Re: AW: NON IRQ DEADLOCK in 2.0.31

Rob Hagopian (hagopiar@vuser.vu.union.edu)
Fri, 14 Nov 1997 14:23:50 -0500 (EST)


On Fri, 14 Nov 1997, Palme, Sven wrote:

> it seems we discovered the same ...

Very similar...

> my SMP Box (2.0.29) runs for a long time,
> but after replacing the 2x133MHz CPU´s with 2x200MHz

I presume those are Pentiums then. I've got PIIs (300s running at 337 on a
75Mhz bus).

> the same Kernel crashed ...
> it was a respond to ping and telnet - but no
> connection will be established ...

Same (for the second crash. the first one was a DEADLOCK, no pings, no
telnet, no nothing).

> and all this happend while a backup via NFS was done!!!!
>
> switching to 2.0.31 was terrible since it locked after 5
> hours of running ...

I'm running 2.0.31...

> BTW : the machine is a heavy loaded WEB-Server
> loadavg 2....20 more than 200 Procs in the list (and more)

Yes, we've peaked at over 270 httpd processes, but this was off hours, I'd
suspect around 50-75 were running at the time of the second crash.

> but NO special hardware inside (EIDE with TYAN-Board)
> and no smp-kernel of the 2.0.xx serie will work without a lookup
> :-(
> hope the >2.1.64 will help

I have the SuperMicro DLE (LX chipset) with 128M SDRAM... The first crash
came during a backup of an IDE drive, the second crash came during the
backup of a RAID-0 MD partition (2 quantum ST SCSI-II drives on an
NCR825a).

I'm also wondering what's in 2.0.32 besides the f00f patch... anyone?
-Rob H.