AppArmor causes hard lock when free vmalloced memory in interrupt context

Bug #562044 reported by John Johansen
12
This bug affects 2 people
Affects Status Importance Assigned to Milestone
linux (Ubuntu)
In Progress
Undecided
John Johansen

Bug Description

AppArmor will cause a hard lockup due to calling vfree in interrupt context, backtrace attached below

Revision history for this message
John Johansen (jjohansen) wrote :
Download full text (8.6 KiB)

[ 80.250147] ------------[ cut here ]------------
[ 80.251233] kernel BUG at /home/jj/ubuntu/lucid/ubuntu-lucid/mm/vmalloc.c:1406!
[ 80.252769] invalid opcode: 0000 [#1] SMP
[ 80.253702] last sysfs file: /sys/devices/pci0000:00/0000:00:04.0/class
[ 80.255107] CPU 0
[ 80.255559] Modules linked in: binfmt_misc ppdev lp parport fbcon tileblit font bitblit softcursor psmouse serio_raw vga16fb vgastate i2c_piix4 virtio_blk virtio_net floppy virtio_pci virtio_ring virtio
[ 80.259814] Pid: 671, comm: Xorg Not tainted 2.6.32-20-generic #30 Bochs
[ 80.260019] RIP: 0010:[<ffffffff8111f65c>] [<ffffffff8111f65c>] vfree+0x2c/0x30
[ 80.260019] RSP: 0018:ffff880001c03d50 EFLAGS: 00010206
[ 80.260019] RAX: 0000000000000100 RBX: 0000000000000000 RCX: ffffffff817af230
[ 80.260019] RDX: 000000000020e000 RSI: ffffffff8127df30 RDI: ffffc9000020e000
[ 80.260019] RBP: ffff880001c03d50 R08: 0000000000000000 R09: 0000000000000000
[ 80.260019] R10: 0000000000000000 R11: 0000000000000001 R12: ffff88001cc2b540
[ 80.260019] R13: ffff88001f8c5840 R14: ffff88001f8c5840 R15: ffff880001c10610
[ 80.260019] FS: 00007f919246c700(0000) GS:ffff880001c00000(0000) knlGS:0000000000000000
[ 80.260019] CS: 0010 DS: 0000 ES: 0000 CR0: 000000008005003b
[ 80.260019] CR2: 00007f0a62c70000 CR3: 000000001085a000 CR4: 00000000000006f0
[ 80.260019] DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
[ 80.260019] DR3: 0000000000000000 DR6: 00000000ffff0ff0 DR7: 0000000000000400
[ 80.260019] Process Xorg (pid: 671, threadinfo ffff880011db0000, task ffff8800127a44d0)
[ 80.260019] Stack:
[ 80.260019] ffff880001c03d60 ffffffff8127dd15 ffff880001c03d80 ffffffff8127df0a
[ 80.260019] <0> ffff88001cc2b540 ffffffff8127df30 ffff880001c03d90 ffffffff8127df3e
[ 80.260019] <0> ffff880001c03db0 ffffffff812b45d7 ffff880012425800 ffffffff81280e80
[ 80.260019] Call Trace:
[ 80.260019] <IRQ>
[ 80.260019] [<ffffffff8127dd15>] free_table+0x45/0x50
[ 80.260019] [<ffffffff8127df0a>] aa_dfa_free+0x2a/0x50
[ 80.260019] [<ffffffff8127df30>] ? aa_dfa_free_kref+0x0/0x10
[ 80.260019] [<ffffffff8127df3e>] aa_dfa_free_kref+0xe/0x10
[ 80.260019] [<ffffffff812b45d7>] kref_put+0x37/0x70
[ 80.260019] [<ffffffff81280e80>] ? aa_free_profile_kref+0x0/0x20
[ 80.260019] [<ffffffff81280dfa>] aa_free_profile+0xaa/0x130
[ 80.260019] [<ffffffff81280e92>] aa_free_profile_kref+0x12/0x20
[ 80.260019] [<ffffffff812b45d7>] kref_put+0x37/0x70
[ 80.260019] [<ffffffff8127d4ce>] aa_free_task_context+0x2e/0x70
[ 80.260019] [<ffffffff812829ea>] apparmor_cred_free+0x1a/0x30
[ 80.260019] [<ffffffff81250d86>] security_cred_free+0x16/0x20
[ 80.260019] [<ffffffff8108b6e6>] put_cred_rcu+0x26/0xb0
[ 80.260019] [<ffffffff810c9f9d>] __rcu_process_callbacks+0x10d/0x340
[ 80.260019] [<ffffffff81038a61>] ? kvm_clock_read+0x21/0x30
[ 80.260019] [<ffffffff810ca1fb>] rcu_process_callbacks+0x2b/0x50
[ 80.260019] [<ffffffff8106e3a7>] __do_softirq+0xb7/0x1e0
[ 80.260019] [<ffffffff8109445a>] ? tick_program_event+0x2a/0x30
[ 80.260019] [<ffffffff810142ec>] call_softirq+0x1c/0x30
[ 80.260019] [<ffffffff81015cb5>] do_softirq+0x65/0xa0...

Read more...

Changed in linux (Ubuntu):
assignee: nobody → John Johansen (jjohansen)
status: New → In Progress
Revision history for this message
Alex Harrington (alexharrington) wrote :

We think we're seeing this problem - very similar console output after the crash on two HP DL320s servers running 10.04 64 bit. They've been rock solid for 3 years running 8.04 but started crashing about once every two weeks after the upgrade to 10.04.

Revision history for this message
John Johansen (jjohansen) wrote :

Alex,

Do you have a log log files, or console output that you can attach? If not test could you take a picture of the your console?

Revision history for this message
Alex Harrington (alexharrington) wrote :

John,

I'm afraid I don't. Next time they go I'll grab a photo. I've disabled and uninstalled AppArmor on them and added a further 4GB of RAM which I'm hoping will work around it. Unfortunately they're serving 600 odd workstations in a production environment so I don't want to do anything that's likely to make it happen.

Alex

To post a comment you must log in.
This report contains Public information  
Everyone can see this information.

Other bug subscribers

Remote bug watches

Bug watches keep track of this bug in other bug trackers.