_absent(u32 insn)
+{
+       int cop_nr;
+       u32 cpucr;
+       if ( (insn & 0xfdf00000) == 0xf1900000 )
+               /* LDC0 */
+               cop_nr = 0;
+       else
+               cop_nr = (insn >> 13) & 0x7;
+
+       /* Try enabling the coprocessor */
+       cpucr = sysreg_read(CPUCR);
+       cpucr |= (1 << (24 + cop_nr));
+       sysreg_write(CPUCR, cpucr);
+
+       cpucr = sysreg_read(CPUCR);
+       if ( !(cpucr & (1 << (24 + cop_nr))) ){
+               printk("Coprocessor #%i not found!\n", cop_nr);
+               return -1;
+       }
+
+       return 0;
+}
+
+#ifdef CONFIG_BUG
+#ifdef CONFIG_DEBUG_BUGVERBOSE
+static inline void do_bug_verbose(struct pt_regs *regs, u32 insn)
+{
+       char *file;
+       u16 line;
+       char c;
+
+       if (__get_user(line, (u16 __user *)(regs->pc + 2)))
+               return;
+       if (__get_user(file, (char * __user *)(regs->pc + 4))
+           || (unsigned long)file < PAGE_OFFSET
+           || __get_user(c, file))
+               file = "<bad filename>";
+
+       printk(KERN_ALERT "kernel BUG at %s:%d!\n", file, line);
+}
+#else
+static inline void do_bug_verbose(struct pt_regs *regs, u32 insn)
+{
+
+}
+#endif
+#endif
+
+asmlinkage void do_illegal_opcode(unsigned long ecr, struct pt_regs *regs)
+{
+       u32 insn;
+       struct undef_hook *hook;
+       siginfo_t info;
+       void __user *pc;
+
+       if (!user_mode(regs))
+               goto kernel_trap;
+
+       local_irq_enable();
+
+       pc = (void __user *)instruction_pointer(regs);
+       if (__get_user(insn, (u32 __user *)pc))
+               goto invalid_area;
+
+        if (ecr == ECR_COPROC_ABSENT) {
+               if (do_cop_absent(insn) == 0)
+                       return;
+        }
+
+       spin_lock_irq(&undef_lock);
+       list_for_each_entry(hook, &undef_hook, node) {
+               if ((insn & hook->insn_mask) == hook->insn_val) {
+                       if (hook->fn(regs, insn) == 0) {
+                               spin_unlock_irq(&undef_lock);
+                               return;
+                       }
+               }
+       }
+       spin_unlock_irq(&undef_lock);
+
+invalid_area:
+
+#ifdef DEBUG
+       printk("Illegal instruction at pc = %08lx\n", regs->pc);
+       if (regs->pc < TASK_SIZE) {
+               unsigned long ptbr, pgd, pte, *p;
+
+               ptbr = sysreg_read(PTBR);
+               p = (unsigned long *)ptbr;
+               pgd = p[regs->pc >> 22];
+               p = (unsigned long *)((pgd & 0x1ffff000) | 0x80000000);
+               pte = p[(regs->pc >> 12) & 0x3ff];
+               printk("page table: 0x%08lx -> 0x%08lx -> 0x%08lx\n", ptbr, pgd, pte);
+       }
+#endif
+
+       info.si_signo = SIGILL;
+       info.si_errno = 0;
+       info.si_addr = (void __user *)regs->pc;
+       switch (ecr) {
+       case ECR_ILLEGAL_OPCODE:
+       case ECR_UNIMPL_INSTRUCTION:
+               info.si_code = ILL_ILLOPC;
+               break;
+       case ECR_PRIVILEGE_VIOLATION:
+               info.si_code = ILL_PRVOPC;
+               break;
+       case ECR_COPROC_ABSENT:
+               info.si_code = ILL_COPROC;
+               break;
+       default:
+               BUG();
+       }
+
+       force_sig_info(SIGILL, &info, current);
+       return;
+
+kernel_trap:
+#ifdef CONFIG_BUG
+       if (__kernel_text_address(instruction_pointer(regs))) {
+               insn = *(u16 *)instruction_pointer(regs);
+               if (insn == AVR32_BUG_OPCODE) {
+                       do_bug_verbose(regs, insn);
+                       die("Kernel BUG", regs, 0);
+                       return;
+               }
+       }
+#endif
+
+       die("Oops: Illegal instruction in kernel code", regs, ecr);
+}
+
+asmlinkage void do_fpe(unsigned long ecr, struct pt_regs *regs)
+{
+       siginfo_t info;
+
+       printk("Floating-point exception at pc = %08lx\n", regs->pc);
+
+       /* We have no FPU... */
+       info.si_signo = SIGILL;
+       info.si_errno = 0;
+       info.si_addr = (void __user *)regs->pc;
+       info.si_code = ILL_COPROC;
+
+       force_sig_info(SIGILL, &info, current);
+}
+
+
+void __init trap_init(void)
+{
+
+}
diff --git a/arch/avr32/kernel/vmlinux.lds.c b/arch/avr32/kernel/vmlinux.lds.c
new file mode 100644 (file)
index 0000000..cdd627c
--- /dev/null
@@ -0,0 +1,139 @@
+/*
+ * AVR32 linker script for the Linux kernel
+ *
+ * Copyright (C) 2004-2006 Atmel Corporation
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ */
+#define LOAD_OFFSET 0x00000000
+#include <asm-generic/vmlinux.lds.h>
+
+OUTPUT_FORMAT("elf32-avr32", "elf32-avr32", "elf32-avr32")
+OUTPUT_ARCH(avr32)
+ENTRY(_start)
+
+/* Big endian */
+jiffies = jiffies_64 + 4;
+
+SECTIONS
+{
+       . = CONFIG_ENTRY_ADDRESS;
+       .init           : AT(ADDR(.init) - LOAD_OFFSET) {
+               _stext = .;
+               __init_begin = .;
+                       _sinittext = .;
+                       *(.text.reset)
+                       *(.init.text)
+                       _einittext = .;
+               . = ALIGN(4);
+               __tagtable_begin = .;
+                       *(.taglist)
+               __tagtable_end = .;
+                       *(.init.data)
+               . = ALIGN(16);
+               __setup_start = .;
+                       *(.init.setup)
+               __setup_end = .;
+               . = ALIGN(4);
+               __initcall_start = .;
+                       *(.initcall1.init)
+                       *(.initcall2.init)
+                       *(.initcall3.init)
+                       *(.initcall4.init)
+                       *(.initcall5.init)
+                       *(.initcall6.init)
+                       *(.initcall7.init)
+               __initcall_end = .;
+               __con_initcall_start = .;
+                       *(.con_initcall.init)
+               __con_initcall_end = .;
+               __security_initcall_start = .;
+                       *(.security_initcall.init)
+               __security_initcall_end = .;
+               . = ALIGN(32);
+               __initramfs_start = .;
+                       *(.init.ramfs)
+               __initramfs_end = .;
+               . = ALIGN(4096);
+               __init_end = .;
+       }
+
+       . = ALIGN(8192);
+       .text           : AT(ADDR(.text) - LOAD_OFFSET) {
+               _evba = .;
+               _text = .;
+               *(.ex.text)
+               . = 0x50;
+               *(.tlbx.ex.text)
+               . = 0x60;
+               *(.tlbr.ex.text)
+               . = 0x70;
+               *(.tlbw.ex.text)
+               . = 0x100;
+               *(.scall.text)
+               *(.irq.text)
+               *(.text)
+               SCHED_TEXT
+               LOCK_TEXT
+               KPROBES_TEXT
+               *(.fixup)
+               *(.gnu.warning)
+               _etext = .;
+       } = 0xd703d703
+
+       . = ALIGN(4);
+       __ex_table      : AT(ADDR(__ex_table) - LOAD_OFFSET) {
+               __start___ex_table = .;
+               *(__ex_table)
+               __stop___ex_table = .;
+       }
+
+       RODATA
+
+       . = ALIGN(8192);
+
+       .data           : AT(ADDR(.data) - LOAD_OFFSET) {
+               _data = .;
+               _sdata = .;
+               /*
+                * First, the init task union, aligned to an 8K boundary.
+                */
+               *(.data.init_task)
+
+               /* Then, the cacheline aligned data */
+               . = ALIGN(32);
+               *(.data.cacheline_aligned)
+
+               /* And the rest... */
+               *(.data.rel*)
+               *(.data)
+               CONSTRUCTORS
+
+               _edata = .;
+       }
+
+
+       . = ALIGN(8);
+       .bss            : AT(ADDR(.bss) - LOAD_OFFSET) {
+               __bss_start = .;
+               *(.bss)
+               *(COMMON)
+               . = ALIGN(8);
+               __bss_stop = .;
+               _end = .;
+       }
+
+       /* When something in the kernel is NOT compiled as a module, the module
+        * cleanup code and data are put into these segments. Both can then be
+        * thrown away, as cleanup code is never called unless it's a module.
+        */
+       /DISCARD/               : {
+               *(.exit.text)
+               *(.exit.data)
+               *(.exitcall.exit)
+       }
+
+       DWARF_DEBUG
+}
diff --git a/arch/avr32/lib/Makefile b/arch/avr32/lib/Makefile
new file mode 100644 (file)
index 0000000..09ac43e
--- /dev/null
@@ -0,0 +1,10 @@
+#
+# Makefile for AVR32-specific library files
+#
+
+lib-y  := copy_user.o clear_user.o
+lib-y  += strncpy_from_user.o strnlen_user.o
+lib-y  += delay.o memset.o memcpy.o findbit.o
+lib-y  += csum_partial.o csum_partial_copy_generic.o
+lib-y  += io-readsw.o io-readsl.o io-writesw.o io-writesl.o
+lib-y  += __avr32_lsl64.o __avr32_lsr64.o __avr32_asr64.o
diff --git a/arch/avr32/lib/__avr32_asr64.S b/arch/avr32/lib/__avr32_asr64.S
new file mode 100644 (file)
index 0000000..368b6bc
--- /dev/null
@@ -0,0 +1,31 @@
+/*
+ * Copyright (C) 2005-2006 Atmel Corporation
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ */
+
+       /*
+        * DWtype __avr32_asr64(DWtype u, word_type b)
+        */
+       .text
+       .global __avr32_asr64
+       .type   __avr32_asr64,@function
+__avr32_asr64:
+       cp.w    r12, 0
+       reteq   r12
+
+       rsub    r9, r12, 32
+       brle    1f
+
+       lsl     r8, r11, r9
+       lsr     r10, r10, r12
+       asr     r11, r11, r12
+       or      r10, r8
+       retal   r12
+
+1:     neg     r9
+       asr     r10, r11, r9
+       asr     r11, 31
+       retal   r12
diff --git a/arch/avr32/lib/__avr32_lsl64.S b/arch/avr32/lib/__avr32_lsl64.S
new file mode 100644 (file)
index 0000000..f1dbc2b
--- /dev/null
@@ -0,0 +1,31 @@
+/*
+ * Copyright (C) 2005-2006 Atmel Corporation
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ */
+
+       /*
+        * DWtype __avr32_lsl64(DWtype u, word_type b)
+        */
+       .text
+       .global __avr32_lsl64
+       .type   __avr32_lsl64,@function
+__avr32_lsl64:
+       cp.w    r12, 0
+       reteq   r12
+
+       rsub    r9, r12, 32
+       brle    1f
+
+       lsr     r8, r10, r9
+       lsl     r10, r10, r12
+       lsl     r11, r11, r12
+       or      r11, r8
+       retal   r12
+
+1:     neg     r9
+       lsl     r11, r10, r9
+       mov     r10, 0
+       retal   r12
diff --git a/arch/avr32/lib/__avr32_lsr64.S b/arch/avr32/lib/__avr32_lsr64.S
new file mode 100644 (file)
index 0000000..e65bb7f
--- /dev/null
@@ -0,0 +1,31 @@
+/*
+ * Copyright (C) 2005-2006 Atmel Corporation
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ */
+
+       /*
+        * DWtype __avr32_lsr64(DWtype u, word_type b)
+        */
+       .text
+       .global __avr32_lsr64
+       .type   __avr32_lsr64,@function
+__avr32_lsr64:
+       cp.w    r12, 0
+       reteq   r12
+
+       rsub    r9, r12, 32
+       brle    1f
+
+       lsl     r8, r11, r9
+       lsr     r11, r11, r12
+       lsr     r10, r10, r12
+       or      r10, r8
+       retal   r12
+
+1:     neg     r9
+       lsr     r10, r11, r9
+       mov     r11, 0
+       retal   r12
diff --git a/arch/avr32/lib/clear_user.S b/arch/avr32/lib/clear_user.S
new file mode 100644 (file)
index 0000000..d8991b6
--- /dev/null
@@ -0,0 +1,76 @@
+/*
+ * Copyright 2004-2006 Atmel Corporation
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ */
+#include <asm/page.h>
+#include <asm/thread_info.h>
+#include <asm/asm.h>
+
+       .text
+       .align  1
+       .global clear_user
+       .type   clear_user, "function"
+clear_user:
+       branch_if_kernel r8, __clear_user
+       ret_if_privileged r8, r12, r11, r11
+
+       .global __clear_user
+       .type   __clear_user, "function"
+__clear_user:
+       mov     r9, r12
+       mov     r8, 0
+       andl    r9, 3, COH
+       brne    5f
+
+1:     sub     r11, 4
+       brlt    2f
+
+10:    st.w    r12++, r8
+       sub     r11, 4
+       brge    10b
+
+2:     sub     r11, -4
+       reteq   0
+
+       /* Unaligned count or address */
+       bld     r11, 1
+       brcc    12f
+11:    st.h    r12++, r8
+       sub     r11, 2
+       reteq   0
+12:    st.b    r12++, r8
+       retal   0
+
+       /* Unaligned address */
+5:     cp.w    r11, 4
+       brlt    2b
+
+       lsl     r9, 2
+       add     pc, pc, r9
+13:    st.b    r12++, r8
+       sub     r11, 1
+14:    st.b    r12++, r8
+       sub     r11, 1
+15:    st.b    r12++, r8
+       sub     r11, 1
+       rjmp    1b
+
+       .size   clear_user, . - clear_user
+       .size   __clear_user, . - __clear_user
+
+       .section .fixup, "ax"
+       .align  1
+18:    sub     r11, -4
+19:    retal   r11
+
+       .section __ex_table, "a"
+       .align  2
+       .long   10b, 18b
+       .long   11b, 19b
+       .long   12b, 19b
+       .long   13b, 19b
+       .long   14b, 19b
+       .long   15b, 19b
diff --git a/arch/avr32/lib/copy_user.S b/arch/avr32/lib/copy_user.S
new file mode 100644 (file)
index 0000000..ea59c04
--- /dev/null
@@ -0,0 +1,119 @@
+/*
+ * Copy to/from userspace with optional address space checking.
+ *
+ * Copyright 2004-2006 Atmel Corporation
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ */
+#include <asm/page.h>
+#include <asm/thread_info.h>
+#include <asm/asm.h>
+
+       /*
+        * __kernel_size_t
+        * __copy_user(void *to, const void *from, __kernel_size_t n)
+        *
+        * Returns the number of bytes not copied. Might be off by
+        * max 3 bytes if we get a fault in the main loop.
+        *
+        * The address-space checking functions simply fall through to
+        * the non-checking version.
+        */
+       .text
+       .align  1
+       .global copy_from_user
+       .type   copy_from_user, @function
+copy_from_user:
+       branch_if_kernel r8, __copy_user
+       ret_if_privileged r8, r11, r10, r10
+       rjmp    __copy_user
+       .size   copy_from_user, . - copy_from_user
+
+       .global copy_to_user
+       .type   copy_to_user, @function
+copy_to_user:
+       branch_if_kernel r8, __copy_user
+       ret_if_privileged r8, r12, r10, r10
+       .size   copy_to_user, . - copy_to_user
+
+       .global __copy_user
+       .type   __copy_user, @function
+__copy_user:
+       mov     r9, r11
+       andl    r9, 3, COH
+       brne    6f
+
+       /* At this point, from is word-aligned */
+1:     sub     r10, 4
+       brlt    3f
+
+2:
+10:    ld.w    r8, r11++
+11:    st.w    r12++, r8
+       sub     r10, 4
+       brge    2b
+
+3:     sub     r10, -4
+       reteq   0
+
+       /*
+        * Handle unaligned count. Need to be careful with r10 here so
+        * that we return the correct value even if we get a fault
+        */
+4:
+20:    ld.ub   r8, r11++
+21:    st.b    r12++, r8
+       sub     r10, 1
+       reteq   0
+22:    ld.ub   r8, r11++
+23:    st.b    r12++, r8
+       sub     r10, 1
+       reteq   0
+24:    ld.ub   r8, r11++
+25:    st.b    r12++, r8
+       retal   0
+
+       /* Handle unaligned from-pointer */
+6:     cp.w    r10, 4
+       brlt    4b
+       rsub    r9, r9, 4
+
+30:    ld.ub   r8, r11++
+31:    st.b    r12++, r8
+       sub     r10, 1
+       sub     r9, 1
+       breq    1b
+32:    ld.ub   r8, r11++
+33:    st.b    r12++, r8
+       sub     r10, 1
+       sub     r9, 1
+       breq    1b
+34:    ld.ub   r8, r11++
+35:    st.b    r12++, r8
+       sub     r10, 1
+       rjmp    1b
+       .size   __copy_user, . - __copy_user
+
+       .section .fixup,"ax"
+       .align  1
+19:    sub     r10, -4
+29:    retal   r10
+
+       .section __ex_table,"a"
+       .align  2
+       .long   10b, 19b
+       .long   11b, 19b
+       .long   20b, 29b
+       .long   21b, 29b
+       .long   22b, 29b
+       .long   23b, 29b
+       .long   24b, 29b
+       .long   25b, 29b
+       .long   30b, 29b
+       .long   31b, 29b
+       .long   32b, 29b
+       .long   33b, 29b
+       .long   34b, 29b
+       .long   35b, 29b
diff --git a/arch/avr32/lib/csum_partial.S b/arch/avr32/lib/csum_partial.S
new file mode 100644 (file)
index 0000000..6a262b5
--- /dev/null
@@ -0,0 +1,47 @@
+/*
+ * Copyright (C) 2004-2006 Atmel Corporation
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ */
+
+       /*
+        * unsigned int csum_partial(const unsigned char *buff,
+        *                           int len, unsigned int sum)
+        */
+       .text
+       .global csum_partial
+       .type   csum_partial,"function"
+       .align  1
+csum_partial:
+       /* checksum complete words, aligned or not */
+3:     sub     r11, 4
+       brlt    5f
+4:     ld.w    r9, r12++
+       add     r10, r9
+       acr     r10
+       sub     r11, 4
+       brge    4b
+
+       /* return if we had a whole number of words */
+5:     sub     r11, -4
+       reteq   r10
+
+       /* checksum any remaining bytes at the end */
+       mov     r9, 0
+       mov     r8, 0
+       cp      r11, 2
+       brlt    6f
+       ld.uh   r9, r12++
+       sub     r11, 2
+       breq    7f
+       lsl     r9, 16
+6:     ld.ub   r8, r12++
+       lsl     r8, 8
+7:     or      r9, r8
+       add     r10, r9
+       acr     r10
+
+       retal   r10
+       .size   csum_partial, . - csum_partial
diff --git a/arch/avr32/lib/csum_partial_copy_generic.S b/arch/avr32/lib/csum_partial_copy_generic.S
new file mode 100644 (file)
index 0000000..a3a0f9b
--- /dev/null
@@ -0,0 +1,99 @@
+/*
+ * Copyright (C) 2004-2006 Atmel Corporation
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ */
+#include <asm/errno.h>
+#include <asm/asm.h>
+
+       /*
+        * unsigned int csum_partial_copy_generic(const char *src, char *dst, int len
+        *                                        int sum, int *src_err_ptr,
+        *                                        int *dst_err_ptr)
+        *
+        * Copy src to dst while checksumming, otherwise like csum_partial.
+        */
+
+       .macro ld_src size, reg, ptr
+9999:  ld.\size \reg, \ptr
+       .section __ex_table, "a"
+       .long   9999b, fixup_ld_src
+       .previous
+       .endm
+
+       .macro st_dst size, ptr, reg
+9999:  st.\size \ptr, \reg
+       .section __ex_table, "a"
+       .long   9999b, fixup_st_dst
+       .previous
+       .endm
+
+       .text
+       .global csum_partial_copy_generic
+       .type   csum_partial_copy_generic,"function"
+       .align  1
+csum_partial_copy_generic:
+       pushm   r4-r7,lr
+
+       /* The inner loop */
+1:     sub     r10, 4
+       brlt    5f
+2:     ld_src  w, r5, r12++
+       st_dst  w, r11++, r5
+       add     r9, r5
+       acr     r9
+       sub     r10, 4
+       brge    2b
+
+       /* return if we had a whole number of words */
+5:     sub     r10, -4
+       brne    7f
+
+6:     mov     r12, r9
+       popm    r4-r7,pc
+
+       /* handle additional bytes at the tail */
+7:     mov     r5, 0
+       mov     r4, 32
+8:     ld_src  ub, r6, r12++
+       st_dst  b, r11++, r6
+       lsl     r5, 8
+       sub     r4, 8
+       bfins   r5, r6, 0, 8
+       sub     r10, 1
+       brne    8b
+
+       lsl     r5, r5, r4
+       add     r9, r5
+       acr     r9
+       rjmp    6b
+
+       /* Exception handler */
+       .section .fixup,"ax"
+       .align  1
+fixup_ld_src:
+       mov     r9, -EFAULT
+       cp.w    r8, 0
+       breq    1f
+       st.w    r8[0], r9
+
+1:     /*
+        * TODO: zero the complete destination - computing the rest
+        * is too much work
+        */
+
+       mov     r9, 0
+       rjmp    6b
+
+fixup_st_dst:
+       mov     r9, -EFAULT
+       lddsp   r8, sp[20]
+       cp.w    r8, 0
+       breq    1f
+       st.w    r8[0], r9
+1:     mov     r9, 0
+       rjmp    6b
+
+       .previous
diff --git a/arch/avr32/lib/delay.c b/arch/avr32/lib/delay.c
new file mode 100644 (file)
index 0000000..462c830
--- /dev/null
@@ -0,0 +1,55 @@
+/*
+ *      Precise Delay Loops for avr32
+ *
+ *      Copyright (C) 1993 Linus Torvalds
+ *      Copyright (C) 1997 Martin Mares <mj@atrey.karlin.mff.cuni.cz>
+ *     Copyright (C) 2005-2006 Atmel Corporation
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ */
+
+#include <linux/delay.h>
+#include <linux/module.h>
+#include <linux/types.h>
+
+#include <asm/delay.h>
+#include <asm/processor.h>
+#include <asm/sysreg.h>
+
+int read_current_timer(unsigned long *timer_value)
+{
+       *timer_value = sysreg_read(COUNT);
+       return 0;
+}
+
+void __delay(unsigned long loops)
+{
+       unsigned bclock, now;
+
+       bclock = sysreg_read(COUNT);
+       do {
+               now = sysreg_read(COUNT);
+       } while ((now - bclock) < loops);
+}
+
+inline void __const_udelay(unsigned long xloops)
+{
+       unsigned long long loops;
+
+       asm("mulu.d %0, %1, %2"
+           : "=r"(loops)
+           : "r"(current_cpu_data.loops_per_jiffy * HZ), "r"(xloops));
+       __delay(loops >> 32);
+}
+
+void __udelay(unsigned long usecs)
+{
+       __const_udelay(usecs * 0x000010c7); /* 2**32 / 1000000 (rounded up) */
+}
+
+void __ndelay(unsigned long nsecs)
+{
+       __const_udelay(nsecs * 0x00005); /* 2**32 / 1000000000 (rounded up) */
+}
diff --git a/arch/avr32/lib/findbit.S b/arch/avr32/lib/findbit.S
new file mode 100644 (file)
index 0000000..2b4856f
--- /dev/null
@@ -0,0 +1,154 @@
+/*
+ * Copyright (C) 2006 Atmel Corporation
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ */
+#include <linux/linkage.h>
+
+       .text
+       /*
+        * unsigned long find_first_zero_bit(const unsigned long *addr,
+        *                                   unsigned long size)
+        */
+ENTRY(find_first_zero_bit)
+       cp.w    r11, 0
+       reteq   r11
+       mov     r9, r11
+1:     ld.w    r8, r12[0]
+       com     r8
+       brne    .L_found
+       sub     r12, -4
+       sub     r9, 32
+       brgt    1b
+       retal   r11
+
+       /*
+        * unsigned long find_next_zero_bit(const unsigned long *addr,
+        *                                  unsigned long size,
+        *                                  unsigned long offset)
+        */
+ENTRY(find_next_zero_bit)
+       lsr     r8, r10, 5
+       sub     r9, r11, r10
+       retle   r11
+
+       lsl     r8, 2
+       add     r12, r8
+       andl    r10, 31, COH
+       breq    1f
+
+       /* offset is not word-aligned. Handle the first (32 - r10) bits */
+       ld.w    r8, r12[0]
+       com     r8
+       sub     r12, -4
+       lsr     r8, r8, r10
+       brne    .L_found
+
+       /* r9 = r9 - (32 - r10) = r9 + r10 - 32 */
+       add     r9, r10
+       sub     r9, 32
+       retle   r11
+
+       /* Main loop. offset must be word-aligned */
+1:     ld.w    r8, r12[0]
+       com     r8
+       brne    .L_found
+       sub     r12, -4
+       sub     r9, 32
+       brgt    1b
+       retal   r11
+
+       /* Common return path for when a bit is actually found. */
+.L_found:
+       brev    r8
+       clz     r10, r8
+       rsub    r9, r11
+       add     r10, r9
+
+       /* XXX: If we don't have to return exactly "size" when the bit
+          is not found, we may drop this "min" thing */
+       min     r12, r11, r10
+       retal   r12
+
+       /*
+        * unsigned long find_first_bit(const unsigned long *addr,
+        *                              unsigned long size)
+        */
+ENTRY(find_first_bit)
+       cp.w    r11, 0
+       reteq   r11
+       mov     r9, r11
+1:     ld.w    r8, r12[0]
+       cp.w    r8, 0
+       brne    .L_found
+       sub     r12, -4
+       sub     r9, 32
+       brgt    1b
+       retal   r11
+
+       /*
+        * unsigned long find_next_bit(const unsigned long *addr,
+        *                             unsigned long size,
+        *                             unsigned long offset)
+        */
+ENTRY(find_next_bit)
+       lsr     r8, r10, 5
+       sub     r9, r11, r10
+       retle   r11
+
+       lsl     r8, 2
+       add     r12, r8
+       andl    r10, 31, COH
+       breq    1f
+
+       /* offset is not word-aligned. Handle the first (32 - r10) bits */
+       ld.w    r8, r12[0]
+       sub     r12, -4
+       lsr     r8, r8, r10
+       brne    .L_found
+
+       /* r9 = r9 - (32 - r10) = r9 + r10 - 32 */
+       add     r9, r10
+       sub     r9, 32
+       retle   r11
+
+       /* Main loop. offset must be word-aligned */
+1:     ld.w    r8, r12[0]
+       cp.w    r8, 0
+       brne    .L_found
+       sub     r12, -4
+       sub     r9, 32
+       brgt    1b
+       retal   r11
+
+ENTRY(generic_find_next_zero_le_bit)
+       lsr     r8, r10, 5
+       sub     r9, r11, r10
+       retle   r11
+
+       lsl     r8, 2
+       add     r12, r8
+       andl    r10, 31, COH
+       breq    1f
+
+       /* offset is not word-aligned. Handle the first (32 - r10) bits */
+       ldswp.w r8, r12[0]
+       sub     r12, -4
+       lsr     r8, r8, r10
+       brne    .L_found
+
+       /* r9 = r9 - (32 - r10) = r9 + r10 - 32 */
+       add     r9, r10
+       sub     r9, 32
+       retle   r11
+
+       /* Main loop. offset must be word-aligned */
+1:     ldswp.w r8, r12[0]
+       cp.w    r8, 0
+       brne    .L_found
+       sub     r12, -4
+       sub     r9, 32
+       brgt    1b
+       retal   r11
diff --git a/arch/avr32/lib/io-readsl.S b/arch/avr32/lib/io-readsl.S
new file mode 100644 (file)
index 0000000..b103511
--- /dev/null
@@ -0,0 +1,24 @@
+/*
+ * Copyright (C) 2004-2006 Atmel Corporation
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ */
+
+       .global __raw_readsl
+       .type   __raw_readsl,@function
+__raw_readsl:
+       cp.w    r10, 0
+       reteq   r12
+
+       /*
+        * If r11 isn't properly aligned, we might get an exception on
+        * some implementations. But there's not much we can do about it.
+        */
+1:     ld.w    r8, r12[0]
+       sub     r10, 1
+       st.w    r11++, r8
+       brne    1b
+
+       retal   r12
diff --git a/arch/avr32/lib/io-readsw.S b/arch/avr32/lib/io-readsw.S
new file mode 100644 (file)
index 0000000..456be99
--- /dev/null
@@ -0,0 +1,43 @@
+/*
+ * Copyright (C) 2004-2006 Atmel Corporation
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ */
+
+.Lnot_word_aligned:
+       /*
+        * Bad alignment will cause a hardware exception, which is as
+        * good as anything. No need for us to check for proper alignment.
+        */
+       ld.uh   r8, r12[0]
+       sub     r10, 1
+       st.h    r11++, r8
+
+       /* fall through */
+
+       .global __raw_readsw
+       .type   __raw_readsw,@function
+__raw_readsw:
+       cp.w    r10, 0
+       reteq   r12
+       mov     r9, 3
+       tst     r11, r9
+       brne    .Lnot_word_aligned
+
+       sub     r10, 2
+       brlt    2f
+
+1:     ldins.h r8:t, r12[0]
+       ldins.h r8:b, r12[0]
+       st.w    r11++, r8
+       sub     r10, 2
+       brge    1b
+
+2:     sub     r10, -2
+       reteq   r12
+
+       ld.uh   r8, r12[0]
+       st.h    r11++, r8
+       retal   r12
diff --git a/arch/avr32/lib/io-writesl.S b/arch/avr32/lib/io-writesl.S
new file mode 100644 (file)
index 0000000..22138b3
--- /dev/null
@@ -0,0 +1,20 @@
+/*
+ * Copyright (C) 2004-2006 Atmel Corporation
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ */
+
+       .global __raw_writesl
+       .type   __raw_writesl,@function
+__raw_writesl:
+       cp.w    r10, 0
+       reteq   r12
+
+1:     ld.w    r8, r11++
+       sub     r10, 1
+       st.w    r12[0], r8
+       brne    1b
+
+       retal   r12
diff --git a/arch/avr32/lib/io-writesw.S b/arch/avr32/lib/io-writesw.S
new file mode 100644 (file)
index 0000000..8c4a53f
--- /dev/null
@@ -0,0 +1,38 @@
+/*
+ * Copyright (C) 2004-2006 Atmel Corporation
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ */
+
+.Lnot_word_aligned:
+       ld.uh   r8, r11++
+       sub     r10, 1
+       st.h    r12[0], r8
+
+       .global __raw_writesw
+       .type   __raw_writesw,@function
+__raw_writesw:
+       cp.w    r10, 0
+       mov     r9, 3
+       reteq   r12
+       tst     r11, r9
+       brne    .Lnot_word_aligned
+
+       sub     r10, 2
+       brlt    2f
+
+1:     ld.w    r8, r11++
+       bfextu  r9, r8, 16, 16
+       st.h    r12[0], r9
+       st.h    r12[0], r8
+       sub     r10, 2
+       brge    1b
+
+2:     sub     r10, -2
+       reteq   r12
+
+       ld.uh   r8, r11++
+       st.h    r12[0], r8
+       retal   r12
diff --git a/arch/avr32/lib/libgcc.h b/arch/avr32/lib/libgcc.h
new file mode 100644 (file)
index 0000000..5a091b5
--- /dev/null
@@ -0,0 +1,33 @@
+/* Definitions for various functions 'borrowed' from gcc-3.4.3 */
+
+#define BITS_PER_UNIT  8
+
+typedef                 int QItype     __attribute__ ((mode (QI)));
+typedef unsigned int UQItype   __attribute__ ((mode (QI)));
+typedef                 int HItype     __attribute__ ((mode (HI)));
+typedef unsigned int UHItype   __attribute__ ((mode (HI)));
+typedef         int SItype     __attribute__ ((mode (SI)));
+typedef unsigned int USItype   __attribute__ ((mode (SI)));
+typedef                 int DItype     __attribute__ ((mode (DI)));
+typedef unsigned int UDItype   __attribute__ ((mode (DI)));
+typedef        float SFtype    __attribute__ ((mode (SF)));
+typedef                float DFtype    __attribute__ ((mode (DF)));
+typedef int word_type __attribute__ ((mode (__word__)));
+
+#define W_TYPE_SIZE (4 * BITS_PER_UNIT)
+#define Wtype  SItype
+#define UWtype USItype
+#define HWtype SItype
+#define UHWtype        USItype
+#define DWtype DItype
+#define UDWtype        UDItype
+#define __NW(a,b)      __ ## a ## si ## b
+#define __NDW(a,b)     __ ## a ## di ## b
+
+struct DWstruct {Wtype high, low;};
+
+typedef union
+{
+  struct DWstruct s;
+  DWtype ll;
+} DWunion;
diff --git a/arch/avr32/lib/longlong.h b/arch/avr32/lib/longlong.h
new file mode 100644 (file)
index 0000000..cd5e369
--- /dev/null
@@ -0,0 +1,98 @@
+/* longlong.h -- definitions for mixed size 32/64 bit arithmetic.
+   Copyright (C) 1991, 1992, 1994, 1995, 1996, 1997, 1998, 1999, 2000
+   Free Software Foundation, Inc.
+
+   This definition file is free software; you can redistribute it
+   and/or modify it under the terms of the GNU General Public
+   License as published by the Free Software Foundation; either
+   version 2, or (at your option) any later version.
+
+   This definition file is distributed in the hope that it will be
+   useful, but WITHOUT ANY WARRANTY; without even the implied
+   warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
+   See the GNU General Public License for more details.
+
+   You should have received a copy of the GNU General Public License
+   along with this program; if not, write to the Free Software
+   Foundation, Inc., 59 Temple Place - Suite 330,
+   Boston, MA 02111-1307, USA.  */
+
+/* Borrowed from gcc-3.4.3 */
+
+#define __BITS4 (W_TYPE_SIZE / 4)
+#define __ll_B ((UWtype) 1 << (W_TYPE_SIZE / 2))
+#define __ll_lowpart(t) ((UWtype) (t) & (__ll_B - 1))
+#define __ll_highpart(t) ((UWtype) (t) >> (W_TYPE_SIZE / 2))
+
+#define count_leading_zeros(count, x) ((count) = __builtin_clz(x))
+
+#define __udiv_qrnnd_c(q, r, n1, n0, d) \
+  do {                                                                 \
+    UWtype __d1, __d0, __q1, __q0;                                     \
+    UWtype __r1, __r0, __m;                                            \
+    __d1 = __ll_highpart (d);                                          \
+    __d0 = __ll_lowpart (d);                                           \
+                                                                       \
+    __r1 = (n1) % __d1;                                                        \
+    __q1 = (n1) / __d1;                                                        \
+    __m = (UWtype) __q1 * __d0;                                                \
+    __r1 = __r1 * __ll_B | __ll_highpart (n0);                         \
+    if (__r1 < __m)                                                    \
+      {                                                                        \
+       __q1--, __r1 += (d);                                            \
+       if (__r1 >= (d)) /* i.e. we didn't get carry when adding to __r1 */\
+         if (__r1 < __m)                                               \
+           __q1--, __r1 += (d);                                        \
+      }                                                                        \
+    __r1 -= __m;                                                       \
+                                                                       \
+    __r0 = __r1 % __d1;                                                        \
+    __q0 = __r1 / __d1;                                                        \
+    __m = (UWtype) __q0 * __d0;                                                \
+    __r0 = __r0 * __ll_B | __ll_lowpart (n0);                          \
+    if (__r0 < __m)                                                    \
+      {                                                                        \
+       __q0--, __r0 += (d);                                            \
+       if (__r0 >= (d))                                                \
+         if (__r0 < __m)                                               \
+           __q0--, __r0 += (d);                                        \
+      }                                                                        \
+    __r0 -= __m;                                                       \
+                                                                       \
+    (q) = (UWtype) __q1 * __ll_B | __q0;                               \
+    (r) = __r0;                                                                \
+  } while (0)
+
+#define udiv_qrnnd __udiv_qrnnd_c
+
+#define sub_ddmmss(sh, sl, ah, al, bh, bl) \
+  do {                                                                 \
+    UWtype __x;                                                                \
+    __x = (al) - (bl);                                                 \
+    (sh) = (ah) - (bh) - (__x > (al));                                 \
+    (sl) = __x;                                                                \
+  } while (0)
+
+#define umul_ppmm(w1, w0, u, v)                                                \
+  do {                                                                 \
+    UWtype __x0, __x1, __x2, __x3;                                     \
+    UHWtype __ul, __vl, __uh, __vh;                                    \
+                                                                       \
+    __ul = __ll_lowpart (u);                                           \
+    __uh = __ll_highpart (u);                                          \
+    __vl = __ll_lowpart (v);                                           \
+    __vh = __ll_highpart (v);                                          \
+                                                                       \
+    __x0 = (UWtype) __ul * __vl;                                       \
+    __x1 = (UWtype) __ul * __vh;                                       \
+    __x2 = (UWtype) __uh * __vl;                                       \
+    __x3 = (UWtype) __uh * __vh;                                       \
+                                                                       \
+    __x1 += __ll_highpart (__x0);/* this can't give carry */           \
+    __x1 += __x2;              /* but this indeed can */               \
+    if (__x1 < __x2)           /* did we get it? */                    \
+      __x3 += __ll_B;          /* yes, add it in the proper pos.  */   \
+                                                                       \
+    (w1) = __x3 + __ll_highpart (__x1);                                        \
+    (w0) = __ll_lowpart (__x1) * __ll_B + __ll_lowpart (__x0);         \
+  } while (0)
diff --git a/arch/avr32/lib/memcpy.S b/arch/avr32/lib/memcpy.S
new file mode 100644 (file)
index 0000000..0abb261
--- /dev/null
@@ -0,0 +1,62 @@
+/*
+ * Copyright (C) 2004-2006 Atmel Corporation
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ */
+
+       /*
+        * void *memcpy(void *to, const void *from, unsigned long n)
+        *
+        * This implementation does word-aligned loads in the main loop,
+        * possibly sacrificing alignment of stores.
+        *
+        * Hopefully, in most cases, both "to" and "from" will be
+        * word-aligned to begin with.
+        */
+       .text
+       .global memcpy
+       .type   memcpy, @function
+memcpy:
+       mov     r9, r11
+       andl    r9, 3, COH
+       brne    1f
+
+       /* At this point, "from" is word-aligned */
+2:     sub     r10, 4
+       mov     r9, r12
+       brlt    4f
+
+3:     ld.w    r8, r11++
+       sub     r10, 4
+       st.w    r12++, r8
+       brge    3b
+
+4:     neg     r10
+       reteq   r9
+
+       /* Handle unaligned count */
+       lsl     r10, 2
+       add     pc, pc, r10
+       ld.ub   r8, r11++
+       st.b    r12++, r8
+       ld.ub   r8, r11++
+       st.b    r12++, r8
+       ld.ub   r8, r11++
+       st.b    r12++, r8
+       retal   r9
+
+       /* Handle unaligned "from" pointer */
+1:     sub     r10, 4
+       brlt    4b
+       add     r10, r9
+       lsl     r9, 2
+       add     pc, pc, r9
+       ld.ub   r8, r11++
+       st.b    r12++, r8
+       ld.ub   r8, r11++
+       st.b    r12++, r8
+       ld.ub   r8, r11++
+       st.b    r12++, r8
+       rjmp    2b
diff --git a/arch/avr32/lib/memset.S b/arch/avr32/lib/memset.S
new file mode 100644 (file)
index 0000000..40da32c
--- /dev/null
@@ -0,0 +1,72 @@
+/*
+ * Copyright (C) 2004-2006 Atmel Corporation
+ *
+ * Based on linux/arch/arm/lib/memset.S
+ *   Copyright (C) 1995-2000 Russell King
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ *
+ * ASM optimised string functions
+ */
+#include <asm/asm.h>
+
+       /*
+        * r12: void *b
+        * r11: int c
+        * r10: size_t len
+        *
+        * Returns b in r12
+        */
+       .text
+       .global memset
+       .type   memset, @function
+       .align  5
+memset:
+       mov     r9, r12
+       mov     r8, r12
+       or      r11, r11, r11 << 8
+       andl    r9, 3, COH
+       brne    1f
+
+2:     or      r11, r11, r11 << 16
+       sub     r10, 4
+       brlt    5f
+
+       /* Let's do some real work */
+4:     st.w    r8++, r11
+       sub     r10, 4
+       brge    4b
+
+       /*
+        * When we get here, we've got less than 4 bytes to set. r10
+        * might be negative.
+        */
+5:     sub     r10, -4
+       reteq   r12
+
+       /* Fastpath ends here, exactly 32 bytes from memset */
+
+       /* Handle unaligned count or pointer */
+       bld     r10, 1
+       brcc    6f
+       st.b    r8++, r11
+       st.b    r8++, r11
+       bld     r10, 0
+       retcc   r12
+6:     st.b    r8++, r11
+       retal   r12
+
+       /* Handle unaligned pointer */
+1:     sub     r10, 4
+       brlt    5b
+       add     r10, r9
+       lsl     r9, 1
+       add     pc, r9
+       st.b    r8++, r11
+       st.b    r8++, r11
+       st.b    r8++, r11
+       rjmp    2b
+
+       .size   memset, . - memset
diff --git a/arch/avr32/lib/strncpy_from_user.S b/arch/avr32/lib/strncpy_from_user.S
new file mode 100644 (file)
index 0000000..72bd505
--- /dev/null
@@ -0,0 +1,60 @@
+/*
+ * Copy to/from userspace with optional address space checking.
+ *
+ * Copyright 2004-2006 Atmel Corporation
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ */
+#include <linux/errno.h>
+
+#include <asm/page.h>
+#include <asm/thread_info.h>
+#include <asm/asm.h>
+
+       /*
+        * long strncpy_from_user(char *dst, const char *src, long count)
+        *
+        * On success, returns the length of the string, not including
+        * the terminating NUL.
+        *
+        * If the string is longer than count, returns count
+        *
+        * If userspace access fails, returns -EFAULT
+        */
+       .text
+       .align  1
+       .global strncpy_from_user
+       .type   strncpy_from_user, "function"
+strncpy_from_user:
+       mov     r9, -EFAULT
+       branch_if_kernel r8, __strncpy_from_user
+       ret_if_privileged r8, r11, r10, r9
+
+       .global __strncpy_from_user
+       .type   __strncpy_from_user, "function"
+__strncpy_from_user:
+       cp.w    r10, 0
+       reteq   0
+
+       mov     r9, r10
+
+1:     ld.ub   r8, r11++
+       st.b    r12++, r8
+       cp.w    r8, 0
+       breq    2f
+       sub     r9, 1
+       brne    1b
+
+2:     sub     r10, r9
+       retal   r10
+
+       .section .fixup, "ax"
+       .align  1
+3:     mov     r12, -EFAULT
+       retal   r12
+
+       .section __ex_table, "a"
+       .align  2
+       .long   1b, 3b
diff --git a/arch/avr32/lib/strnlen_user.S b/arch/avr32/lib/strnlen_user.S
new file mode 100644 (file)
index 0000000..65ce11a
--- /dev/null
@@ -0,0 +1,67 @@
+/*
+ * Copy to/from userspace with optional address space checking.
+ *
+ * Copyright 2004-2006 Atmel Corporation
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ */
+#include <asm/page.h>
+#include <asm/thread_info.h>
+#include <asm/processor.h>
+#include <asm/asm.h>
+
+       .text
+       .align  1
+       .global strnlen_user
+       .type   strnlen_user, "function"
+strnlen_user:
+       branch_if_kernel r8, __strnlen_user
+       sub     r8, r11, 1
+       add     r8, r12
+       retcs   0
+       brmi    adjust_length   /* do a closer inspection */
+
+       .global __strnlen_user
+       .type   __strnlen_user, "function"
+__strnlen_user:
+       mov     r10, r12
+
+10:    ld.ub   r8, r12++
+       cp.w    r8, 0
+       breq    2f
+       sub     r11, 1
+       brne    10b
+
+       sub     r12, -1
+2:     sub     r12, r10
+       retal   r12
+
+
+       .type   adjust_length, "function"
+adjust_length:
+       cp.w    r12, 0          /* addr must always be < TASK_SIZE */
+       retmi   0
+
+       pushm   lr
+       lddpc   lr, _task_size
+       sub     r11, lr, r12
+       mov     r9, r11
+       rcall   __strnlen_user
+       cp.w    r12, r9
+       brgt    1f
+       popm    pc
+1:     popm    pc, r12=0
+
+       .align  2
+_task_size:
+       .long   TASK_SIZE
+
+       .section .fixup, "ax"
+       .align  1
+19:    retal   0
+
+       .section __ex_table, "a"
+       .align  2
+       .long   10b, 19b
diff --git a/arch/avr32/mach-at32ap/Makefile b/arch/avr32/mach-at32ap/Makefile
new file mode 100644 (file)
index 0000000..4b10853
--- /dev/null
@@ -0,0 +1,2 @@
+obj-y                          += at32ap.o clock.o pio.o intc.o extint.o
+obj-$(CONFIG_CPU_AT32AP7000)   += at32ap7000.o
diff --git a/arch/avr32/mach-at32ap/at32ap.c b/arch/avr32/mach-at32ap/at32ap.c
new file mode 100644 (file)
index 0000000..f7cedf5
--- /dev/null
@@ -0,0 +1,90 @@
+/*
+ * Copyright (C) 2006 Atmel Corporation
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ */
+
+#include <linux/clk.h>
+#include <linux/err.h>
+#include <linux/init.h>
+#include <linux/platform_device.h>
+
+#include <asm/io.h>
+
+#include <asm/arch/init.h>
+#include <asm/arch/sm.h>
+
+struct at32_sm system_manager;
+
+static int __init at32_sm_init(void)
+{
+       struct resource *regs;
+       struct at32_sm *sm = &system_manager;
+       int ret = -ENXIO;
+
+       regs = platform_get_resource(&at32_sm_device, IORESOURCE_MEM, 0);
+       if (!regs)
+               goto fail;
+
+       spin_lock_init(&sm->lock);
+       sm->pdev = &at32_sm_device;
+
+       ret = -ENOMEM;
+       sm->regs = ioremap(regs->start, regs->end - regs->start + 1);
+       if (!sm->regs)
+               goto fail;
+
+       return 0;
+
+fail:
+       printk(KERN_ERR "Failed to initialize System Manager: %d\n", ret);
+       return ret;
+}
+
+void __init setup_platform(void)
+{
+       at32_sm_init();
+       at32_clock_init();
+       at32_portmux_init();
+
+       /* FIXME: This doesn't belong here */
+       at32_setup_serial_console(1);
+}
+
+static int __init pdc_probe(struct platform_device *pdev)
+{
+       struct clk *pclk, *hclk;
+
+       pclk = clk_get(&pdev->dev, "pclk");
+       if (IS_ERR(pclk)) {
+               dev_err(&pdev->dev, "no pclk defined\n");
+               return PTR_ERR(pclk);
+       }
+       hclk = clk_get(&pdev->dev, "hclk");
+       if (IS_ERR(hclk)) {
+               dev_err(&pdev->dev, "no hclk defined\n");
+               clk_put(pclk);
+               return PTR_ERR(hclk);
+       }
+
+       clk_enable(pclk);
+       clk_enable(hclk);
+
+       dev_info(&pdev->dev, "Atmel Peripheral DMA Controller enabled\n");
+       return 0;
+}
+
+static struct platform_driver pdc_driver = {
+       .probe          = pdc_probe,
+       .driver         = {
+               .name   = "pdc",
+       },
+};
+
+static int __init pdc_init(void)
+{
+       return platform_driver_register(&pdc_driver);
+}
+arch_initcall(pdc_init);
diff --git a/arch/avr32/mach-at32ap/at32ap7000.c b/arch/avr32/mach-at32ap/at32ap7000.c
new file mode 100644 (file)
index 0000000..e8c6893
--- /dev/null
@@ -0,0 +1,866 @@
+/*
+ * Copyright (C) 2005-2006 Atmel Corporation
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ */
+#include <linux/clk.h>
+#include <linux/init.h>
+#include <linux/platform_device.h>
+
+#include <asm/io.h>
+
+#include <asm/arch/board.h>
+#include <asm/arch/portmux.h>
+#include <asm/arch/sm.h>
+
+#include "clock.h"
+#include "pio.h"
+#include "sm.h"
+
+#define PBMEM(base)                                    \
+       {                                               \
+               .start          = base,                 \
+               .end            = base + 0x3ff,         \
+               .flags          = IORESOURCE_MEM,       \
+       }
+#define IRQ(num)                                       \
+       {                                               \
+               .start          = num,                  \
+               .end            = num,                  \
+               .flags          = IORESOURCE_IRQ,       \
+       }
+#define NAMED_IRQ(num, _name)                          \
+       {                                               \
+               .start          = num,                  \
+               .end            = num,                  \
+               .name           = _name,                \
+               .flags          = IORESOURCE_IRQ,       \
+       }
+
+#define DEFINE_DEV(_name, _id)                                 \
+static struct platform_device _name##_id##_device = {          \
+       .name           = #_name,                               \
+       .id             = _id,                                  \
+       .resource       = _name##_id##_resource,                \
+       .num_resources  = ARRAY_SIZE(_name##_id##_resource),    \
+}
+#define DEFINE_DEV_DATA(_name, _id)                            \
+static struct platform_device _name##_id##_device = {          \
+       .name           = #_name,                               \
+       .id             = _id,                                  \
+       .dev            = {                                     \
+               .platform_data  = &_name##_id##_data,           \
+       },                                                      \
+       .resource       = _name##_id##_resource,                \
+       .num_resources  = ARRAY_SIZE(_name##_id##_resource),    \
+}
+
+#define DEV_CLK(_name, devname, bus, _index)                   \
+static struct clk devname##_##_name = {                                \
+       .name           = #_name,                               \
+       .dev            = &devname##_device.dev,                \
+       .parent         = &bus##_clk,                           \
+       .mode           = bus##_clk_mode,                       \
+       .get_rate       = bus##_clk_get_rate,                   \
+       .index          = _index,                               \
+}
+
+enum {
+       PIOA,
+       PIOB,
+       PIOC,
+       PIOD,
+};
+
+enum {
+       FUNC_A,
+       FUNC_B,
+};
+
+unsigned long at32ap7000_osc_rates[3] = {
+       [0] = 32768,
+       /* FIXME: these are ATSTK1002-specific */
+       [1] = 20000000,
+       [2] = 12000000,
+};
+
+static unsigned long osc_get_rate(struct clk *clk)
+{
+       return at32ap7000_osc_rates[clk->index];
+}
+
+static unsigned long pll_get_rate(struct clk *clk, unsigned long control)
+{
+       unsigned long div, mul, rate;
+
+       if (!(control & SM_BIT(PLLEN)))
+               return 0;
+
+       div = SM_BFEXT(PLLDIV, control) + 1;
+       mul = SM_BFEXT(PLLMUL, control) + 1;
+
+       rate = clk->parent->get_rate(clk->parent);
+       rate = (rate + div / 2) / div;
+       rate *= mul;
+
+       return rate;
+}
+
+static unsigned long pll0_get_rate(struct clk *clk)
+{
+       u32 control;
+
+       control = sm_readl(&system_manager, PM_PLL0);
+
+       return pll_get_rate(clk, control);
+}
+
+static unsigned long pll1_get_rate(struct clk *clk)
+{
+       u32 control;
+
+       control = sm_readl(&system_manager, PM_PLL1);
+
+       return pll_get_rate(clk, control);
+}
+
+/*
+ * The AT32AP7000 has five primary clock sources: One 32kHz
+ * oscillator, two crystal oscillators and two PLLs.
+ */
+static struct clk osc32k = {
+       .name           = "osc32k",
+       .get_rate       = osc_get_rate,
+       .users          = 1,
+       .index          = 0,
+};
+static struct clk osc0 = {
+       .name           = "osc0",
+       .get_rate       = osc_get_rate,
+       .users          = 1,
+       .index          = 1,
+};
+static struct clk osc1 = {
+       .name           = "osc1",
+       .get_rate       = osc_get_rate,
+       .index          = 2,
+};
+static struct clk pll0 = {
+       .name           = "pll0",
+       .get_rate       = pll0_get_rate,
+       .parent         = &osc0,
+};
+static struct clk pll1 = {
+       .name           = "pll1",
+       .get_rate       = pll1_get_rate,
+       .parent         = &osc0,
+};
+
+/*
+ * The main clock can be either osc0 or pll0.  The boot loader may
+ * have chosen one for us, so we don't really know which one until we
+ * have a look at the SM.
+ */
+static struct clk *main_clock;
+
+/*
+ * Synchronous clocks are generated from the main clock. The clocks
+ * must satisfy the constraint
+ *   fCPU >= fHSB >= fPB
+ * i.e. each clock must not be faster than its parent.
+ */
+static unsigned long bus_clk_get_rate(struct clk *clk, unsigned int shift)
+{
+       return main_clock->get_rate(main_clock) >> shift;
+};
+
+static void cpu_clk_mode(struct clk *clk, int enabled)
+{
+       struct at32_sm *sm = &system_manager;
+       unsigned long flags;
+       u32 mask;
+
+       spin_lock_irqsave(&sm->lock, flags);
+       mask = sm_readl(sm, PM_CPU_MASK);
+       if (enabled)
+               mask |= 1 << clk->index;
+       else
+               mask &= ~(1 << clk->index);
+       sm_writel(sm, PM_CPU_MASK, mask);
+       spin_unlock_irqrestore(&sm->lock, flags);
+}
+
+static unsigned long cpu_clk_get_rate(struct clk *clk)
+{
+       unsigned long cksel, shift = 0;
+
+       cksel = sm_readl(&system_manager, PM_CKSEL);
+       if (cksel & SM_BIT(CPUDIV))
+               shift = SM_BFEXT(CPUSEL, cksel) + 1;
+
+       return bus_clk_get_rate(clk, shift);
+}
+
+static void hsb_clk_mode(struct clk *clk, int enabled)
+{
+       struct at32_sm *sm = &system_manager;
+       unsigned long flags;
+       u32 mask;
+
+       spin_lock_irqsave(&sm->lock, flags);
+       mask = sm_readl(sm, PM_HSB_MASK);
+       if (enabled)
+               mask |= 1 << clk->index;
+       else
+               mask &= ~(1 << clk->index);
+       sm_writel(sm, PM_HSB_MASK, mask);
+       spin_unlock_irqrestore(&sm->lock, flags);
+}
+
+static unsigned long hsb_clk_get_rate(struct clk *clk)
+{
+       unsigned long cksel, shift = 0;
+
+       cksel = sm_readl(&system_manager, PM_CKSEL);
+       if (cksel & SM_BIT(HSBDIV))
+               shift = SM_BFEXT(HSBSEL, cksel) + 1;
+
+       return bus_clk_get_rate(clk, shift);
+}
+
+static void pba_clk_mode(struct clk *clk, int enabled)
+{
+       struct at32_sm *sm = &system_manager;
+       unsigned long flags;
+       u32 mask;
+
+       spin_lock_irqsave(&sm->lock, flags);
+       mask = sm_readl(sm, PM_PBA_MASK);
+       if (enabled)
+               mask |= 1 << clk->index;
+       else
+               mask &= ~(1 << clk->index);
+       sm_writel(sm, PM_PBA_MASK, mask);
+       spin_unlock_irqrestore(&sm->lock, flags);
+}
+
+static unsigned long pba_clk_get_rate(struct clk *clk)
+{
+       unsigned long cksel, shift = 0;
+
+       cksel = sm_readl(&system_manager, PM_CKSEL);
+       if (cksel & SM_BIT(PBADIV))
+               shift = SM_BFEXT(PBASEL, cksel) + 1;
+
+       return bus_clk_get_rate(clk, shift);
+}
+
+static void pbb_clk_mode(struct clk *clk, int enabled)
+{
+       struct at32_sm *sm = &system_manager;
+       unsigned long flags;
+       u32 mask;
+
+       spin_lock_irqsave(&sm->lock, flags);
+       mask = sm_readl(sm, PM_PBB_MASK);
+       if (enabled)
+               mask |= 1 << clk->index;
+       else
+               mask &= ~(1 << clk->index);
+       sm_writel(sm, PM_PBB_MASK, mask);
+       spin_unlock_irqrestore(&sm->lock, flags);
+}
+
+static unsigned long pbb_clk_get_rate(struct clk *clk)
+{
+       unsigned long cksel, shift = 0;
+
+       cksel = sm_readl(&system_manager, PM_CKSEL);
+       if (cksel & SM_BIT(PBBDIV))
+               shift = SM_BFEXT(PBBSEL, cksel) + 1;
+
+       return bus_clk_get_rate(clk, shift);
+}
+
+static struct clk cpu_clk = {
+       .name           = "cpu",
+       .get_rate       = cpu_clk_get_rate,
+       .users          = 1,
+};
+static struct clk hsb_clk = {
+       .name           = "hsb",
+       .parent         = &cpu_clk,
+       .get_rate       = hsb_clk_get_rate,
+};
+static struct clk pba_clk = {
+       .name           = "pba",
+       .parent         = &hsb_clk,
+       .mode           = hsb_clk_mode,
+       .get_rate       = pba_clk_get_rate,
+       .index          = 1,
+};
+static struct clk pbb_clk = {
+       .name           = "pbb",
+       .parent         = &hsb_clk,
+       .mode           = hsb_clk_mode,
+       .get_rate       = pbb_clk_get_rate,
+       .users          = 1,
+       .index          = 2,
+};
+
+/* --------------------------------------------------------------------
+ *  Generic Clock operations
+ * -------------------------------------------------------------------- */
+
+static void genclk_mode(struct clk *clk, int enabled)
+{
+       u32 control;
+
+       BUG_ON(clk->index > 7);
+
+       control = sm_readl(&system_manager, PM_GCCTRL + 4 * clk->index);
+       if (enabled)
+               control |= SM_BIT(CEN);
+       else
+               control &= ~SM_BIT(CEN);
+       sm_writel(&system_manager, PM_GCCTRL + 4 * clk->index, control);
+}
+
+static unsigned long genclk_get_rate(struct clk *clk)
+{
+       u32 control;
+       unsigned long div = 1;
+
+       BUG_ON(clk->index > 7);
+
+       if (!clk->parent)
+               return 0;
+
+       control = sm_readl(&system_manager, PM_GCCTRL + 4 * clk->index);
+       if (control & SM_BIT(DIVEN))
+               div = 2 * (SM_BFEXT(DIV, control) + 1);
+
+       return clk->parent->get_rate(clk->parent) / div;
+}
+
+static long genclk_set_rate(struct clk *clk, unsigned long rate, int apply)
+{
+       u32 control;
+       unsigned long parent_rate, actual_rate, div;
+
+       BUG_ON(clk->index > 7);
+
+       if (!clk->parent)
+               return 0;
+
+       parent_rate = clk->parent->get_rate(clk->parent);
+       control = sm_readl(&system_manager, PM_GCCTRL + 4 * clk->index);
+
+       if (rate > 3 * parent_rate / 4) {
+               actual_rate = parent_rate;
+               control &= ~SM_BIT(DIVEN);
+       } else {
+               div = (parent_rate + rate) / (2 * rate) - 1;
+               control = SM_BFINS(DIV, div, control) | SM_BIT(DIVEN);
+               actual_rate = parent_rate / (2 * (div + 1));
+       }
+
+       printk("clk %s: new rate %lu (actual rate %lu)\n",
+              clk->name, rate, actual_rate);
+
+       if (apply)
+               sm_writel(&system_manager, PM_GCCTRL + 4 * clk->index,
+                         control);
+
+       return actual_rate;
+}
+
+int genclk_set_parent(struct clk *clk, struct clk *parent)
+{
+       u32 control;
+
+       BUG_ON(clk->index > 7);
+
+       printk("clk %s: new parent %s (was %s)\n",
+              clk->name, parent->name,
+              clk->parent ? clk->parent->name : "(null)");
+
+       control = sm_readl(&system_manager, PM_GCCTRL + 4 * clk->index);
+
+       if (parent == &osc1 || parent == &pll1)
+               control |= SM_BIT(OSCSEL);
+       else if (parent == &osc0 || parent == &pll0)
+               control &= ~SM_BIT(OSCSEL);
+       else
+               return -EINVAL;
+
+       if (parent == &pll0 || parent == &pll1)
+               control |= SM_BIT(PLLSEL);
+       else
+               control &= ~SM_BIT(PLLSEL);
+
+       sm_writel(&system_manager, PM_GCCTRL + 4 * clk->index, control);
+       clk->parent = parent;
+
+       return 0;
+}
+
+/* --------------------------------------------------------------------
+ *  System peripherals
+ * -------------------------------------------------------------------- */
+static struct resource sm_resource[] = {
+       PBMEM(0xfff00000),
+       NAMED_IRQ(19, "eim"),
+       NAMED_IRQ(20, "pm"),
+       NAMED_IRQ(21, "rtc"),
+};
+struct platform_device at32_sm_device = {
+       .name           = "sm",
+       .id             = 0,
+       .resource       = sm_resource,
+       .num_resources  = ARRAY_SIZE(sm_resource),
+};
+DEV_CLK(pclk, at32_sm, pbb, 0);
+
+static struct resource intc0_resource[] = {
+       PBMEM(0xfff00400),
+};
+struct platform_device at32_intc0_device = {
+       .name           = "intc",
+       .id             = 0,
+       .resource       = intc0_resource,
+       .num_resources  = ARRAY_SIZE(intc0_resource),
+};
+DEV_CLK(pclk, at32_intc0, pbb, 1);
+
+static struct clk ebi_clk = {
+       .name           = "ebi",
+       .parent         = &hsb_clk,
+       .mode           = hsb_clk_mode,
+       .get_rate       = hsb_clk_get_rate,
+       .users          = 1,
+};
+static struct clk hramc_clk = {
+       .name           = "hramc",
+       .parent         = &hsb_clk,
+       .mode           = hsb_clk_mode,
+       .get_rate       = hsb_clk_get_rate,
+       .users          = 1,
+};
+
+static struct platform_device pdc_device = {
+       .name           = "pdc",
+       .id             = 0,
+};
+DEV_CLK(hclk, pdc, hsb, 4);
+DEV_CLK(pclk, pdc, pba, 16);
+
+static struct clk pico_clk = {
+       .name           = "pico",
+       .parent         = &cpu_clk,
+       .mode           = cpu_clk_mode,
+       .get_rate       = cpu_clk_get_rate,
+       .users          = 1,
+};
+
+/* --------------------------------------------------------------------
+ *  PIO
+ * -------------------------------------------------------------------- */
+
+static struct resource pio0_resource[] = {
+       PBMEM(0xffe02800),
+       IRQ(13),
+};
+DEFINE_DEV(pio, 0);
+DEV_CLK(mck, pio0, pba, 10);
+
+static struct resource pio1_resource[] = {
+       PBMEM(0xffe02c00),
+       IRQ(14),
+};
+DEFINE_DEV(pio, 1);
+DEV_CLK(mck, pio1, pba, 11);
+
+static struct resource pio2_resource[] = {
+       PBMEM(0xffe03000),
+       IRQ(15),
+};
+DEFINE_DEV(pio, 2);
+DEV_CLK(mck, pio2, pba, 12);
+
+static struct resource pio3_resource[] = {
+       PBMEM(0xffe03400),
+       IRQ(16),
+};
+DEFINE_DEV(pio, 3);
+DEV_CLK(mck, pio3, pba, 13);
+
+void __init at32_add_system_devices(void)
+{
+       system_manager.eim_first_irq = NR_INTERNAL_IRQS;
+
+       platform_device_register(&at32_sm_device);
+       platform_device_register(&at32_intc0_device);
+       platform_device_register(&pdc_device);
+
+       platform_device_register(&pio0_device);
+       platform_device_register(&pio1_device);
+       platform_device_register(&pio2_device);
+       platform_device_register(&pio3_device);
+}
+
+/* --------------------------------------------------------------------
+ *  USART
+ * -------------------------------------------------------------------- */
+
+static struct resource usart0_resource[] = {
+       PBMEM(0xffe00c00),
+       IRQ(7),
+};
+DEFINE_DEV(usart, 0);
+DEV_CLK(usart, usart0, pba, 4);
+
+static struct resource usart1_resource[] = {
+       PBMEM(0xffe01000),
+       IRQ(7),
+};
+DEFINE_DEV(usart, 1);
+DEV_CLK(usart, usart1, pba, 4);
+
+static struct resource usart2_resource[] = {
+       PBMEM(0xffe01400),
+       IRQ(8),
+};
+DEFINE_DEV(usart, 2);
+DEV_CLK(usart, usart2, pba, 5);
+
+static struct resource usart3_resource[] = {
+       PBMEM(0xffe01800),
+       IRQ(9),
+};
+DEFINE_DEV(usart, 3);
+DEV_CLK(usart, usart3, pba, 6);
+
+static inline void configure_usart0_pins(void)
+{
+       portmux_set_func(PIOA,  8, FUNC_B);     /* RXD  */
+       portmux_set_func(PIOA,  9, FUNC_B);     /* TXD  */
+}
+
+static inline void configure_usart1_pins(void)
+{
+       portmux_set_func(PIOA, 17, FUNC_A);     /* RXD  */
+       portmux_set_func(PIOA, 18, FUNC_A);     /* TXD  */
+}
+
+static inline void configure_usart2_pins(void)
+{
+       portmux_set_func(PIOB, 26, FUNC_B);     /* RXD  */
+       portmux_set_func(PIOB, 27, FUNC_B);     /* TXD  */
+}
+
+static inline void configure_usart3_pins(void)
+{
+       portmux_set_func(PIOB, 18, FUNC_B);     /* RXD  */
+       portmux_set_func(PIOB, 17, FUNC_B);     /* TXD  */
+}
+
+static struct platform_device *setup_usart(unsigned int id)
+{
+       struct platform_device *pdev;
+
+       switch (id) {
+       case 0:
+               pdev = &usart0_device;
+               configure_usart0_pins();
+               break;
+       case 1:
+               pdev = &usart1_device;
+               configure_usart1_pins();
+               break;
+       case 2:
+               pdev = &usart2_device;
+               configure_usart2_pins();
+               break;
+       case 3:
+               pdev = &usart3_device;
+               configure_usart3_pins();
+               break;
+       default:
+               pdev = NULL;
+               break;
+       }
+
+       return pdev;
+}
+
+struct platform_device *__init at32_add_device_usart(unsigned int id)
+{
+       struct platform_device *pdev;
+
+       pdev = setup_usart(id);
+       if (pdev)
+               platform_device_register(pdev);
+
+       return pdev;
+}
+
+struct platform_device *at91_default_console_device;
+
+void __init at32_setup_serial_console(unsigned int usart_id)
+{
+       at91_default_console_device = setup_usart(usart_id);
+}
+
+/* --------------------------------------------------------------------
+ *  Ethernet
+ * -------------------------------------------------------------------- */
+
+static struct eth_platform_data macb0_data;
+static struct resource macb0_resource[] = {
+       PBMEM(0xfff01800),
+       IRQ(25),
+};
+DEFINE_DEV_DATA(macb, 0);
+DEV_CLK(hclk, macb0, hsb, 8);
+DEV_CLK(pclk, macb0, pbb, 6);
+
+struct platform_device *__init
+at32_add_device_eth(unsigned int id, struct eth_platform_data *data)
+{
+       struct platform_device *pdev;
+
+       switch (id) {
+       case 0:
+               pdev = &macb0_device;
+
+               portmux_set_func(PIOC,  3, FUNC_A);     /* TXD0 */
+               portmux_set_func(PIOC,  4, FUNC_A);     /* TXD1 */
+               portmux_set_func(PIOC,  7, FUNC_A);     /* TXEN */
+               portmux_set_func(PIOC,  8, FUNC_A);     /* TXCK */
+               portmux_set_func(PIOC,  9, FUNC_A);     /* RXD0 */
+               portmux_set_func(PIOC, 10, FUNC_A);     /* RXD1 */
+               portmux_set_func(PIOC, 13, FUNC_A);     /* RXER */
+               portmux_set_func(PIOC, 15, FUNC_A);     /* RXDV */
+               portmux_set_func(PIOC, 16, FUNC_A);     /* MDC  */
+               portmux_set_func(PIOC, 17, FUNC_A);     /* MDIO */
+
+               if (!data->is_rmii) {
+                       portmux_set_func(PIOC,  0, FUNC_A);     /* COL  */
+                       portmux_set_func(PIOC,  1, FUNC_A);     /* CRS  */
+                       portmux_set_func(PIOC,  2, FUNC_A);     /* TXER */
+                       portmux_set_func(PIOC,  5, FUNC_A);     /* TXD2 */
+                       portmux_set_func(PIOC,  6, FUNC_A);     /* TXD3 */
+                       portmux_set_func(PIOC, 11, FUNC_A);     /* RXD2 */
+                       portmux_set_func(PIOC, 12, FUNC_A);     /* RXD3 */
+                       portmux_set_func(PIOC, 14, FUNC_A);     /* RXCK */
+                       portmux_set_func(PIOC, 18, FUNC_A);     /* SPD  */
+               }
+               break;
+
+       default:
+               return NULL;
+       }
+
+       memcpy(pdev->dev.platform_data, data, sizeof(struct eth_platform_data));
+       platform_device_register(pdev);
+
+       return pdev;
+}
+
+/* --------------------------------------------------------------------
+ *  SPI
+ * -------------------------------------------------------------------- */
+static struct resource spi0_resource[] = {
+       PBMEM(0xffe00000),
+       IRQ(3),
+};
+DEFINE_DEV(spi, 0);
+DEV_CLK(mck, spi0, pba, 0);
+
+struct platform_device *__init at32_add_device_spi(unsigned int id)
+{
+       struct platform_device *pdev;
+
+       switch (id) {
+       case 0:
+               pdev = &spi0_device;
+               portmux_set_func(PIOA,  0, FUNC_A);     /* MISO  */
+               portmux_set_func(PIOA,  1, FUNC_A);     /* MOSI  */
+               portmux_set_func(PIOA,  2, FUNC_A);     /* SCK   */
+               portmux_set_func(PIOA,  3, FUNC_A);     /* NPCS0 */
+               portmux_set_func(PIOA,  4, FUNC_A);     /* NPCS1 */
+               portmux_set_func(PIOA,  5, FUNC_A);     /* NPCS2 */
+               break;
+
+       default:
+               return NULL;
+       }
+
+       platform_device_register(pdev);
+       return pdev;
+}
+
+/* --------------------------------------------------------------------
+ *  LCDC
+ * -------------------------------------------------------------------- */
+static struct lcdc_platform_data lcdc0_data;
+static struct resource lcdc0_resource[] = {
+       {
+               .start          = 0xff000000,
+               .end            = 0xff000fff,
+               .flags          = IORESOURCE_MEM,
+       },
+       IRQ(1),
+};
+DEFINE_DEV_DATA(lcdc, 0);
+DEV_CLK(hclk, lcdc0, hsb, 7);
+static struct clk lcdc0_pixclk = {
+       .name           = "pixclk",
+       .dev            = &lcdc0_device.dev,
+       .mode           = genclk_mode,
+       .get_rate       = genclk_get_rate,
+       .set_rate       = genclk_set_rate,
+       .set_parent     = genclk_set_parent,
+       .index          = 7,
+};
+
+struct platform_device *__init
+at32_add_device_lcdc(unsigned int id, struct lcdc_platform_data *data)
+{
+       struct platform_device *pdev;
+
+       switch (id) {
+       case 0:
+               pdev = &lcdc0_device;
+               portmux_set_func(PIOC, 19, FUNC_A);     /* CC     */
+               portmux_set_func(PIOC, 20, FUNC_A);     /* HSYNC  */
+               portmux_set_func(PIOC, 21, FUNC_A);     /* PCLK   */
+               portmux_set_func(PIOC, 22, FUNC_A);     /* VSYNC  */
+               portmux_set_func(PIOC, 23, FUNC_A);     /* DVAL   */
+               portmux_set_func(PIOC, 24, FUNC_A);     /* MODE   */
+               portmux_set_func(PIOC, 25, FUNC_A);     /* PWR    */
+               portmux_set_func(PIOC, 26, FUNC_A);     /* DATA0  */
+               portmux_set_func(PIOC, 27, FUNC_A);     /* DATA1  */
+               portmux_set_func(PIOC, 28, FUNC_A);     /* DATA2  */
+               portmux_set_func(PIOC, 29, FUNC_A);     /* DATA3  */
+               portmux_set_func(PIOC, 30, FUNC_A);     /* DATA4  */
+               portmux_set_func(PIOC, 31, FUNC_A);     /* DATA5  */
+               portmux_set_func(PIOD,  0, FUNC_A);     /* DATA6  */
+               portmux_set_func(PIOD,  1, FUNC_A);     /* DATA7  */
+               portmux_set_func(PIOD,  2, FUNC_A);     /* DATA8  */
+               portmux_set_func(PIOD,  3, FUNC_A);     /* DATA9  */
+               portmux_set_func(PIOD,  4, FUNC_A);     /* DATA10 */
+               portmux_set_func(PIOD,  5, FUNC_A);     /* DATA11 */
+               portmux_set_func(PIOD,  6, FUNC_A);     /* DATA12 */
+               portmux_set_func(PIOD,  7, FUNC_A);     /* DATA13 */
+               portmux_set_func(PIOD,  8, FUNC_A);     /* DATA14 */
+               portmux_set_func(PIOD,  9, FUNC_A);     /* DATA15 */
+               portmux_set_func(PIOD, 10, FUNC_A);     /* DATA16 */
+               portmux_set_func(PIOD, 11, FUNC_A);     /* DATA17 */
+               portmux_set_func(PIOD, 12, FUNC_A);     /* DATA18 */
+               portmux_set_func(PIOD, 13, FUNC_A);     /* DATA19 */
+               portmux_set_func(PIOD, 14, FUNC_A);     /* DATA20 */
+               portmux_set_func(PIOD, 15, FUNC_A);     /* DATA21 */
+               portmux_set_func(PIOD, 16, FUNC_A);     /* DATA22 */
+               portmux_set_func(PIOD, 17, FUNC_A);     /* DATA23 */
+
+               clk_set_parent(&lcdc0_pixclk, &pll0);
+               clk_set_rate(&lcdc0_pixclk, clk_get_rate(&pll0));
+               break;
+
+       default:
+               return NULL;
+       }
+
+       memcpy(pdev->dev.platform_data, data,
+              sizeof(struct lcdc_platform_data));
+
+       platform_device_register(pdev);
+       return pdev;
+}
+
+struct clk *at32_clock_list[] = {
+       &osc32k,
+       &osc0,
+       &osc1,
+       &pll0,
+       &pll1,
+       &cpu_clk,
+       &hsb_clk,
+       &pba_clk,
+       &pbb_clk,
+       &at32_sm_pclk,
+       &at32_intc0_pclk,
+       &ebi_clk,
+       &hramc_clk,
+       &pdc_hclk,
+       &pdc_pclk,
+       &pico_clk,
+       &pio0_mck,
+       &pio1_mck,
+       &pio2_mck,
+       &pio3_mck,
+       &usart0_usart,
+       &usart1_usart,
+       &usart2_usart,
+       &usart3_usart,
+       &macb0_hclk,
+       &macb0_pclk,
+       &spi0_mck,
+       &lcdc0_hclk,
+       &lcdc0_pixclk,
+};
+unsigned int at32_nr_clocks = ARRAY_SIZE(at32_clock_list);
+
+void __init at32_portmux_init(void)
+{
+       at32_init_pio(&pio0_device);
+       at32_init_pio(&pio1_device);
+       at32_init_pio(&pio2_device);
+       at32_init_pio(&pio3_device);
+}
+
+void __init at32_clock_init(void)
+{
+       struct at32_sm *sm = &system_manager;
+       u32 cpu_mask = 0, hsb_mask = 0, pba_mask = 0, pbb_mask = 0;
+       int i;
+
+       if (sm_readl(sm, PM_MCCTRL) & SM_BIT(PLLSEL))
+               main_clock = &pll0;
+       else
+               main_clock = &osc0;
+
+       if (sm_readl(sm, PM_PLL0) & SM_BIT(PLLOSC))
+               pll0.parent = &osc1;
+       if (sm_readl(sm, PM_PLL1) & SM_BIT(PLLOSC))
+               pll1.parent = &osc1;
+
+       /*
+        * Turn on all clocks that have at least one user already, and
+        * turn off everything else. We only do this for module
+        * clocks, and even though it isn't particularly pretty to
+        * check the address of the mode function, it should do the
+        * trick...
+        */
+       for (i = 0; i < ARRAY_SIZE(at32_clock_list); i++) {
+               struct clk *clk = at32_clock_list[i];
+
+               if (clk->mode == &cpu_clk_mode)
+                       cpu_mask |= 1 << clk->index;
+               else if (clk->mode == &hsb_clk_mode)
+                       hsb_mask |= 1 << clk->index;
+               else if (clk->mode == &pba_clk_mode)
+                       pba_mask |= 1 << clk->index;
+               else if (clk->mode == &pbb_clk_mode)
+                       pbb_mask |= 1 << clk->index;
+       }
+
+       sm_writel(sm, PM_CPU_MASK, cpu_mask);
+       sm_writel(sm, PM_HSB_MASK, hsb_mask);
+       sm_writel(sm, PM_PBA_MASK, pba_mask);
+       sm_writel(sm, PM_PBB_MASK, pbb_mask);
+}
diff --git a/arch/avr32/mach-at32ap/clock.c b/arch/avr32/mach-at32ap/clock.c
new file mode 100644 (file)
index 0000000..3d0d109
--- /dev/null
@@ -0,0 +1,148 @@
+/*
+ * Clock management for AT32AP CPUs
+ *
+ * Copyright (C) 2006 Atmel Corporation
+ *
+ * Based on arch/arm/mach-at91rm9200/clock.c
+ *   Copyright (C) 2005 David Brownell
+ *   Copyright (C) 2005 Ivan Kokshaysky
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ */
+#include <linux/clk.h>
+#include <linux/err.h>
+#include <linux/device.h>
+#include <linux/string.h>
+
+#include "clock.h"
+
+static spinlock_t clk_lock = SPIN_LOCK_UNLOCKED;
+
+struct clk *clk_get(struct device *dev, const char *id)
+{
+       int i;
+
+       for (i = 0; i < at32_nr_clocks; i++) {
+               struct clk *clk = at32_clock_list[i];
+
+               if (clk->dev == dev && strcmp(id, clk->name) == 0)
+                       return clk;
+       }
+
+       return ERR_PTR(-ENOENT);
+}
+EXPORT_SYMBOL(clk_get);
+
+void clk_put(struct clk *clk)
+{
+       /* clocks are static for now, we can't free them */
+}
+EXPORT_SYMBOL(clk_put);
+
+static void __clk_enable(struct clk *clk)
+{
+       if (clk->parent)
+               __clk_enable(clk->parent);
+       if (clk->users++ == 0 && clk->mode)
+               clk->mode(clk, 1);
+}
+
+int clk_enable(struct clk *clk)
+{
+       unsigned long flags;
+
+       spin_lock_irqsave(&clk_lock, flags);
+       __clk_enable(clk);
+       spin_unlock_irqrestore(&clk_lock, flags);
+
+       return 0;
+}
+EXPORT_SYMBOL(clk_enable);
+
+static void __clk_disable(struct clk *clk)
+{
+       BUG_ON(clk->users == 0);
+
+       if (--clk->users == 0 && clk->mode)
+               clk->mode(clk, 0);
+       if (clk->parent)
+               __clk_disable(clk->parent);
+}
+
+void clk_disable(struct clk *clk)
+{
+       unsigned long flags;
+
+       spin_lock_irqsave(&clk_lock, flags);
+       __clk_disable(clk);
+       spin_unlock_irqrestore(&clk_lock, flags);
+}
+EXPORT_SYMBOL(clk_disable);
+
+unsigned long clk_get_rate(struct clk *clk)
+{
+       unsigned long flags;
+       unsigned long rate;
+
+       spin_lock_irqsave(&clk_lock, flags);
+       rate = clk->get_rate(clk);
+       spin_unlock_irqrestore(&clk_lock, flags);
+
+       return rate;
+}
+EXPORT_SYMBOL(clk_get_rate);
+
+long clk_round_rate(struct clk *clk, unsigned long rate)
+{
+       unsigned long flags, actual_rate;
+
+       if (!clk->set_rate)
+               return -ENOSYS;
+
+       spin_lock_irqsave(&clk_lock, flags);
+       actual_rate = clk->set_rate(clk, rate, 0);
+       spin_unlock_irqrestore(&clk_lock, flags);
+
+       return actual_rate;
+}
+EXPORT_SYMBOL(clk_round_rate);
+
+int clk_set_rate(struct clk *clk, unsigned long rate)
+{
+       unsigned long flags;
+       long ret;
+
+       if (!clk->set_rate)
+               return -ENOSYS;
+
+       spin_lock_irqsave(&clk_lock, flags);
+       ret = clk->set_rate(clk, rate, 1);
+       spin_unlock_irqrestore(&clk_lock, flags);
+
+       return (ret < 0) ? ret : 0;
+}
+EXPORT_SYMBOL(clk_set_rate);
+
+int clk_set_parent(struct clk *clk, struct clk *parent)
+{
+       unsigned long flags;
+       int ret;
+
+       if (!clk->set_parent)
+               return -ENOSYS;
+
+       spin_lock_irqsave(&clk_lock, flags);
+       ret = clk->set_parent(clk, parent);
+       spin_unlock_irqrestore(&clk_lock, flags);
+
+       return ret;
+}
+EXPORT_SYMBOL(clk_set_parent);
+
+struct clk *clk_get_parent(struct clk *clk)
+{
+       return clk->parent;
+}
+EXPORT_SYMBOL(clk_get_parent);
diff --git a/arch/avr32/mach-at32ap/clock.h b/arch/avr32/mach-at32ap/clock.h
new file mode 100644 (file)
index 0000000..f953f04
--- /dev/null
@@ -0,0 +1,30 @@
+/*
+ * Clock management for AT32AP CPUs
+ *
+ * Copyright (C) 2006 Atmel Corporation
+ *
+ * Based on arch/arm/mach-at91rm9200/clock.c
+ *   Copyright (C) 2005 David Brownell
+ *   Copyright (C) 2005 Ivan Kokshaysky
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ */
+#include <linux/clk.h>
+
+struct clk {
+       const char      *name;          /* Clock name/function */
+       struct device   *dev;           /* Device the clock is used by */
+       struct clk      *parent;        /* Parent clock, if any */
+       void            (*mode)(struct clk *clk, int enabled);
+       unsigned long   (*get_rate)(struct clk *clk);
+       long            (*set_rate)(struct clk *clk, unsigned long rate,
+                                   int apply);
+       int             (*set_parent)(struct clk *clk, struct clk *parent);
+       u16             users;          /* Enabled if non-zero */
+       u16             index;          /* Sibling index */
+};
+
+extern struct clk *at32_clock_list[];
+extern unsigned int at32_nr_clocks;
diff --git a/arch/avr32/mach-at32ap/extint.c b/arch/avr32/mach-at32ap/extint.c
new file mode 100644 (file)
index 0000000..7da9c5f
--- /dev/null
@@ -0,0 +1,171 @@
+/*
+ * External interrupt handling for AT32AP CPUs
+ *
+ * Copyright (C) 2006 Atmel Corporation
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ */
+
+#include <linux/errno.h>
+#include <linux/init.h>
+#include <linux/interrupt.h>
+#include <linux/irq.h>
+#include <linux/platform_device.h>
+#include <linux/random.h>
+
+#include <asm/io.h>
+
+#include <asm/arch/sm.h>
+
+#include "sm.h"
+
+static void eim_ack_irq(unsigned int irq)
+{
+       struct at32_sm *sm = get_irq_chip_data(irq);
+       sm_writel(sm, EIM_ICR, 1 << (irq - sm->eim_first_irq));
+}
+
+static void eim_mask_irq(unsigned int irq)
+{
+       struct at32_sm *sm = get_irq_chip_data(irq);
+       sm_writel(sm, EIM_IDR, 1 << (irq - sm->eim_first_irq));
+}
+
+static void eim_mask_ack_irq(unsigned int irq)
+{
+       struct at32_sm *sm = get_irq_chip_data(irq);
+       sm_writel(sm, EIM_ICR, 1 << (irq - sm->eim_first_irq));
+       sm_writel(sm, EIM_IDR, 1 << (irq - sm->eim_first_irq));
+}
+
+static void eim_unmask_irq(unsigned int irq)
+{
+       struct at32_sm *sm = get_irq_chip_data(irq);
+       sm_writel(sm, EIM_IER, 1 << (irq - sm->eim_first_irq));
+}
+
+static int eim_set_irq_type(unsigned int irq, unsigned int flow_type)
+{
+       struct at32_sm *sm = get_irq_chip_data(irq);
+       unsigned int i = irq - sm->eim_first_irq;
+       u32 mode, edge, level;
+       unsigned long flags;
+       int ret = 0;
+
+       flow_type &= IRQ_TYPE_SENSE_MASK;
+
+       spin_lock_irqsave(&sm->lock, flags);
+
+       mode = sm_readl(sm, EIM_MODE);
+       edge = sm_readl(sm, EIM_EDGE);
+       level = sm_readl(sm, EIM_LEVEL);
+
+       switch (flow_type) {
+       case IRQ_TYPE_LEVEL_LOW:
+               mode |= 1 << i;
+               level &= ~(1 << i);
+               break;
+       case IRQ_TYPE_LEVEL_HIGH:
+               mode |= 1 << i;
+               level |= 1 << i;
+               break;
+       case IRQ_TYPE_EDGE_RISING:
+               mode &= ~(1 << i);
+               edge |= 1 << i;
+               break;
+       case IRQ_TYPE_EDGE_FALLING:
+               mode &= ~(1 << i);
+               edge &= ~(1 << i);
+               break;
+       default:
+               ret = -EINVAL;
+               break;
+       }
+
+       sm_writel(sm, EIM_MODE, mode);
+       sm_writel(sm, EIM_EDGE, edge);
+       sm_writel(sm, EIM_LEVEL, level);
+
+       spin_unlock_irqrestore(&sm->lock, flags);
+
+       return ret;
+}
+
+struct irq_chip eim_chip = {
+       .name           = "eim",
+       .ack            = eim_ack_irq,
+       .mask           = eim_mask_irq,
+       .mask_ack       = eim_mask_ack_irq,
+       .unmask         = eim_unmask_irq,
+       .set_type       = eim_set_irq_type,
+};
+
+static void demux_eim_irq(unsigned int irq, struct irq_desc *desc,
+                         struct pt_regs *regs)
+{
+       struct at32_sm *sm = desc->handler_data;
+       struct irq_desc *ext_desc;
+       unsigned long status, pending;
+       unsigned int i, ext_irq;
+
+       spin_lock(&sm->lock);
+
+       status = sm_readl(sm, EIM_ISR);
+       pending = status & sm_readl(sm, EIM_IMR);
+
+       while (pending) {
+               i = fls(pending) - 1;
+               pending &= ~(1 << i);
+
+               ext_irq = i + sm->eim_first_irq;
+               ext_desc = irq_desc + ext_irq;
+               ext_desc->handle_irq(ext_irq, ext_desc, regs);
+       }
+
+       spin_unlock(&sm->lock);
+}
+
+static int __init eim_init(void)
+{
+       struct at32_sm *sm = &system_manager;
+       unsigned int i;
+       unsigned int nr_irqs;
+       unsigned int int_irq;
+       u32 pattern;
+
+       /*
+        * The EIM is really the same module as SM, so register
+        * mapping, etc. has been taken care of already.
+        */
+
+       /*
+        * Find out how many interrupt lines that are actually
+        * implemented in hardware.
+        */
+       sm_writel(sm, EIM_IDR, ~0UL);
+       sm_writel(sm, EIM_MODE, ~0UL);
+       pattern = sm_readl(sm, EIM_MODE);
+       nr_irqs = fls(pattern);
+
+       sm->eim_chip = &eim_chip;
+
+       for (i = 0; i < nr_irqs; i++) {
+               set_irq_chip(sm->eim_first_irq + i, &eim_chip);
+               set_irq_chip_data(sm->eim_first_irq + i, sm);
+       }
+
+       int_irq = platform_get_irq_byname(sm->pdev, "eim");
+
+       set_irq_chained_handler(int_irq, demux_eim_irq);
+       set_irq_data(int_irq, sm);
+
+       printk("EIM: External Interrupt Module at 0x%p, IRQ %u\n",
+              sm->regs, int_irq);
+       printk("EIM: Handling %u external IRQs, starting with IRQ %u\n",
+              nr_irqs, sm->eim_first_irq);
+
+       return 0;
+}
+arch_initcall(eim_init);
diff --git a/arch/avr32/mach-at32ap/intc.c b/arch/avr32/mach-at32ap/intc.c
new file mode 100644 (file)
index 0000000..74f8c9f
--- /dev/null
@@ -0,0 +1,133 @@
+/*
+ * Copyright (C) 2006 Atmel Corporation
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ */
+
+#include <linux/clk.h>
+#include <linux/err.h>
+#include <linux/init.h>
+#include <linux/interrupt.h>
+#include <linux/irq.h>
+#include <linux/platform_device.h>
+
+#include <asm/io.h>
+
+#include "intc.h"
+
+struct intc {
+       void __iomem    *regs;
+       struct irq_chip chip;
+};
+
+extern struct platform_device at32_intc0_device;
+
+/*
+ * TODO: We may be able to implement mask/unmask by setting IxM flags
+ * in the status register.
+ */
+static void intc_mask_irq(unsigned int irq)
+{
+
+}
+
+static void intc_unmask_irq(unsigned int irq)
+{
+
+}
+
+static struct intc intc0 = {
+       .chip = {
+               .name           = "intc",
+               .mask           = intc_mask_irq,
+               .unmask         = intc_unmask_irq,
+       },
+};
+
+/*
+ * All interrupts go via intc at some point.
+ */
+asmlinkage void do_IRQ(int level, struct pt_regs *regs)
+{
+       struct irq_desc *desc;
+       unsigned int irq;
+       unsigned long status_reg;
+
+       local_irq_disable();
+
+       irq_enter();
+
+       irq = intc_readl(&intc0, INTCAUSE0 - 4 * level);
+       desc = irq_desc + irq;
+       desc->handle_irq(irq, desc, regs);
+
+       /*
+        * Clear all interrupt level masks so that we may handle
+        * interrupts during softirq processing.  If this is a nested
+        * interrupt, interrupts must stay globally disabled until we
+        * return.
+        */
+       status_reg = sysreg_read(SR);
+       status_reg &= ~(SYSREG_BIT(I0M) | SYSREG_BIT(I1M)
+                       | SYSREG_BIT(I2M) | SYSREG_BIT(I3M));
+       sysreg_write(SR, status_reg);
+
+       irq_exit();
+}
+
+void __init init_IRQ(void)
+{
+       extern void _evba(void);
+       extern void irq_level0(void);
+       struct resource *regs;
+       struct clk *pclk;
+       unsigned int i;
+       u32 offset, readback;
+
+       regs = platform_get_resource(&at32_intc0_device, IORESOURCE_MEM, 0);
+       if (!regs) {
+               printk(KERN_EMERG "intc: no mmio resource defined\n");
+               goto fail;
+       }
+       pclk = clk_get(&at32_intc0_device.dev, "pclk");
+       if (IS_ERR(pclk)) {
+               printk(KERN_EMERG "intc: no clock defined\n");
+               goto fail;
+       }
+
+       clk_enable(pclk);
+
+       intc0.regs = ioremap(regs->start, regs->end - regs->start + 1);
+       if (!intc0.regs) {
+               printk(KERN_EMERG "intc: failed to map registers (0x%08lx)\n",
+                      (unsigned long)regs->start);
+               goto fail;
+       }
+
+       /*
+        * Initialize all interrupts to level 0 (lowest priority). The
+        * priority level may be changed by calling
+        * irq_set_priority().
+        *
+        */
+       offset = (unsigned long)&irq_level0 - (unsigned long)&_evba;
+       for (i = 0; i < NR_INTERNAL_IRQS; i++) {
+               intc_writel(&intc0, INTPR0 + 4 * i, offset);
+               readback = intc_readl(&intc0, INTPR0 + 4 * i);
+               if (readback == offset)
+                       set_irq_chip_and_handler(i, &intc0.chip,
+                                                handle_simple_irq);
+       }
+
+       /* Unmask all interrupt levels */
+       sysreg_write(SR, (sysreg_read(SR)
+                         & ~(SR_I3M | SR_I2M | SR_I1M | SR_I0M)));
+
+       return;
+
+fail:
+       panic("Interrupt controller initialization failed!\n");
+}
+
diff --git a/arch/avr32/mach-at32ap/intc.h b/arch/avr32/mach-at32ap/intc.h
new file mode 100644 (file)
index 0000000..d289ca2
--- /dev/null
@@ -0,0 +1,327 @@
+/*
+ * Automatically generated by gen-header.xsl
+ */
+#ifndef __ASM_AVR32_PERIHP_INTC_H__
+#define __ASM_AVR32_PERIHP_INTC_H__
+
+#define INTC_NUM_INT_GRPS            33
+
+#define INTC_INTPR0                  0x0
+# define INTC_INTPR0_INTLEV_OFFSET   30
+# define INTC_INTPR0_INTLEV_SIZE     2
+# define INTC_INTPR0_OFFSET_OFFSET   0
+# define INTC_INTPR0_OFFSET_SIZE     24
+#define INTC_INTREQ0                 0x100
+# define INTC_INTREQ0_IREQUEST0_OFFSET 0
+# define INTC_INTREQ0_IREQUEST0_SIZE 1
+# define INTC_INTREQ0_IREQUEST1_OFFSET 1
+# define INTC_INTREQ0_IREQUEST1_SIZE 1
+#define INTC_INTPR1                  0x4
+# define INTC_INTPR1_INTLEV_OFFSET   30
+# define INTC_INTPR1_INTLEV_SIZE     2
+# define INTC_INTPR1_OFFSET_OFFSET   0
+# define INTC_INTPR1_OFFSET_SIZE     24
+#define INTC_INTREQ1                 0x104
+# define INTC_INTREQ1_IREQUEST32_OFFSET 0
+# define INTC_INTREQ1_IREQUEST32_SIZE 1
+# define INTC_INTREQ1_IREQUEST33_OFFSET 1
+# define INTC_INTREQ1_IREQUEST33_SIZE 1
+# define INTC_INTREQ1_IREQUEST34_OFFSET 2
+# define INTC_INTREQ1_IREQUEST34_SIZE 1
+# define INTC_INTREQ1_IREQUEST35_OFFSET 3
+# define INTC_INTREQ1_IREQUEST35_SIZE 1
+# define INTC_INTREQ1_IREQUEST36_OFFSET 4
+# define INTC_INTREQ1_IREQUEST36_SIZE 1
+# define INTC_INTREQ1_IREQUEST37_OFFSET 5
+# define INTC_INTREQ1_IREQUEST37_SIZE 1
+#define INTC_INTPR2                  0x8
+# define INTC_INTPR2_INTLEV_OFFSET   30
+# define INTC_INTPR2_INTLEV_SIZE     2
+# define INTC_INTPR2_OFFSET_OFFSET   0
+# define INTC_INTPR2_OFFSET_SIZE     24
+#define INTC_INTREQ2                 0x108
+# define INTC_INTREQ2_IREQUEST64_OFFSET 0
+# define INTC_INTREQ2_IREQUEST64_SIZE 1
+# define INTC_INTREQ2_IREQUEST65_OFFSET 1
+# define INTC_INTREQ2_IREQUEST65_SIZE 1
+# define INTC_INTREQ2_IREQUEST66_OFFSET 2
+# define INTC_INTREQ2_IREQUEST66_SIZE 1
+# define INTC_INTREQ2_IREQUEST67_OFFSET 3
+# define INTC_INTREQ2_IREQUEST67_SIZE 1
+# define INTC_INTREQ2_IREQUEST68_OFFSET 4
+# define INTC_INTREQ2_IREQUEST68_SIZE 1
+#define INTC_INTPR3                  0xc
+# define INTC_INTPR3_INTLEV_OFFSET   30
+# define INTC_INTPR3_INTLEV_SIZE     2
+# define INTC_INTPR3_OFFSET_OFFSET   0
+# define INTC_INTPR3_OFFSET_SIZE     24
+#define INTC_INTREQ3                 0x10c
+# define INTC_INTREQ3_IREQUEST96_OFFSET 0
+# define INTC_INTREQ3_IREQUEST96_SIZE 1
+#define INTC_INTPR4                  0x10
+# define INTC_INTPR4_INTLEV_OFFSET   30
+# define INTC_INTPR4_INTLEV_SIZE     2
+# define INTC_INTPR4_OFFSET_OFFSET   0
+# define INTC_INTPR4_OFFSET_SIZE     24
+#define INTC_INTREQ4                 0x110
+# define INTC_INTREQ4_IREQUEST128_OFFSET 0
+# define INTC_INTREQ4_IREQUEST128_SIZE 1
+#define INTC_INTPR5                  0x14
+# define INTC_INTPR5_INTLEV_OFFSET   30
+# define INTC_INTPR5_INTLEV_SIZE     2
+# define INTC_INTPR5_OFFSET_OFFSET   0
+# define INTC_INTPR5_OFFSET_SIZE     24
+#define INTC_INTREQ5                 0x114
+# define INTC_INTREQ5_IREQUEST160_OFFSET 0
+# define INTC_INTREQ5_IREQUEST160_SIZE 1
+#define INTC_INTPR6                  0x18
+# define INTC_INTPR6_INTLEV_OFFSET   30
+# define INTC_INTPR6_INTLEV_SIZE     2
+# define INTC_INTPR6_OFFSET_OFFSET   0
+# define INTC_INTPR6_OFFSET_SIZE     24
+#define INTC_INTREQ6                 0x118
+# define INTC_INTREQ6_IREQUEST192_OFFSET 0
+# define INTC_INTREQ6_IREQUEST192_SIZE 1
+#define INTC_INTPR7                  0x1c
+# define INTC_INTPR7_INTLEV_OFFSET   30
+# define INTC_INTPR7_INTLEV_SIZE     2
+# define INTC_INTPR7_OFFSET_OFFSET   0
+# define INTC_INTPR7_OFFSET_SIZE     24
+#define INTC_INTREQ7                 0x11c
+# define INTC_INTREQ7_IREQUEST224_OFFSET 0
+# define INTC_INTREQ7_IREQUEST224_SIZE 1
+#define INTC_INTPR8                  0x20
+# define INTC_INTPR8_INTLEV_OFFSET   30
+# define INTC_INTPR8_INTLEV_SIZE     2
+# define INTC_INTPR8_OFFSET_OFFSET   0
+# define INTC_INTPR8_OFFSET_SIZE     24
+#define INTC_INTREQ8                 0x120
+# define INTC_INTREQ8_IREQUEST256_OFFSET 0
+# define INTC_INTREQ8_IREQUEST256_SIZE 1
+#define INTC_INTPR9                  0x24
+# define INTC_INTPR9_INTLEV_OFFSET   30
+# define INTC_INTPR9_INTLEV_SIZE     2
+# define INTC_INTPR9_OFFSET_OFFSET   0
+# define INTC_INTPR9_OFFSET_SIZE     24
+#define INTC_INTREQ9                 0x124
+# define INTC_INTREQ9_IREQUEST288_OFFSET 0
+# define INTC_INTREQ9_IREQUEST288_SIZE 1
+#define INTC_INTPR10                 0x28
+# define INTC_INTPR10_INTLEV_OFFSET  30
+# define INTC_INTPR10_INTLEV_SIZE    2
+# define INTC_INTPR10_OFFSET_OFFSET  0
+# define INTC_INTPR10_OFFSET_SIZE    24
+#define INTC_INTREQ10                0x128
+# define INTC_INTREQ10_IREQUEST320_OFFSET 0
+# define INTC_INTREQ10_IREQUEST320_SIZE 1
+#define INTC_INTPR11                 0x2c
+# define INTC_INTPR11_INTLEV_OFFSET  30
+# define INTC_INTPR11_INTLEV_SIZE    2
+# define INTC_INTPR11_OFFSET_OFFSET  0
+# define INTC_INTPR11_OFFSET_SIZE    24
+#define INTC_INTREQ11                0x12c
+# define INTC_INTREQ11_IREQUEST352_OFFSET 0
+# define INTC_INTREQ11_IREQUEST352_SIZE 1
+#define&n