Re: 2.6.29.2 - WARNING: at kernel/hrtimer.c:625 hres_timers_resume+0x2f/0x45()

["Rafael J. Wysocki" <rjw@sisk.pl>]

On Sunday 17 May 2009, Fabio Comolli wrote:
> Hi Rafael.
> 
> On Sat, May 16, 2009 at 10:55 PM, Rafael J. Wysocki <rjw@sisk.pl> wrote:
> > On Monday 04 May 2009, Fabio Comolli wrote:
> >> Hi.
> >> After a successful resume from disk I get this warning:
> >>
> >> May  4 22:53:15 hawking kernel: ------------[ cut here ]------------
> >> May  4 22:53:15 hawking kernel: WARNING: at kernel/hrtimer.c:625
> >> hres_timers_resume+0x2f/0x45()
> >> May  4 22:53:15 hawking kernel: Hardware name: Pavilion dv4000 (ES973EA#ABZ)
> >> May  4 22:53:15 hawking kernel: hres_timers_resume() called with IRQs
> >> enabled!Modules linked in: radeon(N) drm(N) michael_mic(N) arc4(N)
> >> ecb(N) lib80211_crypt_tkip(N) af_packet(N) snd_pcm_oss(N)
> >> snd_mixer_oss(N) snd_seq(N) microcode(N) snd_seq_device(N) edd(N)
> >> cpufreq_conservative(N) cpufreq_userspace(N) cpufreq_powersave(N)
> >> acpi_cpufreq(N) speedstep_lib(N) fuse(N) loop(N) dm_mod(N)
> >> snd_intel8x0m(N) snd_intel8x0(N) snd_ac97_codec(N) pcmcia(N)
> >> ac97_bus(N) sdhci_pci(N) sdhci(N) snd_pcm(N) thermal(N) snd_timer(N)
> >> yenta_socket(N) btusb(N) iTCO_wdt(N) mmc_core(N) ipw2200(N)
> >> tifm_7xx1(N) ohci1394(N) rsrc_nonstatic(N) 8139too(N) snd(N)
> >> rtc_cmos(N) iTCO_vendor_support(N) video(N) processor(N) soundcore(N)
> >> thermal_sys(N) i2c_i801(N) intel_agp(N) wmi(N) libipw(N) output(N)
> >> led_class(N) battery(N) ac(N) 8139cp(N) tifm_core(N) ieee1394(N)
> >> pcmcia_core(N) rtc_core(N) sr_mod(N) button(N) snd_page_alloc(N)
> >> pcspkr(N) rtc_lib(N) i2c_core(N) hwmon(N) bluetooth(N) agpgart(N)
> >> cdrom(N) joydev(N) sg(N) ext4(N) mbcache(N) jbd2(N) crc16(
> >> May  4 22:53:15 hawking kernel: N) sd_mod(N) crc_t10dif(N) ata_piix(N)
> >> libata(N) scsi_mod(N) ehci_hcd(N) uhci_hcd(N) usbcore(N)
> >> May  4 22:53:15 hawking kernel: Supported: Yes
> >> May  4 22:53:15 hawking kernel: Pid: 3533, comm: s2disk Tainted: G
> >>      N  2.6.29.2-6-pae #1
> >> May  4 22:53:15 hawking kernel: Call Trace:
> >> May  4 22:53:15 hawking kernel:  [<c010688c>] try_stack_unwind+0x51/0x11d
> >> May  4 22:53:15 hawking kernel:  [<c0105ca3>] dump_trace+0x55/0xca
> >> May  4 22:53:15 hawking kernel:  [<c010665f>] show_trace_log_lvl+0x33/0x3e
> >> May  4 22:53:15 hawking kernel:  [<c010667a>] show_trace+0x10/0x14
> >> May  4 22:53:15 hawking kernel:  [<c03727ab>] dump_stack+0x57/0x61
> >> May  4 22:53:15 hawking kernel:  [<c012f2a2>] warn_slowpath+0x71/0xa8
> >> May  4 22:53:16 hawking kernel:  [<c01431cb>] hres_timers_resume+0x2f/0x45
> >> May  4 22:53:16 hawking kernel:  [<c0147049>] timekeeping_resume+0x11f/0x125
> >> May  4 22:53:16 hawking kernel:  [<c02bdd58>] __sysdev_resume+0x14/0x38
> >> May  4 22:53:16 hawking kernel:  [<c02bdedd>] sysdev_resume+0xcb/0x105
> >> May  4 22:53:16 hawking kernel:  [<c0155e69>] hibernation_snapshot+0xf1/0x189
> >> May  4 22:53:16 hawking kernel:  [<c0159008>] snapshot_ioctl+0x214/0x4c5
> >> May  4 22:53:16 hawking kernel:  [<c01b341f>] vfs_ioctl+0x22/0x67
> >> May  4 22:53:16 hawking kernel:  [<c01b397a>] do_vfs_ioctl+0x444/0x47d
> >> May  4 22:53:16 hawking kernel:  [<c01b3a1a>] sys_ioctl+0x67/0x81
> >> May  4 22:53:16 hawking kernel:  [<c010406b>] sysenter_do_call+0x12/0x2f
> >> May  4 22:53:17 hawking kernel:  [<ffffe430>] 0xffffe430
> >> May  4 22:53:17 hawking kernel: ---[ end trace 8cd3b732941f0f91 ]---
> >>
> >>
> >> This causes a second hibernate-resume cycle to hang solid during
> >> resume. Power cycle is needed.
> >>
> >> I'm not sure if this is a regression wrt 2.6.28.7 (my previous kernel)
> >> because I changed the DRI driver from fglrx to radeon and also
> >> upgraded the X.org userspace. I didn't have any hibernation-related
> >> problem with any previous kernels for sure.
> >>
> >> It may be related to http://lkml.org/lkml/2009/4/19/133 but the
> >> backtrace is somewhat different.
> >>
> >> Please let me know if you need more details.
> >
> > I think this is a cpufreq problem.
> >
> 
> Indeed it is. I managed to "solve" it by rmmod-ing acpi-cpufreq before
> hibernating and reloading it after resume. With this trick I had no
> further problems.

Thanks for this information, it's quite important.

> > Please try to reproduce the problem with the appended patch applied.
> 
> I'm sorry but I don't have access to that laptop anymore.

No problem, we should be able to find the bug in acpi-cpufreq anyway.

Thanks,
Rafael