Messages in this thread | | | Date | Sat, 31 Jul 1999 21:26:53 -0700 (PDT) | From | Robert Dinse <> | Subject | Re: linux-kernel-digest V1 #4229 |
| |
On Sat, 31 Jul 1999 owner-linux-kernel-digest@vger.rutgers.edu wrote: > > From: "Robert de Bath" <rd103978@home-box.demon.co.uk> > Date: Sat, 31 Jul 1999 12:59:15 +0100 (BST) > Subject: Re: 2.2.9+ extreme instability > > The 2.2.X kernels are in general stable, this definitly sounds like a > hardware problem ... tell me has it been nice and sunny recently ? > > - -- > Rob. (Robert de Bath <http://poboxes.com/rdebath>) > <rdebath @ poboxes.com> <http://www.cix.co.uk/~mayday>
Oh how I wish this were true!
Unfortunately, 2.2.x is NOT stable on Sparc multi-CPU platforms, at least not on 32-bit hardware platforms.
I have 2.2.10 running on two SS-10's with quad Ross Hypersparc RTK-625 CPU's, one has 90mhz the other 100mhz CPU's, and I still have the spin_lock problems on a frequent basis (average about once every other day). I had this same problem on a Sun 4/670MP with dual Hypersparc. This problem has existing in some of the 2.1.x kernels and every 2.2.x kernel since 2.2.0. It has gotten better, 2.2.0 would lock with scrolling spin_lock messages sometimes before it even made it all the way up into multi-user mode, whereas 2.2.10 on the average takes a couple of days, but it is STILL bad.
One machine is a news server running INN 1.7.x, the other is a web server running Apache 1.3.3p3.
2.0.35 was completely stable on both machines, except that 2.0.35 would frequently hang during fsck's.
Why don't I go back to 2.0.35? Because 2.2.10 is a MASSIVE improvement in performance, both machines were nearly maxed on CPU before, now they're averaging around 10% CPU occupancy. News could hardly keep up before, now it's mostly idle. In short the performance is a big win but this spin_lock evil bug is really a major frustration.
- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.rutgers.edu Please read the FAQ at http://www.tux.org/lkml/
| |