2005-05-03 18:13:36

by Shawn Starr

[permalink] [raw]
Subject: [2.6.12-rc3][SUSPEND] qla1280 (QLogic 12160 Ultra3) blows up on A7M266-D

I was feeling lucky yesterday and decided to try my
luck on an A7M266-D with suspend-to-disk. I noticed
two things

1) If I use XFS /w the SCSI controller (connected to 2
IBM HD 10K Ultra3 SCSI disks) I can suspend to disk no
problem, but resuming all hell breaks loose. It takes
a half an hour to reload the swap memory dumped to
disk. When it finally does resume. The SCSI controller
throws ISP errors and hangs. When rebooting after, the
partition table is completely gone (I did not try to
repair it).

2) If I use EXT3, suspending to disk is fine resuming
is fine there is no long delay to load the swap memory
back to RAM. But when it finishes resuming I get the
same ISP error and the partition table gets corrupt as
well.


Is it likely this SCSI driver doesn't know how to
handle suspend events?

Shawn.


2005-05-06 22:08:40

by Stefan Seyfried

[permalink] [raw]
Subject: Re: [2.6.12-rc3][SUSPEND] qla1280 (QLogic 12160 Ultra3) blows up on A7M266-D

Shawn Starr wrote:
> I was feeling lucky yesterday and decided to try my
> luck on an A7M266-D with suspend-to-disk. I noticed
> two things

> 1) If I use XFS /w the SCSI controller (connected to 2
> IBM HD 10K Ultra3 SCSI disks) I can suspend to disk no
> problem, but resuming all hell breaks loose. It takes
> a half an hour to reload the swap memory dumped to
> disk.

Known, XFS was broken / breaking wrt suspend. Pavel fixed this with the
XFS guys IIRC and i think those patches were on lkml also, but am not
sure. => this should work soon.

> 2) If I use EXT3, suspending to disk is fine resuming
> is fine there is no long delay to load the swap memory
> back to RAM. But when it finishes resuming I get the
> same ISP error and the partition table gets corrupt as
> well.

> Is it likely this SCSI driver doesn't know how to
> handle suspend events?

Yes. Almost all drivers that are not commonly used in notebooks are
totally ignorant of suspend / resume. Even the brand new SATA driver
stuff (that is actually in almost every new notebook) had no suspend
support until some days ago.
--
Stefan Seyfried
QA / R&D Team Mobile Devices | "Any ideas, John?"
SUSE LINUX Products GmbH, N?rnberg | "Well, surrounding them's out."

2005-05-07 08:40:57

by Christoph Hellwig

[permalink] [raw]
Subject: Re: [2.6.12-rc3][SUSPEND] qla1280 (QLogic 12160 Ultra3) blows up on A7M266-D

On Fri, May 06, 2005 at 11:34:02PM +0200, Stefan Seyfried wrote:
> Known, XFS was broken / breaking wrt suspend. Pavel fixed this with the
> XFS guys IIRC and i think those patches were on lkml also, but am not
> sure. => this should work soon.

Pavel's fix wasn't enough. The fix that has been verified to work is
in 2.6.12-rc4.

> > 2) If I use EXT3, suspending to disk is fine resuming
> > is fine there is no long delay to load the swap memory
> > back to RAM. But when it finishes resuming I get the
> > same ISP error and the partition table gets corrupt as
> > well.
>
> > Is it likely this SCSI driver doesn't know how to
> > handle suspend events?
>
> Yes. Almost all drivers that are not commonly used in notebooks are
> totally ignorant of suspend / resume. Even the brand new SATA driver
> stuff (that is actually in almost every new notebook) had no suspend
> support until some days ago.

qla1280 doesn't handle suspend/resume indeed.

2005-05-07 13:29:32

by Stefan Seyfried

[permalink] [raw]
Subject: Re: [2.6.12-rc3][SUSPEND] qla1280 (QLogic 12160 Ultra3) blows up on A7M266-D

Christoph Hellwig wrote:
> On Fri, May 06, 2005 at 11:34:02PM +0200, Stefan Seyfried wrote:
>> Known, XFS was broken / breaking wrt suspend. Pavel fixed this with the
>> XFS guys IIRC and i think those patches were on lkml also, but am not
>> sure. => this should work soon.
>
> Pavel's fix wasn't enough.

That's what i wanted to tell with "...with the XFS guys..." :-)

> The fix that has been verified to work is
> in 2.6.12-rc4.

Ok, i only knew there was something in the works.

> qla1280 doesn't handle suspend/resume indeed.

As almost all SCSI stuff, which is no surprise to me since suspend /
resume seem rather uncommon on servers where SCSI is mostly used today
;-) I was more baffled to find out that the brand new SATA drivers had
no suspend support.
--
Stefan Seyfried
QA / R&D Team Mobile Devices | "Any ideas, John?"
SUSE LINUX Products GmbH, N?rnberg | "Well, surrounding them's out."

2005-05-09 06:45:29

by Shawn Starr

[permalink] [raw]
Subject: Re: [2.6.12-rc3][SUSPEND] qla1280 (QLogic 12160 Ultra3) blows up on A7M266-D


Well, it makes sense, normally you won't find SCSI on desktops so why bother
with suspend. I don't know if the cards can do it or not, since they need
their firmware loaded at driver init. The firmware would need to be modified
to support such state?

Shawn.

On May 7, 2005 09:27, Stefan Seyfried wrote:
> Christoph Hellwig wrote:
> > On Fri, May 06, 2005 at 11:34:02PM +0200, Stefan Seyfried wrote:
> >> Known, XFS was broken / breaking wrt suspend. Pavel fixed this with the
> >> XFS guys IIRC and i think those patches were on lkml also, but am not
> >> sure. => this should work soon.
> >
> > Pavel's fix wasn't enough.
>
> That's what i wanted to tell with "...with the XFS guys..." :-)
>
> > The fix that has been verified to work is
> > in 2.6.12-rc4.
>
> Ok, i only knew there was something in the works.
>
> > qla1280 doesn't handle suspend/resume indeed.
>
> As almost all SCSI stuff, which is no surprise to me since suspend /
> resume seem rather uncommon on servers where SCSI is mostly used today
> ;-) I was more baffled to find out that the brand new SATA drivers had
> no suspend support.

2005-05-09 12:46:25

by Stefan Seyfried

[permalink] [raw]
Subject: Re: [2.6.12-rc3][SUSPEND] qla1280 (QLogic 12160 Ultra3) blows up on A7M266-D

Shawn Starr wrote:
> Well, it makes sense,

I tend to disagree

> normally you won't find SCSI on

low-end-

> desktops so why bother
> with suspend.

One example: developers often have SCSI systems. They cannot help fixing
their (audio|video|network card) drivers because they cannot test it.

> I don't know if the cards can do it or not, since they need
> their firmware loaded at driver init.

Lots of drivers do this without much problems. Wireless cards are a
prime example.

> The firmware would need to be modified
> to support such state?

I don't believe so.
--
seife
Never trust a computer you can't lift.