Btrfs: fix hang on extent buffer lock caused by the inode_paths ioctl

2016-02-25T19:57:46+00:00

commit 0c0fe3b0fa45082cd752553fdb3a4b42503a118e upstream.

While doing some tests I ran into an hang on an extent buffer's rwlock
that produced the following trace:

[39389.800012] NMI watchdog: BUG: soft lockup - CPU#15 stuck for 22s! [fdm-stress:32166]
[39389.800016] NMI watchdog: BUG: soft lockup - CPU#14 stuck for 22s! [fdm-stress:32165]
[39389.800016] Modules linked in: btrfs dm_mod ppdev xor sha256_generic hmac raid6_pq drbg ansi_cprng aesni_intel i2c_piix4 acpi_cpufreq aes_x86_64 ablk_helper tpm_tis parport_pc i2c_core sg cryptd evdev psmouse lrw tpm parport gf128mul serio_raw pcspkr glue_helper processor button loop autofs4 ext4 crc16 mbcache jbd2 sd_mod sr_mod cdrom ata_generic virtio_scsi ata_piix libata virtio_pci virtio_ring crc32c_intel scsi_mod e1000 virtio floppy [last unloaded: btrfs]
[39389.800016] irq event stamp: 0
[39389.800016] hardirqs last  enabled at (0): [<          (null)>]           (null)
[39389.800016] hardirqs last disabled at (0): [] copy_process+0x638/0x1a35
[39389.800016] softirqs last  enabled at (0): [] copy_process+0x638/0x1a35
[39389.800016] softirqs last disabled at (0): [<          (null)>]           (null)
[39389.800016] CPU: 14 PID: 32165 Comm: fdm-stress Not tainted 4.4.0-rc6-btrfs-next-18+ #1
[39389.800016] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS by qemu-project.org 04/01/2014
[39389.800016] task: ffff880175b1ca40 ti: ffff8800a185c000 task.ti: ffff8800a185c000
[39389.800016] RIP: 0010:[]  [] queued_spin_lock_slowpath+0x57/0x158
[39389.800016] RSP: 0018:ffff8800a185fb80  EFLAGS: 00000202
[39389.800016] RAX: 0000000000000101 RBX: ffff8801710c4e9c RCX: 0000000000000101
[39389.800016] RDX: 0000000000000100 RSI: 0000000000000001 RDI: 0000000000000001
[39389.800016] RBP: ffff8800a185fb98 R08: 0000000000000001 R09: 0000000000000000
[39389.800016] R10: ffff8800a185fb68 R11: 6db6db6db6db6db7 R12: ffff8801710c4e98
[39389.800016] R13: ffff880175b1ca40 R14: ffff8800a185fc10 R15: ffff880175b1ca40
[39389.800016] FS:  00007f6d37fff700(0000) GS:ffff8802be9c0000(0000) knlGS:0000000000000000
[39389.800016] CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[39389.800016] CR2: 00007f6d300019b8 CR3: 0000000037c93000 CR4: 00000000001406e0
[39389.800016] Stack:
[39389.800016]  ffff8801710c4e98 ffff8801710c4e98 ffff880175b1ca40 ffff8800a185fbb0
[39389.800016]  ffffffff81091e11 ffff8801710c4e98 ffff8800a185fbc8 ffffffff81091895
[39389.800016]  ffff8801710c4e98 ffff8800a185fbe8 ffffffff81486c5c ffffffffa067288c
[39389.800016] Call Trace:
[39389.800016]  [] queued_read_lock_slowpath+0x46/0x60
[39389.800016]  [] do_raw_read_lock+0x3e/0x41
[39389.800016]  [] _raw_read_lock+0x3d/0x44
[39389.800016]  [] ? btrfs_tree_read_lock+0x54/0x125 [btrfs]
[39389.800016]  [] btrfs_tree_read_lock+0x54/0x125 [btrfs]
[39389.800016]  [] ? btrfs_find_item+0xa7/0xd2 [btrfs]
[39389.800016]  [] btrfs_ref_to_path+0xd6/0x174 [btrfs]
[39389.800016]  [] inode_to_path+0x53/0xa2 [btrfs]
[39389.800016]  [] paths_from_inode+0x117/0x2ec [btrfs]
[39389.800016]  [] btrfs_ioctl+0xd5b/0x2793 [btrfs]
[39389.800016]  [] ? arch_local_irq_save+0x9/0xc
[39389.800016]  [] ? __this_cpu_preempt_check+0x13/0x15
[39389.800016]  [] ? arch_local_irq_save+0x9/0xc
[39389.800016]  [] ? rcu_read_unlock+0x3e/0x5d
[39389.800016]  [] do_vfs_ioctl+0x42b/0x4ea
[39389.800016]  [] ? __fget_light+0x62/0x71
[39389.800016]  [] SyS_ioctl+0x57/0x79
[39389.800016]  [] entry_SYSCALL_64_fastpath+0x12/0x6f
[39389.800016] Code: b9 01 01 00 00 f7 c6 00 ff ff ff 75 32 83 fe 01 89 ca 89 f0 0f 45 d7 f0 0f b1 13 39 f0 74 04 89 c6 eb e2 ff ca 0f 84 fa 00 00 00 <8b> 03 84 c0 74 04 f3 90 eb f6 66 c7 03 01 00 e9 e6 00 00 00 e8
[39389.800012] Modules linked in: btrfs dm_mod ppdev xor sha256_generic hmac raid6_pq drbg ansi_cprng aesni_intel i2c_piix4 acpi_cpufreq aes_x86_64 ablk_helper tpm_tis parport_pc i2c_core sg cryptd evdev psmouse lrw tpm parport gf128mul serio_raw pcspkr glue_helper processor button loop autofs4 ext4 crc16 mbcache jbd2 sd_mod sr_mod cdrom ata_generic virtio_scsi ata_piix libata virtio_pci virtio_ring crc32c_intel scsi_mod e1000 virtio floppy [last unloaded: btrfs]
[39389.800012] irq event stamp: 0
[39389.800012] hardirqs last  enabled at (0): [<          (null)>]           (null)
[39389.800012] hardirqs last disabled at (0): [] copy_process+0x638/0x1a35
[39389.800012] softirqs last  enabled at (0): [] copy_process+0x638/0x1a35
[39389.800012] softirqs last disabled at (0): [<          (null)>]           (null)
[39389.800012] CPU: 15 PID: 32166 Comm: fdm-stress Tainted: G             L  4.4.0-rc6-btrfs-next-18+ #1
[39389.800012] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS by qemu-project.org 04/01/2014
[39389.800012] task: ffff880179294380 ti: ffff880034a60000 task.ti: ffff880034a60000
[39389.800012] RIP: 0010:[]  [] queued_write_lock_slowpath+0x62/0x72
[39389.800012] RSP: 0018:ffff880034a639f0  EFLAGS: 00000206
[39389.800012] RAX: 0000000000000101 RBX: ffff8801710c4e98 RCX: 0000000000000000
[39389.800012] RDX: 00000000000000ff RSI: 0000000000000000 RDI: ffff8801710c4e9c
[39389.800012] RBP: ffff880034a639f8 R08: 0000000000000001 R09: 0000000000000000
[39389.800012] R10: ffff880034a639b0 R11: 0000000000001000 R12: ffff8801710c4e98
[39389.800012] R13: 0000000000000001 R14: ffff880172cbc000 R15: ffff8801710c4e00
[39389.800012] FS:  00007f6d377fe700(0000) GS:ffff8802be9e0000(0000) knlGS:0000000000000000
[39389.800012] CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[39389.800012] CR2: 00007f6d3d3c1000 CR3: 0000000037c93000 CR4: 00000000001406e0
[39389.800012] Stack:
[39389.800012]  ffff8801710c4e98 ffff880034a63a10 ffffffff81091963 ffff8801710c4e98
[39389.800012]  ffff880034a63a30 ffffffff81486f1b ffffffffa0672cb3 ffff8801710c4e00
[39389.800012]  ffff880034a63a78 ffffffffa0672cb3 ffff8801710c4e00 ffff880034a63a58
[39389.800012] Call Trace:
[39389.800012]  [] do_raw_write_lock+0x72/0x8c
[39389.800012]  [] _raw_write_lock+0x3a/0x41
[39389.800012]  [] ? btrfs_tree_lock+0x119/0x251 [btrfs]
[39389.800012]  [] btrfs_tree_lock+0x119/0x251 [btrfs]
[39389.800012]  [] ? rcu_read_unlock+0x5b/0x5d [btrfs]
[39389.800012]  [] ? btrfs_root_node+0xda/0xe6 [btrfs]
[39389.800012]  [] btrfs_lock_root_node+0x22/0x42 [btrfs]
[39389.800012]  [] btrfs_search_slot+0x1b8/0x758 [btrfs]
[39389.800012]  [] ? time_hardirqs_on+0x15/0x28
[39389.800012]  [] btrfs_lookup_inode+0x31/0x95 [btrfs]
[39389.800012]  [] ? trace_hardirqs_on+0xd/0xf
[39389.800012]  [] ? mutex_lock_nested+0x397/0x3bc
[39389.800012]  [] __btrfs_update_delayed_inode+0x59/0x1c0 [btrfs]
[39389.800012]  [] __btrfs_commit_inode_delayed_items+0x194/0x5aa [btrfs]
[39389.800012]  [] ? _raw_spin_unlock+0x31/0x44
[39389.800012]  [] __btrfs_run_delayed_items+0xa4/0x15c [btrfs]
[39389.800012]  [] btrfs_run_delayed_items+0x11/0x13 [btrfs]
[39389.800012]  [] btrfs_commit_transaction+0x234/0x96e [btrfs]
[39389.800012]  [] btrfs_sync_fs+0x145/0x1ad [btrfs]
[39389.800012]  [] btrfs_ioctl+0x11d2/0x2793 [btrfs]
[39389.800012]  [] ? arch_local_irq_save+0x9/0xc
[39389.800012]  [] ? __might_fault+0x4c/0xa7
[39389.800012]  [] ? __might_fault+0x4c/0xa7
[39389.800012]  [] ? arch_local_irq_save+0x9/0xc
[39389.800012]  [] ? rcu_read_unlock+0x3e/0x5d
[39389.800012]  [] do_vfs_ioctl+0x42b/0x4ea
[39389.800012]  [] ? __fget_light+0x62/0x71
[39389.800012]  [] SyS_ioctl+0x57/0x79
[39389.800012]  [] entry_SYSCALL_64_fastpath+0x12/0x6f
[39389.800012] Code: f0 0f b1 13 85 c0 75 ef eb 2a f3 90 8a 03 84 c0 75 f8 f0 0f b0 13 84 c0 75 f0 ba ff 00 00 00 eb 0a f0 0f b1 13 ff c8 74 0b f3 90 <8b> 03 83 f8 01 75 f7 eb ed c6 43 04 00 5b 5d c3 0f 1f 44 00 00

This happens because in the code path executed by the inode_paths ioctl we
end up nesting two calls to read lock a leaf's rwlock when after the first
call to read_lock() and before the second call to read_lock(), another
task (running the delayed items as part of a transaction commit) has
already called write_lock() against the leaf's rwlock. This situation is
illustrated by the following diagram:

         Task A                       Task B

  btrfs_ref_to_path()               btrfs_commit_transaction()
    read_lock(&eb->lock);

                                      btrfs_run_delayed_items()
                                        __btrfs_commit_inode_delayed_items()
                                          __btrfs_update_delayed_inode()
                                            btrfs_lookup_inode()

                                              write_lock(&eb->lock);
                                                --> task waits for lock

    read_lock(&eb->lock);
    --> makes this task hang
        forever (and task B too
	of course)

So fix this by avoiding doing the nested read lock, which is easily
avoidable. This issue does not happen if task B calls write_lock() after
task A does the second call to read_lock(), however there does not seem
to exist anything in the documentation that mentions what is the expected
behaviour for recursive locking of rwlocks (leaving the idea that doing
so is not a good usage of rwlocks).

Also, as a side effect necessary for this fix, make sure we do not
needlessly read lock extent buffers when the input path has skip_locking
set (used when called from send).

Signed-off-by: Filipe Manana 
Signed-off-by: Greg Kroah-Hartman

Btrfs: fix scrub_print_warning to handle skinny metadata extents

2014-07-01T03:09:46+00:00

commit 6eda71d0c030af0fc2f68aaa676e6d445600855b upstream.

The skinny extents are intepreted incorrectly in scrub_print_warning(),
and end up hitting the BUG() in btrfs_extent_inline_ref_size.

Reported-by: Konstantinos Skarlatos 
Signed-off-by: Liu Bo 
Signed-off-by: Chris Mason 
Signed-off-by: Greg Kroah-Hartman

Btrfs: use right type to get real comparison

2014-07-01T03:09:46+00:00

commit cd857dd6bc2ae9ecea14e75a34e8a8fdc158e307 upstream.

We want to make sure the point is still within the extent item, not to verify
the memory it's pointing to.

Signed-off-by: Liu Bo 
Signed-off-by: Chris Mason 
Signed-off-by: Greg Kroah-Hartman

Btrfs: fix possible memory leak in the find_parent_nodes()

2013-05-18T01:40:17+00:00

In the find_parent_nodes(), if read_tree_block() fails, we can
not return directly, we should free some allocated memory otherwise
memory leak happens.

Signed-off-by: Wang Shilong 
Signed-off-by: Josef Bacik

btrfs: make static code static & remove dead code

2013-05-06T19:55:23+00:00

Big patch, but all it does is add statics to functions which
are in fact static, then remove the associated dead-code fallout.

removed functions:

btrfs_iref_to_path()
__btrfs_lookup_delayed_deletion_item()
__btrfs_search_delayed_insertion_item()
__btrfs_search_delayed_deletion_item()
find_eb_for_page()
btrfs_find_block_group()
range_straddles_pages()
extent_range_uptodate()
btrfs_file_extent_length()
btrfs_scrub_cancel_devid()
btrfs_start_transaction_lflush()

btrfs_print_tree() is left because it is used for debugging.
btrfs_start_transaction_lflush() and btrfs_reada_detach() are
left for symmetry.

ulist.c functions are left, another patch will take care of those.

Signed-off-by: Eric Sandeen 
Signed-off-by: Josef Bacik

Btrfs: fix all callers of read_tree_block

2013-05-06T19:55:07+00:00

We kept leaking extent buffers when mounting a broken file system and it turns
out it's because not everybody uses read_tree_block properly.  You need to check
and make sure the extent_buffer is uptodate before you use it.  This patch fixes
everybody who calls read_tree_block directly to make sure they check that it is
uptodate and free it and return an error if it is not.  With this we no longer
leak EB's when things go horribly wrong.  Thanks,

Signed-off-by: Josef Bacik

Btrfs: remove unused variable in the iterate_extent_inodes()

2013-05-06T19:54:56+00:00

Signed-off-by: Wang Shilong 
Signed-off-by: Josef Bacik

Btrfs: do not continue if out of memory happens

2013-05-06T19:54:50+00:00

If out of memory happens, we should return -ENOMEM directly to the caller
rather than continue the work.

Signed-off-by: Wang Shilong 
Signed-off-by: Josef Bacik

Btrfs: make __merge_refs() return type be void

2013-05-06T19:54:43+00:00

__merge_refs() always return 0, it is unnecessary
for the caller to check the return value.

Signed-off-by: Wang Shilong 
Signed-off-by: Josef Bacik

Btrfs: remove some BUG_ONs() when walking backref tree

2013-05-06T19:54:42+00:00

The only error return value of __add_prelim_ref() is -ENOMEM,
just return errors rather than trigger BUG_ON().

Signed-off-by: Wang Shilong 
Reviewed-by: David Sterba 
Signed-off-by: Josef Bacik

linux-stable.git/fs/btrfs/backref.c, branch linux-3.10.y