Re: kmem leak in tmpmfs?

Iasen Kostov <tbyte@xxxxxxxx> wrote:

On Thu, 2006-05-25 at 16:54 -0400, Kris Kennaway wrote:
On Thu, May 25, 2006 at 06:01:30PM +0200, Arno J. Klaassen wrote:

I get a very easy to reproduce panic on 6.1-STABLE :

/etc/periodic/weekly/310.locate panics with

panic: kmem_malloc(4096): kmem_map too small: 335544320 total
allocated

It looks like you are using a malloc-backed md and you do not have
enough RAM to handle the size. Perhaps tmpmfs does not use swap
backing, as it is supposed to?

First of all if there is not enough kmem (not just plain ram
I think) kernel should not allow disk creation in first place, second
- I think (although there could be some ... reason for that) it's
stupid way to say "I don't have more kmem" by panicing :). Better way
will be just to fail disk operation of that FS with "Disk is full" or
something like that. At home I tried to raise kmem like that:
"vm.kmem_size_max="1073741824" (I got 2G of RAM)
(setting vm.kmem_size directly panices kernel at boot if I remember
correctly).

but for my surprise kernel panices at exact same allocated md disk
space with the same panic as the original poster's. Is it possible
that I should rise KVA_PAGES too ? And I don't think its documented
anywhere (of course I've tried googling and it's always possible that
I've missed something :). All this was on FreeBSD 6.0.

man mdconfig mentions the problem:

malloc Storage for this type of memory disk is allocated with
malloc(9). This limits the size to the malloc bucket
limit in the kernel. If the -o reserve option is not
set, creating and filling a large malloc-backed memory
disk is a very easy way to panic a system.

Use a swap backed disk and the problem will disappear.

Fabian
--
http://www.fabiankeil.de/

Attachment: signature.asc
Description: PGP signature

Relevant Pages

  • Re: kmem leak in tmpmfs?
    ... I think) kernel should not allow disk creation in first place, ... This limits the size to the malloc bucket ... Use a swap backed disk and the problem will disappear. ...
    (freebsd-stable)
  • Re: FreeBSD 5.2R+C: panic: kmem_malloc(4096): kmem_map too small
    ... small allocations. ... The kernel has a rather small malloc pool (this is ... is much more appropriate to use the 'swap' disk type of mdconfig. ... panics. ...
    (freebsd-current)
  • Re: malloc backed md/mfs filesystem swapped?
    ... > of type MD_MALLOC is kernel memory which will not be swapped, ... the relevant fstab entries for a malloc backed disk having a UFS2 ...
    (freebsd-hackers)
  • Re: 6.0 Witness squawk
    ... > There are three entry points into the driver. ... The disk ... AFAIK, if the code that calls malloc() can be reached from 2) or 3), ...
    (freebsd-arch)
  • Re: 6.0 Witness squawk
    ... > There are three entry points into the driver. ... The disk ... AFAIK, if the code that calls malloc() can be reached from 2) or 3), ...
    (freebsd-current)