I had symptoms like this, and that may be part of what you're
remembering.
It turned out that it had nothing to do with "xntpd" or kernel time
interfaces; the "fdomain.c" SCSI driver had huge (300ms+) latencies.
When the SCSI drivers were changed, en masse, to be SMP-friendly by
putting a coarse "io_request_lock" around every SCSI interrupt
handler, "fdomain.c" needed a patch to fine-grain those locks.
My patch for that particular driver went in at 2.1.111 or thereabouts,
but I don't think anyone looked through the rest of the SCSI drivers
systematically. Certainly the drivers where the problem was most
obvious have been patched, but perhaps there are a few SCSI drivers
that lose a few timer ticks here and there, just enough so that only
people using "xntpd" notice the slippage.
I guess this is just a heads-up for people observing time
synchronization problems: if your time is periodically stepped forward
by relatively big (i.e., multisecond) chunks at seemingly random
times, suspect a SCSI (or other) driver dropping interrupts.
Kevin <buhr@stat.wisc.edu>
-
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/