[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Reliability of Linux kernel on sun4d



This is more for the record than anything else, although if anybody could give
me any pointers I'd do my best to put in more work subject to having limited
time.

I'm afraid that 2.4.32 suffers from random lockups on a SPARCserver 1000E, LEDs
show that either CPU0 or CPU1 is locked up and there's no response to the serial
console or to network activity. Sometimes it runs for hours, sometimes it locks
up during startup- no obvious pattern.

Out of curiosity I've enabled IRQ distribution and I believe I have seen SCSI
problems, so I suspect that Jakub's warning still stands. Apart from that
reliability and performance was similar to the case where all IRQs were going to
CPU0.

2.2 can be made to work on (at least some of) these machines either by disabling
speed calibration for CPUs other than the first or by bringing SBUS
initialisation forward. This kernel appears to be reliable with IRQ distribution
(although possibly slightly slower) so I suspect that any problems are being
masked by the Big Kernel Lock.

It looks as though my SPARCserver is going to stay as Woody+2.2 at which it's
rock-solid, although I might try 2.6 once I've got the source on something.

-- 
Mark Morgan Lloyd
markMLl .AT. telemetry.co .DOT. uk

[Opinions above are the author's, not those of his employers or colleagues]



Reply to: