linux-ide.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
* Re: 2.6.21-mm1: many processes end up in D state
@ 2007-05-04 19:19 Mikael Pettersson
  2007-05-05 15:30 ` Tejun Heo
  0 siblings, 1 reply; 6+ messages in thread
From: Mikael Pettersson @ 2007-05-04 19:19 UTC (permalink / raw)
  To: jirislaby; +Cc: akpm, jgarzik, linux-ide, linux-kernel, mikpe, stable

On Fri, 04 May 2007 17:02:10 +0200, Jiri Slaby wrote:
>>>> I have a problem with higher disk loads (e.g. running git-log or yum update).
>>>> Many processes end up in D state and system is unusable -- I'm not able to run
>>>> anything but smooth mouse moving when this happens.
...(boring stack dumps deleted)
>> causes this? When I change this:
>> diff --git a/drivers/ata/sata_promise.c b/drivers/ata/sata_promise.c
>> index f56549b..a0024ae 100644
>> --- a/drivers/ata/sata_promise.c
>> +++ b/drivers/ata/sata_promise.c
>> @@ -668,10 +668,8 @@ static inline unsigned int pdc_host_intr( struct ata_port *ap,
>>         else
>>                 err_mask &= ~PDC2_ERR_MASK;
>>         port_status = readl(port_mmio + PDC_GLOBAL_CTL);
>> -       if (unlikely(port_status & err_mask)) {
>> +       if (unlikely(port_status & err_mask))
>>                 pdc_error_intr(ap, qc, port_status, err_mask);
>> -               return 1;
>> -       }
>> 
>>         switch (qc->tf.protocol) {
>>         case ATA_PROT_DMA:
>> the problem disappears.
>Actually changes:
>ata6.00: exception Emask 0x0 SAct 0x0 SErr 0x0 action 0x2 frozen
>ata6.00: (port_status 0x20080000)
>ata6.00: cmd c8/00:08:03:d3:61/00:00:00:00:00/e0 tag 0 cdb 0x0 data 4096 in
>         res 40/00:01:01:4f:c2/00:00:00:00:00/00 Emask 0x6 (timeout)
>ata6: soft resetting port
>ata6: SATA link up 3.0 Gbps (SStatus 123 SControl 300)
>ata6.00: ata_hpa_resize 1: sectors = 156301488, hpa_sectors = 156301488
>ata6.00: ata_hpa_resize 1: sectors = 156301488, hpa_sectors = 156301488
>ata6.00: configured for UDMA/133
>ata6: EH complete
>sd 5:0:0:0: [sdc] 156301488 512-byte hardware sectors (80026 MB)
>sd 5:0:0:0: [sdc] Write Protect is off
>sd 5:0:0:0: [sdc] Write cache: enabled, read cache: enabled, doesn't support DPO
>or FUA
>
>
>to
>ata5.00: exception Emask 0x0 SAct 0x0 SErr 0x0 action 0x2
>ata5.00: (port_status 0x20080000)
>ata5.00: cmd c8/00:08:bf:cd:4b/00:00:00:00:00/e0 tag 0 cdb 0x0 data 4096 in
>         res 50/00:00:c6:cd:4b/00:00:00:00:00/e0 Emask 0x2 (HSM violation)
>ata5: soft resetting port
>ata5: SATA link up 3.0 Gbps (SStatus 123 SControl 300)
>ata5.00: ata_hpa_resize 1: sectors = 156301488, hpa_sectors = 156301488
>ata5.00: ata_hpa_resize 1: sectors = 156301488, hpa_sectors = 156301488
>ata5.00: configured for UDMA/133
>ata5: EH complete
>sd 4:0:0:0: [sdb] 156301488 512-byte hardware sectors (80026 MB)
>sd 4:0:0:0: [sdb] Write Protect is off
>sd 4:0:0:0: [sdb] Mode Sense: 00 3a 00 00
>sd 4:0:0:0: [sdb] Write cache: enabled, read cache: enabled, doesn't support DPO
>or FUA
>
>
>I.e. no freezing of ports...

Your patch to delete the 'return 1;' on error is correct,
and makes the code match exactly the behaviour of previous
versions of sata_promise, except for the additional error
decoding.

ahci and sata_sil24 do the return in this situation. I don't
yet understand why they can get away with it while sata_promise
cannot, but for now the return should be removed.

/Mikael

^ permalink raw reply	[flat|nested] 6+ messages in thread
* Re: 2.6.21-mm1: many processes end up in D state
@ 2007-05-05 23:46 Mikael Pettersson
  2007-05-07  9:52 ` Jiri Slaby
  0 siblings, 1 reply; 6+ messages in thread
From: Mikael Pettersson @ 2007-05-05 23:46 UTC (permalink / raw)
  To: htejun, mikpe; +Cc: akpm, jgarzik, jirislaby, linux-ide, linux-kernel, stable

On Sat, 05 May 2007 17:30:51 +0200, Tejun Heo wrote:
> Mikael Pettersson wrote:
> >> I.e. no freezing of ports...
> > 
> > Your patch to delete the 'return 1;' on error is correct,
> > and makes the code match exactly the behaviour of previous
> > versions of sata_promise, except for the additional error
> > decoding.
> > 
> > ahci and sata_sil24 do the return in this situation. I don't
> > yet understand why they can get away with it while sata_promise
> > cannot, but for now the return should be removed.
> 
> That's because sata_sil24 and ahci call either ata_port_abort() or
> ata_port_freeze() prior to finishing error_intr routine.  Both functions
> abort all in-flight command and schedule EH.

Ah. Thanks, that clarifies things.

Jiri: please test the patch below instead. That is, revert to
the original code _with_ the 'return 1;', and then add this
patch to it. It should have pretty much the same effect as
removing the 'return 1;', however calling ata_port_abort()
is more in line with libata's new-style error handling.

/Mikael

--- linux-2.6.21-mm1/drivers/ata/sata_promise.c.~1~	2007-05-05 22:24:41.000000000 +0200
+++ linux-2.6.21-mm1/drivers/ata/sata_promise.c	2007-05-05 22:25:21.000000000 +0200
@@ -653,6 +653,8 @@ static void pdc_error_intr(struct ata_po
 	qc->err_mask |= ac_err_mask;
 
 	pdc_reset_port(ap);
+
+	ata_port_abort(ap);
 }
 
 static inline unsigned int pdc_host_intr( struct ata_port *ap,

^ permalink raw reply	[flat|nested] 6+ messages in thread
end of thread, other threads:[~2007-05-07  9:52 UTC | newest]

Thread overview: 6+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
     [not found] <46360DA7.6040003@gmail.com>
     [not found] ` <46371128.4060101@gmail.com>
2007-05-04 14:58   ` 2.6.21-mm1: many processes end up in D state Jiri Slaby
2007-05-04 15:02     ` Jiri Slaby
2007-05-04 19:19 Mikael Pettersson
2007-05-05 15:30 ` Tejun Heo
  -- strict thread matches above, loose matches on Subject: below --
2007-05-05 23:46 Mikael Pettersson
2007-05-07  9:52 ` Jiri Slaby

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).