limit on locked memory
Manuel Bouyer
I'm trying to run memtest (pkgsrc/sysutils/memtester) on a box with
1GB RAM (i386, if that matters). memtest locks the memory it will work on,
but it seems that it's not possible to have more than about 332MB ram locked.
memtest will find out how much memory it can lock:
Allocated 352321536 bytes...trying mlock...failed: insufficient resources.
Allocated 351272960 bytes...trying mlock...failed: insufficient resources.
Allocated 350224384 bytes...trying mlock...failed: insufficient resources.
Allocated 349175808 bytes...trying mlock...failed: insufficient resources.
Allocated 348127232 bytes...trying mlock...success. Starting tests...
starting a second instance of memtest it can lock a very small amout of
ram:
Allocated 950272 bytes...trying mlock...failed: insufficient resources.
Allocated 933888 bytes...trying mlock...failed: insufficient resources.
Allocated 917504 bytes...trying mlock...failed: insufficient resources.
Allocated 901120 bytes...trying mlock...failed: insufficient resources.
Allocated 884736 bytes...trying mlock...success. Starting tests...
A third instance can't lock anything:
Allocated 6144 bytes...trying mlock...failed: insufficient resources.
Allocated 5120 bytes...trying mlock...failed: insufficient resources.
Allocated 4096 bytes...trying mlock...failed: insufficient resources.
Allocated 3072 bytes...trying mlock...failed to align to pagesize of 4096 bytes.
Failed to get sufficient memory. Exiting...
of course I raised all limits to the max, memorylocked and memoryuse are
unlimited, datasize is at 1GB.
Any idea of what is the limiting factor and if it's easy to raise ?
With this limit, I can only test 1/3 of the available RAM ...
--
Manuel Bouyer <bou...@antioche.eu.org>
NetBSD: 26 ans d'experience feront toujours la difference
--
Jason Thorpe
On Jul 29, 2004, at 9:42 AM, Manuel Bouyer wrote:
> of course I raised all limits to the max, memorylocked and memoryuse
> are
> unlimited, datasize is at 1GB.
> Any idea of what is the limiting factor and if it's easy to raise ?
> With this limit, I can only test 1/3 of the available RAM ...
I think there's a system-wide limit that applies, also. Not sure how
you change it.
-- Jason R. Thorpe <tho...@wasabisystems.com>
Eric Haszlakiewicz
> On Jul 29, 2004, at 9:42 AM, Manuel Bouyer wrote:
>
> >of course I raised all limits to the max, memorylocked and memoryuse
> >are
> >unlimited, datasize is at 1GB.
> >Any idea of what is the limiting factor and if it's easy to raise ?
> >With this limit, I can only test 1/3 of the available RAM ...
>
> I think there's a system-wide limit that applies, also. Not sure how
> you change it.
It doesn't look like it's dynamically adjustable, but it wouldn't be too hard
to add to go with the other vm.*max sysctls. (uvm_meter.c)
eric
Manuel Bouyer
> uvm_pdaemon.c: uvmexp.wiredmax = uvmexp.npages / 3;
> It doesn't look like it's dynamically adjustable, but it wouldn't be too hard
> to add to go with the other vm.*max sysctls. (uvm_meter.c)
Yes ! changed this to
uvmexp.wiredmax = uvmexp.npages - 32768, and now:
Memory: 12M Act, 872M Wired, 3712K Exec, 3184K File, 89M Free
Swap: 1201M Total, 1201M Free
PID USERNAME PRI NICE SIZE RES STATE TIME WCPU CPU COMMAND
229 root 64 0 128M 129M RUN/0 2:39 29.98% 29.98% memtest
628 root 64 0 128M 129M RUN/0 2:35 29.30% 29.30% memtest
524 root 64 0 128M 129M RUN/1 2:36 29.00% 29.00% memtest
227 root 64 0 128M 129M RUN/1 2:36 28.56% 28.56% memtest
331 root 64 0 128M 129M CPU/1 2:26 26.86% 26.86% memtest
694 root 64 0 128M 129M RUN/0 2:29 26.76% 26.76% memtest
533 root 64 0 105M 106M RUN/0 2:31 25.73% 25.73% memtest
(I could probably run a single memtest with 872M, but I also want to exercise
the data path to both CPUs :)
thanks !