2007-10-18 19:29:26

by Ingo Molnar

[permalink] [raw]
Subject: Re: libata crash


* Jens Axboe <[email protected]> wrote:

> Hi Ingo,
>
> Can you try this rolled up patch against current -git and see if it
> works for you?

update: booted up a couple of times with your patch (with
DEBUG_PAGEALLOC enabled) and it's working all fine, no crashes. So i
think you nailed it!

Ingo


2007-10-18 19:32:50

by Jens Axboe

[permalink] [raw]
Subject: Re: libata crash

On Thu, Oct 18 2007, Ingo Molnar wrote:
>
> * Jens Axboe <[email protected]> wrote:
>
> > Hi Ingo,
> >
> > Can you try this rolled up patch against current -git and see if it
> > works for you?
>
> update: booted up a couple of times with your patch (with
> DEBUG_PAGEALLOC enabled) and it's working all fine, no crashes. So i
> think you nailed it!

Awesome, thanks a lot for testing it Ingo!

I'll see if I can spend some time tonight chopping the patch into
digestable pieces, unfortunately I'll be away the next three days. And
it would be nice to get this thing merged asap.

--
Jens Axboe

2007-10-18 19:59:31

by Ingo Molnar

[permalink] [raw]
Subject: Re: libata crash

* Jens Axboe <[email protected]> wrote:

> I'll see if I can spend some time tonight chopping the patch into
> digestable pieces, unfortunately I'll be away the next three days. And
> it would be nice to get this thing merged asap.

and make sure it builds:

drivers/scsi/fd_mcs.c: In function 'fd_mcs_intr':
drivers/scsi/fd_mcs.c:976: error: 'struct scatterlist' has no member named 'page'
drivers/scsi/fd_mcs.c:1009: error: 'struct scatterlist' has no member named 'page'
drivers/scsi/fd_mcs.c: In function 'fd_mcs_queue':
drivers/scsi/fd_mcs.c:1112: error: 'struct scatterlist' has no member named 'page'
drivers/scsi/in2000.c: In function 'in2000_queuecommand':
drivers/scsi/in2000.c:375: error: 'struct scatterlist' has no member named 'page'
drivers/scsi/in2000.c: In function 'transfer_bytes':
drivers/scsi/in2000.c:767: error: 'struct scatterlist' has no member named 'page'

;-)

Ingo

2007-10-18 20:41:24

by Jens Axboe

[permalink] [raw]
Subject: Re: libata crash

On Thu, Oct 18 2007, Ingo Molnar wrote:
> * Jens Axboe <[email protected]> wrote:
>
> > I'll see if I can spend some time tonight chopping the patch into
> > digestable pieces, unfortunately I'll be away the next three days. And
> > it would be nice to get this thing merged asap.
>
> and make sure it builds:
>
> drivers/scsi/fd_mcs.c: In function 'fd_mcs_intr':
> drivers/scsi/fd_mcs.c:976: error: 'struct scatterlist' has no member named 'page'
> drivers/scsi/fd_mcs.c:1009: error: 'struct scatterlist' has no member named 'page'
> drivers/scsi/fd_mcs.c: In function 'fd_mcs_queue':
> drivers/scsi/fd_mcs.c:1112: error: 'struct scatterlist' has no member named 'page'
> drivers/scsi/in2000.c: In function 'in2000_queuecommand':
> drivers/scsi/in2000.c:375: error: 'struct scatterlist' has no member named 'page'
> drivers/scsi/in2000.c: In function 'transfer_bytes':
> drivers/scsi/in2000.c:767: error: 'struct scatterlist' has no member named 'page'

Oops, wonder why they were missed in the allyes/modconfig. Fixed now,
thanks!

--
Jens Axboe

2007-10-18 20:52:33

by Jeff Garzik

[permalink] [raw]
Subject: Re: libata crash

Jens Axboe wrote:
> On Thu, Oct 18 2007, Ingo Molnar wrote:
>> * Jens Axboe <[email protected]> wrote:
>>
>>> I'll see if I can spend some time tonight chopping the patch into
>>> digestable pieces, unfortunately I'll be away the next three days. And
>>> it would be nice to get this thing merged asap.
>> and make sure it builds:
>>
>> drivers/scsi/fd_mcs.c: In function 'fd_mcs_intr':
>> drivers/scsi/fd_mcs.c:976: error: 'struct scatterlist' has no member named 'page'
>> drivers/scsi/fd_mcs.c:1009: error: 'struct scatterlist' has no member named 'page'
>> drivers/scsi/fd_mcs.c: In function 'fd_mcs_queue':
>> drivers/scsi/fd_mcs.c:1112: error: 'struct scatterlist' has no member named 'page'
>> drivers/scsi/in2000.c: In function 'in2000_queuecommand':
>> drivers/scsi/in2000.c:375: error: 'struct scatterlist' has no member named 'page'
>> drivers/scsi/in2000.c: In function 'transfer_bytes':
>> drivers/scsi/in2000.c:767: error: 'struct scatterlist' has no member named 'page'
>
> Oops, wonder why they were missed in the allyes/modconfig. Fixed now,
> thanks!

Building on x86-64, I'm betting? :)

I fell victim to the same thing a few days ago, missing some compile
breakage that only appeared with

make ARCH=i386 allmodconfig && make ARCH=i386 -sj9

Though am I alone in dreaming of a kernel.org service that would permit
all-arch build testing of a git URL?

Regards,

Jeff



2007-10-19 06:48:54

by Ingo Molnar

[permalink] [raw]
Subject: Re: libata crash


* Jens Axboe <[email protected]> wrote:

> > and make sure it builds:
> >
> > drivers/scsi/fd_mcs.c: In function 'fd_mcs_intr':
> > drivers/scsi/fd_mcs.c:976: error: 'struct scatterlist' has no member named 'page'
> > drivers/scsi/fd_mcs.c:1009: error: 'struct scatterlist' has no member named 'page'
> > drivers/scsi/fd_mcs.c: In function 'fd_mcs_queue':
> > drivers/scsi/fd_mcs.c:1112: error: 'struct scatterlist' has no member named 'page'
> > drivers/scsi/in2000.c: In function 'in2000_queuecommand':
> > drivers/scsi/in2000.c:375: error: 'struct scatterlist' has no member named 'page'
> > drivers/scsi/in2000.c: In function 'transfer_bytes':
> > drivers/scsi/in2000.c:767: error: 'struct scatterlist' has no member named 'page'
>
> Oops, wonder why they were missed in the allyes/modconfig. Fixed now,
> thanks!

where can i pick the latest sg patchset up from? These build failures
are ruining my make randconfig build tests :)

Ingo

2007-10-19 07:21:07

by Jan Dittmer

[permalink] [raw]
Subject: Re: libata crash

Jeff Garzik wrote:
> Building on x86-64, I'm betting? :)
>
> I fell victim to the same thing a few days ago, missing some compile
> breakage that only appeared with
>
> make ARCH=i386 allmodconfig && make ARCH=i386 -sj9
>
> Though am I alone in dreaming of a kernel.org service that would permit
> all-arch build testing of a git URL?

I could add that to the service at http://l4x.org/k/ if there is
sufficient (>1) interest.

Jan

2007-10-19 10:12:59

by Denys Vlasenko

[permalink] [raw]
Subject: Re: libata crash

On Friday 19 October 2007 08:05, Jan Dittmer wrote:
> Jeff Garzik wrote:
> > Building on x86-64, I'm betting? :)
> >
> > I fell victim to the same thing a few days ago, missing some compile
> > breakage that only appeared with
> >
> > make ARCH=i386 allmodconfig && make ARCH=i386 -sj9
> >
> > Though am I alone in dreaming of a kernel.org service that would permit
> > all-arch build testing of a git URL?
>
> I could add that to the service at http://l4x.org/k/ if there is
> sufficient (>1) interest.

Wow.

When you click on Ok/Fail, you get a build log. One part of it
(make oldconfig output) is not usable much, verbatim .config would be more useful.

But this is a minor nitpicking. Overall looks awesome.
--
vda