Re: memory leak detector, 2.0.32, patch

[Chris Adams]

Once upon a time, Chris Adams wrote
> Well, this patch slowed down my system a lot.  100% in fact.  I got this
> during boot:
> 
> ---
> Memory: 281056k/286720k available (540k kernel code, 384k reserved, 4740k data)
> Bad pmd in pte_alloc: 120000e7
> Kernel panic: Failed to allocate buffer hash table
> 
> In swapper task - not syncing
> ---
> 
> and then it stopped.
> 
> This is a server with 320M of RAM (have to pass mem=320M from LILO).

To follow up to my own message, I did some more experimenting.  I found
out the above crash is sensitive to the amount of memory available.  If
I booted with mem=64M, I could boot up okay.  If I booted with mem=128M
the system crashed similar to above (the Memory: line was different and
the Bad pmd in pte_alloc: number was different).

I poked around a bit (I'm not exactly an experienced kernel hacker
obviously) and this error message comes from buffer_init in fs/buffer.c,
which does a vmalloc to get its memory.  I notice that the memleak patch
doesn't handle vmalloc; could it be interfering with it somehow in a way
that only shows up with > 64M RAM?
-- 
Chris Adams - [e-mail]
System Administrator - Renaissance Internet Services
I don't speak for anybody but myself - that's enough trouble.