linux-stable.git/fs/sysfs/dir.c, branch v3.5

sysfs: get rid of some lockdep false positives

2012-05-14T19:19:56+00:00

This patch (as1554) fixes a lockdep false-positive report.  The
problem arises because lockdep is unable to deal with the
tree-structured locks created by the device core and sysfs.

This particular problem involves a sysfs attribute method that
unregisters itself, not from the device it was called for, but from a
descendant device.  Lockdep doesn't understand the distinction and
reports a possible deadlock, even though the operation is safe.

This is the sort of thing that would normally be handled by using a
nested lock annotation; unfortunately it's not feasible to do that
here.  There's no sensible way to tell sysfs when attribute removal
occurs in the context of a parent attribute method.

As a workaround, the patch adds a new flag to struct attribute
telling sysfs not to inform lockdep when it acquires a readlock on a
sysfs_dirent instance for the attribute.  The readlock is still
acquired, but lockdep doesn't know about it and hence does not
complain about impossible deadlock scenarios.

Also added are macros for static initialization of attribute
structures with the ignore_lockdep flag set.  The three offending
attributes in the USB subsystem are converted to use the new macros.

Signed-off-by: Alan Stern 
Acked-by: Tejun Heo 
CC: Eric W. Biederman 
CC: Peter Zijlstra 
Signed-off-by: Greg Kroah-Hartman

sysfs: Removed dup_name entirely in sysfs_rename

2012-05-02T21:55:09+00:00

Since no one using "dup_name", removed it completely in sysfs_rename.

Signed-off-by: Sasikantha babu 
Signed-off-by: Greg Kroah-Hartman

sysfs: handle 'parent deleted before child added'

2012-04-10T21:48:51+00:00

In scsi at least two cases of the parent device being deleted before the
child is added have been observed.

1/ scsi is performing async scans and the device is removed prior to the
   async can thread running (can happen with an in-opportune / unlikely
   unplug during initial scan).

2/ libsas discovery event running after the parent port has been torn
   down (this is a bug in libsas).

Result in crash signatures like:
 BUG: unable to handle kernel NULL pointer dereference at 0000000000000098
 IP: [] sysfs_create_dir+0x32/0xb6
 ...
 Process scsi_scan_8 (pid: 5417, threadinfo ffff88080bd16000, task ffff880801b8a0b0)
 Stack:
  00000000fffffffe ffff880813470628 ffff88080bd17cd0 ffff88080614b7e8
  ffff88080b45c108 00000000fffffffe ffff88080bd17d20 ffffffff8125e4a8
  ffff88080bd17cf0 ffffffff81075149 ffff88080bd17d30 ffff88080614b7e8
 Call Trace:
  [] kobject_add_internal+0x120/0x1e3
  [] ? trace_hardirqs_on+0xd/0xf
  [] kobject_add_varg+0x41/0x50
  [] kobject_add+0x64/0x66
  [] device_add+0x12d/0x63a

In this scenario the parent is still valid (because we have a
reference), but it has been device_del()'d which means its kobj->sd
pointer is NULL'd via:

 device_del()->kobject_del()->sysfs_remove_dir()

...and then sysfs_create_dir() (without this fix) goes ahead and
de-references parent_sd via sysfs_ns_type():

 return (sd->s_flags & SYSFS_NS_TYPE_MASK) >> SYSFS_NS_TYPE_SHIFT;

This scenario is being fixed in scsi/libsas, but if other subsystems
present the same ordering the system need not immediately crash.

Cc: Greg Kroah-Hartman 
Cc: James Bottomley 
Signed-off-by: Dan Williams 
Signed-off-by: Greg Kroah-Hartman

sysfs: Update the name hash for an entry after changing the namespace

2012-04-09T22:33:00+00:00

This is needed to allow renaming network devices that have been moved
to another network namespace.

Signed-off-by: Tom Goff 
Signed-off-by: Greg Kroah-Hartman

Revert "sysfs: Kill nlink counting."

2012-03-08T21:03:10+00:00

This reverts commit 524b6c5b39b931311dfe5a2f5abae2f5c9731676.

It has shown to break userspace tools, which is not acceptable.

Reported-by: Jiri Slaby 
Cc: Eric W. Biederman 
Cc: Linus Torvalds 
Signed-off-by: Greg Kroah-Hartman

sysfs: Update the name hash when renaming sysfs entries

2012-01-31T19:30:48+00:00

This fixes a bug introduced with sysfs name hashes where renaming a
network device appears to succeed but silently makes the sysfs files for
that network device inaccessible.

In at least one configuration this bug has stopped networking from
coming up during boot.

Signed-off-by: Eric W. Biederman 
Tested-by: Jiri Slaby 
Signed-off-by: Greg Kroah-Hartman

sysfs: Kill nlink counting.

2012-01-24T20:41:46+00:00

Tracking the number of subdirectories requires an extra field that increases
the size of sysfs_dirent.  nlinks are not particularly interesting for sysfs
and the nlink counts are wrong when network namespaces are involved so stop
counting them, and always return nlink == 1.  Userspace already knows that
directories with nlink == 1 have an nlink count they can't use to count
subdirectories.

This reduces the size of sysfs_dirent by 8 bytes on 64bit platforms.

Signed-off-by: Eric W. Biederman 
Signed-off-by: Greg Kroah-Hartman

sysfs: Store the sysfs inode in an unsigned int.

2012-01-24T20:41:46+00:00

Store the sysfs inode number in an unsided int because
ida inode allocator can return at most a 31 bit number,
reducing the size of struct sysfs_dirent by 8 bytes
on 64bit platforms.

Signed-off-by: Eric W. Biederman 
Signed-off-by: Greg Kroah-Hartman

sysfs: Add s_hash to sysfs_dirent and order directory entries by hash

2012-01-24T20:36:17+00:00

Compute a 31 bit hash of directory entries (that can fit in a signed
32bit off_t) and index the sysfs directory entries by that hash,
replacing the per directory indexes by name and by inode.  Because we
now only use a single rbtree this reduces the size of sysfs_dirent by 2
pointers.  Because we have fewer cases to deal with the code is now
simpler.

For now I use the simple hash that the dcache uses as that is easy to
use and seems simple enough.

In addition to makeing the code simpler using a hash for the file
position in readdir brings sysfs in line with other filesystems that
have non-trivial directory structures.

Signed-off-by: Eric W. Biederman 
Signed-off-by: Greg Kroah-Hartman

sysfs: Make sysfs_rename safe with sysfs_dirents in rbtrees.

2011-11-01T16:16:14+00:00

In sysfs_rename we need to remove the optimization of not calling
sysfs_unlink_sibling and sysfs_link_sibling if the renamed parent
directory is not changing.  This optimization is no longer valid now
that sysfs dirents are stored in an rbtree sorted by name.

Move the assignment of s_ns before the call of sysfs_link_sibling.  With
no sysfs_dirent fields changing after the call of sysfs_link_sibling
this allows sysfs_link_sibling to take any of the directory entries into
account when it builds the rbtrees, and s_ns looks like a prime canidate
to be used in the rbtree in the future.

Signed-off-by: Eric W. Biederman 
Cc: Jiri Slaby 
Cc: Greg KH 
Cc: David Miller 
Cc: Mikulas Patocka 
Cc: Andrew Morton 
Signed-off-by: Linus Torvalds