16 years ago[PATCH] ppc32: make usage of CONFIG_PTE_64BIT & CONFIG_PHYS_64BIT consistent
Kumar Gala [Sat, 16 Apr 2005 22:24:21 +0000]
[PATCH] ppc32: make usage of CONFIG_PTE_64BIT & CONFIG_PHYS_64BIT consistent

CONFIG_PTE_64BIT & CONFIG_PHYS_64BIT are not currently consistently used in
the code base.  Fixed up the usage such that CONFIG_PTE_64BIT is used when we
have a 64-bit PTE regardless of physical address width.  CONFIG_PHYS_64BIT is
used if the physical address width is larger than 32-bits, regardless of PTE
size.

These changes required a few sub-arch specific ifdef's to be fixed and the
introduction of a physical address format string.

Signed-off-by: Kumar Gala <kumar.gala@freescale.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

16 years ago[PATCH] ppc32: Fix pte_update for 64-bit PTEs
Kumar Gala [Sat, 16 Apr 2005 22:24:20 +0000]
[PATCH] ppc32: Fix pte_update for 64-bit PTEs

While the existing pte_update code handled atomically modifying a 64-bit PTE,
it did not return all 64-bits of the PTE before it was modified.  This causes
problems in some places that expect the full PTE to be returned, like
ptep_get_and_clear().

Created a new pte_update function that is conditional on CONFIG_PTE_64BIT.  It
atomically reads the low PTE word which all PTE flags are required to be in
and returns a premodified full 64-bit PTE.

Since we now have an explicit 64-bit PTE version of pte_update we can also
remove the hack that existed to get the low PTE word regardless of size.

Signed-off-by: Kumar Gala <kumar.gala@freescale.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

16 years ago[PATCH] ppc32: Fix AGP and sleep again
Benjamin Herrenschmidt [Sat, 16 Apr 2005 22:24:19 +0000]
[PATCH] ppc32: Fix AGP and sleep again

My previous patch that added sleep support for uninorth-agp and some AGP
"off" stuff in radeonfb and aty128fb is breaking some configs.  More
specifically, it has problems with rage128 setups since the DRI code for
these in X doesn't properly re-enable AGP on wakeup or console switch
(unlike the radeon DRM).

This patch fixes the problem for pmac once for all by using a different
approach.  The AGP driver "registers" special suspend/resume callbacks with
some arch code that the fbdev's can later on call to suspend and resume
AGP, making sure it's resumed back in the same state it was when suspended.
 This is platform specific for now.  It would be too complicated to try to
do a generic implementation of this at this point due to all sort of weird
things going on with AGP on other architectures.  We'll re-work that whole
problem cleanly once we finally merge fbdev's and DRI.

In the meantime, please apply this patch which brings back some r128 based
laptops into working condition as far as system sleep is concerned.

Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

16 years ago[PATCH] ppc32: Fix cpufreq problems
Benjamin Herrenschmidt [Sat, 16 Apr 2005 22:24:18 +0000]
[PATCH] ppc32: Fix cpufreq problems

This patch updates the PowerMac cpufreq driver.  It depends on the addition
of the suspend() method (my previous patch) and on the new flag I defined
to silence some warnings that are normal for us.

It fixes various issues related to cpufreq on pmac, including some crashes
on some models when sleeping the machine while in low speed, proper voltage
control on some newer machines, and adds voltage control on 750FX based G3
laptops.

Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

16 years ago[PATCH] ppc32: fix single-stepping of emulated instructions
Paul Mackerras [Sat, 16 Apr 2005 22:24:17 +0000]
[PATCH] ppc32: fix single-stepping of emulated instructions

On ppc, we emulate instructions that cause alignment exceptions.  If we are
single-stepping an instruction and it causes an alignment exception, we
will currently do the next instruction as well before taking the
single-step exception.  This patch fixes that, so we take the single-step
exception after emulating the instruction.

Signed-off-by: Paul Mackerras <paulus@samba.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

16 years ago[PATCH] ppc32: oops on kernel altivec assist exceptions
Paul Mackerras [Sat, 16 Apr 2005 22:24:17 +0000]
[PATCH] ppc32: oops on kernel altivec assist exceptions

If we should happen to get an altivec assist exception while executing in
the kernel, we will currently try to handle it and fail, and end up oopsing
with (apparently) a segfault.  (An altivec assist exception occurs for
floating-point altivec instructions with denormalized inputs or outputs if
the altivec unit is in java mode.)

This patch checks explicitly if we are in user mode and prints a useful
message if not.

Signed-off-by: Paul Mackerras <paulus@samba.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

16 years ago[PATCH] ppc32: improve timebase sync for SMP
Paul Mackerras [Sat, 16 Apr 2005 22:24:16 +0000]
[PATCH] ppc32: improve timebase sync for SMP

Currently the procedure in the ppc32 kernel that synchronizes the timebase
registers across an SMP powermac system does so by setting both timebases
to zero.  That is OK at boot but causes problems if done later.  So that we
can do hotplug CPU on these machines, this patch changes the code so it
reads the timebase from one CPU and transfers the value to the other CPU.
(Hotplug CPU is needed for sleep (aka suspend to RAM) to work.)

Signed-off-by: Paul Mackerras <paulus@samba.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

16 years ago[PATCH] ppc32: ppc4xx_pic - add acknowledge when enabling level-sensitive IRQ
Eugene Surovegin [Sat, 16 Apr 2005 22:24:15 +0000]
[PATCH] ppc32: ppc4xx_pic - add acknowledge when enabling level-sensitive IRQ

This patch adds interrupt acknowledge to the PPC4xx PIC enable_irq
implementation for level-sensitive IRQ sources.  This helps in cases when
enable/disable_irq is used in interrupt handlers for hardware, which
requires IRQ acknowledge to be issued from non-interrupt context (e.g.
when actual ACK in device needs an I2C transaction).  For such strange
hardware, interrupt handler disables IRQ and defers actual ACK to some
other context.  When this happens, IRQ is enabled again.  For
level-sensitive sources we get spurious triggering right after IRQ is
enabled.  This patch fixes this.

Suggested by Tolunay Orkun <listmember@orkun.us>.

Signed-off-by: Eugene Surovegin <ebs@ebshome.net>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

16 years ago[PATCH] ppc32: fix bogosity in process-freezing code
Paul Mackerras [Sat, 16 Apr 2005 22:24:14 +0000]
[PATCH] ppc32: fix bogosity in process-freezing code

The code that went into arch/ppc/kernel/signal.c recently to handle process
freezing seems to contain a dubious assumption: that a process that calls
do_signal when PF_FREEZE is set will have entered the kernel because of a
system call.  This patch removes that assumption.

Signed-off-by: Paul Mackerras <paulus@samba.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

16 years ago[PATCH] SELinux: add support for NETLINK_KOBJECT_UEVENT
James Morris [Sat, 16 Apr 2005 22:24:13 +0000]
[PATCH] SELinux: add support for NETLINK_KOBJECT_UEVENT

This patch adds SELinux support for the KOBJECT_UEVENT Netlink family, so
that SELinux can apply finer grained controls to it.  For example, security
policy for hald can be locked down to the KOBJECT_UEVENT Netlink family
only.  Currently, this family simply defaults to the default Netlink socket
class.

Note that some new permission definitions are added to sync with changes in
the core userspace policy package, which auto-generates header files.

Signed-off-by: James Morris <jmorris@redhat.com>
Signed-off-by: Stephen Smalley <sds@tycho.nsa.gov>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

16 years ago[PATCH] irda_device() oops fix
Jean Tourrilhes [Sat, 16 Apr 2005 22:24:11 +0000]
[PATCH] irda_device() oops fix

Acked-by: "David S. Miller" <davem@davemloft.net>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

16 years ago[PATCH] Fix dst_destroy() race
Herbert Xu [Sat, 16 Apr 2005 22:24:10 +0000]
[PATCH] Fix dst_destroy() race

When we are not the real parent of the dst (e.g., when we're xfrm_dst and
the child is an rtentry), it may already be on the GC list.

In fact the current code is buggy to, we need to check dst->flags before
the dec as dst may no longer be valid afterwards.

Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

16 years ago[PATCH] net: don't call kmem_cache_create with a spinlock held
Arnaldo Carvalho de Melo [Sat, 16 Apr 2005 22:24:09 +0000]
[PATCH] net: don't call kmem_cache_create with a spinlock held

This fixes the warning reported by Marcel Holtmann (Thanks!).

Signed-off-by: Arnaldo Carvalho de Melo <acme@conectiva.com.br>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

16 years ago[PATCH] Fix linux/atalk.h header
David S. Miller [Sat, 16 Apr 2005 22:24:09 +0000]
[PATCH] Fix linux/atalk.h header

This recently got changed to include a lot of kernel internal stuff in the
non-__KERNEL__ area of the header, which isn't so kosher and breaks libc
builds.

The fix is pretty simple.

Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

16 years ago[PATCH] meminfo: add Cached underflow check
Martin Hicks [Sat, 16 Apr 2005 22:24:08 +0000]
[PATCH] meminfo: add Cached underflow check

Working on some code lately I've been getting huge values for "Cached".
The cause is that get_page_cache_size() is an approximate value, and for a
sufficiently small returned value of get_page_cache_size() the value
underflows.

Signed-off-by: Martin Hicks <mort@sgi.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

16 years ago[PATCH] end_buffer_write_sync() avoid pointless assignments
akpm@osdl.org [Sat, 16 Apr 2005 22:24:07 +0000]
[PATCH] end_buffer_write_sync() avoid pointless assignments

Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

16 years ago[PATCH] vmscan: pageout(): remove unneeded test
akpm@osdl.org [Sat, 16 Apr 2005 22:24:06 +0000]
[PATCH] vmscan: pageout(): remove unneeded test

\r)

We only call pageout() for dirty pages, so this test is redundant.

Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

16 years ago[PATCH] oom-killer disable for iscsi/lvm2/multipath userland critical sections
Andrea Arcangeli [Sat, 16 Apr 2005 22:24:05 +0000]
[PATCH] oom-killer disable for iscsi/lvm2/multipath userland critical sections

iscsi/lvm2/multipath needs guaranteed protection from the oom-killer, so
make the magical value of -17 in /proc/<pid>/oom_adj defeat the oom-killer
altogether.

(akpm: we still need to document oom_adj and friends in
Documentation/filesystems/proc.txt!)

Signed-off-by: Andrea Arcangeli <andrea@suse.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

16 years ago[PATCH] filemap_getpage can block when MAP_NONBLOCK specified
Jeff Moyer [Sat, 16 Apr 2005 22:24:05 +0000]
[PATCH] filemap_getpage can block when MAP_NONBLOCK specified

We will return NULL from filemap_getpage when a page does not exist in the
page cache and MAP_NONBLOCK is specified, here:

page = find_get_page(mapping, pgoff);
if (!page) {
if (nonblock)
return NULL;
goto no_cached_page;
}

But we forget to do so when the page in the cache is not uptodate.  The
following could result in a blocking call:

/*
 * Ok, found a page in the page cache, now we need to check
 * that it's up-to-date.
 */
if (!PageUptodate(page))
goto page_not_uptodate;

Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

16 years ago[PATCH] r128_state.c: break missing in switch statement
Dave Airlie [Sat, 16 Apr 2005 22:24:04 +0000]
[PATCH] r128_state.c: break missing in switch statement

drm: fix r128_state.c switch statements..  in drivers/char/drm/r128_state.c
(linux-2.6.12-rc2), some breaks are missing in the switch statement.  See
trivial fix below.

Signed-off-by: Hansjoerg Lipp <hjlipp@web.de>
Signed-off-by: Dave Airlie <airlied@linux.ie>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

16 years ago[PATCH] SELinux: fix bug in Netlink message type detection
James Morris [Sat, 16 Apr 2005 22:24:03 +0000]
[PATCH] SELinux: fix bug in Netlink message type detection

This patch fixes a bug in the SELinux Netlink message type detection code,
where the wrong constant was being used in a case statement.  The incorrect
value is not valid for this class of object so it would not have been
reached, and fallen through to a default handler for all Netlink messages.

Signed-off-by: James Morris <jmorris@redhat.com>
Signed-off-by: Stephen Smalley <sds@tycho.nsa.gov>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

16 years ago[PATCH] fix Bug 4395: modprobe bttv freezes the computer
akpm@osdl.org [Sat, 16 Apr 2005 22:24:02 +0000]
[PATCH] fix Bug 4395: modprobe bttv freezes the computer

\r)

From: Johannes Stezenbach <js@linuxtv.org>

Fix http://bugme.osdl.org/show_bug.cgi?id=4395.

Patch by Manu Abraham and Gerd Knorr:

Remove redundant bttv_reset_audio() which caused the computer to freeze
with some bt8xx based DVB cards when loading the bttv driver.

Signed-off-by: Johannes Stezenbach <js@linuxtv.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

16 years ago[PATCH] Fix get_compat_sigevent()
David S. Miller [Sat, 16 Apr 2005 22:24:01 +0000]
[PATCH] Fix get_compat_sigevent()

I have no idea how a bug like this lasted so long.  Anyways, obvious
memset()'ing of incorrect pointer.

Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

16 years ago[PATCH] fix crash in entry.S restore_all
Stas Sergeev [Sat, 16 Apr 2005 22:24:01 +0000]
[PATCH] fix crash in entry.S restore_all

Fix the access-above-bottom-of-stack crash.

1. Allows to preserve the valueable optimization

2. Works for NMIs

3.  Doesn't care whether or not there are more of the like instances
   where the stack is left empty.

4. Seems to work for me without the crashes:)

(akpm: this is still under discussion, although I _think_ it's OK.  You might
want to hold off)

Signed-off-by: Stas Sergeev <stsp@aknet.ru>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

16 years ago[PATCH] Fix acl Oops
akpm@osdl.org [Sat, 16 Apr 2005 22:24:00 +0000]
[PATCH] Fix acl Oops

\r)

From: Andreas Gruenbacher <agruen@suse.de>

ext[23]_get_acl will return an error when reading the attribute fails or
out-of-memory occurs.  Catch this case.

Signed-off-by: Andreas Gruenbacher <agruen@suse.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

16 years ago[PATCH] re-export cancel_rearming_delayed_workqueue
James Bottomley [Sat, 16 Apr 2005 22:23:59 +0000]
[PATCH] re-export cancel_rearming_delayed_workqueue

This was unexported by Arjan because we have no current users.

However, during a conversion from tasklets to workqueues of the parisc led
functions, we ran across a case where this was needed.  In particular, the
open coded equivalent of cancel_rearming_delayed_workqueue was implemented
incorrectly, which is, I think, all the evidence necessary that this is a
useful API.

Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

16 years ago[PATCH] crypto: call zlib end functions on deflate exit path
Artem B. Bityuckiy [Sat, 16 Apr 2005 22:23:58 +0000]
[PATCH] crypto: call zlib end functions on deflate exit path

In the deflate_[compress|uncompress|pcompress] functions we call the
zlib_[in|de]flateReset function at the beginning.  This is OK.  But when we
unload the deflate module we don't call zlib_[in|de]flateEnd to free all
the zlib internal data.  It looks like a bug for me.  Please, consider the
attached patch.

Signed-off-by: Artem B. Bityuckiy <dedekind@infradead.org>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Cc: "David S. Miller" <davem@davemloft.net>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

16 years ago[PATCH] arm: add comment about max_low_pfn/max_pfn
akpm@osdl.org [Sat, 16 Apr 2005 22:23:57 +0000]
[PATCH] arm: add comment about max_low_pfn/max_pfn

\r)

From: Russell King <rmk+lkml@arm.linux.org.uk>

Oddly, max_low_pfn/max_pfn end up being the number of pages in the system,
rather than the maximum PFN on ARM.  This doesn't seem to cause any problems,
so just add a note about it.

Signed-off-by: Russell King <rmk@arm.linux.org.uk>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

16 years ago[PATCH] arm: add comment about dma_supported()
akpm@osdl.org [Sat, 16 Apr 2005 22:23:57 +0000]
[PATCH] arm: add comment about dma_supported()

\r)

From: Russell King <rmk+lkml@arm.linux.org.uk>

The ARM dma_supported() is rather basic, and I don't think it takes into
account everything that it should do (eg, whether the mask agrees with what
we'd return for GFP_DMA allocations).  Note this.

Signed-off-by: Russell King <rmk@arm.linux.org.uk>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

16 years ago[PATCH] arm: fix help text for ixdp465
akpm@osdl.org [Sat, 16 Apr 2005 22:23:56 +0000]
[PATCH] arm: fix help text for ixdp465

\r)

From: Russell King <rmk+lkml@arm.linux.org.uk>

For some reason, this help text was missed when the file was last audited
by the documentation referencing folk.  Fix this incorrect documentation
reference.

Signed-off-by: Russell King <rmk@arm.linux.org.uk>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

16 years ago[PATCH] arm: fix SIGBUS handling
akpm@osdl.org [Sat, 16 Apr 2005 22:23:55 +0000]
[PATCH] arm: fix SIGBUS handling

\r)

From: Russell King <rmk+lkml@arm.linux.org.uk>

ARM wasn't raising a SIGBUS with a siginfo structure.  Fix
__do_user_fault() to allow us to use it for SIGBUS conditions, and arrange
for the sigbus path to use this.

We need to prevent the siginfo code being called if we do not have a user
space context to call it, so consolidate the "user_mode()" tests.

Thanks to Ian Campbell who spotted this oversight.

Signed-off-by: Russell King <rmk@arm.linux.org.uk>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

16 years ago[PATCH] Avoid deadlock in sync_page_io by using GFP_NOIO
Neil Brown [Sat, 16 Apr 2005 22:23:54 +0000]
[PATCH] Avoid deadlock in sync_page_io by using GFP_NOIO

..as sync_page_io can be called on the write-out path.

Signed-off-by: Neil Brown <neilb@cse.unsw.edu.au>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

16 years ago[PATCH] mmtimer build fix
Christoph Lameter [Sat, 16 Apr 2005 22:23:53 +0000]
[PATCH] mmtimer build fix

Signed-off-by: Christoph Lameter <clameter@sgi.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

16 years agoLinux-2.6.12-rc2 master
Linus Torvalds [Sat, 16 Apr 2005 22:20:36 +0000]
Linux-2.6.12-rc2

Initial git repository build. I'm not bothering with the full history,
even though we have it. We can create a separate "historical" git
archive of that later if we want to, and in the meantime it's about
3.2GB when imported into git - space that would just make the early
git days unnecessarily complicated, when we don't have a lot of good
infrastructure for it.

Let it rip!