2.6.16-rc5-mm2 - yet another memory leak...

[Valdis.Kletnieks at vt.edu]

On Sun, 05 Mar 2006 19:38:28 EST, Valdis.Kletnieks at vt.edu said:

> Yet another leak - this time it seems to be the audit_socketcall()
> in sys_socketcall().

>  199154  c02e72b2 <sys_socketcall+0x4d/0x186>

Still present in 2.6.16-rc6-mm1:

% grep socket /proc/slab_allocators 
size-1024: 3 yenta_probe+0x40/0x49f [yenta_socket]
size-64: 354 audit_socketcall+0x3e/0x84
size-32: 108634 audit_socketcall+0x3e/0x84

Uptime of just over an hour.  However, this one might not be audit's fault -
looks like there's something else leaking, and this is just along for the ride:

% grep inode /proc/slab_allocators
mqueue_inode_cache: 1 mqueue_alloc_inode+0x12/0x14
ext3_inode_cache: 13280 ext3_alloc_inode+0x12/0x41
shmem_inode_cache: 418 shmem_alloc_inode+0x12/0x1f
sock_inode_cache: 138577 sock_alloc_inode+0x12/0x51
proc_inode_cache: 403 proc_alloc_inode+0x16/0x67
bdev_cache: 33 bdev_alloc_inode+0x12/0x22
inode_cache: 967 alloc_inode+0x26/0x17a
size-2048: 10 journal_init_inode+0x81/0x118
size-512: 2 journal_init_inode+0x81/0x118
size-64: 153679 selinux_inode_alloc_security+0x25/0x71

Lots of sock_alloc_inode's left dangling, which is probably dragging the
selinux_inode_alloc_security and audit context along with it.

lsof | grep sock | wc -l  says there's only 62 sockets open on the system.

Looks like time to go wade through and figure out where the sock_alloc_inode
stuff is going astray.  Yee. Hah.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 228 bytes
Desc: not available
URL: <http://listman.redhat.com/archives/linux-audit/attachments/20060315/26333487/attachment.sig>