-----BEGIN PGP SIGNATURE-----
iHUEABYKAB0WIQRAhzRXHqcMeLMyaSiRxhvAZXjcogUCZuQEGwAKCRCRxhvAZXjc
ojIuAQC433+hBkvjvmQ7H0r5rgZSjUuCTG3bSmdU7RJmPHUHhwEA85v/NGq53f+W
IhandK6t+Cf0JYpFZ3N0bT88hDYVhQQ=
=9zGL
-----END PGP SIGNATURE-----
Merge tag 'vfs-6.12.misc' of gitolite.kernel.org:pub/scm/linux/kernel/git/vfs/vfs
Pull misc vfs updates from Christian Brauner:
"This contains the usual pile of misc updates:
Features:
- Add F_CREATED_QUERY fcntl() that allows userspace to query whether
a file was actually created. Often userspace wants to know whether
an O_CREATE request did actually create a file without using
O_EXCL. The current logic is that to first attempts to open the
file without O_CREAT | O_EXCL and if ENOENT is returned userspace
tries again with both flags. If that succeeds all is well. If it
now reports EEXIST it retries.
That works fairly well but some corner cases make this more
involved. If this operates on a dangling symlink the first openat()
without O_CREAT | O_EXCL will return ENOENT but the second openat()
with O_CREAT | O_EXCL will fail with EEXIST.
The reason is that openat() without O_CREAT | O_EXCL follows the
symlink while O_CREAT | O_EXCL doesn't for security reasons. So
it's not something we can really change unless we add an explicit
opt-in via O_FOLLOW which seems really ugly.
All available workarounds are really nasty (fanotify, bpf lsm etc)
so add a simple fcntl().
- Try an opportunistic lookup for O_CREAT. Today, when opening a file
we'll typically do a fast lookup, but if O_CREAT is set, the kernel
always takes the exclusive inode lock. This was likely done with
the expectation that O_CREAT means that we always expect to do the
create, but that's often not the case. Many programs set O_CREAT
even in scenarios where the file already exists (see related
F_CREATED_QUERY patch motivation above).
The series contained in the pr rearranges the pathwalk-for-open
code to also attempt a fast_lookup in certain O_CREAT cases. If a
positive dentry is found, the inode_lock can be avoided altogether
and it can stay in rcuwalk mode for the last step_into.
- Expose the 64 bit mount id via name_to_handle_at()
Now that we provide a unique 64-bit mount ID interface in statx(2),
we can now provide a race-free way for name_to_handle_at(2) to
provide a file handle and corresponding mount without needing to
worry about racing with /proc/mountinfo parsing or having to open a
file just to do statx(2).
While this is not necessary if you are using AT_EMPTY_PATH and
don't care about an extra statx(2) call, users that pass full paths
into name_to_handle_at(2) need to know which mount the file handle
comes from (to make sure they don't try to open_by_handle_at a file
handle from a different filesystem) and switching to AT_EMPTY_PATH
would require allocating a file for every name_to_handle_at(2) call
- Add a per dentry expire timeout to autofs
There are two fairly well known automounter map formats, the autofs
format and the amd format (more or less System V and Berkley).
Some time ago Linux autofs added an amd map format parser that
implemented a fair amount of the amd functionality. This was done
within the autofs infrastructure and some functionality wasn't
implemented because it either didn't make sense or required extra
kernel changes. The idea was to restrict changes to be within the
existing autofs functionality as much as possible and leave changes
with a wider scope to be considered later.
One of these changes is implementing the amd options:
1) "unmount", expire this mount according to a timeout (same as
the current autofs default).
2) "nounmount", don't expire this mount (same as setting the
autofs timeout to 0 except only for this specific mount) .
3) "utimeout=<seconds>", expire this mount using the specified
timeout (again same as setting the autofs timeout but only for
this mount)
To implement these options per-dentry expire timeouts need to be
implemented for autofs indirect mounts. This is because all map
keys (mounts) for autofs indirect mounts use an expire timeout
stored in the autofs mount super block info. structure and all
indirect mounts use the same expire timeout.
Fixes:
- Fix missing fput for FSCONFIG_SET_FD in autofs
- Use param->file for FSCONFIG_SET_FD in coda
- Delete the 'fs/netfs' proc subtreee when netfs module exits
- Make sure that struct uid_gid_map fits into a single cacheline
- Don't flush in-flight wb switches for superblocks without cgroup
writeback
- Correcting the idmapping mount example in the idmapping
documentation
- Fix a race between evice_inodes() and find_inode() and iput()
- Refine the show_inode_state() macro definition in writeback code
- Prevent dump_mapping() from accessing invalid dentry.d_name.name
- Show actual source for debugfs in /proc/mounts
- Annotate data-race of busy_poll_usecs in eventpoll
- Don't WARN for racy path_noexec check in exec code
- Handle OOM on mnt_warn_timestamp_expiry()
- Fix some spelling in the iomap design documentation
- Fix typo in procfs comment
- Fix typo in fs/namespace.c comment
Cleanups:
- Add the VFS git tree to the MAINTAINERS file
- Move FMODE_UNSIGNED_OFFSET to fop_flags freeing up another f_mode
bit in struct file bringing us to 5 free f_mode bits
- Remove the __I_DIO_WAKEUP bit from i_state flags as we can simplify
the wait mechanism
- Remove the unused path_put_init() helper
- Replace a __u32 with u32 for s_fsnotify_mask as __u32 is uapi
specific
- Replace the unsigned long i_state member with a u32 i_state member
in struct inode freeing up 4 bytes in struct inode. Instead of
using the bit based wait apis we're now using the var event apis
and using the individual bytes of the i_state member to wait on
state changes
- Explain how per-syscall AT_* flags should be allocated
- Use in_group_or_capable() helper to simplify the posix acl mode
update code
- Switch to LIST_HEAD() in fsync_buffers_list() to simplify the code
- Removed comment about d_rcu_to_refcount() as that function doesn't
exist anymore
- Add kernel documentation for lookup_fast()
- Don't re-zero evenpoll fields
- Remove outdated comment after close_fd()
- Fix imprecise wording in comment about the pipe filesystem
- Drop GFP_NOFAIL mode from alloc_page_buffers
- Missing blank line warnings and struct declaration improved in
file_table
- Annotate struct poll_list with __counted_by()
- Remove the unused read parameter in percpu-rwsem
- Remove linux/prefetch.h include from direct-io code
- Use kmemdup_array instead of kmemdup for multiple allocation in
mnt_idmapping code
- Remove unused mnt_cursor_del() declaration
Performance tweaks:
- Dodge smp_mb in break_lease and break_deleg in the common case
- Only read fops once in fops_{get,put}()
- Use RCU in ilookup()
- Elide smp_mb in iversion handling in the common case
- Drop one lock trip in evict()"
* tag 'vfs-6.12.misc' of gitolite.kernel.org:pub/scm/linux/kernel/git/vfs/vfs: (58 commits)
uidgid: make sure we fit into one cacheline
proc: Fix typo in the comment
fs/pipe: Correct imprecise wording in comment
fhandle: expose u64 mount id to name_to_handle_at(2)
uapi: explain how per-syscall AT_* flags should be allocated
fs: drop GFP_NOFAIL mode from alloc_page_buffers
writeback: Refine the show_inode_state() macro definition
fs/inode: Prevent dump_mapping() accessing invalid dentry.d_name.name
mnt_idmapping: Use kmemdup_array instead of kmemdup for multiple allocation
netfs: Delete subtree of 'fs/netfs' when netfs module exits
fs: use LIST_HEAD() to simplify code
inode: make i_state a u32
inode: port __I_LRU_ISOLATING to var event
vfs: fix race between evice_inodes() and find_inode()&iput()
inode: port __I_NEW to var event
inode: port __I_SYNC to var event
fs: reorder i_state bits
fs: add i_state helpers
MAINTAINERS: add the VFS git tree
fs: s/__u32/u32/ for s_fsnotify_mask
...
The runtime constant feature removes all the users of these variables,
allowing the compiler to optimize them away. It's quite difficult to
extract their values from the kernel text, and the memory saved by
removing them is tiny, and it was never the point of this optimization.
Since the dentry_hashtable is a core data structure, it's valuable for
debugging tools to be able to read it easily. For instance, scripts
built on drgn, like the dentrycache script[1], rely on it to be able to
perform diagnostics on the contents of the dcache. Annotate it as used,
so the compiler doesn't discard it.
Link: 3afc56146f/drgn_tools/dentry.py (L325-L355) [1]
Fixes: e3c92e8171 ("runtime constants: add x86 architecture support")
Signed-off-by: Stephen Brennan <stephen.s.brennan@oracle.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
const qualify the struct ctl_table argument in the proc_handler function
signatures. This is a prerequisite to moving the static ctl_table
structs into .rodata data which will ensure that proc_handler function
pointers cannot be modified.
This patch has been generated by the following coccinelle script:
```
virtual patch
@r1@
identifier ctl, write, buffer, lenp, ppos;
identifier func !~ "appldata_(timer|interval)_handler|sched_(rt|rr)_handler|rds_tcp_skbuf_handler|proc_sctp_do_(hmac_alg|rto_min|rto_max|udp_port|alpha_beta|auth|probe_interval)";
@@
int func(
- struct ctl_table *ctl
+ const struct ctl_table *ctl
,int write, void *buffer, size_t *lenp, loff_t *ppos);
@r2@
identifier func, ctl, write, buffer, lenp, ppos;
@@
int func(
- struct ctl_table *ctl
+ const struct ctl_table *ctl
,int write, void *buffer, size_t *lenp, loff_t *ppos)
{ ... }
@r3@
identifier func;
@@
int func(
- struct ctl_table *
+ const struct ctl_table *
,int , void *, size_t *, loff_t *);
@r4@
identifier func, ctl;
@@
int func(
- struct ctl_table *ctl
+ const struct ctl_table *ctl
,int , void *, size_t *, loff_t *);
@r5@
identifier func, write, buffer, lenp, ppos;
@@
int func(
- struct ctl_table *
+ const struct ctl_table *
,int write, void *buffer, size_t *lenp, loff_t *ppos);
```
* Code formatting was adjusted in xfs_sysctl.c to comply with code
conventions. The xfs_stats_clear_proc_handler,
xfs_panic_mask_proc_handler and xfs_deprecated_dointvec_minmax where
adjusted.
* The ctl_table argument in proc_watchdog_common was const qualified.
This is called from a proc_handler itself and is calling back into
another proc_handler, making it necessary to change it as part of the
proc_handler migration.
Co-developed-by: Thomas Weißschuh <linux@weissschuh.net>
Signed-off-by: Thomas Weißschuh <linux@weissschuh.net>
Co-developed-by: Joel Granados <j.granados@samsung.com>
Signed-off-by: Joel Granados <j.granados@samsung.com>
-----BEGIN PGP SIGNATURE-----
iHUEABYKAB0WIQRAhzRXHqcMeLMyaSiRxhvAZXjcogUCZpEF0AAKCRCRxhvAZXjc
oq0TAQDjfTLN75RwKQ34RIFtRun2q+OMfBQtSegtaccqazghyAD/QfmPuZDxB5DL
rsI/5k5O4VupIKrEdIaqvNxmkmDsSAc=
=bf7E
-----END PGP SIGNATURE-----
Merge tag 'vfs-6.11.misc' of git://git.kernel.org/pub/scm/linux/kernel/git/vfs/vfs
Pull misc vfs updates from Christian Brauner:
"Features:
- Support passing NULL along AT_EMPTY_PATH for statx().
NULL paths with any flag value other than AT_EMPTY_PATH go the
usual route and end up with -EFAULT to retain compatibility (Rust
is abusing calls of the sort to detect availability of statx)
This avoids path lookup code, lockref management, memory allocation
and in case of NULL path userspace memory access (which can be
quite expensive with SMAP on x86_64)
- Don't block i_writecount during exec. Remove the
deny_write_access() mechanism for executables
- Relax open_by_handle_at() permissions in specific cases where we
can prove that the caller had sufficient privileges to open a file
- Switch timespec64 fields in struct inode to discrete integers
freeing up 4 bytes
Fixes:
- Fix false positive circular locking warning in hfsplus
- Initialize hfs_inode_info after hfs_alloc_inode() in hfs
- Avoid accidental overflows in vfs_fallocate()
- Don't interrupt fallocate with EINTR in tmpfs to avoid constantly
restarting shmem_fallocate()
- Add missing quote in comment in fs/readdir
Cleanups:
- Don't assign and test in an if statement in mqueue. Move the
assignment out of the if statement
- Reflow the logic in may_create_in_sticky()
- Remove the usage of the deprecated ida_simple_xx() API from procfs
- Reject FSCONFIG_CMD_CREATE_EXCL requets that depend on the new
mount api early
- Rename variables in copy_tree() to make it easier to understand
- Replace WARN(down_read_trylock, ...) abuse with proper asserts in
various places in the VFS
- Get rid of user_path_at_empty() and drop the empty argument from
getname_flags()
- Check for error while copying and no path in one branch in
getname_flags()
- Avoid redundant smp_mb() for THP handling in do_dentry_open()
- Rename parent_ino to d_parent_ino and make it use RCU
- Remove unused header include in fs/readdir
- Export in_group_capable() helper and switch f2fs and fuse over to
it instead of open-coding the logic in both places"
* tag 'vfs-6.11.misc' of git://git.kernel.org/pub/scm/linux/kernel/git/vfs/vfs: (27 commits)
ipc: mqueue: remove assignment from IS_ERR argument
vfs: rename parent_ino to d_parent_ino and make it use RCU
vfs: support statx(..., NULL, AT_EMPTY_PATH, ...)
stat: use vfs_empty_path() helper
fs: new helper vfs_empty_path()
fs: reflow may_create_in_sticky()
vfs: remove redundant smp_mb for thp handling in do_dentry_open
fuse: Use in_group_or_capable() helper
f2fs: Use in_group_or_capable() helper
fs: Export in_group_or_capable()
vfs: reorder checks in may_create_in_sticky
hfs: fix to initialize fields of hfs_inode_info after hfs_alloc_inode()
proc: Remove usage of the deprecated ida_simple_xx() API
hfsplus: fix to avoid false alarm of circular locking
Improve readability of copy_tree
vfs: shave a branch in getname_flags
vfs: retire user_path_at_empty and drop empty arg from getname_flags
vfs: stop using user_path_at_empty in do_readlinkat
tmpfs: don't interrupt fallocate with EINTR
fs: don't block i_writecount during exec
...
Merge runtime constants infrastructure with implementations for x86 and
arm64.
This is one of four branches that came out of me looking at profiles of
my kernel build filesystem load on my 128-core Altra arm64 system, where
pathname walking and the user copies (particularly strncpy_from_user()
for fetching the pathname from user space) is very hot.
This is a very specialized "instruction alternatives" model where the
dentry hash pointer and hash count will be constants for the lifetime of
the kernel, but the allocation are not static but done early during the
kernel boot. In order to avoid the pointer load and dynamic shift, we
just rewrite the constants in the instructions in place.
We can't use the "generic" alternative instructions infrastructure,
because different architectures do it very differently, and it's
actually simpler to just have very specific helpers, with a fallback to
the generic ("old") model of just using variables for architectures that
do not implement the runtime constant patching infrastructure.
Link: https://lore.kernel.org/all/CAHk-=widPe38fUNjUOmX11ByDckaeEo9tN4Eiyke9u1SAtu9sA@mail.gmail.com/
* runtime-constants:
arm64: add 'runtime constant' support
runtime constants: add x86 architecture support
runtime constants: add default dummy infrastructure
vfs: dcache: move hashlen_hash() from callers into d_hash()
-----BEGIN PGP SIGNATURE-----
iHUEABYKAB0WIQRAhzRXHqcMeLMyaSiRxhvAZXjcogUCZo9dYAAKCRCRxhvAZXjc
omYQAP4wELNW5StzljRReC6s/Kzu6IANJQlfFpuGnPIl23iRmwD+Pq433xQqSy5f
uonMBEdxqbOrJM7A6KeHKCyuAKYpNg0=
=zg3n
-----END PGP SIGNATURE-----
Merge tag 'vfs-6.10-rc8.fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/vfs/vfs
Pull vfs fixes from Christian Brauner:
"cachefiles:
- Export an existing and add a new cachefile helper to be used in
filesystems to fix reference count bugs
- Use the newly added fscache_ty_get_volume() helper to get a
reference count on an fscache_volume to handle volumes that are
about to be removed cleanly
- After withdrawing a fscache_cache via FSCACHE_CACHE_IS_WITHDRAWN
wait for all ongoing cookie lookups to complete and for the object
count to reach zero
- Propagate errors from vfs_getxattr() to avoid an infinite loop in
cachefiles_check_volume_xattr() because it keeps seeing ESTALE
- Don't send new requests when an object is dropped by raising
CACHEFILES_ONDEMAND_OJBSTATE_DROPPING
- Cancel all requests for an object that is about to be dropped
- Wait for the ondemand_boject_worker to finish before dropping a
cachefiles object to prevent use-after-free
- Use cyclic allocation for message ids to better handle id recycling
- Add missing lock protection when iterating through the xarray when
polling
netfs:
- Use standard logging helpers for debug logging
VFS:
- Fix potential use-after-free in file locks during
trace_posix_lock_inode(). The tracepoint could fire while another
task raced it and freed the lock that was requested to be traced
- Only increment the nr_dentry_negative counter for dentries that are
present on the superblock LRU. Currently, DCACHE_LRU_LIST list is
used to detect this case. However, the flag is also raised in
combination with DCACHE_SHRINK_LIST to indicate that dentry->d_lru
is used. So checking only DCACHE_LRU_LIST will lead to wrong
nr_dentry_negative count. Fix the check to not count dentries that
are on a shrink related list
Misc:
- hfsplus: fix an uninitialized value issue in copy_name
- minix: fix minixfs_rename with HIGHMEM. It still uses kunmap() even
though we switched it to kmap_local_page() a while ago"
* tag 'vfs-6.10-rc8.fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/vfs/vfs:
minixfs: Fix minixfs_rename with HIGHMEM
hfsplus: fix uninit-value in copy_name
vfs: don't mod negative dentry count when on shrinker list
filelock: fix potential use-after-free in posix_lock_inode
cachefiles: add missing lock protection when polling
cachefiles: cyclic allocation of msg_id to avoid reuse
cachefiles: wait for ondemand_object_worker to finish when dropping object
cachefiles: cancel all requests for the object that is being dropped
cachefiles: stop sending new request when dropping object
cachefiles: propagate errors from vfs_getxattr() to avoid infinite loop
cachefiles: fix slab-use-after-free in cachefiles_withdraw_cookie()
cachefiles: fix slab-use-after-free in fscache_withdraw_volume()
netfs, fscache: export fscache_put_volume() and add fscache_try_get_volume()
netfs: Switch debug logging to pr_debug()
The nr_dentry_negative counter is intended to only account negative
dentries that are present on the superblock LRU. Therefore, the LRU
add, remove and isolate helpers modify the counter based on whether
the dentry is negative, but the shrinker list related helpers do not
modify the counter, and the paths that change a dentry between
positive and negative only do so if DCACHE_LRU_LIST is set.
The problem with this is that a dentry on a shrinker list still has
DCACHE_LRU_LIST set to indicate ->d_lru is in use. The additional
DCACHE_SHRINK_LIST flag denotes whether the dentry is on LRU or a
shrink related list. Therefore if a relevant operation (i.e. unlink)
occurs while a dentry is present on a shrinker list, and the
associated codepath only checks for DCACHE_LRU_LIST, then it is
technically possible to modify the negative dentry count for a
dentry that is off the LRU. Since the shrinker list related helpers
do not modify the negative dentry count (because non-LRU dentries
should not be included in the count) when the dentry is ultimately
removed from the shrinker list, this can cause the negative dentry
count to become permanently inaccurate.
This problem can be reproduced via a heavy file create/unlink vs.
drop_caches workload. On an 80xcpu system, I start 80 tasks each
running a 1k file create/delete loop, and one task spinning on
drop_caches. After 10 minutes or so of runtime, the idle/clean cache
negative dentry count increases from somewhere in the range of 5-10
entries to several hundred (and increasingly grows beyond
nr_dentry_unused).
Tweak the logic in the paths that turn a dentry negative or positive
to filter out the case where the dentry is present on a shrink
related list. This allows the above workload to maintain an accurate
negative dentry count.
Fixes: af0c9af1b3 ("fs/dcache: Track & report number of negative dentries")
Signed-off-by: Brian Foster <bfoster@redhat.com>
Link: https://lore.kernel.org/r/20240703121301.247680-1-bfoster@redhat.com
Acked-by: Ian Kent <ikent@redhat.com>
Reviewed-by: Josef Bacik <josef@toxicpanda.com>
Reviewed-by: Waiman Long <longman@redhat.com>
Signed-off-by: Christian Brauner <brauner@kernel.org>
Jan reported that 'cd ..' may take a long time in deep directory
hierarchies under a bind-mount. If concurrent renames happen it is
possible to livelock in is_subdir() because it will keep retrying.
Change is_subdir() from simply retrying over and over to retry once and
then acquire the rename lock to handle deep ancestor chains better. The
list of alternatives to this approach were less then pleasant. Change
the scope of rcu lock to cover the whole walk while at it.
A big thanks to Jan and Linus. Both Jan and Linus had proposed
effectively the same thing just that one version ended up being slightly
more elegant.
Reported-by: Jan Kara <jack@suse.cz>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Christian Brauner <brauner@kernel.org>
The routine is used by procfs through dir_emit_dots.
The combined RCU and lock fallback implementation is too big for an
inline. Given that the routine takes a dentry argument fs/dcache.c seems
like the place to put it in.
Signed-off-by: Mateusz Guzik <mjguzik@gmail.com>
Link: https://lore.kernel.org/r/20240627161152.802567-1-mjguzik@gmail.com
Signed-off-by: Christian Brauner <brauner@kernel.org>
This adds the initial dummy support for 'runtime constants' for when
an architecture doesn't actually support an implementation of fixing
up said runtime constants.
This ends up being the fallback to just using the variables as regular
__ro_after_init variables, and changes the dcache d_hash() function to
use this model.
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Both __d_lookup_rcu() and __d_lookup_rcu_op_compare() have the full
'name_hash' value of the qstr that they want to look up, and mask it off
to just the low 32-bit hash before calling down to d_hash().
Other callers just load the 32-bit hash and pass it as the argument.
If we move the masking into d_hash() itself, it simplifies the two
callers that currently do the masking, and is a no-op for the other
cases. It doesn't actually change the generated code since the compiler
will inline d_hash() and see that the end result is the same.
[ Technically, since the parse tree changes, the code generation may not
be 100% the same, and for me on x86-64, this does result in gcc
switching the operands around for one 'cmpl' instruction. So not
necessarily the exact same code generation, but equivalent ]
However, this does encapsulate the 'd_hash()' operation more, and makes
the shift operation in particular be a "shift 32 bits right, return full
word". Which matches the instruction semantics on both x86-64 and arm64
better, since a 32-bit shift will clear the upper bits.
That makes the next step of introducing a "shift by runtime constant"
more obvious and generates the shift with no extraneous type masking.
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
This reverts commit 681ce86235.
We gave it a try, but it turns out the kernel test robot did in fact
find performance regressions for it, so we'll have to look at the more
involved alternative fixes for Yafang Shao's Elasticsearch load issue.
There were several alternatives discussed, they just weren't as simple
as this first attempt.
The report is of a -7.4% regression of filebench.sum_operations/s, which
appears significant enough to trigger my "this patch may get reverted if
somebody finds a performance regression on some other load" rule.
So it's still the case that we should end up deleting dentries more
aggressively - or just be better at pruning them later - but it needs a
bit more finesse than this simple thing.
Link: https://lore.kernel.org/all/202405291318.4dfbb352-oliver.sang@intel.com/
Cc: Yafang Shao <laoar.shao@gmail.com>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Cc: Christian Brauner <brauner@kernel.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Our applications, built on Elasticsearch[0], frequently create and
delete files. These applications operate within containers, some with a
memory limit exceeding 100GB. Over prolonged periods, the accumulation
of negative dentries within these containers can amount to tens of
gigabytes.
Upon container exit, directories are deleted. However, due to the
numerous associated dentries, this process can be time-consuming. Our
users have expressed frustration with this prolonged exit duration,
which constitutes our first issue.
Simultaneously, other processes may attempt to access the parent
directory of the Elasticsearch directories. Since the task responsible
for deleting the dentries holds the inode lock, processes attempting
directory lookup experience significant delays. This issue, our second
problem, is easily demonstrated:
- Task 1 generates negative dentries:
$ pwd
~/test
$ mkdir es && cd es/ && ./create_and_delete_files.sh
[ After generating tens of GB dentries ]
$ cd ~/test && rm -rf es
[ It will take a long duration to finish ]
- Task 2 attempts to lookup the 'test/' directory
$ pwd
~/test
$ ls
The 'ls' command in Task 2 experiences prolonged execution as Task 1
is deleting the dentries.
We've devised a solution to address both issues by deleting associated
dentry when removing a file. Interestingly, we've noted that a similar
patch was proposed years ago[1], although it was rejected citing the
absence of tangible issues caused by negative dentries. Given our
current challenges, we're resubmitting the proposal. All relevant
stakeholders from previous discussions have been included for reference.
Some alternative solutions are also under discussion[2][3], such as
shrinking child dentries outside of the parent inode lock or even
asynchronously shrinking child dentries. However, given the
straightforward nature of the current solution, I believe this approach
is still necessary.
[ NOTE! This is a pretty fundamental change in how we deal with
unlinking dentries, and it doesn't change the fact that you can have
lots of negative dentries from just doing negative lookups.
But the kernel test robot is at least initially happy with this from a
performance angle, so I'm applying this ASAP just to get more testing
and as a "known fix for an issue people hit in real life".
Put another way: we should still look at the alternatives, and this
patch may get reverted if somebody finds a performance regression on
some other load. - Linus ]
Suggested-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Yafang Shao <laoar.shao@gmail.com>
Link: https://github.com/elastic/elasticsearch [0]
Link: https://patchwork.kernel.org/project/linux-fsdevel/patch/1502099673-31620-1-git-send-email-wangkai86@huawei.com [1]
Link: https://lore.kernel.org/linux-fsdevel/20240511200240.6354-2-torvalds@linux-foundation.org/ [2]
Link: https://lore.kernel.org/linux-fsdevel/CAHk-=wjEMf8Du4UFzxuToGDnF3yLaMcrYeyNAaH1NJWa6fwcNQ@mail.gmail.com/ [3]
Cc: Al Viro <viro@zeniv.linux.org.uk>
Cc: Christian Brauner <brauner@kernel.org>
Cc: Jan Kara <jack@suse.cz>
Cc: Waiman Long <longman@redhat.com>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Wangkai <wangkai86@huawei.com>
Cc: Colin Walters <walters@verbum.org>
Tested-by: kernel test robot <oliver.sang@intel.com>
Link: https://lore.kernel.org/all/202405221518.ecea2810-oliver.sang@intel.com/
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Currently, the __d_clear_type_and_inode() writes the value flags to
dentry->d_flags, then immediately re-reads it in order to use it in a if
statement. This re-read is useless because no other update to
dentry->d_flags can occur at this point.
This commit therefore re-use flags in the if statement instead of
re-reading dentry->d_flags.
Signed-off-by: linke li <lilinke99@qq.com>
Link: https://lore.kernel.org/r/tencent_5E187BD0A61BA28605E85405F15228254D0A@qq.com
Reviewed-by: Jan Kara <jack@suse.cz>
Signed-off-by: Christian Brauner <brauner@kernel.org>
-----BEGIN PGP SIGNATURE-----
iHUEABYKAB0WIQRAhzRXHqcMeLMyaSiRxhvAZXjcogUCZem3wQAKCRCRxhvAZXjc
otRMAQDeo8qsuuIAcS2KUicKqZR5yMVvrY9r4sQzf7YRcJo5HQD+NQXkKwQuv1VO
OUeScsic/+I+136AgdjWnlEYO5dp0go=
=4WKU
-----END PGP SIGNATURE-----
Merge tag 'vfs-6.9.misc' of git://git.kernel.org/pub/scm/linux/kernel/git/vfs/vfs
Pull misc vfs updates from Christian Brauner:
"Misc features, cleanups, and fixes for vfs and individual filesystems.
Features:
- Support idmapped mounts for hugetlbfs.
- Add RWF_NOAPPEND flag for pwritev2(). This allows us to fix a bug
where the passed offset is ignored if the file is O_APPEND. The new
flag allows a caller to enforce that the offset is honored to
conform to posix even if the file was opened in append mode.
- Move i_mmap_rwsem in struct address_space to avoid false sharing
between i_mmap and i_mmap_rwsem.
- Convert efs, qnx4, and coda to use the new mount api.
- Add a generic is_dot_dotdot() helper that's used by various
filesystems and the VFS code instead of open-coding it multiple
times.
- Recently we've added stable offsets which allows stable ordering
when iterating directories exported through NFS on e.g., tmpfs
filesystems. Originally an xarray was used for the offset map but
that caused slab fragmentation issues over time. This switches the
offset map to the maple tree which has a dense mode that handles
this scenario a lot better. Includes tests.
- Finally merge the case-insensitive improvement series Gabriel has
been working on for a long time. This cleanly propagates case
insensitive operations through ->s_d_op which in turn allows us to
remove the quite ugly generic_set_encrypted_ci_d_ops() operations.
It also improves performance by trying a case-sensitive comparison
first and then fallback to case-insensitive lookup if that fails.
This also fixes a bug where overlayfs would be able to be mounted
over a case insensitive directory which would lead to all sort of
odd behaviors.
Cleanups:
- Make file_dentry() a simple accessor now that ->d_real() is
simplified because of the backing file work we did the last two
cycles.
- Use the dedicated file_mnt_idmap helper in ntfs3.
- Use smp_load_acquire/store_release() in the i_size_read/write
helpers and thus remove the hack to handle i_size reads in the
filemap code.
- The SLAB_MEM_SPREAD is a nop now. Remove it from various places in
fs/
- It's no longer necessary to perform a second built-in initramfs
unpack call because we retain the contents of the previous
extraction. Remove it.
- Now that we have removed various allocators kfree_rcu() always
works with kmem caches and kmalloc(). So simplify various places
that only use an rcu callback in order to handle the kmem cache
case.
- Convert the pipe code to use a lockdep comparison function instead
of open-coding the nesting making lockdep validation easier.
- Move code into fs-writeback.c that was located in a header but can
be made static as it's only used in that one file.
- Rewrite the alignment checking iterators for iovec and bvec to be
easier to read, and also significantly more compact in terms of
generated code. This saves 270 bytes of text on x86-64 (with
clang-18) and 224 bytes on arm64 (with gcc-13). In profiles it also
saves a bit of time for the same workload.
- Switch various places to use KMEM_CACHE instead of
kmem_cache_create().
- Use inode_set_ctime_to_ts() in inode_set_ctime_current()
- Use kzalloc() in name_to_handle_at() to avoid kernel infoleak.
- Various smaller cleanups for eventfds.
Fixes:
- Fix various comments and typos, and unneeded initializations.
- Fix stack allocation hack for clang in the select code.
- Improve dump_mapping() debug code on a best-effort basis.
- Fix build errors in various selftests.
- Avoid wrap-around instrumentation in various places.
- Don't allow user namespaces without an idmapping to be used for
idmapped mounts.
- Fix sysv sb_read() call.
- Fix fallback implementation of the get_name() export operation"
* tag 'vfs-6.9.misc' of git://git.kernel.org/pub/scm/linux/kernel/git/vfs/vfs: (70 commits)
hugetlbfs: support idmapped mounts
qnx4: convert qnx4 to use the new mount api
fs: use inode_set_ctime_to_ts to set inode ctime to current time
libfs: Drop generic_set_encrypted_ci_d_ops
ubifs: Configure dentry operations at dentry-creation time
f2fs: Configure dentry operations at dentry-creation time
ext4: Configure dentry operations at dentry-creation time
libfs: Add helper to choose dentry operations at mount-time
libfs: Merge encrypted_ci_dentry_ops and ci_dentry_ops
fscrypt: Drop d_revalidate once the key is added
fscrypt: Drop d_revalidate for valid dentries during lookup
fscrypt: Factor out a helper to configure the lookup dentry
ovl: Always reject mounting over case-insensitive directories
libfs: Attempt exact-match comparison first during casefolded lookup
efs: remove SLAB_MEM_SPREAD flag usage
jfs: remove SLAB_MEM_SPREAD flag usage
minix: remove SLAB_MEM_SPREAD flag usage
openpromfs: remove SLAB_MEM_SPREAD flag usage
proc: remove SLAB_MEM_SPREAD flag usage
qnx6: remove SLAB_MEM_SPREAD flag usage
...
This reverts commit 5785160732.
Unfortunately, while we only call that thing once, the callback
*can* be called more than once for the same dentry - all it
takes is rename_lock being touched while we are in d_walk().
For now let's revert it.
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
change of locking rules for __dentry_kill(), regularized refcounting
rules in that area, assorted cleanups and removal of weird corner
cases (e.g. now ->d_iput() on child is always called before the parent
might hit __dentry_kill(), etc.)
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
-----BEGIN PGP SIGNATURE-----
iHUEABYIAB0WIQQqUNBr3gm4hGXdBJlZ7Krx/gZQ6wUCZZ+sQQAKCRBZ7Krx/gZQ
6ybjAQDM5jiS93IUzfHjCWq0nVBX5YGbDAkZOeqxbmIdQb+2UAEA6elP5r0fBBcA
seo3bry4DirQMDaA/Cjh4+8r71YSOQs=
=7+Hk
-----END PGP SIGNATURE-----
Merge tag 'pull-dcache' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/vfs
Pull dcache updates from Al Viro:
"Change of locking rules for __dentry_kill(), regularized refcounting
rules in that area, assorted cleanups and removal of weird corner
cases (e.g. now ->d_iput() on child is always called before the parent
might hit __dentry_kill(), etc)"
* tag 'pull-dcache' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/vfs: (40 commits)
dcache: remove unnecessary NULL check in dget_dlock()
kill DCACHE_MAY_FREE
__d_unalias() doesn't use inode argument
d_alloc_parallel(): in-lookup hash insertion doesn't need an RCU variant
get rid of DCACHE_GENOCIDE
d_genocide(): move the extern into fs/internal.h
simple_fill_super(): don't bother with d_genocide() on failure
nsfs: use d_make_root()
d_alloc_pseudo(): move setting ->d_op there from the (sole) caller
kill d_instantate_anon(), fold __d_instantiate_anon() into remaining caller
retain_dentry(): introduce a trimmed-down lockless variant
__dentry_kill(): new locking scheme
d_prune_aliases(): use a shrink list
switch select_collect{,2}() to use of to_shrink_list()
to_shrink_list(): call only if refcount is 0
fold dentry_kill() into dput()
don't try to cut corners in shrink_lock_dentry()
fold the call of retain_dentry() into fast_dput()
Call retain_dentry() with refcount 0
dentry_kill(): don't bother with retain_dentry() on slow path
...
To help make the move of sysctls out of kernel/sysctl.c not incur a size
penalty sysctl has been changed to allow us to not require the sentinel, the
final empty element on the sysctl array. Joel Granados has been doing all this
work. On the v6.6 kernel we got the major infrastructure changes required to
support this. For v6.7 we had all arch/ and drivers/ modified to remove
the sentinel. For v6.8-rc1 we get a few more updates for fs/ directory only.
The kernel/ directory is left but we'll save that for v6.9-rc1 as those patches
are still being reviewed. After that we then can expect also the removal of the
no longer needed check for procname == NULL.
Let us recap the purpose of this work:
- this helps reduce the overall build time size of the kernel and run time
memory consumed by the kernel by about ~64 bytes per array
- the extra 64-byte penalty is no longer inncurred now when we move sysctls
out from kernel/sysctl.c to their own files
Thomas Weißschuh also sent a few cleanups, for v6.9-rc1 we expect to see further
work by Thomas Weißschuh with the constificatin of the struct ctl_table.
Due to Joel Granados's work, and to help bring in new blood, I have suggested
for him to become a maintainer and he's accepted. So for v6.9-rc1 I look forward
to seeing him sent you a pull request for further sysctl changes. This also
removes Iurii Zaikin as a maintainer as he has moved on to other projects and
has had no time to help at all.
-----BEGIN PGP SIGNATURE-----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=2qBt
-----END PGP SIGNATURE-----
Merge tag 'sysctl-6.8-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/mcgrof/linux
Pull sysctl updates from Luis Chamberlain:
"To help make the move of sysctls out of kernel/sysctl.c not incur a
size penalty sysctl has been changed to allow us to not require the
sentinel, the final empty element on the sysctl array. Joel Granados
has been doing all this work.
In the v6.6 kernel we got the major infrastructure changes required to
support this. For v6.7 we had all arch/ and drivers/ modified to
remove the sentinel. For v6.8-rc1 we get a few more updates for fs/
directory only.
The kernel/ directory is left but we'll save that for v6.9-rc1 as
those patches are still being reviewed. After that we then can expect
also the removal of the no longer needed check for procname == NULL.
Let us recap the purpose of this work:
- this helps reduce the overall build time size of the kernel and run
time memory consumed by the kernel by about ~64 bytes per array
- the extra 64-byte penalty is no longer inncurred now when we move
sysctls out from kernel/sysctl.c to their own files
Thomas Weißschuh also sent a few cleanups, for v6.9-rc1 we expect to
see further work by Thomas Weißschuh with the constificatin of the
struct ctl_table.
Due to Joel Granados's work, and to help bring in new blood, I have
suggested for him to become a maintainer and he's accepted. So for
v6.9-rc1 I look forward to seeing him sent you a pull request for
further sysctl changes. This also removes Iurii Zaikin as a maintainer
as he has moved on to other projects and has had no time to help at
all"
* tag 'sysctl-6.8-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/mcgrof/linux:
sysctl: remove struct ctl_path
sysctl: delete unused define SYSCTL_PERM_EMPTY_DIR
coda: Remove the now superfluous sentinel elements from ctl_table array
sysctl: Remove the now superfluous sentinel elements from ctl_table array
fs: Remove the now superfluous sentinel elements from ctl_table array
cachefiles: Remove the now superfluous sentinel element from ctl_table array
sysclt: Clarify the results of selftest run
sysctl: Add a selftest for handling empty dirs
sysctl: Fix out of bounds access for empty sysctl registers
MAINTAINERS: Add Joel Granados as co-maintainer for proc sysctl
MAINTAINERS: remove Iurii Zaikin from proc sysctl
This commit comes at the tail end of a greater effort to remove the
empty elements at the end of the ctl_table arrays (sentinels) which
will reduce the overall build time size of the kernel and run time
memory bloat by ~64 bytes per sentinel (further information Link :
https://lore.kernel.org/all/ZO5Yx5JFogGi%2FcBo@bombadil.infradead.org/)
Remove sentinel elements ctl_table struct. Special attention was placed in
making sure that an empty directory for fs/verity was created when
CONFIG_FS_VERITY_BUILTIN_SIGNATURES is not defined. In this case we use the
register sysctl call that expects a size.
Signed-off-by: Joel Granados <j.granados@samsung.com>
Reviewed-by: Jan Kara <jack@suse.cz>
Reviewed-by: "Darrick J. Wong" <djwong@kernel.org>
Acked-by: Christian Brauner <brauner@kernel.org>
Signed-off-by: Luis Chamberlain <mcgrof@kernel.org>
Patch series "workload-specific and memory pressure-driven zswap
writeback", v8.
There are currently several issues with zswap writeback:
1. There is only a single global LRU for zswap, making it impossible to
perform worload-specific shrinking - an memcg under memory pressure
cannot determine which pages in the pool it owns, and often ends up
writing pages from other memcgs. This issue has been previously
observed in practice and mitigated by simply disabling
memcg-initiated shrinking:
https://lore.kernel.org/all/20230530232435.3097106-1-nphamcs@gmail.com/T/#u
But this solution leaves a lot to be desired, as we still do not
have an avenue for an memcg to free up its own memory locked up in
the zswap pool.
2. We only shrink the zswap pool when the user-defined limit is hit.
This means that if we set the limit too high, cold data that are
unlikely to be used again will reside in the pool, wasting precious
memory. It is hard to predict how much zswap space will be needed
ahead of time, as this depends on the workload (specifically, on
factors such as memory access patterns and compressibility of the
memory pages).
This patch series solves these issues by separating the global zswap LRU
into per-memcg and per-NUMA LRUs, and performs workload-specific (i.e
memcg- and NUMA-aware) zswap writeback under memory pressure. The new
shrinker does not have any parameter that must be tuned by the user, and
can be opted in or out on a per-memcg basis.
As a proof of concept, we ran the following synthetic benchmark: build the
linux kernel in a memory-limited cgroup, and allocate some cold data in
tmpfs to see if the shrinker could write them out and improved the overall
performance. Depending on the amount of cold data generated, we observe
from 14% to 35% reduction in kernel CPU time used in the kernel builds.
This patch (of 6):
The interface of list_lru is based on the assumption that the list node
and the data it represents belong to the same allocated on the correct
node/memcg. While this assumption is valid for existing slab objects LRU
such as dentries and inodes, it is undocumented, and rather inflexible for
certain potential list_lru users (such as the upcoming zswap shrinker and
the THP shrinker). It has caused us a lot of issues during our
development.
This patch changes list_lru interface so that the caller must explicitly
specify numa node and memcg when adding and removing objects. The old
list_lru_add() and list_lru_del() are renamed to list_lru_add_obj() and
list_lru_del_obj(), respectively.
It also extends the list_lru API with a new function, list_lru_putback,
which undoes a previous list_lru_isolate call. Unlike list_lru_add, it
does not increment the LRU node count (as list_lru_isolate does not
decrement the node count). list_lru_putback also allows for explicit
memcg and NUMA node selection.
Link: https://lkml.kernel.org/r/20231130194023.4102148-1-nphamcs@gmail.com
Link: https://lkml.kernel.org/r/20231130194023.4102148-2-nphamcs@gmail.com
Signed-off-by: Nhat Pham <nphamcs@gmail.com>
Suggested-by: Johannes Weiner <hannes@cmpxchg.org>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Tested-by: Bagas Sanjaya <bagasdotme@gmail.com>
Cc: Chris Li <chrisl@kernel.org>
Cc: Dan Streetman <ddstreet@ieee.org>
Cc: Domenico Cerasuolo <cerasuolodomenico@gmail.com>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Muchun Song <muchun.song@linux.dev>
Cc: Roman Gushchin <roman.gushchin@linux.dev>
Cc: Seth Jennings <sjenning@redhat.com>
Cc: Shakeel Butt <shakeelb@google.com>
Cc: Shuah Khan <shuah@kernel.org>
Cc: Vitaly Wool <vitaly.wool@konsulko.com>
Cc: Yosry Ahmed <yosryahmed@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
dget_dlock() requires dentry->d_lock to be held when called, yet
contains a NULL check for dentry.
An audit of all calls to dget_dlock() shows that it is never called
with a NULL pointer (as spin_lock()/spin_unlock() would crash in these
cases):
$ git grep -W '\<dget_dlock\>'
arch/powerpc/platforms/cell/spufs/inode.c- spin_lock(&dentry->d_lock);
arch/powerpc/platforms/cell/spufs/inode.c- if (simple_positive(dentry)) {
arch/powerpc/platforms/cell/spufs/inode.c: dget_dlock(dentry);
fs/autofs/expire.c- spin_lock_nested(&child->d_lock, DENTRY_D_LOCK_NESTED);
fs/autofs/expire.c- if (simple_positive(child)) {
fs/autofs/expire.c: dget_dlock(child);
fs/autofs/root.c: dget_dlock(active);
fs/autofs/root.c- spin_unlock(&active->d_lock);
fs/autofs/root.c: dget_dlock(expiring);
fs/autofs/root.c- spin_unlock(&expiring->d_lock);
fs/ceph/dir.c- if (!spin_trylock(&dentry->d_lock))
fs/ceph/dir.c- continue;
[...]
fs/ceph/dir.c: dget_dlock(dentry);
fs/ceph/mds_client.c- spin_lock(&alias->d_lock);
[...]
fs/ceph/mds_client.c: dn = dget_dlock(alias);
fs/configfs/inode.c- spin_lock(&dentry->d_lock);
fs/configfs/inode.c- if (simple_positive(dentry)) {
fs/configfs/inode.c: dget_dlock(dentry);
fs/libfs.c: found = dget_dlock(d);
fs/libfs.c- spin_unlock(&d->d_lock);
fs/libfs.c: found = dget_dlock(child);
fs/libfs.c- spin_unlock(&child->d_lock);
fs/libfs.c: child = dget_dlock(d);
fs/libfs.c- spin_unlock(&d->d_lock);
fs/ocfs2/dcache.c: dget_dlock(dentry);
fs/ocfs2/dcache.c- spin_unlock(&dentry->d_lock);
include/linux/dcache.h:static inline struct dentry *dget_dlock(struct dentry *dentry)
After taking out the NULL check, dget_dlock() becomes almost identical
to __dget_dlock(); the only difference is that dget_dlock() returns the
dentry that was passed in. These are static inline helpers, so we can
rely on the compiler to discard unused return values. We can therefore
also remove __dget_dlock() and replace calls to it by dget_dlock().
Also fix up and improve the kerneldoc comments while we're at it.
Al Viro pointed out that we can also clean up some of the callers to
make use of the returned value and provided a bit more info for the
kerneldoc.
While preparing v2 I also noticed that the tabs used in the kerneldoc
comments were causing the kerneldoc to get parsed incorrectly so I also
fixed this up (including for d_unhashed, which is otherwise unrelated).
Testing: x86 defconfig build + boot; make htmldocs for the kerneldoc
warning. objdump shows there are code generation changes.
Link: https://lore.kernel.org/all/20231022164520.915013-1-vegard.nossum@oracle.com/
Cc: Alexander Viro <viro@zeniv.linux.org.uk>
Cc: Christian Brauner <brauner@kernel.org>
Cc: linux-fsdevel@vger.kernel.org
Cc: Nick Piggin <npiggin@kernel.dk>
Cc: Waiman Long <Waiman.Long@hp.com>
Cc: linux-doc@vger.kernel.org
Signed-off-by: Vegard Nossum <vegard.nossum@oracle.com>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
With the new ordering in __dentry_kill() it has become redundant -
it's set if and only if both DCACHE_DENTRY_KILLED and DCACHE_SHRINK_LIST
are set.
We set it in __dentry_kill(), after having set DCACHE_DENTRY_KILLED
with the only condition being that DCACHE_SHRINK_LIST is there;
all of that is done without dropping ->d_lock and the only place
that checks that flag (shrink_dentry_list()) does so under ->d_lock,
after having found the victim on its shrink list. Since DCACHE_SHRINK_LIST
is set only when placing dentry into shrink list and removed only by
shrink_dentry_list() itself, a check for DCACHE_DENTRY_KILLED in
there would be equivalent to check for DCACHE_MAY_FREE.
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
We only search in the damn thing under hlist_bl_lock(); RCU variant
of insertion was, IIRC, pretty much cargo-culted - mea culpa...
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
now that the only user of d_instantiate_anon() is gone...
[braino fix folded - kudos to Dan Carpenter]
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
fast_dput() contains a small piece of code, preceded by scary
comments about 5 times longer than it. What is actually done there is
a trimmed-down subset of retain_dentry() - in some situations we can
tell that retain_dentry() would have returned true without ever needing
->d_lock and that's what that code checks. If these checks come true
fast_dput() can declare that we are done, without bothering with ->d_lock;
otherwise it has to take the lock and do full variant of retain_dentry()
checks.
Trimmed-down variant of the checks is hard to follow and
it's asking for trouble - if we ever decide to change the rules in
retain_dentry(), we'll have to remember to update that code. It turns
out that an equivalent variant of these checks more obviously parallel
to retain_dentry() is not just possible, but easy to unify with
retain_dentry() itself, passing it a new boolean argument ('locked')
to distinguish between the full semantics and trimmed down one.
Note that in lockless case true is returned only when locked
variant would have returned true without ever needing the lock; false
means "punt to the locking path and recheck there".
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Currently we enter __dentry_kill() with parent (along with the victim
dentry and victim's inode) held locked. Then we
mark dentry refcount as dead
call ->d_prune()
remove dentry from hash
remove it from the parent's list of children
unlock the parent, don't need it from that point on
detach dentry from inode, unlock dentry and drop the inode
(via ->d_iput())
call ->d_release()
regain the lock on dentry
check if it's on a shrink list (in which case freeing its empty husk
has to be left to shrink_dentry_list()) or not (in which case we can free it
ourselves). In the former case, mark it as an empty husk, so that
shrink_dentry_list() would know it can free the sucker.
drop the lock on dentry
... and usually the caller proceeds to drop a reference on the parent,
possibly retaking the lock on it.
That is painful for a bunch of reasons, starting with the need to take locks
out of order, but not limited to that - the parent of positive dentry can
change if we drop its ->d_lock, so getting these locks has to be done with
care. Moreover, as soon as dentry is out of the parent's list of children,
shrink_dcache_for_umount() won't see it anymore, making it appear as if
the parent is inexplicably busy. We do work around that by having
shrink_dentry_list() decrement the parent's refcount first and put it on
shrink list to be evicted once we are done with __dentry_kill() of child,
but that may in some cases lead to ->d_iput() on child called after the
parent got killed. That doesn't happen in cases where in-tree ->d_iput()
instances might want to look at the parent, but that's brittle as hell.
Solution: do removal from the parent's list of children in the very
end of __dentry_kill(). As the result, the callers do not need to
lock the parent and by the time we really need the parent locked,
dentry is negative and is guaranteed not to be moved around.
It does mean that ->d_prune() will be called with parent not locked.
It also means that we might see dentries in process of being torn
down while going through the parent's list of children; those dentries
will be unhashed, negative and with refcount marked dead. In practice,
that's enough for in-tree code that looks through the list of children
to do the right thing as-is. Out-of-tree code might need to be adjusted.
Calling conventions: __dentry_kill(dentry) is called with dentry->d_lock
held, along with ->i_lock of its inode (if any). It either returns
the parent (locked, with refcount decremented to 0) or NULL (if there'd
been no parent or if refcount decrement for parent hadn't reached 0).
lock_for_kill() is adjusted for new requirements - it doesn't touch
the parent's ->d_lock at all.
Callers adjusted. Note that for dput() we don't need to bother with
fast_dput() for the parent - we just need to check retain_dentry()
for it, since its ->d_lock is still held since the moment when
__dentry_kill() had taken it to remove the victim from the list of
children.
The kludge with early decrement of parent's refcount in
shrink_dentry_list() is no longer needed - shrink_dcache_for_umount()
sees the half-killed dentries in the list of children for as long
as they are pinning the parent. They are easily recognized and
accounted for by select_collect(), so we know we are not done yet.
As the result, we always have the expected ordering for ->d_iput()/->d_release()
vs. __dentry_kill() of the parent, no exceptions. Moreover, the current
rules for shrink lists (one must make sure that shrink_dcache_for_umount()
won't happen while any dentries from the superblock in question are on
any shrink lists) are gone - shrink_dcache_for_umount() will do the
right thing in all cases, taking such dentries out. Their empty
husks (memory occupied by struct dentry itself + its external name,
if any) will remain on the shrink lists, but they are no obstacles
to filesystem shutdown. And such husks will get freed as soon as
shrink_dentry_list() of the list they are on gets to them.
Reviewed-by: Christian Brauner <brauner@kernel.org>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Instead of dropping aliases one by one, restarting, etc., just
collect them into a shrink list and kill them off in one pass.
We don't really need the restarts - one alias can't pin another
(directory has only one alias, and couldn't be its own ancestor
anyway), so collecting everything that is not busy and taking it
out would take care of everything evictable that had been there
as we entered the function. And new aliases added while we'd
been dropping old ones could just as easily have appeared right
as we return to caller...
Reviewed-by: Christian Brauner <brauner@kernel.org>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
The only thing it does if refcount is not zero is d_lru_del(); no
point, IMO, seeing that plain dput() does nothing of that sort...
Note that 2 of 3 current callers are guaranteed that refcount is 0.
Acked-by: Christian Brauner <brauner@kernel.org>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
That is to say, do *not* treat the ->d_inode or ->d_parent changes
as "it's hard, return false; somebody must have grabbed it, so
even if has zero refcount, we don't need to bother killing it -
final dput() from whoever grabbed it would've done everything".
First of all, that is not guaranteed. It might have been dropped
by shrink_kill() handling of victim's parent, which would've found
it already on a shrink list (ours) and decided that they don't need
to put it on their shrink list.
What's more, dentry_kill() is doing pretty much the same thing,
cutting its own set of corners (it assumes that dentry can't
go from positive to negative, so its inode can change but only once
and only in one direction).
Doing that right allows to get rid of that not-quite-duplication
and removes the only reason for re-incrementing refcount before
the call of dentry_kill().
Replacement is called lock_for_kill(); called under rcu_read_lock
and with ->d_lock held. If it returns false, dentry has non-zero
refcount and the same locks are held. If it returns true,
dentry has zero refcount and all locks required by __dentry_kill()
are taken.
Part of __lock_parent() had been lifted into lock_parent() to
allow its reuse. Now it's called with rcu_read_lock already
held and dentry already unlocked.
Note that this is not the final change - locking requirements for
__dentry_kill() are going to change later in the series and the
set of locks taken by lock_for_kill() will be adjusted. Both
lock_parent() and __lock_parent() will be gone once that happens.
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Calls of retain_dentry() happen immediately after getting false
from fast_dput() and getting true from retain_dentry() is
treated the same way as non-zero refcount would be treated by
fast_dput() - unlock dentry and bugger off.
Doing that in fast_dput() itself is simpler.
Reviewed-by: Christian Brauner <brauner@kernel.org>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Instead of bumping it from 0 to 1, calling retain_dentry(), then
decrementing it back to 0 (with ->d_lock held all the way through),
just leave refcount at 0 through all of that.
It will have a visible effect for ->d_delete() - now it can be
called with refcount 0 instead of 1 and it can no longer play
silly buggers with dropping/regaining ->d_lock. Not that any
in-tree instances tried to (it's pretty hard to get right).
Any out-of-tree ones will have to adjust (assuming they need any
changes).
Note that we do not need to extend rcu-critical area here - we have
verified that refcount is non-negative after having grabbed ->d_lock,
so nobody will be able to free dentry until they get into __dentry_kill(),
which won't happen until they manage to grab ->d_lock.
Reviewed-by: Christian Brauner <brauner@kernel.org>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
We have already checked it and dentry used to look not worthy
of keeping. The only hard obstacle to evicting dentry is
non-zero refcount; everything else is advisory - e.g. memory
pressure could evict any dentry found with refcount zero.
On the slow path in dentry_kill() we had dropped and regained
->d_lock; we must recheck the refcount, but everything else
is not worth bothering with.
Note that filesystem can not count upon ->d_delete() being
called for dentry - not even once. Again, memory pressure
(as well as d_prune_aliases(), or attempted rmdir() of ancestor,
or...) will not call ->d_delete() at all.
So from the correctness point of view we are fine doing the
check only once. And it makes things simpler down the road.
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Currently we call it with refcount equal to 1 when called from
dentry_kill(); all other callers have it equal to 0.
Make it always be called with zero refcount; on this step we
just decrement it before the calls in dentry_kill(). That is
safe, since all places that care about the value of refcount
either do that under ->d_lock or hold a reference to dentry
in question. Either is sufficient to prevent observing a
dentry immediately prior to __dentry_kill() getting called
from dentry_kill().
Reviewed-by: Christian Brauner <brauner@kernel.org>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
retain_dentry() used to decrement refcount if and only if it returned
true. Lift those decrements into the callers.
Reviewed-by: Christian Brauner <brauner@kernel.org>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
... and rename it to to_shrink_list(), seeing that it no longer
does dropping any references
Reviewed-by: Christian Brauner <brauner@kernel.org>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
By now there is only one place in entire fast_dput() where we return
false; that happens after refcount had been decremented and found (under
->d_lock) to be zero. In that case, just prior to returning false to
caller, fast_dput() forcibly changes the refcount from 0 to 1.
Lift that resetting refcount to 1 into the callers; later in the series
it will be massaged out of existence.
Reviewed-by: Christian Brauner <brauner@kernel.org>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
If refcount is less than 1, we should just warn, unlock dentry and
return true, so that the caller doesn't try to do anything else.
Taking care of that leaves the rest of "lockref_put_return() has
failed" case equivalent to "decrement refcount and rejoin the
normal slow path after the point where we grab ->d_lock".
NOTE: lockref_put_return() is strictly a fastpath thing - unlike
the rest of lockref primitives, it does not contain a fallback.
Caller (and it looks like fast_dput() is the only legitimate one
in the entire kernel) has to do that itself. Reasons for
lockref_put_return() failures:
* ->d_lock held by somebody
* refcount <= 0
* ... or an architecture not supporting lockref use of
cmpxchg - sparc, anything non-SMP, config with spinlock debugging...
We could add a fallback, but it would be a clumsy API - we'd have
to distinguish between:
(1) refcount > 1 - decremented, lock not held on return
(2) refcount < 1 - left alone, probably no sense to hold the lock
(3) refcount is 1, no cmphxcg - decremented, lock held on return
(4) refcount is 1, cmphxcg supported - decremented, lock *NOT* held
on return.
We want to return with no lock held in case (4); that's the whole point of that
thing. We very much do not want to have the fallback in case (3) return without
a lock, since the caller might have to retake it in that case.
So it wouldn't be more convenient than doing the fallback in the caller and
it would be very easy to screw up, especially since the test coverage would
suck - no way to test (3) and (4) on the same kernel build.
Reviewed-by: Christian Brauner <brauner@kernel.org>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
->d_delete() is a way for filesystem to tell that dentry is not worth
keeping cached. It is not guaranteed to be called every time a dentry
has refcount drop down to zero; it is not guaranteed to be called before
dentry gets evicted. In other words, it is not suitable for any kind
of keeping track of dentry state.
None of the in-tree filesystems attempt to use it that way, fortunately.
So the contortions done by fast_dput() (as well as dentry_kill()) are
not warranted. fast_dput() certainly should treat having ->d_delete()
instance as "can't assume we'll be keeping it", but that's not different
from the way we treat e.g. DCACHE_DONTCACHE (which is rather similar
to making ->d_delete() returns true when called).
Reviewed-by: Christian Brauner <brauner@kernel.org>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
... we won't see DCACHE_MAY_FREE on anything that is *not* dead
and checking d_flags is just as cheap as checking refcount.
Reviewed-by: Christian Brauner <brauner@kernel.org>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
new helper unifying identical bits of shrink_dentry_list() and
shring_dcache_for_umount()
Reviewed-by: Christian Brauner <brauner@kernel.org>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
