Noticed lately this warnings at syslog, from what i've seen its seems to be a kernel bug. Is this fixable?
Printable View
Noticed lately this warnings at syslog, from what i've seen its seems to be a kernel bug. Is this fixable?
transmission taking up too much network buffer I believe.Quote:
Originally Posted by hggomes
see my post earlier about this: http://wl500g.info/showthread.php?18...544#post251544
also do you set /proc/sys/vm/min_free_kbytes ?
Quote:
echo 8192 > /proc/sys/vm/min_free_kbytes
I don't use Transmission neither other kind of torrent clients ;)
/proc/sys/vm/min_free_kbytes = 1442 so thats maybe why message was showing like you said. I have change it already now, so i will get in touch in a couple of days and confirm it.
Thanks alot.
Yes, it is a problem inside kernel, but:Quote:
Originally Posted by hggomes
- you omit previous messages - dump_stack() usually called when memory exhausted.
- it needs stable reproducible test-case
Anyway - at which build problem observed? Was it seen on earlier builds?
P.S. Please never post screenshots whenever is is possible to provide plain text info.
Well, i believe this was always happening in previous builds also because i never analyse my syslog in real time but now im doing it.
The main reason why i have posted a screenshot instead of plain text it's because i was unable to copy it directly from the application, today i found a way to export it.
I think that's related with Squid usage somehow, but i'm not sure.
PS: That's exactly the problem, it's NOT POSSIBLE to paste in here more than 10K caracters, kernel log has 73K, i hope you can understand that.
Thanks
Root of this problem is in following line:
Thus, wpte guess is right - some application eaten memory.Code:swapper:*page*allocation*failure.*order:3,*mode:0x4020*
Yup i found it already, that was Squid Proxy.
aug*28*00:31:43 32768*pages*of*ram*
aug*28*00:31:43 kernel:*0*pages*of*highmem*
aug*28*00:31:43 kernel:*1067*reserved*pages*
aug*28*00:31:43 kernel:*18945*pages*shared*
aug*28*00:31:43 kernel:*16*pages*swap*cached*
aug*28*00:31:43 kernel:*swapper:*page*allocation*failure.*order:3, *mode:0x4020
Thanks for reply.
Maybe it's an idea to increase min_free_kbytes a bit by default?Quote:
Originally Posted by lly
I mean, this error is probably only because the kernel wants to write so memory to swap, but is unable to because not enough ram is left?
On my PC the value is 67MB, obviously a bit overkill for a router, but 8MB should be do-able for the modern routers.
I guess you could run short in memory for older models without swap tho.
Generally, your idea is right. But it's value must vary on router model, and has to be good approximation.Quote:
Originally Posted by wpte
Additionally, I have some untested backports of Lumpy reclaim from upstream - it should improve swapping. Unfortunately, I hasn't enough time to test it thoroughly...
After this "echo 8192 > /proc/sys/vm/min_free_kbytes" i have never seen it again, i think it solved the problem.
Btw router model is RT-N16, probably the default value on a better model (ex: RT-N66U) could be higher ah?
Thank you both.