8 years agocrypto: rfc4106 - Extending the RC4106 AES-GCM test vectors
Adrian Hoban [Thu, 4 Nov 2010 19:02:04 +0000]
crypto: rfc4106 - Extending the RC4106 AES-GCM test vectors

Updated RFC4106 AES-GCM testing. Some test vectors were taken from
http://csrc.nist.gov/groups/ST/toolkit/BCM/documents/proposedmodes/
gcm/gcm-test-vectors.tar.gz

Signed-off-by: Adrian Hoban <adrian.hoban@intel.com>
Signed-off-by: Tadeusz Struk <tadeusz.struk@intel.com>
Signed-off-by: Gabriele Paoloni <gabriele.paoloni@intel.com>
Signed-off-by: Aidan O'Mahony <aidan.o.mahony@intel.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>

8 years agocrypto: aesni-intel - RFC4106 AES-GCM Driver Using Intel New Instructions
Tadeusz Struk [Thu, 4 Nov 2010 19:00:45 +0000]
crypto: aesni-intel - RFC4106 AES-GCM Driver Using Intel New Instructions

This patch adds an optimized RFC4106 AES-GCM implementation for 64-bit
kernels. It supports 128-bit AES key size. This leverages the crypto
AEAD interface type to facilitate a combined AES & GCM operation to
be implemented in assembly code. The assembly code leverages Intel(R)
AES New Instructions and the PCLMULQDQ instruction.

Signed-off-by: Adrian Hoban <adrian.hoban@intel.com>
Signed-off-by: Tadeusz Struk <tadeusz.struk@intel.com>
Signed-off-by: Gabriele Paoloni <gabriele.paoloni@intel.com>
Signed-off-by: Aidan O'Mahony <aidan.o.mahony@intel.com>
Signed-off-by: Erdinc Ozturk <erdinc.ozturk@intel.com>
Signed-off-by: James Guilford <james.guilford@intel.com>
Signed-off-by: Wajdi Feghali <wajdi.k.feghali@intel.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>

8 years agocrypto: cast5 - simplify if-statements
Nicolas Kaiser [Thu, 4 Nov 2010 18:58:12 +0000]
crypto: cast5 - simplify if-statements

I noticed that by factoring out common rounds from the
branches of the if-statements in the encryption and
decryption functions, the executable file size goes down
significantly, for crypto/cast5.ko from 26688 bytes
to 24336 bytes (amd64).

On my test system, I saw a slight speedup. This is the
first time I'm doing such a benchmark - I found a similar
one on the crypto mailing list, and I hope I did it right?

Before:
# cryptsetup create dm-test /dev/hda2 -c cast5-cbc-plain -s 128
Passsatz eingeben:
# dd if=/dev/zero of=/dev/mapper/dm-test bs=1M count=50
52428800 Bytes (52 MB) kopiert, 2,43484 s, 21,5 MB/s
# dd if=/dev/zero of=/dev/mapper/dm-test bs=1M count=50
52428800 Bytes (52 MB) kopiert, 2,4089 s, 21,8 MB/s
# dd if=/dev/zero of=/dev/mapper/dm-test bs=1M count=50
52428800 Bytes (52 MB) kopiert, 2,41091 s, 21,7 MB/s

After:
# cryptsetup create dm-test /dev/hda2 -c cast5-cbc-plain -s 128
Passsatz eingeben:
# dd if=/dev/zero of=/dev/mapper/dm-test bs=1M count=50
52428800 Bytes (52 MB) kopiert, 2,38128 s, 22,0 MB/s
# dd if=/dev/zero of=/dev/mapper/dm-test bs=1M count=50
52428800 Bytes (52 MB) kopiert, 2,29486 s, 22,8 MB/s
# dd if=/dev/zero of=/dev/mapper/dm-test bs=1M count=50
52428800 Bytes (52 MB) kopiert, 2,37162 s, 22,1 MB/s

Signed-off-by: Nicolas Kaiser <nikai@nikai.net>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>

8 years agocrypto: hash - Fix async import on shash algorithm
Herbert Xu [Thu, 4 Nov 2010 18:00:22 +0000]
crypto: hash - Fix async import on shash algorithm

The function shash_async_import did not initialise the descriptor
correctly prior to calling the underlying shash import function.

This patch adds the required initialisation.

Reported-by: Miloslav Trmac <mitr@redhat.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>

8 years agoMerge branch 'upstream-merge' of git://git.kernel.org/pub/scm/linux/kernel/git/tytso...
Linus Torvalds [Thu, 28 Oct 2010 04:54:31 +0000]
Merge branch 'upstream-merge' of git://git./linux/kernel/git/tytso/ext4

* 'upstream-merge' of git://git.kernel.org/pub/scm/linux/kernel/git/tytso/ext4: (50 commits)
  ext4,jbd2: convert tracepoints to use major/minor numbers
  ext4: optimize orphan_list handling for ext4_setattr
  ext4: fix unbalanced mutex unlock in error path of ext4_li_request_new
  ext4: fix compile error in ext4_fallocate()
  ext4: move ext4_mb_{get,put}_buddy_cache_lock and make them static
  ext4: rename mark_bitmap_end() to ext4_mark_bitmap_end()
  ext4: move flush_completed_IO to fs/ext4/fsync.c and make it static
  ext4: rename {ext,idx}_pblock and inline small extent functions
  ext4: make various ext4 functions be static
  ext4: rename {exit,init}_ext4_*() to ext4_{exit,init}_*()
  ext4: fix kernel oops if the journal superblock has a non-zero j_errno
  ext4: update writeback_index based on last page scanned
  ext4: implement writeback livelock avoidance using page tagging
  ext4: tidy up a void argument in inode.c
  ext4: add batched_discard into ext4 feature list
  ext4: Add batched discard support for ext4
  fs: Add FITRIM ioctl
  ext4: Use return value from sb_issue_discard()
  ext4: Check return value of sb_getblk() and friends
  ext4: use bio layer instead of buffer layer in mpage_da_submit_io
  ...

8 years agoMerge branch 'next' into upstream-merge
Theodore Ts'o [Thu, 28 Oct 2010 03:44:47 +0000]
Merge branch 'next' into upstream-merge

Conflicts:
fs/ext4/inode.c
fs/ext4/mballoc.c
include/trace/events/ext4.h

8 years agoMerge branch 'drm-core-next' of git://git.kernel.org/pub/scm/linux/kernel/git/airlied...
Linus Torvalds [Thu, 28 Oct 2010 03:37:06 +0000]
Merge branch 'drm-core-next' of git://git./linux/kernel/git/airlied/drm-2.6

* 'drm-core-next' of git://git.kernel.org/pub/scm/linux/kernel/git/airlied/drm-2.6:
  drm/radeon/kms: enable unmappable vram for evergreen
  drm/radeon/kms: fix tiled db height calculation on 6xx/7xx
  drm/radeon/kms: fix handling of tex lookup disable in cs checker on r2xx

8 years agoMerge branch 'for_linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jack/linux...
Linus Torvalds [Thu, 28 Oct 2010 03:13:18 +0000]
Merge branch 'for_linus' of git://git./linux/kernel/git/jack/linux-fs-2.6

* 'for_linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jack/linux-fs-2.6: (24 commits)
  quota: Fix possible oops in __dquot_initialize()
  ext3: Update kernel-doc comments
  jbd/2: fixed typos
  ext2: fixed typo.
  ext3: Fix debug messages in ext3_group_extend()
  jbd: Convert atomic_inc() to get_bh()
  ext3: Remove misplaced BUFFER_TRACE() in ext3_truncate()
  jbd: Fix debug message in do_get_write_access()
  jbd: Check return value of __getblk()
  ext3: Use DIV_ROUND_UP() on group desc block counting
  ext3: Return proper error code on ext3_fill_super()
  ext3: Remove unnecessary casts on bh->b_data
  ext3: Cleanup ext3_setup_super()
  quota: Fix issuing of warnings from dquot_transfer
  quota: fix dquot_disable vs dquot_transfer race v2
  jbd: Convert bitops to buffer fns
  ext3/jbd: Avoid WARN() messages when failing to write the superblock
  jbd: Use offset_in_page() instead of manual calculation
  jbd: Remove unnecessary goto statement
  jbd: Use printk_ratelimited() in journal_alloc_journal_head()
  ...

8 years agoext4,jbd2: convert tracepoints to use major/minor numbers
Theodore Ts'o [Thu, 28 Oct 2010 02:08:50 +0000]
ext4,jbd2: convert tracepoints to use major/minor numbers

Unfortunately perf can't deal with anything other than direct structure
accesses in the TP_printk() section.  It will drop dead when it sees
jbd2_dev_to_name() in the "print fmt" section of the tracepoint.

Addresses-Google-Bug: 3138508

Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>

8 years agoext4: optimize orphan_list handling for ext4_setattr
Dmitry Monakhov [Thu, 28 Oct 2010 02:08:46 +0000]
ext4: optimize orphan_list handling for ext4_setattr

Surprisingly chown() on ext4 is not SMP scalable operation.
Due to unconditional orphan_del(NULL, inode) in ext4_setattr()
result in significant performance overhead because of global orphan
mutex, especially in no-journal mode (where orphan_add() is noop).
It is possible to skip explicit orphan_del if possible.
Results of fchown() micro-benchmark in no-journal mode
while (1) {
   iteration++;
   fchown(fd, uid, gid);
   fchown(fd, uid + 1, gid + 1)
}
measured: iterations per millisecond
| nr_tasks | w/o patch | with patch |
|        1 |       142 |        185 |
|        4 |       109 |        642 |

Signed-off-by: Dmitry Monakhov <dmonakhov@openvz.org>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>

8 years agoext4: fix unbalanced mutex unlock in error path of ext4_li_request_new
Nicolas Kaiser [Thu, 28 Oct 2010 02:08:42 +0000]
ext4: fix unbalanced mutex unlock in error path of ext4_li_request_new

Signed-off-by: Nicolas Kaiser <nikai@nikai.net>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>

8 years agoMerge branch 'next' of git://git.kernel.org/pub/scm/linux/kernel/git/djbw/async_tx
Linus Torvalds [Thu, 28 Oct 2010 02:04:36 +0000]
Merge branch 'next' of git://git./linux/kernel/git/djbw/async_tx

* 'next' of git://git.kernel.org/pub/scm/linux/kernel/git/djbw/async_tx: (48 commits)
  DMAENGINE: move COH901318 to arch_initcall
  dma: imx-dma: fix signedness bug
  dma/timberdale: simplify conditional
  ste_dma40: remove channel_type
  ste_dma40: remove enum for endianess
  ste_dma40: remove TIM_FOR_LINK option
  ste_dma40: move mode_opt to separate config
  ste_dma40: move channel mode to a separate field
  ste_dma40: move priority to separate field
  ste_dma40: add variable to indicate valid dma_cfg
  async_tx: make async_tx channel switching opt-in
  move async raid6 test to lib/Kconfig.debug
  dmaengine: Add Freescale i.MX1/21/27 DMA driver
  intel_mid_dma: change the slave interface
  intel_mid_dma: fix the WARN_ONs
  intel_mid_dma: Add sg list support to DMA driver
  intel_mid_dma: Allow DMAC2 to share interrupt
  intel_mid_dma: Allow IRQ sharing
  intel_mid_dma: Add runtime PM support
  DMAENGINE: define a dummy filter function for ste_dma40
  ...

8 years agoMerge branch 'viafb-next' of git://github.com/schandinat/linux-2.6
Linus Torvalds [Thu, 28 Oct 2010 02:02:41 +0000]
Merge branch 'viafb-next' of git://github.com/schandinat/linux-2.6

* 'viafb-next' of git://github.com/schandinat/linux-2.6: (29 commits)
  viafb: add initial VX900 support
  viafb: fix hardware acceleration for suspend & resume
  viafb: make suspend and resume work (on all machines?)
  viafb: restore display on resume
  Minimal support for viafb suspend/resume
  viafb: use proper register for colour when doing fill ops
  viafb: add documentation for proc interface
  viafb: rename output devices
  viafb: add a mapping of supported output devices
  viafb: set sync polarity for all output devices
  viafb: add function to change sync polarity per device
  viafb: reduce I2C timeout and delay
  viafb: enable I2C for CRT
  viafb: fix i2c_transfer error handling
  viafb: vt1636 cleanup
  viafb: introduce per output device power management
  viafb: limit LCD code impact
  viafb: add interface for output device configuration
  viafb: merge the remaining output path with enable functions
  viafb: use new device routing
  ...

8 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/dhowells/linux-2.6-mn10300
Linus Torvalds [Thu, 28 Oct 2010 01:53:26 +0000]
Merge git://git./linux/kernel/git/dhowells/linux-2.6-mn10300

* git://git.kernel.org/pub/scm/linux/kernel/git/dhowells/linux-2.6-mn10300: (44 commits)
  MN10300: Save frame pointer in thread_info struct rather than global var
  MN10300: Change "Matsushita" to "Panasonic".
  MN10300: Create a defconfig for the ASB2364 board
  MN10300: Update the ASB2303 defconfig
  MN10300: ASB2364: Add support for SMSC911X and SMC911X
  MN10300: ASB2364: Handle the IRQ multiplexer in the FPGA
  MN10300: Generic time support
  MN10300: Specify an ELF HWCAP flag for MN10300 Atomic Operations Unit support
  MN10300: Map userspace atomic op regs as a vmalloc page
  MN10300: And Panasonic AM34 subarch and implement SMP
  MN10300: Delete idle_timestamp from irq_cpustat_t
  MN10300: Make various interrupt priority settings configurable
  MN10300: Optimise do_csum()
  MN10300: Implement atomic ops using atomic ops unit
  MN10300: Make the FPU operate in non-lazy mode under SMP
  MN10300: SMP TLB flushing
  MN10300: Use the [ID]PTEL2 registers rather than [ID]PTEL for TLB control
  MN10300: Make the use of PIDR to mark TLB entries controllable
  MN10300: Rename __flush_tlb*() to local_flush_tlb*()
  MN10300: AM34 erratum requires MMUCTR read and write on exception entry
  ...

8 years agoMerge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tiwai/sound-2.6
Linus Torvalds [Thu, 28 Oct 2010 01:52:49 +0000]
Merge branch 'for-linus' of git://git./linux/kernel/git/tiwai/sound-2.6

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tiwai/sound-2.6:
  ALSA: usb-audio: automatically detect feedback format
  ASoC: sound/wm9090: add missing __devexit marker
  ASoC: sound/max98088: add missing __devexit marker
  ASoC: sound/ad73311: add missing __devexit marker
  ASoC: fsl - fix build error in pcm030-audio-fabric.c
  sound/oss/sb_ess.c: delete double assignment
  ALSA: hda - Change BTL amp level on some HP notebooks

8 years agoMerge branch 'perf-fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Thu, 28 Oct 2010 01:48:00 +0000]
Merge branch 'perf-fixes-for-linus' of git://git./linux/kernel/git/tip/linux-2.6-tip

* 'perf-fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip: (50 commits)
  perf python scripting: Add futex-contention script
  perf python scripting: Fixup cut'n'paste error in sctop script
  perf scripting: Shut up 'perf record' final status
  perf record: Remove newline character from perror() argument
  perf python scripting: Support fedora 11 (audit 1.7.17)
  perf python scripting: Improve the syscalls-by-pid script
  perf python scripting: print the syscall name on sctop
  perf python scripting: Improve the syscalls-counts script
  perf python scripting: Improve the failed-syscalls-by-pid script
  kprobes: Remove redundant text_mutex lock in optimize
  x86/oprofile: Fix uninitialized variable use in debug printk
  tracing: Fix 'faild' -> 'failed' typo
  perf probe: Fix format specified for Dwarf_Off parameter
  perf trace: Fix detection of script extension
  perf trace: Use $PERF_EXEC_PATH in canned report scripts
  perf tools: Document event modifiers
  perf tools: Remove direct slang.h include
  perf_events: Fix for transaction recovery in group_sched_in()
  perf_events: Revert: Fix transaction recovery in group_sched_in()
  perf, x86: Use NUMA aware allocations for PEBS/BTS/DS allocations
  ...

8 years agoMerge branch 'module' of git://git.kernel.org/pub/scm/linux/kernel/git/rusty/linux...
Linus Torvalds [Thu, 28 Oct 2010 01:47:39 +0000]
Merge branch 'module' of git://git./linux/kernel/git/rusty/linux-2.6-for-linus

* 'module' of git://git.kernel.org/pub/scm/linux/kernel/git/rusty/linux-2.6-for-linus:
  NULL-terminate all pci_device_id tables
  (trivial) Fix compiler warning in kernel/modules.c

8 years agoMerge branch 'akpm-incoming-2'
Linus Torvalds [Thu, 28 Oct 2010 01:42:52 +0000]
Merge branch 'akpm-incoming-2'

* akpm-incoming-2: (139 commits)
  epoll: make epoll_wait() use the hrtimer range feature
  select: rename estimate_accuracy() to select_estimate_accuracy()
  Remove duplicate includes from many files
  ramoops: use the platform data structure instead of module params
  kernel/resource.c: handle reinsertion of an already-inserted resource
  kfifo: fix kfifo_alloc() to return a signed int value
  w1: don't allow arbitrary users to remove w1 devices
  alpha: remove dma64_addr_t usage
  mips: remove dma64_addr_t usage
  sparc: remove dma64_addr_t usage
  fuse: use release_pages()
  taskstats: use real microsecond granularity for CPU times
  taskstats: split fill_pid function
  taskstats: separate taskstats commands
  delayacct: align to 8 byte boundary on 64-bit systems
  delay-accounting: reimplement -c for getdelays.c to report information on a target command
  namespaces Kconfig: move namespace menu location after the cgroup
  namespaces Kconfig: remove the cgroup device whitelist experimental tag
  namespaces Kconfig: remove pointless cgroup dependency
  namespaces Kconfig: make namespace a submenu
  ...

8 years agoMerge branch 'x86-fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Thu, 28 Oct 2010 01:38:55 +0000]
Merge branch 'x86-fixes-for-linus' of git://git./linux/kernel/git/tip/linux-2.6-tip

* 'x86-fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip:
  percpu: Remove the multi-page alignment facility
  x86-32: Allocate irq stacks seperate from percpu area
  x86-32, mm: Remove duplicated #include
  x86, printk: Get rid of <0> from stack output
  x86, kexec: Make sure to stop all CPUs before exiting the kernel
  x86/vsmp: Eliminate kconfig dependency warning

8 years agoproc_bus_pci_ioctl: remove pointless BKL usage
Linus Torvalds [Thu, 28 Oct 2010 01:34:59 +0000]
proc_bus_pci_ioctl: remove pointless BKL usage

The BKL was pushed into this function when it was converted to use the
unlocked_ioctl interface, but nothing that the function touches is
actually protected by the BKL.  So just remove the BKL entirely, so that
we finally can get a realistic system build without the BKL being
enabled at all.

Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

8 years agoext4: fix compile error in ext4_fallocate()
Kazuya Mio [Thu, 28 Oct 2010 01:30:15 +0000]
ext4: fix compile error in ext4_fallocate()

When I compiled 2.6.36-rc3 kernel with EXT4FS_DEBUG definition, I got
the following compile error.

CC [M] fs/ext4/extents.o
fs/ext4/extents.c: In function 'ext4_fallocate':
fs/ext4/extents.c:3772: error: 'block' undeclared (first use in this function)
fs/ext4/extents.c:3772: error: (Each undeclared identifier is reported only once
fs/ext4/extents.c:3772: error: for each function it appears in.)
make[2]: *** [fs/ext4/extents.o] Error 1

The patch fixes this problem.

Signed-off-by: Kazuya Mio <k-mio@sx.jp.nec.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>

8 years agoext4: move ext4_mb_{get,put}_buddy_cache_lock and make them static
Eric Sandeen [Thu, 28 Oct 2010 01:30:15 +0000]
ext4: move ext4_mb_{get,put}_buddy_cache_lock and make them static

These functions are only used within fs/ext4/mballoc.c, so move them
so they are used after they are defined, and then make them be static.

Signed-off-by: Eric Sandeen <sandeen@redhat.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>

8 years agoext4: rename mark_bitmap_end() to ext4_mark_bitmap_end()
Theodore Ts'o [Thu, 28 Oct 2010 01:30:15 +0000]
ext4: rename mark_bitmap_end() to ext4_mark_bitmap_end()

Fix a namespace leak from fs/ext4

Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>

8 years agoext4: move flush_completed_IO to fs/ext4/fsync.c and make it static
Theodore Ts'o [Thu, 28 Oct 2010 01:30:14 +0000]
ext4: move flush_completed_IO to fs/ext4/fsync.c and make it static

Fix a namespace leak by moving the function to the file where it is
used and making it static.

Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>

8 years agoext4: rename {ext,idx}_pblock and inline small extent functions
Theodore Ts'o [Thu, 28 Oct 2010 01:30:14 +0000]
ext4: rename {ext,idx}_pblock and inline small extent functions

Cleanup namespace leaks from fs/ext4 and the inline trivial functions
ext4_{ext,idx}_pblock() and ext4_{ext,idx}_store_pblock() since the
code size actually shrinks when we make these functions inline,
they're so trivial.

Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>

8 years agoext4: make various ext4 functions be static
Theodore Ts'o [Thu, 28 Oct 2010 01:30:14 +0000]
ext4: make various ext4 functions be static

These functions have no need to be exported beyond file context.

No functions needed to be moved for this commit; just some function
declarations changed to be static and removed from header files.

(A similar patch was submitted by Eric Sandeen, but I wanted to handle
code movement in separate patches to make sure code changes didn't
accidentally get dropped.)

Signed-off-by: Eric Sandeen <sandeen@redhat.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>

8 years agoext4: rename {exit,init}_ext4_*() to ext4_{exit,init}_*()
Theodore Ts'o [Thu, 28 Oct 2010 01:30:14 +0000]
ext4: rename {exit,init}_ext4_*() to ext4_{exit,init}_*()

This is a cleanup to avoid namespace leaks out of fs/ext4

Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>

8 years agoext4: fix kernel oops if the journal superblock has a non-zero j_errno
Theodore Ts'o [Thu, 28 Oct 2010 01:30:13 +0000]
ext4: fix kernel oops if the journal superblock has a non-zero j_errno

Commit 84061e0 fixed an accounting bug only to introduce the
possibility of a kernel OOPS if the journal has a non-zero j_errno
field indicating that the file system had detected a fs inconsistency.
After the journal replay, if the journal superblock indicates that the
file system has an error, this indication is transfered to the file
system and then ext4_commit_super() is called to write this to the
disk.

But since the percpu counters are now initialized after the journal
replay, the call to ext4_commit_super() will cause a kernel oops since
it needs to use the percpu counters the ext4 superblock structure.

The fix is to skip setting the ext4 free block and free inode fields
if the percpu counter has not been set.

Thanks to Ken Sumrall for reporting and analyzing the root causes of
this bug.

Addresses-Google-Bug: #3054080

Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>

8 years agoext4: update writeback_index based on last page scanned
Eric Sandeen [Thu, 28 Oct 2010 01:30:13 +0000]
ext4: update writeback_index based on last page scanned

As pointed out in a prior patch, updating the mapping's
writeback_index based on pages written isn't quite right;
what the writeback index is really supposed to reflect is
the next page which should be scanned for writeback during
periodic flush.

As in write_cache_pages(), write_cache_pages_da() does
this scanning for us as we assemble the mpd for later
writeout.  If we keep track of the next page after the
current scan, we can easily update writeback_index without
worrying about pages written vs. pages skipped, etc.

Without this, an fsync will reset writeback_index to
0 (its starting index) + however many pages it wrote, which
can mess up the progress of periodic flush.

Signed-off-by: Eric Sandeen <sandeen@redhat.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>

8 years agoext4: implement writeback livelock avoidance using page tagging
Eric Sandeen [Thu, 28 Oct 2010 01:30:13 +0000]
ext4: implement writeback livelock avoidance using page tagging

This is analogous to Jan Kara's commit,
f446daaea9d4a420d16c606f755f3689dcb2d0ce
mm: implement writeback livelock avoidance using page tagging

but since we forked write_cache_pages, we need to reimplement
it there (and in ext4_da_writepages, since range_cyclic handling
was moved to there)

If you start a large buffered IO to a file, and then set
fsync after it, you'll find that fsync does not complete
until the other IO stops.

If you continue re-dirtying the file (say, putting dd
with conv=notrunc in a loop), when fsync finally completes
(after all IO is done), it reports via tracing that
it has written many more pages than the file contains;
in other words it has synced and re-synced pages in
the file multiple times.

This then leads to problems with our writeback_index
update, since it advances it by pages written, and
essentially sets writeback_index off the end of the
file...

With the following patch, we only sync as much as was
dirty at the time of the sync.

Signed-off-by: Eric Sandeen <sandeen@redhat.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>

8 years agoext4: tidy up a void argument in inode.c
Eric Sandeen [Thu, 28 Oct 2010 01:30:12 +0000]
ext4: tidy up a void argument in inode.c

This doesn't fix anything at all, it just removes a vestige
of prior use from __mpage_da_writepage()

__mpage_da_writepage() had a *void argument leftover from
its previous life as a callback; make it reflect the actual type.

Fixing this up makes it slightly more obvious to read, and
enables proper typechecking.

Signed-off-by: Eric Sandeen <sandeen@redhat.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>

8 years agoext4: add batched_discard into ext4 feature list
Lukas Czerner [Thu, 28 Oct 2010 01:30:12 +0000]
ext4: add batched_discard into ext4 feature list

Should be applied on the top of "lazy inode table initialization"
and "batched discard support" patch-sets.

Signed-off-by: Lukas Czerner <lczerner@redhat.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>

8 years agoext4: Add batched discard support for ext4
Lukas Czerner [Thu, 28 Oct 2010 01:30:12 +0000]
ext4: Add batched discard support for ext4

Walk through allocation groups and trim all free extents. It can be
invoked through FITRIM ioctl on the file system. The main idea is to
provide a way to trim the whole file system if needed, since some SSD's
may suffer from performance loss after the whole device was filled (it
does not mean that fs is full!).

It search for free extents in allocation groups specified by Byte range
start -> start+len. When the free extent is within this range, blocks
are marked as used and then trimmed. Afterwards these blocks are marked
as free in per-group bitmap.

Since fstrim is a long operation it is good to have an ability to
interrupt it by a signal. This was added by Dmitry Monakhov.
Thanks Dimitry.

Signed-off-by: Lukas Czerner <lczerner@redhat.com>
Signed-off-by: Dmitry Monakhov <dmonakhov@openvz.org>
Reviewed-by: Jan Kara <jack@suse.cz>
Reviewed-by: Dmitry Monakhov <dmonakhov@openvz.org>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>

8 years agofs: Add FITRIM ioctl
Lukas Czerner [Thu, 28 Oct 2010 01:30:11 +0000]
fs: Add FITRIM ioctl

Adds an filesystem independent ioctl to allow implementation of file
system batched discard support. I takes fstrim_range structure as an
argument. fstrim_range is definec in the include/fs.h and its
definition is as follows.

struct fstrim_range {
start;
len;
minlen;
}

start - first Byte to trim
len - number of Bytes to trim from start
minlen - minimum extent length to trim, free extents shorter than this
  number of Bytes will be ignored. This will be rounded up to fs
  block size.

It is also possible to specify NULL as an argument. In this case the
arguments will set itself as follows:

start = 0;
len = ULLONG_MAX;
minlen = 0;

So it will trim the whole file system at one run.

After the FITRIM is done, the number of actually discarded Bytes is stored
in fstrim_range.len to give the user better insight on how much storage
space has been really released for wear-leveling.

Signed-off-by: Lukas Czerner <lczerner@redhat.com>
Reviewed-by: Dmitry Monakhov <dmonakhov@openvz.org>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>

8 years agoext4: Use return value from sb_issue_discard()
Lukas Czerner [Thu, 28 Oct 2010 01:30:11 +0000]
ext4: Use return value from sb_issue_discard()

Use return value from sb_issue_discard() as return value in
ext4_issue_discard(). Since sb_issue_discard() may result in more
serious errors than just -EOPNOTSUPP it is worth to inform user of this
function about them to handle error cases properly.

Signed-off-by: Lukas Czerner <lczerner@redhat.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>

8 years agoext4: Check return value of sb_getblk() and friends
Namhyung Kim [Thu, 28 Oct 2010 01:30:11 +0000]
ext4: Check return value of sb_getblk() and friends

Fail block allocation if sb_getblk() returns NULL. In that case,
sb_find_get_block() also likely to fail so that it should skip
calling ext4_forget().

Signed-off-by: Namhyung Kim <namhyung@gmail.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>

8 years agoext4: use bio layer instead of buffer layer in mpage_da_submit_io
Theodore Ts'o [Thu, 28 Oct 2010 01:30:10 +0000]
ext4: use bio layer instead of buffer layer in mpage_da_submit_io

Call the block I/O layer directly instad of going through the buffer
layer.  This should give us much better performance and scalability,
as well as lowering our CPU utilization when doing buffered writeback.

Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>

8 years agoext4: move mpage_put_bnr_to_bhs()'s functionality to mpage_da_submit_io()
Theodore Ts'o [Thu, 28 Oct 2010 01:30:10 +0000]
ext4: move mpage_put_bnr_to_bhs()'s functionality to mpage_da_submit_io()

This massively simplifies the ext4_da_writepages() code path by
completely removing mpage_put_bnr_bhs(), which is almost 100 lines of
code iterating over a set of pages using pagevec_lookup(), and folds
that functionality into mpage_da_submit_io()'s existing
pagevec_lookup() loop.

Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>

8 years agoext4: inline walk_page_buffers() into mpage_da_submit_io
Theodore Ts'o [Thu, 28 Oct 2010 01:30:10 +0000]
ext4: inline walk_page_buffers() into mpage_da_submit_io

Expand the call:

  if (walk_page_buffers(NULL, page_bufs, 0, len, NULL,
                        ext4_bh_delay_or_unwritten))
goto redirty_page

into mpage_da_submit_io().

This will allow us to merge in mpage_put_bnr_to_bhs() in the next
patch.

Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>

8 years agoext4: inline ext4_writepage() into mpage_da_submit_io()
Theodore Ts'o [Thu, 28 Oct 2010 01:30:09 +0000]
ext4: inline ext4_writepage() into mpage_da_submit_io()

As a prepratory step to switching to bio_submit, inline
ext4_writepage() into mpage_da_submit() and then simplify things a
bit.  This makes it clearer what mpage_da_submit needs to do.

Also, move the ClearPageChecked(page) call into
__ext4_journalled_writepage(), as a minor bit of cleanup refactoring.

This also allows us to pull i_size_read() and
ext4_should_journal_data() out of the loop, which should be a very
minor CPU savings.

Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>

8 years agoext4: simplify ext4_writepage()
Theodore Ts'o [Thu, 28 Oct 2010 01:30:09 +0000]
ext4: simplify ext4_writepage()

The actual code in ext4_writepage() is unnecessarily convoluted.
Simplify it so it is easier to understand, but otherwise logically
equivalent.

Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>

8 years agoext4: call mpage_da_submit_io() from mpage_da_map_blocks()
Theodore Ts'o [Thu, 28 Oct 2010 01:30:09 +0000]
ext4: call mpage_da_submit_io() from mpage_da_map_blocks()

Eventually we need to completely reorganize the ext4 writepage
callpath, but for now, we simplify things a little by calling
mpage_da_submit_io() from mpage_da_map_blocks(), since all of the
places where we call mpage_da_map_blocks() it is followed up by a call
to mpage_da_submit_io().

We're also a wee bit better with respect to error handling, but there
are still a number of issues where it's not clear what the right thing
is to do with ext4 functions deep in the writeback codepath fails.

Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>

8 years agoext4: use KMEM_CACHE instead of kmem_cache_create
Theodore Ts'o [Thu, 28 Oct 2010 01:30:09 +0000]
ext4: use KMEM_CACHE instead of kmem_cache_create

Also remove the SLAB_RECLAIM_ACCOUNT flag from the system zone kmem
cache.  This slab tends to be fairly static, so it shouldn't be marked
as likely to have free pages that can be reclaimed.

Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>

8 years agoext4: use search_dirblock() in ext4_dx_find_entry()
Theodore Ts'o [Thu, 28 Oct 2010 01:30:08 +0000]
ext4: use search_dirblock() in ext4_dx_find_entry()

Use the search_dirblock() in ext4_dx_find_entry().  It makes the code
easier to read, and it takes advantage of common code.  It also saves
100 bytes or so of text space.

Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
Cc: Brad Spengler <spender@grsecurity.net>

8 years agoext4: avoid uninitialized memory references in ext3_htree_next_block()
Theodore Ts'o [Thu, 28 Oct 2010 01:30:08 +0000]
ext4: avoid uninitialized memory references in ext3_htree_next_block()

If the first block of htree directory is missing '.' or '..' but is
otherwise a valid directory, and we do a lookup for '.' or '..', it's
possible to dereference an uninitialized memory pointer in
ext4_htree_next_block().

We avoid this by moving the special case from ext4_dx_find_entry() to
ext4_find_entry(); this also means we can optimize ext4_find_entry()
slightly when NFS looks up "..".

Thanks to Brad Spengler for pointing a Clang warning that led me to
look more closely at this code.  The warning was harmless, but it was
useful in pointing out code that was too ugly to live.  This warning was
also reported by Roman Borisov.

Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
Cc: Brad Spengler <spender@grsecurity.net>

8 years agoext4: remove unused ext4_sb_info members
Eric Sandeen [Thu, 28 Oct 2010 01:30:08 +0000]
ext4: remove unused ext4_sb_info members

Not that these take up a lot of room, but the structure is long enough
as it is, and there's no need to confuse people with these various
undocumented & unused structure members...

Signed-off-by: Eric Sandeen <sandeen@redaht.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>

8 years agoext4: queue conversion after adding to inode's completed IO list
Eric Sandeen [Thu, 28 Oct 2010 01:30:07 +0000]
ext4: queue conversion after adding to inode's completed IO list

By queuing the io end on the unwritten workqueue before adding it
to our inode's list of completed IOs, I think we run the risk
of the work getting completed, and the IO freed, before we try
to add it to the inode's i_completed_io_list.

It should be safe to add it to the inode's list of completed
IOs, and -then- queue it for completion, I think.

Thanks to Dave Chinner for pointing out the race.

Signed-off-by: Eric Sandeen <sandeen@redhat.com>
Reviewed-by: Jiaying Zhang <jiayingz@google.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>

8 years agoext4: don't use ext4_allocation_contexts for tracing
Eric Sandeen [Thu, 28 Oct 2010 01:30:07 +0000]
ext4: don't use ext4_allocation_contexts for tracing

Many tracepoints were populating an ext4_allocation_context
to pass in, but this requires a slab allocation even when
tracepoints are off.  In fact, 4 of 5 of these allocations
were only for tracing.  In addition, we were only using a
small fraction of the 144 bytes of this structure for this
purpose.

We can do away with all these alloc/frees of the ac and
simply pass in the bits we care about, instead.

I tested this by turning on tracing and running through
xfstests on x86_64.  I did not actually do anything with
the trace output, however.

Signed-off-by: Eric Sandeen <sandeen@redhat.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>

8 years agoext4: fix oops in trace_ext4_mb_release_group_pa
Eric Sandeen [Thu, 28 Oct 2010 01:30:07 +0000]
ext4: fix oops in trace_ext4_mb_release_group_pa

Our QA reported an oops in the ext4_mb_release_group_pa tracing,
and Josef Bacik pointed out that it was because we may have a
non-null but uninitialized ac_inode in the allocation context.

I can reproduce it when running xfstests with ext4 tracepoints on,
on a CONFIG_SLAB_DEBUG kernel.

We call trace_ext4_mb_release_group_pa from 2 places,
ext4_mb_discard_group_preallocations and
ext4_mb_discard_lg_preallocations

In both cases we allocate an ac as a container just for tracing (!)
and never fill in the ac_inode.  There's no reason to be assigning,
testing, or printing it as far as I can see, so just remove it from
the tracepoint.

Signed-off-by: Eric Sandeen <sandeen@redhat.com>
Reviewed-by: Josef Bacik <josef@redhat.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>

8 years agoext4: fix potential infinite loop in ext4_da_writepages()
Toshiyuki Okajima [Thu, 28 Oct 2010 01:30:07 +0000]
ext4: fix potential infinite loop in ext4_da_writepages()

On linux-2.6.36-rc2, if we execute the following script, we can hang
the system when the /bin/sync command is executed:

========================================================================
#!/bin/sh

echo -n "HANG UP TEST: "
/bin/dd if=/dev/zero of=/tmp/img bs=1k count=1 seek=1M 2> /dev/null
/sbin/mkfs.ext4 -Fq /tmp/img
/bin/mount -o loop -t ext4 /tmp/img /mnt
/bin/dd if=/dev/zero of=/mnt/file bs=1 count=1 \
seek=$((16*1024*1024*1024*1024-4096)) 2> /dev/null
/bin/sync
/bin/umount /mnt
echo "DONE"
exit 0
========================================================================

We can see the following backtrace if we get the kdump when this
hangup occurs:

======================================================================
kthread()
=> bdi_writeback_thread()
   => wb_do_writeback()
      => wb_writeback()
         => writeback_inodes_wb()
            => writeback_sb_inodes()
               => writeback_single_inode()
                  => ext4_da_writepages()  ---+
                                ^ infinite    |
                                |   loop      |
                                +-------------+
======================================================================

The reason why this hangup happens is described as follows:
1) We write the last extent block of the file whose size is the filesystem
   maximum size.
2) "BH_Delay" flag is set on the buffer_head of its block.
3) - the member, "m_lblk" of struct mpage_da_data is 4294967295 (UINT_MAX)
   - the member, "m_len" of struct mpage_da_data is 1
  mpage_put_bnr_to_bhs() which is called via ext4_da_writepages()
  cannot clear "BH_Delay" flag of the buffer_head because the type of
  m_lblk is ext4_lblk_t and then m_lblk + m_len is overflow.

  Therefore an infinite loop occurs because ext4_da_writepages()
  cannot write the page (which corresponds to the block) since
  "BH_Delay" flag isn't cleared.
----------------------------------------------------------------------
static void mpage_put_bnr_to_bhs(struct mpage_da_data *mpd,
struct ext4_map_blocks *map)
{
...
int blocks = map->m_len;
...
do {
// cur_logical = 4294967295
// map->m_lblk = 4294967295
// blocks = 1
// *** map->m_lblk + blocks == 0 (OVERFLOW!) ***
// (cur_logical >= map->m_lblk + blocks) => true
if (cur_logical >= map->m_lblk + blocks)
break;
----------------------------------------------------------------------

NOTE: Mounting with the nodelalloc option will avoid this codepath,
and thus, avoid this hang

Signed-off-by: Toshiyuki Okajima <toshi.okajima@jp.fujitsu.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>

8 years agoext4: improve llseek error handling for overly large seek offsets
Toshiyuki Okajima [Thu, 28 Oct 2010 01:30:06 +0000]
ext4: improve llseek error handling for overly large seek offsets

The llseek system call should return EINVAL if passed a seek offset
which results in a write error.  What this maximum offset should be
depends on whether or not the huge_file file system feature is set,
and whether or not the file is extent based or not.

If the file has no "EXT4_EXTENTS_FL" flag, the maximum size which can be
written (write systemcall) is different from the maximum size which can be
sought (lseek systemcall).

For example, the following 2 cases demonstrates the differences
between the maximum size which can be written, versus the seek offset
allowed by the llseek system call:

#1: mkfs.ext3 <dev>; mount -t ext4 <dev>
#2: mkfs.ext3 <dev>; tune2fs -Oextent,huge_file <dev>; mount -t ext4 <dev>

Table. the max file size which we can write or seek
       at each filesystem feature tuning and file flag setting
+============+===============================+===============================+
| \ File flag|                               |                               |
|      \     |     !EXT4_EXTENTS_FL          |        EXT4_EXTETNS_FL        |
|case       \|                               |                               |
+------------+-------------------------------+-------------------------------+
| #1         |   write:      2194719883264   | write:       --------------   |
|            |   seek:       2199023251456   | seek:        --------------   |
+------------+-------------------------------+-------------------------------+
| #2         |   write:      4402345721856   | write:       17592186044415   |
|            |   seek:      17592186044415   | seek:        17592186044415   |
+------------+-------------------------------+-------------------------------+

The differences exist because ext4 has 2 maxbytes which are sb->s_maxbytes
(= extent-mapped maxbytes) and EXT4_SB(sb)->s_bitmap_maxbytes (= block-mapped
maxbytes).  Although generic_file_llseek uses only extent-mapped maxbytes.
(llseek of ext4_file_operations is generic_file_llseek which uses
sb->s_maxbytes.)

Therefore we create ext4 llseek function which uses 2 maxbytes.

The new own function originates from generic_file_llseek().
If the file flag, "EXT4_EXTENTS_FL" is not set, the function alters
inode->i_sb->s_maxbytes into EXT4_SB(inode->i_sb)->s_bitmap_maxbytes.

Signed-off-by: Toshiyuki Okajima <toshi.okajima@jp.fujitsu.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
Cc: Andreas Dilger <adilger.kernel@dilger.ca>

8 years agoext4: don't update sb journal_devnum when RO dev
Maciej ┼╗enczykowski [Thu, 28 Oct 2010 01:30:06 +0000]
ext4: don't update sb journal_devnum when RO dev

An ext4 filesystem on a read-only device, with an external journal
which is at a different device number then recorded in the superblock
will fail to honor the read-only setting of the device and trigger
a superblock update (write).

For example:
  - ext4 on a software raid which is in read-only mode
  - external journal on a read-write device which has changed device num
  - attempt to mount with -o journal_dev=<new_number>
  - hits BUG_ON(mddev->ro = 1) in md.c

Cc: Theodore Ts'o <tytso@mit.edu>
Signed-off-by: Maciej ┼╗enczykowski <zenczykowski@gmail.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>

8 years agoext4: use sb_issue_zeroout in ext4_ext_zeroout
Lukas Czerner [Thu, 28 Oct 2010 01:30:06 +0000]
ext4: use sb_issue_zeroout in ext4_ext_zeroout

Change ext4_ext_zeroout to use sb_issue_zeroout instead of its
own approach to zero out extents.

Signed-off-by: Lukas Czerner <lczerner@redhat.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>

8 years agoext4: use sb_issue_zeroout in setup_new_group_blocks
Lukas Czerner [Thu, 28 Oct 2010 01:30:05 +0000]
ext4: use sb_issue_zeroout in setup_new_group_blocks

Use sb_issue_zeroout to zero out inode table and descriptor table
blocks instead of old approach which involves journaling.

Signed-off-by: Lukas Czerner <lczerner@redhat.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>

8 years agoext4: add interface to advertise ext4 features in sysfs
Lukas Czerner [Thu, 28 Oct 2010 01:30:05 +0000]
ext4: add interface to advertise ext4 features in sysfs

User-space should have the opportunity to check what features doest ext4
support in each particular copy. This adds easy interface by creating new
"features" directory in sys/fs/ext4/. In that directory files
advertising feature names can be created.

Add lazy_itable_init to the feature list.

Signed-off-by: Lukas Czerner <lczerner@redhat.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>

8 years agoext4: add support for lazy inode table initialization
Lukas Czerner [Thu, 28 Oct 2010 01:30:05 +0000]
ext4: add support for lazy inode table initialization

When the lazy_itable_init extended option is passed to mke2fs, it
considerably speeds up filesystem creation because inode tables are
not zeroed out.  The fact that parts of the inode table are
uninitialized is not a problem so long as the block group descriptors,
which contain information regarding how much of the inode table has
been initialized, has not been corrupted However, if the block group
checksums are not valid, e2fsck must scan the entire inode table, and
the the old, uninitialized data could potentially cause e2fsck to
report false problems.

Hence, it is important for the inode tables to be initialized as soon
as possble.  This commit adds this feature so that mke2fs can safely
use the lazy inode table initialization feature to speed up formatting
file systems.

This is done via a new new kernel thread called ext4lazyinit, which is
created on demand and destroyed, when it is no longer needed.  There
is only one thread for all ext4 filesystems in the system. When the
first filesystem with inititable mount option is mounted, ext4lazyinit
thread is created, then the filesystem can register its request in the
request list.

This thread then walks through the list of requests picking up
scheduled requests and invoking ext4_init_inode_table(). Next schedule
time for the request is computed by multiplying the time it took to
zero out last inode table with wait multiplier, which can be set with
the (init_itable=n) mount option (default is 10).  We are doing
this so we do not take the whole I/O bandwidth. When the thread is no
longer necessary (request list is empty) it frees the appropriate
structures and exits (and can be created later later by another
filesystem).

We do not disturb regular inode allocations in any way, it just do not
care whether the inode table is, or is not zeroed. But when zeroing, we
have to skip used inodes, obviously. Also we should prevent new inode
allocations from the group, while zeroing is on the way. For that we
take write alloc_sem lock in ext4_init_inode_table() and read alloc_sem
in the ext4_claim_inode, so when we are unlucky and allocator hits the
group which is currently being zeroed, it just has to wait.

This can be suppresed using the mount option no_init_itable.

Signed-off-by: Lukas Czerner <lczerner@redhat.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>

8 years agoAdd helper function for blkdev_issue_zeroout (sb_issue_discard)
Lukas Czerner [Thu, 28 Oct 2010 01:30:04 +0000]
Add helper function for blkdev_issue_zeroout (sb_issue_discard)

This is done the same way as helper sb_issue_discard for
blkdev_issue_discard.

Signed-off-by: Lukas Czerner <lczerner@redhat.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>

8 years agojbd2: Add sanity check for attempts to start handle during umount
Theodore Ts'o [Thu, 28 Oct 2010 01:30:04 +0000]
jbd2: Add sanity check for attempts to start handle during umount

An attempt to modify the file system during the call to
jbd2_destroy_journal() can lead to a system lockup.  So add some
checking to make it much more obvious when this happens to and to
determine where the offending code is located.

Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>

8 years agoext4: fix NULL pointer dereference in print_daily_error_info()
Sergey Senozhatsky [Thu, 28 Oct 2010 01:30:04 +0000]
ext4: fix NULL pointer dereference in print_daily_error_info()

Fix NULL pointer dereference in print_daily_error_info, when
called on unmounted fs (EXT4_SB(sb) returns NULL), by removing error
reporting timer in ext4_put_super.

Google-Bug-Id: 3017663

Signed-off-by: Sergey Senozhatsky <sergey.senozhatsky@gmail.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>

8 years agoext4: don't hold spinlock while calling ext4_issue_discard()
Lukas Czerner [Thu, 28 Oct 2010 01:30:04 +0000]
ext4: don't hold spinlock while calling ext4_issue_discard()

We can't hold the block group spinlock because we ext4_issue_discard()
calls wait and hence can get rescheduled.

Google-Bug-Id: 3017678

Signed-off-by: Lukas Czerner <lczerner@redhat.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>

8 years agoext4: check for negative error code from sb_issue_discard
Lukas Czerner [Thu, 28 Oct 2010 01:30:03 +0000]
ext4: check for negative error code from sb_issue_discard

sb_issue_discard() is returning negative error code, so check for
-EOPNOTSUPP.

Signed-off-by: Lukas Czerner <lczerner@redhat.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>

8 years agoext4: don't bump up LONG_MAX nr_to_write by a factor of 8
Eric Sandeen [Thu, 28 Oct 2010 01:30:03 +0000]
ext4: don't bump up LONG_MAX nr_to_write by a factor of 8

I'm uneasy with lots of stuff going on in ext4_da_writepages(),
but bumping nr_to_write from LLONG_MAX to -8 clearly isn't
making anything better, so avoid the multiplier in that case.

Signed-off-by: Eric Sandeen <sandeen@redhat.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>

8 years agoext4: stop looping in ext4_num_dirty_pages when max_pages reached
Eric Sandeen [Thu, 28 Oct 2010 01:30:03 +0000]
ext4: stop looping in ext4_num_dirty_pages when max_pages reached

Today we simply break out of the inner loop when we have accumulated
max_pages; this keeps scanning forwad and doing pagevec_lookup_tag()
in the while (!done) loop, this does potentially a lot of work
with no net effect.

When we have accumulated max_pages, just clean up and return.

Signed-off-by: Eric Sandeen <sandeen@redhat.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>

8 years agoext4: use dedicated slab caches for group_info structures
Curt Wohlgemuth [Thu, 28 Oct 2010 01:29:12 +0000]
ext4: use dedicated slab caches for group_info structures

ext4_group_info structures are currently allocated with kmalloc().
With a typical 4K block size, these are 136 bytes each -- meaning
they'll each consume a 256-byte slab object.  On a system with many
ext4 large partitions, that's a lot of wasted kernel slab space.
(E.g., a single 1TB partition will have about 8000 block groups, using
about 2MB of slab, of which nearly 1MB is wasted.)

This patch creates an array of slab pointers created as needed --
depending on the superblock block size -- and uses these slabs to
allocate the group info objects.

Google-Bug-Id: 2980809

Signed-off-by: Curt Wohlgemuth <curtw@google.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>

8 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/davem/net-2.6
Linus Torvalds [Thu, 28 Oct 2010 01:28:00 +0000]
Merge git://git./linux/kernel/git/davem/net-2.6

* git://git.kernel.org/pub/scm/linux/kernel/git/davem/net-2.6: (108 commits)
  ehea: Fixing statistics
  bonding: Fix lockdep warning after bond_vlan_rx_register()
  tunnels: Fix tunnels change rcu protection
  caif-u5500: Build config for CAIF shared mem driver
  caif-u5500: CAIF shared memory mailbox interface
  caif-u5500: CAIF shared memory transport protocol
  caif-u5500: Adding shared memory include
  drivers/isdn: delete double assignment
  drivers/net/typhoon.c: delete double assignment
  drivers/net/sb1000.c: delete double assignment
  qlcnic: define valid vlan id range
  qlcnic: reduce rx ring size
  qlcnic: fix mac learning
  ehea: fix use after free
  inetpeer: __rcu annotations
  fib_rules: __rcu annotates ctarget
  tunnels: add __rcu annotations
  net: add __rcu annotations to protocol
  ipv4: add __rcu annotations to routes.c
  qlge: bugfix: Restoring the vlan setting.
  ...

8 years agodrm/radeon/kms: enable unmappable vram for evergreen
Alex Deucher [Tue, 19 Oct 2010 04:36:19 +0000]
drm/radeon/kms: enable unmappable vram for evergreen

Evergreen now has blit support, but unmappable vram support
was disabled in c919b371cb734f42b1130e706ecee262f8d9261d
(drm/radeon/kms: avoid corner case issue with unmappable vram V2)
due to merge ordering.  This re-enables unmappable vram on
evergreen.

Signed-off-by: Alex Deucher <alexdeucher@gmail.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>

8 years agoext4: avoid null dereference in trace_ext4_mballoc_discard
Wen Congyang [Thu, 28 Oct 2010 01:27:12 +0000]
ext4: avoid null dereference in trace_ext4_mballoc_discard

ac->inode is set to null in function ext4_mb_release_group_pa(),
and then trace_ext4_mballoc_discard(ac) is called, the kernel
will panic.

BUG: unable to handle kernel NULL pointer dereference at 000000a4
IP: [<f87e1714>] ftrace_raw_event_ext4__mballoc+0x54/0xc0 [ext4]
*pdpt = 0000000000abd001 *pde = 0000000000000000
Oops: 0000 [#1] SMP

Pid: 550, comm: flush-8:16 Not tainted 2.6.36-rc1 #1 SE7320EP2/Altos G530
EIP: 0060:[<f87e1714>] EFLAGS: 00010206 CPU: 1
EIP is at ftrace_raw_event_ext4__mballoc+0x54/0xc0 [ext4]
EAX: f32ac840 EBX: f3f1cf88 ECX: f32ac840 EDX: 00000000
ESI: f32ac83c EDI: f880b9d8 EBP: 00000000 ESP: f4b77ae4
 DS: 007b ES: 007b FS: 00d8 GS: 00e0 SS: 0068
Process flush-8:16 (pid: 550, ti=f4b76000 task=f613e540 task.ti=f4b76000)
Call Trace:
 [<f87f5ac1>] ? ext4_mb_release_group_pa+0x121/0x150 [ext4]
 [<f87f8356>] ? ext4_mb_discard_group_preallocations+0x336/0x400 [ext4]
 [<f87fb7f1>] ? ext4_mb_new_blocks+0x3d1/0x4f0 [ext4]
 [<c05a6c5b>] ? __make_request+0x10b/0x440
 [<f87f1fb4>] ? ext4_ext_map_blocks+0x1334/0x1980 [ext4]
 [<c04ac78a>] ? rb_reserve_next_event+0xaa/0x3b0
 [<f87d18d6>] ? ext4_map_blocks+0xd6/0x1d0 [ext4]
 [<f87d2da7>] ? mpage_da_map_blocks+0xc7/0x8a0 [ext4]
 [<c04c8a68>] ? find_get_pages_tag+0x38/0x110
 [<c04d23a5>] ? __pagevec_release+0x15/0x20
 [<f87d3ca5>] ? ext4_da_writepages+0x2b5/0x5d0 [ext4]
 [<c04cfbe0>] ? __writepage+0x0/0x30
 [<c04d0e34>] ? do_writepages+0x14/0x30
 [<c0526600>] ? writeback_single_inode+0xa0/0x240
 [<c0526971>] ? writeback_sb_inodes+0xc1/0x180
 [<c0526ab8>] ? writeback_inodes_wb+0x88/0x140
 [<c0526d7b>] ? wb_writeback+0x20b/0x320
 [<c045aca7>] ? lock_timer_base+0x27/0x50
 [<c0526fe0>] ? wb_do_writeback+0x150/0x190
 [<c05270a8>] ? bdi_writeback_thread+0x88/0x1f0
 [<c043b680>] ? complete+0x40/0x60
 [<c0527020>] ? bdi_writeback_thread+0x0/0x1f0
 [<c0469474>] ? kthread+0x74/0x80
 [<c0469400>] ? kthread+0x0/0x80
 [<c040a23e>] ? kernel_thread_helper+0x6/0x10

Signed-off-by: Wen Congyang <wency@cn.fujitsu.com>
Acked-by: Steven Rostedt <rostedt@goodmis.org>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>

8 years agodrm/radeon/kms: fix tiled db height calculation on 6xx/7xx
Alex Deucher [Wed, 27 Oct 2010 05:44:35 +0000]
drm/radeon/kms: fix tiled db height calculation on 6xx/7xx

Calculate height based on the slice bitfield rather than the size.
Same as Dave's CB fix.

Signed-off-by: Alex Deucher <alexdeucher@gmail.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>

8 years agodrm/radeon/kms: fix handling of tex lookup disable in cs checker on r2xx
Alex Deucher [Wed, 27 Oct 2010 05:02:35 +0000]
drm/radeon/kms: fix handling of tex lookup disable in cs checker on r2xx

There are cases when multiple texture units have to be enabled,
but not actually used to sample.  This patch checks to see if
the lookup_disable bit is set and if so, skips the texture check.

Fixes:
https://bugs.freedesktop.org/show_bug.cgi?id=25544

Signed-off-by: Alex Deucher <alexdeucher@gmail.com>
Cc: stable@kernel.org
Signed-off-by: Dave Airlie <airlied@redhat.com>

8 years agojbd2: Fix I/O hang in jbd2_journal_release_jbd_inode
Brian King [Thu, 28 Oct 2010 01:25:12 +0000]
jbd2: Fix I/O hang in jbd2_journal_release_jbd_inode

This fixes a hang seen in jbd2_journal_release_jbd_inode
on a lot of Power 6 systems running with ext4. When we get
in the hung state, all I/O to the disk in question gets blocked
where we stay indefinitely. Looking at the task list, I can see
we are stuck in jbd2_journal_release_jbd_inode waiting on a
wake up. I added some debug code to detect this scenario and
dump additional data if we were stuck in jbd2_journal_release_jbd_inode
for longer than 30 minutes. When it hit, I was able to see that
i_flags was 0, suggesting we missed the wake up.

This patch changes i_flags to be an unsigned long, uses bit operators
to access it, and adds barriers around the accesses. Prior to applying
this patch, we were regularly hitting this hang on numerous systems
in our test environment. After applying the patch, the hangs no longer
occur.

Signed-off-by: Brian King <brking@linux.vnet.ibm.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>

8 years agoext4: fix EOFBLOCKS_FL handling
Theodore Ts'o [Thu, 28 Oct 2010 01:23:12 +0000]
ext4: fix EOFBLOCKS_FL handling

It turns out we have several problems with how EOFBLOCKS_FL is
handled.  First of all, there was a fencepost error where we were not
clearing the EOFBLOCKS_FL when fill in the last uninitialized block,
but rather when we allocate the next block _after_ the uninitalized
block.  Secondly we were not testing to see if we needed to clear the
EOFBLOCKS_FL when writing to the file O_DIRECT or when were converting
an uninitialized block (which is the most common case).

Google-Bug-Id: 2928259

Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>

8 years agofasync: Fix placement of FASYNC flag comment
Linus Torvalds [Thu, 28 Oct 2010 01:17:02 +0000]
fasync: Fix placement of FASYNC flag comment

In commit f7347ce4ee7c ("fasync: re-organize fasync entry insertion to
allow it under a spinlock") Arnd took an earlier patch of mine that had
the comment about the FASYNC flag above the wrong function.

When the fasync_add_entry() function was split to introduce the new
fasync_insert_entry() helper function, the code that actually cares
about the FASYNC bit moved to that new helper.

So just move the comment to the right point.

Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

8 years agoMerge branch 'flock' of git://git.kernel.org/pub/scm/linux/kernel/git/arnd/bkl
Linus Torvalds [Thu, 28 Oct 2010 01:13:34 +0000]
Merge branch 'flock' of git://git./linux/kernel/git/arnd/bkl

* 'flock' of git://git.kernel.org/pub/scm/linux/kernel/git/arnd/bkl:
  locks: turn lock_flocks into a spinlock
  fasync: re-organize fasync entry insertion to allow it under a spinlock
  locks/nfsd: allocate file lock outside of spinlock
  lockd: fix nlmsvc_notify_blocked locking
  lockd: push lock_flocks down

8 years agoepoll: make epoll_wait() use the hrtimer range feature
Shawn Bohrer [Wed, 27 Oct 2010 22:34:54 +0000]
epoll: make epoll_wait() use the hrtimer range feature

This make epoll use hrtimers for the timeout value which prevents
epoll_wait() from timing out up to a millisecond early.

This mirrors the behavior of select() and poll().

Signed-off-by: Shawn Bohrer <shawn.bohrer@gmail.com>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Acked-by: Davide Libenzi <davidel@xmailserver.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

8 years agoselect: rename estimate_accuracy() to select_estimate_accuracy()
Andrew Morton [Wed, 27 Oct 2010 22:34:53 +0000]
select: rename estimate_accuracy() to select_estimate_accuracy()

Make it a subsystem-specific identifier because we wish to amke it
non-static in the next patch ("epoll: make epoll_wait() use the hrtimer
range feature").

Cc: Shawn Bohrer <shawn.bohrer@gmail.com>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Cc: Davide Libenzi <davidel@xmailserver.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

8 years agoRemove duplicate includes from many files
Zimny Lech [Wed, 27 Oct 2010 22:34:53 +0000]
Remove duplicate includes from many files

Signed-off-by: Zimny Lech <napohybelskurwysynom2010@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

8 years agoramoops: use the platform data structure instead of module params
Kyungmin Park [Wed, 27 Oct 2010 22:34:52 +0000]
ramoops: use the platform data structure instead of module params

As each board and system has different memory for ramoops.  It's better to
define the platform data instead of module params.

[akpm@linux-foundation.org: fix ramoops_remove() return type]
Signed-off-by: Kyungmin Park <kyungmin.park@samsung.com>
Cc: Marco Stornelli <marco.stornelli@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

8 years agokernel/resource.c: handle reinsertion of an already-inserted resource
Huang Shijie [Wed, 27 Oct 2010 22:34:52 +0000]
kernel/resource.c: handle reinsertion of an already-inserted resource

If the same resource is inserted to the resource tree (maybe not on
purpose), a dead loop will be created.  In this situation, The kernel does
not report any warning or error :(

  The command below will show a endless print.
  #cat /proc/iomem

[akpm@linux-foundation.org: add WARN_ON()]
Signed-off-by: Huang Shijie <shijie8@gmail.com>
Cc: Jesse Barnes <jbarnes@virtuousgeek.org>
Cc: Bjorn Helgaas <bjorn.helgaas@hp.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

8 years agokfifo: fix kfifo_alloc() to return a signed int value
Stefani Seibold [Wed, 27 Oct 2010 22:34:50 +0000]
kfifo: fix kfifo_alloc() to return a signed int value

Add a new __kfifo_int_must_check_helper() helper function, which is needed
for kfifo_alloc() to return the right signed integer value.

The origin __kfifo_must_check_helper() helper was renamed into
__kfifo_uint_must_check_helper() to show the sign which is expected and
returned.

(And revert the temporary disabling of __kfifo_must_check_helper())

Signed-off-by: Stefani Seibold <stefani@seibold.net>
Acked-by: Randy Dunlap <randy.dunlap@oracle.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

8 years agow1: don't allow arbitrary users to remove w1 devices
Brian Swetland [Wed, 27 Oct 2010 22:34:49 +0000]
w1: don't allow arbitrary users to remove w1 devices

The search/pullup/add/remove device attributes were 0666 which would allow
arbitrary users to affect the 1 wire bus.  Change to 0664 to prevent that.

I found this patch in the Android tree, apparently this has never been
sent upstream so doing it now.

Signed-off-by: Brian Swetland <swetland@google.com>
Signed-off-by: Linus Walleij <linus.walleij@stericsson.com>
Cc: Evgeniy Polyakov <johnpol@2ka.mipt.ru>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

8 years agoalpha: remove dma64_addr_t usage
FUJITA Tomonori [Wed, 27 Oct 2010 22:34:48 +0000]
alpha: remove dma64_addr_t usage

dma_addr_t is always 64 bit on alpha. So let's use dma_addr_t instead.

Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
Cc: Richard Henderson <rth@twiddle.net>
Cc: Ivan Kokshaysky <ink@jurassic.park.msu.ru>
Cc: Matt Turner <mattst88@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

8 years agomips: remove dma64_addr_t usage
FUJITA Tomonori [Wed, 27 Oct 2010 22:34:47 +0000]
mips: remove dma64_addr_t usage

dma64_addr_t looks pointless (at least there is no point that an
architecture has the own dma64_addr_t typedef).

dma_addr_t is set to 32 or 64 bits appropriately.  You can use u64 at
places where you know that 64 bit address is always necessary.

Let's use u64 instead for mips.

Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
Cc: Ralf Baechle <ralf@linux-mips.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

8 years agosparc: remove dma64_addr_t usage
FUJITA Tomonori [Wed, 27 Oct 2010 22:34:47 +0000]
sparc: remove dma64_addr_t usage

dma64_addr_t looks pointless (at least there is no point that an
architecture has the own dma64_addr_t typedef).

dma_addr_t is set to 32 or 64 bits appropriately. You can use u64
at places where you know that 64 bit address is always necessary.

Let's use u64 instead for sparc32.

Looks like PCI654_REQUIRED_MASK or PCI64_ADR_BASE isn't used. They can
be removed?

Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
Acked-by: David Miller <davem@davemloft.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

8 years agofuse: use release_pages()
Miklos Szeredi [Wed, 27 Oct 2010 22:34:46 +0000]
fuse: use release_pages()

Replace iterated page_cache_release() with release_pages(), which is
faster and shorter.

Needs release_pages() to be exported to modules.

Suggested-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Miklos Szeredi <mszeredi@suse.cz>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

8 years agotaskstats: use real microsecond granularity for CPU times
Michael Holzheu [Wed, 27 Oct 2010 22:34:45 +0000]
taskstats: use real microsecond granularity for CPU times

The taskstats interface uses microsecond granularity for the user and
system time values.  The conversion from cputime to the taskstats values
uses the cputime_to_msecs primitive which effectively limits the
granularity to milliseconds.  Add the cputime_to_usecs primitive for
architectures that have better, more precise CPU time values.  Remove
cputime_to_msecs primitive because there are no more users left.

Signed-off-by: Michael Holzheu <holzheu@linux.vnet.ibm.com>
Acked-by: Balbir Singh <balbir@linux.vnet.ibm.com>
Cc: Luck Tony <tony.luck@intel.com>
Cc: Shailabh Nagar <nagar1234@in.ibm.com>
Cc: Martin Schwidefsky <schwidefsky@de.ibm.com>
Cc: Oleg Nesterov <oleg@redhat.com>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Heiko Carstens <heiko.carstens@de.ibm.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Shailabh Nagar <nagar@us.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

8 years agotaskstats: split fill_pid function
Michael Holzheu [Wed, 27 Oct 2010 22:34:44 +0000]
taskstats: split fill_pid function

Separate the finding of a task_struct by pid or tgid from filling the
taskstats data. This makes the code more readable.

Signed-off-by: Michael Holzheu <holzheu@linux.vnet.ibm.com>
Acked-by: Balbir Singh <balbir@linux.vnet.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

8 years agotaskstats: separate taskstats commands
Michael Holzheu [Wed, 27 Oct 2010 22:34:44 +0000]
taskstats: separate taskstats commands

Move each taskstats command into a single function.  This makes the code
more readable and makes it easier to add new commands.

Signed-off-by: Michael Holzheu <holzheu@linux.vnet.ibm.com>
Acked-by: Balbir Singh <balbir@linux.vnet.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

8 years agodelayacct: align to 8 byte boundary on 64-bit systems
Jeff Mahoney [Wed, 27 Oct 2010 22:34:43 +0000]
delayacct: align to 8 byte boundary on 64-bit systems

prepare_reply() sets up an skb for the response.  The payload contains:

 +--------------------------------+
 | genlmsghdr - 4 bytes           |
 +--------------------------------+
 | NLA header - 4 bytes           | /* Aggregate header */
 +-+------------------------------+
 | | NLA header - 4 bytes         | /* PID header */
 | +------------------------------+
 | | pid/tgid   - 4 bytes         |
 | +------------------------------+
 | | NLA header - 4 bytes         | /* stats header */
 | + -----------------------------+ <- oops. aligned on 4 byte boundary
 | | struct taskstats - 328 bytes |
 +-+------------------------------+

The start of the taskstats struct must be 8 byte aligned on IA64 (and
other systems with 8 byte alignment rules for 64-bit types) or runtime
alignment warnings will be issued.

This patch pads the pid/tgid field out to sizeof(long), which forces the
alignment of taskstats.  The getdelays userspace code is ok with this
since it assumes 32-bit pid/tgid and then honors that header's length
field.

An array is used to avoid exposing kernel memory contents to userspace in
the response.

Signed-off-by: Jeff Mahoney <jeffm@suse.com>
Cc: Balbir Singh <balbir@in.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

8 years agodelay-accounting: reimplement -c for getdelays.c to report information on a target...
Mel Gorman [Wed, 27 Oct 2010 22:34:42 +0000]
delay-accounting: reimplement -c for getdelays.c to report information on a target command

Task delay-accounting was identified as one means of determining how long
a process spends in congestion_wait() without adding new statistics.  For
example, if the workload should not be doing IO, delay-accounting could
reveal how long it was spending in unexpected IO or delays.
Unfortunately, on closer examination it was clear that getdelays does not
act as documented.

Commit a3baf649 ("per-task-delay-accounting: documentation") added
Documentation/accounting/getdelays.c with a -c switch that was documented
to fork/exec a child and report statistics on it but for reasons that are
unclear to me, commit 9e06d3f9 deleted support for this switch but did not
update the documentation.  It might be an oversight or it might be because
the control flow of the program meant that accounting information would be
printed once early in the lifetime of the program making it of limited
use.

This patch reimplements -c for getdelays.c to act as documented.  Unlike
the original version, it waits until the command completes before printing
any information on it.  An example of it being used looks like

$ ./getdelays -d -c find /home/mel -name mel
print delayacct stats ON
/home/mel
/home/mel/.notes-wine/drive_c/windows/profiles/mel
/home/mel/.wine/drive_c/windows/profiles/mel
/home/mel/git-configs/dot.kde/share/apps/konqueror/home/mel
PID 5923

CPU             count     real total  virtual total    delay total
                42779     5051232096     5164722692      564207988
IO              count    delay total
                41727    97804147758
SWAP            count    delay total
                    0              0
RECLAIM         count    delay total
                    0              0

[akpm@linux-foundation.org: coding-style fixes]
Signed-off-by: Mel Gorman <mel@csn.ul.ie>
Acked-by: Balbir Singh <balbir@in.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

8 years agonamespaces Kconfig: move namespace menu location after the cgroup
Daniel Lezcano [Wed, 27 Oct 2010 22:34:41 +0000]
namespaces Kconfig: move namespace menu location after the cgroup

We have the namespaces as a menuconfig like the cgroup.  The cgroup and
the namespace are two base bricks for the containers.

It is more logical to put the namespace menu right after the cgroup menu.

Signed-off-by: Daniel Lezcano <daniel.lezcano@free.fr>
Cc: Li Zefan <lizf@cn.fujitsu.com>
Cc: "Serge E. Hallyn" <serue@us.ibm.com>
Cc: "Eric W. Biederman" <ebiederm@xmission.com>
Cc: Paul Menage <menage@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

8 years agonamespaces Kconfig: remove the cgroup device whitelist experimental tag
Daniel Lezcano [Wed, 27 Oct 2010 22:34:40 +0000]
namespaces Kconfig: remove the cgroup device whitelist experimental tag

This subsystem is merged since a long time now, I think we can consider it
mature enough.

Signed-off-by: Daniel Lezcano <daniel.lezcano@free.fr>
Cc: Li Zefan <lizf@cn.fujitsu.com>
Cc: "Serge E. Hallyn" <serue@us.ibm.com>
Cc: "Eric W. Biederman" <ebiederm@xmission.com>
Cc: Paul Menage <menage@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

8 years agonamespaces Kconfig: remove pointless cgroup dependency
Daniel Lezcano [Wed, 27 Oct 2010 22:34:39 +0000]
namespaces Kconfig: remove pointless cgroup dependency

The different cgroup subsystems are under the cgroup submenu.  The
dependency between the cgroups and the menu subsystems is pointless.

Signed-off-by: Daniel Lezcano <daniel.lezcano@free.fr>
Acked-by: Li Zefan <lizf@cn.fujitsu.com>
Cc: "Serge E. Hallyn" <serue@us.ibm.com>
Cc: "Eric W. Biederman" <ebiederm@xmission.com>
Cc: Paul Menage <menage@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

8 years agonamespaces Kconfig: make namespace a submenu
Daniel Lezcano [Wed, 27 Oct 2010 22:34:38 +0000]
namespaces Kconfig: make namespace a submenu

Make the namespaces config option a submenu.

Signed-off-by: Daniel Lezcano <daniel.lezcano@free.fr>
Cc: Li Zefan <lizf@cn.fujitsu.com>
Cc: "Serge E. Hallyn" <serue@us.ibm.com>
Cc: "Eric W. Biederman" <ebiederm@xmission.com>
Cc: Paul Menage <menage@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

8 years agonamespaces: default all the namespaces to 'yes' when CONFIG_NAMESPACES is selected
Daniel Lezcano [Wed, 27 Oct 2010 22:34:37 +0000]
namespaces: default all the namespaces to 'yes' when CONFIG_NAMESPACES is selected

As the different namespaces depend on 'CONFIG_NAMESPACES', it is logical
to enable all the namespaces when we enable NAMESPACES.

Signed-off-by: Daniel Lezcano <daniel.lezcano@free.fr>
Cc: "Eric W. Biederman" <ebiederm@xmission.com>
Cc: David Miller <davem@davemloft.net>
Acked-By: Matt Helsley <matthltc@us.ibm.com>
Acked-by: Serge Hallyn <serge.hallyn@canonical.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

8 years agonamespaces: remove pid_ns and net_ns experimental status
Daniel Lezcano [Wed, 27 Oct 2010 22:34:37 +0000]
namespaces: remove pid_ns and net_ns experimental status

The pid namespace is in the kernel since 2.6.27 and the net_ns since
2.6.29.  They are enabled in the distro by default and used by userspace
component.  They are mature enough to remove the 'experimental' label.

Signed-off-by: Daniel Lezcano <daniel.lezcano@free.fr>
Cc: "Eric W. Biederman" <ebiederm@xmission.com>
Cc: David Miller <davem@davemloft.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

8 years agorapidio: fix destructive port EM initialization for Tsi568
Alexandre Bounine [Wed, 27 Oct 2010 22:34:36 +0000]
rapidio: fix destructive port EM initialization for Tsi568

Replace possibly damaging broadcast writes into the per-port SP_MODE
registers with individual writes for each port.  This will preserve
individual port configurations in case if ports are configured
differently.

Signed-off-by: Alexandre Bounine <alexandre.bounine@idt.com>
Cc: Kumar Gala <galak@kernel.crashing.org>
Cc: Matt Porter <mporter@kernel.crashing.org>
Cc: Li Yang <leoli@freescale.com>
Cc: Thomas Moll <thomas.moll@sysgo.com>
Cc: Micha Nelissen <micha@neli.hopto.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

8 years agorapidio: fix maximum port number in tsi57x EM initialization
Alexandre Bounine [Wed, 27 Oct 2010 22:34:35 +0000]
rapidio: fix maximum port number in tsi57x EM initialization

Replace hardcoded maximum port number with actual reported number of
switch ports.

Signed-off-by: Alexandre Bounine <alexandre.bounine@idt.com>
Cc: Kumar Gala <galak@kernel.crashing.org>
Cc: Matt Porter <mporter@kernel.crashing.org>
Cc: Li Yang <leoli@freescale.com>
Cc: Thomas Moll <thomas.moll@sysgo.com>
Cc: Micha Nelissen <micha@neli.hopto.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

8 years agorapidio: fix IDLE2 bits corruption
Alexandre Bounine [Wed, 27 Oct 2010 22:34:35 +0000]
rapidio: fix IDLE2 bits corruption

RapidIO spec v.2.1 adds Idle Sequence 2 into LP-Serial Physical Layer.
The fix ensures that corresponding bits are not corrupted during error
handling.

Signed-off-by: Alexandre Bounine <alexandre.bounine@idt.com>
Cc: Thomas Moll <thomas.moll@sysgo.com>
Cc: Matt Porter <mporter@kernel.crashing.org>
Cc: Li Yang <leoli@freescale.com>
Cc: Kumar Gala <galak@kernel.crashing.org>
Cc: Micha Nelissen <micha@neli.hopto.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

8 years agorapidio: add handling of redundant routes
Alexandre Bounine [Wed, 27 Oct 2010 22:34:34 +0000]
rapidio: add handling of redundant routes

Detects RIO link to the already enumerated device and properly sets links
between device objects.  Changes to the enumeration/discovery logic:

1. Use Master Enable bit to signal end of the enumeration - agents may
   start their discovery process as soon as they see this bit set
   (Component Tag register was used before for this purpose).

2. Enumerator sets Component Tag (!= 0) immediately during device
   setup.  This allows to identify the device if the redundant route
   exists in a RIO system.

Signed-off-by: Alexandre Bounine <alexandre.bounine@idt.com>
Cc: Thomas Moll <thomas.moll@sysgo.com>
Cc: Matt Porter <mporter@kernel.crashing.org>
Cc: Li Yang <leoli@freescale.com>
Cc: Kumar Gala <galak@kernel.crashing.org>
Cc: Micha Nelissen <micha@neli.hopto.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

8 years agorapidio: add support for IDT CPS Gen2 switches
Alexandre Bounine [Wed, 27 Oct 2010 22:34:33 +0000]
rapidio: add support for IDT CPS Gen2 switches

Add the RIO switch driver and definitions for IDT CPS-1848 and CPS-1616
Gen2 devices.

Signed-off-by: Alexandre Bounine <alexandre.bounine@idt.com>
Cc: Thomas Moll <thomas.moll@sysgo.com>
Cc: Matt Porter <mporter@kernel.crashing.org>
Cc: Li Yang <leoli@freescale.com>
Cc: Kumar Gala <galak@kernel.crashing.org>
Cc: Micha Nelissen <micha@neli.hopto.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>