* Re: Burning multiple DVDs at one time
[not found] <20120315225310.GB3152@animx.eu.org>
@ 2012-03-17 8:56 ` Stefan Richter
2012-03-17 10:53 ` Emmanuel Florac
2012-03-17 14:22 ` Wakko Warner
0 siblings, 2 replies; 9+ messages in thread
From: Stefan Richter @ 2012-03-17 8:56 UTC (permalink / raw)
To: Wakko Warner; +Cc: linux-kernel, linux-scsi
On Mar 15 Wakko Warner wrote:
> I'm having problems doing this.
>
> When I burn a single disk, wodim shows the drive buf @ 99% consistently.
> The instant that a 2nd disk is being burned, the drive buf on the first one
> starts to drop and data stops when the 2nd wodim is performing OPC.
>
> During the burn of both discs, the drive buf will drop on both until one of
> them finishes. Both drives see under runs.
>
> When one starts fixating, the other will hang until the fixation is
> completed.
>
> During the burns, the fifo of both never drop below 99%
>
> There are no logs that are produced.
>
> My burners are:
> [6:0:0:0] cd/dvd ATAPI iHAS422 8 4L11 /dev/sr7
> [7:0:0:0] cd/dvd ATAPI iHAS224 B GL05 /dev/sr8
>
> Both are SATA drives attached to the onboard sata controller
> 00:1f.2 SATA controller: Intel Corporation 631xESB/632xESB SATA AHCI Controller (rev 09)
> 00:1f.2 0106: 8086:2681 (rev 09)
>
> The kernel is a vanilla kernel v3.0.0. (This also happened with 2.6.35)
>
> I don't believe it matters, but the data is coming over NFS via gigabit
> ethernet. Each burn process uses a 768M host buffer. As I stated, the host
> fifo on wodim never goes below 99%
>
> The system motherboard is a SuperMicro X7DA8
This is likely due to serialization by a global mutex in the sr driver.
Have a look at thread "[PATCH] [SCSI] sr: fix multi-drive performance,
remove BKL replacement" from February. https://lkml.org/lkml/2012/2/28/230
--
Stefan Richter
-=====-===-- --== =---=
http://arcgraph.de/sr/
^ permalink raw reply [flat|nested] 9+ messages in thread
* Re: Burning multiple DVDs at one time
2012-03-17 8:56 ` Burning multiple DVDs at one time Stefan Richter
@ 2012-03-17 10:53 ` Emmanuel Florac
2012-03-17 12:47 ` Stefan Richter
2012-03-17 14:22 ` Wakko Warner
1 sibling, 1 reply; 9+ messages in thread
From: Emmanuel Florac @ 2012-03-17 10:53 UTC (permalink / raw)
To: Stefan Richter; +Cc: Wakko Warner, linux-kernel, linux-scsi
Le Sat, 17 Mar 2012 09:56:44 +0100 vous écriviez:
> This is likely due to serialization by a global mutex in the sr
> driver. Have a look at thread "[PATCH] [SCSI] sr: fix multi-drive
> performance, remove BKL replacement" from February.
> https://lkml.org/lkml/2012/2/28/230
When was this regression introduced? I have a dedicated cd burning
machine running 2.6.35.x, and it works great with 4 drives at a time.
--
------------------------------------------------------------------------
Emmanuel Florac | Direction technique
| Intellique
| <eflorac@intellique.com>
| +33 1 78 94 84 02
------------------------------------------------------------------------
--
To unsubscribe from this list: send the line "unsubscribe linux-scsi" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
^ permalink raw reply [flat|nested] 9+ messages in thread
* Re: Burning multiple DVDs at one time
2012-03-17 10:53 ` Emmanuel Florac
@ 2012-03-17 12:47 ` Stefan Richter
0 siblings, 0 replies; 9+ messages in thread
From: Stefan Richter @ 2012-03-17 12:47 UTC (permalink / raw)
To: Emmanuel Florac; +Cc: Wakko Warner, linux-kernel, linux-scsi
On Mar 17 Emmanuel Florac wrote:
> Le Sat, 17 Mar 2012 09:56:44 +0100 vous écriviez:
>
> > This is likely due to serialization by a global mutex in the sr
> > driver. Have a look at thread "[PATCH] [SCSI] sr: fix multi-drive
> > performance, remove BKL replacement" from February.
> > https://lkml.org/lkml/2012/2/28/230
>
> When was this regression introduced? I have a dedicated cd burning
> machine running 2.6.35.x, and it works great with 4 drives at a time.
BKL was replaced by sr_mutex in 2.6.37-rc1.
Wakko, you could try the patch from the mentioned thread and see whether
it fixes your problem. But be aware that the patch opens possible race
conditions between processes that access the same optical drive, e.g.
applications vs. udisks-daemon. (I haven't noticed anything bad during the
period of using the patch though.)
--
Stefan Richter
-=====-===-- --== =---=
http://arcgraph.de/sr/
--
To unsubscribe from this list: send the line "unsubscribe linux-scsi" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
^ permalink raw reply [flat|nested] 9+ messages in thread
* Re: Burning multiple DVDs at one time
2012-03-17 8:56 ` Burning multiple DVDs at one time Stefan Richter
2012-03-17 10:53 ` Emmanuel Florac
@ 2012-03-17 14:22 ` Wakko Warner
2012-03-17 15:28 ` Stefan Richter
1 sibling, 1 reply; 9+ messages in thread
From: Wakko Warner @ 2012-03-17 14:22 UTC (permalink / raw)
To: Stefan Richter; +Cc: linux-kernel, linux-scsi
Stefan Richter wrote:
> On Mar 15 Wakko Warner wrote:
> > I'm having problems doing this.
> >
> > When I burn a single disk, wodim shows the drive buf @ 99% consistently.
> > The instant that a 2nd disk is being burned, the drive buf on the first one
> > starts to drop and data stops when the 2nd wodim is performing OPC.
> >
> > During the burn of both discs, the drive buf will drop on both until one of
> > them finishes. Both drives see under runs.
> >
> > When one starts fixating, the other will hang until the fixation is
> > completed.
> >
> > During the burns, the fifo of both never drop below 99%
> >
> > There are no logs that are produced.
> >
> > My burners are:
> > [6:0:0:0] cd/dvd ATAPI iHAS422 8 4L11 /dev/sr7
> > [7:0:0:0] cd/dvd ATAPI iHAS224 B GL05 /dev/sr8
> >
> > Both are SATA drives attached to the onboard sata controller
> > 00:1f.2 SATA controller: Intel Corporation 631xESB/632xESB SATA AHCI Controller (rev 09)
> > 00:1f.2 0106: 8086:2681 (rev 09)
> >
> > The kernel is a vanilla kernel v3.0.0. (This also happened with 2.6.35)
> >
> > I don't believe it matters, but the data is coming over NFS via gigabit
> > ethernet. Each burn process uses a 768M host buffer. As I stated, the host
> > fifo on wodim never goes below 99%
> >
> > The system motherboard is a SuperMicro X7DA8
>
> This is likely due to serialization by a global mutex in the sr driver.
> Have a look at thread "[PATCH] [SCSI] sr: fix multi-drive performance,
> remove BKL replacement" from February. https://lkml.org/lkml/2012/2/28/230
Thanks. I looked at the patch. I would just like to confirm that I can
patch my 3.0.0 vanilla kernel, compile the sr module, unload the current and
load the patched one without the need to reboot.
I do not use udisks.
Thanks for the responce, I'll try this when I have more discs to burn.
--
Microsoft has beaten Volkswagen's world record. Volkswagen only created 22
million bugs.
^ permalink raw reply [flat|nested] 9+ messages in thread
* Re: Burning multiple DVDs at one time
2012-03-17 14:22 ` Wakko Warner
@ 2012-03-17 15:28 ` Stefan Richter
2012-03-18 13:30 ` Wakko Warner
0 siblings, 1 reply; 9+ messages in thread
From: Stefan Richter @ 2012-03-17 15:28 UTC (permalink / raw)
To: Wakko Warner; +Cc: linux-kernel, linux-scsi
On Mar 17 Wakko Warner wrote:
> Stefan Richter wrote:
> > On Mar 15 Wakko Warner wrote:
> > > I'm having problems doing this.
> > >
> > > When I burn a single disk, wodim shows the drive buf @ 99% consistently.
> > > The instant that a 2nd disk is being burned, the drive buf on the first one
> > > starts to drop and data stops when the 2nd wodim is performing OPC.
> > >
> > > During the burn of both discs, the drive buf will drop on both until one of
> > > them finishes. Both drives see under runs.
> > >
> > > When one starts fixating, the other will hang until the fixation is
> > > completed.
> > >
> > > During the burns, the fifo of both never drop below 99%
> > >
> > > There are no logs that are produced.
> > >
> > > My burners are:
> > > [6:0:0:0] cd/dvd ATAPI iHAS422 8 4L11 /dev/sr7
> > > [7:0:0:0] cd/dvd ATAPI iHAS224 B GL05 /dev/sr8
> > >
> > > Both are SATA drives attached to the onboard sata controller
> > > 00:1f.2 SATA controller: Intel Corporation 631xESB/632xESB SATA AHCI Controller (rev 09)
> > > 00:1f.2 0106: 8086:2681 (rev 09)
[...]
> > This is likely due to serialization by a global mutex in the sr driver.
> > Have a look at thread "[PATCH] [SCSI] sr: fix multi-drive performance,
> > remove BKL replacement" from February. https://lkml.org/lkml/2012/2/28/230
>
> Thanks. I looked at the patch. I would just like to confirm that I can
> patch my 3.0.0 vanilla kernel, compile the sr module, unload the current and
> load the patched one without the need to reboot.
Yes, this should be possible.
Oh, I only noticed just know that you also wrote:
> > > The kernel is a vanilla kernel v3.0.0. (This also happened with 2.6.35)
In 2.6.35, the Big Kernel Lock was still in place there. That lock
behaved differently from a plain mutex --- notably it was released when a
thread went to sleep --- so maybe there is more to your problem than just
sr_mutex blocking unrelated sr accesses.
--
Stefan Richter
-=====-===-- --== =---=
http://arcgraph.de/sr/
^ permalink raw reply [flat|nested] 9+ messages in thread
* Re: Burning multiple DVDs at one time
2012-03-17 15:28 ` Stefan Richter
@ 2012-03-18 13:30 ` Wakko Warner
2012-03-18 14:01 ` Stefan Richter
0 siblings, 1 reply; 9+ messages in thread
From: Wakko Warner @ 2012-03-18 13:30 UTC (permalink / raw)
To: Stefan Richter; +Cc: linux-kernel, linux-scsi
Stefan Richter wrote:
> On Mar 17 Wakko Warner wrote:
> > Thanks. I looked at the patch. I would just like to confirm that I can
> > patch my 3.0.0 vanilla kernel, compile the sr module, unload the current and
> > load the patched one without the need to reboot.
>
> Yes, this should be possible.
>
> Oh, I only noticed just know that you also wrote:
>
> > > > The kernel is a vanilla kernel v3.0.0. (This also happened with 2.6.35)
>
> In 2.6.35, the Big Kernel Lock was still in place there. That lock
> behaved differently from a plain mutex --- notably it was released when a
> thread went to sleep --- so maybe there is more to your problem than just
> sr_mutex blocking unrelated sr accesses.
There is a good possibility that I was mistaken on the 2.6.35.
I manually applied your patch to 3.0.0, compiled, removed sr-mod module and
inserted the newly compiled one.
With the vanilla one, I would see the buf % in wodim drop while the fifo was
full. With the patch (removing the mutex), buf stayed above 98% and the
burn speed was 16x at the end (10x avg).
Seems to have fixed my problem.
Thanks.
--
Microsoft has beaten Volkswagen's world record. Volkswagen only created 22
million bugs.
^ permalink raw reply [flat|nested] 9+ messages in thread
* Re: Burning multiple DVDs at one time
2012-03-18 13:30 ` Wakko Warner
@ 2012-03-18 14:01 ` Stefan Richter
2012-04-28 16:02 ` Wakko Warner
0 siblings, 1 reply; 9+ messages in thread
From: Stefan Richter @ 2012-03-18 14:01 UTC (permalink / raw)
To: Wakko Warner; +Cc: linux-kernel, linux-scsi
On Mar 18 Wakko Warner wrote:
> Stefan Richter wrote:
> > On Mar 17 Wakko Warner wrote:
> > > Thanks. I looked at the patch. I would just like to confirm that I can
> > > patch my 3.0.0 vanilla kernel, compile the sr module, unload the current and
> > > load the patched one without the need to reboot.
> >
> > Yes, this should be possible.
> >
> > Oh, I only noticed just know that you also wrote:
> >
> > > > > The kernel is a vanilla kernel v3.0.0. (This also happened with 2.6.35)
> >
> > In 2.6.35, the Big Kernel Lock was still in place there. That lock
> > behaved differently from a plain mutex --- notably it was released when a
> > thread went to sleep --- so maybe there is more to your problem than just
> > sr_mutex blocking unrelated sr accesses.
>
> There is a good possibility that I was mistaken on the 2.6.35.
>
> I manually applied your patch to 3.0.0, compiled, removed sr-mod module and
> inserted the newly compiled one.
>
> With the vanilla one, I would see the buf % in wodim drop while the fifo was
> full. With the patch (removing the mutex), buf stayed above 98% and the
> burn speed was 16x at the end (10x avg).
>
> Seems to have fixed my problem.
Very nice; so there is an almost easy fix. Now I or somebody else need to
work out the required cdrom_device_info protection on which Arnd and James
commented.
--
Stefan Richter
-=====-===-- --== =--=-
http://arcgraph.de/sr/
^ permalink raw reply [flat|nested] 9+ messages in thread
* Re: Burning multiple DVDs at one time
2012-03-18 14:01 ` Stefan Richter
@ 2012-04-28 16:02 ` Wakko Warner
2012-04-28 16:55 ` Stefan Richter
0 siblings, 1 reply; 9+ messages in thread
From: Wakko Warner @ 2012-04-28 16:02 UTC (permalink / raw)
To: Stefan Richter; +Cc: linux-kernel, linux-scsi
Stefan Richter wrote:
> On Mar 18 Wakko Warner wrote:
> > Stefan Richter wrote:
> > > On Mar 17 Wakko Warner wrote:
> > > > Thanks. I looked at the patch. I would just like to confirm that I can
> > > > patch my 3.0.0 vanilla kernel, compile the sr module, unload the current and
> > > > load the patched one without the need to reboot.
> > >
> > > Yes, this should be possible.
> > >
> > > Oh, I only noticed just know that you also wrote:
> > >
> > > > > > The kernel is a vanilla kernel v3.0.0. (This also happened with 2.6.35)
> > >
> > > In 2.6.35, the Big Kernel Lock was still in place there. That lock
> > > behaved differently from a plain mutex --- notably it was released when a
> > > thread went to sleep --- so maybe there is more to your problem than just
> > > sr_mutex blocking unrelated sr accesses.
> >
> > There is a good possibility that I was mistaken on the 2.6.35.
> >
> > I manually applied your patch to 3.0.0, compiled, removed sr-mod module and
> > inserted the newly compiled one.
> >
> > With the vanilla one, I would see the buf % in wodim drop while the fifo was
> > full. With the patch (removing the mutex), buf stayed above 98% and the
> > burn speed was 16x at the end (10x avg).
> >
> > Seems to have fixed my problem.
>
> Very nice; so there is an almost easy fix. Now I or somebody else need to
> work out the required cdrom_device_info protection on which Arnd and James
> commented.
Has anything come of this as far as the fix going into the kernel?
--
Microsoft has beaten Volkswagen's world record. Volkswagen only created 22
million bugs.
^ permalink raw reply [flat|nested] 9+ messages in thread
* Re: Burning multiple DVDs at one time
2012-04-28 16:02 ` Wakko Warner
@ 2012-04-28 16:55 ` Stefan Richter
0 siblings, 0 replies; 9+ messages in thread
From: Stefan Richter @ 2012-04-28 16:55 UTC (permalink / raw)
To: Wakko Warner; +Cc: linux-kernel, linux-scsi
On Apr 28 Wakko Warner wrote:
> Stefan Richter wrote:
> > On Mar 18 Wakko Warner wrote:
[sr_mutex removal]
> > > Seems to have fixed my problem.
> >
> > Very nice; so there is an almost easy fix. Now I or somebody else need to
> > work out the required cdrom_device_info protection on which Arnd and James
> > commented.
>
> Has anything come of this as far as the fix going into the kernel?
I haven't been working on it in the meantime; don't know if anybode else
did.
--
Stefan Richter
-=====-===-- -=-- ===--
http://arcgraph.de/sr/
^ permalink raw reply [flat|nested] 9+ messages in thread
end of thread, other threads:[~2012-04-28 16:55 UTC | newest]
Thread overview: 9+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
[not found] <20120315225310.GB3152@animx.eu.org>
2012-03-17 8:56 ` Burning multiple DVDs at one time Stefan Richter
2012-03-17 10:53 ` Emmanuel Florac
2012-03-17 12:47 ` Stefan Richter
2012-03-17 14:22 ` Wakko Warner
2012-03-17 15:28 ` Stefan Richter
2012-03-18 13:30 ` Wakko Warner
2012-03-18 14:01 ` Stefan Richter
2012-04-28 16:02 ` Wakko Warner
2012-04-28 16:55 ` Stefan Richter
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).