Re: Suspend to memory is freezing my machine

[Robert Hancock <hancockr@shaw.ca>]

Jacek Luczak wrote:
> Jacek Luczak pisze:
>> Robert Hancock pisze:
>>> Jacek Luczak wrote:
>>>> Robert Hancock pisze:
>>>>> Jacek Luczak wrote:
>>>>>> Rafael J. Wysocki pisze:
>>>>>>> On Sunday, 4 of May 2008, Zdenek Kabelac wrote:
>>>>>>>> Hello
>>>>>>> Hi,
>>>>>>>
>>>>>>>> With recent 2.6.25 & 2.6.26-rc1 git (around 1 week) I get
>>>>>>>> occasionally
>>>>>>>> complete freeze of my T61 during suspend. (dual core, 2GB).
>>>>>>> How reproducible is this?
>>>>>>>
>>>>>>>> I'm running kernel with no_console_suspend - but all I can see is
>>>>>>>> blinking cursor on an empty screen - thus even when I run kernel with
>>>>>>>> most debug options turned on, I can't pass more details so far.  I
>>>>>>>> run
>>>>>>>> suspend with with SD card in - so maybe some update in the MMC driver
>>>>>>>> might be responsible for this ?
>>>>>>>>
>>>>>>>> Also - I think that option no_console_suspend doens't work
>>>>>>>> correctly -
>>>>>>>> as many times with suspend I do not see any log message on my console
>>>>>>>> screen. However sometimes the log is shown.
>>>>>>> It would be helpful if you could verify if:
>>>>>>>
>>>>>>> (1) The problem occurs without no_console_suspend.
>>>>>>> (2) The problem occurs without the SD card.
>>>>>>>
>>>>>> Hi Rafael,
>>>>>>
>>>>>> same problem here, although I was able to resume system (it's
>>>>>> basically Intel
>>>>>> machine) , but it was unusable - I was able to switch between
>>>>>> terminals and see
>>>>>> output from kernel. So there was:
>>>>>>     - Disabling irq #19;
>>>>>>     - some kind of lock spinning on disk:
>>>>>> IDE interface: Intel Corporation 82801GBM/GHM (ICH7 Family) Serial ATA
>>>>>> Storage
>>>>>> Controller IDE (rev 02)
>>>>>>     but I can't provide more output of that lock now - no sign in logs.
>>>>>>
>>>>>> I've made some successful suspend/resume all without sound card active
>>>>>> without
>>>>>> problem. Those appear with sound card active, but I must take closer
>>>>>> look - will
>>>>>> send info later.
>>>>> Can you post your dmesg and /proc/interrupts output from normal bootup ?
>>>> Sure I can ;)
>>>>
>>>> 1) /proc/interrupts
>>>>
>>>>            CPU0       CPU1
>>>>   0:   11846981          0   IO-APIC-edge      timer
>>>>   1:      30098          0   IO-APIC-edge      i8042
>>>>   8:          3          0   IO-APIC-edge      rtc
>>>>   9:         13          0   IO-APIC-fasteoi   acpi
>>>>  12:    1776540          0   IO-APIC-edge      i8042
>>>>  14:         39          0   IO-APIC-edge      ata_piix
>>>>  15:          0          0   IO-APIC-edge      ata_piix
>>>>  16:      54570      44642   IO-APIC-fasteoi   i915@pci:0000:00:02.0
>>>>  17:          0          0   IO-APIC-fasteoi   uhci_hcd:usb3
>>>>  18:          0          0   IO-APIC-fasteoi   uhci_hcd:usb4
>>>>  19:      98243          0   IO-APIC-fasteoi   ata_piix, uhci_hcd:usb5
>>>>  21:    1650574          0   IO-APIC-fasteoi   HDA Intel
>>>>  23:          0          0   IO-APIC-fasteoi   ehci_hcd:usb1,
>>>> uhci_hcd:usb2
>>>> 220:      14263          0   PCI-MSI-edge      iwl3945
>>>> 221:    1166041    1333296   PCI-MSI-edge      eth0
>>>> NMI:          0          0   Non-maskable interrupts
>>>> LOC:    1104887    7534969   Local timer interrupts
>>>> RES:     633378     701351   Rescheduling interrupts
>>>> CAL:         16      28315   function call interrupts
>>>> TLB:       1721       2620   TLB shootdowns
>>>> TRM:          0          0   Thermal event interrupts
>>>> SPU:          0          0   Spurious interrupts
>>>> ERR:          0
>>>> MIS:          0
>>>>
>>>> 2) dmesg can here -> http://212.109.128.251/~difrost/linux-next/dmesg.log
>>>> 3) Kernel:
>>>> Linux difrost 2.6.25-07422-gb66e1f1-dirty #14 SMP Fri May 2 22:04:17
>>>> CEST 2008
>>>> i686 i686 i386 GNU/Linux
>>>> It's marked dirty because due to http://lkml.org/lkml/2008/5/2/405
>>>> patch applied.
>>>>
>>>> -Jacek
>>>>
>>> Well, if IRQ 19 got disabled, that's your SATA controller, so resume
>>> likely isn't going to work. Could be a libata problem? CCing linux-ide.
>> Yep, I know, that's why I pointed that out. Irq was disabled somehow in suspend
>> or resume process.
>>
>>> BTW, if your BIOS offers an option to enable AHCI on your SATA
>>> controller then that would be a more optimal configuration (could get
>>> NCQ support), but that is an aside.
>> With AHCI I've got pretty bad timings (and I don't really know why!):
>>
>> [root|20:49|~]$ cat sda_ahci_t
>>
>> /dev/sda:
>>  Timing cached reads:   1560 MB in  2.00 seconds = 780.51 MB/sec
>>  Timing buffered disk reads:  102 MB in  3.02 seconds =  33.74 MB/sec
>> [root|20:49|~]$ cat sda_piix_t
>>
>> /dev/sda:
>>  Timing cached reads:   1544 MB in  2.00 seconds = 772.35 MB/sec
>>  Timing buffered disk reads:  134 MB in  3.04 seconds =  44.05 MB/sec
> 
> Here's the latest report (all on latest git):
> 	1) I've switched to AHCI mode and suspend/resume works OK (because SATA
> controller irq is not disabled).
> 
> 	2) now /proc/interrupts look like that:
>            CPU0       CPU1
>   0:     110708          0   IO-APIC-edge      timer
>   1:       4008          0   IO-APIC-edge      i8042
>   8:          3          0   IO-APIC-edge      rtc
>   9:      15091          0   IO-APIC-fasteoi   acpi
>  12:      77467          0   IO-APIC-edge      i8042
>  14:         44          0   IO-APIC-edge      ata_piix
>  15:          0          0   IO-APIC-edge      ata_piix
>  16:          0          0   IO-APIC-fasteoi   i915@pci:0000:00:02.0
>  17:          0          0   IO-APIC-fasteoi   uhci_hcd:usb3
>  18:          0          0   IO-APIC-fasteoi   uhci_hcd:usb4
>  19:     100001          0   IO-APIC-fasteoi   uhci_hcd:usb5
>  21:        282          0   IO-APIC-fasteoi   HDA Intel
>  23:          1          0   IO-APIC-fasteoi   ehci_hcd:usb1, uhci_hcd:usb2
> 219:        858          0   PCI-MSI-edge      iwl3945
> 220:       8616          0   PCI-MSI-edge      eth0
> 221:       6423          0   PCI-MSI-edge      ahci
> NMI:          0          0   Non-maskable interrupts
> LOC:      15777      64510   Local timer interrupts
> RES:       9045      24560   Rescheduling interrupts
> CAL:         30      28255   function call interrupts
> TLB:        341        145   TLB shootdowns
> TRM:          0          0   Thermal event interrupts
> SPU:          0          0   Spurious interrupts
> ERR:          0
> MIS:          0
> 
> 	3) The IRQ #19 remains disabled after resume and produce:
> irq 19: nobody cared (try booting with the "irqpoll" option)
> Pid: 13, comm: kacpi_notify Not tainted 2.6.26-rc1-07561-gafa26be-dirty #16
>  [<c013ea27>] __report_bad_irq+0x24/0x69
>  [<c013ea2e>] __report_bad_irq+0x2b/0x69
>  [<c013ec25>] note_interrupt+0x1b9/0x210
>  [<c013e36c>] handle_IRQ_event+0x1a/0x3f
>  [<c013f195>] handle_fasteoi_irq+0x84/0xa2
>  [<c0104fde>] do_IRQ+0x4f/0x65
>  [<c01034ff>] common_interrupt+0x23/0x28
>  [<c013007b>] timekeeping_resume+0x9b/0x127
>  [<c020b090>] acpi_os_read_port+0x29/0x44
>  [<c02177c9>] acpi_hw_register_read+0x61/0x119
>  [<c020f76e>] acpi_ev_fixed_event_detect+0x2a/0xa0
>  [<c021001a>] acpi_ev_sci_xrupt_handler+0x9/0x17
>  [<c020b053>] acpi_irq+0xb/0x1f
>  [<c013e36c>] handle_IRQ_event+0x1a/0x3f
>  [<c013f181>] handle_fasteoi_irq+0x70/0xa2
>  [<c0104fde>] do_IRQ+0x4f/0x65
>  [<c020b623>] acpi_os_execute_deferred+0x0/0x25
>  [<c01034ff>] common_interrupt+0x23/0x28
>  [<c020b623>] acpi_os_execute_deferred+0x0/0x25
>  [<c020b0b8>] acpi_os_write_port+0xd/0x2c
>  [<c020b640>] acpi_os_execute_deferred+0x1d/0x25
>  [<c01290fa>] run_workqueue+0x69/0xda
>  [<c0129221>] worker_thread+0xb6/0xc2
>  [<c012bca6>] autoremove_wake_function+0x0/0x2d
>  [<c012916b>] worker_thread+0x0/0xc2
>  [<c012ba42>] kthread+0x38/0x5d
>  [<c012ba0a>] kthread+0x0/0x5d
>  [<c010370f>] kernel_thread_helper+0x7/0x10
>  =======================
> handlers:
> [<c027d100>] (usb_hcd_irq+0x0/0x53)
> Disabling IRQ #19

Hmm, so either it's the SATA controller still generating that IRQ even 
when it's in AHCI mode, or else it's USB that's the real problem..

> 
> This might happen due to "ACPI: EC: GPE storm detected, disabling EC GPE", but
> here it should revert to polling mode (which is done during boot, but not during
> resume). I'm not expert here.

That does seem unusual, but it doesn't seem directly related (ACPI is on 
IRQ9).

> 
> Full dmesg here -> http://212.109.128.251/~difrost/linux-next/dmesg_ahci.log
> 
> -Jacek
> 
> PS: Site note: Why there's such big difference on hdparm timings with PATA and
> AHCI mode?

You can narrow that down by doing this with AHCI in use:

echo 1 > /sys/block/sda/device/queue_depth

which will disable NCQ but keep AHCI. If that brings the performance 
back up, then quite likely your drive's NCQ implementation isn't really 
optimized for sequential reads..