Fix afs_lookup() to not clobber the version set on a new dentry by
afs_do_lookup() - especially as it's using the wrong version of the
version (we need to use the one given to us by whatever op the dir
contents correspond to rather than what's in the afs_vnode).

Fixes: 9dd0b82ef530 ("afs: Fix missing dentry data version updating")
Signed-off-by: David Howells <dhowells@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

afs_lookup() has a tracepoint to indicate the outcome of
d_splice_alias(), passing it the inode to retrieve the fid from.
However, the function gave up its ref on that inode when it called
d_splice_alias(), which may have failed and dropped the inode.

Fix this by caching the fid.

Fixes: 80548b03991f ("afs: Add more tracepoints")
Reported-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: David Howells <dhowells@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

When a lookup is done, the afs filesystem will perform a bulk status-fetch
operation on the requested vnode (file) plus the next 49 other vnodes from
the directory list (in AFS, directory contents are downloaded as blobs and
parsed locally).  When the results are received, it will speculatively
populate the inode cache from the extra data.

However, if the lookup races with another lookup on the same directory, but
for a different file - one that's in the 49 extra fetches, then if the bulk
status-fetch operation finishes first, it will try and update the inode
from the other lookup.

If this other inode is still in the throes of being created, however, this
will cause an assertion failure in afs_apply_status():

	BUG_ON(test_bit(AFS_VNODE_UNSET, &vnode->flags));

on or about fs/afs/inode.c:175 because it expects data to be there already
that it can compare to.

Fix this by skipping the update if the inode is being created as the
creator will presumably set up the inode with the same information.

Fixes: 39db9815da48 ("afs: Fix application of the results of a inline bulk status fetch")
Signed-off-by: David Howells <dhowells@redhat.com>
Reviewed-by: Marc Dionne <marc.dionne@auristor.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

Make afs_permission() and afs_d_revalidate() do initial checks in RCU-mode
pathwalk to reduce latency in pathwalk elements that get done multiple
times.  We don't need to query the server unless we've received a
notification from it that something has changed or the callback has
expired.

This requires that we can request a key and check permits under RCU
conditions if we need to.

Signed-off-by: David Howells <dhowells@redhat.com>

The afs_lookup trace event can cause the following:

[  216.576777] BUG: kernel NULL pointer dereference, address: 000000000000023b
[  216.576803] #PF: supervisor read access in kernel mode
[  216.576813] #PF: error_code(0x0000) - not-present page
...
[  216.576913] RIP: 0010:trace_event_raw_event_afs_lookup+0x9e/0x1c0 [kafs]

If the inode from afs_do_lookup() is an error other than ENOENT, or if it
is ENOENT and afs_try_auto_mntpt() returns an error, the trace event will
try to dereference the error pointer as a valid pointer.

Use IS_ERR_OR_NULL to only pass a valid pointer for the trace, or NULL.

Ideally the trace would include the error value, but for now just avoid
the oops.

Fixes: 80548b03991f ("afs: Add more tracepoints")
Signed-off-by: Marc Dionne <marc.dionne@auristor.com>
Signed-off-by: David Howells <dhowells@redhat.com>

In the in-kernel afs filesystem, the d_fsdata dentry field is used to hold
the data version of the parent directory when it was created or when
d_revalidate() last caused it to be updated.  This is compared to the
->invalid_before field in the directory inode, rather than the actual data
version number, thereby allowing changes due to local edits to be ignored.
Only if the server data version gets bumped unexpectedly (eg. by a
competing client), do we need to revalidate stuff.

However, the d_fsdata field should also be updated if an rpc op is
performed that modifies that particular dentry.  Such ops return the
revised data version of the directory(ies) involved, so we should use that.

This is particularly problematic for rename, since a dentry from one
directory may be moved directly into another directory (ie. mv a/x b/x).
It would then be sporting the wrong data version - and if this is in the
future, for the destination directory, revalidations would be missed,
leading to foreign renames and hard-link deletion being missed.

Fix this by the following means:

 (1) Return the data version number from operations that read the directory
     contents - if they issue the read.  This starts in afs_dir_iterate()
     and is used, ignored or passed back by its callers.

 (2) In afs_lookup*(), set the dentry version to the version returned by
     (1) before d_splice_alias() is called and the dentry published.

 (3) In afs_d_revalidate(), set the dentry version to that returned from
     (1) if an rpc call was issued.  This means that if a parallel
     procedure, such as mkdir(), modifies the directory, we won't
     accidentally use the data version from that.

 (4) In afs_{mkdir,create,link,symlink}(), set the new dentry's version to
     the directory data version before d_instantiate() is called.

 (5) In afs_{rmdir,unlink}, update the target dentry's version to the
     directory data version as soon as we've updated the directory inode.

 (6) In afs_rename(), we need to unhash the old dentry before we start so
     that we don't get afs_d_revalidate() reverting the version change in
     cross-directory renames.

     We then need to set both the old and the new dentry versions the data
     version of the new directory before we call d_move() as d_move() will
     rehash them.

Fixes: 1da177e4c3f4 ("Linux-2.6.12-rc2")
Signed-off-by: David Howells <dhowells@redhat.com>

In the in-kernel afs filesystem, d_fsdata is set with the data version of
the parent directory.  afs_d_revalidate() will update this to the current
directory version, but it shouldn't do this if it the value it read from
d_fsdata is the same as no lock is held and cmpxchg() is not used.

Fix the code to only change the value if it is different from the current
directory version.

Fixes: 260a980317da ("[AFS]: Add "directory write" support.")
Signed-off-by: David Howells <dhowells@redhat.com>

When afs_rename() calculates the expected data version of the target
directory in a cross-directory rename, it doesn't increment it as it
should, so it always thinks that the target inode is unexpectedly modified
on the server.

Fixes: a58823ac4589 ("afs: Fix application of status and callback to be under same lock")
Signed-off-by: David Howells <dhowells@redhat.com>

Pull afs updates from David Howells:
 "A set of minor changes for AFS:

   - Remove an unnecessary check in afs_unlink()

   - Add a tracepoint for tracking callback management

   - Add a tracepoint for afs_server object usage

   - Use struct_size()

   - Add mappings for AFS UAE abort codes to Linux error codes, using
     symbolic names rather than hex numbers in the .c file"

* tag 'afs-next-20190628' of git://git.kernel.org/pub/scm/linux/kernel/git/dhowells/linux-fs:
  afs: Add support for the UAE error table
  fs/afs: use struct_size() in kzalloc()
  afs: Trace afs_server usage
  afs: Add some callback management tracepoints
  afs: afs_unlink() doesn't need to check dentry->d_inode

As Gustavo said in other patches doing the same replace, we can now
use the new struct_size() helper to avoid leaving these open-coded and
prone to type mistake.

Signed-off-by: Zhengyuan Liu <liuzhengyuan@kylinos.cn>
Signed-off-by: David Howells <dhowells@redhat.com>