[9198] in linux-scsi channel archive
Re: Feature in ncr53c8xx/sym53c8xx driver
daemon@ATHENA.MIT.EDU (=?ISO-8859-1?Q?G=E9rard_Roudier?=)
Fri Jul 14 08:24:24 2000
Date: Fri, 14 Jul 2000 14:01:52 +0200 (CEST)
From: =?ISO-8859-1?Q?G=E9rard_Roudier?= <groudier@club-internet.fr>
To: Miquel van Smoorenburg <miquels@cistron.nl>
cc: linux-scsi@vger.rutgers.edu
In-Reply-To: <8kmsfi$8t6$1@enterprise.cistron.net>
Message-ID: <Pine.LNX.4.10.10007141350520.1227-100000@linux.local>
MIME-Version: 1.0
Content-Type: TEXT/PLAIN; charset=ISO-8859-1
Content-Transfer-Encoding: 8BIT
PCI interrupt lines are REQUIRED be configured for LEVEL SENSITIVE.
(level-triggerred is sometimes used in lieu of level-sensitive)
This applies to ALL PCI devices that support interrupts.
You IRQ configuration for the SYMBIOS devices is _seriously_ broken.
A system that works more than a couple of seconds with PCI interrupt
misconfigured to EDGE is kind of miracle that always surprises me.
Here are the things that are to be fixed.
> 10: 41954 47456 IO-APIC-edge sym53c8xx
> 15: 1684 3355 IO-APIC-edge sym53c8xx
Gérard.
On 14 Jul 2000, Miquel van Smoorenburg wrote:
> In article <cistron.Pine.LNX.4.10.10007132135290.2177-100000@linux.local>,
> =?ISO-8859-1?Q?G=E9rard_Roudier?= <groudier@club-internet.fr> wrote:
> >Could you please update your sym53c8xx driver version and let me know if
> >it makes differences.
>
> I've updated to your latest driver, but I still have the same problem.
> I have managed to create a test suite that locks up the system
> within one minute of disk activity. What happens is that the
> sym chipset stops producing interrupts.
>
> Short description of test setup:
>
> The machine has 5 disks. One the first disk is the OS, on the
> four other disks I have created one raid0 partition of 6 GB,
> the rest of the disks consist of empty partitions (that can be
> used by INN directly). I start a read 4K block - lseek -4096 - write
> sequence on each disk on the empty partitions and run a bonnie
> on the raid0 ext2 filesystem - lockup.
>
> The only difference is that with the new driver, it seems that it
> also polls the chipset every so often through a timer, so throughput
> gets dead slow (disk leds blink 1 time/sec) but will keep on "working"-
> though sync()ing the cache on this 1 GB RAM machine takes forever.
>
> Value of /proc/interrupts:
>
> % cat /proc/interrupts
> CPU0 CPU1
> 0: 129420 141962 IO-APIC-edge timer
> 1: 826 950 IO-APIC-edge keyboard
> 2: 0 0 XT-PIC cascade
> 4: 7 0 IO-APIC-edge serial
> 8: 1 0 IO-APIC-edge rtc
> 10: 41954 47456 IO-APIC-edge sym53c8xx
> 13: 1 0 XT-PIC fpu
> 15: 1684 3355 IO-APIC-edge sym53c8xx
> 19: 595 611 IO-APIC-level eth0
> NMI: 0
> ERR: 0
>
> Interrupt 10 and 15 don't change anymore.
>
> System is AMI Megarum II, 2xPIII/450, 1GB RAM, onboard dual 53c896.
>
> Mike.
-
To unsubscribe from this list: send the line "unsubscribe linux-scsi" in
the body of a message to majordomo@vger.rutgers.edu
