linux.git/fs/dcache.c, branch v3.10.76

dcache: Fix locking bugs in backported "deal with deadlock in d_walk()"

2015-04-29T08:34:02+00:00

commit 20defcec264ceab2630356fb9d397f3d237b5e6d upstream in 3.2-stable

Steven Rostedt reported:
> Porting -rt to the latest 3.2 stable tree I triggered this bug:
>
> =====================================
> [ BUG: bad unlock balance detected! ]
> -------------------------------------
> rm/1638 is trying to release lock (rcu_read_lock) at:
> [] rcu_read_unlock+0x0/0x23
> but there are no more locks to release!
>
> other info that might help us debug this:
> 2 locks held by rm/1638:
>  #0:  (&sb->s_type->i_mutex_key#9/1){+.+.+.}, at: [] do_rmdir+0x5f/0xd2
>  #1:  (&sb->s_type->i_mutex_key#9){+.+.+.}, at: [] vfs_rmdir+0x49/0xac
>
> stack backtrace:
> Pid: 1638, comm: rm Not tainted 3.2.66-test-rt96+ #2
> Call Trace:
>  [] ? printk+0x1d/0x1f
>  [] print_unlock_inbalance_bug+0xc3/0xcd
>  [] lock_release_non_nested+0x98/0x1ec
>  [] ? trace_hardirqs_off_caller+0x18/0x90
>  [] ? local_clock+0x2d/0x50
>  [] ? d_hash+0x2f/0x2f
>  [] ? d_hash+0x2f/0x2f
>  [] lock_release+0x192/0x1ad
>  [] rcu_read_unlock+0x17/0x23
>  [] shrink_dcache_parent+0x227/0x270
>  [] vfs_rmdir+0x68/0xac
>  [] do_rmdir+0x98/0xd2
>  [] ? fput+0x1a3/0x1ab
>  [] ? sysenter_exit+0xf/0x1a
>  [] ? trace_hardirqs_on_caller+0x118/0x149
>  [] sys_unlinkat+0x2b/0x35
>  [] sysenter_do_call+0x12/0x12
>
>
>
>
> There's a path to calling rcu_read_unlock() without calling
> rcu_read_lock() in have_submounts().
>
> 	goto positive;
>
> positive:
> 	if (!locked && read_seqretry(&rename_lock, seq))
> 		goto rename_retry;
>
> rename_retry:
> 	rcu_read_unlock();
>
> in the above path, rcu_read_lock() is never done before calling
> rcu_read_unlock();

I reviewed locking contexts in all three functions that I changed when
backporting "deal with deadlock in d_walk()".  It's actually worse
than this:

- We don't hold this_parent->d_lock at the 'positive' label in
  have_submounts(), but it is unlocked after 'rename_retry'.
- There is an rcu_read_unlock() after the 'out' label in
  select_parent(), but it's not held at the 'goto out'.

Fix all three lock imbalances.

Reported-by: Steven Rostedt 
Signed-off-by: Ben Hutchings 
Tested-by: Steven Rostedt 
Signed-off-by: Greg Kroah-Hartman

deal with deadlock in d_walk()

2015-04-29T08:34:00+00:00

commit ca5358ef75fc69fee5322a38a340f5739d997c10 upstream.

... by not hitting rename_retry for reasons other than rename having
happened.  In other words, do _not_ restart when finding that
between unlocking the child and locking the parent the former got
into __dentry_kill().  Skip the killed siblings instead...

Signed-off-by: Al Viro 
Cc: Ben Hutchings 
[hujianyang: Backported to 3.10 refer to the work of Ben Hutchings in 3.2:
 - As we only have try_to_ascend() and not d_walk(), apply this
   change to all callers of try_to_ascend()
 - Adjust context to make __dentry_kill() apply to d_kill()]
Signed-off-by: hujianyang 
Signed-off-by: Greg Kroah-Hartman

move d_rcu from overlapping d_child to overlapping d_alias

2015-04-29T08:34:00+00:00

commit 946e51f2bf37f1656916eb75bd0742ba33983c28 upstream.

Signed-off-by: Al Viro 
Cc: Ben Hutchings 
[hujianyang: Backported to 3.10 refer to the work of Ben Hutchings in 3.2:
 - Apply name changes in all the different places we use d_alias and d_child
 - Move the WARN_ON() in __d_free() to d_free() as we don't have dentry_free()]
Signed-off-by: hujianyang 
Signed-off-by: Greg Kroah-Hartman

vfs: fix bad hashing of dentries

2014-09-17T16:04:02+00:00

commit 99d263d4c5b2f541dfacb5391e22e8c91ea982a6 upstream.

Josef Bacik found a performance regression between 3.2 and 3.10 and
narrowed it down to commit bfcfaa77bdf0 ("vfs: use 'unsigned long'
accesses for dcache name comparison and hashing"). He reports:

 "The test case is essentially

      for (i = 0; i < 1000000; i++)
              mkdir("a$i");

  On xfs on a fio card this goes at about 20k dir/sec with 3.2, and 12k
  dir/sec with 3.10.  This is because we spend waaaaay more time in
  __d_lookup on 3.10 than in 3.2.

  The new hashing function for strings is suboptimal for <
  sizeof(unsigned long) string names (and hell even > sizeof(unsigned
  long) string names that I've tested).  I broke out the old hashing
  function and the new one into a userspace helper to get real numbers
  and this is what I'm getting:

      Old hash table had 1000000 entries, 0 dupes, 0 max dupes
      New hash table had 12628 entries, 987372 dupes, 900 max dupes
      We had 11400 buckets with a p50 of 30 dupes, p90 of 240 dupes, p99 of 567 dupes for the new hash

  My test does the hash, and then does the d_hash into a integer pointer
  array the same size as the dentry hash table on my system, and then
  just increments the value at the address we got to see how many
  entries we overlap with.

  As you can see the old hash function ended up with all 1 million
  entries in their own bucket, whereas the new one they are only
  distributed among ~12.5k buckets, which is why we're using so much
  more CPU in __d_lookup".

The reason for this hash regression is two-fold:

 - On 64-bit architectures the down-mixing of the original 64-bit
   word-at-a-time hash into the final 32-bit hash value is very
   simplistic and suboptimal, and just adds the two 32-bit parts
   together.

   In particular, because there is no bit shuffling and the mixing
   boundary is also a byte boundary, similar character patterns in the
   low and high word easily end up just canceling each other out.

 - the old byte-at-a-time hash mixed each byte into the final hash as it
   hashed the path component name, resulting in the low bits of the hash
   generally being a good source of hash data.  That is not true for the
   word-at-a-time case, and the hash data is distributed among all the
   bits.

The fix is the same in both cases: do a better job of mixing the bits up
and using as much of the hash data as possible.  We already have the
"hash_32|64()" functions to do that.

Reported-by: Josef Bacik 
Cc: Al Viro 
Cc: Christoph Hellwig 
Cc: Chris Mason 
Cc: linux-fsdevel@vger.kernel.org
Signed-off-by: Linus Torvalds 
Signed-off-by: Greg Kroah-Hartman

dcache.c: get rid of pointless macros

2014-09-17T16:04:02+00:00

commit 482db9066199813d6b999b65a3171afdbec040b6 upstream.

D_HASH{MASK,BITS} are used once each, both in the same function (d_hash()).
At this point they are actively misguiding - they imply that values are
compiler constants, which is no longer true.

Signed-off-by: Al Viro 
Signed-off-by: Greg Kroah-Hartman

vfs: In d_path don't call d_dname on a mount point

2014-01-25T16:27:11+00:00

commit f48cfddc6729ef133933062320039808bafa6f45 upstream.

Aditya Kali (adityakali@google.com) wrote:
> Commit bf056bfa80596a5d14b26b17276a56a0dcb080e5:
> "proc: Fix the namespace inode permission checks." converted
> the namespace files into symlinks. The same commit changed
> the way namespace bind mounts appear in /proc/mounts:
>   $ mount --bind /proc/self/ns/ipc /mnt/ipc
> Originally:
>   $ cat /proc/mounts | grep ipc
>   proc /mnt/ipc proc rw,nosuid,nodev,noexec 0 0
>
> After commit bf056bfa80596a5d14b26b17276a56a0dcb080e5:
>   $ cat /proc/mounts | grep ipc
>   proc ipc:[4026531839] proc rw,nosuid,nodev,noexec 0 0
>
> This breaks userspace which expects the 2nd field in
> /proc/mounts to be a valid path.

The symlink /proc//ns/{ipc,mnt,net,pid,user,uts} point to
dentries allocated with d_alloc_pseudo that we can mount, and
that have interesting names printed out with d_dname.

When these files are bind mounted /proc/mounts is not currently
displaying the mount point correctly because d_dname is called instead
of just displaying the path where the file is mounted.

Solve this by adding an explicit check to distinguish mounted pseudo
inodes and unmounted pseudo inodes.  Unmounted pseudo inodes always
use mount of their filesstem as the mnt_root  in their path making
these two cases easy to distinguish.

Acked-by: Serge Hallyn 
Reported-by: Aditya Kali 
Signed-off-by: "Eric W. Biederman" 
Signed-off-by: Greg Kroah-Hartman

cope with potentially long ->d_dname() output for shmem/hugetlb

2013-10-18T14:45:45+00:00

commit 118b23022512eb2f41ce42db70dc0568d00be4ba upstream.

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 
Cc: Colin Cross 
Signed-off-by: Greg Kroah-Hartman

vfs: use list_move instead of list_del/list_add

2013-05-04T19:43:02+00:00

Using list_move() instead of list_del() + list_add().

Signed-off-by: Wei Yongjun 
Signed-off-by: Al Viro

fs: remove dentry_lru_prune()

2013-05-04T19:04:01+00:00

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 
Signed-off-by: Al Viro

fs/dcache.c: add cond_resched() to shrink_dcache_parent()

2013-05-01T00:04:00+00:00

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 
Signed-off-by: Dave Chinner 
Cc: 
Signed-off-by: Andrew Morton 
Signed-off-by: Linus Torvalds