* [resend] ioapic+shared interrupt+concurrent access=ide lockup
@ 2004-02-26 15:09 Andreas Steinmetz
0 siblings, 0 replies; 2+ messages in thread
From: Andreas Steinmetz @ 2004-02-26 15:09 UTC (permalink / raw)
To: linux-kernel, mingo, discuss
I did send a similar message about 3 weeks ago to lkml but there was no
reply. The following is related to 2.6.2 (x86-64) though 2.4 shows
similar symptoms.
Summary:
--------
The ide subsystem locks up on a dual Opteron system when io-apic is
enabled and there are concurrent accesses to both ide channels of a pci
ide controller card. The lockup occurs on different pci ide controller
cards and there are boot parameters that can be used as workarounds.
Details are given below. If additional information is required please
contact me. I can currently run any tests required to fix this problem.
Symptoms:
---------
Execute: "dd if=/dev/hde of=/dev/null & ; dd if=/dev/hdg of=/dev/null &"
This results after a short time in:
hde: dma_timer_expiry: dma status == 0x24
hdg: dma_timer_expiry: dma status == 0x24
hde: DMA interrupt recovery
hde: lost interrupt
hdg: DMA interrupt recovery
hdg: lost interrupt
hde: dma_timer_expiry: dma status == 0x24
hdg: dma_timer_expiry: dma status == 0x24
hde: DMA interrupt recovery
hde: lost interrupt
hdg: DMA interrupt recovery
hdg: lost interrupt
...
Further accesses to hde and hdg always result then in the same messages
as stated above.
The MIS count of /proc/interrupts steadily increases prior to the
lockup. It always has the same value as the ide2/ide3 interrupt count on
cpu1. Note that no irq balancing is active so there should be no
interrupts on cpu1.
During the tests causing the lockup the system is otherwise idle.
Workarounds:
------------
In the cases listed below the lockup does not happen and the MIS count
of /proc/interrupts stays at zero.
1. Boot with "noapic"
2. Boot with "ide2=serialize"
3. Boot with "ide3=serialize"
System:
-------
Tyan S2885 dual Opteron 246
IDE config:
-----------
ide0/hda WD2500JB
ide1/hdc WD2500JB
ide2/hde WD2500JB
ide3/hdg DVD-RW
ide2/ide3 are were tested with the following cards: HPT302, CMD680 and
PDC20268 (the latter being now the one in the system). All cards show
the same behaviour so it is probably not a driver problem.
It doesn't matter if hdg is a DVD-RW or a hard disk, the problem is the
same.
--
Andreas Steinmetz
^ permalink raw reply [flat|nested] 2+ messages in thread
* Re: [resend] ioapic+shared interrupt+concurrent access=ide lockup
@ 2004-02-26 19:41 David Whysong
0 siblings, 0 replies; 2+ messages in thread
From: David Whysong @ 2004-02-26 19:41 UTC (permalink / raw)
To: linux-kernel, discuss
Andreas Steinmetz wrote:
> The ide subsystem locks up on a dual Opteron system when io-apic is
> enabled and there are concurrent accesses to both ide channels of a pci
> ide controller card.
[...]
I'll confirm this, I had the same problem with a Promise 20269 controller,
Tyan S2885 motherboard, dual Opteron 240.
--
David Whysong dwhysong@physics.ucsb.edu
Astrophysics graduate student University of California, Santa Barbara
My public PGP keys are on my web page - http://www.physics.ucsb.edu/~dwhysong
DSS PGP Key 0x903F5BD6 : FE78 91FE 4508 106F 7C88 1706 B792 6995 903F 5BD6
D-H PGP key 0x5DAB0F91 : BC33 0F36 FCCD E72C 441F 663A 72ED 7FB7 5DAB 0F91
^ permalink raw reply [flat|nested] 2+ messages in thread
end of thread, other threads:[~2004-02-26 19:43 UTC | newest]
Thread overview: 2+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2004-02-26 15:09 [resend] ioapic+shared interrupt+concurrent access=ide lockup Andreas Steinmetz
-- strict thread matches above, loose matches on Subject: below --
2004-02-26 19:41 David Whysong
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox