recursive lock-enter-deadlock

recursive lock-enter-deadlock

[R.Oehler]

A short question (I don't have a recent 2.4.x at hand, currently):

Is this recursive lock-enter-deadlock (2.4.0) fixed in newer kernels?

Regards,
        Ralf





void truncate_inode_pages(struct address_space * mapping, loff_t lstart) 
{
        unsigned long start = (lstart + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT;
        unsigned partial = lstart & (PAGE_CACHE_SIZE - 1);

repeat:
        spin_lock(&pagecache_lock);
        if (truncate_list_pages(&mapping->clean_pages, start, &partial))
                goto repeat;
        if (truncate_list_pages(&mapping->dirty_pages, start, &partial))
                goto repeat;
        if (truncate_list_pages(&mapping->locked_pages, start, &partial))
                goto repeat;
        spin_unlock(&pagecache_lock);
}

 -----------------------------------------------------------------
|  Ralf Oehler
|  GDI - Gesellschaft fuer Digitale Informationstechnik mbH
|
|  E-Mail:      R.Oehler@GDImbH.com
|  Tel.:        +49 6182-9271-23 
|  Fax.:        +49 6182-25035           
|  Mail:        GDI, Bensbruchstrasse 11, D-63533 Mainhausen
|  HTTP:        www.GDImbH.com
 -----------------------------------------------------------------

time is a funny concept

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/

Re: recursive lock-enter-deadlock

[Stephen C. Tweedie]

Hi,

On Wed, Nov 21, 2001 at 11:19:13AM +0100, R.Oehler@GDImbH.com wrote:
> A short question (I don't have a recent 2.4.x at hand, currently):
> 
> Is this recursive lock-enter-deadlock (2.4.0) fixed in newer kernels?

Yes.  Seriously, 2.4.0 is so old and so full of bugs like this that
it's really not worth spending any effort looking for problems like
that in it.

Cheers,
 Stephen
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/

Re: recursive lock-enter-deadlock

[R.Oehler]

On 21-Nov-2001 Stephen C. Tweedie wrote:
> Hi,
> 
> On Wed, Nov 21, 2001 at 11:19:13AM +0100, R.Oehler@GDImbH.com wrote:
>> A short question (I don't have a recent 2.4.x at hand, currently):
>> Is this recursive lock-enter-deadlock (2.4.0) fixed in newer kernels?
> 
> Yes.  Seriously, 2.4.0 is so old and so full of bugs like this that
> it's really not worth spending any effort looking for problems like
> that in it.
> 
Well, maybe, but it's the one distributed in SuSE-71. And it supports
block media with sectorsizes >1k. SuSE-73 shipped 2.4.10, which seems
to have a bug in the block layer which prevents me (and out commercial
product) from using 2k-sector and 4k-sector SCSI-media. 
I have to use a kernel from a SuSE-distribution.

The bug is easy to trigger with 
"dd if=/dev/zero of=/dev/sda bs=1M count=1" 
and an MO-drive with 2k-sector-medium.
The symptom is, that sd.c gets misaligned (only 1k-aligned) 
requests and complains loudly to the syslog.)

By the way: 2.4.10-ac works, as Alan says, so what changed in the linus'
kernel and didn't change in the -ac kernel between 2.4.0 and 2.4.10 ?

Regards,
        Ralf


 -----------------------------------------------------------------
|  Ralf Oehler
|  GDI - Gesellschaft fuer Digitale Informationstechnik mbH
|
|  E-Mail:      R.Oehler@GDImbH.com
|  Tel.:        +49 6182-9271-23 
|  Fax.:        +49 6182-25035           
|  Mail:        GDI, Bensbruchstrasse 11, D-63533 Mainhausen
|  HTTP:        www.GDImbH.com
 -----------------------------------------------------------------

time is a funny concept

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/

Re: recursive lock-enter-deadlock

[Stephen C. Tweedie]

Hi,

On Wed, Nov 21, 2001 at 12:52:11PM +0100, R.Oehler@GDImbH.com wrote:
 
> On 21-Nov-2001 Stephen C. Tweedie wrote:
> > Hi,
> > 
> > On Wed, Nov 21, 2001 at 11:19:13AM +0100, R.Oehler@GDImbH.com wrote:
> >> A short question (I don't have a recent 2.4.x at hand, currently):
> >> Is this recursive lock-enter-deadlock (2.4.0) fixed in newer kernels?
> > 
> > Yes.  Seriously, 2.4.0 is so old and so full of bugs like this that
> > it's really not worth spending any effort looking for problems like
> > that in it.

> Well, maybe, but it's the one distributed in SuSE-71.

Isn't that the one that only shipped 2.4 as a preview kernel?

For production use, it's common for distributions to apply their own
patches on top of the basic kernel --- mostly bugfixes back-ported
from later official kernel releases.  I'd be enormously surprised if
SuSE shipped a completely unpatched 2.4.0 as a production kernel, so
it's quite possible that the SuSE kernel has that fix applied.

> By the way: 2.4.10-ac works, as Alan says, so what changed in the linus'
> kernel and didn't change in the -ac kernel between 2.4.0 and 2.4.10 ?

2.4.10 had an absolute ton of block device layer changes which Alan
didn't apply to 2.4.10-ac.  Your bug isn't the only nasty in 2.4.10:
e2fsprogs gets bitten too, and things like tune2fs on mounted
filesystems are broken in that kernel.

Cheers,  
  Stephen
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/