You are not logged in.
Pages: 1
I just installed the latest MM kernel, and discovered that, after half an hour of uptime or so, the CPU usage jumps to 100% and refuses to come back down. Listing the running processes shows me that none of them are eating up that much CPU time.
(Yes, I know that the MM kernel isn't stable. I'm just pointing out what I'm guessing is a bug.)
Offline
point it out on the mailing list not here.
Offline
the cause of this 100% is kswapd0 that do not go idle but stays running. i already posted this to the kernel ML and the devs are investigating the issue. funny, but enabling some minimal debug-options, it seems to solve this issue (i will upload -2 as soon as it is ready solving this mystery)
The impossible missions are the only ones which succeed.
Offline
see also
http://www.ussg.iu.edu/hypermail/linux/ … /0043.html
The impossible missions are the only ones which succeed.
Offline
i was wrong: it does not solve this issue, but it makes it appear less frequent (no idea why)
i'm uploading a -2 release of the kernel26mm package that has magic keys enabled.
if you see kswapd0 running without idling,press [ALT]-[SysRq]-[p] (SysRq is same as PrtScrn (print screen button) - maybe you will need to go to VC first, if you WM catches such combinations) and then do dmesg and see the last part that was added. if it is different to the one i get here:
http://www.ussg.iu.edu/hypermail/linux/ … /0233.html
feel free to report to me or directly to the kernel ML to the thread about it to help the devs find out what's happening.
The impossible missions are the only ones which succeed.
Offline
update:
it's cachefs that is causing kswapd0 this sleepless nights .... therefore i disabled cachefs in rc3-mm3-2 and since then didn't get this issue
The impossible missions are the only ones which succeed.
Offline
What's this cachefs? Is this your standard HD caching in RAM, or is it a new feature?
Offline
the idea of cachefs is old:
http://www.starlink.rl.ac.uk/star/docs/ … node9.html
but i don't know how long it is in linux enabled (i only know that i almost always use it because it makes dirsync much faster for NFS shares ;-) )
The impossible missions are the only ones which succeed.
Offline
the cause of this 100% is kswapd0 that do not go idle but stays running. i already posted this to the kernel ML and the devs are investigating the issue. funny, but enabling some minimal debug-options, it seems to solve this issue (i will upload -2 as soon as it is ready solving this mystery)
I had this problem too, but it is fixed in the newer rc's.
Offline
dp wrote:the cause of this 100% is kswapd0 that do not go idle but stays running. i already posted this to the kernel ML and the devs are investigating the issue. funny, but enabling some minimal debug-options, it seems to solve this issue (i will upload -2 as soon as it is ready solving this mystery)
I had this problem too, but it is fixed in the newer rc's.
yes, because cachefs i disabled for now up untill we know for sure it is fixed
The impossible missions are the only ones which succeed.
Offline
What about 2.6.12-rc4-mm2?
Offline
What about 2.6.12-rc4-mm2?
will be up, as soon my cpu is free to compile it (university is more important than arch, sorry) ... if you want to try it, use ABS (my PKGBUILD should work by simply changing rc and mm versions)
if you ask about cachefs and kswapd0: i don't know and as cachefs is not essential to me, i don't want to spend time to know that it is still broken. if someone reports that it is fixed, i can enable cachefs back again ;-)
The impossible missions are the only ones which succeed.
Offline
I don't use the -mm from the arch repo, I roll my own, and can tell you that it is fixed for me with -mm2.
Offline
Woo-hoo! Back to the wild side of computing I go...
Offline
I don't use the -mm from the arch repo, I roll my own, and can tell you that it is fixed for me with -mm2.
good news ... i will check it myself, thanx for info!
The impossible missions are the only ones which succeed.
Offline
Pages: 1