64GB option broken in 2.4.2

From: Rico Tudor (rico@patrec.com)
Date: Sun Feb 25 2001 - 20:21:12 EST


Subject: 64GB option broken in 2.4.2

Problem:
        The 64GB option causes processes to spin in the kernel.

Setup:
        Linux 2.4.2
        Slack current 20-feb-2001
        ServerWorks III HE
        4GB main memory
        dual 933 Piii: SL47Q (stepping 3), SL4KK (stepping 6)
        gcc 2.95.2 19991024 (release)

Test procedure:
        On /dev/tty1, type

                cat /dev/sda >/dev/null # or any fast disk

        On /dev/tty2, type

                while sleep 0; do date; done

        On /dev/tty3, type

                top

        The purpose of the disk read is to provide some nondeterminism,
        which triggers the bug. `top' lets you watch the debacle.

Observation:
        After a short time, dates will appear with increasing
        irregularity. `top' will show `sleep' and `date' processes
        racking up many seconds of CPU time. At this point, terminate
        the disk read, and one of these processes will go completely
        CPU-bound, and uninterruptible. In this state, the system is
        unusable, since any exec'd process has a good chance of being
        bushwhacked in the kernel.

Variations:
        noapic problem remains
        mem=1200m problem remains
        nosmp noapic problem fixed
        mem=800m problem fixed
        4GB option problem fixed

Hypothesis:
        Code to handle PAE has buggy spinlock management.
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/



This archive was generated by hypermail 2b29 : Wed Feb 28 2001 - 21:00:10 EST