Messages in this thread | | | Date | Tue, 16 Sep 1997 22:20:44 +0200 | From | Andrej Presern <> | Subject | 2.0.31-pre9 + QNX + NoStackExec weirdness |
| |
Ok, here's the deal.
I'm experiencing some weird things using the 2.0.31-pre9 together with a QNX scheduler (1.09), the nonexecutable stack patch, the delay fix and the cyrix patch.
When the screen blanker goes on and is on for some time, I can't login to the system anymore and I can't even shut down the system (hard reset does it:/). The first thing I notice when the screenblanker pops off is that the load is absolutely too high (1.5 even though NOTHING is running - the screen blanker just turns the screen off, so I doubt that's the reason for the load), then I notice that diald's dctrl is in run queue 19 (? and that's ) and is locked (no activity.. as if it was starved?) yet top and hifs say the system is 95% idle..
The loadavg then slowly goes back to normal, but the dctrl is still not responding to anything, the diald is not working and I can't login anymore (except through the xdm which makes the thing even stranger).
Now you call this weird? I call this 'the report #1'... Let's go on to #2 :)
I saw there was a noexec stack patch on the 'mama' homepage, downloaded it and patched the kernel. Now the things get weirder - X server (and only the X server) starts LEAKING memory like hell (that's like nothing ever gets freed - the thing is 70 megs and rising in only a few minutes if I make a mistake and start netscape).
I got the first noexec patch from linuxhq when it was still up and running and I asumed the two should be more or less identical except that the one that I got from the 'mama' homepage was made explicitly for the pre9 while the first one was for 2.0.30 (i think). All the patches went in cleanly.
Any suggestions? I think I can reproduce this at will.
Andrej
PS: I just checked the dctrl now (after a fresh boot): [root@soda /usr/src]# qsched 323 pid: 323 policy: 0 (Adaptive) timeslice: 200ms fork_thresh: 15s strv_thresh: 1000ms ts_max: 1 run_q: 19 run_q_min: 19 run_q_max: 31
Hmm.. so I guess it IS that cpu intensive? Let's see the load..
[root@soda /usr/src]# uptime 10:10pm up 24 min, 2 users, load average: 0.08, 0.02, 0.06
Then again.. I guess not.. I wouldn't mind it droping into run queue 19 due to high cpu usage but what in the hell got it to MIN run queue 19???
PPS: One more thing ('the report #3').. to get the pid of dctrl I did a 'ps -aux | grep dctrl' and got this: [root@soda /usr/src]# ps -aux | grep dctrl andrejp 323 0.2 2.8 2768 1796 ? S N 21:49 0:03 wish /usr/bin/dctrl -
Now what I found to be weird here is that when I did this exact same command before the reboot (when I couldn't login), I got the 'grep dctrl' command in the listing too.. why is that? One time I get both commands and the other I get just one? What's the trick here?
| |