[Qemu-devel] Sparc guest - SCSI

[Qemu-devel] Sparc guest - SCSI

[Nigel Horne]

[-- Attachment #1: Type: text/plain, Size: 744 bytes --]

Guest: Linux/sparc
Host: Linux/x86

>From time to time I am getting SCSI errors from the guest using the
latest CVS:

scsi : aborting command due to timeout : pid 50803, scsi0, channel 0, id 0, lun 0 0x2a 00 00 3d be 84 00 00 08 00 
esp0: Aborting command
esp0: dumping state
esp0: dma -- cond_reg<a4000211> addr<f292b000>
esp0: SW [sreg<00> sstep<04> ireg<18>]
esp0: HW reread [sreg<03> sstep<00> ireg<10>]
esp0: current command [tgt<00> lun<00> pphase<DATAOUT> cphase<DATAOUT>]
esp0: disconnected 
SCSI host 0 abort (pid 50803) timed out - resetting
SCSI bus is being reset for host 0 channel 0.
esp0: Resetting scsi bus
esp0: SCSI bus reset interrupt

As far as I can tell the guest recovers fine, but it may be worth investigating.

-Nigel

[-- Attachment #2: njh.vcf --]
[-- Type: text/x-vcard, Size: 279 bytes --]

begin:vcard
fn:Nigel Horne
n:Horne;Nigel
org:NJH Music
adr:Pogmoor;;28 Wharfedale Road;Barnsley;South Yorkshire;S75 2LJ;UK
email;internet:njh@bandsman.co.uk
tel;fax:+44 870 705 9334
note:Skype: nigelhorne
x-mozilla-html:FALSE
url:http://www.bandsman.co.uk
version:2.1
end:vcard

Re: [Qemu-devel] Sparc guest - SCSI

[Andreas Färber]

Am 03.08.2007 um 10:45 schrieb Nigel Horne:

> Guest: Linux/sparc
> Host: Linux/x86
>
>> From time to time I am getting SCSI errors from the guest using the
> latest CVS:
>
> scsi : aborting command due to timeout : pid 50803, scsi0, channel  
> 0, id 0, lun 0 0x2a 00 00 3d be 84 00 00 08 00 esp0: Aborting command
> esp0: dumping state
> esp0: dma -- cond_reg<a4000211> addr<f292b000>
> esp0: SW [sreg<00> sstep<04> ireg<18>]
> esp0: HW reread [sreg<03> sstep<00> ireg<10>]
> esp0: current command [tgt<00> lun<00> pphase<DATAOUT>  
> cphase<DATAOUT>]
> esp0: disconnected SCSI host 0 abort (pid 50803) timed out - resetting
> SCSI bus is being reset for host 0 channel 0.
> esp0: Resetting scsi bus
> esp0: SCSI bus reset interrupt
>
> As far as I can tell the guest recovers fine, but it may be worth  
> investigating.

I reported similar errors recently (OS X ppc host) - for me it panics  
afterwards and does not recover fine. I got the impression that it  
happens later if I disable the lance NIC.

Andreas

Re: [Qemu-devel] Sparc guest - SCSI

[Nigel Horne]

[-- Attachment #1: Type: text/plain, Size: 1325 bytes --]

Andreas Färber wrote:
> 
> Am 03.08.2007 um 10:45 schrieb Nigel Horne:
> 
>> Guest: Linux/sparc
>> Host: Linux/x86
>>
>>> From time to time I am getting SCSI errors from the guest using the
>> latest CVS:
>>
>> scsi : aborting command due to timeout : pid 50803, scsi0, channel 0, 
>> id 0, lun 0 0x2a 00 00 3d be 84 00 00 08 00 esp0: Aborting command
>> esp0: dumping state
>> esp0: dma -- cond_reg<a4000211> addr<f292b000>
>> esp0: SW [sreg<00> sstep<04> ireg<18>]
>> esp0: HW reread [sreg<03> sstep<00> ireg<10>]
>> esp0: current command [tgt<00> lun<00> pphase<DATAOUT> cphase<DATAOUT>]
>> esp0: disconnected SCSI host 0 abort (pid 50803) timed out - resetting
>> SCSI bus is being reset for host 0 channel 0.
>> esp0: Resetting scsi bus
>> esp0: SCSI bus reset interrupt
>>
>> As far as I can tell the guest recovers fine, but it may be worth 
>> investigating.
> 
> I reported similar errors recently (OS X ppc host) - for me it panics 
> afterwards and does not recover fine. I got the impression that it 
> happens later if I disable the lance NIC.

I have the lance NIC enabled in the guest.

> 
> Andreas
> 

-- 
Nigel Horne. Arranger, Adjudicator, Band Trainer, Composer, Tutor, Typesetter.
NJH Music, Barnsley, UK.  ICQ#20252325
njh@bandsman.co.uk http://www.bandsman.co.uk

[-- Attachment #2: njh.vcf --]
[-- Type: text/x-vcard, Size: 279 bytes --]

begin:vcard
fn:Nigel Horne
n:Horne;Nigel
org:NJH Music
adr:Pogmoor;;28 Wharfedale Road;Barnsley;South Yorkshire;S75 2LJ;UK
email;internet:njh@bandsman.co.uk
tel;fax:+44 870 705 9334
note:Skype: nigelhorne
x-mozilla-html:FALSE
url:http://www.bandsman.co.uk
version:2.1
end:vcard

Re: [Qemu-devel] Sparc guest - SCSI

[Andreas Färber]

Am 03.08.2007 um 12:42 schrieb Nigel Horne:

> Andreas Färber wrote:
>> Am 03.08.2007 um 10:45 schrieb Nigel Horne:
>>> Guest: Linux/sparc
>>> Host: Linux/x86
>>>
>>>> From time to time I am getting SCSI errors from the guest using the
>>> latest CVS:
>>>
>>> scsi : aborting command due to timeout : pid 50803, scsi0,  
>>> channel 0, id 0, lun 0 0x2a 00 00 3d be 84 00 00 08 00 esp0:  
>>> Aborting command
>>> esp0: dumping state
>>> esp0: dma -- cond_reg<a4000211> addr<f292b000>
>>> esp0: SW [sreg<00> sstep<04> ireg<18>]
>>> esp0: HW reread [sreg<03> sstep<00> ireg<10>]
>>> esp0: current command [tgt<00> lun<00> pphase<DATAOUT>  
>>> cphase<DATAOUT>]
>>> esp0: disconnected SCSI host 0 abort (pid 50803) timed out -  
>>> resetting
>>> SCSI bus is being reset for host 0 channel 0.
>>> esp0: Resetting scsi bus
>>> esp0: SCSI bus reset interrupt
>>>
>>> As far as I can tell the guest recovers fine, but it may be worth  
>>> investigating.
>> I reported similar errors recently (OS X ppc host) - for me it  
>> panics afterwards and does not recover fine. I got the impression  
>> that it happens later if I disable the lance NIC.
>
> I have the lance NIC enabled in the guest.

So what? I usually have it enabled, too. :)

Re: [Qemu-devel] Sparc guest - SCSI

[Aurelien Jarno]

Andreas Färber a écrit :
> Am 03.08.2007 um 10:45 schrieb Nigel Horne:
> 
>> Guest: Linux/sparc
>> Host: Linux/x86
>>
>>> From time to time I am getting SCSI errors from the guest using the
>> latest CVS:
>>
>> scsi : aborting command due to timeout : pid 50803, scsi0, channel  
>> 0, id 0, lun 0 0x2a 00 00 3d be 84 00 00 08 00 esp0: Aborting command
>> esp0: dumping state
>> esp0: dma -- cond_reg<a4000211> addr<f292b000>
>> esp0: SW [sreg<00> sstep<04> ireg<18>]
>> esp0: HW reread [sreg<03> sstep<00> ireg<10>]
>> esp0: current command [tgt<00> lun<00> pphase<DATAOUT>  
>> cphase<DATAOUT>]
>> esp0: disconnected SCSI host 0 abort (pid 50803) timed out - resetting
>> SCSI bus is being reset for host 0 channel 0.
>> esp0: Resetting scsi bus
>> esp0: SCSI bus reset interrupt
>>
>> As far as I can tell the guest recovers fine, but it may be worth  
>> investigating.
> 
> I reported similar errors recently (OS X ppc host) - for me it panics  
> afterwards and does not recover fine. I got the impression that it  
> happens later if I disable the lance NIC.
> 

The problem appears when one interrupt on the SCSI controller is
triggered twice by the interrupt controller, the second interrupt
disturb the driver.

I have fixed one problem like that a few months ago in the CVS, so it
happens less often, but there is still a problem somewhere.

-- 
  .''`.  Aurelien Jarno	            | GPG: 1024D/F1BCDB73
 : :' :  Debian developer           | Electrical Engineer
 `. `'   aurel32@debian.org         | aurelien@aurel32.net
   `-    people.debian.org/~aurel32 | www.aurel32.net

RE: [Qemu-devel] Sparc guest - SCSI

[Nigel Horne]

> Am 03.08.2007 um 12:42 schrieb Nigel Horne:
>
> > Andreas Färber wrote:
> >> Am 03.08.2007 um 10:45 schrieb Nigel Horne:
> >>> Guest: Linux/sparc
> >>> Host: Linux/x86
> >>>
> >>>> From time to time I am getting SCSI errors from the guest using the
> >>> latest CVS:
> >>>
> >>> scsi : aborting command due to timeout : pid 50803, scsi0,
> >>> channel 0, id 0, lun 0 0x2a 00 00 3d be 84 00 00 08 00 esp0:
> >>> Aborting command
> >>> esp0: dumping state
> >>> esp0: dma -- cond_reg<a4000211> addr<f292b000>
> >>> esp0: SW [sreg<00> sstep<04> ireg<18>]
> >>> esp0: HW reread [sreg<03> sstep<00> ireg<10>]
> >>> esp0: current command [tgt<00> lun<00> pphase<DATAOUT>
> >>> cphase<DATAOUT>]
> >>> esp0: disconnected SCSI host 0 abort (pid 50803) timed out -
> >>> resetting
> >>> SCSI bus is being reset for host 0 channel 0.
> >>> esp0: Resetting scsi bus
> >>> esp0: SCSI bus reset interrupt
> >>>
> >>> As far as I can tell the guest recovers fine, but it may be worth
> >>> investigating.
> >> I reported similar errors recently (OS X ppc host) - for me it
> >> panics afterwards and does not recover fine. I got the impression
> >> that it happens later if I disable the lance NIC.
> >
> > I have the lance NIC enabled in the guest.
>
> So what? I usually have it enabled, too. :)
>

No need to be so rude/agressive, I just thought I'd add the information in
case
some one found it helpful.

Re: [Qemu-devel] Sparc guest - SCSI

[Andreas Färber]

Am 03.08.2007 um 13:52 schrieb Aurelien Jarno:

> The problem appears when one interrupt on the SCSI controller is
> triggered twice by the interrupt controller, the second interrupt
> disturb the driver.
>
> I have fixed one problem like that a few months ago in the CVS, so it
> happens less often, but there is still a problem somewhere.

Do you know under which circumstances this happens? And is there any  
way to avoid it?

Andreas

Re: [Qemu-devel] Sparc guest - SCSI

[Aurelien Jarno]

Andreas Färber a écrit :
> Am 03.08.2007 um 13:52 schrieb Aurelien Jarno:
> 
>> The problem appears when one interrupt on the SCSI controller is
>> triggered twice by the interrupt controller, the second interrupt
>> disturb the driver.
>>
>> I have fixed one problem like that a few months ago in the CVS, so it
>> happens less often, but there is still a problem somewhere.
> 
> Do you know under which circumstances this happens? And is there any  
> way to avoid it?

I suggest to read the description of the problem from last time:
http://www.mail-archive.com/qemu-devel%40nongnu.org/msg08828.html

I guess there is still one case when this can happen. The only solution
I see to avoid that is to fix the bug (the problem doesn't occurs on
real hardware, so the emulation is buggy).

-- 
  .''`.  Aurelien Jarno	            | GPG: 1024D/F1BCDB73
 : :' :  Debian developer           | Electrical Engineer
 `. `'   aurel32@debian.org         | aurelien@aurel32.net
   `-    people.debian.org/~aurel32 | www.aurel32.net

Re: [Qemu-devel] Sparc guest - SCSI

[Andreas Färber]

Am 03.08.2007 um 14:02 schrieb Nigel Horne:

>>>>> Guest: Linux/sparc
>>>>> Host: Linux/x86
>>>>>
>>>>>> From time to time I am getting SCSI errors from the guest  
>>>>>> using the
>>>>> latest CVS:
>>>>>
>>>>> scsi : aborting command due to timeout : pid 50803, scsi0,
>>>>> channel 0, id 0, lun 0 0x2a 00 00 3d be 84 00 00 08 00 esp0:
>>>>> Aborting command
>>>>> esp0: dumping state
>>>>> esp0: dma -- cond_reg<a4000211> addr<f292b000>
>>>>> esp0: SW [sreg<00> sstep<04> ireg<18>]
>>>>> esp0: HW reread [sreg<03> sstep<00> ireg<10>]
>>>>> esp0: current command [tgt<00> lun<00> pphase<DATAOUT>
>>>>> cphase<DATAOUT>]
>>>>> esp0: disconnected SCSI host 0 abort (pid 50803) timed out -
>>>>> resetting
>>>>> SCSI bus is being reset for host 0 channel 0.
>>>>> esp0: Resetting scsi bus
>>>>> esp0: SCSI bus reset interrupt
>>>>>
>>>>> As far as I can tell the guest recovers fine, but it may be worth
>>>>> investigating.
>>>> I reported similar errors recently (OS X ppc host) - for me it
>>>> panics afterwards and does not recover fine. I got the impression
>>>> that it happens later if I disable the lance NIC.
>>>
>>> I have the lance NIC enabled in the guest.
>>
>> So what? I usually have it enabled, too. :)
>>
>
> No need to be so rude/agressive, I just thought I'd add the  
> information in
> case
> some one found it helpful.

Hey, asking what you meant and a smiley are most certainly not rude  
or aggressive. You replied to my suggestion with a one-liner that was  
about as helpful as replying with your screen resolution, as it  
happens with and without NIC. So I wanted to know what you wanted to  
imply with that. Apparently nothing - question answered, thank you.

But on the subject of being "rude", I reported this a month ago in  
your "sparc guest" thread and it was ignored by everyone, you said  
nothing of esp0 issues back then and now it's suddenly a new issue...
http://lists.gnu.org/archive/html/qemu-devel/2007-06/msg00481.html

Andreas

Re: [Qemu-devel] Sparc guest - SCSI

[Andreas Färber]

Am 03.08.2007 um 16:15 schrieb Aurelien Jarno:

> Andreas Färber a écrit :
>> Am 03.08.2007 um 13:52 schrieb Aurelien Jarno:
>>
>>> The problem appears when one interrupt on the SCSI controller is
>>> triggered twice by the interrupt controller, the second interrupt
>>> disturb the driver.
>>>
>>> I have fixed one problem like that a few months ago in the CVS,  
>>> so it
>>> happens less often, but there is still a problem somewhere.
>>
>> Do you know under which circumstances this happens? And is there any
>> way to avoid it?
>
> I suggest to read the description of the problem from last time:
> http://www.mail-archive.com/qemu-devel%40nongnu.org/msg08828.html
>
> I guess there is still one case when this can happen. The only  
> solution
> I see to avoid that is to fix the bug (the problem doesn't occurs on
> real hardware, so the emulation is buggy).

I tried that patch before ever posting to this list. Like I reported  
earlier, if applied I don't get a kernel panic any more but instead  
qemu crashes and is "gone". So sadly it didn't fix the issue for me  
as is.

And without knowing about interrupt controllers it is hard to fix a  
bug no-one knows what or where exactly it is and how it should be  
fixed... :(

Andreas

http://lists.gnu.org/archive/html/qemu-devel/2007-06/msg00086.html

Re: [Qemu-devel] Sparc guest - SCSI

[Aurelien Jarno]

Andreas Färber a écrit :
> Am 03.08.2007 um 16:15 schrieb Aurelien Jarno:
> 
>> Andreas Färber a écrit :
>>> Am 03.08.2007 um 13:52 schrieb Aurelien Jarno:
>>>
>>>> The problem appears when one interrupt on the SCSI controller is
>>>> triggered twice by the interrupt controller, the second interrupt
>>>> disturb the driver.
>>>>
>>>> I have fixed one problem like that a few months ago in the CVS,  
>>>> so it
>>>> happens less often, but there is still a problem somewhere.
>>> Do you know under which circumstances this happens? And is there any
>>> way to avoid it?
>> I suggest to read the description of the problem from last time:
>> http://www.mail-archive.com/qemu-devel%40nongnu.org/msg08828.html
>>
>> I guess there is still one case when this can happen. The only  
>> solution
>> I see to avoid that is to fix the bug (the problem doesn't occurs on
>> real hardware, so the emulation is buggy).
> 
> I tried that patch before ever posting to this list. Like I reported  
> earlier, if applied I don't get a kernel panic any more but instead  
> qemu crashes and is "gone". So sadly it didn't fix the issue for me  
> as is.

Which version are you using? This patch is supposed to be in the CVS
already.


-- 
  .''`.  Aurelien Jarno	            | GPG: 1024D/F1BCDB73
 : :' :  Debian developer           | Electrical Engineer
 `. `'   aurel32@debian.org         | aurelien@aurel32.net
   `-    people.debian.org/~aurel32 | www.aurel32.net

Re: [Qemu-devel] Sparc guest - SCSI

[Andreas Färber]

Am 03.08.2007 um 16:54 schrieb Aurelien Jarno:

>>> I suggest to read the description of the problem from last time:
>>> http://www.mail-archive.com/qemu-devel%40nongnu.org/msg08828.html
>>>
>>> I guess there is still one case when this can happen. The only
>>> solution
>>> I see to avoid that is to fix the bug (the problem doesn't occurs on
>>> real hardware, so the emulation is buggy).
>>
>> I tried that patch before ever posting to this list. Like I reported
>> earlier, if applied I don't get a kernel panic any more but instead
>> qemu crashes and is "gone". So sadly it didn't fix the issue for me
>> as is.
>
> Which version are you using?

At the time both 0.9.0 and CVS HEAD, today CVS HEAD.

> This patch is supposed to be in the CVS
> already.

Negative, this is the excerpt from today's hw/slavio_intctl.c around  
line 293:

static void slavio_set_irq(void *opaque, int irq, int level)
{
     SLAVIO_INTCTLState *s = opaque;
     uint32_t mask = 1 << irq;
     uint32_t pil = s->intbit_to_level[irq];

     DPRINTF("Set cpu %d irq %d -> pil %d level %d\n", s->target_cpu,  
irq, pil,
             level);
     if (pil > 0) {
         if (level) {
             s->intregm_pending |= mask;
             s->intreg_pending[s->target_cpu] |= 1 << pil;
         } else {
             s->intregm_pending &= ~mask;
             s->intreg_pending[s->target_cpu] &= ~(1 << pil);
         }
         slavio_check_interrupts(s);
     }
}

Your referenced patch moves slavio_check_interrupts(s); four lines up  
so is not in CVS.

Andreas

Re: [Qemu-devel] Sparc guest - SCSI

[Blue Swirl]

I seem to have reverted the patch here:
http://cvs.savannah.gnu.org/viewvc/qemu/qemu/hw/slavio_intctl.c?r1=1.15&r2=1.16

But I still get SCSI errors when using e2fsck with this change.

On 8/3/07, Andreas Färber <andreas.faerber@web.de> wrote:
>
> Am 03.08.2007 um 16:54 schrieb Aurelien Jarno:
>
> >>> I suggest to read the description of the problem from last time:
> >>> http://www.mail-archive.com/qemu-devel%40nongnu.org/msg08828.html
> >>>
> >>> I guess there is still one case when this can happen. The only
> >>> solution
> >>> I see to avoid that is to fix the bug (the problem doesn't occurs on
> >>> real hardware, so the emulation is buggy).
> >>
> >> I tried that patch before ever posting to this list. Like I reported
> >> earlier, if applied I don't get a kernel panic any more but instead
> >> qemu crashes and is "gone". So sadly it didn't fix the issue for me
> >> as is.
> >
> > Which version are you using?
>
> At the time both 0.9.0 and CVS HEAD, today CVS HEAD.
>
> > This patch is supposed to be in the CVS
> > already.
>
> Negative, this is the excerpt from today's hw/slavio_intctl.c around
> line 293:
>
> static void slavio_set_irq(void *opaque, int irq, int level)
> {
>      SLAVIO_INTCTLState *s = opaque;
>      uint32_t mask = 1 << irq;
>      uint32_t pil = s->intbit_to_level[irq];
>
>      DPRINTF("Set cpu %d irq %d -> pil %d level %d\n", s->target_cpu,
> irq, pil,
>              level);
>      if (pil > 0) {
>          if (level) {
>              s->intregm_pending |= mask;
>              s->intreg_pending[s->target_cpu] |= 1 << pil;
>          } else {
>              s->intregm_pending &= ~mask;
>              s->intreg_pending[s->target_cpu] &= ~(1 << pil);
>          }
>          slavio_check_interrupts(s);
>      }
> }
>
> Your referenced patch moves slavio_check_interrupts(s); four lines up
> so is not in CVS.
>
> Andreas
>
>
>

Re: [Qemu-devel] Sparc guest - SCSI

[Blue Swirl]

I reworked the interrupt handling so that both e2fsck and SMP work.
Previously there may have been lost or spurious interrupts, now they
should be fixed. I pushed the interrupt priority handling   from
SLAVIO to CPU.