Re: PROBLEM: 2.6.0-test9: SCSI mid layer tells a lie.

[Hironobu Ishii <ishii.hironobu@jp.fujitsu.com>]

Hi all,

I also tested this probem with 2.6.0-test11.
This problem has not been fixed yet.

Thanks, 
Hironobu Ishii
----- Original Message ----- 
From: "Hironobu Ishii" <ishii.hironobu@jp.fujitsu.com>
To: "linux-scsi" <linux-scsi@vger.kernel.org>
Sent: Wednesday, December 03, 2003 3:43 PM
Subject: PROBLEM: 2.6.0-test9: SCSI mid layer tells a lie.


> Hi all,
> 
> I am verifying error recovery logics of SCSI mid layer with
> pseudo target device.
> In my test, I found a data corruption problem.
> Please see bellow.
> 
> Thanks,
> Hironobu Ishii.
> ---------------
> [1.] One line summary of the problem:
> 2.6.0-test9: SCSI mid layer tells a lie.
> 
> [2.] Full description of the problem/report:
> Kernel:  2.6.0-test9 vanilla
> 
> Problem:
>  SCSI mid layer failed to read(or write) the device,
>  but it returns normal completion to the application.
> 
>  The sequence is as follows.
>  SCSI mid layer repeats (a) part 5 times(SD_MAX_RETRIES).
>  I found this problem occurs with either READ or WRITE command.
> 
>   Initiator              LLD(Fusion MPT)           Target
>   -----------------------------------------------------------
> +- READ(or WRITE)        --------------------->   (Time out)
> |
> | eh_abort               --------------------->
> |                        LLD issues abort msg,
> |                        but it doesn't wait for its completion
> |                        and eh_aobrt_handler returns 0x2003(FAILED).
> |
> | eh_device_reset_handler
> |                        LLD issues nothing on the SCSI BUS
> (a)                      and returns 0x2003(FAILED)
> |
> | eh_device_bus_reset_handler
> |                        --------------------->   BUS RESET
> |                        LLD returns 0x2002(SUCCESS)
> |
> | TEST UNIT READY         -------------------->
> |                         <--------------------   CHK(06/0000)
> |
> | TEST UNIT READY         --------------------->
> +-                        <---------------------  GOOD
> 
> The purpose of this test is to verify operation when there is
> a medium error in disk.
> 
> I tested this problem with test6 and test9.
> I got the same result with either.
> I'm going to re-test with test11. But it takes for a while.
> (I looked at the diff between test6 and test11, but I can't
>  find a fix relating to this problem.)
> 
> Environments:
> Initiator HBA: LSI Logic 53c1030(Fusion MPT)
> Target:        Pseudo target device
> Operation:     dd  if=/dev/sde of=/tmp/read_data count=1
>               (or dd if=/tmp/data of=/dev/sde count=1)
> 
> [3.] Keywords (i.e., modules, networking, kernel):
>  scsi_mod, time out
> 
> [4.] Kernel version (from /proc/version):
> Linux version 2.6.0-test9 (root@lsd6129) (gcc version 3.2.2 20030222 (Red Hat
>  Linux 3.2.2-5)) #2 SMP Mon Nov 10 15:48:58 JST 2003
> 
> [5.] Output of Oops.. message (if applicable) with symbolic information
>      resolved (see Documentation/oops-tracing.txt)
> [6.] A small shell script or example program which triggers the
>      problem (if possible)
> [7.] Environment
>  See above.