We recast the replace return status
BTRFS_IOCTL_DEV_REPLACE_RESULT_SCRUB_INPROGRESS to 0, to indicate no
error.
And since BTRFS_IOCTL_DEV_REPLACE_RESULT_NO_ERROR should also return 0,
which is also declared as 0, so we just return. Instead add it to the if
statement so that there is enough clarity while reading the code.
Signed-off-by: Anand Jain <anand.jain@oracle.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
When the replace state is in the suspended state, btrfs_scrub_cancel()
should fail with -ENOTCONN as there is no scrub running. As a safety
catch check if btrfs_scrub_cancel() returns -ENOTCONN and assert if it
doesn't.
Signed-off-by: Anand Jain <anand.jain@oracle.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
The device-replace needs to check the result code of the scrub workers
in btrfs_dev_replace_cancel and distinguish if successful cancel
operation and when the there was no operation running.
If btrfs_scrub_cancel() fails, return
BTRFS_IOCTL_DEV_REPLACE_RESULT_NOT_STARTED so that user can try
to cancel the replace again.
Signed-off-by: Anand Jain <anand.jain@oracle.com>
Reviewed-by: David Sterba <dsterba@suse.com>
[ update changelog ]
Signed-off-by: David Sterba <dsterba@suse.com>
The device replace cancel thread can race with the replace start thread
and if fs_info::scrubs_running is not yet set, btrfs_scrub_cancel() will
fail to stop the scrub thread.
The scrub thread continues with the scrub for replace which then will
try to write to the target device and which is already freed by the
cancel thread.
scrub_setup_ctx() warns as tgtdev is NULL.
struct scrub_ctx *scrub_setup_ctx(struct btrfs_device *dev, int is_dev_replace)
{
...
if (is_dev_replace) {
WARN_ON(!fs_info->dev_replace.tgtdev); <===
sctx->pages_per_wr_bio = SCRUB_PAGES_PER_WR_BIO;
sctx->wr_tgtdev = fs_info->dev_replace.tgtdev;
sctx->flush_all_writes = false;
}
[ 6724.497655] BTRFS info (device sdb): dev_replace from /dev/sdb (devid 1) to /dev/sdc started
[ 6753.945017] BTRFS info (device sdb): dev_replace from /dev/sdb (devid 1) to /dev/sdc canceled
[ 6852.426700] WARNING: CPU: 0 PID: 4494 at fs/btrfs/scrub.c:622 scrub_setup_ctx.isra.19+0x220/0x230 [btrfs]
...
[ 6852.428928] RIP: 0010:scrub_setup_ctx.isra.19+0x220/0x230 [btrfs]
...
[ 6852.432970] Call Trace:
[ 6852.433202] btrfs_scrub_dev+0x19b/0x5c0 [btrfs]
[ 6852.433471] btrfs_dev_replace_start+0x48c/0x6a0 [btrfs]
[ 6852.433800] btrfs_dev_replace_by_ioctl+0x3a/0x60 [btrfs]
[ 6852.434097] btrfs_ioctl+0x2476/0x2d20 [btrfs]
[ 6852.434365] ? do_sigaction+0x7d/0x1e0
[ 6852.434623] do_vfs_ioctl+0xa9/0x6c0
[ 6852.434865] ? syscall_trace_enter+0x1c8/0x310
[ 6852.435124] ? syscall_trace_enter+0x1c8/0x310
[ 6852.435387] ksys_ioctl+0x60/0x90
[ 6852.435663] __x64_sys_ioctl+0x16/0x20
[ 6852.435907] do_syscall_64+0x50/0x180
[ 6852.436150] entry_SYSCALL_64_after_hwframe+0x49/0xbe
Further, as the replace thread enters scrub_write_page_to_dev_replace()
without the target device it panics:
static int scrub_add_page_to_wr_bio(struct scrub_ctx *sctx,
struct scrub_page *spage)
{
...
bio_set_dev(bio, sbio->dev->bdev); <======
[ 6929.715145] BUG: unable to handle kernel NULL pointer dereference at 00000000000000a0
..
[ 6929.717106] Workqueue: btrfs-scrub btrfs_scrub_helper [btrfs]
[ 6929.717420] RIP: 0010:scrub_write_page_to_dev_replace+0xb4/0x260
[btrfs]
..
[ 6929.721430] Call Trace:
[ 6929.721663] scrub_write_block_to_dev_replace+0x3f/0x60 [btrfs]
[ 6929.721975] scrub_bio_end_io_worker+0x1af/0x490 [btrfs]
[ 6929.722277] normal_work_helper+0xf0/0x4c0 [btrfs]
[ 6929.722552] process_one_work+0x1f4/0x520
[ 6929.722805] ? process_one_work+0x16e/0x520
[ 6929.723063] worker_thread+0x46/0x3d0
[ 6929.723313] kthread+0xf8/0x130
[ 6929.723544] ? process_one_work+0x520/0x520
[ 6929.723800] ? kthread_delayed_work_timer_fn+0x80/0x80
[ 6929.724081] ret_from_fork+0x3a/0x50
Fix this by letting the btrfs_dev_replace_finishing() to do the job of
cleaning after the cancel, including freeing of the target device.
btrfs_dev_replace_finishing() is called when btrfs_scub_dev() returns
along with the scrub return status.
Signed-off-by: Anand Jain <anand.jain@oracle.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
In a secnario where balance and replace co-exists as below,
- start balance
- pause balance
- start replace
- reboot
and when system restarts, balance resumes first. Then the replace is
attempted to restart but will fail as the EXCL_OP lock is already held
by the balance. If so place the replace state back to
BTRFS_IOCTL_DEV_REPLACE_STATE_SUSPENDED state.
Fixes: 010a47bde9 ("btrfs: add proper safety check before resuming dev-replace")
CC: stable@vger.kernel.org # 4.18+
Signed-off-by: Anand Jain <anand.jain@oracle.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
At the time of forced unmount we place the running replace to
BTRFS_IOCTL_DEV_REPLACE_STATE_SUSPENDED state, so when the system comes
back and expect the target device is missing.
Then let the replace state continue to be in
BTRFS_IOCTL_DEV_REPLACE_STATE_SUSPENDED state instead of
BTRFS_IOCTL_DEV_REPLACE_STATE_STARTED as there isn't any matching scrub
running as part of replace.
Fixes: e93c89c1aa ("Btrfs: add new sources for device replace code")
CC: stable@vger.kernel.org # 4.4+
Signed-off-by: Anand Jain <anand.jain@oracle.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
There isn't any other consumer other than in its own file dev-replace.c.
Signed-off-by: Anand Jain <anand.jain@oracle.com>
Reviewed-by: Nikolay Borisov <nborisov@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
dev_replace::replace_state has been set to
BTRFS_DEV_REPLACE_ITEM_STATE_NEVER_STARTED (0) in the same function,
So delete the line which sets replace_state = 0;
Reviewed-by: Nikolay Borisov <nborisov@suse.com>
Signed-off-by: Anand Jain <anand.jain@oracle.com>
Signed-off-by: David Sterba <dsterba@suse.com>
A later patch will implement swap file support for Btrfs, but before we
do that, we need to make sure that the various Btrfs ioctls cannot
change a swap file.
When a swap file is active, we must make sure that the extents of the
file are not moved and that they don't become shared. That means that
the following are not safe:
- chattr +c (enable compression)
- reflink
- dedupe
- snapshot
- defrag
Don't allow those to happen on an active swap file.
Additionally, balance, resize, device remove, and device replace are
also unsafe if they affect an active swapfile. Add a red-black tree of
block groups and devices which contain an active swapfile. Relocation
checks each block group against this tree and skips it or errors out for
balance or resize, respectively. Device remove and device replace check
the tree for the device they will operate on.
Note that we don't have to worry about chattr -C (disable nocow), which
we ignore for non-empty files, because an active swapfile must be
non-empty and can't be truncated. We also don't have to worry about
autodefrag because it's only done on COW files. Truncate and fallocate
are already taken care of by the generic code. Device add doesn't do
relocation so it's not an issue, either.
Signed-off-by: Omar Sandoval <osandov@fb.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
The value of blocking_readers is increased only when the lock is taken
for read, no way we can fail the condition with the write lock.
Reviewed-by: Anand Jain <anand.jain@oracle.com>
Signed-off-by: David Sterba <dsterba@suse.com>
The replace_wait and bio_counter were mistakenly added to fs_info in
commit c404e0dc2c ("Btrfs: fix use-after-free in the finishing
procedure of the device replace"), but they logically belong to
fs_info::dev_replace. Besides, bio_counter is a very generic name and is
confusing in bare fs_info context.
Reviewed-by: Anand Jain <anand.jain@oracle.com>
Signed-off-by: David Sterba <dsterba@suse.com>
The exit sequence in btrfs_dev_replace_start does not allow to simply
add a label to the right place so the error handling after starting
transaction failure jumps there. Currently there's a lock that pairs
with the unlock in the section, which is unnecessary and only raises
questions. Add a variable to track the locking status and avoid the
extra locking.
Reviewed-by: Omar Sandoval <osandov@fb.com>
Reviewed-by: Anand Jain <anand.jain@oracle.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Too trivial, the purpose can be simply documented in a comment.
Reviewed-by: Omar Sandoval <osandov@fb.com>
Reviewed-by: Anand Jain <anand.jain@oracle.com>
Signed-off-by: David Sterba <dsterba@suse.com>
There's a single caller and the function name does not say it's actually
taking the lock, so open coding makes it more explicit.
For now, btrfs_dev_replace_read_lock is used instead of read_lock so
it's paired with the unlocking wrapper in the same block.
Reviewed-by: Anand Jain <anand.jain@oracle.com>
Signed-off-by: David Sterba <dsterba@suse.com>
This member seems to be copied from the extent_buffer locking scheme and
is at least used to assert that the read lock/unlock is properly nested.
In some way. While the _inc/_dec are called inside the read lock
section, the asserts are both inside and outside, so the ordering is not
guaranteed and we can see read/inc/dec ordered in any way
(theoretically).
A missing call of btrfs_dev_replace_clear_lock_blocking could cause
unexpected read_locks count, so this at least looks like a valid
assertion, but this will become unnecessary with later updates.
Reviewed-by: Anand Jain <anand.jain@oracle.com>
Signed-off-by: David Sterba <dsterba@suse.com>
When we fail to start a transaction in btrfs_dev_replace_start, we leave
dev_replace->replace_start set to STARTED but clear ->srcdev and
->tgtdev. Later, that can result in an Oops in
btrfs_dev_replace_progress when having state set to STARTED or SUSPENDED
implies that ->srcdev is valid.
Also fix error handling when the state is already STARTED or SUSPENDED
while starting. That, too, will clear ->srcdev and ->tgtdev even though
it doesn't own them. This should be an impossible case to hit since we
should be protected by the BTRFS_FS_EXCL_OP bit being set. Let's add an
ASSERT there while we're at it.
Fixes: e93c89c1aa (Btrfs: add new sources for device replace code)
CC: stable@vger.kernel.org # 4.4+
Signed-off-by: Jeff Mahoney <jeffm@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Instead of returning an error value and using one of the parameters for
returning the actual object we are interested in just refactor the
function to directly return btrfs_device *. Also bubble up the error
handling for the special BTRFS_ERROR_DEV_MISSING_NOT_FOUND value into
btrfs_rm_device. No functional changes.
Signed-off-by: Nikolay Borisov <nborisov@suse.com>
Reviewed-by: Qu Wenruo <wqu@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
on-disk devs stats value is updated in btrfs_run_dev_stats(),
which is called during commit transaction, if device->dev_stats_ccnt
is not zero.
Since current replace operation does not touch dev_stats_ccnt,
on-disk dev stats value is not updated. Therefore "btrfs device stats"
may return old device's value after umount/mount
(Example: See "btrfs ins dump-t -t DEV $DEV" after btrfs/100 finish).
Fix this by just incrementing dev_stats_ccnt in
btrfs_dev_replace_finishing() when replace is succeeded and this will
update the values.
Signed-off-by: Misono Tomohiro <misono.tomohiro@jp.fujitsu.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
This function is always passed a well-formed tgtdevice so the fs_info
can be referenced from there.
Signed-off-by: Nikolay Borisov <nborisov@suse.com>
Reviewed-by: Lu Fengqi <lufq.fnst@cn.fujitsu.com>
Signed-off-by: David Sterba <dsterba@suse.com>
It can be referenced from the passed 'device' argument which is always
a well-formed device.
Signed-off-by: Nikolay Borisov <nborisov@suse.com>
Reviewed-by: Lu Fengqi <lufq.fnst@cn.fujitsu.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
It can be referenced from the passed srcdev argument.
Signed-off-by: Nikolay Borisov <nborisov@suse.com>
Reviewed-by: Lu Fengqi <lufq.fnst@cn.fujitsu.com>
Signed-off-by: David Sterba <dsterba@suse.com>
The get_seconds() function is deprecated as it truncates the timestamp
to 32 bits. Change it to or ktime_get_real_seconds().
Signed-off-by: Allen Pais <allen.lkml@gmail.com>
Reviewed-by: David Sterba <dsterba@suse.com>
[ update changelog ]
Signed-off-by: David Sterba <dsterba@suse.com>
btrfs_dev_replace_finishing updates devices (soruce and target) which
are within the btrfs_fs_devices::devices or withint the cloned seed
devices (btrfs_fs_devices::seed::devices), so we don't need the global
uuid_mutex.
The device replace context is also locked by its own locks.
Signed-off-by: Anand Jain <anand.jain@oracle.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Use the wrappers and reduce the amount of low-level details about the
waitqueue management.
Reviewed-by: Nikolay Borisov <nborisov@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
This parameter was introduced alongside the function in
eb73c1b7ce ("Btrfs: introduce per-subvolume delalloc inode list") to
avoid deadlocks since this function was used in the transaction commit
path. However, commit 8d875f95da ("btrfs: disable strict file flushes
for renames and truncates") removed that usage, rendering the parameter
obsolete.
Signed-off-by: Nikolay Borisov <nborisov@suse.com>
Reviewed-by: Qu Wenruo <wqu@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
The volume mutex does not protect against anything in this case, the
comment about scrub is right but not related to locking and looks
confusing. The comment in btrfs_find_device_missing_or_by_path is wrong
and confusing too.
The device_list_mutex is not held here to protect device lookup, but in
this case device replace cannot run in parallel with device removal (due
to exclusive op protection), so we don't need further locking here.
Reviewed-by: Anand Jain <anand.jain@oracle.com>
Signed-off-by: David Sterba <dsterba@suse.com>
The device replace is paused by unmount or read only remount, and
resumed on next mount or write remount.
The exclusive status should be checked properly as it's a global
invariant and we must not allow 2 operations run. In this case, the
balance can be also paused and resumed under same conditions. It's
always checked first so dev-replace could see the EXCL_OP already taken,
BUT, the ioctl would never let start both at the same time.
Replace the WARN_ON with message and return 0, indicating no error as
this is purely theoretical and the user will be informed. Resolving that
manually should be possible by waiting for the other operation to finish
or cancel the paused state.
Reviewed-by: Anand Jain <anand.jain@oracle.com>
Reviewed-by: Nikolay Borisov <nborisov@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
The function logically belongs there and there's only a single caller,
no need to export it. No code changes.
Reviewed-by: Anand Jain <anand.jain@oracle.com>
Signed-off-by: David Sterba <dsterba@suse.com>
This is a preparatory cleanup that will make clear that the only
successful way out of btrfs_init_dev_replace_tgtdev will also set the
device_out to a valid pointer. With this guarantee, the callers can be
simplified.
Reviewed-by: Anand Jain <anand.jain@oracle.com>
Signed-off-by: David Sterba <dsterba@suse.com>
The function is called once and is fairly small, we can merge it with
the caller.
Reviewed-by: Anand Jain <anand.jain@oracle.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Remove GPL boilerplate text (long, short, one-line) and keep the rest,
ie. personal, company or original source copyright statements. Add the
SPDX header.
Signed-off-by: David Sterba <dsterba@suse.com>
The current calls are unclear in what way btrfs_dev_replace_lock takes
the locks, so drop the argument, split the helpers and use similar
naming as for read and write locks.
Signed-off-by: David Sterba <dsterba@suse.com>
For debugging or administration purposes, we would want to know if and
when the user cancels the replace, to complement the existing messages
when dev-replace starts or finishes.
Signed-off-by: Anand Jain <anand.jain@oracle.com>
Reviewed-by: David Sterba <dsterba@suse.com>
[ update changelog, fold fix for RCU warning from Nikolay ]
Signed-off-by: David Sterba <dsterba@suse.com>
The replace target device can be missing when mounted with -o degraded,
but we wont allocate a missing btrfs_device to it. So check the device
before accessing.
BUG: unable to handle kernel NULL pointer dereference at 00000000000000b0
IP: btrfs_destroy_dev_replace_tgtdev+0x43/0xf0 [btrfs]
Call Trace:
btrfs_dev_replace_cancel+0x15f/0x180 [btrfs]
btrfs_ioctl+0x2216/0x2590 [btrfs]
do_vfs_ioctl+0x625/0x650
SyS_ioctl+0x4e/0x80
do_syscall_64+0x5d/0x160
entry_SYSCALL64_slow_path+0x25/0x25
This patch has been moved in front of patch "btrfs: log, when replace,
is canceled by the user" that could reproduce the crash if the system
reboots inside btrfs_dev_replace_start before the
btrfs_dev_replace_finishing call.
$ mkfs /dev/sda
$ mount /dev/sda mnt
$ btrfs replace start /dev/sda /dev/sdb
<insert reboot>
$ mount po degraded /dev/sdb mnt
<crash>
Signed-off-by: Anand Jain <anand.jain@oracle.com>
[ added reproducer description from mail ]
Signed-off-by: David Sterba <dsterba@suse.com>
btrfs_init_dev_replace_tgtdev_for_resume() initializes replace
target device in a few simple steps, so do it at the parent function.
Moreover, there isn't any other caller so just open code it.
Signed-off-by: Anand Jain <anand.jain@oracle.com>
Reviewed-by: Qu Wenruo <wqu@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Current u64 return from btrfs_dev_replace_cancel() was probably done
to match the btrfs_ioctl_dev_replace_args::result. However as our
actual return value fits in int, and it further gets typecast to u64,
so just return int.
Signed-off-by: Anand Jain <anand.jain@oracle.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Remove __ which is for the special functions.
Signed-off-by: Anand Jain <anand.jain@oracle.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
btrfs_dev_replace_cancel() calls __btrfs_dev_replace_cancel() for the
actual cancel so just open code it.
Signed-off-by: Anand Jain <anand.jain@oracle.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Currently device state is being managed by each individual int
variable such as struct btrfs_device::is_tgtdev_for_dev_replace.
Instead of that declare btrfs_device::dev_state
BTRFS_DEV_STATE_MISSING and use the bit operations.
Signed-off-by: Anand Jain <anand.jain@oracle.com>
[ whitespace adjustments ]
Signed-off-by: David Sterba <dsterba@suse.com>
Currently device state is being managed by each individual int
variable such as struct btrfs_device::missing. Instead of that
declare btrfs_device::dev_state BTRFS_DEV_STATE_MISSING and use
the bit operations.
Signed-off-by: Anand Jain <anand.jain@oracle.com>
Reviewed-by : Nikolay Borisov <nborisov@suse.com>
[ whitespace adjustments ]
Signed-off-by: David Sterba <dsterba@suse.com>
This patch creates a helper function to get either the rcu device path
or missing.
Signed-off-by: Anand Jain <anand.jain@oracle.com>
[ rename to btrfs_dev_name, switch to if/else ]
Signed-off-by: David Sterba <dsterba@suse.com>
Pull mount flag updates from Al Viro:
"Another chunk of fmount preparations from dhowells; only trivial
conflicts for that part. It separates MS_... bits (very grotty
mount(2) ABI) from the struct super_block ->s_flags (kernel-internal,
only a small subset of MS_... stuff).
This does *not* convert the filesystems to new constants; only the
infrastructure is done here. The next step in that series is where the
conflicts would be; that's the conversion of filesystems. It's purely
mechanical and it's better done after the merge, so if you could run
something like
list=$(for i in MS_RDONLY MS_NOSUID MS_NODEV MS_NOEXEC MS_SYNCHRONOUS MS_MANDLOCK MS_DIRSYNC MS_NOATIME MS_NODIRATIME MS_SILENT MS_POSIXACL MS_KERNMOUNT MS_I_VERSION MS_LAZYTIME; do git grep -l $i fs drivers/staging/lustre drivers/mtd ipc mm include/linux; done|sort|uniq|grep -v '^fs/namespace.c$')
sed -i -e 's/\<MS_RDONLY\>/SB_RDONLY/g' \
-e 's/\<MS_NOSUID\>/SB_NOSUID/g' \
-e 's/\<MS_NODEV\>/SB_NODEV/g' \
-e 's/\<MS_NOEXEC\>/SB_NOEXEC/g' \
-e 's/\<MS_SYNCHRONOUS\>/SB_SYNCHRONOUS/g' \
-e 's/\<MS_MANDLOCK\>/SB_MANDLOCK/g' \
-e 's/\<MS_DIRSYNC\>/SB_DIRSYNC/g' \
-e 's/\<MS_NOATIME\>/SB_NOATIME/g' \
-e 's/\<MS_NODIRATIME\>/SB_NODIRATIME/g' \
-e 's/\<MS_SILENT\>/SB_SILENT/g' \
-e 's/\<MS_POSIXACL\>/SB_POSIXACL/g' \
-e 's/\<MS_KERNMOUNT\>/SB_KERNMOUNT/g' \
-e 's/\<MS_I_VERSION\>/SB_I_VERSION/g' \
-e 's/\<MS_LAZYTIME\>/SB_LAZYTIME/g' \
$list
and commit it with something along the lines of 'convert filesystems
away from use of MS_... constants' as commit message, it would save a
quite a bit of headache next cycle"
* 'work.mount' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/vfs:
VFS: Differentiate mount flags (MS_*) from internal superblock flags
VFS: Convert sb->s_flags & MS_RDONLY to sb_rdonly(sb)
vfs: Add sb_rdonly(sb) to query the MS_RDONLY flag on s_flags
This function prints an informative message and then continues
dev-replace. The message contains a progress percentage which is read
from the status. The status is allocated dynamically, about 2600 bytes,
just to read the single value. That's an overkill. We'll use the new
helper and drop the allocation.
Signed-off-by: David Sterba <dsterba@suse.com>
We'll want to read the percentage value from dev_replace elsewhere, move
the logic to a separate helper.
Signed-off-by: David Sterba <dsterba@suse.com>
Firstly by applying the following with coccinelle's spatch:
@@ expression SB; @@
-SB->s_flags & MS_RDONLY
+sb_rdonly(SB)
to effect the conversion to sb_rdonly(sb), then by applying:
@@ expression A, SB; @@
(
-(!sb_rdonly(SB)) && A
+!sb_rdonly(SB) && A
|
-A != (sb_rdonly(SB))
+A != sb_rdonly(SB)
|
-A == (sb_rdonly(SB))
+A == sb_rdonly(SB)
|
-!(sb_rdonly(SB))
+!sb_rdonly(SB)
|
-A && (sb_rdonly(SB))
+A && sb_rdonly(SB)
|
-A || (sb_rdonly(SB))
+A || sb_rdonly(SB)
|
-(sb_rdonly(SB)) != A
+sb_rdonly(SB) != A
|
-(sb_rdonly(SB)) == A
+sb_rdonly(SB) == A
|
-(sb_rdonly(SB)) && A
+sb_rdonly(SB) && A
|
-(sb_rdonly(SB)) || A
+sb_rdonly(SB) || A
)
@@ expression A, B, SB; @@
(
-(sb_rdonly(SB)) ? 1 : 0
+sb_rdonly(SB)
|
-(sb_rdonly(SB)) ? A : B
+sb_rdonly(SB) ? A : B
)
to remove left over excess bracketage and finally by applying:
@@ expression A, SB; @@
(
-(A & MS_RDONLY) != sb_rdonly(SB)
+(bool)(A & MS_RDONLY) != sb_rdonly(SB)
|
-(A & MS_RDONLY) == sb_rdonly(SB)
+(bool)(A & MS_RDONLY) == sb_rdonly(SB)
)
to make comparisons against the result of sb_rdonly() (which is a bool)
work correctly.
Signed-off-by: David Howells <dhowells@redhat.com>
Dave Jones hit a WARN_ON(nr < 0) in btrfs_wait_ordered_roots() with
v4.12-rc6. This was because commit 70e7af244 made it possible for
calc_reclaim_items_nr() to return a negative number. It's not really a
bug in that commit, it just didn't go far enough down the stack to find
all the possible 64->32 bit overflows.
This switches calc_reclaim_items_nr() to return a u64 and changes everyone
that uses the results of that math to u64 as well.
Reported-by: Dave Jones <davej@codemonkey.org.uk>
Fixes: 70e7af2 ("Btrfs: fix delalloc accounting leak caused by u32 overflow")
Signed-off-by: Chris Mason <clm@fb.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
This is fixing code pieces where we use div_u64 when passing a u64 divisor.
Cc: David Sterba <dsterba@suse.cz>
Signed-off-by: Liu Bo <bo.li.liu@oracle.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
When raid56 dev-replace is cancelled by running scrub, we will free
target device without waiting for in-flight bios, causing the following
NULL pointer deference or general protection failure.
BUG: unable to handle kernel NULL pointer dereference at 00000000000005e0
IP: generic_make_request_checks+0x4d/0x610
CPU: 1 PID: 11676 Comm: kworker/u4:14 Tainted: G O 4.11.0-rc2 #72
Hardware name: QEMU Standard PC (Q35 + ICH9, 2009), BIOS 1.10.2-20170228_101828-anatol 04/01/2014
Workqueue: btrfs-endio-raid56 btrfs_endio_raid56_helper [btrfs]
task: ffff88002875b4c0 task.stack: ffffc90001334000
RIP: 0010:generic_make_request_checks+0x4d/0x610
Call Trace:
? generic_make_request+0xc7/0x360
generic_make_request+0x24/0x360
? generic_make_request+0xc7/0x360
submit_bio+0x64/0x120
? page_in_rbio+0x4d/0x80 [btrfs]
? rbio_orig_end_io+0x80/0x80 [btrfs]
finish_rmw+0x3f4/0x540 [btrfs]
validate_rbio_for_rmw+0x36/0x40 [btrfs]
raid_rmw_end_io+0x7a/0x90 [btrfs]
bio_endio+0x56/0x60
end_workqueue_fn+0x3c/0x40 [btrfs]
btrfs_scrubparity_helper+0xef/0x620 [btrfs]
btrfs_endio_raid56_helper+0xe/0x10 [btrfs]
process_one_work+0x2af/0x720
? process_one_work+0x22b/0x720
worker_thread+0x4b/0x4f0
kthread+0x10f/0x150
? process_one_work+0x720/0x720
? kthread_create_on_node+0x40/0x40
ret_from_fork+0x2e/0x40
RIP: generic_make_request_checks+0x4d/0x610 RSP: ffffc90001337bb8
In btrfs_dev_replace_finishing(), we will call
btrfs_rm_dev_replace_blocked() to wait bios before destroying the target
device when scrub is finished normally.
However when dev-replace is aborted, either due to error or cancelled by
scrub, we didn't wait for bios, this can lead to use-after-free if there
are bios holding the target device.
Furthermore, for raid56 scrub, at least 2 places are calling
btrfs_map_sblock() without protection of bio_counter, leading to the
problem.
This patch fixes the problem:
1) Wait for bio_counter before freeing target device when canceling
replace
2) When calling btrfs_map_sblock() for raid56, use bio_counter to
protect the call.
Cc: Liu Bo <bo.li.liu@oracle.com>
Signed-off-by: Qu Wenruo <quwenruo@cn.fujitsu.com>
Reviewed-by: Liu Bo <bo.li.liu@oracle.com>
Signed-off-by: David Sterba <dsterba@suse.com>
There are several operations, usually started from ioctls, that cannot
run concurrently. The status is tracked in
mutually_exclusive_operation_running as an atomic_t. We can easily track
the status as one of the per-filesystem flag bits with same
synchronization guarantees.
The conversion replaces:
* atomic_xchg(..., 1) -> test_and_set_bit(FLAG, ...)
* atomic_set(..., 0) -> clear_bit(FLAG, ...)
Reviewed-by: Anand Jain <anand.jain@oracle.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Now we only use the root parameter to print the root objectid in
a tracepoint. We can use the root parameter from the transaction
handle for that. It's also used to join the transaction with
async commits, so we remove the comment that it's just for checking.
Signed-off-by: Jeff Mahoney <jeffm@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
There are loads of functions in btrfs that accept a root parameter
but only use it to obtain an fs_info pointer. Let's convert those to
just accept an fs_info pointer directly.
Signed-off-by: Jeff Mahoney <jeffm@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
In routines where someptr->fs_info is referenced multiple times, we
introduce a convenience variable. This makes the code considerably
more readable.
Signed-off-by: Jeff Mahoney <jeffm@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
For many printks, we want to know which file system issued the message.
This patch converts most pr_* calls to use the btrfs_* versions instead.
In some cases, this means adding plumbing to allow call sites access to
an fs_info pointer.
fs/btrfs/check-integrity.c is left alone for another day.
Signed-off-by: Jeff Mahoney <jeffm@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
CodingStyle chapter 2:
"[...] never break user-visible strings such as printk messages,
because that breaks the ability to grep for them."
This patch unsplits user-visible strings.
Signed-off-by: Jeff Mahoney <jeffm@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
btrfs_test_opt and friends only use the root pointer to access
the fs_info. Let's pass the fs_info directly in preparation to
eliminate similar patterns all over btrfs.
Signed-off-by: Jeff Mahoney <jeffm@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Before the relocation process of a block group starts, it sets the block
group to readonly mode, then flushes all delalloc writes and then finally
it waits for all ordered extents to complete. This last step includes
waiting for ordered extents destinated at extents allocated in other block
groups, making us waste unecessary time.
So improve this by waiting only for ordered extents that fall into the
block group's range.
Signed-off-by: Filipe Manana <fdmanana@suse.com>
Reviewed-by: Josef Bacik <jbacik@fb.com>
Reviewed-by: Liu Bo <bo.li.liu@oracle.com>
Creates helper fucntion as needed by the device delete
and replace operations. Also now it checks if the next
device being assigned is an active device.
Signed-off-by: Anand Jain <anand.jain@oracle.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Yauhen reported in the ML that s_bdev is null at mount, and
s_bdev gets updated to some device when missing device is
replaced, as because bdev is null for missing device, things
gets matched up. Fix this by checking if s_bdev is set. I
didn't want to completely remove updating s_bdev because
the future multi device support at vfs layer may need it.
Signed-off-by: Anand Jain <anand.jain@oracle.com>
Reported-by: Yauhen Kharuzhy <yauhen.kharuzhy@zavadatar.com>
Signed-off-by: David Sterba <dsterba@suse.com>
A refactor patch, and avoids user input verification in the
btrfs_dev_replace_start(), and so this function can be reused.
Signed-off-by: Anand Jain <anand.jain@oracle.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Local variable fs_info, contains root->fs_info, use it.
Signed-off-by: Anand Jain <anand.jain@oracle.com>
Signed-off-by: David Sterba <dsterba@suse.com>
For clarity how we are going to find the device, let's call it a device
specifier, devspec for short. Also rename the arguments that are a
leftover from previous function purpose.
Reviewed-by: Anand Jain <anand.jain@oracle.com>
Signed-off-by: David Sterba <dsterba@suse.com>
The operation of device replace and device delete follows same steps upto
some depth with in btrfs kernel, however they don't share codes. This
enhancement will help replace and delete to share codes.
Signed-off-by: Anand Jain <anand.jain@oracle.com>
Signed-off-by: David Sterba <dsterba@suse.com>
The patch renames btrfs_dev_replace_find_srcdev() to
btrfs_find_device_by_user_input() and moves it to volumes.c, so that
delete device can use it.
Signed-off-by: Anand Jain <anand.jain@oracle.com>
Signed-off-by: David Sterba <dsterba@suse.com>
If device replace entry was found on disk at mounting and its num_write_errors
stats counter has non-NULL value, then replace operation will never be
finished and -EIO error will be reported by btrfs_scrub_dev() because
this counter is never reset.
# mount -o degraded /media/a4fb5c0a-21c5-4fe7-8d0e-fdd87d5f71ee/
# btrfs replace status /media/a4fb5c0a-21c5-4fe7-8d0e-fdd87d5f71ee/
Started on 25.Mar 07:28:00, canceled on 25.Mar 07:28:01 at 0.0%, 40 write errs, 0 uncorr. read errs
# btrfs replace start -B 4 /dev/sdg /media/a4fb5c0a-21c5-4fe7-8d0e-fdd87d5f71ee/
ERROR: ioctl(DEV_REPLACE_START) failed on "/media/a4fb5c0a-21c5-4fe7-8d0e-fdd87d5f71ee/": Input/output error, no error
Reset num_write_errors and num_uncorrectable_read_errors counters in the
dev_replace structure before start of replacing.
Signed-off-by: Yauhen Kharuzhy <yauhen.kharuzhy@zavadatar.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Xfstests btrfs/011 complains about a deadlock warning,
[ 1226.649039] =========================================================
[ 1226.649039] [ INFO: possible irq lock inversion dependency detected ]
[ 1226.649039] 4.1.0+ #270 Not tainted
[ 1226.649039] ---------------------------------------------------------
[ 1226.652955] kswapd0/46 just changed the state of lock:
[ 1226.652955] (&delayed_node->mutex){+.+.-.}, at: [<ffffffff81458735>] __btrfs_release_delayed_node+0x45/0x1d0
[ 1226.652955] but this lock took another, RECLAIM_FS-unsafe lock in the past:
[ 1226.652955] (&fs_info->dev_replace.lock){+.+.+.}
and interrupts could create inverse lock ordering between them.
[ 1226.652955]
other info that might help us debug this:
[ 1226.652955] Chain exists of:
&delayed_node->mutex --> &found->groups_sem --> &fs_info->dev_replace.lock
[ 1226.652955] Possible interrupt unsafe locking scenario:
[ 1226.652955] CPU0 CPU1
[ 1226.652955] ---- ----
[ 1226.652955] lock(&fs_info->dev_replace.lock);
[ 1226.652955] local_irq_disable();
[ 1226.652955] lock(&delayed_node->mutex);
[ 1226.652955] lock(&found->groups_sem);
[ 1226.652955] <Interrupt>
[ 1226.652955] lock(&delayed_node->mutex);
[ 1226.652955]
*** DEADLOCK ***
Commit 084b6e7c76 ("btrfs: Fix a lockdep warning when running xfstest.") tried
to fix a similar one that has the exactly same warning, but with that, we still
run to this.
The above lock chain comes from
btrfs_commit_transaction
->btrfs_run_delayed_items
...
->__btrfs_update_delayed_inode
...
->__btrfs_cow_block
...
->find_free_extent
->cache_block_group
->load_free_space_cache
->btrfs_readpages
->submit_one_bio
...
->__btrfs_map_block
->btrfs_dev_replace_lock
However, with high memory pressure, tasks which hold dev_replace.lock can
be interrupted by kswapd and then kswapd is intended to release memory occupied
by superblock, inodes and dentries, where we may call evict_inode, and it comes
to
[ 1226.652955] [<ffffffff81458735>] __btrfs_release_delayed_node+0x45/0x1d0
[ 1226.652955] [<ffffffff81459e74>] btrfs_remove_delayed_node+0x24/0x30
[ 1226.652955] [<ffffffff8140c5fe>] btrfs_evict_inode+0x34e/0x700
delayed_node->mutex may be acquired in __btrfs_release_delayed_node(), and it leads
to a ABBA deadlock.
To fix this, we can use "blocking rwlock" used in the case of extent_buffer, but
things are simpler here since we only needs read's spinlock to blocking lock.
With this, btrfs/011 no more produces warnings in dmesg.
Signed-off-by: Liu Bo <bo.li.liu@oracle.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Scrub is not on the critical writeback path we don't need to use
GFP_NOFS for all allocations. The failures are handled and stats passed
back to userspace.
Let's use GFP_KERNEL on the paths where everything is ok, ie. setup the
global structures and the IO submission paths.
Functions that do the repair and fixups still use GFP_NOFS as we might
want to skip any other filesystem activity if we encounter an error.
This could turn out to be unnecessary, but requires more review compared
to the easy cases in this patch.
Signed-off-by: David Sterba <dsterba@suse.com>
Overloading extent_map->bdev to struct map_lookup * might have started out
as a means to an end, but it's a pattern that's used all over the place
now. Let's get rid of the casting and just add a union instead.
Signed-off-by: Jeff Mahoney <jeffm@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Normally the waitqueue_active would need a barrier, but this is not
necessary here because it's not a performance sensitive context and we
can call wake_up directly.
Suggested-by: Chris Mason <clm@fb.com>
Signed-off-by: David Sterba <dsterba@suse.com>
To avoid deadlock described in commit 084b6e7c76 ("btrfs: Fix a
lockdep warning when running xfstest."), we should move kobj stuff out
of dev_replace lock range.
"It is because the btrfs_kobj_{add/rm}_device() will call memory
allocation with GFP_KERNEL,
which may flush fs page cache to free space, waiting for it self to do
the commit, causing the deadlock.
To solve the problem, move btrfs_kobj_{add/rm}_device() out of the
dev_replace lock range, also involing split the
btrfs_rm_dev_replace_srcdev() function into remove and free parts.
Now only btrfs_rm_dev_replace_remove_srcdev() is called in dev_replace
lock range, and kobj_{add/rm} and btrfs_rm_dev_replace_free_srcdev() are
called out of the lock range."
Signed-off-by: Liu Bo <bo.li.liu@oracle.com>
Signed-off-by: Anand Jain <anand.jain@oracle.com>
[added lockup description]
Signed-off-by: David Sterba <dsterba@suse.com>
By general rule of thumb there shouldn't be any way that user land
could trigger a kernel operation just by sending wrong arguments.
Here do commit cleanups after user input has been verified.
Signed-off-by: Anand Jain <anand.jain@oracle.com>
Signed-off-by: David Sterba <dsterba@suse.com>
We need not check path before btrfs_free_path() is called because
path is checked in btrfs_free_path().
Signed-off-by: Tsutomu Itoh <t-itoh@jp.fujitsu.com>
Reviewed-by: Qu Wenruo <quwenruo@cn.fujitsu.com>
Signed-off-by: Chris Mason <clm@fb.com>
sorry I indented to use btrfs_err() and I have no idea
how btrfs_error() got there.
infact I was thinking about these kind of oversights
since these two func are too closely named.
Signed-off-by: Anand Jain <anand.jain@oracle.com>
Reviewed-by: Liu Bo <bo.li.liu@oracle.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: Chris Mason <clm@fb.com>
This patch will add support to show the replacing target in sysfs
during the process of replacement.
Signed-off-by: Anand Jain <anand.jain@oracle.com>
Signed-off-by: David Sterba <dsterba@suse.cz>
btrfs_kobj_add_device() does not need fs_info any more.
Signed-off-by: Anand Jain <anand.jain@oracle.com>
Signed-off-by: David Sterba <dsterba@suse.cz>
Switch to div_u64 if the divisor is a numeric constant or sum of
sizeof()s. We can remove a few instances of do_div that has the hidden
semtantics of changing the 1st argument.
Small power-of-two divisors are converted to bitshifts, large values are
kept intact for clarity.
Signed-off-by: David Sterba <dsterba@suse.cz>
1: Remove no-need DEFINE_WAIT(wait)
2: Add likely() for BTRFS_FS_STATE_DEV_REPLACING condition
3: Use while loop instead of goto
Signed-off-by: Zhao Lei <zhaolei@cn.fujitsu.com>
Signed-off-by: Miao Xie <miaox@cn.fujitsu.com>
Signed-off-by: Chris Mason <clm@fb.com>
The commit c404e0dc (Btrfs: fix use-after-free in the finishing
procedure of the device replace) fixed a use-after-free problem
which happened when removing the source device at the end of device
replace, but at that time, btrfs didn't support device replace
on raid56, so we didn't fix the problem on the raid56 profile.
Currently, we implemented device replace for raid56, so we need
kick that problem out before we enable that function for raid56.
The fix method is very simple, we just increase the bio per-cpu
counter before we submit a raid56 io, and decrease the counter
when the raid56 io ends.
Signed-off-by: Miao Xie <miaox@cn.fujitsu.com>
The following lockdep warning is triggered during xfstests:
[ 1702.980872] =========================================================
[ 1702.981181] [ INFO: possible irq lock inversion dependency detected ]
[ 1702.981482] 3.18.0-rc1 #27 Not tainted
[ 1702.981781] ---------------------------------------------------------
[ 1702.982095] kswapd0/77 just changed the state of lock:
[ 1702.982415] (&delayed_node->mutex){+.+.-.}, at: [<ffffffffa03b0b51>] __btrfs_release_delayed_node+0x41/0x1f0 [btrfs]
[ 1702.982794] but this lock took another, RECLAIM_FS-unsafe lock in the past:
[ 1702.983160] (&fs_info->dev_replace.lock){+.+.+.}
and interrupts could create inverse lock ordering between them.
[ 1702.984675]
other info that might help us debug this:
[ 1702.985524] Chain exists of:
&delayed_node->mutex --> &found->groups_sem --> &fs_info->dev_replace.lock
[ 1702.986799] Possible interrupt unsafe locking scenario:
[ 1702.987681] CPU0 CPU1
[ 1702.988137] ---- ----
[ 1702.988598] lock(&fs_info->dev_replace.lock);
[ 1702.989069] local_irq_disable();
[ 1702.989534] lock(&delayed_node->mutex);
[ 1702.990038] lock(&found->groups_sem);
[ 1702.990494] <Interrupt>
[ 1702.990938] lock(&delayed_node->mutex);
[ 1702.991407]
*** DEADLOCK ***
It is because the btrfs_kobj_{add/rm}_device() will call memory
allocation with GFP_KERNEL,
which may flush fs page cache to free space, waiting for it self to do
the commit, causing the deadlock.
To solve the problem, move btrfs_kobj_{add/rm}_device() out of the
dev_replace lock range, also involing split the
btrfs_rm_dev_replace_srcdev() function into remove and free parts.
Now only btrfs_rm_dev_replace_remove_srcdev() is called in dev_replace
lock range, and kobj_{add/rm} and btrfs_rm_dev_replace_free_srcdev() are
called out of the lock range.
Signed-off-by: Qu Wenruo <quwenruo@cn.fujitsu.com>
Signed-off-by: Chris Mason <clm@fb.com>
device replace could fail due to another running scrub process or any
other errors btrfs_scrub_dev() may hit, but this failure doesn't get
returned to userspace.
The following steps could reproduce this issue
mkfs -t btrfs -f /dev/sdb1 /dev/sdb2
mount /dev/sdb1 /mnt/btrfs
while true; do btrfs scrub start -B /mnt/btrfs >/dev/null 2>&1; done &
btrfs replace start -Bf /dev/sdb2 /dev/sdb3 /mnt/btrfs
# if this replace succeeded, do the following and repeat until
# you see this log in dmesg
# BTRFS: btrfs_scrub_dev(/dev/sdb2, 2, /dev/sdb3) failed -115
#btrfs replace start -Bf /dev/sdb3 /dev/sdb2 /mnt/btrfs
# once you see the error log in dmesg, check return value of
# replace
echo $?
Introduce a new dev replace result
BTRFS_IOCTL_DEV_REPLACE_RESULT_SCRUB_INPROGRESS
to catch -EINPROGRESS explicitly and return other errors directly to
userspace.
Signed-off-by: Eryu Guan <guaneryu@gmail.com>
Signed-off-by: Chris Mason <clm@fb.com>
The problem is:
Task0(device scan task) Task1(device replace task)
scan_one_device()
mutex_lock(&uuid_mutex)
device = find_device()
mutex_lock(&device_list_mutex)
lock_chunk()
rm_and_free_source_device
unlock_chunk()
mutex_unlock(&device_list_mutex)
check device
Destroying the target device if device replace fails also has the same problem.
We fix this problem by locking uuid_mutex during destroying source device or
target device, just like the device remove operation.
It is a temporary solution, we can fix this problem and make the code more
clear by atomic counter in the future.
Signed-off-by: Miao Xie <miaox@cn.fujitsu.com>
Signed-off-by: Chris Mason <clm@fb.com>
There were several problems about chunk mutex usage:
- Lock chunk mutex when updating metadata. It would cause the nested
deadlock because updating metadata might need allocate new chunks
that need acquire chunk mutex. We remove chunk mutex at this case,
because b-tree lock and other lock mechanism can help us.
- ABBA deadlock occured between device_list_mutex and chunk_mutex.
When we update device status, we must acquire device_list_mutex at the
beginning, and then we might get chunk_mutex during the device status
update because we need allocate new chunks for metadata COW. But at
most place, we acquire chunk_mutex at first and then acquire device list
mutex. We need change the lock order.
- Some place we needn't acquire chunk_mutex. For example we needn't get
chunk_mutex when we free a empty seed fs_devices structure.
Signed-off-by: Miao Xie <miaox@cn.fujitsu.com>
Signed-off-by: Chris Mason <clm@fb.com>
->total_bytes,->disk_total_bytes,->bytes_used is protected by chunk
lock when we change them, but sometimes we read them without any lock,
and we might get unexpected value. We fix this problem like inode's
i_size.
Signed-off-by: Miao Xie <miaox@cn.fujitsu.com>
Signed-off-by: Chris Mason <clm@fb.com>
device->bytes_used will be changed when allocating a new chunk, and
disk_total_size will be changed if resizing is successful.
Meanwhile, the on-disk super blocks of the previous transaction
might not be updated. Considering the consistency of the metadata
in the previous transaction, We should use the size in the previous
transaction to check if the super block is beyond the boundary
of the device.
Though it is not big problem because we don't use it now, but anyway
it is better that we make it be consistent with the common metadata,
maybe we will use it in the future.
Signed-off-by: Miao Xie <miaox@cn.fujitsu.com>
Signed-off-by: Chris Mason <clm@fb.com>
