ext4: endless truncate due to nonlocked dio readers
If we have enough aggressive DIO readers, truncate and other dio
waiters will wait forever inside inode_dio_wait(). It is reasonable
to disable nonlock DIO read optimization during truncate.
Reviewed-by: Jan Kara <jack@suse.cz> Signed-off-by: Dmitry Monakhov <dmonakhov@openvz.org> Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
In order to serialize with unlocked DIO reads we have to
rearrange wait sequence
1) update i_size first
2) if i_size about to be reduced wait for outstanding DIO requests
3) and only after that truncate inode blocks
Reviewed-by: Jan Kara <jack@suse.cz> Signed-off-by: Dmitry Monakhov <dmonakhov@openvz.org> Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
Current unwritten extent conversion state-machine is very fuzzy.
- For unknown reason it performs conversion under i_mutex. What for?
My diagnosis:
We already protect extent tree with i_data_sem, truncate and punch_hole
should wait for DIO, so the only data we have to protect is end_io->flags
modification, but only flush_completed_IO and end_io_work modified this
flags and we can serialize them via i_completed_io_lock.
Currently all these games with mutex_trylock result in the following deadlock
truncate: kworker:
ext4_setattr ext4_end_io_work
mutex_lock(i_mutex)
inode_dio_wait(inode) ->BLOCK
DEADLOCK<- mutex_trylock()
inode_dio_done()
#TEST_CASE1_BEGIN
MNT=/mnt_scrach
unlink $MNT/file
fallocate -l $((1024*1024*1024)) $MNT/file
aio-stress -I 100000 -O -s 100m -n -t 1 -c 10 -o 2 -o 3 $MNT/file
sleep 2
truncate -s 0 $MNT/file
#TEST_CASE1_END
Or use 286's xfstests https://github.com/dmonakhov/xfstests/blob/devel/286
This patch makes state machine simple and clean:
(1) xxx_end_io schedule final extent conversion simply by calling
ext4_add_complete_io(), which append it to ei->i_completed_io_list
NOTE1: because of (2A) work should be queued only if
->i_completed_io_list was empty, otherwise the work is scheduled already.
(2) ext4_flush_completed_IO is responsible for handling all pending
end_io from ei->i_completed_io_list
Flushing sequence consists of following stages:
A) LOCKED: Atomically drain completed_io_list to local_list
B) Perform extents conversion
C) LOCKED: move converted io's to to_free list for final deletion
This logic depends on context which we was called from.
D) Final end_io context destruction
NOTE1: i_mutex is no longer required because end_io->flags modification
is protected by ei->ext4_complete_io_lock
Full list of changes:
- Move all completion end_io related routines to page-io.c in order to improve
logic locality
- Move open coded logic from various xx_end_xx routines to ext4_add_complete_io()
- remove EXT4_IO_END_FSYNC
- Improve SMP scalability by removing useless i_mutex which does not
protect io->flags anymore.
- Reduce lock contention on i_completed_io_lock by optimizing list walk.
- Rename ext4_end_io_nolock to end4_end_io and make it static
- Check flush completion status to ext4_ext_punch_hole(). Because it is
not good idea to punch blocks from corrupted inode.
Changes since V3 (in request to Jan's comments):
Fall back to active flush_completed_IO() approach in order to prevent
performance issues with nolocked DIO reads.
Changes since V2:
Fix use-after-free caused by race truncate vs end_io_work
ext4_set_io_unwritten_flag() will increment i_unwritten counter, so
once we mark end_io with EXT4_END_IO_UNWRITTEN we have to revert it back
on error path.
- add missed error checks to prevent counter leakage
- ext4_end_io_nolock() will clear EXT4_END_IO_UNWRITTEN flag to signal
that conversion finished.
- add BUG_ON to ext4_free_end_io() to prevent similar leakage in future.
Visible effect of this bug is that unaligned aio_stress may deadlock
Reviewed-by: Jan Kara <jack@suse.cz> Signed-off-by: Dmitry Monakhov <dmonakhov@openvz.org> Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
Carlos Maiolino [Thu, 27 Sep 2012 13:31:33 +0000 (09:31 -0400)]
ext4: ext4_bread usage audit
When ext4_bread() returns NULL and err is set to zero, this means
there is no phyical block mapped to the specified logical block
number. (Previous to commit 90b0a97323, err was uninitialized in this
case, which caused other problems.)
The directory handling routines use ext4_bread() in many places, the
fact that ext4_bread() now returns NULL with err set to zero could
cause problems since a number of these functions will simply return
the value of err if the result of ext4_bread() was the NULL pointer,
causing the caller of the function to think that the function was
successful.
Since directories should never contain holes, this case can only
happen if the file system is corrupted. This commit audits all of the
callers of ext4_bread(), and makes sure they do the right thing if a
hole in a directory is found by ext4_bread().
Some ext4_bread() callers did not need any changes either because they
already had its own hole detector paths.
Signed-off-by: Carlos Maiolino <cmaiolino@redhat.com> Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
As discussed at the Plumber's Conference, reserve the bit 0x04 in
fallocate() to prevent collisions with a commonly used out-of-tree
patch which implements the no-hide-stale feature.
Wang Sheng-Hui [Thu, 27 Sep 2012 12:00:01 +0000 (08:00 -0400)]
ext4: remove redundant offset check in mext_check_arguments()
In the check code above, if orig_start != donor_start, we would
return -EINVAL. So here, orig_start should be equal with donor_start.
Remove the redundant check here.
Signed-off-by: Wang Sheng-Hui <shhuiw@gmail.com> Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
Eric Sandeen [Thu, 27 Sep 2012 03:30:12 +0000 (23:30 -0400)]
ext4: don't clear orphan list on ro mount with errors
If the file system contains errors and it is being mounted read-only,
don't clear the orphan list. We should minimize changes to the file
system if it is mounted read-only.
Signed-off-by: Eric Sandeen <sandeen@redhat.com> Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
Jan Kara [Thu, 27 Sep 2012 03:11:13 +0000 (23:11 -0400)]
jbd2: fix assertion failure in commit code due to lacking transaction credits
ext4 users of data=journal mode with blocksize < pagesize were
occasionally hitting assertion failure in
jbd2_journal_commit_transaction() checking whether the transaction has
at least as many credits reserved as buffers attached. The core of the
problem is that when a file gets truncated, buffers that still need
checkpointing or that are attached to the committing transaction are
left with buffer_mapped set. When this happens to buffers beyond i_size
attached to a page stradding i_size, subsequent write extending the file
will see these buffers and as they are mapped (but underlying blocks
were freed) things go awry from here.
The assertion failure just coincidentally (and in this case luckily as
we would start corrupting filesystem) triggers due to journal_head not
being properly cleaned up as well.
We fix the problem by unmapping buffers if possible (in lots of cases we
just need a buffer attached to a transaction as a place holder but it
must not be written out anyway). And in one case, we just have to bite
the bullet and wait for transaction commit to finish.
CC: Josef Bacik <jbacik@fusionio.com> Signed-off-by: Jan Kara <jack@suse.cz>
With a minor tweaks regarding minimum extent size to discard and
discarded bytes reporting the FITRIM can be enabled on bigalloc file
system and it works without any problem.
This patch fixes minlen handling and discarded bytes reporting to
take into consideration bigalloc enabled file systems and finally
removes the restriction and allow FITRIM to be used on file system with
bigalloc feature enabled.
Jan Kara [Thu, 27 Sep 2012 01:52:20 +0000 (21:52 -0400)]
ext4: fix fdatasync() for files with only i_size changes
Code tracking when transaction needs to be committed on fdatasync(2) forgets
to handle a situation when only inode's i_size is changed. Thus in such
situations fdatasync(2) doesn't force transaction with new i_size to disk
and that can result in wrong i_size after a crash.
Fix the issue by updating inode's i_datasync_tid whenever its size is
updated.
CC: <stable@vger.kernel.org> # >= 2.6.32 Reported-by: Kristian Nielsen <knielsen@knielsen-hq.org> Signed-off-by: Jan Kara <jack@suse.cz>
ext4_special_inode_operations have their own ifdef CONFIG_EXT4_FS_XATTR
to mask those methods. And ext4_iget also always sets it, so there is
an inconsistency.
ext4: reimplement uninit extent optimization for move_extent_per_page()
Uninitialized extent may became initialized(parallel writeback task)
at any moment after we drop i_data_sem, so we have to recheck extent's
state after we hold page's lock and i_data_sem.
If we about to change page's mapping we must hold page's lock in order to
serialize other users.
ext4: clean up online defrag bugs in move_extent_per_page()
Non-full list of bugs:
1) uninitialized extent optimization does not hold page's lock,
and simply replace brunches after that writeback code goes
crazy because block mapping changed under it's feets
kernel BUG at fs/ext4/inode.c:1434! ( 288'th xfstress)
2) uninitialized extent may became initialized right after we
drop i_data_sem, so extent state must be rechecked
3) Locked pages goes uptodate via following sequence:
->readpage(page); lock_page(page); use_that_page(page)
But after readpage() one may invalidate it because it is
uptodate and unlocked (reclaimer does that)
As result kernel bug at include/linux/buffer_head.c:133!
4) We call write_begin() with already opened stansaction which
result in following deadlock:
->move_extent_per_page()
->ext4_journal_start()-> hold journal transaction
->write_begin()
->ext4_da_write_begin()
->ext4_nonda_switch()
->writeback_inodes_sb_if_idle() --> will wait for journal_stop()
5) try_to_release_page() may fail and it does fail if one of page's bh was
pinned by journal
6) If we about to change page's mapping we MUST hold it's lock during entire
remapping procedure, this is true for both pages(original and donor one)
Fixes:
- Avoid (1) and (2) simply by temproraly drop uninitialized extent handling
optimization, this will be reimplemented later.
- Fix (3) by manually forcing page to uptodate state w/o dropping it's lock
- Fix (4) by rearranging existing locking:
from: journal_start(); ->write_begin
to: write_begin(); journal_extend()
- Fix (5) simply by checking retvalue
- Fix (6) by locking both (original and donor one) pages during extent swap
with help of mext_page_double_lock()
ext4: online defrag is not supported for journaled files
Proper block swap for inodes with full journaling enabled is
truly non obvious task. In order to be on a safe side let's
explicitly disable it for now.
- Remove usless checks, because it is too late to check that inode != NULL
at the moment it was referenced several times.
- Double lock routines looks very ugly and locking ordering relays on
order of i_ino, but other kernel code rely on order of pointers.
Let's make them simple and clean.
- check that inodes belongs to the same SB as soon as possible.
Tao Ma [Wed, 26 Sep 2012 04:08:57 +0000 (00:08 -0400)]
ext4: don't call update_backups() multiple times for the same bg
When performing an online resize, we add a bunch of groups at one time
in ext4_flex_group_add, so in most cases a lot of group descriptors
will be in the same group block. But in the end of this function,
update_backups will be called for every group descriptor and the same
block will be copied and journalled again and again. It is really a
waste.
Fix things so we only update a particular bg descriptor block once and
skip subsequent updates of the same block.
Signed-off-by: Tao Ma <boyu.mt@taobao.com> Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
Yongqiang Yang [Mon, 24 Sep 2012 03:16:03 +0000 (23:16 -0400)]
ext4: check free inode count before allocating an inode
Recently, I ecountered some corrupted filesystems in which some
groups' free inode counts were 65535, it seemed that free inode
count was overflow. This patch teaches ext4 to check free inode
count before allocaing an inode.
Signed-off-by: Yongqiang Yang <xiaoqiangnk@gmail.com> Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
Tao Ma [Thu, 20 Sep 2012 15:35:38 +0000 (11:35 -0400)]
ext4: remove erroneous ext4_superblock_csum_set() in update_backups()
The update_backups() function is used to backup all the metadata
blocks, so we should not take it for granted that 'data' is pointed to
a super block and use ext4_superblock_csum_set to calculate the
checksum there. In case where the data is a group descriptor block,
it will corrupt the last group descriptor, and then e2fsck will
complain about it it.
As all the metadata checksums should already be OK when we do the
backup, remove the wrong ext4_superblock_csum_set and it should be
just fine.
Reported-by: "Theodore Ts'o" <tytso@mit.edu> Signed-off-by: Tao Ma <boyu.mt@taobao.com> Signed-off-by: "Theodore Ts'o" <tytso@mit.edu> Cc: stable@vger.kernel.org
ext4: fix potential deadlock in ext4_nonda_switch()
In ext4_nonda_switch(), if the file system is getting full we used to
call writeback_inodes_sb_if_idle(). The problem is that we can be
holding i_mutex already, and this causes a potential deadlock when
writeback_inodes_sb_if_idle() when it tries to take s_umount. (See
lockdep output below).
As it turns out we don't need need to hold s_umount; the fact that we
are in the middle of the write(2) system call will keep the superblock
pinned. Unfortunately writeback_inodes_sb() checks to make sure
s_umount is taken, and the VFS uses a different mechanism for making
sure the file system doesn't get unmounted out from under us. The
simplest way of dealing with this is to just simply grab s_umount
using a trylock, and skip kicking the writeback flusher thread in the
very unlikely case that we can't take a read lock on s_umount without
blocking.
Also, we now check the cirteria for kicking the writeback thread
before we decide to whether to fall back to non-delayed writeback, so
if there are any outstanding delayed allocation writes, we try to get
them resolved as soon as possible.
[ INFO: possible circular locking dependency detected ] 3.6.0-rc1-00042-gce894ca #367 Not tainted
-------------------------------------------------------
dd/8298 is trying to acquire lock:
(&type->s_umount_key#18){++++..}, at: [<c02277d4>] writeback_inodes_sb_if_idle+0x28/0x46
but task is already holding lock:
(&sb->s_type->i_mutex_key#8){+.+...}, at: [<c01ddcce>] generic_file_aio_write+0x5f/0xd3
which lock already depends on the new lock.
2 locks held by dd/8298:
#0: (sb_writers#2){.+.+.+}, at: [<c01ddcc5>] generic_file_aio_write+0x56/0xd3
#1: (&sb->s_type->i_mutex_key#8){+.+...}, at: [<c01ddcce>] generic_file_aio_write+0x5f/0xd3
ext4: speed up truncate/unlink by not using bforget() unless needed
Do not iterate over data blocks scanning for bh's to forget as they're
never exist. This improves time taken by unlink / truncate syscall.
Tested by continuously truncating file that is being written by dd.
Another test is rm -rf of linux tree while tar unpacks it. With
ordered data mode condition unlikely(!tbh) was always met in
ext4_free_blocks. With journal data mode tbh was found only few times,
so optimisation is also possible.
Unlinking fallocated 60G file after doing sync && echo 3 >
/proc/sys/vm/drop_caches && time rm --help
X86 before (linux 3.6-rc4):
# time rm -f test1
real 0m2.710s
user 0m0.000s
sys 0m1.530s
X86 after:
# time rm -f test1
real 0m0.644s
user 0m0.003s
sys 0m0.060s
MIPS before (linux 2.6.37):
# time rm -f test1
real 0m 4.93s
user 0m 0.00s
sys 0m 4.61s
MIPS after:
# time rm -f test1
real 0m 0.16s
user 0m 0.00s
sys 0m 0.06s
ext4: fix online resizing when the # of block groups is constant
Commit 1c6bd7173d66b3 introduced a regression where an online resize
operation which did not change the number of block groups would fail,
i.e:
mke2fs -t /dev/vdc 60000
mount /dev/vdc
resize2fs /dev/vdc 60001
This was due to a bug in the logic regarding when to try converting
the filesystem to use meta_bg.
Also fix up a number of other minor issues with the online resizing
code: (a) Fix a sparse warning; (b) only check to make sure the device
is large enough once, instead of multiple times through the resize
loop.
ext4: make orphan functions be no-op in no-journal mode
Instead of checking whether the handle is valid, we check if journal
is enabled. This avoids taking the s_orphan_lock mutex in all cases
when there is no journal in use, including the error paths where
ext4_orphan_del() is called with a handle set to NULL.
ext4: re-enable -o discard functionality in no-journal mode
This is a revert of commit b56ff9d397ce, which removed the call to
ext4_issue_discard() to fix a BUG reported because
ext4_issue_discard() was being called from inside a block group
spinlock. As it turns out this bug had already been fixed by Lukas
Czerner in commit 53fdcf992d61 by the simple expedient of moving when
we call ext4_issue_discard() outside the spinlock.
So it should be safe to re-enable this functionality, which I tested
by putting an BUG_ON(in_atomic) just after the restored callsite to
ext4_issue_discard().
Carlos Maiolino [Tue, 18 Sep 2012 03:39:12 +0000 (23:39 -0400)]
ext4: fix possible non-initialized variable in htree_dirblock_to_tree()
htree_dirblock_to_tree() declares a non-initialized 'err' variable,
which is passed as a reference to another functions expecting them to
set this variable with their error codes.
It's passed to ext4_bread(), which then passes it to ext4_getblk(). If
ext4_map_blocks() returns 0 due to a lookup failure, leaving the
ext4_getblk() buffer_head uninitialized, it will make ext4_getblk()
return to ext4_bread() without initialize the 'err' variable, and
ext4_bread() will return to htree_dirblock_to_tree() with this variable
still uninitialized. htree_dirblock_to_tree() will pass this variable
with garbage back to ext4_htree_fill_tree(), which expects a number of
directory entries added to the rb-tree. which, in case, might return a
fake non-zero value due the garbage left in the 'err' variable, leading
the kernel to an Oops in ext4_dx_readdir(), once this is expecting a
filled rb-tree node, when in turn it will have a NULL-ed one, causing an
invalid page request when trying to get a fname struct from this NULL-ed
rb-tree node in this line:
The patch itself initializes the err variable in
htree_dirblock_to_tree() to avoid usage mistakes by the called
functions, and also fix ext4_getblk() to return a initialized 'err'
variable when ext4_map_blocks() fails a lookup.
Signed-off-by: Carlos Maiolino <cmaiolino@redhat.com> Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
ext4: convert file system to meta_bg if needed during resizing
If we have run out of reserved gdt blocks, then clear the resize_inode
feature and enable the meta_bg feature, so that we can continue
resizing the file system seamlessly.
Yongqiang Yang [Wed, 5 Sep 2012 05:33:50 +0000 (01:33 -0400)]
ext4: add online resizing support for meta_bg and 64-bit file systems
This patch adds support for resizing file systems with the meta_bg and
64bit features.
[ Added a fix by tytso to fix a divide by zero when resizing a
filesystem from 14 TB to 18TB. Also fixed overhead accounting for
meta_bg file systems.]
Signed-off-by: Yongqiang Yang <xiaoqiangnk@gmail.com> Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
Previously we allocated the s_group_info array with enough space for
any future possible growth of the file system via online resize. This
is unfortunate because it wastes memory, and it doesn't work for the
meta_bg scheme, since there is no limit based on the number of
reserved gdt blocks. So add the code to grow the s_group_info array
as needed.
ext4: grow the s_flex_groups array as needed when resizing
Previously, we allocated the s_flex_groups array to the maximum size
that the file system could be resized. There was two problems with
this approach. First, it wasted memory in the common case where the
file system was not resized. Secondly, once we start allowing online
resizing using the meta_bg scheme, there is no maximum size that the
file system can be resized. So instead, we need to grow the
s_flex_groups at inline resize time.
Yongqiang Yang [Wed, 5 Sep 2012 05:25:50 +0000 (01:25 -0400)]
ext4: don't copy non-existent gdt blocks when resizing
The resize code was copying blocks at the beginning of each block
group in order to copy the superblock and block group descriptor table
(gdt) blocks. This was, unfortunately, being done even for block
groups that did not have super blocks or gdt blocks. This is a
complete waste of perfectly good I/O bandwidth, to skip writing those
blocks for sparse bg's.
Signed-off-by: Yongqiang Yang <xiaoqiangnk@gmail.com> Signed-off-by: "Theodore Ts'o" <tytso@mit.edu> Cc: stable@vger.kernel.org
Zheng Liu [Sun, 19 Aug 2012 22:07:40 +0000 (18:07 -0400)]
ext4: remove duplicated declarations in inode.c
In patch cb20d5188366f04d96d2e07b1240cc92170ade40, ext4_set_bh_endio
and ext4_end_io_buffer_write are declared at the beginning of inode.c,
and again later on in the middle of the file. Remove the second set
of duplicated function declarations.
Signed-off-by: Zheng Liu <wenqing.lz@taobao.com> Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
results in an IO error when unmounting the RO filesystem:
[ 318.020828] Buffer I/O error on device loop1, logical block 196608
[ 318.027024] lost page write due to I/O error on loop1
[ 318.032088] JBD2: Error -5 detected when updating journal superblock for loop1-8.
This was a regression introduced by commit 24bcc89c7e7c: "jbd2: split
updating of journal superblock and marking journal empty".
Signed-off-by: Eric Sandeen <sandeen@redhat.com> Signed-off-by: "Theodore Ts'o" <tytso@mit.edu> Cc: stable@vger.kernel.org
Theodore Ts'o [Fri, 17 Aug 2012 14:04:17 +0000 (10:04 -0400)]
ext4: return an error if kset_create_and_add fails in ext4_init_fs()
In the very unlikely case that kset_create_and_add() fails when the
ext4.ko module is being loaded (or during kernel startup) set err so
that it's clear that the module load failed.
Zheng Liu [Fri, 17 Aug 2012 13:54:17 +0000 (09:54 -0400)]
ext4: make the zero-out chunk size tunable
Currently in ext4 the length of zero-out chunk is set to 7 file system
blocks. But if an inode has uninitailized extents from using
fallocate to preallocate space, and the workload issues many random
writes, this can cause a fragmented extent tree that will
unnecessarily grow the extent tree.
So create a new sysfs tunable, extent_max_zeroout_kb, which controls
the maximum size where blocks will be zeroed out instead of creating a
new uninitialized extent. The default of this has been sent to 32kb.
CC: Zach Brown <zab@zabbo.net> CC: Andreas Dilger <adilger@dilger.ca> Signed-off-by: Zheng Liu <wenqing.lz@taobao.com> Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
Anatol Pomozov [Fri, 17 Aug 2012 13:50:17 +0000 (09:50 -0400)]
ext4: realign trace events structs to make it smaller
Most hardware architectures require that data (including struct fields)
have to be aligned in memory. To make it happen compiler inserts padding
between struct fields if they are not aligned correctly.
Reorder fields to remove paddings and make structures denser. Making data
smaller saves some memory that is very important for trace events.
Tracing buffer has limited size and making objects smaller we can put more
of them without overflowing the tracing buffer.
To find data struct holes I used 'pahole -H 1 -E -I vmlinux.o' from
'dwarves' package.
Theodore Ts'o [Fri, 17 Aug 2012 13:48:17 +0000 (09:48 -0400)]
ext4: add max_dir_size_kb mount option
Very large directories can cause significant performance problems, or
perhaps even invoke the OOM killer, if the process is running in a
highly constrained memory environment (whether it is VM's with a small
amount of memory or in a small memory cgroup).
So it is useful, in cloud server/data center environments, to be able
to set a filesystem-wide cap on the maximum size of a directory, to
ensure that directories never get larger than a sane size. We do this
via a new mount option, max_dir_size_kb. If there is an attempt to
grow the directory larger than max_dir_size_kb, the system call will
return ENOSPC instead.
Theodore Ts'o [Fri, 17 Aug 2012 13:46:17 +0000 (09:46 -0400)]
ext4: don't load the block bitmap for block groups which have no space
Add a short circuit check to ext4_mb_group_group() so that we don't
bother to load the block bitmap for a block group which does not have
any space available. (Or which does not have enough space until we
are in desperation mode, i.e., when cr == 3.)
Theodore Ts'o [Fri, 17 Aug 2012 13:44:17 +0000 (09:44 -0400)]
ext4: collapse a single extent tree block into the inode if possible
If an inode has more than 4 extents, but then later some of the
extents are merged together, we can optimize the file system by moving
the extents up into the inode, and discarding the extent tree block.
This is important, because if there are a large number of inodes with
an external extent tree blocks where the contents could fit in the
inode, this can significantly increase the fsck time of the file
system.
Theodore Ts'o [Fri, 17 Aug 2012 12:54:52 +0000 (08:54 -0400)]
ext4: fix kernel BUG on large-scale rm -rf commands
Commit 968dee7722: "ext4: fix hole punch failure when depth is greater
than 0" introduced a regression in v3.5.1/v3.6-rc1 which caused kernel
crashes when users ran run "rm -rf" on large directory hierarchy on
ext4 filesystems on RAID devices:
BUG: unable to handle kernel NULL pointer dereference at 0000000000000028
The problem in commit 968dee7722 was that caused the variable 'i' to
be left uninitialized if the truncate required more space than was
available in the journal. This resulted in the function
ext4_ext_truncate_extend_restart() returning -EAGAIN, which caused
ext4_ext_remove_space() to restart the truncate operation after
starting a new jbd2 handle.
Theodore Ts'o [Thu, 16 Aug 2012 15:59:04 +0000 (11:59 -0400)]
ext4: fix long mount times on very big file systems
Commit 8aeb00ff85a: "ext4: fix overhead calculation used by
ext4_statfs()" introduced a O(n**2) calculation which makes very large
file systems take forever to mount. Fix this with an optimization for
non-bigalloc file systems. (For bigalloc file systems the overhead
needs to be set in the the superblock.)
Theodore Ts'o [Fri, 10 Aug 2012 17:57:52 +0000 (13:57 -0400)]
ext4: don't call ext4_error while block group is locked
While in ext4_validate_block_bitmap(), if an block allocation bitmap
is found to be invalid, we call ext4_error() while the block group is
still locked. This causes ext4_commit_super() to call a function
which might sleep while in an atomic context.
There's no need to keep the block group locked at this point, so hoist
the ext4_error() call up to ext4_validate_block_bitmap() and release
the block group spinlock before calling ext4_error().
Theodore Ts'o [Mon, 6 Aug 2012 03:28:16 +0000 (23:28 -0400)]
ext4: avoid kmemcheck complaint from reading uninitialized memory
Commit 03179fe923 introduced a kmemcheck complaint in
ext4_da_get_block_prep() because we save and restore
ei->i_da_metadata_calc_last_lblock even though it is left
uninitialized in the case where i_da_metadata_calc_len is zero.
This doesn't hurt anything, but silencing the kmemcheck complaint
makes it easier for people to find real bugs.
Addresses https://bugzilla.kernel.org/show_bug.cgi?id=45631
(which is marked as a regression).
Theodore Ts'o [Sun, 5 Aug 2012 23:04:57 +0000 (19:04 -0400)]
ext4: make sure the journal sb is written in ext4_clear_journal_err()
After we transfer set the EXT4_ERROR_FS bit in the file system
superblock, it's not enough to call jbd2_journal_clear_err() to clear
the error indication from journal superblock --- we need to call
jbd2_journal_update_sb_errno() as well. Otherwise, when the root file
system is mounted read-only, the journal is replayed, and the error
indicator is transferred to the superblock --- but the s_errno field
in the jbd2 superblock is left set (since although we cleared it in
memory, we never flushed it out to disk).
This can end up confusing e2fsck. We should make e2fsck more robust
in this case, but the kernel shouldn't be leaving things in this
confused state, either.
Linus Torvalds [Thu, 2 Aug 2012 18:52:39 +0000 (11:52 -0700)]
Merge branch 'for-linus-3.6' of git://dev.laptop.org/users/dilinger/linux-olpc
Pull OLPC platform updates from Andres Salomon:
"These move the OLPC Embedded Controller driver out of
arch/x86/platform and into drivers/platform/olpc.
OLPC machines are now ARM-based (which means lots of x86 and ARM
changes), but are typically pretty self-contained.. so it makes more
sense to go through a separate OLPC tree after getting the appropriate
review/ACKs."
* 'for-linus-3.6' of git://dev.laptop.org/users/dilinger/linux-olpc:
x86: OLPC: move s/r-related EC cmds to EC driver
Platform: OLPC: move global variables into priv struct
Platform: OLPC: move debugfs support from x86 EC driver
x86: OLPC: switch over to using new EC driver on x86
Platform: OLPC: add a suspended flag to the EC driver
Platform: OLPC: turn EC driver into a platform_driver
Platform: OLPC: allow EC cmd to be overridden, and create a workqueue to call it
drivers: OLPC: update various drivers to include olpc-ec.h
Platform: OLPC: add a stub to drivers/platform/ for the OLPC EC driver
Linus Torvalds [Thu, 2 Aug 2012 18:50:24 +0000 (11:50 -0700)]
Merge tag 'dt2' of git://git.kernel.org/pub/scm/linux/kernel/git/arm/arm-soc
Pull arm-soc Marvell Orion device-tree updates from Olof Johansson:
"This contains a set of device-tree conversions for Marvell Orion
platforms that were staged early but took a few tries to get the
branch into a format where it was suitable for us to pick up.
Given that most people working on these platforms are hobbyists with
limited time, we were a bit more flexible with merging it even though
it came in late."
* tag 'dt2' of git://git.kernel.org/pub/scm/linux/kernel/git/arm/arm-soc: (21 commits)
ARM: Kirkwood: Replace mrvl with marvell
ARM: Kirkwood: Describe GoFlex Net LEDs and SATA in DT.
ARM: Kirkwood: Describe Dreamplug LEDs in DT.
ARM: Kirkwood: Describe iConnects LEDs in DT.
ARM: Kirkwood: Describe iConnects temperature sensor in DT.
ARM: Kirkwood: Describe IB62x0 LEDs in DT.
ARM: Kirkwood: Describe IB62x0 gpio-keys in DT.
ARM: Kirkwood: Describe DNS32? gpio-keys in DT.
ARM: Kirkwood: Move common portions into a kirkwood-dnskw.dtsi
ARM: Kirkwood: Replace DNS-320/DNS-325 leds with dt bindings
ARM: Kirkwood: Describe DNS325 temperature sensor in DT.
ARM: Kirkwood: Use DT to configure SATA device.
ARM: kirkwood: use devicetree for SPI on dreamplug
ARM: kirkwood: Add LS-XHL and LS-CHLv2 support
ARM: Kirkwood: Initial DTS support for Kirkwood GoFlex Net
ARM: Kirkwood: Add basic device tree support for QNAP TS219.
ATA: sata_mv: Add device tree support
ARM: Orion: DTify the watchdog timer.
ARM: Orion: Add arch support needed for I2C via DT.
ARM: kirkwood: use devicetree for orion-spi
...
Linus Torvalds [Thu, 2 Aug 2012 18:48:54 +0000 (11:48 -0700)]
Merge tag 'pm2' of git://git.kernel.org/pub/scm/linux/kernel/git/arm/arm-soc
Pull arm-soc cpuidle enablement for OMAP from Olof Johansson:
"Coupled cpuidle was meant to merge for 3.5 through Len Brown's tree,
but didn't go in because the pull request ended up rejected. So it
just got merged, and we got this staged branch that enables the
coupled cpuidle code on OMAP.
With a stable git workflow from the other maintainer we could have
staged this earlier, but that wasn't the case so we have had to merge
it late.
The alternative is to hold it off until 3.7 but given that the code is
well-isolated to OMAP and they are eager to see it go in, I didn't
push back hard in that direction."
* tag 'pm2' of git://git.kernel.org/pub/scm/linux/kernel/git/arm/arm-soc:
ARM: OMAP4: CPUidle: Open broadcast clock-event device.
ARM: OMAP4: CPUidle: add synchronization for coupled idle states
ARM: OMAP4: CPUidle: Use coupled cpuidle states to implement SMP cpuidle.
ARM: OMAP: timer: allow gp timer clock-event to be used on both cpus
Linus Torvalds [Thu, 2 Aug 2012 18:48:20 +0000 (11:48 -0700)]
Merge tag 'fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/arm/arm-soc
Pull ARM SoC fixes from Olof Johansson:
"A few fixes for merge window fallout, and a bugfix for timer resume on
PRIMA2."
* tag 'fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/arm/arm-soc:
ARM: mmp: add missing irqs.h
arm: mvebu: fix typo in .dtsi comment for Armada XP SoCs
ARM: PRIMA2: delete redundant codes to restore LATCHED when timer resumes
ARM: mxc: Include missing irqs.h header
Linus Torvalds [Thu, 2 Aug 2012 18:45:42 +0000 (11:45 -0700)]
Merge tag 'sh-for-linus' of git://github.com/pmundt/linux-sh
Pull SuperH fixes from Paul Mundt.
* tag 'sh-for-linus' of git://github.com/pmundt/linux-sh: (24 commits)
sh: explicitly include sh_dma.h in setup-sh7722.c
sh: ecovec: care CN5 VBUS if USB host mode
sh: sh7724: fixup renesas_usbhs clock settings
sh: intc: initial irqdomain support.
sh: pfc: Fix up init ordering mess.
serial: sh-sci: fix compilation breakage, when DMA is enabled
dmaengine: shdma: restore partial transfer calculation
sh: modify the sh_dmae_slave_config for RSPI in setup-sh7757
sh: Fix up recursive fault in oops with unset TTB.
sh: pfc: Build fix for pinctrl_remove_gpio_range() changes.
sh: select the fixed regulator driver on several boards
sh: ecovec: switch MMC power control to regulators
sh: add fixed voltage regulators to se7724
sh: add fixed voltage regulators to sdk7786
sh: add fixed voltage regulators to rsk
sh: add fixed voltage regulators to migor
sh: add fixed voltage regulators to kfr2r09
sh: add fixed voltage regulators to ap325rxa
sh: add fixed voltage regulators to sh7757lcr
sh: add fixed voltage regulators to sh2007
...
Linus Torvalds [Thu, 2 Aug 2012 18:34:40 +0000 (11:34 -0700)]
Merge tag 'md-3.6' of git://neil.brown.name/md
Pull additional md update from NeilBrown:
"This contains a few patches that depend on plugging changes in the
block layer so needed to wait for those.
It also contains a Kconfig fix for the new RAID10 support in dm-raid."
* tag 'md-3.6' of git://neil.brown.name/md:
md/dm-raid: DM_RAID should select MD_RAID10
md/raid1: submit IO from originating thread instead of md thread.
raid5: raid5d handle stripe in batch way
raid5: make_request use batch stripe release
Linus Torvalds [Thu, 2 Aug 2012 17:57:31 +0000 (10:57 -0700)]
Merge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/sage/ceph-client
Pull two ceph fixes from Sage Weil:
"The first patch fixes up the old crufty open intent code to use the
atomic_open stuff properly, and the second fixes a possible null deref
and memory leak with the crypto keys."
Linus Torvalds [Thu, 2 Aug 2012 17:56:34 +0000 (10:56 -0700)]
Merge tag 'ecryptfs-3.6-rc1-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/tyhicks/ecryptfs
Pull ecryptfs fixes from Tyler Hicks:
- Fixes a bug when the lower filesystem mount options include 'acl',
but the eCryptfs mount options do not
- Cleanups in the messaging code
- Better handling of empty files in the lower filesystem to improve
usability. Failed file creations are now cleaned up and empty lower
files are converted into eCryptfs during open().
- The write-through cache changes are being reverted due to bugs that
are not easy to fix. Stability outweighs the performance
enhancements here.
- Improvement to the mount code to catch unsupported ciphers specified
in the mount options
* tag 'ecryptfs-3.6-rc1-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/tyhicks/ecryptfs:
eCryptfs: check for eCryptfs cipher support at mount
eCryptfs: Revert to a writethrough cache model
eCryptfs: Initialize empty lower files when opening them
eCryptfs: Unlink lower inode when ecryptfs_create() fails
eCryptfs: Make all miscdev functions use daemon ptr in file private_data
eCryptfs: Remove unused messaging declarations and function
eCryptfs: Copy up POSIX ACL and read-only flags from lower mount
Linus Torvalds [Thu, 2 Aug 2012 17:54:11 +0000 (10:54 -0700)]
Merge branch 'for-next' of git://git.samba.org/sfrench/cifs-2.6
Pull CIFS update from Steve French:
"Adds SMB2 rmdir/mkdir capability to the SMB2/SMB2.1 support in cifs.
I am holding up a few more days on merging the remainder of the
SMB2/SMB2.1 enablement although it is nearing review completion, in
order to address some review comments from Jeff Layton on a few of the
subsequent SMB2 patches, and also to debug an unrelated cifs problem
that Pavel discovered."
* 'for-next' of git://git.samba.org/sfrench/cifs-2.6:
CIFS: Add SMB2 support for rmdir
CIFS: Move rmdir code to ops struct
CIFS: Add SMB2 support for mkdir operation
CIFS: Separate protocol specific part from mkdir
CIFS: Simplify cifs_mkdir call
Linus Torvalds [Thu, 2 Aug 2012 17:37:03 +0000 (10:37 -0700)]
mm: remove node_start_pfn checking in new WARN_ON for now
Borislav Petkov reports that the new warning added in commit 88fdf75d1bb5 ("mm: warn if pg_data_t isn't initialized with zero")
triggers for him, and it is the node_start_pfn field that has already
been initialized once.
and (with the warning replaced by debug output), Borislav sees
On node 0 totalpages: 4193848
DMA zone: 64 pages used for memmap
DMA zone: 6 pages reserved
DMA zone: 3890 pages, LIFO batch:0
DMA32 zone: 16320 pages used for memmap
DMA32 zone: 798464 pages, LIFO batch:31
Normal zone: 52736 pages used for memmap
Normal zone: 3322368 pages, LIFO batch:31
free_area_init_node: pgdat->node_start_pfn: 4423680 <----
On node 1 totalpages: 4194304
Normal zone: 65536 pages used for memmap
Normal zone: 4128768 pages, LIFO batch:31
free_area_init_node: pgdat->node_start_pfn: 8617984 <----
On node 2 totalpages: 4194304
Normal zone: 65536 pages used for memmap
Normal zone: 4128768 pages, LIFO batch:31
free_area_init_node: pgdat->node_start_pfn: 12812288 <----
On node 3 totalpages: 4194304
Normal zone: 65536 pages used for memmap
Normal zone: 4128768 pages, LIFO batch:31
so remove the bogus warning for now to avoid annoying people. Minchan
Kim is looking at it.
Reported-by: Borislav Petkov <bp@amd64.org> Cc: Minchan Kim <minchan@kernel.org> Cc: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Barry Song [Mon, 30 Jul 2012 05:29:30 +0000 (13:29 +0800)]
ARM: PRIMA2: delete redundant codes to restore LATCHED when timer resumes
The only way to write LATCHED registers to write LATCH_BIT to LATCH register,
that will latch COUNTER into LATCHED.e.g.
writel_relaxed(SIRFSOC_TIMER_LATCH_BIT, sirfsoc_timer_base +
SIRFSOC_TIMER_LATCH);
Writing values to LATCHED registers directly is useless at all.
Signed-off-by: Barry Song <Baohua.Song@csr.com> Signed-off-by: Olof Johansson <olof@lixom.net>
Sage Weil [Tue, 31 Jul 2012 18:27:36 +0000 (11:27 -0700)]
ceph: simplify+fix atomic_open
The initial ->atomic_open op was carried over from the old intent code,
which was incomplete and didn't really work. Replace it with a fresh
method. In particular:
* always attempt to do an atomic open+lookup, both for the create case
and for lookups of existing files.
* fix symlink handling by returning 1 to the VFS so that we can follow
the link to its destination. This fixes a longstanding ceph bug (#2392).
Linus Torvalds [Wed, 1 Aug 2012 23:47:15 +0000 (16:47 -0700)]
Merge branch 'upstream' of git://git.linux-mips.org/pub/scm/ralf/upstream-linus
Pull MIPS updates from Ralf Baechle:
"The lion share of this pull request are fixes for clk-related breakage
caused by other changes during this merge window. For some platforms
the fix was as simple as selecting HAVE_CLK, for others like the
Loongson 2 significant restructuring was required.
The remainder are changes required to get the Lantiq code to work
again."
* 'upstream' of git://git.linux-mips.org/pub/scm/ralf/upstream-linus:
MIPS: Loongson 2: Sort out clock managment.
MIPS: Loongson 1: more clk support and add select HAVE_CLK
MIPS: txx9: Fix redefinition of clk_* by adding select HAVE_CLK
MIPS: BCM63xx: Fix redefinition of clk_* by adding select HAVE_CLK
MIPS: AR7: Fix redefinition of clk_* by adding select HAVE_CLK
MIPS: Lantiq: Platform specific CLK fixup
MIPS: Lantiq: Add device_tree_init function
MIPS: Lantiq: Fix interface clock and PCI control register offset
Linus Torvalds [Wed, 1 Aug 2012 23:45:02 +0000 (16:45 -0700)]
Merge branch 'for-linus-3.6-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/rw/uml
Pull UML fixes from Richard Weinberger:
"This patch set contains mostly fixes and cleanups. The UML tty driver
uses now tty_port and is no longer broken like hell :-)"
* 'for-linus-3.6-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/rw/uml:
um: Add arch/x86/um to MAINTAINERS
um: pass siginfo to guest process
um: fix ubd_file_size for read-only files
um: pull interrupt_end() into userspace()
um: split syscall_trace(), pass pt_regs to it
um: switch UPT_SET_RETURN_VALUE and regs_return_value to pt_regs
um: set BLK_CGROUP=y in defconfig
um: remove count_lock
um: fully use tty_port
um: Remove dead code
um: remove line_ioctl()
TTY: um/line, use tty from tty_port
TTY: um/line, add tty_port
Linus Torvalds [Wed, 1 Aug 2012 23:41:07 +0000 (16:41 -0700)]
Merge branch 'dmaengine' of git://git.linaro.org/people/rmk/linux-arm
Pull ARM DMA engine updates from Russell King:
"This looks scary at first glance, but what it is is:
- a rework of the sa11x0 DMA engine driver merged during the previous
cycle, to extract a common set of helper functions for DMA engine
implementations.
- conversion of amba-pl08x.c to use these helper functions.
- addition of OMAP DMA engine driver (using these helper functions),
and conversion of some of the OMAP DMA users to use DMA engine.
Nothing in the helper functions is ARM specific, so I hope that other
implementations can consolidate some of their code by making use of
these helpers.
This has been sitting in linux-next most of the merge cycle, and has
been tested by several OMAP folk. I've tested it on sa11x0 platforms,
and given it my best shot on my broken platforms which have the
amba-pl08x controller.
The last point is the addition to feature-removal-schedule.txt, which
will have a merge conflict. Between myself and TI, we're planning to
remove the old TI DMA implementation next year."
Fix up trivial add/add conflicts in Documentation/feature-removal-schedule.txt
and drivers/dma/{Kconfig,Makefile}
* 'dmaengine' of git://git.linaro.org/people/rmk/linux-arm: (53 commits)
ARM: 7481/1: OMAP2+: omap2plus_defconfig: enable OMAP DMA engine
ARM: 7464/1: mmc: omap_hsmmc: ensure probe returns error if DMA channel request fails
Add feature removal of old OMAP private DMA implementation
mtd: omap2: remove private DMA API implementation
mtd: omap2: add DMA engine support
spi: omap2-mcspi: remove private DMA API implementation
spi: omap2-mcspi: add DMA engine support
ARM: omap: remove mmc platform data dma_mask and initialization
mmc: omap: remove private DMA API implementation
mmc: omap: add DMA engine support
mmc: omap_hsmmc: remove private DMA API implementation
mmc: omap_hsmmc: add DMA engine support
dmaengine: omap: add support for cyclic DMA
dmaengine: omap: add support for setting fi
dmaengine: omap: add support for returning residue in tx_state method
dmaengine: add OMAP DMA engine driver
dmaengine: sa11x0-dma: add cyclic DMA support
dmaengine: sa11x0-dma: fix DMA residue support
dmaengine: PL08x: ensure all descriptors are freed when channel is released
dmaengine: PL08x: get rid of write only pool_ctr and free_txd locking
...
Linus Torvalds [Wed, 1 Aug 2012 23:35:37 +0000 (16:35 -0700)]
Merge branch 'audit' of git://git.linaro.org/people/rmk/linux-arm
Pull ARM audit/signal updates from Russell King:
"ARM audit/signal handling updates from Al and Will. This improves on
the work Viro did last merge window, and sorts out some of the issues
found with that work."
* 'audit' of git://git.linaro.org/people/rmk/linux-arm:
ARM: 7475/1: sys_trace: allow all syscall arguments to be updated via ptrace
ARM: 7474/1: get rid of TIF_SYSCALL_RESTARTSYS
ARM: 7473/1: deal with handlerless restarts without leaving the kernel
ARM: 7472/1: pull all work_pending logics into C function
ARM: 7471/1: Revert "7442/1: Revert "remove unused restart trampoline""
ARM: 7470/1: Revert "7443/1: Revert "new way of handling ERESTART_RESTARTBLOCK""
Linus Torvalds [Wed, 1 Aug 2012 23:30:45 +0000 (16:30 -0700)]
Merge branch 'fixes' of git://git.linaro.org/people/rmk/linux-arm
Pull ARM fixes from Russell King:
"This fixes various issues found during July"
* 'fixes' of git://git.linaro.org/people/rmk/linux-arm:
ARM: 7479/1: mm: avoid NULL dereference when flushing gate_vma with VIVT caches
ARM: Fix undefined instruction exception handling
ARM: 7480/1: only call smp_send_stop() on SMP
ARM: 7478/1: errata: extend workaround for erratum #720789
ARM: 7477/1: vfp: Always save VFP state in vfp_pm_suspend on UP
ARM: 7476/1: vfp: only clear vfp state for current cpu in vfp_pm_suspend
ARM: 7468/1: ftrace: Trace function entry before updating index
ARM: 7467/1: mutex: use generic xchg-based implementation for ARMv6+
ARM: 7466/1: disable interrupt before spinning endlessly
ARM: 7465/1: Handle >4GB memory sizes in device tree and mem=size@start option
Martin Pärtel [Wed, 1 Aug 2012 22:49:17 +0000 (00:49 +0200)]
um: pass siginfo to guest process
UML guest processes now get correct siginfo_t for SIGTRAP, SIGFPE,
SIGILL and SIGBUS. Specifically, si_addr and si_code are now correct
where previously they were si_addr = NULL and si_code = 128.
Signed-off-by: Martin Pärtel <martin.partel@gmail.com> Signed-off-by: Richard Weinberger <richard@nod.at>
NeilBrown [Wed, 1 Aug 2012 22:33:20 +0000 (08:33 +1000)]
md/raid1: submit IO from originating thread instead of md thread.
queuing writes to the md thread means that all requests go through the
one processor which may not be able to keep up with very high request
rates.
So use the plugging infrastructure to submit all requests on unplug.
If a 'schedule' is needed, we fall back on the old approach of handing
the requests to the thread for it to handle.
Shaohua Li [Wed, 1 Aug 2012 22:33:00 +0000 (08:33 +1000)]
raid5: make_request use batch stripe release
make_request() does stripe release for every stripe and the stripe usually has
count 1, which makes previous release_stripe() optimization not work. In my
test, this release_stripe() becomes the heaviest pleace to take
conf->device_lock after previous patches applied.
Below patch makes stripe release batch. All the stripes will be released in
unplug. The STRIPE_ON_UNPLUG_LIST bit is to protect concurrent access stripe
lru.
Signed-off-by: Shaohua Li <shli@fusionio.com> Signed-off-by: NeilBrown <neilb@suse.de>
projects / linux-beck.git / log