]> pilppa.com Git - linux-2.6-omap-h63xx.git/log
linux-2.6-omap-h63xx.git
16 years ago[XFS] Sanitise xfs_log_force error checking.
David Chinner [Thu, 10 Apr 2008 02:24:30 +0000 (12:24 +1000)]
[XFS] Sanitise xfs_log_force error checking.

xfs_log_force() is declared to return an error, but we almost never check
it. We don't need to check it in most cases; if there's a log I/O error
then we'll be shutting down the filesystem anyway and that means we'll
catch the error somewhere else.

However, on certain calls we should be returning an error - sync
transactions, fsync, sync writes, etc. so this isn't a pure black and
white distinction. Hence make xfs_log_force() a void function that issues
a warning to the syslog on error, and call _xfs_log_force() in all the
places where we actually care about the error status returned.

SGI-PV: 980084
SGI-Modid: xfs-linux-melb:xfs-kern:30832a

Signed-off-by: David Chinner <dgc@sgi.com>
Signed-off-by: Niv Sardi <xaiki@sgi.com>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
16 years ago[XFS] Check for errors when changing buffer pointers.
David Chinner [Thu, 10 Apr 2008 02:24:24 +0000 (12:24 +1000)]
[XFS] Check for errors when changing buffer pointers.

xfs_buf_associate_memory() can fail, but the return is never checked.
Propagate the error through XFS_BUF_SET_PTR() so that failures are
detected.

SGI-PV: 980084
SGI-Modid: xfs-linux-melb:xfs-kern:30831a

Signed-off-by: David Chinner <dgc@sgi.com>
Signed-off-by: Niv Sardi <xaiki@sgi.com>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
16 years ago[XFS] Don't allow silent errors in xfs_inactive().
David Chinner [Thu, 10 Apr 2008 02:24:17 +0000 (12:24 +1000)]
[XFS] Don't allow silent errors in xfs_inactive().

xfs_inactive() fails to report errors when committing the inactive
transaction. Hence we can get silent failures either finishing off the
truncation or committing the transaction. Even if we get errors, we need
to continue, so simply warn loudly to the system if we get errors here.

SGI-PV: 980084
SGI-Modid: xfs-linux-melb:xfs-kern:30830a

Signed-off-by: David Chinner <dgc@sgi.com>
Signed-off-by: Niv Sardi <xaiki@sgi.com>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
16 years ago[XFS] Catch errors from xfs_imap().
David Chinner [Thu, 10 Apr 2008 02:24:10 +0000 (12:24 +1000)]
[XFS] Catch errors from xfs_imap().

Catch errors from xfs_imap() in log recovery when we might be trying to
map an invalid inode number due to a corrupted log.

SGI-PV: 980084
SGI-Modid: xfs-linux-melb:xfs-kern:30829a

Signed-off-by: David Chinner <dgc@sgi.com>
Signed-off-by: Niv Sardi <xaiki@sgi.com>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
16 years ago[XFS] xfs_bulkstat_one_dinode() never returns an error.
David Chinner [Thu, 10 Apr 2008 02:24:04 +0000 (12:24 +1000)]
[XFS] xfs_bulkstat_one_dinode() never returns an error.

Mark it void.

SGI-PV: 980084
SGI-Modid: xfs-linux-melb:xfs-kern:30828a

Signed-off-by: David Chinner <dgc@sgi.com>
Signed-off-by: Niv Sardi <xaiki@sgi.com>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
16 years ago[XFS] xfs_iflush_fork() never returns an error.
David Chinner [Thu, 10 Apr 2008 02:23:58 +0000 (12:23 +1000)]
[XFS] xfs_iflush_fork() never returns an error.

xfs_iflush_fork() never returns an error. Mark it void and clean up the
code calling it that checks for errors.

SGI-PV: 980084
SGI-Modid: xfs-linux-melb:xfs-kern:30827a

Signed-off-by: David Chinner <dgc@sgi.com>
Signed-off-by: Niv Sardi <xaiki@sgi.com>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
16 years ago[XFS] Catch unwritten extent conversion errors.
David Chinner [Thu, 10 Apr 2008 02:23:52 +0000 (12:23 +1000)]
[XFS] Catch unwritten extent conversion errors.

On unwritten I/O completion, we fail to propagate an error when converting
the extent to a written extent. This means that the I/O silently fails.
propagate the error onto the ioend so that the inode is marked with an
error appropriately.

SGI-PV: 980084
SGI-Modid: xfs-linux-melb:xfs-kern:30826a

Signed-off-by: David Chinner <dgc@sgi.com>
Signed-off-by: Niv Sardi <xaiki@sgi.com>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
16 years ago[XFS] xfs_bdwrite() does not return errors.
David Chinner [Thu, 10 Apr 2008 02:23:46 +0000 (12:23 +1000)]
[XFS] xfs_bdwrite() does not return errors.

xfs_bdwrite() cannot return an error; it only queues buffers to the
delayed write list and as such never encounters anything that can fail.
Mark it void.

SGI-PV: 980084
SGI-Modid: xfs-linux-melb:xfs-kern:30825a

Signed-off-by: David Chinner <dgc@sgi.com>
Signed-off-by: Niv Sardi <xaiki@sgi.com>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
16 years ago[XFS] Ensure xfs_bawrite() errors are checked.
David Chinner [Thu, 10 Apr 2008 02:22:24 +0000 (12:22 +1000)]
[XFS] Ensure xfs_bawrite() errors are checked.

xfs_bawrite() can return immediate error status on async writes. Unlike
xfsbdstrat() we don't ever check the error on the buffer after the call,
so we currently do not catch errors at all here. Ensure we catch and
propagate or warn to the syslog about up-front async write errors.

SGI-PV: 980084
SGI-Modid: xfs-linux-melb:xfs-kern:30824a

Signed-off-by: David Chinner <dgc@sgi.com>
Signed-off-by: Niv Sardi <xaiki@sgi.com>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
16 years ago[XFS] Ensure errors from xfs_bdstrat() are correctly checked.
David Chinner [Thu, 10 Apr 2008 02:22:17 +0000 (12:22 +1000)]
[XFS] Ensure errors from xfs_bdstrat() are correctly checked.

xfsbdstrat() is declared to return an error. That is never checked because
the error is propagated by the xfs_buf_t that is passed through the
function.

Mark xfsbdstrat() as returning void and comment the prototype on the
methods needed for error checking.

SGI-PV: 980084
SGI-Modid: xfs-linux-melb:xfs-kern:30823a

Signed-off-by: David Chinner <dgc@sgi.com>
Signed-off-by: Niv Sardi <xaiki@sgi.com>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
16 years ago[XFS] remove bhv_vname_t and xfs_rename code
Barry Naujok [Thu, 10 Apr 2008 02:22:07 +0000 (12:22 +1000)]
[XFS] remove bhv_vname_t and xfs_rename code

SGI-PV: 976035
SGI-Modid: xfs-linux-melb:xfs-kern:30804a

Signed-off-by: Barry Naujok <bnaujok@sgi.com>
Signed-off-by: Christoph Hellwig <hch@infradead.org>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
16 years ago[XFS] Catch errors returned from xfs_bmap_last_offset().
David Chinner [Thu, 10 Apr 2008 02:21:59 +0000 (12:21 +1000)]
[XFS] Catch errors returned from xfs_bmap_last_offset().

xfs_bmap_last_offset() can fail and return an error.
xfs_iomap_write_allocate() fails to detect and propagate the error.

SGI-PV: 980084
SGI-Modid: xfs-linux-melb:xfs-kern:30802a

Signed-off-by: David Chinner <dgc@sgi.com>
Signed-off-by: Niv Sardi <xaiki@sgi.com>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
16 years ago[XFS] Check for xfs_free_extent() failing.
David Chinner [Thu, 10 Apr 2008 02:21:53 +0000 (12:21 +1000)]
[XFS] Check for xfs_free_extent() failing.

xfs_free_extent() can fail, but log recovery never bothers to check if it
successfully free the extent it was supposed to. This could lead to silent
corruption during log recovery. Abort log recovery if we fail to free an
extent.

SGI-PV: 980084
SGI-Modid: xfs-linux-melb:xfs-kern:30801a

Signed-off-by: David Chinner <dgc@sgi.com>
Signed-off-by: Niv Sardi <xaiki@sgi.com>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
16 years ago[XFS] Warn if errors come from block_truncate_page().
David Chinner [Thu, 10 Apr 2008 02:21:46 +0000 (12:21 +1000)]
[XFS] Warn if errors come from block_truncate_page().

block_truncate_page() can return errors that we currently ignore and
silently discard. We should not ever get errors reported here - an error
indicates a bug somewhere else. Hence catch the error and issue a stack
dump to the syslog because we cannot propagate the error any further up
the call chain.

SGI-PV: 980084
SGI-Modid: xfs-linux-melb:xfs-kern:30800a

Signed-off-by: David Chinner <dgc@sgi.com>
Signed-off-by: Niv Sardi <xaiki@sgi.com>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
16 years ago[XFS] xfs_bmap_adjacent() never returns an error.
David Chinner [Thu, 10 Apr 2008 02:21:40 +0000 (12:21 +1000)]
[XFS] xfs_bmap_adjacent() never returns an error.

Mark it void.

SGI-PV: 980084
SGI-Modid: xfs-linux-melb:xfs-kern:30798a

Signed-off-by: David Chinner <dgc@sgi.com>
Signed-off-by: Niv Sardi <xaiki@sgi.com>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
16 years ago[XFS] Make xfs_alloc_compute_aligned() void.
David Chinner [Thu, 10 Apr 2008 02:21:32 +0000 (12:21 +1000)]
[XFS] Make xfs_alloc_compute_aligned() void.

xfs_alloc_compute_aligned() returns a value based on a comparison of the
computed extent length and the minimum length allowed. This is only used
by some callers - the other four return parameters are used more often.
Hence move the comparison to the code that actually needs to do it and
make xfs_alloc_compute_aligned() a void function.

SGI-PV: 980084
SGI-Modid: xfs-linux-melb:xfs-kern:30797a

Signed-off-by: David Chinner <dgc@sgi.com>
Signed-off-by: Niv Sardi <xaiki@sgi.com>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
16 years ago[XFS] Clean up xfs_alloc_search_busy() return values.
David Chinner [Thu, 10 Apr 2008 02:21:25 +0000 (12:21 +1000)]
[XFS] Clean up xfs_alloc_search_busy() return values.

xfs_alloc_search_busy() returns an index into the busy array if the extent
was found in the array. This is never checked, and the
xfs_alloc_search_busy() does a log force to prevent reuse of the extent
before the free transaction hits the disk. Hence the return value is
useless. Declare the function void and remove the slot number from the
tracing as well.

SGI-PV: 980084
SGI-Modid: xfs-linux-melb:xfs-kern:30796a

Signed-off-by: David Chinner <dgc@sgi.com>
Signed-off-by: Niv Sardi <xaiki@sgi.com>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
16 years ago[XFS] Propagate errors from xfs_trans_commit().
David Chinner [Thu, 10 Apr 2008 02:21:18 +0000 (12:21 +1000)]
[XFS] Propagate errors from xfs_trans_commit().

xfs_trans_commit() can return errors when there are problems in the
transaction subsystem. They are indicative that the entire transaction may
be incomplete, and hence the error should be propagated as there is a good
possibility that there is something fatally wrong in the filesystem. Catch
and propagate or warn about commit errors in the places where they are
currently ignored.

SGI-PV: 980084
SGI-Modid: xfs-linux-melb:xfs-kern:30795a

Signed-off-by: David Chinner <dgc@sgi.com>
Signed-off-by: Niv Sardi <xaiki@sgi.com>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
16 years ago[XFS] Propagate xfs_trans_reserve() errors.
David Chinner [Thu, 10 Apr 2008 02:21:11 +0000 (12:21 +1000)]
[XFS] Propagate xfs_trans_reserve() errors.

xfs_trans_reserve() reports errors that should not be ignored. For
example, a shutdown filesystem will report errors through
xfs_trans_reserve() to prevent further changes from being attempted on a
damaged filesystem. Catch and propagate all error conditions from
xfs_trans_reserve().

SGI-PV: 980084
SGI-Modid: xfs-linux-melb:xfs-kern:30794a

Signed-off-by: David Chinner <dgc@sgi.com>
Signed-off-by: Niv Sardi <xaiki@sgi.com>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
16 years ago[XFS] Catch errors from xfs_acl_vremove().
David Chinner [Thu, 10 Apr 2008 02:21:04 +0000 (12:21 +1000)]
[XFS] Catch errors from xfs_acl_vremove().

Removing an ACL can return an error. Propagate it.

SGI-PV: 980084
SGI-Modid: xfs-linux-melb:xfs-kern:30793a

Signed-off-by: David Chinner <dgc@sgi.com>
Signed-off-by: Niv Sardi <xaiki@sgi.com>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
16 years ago[XFS] Catch errors from xfs_acl_setmode().
David Chinner [Thu, 10 Apr 2008 02:20:58 +0000 (12:20 +1000)]
[XFS] Catch errors from xfs_acl_setmode().

Propagate the error status from xfs_acl_setmode() so that callers know if
the ACl was set correctly or not.

SGI-PV: 980084
SGI-Modid: xfs-linux-melb:xfs-kern:30792a

Signed-off-by: David Chinner <dgc@sgi.com>
Signed-off-by: Niv Sardi <xaiki@sgi.com>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
16 years ago[XFS] Propagate quota file truncation errors.
David Chinner [Thu, 10 Apr 2008 02:20:51 +0000 (12:20 +1000)]
[XFS] Propagate quota file truncation errors.

Truncating the quota files can silently fail. Ensure that truncation
errors are propagated to the callers.

SGI-PV: 980084
SGI-Modid: xfs-linux-melb:xfs-kern:30791a

Signed-off-by: David Chinner <dgc@sgi.com>
Signed-off-by: Niv Sardi <xaiki@sgi.com>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
16 years ago[XFS] Catch errors when turning off quotas.
David Chinner [Thu, 10 Apr 2008 02:20:45 +0000 (12:20 +1000)]
[XFS] Catch errors when turning off quotas.

When turning off quota, we need to write various transactions to the log
to ensure that they are cleanly removed in the case of a crash. We need to
check that the transactions hit the disk correctly. If we fail to write
the final quota off transaction, we are corrupt in memory and so the only
option is to shut the filesystem down at this point.

SGI-PV: 980084
SGI-Modid: xfs-linux-melb:xfs-kern:30790a

Signed-off-by: David Chinner <dgc@sgi.com>
Signed-off-by: Niv Sardi <xaiki@sgi.com>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
16 years ago[XFS] Catch errors resetting quota flags.
David Chinner [Thu, 10 Apr 2008 02:20:38 +0000 (12:20 +1000)]
[XFS] Catch errors resetting quota flags.

Warn to the syslog if we fail to reset the quota flags in the superblock
when a quota check fails.

SGI-PV: 980084
SGI-Modid: xfs-linux-melb:xfs-kern:30789a

Signed-off-by: David Chinner <dgc@sgi.com>
Signed-off-by: Niv Sardi <xaiki@sgi.com>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
16 years ago[XFS] Clean up quotamount error handling.
David Chinner [Thu, 10 Apr 2008 02:20:31 +0000 (12:20 +1000)]
[XFS] Clean up quotamount error handling.

xfs_qm_mount_quotas() returns an error status that is ignored. If we fail
to mount quotas, we continue with quota's turned off, which is all handled
inside xfs_qm_mount_quotas(). Mark it as void to indicate that errors need
not be returned to the callers.

SGI-PV: 980084
SGI-Modid: xfs-linux-melb:xfs-kern:30788a

Signed-off-by: David Chinner <dgc@sgi.com>
Signed-off-by: Niv Sardi <xaiki@sgi.com>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
16 years ago[XFS] Check for dquot flush errors
David Chinner [Thu, 10 Apr 2008 02:20:24 +0000 (12:20 +1000)]
[XFS] Check for dquot flush errors

xfs_qm_dqflush() can fail, but the return is not checked anywhere. Hence
we never know if we've failed to flush a dquot to disk. Propagate the
error and warn to the syslog if a flush ever fails.

SGI-PV: 980084
SGI-Modid: xfs-linux-melb:xfs-kern:30787a

Signed-off-by: David Chinner <dgc@sgi.com>
Signed-off-by: Niv Sardi <xaiki@sgi.com>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
16 years ago[XFS] Propagate xfs_qm_dqflush_all() errors.
David Chinner [Thu, 10 Apr 2008 02:20:17 +0000 (12:20 +1000)]
[XFS] Propagate xfs_qm_dqflush_all() errors.

xfs_qm_dqflush_all() can return flush errors. Ensure they are propagated
into the quotacheck code to determine if the quotacheck succeeded or not.

SGI-PV: 980084
SGI-Modid: xfs-linux-melb:xfs-kern:30786a

Signed-off-by: David Chinner <dgc@sgi.com>
Signed-off-by: Niv Sardi <xaiki@sgi.com>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
16 years ago[XFS] xfs_qm_reset_dqcounts() does not return errors.
David Chinner [Thu, 10 Apr 2008 02:20:10 +0000 (12:20 +1000)]
[XFS] xfs_qm_reset_dqcounts() does not return errors.

Declare it void.

SGI-PV: 980084
SGI-Modid: xfs-linux-melb:xfs-kern:30785a

Signed-off-by: David Chinner <dgc@sgi.com>
Signed-off-by: Niv Sardi <xaiki@sgi.com>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
16 years ago[XFS] Report errors from xfs_reserve_blocks().
David Chinner [Thu, 10 Apr 2008 02:20:03 +0000 (12:20 +1000)]
[XFS] Report errors from xfs_reserve_blocks().

xfs_reserve_blocks() can fail in interesting ways. In neither case is it a
fatal error, but the result can lead to sub-optimal behaviour. Warn to the
syslog if the call fails but otherwise continue.

SGI-PV: 980084
SGI-Modid: xfs-linux-melb:xfs-kern:30784a

Signed-off-by: David Chinner <dgc@sgi.com>
Signed-off-by: Niv Sardi <xaiki@sgi.com>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
16 years ago[XFS] xfs_icsb_counter_disabled() never returns an error.
David Chinner [Thu, 10 Apr 2008 02:19:56 +0000 (12:19 +1000)]
[XFS] xfs_icsb_counter_disabled() never returns an error.

Mark it void.

SGI-PV: 980084
SGI-Modid: xfs-linux-melb:xfs-kern:30782a

Signed-off-by: David Chinner <dgc@sgi.com>
Signed-off-by: Niv Sardi <xaiki@sgi.com>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
16 years ago[XFS] Remove useless whitespace in function prototypes
David Chinner [Thu, 10 Apr 2008 02:19:47 +0000 (12:19 +1000)]
[XFS] Remove useless whitespace in function prototypes

Makes it simpler to annotate function prototypes with __must_check via sed
scripts.

SGI-PV: 980084
SGI-Modid: xfs-linux-melb:xfs-kern:30781a

Signed-off-by: David Chinner <dgc@sgi.com>
Signed-off-by: Niv Sardi <xaiki@sgi.com>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
16 years ago[XFS] xfs_quiesce_fs() never returns an error. Mark it void.
David Chinner [Thu, 10 Apr 2008 02:19:40 +0000 (12:19 +1000)]
[XFS] xfs_quiesce_fs() never returns an error. Mark it void.

SGI-PV: 980084
SGI-Modid: xfs-linux-melb:xfs-kern:30780a

Signed-off-by: David Chinner <dgc@sgi.com>
Signed-off-by: Niv Sardi <xaiki@sgi.com>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
16 years ago[XFS] Don't validate symlink target component length
Christoph Hellwig [Thu, 10 Apr 2008 02:19:27 +0000 (12:19 +1000)]
[XFS] Don't validate symlink target component length

This target component validation is not POSIX conformant and it is not
done by any other Linux filesystem so remove it from XFS.

SGI-PV: 980080
SGI-Modid: xfs-linux-melb:xfs-kern:30776a

Signed-off-by: Christoph Hellwig <hch@infradead.org>
Signed-off-by: David Chinner <dgc@sgi.com>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
16 years ago[XFS] replace remaining __FUNCTION__ occurrences
Harvey Harrison [Thu, 10 Apr 2008 02:19:21 +0000 (12:19 +1000)]
[XFS] replace remaining __FUNCTION__ occurrences

__FUNCTION__ is gcc-specific, use __func__

SGI-PV: 976035
SGI-Modid: xfs-linux-melb:xfs-kern:30775a

Signed-off-by: Harvey Harrison <harvey.harrison@gmail.com>
Signed-off-by: David Chinner <dgc@sgi.com>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
16 years ago[XFS] Replace __inline with inline
Harvey Harrison [Thu, 10 Apr 2008 02:19:10 +0000 (12:19 +1000)]
[XFS] Replace __inline with inline

Remove the remaining uses of __inline in the XFS code base.

SGI-PV: 976035
SGI-Modid: xfs-linux-melb:xfs-kern:30774a

Signed-off-by: Harvey Harrison <harvey.harrison@gmail.com>
Signed-off-by: David Chinner <dgc@sgi.com>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
16 years ago[XFS] Fix lock inversion in forced shutdown.
David Chinner [Thu, 10 Apr 2008 02:19:02 +0000 (12:19 +1000)]
[XFS] Fix lock inversion in forced shutdown.

Recent changes to xlog_state_release_iclog() placed the grant_lock inside
the icloglock. forced unmount of the log does this the opposite way
around, but does not depend on the order for correct working. Fix the
inversion by changing the order locks are gained in
xfs_log_force_umount().

SGI-PV: 979661
SGI-Modid: xfs-linux-melb:xfs-kern:30773a

Signed-off-by: David Chinner <dgc@sgi.com>
Signed-off-by: Christoph Hellwig <hch@infradead.org>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
16 years ago[XFS] Reorganise xlog_t for better cacheline isolation of contention
David Chinner [Thu, 10 Apr 2008 02:18:54 +0000 (12:18 +1000)]
[XFS] Reorganise xlog_t for better cacheline isolation of contention

To reduce contention on the log in large CPU count, separate out different
parts of the xlog_t structure onto different cachelines. Move each lock
onto a different cacheline along with all the members that are
accessed/modified while that lock is held.

Also, move the debugging code into debug code.

SGI-PV: 978729
SGI-Modid: xfs-linux-melb:xfs-kern:30772a

Signed-off-by: David Chinner <dgc@sgi.com>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
16 years ago[XFS] Remove the xlog_ticket allocator
David Chinner [Thu, 10 Apr 2008 02:18:46 +0000 (12:18 +1000)]
[XFS] Remove the xlog_ticket allocator

The ticket allocator is just a simple slab implementation internal to the
log. It requires the icloglock to be held when manipulating it and this
contributes to contention on that lock.

Just kill the entire allocator and use a memory zone instead. While there,
allow us to gracefully fail allocation with ENOMEM.

SGI-PV: 978729
SGI-Modid: xfs-linux-melb:xfs-kern:30771a

Signed-off-by: David Chinner <dgc@sgi.com>
Signed-off-by: Christoph Hellwig <hch@infradead.org>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
16 years ago[XFS] Per iclog callback chain lock
David Chinner [Thu, 10 Apr 2008 02:18:39 +0000 (12:18 +1000)]
[XFS] Per iclog callback chain lock

Rather than use the icloglock for protecting the iclog completion callback
chain, use a new per-iclog lock so that walking the callback chain doesn't
require holding a global lock.

This reduces contention on the icloglock during transaction commit and log
I/O completion by reducing the number of times we need to hold the global
icloglock during these operations.

SGI-PV: 978729
SGI-Modid: xfs-linux-melb:xfs-kern:30770a

Signed-off-by: David Chinner <dgc@sgi.com>
Signed-off-by: Christoph Hellwig <hch@infradead.org>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
16 years ago[XFS] Prevent xfs_bmap_check_leaf_extents() referencing unmapped memory.
Lachlan McIlroy [Thu, 27 Mar 2008 07:01:14 +0000 (18:01 +1100)]
[XFS] Prevent xfs_bmap_check_leaf_extents() referencing unmapped memory.

While investigating the extent corruption bug I ran into this bug in debug
only code. xfs_bmap_check_leaf_extents() loops through the leaf blocks of
the extent btree checking that every extent is entirely before the next
extent. It also compares the last extent in the previous block to the
first extent in the current block when the previous block has been
released and potentially unmapped. So take a copy of the last extent
instead of a pointer. Also move the last extent check out of the loop
because we only need to do it once.

SGI-PV: 976035
SGI-Modid: xfs-linux-melb:xfs-kern:30718a

Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
Signed-off-by: Christoph Hellwig <hch@infradead.org>
16 years ago[XFS] remove most calls to VN_RELE
Christoph Hellwig [Thu, 27 Mar 2008 07:01:08 +0000 (18:01 +1100)]
[XFS] remove most calls to VN_RELE

Most VN_RELE calls either directly contain a XFS_ITOV or have the
corresponding xfs_inode already in scope. Use the IRELE helper instead of
VN_RELE to clarify the code. With a little more work we can kill VN_RELE
altogether and define IRELE in terms of iput directly.

SGI-PV: 976035
SGI-Modid: xfs-linux-melb:xfs-kern:30710a

Signed-off-by: Christoph Hellwig <hch@infradead.org>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
16 years ago[XFS] split xfs_ioc_xattr
Lachlan McIlroy [Fri, 18 Apr 2008 01:44:03 +0000 (11:44 +1000)]
[XFS] split xfs_ioc_xattr

The three subcases of xfs_ioc_xattr don't share any semantics and almost
no code, so split it into three separate helpers.

SGI-PV: 976035
SGI-Modid: xfs-linux-melb:xfs-kern:30709a

Signed-off-by: Christoph Hellwig <hch@infradead.org>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
16 years ago[XFS] cleanup root inode handling in xfs_fs_fill_super
Christoph Hellwig [Thu, 27 Mar 2008 07:00:54 +0000 (18:00 +1100)]
[XFS] cleanup root inode handling in xfs_fs_fill_super

- rename rootvp to root for clarify
- remove useless vn_to_inode call
- check is_bad_inode before calling d_alloc_root
- use iput instead of VN_RELE in the error case

SGI-PV: 976035
SGI-Modid: xfs-linux-melb:xfs-kern:30708a

Signed-off-by: Christoph Hellwig <hch@infradead.org>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
16 years ago[XFS] Ensure a btree insert returns a valid cursor.
David Chinner [Thu, 27 Mar 2008 07:00:45 +0000 (18:00 +1100)]
[XFS] Ensure a btree insert returns a valid cursor.

When writing into preallocated regions there is a case where XFS can oops
or hang doing the unwritten extent conversion on I/O completion. It turns
out that the problem is related to the btree cursor being invalid.

When we do an insert into the tree, we may need to split blocks in the
tree. When we only split at the leaf level (i.e. level 0), everything
works just fine. However, if we have a multi-level split in the btreee,
the cursor passed to the insert function is no longer valid once the
insert is complete.

The leaf level split is handled correctly because all the operations at
level 0 are done using the original cursor, hence it is updated correctly.
However, when we need to update the next level up the tree, we don't use
that cursor - we use a cloned cursor that points to the index in the next
level up where we need to do the insert.

Hence if we need to split a second level, the changes to the tree are
reflected in the cloned cursor and not the original cursor. This
clone-and-move-up-a-level-on-split behaviour recurses all the way to the
top of the tree.

The complexity here is that these cloned cursors do not point to the
original index that was inserted - they point to the newly allocated block
(the right block) and the original cursor pointer to that level may still
point to the left block. Hence, without deep examination of the cloned
cursor and buffers, we cannot update the original cursor with the new path
from the cloned cursor.

In these cases the original cursor could be pointing to the wrong block(s)
and hence a subsequent modification to the tree using that cursor will
lead to corruption of the tree.

The crash case occurs when the tree changes height - we insert a new level
in the tree, and the cursor does not have a buffer in it's path for that
level. Hence any attempt to walk back up the cursor to the root block will
result in a null pointer dereference.

To make matters even more complex, the BMAP BT is rooted in an inode, so
we can have a change of height in the btree *without a root split*. That
is, if the root block in the inode is full when we split a leaf node, we
cannot fit the pointer to the new block in the root, so we allocate a new
block, migrate all the ptrs out of the inode into the new block and point
the inode root block at the newly allocated block. This changes the height
of the tree without a root split having occurred and hence invalidates the
path in the original cursor.

The patch below prevents xfs_bmbt_insert() from returning with an invalid
cursor by detecting the cases that invalidate the original cursor and
refresh it by do a lookup into the btree for the original index we were
inserting at.

Note that the INOBT, AGFBNO and AGFCNT btree implementations also have
this bug, but the cursor is currently always destroyed or revalidated
after an insert for those trees. Hence this patch only address the problem
in the BMBT code.

SGI-PV: 979339
SGI-Modid: xfs-linux-melb:xfs-kern:30701a

Signed-off-by: David Chinner <dgc@sgi.com>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
16 years ago[XFS] Account for inode cluster alignment in all allocations
David Chinner [Thu, 27 Mar 2008 07:00:38 +0000 (18:00 +1100)]
[XFS] Account for inode cluster alignment in all allocations

At ENOSPC, we can get a filesystem shutdown due to a cancelling a dirty
transaction in xfs_mkdir or xfs_create. This is due to the initial
allocation attempt not taking into account inode alignment and hence we
can prepare the AGF freelist for allocation when it's not actually
possible to do an allocation. This results in inode allocation returning
ENOSPC with a dirty transaction, and hence we shut down the filesystem.

Because the first allocation is an exact allocation attempt, we must tell
the allocator that the alignment does not affect the allocation attempt.
i.e. we will accept any extent alignment as long as the extent starts at
the block we want. Unfortunately, this means that if the longest free
extent is less than the length + alignment necessary for fallback
allocation attempts but is long enough to attempt a non-aligned
allocation, we will modify the free list.

If we then have the exact allocation fail, all other allocation attempts
will also fail due to the alignment constraint being taken into account.
Hence the initial attempt needs to set the "alignment slop" field so that
alignment, while not required, must be taken into account when determining
if there is enough space left in the AG to do the allocation.

That means if the exact allocation fails, we will not dirty the freelist
if there is not enough space available fo a subsequent allocation to
succeed. Hence we get an ENOSPC error back to userspace without shutting
down the filesystem.

SGI-PV: 978886
SGI-Modid: xfs-linux-melb:xfs-kern:30699a

Signed-off-by: David Chinner <dgc@sgi.com>
Signed-off-by: Christoph Hellwig <hch@infradead.org>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
16 years ago[XFS] Replace custom AIL linked-list code with struct list_head
Josef 'Jeff' Sipek [Thu, 27 Mar 2008 06:58:27 +0000 (17:58 +1100)]
[XFS] Replace custom AIL linked-list code with struct list_head

Replace the xfs_ail_entry_t with a struct list_head and clean the
surrounding code up. Also fixes a livelock in xfs_trans_first_push_ail()
by terminating the loop at the head of the list correctly.

SGI-PV: 978682
SGI-Modid: xfs-linux-melb:xfs-kern:30636a

Signed-off-by: Josef 'Jeff' Sipek <jeffpc@josefsipek.net>
Signed-off-by: David Chinner <dgc@sgi.com>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
16 years ago[XFS] Remove superflous xfs_readsb call in xfs_mountfs.
Christoph Hellwig [Thu, 6 Mar 2008 02:49:36 +0000 (13:49 +1100)]
[XFS] Remove superflous xfs_readsb call in xfs_mountfs.

When xfs_mountfs is called by xfs_mount xfs_readsb was called 35 lines
above unconditionally, so there is no need to try to read the superblock
if it's not present. If any other port doesn't have the superblock read at
this point it should just call it directly from it's xfs_mount equivalent.

SGI-PV: 976035
SGI-Modid: xfs-linux-melb:xfs-kern:30603a

Signed-off-by: Christoph Hellwig <hch@infradead.org>
Signed-off-by: Donald Douwsma <donaldd@sgi.com>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
16 years ago[XFS] kill t_sema member of struct xfs_trans
Niv Sardi [Thu, 6 Mar 2008 02:49:26 +0000 (13:49 +1100)]
[XFS] kill t_sema member of struct xfs_trans

It's completely unused so we might aswell kill it. Note that there is
another t_sema in struct xlog_ticket, which is used and actually an sv_t
despite the name. That one is left untouched by this patch.

SGI-PV: 971186
SGI-Modid: xfs-linux-melb:xfs-kern:30591a

Signed-off-by: Niv Sardi <xaiki@sgi.com>
Signed-off-by: Christoph Hellwig <hch@infradead.org>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
16 years ago[XFS] cleanup vnode use in xfs_bmap.c
Christoph Hellwig [Thu, 6 Mar 2008 02:46:49 +0000 (13:46 +1100)]
[XFS] cleanup vnode use in xfs_bmap.c

SGI-PV: 976035
SGI-Modid: xfs-linux-melb:xfs-kern:30553a

Signed-off-by: Christoph Hellwig <hch@infradead.org>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
16 years ago[XFS] cleanup vnode use in xfs_iops.c
Christoph Hellwig [Thu, 6 Mar 2008 02:46:43 +0000 (13:46 +1100)]
[XFS] cleanup vnode use in xfs_iops.c

SGI-PV: 976035
SGI-Modid: xfs-linux-melb:xfs-kern:30552a

Signed-off-by: Christoph Hellwig <hch@infradead.org>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
16 years ago[XFS] cleanup vnode use in xfs_lrw.c
Christoph Hellwig [Thu, 6 Mar 2008 02:46:37 +0000 (13:46 +1100)]
[XFS] cleanup vnode use in xfs_lrw.c

SGI-PV: 976035
SGI-Modid: xfs-linux-melb:xfs-kern:30551a

Signed-off-by: Christoph Hellwig <hch@infradead.org>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
16 years ago[XFS] cleanup vnode use in xfs_lookup
Christoph Hellwig [Thu, 6 Mar 2008 02:46:25 +0000 (13:46 +1100)]
[XFS] cleanup vnode use in xfs_lookup

SGI-PV: 976035
SGI-Modid: xfs-linux-melb:xfs-kern:30550a

Signed-off-by: Christoph Hellwig <hch@infradead.org>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
16 years ago[XFS] cleanup vnode use in xfs_symlink and xfs_rename
Christoph Hellwig [Thu, 6 Mar 2008 02:46:19 +0000 (13:46 +1100)]
[XFS] cleanup vnode use in xfs_symlink and xfs_rename

SGI-PV: 976035
SGI-Modid: xfs-linux-melb:xfs-kern:30548a

Signed-off-by: Christoph Hellwig <hch@infradead.org>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
16 years ago[XFS] cleanup vnode use in xfs_link
Christoph Hellwig [Thu, 6 Mar 2008 02:46:12 +0000 (13:46 +1100)]
[XFS] cleanup vnode use in xfs_link

SGI-PV: 976035
SGI-Modid: xfs-linux-melb:xfs-kern:30547a

Signed-off-by: Christoph Hellwig <hch@infradead.org>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
16 years ago[XFS] cleanup vnode use in xfs_create/mknod/mkdir
Christoph Hellwig [Thu, 6 Mar 2008 02:46:05 +0000 (13:46 +1100)]
[XFS] cleanup vnode use in xfs_create/mknod/mkdir

SGI-PV: 976035
SGI-Modid: xfs-linux-melb:xfs-kern:30546a

Signed-off-by: Christoph Hellwig <hch@infradead.org>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
16 years ago[XFS] cleanup vnode use in dmapi calls
Christoph Hellwig [Thu, 6 Mar 2008 02:45:58 +0000 (13:45 +1100)]
[XFS] cleanup vnode use in dmapi calls

SGI-PV: 976035
SGI-Modid: xfs-linux-melb:xfs-kern:30545a

Signed-off-by: Christoph Hellwig <hch@infradead.org>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
16 years ago[XFS] Use power-of-2 sized buffers to reduce overhead
David Chinner [Thu, 6 Mar 2008 02:45:43 +0000 (13:45 +1100)]
[XFS] Use power-of-2 sized buffers to reduce overhead

Now that the ktrace_enter() code is using atomics, the non-power-of-2
buffer sizes - which require modulus operations to get the index - are
showing up as using substantial CPU in the profiles.

Force the buffer sizes to be rounded up to the nearest power of two and
use masking rather than modulus operations to convert the index counter to
the buffer index. This reduces ktrace_enter overhead to 8% of a CPU time,
and again almost halves the trace intensive test runtime.

SGI-PV: 977546
SGI-Modid: xfs-linux-melb:xfs-kern:30538a

Signed-off-by: David Chinner <dgc@sgi.com>
Signed-off-by: Christoph Hellwig <hch@infradead.org>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
16 years ago[XFS] Use atomic counters for ktrace buffer indexes
David Chinner [Thu, 6 Mar 2008 02:45:35 +0000 (13:45 +1100)]
[XFS] Use atomic counters for ktrace buffer indexes

ktrace_enter() is consuming vast amounts of CPU time due to the use of a
single global lock for protecting buffer index increments. Change it to
use per-buffer atomic counters - this reduces ktrace_enter() overhead
during a trace intensive test on a 4p machine from 58% of all CPU time to
12% and halves test runtime.

SGI-PV: 977546
SGI-Modid: xfs-linux-melb:xfs-kern:30537a

Signed-off-by: David Chinner <dgc@sgi.com>
Signed-off-by: Christoph Hellwig <hch@infradead.org>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
16 years ago[XFS] Update c/mtime correctly on truncates
David Chinner [Thu, 6 Mar 2008 02:45:29 +0000 (13:45 +1100)]
[XFS] Update c/mtime correctly on truncates

XFS changes the c/mtime of an inode when truncating it to the same size.
The c/mtime is only supposed to change if the size is changed. Not to be
confused with ftruncate, where the c/mtime is supposed to be changed even
if the size is not changed.

The Linux VFS encodes this semantic difference in the flags it sends down
to ->setattr, which XFS currently ignores. We need to make XFS pay
attention to the VFS flags and hence Do The Right Thing.

SGI-PV: 977547
SGI-Modid: xfs-linux-melb:xfs-kern:30536a

Signed-off-by: David Chinner <dgc@sgi.com>
Signed-off-by: Christoph Hellwig <hch@infradead.org>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
16 years ago[XFS] don't encode parent in nfs filehandles unless nessecary
Christoph Hellwig [Thu, 6 Mar 2008 02:45:16 +0000 (13:45 +1100)]
[XFS] don't encode parent in nfs filehandles unless nessecary

As Dave pointed out after the export ops changes we now always encode the
parent into the filehandle for regular files, but it's not actually needed
when the filesystem is export with no_subtree_check. This one-liner fixes
xfs_fs_encode_fh to skip encoding the parent unless nessecary.

SGI-PV: 976035
SGI-Modid: xfs-linux-melb:xfs-kern:30535a

Signed-off-by: Christoph Hellwig <hch@infradead.org>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
16 years ago[XFS] kill xfs_rwlock/xfs_rwunlock
Christoph Hellwig [Thu, 6 Mar 2008 02:44:57 +0000 (13:44 +1100)]
[XFS] kill xfs_rwlock/xfs_rwunlock

We can just use xfs_ilock/xfs_iunlock instead and get rid of the ugly
bhv_vrwlock_t.

SGI-PV: 976035
SGI-Modid: xfs-linux-melb:xfs-kern:30533a

Signed-off-by: Christoph Hellwig <hch@infradead.org>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
16 years ago[XFS] kill xfs_get_dir_entry
Christoph Hellwig [Thu, 6 Mar 2008 02:44:50 +0000 (13:44 +1100)]
[XFS] kill xfs_get_dir_entry

Instead of of xfs_get_dir_entry use a macro to get the xfs_inode from the
dentry in the callers and grab the reference manually.

Only grab the reference once as it's fine to keep it over the dmapi calls.
(And even that reference is actually superflous in Linux but I'll leave
that for another patch)

SGI-PV: 976035
SGI-Modid: xfs-linux-melb:xfs-kern:30531a

Signed-off-by: Christoph Hellwig <hch@infradead.org>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
16 years ago[XFS] vnode cleanup in xfs_fs_subr.c
Christoph Hellwig [Thu, 6 Mar 2008 02:44:41 +0000 (13:44 +1100)]
[XFS] vnode cleanup in xfs_fs_subr.c

Cleanup the unneeded intermediate vnode step in the flushing helpers and
go directly from the xfs_inode to the struct address_space.

SGI-PV: 976035
SGI-Modid: xfs-linux-melb:xfs-kern:30530a

Signed-off-by: Christoph Hellwig <hch@infradead.org>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
16 years ago[XFS] cleanup xfs_vn_mknod
Christoph Hellwig [Thu, 6 Mar 2008 02:44:35 +0000 (13:44 +1100)]
[XFS] cleanup xfs_vn_mknod

- use proper goto based unwinding instead of the current mess of
  multiple conditionals
- rename ip to inode because that's the normal convention for Linux
  inodes while ip is the convention for xfs_inodes
- remove unlikely checks for the default_acl - branches marked unlikely
  might lead to extreme branch bredictor slowdons if taken and for some
  workloads a default acl is quite common
- properly indent the switch statements
- remove xfs_has_fs_struct as nfsd has a fs_struct in any semi-recent
  kernel

SGI-PV: 976035
SGI-Modid: xfs-linux-melb:xfs-kern:30529a

Signed-off-by: Christoph Hellwig <hch@infradead.org>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
16 years ago[XFS] Use atomics for iclog reference counting
David Chinner [Thu, 6 Mar 2008 02:44:14 +0000 (13:44 +1100)]
[XFS] Use atomics for iclog reference counting

Now that we update the log tail LSN less frequently on transaction
completion, we pass the contention straight to the global log state lock
(l_iclog_lock) during transaction completion.

We currently have to take this lock to decrement the iclog reference
count. there is a reference count on each iclog, so we need to take Ã¾he
global lock for all refcount changes.

When large numbers of processes are all doing small trnasctions, the iclog
reference counts will be quite high, and the state change that absolutely
requires the l_iclog_lock is the except rather than the norm.

Change the reference counting on the iclogs to use atomic_inc/dec so that
we can use atomic_dec_and_lock during transaction completion and avoid the
need for grabbing the l_iclog_lock for every reference count decrement
except the one that matters - the last.

SGI-PV: 975671
SGI-Modid: xfs-linux-melb:xfs-kern:30505a

Signed-off-by: David Chinner <dgc@sgi.com>
Signed-off-by: Tim Shimmin <tes@sgi.com>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
16 years ago[XFS] Prevent AIL lock contention during transaction completion
David Chinner [Thu, 6 Mar 2008 02:44:06 +0000 (13:44 +1100)]
[XFS] Prevent AIL lock contention during transaction completion

When hundreds of processors attempt to commit transactions at the same
time, they can contend on the AIL lock when updating the tail LSN held in
the in-core log structure.

At the moment, the tail LSN is only needed when actually writing out an
iclog, so it really does not need to be updated on every single
transaction completion - only those that result in switching iclogs and
flushing them to disk.

The result is that we reduce the number of times we need to grab the AIL
lock and the log grant lock by up to two orders of magnitude on large
processor count machines. The problem has previously been hidden by AIL
lock contention walking the AIL list which was recently solved and
uncovered this issue.

SGI-PV: 975671
SGI-Modid: xfs-linux-melb:xfs-kern:30504a

Signed-off-by: David Chinner <dgc@sgi.com>
Signed-off-by: Tim Shimmin <tes@sgi.com>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
16 years ago[XFS] Use xfs_inode_clean() in more places
David Chinner [Thu, 6 Mar 2008 02:43:59 +0000 (13:43 +1100)]
[XFS] Use xfs_inode_clean() in more places

Remove open coded checks for the whether the inode is clean and replace
them with an inlined function.

SGI-PV: 977461
SGI-Modid: xfs-linux-melb:xfs-kern:30503a

Signed-off-by: David Chinner <dgc@sgi.com>
Signed-off-by: Christoph Hellwig <hch@infradead.org>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
16 years ago[XFS] Remove the xfs_icluster structure
David Chinner [Thu, 6 Mar 2008 02:43:49 +0000 (13:43 +1100)]
[XFS] Remove the xfs_icluster structure

Remove the xfs_icluster structure and replace with a radix tree lookup.

We don't need to keep a list of inodes in each cluster around anymore as
we can look them up quickly when we need to. The only time we need to do
this now is during inode writeback.

Factor the inode cluster writeback code out of xfs_iflush and convert it
to use radix_tree_gang_lookup() instead of walking a list of inodes built
when we first read in the inodes.

This remove 3 pointers from each xfs_inode structure and the xfs_icluster
structure per inode cluster. Hence we reduce the cache footprint of the
xfs_inodes by between 5-10% depending on cluster sparseness.

To be truly efficient we need a radix_tree_gang_lookup_range() call to
stop searching once we are past the end of the cluster instead of trying
to find a full cluster's worth of inodes.

Before (ia64):

$ cat /sys/slab/xfs_inode/object_size 536

After:

$ cat /sys/slab/xfs_inode/object_size 512

SGI-PV: 977460
SGI-Modid: xfs-linux-melb:xfs-kern:30502a

Signed-off-by: David Chinner <dgc@sgi.com>
Signed-off-by: Christoph Hellwig <hch@infradead.org>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
16 years ago[XFS] Don't block pdflush when writing back inodes
David Chinner [Thu, 6 Mar 2008 02:43:42 +0000 (13:43 +1100)]
[XFS] Don't block pdflush when writing back inodes

When pdflush is writing back inodes, it can get stuck on inode cluster
buffers that are currently under I/O. This occurs when we write data to
multiple inodes in the same inode cluster at the same time.

Effectively, delayed allocation marks the inode dirty during the data
writeback. Hence if the inode cluster was flushed during the writeback of
the first inode, the writeback of the second inode will block waiting for
the inode cluster write to complete before writing it again for the newly
dirtied inode.

Basically, we want to avoid this from happening so we don't block pdflush
and slow down all of writeback. Hence we introduce a non-blocking async
inode flush flag that pdflush uses. If this flag is set, we use
non-blocking operations (e.g. try locks) whereever we can to avoid
blocking or extra I/O being issued.

SGI-PV: 970925
SGI-Modid: xfs-linux-melb:xfs-kern:30501a

Signed-off-by: David Chinner <dgc@sgi.com>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
16 years ago[XFS] Factor xfs_itobp() and xfs_inotobp().
David Chinner [Thu, 6 Mar 2008 02:43:34 +0000 (13:43 +1100)]
[XFS] Factor xfs_itobp() and xfs_inotobp().

The only difference between the functions is one passes an inode for the
lookup, the other passes an inode number. However, they don't do the same
validity checking or set all the same state on the buffer that is returned
yet they should.

Factor the functions into a common implementation.

SGI-PV: 970925
SGI-Modid: xfs-linux-melb:xfs-kern:30500a

Signed-off-by: David Chinner <dgc@sgi.com>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
16 years ago[XFS] Fix regression due to refcache removal
Lachlan McIlroy [Thu, 6 Mar 2008 02:43:27 +0000 (13:43 +1100)]
[XFS] Fix regression due to refcache removal

SGI-PV: 971186
SGI-Modid: xfs-linux-melb:xfs-kern:30490a

Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
Signed-off-by: Donald Douwsma <donaldd@sgi.com>
16 years ago[XFS] Remove the xfs_refcache
Donald Douwsma [Thu, 6 Mar 2008 02:43:20 +0000 (13:43 +1100)]
[XFS] Remove the xfs_refcache

Remove the xfs_refcache, it was only needed while we were still
building for 2.4 kernels.

SGI-PV: 971186
SGI-Modid: xfs-linux-melb:xfs-kern:30472a

Signed-off-by: Donald Douwsma <donaldd@sgi.com>
Signed-off-by: Christoph Hellwig <hch@infradead.org>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
16 years ago[XFS] make inode reclaim synchronise with xfs_iflush_done()
Lachlan McIlroy [Thu, 6 Mar 2008 02:43:11 +0000 (13:43 +1100)]
[XFS] make inode reclaim synchronise with xfs_iflush_done()

On a forced shutdown, xfs_finish_reclaim() will skip flushing the inode.
If the inode flush lock is not already held and there is an outstanding
xfs_iflush_done() then we might free the inode prematurely. By acquiring
and releasing the flush lock we will synchronise with xfs_iflush_done().

SGI-PV: 909874
SGI-Modid: xfs-linux-melb:xfs-kern:30468a

Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
Signed-off-by: David Chinner <dgc@sgi.com>
16 years ago[XFS] actually check error returned by xfs_flush_pages, clean up and
Niv Sardi [Thu, 6 Mar 2008 02:43:03 +0000 (13:43 +1100)]
[XFS] actually check error returned by xfs_flush_pages, clean up and
bailout if fails.

SGI-PV: 973041
SGI-Modid: xfs-linux-melb:xfs-kern:30462a

Signed-off-by: Niv Sardi <xaiki@sgi.com>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
16 years agoLinux 2.6.25 v2.6.25
Linus Torvalds [Thu, 17 Apr 2008 02:49:44 +0000 (19:49 -0700)]
Linux 2.6.25

16 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/bart/ide-2.6
Linus Torvalds [Thu, 17 Apr 2008 01:58:37 +0000 (18:58 -0700)]
Merge git://git.kernel.org/pub/scm/linux/kernel/git/bart/ide-2.6

* git://git.kernel.org/pub/scm/linux/kernel/git/bart/ide-2.6:
  it821x: do not describe noraid parameter with its value
  Pb1200/DBAu1200: fix bad IDE resource size
  Au1200: IDE driver build fix
  Au1200: kill IDE driver function prototypes
  avr32 mustn't select HAVE_IDE

16 years agoit821x: do not describe noraid parameter with its value
Paul Bolle [Wed, 16 Apr 2008 23:14:33 +0000 (01:14 +0200)]
it821x: do not describe noraid parameter with its value

Describe noraid parameter with its name (and not its value).

Signed-off-by: Paul Bolle <pebolle@tiscali.nl>
Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>
16 years agoPb1200/DBAu1200: fix bad IDE resource size
Sergei Shtylyov [Wed, 16 Apr 2008 23:14:33 +0000 (01:14 +0200)]
Pb1200/DBAu1200: fix bad IDE resource size

The header files for the Pb1200/DBAu1200 boards have wrong definition for the
IDE interface's decoded range length -- it should be 512 bytes according to
what the IDE driver does.  In addition, the IDE platform device claims 1 byte
too many for its memory resource -- fix the platform code and the IDE driver
in accordance.

Signed-off-by: Sergei Shtylyov <sshtylyov@ru.mvista.com>
Cc: Ralf Baechle <ralf@linux-mips.org>
Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>
16 years agoAu1200: IDE driver build fix
Sergei Shtylyov [Wed, 16 Apr 2008 23:14:33 +0000 (01:14 +0200)]
Au1200: IDE driver build fix

The driver fails to compile with CONFIG_BLK_DEV_IDE_AU1XXX_MDMA2_DBDMA enabled:

drivers/ide/mips/au1xxx-ide.c: In function `auide_build_dmatable':
drivers/ide/mips/au1xxx-ide.c:256: error: implicit declaration of function
`sg_virt'
drivers/ide/mips/au1xxx-ide.c:275: error: implicit declaration of function
`sg_next'
drivers/ide/mips/au1xxx-ide.c:275: warning: assignment makes pointer from
integer without a cast

Fix this by including <linux/scatterlist.h>. While at it, remove the #include's
without which the driver happily builds.

Signed-off-by: Sergei Shtylyov <sshtylyov@ru.mvista.com>
Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>
16 years agoAu1200: kill IDE driver function prototypes
Sergei Shtylyov [Wed, 16 Apr 2008 23:14:33 +0000 (01:14 +0200)]
Au1200: kill IDE driver function prototypes

Fix these warnings emitted when compiling drivers/ide/mips/au1xxx-ide.c:

include/asm/mach-au1x00/au1xxx_ide.h:137: warning: 'auide_tune_drive' declared
`static' but never defined
include/asm/mach-au1x00/au1xxx_ide.h:138: warning: 'auide_tune_chipset' declared
 `static' but never defined

by wiping out the whole "function prototyping" section from the header file
<asm-mips/mach-au1x00/au1xxx_ide.h> as it mostly declared functions that are
already dead in the IDE driver; move the only useful prototype into the driver.

Signed-off-by: Sergei Shtylyov <sshtylyov@ru.mvista.com>
Cc: Ralf Baechle <ralf@linux-mips.org>
Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>
16 years agoavr32 mustn't select HAVE_IDE
Adrian Bunk [Wed, 16 Apr 2008 23:14:32 +0000 (01:14 +0200)]
avr32 mustn't select HAVE_IDE

There's a libata based PATA driver for avr32, but no support for
drivers/ide/ on avr32.

This patch fixes the following compile error:

<--  snip  -->

...
  CC [M]  drivers/ide/ide-cd.o
In file included from /home/bunk/linux/kernel-2.6/git/linux-2.6/drivers/ide/ide-cd.c:37:
/home/bunk/linux/kernel-2.6/git/linux-2.6/include/linux/ide.h:209:21: error: asm/ide.h: No such file or directory
make[3]: *** [drivers/ide/ide-cd.o] Error 1

<--  snip  -->

Signed-off-by: Adrian Bunk <bunk@kernel.org>
Cc: Haavard Skinnemoen <hskinnemoen@atmel.com>
Acked-by: Sam Ravnborg <sam@ravnborg.org>
Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>
16 years agoMerge branch 'for-linus' of git://git.kernel.dk/linux-2.6-block
Linus Torvalds [Wed, 16 Apr 2008 14:45:45 +0000 (07:45 -0700)]
Merge branch 'for-linus' of git://git.kernel.dk/linux-2.6-block

* 'for-linus' of git://git.kernel.dk/linux-2.6-block:
  block: update git url for blktrace
  io context: increment task attachment count in ioc_task_link()

16 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/usb-2.6
Linus Torvalds [Wed, 16 Apr 2008 14:45:05 +0000 (07:45 -0700)]
Merge git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/usb-2.6

* git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/usb-2.6:
  USB: remove broken usb-serial num_endpoints check
  USB: option: Add new vendor ID and device ID for AMOI HSDPA modem
  USB: support more Huawei data card product IDs
  USB: option.c: add more device IDs
  USB: Obscure Maxon BP3-USB Device Support 16d8:6280 for option driver

16 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/davem/net-2.6
Linus Torvalds [Wed, 16 Apr 2008 14:44:27 +0000 (07:44 -0700)]
Merge git://git.kernel.org/pub/scm/linux/kernel/git/davem/net-2.6

* git://git.kernel.org/pub/scm/linux/kernel/git/davem/net-2.6:
  [TCP]: Add return value indication to tcp_prune_ofo_queue().
  PS3: gelic: fix the oops on the broken IE returned from the hypervisor
  b43legacy: fix DMA mapping leakage
  mac80211: remove message on receiving unexpected unencrypted frames
  Update rt2x00 MAINTAINERS entry
  Add rfkill to MAINTAINERS file
  rfkill: Fix device type check when toggling states
  b43legacy: Fix usage of struct device used for DMAing
  ssb: Fix usage of struct device used for DMAing
  MAINTAINERS: move to generic repository for iwlwifi
  b43legacy: fix initvals loading on bcm4303
  rtl8187: Add missing priv->vif assignments
  netconsole: only set CON_PRINTBUFFER if the user specifies a netconsole
  [CAN]: Update documentation of struct sockaddr_can
  MAINTAINERS: isdn4linux@listserv.isdn4linux.de is subscribers-only
  [TCP]: Fix never pruned tcp out-of-order queue.
  [NET_SCHED] sch_api: fix qdisc_tree_decrease_qlen() loop

16 years agoAFS: Do not describe debug parameters with their value
Paul Bolle [Wed, 16 Apr 2008 10:08:22 +0000 (11:08 +0100)]
AFS: Do not describe debug parameters with their value

Describe debug parameters with their names (and not their values).

Signed-off-by: Paul Bolle <pebolle@tiscali.nl>
Signed-off-by: David Howells <dhowells@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agoUSB: remove broken usb-serial num_endpoints check
Greg Kroah-Hartman [Mon, 14 Apr 2008 21:17:29 +0000 (14:17 -0700)]
USB: remove broken usb-serial num_endpoints check

The num_interrupt_in, num_bulk_in, and other checks in the usb-serial
code are just wrong, there are too many different devices out there with
different numbers of endpoints.  We need to just be sticking with the
device ids instead of trying to catch this kind of thing.  It broke too
many different devices.

This fixes a large number of usb-serial devices to get them working
properly again.

Cc: Oliver Neukum <oliver@neukum.org>
Cc: stable <stable@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
16 years agoUSB: option: Add new vendor ID and device ID for AMOI HSDPA modem
tang kai [Mon, 14 Apr 2008 02:06:35 +0000 (10:06 +0800)]
USB: option: Add new vendor ID and device ID for AMOI HSDPA modem

This patch add new vendor ID and device ID  for AMOI HSDPA modem.

From: tang kai <tangk73@hotmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
16 years agoUSB: support more Huawei data card product IDs
fangxiaozhi [Thu, 10 Apr 2008 06:51:06 +0000 (14:51 +0800)]
USB: support more Huawei data card product IDs

 - declare the unusal device for Huawei data card devices in
   unusual_devs.h
 - disable the product ID matching for Huawei data card devices in
   usb_match_device function of driver.c
 - declare the product IDs in option.c.

Signed-off-by: fangxiaozhi <huananhu@huawei.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
16 years agoUSB: option.c: add more device IDs
Matthias Urlichs [Thu, 10 Apr 2008 08:13:32 +0000 (10:13 +0200)]
USB: option.c: add more device IDs

Add devices by AMOI and NovatelWireless.

Signed-Off-By: Matthias Urlichs <matthias@urlichs.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
16 years agoUSB: Obscure Maxon BP3-USB Device Support 16d8:6280 for option driver
James Cameron [Wed, 9 Apr 2008 08:59:13 +0000 (18:59 +1000)]
USB: Obscure Maxon BP3-USB Device Support 16d8:6280 for option driver

The modem was detected, the ttyUSB{0,1,2} appeared, a call could be
made, and the expected data rate was achieved.  Tested for an hour or
two, total of 100Mb.  I shall do more testing.

Signed-off-by: James Cameron <quozl@laptop.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
16 years ago[TCP]: Add return value indication to tcp_prune_ofo_queue().
Vitaliy Gusev [Wed, 16 Apr 2008 03:26:34 +0000 (20:26 -0700)]
[TCP]: Add return value indication to tcp_prune_ofo_queue().

Returns non-zero if tp->out_of_order_queue was seen non-empty.
This allows tcp_try_rmem_schedule() to return early.

Signed-off-by: Vitaliy Gusev <vgusev@openvz.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
16 years agoacpi: unneccessary to scan the PCI bus already scanned
yakui.zhao@intel.com [Tue, 15 Apr 2008 21:34:49 +0000 (14:34 -0700)]
acpi: unneccessary to scan the PCI bus already scanned

http://bugzilla.kernel.org/show_bug.cgi?id=10124

this change:

      commit 08f1c192c3c32797068bfe97738babb3295bbf42
      Author: Muli Ben-Yehuda <muli@il.ibm.com>
      Date:   Sun Jul 22 00:23:39 2007 +0300

         x86-64: introduce struct pci_sysdata to facilitate sharing of ->sysdata

         This patch introduces struct pci_sysdata to x86 and x86-64, and
         converts the existing two users (NUMA, Calgary) to use it.

         This lays the groundwork for having other users of sysdata, such as
         the PCI domains work.

         The Calgary bits are tested, the NUMA bits just look ok.

replaces pcibios_scan_root by pci_scan_bus_parented...

but in pcibios_scan_root we have a check about scanned busses.

Cc: <yakui.zhao@intel.com>
Cc: Stian Jordet <stian@jordet.net>
Cc: Len Brown <lenb@kernel.org>
Cc: Greg KH <greg@kroah.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: "Yinghai Lu" <yhlu.kernel@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agoacpi thermal trip points increased to 12
Krzysztof Helt [Tue, 15 Apr 2008 21:34:47 +0000 (14:34 -0700)]
acpi thermal trip points increased to 12

The THERMAL_MAX_TRIPS value is set to 10.  It is too few for the Compaq AP550
machine which has 12 trip points.

Signed-off-by: Krzysztof Helt <krzysztof.h1@wp.pl>
Cc: Len Brown <lenb@kernel.org>
Cc: Zhang Rui <rui.zhang@intel.com>
Cc: "Rafael J. Wysocki" <rjw@sisk.pl>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agoacpi: bus: check once more for an empty list after locking it
Chuck Ebbert [Tue, 15 Apr 2008 21:34:47 +0000 (14:34 -0700)]
acpi: bus: check once more for an empty list after locking it

List could have become empty after the unlocked check that was made earlier,
so check again inside the lock.

Should fix https://bugzilla.redhat.com/show_bug.cgi?id=427765

Signed-off-by: Chuck Ebbert <cebbert@redhat.com>
Cc: <stable@kernel.org>
Cc: Len Brown <lenb@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agospi: spi_s3c24xx must initialize num_chipselect
Ben Dooks [Tue, 15 Apr 2008 21:34:46 +0000 (14:34 -0700)]
spi: spi_s3c24xx must initialize num_chipselect

The SPI core now expects num_chipselect to be set correctly as due to added
checks on the chip being selected before an transfer is allowed.  This patch
adds a num_cs field to the platform data which needs to be set correctly
before adding the SPI platform device.

Signed-off-by: Ben Dooks <ben-linux@fluff.org>
Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
Cc: Russell King <rmk@arm.linux.org.uk>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agospi: spi_s3c24xx must initialize bus_num
Ben Dooks [Tue, 15 Apr 2008 21:34:45 +0000 (14:34 -0700)]
spi: spi_s3c24xx must initialize bus_num

Pass the bus number we expect the S3C24XX SPI driver to attach to via the
platform data.

Signed-off-by: Ben Dooks <ben-linux@fluff.org>
Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agospi: spi_s3c24xx driver must init completion
Ben Dooks [Tue, 15 Apr 2008 21:34:44 +0000 (14:34 -0700)]
spi: spi_s3c24xx driver must init completion

The s3c24xx_spi_txrx() function should initialise the completion each time
before using it, otherwise we end up with the possibility of returning success
before the interrupt handler has processed all the data.

Signed-off-by: Ben Dooks <ben-linux@fluff.org>
Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agovfs: fix possible deadlock in ext2, ext3, ext4 when using xattrs
Jan Kara [Tue, 15 Apr 2008 21:34:43 +0000 (14:34 -0700)]
vfs: fix possible deadlock in ext2, ext3, ext4 when using xattrs

mb_cache_entry_alloc() was allocating cache entries with GFP_KERNEL.  But
filesystems are calling this function while holding xattr_sem so possible
recursion into the fs violates locking ordering of xattr_sem and transaction
start / i_mutex for ext2-4.  Change mb_cache_entry_alloc() so that filesystems
can specify desired gfp mask and use GFP_NOFS from all of them.

Signed-off-by: Jan Kara <jack@suse.cz>
Reported-by: Dave Jones <davej@redhat.com>
Cc: <linux-ext4@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agoDocumentation: correct overcommit caveat in hugetlbpage.txt
Nishanth Aravamudan [Tue, 15 Apr 2008 21:34:43 +0000 (14:34 -0700)]
Documentation: correct overcommit caveat in hugetlbpage.txt

As shown by Gurudas Pai recently, we can put hugepages into the surplus
state (by echo 0 > /proc/sys/vm/nr_hugepages), even when
/proc/sys/vm/nr_overcommit_hugepages is 0. This is actually correct, to
allow the original goal (shrink the static pool to 0) to succeed (we are
converting hugepages to surplus because they are in use). However, the
documentation does not accurately reflect this case. Update it.

Signed-off-by: Nishanth Aravamudan <nacc@us.ibm.com>
Acked-by: Andy Whitcroft <apw@shadowen.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agoadd "Isolate" migratetype name to /proc/pagetypeinfo
KOSAKI Motohiro [Tue, 15 Apr 2008 21:34:42 +0000 (14:34 -0700)]
add "Isolate" migratetype name to /proc/pagetypeinfo

In a5d76b54a3f3a40385d7f76069a2feac9f1bad63 (memory unplug: page isolation by
KAMEZAWA Hiroyuki), "isolate" migratetype added.  but unfortunately, it
doesn't treat /proc/pagetypeinfo display logic.

this patch add "Isolate" to pagetype name field.

/proc/pagetype
before:
------------------------------------------------------------------------------------------------------------------------
Free pages count per migrate type at order       0      1      2      3      4      5      6      7      8      9     10
Node    0, zone      DMA, type    Unmovable      1      2      2      2      1      2      2      1      1      0      0
Node    0, zone      DMA, type  Reclaimable      0      0      0      0      0      0      0      0      0      0      0
Node    0, zone      DMA, type      Movable      2      3      3      1      3      3      2      0      0      0      0
Node    0, zone      DMA, type      Reserve      0      0      0      0      0      0      0      0      0      0      1
Node    0, zone      DMA, type       <NULL>      0      0      0      0      0      0      0      0      0      0      0
Node    0, zone   Normal, type    Unmovable      1      9      7      4      1      1      1      1      0      0      0
Node    0, zone   Normal, type  Reclaimable      5      2      0      0      1      1      0      0      0      1      0
Node    0, zone   Normal, type      Movable      0      1      1      0      0      0      1      0      0      1     60
Node    0, zone   Normal, type      Reserve      0      0      0      0      0      0      0      0      0      0      1
Node    0, zone   Normal, type       <NULL>      0      0      0      0      0      0      0      0      0      0      0
Node    0, zone  HighMem, type    Unmovable      0      0      1      1      1      0      1      1      2      2      0
Node    0, zone  HighMem, type  Reclaimable      0      0      0      0      0      0      0      0      0      0      0
Node    0, zone  HighMem, type      Movable    236     62      6      2      2      1      1      0      1      1     16
Node    0, zone  HighMem, type      Reserve      0      0      0      0      0      0      0      0      0      0      1
Node    0, zone  HighMem, type       <NULL>      0      0      0      0      0      0      0      0      0      0      0

Number of blocks type     Unmovable  Reclaimable      Movable      Reserve       <NULL>
Node 0, zone      DMA            1            0            2       1            0
Node 0, zone   Normal           10           40          169       1            0
Node 0, zone  HighMem            2            0          283       1            0

after:
------------------------------------------------------------------------------------------------------------------------
Free pages count per migrate type at order       0      1      2      3      4      5      6      7      8      9     10
Node    0, zone      DMA, type    Unmovable      1      2      2      2      1      2      2      1      1      0      0
Node    0, zone      DMA, type  Reclaimable      0      0      0      0      0      0      0      0      0      0      0
Node    0, zone      DMA, type      Movable      2      3      3      1      3      3      2      0      0      0      0
Node    0, zone      DMA, type      Reserve      0      0      0      0      0      0      0      0      0      0      1
Node    0, zone      DMA, type      Isolate      0      0      0      0      0      0      0      0      0      0      0
Node    0, zone   Normal, type    Unmovable      0      2      1      1      0      1      0      0      0      0      0
Node    0, zone   Normal, type  Reclaimable      1      1      1      1      1      0      1      1      1      0      0
Node    0, zone   Normal, type      Movable      0      1      1      1      0      1      0      1      0      0    196
Node    0, zone   Normal, type      Reserve      0      0      0      0      0      0      0      0      0      0      1
Node    0, zone   Normal, type      Isolate      0      0      0      0      0      0      0      0      0      0      0
Node    0, zone  HighMem, type    Unmovable      0      1      0      0      0      1      1      1      2      2      0
Node    0, zone  HighMem, type  Reclaimable      0      0      0      0      0      0      0      0      0      0      0
Node    0, zone  HighMem, type      Movable      1      0      1      1      0      0      0      0      1      0    200
Node    0, zone  HighMem, type      Reserve      0      0      0      0      0      0      0      0      0      0      1
Node    0, zone  HighMem, type      Isolate      0      0      0      0      0      0      0      0      0      0      0

Number of blocks type     Unmovable  Reclaimable      Movable      Reserve      Isolate
Node 0, zone      DMA            1            0            2       1            0
Node 0, zone   Normal            8            4          207       1            0
Node 0, zone  HighMem            2            0          283       1            0

Signed-off-by: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Acked-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Acked-by: Mel Gorman <mel@csn.ul.ie>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>