When running the AIM7's short workload, Linus' lockref patch eliminated
most of the spinlock contention. However, there were still some left:
8.46% reaim [kernel.kallsyms] [k] _raw_spin_lock
|--42.21%-- d_path
| proc_pid_readlink
| SyS_readlinkat
| SyS_readlink
| system_call
| __GI___readlink
|
|--40.97%-- sys_getcwd
| system_call
| __getcwd
The big one here is the rename_lock (seqlock) contention in d_path()
and the getcwd system call. This patch will eliminate the need to take
the rename_lock while translating dentries into the full pathnames.
The need to take the rename_lock is to make sure that no rename
operation can be ongoing while the translation is in progress. However,
only one thread can take the rename_lock thus blocking all the other
threads that need it even though the translation process won't make
any change to the dentries.
This patch will replace the writer's write_seqlock/write_sequnlock
sequence of the rename_lock of the callers of the prepend_path() and
__dentry_path() functions with the reader's read_seqbegin/read_seqretry
sequence within these 2 functions. As a result, the code will have to
retry if one or more rename operations had been performed. In addition,
RCU read lock will be taken during the translation process to make sure
that no dentries will go away. To prevent live-lock from happening,
the code will switch back to take the rename_lock if read_seqretry()
fails for three times.
To further reduce spinlock contention, this patch does not take the
dentry's d_lock when copying the filename from the dentries. Instead,
it treats the name pointer and length as unreliable and just copy
the string byte-by-byte over until it hits a null byte or the end of
string as specified by the length. This should avoid stepping into
invalid memory address. The error cases are left to be handled by
the sequence number check.
The following code re-factoring are also made:
1. Move prepend('/') into prepend_name() to remove one conditional
check.
2. Move the global root check in prepend_path() back to the top of
the while loop.
With this patch, the _raw_spin_lock will now account for only 1.2%
of the total CPU cycles for the short workload. This patch also has
the effect of reducing the effect of running perf on its profile
since the perf command itself can be a heavy user of the d_path()
function depending on the complexity of the workload.
When taking the perf profile of the high-systime workload, the amount
of spinlock contention contributed by running perf without this patch
was about 16%. With this patch, the spinlock contention caused by
the running of perf will go away and we will have a more accurate
perf profile.
Signed-off-by: Waiman Long <Waiman.Long@hp.com>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
This simplifies the RCU to refcounting code in particular.
I was originally intending to leave this for later, but walking through
all the dput() logic (see previous commit), I realized that the dput()
"might_sleep()" check was misleadingly weak. And I removed it as
misleading, both for performance profiling and for debugging.
However, the might_sleep() debugging case is actually true: the final
dput() can indeed sleep, if the inode of the dentry that you are
releasing ends up sleeping at iput time (see dentry_iput()). So the
problem with the might_sleep() in dput() wasn't that it wasn't true, it
was that it wasn't actually testing and triggering on the interesting
case.
In particular, just about *any* dput() can indeed sleep, if you happen
to race with another thread deleting the file in question, and you then
lose the race to the be the last dput() for that file. But because it's
a very rare race, the debugging code would never trigger it in practice.
Why is this problematic? The new d_rcu_to_refcount() (see commit
15570086b5: "vfs: reimplement d_rcu_to_refcount() using
lockref_get_or_lock()") does a dput() for the failure case, and it does
it under the RCU lock. So potentially sleeping really is a bug.
But there's no way I'm going to fix this with the previous complicated
"lockref_get_or_lock()" interface. And rather than revert to the old
and crufty nested dentry locking code (which did get this right by
delaying the reference count updates until they were verified to be
safe), let's make forward progress.
Cc: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
This is me being a bit OCD after all the dentry optimization work this
merge window: profiles end up showing 'dput()' as a rather expensive
operation, and there were two unrelated bad reasons for that.
The first reason was reading d_lockref.count for debugging purposes,
which touches the lockref cacheline (for reads) before really need to.
More importantly, the debugging test in question is _wrong_, and has
hidden bugs. It's true that we can only sleep when the count goes down
to zero, but the test as-is hides the much more subtle bug that happens
if we race with somebody else deleting the file.
Anyway we _will_ touch that cacheline, but let's do it for a write and
in the right routine (ie in "lockref_put_or_lock()") which annotates the
costs better. So remove the misleading debug code.
The other was an unnecessary access to the cacheline that contains the
d_lru list, just to check whether we already were on the LRU list or
not. This is exactly what we have d_flags for, so that we can avoid
touching extra cache lines for the common case. So just add another bit
for "is this dentry on the LRU".
Finally, mark the tests properly likely/unlikely, so that the common
fast-paths are dense in the instruction stream.
This makes the profiles look much saner.
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
We check submounts before doing d_drop() on a non-empty directory dentry in
NFS (have_submounts()), but we do not exclude a racing mount. Nor do we
prevent mounts to be added to the disconnected subtree using relative paths
after the d_drop().
This patch fixes these issues by checking for unlinked (unhashed, non-root)
ancestors before proceeding with the mount. This is done with rename
seqlock taken for write and with ->d_lock grabbed on each ancestor in turn,
including our dentry itself. This ensures that the only one of
check_submounts_and_drop() or has_unlinked_ancestor() can succeed.
Signed-off-by: Miklos Szeredi <miklos@szeredi.hu>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
We check submounts before doing d_drop() on a non-empty directory dentry in
NFS (have_submounts()), but we do not exclude a racing mount.
Process A: have_submounts() -> returns false
Process B: mount() -> success
Process A: d_drop()
This patch prepares the ground for the fix by doing the following
operations all under the same rename lock:
have_submounts()
shrink_dcache_parent()
d_drop()
This is actually an optimization since have_submounts() and
shrink_dcache_parent() both traverse the same dentry tree separately.
Signed-off-by: Miklos Szeredi <mszeredi@suse.cz>
CC: David Howells <dhowells@redhat.com>
CC: Steven Whitehouse <swhiteho@redhat.com>
CC: Trond Myklebust <Trond.Myklebust@netapp.com>
CC: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
This one replaces three instances open coded tree walking (have_submounts,
select_parent, d_genocide) with a common helper.
In addition to slightly reducing the kernel size, this simplifies the
callers and makes them less bug prone.
Signed-off-by: Miklos Szeredi <mszeredi@suse.cz>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
It shouldn't matter when we decrement the refcount during the walk as long
as we do it exactly once.
Restructure d_genocide() to do the killing on entering the dentry instead
of when leaving it. This helps creating a common helper for tree walking.
Signed-off-by: Miklos Szeredi <mszeredi@suse.cz>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Pull vfs pile 1 from Al Viro:
"Unfortunately, this merge window it'll have a be a lot of small piles -
my fault, actually, for not keeping #for-next in anything that would
resemble a sane shape ;-/
This pile: assorted fixes (the first 3 are -stable fodder, IMO) and
cleanups + %pd/%pD formats (dentry/file pathname, up to 4 last
components) + several long-standing patches from various folks.
There definitely will be a lot more (starting with Miklos'
check_submount_and_drop() series)"
* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/vfs: (26 commits)
direct-io: Handle O_(D)SYNC AIO
direct-io: Implement generic deferred AIO completions
add formats for dentry/file pathnames
kvm eventfd: switch to fdget
powerpc kvm: use fdget
switch fchmod() to fdget
switch epoll_ctl() to fdget
switch copy_module_from_fd() to fdget
git simplify nilfs check for busy subtree
ibmasmfs: don't bother passing superblock when not needed
don't pass superblock to hypfs_{mkdir,create*}
don't pass superblock to hypfs_diag_create_files
don't pass superblock to hypfs_vm_create_files()
oprofile: get rid of pointless forward declarations of struct super_block
oprofilefs_create_...() do not need superblock argument
oprofilefs_mkdir() doesn't need superblock argument
don't bother with passing superblock to oprofile_create_stats_files()
oprofile: don't bother with passing superblock to ->create_files()
don't bother passing sb to oprofile_create_files()
coh901318: don't open-code simple_read_from_buffer()
...
The d_prune dentry operation is used to notify filesystem when VFS
about to prune a hashed dentry from the dcache. There are three
code paths that prune dentries: shrink_dcache_for_umount_subtree(),
prune_dcache_sb() and d_prune_aliases(). For the d_prune_aliases()
case, VFS unhashes the dentry first, then call the d_prune dentry
operation. This confuses ceph_d_prune() (ceph uses the d_prune
dentry operation to maintain a flag indicating whether the complete
contents of a directory are in the dcache, pruning unhashed dentry
does not affect dir's completeness)
This patch fixes the issue by calling the d_prune dentry operation
in d_prune_aliases(), before unhashing the dentry. Also make VFS
only call the d_prune dentry operation for hashed dentry, to avoid
calling the d_prune dentry operation twice when dentry is pruned
by d_prune_aliases().
Signed-off-by: Yan, Zheng <zheng.z.yan@intel.com>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
This moves __d_rcu_to_refcount() from <linux/dcache.h> into fs/namei.c
and re-implements it using the lockref infrastructure instead. It also
adds a lot of comments about what is actually going on, because turning
a dentry that was looked up using RCU into a long-lived reference
counted entry is one of the more subtle parts of the rcu walk.
We also used to be _particularly_ subtle in unlazy_walk() where we
re-validate both the dentry and its parent using the same sequence
count. We used to do it by nesting the locks and then verifying the
sequence count just once.
That was silly, because nested locking is expensive, but the sequence
count check is not. So this just re-validates the dentry and the parent
separately, avoiding the nested locking, and making the lockref lookup
possible.
Acked-by: Waiman Long <waiman.long@hp.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
A valid parent pointer is always going to have a non-zero reference
count, but if we look up the parent optimistically without locking, we
have to protect against the (very unlikely) race against renaming
changing the parent from under us.
We do that by using lockref_get_not_zero(), and then re-checking the
parent pointer after getting a valid reference.
[ This is a re-implementation of a chunk from the original patch by
Waiman Long: "dcache: Enable lockless update of dentry's refcount".
I've completely rewritten the patch-series and split it up, but I'm
attributing this part to Waiman as it's close enough to his earlier
patch - Linus ]
Signed-off-by: Waiman Long <Waiman.Long@hp.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
This just replaces the dentry count/lock combination with the lockref
structure that contains both a count and a spinlock, and does the
mechanical conversion to use the lockref infrastructure.
There are no semantic changes here, it's purely syntactic. The
reference lockref implementation uses the spinlock exactly the same way
that the old dcache code did, and the bulk of this patch is just
expanding the internal "d_count" use in the dcache code to use
"d_lockref.count" instead.
This is purely preparation for the real change to make the reference
count updates be lockless during the 3.12 merge window.
[ As with the previous commit, this is a rewritten version of a concept
originally from Waiman, so credit goes to him, blame for any errors
goes to me.
Waiman's patch had some semantic differences for taking advantage of
the lockless update in dget_parent(), while this patch is
intentionally a pure search-and-replace change with no semantic
changes. - Linus ]
Signed-off-by: Waiman Long <Waiman.Long@hp.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
dynamic_dname() is both too much and too little for those - the
output may be well in excess of 64 bytes dynamic_dname() assumes
to be enough (thanks to ashmem feeding really long names to
shmem_file_setup()) and vsnprintf() is an overkill for those
guys.
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Pull second set of VFS changes from Al Viro:
"Assorted f_pos race fixes, making do_splice_direct() safe to call with
i_mutex on parent, O_TMPFILE support, Jeff's locks.c series,
->d_hash/->d_compare calling conventions changes from Linus, misc
stuff all over the place."
* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/vfs: (63 commits)
Document ->tmpfile()
ext4: ->tmpfile() support
vfs: export lseek_execute() to modules
lseek_execute() doesn't need an inode passed to it
block_dev: switch to fixed_size_llseek()
cpqphp_sysfs: switch to fixed_size_llseek()
tile-srom: switch to fixed_size_llseek()
proc_powerpc: switch to fixed_size_llseek()
ubi/cdev: switch to fixed_size_llseek()
pci/proc: switch to fixed_size_llseek()
isapnp: switch to fixed_size_llseek()
lpfc: switch to fixed_size_llseek()
locks: give the blocked_hash its own spinlock
locks: add a new "lm_owner_key" lock operation
locks: turn the blocked_list into a hashtable
locks: convert fl_link to a hlist_node
locks: avoid taking global lock if possible when waking up blocked waiters
locks: protect most of the file_lock handling with i_lock
locks: encapsulate the fl_link list handling
locks: make "added" in __posix_lock_file a bool
...
Instances either don't look at it at all (the majority of cases) or
only want it to find the superblock (which can be had as dentry->d_sb).
A few cases that want more are actually safe with dentry->d_inode -
the only precaution needed is the check that it hadn't been replaced with
NULL by rmdir() or by overwriting rename(), which case should be simply
treated as cache miss.
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
the only remaining caller (in ncpfs) is guaranteed to return 0 -
we only hit it if we'd just checked that there's no dentry with
such name.
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
I've restricted atomic_open to only operate on regular files, although
I still don't understand why atomic_open should not be possible also for
directories on GFS2. That can always be added in later though, if it
makes sense.
The ->atomic_open function can be passed negative dentries, which
in most cases means either ENOENT (->lookup) or a call to d_instantiate
(->create). In the GFS2 case though, we need to actually perform the
look up, since we do not know whether there has been a new inode created
on another node. The look up calls d_splice_alias which then tries to
rehash the dentry - so the solution here is to simply check for that
in d_splice_alias. The same issue is likely to affect any other cluster
filesystem implementing ->atomic_open
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Cc: "J. Bruce Fields" <bfields fieldses org>
Cc: Jeff Layton <jlayton@redhat.com>
Using list_move() instead of list_del() + list_add().
Signed-off-by: Wei Yongjun <yongjun_wei@trendmicro.com.cn>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
When pruning a dentry, its ancestor dentry can also be pruned. But
the ancestor dentry does not go through dput(), so it does not get
put on the dentry LRU. Hence associating d_prune with removing the
dentry from the LRU is the wrong.
The fix is remove dentry_lru_prune(). Call file system's d_prune()
callback directly when pruning dentries.
Signed-off-by: Yan, Zheng <zheng.z.yan@intel.com>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Call cond_resched() in shrink_dcache_parent() to maintain interactivity.
Before this patch:
void shrink_dcache_parent(struct dentry * parent)
{
while ((found = select_parent(parent, &dispose)) != 0)
shrink_dentry_list(&dispose);
}
select_parent() populates the dispose list with dentries which
shrink_dentry_list() then deletes. select_parent() carefully uses
need_resched() to avoid doing too much work at once. But neither
shrink_dcache_parent() nor its called functions call cond_resched(). So
once need_resched() is set select_parent() will return single dentry
dispose list which is then deleted by shrink_dentry_list(). This is
inefficient when there are a lot of dentry to process. This can cause
softlockup and hurts interactivity on non preemptable kernels.
This change adds cond_resched() in shrink_dcache_parent(). The benefit
of this is that need_resched() is quickly cleared so that future calls
to select_parent() are able to efficiently return a big batch of dentry.
These additional cond_resched() do not seem to impact performance, at
least for the workload below.
Here is a program which can cause soft lockup if other system activity
sets need_resched().
int main()
{
struct rlimit rlim;
int i;
int f[100000];
char buf[20];
struct timeval t1, t2;
double diff;
/* cleanup past run */
system("rm -rf x");
/* boost nfile rlimit */
rlim.rlim_cur = 200000;
rlim.rlim_max = 200000;
if (setrlimit(RLIMIT_NOFILE, &rlim))
err(1, "setrlimit");
/* make directory for files */
if (mkdir("x", 0700))
err(1, "mkdir");
if (gettimeofday(&t1, NULL))
err(1, "gettimeofday");
/* populate directory with open files */
for (i = 0; i < 100000; i++) {
snprintf(buf, sizeof(buf), "x/%d", i);
f[i] = open(buf, O_CREAT);
if (f[i] == -1)
err(1, "open");
}
/* close some of the files */
for (i = 0; i < 85000; i++)
close(f[i]);
/* unlink all files, even open ones */
system("rm -rf x");
if (gettimeofday(&t2, NULL))
err(1, "gettimeofday");
diff = (((double)t2.tv_sec * 1000000 + t2.tv_usec) -
((double)t1.tv_sec * 1000000 + t1.tv_usec));
printf("done: %g elapsed\n", diff/1e6);
return 0;
}
Signed-off-by: Greg Thelen <gthelen@google.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>
Cc: <stable@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
... lest we get livelocks between path_is_under() and d_path() and friends.
The thing is, wrt fairness lglocks are more similar to rwsems than to rwlocks;
it is possible to have thread B spin on attempt to take lock shared while thread
A is already holding it shared, if B is on lower-numbered CPU than A and there's
a thread C spinning on attempt to take the same lock exclusive.
As the result, we need consistent ordering between vfsmount_lock (lglock) and
rename_lock (seq_lock), even though everything that takes both is going to take
vfsmount_lock only shared.
Spotted-by: Brad Spengler <spender@grsecurity.net>
Cc: stable@vger.kernel.org
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
I'm not sure why, but the hlist for each entry iterators were conceived
list_for_each_entry(pos, head, member)
The hlist ones were greedy and wanted an extra parameter:
hlist_for_each_entry(tpos, pos, head, member)
Why did they need an extra pos parameter? I'm not quite sure. Not only
they don't really need it, it also prevents the iterator from looking
exactly like the list iterator, which is unfortunate.
Besides the semantic patch, there was some manual work required:
- Fix up the actual hlist iterators in linux/list.h
- Fix up the declaration of other iterators based on the hlist ones.
- A very small amount of places were using the 'node' parameter, this
was modified to use 'obj->member' instead.
- Coccinelle didn't handle the hlist_for_each_entry_safe iterator
properly, so those had to be fixed up manually.
The semantic patch which is mostly the work of Peter Senna Tschudin is here:
@@
iterator name hlist_for_each_entry, hlist_for_each_entry_continue, hlist_for_each_entry_from, hlist_for_each_entry_rcu, hlist_for_each_entry_rcu_bh, hlist_for_each_entry_continue_rcu_bh, for_each_busy_worker, ax25_uid_for_each, ax25_for_each, inet_bind_bucket_for_each, sctp_for_each_hentry, sk_for_each, sk_for_each_rcu, sk_for_each_from, sk_for_each_safe, sk_for_each_bound, hlist_for_each_entry_safe, hlist_for_each_entry_continue_rcu, nr_neigh_for_each, nr_neigh_for_each_safe, nr_node_for_each, nr_node_for_each_safe, for_each_gfn_indirect_valid_sp, for_each_gfn_sp, for_each_host;
type T;
expression a,c,d,e;
identifier b;
statement S;
@@
-T b;
<+... when != b
(
hlist_for_each_entry(a,
- b,
c, d) S
|
hlist_for_each_entry_continue(a,
- b,
c) S
|
hlist_for_each_entry_from(a,
- b,
c) S
|
hlist_for_each_entry_rcu(a,
- b,
c, d) S
|
hlist_for_each_entry_rcu_bh(a,
- b,
c, d) S
|
hlist_for_each_entry_continue_rcu_bh(a,
- b,
c) S
|
for_each_busy_worker(a, c,
- b,
d) S
|
ax25_uid_for_each(a,
- b,
c) S
|
ax25_for_each(a,
- b,
c) S
|
inet_bind_bucket_for_each(a,
- b,
c) S
|
sctp_for_each_hentry(a,
- b,
c) S
|
sk_for_each(a,
- b,
c) S
|
sk_for_each_rcu(a,
- b,
c) S
|
sk_for_each_from
-(a, b)
+(a)
S
+ sk_for_each_from(a) S
|
sk_for_each_safe(a,
- b,
c, d) S
|
sk_for_each_bound(a,
- b,
c) S
|
hlist_for_each_entry_safe(a,
- b,
c, d, e) S
|
hlist_for_each_entry_continue_rcu(a,
- b,
c) S
|
nr_neigh_for_each(a,
- b,
c) S
|
nr_neigh_for_each_safe(a,
- b,
c, d) S
|
nr_node_for_each(a,
- b,
c) S
|
nr_node_for_each_safe(a,
- b,
c, d) S
|
- for_each_gfn_sp(a, c, d, b) S
+ for_each_gfn_sp(a, c, d) S
|
- for_each_gfn_indirect_valid_sp(a, c, d, b) S
+ for_each_gfn_indirect_valid_sp(a, c, d) S
|
for_each_host(a,
- b,
c) S
|
for_each_host_safe(a,
- b,
c, d) S
|
for_each_mesh_entry(a,
- b,
c, d) S
)
...+>
[akpm@linux-foundation.org: drop bogus change from net/ipv4/raw.c]
[akpm@linux-foundation.org: drop bogus hunk from net/ipv6/raw.c]
[akpm@linux-foundation.org: checkpatch fixes]
[akpm@linux-foundation.org: fix warnings]
[akpm@linux-foudnation.org: redo intrusive kvm changes]
Tested-by: Peter Senna Tschudin <peter.senna@gmail.com>
Acked-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Signed-off-by: Sasha Levin <sasha.levin@oracle.com>
Cc: Wu Fengguang <fengguang.wu@intel.com>
Cc: Marcelo Tosatti <mtosatti@redhat.com>
Cc: Gleb Natapov <gleb@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
The following set of operations on a NFS client and server will cause
server# mkdir a
client# cd a
server# mv a a.bak
client# sleep 30 # (or whatever the dir attrcache timeout is)
client# stat .
stat: cannot stat `.': Stale NFS file handle
Obviously, we should not be getting an ESTALE error back there since the
inode still exists on the server. The problem is that the lookup code
will call d_revalidate on the dentry that "." refers to, because NFS has
FS_REVAL_DOT set.
nfs_lookup_revalidate will see that the parent directory has changed and
will try to reverify the dentry by redoing a LOOKUP. That of course
fails, so the lookup code returns ESTALE.
The problem here is that d_revalidate is really a bad fit for this case.
What we really want to know at this point is whether the inode is still
good or not, but we don't really care what name it goes by or whether
the dcache is still valid.
Add a new d_op->d_weak_revalidate operation and have complete_walk call
that instead of d_revalidate. The intent there is to allow for a
"weaker" d_revalidate that just checks to see whether the inode is still
good. This is also gives us an opportunity to kill off the FS_REVAL_DOT
special casing.
[AV: changed method name, added note in porting, fixed confusion re
having it possibly called from RCU mode (it won't be)]
Cc: NeilBrown <neilb@suse.de>
Signed-off-by: Jeff Layton <jlayton@redhat.com>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
* calling conventions change - ERR_PTR() is returned on ->d_hash() errors;
NULL is just for dcache miss now.
* exported, open-coded instances in ncpfs and cifs converted.
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
The last caller was removed >2 years ago in commit 7b2a69ba7.
Signed-off-by: Jeff Layton <jlayton@redhat.com>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
NFS appears to use d_obtain_alias() to create the root dentry rather than
d_make_root. This can cause 'prepend_path()' to complain that the root
has a weird name if an NFS filesystem is lazily unmounted. e.g. if
"/mnt" is an NFS mount then
{ cd /mnt; umount -l /mnt ; ls -l /proc/self/cwd; }
will cause a WARN message like
WARNING: at /home/git/linux/fs/dcache.c:2624 prepend_path+0x1d7/0x1e0()
...
Root dentry has weird name <>
to appear in kernel logs.
So change d_obtain_alias() to use "/" rather than "" as the anonymous
name.
Signed-off-by: NeilBrown <neilb@suse.de>
Cc: Trond Myklebust <Trond.Myklebust@netapp.com>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
The code that relied on that flag was ripped out of btrfs quite some
time ago, and never added back. Josef indicated that he was going to
take a different approach to the problem in btrfs, and that we
could just eliminate this flag.
Cc: Josef Bacik <jbacik@fusionio.com>
Signed-off-by: Jeff Layton <jlayton@redhat.com>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Pull vfs update from Al Viro:
- big one - consolidation of descriptor-related logics; almost all of
that is moved to fs/file.c
(BTW, I'm seriously tempted to rename the result to fd.c. As it is,
we have a situation when file_table.c is about handling of struct
file and file.c is about handling of descriptor tables; the reasons
are historical - file_table.c used to be about a static array of
struct file we used to have way back).
A lot of stray ends got cleaned up and converted to saner primitives,
disgusting mess in android/binder.c is still disgusting, but at least
doesn't poke so much in descriptor table guts anymore. A bunch of
relatively minor races got fixed in process, plus an ext4 struct file
leak.
- related thing - fget_light() partially unuglified; see fdget() in
there (and yes, it generates the code as good as we used to have).
- also related - bits of Cyrill's procfs stuff that got entangled into
that work; _not_ all of it, just the initial move to fs/proc/fd.c and
switch of fdinfo to seq_file.
- Alex's fs/coredump.c spiltoff - the same story, had been easier to
take that commit than mess with conflicts. The rest is a separate
pile, this was just a mechanical code movement.
- a few misc patches all over the place. Not all for this cycle,
there'll be more (and quite a few currently sit in akpm's tree)."
Fix up trivial conflicts in the android binder driver, and some fairly
simple conflicts due to two different changes to the sock_alloc_file()
interface ("take descriptor handling from sock_alloc_file() to callers"
vs "net: Providing protocol type via system.sockprotoname xattr of
/proc/PID/fd entries" adding a dentry name to the socket)
* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/vfs: (72 commits)
MAX_LFS_FILESIZE should be a loff_t
compat: fs: Generic compat_sys_sendfile implementation
fs: push rcu_barrier() from deactivate_locked_super() to filesystems
btrfs: reada_extent doesn't need kref for refcount
coredump: move core dump functionality into its own file
coredump: prevent double-free on an error path in core dumper
usb/gadget: fix misannotations
fcntl: fix misannotations
ceph: don't abuse d_delete() on failure exits
hypfs: ->d_parent is never NULL or negative
vfs: delete surplus inode NULL check
switch simple cases of fget_light to fdget
new helpers: fdget()/fdput()
switch o2hb_region_dev_write() to fget_light()
proc_map_files_readdir(): don't bother with grabbing files
make get_file() return its argument
vhost_set_vring(): turn pollstart/pollstop into bool
switch prctl_set_mm_exe_file() to fget_light()
switch xfs_find_handle() to fget_light()
switch xfs_swapext() to fget_light()
...
IBM reported a deadlock in select_parent(). This was found to be caused
by taking rename_lock when already locked when restarting the tree
traversal.
There are two cases when the traversal needs to be restarted:
1) concurrent d_move(); this can only happen when not already locked,
since taking rename_lock protects against concurrent d_move().
2) racing with final d_put() on child just at the moment of ascending
to parent; rename_lock doesn't protect against this rare race, so it
can happen when already locked.
Because of case 2, we need to be able to handle restarting the traversal
when rename_lock is already held. This patch fixes all three callers of
try_to_ascend().
IBM reported that the deadlock is gone with this patch.
[ I rewrote the patch to be smaller and just do the "goto again" if the
lock was already held, but credit goes to Miklos for the real work.
- Linus ]
Signed-off-by: Miklos Szeredi <mszeredi@suse.cz>
Cc: Al Viro <viro@ZenIV.linux.org.uk>
Cc: stable@vger.kernel.org
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
"Search list for X" sounds like you're trying to find X on a list.
Signed-off-by: J. Bruce Fields <bfields@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Each iteration of d_delete we reload inode from dentry->d_inode and
then call S_ISDIR(inode-i_mode), so inode cannot possibly be NULL
shortly afterwards unless something went horribly wrong.
Signed-off-by: Alan Cox <alan@linux.intel.com>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
IBM reported a soft lockup after applying the fix for the rename_lock
deadlock. Commit c83ce989cb ("VFS: Fix the nfs sillyrename regression
in kernel 2.6.38") was found to be the culprit.
The nfs sillyrename fix used DCACHE_DISCONNECTED to indicate that the
dentry was killed. This flag can be set on non-killed dentries too,
which results in infinite retries when trying to traverse the dentry
tree.
This patch introduces a separate flag: DCACHE_DENTRY_KILLED, which is
only set in d_kill() and makes try_to_ascend() test only this flag.
IBM reported successful test results with this patch.
Signed-off-by: Miklos Szeredi <mszeredi@suse.cz>
Cc: Trond Myklebust <Trond.Myklebust@netapp.com>
Cc: stable@vger.kernel.org
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
it's enough to set ->mnt_ns of internal vfsmounts to something
distinct from all struct mnt_namespace out there; then we can
just use the check for ->mnt_ns != NULL in the fast path of
mntput_no_expire()
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
This reverts commit 7732a557b1 (and commit
3f50fff4da, which was a follow-up
cleanup).
We're chasing an elusive bug that Dave Jones can apparently reproduce
using his system call fuzzer tool, and that looks like some kind of
locking ordering problem on the directory i_mutex chain. Our i_mutex
locking is rather complex, and depends on the topological ordering of
the directories, which is why we have been very wary of splicing
directory entries around.
Of course, we really don't want to ever see aliased unconnected
directories anyway, so none of this should ever happen, but this revert
aims to basically get us back to a known older state.
Bruce points to some of the previous discussion at
http://marc.info/?i=<20110310105821.GE22723@ZenIV.linux.org.uk>
and in particular a long post from Neil:
http://marc.info/?i=<20110311150749.2fa2be66@notabene.brown>
It should be noted that it's possible that Dave's problems come from
other changes altohgether, including possibly just the fact that Dave
constantly is teachning his fuzzer new tricks. So what appears to be a
new bug could in fact be an old one that just gets newly triggered, but
reverting these patches as "still under heavy discussion" is the right
thing regardless.
Requested-by: Al Viro <viro@zeniv.linux.org.uk>
Acked-by: J. Bruce Fields <bfields@fieldses.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Nobody sets want_disconn any more.
Reported-by: Peng Tao <bergwolf@gmail.com>
Signed-off-by: J. Bruce Fields <bfields@redhat.com>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
A directory should never have more than one dentry pointing to it.
But d_splice_alias() will add one if it finds a directory with an
already-existing non-DISCONNECTED dentry.
I can't find an obvious reproducer, but I also can't see what prevents
d_splice_alias() from encountering such a case.
It therefore seems safest to allow d_splice_alias to use any dentry it
finds.
(Prior to the removal of dentry_unhash() from vfs_rmdir(), around v3.0,
this could cause an nfsd deadlock like this:
- Somebody attempts to remove a non-empty directory.
- The dentry_unhash() in vfs_rmdir() unhashes the dentry
pointing to the non-empty directory.
- ->rmdir() then fails with -ENOTEMPTY
- Before the vfs_rmdir() caller reaches dput(), an nfsd process
in rename looks up the directory by filehandle; at the end of
that lookup, this dentry is found by d_alloc_anon(), and a
reference is taken on it, preventing dput() from removing it.
- A regular lookup of the directory calls d_splice_alias(),
finds only an unhashed (not a DISCONNECTED) dentry, and
insteads adds a new one, so the directory now has two
dentries.
- The nfsd process in rename, which was previously looking up
the source directory of the rename, now looks up the target
directory (which is the same), and gets the dentry newly
created by the previous lookup.
- The rename, seeing two different dentries, assumes this is a
cross-directory rename and attempts to take the i_mutex on the
directory twice.
That reproducer no longer exists, but I don't think there was anything
fundamentally incorrect about the vfs_rmdir() behavior there, so I think
the real fault was here in d_splice_alias().)
Signed-off-by: J. Bruce Fields <bfields@redhat.com>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
lglocks and brlocks are currently generated with some complicated macros
in lglock.h. But there's no reason to not just use common utility
functions and put all the data into a common data structure.
In preparation, this patch changes the API to look more like normal
function calls with pointers, not magic macros.
The patch is rather large because I move over all users in one go to keep
it bisectable. This impacts the VFS somewhat in terms of lines changed.
But no actual behaviour change.
[akpm@linux-foundation.org: checkpatch fixes]
Signed-off-by: Andi Kleen <ak@linux.intel.com>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Cc: Rusty Russell <rusty@rustcorp.com.au>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
UDP stack needs a minimum hash size value for proper operation and also
uses alloc_large_system_hash() for proper NUMA distribution of its hash
tables and automatic sizing depending on available system memory.
On some low memory situations, udp_table_init() must ignore the
alloc_large_system_hash() result and reallocs a bigger memory area.
As we cannot easily free old hash table, we leak it and kmemleak can
issue a warning.
This patch adds a low limit parameter to alloc_large_system_hash() to
solve this problem.
We then specify UDP_HTABLE_SIZE_MIN for UDP/UDPLite hash table
allocation.
Reported-by: Mark Asselstine <mark.asselstine@windriver.com>
Reported-by: Tim Bird <tim.bird@am.sony.com>
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Cc: Paul Gortmaker <paul.gortmaker@windriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This reverts commit 8c01a529b8.
It turns out the d_unhashed() check isn't unnecessary after all: while
it's true that unhashing will increment the sequence numbers, that does
not necessarily invalidate the RCU lookup, because it might have seen
the dentry pointer (before it got unhashed), but by the time it loaded
the sequence number, it could have seen the *new* sequence number (after
it got unhashed).
End result: we might look up an unhashed dentry that is about to be
freed, with the sequence number never indicating anything bad about it.
So checking that the dentry is still hashed (*after* reading the sequence
number) is indeed the proper fix, and was never unnecessary.
Reported-by: Dave Jones <davej@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Miklos Szeredi points out that we need to also worry about memory
odering when doing the dentry name comparison asynchronously with RCU.
In particular, doing a rename can do a memcpy() of one dentry name over
another, and we want to make sure that any unlocked reader will always
see the proper terminating NUL character, so that it won't ever run off
the allocation.
Rather than having to be extra careful with the name copy or at lookup
time for each character, this resolves the issue by making sure that all
names that are inlined in the dentry always have a NUL character at the
end of the name allocation. If we do that at dentry allocation time, we
know that no future name copy will ever change that final NUL to
anything else, so there are no memory ordering issues.
So even if a concurrent rename ends up overwriting the NUL character
that terminates the original name, we always know that there is one
final NUL at the end, and there is no worry about the lockless RCU
lookup traversing the name too far.
The out-of-line allocations are never copied over, so we can just make
sure that we write the name (with terminating NULL) and do a write
barrier before we expose the name to anything else by setting it in the
dentry.
Reported-by: Miklos Szeredi <mszeredi@suse.cz>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Cc: Nick Piggin <npiggin@gmail.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
This allows comparing hash and len in one operation on 64-bit
architectures. Right now only __d_lookup_rcu() takes advantage of this,
since that is the case we care most about.
The use of anonymous struct/unions hides the alternate 64-bit approach
from most users, the exception being a few cases where we initialize a
'struct qstr' with a static initializer. This makes the problematic
cases use a new QSTR_INIT() helper function for that (but initializing
just the name pointer with a "{ .name = xyzzy }" initializer remains
valid, as does just copying another qstr structure).
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
All callers do want to check the dentry length, but some of them can
check the length and the hash together, so doing it in dentry_cmp() can
be counter-productive.
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Commit 12f8ad4b05 ("vfs: clean up __d_lookup_rcu() and dentry_cmp()
interfaces") did the careful ACCESS_ONCE() of the dentry name only for
the word-at-a-time case, even though the issue is generic.
Admittedly I don't really see gcc ever reloading the value in the middle
of the loop, so the ACCESS_ONCE() protects us from a fairly theoretical
issue. But better safe than sorry.
Also, this consolidates the common parts of the word-at-a-time and
bytewise logic, which includes checking the length. We'll be changing
that later.
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
The check for d_unhashed() is not strictly incorrect, but at the same
time it is also not sensible. The actual dentry removal from the dentry
hash chains is totally asynchronous to the __d_lookup_rcu() logic, and
we depend on __d_drop() updating the sequence number to invalidate any
lookup of an unhashed dentry.
So checking d_unhashed() is not incorrect, but it's not useful either:
the code has to work correctly even without it. So just remove it.
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
The calling conventions for __d_lookup_rcu() and dentry_cmp() are
annoying in different ways, and there is actually one single underlying
reason for both of the annoyances.
The fundamental reason is that we do the returned dentry sequence number
check inside __d_lookup_rcu() instead of doing it in the caller. This
results in two annoyances:
- __d_lookup_rcu() now not only needs to return the dentry and the
sequence number that goes along with the lookup, it also needs to
return the inode pointer that was validated by that sequence number
check.
- and because we did the sequence number check early (to validate the
name pointer and length) we also couldn't just pass the dentry itself
to dentry_cmp(), we had to pass the counted string that contained the
name.
So that sequence number decision caused two separate ugly calling
conventions.
Both of these problems would be solved if we just did the sequence
number check in the caller instead. There's only one caller, and that
caller already has to do the sequence number check for the parent
anyway, so just do that.
That allows us to stop returning the dentry->d_inode in that in-out
argument (pointer-to-pointer-to-inode), so we can make the inode
argument just a regular input inode pointer. The caller can just load
the inode from dentry->d_inode, and then do the sequence number check
after that to make sure that it's synchronized with the name we looked
up.
And it allows us to just pass in the dentry to dentry_cmp(), which is
what all the callers really wanted. Sure, dentry_cmp() has to be a bit
careful about the dentry (which is not stable during RCU lookup), but
that's actually very simple.
And now that dentry_cmp() can clearly see that the first string argument
is a dentry, we can use the direct word access for that, instead of the
careful unaligned zero-padding. The dentry name is always properly
aligned, since it is a single path component that is either embedded
into the dentry itself, or was allocated with kmalloc() (see __d_alloc).
Finally, this also uninlines the nasty slow-case for dentry comparisons:
that one *does* need to do a sequence number check, since it will call
in to the low-level filesystems, and we want to give those a stable
inode pointer and path component length/start arguments. Doing an extra
sequence check for that slow case is not a problem, though.
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
It turns out that there are more cases than CONFIG_DEBUG_PAGEALLOC that
can have holes in the kernel address space: it seems to happen easily
with Xen, and it looks like the AMD gart64 code will also punch holes
dynamically.
Actually hitting that case is still very unlikely, so just do the
access, and take an exception and fix it up for the very unlikely case
of it being a page-crosser with no next page.
And hey, this abstraction might even help other architectures that have
other issues with unaligned word accesses than the possible missing next
page. IOW, this could do the byte order magic too.
Peter Anvin fixed a thinko in the shifting for the exception case.
Reported-and-tested-by: Jana Saout <jana@saout.de>
Cc: Peter Anvin <hpa@zytor.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
In d_materialise_unique() there are 3 subcases to the 'aliased dentry'
case; in two subcases the inode i_lock is properly released but this
does not occur in the -ELOOP subcase.
This seems to have been introduced by commit 1836750115 ("fix loop
checks in d_materialise_unique()").
Signed-off-by: Michel Lespinasse <walken@google.com>
Cc: stable@vger.kernel.org # v3.0+
[ Added a comment, and moved the unlock to where we generate the -ELOOP,
which seems to be more natural.
You probably can't actually trigger this without a buggy network file
server - d_materialize_unique() is for finding aliases on non-local
filesystems, and the d_ancestor() case is for a hardlinked directory
loop.
But we should be robust in the case of such buggy servers anyway. ]
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
"[PATCH 0/3] RFC - module.h usage cleanups in fs/ and lib/"
https://lkml.org/lkml/2012/2/29/589
--
Fix up files in fs/ and lib/ dirs to only use module.h if they really
need it.
These are trivial in scope vs. the work done previously. We now have
things where any few remaining cleanups can be farmed out to arch or
subsystem maintainers, and I have done so when possible. What is
remaining here represents the bits that don't clearly lie within a
single arch/subsystem boundary, like the fs dir and the lib dir.
Some duplicate includes arising from overlapping fixes from
independent subsystem maintainer submissions are also quashed.
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.11 (GNU/Linux)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=Z4w0
-----END PGP SIGNATURE-----
Merge tag 'module-for-3.4' of git://git.kernel.org/pub/scm/linux/kernel/git/paulg/linux
Pull cleanup of fs/ and lib/ users of module.h from Paul Gortmaker:
"Fix up files in fs/ and lib/ dirs to only use module.h if they really
need it.
These are trivial in scope vs the work done previously. We now have
things where any few remaining cleanups can be farmed out to arch or
subsystem maintainers, and I have done so when possible. What is
remaining here represents the bits that don't clearly lie within a
single arch/subsystem boundary, like the fs dir and the lib dir.
Some duplicate includes arising from overlapping fixes from
independent subsystem maintainer submissions are also quashed."
Fix up trivial conflicts due to clashes with other include file cleanups
(including some due to the previous bug.h cleanup pull).
* tag 'module-for-3.4' of git://git.kernel.org/pub/scm/linux/kernel/git/paulg/linux:
lib: reduce the use of module.h wherever possible
fs: reduce the use of module.h wherever possible
includecheck: delete any duplicate instances of module.h
Fix kernel-doc warnings in fs/dcache.c:
Warning(fs/dcache.c:1743): No description found for parameter 'seqp'
Warning(fs/dcache.c:1743): Excess function parameter 'seq' description in '__d_lookup_rcu'
Signed-off-by: Randy Dunlap <rdunlap@xenotime.net>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Pull vfs pile 1 from Al Viro:
"This is _not_ all; in particular, Miklos' and Jan's stuff is not there
yet."
* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/vfs: (64 commits)
ext4: initialization of ext4_li_mtx needs to be done earlier
debugfs-related mode_t whack-a-mole
hfsplus: add an ioctl to bless files
hfsplus: change finder_info to u32
hfsplus: initialise userflags
qnx4: new helper - try_extent()
qnx4: get rid of qnx4_bread/qnx4_getblk
take removal of PF_FORKNOEXEC to flush_old_exec()
trim includes in inode.c
um: uml_dup_mmap() relies on ->mmap_sem being held, but activate_mm() doesn't hold it
um: embed ->stub_pages[] into mmu_context
gadgetfs: list_for_each_safe() misuse
ocfs2: fix leaks on failure exits in module_init
ecryptfs: make register_filesystem() the last potential failure exit
ntfs: forgets to unregister sysctls on register_filesystem() failure
logfs: missing cleanup on register_filesystem() failure
jfs: mising cleanup on register_filesystem() failure
make configfs_pin_fs() return root dentry on success
configfs: configfs_create_dir() has parent dentry in dentry->d_parent
configfs: sanitize configfs_create()
...
* branch 'dcache-word-accesses':
vfs: use 'unsigned long' accesses for dcache name comparison and hashing
This does the name hashing and lookup using word-sized accesses when
that is efficient, namely on x86 (although any little-endian machine
with good unaligned accesses would do).
It does very much depend on little-endian logic, but it's a very hot
couple of functions under some real loads, and this patch improves the
performance of __d_lookup_rcu() and link_path_walk() by up to about 30%.
Giving a 10% improvement on some very pathname-heavy benchmarks.
Because we do make unaligned accesses past the filename, the
optimization is disabled when CONFIG_DEBUG_PAGEALLOC is active, and we
effectively depend on the fact that on x86 we don't really ever have the
last page of usable RAM followed immediately by any IO memory (due to
ACPI tables, BIOS buffer areas etc).
Some of the bit operations we do are a bit "subtle". It's commented,
but you do need to really think about the code. Or just consider it
black magic.
Thanks to people on G+ for some of the optimized bit tricks.
For some odd historical reason, the final mixing round for the dentry
cache hash table lookup had an insane "xor with big constant" logic. In
two places.
The big constant that is being xor'ed is GOLDEN_RATIO_PRIME, which is a
fairly random-looking number that is designed to be *multiplied* with so
that the bits get spread out over a whole long-word.
But xor'ing with it is insane. It doesn't really even change the hash -
it really only shifts the hash around in the hash table. To make
matters worse, the insane big constant is different on 32-bit and 64-bit
builds, even though the name hash bits we use are always 32-bit (and the
bits from the pointer we mix in effectively are too).
It's all total voodoo programming, in other words.
Now, some testing and analysis of the hash chains shows that the rest of
the hash function seems to be fairly good. It does pick the right bits
of the parent dentry pointer, for example, and while it's generally a
bad idea to use an xor to mix down the upper bits (because if there is a
repeating pattern, the xor can cause "destructive interference"), it
seems to not have been a disaster.
For example, replacing the hash with the normal "hash_long()" code (that
uses the GOLDEN_RATIO_PRIME constant correctly, btw) actually just makes
the hash worse. The hand-picked hash knew which bits of the pointer had
the highest entropy, and hash_long() ends up mixing bits less optimally
at least in some trivial tests.
So the hash function overall seems fine, it just has that really odd
"shift result around by a constant xor".
So get rid of the silly xor, and replace the down-mixing of the bits
with an add instead of an xor that tends to not have the same kind of
destructive interference issues. Some stats on the resulting hash
chains shows that they look statistically identical before and after,
but the code is simpler and no longer makes you go "WTF?".
Also, the incoming hash really is just "unsigned int", not a long, and
there's no real point to worry about the high 26 bits of the dentry
pointer for the 64-bit case, because they are all going to be identical
anyway.
So also change the hashing to be done in the more natural 'unsigned int'
that is the real size of the actual hashed data anyway.
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Ok, this is hacky, and only works on little-endian machines with goo
unaligned handling. And even then only with CONFIG_DEBUG_PAGEALLOC
disabled, since it can access up to 7 bytes after the pathname.
But it runs like a bat out of hell.
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
It's only used inside fs/dcache.c, and we're going to play games with it
for the word-at-a-time patches. This time we really don't even want to
export it, because it really is an internal function to fs/dcache.c, and
has been since it was introduced.
Having it in that extremely hot header file (it's included in pretty
much everything, thanks to <linux/fs.h>) is a disaster for testing
different versions, and is utterly pointless.
We really should have some kind of header file diet thing, where we
figure out which parts of header files are really better off private and
only result in more expensive compiles.
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
These don't change any semantics, but they clean up the code a bit and
mark some arguments appropriately 'const'.
They came up as I was doing the word-at-a-time dcache name accessor
code, and cleaning this up now allows me to send out a smaller relevant
interesting patch for the experimental stuff.
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
For files only using THIS_MODULE and/or EXPORT_SYMBOL, map
them onto including export.h -- or if the file isn't even
using those, then just delete the include. Fix up any implicit
include dependencies that were being masked by module.h along
the way.
Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
When the number of dentry cache hash table entries gets too high
(2147483648 entries), as happens by default on a 16TB system, use of a
signed integer in the dcache_init() initialization loop prevents the
dentry_hashtable from getting initialized, causing a panic in
__d_lookup(). Fix this in dcache_init() and similar areas.
Signed-off-by: Dimitri Sivanich <sivanich@sgi.com>
Acked-by: David S. Miller <davem@davemloft.net>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/sage/ceph-client:
ceph: ensure prealloc_blob is in place when removing xattr
rbd: initialize snap_rwsem in rbd_add()
ceph: enable/disable dentry complete flags via mount option
vfs: export symbol d_find_any_alias()
ceph: always initialize the dentry in open_root_dentry()
libceph: remove useless return value for osd_client __send_request()
ceph: avoid iput() while holding spinlock in ceph_dir_fsync
ceph: avoid useless dget/dput in encode_fh
ceph: dereference pointer after checking for NULL
crush: fix force for non-root TAKE
ceph: remove unnecessary d_fsdata conditional checks
ceph: Use kmemdup rather than duplicating its implementation
Fix up conflicts in fs/ceph/super.c (d_alloc_root() failure handling vs
always initialize the dentry in open_root_dentry)
Two (or more) concurrent calls of shrink_dcache_parent() on the same dentry may
cause shrink_dcache_parent() to loop forever.
Here's what appears to happen:
1 - CPU0: select_parent(P) finds C and puts it on dispose list, returns 1
2 - CPU1: select_parent(P) locks P->d_lock
3 - CPU0: shrink_dentry_list() locks C->d_lock
dentry_kill(C) tries to lock P->d_lock but fails, unlocks C->d_lock
4 - CPU1: select_parent(P) locks C->d_lock,
moves C from dispose list being processed on CPU0 to the new
dispose list, returns 1
5 - CPU0: shrink_dentry_list() finds dispose list empty, returns
6 - Goto 2 with CPU0 and CPU1 switched
Basically select_parent() steals the dentry from shrink_dentry_list() and thinks
it found a new one, causing shrink_dentry_list() to think it's making progress
and loop over and over.
One way to trigger this is to make udev calls stat() on the sysfs file while it
is going away.
Having a file in /lib/udev/rules.d/ with only this one rule seems to the trick:
ATTR{vendor}=="0x8086", ATTR{device}=="0x10ca", ENV{PCI_SLOT_NAME}="%k", ENV{MATCHADDR}="$attr{address}", RUN+="/bin/true"
Then execute the following loop:
while true; do
echo -bond0 > /sys/class/net/bonding_masters
echo +bond0 > /sys/class/net/bonding_masters
echo -bond1 > /sys/class/net/bonding_masters
echo +bond1 > /sys/class/net/bonding_masters
done
One fix would be to check all callers and prevent concurrent calls to
shrink_dcache_parent(). But I think a better solution is to stop the
stealing behavior.
This patch adds a new dentry flag that is set when the dentry is added to the
dispose list. The flag is cleared in dentry_lru_del() in case the dentry gets a
new reference just before being pruned.
If the dentry has this flag, select_parent() will skip it and let
shrink_dentry_list() retry pruning it. With select_parent() skipping those
dentries there will not be the appearance of progress (new dentries found) when
there is none, hence shrink_dcache_parent() will not loop forever.
Set the flag is also set in prune_dcache_sb() for consistency as suggested by
Linus.
Signed-off-by: Miklos Szeredi <mszeredi@suse.cz>
CC: stable@vger.kernel.org
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
select_parent currently abuses the dentry cache LRU to provide
cleanup features for child dentries that need to be freed. It moves
them to the tail of the LRU, then tells shrink_dcache_parent() to
calls __shrink_dcache_sb to unconditionally move them to a dispose
list (as DCACHE_REFERENCED is ignored). __shrink_dcache_sb() has to
relock the dentries to move them off the LRU onto the dispose list,
but otherwise does not touch the dentries that select_parent() moved
to the tail of the LRU. It then passses the dispose list to
shrink_dentry_list() which tries to free the dentries.
IOWs, the use of __shrink_dcache_sb() is superfluous - we can build
exactly the same list of dentries for disposal directly in
select_parent() and call shrink_dentry_list() instead of calling
__shrink_dcache_sb() to do that. This means that we avoid long holds
on the lru lock walking the LRU moving dentries to the dispose list
We also avoid the need to relock each dentry just to move it off the
LRU, reducing the numebr of times we lock each dentry to dispose of
them in shrink_dcache_parent() from 3 to 2 times.
Further, we remove one of the two callers of __shrink_dcache_sb().
This also means that __shrink_dcache_sb can be moved into back into
prune_dcache_sb() and we no longer have to handle referenced
dentries conditionally, simplifying the code.
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
a) mount --move is checking that ->mnt_parent is non-NULL before
looking if that parent happens to be shared; ->mnt_parent is never
NULL and it's not even an misspelled !mnt_has_parent()
b) pivot_root open-codes is_path_reachable(), poorly.
c) so does path_is_under(), while we are at it.
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
vfsmounts have ->mnt_parent pointing either to a different vfsmount
or to itself; it's never NULL and termination condition in loops
traversing the tree towards root is mnt == mnt->mnt_parent. At least
one place (see the next patch) is confused about what's going on;
let's add an explicit helper checking it right way and use it in
all places where we need it. Not that there had been too many,
but...
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
__d_path() API is asking for trouble and in case of apparmor d_namespace_path()
getting just that. The root cause is that when __d_path() misses the root
it had been told to look for, it stores the location of the most remote ancestor
in *root. Without grabbing references. Sure, at the moment of call it had
been pinned down by what we have in *path. And if we raced with umount -l, we
could have very well stopped at vfsmount/dentry that got freed as soon as
prepend_path() dropped vfsmount_lock.
It is safe to compare these pointers with pre-existing (and known to be still
alive) vfsmount and dentry, as long as all we are asking is "is it the same
address?". Dereferencing is not safe and apparmor ended up stepping into
that. d_namespace_path() really wants to examine the place where we stopped,
even if it's not connected to our namespace. As the result, it looked
at ->d_sb->s_magic of a dentry that might've been already freed by that point.
All other callers had been careful enough to avoid that, but it's really
a bad interface - it invites that kind of trouble.
The fix is fairly straightforward, even though it's bigger than I'd like:
* prepend_path() root argument becomes const.
* __d_path() is never called with NULL/NULL root. It was a kludge
to start with. Instead, we have an explicit function - d_absolute_root().
Same as __d_path(), except that it doesn't get root passed and stops where
it stops. apparmor and tomoyo are using it.
* __d_path() returns NULL on path outside of root. The main
caller is show_mountinfo() and that's precisely what we pass root for - to
skip those outside chroot jail. Those who don't want that can (and do)
use d_path().
* __d_path() root argument becomes const. Everyone agrees, I hope.
* apparmor does *NOT* try to use __d_path() or any of its variants
when it sees that path->mnt is an internal vfsmount. In that case it's
definitely not mounted anywhere and dentry_path() is exactly what we want
there. Handling of sysctl()-triggered weirdness is moved to that place.
* if apparmor is asked to do pathname relative to chroot jail
and __d_path() tells it we it's not in that jail, the sucker just calls
d_absolute_path() instead. That's the other remaining caller of __d_path(),
BTW.
* seq_path_root() does _NOT_ return -ENAMETOOLONG (it's stupid anyway -
the normal seq_file logics will take care of growing the buffer and redoing
the call of ->show() just fine). However, if it gets path not reachable
from root, it returns SEQ_SKIP. The only caller adjusted (i.e. stopped
ignoring the return value as it used to do).
Reviewed-by: John Johansen <john.johansen@canonical.com>
ACKed-by: John Johansen <john.johansen@canonical.com>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Cc: stable@vger.kernel.org
To prevent an NFS server from being used to create a directory loop in an NFS
superblock on the client, the following patch was committed:
commit 1836750115
Author: Al Viro <viro@zeniv.linux.org.uk>
Date: Tue Jul 12 21:42:24 2011 -0400
Subject: fix loop checks in d_materialise_unique()
This causes ELOOP to be reported to anyone trying to access the dentry that
would otherwise cause the kernel to complete the loop.
However, no indication is given to the caller as to why an operation that ought
to work doesn't. The fault is with the kernel, which doesn't want to try and
solve the problem as it gets horrendously messy if there's another mountpoint
somewhere in the trees being spliced that can't be moved[*].
[*] The real problem is that we don't handle the excision of a subtree that
gets moved _out_ of what we can see. This can happen on the server where a
directory is merely moved between two other dirs on the same filesystem, but
where destination dir is not accessible by the client.
So, given the choice to return ELOOP rather than trying to reconfigure the
dentry tree, we should give the caller some indication of why they aren't being
allowed to make what should be a legitimate request and log a message.
Signed-off-by: David Howells <dhowells@redhat.com>
Acked-by: Sachin Prabhu <sprabhu@redhat.com>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
This adds a d_prune dentry operation that is called by the VFS prior to
pruning (i.e. unhashing and killing) a hashed dentry from the dcache.
Wrap dentry_lru_del() and use the new _prune() helper in the cases where we
are about to unhash and kill the dentry.
This will be used by Ceph to maintain a flag indicating whether the
complete contents of a directory are contained in the dcache, allowing it
to satisfy lookups and readdir without addition server communication.
Renumber a few DCACHE_* #defines to group DCACHE_OP_PRUNE with the other
DCACHE_OP_ bits.
Signed-off-by: Sage Weil <sage@newdream.net>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Gcc tends to generate better code with small integers, including the
DCACHE_xyz flag tests - so move the common ones to be first in the list.
Also just remove the unused DCACHE_INOTIFY_PARENT_WATCHED and
DCACHE_AUTOFS_PENDING values, their users no longer exists in the source
tree.
And add a "unlikely()" to the DCACHE_OP_COMPARE test, since we want the
common case to be a nice straight-line fall-through.
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Fix new kernel-doc warning in fs/dcache.c:
Warning(fs/dcache.c:797): No description found for parameter 'sb'
Signed-off-by: Randy Dunlap <rdunlap@xenotime.net>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Reorganise shrink_dcache_for_umount_subtree() in light of the demise of
dcache_lock. Without that dcache_lock, there is no need for the batching of
removal of dentries from the system under it (we wanted to make intensive use
of the locked data whilst we held it, but didn't want to hold it for long at a
time).
This works, provided the preceding patch is correct in its removal of locking
on dentry->d_lock on the basis that no one should be locking these dentries any
more as the whole superblock is defunct.
With this patch, the calls to dentry_lru_del() and __d_shrink() are placed at
the point where each dentry is detached handled.
It is possible that, as an alternative, the batching should still be done -
but only for dentry_lru_del() of all a dentry's children in one go. In such a
case, the batching would be done under dcache_lru_lock.
Signed-off-by: David Howells <dhowells@redhat.com>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Locks of the dcache_lock were replaced by locks of dentry->d_lock in commits
such as:
23044507832fd6b7f507
as part of the RCU-based pathwalk changes, despite the fact that the caller
(shrink_dcache_for_umount()) notes in the banner comment the reasons that
d_lock is not necessary in these functions:
/*
* destroy the dentries attached to a superblock on unmounting
* - we don't need to use dentry->d_lock because:
* - the superblock is detached from all mountings and open files, so the
* dentry trees will not be rearranged by the VFS
* - s_umount is write-locked, so the memory pressure shrinker will ignore
* any dentries belonging to this superblock that it comes across
* - the filesystem itself is no longer permitted to rearrange the dentries
* in this superblock
*/
So remove these locks. If the locks are actually necessary, then this banner
comment should be altered instead.
The hash table chains are protected by 1-bit locks in the hash table heads, so
those shouldn't be a problem.
Note that to make this work, __d_drop() has to be split so that the RCUwalk
barrier can be avoided. This causes problems otherwise as it has an assertion
that dentry->d_lock is locked - but there is no need for that as no one else
can be trying to access this dentry, except to step over it (and that should
be handled by d_free(), I think).
Signed-off-by: David Howells <dhowells@redhat.com>
Cc: Nick Piggin <npiggin@kernel.dk>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Remove the detached-dentry counter from shrink_dcache_for_umount_subtree() as
the value it computes is no longer used as of commit
312d3ca856 which made the nr_dentry counters
summed per-CPU rather than global atomic.
Signed-off-by: David Howells <dhowells@redhat.com>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Adding a comment to d_materialise_unique per Al's request...
d_move and __d_move have some pretty substantial locking requirements,
but they are not clearly documented. Add some comments spelling them
out. Also, document the requirement for the i_mutex of the parent in
d_materialise_unique.
Cc: Al Viro <viro@ZenIV.linux.org.uk>
Signed-off-by: Jeff Layton <jlayton@redhat.com>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/vfs-2.6: (107 commits)
vfs: use ERR_CAST for err-ptr tossing in lookup_instantiate_filp
isofs: Remove global fs lock
jffs2: fix IN_DELETE_SELF on overwriting rename() killing a directory
fix IN_DELETE_SELF on overwriting rename() on ramfs et.al.
mm/truncate.c: fix build for CONFIG_BLOCK not enabled
fs:update the NOTE of the file_operations structure
Remove dead code in dget_parent()
AFS: Fix silly characters in a comment
switch d_add_ci() to d_splice_alias() in "found negative" case as well
simplify gfs2_lookup()
jfs_lookup(): don't bother with . or ..
get rid of useless dget_parent() in btrfs rename() and link()
get rid of useless dget_parent() in fs/btrfs/ioctl.c
fs: push i_mutex and filemap_write_and_wait down into ->fsync() handlers
drivers: fix up various ->llseek() implementations
fs: handle SEEK_HOLE/SEEK_DATA properly in all fs's that define their own llseek
Ext4: handle SEEK_HOLE/SEEK_DATA generically
Btrfs: implement our own ->llseek
fs: add SEEK_HOLE and SEEK_DATA flags
reiserfs: make reiserfs default to barrier=flush
...
Fix up trivial conflicts in fs/xfs/linux-2.6/xfs_super.c due to the new
shrinker callout for the inode cache, that clashed with the xfs code to
start the periodic workers later.
It seems to hurt performance in real life. Yes, the inode will be used
later, but the conditional doesn't seem to predict all that well
(negative dentries are not uncommon) and it looks like the cost of
prefetching is simply higher than depending on the cache doing the right
thing.
As usual.
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
With context based shrinkers, we can implement a per-superblock
shrinker that shrinks the caches attached to the superblock. We
currently have global shrinkers for the inode and dentry caches that
split up into per-superblock operations via a coarse proportioning
method that does not batch very well. The global shrinkers also
have a dependency - dentries pin inodes - so we have to be very
careful about how we register the global shrinkers so that the
implicit call order is always correct.
With a per-sb shrinker callout, we can encode this dependency
directly into the per-sb shrinker, hence avoiding the need for
strictly ordering shrinker registrations. We also have no need for
any proportioning code for the shrinker subsystem already provides
this functionality across all shrinkers. Allowing the shrinker to
operate on a single superblock at a time means that we do less
superblock list traversals and locking and reclaim should batch more
effectively. This should result in less CPU overhead for reclaim and
potentially faster reclaim of items from each filesystem.
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
New helper (non-exported, fs/internal.h-only): __d_alloc(sb, name).
Allocates dentry, sets its ->d_sb to given superblock and sets
->d_op accordingly. Old d_alloc(NULL, name) callers are converted
to that (all of them know what superblock they want). d_alloc()
itself is left only for parent != NULl case; uses __d_alloc(),
inserts result into the list of parent's children.
Note that now ->d_sb is assign-once and never NULL *and*
->d_parent is never NULL either.
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Btrfs (and I'd venture most other fs's) stores its indexes in nice disk order
for readdir, but unfortunately in the case of anything that stats the files in
order that readdir spits back (like oh say ls) that means we still have to do
the normal lookup of the file, which means looking up our other index and then
looking up the inode. What I want is a way to create dummy dentries when we
find them in readdir so that when ls or anything else subsequently does a
stat(), we already have the location information in the dentry and can go
straight to the inode itself. The lookup stuff just assumes that if it finds a
dentry it is done, it doesn't perform a lookup. So add a DCACHE_NEED_LOOKUP
flag so that the lookup code knows it still needs to run i_op->lookup() on the
parent to get the inode for the dentry. I have tested this with btrfs and I
went from something that looks like this
http://people.redhat.com/jwhiter/ls-noreada.png
To this
http://people.redhat.com/jwhiter/ls-good.png
Thats a savings of 1300 seconds, or 22 minutes. That is a significant savings.
Thanks,
Signed-off-by: Josef Bacik <josef@redhat.com>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Both __d_unalias() and __d_materialise_dentry() need loop prevention.
Grab rename_lock in caller, check for loops there...
As a side benefit, we have dentry_lock_for_move() called only under
rename_lock, which seriously reduces deadlock potential of the
execrable "locking order" used for ->d_lock.
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Change each shrinker's API by consolidating the existing parameters into
shrink_control struct. This will simplify any further features added w/o
touching each file of shrinker.
[akpm@linux-foundation.org: fix build]
[akpm@linux-foundation.org: fix warning]
[kosaki.motohiro@jp.fujitsu.com: fix up new shrinker API]
[akpm@linux-foundation.org: fix xfs warning]
[akpm@linux-foundation.org: update gfs2]
Signed-off-by: Ying Han <yinghan@google.com>
Cc: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Cc: Minchan Kim <minchan.kim@gmail.com>
Acked-by: Pavel Emelyanov <xemul@openvz.org>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Mel Gorman <mel@csn.ul.ie>
Acked-by: Rik van Riel <riel@redhat.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Hugh Dickins <hughd@google.com>
Cc: Dave Hansen <dave@linux.vnet.ibm.com>
Cc: Steven Whitehouse <swhiteho@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Commit e66eed651f ("list: remove prefetching from regular list
iterators") removed the include of prefetch.h from list.h, which
uncovered several cases that had apparently relied on that rather
obscure header file dependency.
So this fixes things up a bit, using
grep -L linux/prefetch.h $(git grep -l '[^a-z_]prefetchw*(' -- '*.[ch]')
grep -L 'prefetchw*(' $(git grep -l 'linux/prefetch.h' -- '*.[ch]')
to guide us in finding files that either need <linux/prefetch.h>
inclusion, or have it despite not needing it.
There are more of them around (mostly network drivers), but this gets
many core ones.
Reported-by: Stephen Rothwell <sfr@canb.auug.org.au>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Now that the whole dcache_hash_bucket crap is gone, go all the way and
also remove the weird locking layering violations for locking the hash
buckets. Add hlist_bl_lock/unlock helpers to move the locking into the
list abstraction instead of requiring each caller to open code it.
After all allowing for the bit locks is the whole point of these helpers
over the plain hlist variant.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
The dentry hashing rules have been really quite complicated for a long
while, in odd ways. That made functions like __d_drop() very fragile
and non-obvious.
In particular, whether a dentry was hashed or not was indicated with an
explicit DCACHE_UNHASHED bit. That's despite the fact that the hash
abstraction that the dentries use actually have a 'is this entry hashed
or not' model (which is a simple test of the 'pprev' pointer).
The reason that was done is because we used the normal 'is this entry
unhashed' model to mark whether the dentry had _ever_ been hashed in the
dentry hash tables, and that logic goes back many years (commit
b3423415fb: "dcache: avoid RCU for never-hashed dentries").
That, in turn, meant that __d_drop had totally different unhashing logic
for the dentry hash table case and for the anonymous dcache case,
because in order to use the "is this dentry hashed" logic as a flag for
whether it had ever been on the RCU hash table, we had to unhash such a
dentry differently so that we'd never think that it wasn't 'unhashed'
and wouldn't be free'd correctly.
That's just insane. It made the logic really hard to follow, when there
were two different kinds of "unhashed" states, and one of them (the one
that used "list_bl_unhashed()") really had nothing at all to do with
being unhashed per se, but with a very subtle lifetime rule instead.
So turn all of it around, and make it logical.
Instead of having a DENTRY_UNHASHED bit in d_flags to indicate whether
the dentry is on the hash chains or not, use the hash chain unhashed
logic for that. Suddenly "d_unhashed()" just uses "list_bl_unhashed()",
and everything makes sense.
And for the lifetime rule, just use an explicit DENTRY_RCUACCEES bit.
If we ever insert the dentry into the dentry hash table so that it is
visible to RCU lookup, we mark it DENTRY_RCUACCESS to show that it now
needs the RCU lifetime rules. Now suddently that test at dentry free
time makes sense too.
And because unhashing now is sane and doesn't depend on where the dentry
got unhashed from (because the dentry hash chain details doesn't have
some subtle side effects), we can re-unify the __d_drop() logic and use
common code for the unhashing.
Also fix one more open-coded hash chain bit_spin_lock() that I missed in
the previous chain locking cleanup commit.
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
It's a useless abstraction for 'hlist_bl_head', and it doesn't actually
help anything - quite the reverse. All the users end up having to know
about the hlist_bl_head details anyway, using 'struct hlist_bl_node *'
etc. So it just makes the code look confusing.
And the cost of it is extra '&b->head' syntactic noise, but more
importantly it spuriously makes the hash table dentry list look
different from the per-superblock DCACHE_DISCONNECTED dentry list.
As a result, the code ended up using ad-hoc locking for one case and
special helper functions for what is really another totally identical
case in the very same function.
Make it all look and work the same.
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
The case we should be verifying when updating the dentry name is that
the _parent_ inode (the directory) semaphore is held, not the semaphore
for the dentry itself. It's the directory locking that rename and
readdir() etc all care about.
The comment just above even says so - but then the BUG_ON() still
checked the dentry inode itself.
Very few people noticed, because this helper function really isn't used
for very much, so you had to be using ncpfs to ever hit it.
I think I should just remove the BUG_ON (the function really has just
one user), but let's run with it fixed for a while before getting rid of
it entirely.
Reported-and-tested-by: Bongani Hlope <bonganih@bankservafrica.com>
Reported-and-tested-by: Bernd Feige <bernd.feige@uniklinik-freiburg.de>
Cc: Petr Vandrovec <petr@vandrovec.name>,
Cc: Arnd Bergmann <arnd@arndb.de>
Cc: Christoph Hellwig <hch@lst.de>
Cc: Nick Piggin <npiggin@kernel.dk>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/vfs-2.6:
fs: call security_d_instantiate in d_obtain_alias V2
lose 'mounting_here' argument in ->d_manage()
don't pass 'mounting_here' flag to follow_down()
change the locking order for namespace_sem
fix deadlock in pivot_root()
vfs: split off vfsmount-related parts of vfs_kern_mount()
Some fixes for pstore
kill simple_set_mnt()
* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jikos/trivial: (47 commits)
doc: CONFIG_UNEVICTABLE_LRU doesn't exist anymore
Update cpuset info & webiste for cgroups
dcdbas: force SMI to happen when expected
arch/arm/Kconfig: remove one to many l's in the word.
asm-generic/user.h: Fix spelling in comment
drm: fix printk typo 'sracth'
Remove one to many n's in a word
Documentation/filesystems/romfs.txt: fixing link to genromfs
drivers:scsi Change printk typo initate -> initiate
serial, pch uart: Remove duplicate inclusion of linux/pci.h header
fs/eventpoll.c: fix spelling
mm: Fix out-of-date comments which refers non-existent functions
drm: Fix printk typo 'failled'
coh901318.c: Change initate to initiate.
mbox-db5500.c Change initate to initiate.
edac: correct i82975x error-info reported
edac: correct i82975x mci initialisation
edac: correct commented info
fs: update comments to point correct document
target: remove duplicate include of target/target_core_device.h from drivers/target/target_core_hba.c
...
Trivial conflict in fs/eventpoll.c (spelling vs addition)
While trying to track down some NFS problems with BTRFS, I kept noticing I was
getting -EACCESS for no apparent reason. Eric Paris and printk() helped me
figure out that it was SELinux that was giving me grief, with the following
denial
type=AVC msg=audit(1290013638.413:95): avc: denied { 0x800000 } for pid=1772
comm="nfsd" name="" dev=sda1 ino=256 scontext=system_u:system_r:kernel_t:s0
tcontext=system_u:object_r:unlabeled_t:s0 tclass=file
Turns out this is because in d_obtain_alias if we can't find an alias we create
one and do all the normal instantiation stuff, but we don't do the
security_d_instantiate.
Usually we are protected from getting a hashed dentry that hasn't yet run
security_d_instantiate() by the parent's i_mutex, but obviously this isn't an
option there, so in order to deal with the case that a second thread comes in
and finds our new dentry before we get to run security_d_instantiate(), we go
ahead and call it if we find a dentry already. Eric assures me that this is ok
as the code checks to see if the dentry has been initialized already so calling
security_d_instantiate() against the same dentry multiple times is ok. With
this patch I'm no longer getting errant -EACCESS values.
Signed-off-by: Josef Bacik <josef@redhat.com>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
The new vfs locking scheme introduced in 2.6.38 breaks NFS sillyrename
because the latter relies on being able to determine the parent
directory of the dentry in the ->iput() callback in order to send the
appropriate unlink rpc call.
Looking at the code that cares about races with dput(), there doesn't
seem to be anything that specifically uses d_parent as a test for
whether or not there is a race:
- __d_lookup_rcu(), __d_lookup() all test for d_hashed() after d_parent
- shrink_dcache_for_umount() is safe since nothing else can rearrange
the dentries in that super block.
- have_submount(), select_parent() and d_genocide() can test for a
deletion if we set the DCACHE_DISCONNECTED flag when the dentry
is removed from the parent's d_subdirs list.
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
Cc: stable@kernel.org (2.6.38, needs commit c826cb7dfc "dcache.c:
create helper function for duplicated functionality" )
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
This creates a helper function for he "try to ascend into the parent
directory" case, which was written out in triplicate before. With all
the locking and subtle sequence number stuff, we really don't want to
duplicate that kind of code.
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Without this patch, inodes are not promptly freed on last close of an
unlinked file by an nfs client:
client$ mount -tnfs4 server:/export/ /mnt/
client$ tail -f /mnt/FOO
...
server$ df -i /export
server$ rm /export/FOO
(^C the tail -f)
server$ df -i /export
server$ echo 2 >/proc/sys/vm/drop_caches
server$ df -i /export
the df's will show that the inode is not freed on the filesystem until
the last step, when it could have been freed after killing the client's
tail -f. On-disk data won't be deallocated either, leading to possible
spurious ENOSPC.
This occurs because when the client does the close, it arrives in a
compound with a putfh and a close, processed like:
- putfh: look up the filehandle. The only alias found for the
inode will be DCACHE_UNHASHED alias referenced by the filp
this, so it creates a new DCACHE_DISCONECTED dentry and
returns that instead.
- close: closes the existing filp, which is destroyed
immediately by dput() since it's DCACHE_UNHASHED.
- end of the compound: release the reference
to the current filehandle, and dput() the new
DCACHE_DISCONECTED dentry, which gets put on the
unused list instead of being destroyed immediately.
Nick Piggin suggested fixing this by allowing d_obtain_alias to return
the unhashed dentry that is referenced by the filp, instead of making it
create a new dentry.
Leave __d_find_alias() alone to avoid changing behavior of other
callers.
Also nfsd doesn't need all the checks of __d_find_alias(); any dentry,
hashed or unhashed, disconnected or not, should work.
Signed-off-by: J. Bruce Fields <bfields@redhat.com>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
dcache-locking.txt is not exist any more, and the path was not
correct anyway. Fix it.
Signed-off-by: Namhyung Kim <namhyung@gmail.com>
Signed-off-by: Jiri Kosina <jkosina@suse.cz>
Fix new fs/dcache.c kernel-doc warnings:
Warning(fs/dcache.c:184): No description found for parameter 'dentry'
Warning(fs/dcache.c:296): No description found for parameter 'parent'
Warning(fs/dcache.c:1985): No description found for parameter 'dparent'
Warning(fs/dcache.c:1985): Excess function parameter 'parent' description in 'd_validate'
Signed-off-by: Randy Dunlap <randy.dunlap@oracle.com>
Cc: Alexander Viro <viro@zeniv.linux.org.uk>
Cc: Nick Piggin <npiggin@kernel.dk>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/vfs-2.6: (23 commits)
sanitize vfsmount refcounting changes
fix old umount_tree() breakage
autofs4: Merge the remaining dentry ops tables
Unexport do_add_mount() and add in follow_automount(), not ->d_automount()
Allow d_manage() to be used in RCU-walk mode
Remove a further kludge from __do_follow_link()
autofs4: Bump version
autofs4: Add v4 pseudo direct mount support
autofs4: Fix wait validation
autofs4: Clean up autofs4_free_ino()
autofs4: Clean up dentry operations
autofs4: Clean up inode operations
autofs4: Remove unused code
autofs4: Add d_manage() dentry operation
autofs4: Add d_automount() dentry operation
Remove the automount through follow_link() kludge code from pathwalk
CIFS: Use d_automount() rather than abusing follow_link()
NFS: Use d_automount() rather than abusing follow_link()
AFS: Use d_automount() rather than abusing follow_link()
Add an AT_NO_AUTOMOUNT flag to suppress terminal automount
...
Add a dentry op (d_automount) to handle automounting directories rather than
abusing the follow_link() inode operation. The operation is keyed off a new
dentry flag (DCACHE_NEED_AUTOMOUNT).
This also makes it easier to add an AT_ flag to suppress terminal segment
automount during pathwalk and removes the need for the kludge code in the
pathwalk algorithm to handle directories with follow_link() semantics.
The ->d_automount() dentry operation:
struct vfsmount *(*d_automount)(struct path *mountpoint);
takes a pointer to the directory to be mounted upon, which is expected to
provide sufficient data to determine what should be mounted. If successful, it
should return the vfsmount struct it creates (which it should also have added
to the namespace using do_add_mount() or similar). If there's a collision with
another automount attempt, NULL should be returned. If the directory specified
by the parameter should be used directly rather than being mounted upon,
-EISDIR should be returned. In any other case, an error code should be
returned.
The ->d_automount() operation is called with no locks held and may sleep. At
this point the pathwalk algorithm will be in ref-walk mode.
Within fs/namei.c itself, a new pathwalk subroutine (follow_automount()) is
added to handle mountpoints. It will return -EREMOTE if the automount flag was
set, but no d_automount() op was supplied, -ELOOP if we've encountered too many
symlinks or mountpoints, -EISDIR if the walk point should be used without
mounting and 0 if successful. The path will be updated to point to the mounted
filesystem if a successful automount took place.
__follow_mount() is replaced by follow_managed() which is more generic
(especially with the patch that adds ->d_manage()). This handles transits from
directories during pathwalk, including automounting and skipping over
mountpoints (and holding processes with the next patch).
__follow_mount_rcu() will jump out of RCU-walk mode if it encounters an
automount point with nothing mounted on it.
follow_dotdot*() does not handle automounts as you don't want to trigger them
whilst following "..".
I've also extracted the mount/don't-mount logic from autofs4 and included it
here. It makes the mount go ahead anyway if someone calls open() or creat(),
tries to traverse the directory, tries to chdir/chroot/etc. into the directory,
or sticks a '/' on the end of the pathname. If they do a stat(), however,
they'll only trigger the automount if they didn't also say O_NOFOLLOW.
I've also added an inode flag (S_AUTOMOUNT) so that filesystems can mark their
inodes as automount points. This flag is automatically propagated to the
dentry as DCACHE_NEED_AUTOMOUNT by __d_instantiate(). This saves NFS and could
save AFS a private flag bit apiece, but is not strictly necessary. It would be
preferable to do the propagation in d_set_d_op(), but that doesn't normally
have access to the inode.
[AV: fixed breakage in case if __follow_mount_rcu() fails and nameidata_drop_rcu()
succeeds in RCU case of do_lookup(); we need to fall through to non-RCU case after
that, rather than just returning with ungrabbed *path]
Signed-off-by: David Howells <dhowells@redhat.com>
Was-Acked-by: Ian Kent <raven@themaw.net>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
It's indicative of a real problem, and it actually triggers with
autofs4, but the BUG_ON() is excessive. The autofs4 case is being fixed
(to only set d_op in the ->lookup method) but not merged yet. In the
meantime this gets the code limping along.
Reported-by: Alex Elder <aelder@sgi.com>
Cc: Ian Kent <raven@themaw.net>
Cc: Nick Piggin <npiggin@kernel.dk>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Fix function kernel-doc warning for prepend_path():
Warning(fs/dcache.c:1924): missing initial short description on line:
Signed-off-by: Randy Dunlap <randy.dunlap@oracle.com>
Cc: Alexander Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Fix function parameter kernel-doc for d_validate():
Warning(fs/dcache.c:1495): No description found for parameter 'parent'
Warning(fs/dcache.c:1495): Excess function parameter 'dparent' description in 'd_validate'
Signed-off-by: Randy Dunlap <randy.dunlap@oracle.com>
Cc: Alexander Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
The standard memcmp function on a Westmere system shows up hot in
profiles in the `git diff` workload (both parallel and single threaded),
and it is likely due to the costs associated with trapping into
microcode, and little opportunity to improve memory access (dentry
name is not likely to take up more than a cacheline).
So replace it with an open-coded byte comparison. This increases code
size by 8 bytes in the critical __d_lookup_rcu function, but the
speedup is huge, averaging 10 runs of each:
git diff st user sys elapsed CPU
before 1.15 2.57 3.82 97.1
after 1.14 2.35 3.61 96.8
git diff mt user sys elapsed CPU
before 1.27 3.85 1.46 349
after 1.26 3.54 1.43 333
Elapsed time for single threaded git diff at 95.0% confidence:
-0.21 +/- 0.01
-5.45% +/- 0.24%
It's -0.66% +/- 0.06% elapsed time on my Opteron, so rep cmp costs on the
fam10h seem to be relatively smaller, but there is still a win.
Signed-off-by: Nick Piggin <npiggin@kernel.dk>
This makes single threaded git diff -1.25% +/- 0.05% elapsed time on my
2s12c24t Westmere system, and -0.86% +/- 0.05% on my 2s8c Barcelona, by
prefetching the important first cacheline of the inode in while we do the
actual name compare and other operations on the dentry.
There was no measurable slowdown in the single file stat case, or the creat
case (where negative dentries would be common).
Signed-off-by: Nick Piggin <npiggin@kernel.dk>
Regardless of how much we possibly try to scale dcache, there is likely
always going to be some fundamental contention when adding or removing children
under the same parent. Pseudo filesystems do not seem need to have connected
dentries because by definition they are disconnected.
Signed-off-by: Nick Piggin <npiggin@kernel.dk>
dcache_inode_lock can be replaced with per-inode locking. Use existing
inode->i_lock for this. This is slightly non-trivial because we sometimes
need to find the inode from the dentry, which requires d_inode to be
stabilised (either with refcount or d_lock).
Signed-off-by: Nick Piggin <npiggin@kernel.dk>
Put dentry and inode fields into top of data structure. This allows RCU path
traversal to perform an RCU dentry lookup in a path walk by touching only the
first 56 bytes of the dentry.
We also fit in 8 bytes of inline name in the first 64 bytes, so for short
names, only 64 bytes needs to be touched to perform the lookup. We should
get rid of the hash->prev pointer from the first 64 bytes, and fit 16 bytes
of name in there, which will take care of 81% rather than 32% of the kernel
tree.
inode is also rearranged so that RCU lookup will only touch a single cacheline
in the inode, plus one in the i_ops structure.
This is important for directory component lookups in RCU path walking. In the
kernel source, directory names average is around 6 chars, so this works.
When we reach the last element of the lookup, we need to lock it and take its
refcount which requires another cacheline access.
Align dentry and inode operations structs, so members will be at predictable
offsets and we can group common operations into head of structure.
Signed-off-by: Nick Piggin <npiggin@kernel.dk>
Reduce some branches and memory accesses in dcache lookup by adding dentry
flags to indicate common d_ops are set, rather than having to check them.
This saves a pointer memory access (dentry->d_op) in common path lookup
situations, and saves another pointer load and branch in cases where we
have d_op but not the particular operation.
Patched with:
git grep -E '[.>]([[:space:]])*d_op([[:space:]])*=' | xargs sed -e 's/\([^\t ]*\)->d_op = \(.*\);/d_set_d_op(\1, \2);/' -e 's/\([^\t ]*\)\.d_op = \(.*\);/d_set_d_op(\&\1, \2);/' -i
Signed-off-by: Nick Piggin <npiggin@kernel.dk>
Rather than keep a d_mounted count in the dentry, set a dentry flag instead.
The flag can be cleared by checking the hash table to see if there are any
mounts left, which is not time critical because it is performed at detach time.
The mounted state of a dentry is only used to speculatively take a look in the
mount hash table if it is set -- before following the mount, vfsmount lock is
taken and mount re-checked without races.
This saves 4 bytes on 32-bit, nothing on 64-bit but it does provide a hole I
might use later (and some configs have larger than 32-bit spinlocks which might
make use of the hole).
Autofs4 conversion and changelog by Ian Kent <raven@themaw.net>:
In autofs4, when expring direct (or offset) mounts we need to ensure that we
block user path walks into the autofs mount, which is covered by another mount.
To do this we clear the mounted status so that follows stop before walking into
the mount and are essentially blocked until the expire is completed. The
automount daemon still finds the correct dentry for the umount due to the
follow mount logic in fs/autofs4/root.c:autofs4_follow_link(), which is set as
an inode operation for direct and offset mounts only and is called following
the lookup that stopped at the covered mount.
At the end of the expire the covering mount probably has gone away so the
mounted status need not be restored. But we need to check this and only restore
the mounted status if the expire failed.
XXX: autofs may not work right if we have other mounts go over the top of it?
Signed-off-by: Nick Piggin <npiggin@kernel.dk>
Perform common cases of path lookups without any stores or locking in the
ancestor dentry elements. This is called rcu-walk, as opposed to the current
algorithm which is a refcount based walk, or ref-walk.
This results in far fewer atomic operations on every path element,
significantly improving path lookup performance. It also avoids cacheline
bouncing on common dentries, significantly improving scalability.
The overall design is like this:
* LOOKUP_RCU is set in nd->flags, which distinguishes rcu-walk from ref-walk.
* Take the RCU lock for the entire path walk, starting with the acquiring
of the starting path (eg. root/cwd/fd-path). So now dentry refcounts are
not required for dentry persistence.
* synchronize_rcu is called when unregistering a filesystem, so we can
access d_ops and i_ops during rcu-walk.
* Similarly take the vfsmount lock for the entire path walk. So now mnt
refcounts are not required for persistence. Also we are free to perform mount
lookups, and to assume dentry mount points and mount roots are stable up and
down the path.
* Have a per-dentry seqlock to protect the dentry name, parent, and inode,
so we can load this tuple atomically, and also check whether any of its
members have changed.
* Dentry lookups (based on parent, candidate string tuple) recheck the parent
sequence after the child is found in case anything changed in the parent
during the path walk.
* inode is also RCU protected so we can load d_inode and use the inode for
limited things.
* i_mode, i_uid, i_gid can be tested for exec permissions during path walk.
* i_op can be loaded.
When we reach the destination dentry, we lock it, recheck lookup sequence,
and increment its refcount and mountpoint refcount. RCU and vfsmount locks
are dropped. This is termed "dropping rcu-walk". If the dentry refcount does
not match, we can not drop rcu-walk gracefully at the current point in the
lokup, so instead return -ECHILD (for want of a better errno). This signals the
path walking code to re-do the entire lookup with a ref-walk.
Aside from the final dentry, there are other situations that may be encounted
where we cannot continue rcu-walk. In that case, we drop rcu-walk (ie. take
a reference on the last good dentry) and continue with a ref-walk. Again, if
we can drop rcu-walk gracefully, we return -ECHILD and do the whole lookup
using ref-walk. But it is very important that we can continue with ref-walk
for most cases, particularly to avoid the overhead of double lookups, and to
gain the scalability advantages on common path elements (like cwd and root).
The cases where rcu-walk cannot continue are:
* NULL dentry (ie. any uncached path element)
* parent with d_inode->i_op->permission or ACLs
* dentries with d_revalidate
* Following links
In future patches, permission checks and d_revalidate become rcu-walk aware. It
may be possible eventually to make following links rcu-walk aware.
Uncached path elements will always require dropping to ref-walk mode, at the
very least because i_mutex needs to be grabbed, and objects allocated.
Signed-off-by: Nick Piggin <npiggin@kernel.dk>
The tricky locking for disposing of a dentry is duplicated 3 times in the
dcache (dput, pruning a dentry from the LRU, and pruning its ancestors).
Consolidate them all into a single function dentry_kill.
Signed-off-by: Nick Piggin <npiggin@kernel.dk>
prune_one_dentry can avoid quite a bit of locking in the common case where
ancestors have an elevated refcount. Alternatively, we could have gone the
other way and made fewer trylocks in the case where d_count goes to zero, but
is probably less common.
Signed-off-by: Nick Piggin <npiggin@kernel.dk>
dget_locked was a shortcut to avoid the lazy lru manipulation when we already
held dcache_lock (lru manipulation was relatively cheap at that point).
However, how that the lru lock is an innermost one, we never hold it at any
caller, so the lock cost can now be avoided. We already have well working lazy
dcache LRU, so it should be fine to defer LRU manipulations to scan time.
Signed-off-by: Nick Piggin <npiggin@kernel.dk>
dcache_inode_lock can be avoided in d_delete() and d_materialise_unique()
in cases where it is not required.
Signed-off-by: Nick Piggin <npiggin@kernel.dk>
It is possible to run dput without taking data structure locks up-front. In
many cases where we don't kill the dentry anyway, these locks are not required.
Signed-off-by: Nick Piggin <npiggin@kernel.dk>
Long lived dcache "multi-step" operations which retry on rename seq can
be starved with a lot of rename activity. If they fail after the 1st pass,
take the rename_lock for writing to avoid further starvation.
Signed-off-by: Nick Piggin <npiggin@kernel.dk>
The remaining usages for dcache_lock is to allow atomic, multi-step read-side
operations over the directory tree by excluding modifications to the tree.
Also, to walk in the leaf->root direction in the tree where we don't have
a natural d_lock ordering.
This could be accomplished by taking every d_lock, but this would mean a
huge number of locks and actually gets very tricky.
Solve this instead by using the rename seqlock for multi-step read-side
operations, retry in case of a rename so we don't walk up the wrong parent.
Concurrent dentry insertions are not serialised against. Concurrent deletes
are tricky when walking up the directory: our parent might have been deleted
when dropping locks so also need to check and retry for that.
We can also use the rename lock in cases where livelock is a worry (and it
is introduced in subsequent patch).
Signed-off-by: Nick Piggin <npiggin@kernel.dk>
Add a new lock, dcache_inode_lock, to protect the inode's i_dentry list
from concurrent modification. d_alias is also protected by d_lock.
Signed-off-by: Nick Piggin <npiggin@kernel.dk>
Protect d_subdirs and d_child with d_lock, except in filesystems that aren't
using dcache_lock for these anyway (eg. using i_mutex).
Note: if we change the locking rule in future so that ->d_child protection is
provided only with ->d_parent->d_lock, it may allow us to reduce some locking.
But it would be an exception to an otherwise regular locking scheme, so we'd
have to see some good results. Probably not worthwhile.
Signed-off-by: Nick Piggin <npiggin@kernel.dk>
Protect d_unhashed(dentry) condition with d_lock. This means keeping
DCACHE_UNHASHED bit in synch with hash manipulations.
Signed-off-by: Nick Piggin <npiggin@kernel.dk>
Make d_count non-atomic and protect it with d_lock. This allows us to ensure a
0 refcount dentry remains 0 without dcache_lock. It is also fairly natural when
we start protecting many other dentry members with d_lock.
Signed-off-by: Nick Piggin <npiggin@kernel.dk>
Add a new lock, dcache_lru_lock, to protect the dcache LRU list from concurrent
modification. d_lru is also protected by d_lock, which allows LRU lists to be
accessed without the lru lock, using RCU in future patches.
Signed-off-by: Nick Piggin <npiggin@kernel.dk>
Add a new lock, dcache_hash_lock, to protect the dcache hash table from
concurrent modification. d_hash is also protected by d_lock.
Signed-off-by: Nick Piggin <npiggin@kernel.dk>
Remove dcache_lock locking from hostfs filesystem, and move it into dcache
helpers. All that is required is a coherent path name. Protection from
concurrent modification of the namespace after path name generation is not
provided in current code, because dcache_lock is dropped before the path is
used.
Signed-off-by: Nick Piggin <npiggin@kernel.dk>
Change d_hash so it may be called from lock-free RCU lookups. See similar
patch for d_compare for details.
For in-tree filesystems, this is just a mechanical change.
Signed-off-by: Nick Piggin <npiggin@kernel.dk>
Change d_compare so it may be called from lock-free RCU lookups. This
does put significant restrictions on what may be done from the callback,
however there don't seem to have been any problems with in-tree fses.
If some strange use case pops up that _really_ cannot cope with the
rcu-walk rules, we can just add new rcu-unaware callbacks, which would
cause name lookup to drop out of rcu-walk mode.
For in-tree filesystems, this is just a mechanical change.
Signed-off-by: Nick Piggin <npiggin@kernel.dk>
smpfs and ncpfs want to update a live dentry name in-place. Rather than
have them open code the locking, provide a documented dcache API.
Signed-off-by: Nick Piggin <npiggin@kernel.dk>
Change d_delete from a dentry deletion notification to a dentry caching
advise, more like ->drop_inode. Require it to be constant and idempotent,
and not take d_lock. This is how all existing filesystems use the callback
anyway.
This makes fine grained dentry locking of dput and dentry lru scanning
much simpler.
Signed-off-by: Nick Piggin <npiggin@kernel.dk>
The nr_unused counters count the number of objects on an LRU, and as such they
are synchronized with LRU object insertion and removal and scanning, and
protected under the LRU lock.
Making it per-cpu does not actually get any concurrency improvements because of
this lock, and summing the counter is much slower, and
incrementing/decrementing it costs more code size and is slower too.
These counters should stay per-LRU, which currently means global.
Signed-off-by: Nick Piggin <npiggin@kernel.dk>
