lkml.org 
[lkml]   [2011]   [Apr]   [18]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
SubjectRe: query: [PATCH 2/2] cgroup: Remove call to synchronize_rcu in cgroup_attach_task
From
Date
On Wed, 2011-04-13 at 15:16 +0200, Paul Menage wrote:

> Also, I'd be concerned that subsystems might get confused by the fact
> that a new group called 'foo' could be created before the old 'foo'
> has been cleaned up?

Decided to try beating on that without my hack. Seems these patches...
patches/cgroup-Set-CGRP_RELEASABLE-when-adding-to-a-group
patches/cgroup-Remove-call-to-synchronize_rcu-in-cgroup_attach_task
...either introduced a bug, or made an existing one easier to hit. With
only those two applied, and several tasks doing - create, attach,
detach, fork off a rmdir - all for the same cgroup, box griped.

With my hack on top, I'd either see rcu workers explode, or list
corruption in cgroup_attach_task(), depending on phase-of-moon.

(hohum, maybe it'll turn out to be a defenseless _baby_ dragon..)

v2.6.39-rc2

[ 5009.808292] general protection fault: 0000 [#1] SMP
[ 5009.812007] last sysfs file: /sys/devices/system/cpu/cpu3/cache/index2/shared_cpu_map
[ 5009.812007] CPU 2
[ 5009.812007] Modules linked in: snd_pcm_oss snd_mixer_oss snd_seq snd_seq_device edd nfsd lockd nfs_acl auth_rpcgss sunrpc parport_pc parport bridge stp cpufreq_conservative cpufreq_ondemand microcode cpufreq_userspace cpufreq_powersave acpi_cpufreq mperf nls_iso8859_1 nls_cp437 vfat fat fuse ext3 jbd dm_mod snd_hda_codec_realtek snd_hda_intel snd_hda_codec snd_hwdep snd_pcm sr_mod sg usb_storage usbmouse cdrom snd_timer usbhid hid firewire_ohci firewire_core snd usb_libusual i2c_i801 soundcore snd_page_alloc e1000e crc_itu_t button ext4 mbcache jbd2 crc16 uhci_hcd ehci_hcd sd_mod rtc_cmos usbcore rtc_core rtc_lib ahci libahci libata scsi_mod fan processor thermal
[ 5009.812007]
[ 5009.812007] Pid: 21021, comm: rmdir Not tainted 2.6.39-smpx #1906 MEDIONPC MS-7502/MS-7502
[ 5009.812007] RIP: 0010:[<ffffffff81079bb5>] [<ffffffff81079bb5>] cgroup_css_sets_empty+0x25/0x50
[ 5009.812007] RSP: 0018:ffff88017dfe5dd8 EFLAGS: 00010246
[ 5009.812007] RAX: dead000000100100 RBX: 0000000000000000 RCX: ffff880228c8ce50
[ 5009.812007] RDX: dead000000100100 RSI: ffff88022db9b900 RDI: ffff880228c8cc00
[ 5009.812007] RBP: ffff88017dfe5dd8 R08: 0000000000000005 R09: ffff88022db9b93c
[ 5009.812007] R10: ffff88017dfe5d18 R11: ffff88017dfe5c84 R12: ffff880228c8cc00
[ 5009.812007] R13: ffff88017dfe5df8 R14: ffff880228c8cc20 R15: ffff8802161cd240
[ 5009.812007] FS: 00007fefea0a1700(0000) GS:ffff88022fd00000(0000) knlGS:0000000000000000
[ 5009.812007] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[ 5009.955702] CR2: 00007fefe9c1acb0 CR3: 000000019107c000 CR4: 00000000000006e0
[ 5009.955702] DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
[ 5009.955702] DR3: 0000000000000000 DR6: 00000000ffff0ff0 DR7: 0000000000000400
[ 5009.955702] Process rmdir (pid: 21021, threadinfo ffff88017dfe4000, task ffff8802161cd240)
[ 5009.955702] Stack:
[ 5009.955702] ffff88017dfe5e58 ffffffff8107a514 ffff88022db9b9a0 ffff88022db9badc
[ 5009.955702] 0000000000000000 ffff8802161cd240 ffffffff81057ee0 ffff88017dfe5e10
[ 5009.955702] ffff88017dfe5e10 0000000000000000 0000000000000001 ffff88022db9b900
[ 5009.955702] Call Trace:
[ 5009.955702] [<ffffffff8107a514>] cgroup_rmdir+0x64/0x510
[ 5009.955702] [<ffffffff81057ee0>] ? wake_up_bit+0x40/0x40
[ 5009.955702] [<ffffffff810d46c6>] vfs_rmdir+0x96/0xe0
[ 5009.955702] [<ffffffff810d6313>] do_rmdir+0x103/0x120
[ 5009.955702] [<ffffffff810d6e21>] sys_rmdir+0x11/0x20
[ 5009.955702] [<ffffffff8134db7b>] system_call_fastpath+0x16/0x1b
[ 5009.955702] Code: 84 00 00 00 00 00 55 48 8b 87 50 02 00 00 48 89 e5 48 8d 8f 50 02 00 00 eb 11 0f 1f 40 00 48 8b 40 28 8b 00 85 c0 7f 1e 48 89 d0
[ 5009.955702] 8b 10 48 39 c8 0f 18 0a 75 e8 b8 01 00 00 00 c9 c3 66 0f 1f
[ 5009.955702] RIP [<ffffffff81079bb5>] cgroup_css_sets_empty+0x25/0x50
[ 5009.955702] RSP <ffff88017dfe5dd8>
[ 5010.075058] hpet1: lost 12 rtc interrupts
[ 5010.079153] ---[ end trace df95d1af02295a25 ]---

(gdb) list *cgroup_css_sets_empty+0x25
0xffffffff81079df5 is in cgroup_css_sets_empty (kernel/cgroup.c:3589).
3584 * Must be called with css_set_lock held */
3585 static int cgroup_css_sets_empty(struct cgroup *cgrp)
3586 {
3587 struct cg_cgroup_link *link;
3588
3589 list_for_each_entry(link, &cgrp->css_sets, cgrp_link_list) {
3590 struct css_set *cg = link->cg;
3591 if (atomic_read(&cg->refcount) > 0)
3592 return 0;
3593 }
(gdb)




\
 
 \ /
  Last update: 2011-04-18 16:23    [W:0.114 / U:0.340 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site