| Age | Commit message (Collapse) | Author |
|---|
|
Since the Linux kernel's sprintf() has conversion to hex built-in via
"%*phN", delete md5_to_hex() and just use that. Also add an explicit
array bound to the dname parameter of nfs4_make_rec_clidname() to make
its size clear. No functional change.
Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Eric Biggers <ebiggers@kernel.org>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
|
|
There is an extraneous space before a newline in a dprintk message.
Remove the space.
Signed-off-by: Colin Ian King <colin.i.king@gmail.com>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
|
|
Instead of allowing the ctime to roll backward with a WRITE_ATTRS
delegation, set FMODE_NOCMTIME on the file and have it skip mtime and
ctime updates.
It is possible that the client will never send a SETATTR to set the
times before returning the delegation. Add two new bools to struct
nfs4_delegation:
dl_written: tracks whether the file has been written since the
delegation was granted. This is set in the WRITE and LAYOUTCOMMIT
handlers.
dl_setattr: tracks whether the client has sent at least one valid
mtime that can also update the ctime in a SETATTR.
When unlocking the lease for the delegation, clear FMODE_NOCMTIME. If
the file has been written, but no setattr for the delegated mtime and
ctime has been done, update the timestamps to current_time().
Suggested-by: NeilBrown <neil@brown.name>
Signed-off-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
|
|
When updating the local timestamps from CB_GETATTR, the updated values
are not being properly vetted.
Compare the update times vs. the saved times in the delegation rather
than the current times in the inode. Also, ensure that the ctime is
properly vetted vs. its original value.
Fixes: 6ae30d6eb26b ("nfsd: add support for delegated timestamps")
Signed-off-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
|
|
SETATTRs containing delegated timestamp updates are currently not being
vetted properly. Since we no longer need to compare the timestamps vs.
the current timestamps, move the vetting of delegated timestamps wholly
into nfsd.
Rename the set_cb_time() helper to nfsd4_vet_deleg_time(), and make it
non-static. Add a new vet_deleg_attrs() helper that is called from
nfsd4_setattr that uses nfsd4_vet_deleg_time() to properly validate the
all the timestamps. If the validation indicates that the update should
be skipped, unset the appropriate flags in ia_valid.
Fixes: 7e13f4f8d27d ("nfsd: handle delegated timestamps in SETATTR")
Signed-off-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
|
|
As Trond points out [1], the "original time" mentioned in RFC 9754
refers to the timestamps on the files at the time that the delegation
was granted, and not the current timestamp of the file on the server.
Store the current timestamps for the file in the nfs4_delegation when
granting one. Add STATX_ATIME and STATX_MTIME to the request mask in
nfs4_delegation_stat(). When granting OPEN_DELEGATE_READ_ATTRS_DELEG, do
a nfs4_delegation_stat() and save the correct atime. If the stat() fails
for any reason, fall back to granting a normal read deleg.
[1]: https://lore.kernel.org/linux-nfs/47a4e40310e797f21b5137e847b06bb203d99e66.camel@kernel.org/
Fixes: 7e13f4f8d27d ("nfsd: handle delegated timestamps in SETATTR")
Signed-off-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
|
|
Ensure that notify_change() doesn't clobber a delegated ctime update
with current_time() by setting ATTR_CTIME_SET for those updates.
Don't bother setting the timestamps in cb_getattr_update_times() in the
non-delegated case. notify_change() will do that itself.
Fixes: 7e13f4f8d27d ("nfsd: handle delegated timestamps in SETATTR")
Signed-off-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
|
|
When ATTR_ATIME_SET and ATTR_MTIME_SET are set in the ia_valid mask, the
notify_change() logic takes that to mean that the request should set
those values explicitly, and not override them with "now".
With the advent of delegated timestamps, similar functionality is needed
for the ctime. Add a ATTR_CTIME_SET flag, and use that to indicate that
the ctime should be accepted as-is. Also, clean up the if statements to
eliminate the extra negatives.
In setattr_copy() and setattr_copy_mgtime() use inode_set_ctime_deleg()
when ATTR_CTIME_SET is set, instead of basing the decision on ATTR_DELEG.
Signed-off-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
|
|
If the only flag left is ATTR_DELEG, then there are no changes to be
made.
Fixes: 7e13f4f8d27d ("nfsd: handle delegated timestamps in SETATTR")
Signed-off-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
|
|
The ia_ctime.tv_nsec field should be set to modify.nseconds.
Fixes: 7e13f4f8d27d ("nfsd: handle delegated timestamps in SETATTR")
Signed-off-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
|
|
The data type of loca_last_write_offset is newoffset4 and is switched
on a boolean value, no_newoffset, that indicates if a previous write
occurred or not. If no_newoffset is FALSE, an offset is not given.
This means that client does not try to update the file size. Thus,
server should not try to calculate new file size and check if it fits
into the segment range. See RFC 8881, section 12.5.4.2.
Sometimes the current incorrect logic may cause clients to hang when
trying to sync an inode. If layoutcommit fails, the client marks the
inode as dirty again.
Fixes: 9cf514ccfacb ("nfsd: implement pNFS operations")
Cc: stable@vger.kernel.org
Co-developed-by: Konstantin Evtushenko <koevtushenko@yandex.com>
Signed-off-by: Konstantin Evtushenko <koevtushenko@yandex.com>
Signed-off-by: Sergey Bashirov <sergeybashirov@gmail.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
|
|
When pNFS client in the block or scsi layout mode sends layoutcommit
to MDS, a variable length array of modified extents is supplied within
the request. This patch allows the server to accept such extent arrays
if they do not fit within single memory page.
The issue can be reproduced when writing to a 1GB file using FIO with
O_DIRECT, 4K block and large I/O depth without preallocation of the
file. In this case, the server returns NFSERR_BADXDR to the client.
Co-developed-by: Konstantin Evtushenko <koevtushenko@yandex.com>
Signed-off-by: Konstantin Evtushenko <koevtushenko@yandex.com>
Signed-off-by: Sergey Bashirov <sergeybashirov@gmail.com>
Reviewed-by: Jeff Layton <jlayton@kernel.org>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
|
|
Use the appropriate xdr function to decode the lc_newoffset field,
which is a boolean value. See RFC 8881, section 18.42.1.
Signed-off-by: Sergey Bashirov <sergeybashirov@gmail.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
|
|
Remove dprintk in nfsd4_layoutcommit. These are not needed
in day to day usage, and the information is also available
in Wireshark when capturing NFS traffic.
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Sergey Bashirov <sergeybashirov@gmail.com>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
|
|
Compilers may optimize the layout of C structures, so we should not rely
on sizeof struct and memcpy to encode and decode XDR structures. The byte
order of the fields should also be taken into account.
This patch adds the correct functions to handle the deviceid4 structure
and removes the pad field, which is currently not used by NFSD, from the
runtime state. The server's byte order is preserved because the deviceid4
blob on the wire is only used as a cookie by the client.
Signed-off-by: Sergey Bashirov <sergeybashirov@gmail.com>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
|
|
This interface was deprecated by commit e6f7e1487ab5 ("nfs_localio:
simplify interface to nfsd for getting nfsd_file") and is now
unused. So let's remove it.
Signed-off-by: NeilBrown <neil@brown.name>
Reviewed-by: Mike Snitzer <snitzer@kernel.org>
Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
|
|
Clean up: The fh_getattr() function is part of NFSD's file handle
API, so relocate it.
I've made it an un-inlined function so that trace points and new
functionality can easily be introduced. That increases the size of
nfsd.ko by about a page on my x86_64 system (out of 26MB; compiled
with -O2).
Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
|
|
Clean up: these helpers are part of the NFSD file handle API.
Relocate them to fs/nfsd/nfsfh.h.
Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
|
|
mm/memory-failure.c defines and uses hwpoison_filter_* parameters but the
values of those parameters can only be modified via mm/hwpoison-inject.c
from userspace. They have a potentially different life time. Decouple
those parameters from mm/memory-failure.c to fix this broken layering.
Link: https://lkml.kernel.org/r/20250904062258.3336092-1-linmiaohe@huawei.com
Signed-off-by: Miaohe Lin <linmiaohe@huawei.com>
Suggested-by: Michal Hocko <mhocko@suse.com>
Cc: David Hildenbrand <david@redhat.com>
Cc: Naoya Horiguchi <nao.horiguchi@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
|
|
The return value of do_mpage_readpage() is arg->bio, which is already set
in the arg structure. Returning it again is redundant.
This patch changes the return type to void since the caller doesn't care
about the return value.
Link: https://lkml.kernel.org/r/20250829023659.688649-2-chizhiling@163.com
Signed-off-by: Chi Zhiling <chizhiling@kylinos.cn>
Reviewed-by: Jan Kara <jack@suse.cz>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Cc: Christian Brauner <brauner@kernel.org>
Cc: Matthew Wilcox (Oracle) <willy@infradead.org>
Cc: Namjae Jeon <linkinjeon@kernel.org>
Cc: Sungjong Seo <sj1557.seo@samsung.com>
Cc: Yuezhang Mo <Yuezhang.Mo@sony.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
|
|
For exFAT filesystems with 4MB read_ahead_size, removing the storage
device during read operations can delay EIO error reporting by several
minutes. This occurs because the read-ahead implementation in mpage
doesn't handle errors.
Another reason for the delay is that the filesystem requires metadata to
issue file read request. When the storage device is removed, the metadata
buffers are invalidated, causing mpage to repeatedly attempt to fetch
metadata during each get_block call.
The original purpose of this patch is terminate read ahead when we fail to
get metadata, to make the patch more generic, implement it by checking
folio status, instead of checking the return of get_block().
So, if a folio is synchronously unlocked and non-uptodate, should we quit
the read ahead?
I think it depends on whether the error is permanent or temporary, and
whether further read ahead might succeed. A device being unplugged is one
reason for returning such a folio, but we could return it for many other
reasons (e.g., metadata errors). I think most errors won't be restored in
a short time, so we should quit read ahead when they occur.
Link: https://lkml.kernel.org/r/20250829023659.688649-1-chizhiling@163.com
Signed-off-by: Chi Zhiling <chizhiling@kylinos.cn>
Reviewed-by: Jan Kara <jack@suse.cz>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Cc: Christian Brauner <brauner@kernel.org>
Cc: Matthew Wilcox (Oracle) <willy@infradead.org>
Cc: Namjae Jeon <linkinjeon@kernel.org>
Cc: Sungjong Seo <sj1557.seo@samsung.com>
Cc: Yuezhang Mo <Yuezhang.Mo@sony.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
|
|
Let's cleanup and simplify the function a bit.
Link: https://lkml.kernel.org/r/20250901150359.867252-17-david@redhat.com
Signed-off-by: David Hildenbrand <david@redhat.com>
Reviewed-by: Zi Yan <ziy@nvidia.com>
Reviewed-by: Lorenzo Stoakes <lorenzo.stoakes@oracle.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
|
|
The nth_page() is not really required anymore, so let's remove it.
Link: https://lkml.kernel.org/r/20250901150359.867252-16-david@redhat.com
Signed-off-by: David Hildenbrand <david@redhat.com>
Reviewed-by: Zi Yan <ziy@nvidia.com>
Reviewed-by: Lorenzo Stoakes <lorenzo.stoakes@oracle.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
|
|
changes required by mm-stable material: hugetlb and damon.
|
|
git://git.kernel.org/pub/scm/linux/kernel/git/kdave/linux
Pull a few more btrfs fixes from David Sterba:
- in tree-checker, fix wrong size of check for inode ref item
- in ref-verify, handle combination of mount options that allow
partially damaged extent tree (reported by syzbot)
- additional validation of compression mount option to catch invalid
string as level
* tag 'for-6.17-rc6-tag' of git://git.kernel.org/pub/scm/linux/kernel/git/kdave/linux:
btrfs: reject invalid compression level
btrfs: ref-verify: handle damaged extent root tree
btrfs: tree-checker: fix the incorrect inode ref size check
|
|
Pull smb client fixes from Steve French:
- Two unlink fixes: one for rename and one for deferred close
- Four smbdirect/RDMA fixes: fix buffer leak in negotiate, two fixes
for races in smbd_destroy, fix offset and length checks in recv_done
* tag '6.17-rc6-smb3-client-fixes' of git://git.samba.org/sfrench/cifs-2.6:
smb: client: fix smbdirect_recv_io leak in smbd_negotiate() error path
smb: client: fix file open check in __cifs_unlink()
smb: client: let smbd_destroy() call disable_work_sync(&info->post_send_credits_work)
smb: client: use disable[_delayed]_work_sync in smbdirect.c
smb: client: fix filename matching of deferred files
smb: client: let recv_done verify data_offset, data_length and remaining_data_length
|
|
This table is never modified, so mark it const.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Darrick J. Wong <djwong@kernel.org>
Signed-off-by: Carlos Maiolino <cem@kernel.org>
|
|
Just use the common helper we have.
Signed-off-by: Christian Brauner <brauner@kernel.org>
|
|
Make it easier to grep and rename to ns_count.
Reviewed-by: Jan Kara <jack@suse.cz>
Signed-off-by: Christian Brauner <brauner@kernel.org>
|
|
Stop accessing ns.count directly.
Reviewed-by: Jan Kara <jack@suse.cz>
Signed-off-by: Christian Brauner <brauner@kernel.org>
|
|
Stop accessing ns.count directly.
Reviewed-by: Jan Kara <jack@suse.cz>
Signed-off-by: Christian Brauner <brauner@kernel.org>
|
|
And drop ns_free_inum(). Anything common that can be wasted centrally
should be wasted in the new common helper.
Reviewed-by: Jan Kara <jack@suse.cz>
Signed-off-by: Christian Brauner <brauner@kernel.org>
|
|
Currently if a user enqueue a work item using schedule_delayed_work() the
used wq is "system_wq" (per-cpu wq) while queue_delayed_work() use
WORK_CPU_UNBOUND (used when a cpu is not specified). The same applies to
schedule_work() that is using system_wq and queue_work(), that makes use
again of WORK_CPU_UNBOUND.
This lack of consistentcy cannot be addressed without refactoring the API.
alloc_workqueue() treats all queues as per-CPU by default, while unbound
workqueues must opt-in via WQ_UNBOUND.
This default is suboptimal: most workloads benefit from unbound queues,
allowing the scheduler to place worker threads where they’re needed and
reducing noise when CPUs are isolated.
This patch adds a new WQ_PERCPU flag to all the fs subsystem users to
explicitly request the use of the per-CPU behavior. Both flags coexist
for one release cycle to allow callers to transition their calls.
Once migration is complete, WQ_UNBOUND can be removed and unbound will
become the implicit default.
With the introduction of the WQ_PERCPU flag (equivalent to !WQ_UNBOUND),
any alloc_workqueue() caller that doesn’t explicitly specify WQ_UNBOUND
must now use WQ_PERCPU.
All existing users have been updated accordingly.
Suggested-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Marco Crivellari <marco.crivellari@suse.com>
Link: https://lore.kernel.org/20250916082906.77439-4-marco.crivellari@suse.com
Signed-off-by: Christian Brauner <brauner@kernel.org>
|
|
Currently if a user enqueue a work item using schedule_delayed_work() the
used wq is "system_wq" (per-cpu wq) while queue_delayed_work() use
WORK_CPU_UNBOUND (used when a cpu is not specified). The same applies to
schedule_work() that is using system_wq and queue_work(), that makes use
again of WORK_CPU_UNBOUND.
This lack of consistentcy cannot be addressed without refactoring the API.
system_wq is a per-CPU worqueue, yet nothing in its name tells about that
CPU affinity constraint, which is very often not required by users.
Make it clear by adding a system_percpu_wq to all the fs subsystem.
The old wq will be kept for a few release cylces.
Suggested-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Marco Crivellari <marco.crivellari@suse.com>
Link: https://lore.kernel.org/20250916082906.77439-3-marco.crivellari@suse.com
Signed-off-by: Christian Brauner <brauner@kernel.org>
|
|
Currently if a user enqueue a work item using schedule_delayed_work() the
used wq is "system_wq" (per-cpu wq) while queue_delayed_work() use
WORK_CPU_UNBOUND (used when a cpu is not specified). The same applies to
schedule_work() that is using system_wq and queue_work(), that makes use
again of WORK_CPU_UNBOUND.
This lack of consistentcy cannot be addressed without refactoring the API.
system_unbound_wq should be the default workqueue so as not to enforce
locality constraints for random work whenever it's not required.
Adding system_dfl_wq to encourage its use when unbound work should be used.
The old system_unbound_wq will be kept for a few release cycles.
Suggested-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Marco Crivellari <marco.crivellari@suse.com>
Link: https://lore.kernel.org/20250916082906.77439-2-marco.crivellari@suse.com
Signed-off-by: Christian Brauner <brauner@kernel.org>
|
|
There's a lot of information that namespace implementers don't need to
know about at all. Encapsulate this all in the initialization helper.
Reviewed-by: Jan Kara <jack@suse.cz>
Signed-off-by: Christian Brauner <brauner@kernel.org>
|
|
Assign the reserved MNT_NS_ANON_INO sentinel to anonymous mount
namespaces and cleanup the initial mount ns allocation. This is just a
preparatory patch and the ns->inum check in ns_common_init() will be
dropped in the next patch.
Reviewed-by: Jan Kara <jack@suse.cz>
Signed-off-by: Christian Brauner <brauner@kernel.org>
|
|
There's various scenarios where we need to know whether we are in the
initial set of namespaces or not to e.g., shortcut permission checking.
All namespaces expose that information. Let's do that too.
Reviewed-by: Jan Kara <jack@suse.cz>
Signed-off-by: Christian Brauner <brauner@kernel.org>
|
|
The mount namespace has supported id retrieval for a while already.
Add support for the other types as well.
Signed-off-by: Christian Brauner <brauner@kernel.org>
|
|
Pidfd file handles are exhaustive meaning they don't require a handle on
another pidfd to pass to open_by_handle_at() so it can derive the
filesystem to decode in. Instead it can be derived from the file
handle itself. The same is possible for namespace file handles.
Reviewed-by: Amir Goldstein <amir73il@gmail.com>
Signed-off-by: Christian Brauner <brauner@kernel.org>
|
|
A while ago we added support for file handles to pidfs so pidfds can be
encoded and decoded as file handles. Userspace has adopted this quickly
and it's proven very useful. Implement file handles for namespaces as
well.
A process is not always able to open /proc/self/ns/. That requires
procfs to be mounted and for /proc/self/ or /proc/self/ns/ to not be
overmounted. However, userspace can always derive a namespace fd from
a pidfd. And that always works for a task's own namespace.
There's no need to introduce unnecessary behavioral differences between
/proc/self/ns/ fds, pidfd-derived namespace fds, and file-handle-derived
namespace fds. So namespace file handles are always decodable if the
caller is located in the namespace the file handle refers to.
This also allows a task to e.g., store a set of file handles to its
namespaces in a file on-disk so it can verify when it gets rexeced that
they're still valid and so on. This is akin to the pidfd use-case.
Or just plainly for namespace comparison reasons where a file handle to
the task's own namespace can be easily compared against others.
Reviewed-by: Amir Goldstein <amir73il@gmail.com>
Signed-off-by: Christian Brauner <brauner@kernel.org>
|
|
Move the mount namespace to the generic ns lookup infrastructure.
This allows us to drop a bunch of members from struct mnt_namespace.
Signed-off-by: Christian Brauner <brauner@kernel.org>
|
|
Bring in the fix for removing a mount namespace from the mount namespace
rbtree and list.
Signed-off-by: Christian Brauner <brauner@kernel.org>
|
|
Don't cargo-cult the same thing over and over.
Reviewed-by: Jan Kara <jack@suse.cz>
Signed-off-by: Christian Brauner <brauner@kernel.org>
|
|
Return IO errors if an ->iomap_begin implementation returns an
IOMAP_INLINE buffer but forgets to set the inline_data pointer.
Filesystems should never do this, but we could help fs developers (me)
fix their bugs by handling this more gracefully than crashing the
kernel.
Signed-off-by: Darrick J. Wong <djwong@kernel.org>
Link: https://lore.kernel.org/175803480324.966383.7414345025943296442.stgit@frogsfrogsfrogs
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Christian Brauner <brauner@kernel.org>
|
|
Trace which bytes actually get zeroed.
Signed-off-by: Darrick J. Wong <djwong@kernel.org>
Link: https://lore.kernel.org/175803480303.966383.2380024013746734540.stgit@frogsfrogsfrogs
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Christian Brauner <brauner@kernel.org>
|
|
When iput() drops the reference counter to zero, it may sleep via
inode_wait_for_writeback(). This happens rarely because it's usually
the dcache which evicts inodes, but really iput() should only ever be
called in contexts where sleeping is allowed. This annotation allows
finding buggy callers.
Additionally, this patch annotates a few low-level functions that can
call iput() conditionally.
Cc: Mateusz Guzik <mjguzik@gmail.com>
Signed-off-by: Max Kellermann <max.kellermann@ionos.com>
Link: https://lore.kernel.org/20250917153632.2228828-1-max.kellermann@ionos.com
Reviewed-by: Jan Kara <jack@suse.cz>
Signed-off-by: Christian Brauner <brauner@kernel.org>
|
|
Add trace_inode_switch_wbs_queue tracepoint to allow insight into how
many inodes are queued to switch their bdi_writeback structure.
Acked-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Jan Kara <jack@suse.cz>
Signed-off-by: Christian Brauner <brauner@kernel.org>
|
|
With lazytime mount option enabled we can be switching many dirty inodes
on cgroup exit to the parent cgroup. The numbers observed in practice
when systemd slice of a large cron job exits can easily reach hundreds
of thousands or millions. The logic in inode_do_switch_wbs() which sorts
the inode into appropriate place in b_dirty list of the target wb
however has linear complexity in the number of dirty inodes thus overall
time complexity of switching all the inodes is quadratic leading to
workers being pegged for hours consuming 100% of the CPU and switching
inodes to the parent wb.
Simple reproducer of the issue:
FILES=10000
# Filesystem mounted with lazytime mount option
MNT=/mnt/
echo "Creating files and switching timestamps"
for (( j = 0; j < 50; j ++ )); do
mkdir $MNT/dir$j
for (( i = 0; i < $FILES; i++ )); do
echo "foo" >$MNT/dir$j/file$i
done
touch -a -t 202501010000 $MNT/dir$j/file*
done
wait
echo "Syncing and flushing"
sync
echo 3 >/proc/sys/vm/drop_caches
echo "Reading all files from a cgroup"
mkdir /sys/fs/cgroup/unified/mycg1 || exit
echo $$ >/sys/fs/cgroup/unified/mycg1/cgroup.procs || exit
for (( j = 0; j < 50; j ++ )); do
cat /mnt/dir$j/file* >/dev/null &
done
wait
echo "Switching wbs"
# Now rmdir the cgroup after the script exits
We need to maintain b_dirty list ordering to keep writeback happy so
instead of sorting inode into appropriate place just append it at the
end of the list and clobber dirtied_time_when. This may result in inode
writeback starting later after cgroup switch however cgroup switches are
rare so it shouldn't matter much. Since the cgroup had write access to
the inode, there are no practical concerns of the possible DoS issues.
Acked-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Jan Kara <jack@suse.cz>
Signed-off-by: Christian Brauner <brauner@kernel.org>
|
|
process_inode_switch_wbs_work() can be switching over 100 inodes to a
different cgroup. Since switching an inode requires counting all dirty &
under-writeback pages in the address space of each inode, this can take
a significant amount of time. Add a possibility to reschedule after
processing each inode to avoid softlockups.
Acked-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Jan Kara <jack@suse.cz>
Signed-off-by: Christian Brauner <brauner@kernel.org>
|
