Re: mac80211 breaks suspend to disk

[Maxim Levitsky <maximlevitsky@gmail.com>]

On Wed, 2009-11-18 at 15:28 -0500, John W. Linville wrote: 
> On Wed, Nov 18, 2009 at 08:27:16PM +0200, Maxim Levitsky wrote:
> > Hi,
> > 
> > I finally managed to track down a rare hang on resume from disk I see
> > for some prolonged time.
> > Symptoms are that sometimes on resume, system almost hangs.
> > This is I can switch VTs and execute SysRQ sequences.
> > This is typical when oops happens.
> > 
> > I tracked this to BUG_ON in add_timer.
> > I replaced this with WARN_ON, and here is the backtrace:
> > 
> > 
> >   331.334935] WARNING: at /home/maxim/software/kernel/linux-2.6/kernel/timer.c:791 add_timer+0x36/0x40()
> > [  331.347374] Hardware name: Aspire 5720     
> > [  331.359725] Modules linked in: nvidia(P)  af_packet usb_storage usb_libusual cpufreq_powersave iwl3945 nfsd snd_hda_codec_realtek cpufreq_conservative exportfs iwlcore cpufreq_userspace uvcvideo snd_hda_intel acpi_cpufreq nfs videodev mac80211 coretemp lockd snd_hda_codec joydev v4l1_compat tg3 nfs_acl ohci1394 v4l2_compat_ioctl32 snd_hwdep uhci_hcd sbp2 cfg80211 psmouse video ehci_hcd auth_rpcgss libphy lirc_ene0100 ieee1394 output snd_pcm usbcore serio_raw rfkill evdev snd_page_alloc sunrpc fuse lzo lzo_decompress lzo_compress
> > [  331.428619] Pid: 4394, comm: pm-hibernate Tainted: P           2.6.32-rc7-wl #183
> > [  331.442747] Call Trace:
> > [  331.456760]  [<ffffffff81043a98>] warn_slowpath_common+0x78/0xb0
> > [  331.471119]  [<ffffffff81043adf>] warn_slowpath_null+0xf/0x20
> > [  331.485475]  [<ffffffff8104f996>] add_timer+0x36/0x40
> > [  331.499814]  [<ffffffffa01fb34c>] ieee80211_sta_restart+0x4c/0x50 [mac80211]
> > [  331.514314]  [<ffffffffa020c86c>] ieee80211_reconfig+0x36c/0x420 [mac80211]
> > [  331.528678]  [<ffffffffa020c65c>] ? ieee80211_reconfig+0x15c/0x420 [mac80211]
> > [  331.543094]  [<ffffffffa0202fa5>] ieee80211_resume+0x15/0x20 [mac80211]
> > [  331.557525]  [<ffffffffa01527a5>] wiphy_resume+0x75/0x90 [cfg80211]
> > [  331.572001]  [<ffffffff81273fee>] dpm_resume_end+0x47e/0x4b0
> > [  331.586325]  [<ffffffff8107d5e9>] hibernation_snapshot+0xc9/0x280
> > [  331.600377]  [<ffffffff8107d88d>] hibernate+0xed/0x1f0
> > [  331.613980]  [<ffffffff8107c00c>] state_store+0xec/0x100
> > [  331.627577]  [<ffffffff811e05f7>] kobj_attr_store+0x17/0x20
> > [  331.641242]  [<ffffffff811345a4>] sysfs_write_file+0xd4/0x150
> > [  331.655056]  [<ffffffff810d02f8>] vfs_write+0xb8/0x1a0
> > [  331.668718]  [<ffffffff810d04bc>] sys_write+0x4c/0x80
> > [  331.682153]  [<ffffffff8100beeb>] system_call_fastpath+0x16/0x1b
> > [  331.695613] ---[ end trace 85c754c80d7debe8 ]---
> > [  331.709557] PM: Image restored successfully.
> > 
> > Now, how the ifmgd->timer could still be pending I have no idea yet, it seems to be explictly
> > del_timer_sync'ed and timer routine also checks this condition, and doesn't restart itself....
> > 
> > Also note that this was seen happening both on resume, and when the suspend image is written (sort of resume too).
> 
> Quick glance suggest some hole in the management of the
> TMR_RUNNING_CHANSW bit?

I guess no, I have futher information.

I forgot to pay attention to few lines before the oops message:


[  331.307139] No probe response from AP 00:1b:9e:d8:77:02 after 500ms, try 1
[  331.320121] ------------[ cut here ]------------
[  331.334935] WARNING: at /home/maxim/software/kernel/linux-2.6/kernel/timer.c:791 add_timer+0x36/0x40()
[  331.347374] Hardware name: Aspire 5720     
.....

This reveals that somehow the ieee80211_sta_work manages to run before ieee80211_sta_restart and it sets the timer....

looking at ieee80211_reconfig it seems suspicious that 'local->suspended = false;' is set so early.
In fact it is set again to false just prior to 

list_for_each_entry(sdata, &local->interfaces, list) {....

So, I would suspect some lines in this function trigger the work queue.

Seems reasonable?

Best regards,
	Maxim Levitsky