Dropped interrupts

[Ben Laurie]

On 8 January 2014 06:44, Ben Laurie <ben at links.org> wrote:
> On 7 January 2014 18:11, Justin T. Gibbs <gibbs at scsiguy.com> wrote:
>> On Jan 7, 2014, at 12:36 AM, Ben Laurie <ben at links.org> wrote:
>>
>>> Attached.
>>>
>>> On 7 January 2014 05:46, Justin T. Gibbs <gibbs at scsiguy.com> wrote:
>>>> On Jan 6, 2014, at 3:01 PM, Ben Laurie <ben at links.org> wrote:
>>>>
>>>>> Not subscribed to the list, so please cc on replies.
>>>>>
>>>>> I'm using Bacula with an LTO-2 SCSI drive.
>>>>>
>>>>> With increasing frequency lately, I've been getting errors like this
>>>>> from bacula:
>>>>>
>>>>> backup-sd JobId 13092: Error: block.c:608 Write error at 23:6772 on
>>>>> device "Ultrium" (/dev/nsa0). ERR=Operation not permitted.
>>>>>
>>>>> Associated with this, I see in dmesg:
>>>>>
>>>>> ahc0: Recovery Initiated
>>>>>
>>>>> [a lot of dump info, including…]
>>>>
>>>> If you provide the dump info, I may be able to tell you why recovery is starting.
>>>>
>>>> The dmesg information from a boot of the system would be good to have too.
>>>>
>>>> —
>>>> Justin
>>
>> The target is keeping us in command phase for some reason.  No parity or other
>> errors are being reported.  My guess is that the tape drive does not like the command
>> that was issued for some reason.
>>
>> Attached are two totally untested/uncompiled changes for you to try out.  The first
>> should give more information about the command that timed out so we can better
>> determine if it is well formed.  The second is an attempted fix for spurious
>> “Interrupts may not be functioning” warnings.  Can you attempt to replicate this
>> again with these changes?
>
> Rebuilding now - you had a ; missing in the patch :-)
>
> Of course, now I've done this, it'll not fail for a month (its been
> failing multiple times per day recently, but on average its a lot
> rarer than that!).
>
> Will let you know when I get a fresh failure.

As predicted, it has now done 3 complete tapes with no problems, and
is on the fourth.