Recipe:
- reboot into 2.2.0-pre8 or pre9.
- log into two VC's
- in one vc run 'strace vmstat 1' (or vmstat 1 if you don't want to see
where it's hanging)
- in the other run 11 or 12 copies of a program that grabs 10MB (on a
96MB system :-) and each one random accesses its 10mb space
continuously.
- switch back to the other VC to see vmstat hang for minutes reading
/proc/NNN/stat
NOTE: a single process using 100MB (on my 96MB system) does not trigger
this!!!
Additionally, ps and killall also hang in D state reading from proc.
The system will eventually perform the intended action... but not without
horrendous delays.
After killing all of the memory hogs, the system works fine again. In
fact, the swap sub-system performs great as far as the memory hogs are
concerned. Other actions are not affected during the test (i.e. logging
into a new vc and running ls -lR performs as expected, slow but not
pathologically so).
Doing this on kernels before pre8 works fine. In fact, I have been
running such a test since 2.1.115 and the MM is flying with respect to
swap perf. and interactivity while the test is running, except for this
new wrinkle.
David
-- /==============================\ | David Mansfield | | david@cobite.com | \==============================/
- 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/