[Devel] vmalloc size (trobles with nvidia-glx)

[Vasily Averin]

Hi Alexander,

Alexander GQ Gerasiov wrote:
> Hello there.
> 
> I'm using OpenVZ on my desktop, where I use nvidia-glx. When I tried to
> switch to OpenVZ I met strange problem: Xserver fails to start with
> message
> 
> (EE) NVIDIA(0): Failed to initialize the NVIDIA graphics device PCI:1:0:0.
> (EE) NVIDIA(0):     Please see the COMMON PROBLEMS section in the README for
> (EE) NVIDIA(0):     additional information.
> (EE) NVIDIA(0): Failed to initialize the NVIDIA graphics device!
> 
> After digging a bit, I found that the problem is in the failed vmalloc
> call. nvidia want ~70M of memory and together with openvz they ran out
> of default limit in 128M.

Thank You very much for Your investigation. We have got several reports about
huge vmalloc allocations but did not have access to the customers nodes and were
not able to investigate such cases.

> So the 1st question: is that normal, that OpenVZ without any VE eats
> 45Mb more memory from vmalloc space, that default kernel?
> If that's normal, what for? (Just interesting.)

It is unexpected for me.
OpenVz kernel can use  a bit more vmalloc than ususal kernel -- but 45Mb is too
high usage IMHO.

That is  vmalloc usage from our test node where 19 VEs are running
(rhel5 based kernel 2.6.18-028stab053.10-ent, node with 512 Mb RAM)
[root at vzlin30 ~]# cat /proc/meminfo | grep Vmalloc
VmallocTotal:  3637240 kB
VmallocUsed:     19224 kB
VmallocChunk:  3617844 kB

As you can see only 19 Mb Vmalloc is used.

> And if that's normal, may be you should override default limit in
> OpenVZ patch? For now I used vmalloc=256m in kernel's command line, but
> this is not trivial to found out the problem for normal not skilled
> user.

Could You please tell me a bit more details about Your node/kernel?

thank you,
	Vasily Averin