IO stall in FreeBSD 9.1

[Sibananda Sahu]

Is this a known issue on FreeBSD-9.1-RELEASE ?

I went through the ERRATA and RELEASE NOTES of FreeBSD-9.1-RELEASE (where I
observe the issue consistently) and FreeBSD-8.4-RELEASE (where this issue
is not observed), but I really did not find any info on issues related to
CAM or GEOM.



Thanks,

Sibananda Sahu





*From:* Sibananda Sahu [mailto:sibananda.sahu at avagotech.com]
*Sent:* Monday, June 08, 2015 7:43 PM
*To:* 'freebsd-scsi at freebsd.org'
*Cc:* 'Kenneth D. Merry'; 'scottl at freebsd.org'; 'Scott Long'; 'Konstantin
Belousov'
*Subject:* IO stall in FreeBSD 9.1



Hi all,



I am working on mrsas(4) driver for AVAGO’s 3108 Invader controllers in
FreeBSD.

Recently while testing on FreeBSD-9.1-RELEASE I have observed that IOs
running on VDs behind AVAGO 3108 controller, stalls after a while.

I have two VDs, both configured of RAID0.

I use fio(1) for to fire IOs to the disks with posixaio as the ioengine.





POST ISSUE ANALYSIS DATA:



In the gstat(1) utility it shows some pending items in the Queue and there
is no activity of IO but fio(1) is running in the background.



dT: 1.002s  w: 1.000s

L(q)        ops/s    r/s   kBps             ms/r      w/s   kBps
ms/w   %busy Name

   11        0              0      0    0.0          0              0
          0.0          0.0|da0

   32        0              0      0    0.0          0              0
          0.0          0.0| da1

    0          0              0      0    0.0          0              0
           0.0          0.0| ada0

    0          0              0      0    0.0          0              0
           0.0          0.0| ada0p1

    0          0              0      0    0.0          0              0
          0.0          0.0| ada0p2

    0          0              0      0    0.0          0              0
          0.0          0.0| ada0p3



I have also used the same configuration on FreeBSD-8.4-RELEASE and
FreeBSD-10.0-RELEASE and the issue is not seen.

IOs were running consistently and there was no IO stall at all.



Furthermore, I have also checked that if mrsas(4) driver is causing this
issue, but to my surprise mrsas(4) is not at all called after that IO stall.

I mean to say, the SIM driver’s action callback (mrsas_action() in this
case) is not at all called.



Below is a dmesg log of camcontrol debug –TSX [bus:target:lun] after IO
stall:

http://www.infysim.org/wp-content/dmesg_camcontrol_debug





Can anyone provide any solution/suggestion to this issue?





Thanks,

Sibananda Sahu