Pinpointed start of OOM 'stuck on TLB IPI wait'

George (greerga@nidhogg.ham.muohio.edu)
Mon, 5 Apr 1999 02:32:18 -0400 (EDT)


>= 2.2.0-pre7 'stuck on TLB IPI wait'
- 556kB swap used by 2.2.0-pre7 before death, others varied.
<= 2.2.0-pre6 no problem at all
- 127MB swap used in all cases

Dual processor Pentium 233/MMX, Tyan Tomcat IV, configuration used for
testing was more or less the below (due to configuration changes every
patchlevel) for every kernel 2.1.120 - 2.2.5

CONFIG_EXPERIMENTAL=y
CONFIG_M586TSC=y
CONFIG_X86_WP_WORKS_OK=y
CONFIG_X86_INVLPG=y
CONFIG_X86_BSWAP=y
CONFIG_X86_POPAD_OK=y
CONFIG_X86_TSC=y
CONFIG_SMP=y
CONFIG_NET=y
CONFIG_PCI=y
CONFIG_PCI_GOANY=y
CONFIG_PCI_BIOS=y
CONFIG_PCI_DIRECT=y
CONFIG_PCI_QUIRKS=y
CONFIG_PCI_OLD_PROC=y
CONFIG_SYSVIPC=y
CONFIG_SYSCTL=y
CONFIG_BINFMT_ELF=y
CONFIG_BLK_DEV_IDE=y
CONFIG_BLK_DEV_IDEDISK=y
CONFIG_PARIDE_PARPORT=y
CONFIG_PACKET=y
CONFIG_UNIX=y
CONFIG_INET=y
CONFIG_IP_NOSR=y
CONFIG_SKB_LARGE=y
CONFIG_VT=y
CONFIG_VT_CONSOLE=y
CONFIG_SERIAL=y
CONFIG_PROC_FS=y
CONFIG_EXT2_FS=y
CONFIG_VGA_CONSOLE=y
CONFIG_VIDEO_SELECT=y
CONFIG_MAGIC_SYSRQ=y

Testing procedure:

1) Boot init=/bin/sh

2) Run start-up script below as rot:

/sbin/e2fsck /dev/hdc1
mount -o remount,rw /
rm -f /tmp/.X0-lock
mount -o ro /dev/hdc4 /usr
mount /proc
source /etc/profile
swapon -a
ifconfig lo 127.0.0.1
route add -host 127.0.0.1 lo
/usr/local/bin/xfsft-1.0.1-glibc &

3) Log in as normal user, boot up X, run 4 copies of multi-threaded GIMP
1.1.4 on a 778kB JPEG file.

4) Watch on console with 'top' until swap is filled or machine dies with
TLB IPI errors.

I'm not kernel-proficient enough to see what broke it in that patchlevel
(though there were VM and spinlock changes) but I can try out any patches
or changes people send.

-George Greer

-
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/