Date: Sat, 26 Jul 1997 10:43:57 -0500 (CDT) From: dave adkins <adkin003@gold.tc.umn.edu> To: smp@FreeBSD.ORG Cc: David A Adkins <adkin003@gold.tc.umn.edu> Subject: Re: ide and SMP current (07/26/97) Message-ID: <Pine.NEB.3.96.970726104109.5589A-100000@samthedog.gold.tc.umn.edu> In-Reply-To: <Pine.NEB.3.96.970726101129.5506A-100000@samthedog.gold.tc.umn.edu>
next in thread | previous in thread | raw e-mail | index | archive | help
Hi, Looks like some minor problems with wd when used with current as of 07/26/97 (apic_vector.s v1.11). I'm configured with / on an IDE drive and all other partions on a NCR53c810. When the IDE controller is configured 0x80008000 I see the following: Jul 26 11:18:50 samthedog /kernel.test: wd2: interrupt timeout: Jul 26 11:18:55 samthedog /kernel.test: wd2: status 58<rdy,seekdone,drq> error 0 Jul 26 11:19:04 samthedog /kernel.test: wd2: interrupt timeout: Jul 26 11:19:04 samthedog /kernel.test: wd2: status 58<rdy,seekdone,drq> error 1<no_dam> The system continues to run but stalls periodically on wd2 accesses. During the stalls systat shows the interrupt load go to nearly 50%. I assume thats because the system is stalled in the wd driver. If I reconfigure to use multi-sector transfers (0x80ff80ff) then I can run normally with very infrequent timeout messages. The problem still exists with multi-sector transfers enabled, its just less frequent. Possibly a race in the driver? I've got drives on wcd0 and wcd1 but I'm only mounting wd2. When I mount the drives on wcd0 they have the same problem in single-sector transfer mode. I originally switched to single-sector in an attempt to minimize silo overflows and never changed it back. The motherboard is a tyan S1563D. There were no timeouts with apic_vector.s V1.10. dave
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?Pine.NEB.3.96.970726104109.5589A-100000>