I've seen such thing happen occasionally on i386 box too, so far
sync(8) always made the system to get out of this in a while. I've
attributed this to softdeps.

Jaromir

Manuel Bouyer wrote:
> Hi,
> for a long time (since about end of july) my sparc IPX dies under heavy I/O
> load (such as a sup -o). I/O seems blocked and the disk keep trashing.
> I suspect that the same request is sent to the disk again and again,
> maybe the VM trying to flush the same dirty page(s) to disk.
> Running processes keep running until they try a disk access.
> This is not the usual out of swap problem, there is no big process and plenty
> of free swap when this happens.
> 
> This machine has 2 disks:
> scsibus0: waiting 2 seconds for devices to settle...
> sd1 at scsibus0 target 1 lun 0: <IMPRIMIS, 94601-15, 4614> SCSI1 0/direct fixed
> sd1: 1001 MB, 1931 cyl, 15 head, 70 sec, 512 bytes/sect x 2050995 sectors
> sd1: async, 8-bit transfers
> sd0 at scsibus0 target 3 lun 0: <CONNER, CP30200  SUN0207, 4544> SCSI2 0/direct fixed
> sd0: 203 MB, 2123 cyl, 4 head, 49 sec, 512 bytes/sect x 416108 sectors
> sd0: sync (200.0ns offset 15), 8-bit (5.000MB/s) transfers
> 
> System is on sd0, swap and sources on sd1. sd1 is really slow, maybe this is
> what triggers the problem (race condition somewhere ?).
> 
> Anyone else seen this ?
> 
> --
> Manuel Bouyer, LIP6, Universite Paris VI.           Manuel.Bouyer@lip6.fr
> --
> 


-- 
Jaromir Dolecek <jdolecek@NetBSD.org> http://www.NetBSD.org/Ports/i386/ps2.html
-=  Those who would give up liberty for a little temporary safety deserve  =-
-=  neither liberty nor safety, and will lose both.  -- Benjamin Franklin  =-