The 'mst_mutex' is not needed since because 'ubifs_write_master()' is only
called on the mount path and commit path. The mount path is sequential and
there is no parallelism, and the commit path is also serialized - there is only
one commit going on at a time.
Signed-off-by: Artem Bityutskiy <artem.bityutskiy@linux.intel.com>
Fix checkpatch warnings:
"WARNING: Prefer seq_puts to seq_printf"
Andrew Morton wrote:
"
- puts is presumably faster
- puts doesn't go rogue if you accidentally pass it a "%".
- this patch actually made fs/ubifs/super.o 12 bytes smaller.
Perhaps because seq_printf() is a varargs function, forcing the
caller to pass args on the stack instead of in registers.
"
Signed-off-by: Fabian Frederick <fabf@skynet.be>
Signed-off-by: Artem Bityutskiy <artem.bityutskiy@linux.intel.com>
In the end of 'create_default_filesystem()' we need to check
the return value of 'ubifs_write_node()' to ensure that we have
successfully written the 'cs_node'.
Signed-off-by: hujianyang <hujianyang@huawei.com>
Signed-off-by: Artem Bityutskiy <artem.bityutskiy@linux.intel.com>
Randy Dunlap pointed that we should use "scanned" instead of "scaned". This
patch makes the correction.
Signed-off-by: Artem Bityutskiy <artem.bityutskiy@linux.intel.com>
We set @ecc in ubifs_scan_leb only if leb_read returns EBADMSG and
do not use it any more. This patch removes this variable and adds
comments about EBADMSG handling.
Artem: re-phrase commentaries
Signed-off-by: hujianyang <hujianyang@huawei.com>
Signed-off-by: Artem Bityutskiy <artem.bityutskiy@linux.intel.com>
This is a minor fix. These two branches in 'dbg_chk_pnode()'
are dealing with different conditions. Although there is
no fault in current state, I think adding "break"s in
each end of branch is better.
Signed-off-by: hujianyang <hujianyang@huawei.com>
Signed-off-by: Artem Bityutskiy <artem.bityutskiy@linux.intel.com>
This patch checks the return value of 'ubifs_unpack_nnode()'.
If this function returns an error, 'nnode' may not be
initialized, so just print an error message and break.
Signed-off-by: hujianyang <hujianyang@huawei.com>
Signed-off-by: Artem Bityutskiy <artem.bityutskiy@linux.intel.com>
Applying restrictive seccomp filter programs to large or diverse
codebases often requires handling threads which may be started early in
the process lifetime (e.g., by code that is linked in). While it is
possible to apply permissive programs prior to process start up, it is
difficult to further restrict the kernel ABI to those threads after that
point.
This change adds a new seccomp syscall flag to SECCOMP_SET_MODE_FILTER for
synchronizing thread group seccomp filters at filter installation time.
When calling seccomp(SECCOMP_SET_MODE_FILTER, SECCOMP_FILTER_FLAG_TSYNC,
filter) an attempt will be made to synchronize all threads in current's
threadgroup to its new seccomp filter program. This is possible iff all
threads are using a filter that is an ancestor to the filter current is
attempting to synchronize to. NULL filters (where the task is running as
SECCOMP_MODE_NONE) are also treated as ancestors allowing threads to be
transitioned into SECCOMP_MODE_FILTER. If prctrl(PR_SET_NO_NEW_PRIVS,
...) has been set on the calling thread, no_new_privs will be set for
all synchronized threads too. On success, 0 is returned. On failure,
the pid of one of the failing threads will be returned and no filters
will have been applied.
The race conditions against another thread are:
- requesting TSYNC (already handled by sighand lock)
- performing a clone (already handled by sighand lock)
- changing its filter (already handled by sighand lock)
- calling exec (handled by cred_guard_mutex)
The clone case is assisted by the fact that new threads will have their
seccomp state duplicated from their parent before appearing on the tasklist.
Holding cred_guard_mutex means that seccomp filters cannot be assigned
while in the middle of another thread's exec (potentially bypassing
no_new_privs or similar). The call to de_thread() may kill threads waiting
for the mutex.
Changes across threads to the filter pointer includes a barrier.
Based on patches by Will Drewry.
Suggested-by: Julien Tinnes <jln@chromium.org>
Signed-off-by: Kees Cook <keescook@chromium.org>
Reviewed-by: Oleg Nesterov <oleg@redhat.com>
Reviewed-by: Andy Lutomirski <luto@amacapital.net>
Since seccomp transitions between threads requires updates to the
no_new_privs flag to be atomic, the flag must be part of an atomic flag
set. This moves the nnp flag into a separate task field, and introduces
accessors.
Signed-off-by: Kees Cook <keescook@chromium.org>
Reviewed-by: Oleg Nesterov <oleg@redhat.com>
Reviewed-by: Andy Lutomirski <luto@amacapital.net>
Pull gfs2 fixes from Steven Whitehouse:
"This patch set contains two minor docs/spelling fixes, some fixes for
flock, a change to use GFP_NOFS to avoid recursion on a rarely used
code path and a fix for a race relating to the glock lru"
* tag 'gfs2-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/steve/gfs2-3.0-fixes:
GFS2: fs/gfs2/rgrp.c: kernel-doc warning fixes
GFS2: memcontrol: Spelling s/invlidate/invalidate/
GFS2: Allow caching of glocks for flock
GFS2: Allow flocks to use normal glock dq rather than dq_wait
GFS2: replace count*size kzalloc by kcalloc
GFS2: Use GFP_NOFS when allocating glocks
GFS2: Fix race in glock lru glock disposal
GFS2: Only wait for demote when last holder is dequeued
Pull xfs fixes from Dave Chinner:
"Fixes for low memory perforamnce regressions and a quota inode
handling regression.
These are regression fixes for issues recently introduced - the change
in the stack switch location is fairly important, so I've held off
sending this update until I was sure that it still addresses the stack
usage problem the original solved. So while the commits in the xfs
tree are recent, it has been under tested for several weeks now"
* tag 'xfs-for-linus-3.16-rc5' of git://oss.sgi.com/xfs/xfs:
xfs: null unused quota inodes when quota is on
xfs: refine the allocation stack switch
Revert "xfs: block allocation work needs to be kswapd aware"
The current code always selects XPRT_TRANSPORT_BC_TCP for the back
channel, even when the forward channel was not TCP (eg, RDMA). When
a 4.1 mount is attempted with RDMA, the server panics in the TCP BC
code when trying to send CB_NULL.
Instead, construct the transport protocol number from the forward
channel transport or'd with XPRT_TRANSPORT_BC. Transports that do
not support bi-directional RPC will not have registered a "BC"
transport, causing create_backchannel_client() to fail immediately.
Fixes: https://bugzilla.linux-nfs.org/show_bug.cgi?id=265
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Signed-off-by: J. Bruce Fields <bfields@redhat.com>
This patch removes the GLF_NOCACHE flag from the glocks associated with
flocks. There should be no good reason not to cache glocks for flocks:
they only force the glock to be demoted before they can be reacquired,
which can slow down performance and even cause glock hangs, especially
in cases where the flocks are held in Shared (SH) mode.
Signed-off-by: Bob Peterson <rpeterso@redhat.com>
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
This patch allows flock glocks to use a non-blocking dequeue rather
than dq_wait. It also reverts the previous patch I had posted regarding
dq_wait. The reverted patch isn't necessarily a bad idea, but I decided
this might avoid unforeseen side effects, and was therefore safer.
Signed-off-by: Bob Peterson <rpeterso@redhat.com>
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
Normally GFP_KERNEL is ok here, but there is now a rarely used code path
relating to deallocation of unlinked inodes (in certain corner cases)
which if hit at times of memory shortage can cause recursion while
trying to free memory.
One solution would be to try and move the gfs2_glock_get() call so
that it is no longer called while another glock is held, but that
doesn't look at all easy, so GFP_NOFS is the best solution for the
time being.
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
We must not leave items on the LRU list with GLF_LOCK set, since
they can be removed if the glock is brought back into use, which
may then potentially result in a hang, waiting for GLF_LOCK to
clear.
It doesn't happen very often, since it requires a glock that has
not been used for a long time to be brought back into use at the
same moment that the shrinker is part way through disposing of
glocks.
The fix is to set GLF_LOCK at a later time, when we already know
that the other locks can be obtained. Also, we now only release
the lru_lock in case a resched is needed, rather than on every
iteration.
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
Function gfs2_glock_dq_wait is supposed to dequeue a glock and then
wait for the lock to be demoted. The problem is, if this is a shared
lock, its demote will depend on the other holders, which means you
might end up waiting forever because the other process is blocked.
This problem is especially apparent when dealing with nested flocks.
Signed-off-by: Bob Peterson <rpeterso@redhat.com>
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
The read() of timerfd files allows to fetch the number of timer ticks
while there is no way to set it back from userspace.
To restore the timer's state as it was at checkpoint moment we need
a path to bring @ticks back. Initially I thought about writing ticks
back via write() interface but it seems such API is somehow obscure.
Instead implement timerfd_ioctl() method with TFD_IOC_SET_TICKS
command which allows to adjust @ticks into non-zero value waking
up the waiters.
I wrapped code with CONFIG_CHECKPOINT_RESTORE which can be
dropped off if there users except c/r camp appear.
v2 (by akpm@):
- Use define timerfd_ioctl NULL for non c/r config
v3:
- Use copy_from_user for @ticks fetching since
not all arch support get_user for 8 byte argument
Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Michael Kerrisk <mtk.manpages@gmail.com>
Cc: Andrey Vagin <avagin@openvz.org>
Cc: Arnd Bergmann <arnd@arndb.de>
Cc: Christopher Covington <cov@codeaurora.org>
Cc: Pavel Emelyanov <xemul@parallels.com>
Cc: Vladimir Davydov <vdavydov@parallels.com>
Link: http://lkml.kernel.org/r/20140715215703.285617923@openvz.org
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
For checkpoint/restore of timerfd files we need to know how exactly
the timer were armed, to be able to recreate it on restore stage.
Thus implement show_fdinfo method which provides enough information
for that.
One of significant changes I think is the addition of @settime_flags
member. Currently there are two flags TFD_TIMER_ABSTIME and
TFD_TIMER_CANCEL_ON_SET, and the second can be found from
@might_cancel variable but in case if the flags will be extended
in future we most probably will have to somehow remember them
explicitly anyway so I guss doing that right now won't hurt.
To not bloat the timerfd_ctx structure I've converted @expired
to short integer and defined @settime_flags as short too.
v2 (by avagin@, vdavydov@ and tglx@):
- Add it_value/it_interval fields
- Save flags being used in timerfd_setup in context
v3 (by tglx@):
- don't forget to use CONFIG_PROC_FS
v4 (by akpm@):
-Use define timerfd_show NULL for non c/r config
Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Michael Kerrisk <mtk.manpages@gmail.com>
Cc: Andrey Vagin <avagin@openvz.org>
Cc: Pavel Emelyanov <xemul@parallels.com>
Cc: Vladimir Davydov <vdavydov@parallels.com>
Link: http://lkml.kernel.org/r/20140715215703.114365649@openvz.org
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
The first 8 ops of the compound are zeroed since they're a part of the
argument that's zeroed by the
memset(rqstp->rq_argp, 0, procp->pc_argsize);
in svc_process_common(). But we handle larger compounds by allocating
the memory on the fly in nfsd4_decode_compound(). Other than code
recently fixed by 01529e3f81 "NFSD: Fix memory leak in encoding denied
lock", I don't know of any examples of code depending on this
initialization. But it definitely seems possible, and I'd rather be
safe.
Compounds this long are unusual so I'm much more worried about failure
in this poorly tested cases than about an insignificant performance hit.
Signed-off-by: J. Bruce Fields <bfields@redhat.com>
sparse says:
fs/nfsd/auth.c:31:38: warning: incorrect type in argument 1 (different address spaces)
fs/nfsd/auth.c:31:38: expected struct cred const *cred
fs/nfsd/auth.c:31:38: got struct cred const [noderef] <asn:4>*real_cred
Add a new accessor for the ->real_cred and use that to fetch the
pointer. Accessing current->real_cred directly is actually quite safe
since we know that they can't go away so this is mostly a cosmetic fixup
to silence sparse.
Signed-off-by: Jeff Layton <jlayton@primarydata.com>
Signed-off-by: J. Bruce Fields <bfields@redhat.com>
Special kernel keys, such as those used to hold DNS results for AFS, CIFS and
NFS and those used to hold idmapper results for NFS, used to be
'invalidateable' with key_revoke(). However, since the default permissions for
keys were reduced:
Commit: 96b5c8fea6
KEYS: Reduce initial permissions on keys
it has become impossible to do this.
Add a key flag (KEY_FLAG_ROOT_CAN_INVAL) that will permit a key to be
invalidated by root. This should not be used for system keyrings as the
garbage collector will try and remove any invalidate key. For system keyrings,
KEY_FLAG_ROOT_CAN_CLEAR can be used instead.
After this, from userspace, keyctl_invalidate() and "keyctl invalidate" can be
used by any possessor of CAP_SYS_ADMIN (typically root) to invalidate DNS and
idmapper keys. Invalidated keys are immediately garbage collected and will be
immediately rerequested if needed again.
Signed-off-by: David Howells <dhowells@redhat.com>
Tested-by: Steve Dickson <steved@redhat.com>
Add an extra delegation state to allow the stateid to remain in the idr
tree until the last reference has been released. This will be necessary
to ensure uniqueness once the client_mutex is removed.
[jlayton: reset the sc_type under the state_lock in unhash_delegation]
Signed-off-by: Trond Myklebust <trond.myklebust@primarydata.com>
Signed-off-by: Jeff Layton <jlayton@primarydata.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: J. Bruce Fields <bfields@redhat.com>
No need to pass the delegation pointer in here as it's only used to get
the nfs4_file pointer.
Signed-off-by: Jeff Layton <jlayton@primarydata.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: J. Bruce Fields <bfields@redhat.com>
state_lock is a heavily contended global lock. We don't want to grab
that while simultaneously holding the inode->i_lock.
Add a new per-nfs4_file lock that we can use to protect the
per-nfs4_file delegation list. Hold that while walking the list in the
break_deleg callback and queue the workqueue job for each one.
The workqueue job can then take the state_lock and do the list
manipulations without the i_lock being held prior to starting the
rpc call.
Signed-off-by: Trond Myklebust <trond.myklebust@primarydata.com>
Signed-off-by: Jeff Layton <jlayton@primarydata.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: J. Bruce Fields <bfields@redhat.com>
It's just an obfuscated INIT_WORK call. Just make the work_func_t a
non-static symbol and use a normal INIT_WORK call.
Signed-off-by: Jeff Layton <jlayton@primarydata.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: J. Bruce Fields <bfields@redhat.com>
It is currently not possible for various wait_on_bit functions
to implement a timeout.
While the "action" function that is called to do the waiting
could certainly use schedule_timeout(), there is no way to carry
forward the remaining timeout after a false wake-up.
As false-wakeups a clearly possible at least due to possible
hash collisions in bit_waitqueue(), this is a real problem.
The 'action' function is currently passed a pointer to the word
containing the bit being waited on. No current action functions
use this pointer. So changing it to something else will be a
little noisy but will have no immediate effect.
This patch changes the 'action' function to take a pointer to
the "struct wait_bit_key", which contains a pointer to the word
containing the bit so nothing is really lost.
It also adds a 'private' field to "struct wait_bit_key", which
is initialized to zero.
An action function can now implement a timeout with something
like
static int timed_out_waiter(struct wait_bit_key *key)
{
unsigned long waited;
if (key->private == 0) {
key->private = jiffies;
if (key->private == 0)
key->private -= 1;
}
waited = jiffies - key->private;
if (waited > 10 * HZ)
return -EAGAIN;
schedule_timeout(waited - 10 * HZ);
return 0;
}
If any other need for context in a waiter were found it would be
easy to use ->private for some other purpose, or even extend
"struct wait_bit_key".
My particular need is to support timeouts in nfs_release_page()
to avoid deadlocks with loopback mounted NFS.
While wait_on_bit_timeout() would be a cleaner interface, it
will not meet my need. I need the timeout to be sensitive to
the state of the connection with the server, which could change.
So I need to use an 'action' interface.
Signed-off-by: NeilBrown <neilb@suse.de>
Acked-by: Peter Zijlstra <peterz@infradead.org>
Cc: Oleg Nesterov <oleg@redhat.com>
Cc: Steve French <sfrench@samba.org>
Cc: David Howells <dhowells@redhat.com>
Cc: Steven Whitehouse <swhiteho@redhat.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Link: http://lkml.kernel.org/r/20140707051604.28027.41257.stgit@notabene.brown
Signed-off-by: Ingo Molnar <mingo@kernel.org>
The current "wait_on_bit" interface requires an 'action'
function to be provided which does the actual waiting.
There are over 20 such functions, many of them identical.
Most cases can be satisfied by one of just two functions, one
which uses io_schedule() and one which just uses schedule().
So:
Rename wait_on_bit and wait_on_bit_lock to
wait_on_bit_action and wait_on_bit_lock_action
to make it explicit that they need an action function.
Introduce new wait_on_bit{,_lock} and wait_on_bit{,_lock}_io
which are *not* given an action function but implicitly use
a standard one.
The decision to error-out if a signal is pending is now made
based on the 'mode' argument rather than being encoded in the action
function.
All instances of the old wait_on_bit and wait_on_bit_lock which
can use the new version have been changed accordingly and their
action functions have been discarded.
wait_on_bit{_lock} does not return any specific error code in the
event of a signal so the caller must check for non-zero and
interpolate their own error code as appropriate.
The wait_on_bit() call in __fscache_wait_on_invalidate() was
ambiguous as it specified TASK_UNINTERRUPTIBLE but used
fscache_wait_bit_interruptible as an action function.
David Howells confirms this should be uniformly
"uninterruptible"
The main remaining user of wait_on_bit{,_lock}_action is NFS
which needs to use a freezer-aware schedule() call.
A comment in fs/gfs2/glock.c notes that having multiple 'action'
functions is useful as they display differently in the 'wchan'
field of 'ps'. (and /proc/$PID/wchan).
As the new bit_wait{,_io} functions are tagged "__sched", they
will not show up at all, but something higher in the stack. So
the distinction will still be visible, only with different
function names (gds2_glock_wait versus gfs2_glock_dq_wait in the
gfs2/glock.c case).
Since first version of this patch (against 3.15) two new action
functions appeared, on in NFS and one in CIFS. CIFS also now
uses an action function that makes the same freezer aware
schedule call as NFS.
Signed-off-by: NeilBrown <neilb@suse.de>
Acked-by: David Howells <dhowells@redhat.com> (fscache, keys)
Acked-by: Steven Whitehouse <swhiteho@redhat.com> (gfs2)
Acked-by: Peter Zijlstra <peterz@infradead.org>
Cc: Oleg Nesterov <oleg@redhat.com>
Cc: Steve French <sfrench@samba.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Link: http://lkml.kernel.org/r/20140707051603.28027.72349.stgit@notabene.brown
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Pull quota fix from Jan Kara:
"Fix locking of dquot shrinker"
* 'for_linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jack/linux-fs:
quota: missing lock in dqcache_shrink_scan()
In __set_test_and_free we will check whether all segment are free in one section
When free one segment, in order to set section to free status.
But the searching region of segmap is from start segno to last segno of f2fs,
it's not necessary. So let's just only check all segment bitmap of target
section.
Signed-off-by: Chao Yu <chao2.yu@samsung.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
Valid data within i_size in page cache will be copied to ICB cache when we
writeback the page by invoking udf_adinicb_writepage, so the copy in
udf_adinicb_write_end is redundant.
After we remove the copy, it's better to use simple_write_end directly in
udf_adinicb_aops instead of udf_adinicb_write_end.
Signed-off-by: Chao Yu <chao2.yu@samsung.com>
Signed-off-by: Jan Kara <jack@suse.cz>
This patch cleans up udf_translate_to_linux() a bit by using globally defined
macros instead of custom code.
We can use sprintf(buf, "%04X", ...) there as well, but this one faster.
Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Signed-off-by: Jan Kara <jack@suse.cz>
type and id were removed and qid added to quota_send_warning in commit
431f19744d
("userns: Convert quota netlink aka quota_send_warning")
Cc: Jan Kara <jack@suse.cz>
Cc: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Fabian Frederick <fabf@skynet.be>
Signed-off-by: Jan Kara <jack@suse.cz>
Fix checkpatch warning
WARNING: Use #include <linux/uaccess.h> instead of <asm/uaccess.h>
Cc: Jan Kara <jack@suse.cz>
Signed-off-by: Fabian Frederick <fabf@skynet.be>
Signed-off-by: Jan Kara <jack@suse.cz>
Drop cast on the result of kmem_cache_alloc.
The semantic patch that makes this change is as follows:
// <smpl>
@@
type T;
@@
- (T *)
(\(kmalloc\|kzalloc\|kcalloc\|kmem_cache_alloc\|kmem_cache_zalloc\|
kmem_cache_alloc_node\|kmalloc_node\|kzalloc_node\)(...))
// </smpl>
Signed-off-by: Himangi Saraogi <himangi774@gmail.com>
Acked-by: Julia Lawall <julia.lawall@lip6.fr>
Signed-off-by: Jan Kara <jack@suse.cz>
Remove dqptr_sem to make quota code scalable: Remove the dqptr_sem,
accessing inode->i_dquot now protected by dquot_srcu, and changing
inode->i_dquot is now serialized by dq_data_lock.
Signed-off-by: Lai Siyao <lai.siyao@intel.com>
Signed-off-by: Niu Yawei <yawei.niu@intel.com>
Signed-off-by: Jan Kara <jack@suse.cz>
Simplify the remove_inode_dquot_ref() to make it more obvious
that now we keep one reference for each dquot from inodes.
Signed-off-by: Jan Kara <jack@suse.cz>
Signed-off-by: Niu Yawei <yawei.niu@intel.com>
Signed-off-by: Jan Kara <jack@suse.cz>
Avoid unnecessary dqget()/dqput() calls in __dquot_initialize(),
that will introduce global lock contention otherwise.
Signed-off-by: Lai Siyao <lai.siyao@intel.com>
Signed-off-by: Niu Yawei <yawei.niu@intel.com>
Signed-off-by: Jan Kara <jack@suse.cz>
dqptr_sem will go away. Protect Q_GETFMT quotactl by
dqonoff_mutex instead. This is also enough to make sure
quota info will not go away while we are looking at it.
Signed-off-by: Lai Siyao <lai.siyao@intel.com>
Signed-off-by: Niu Yawei <yawei.niu@intel.com>
Signed-off-by: Jan Kara <jack@suse.cz>
Commit 1ab6c4997e (fs: convert fs shrinkers to new scan/count API)
accidentally removed locking from quota shrinker. Fix it -
dqcache_shrink_scan() should use dq_list_lock to protect the
scan on free_dquots list.
CC: stable@vger.kernel.org
Fixes: 1ab6c4997e
Signed-off-by: Niu Yawei <yawei.niu@intel.com>
Signed-off-by: Jan Kara <jack@suse.cz>
Pull fuse fixes from Miklos Szeredi:
"This contains miscellaneous fixes"
* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mszeredi/fuse:
fuse: replace count*size kzalloc by kcalloc
fuse: release temporary page if fuse_writepage_locked() failed
fuse: restructure ->rename2()
fuse: avoid scheduling while atomic
fuse: handle large user and group ID
fuse: inode: drop cast
fuse: ignore entry-timeout on LOOKUP_REVAL
fuse: timeout comparison fix
Now ext4_has_inline_data() is used in wide spread codepaths. So we need
to make it as a inline function to avoid burning some CPU cycles.
Change in text size:
text data bss dec hex filename
before: 326110 19258 5528 350896 55ab0 fs/ext4/ext4.o
after: 326227 19258 5528 351013 55b25 fs/ext4/ext4.o
I use the following script to measure the CPU usage.
#!/bin/bash
shm_base='/dev/shm'
img=${shm_base}/ext4-img
mnt=/mnt/loop
e2fsprgs_base=$HOME/e2fsprogs
mkfs=${e2fsprgs_base}/misc/mke2fs
fsck=${e2fsprgs_base}/e2fsck/e2fsck
sudo umount $mnt
dd if=/dev/zero of=$img bs=4k count=3145728
${mkfs} -t ext4 -O inline_data -F $img
sudo mount -t ext4 -o loop $img $mnt
# start testing...
testdir="${mnt}/testdir"
mkdir $testdir
cd $testdir
echo "start testing..."
for ((cnt=0;cnt<100;cnt++)); do
for ((i=0;i<5;i++)); do
for ((j=0;j<5;j++)); do
for ((k=0;k<5;k++)); do
for ((l=0;l<5;l++)); do
mkdir -p $i/$j/$k/$l
echo "$i-$j-$k-$l" > $i/$j/$k/$l/testfile
done
done
done
done
ls -R $testdir > /dev/null
rm -rf $testdir/*
done
The result of `perf top -G -U` is as below.
vanilla:
13.92% [ext4] [k] ext4_do_update_inode
9.36% [ext4] [k] __ext4_get_inode_loc
4.07% [ext4] [k] ftrace_define_fields_ext4_writepages
3.83% [ext4] [k] __ext4_handle_dirty_metadata
3.42% [ext4] [k] ext4_get_inode_flags
2.71% [ext4] [k] ext4_mark_iloc_dirty
2.46% [ext4] [k] ftrace_define_fields_ext4_direct_IO_enter
2.26% [ext4] [k] ext4_get_inode_loc
2.22% [ext4] [k] ext4_has_inline_data
[...]
After applied the patch, we don't see ext4_has_inline_data() because it
has been inlined and perf couldn't sample it. Although it doesn't mean
that the CPU cycles can be saved but at least the overhead of function
calls can be eliminated. So IMHO we'd better inline this function.
Cc: Andreas Dilger <adilger.kernel@dilger.ca>
Signed-off-by: Zheng Liu <wenqing.lz@taobao.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
