Hello,
Using 3.12.0 and ext4fs with 2 x SSDs in a RAID-1 configuration on a
3ware HW RAID card, no md/dm, I noticed the following recently:
[178339.353565] sda2: WRITE SAME failed. Manually zeroing.
It seems to be similar to this issue here:
http://permalink.gmane.org/gmane.linux.kernel/1494512
Relevant logs:
[178339.353519] 3w-sas: scsi0: ERROR: (0x03:0x0101): Invalid command
opcode:opcode=0x41.
[178339.353565] sda2: WRITE SAME failed. Manually zeroing.
Is this something I should be worried about?
I am running a database on the host, could that be the cause of it?
(in the related post, he was also running a DB).
Justin.
>>>>> "Justin" == Justin Piszcz <[email protected]> writes:
Justin> Relevant logs: [178339.353519] 3w-sas: scsi0: ERROR:
Justin> (0x03:0x0101): Invalid command opcode:opcode=0x41.
Justin> [178339.353565] sda2: WRITE SAME failed. Manually zeroing.
Justin> Is this something I should be worried about?
No, it just means that your 3ware controller does not handle the WRITE
SAME command and we fall back to manually writing out zeroes.
I posted a patch a while back that prevents WRITE SAME being issued for
controllers that do not pass the commands through to disk directly:
http://marc.info/?l=linux-scsi&m=138252394614920&w=2
--
Martin K. Petersen Oracle Linux Engineering
-----Original Message-----
From: Martin K. Petersen [mailto:[email protected]]
Sent: Thursday, November 28, 2013 8:00 AM
To: Justin Piszcz
Cc: open list; [email protected]; [email protected]
Subject: Re: 3.12.0: sda2: WRITE SAME failed. Manually zeroing. with 3w-xxxx
driver
>>>>> "Justin" == Justin Piszcz <[email protected]> writes:
Justin> Relevant logs: [178339.353519] 3w-sas: scsi0: ERROR:
Justin> (0x03:0x0101): Invalid command opcode:opcode=0x41.
Justin> [178339.353565] sda2: WRITE SAME failed. Manually zeroing.
Justin> Is this something I should be worried about?
No, it just means that your 3ware controller does not handle the WRITE
SAME command and we fall back to manually writing out zeroes.
I posted a patch a while back that prevents WRITE SAME being issued for
controllers that do not pass the commands through to disk directly:
http://marc.info/?l=linux-scsi&m=138252394614920&w=2
[ .. ]
Awesome, thank you! This patch is over a month old, do you know if is
currently in mainline? I'll also update my case with 3ware/LSI.
Justin.
>>>>> "Justin" == Justin Piszcz <[email protected]> writes:
Martin> I posted a patch a while back that prevents WRITE SAME being
Martin> issued for controllers that do not pass the commands through to
Martin> disk directly:
Martin> http://marc.info/?l=linux-scsi&m=138252394614920&w=2
Justin> Awesome, thank you! This patch is over a month old, do you know
Justin> if is currently in mainline?
Nope. - James?
--
Martin K. Petersen Oracle Linux Engineering
-----Original Message-----
From: Martin K. Petersen [mailto:[email protected]]
Sent: Thursday, November 28, 2013 8:13 AM
[ .. ]
Martin> http://marc.info/?l=linux-scsi&m=138252394614920&w=2
Justin> Awesome, thank you! This patch is over a month old, do you know
Justin> if is currently in mainline?
Nope. - James?
[ .. ]
Could it possibly get added to 3.12.x -stable as well? I've patched against
3.12.0 for now -- I'll let you know if there are any issues-- so far so
good, thanks again.
Justin.
> -----Original Message-----
> From: Martin K. Petersen [mailto:[email protected]]
> Sent: Thursday, November 28, 2013 8:13 AM
>
>
> Martin> http://marc.info/?l=linux-scsi&m=138252394614920&w=2
>
> Justin> Awesome, thank you! This patch is over a month old, do you know
> Justin> if is currently in mainline?
>
> Nope. - James?
>
> Could it possibly get added to 3.12.x -stable as well? I've patched
against
> 3.12.0 for now -- I'll let you know if there are any issues-- so far so
> good, thanks again.
>
> Justin.
Update:
System uptime with the patch is 7+ days under extensive load at times and
I've had no issues to date.
I just checked the current linux git repo (3.13-rc2) and found this patch
has not been included.
Please advise if any further testing or validation needs to be completed to
help get this into -stable/-mainline.
Justin.
On Thu, Dec 05, 2013 at 06:43:39PM -0500, Justin Piszcz wrote:
>
>
> > -----Original Message-----
> > From: Martin K. Petersen [mailto:[email protected]]
> > Sent: Thursday, November 28, 2013 8:13 AM
> >
> >
> > Martin> http://marc.info/?l=linux-scsi&m=138252394614920&w=2
> >
> > Justin> Awesome, thank you! This patch is over a month old, do you know
> > Justin> if is currently in mainline?
> >
> > Nope. - James?
> >
> > Could it possibly get added to 3.12.x -stable as well? I've patched
> against
> > 3.12.0 for now -- I'll let you know if there are any issues-- so far so
> > good, thanks again.
> >
> > Justin.
>
> Update:
> System uptime with the patch is 7+ days under extensive load at times and
> I've had no issues to date.
> I just checked the current linux git repo (3.13-rc2) and found this patch
> has not been included.
>
> Please advise if any further testing or validation needs to be completed to
> help get this into -stable/-mainline.
It needs to get into Linus's tree first, after that, let
[email protected] know the git commit id and I'll be glad to queue
it up for a stable kernel release.
thanks,
greg k-h
>>>>> "Justin" == Justin Piszcz <[email protected]> writes:
Justin> Update: System uptime with the patch is 7+ days under extensive
Justin> load at times and I've had no issues to date. I just checked
Justin> the current linux git repo (3.13-rc2) and found this patch has
Justin> not been included.
James has it queued in scsi/fixes. It'll make it to Linus on the next
pull and it has been tagged for inclusion in stable.
--
Martin K. Petersen Oracle Linux Engineering