RE: Hangs seen with the 3ware controller and the 2.4.17 kernel ..

[Adam Radford]

You're using an old driver before the conversion to
documentation/DMA-mapping.  This
will cause raw io to 'not work'.   If you upgrade to the latest driver
1.02.00.031
driver, you should be fine.

-Adam

-----Original Message-----
From: Manish Lachwani [mailto:[e-mail]]
Sent: Monday, November 11, 2002 11:54 PM
To: [e-mail]
Cc: Manish Lachwani
Subject: Hangs seen with the 3ware controller and the 2.4.17 kernel ...

Hello,

I am using a 2.4.17 SMP kernel and .018 version of the 3ware driver. This
happens when we have two controllers (8-port and 4-port), IO is going on
with both the controllers and on one controller (4-port in my experiment),
there is  command timeout and the reset sequence fails. This is a hard
kernel hang. The last message on the window is "reset sequence failed"

kdb for the eh_1 shows:

scsi_error_handler -> scsi_unjam_host -> scsi_try_host_reset -> schedule

I do know that the scsi_try_host_reset(..) calls scsi_sleep for 10*HZ. 

Anyway, another scenario that causes a hang:

scsi_error_handler -> scsi_unjam_host -> scsi_try_to_abort_command ->
schedule

Also, This hang seems to occur when there are two controllers only. When I
tried with 
one controller numerous times, I could not reproduce this problem. Is it
possible that scsi_unjam_host is getting confused with two devices and when
reset fails on one host?

Any help is appreciated ...