Skip to contentSkip to conversation

2.4.5-ac3: qlogic corruption on alpha

DISCUSSION Tom Vier 2 participants 5 messages May 30, 2001

lore
Share & tools
lore

Tom Vier

author
#0 lore raw

i narrowed down some corruption i was having. it only happens on drives attached to my qlogic isp card. 2.2 has no problem, and in 2.4.5-ac3 my sym53c875 works fine. this machine is an alpha miata. it only happens when writing out a lot to disk. eg, untarring a kernel tarball, restoring a backup. anyone else see this?

Signature · 2 lines
Tom Vier <[email protected]>
DSA Key id 0x27371A2C
#1 lore raw
Quoted 5 lines · Tom Vier

i narrowed down some corruption i was having. it only happens on drives attached to my qlogic isp card. 2.2 has no problem, and in 2.4.5-ac3 my sym53c875 works fine. this machine is an alpha miata. it only happens when writing out a lot to disk. eg, untarring a kernel tarball, restoring a backup. anyone else see this?

Is this reproducible?

A patch went into ac3 which attempts to address qlogicisp problems on Alpha. It originated from the nether regions deep inside Compaq, original source unknown. There is a -extraordinarily slight- possibility this patch (which simply adds "mb()") causes problems. However, it is more likely that the patch is harmless but also useless, not fixing the real problem.

Other questions - is your machine SMP? How much RAM?

Signature · 3 lines
Jeff Garzik      | Disbelief, that's why you fail.
Building 1024    |
MandrakeSoft     |
#2 lore raw
In reply to Tom Vier
Quoted 6 lines · Tom Vier

i narrowed down some corruption i was having. it only happens on drives attached to my qlogic isp card. 2.2 has no problem, and in 2.4.5-ac3 my sym53c875 works fine. this machine is an alpha miata. it only happens when writing out a lot to disk. eg, untarring a kernel tarball, restoring a backup. anyone else see this?

Also, what compiler are you using? Depending on your current compiler, switching to another compiler according to one of the following permutations would be very instructive for us debugging the problem, at least.

	gcc-2.96-RH -> gcc-2.95.3
	gcc-2.95.2 -> gcc-2.95.3
	gcc-2.95.3 -> egcs-1.1.2

Regards,

Jeff

Signature · 3 lines
Jeff Garzik      | Disbelief, that's why you fail.
Building 1024    |
MandrakeSoft     |
author
#3 lore raw
In reply to Tom Vier
Quoted 8 lines · Jeff Garzik
Quoted 5 lines · Tom Vier

i narrowed down some corruption i was having. it only happens on drives attached to my qlogic isp card. 2.2 has no problem, and in 2.4.5-ac3 my sym53c875 works fine. this machine is an alpha miata. it only happens when writing out a lot to disk. eg, untarring a kernel tarball, restoring a backup. anyone else see this?

Is this reproducible?

yes. by restoring a backup from tape. i tried cat /dev/zero > /dev/sdb1, but that wasn't enough to trigger any corruption (i correctly wrote all zeros). actually, now that i think about it, a good chunk of /dev/sdb2 was cached when i read it back, so that wasn't a very good test. i definetly get corruption under 2.4.5-ac4.

<snip>

Quoted 1 line

Other questions - is your machine SMP? How much RAM?

UP, half gig.

Signature · 2 lines
Tom Vier <[email protected]>
DSA Key id 0x27371A2C
author
#4 lore raw
In reply to Tom Vier
Quoted 4 lines · Jeff Garzik

Also, what compiler are you using? Depending on your current compiler, switching to another compiler according to one of the following permutations would be very instructive for us debugging the problem, at least.

debian 2.25.2 20000220

Quoted 3 lines
	gcc-2.96-RH -> gcc-2.95.3
	gcc-2.95.2 -> gcc-2.95.3
	gcc-2.95.3 -> egcs-1.1.2

i'll give another one a try, tomorrow.

Signature · 2 lines
Tom Vier <[email protected]>
DSA Key id 0x27371A2C