On Fri, Sep 25, 2009 at 06:06:35PM +0100, Martin Michlmayr wrote: > * Steve Langasek <vorlon@debian.org> [2009-09-22 23:28]: > > After building myself a 2.6.30 kernel with the iop dma patches for my Thecus, > > I started seeing reproducible kernel oopses on large NFS transfers, such as > > the one included below. After some prodding at the source for other uses of > > down_read(), I concluded that we're not supposed to call > > down_read(¤t->mm->mmap_sem) if in_atomic() is true. Updating the dma1 > > patch from http://people.debian.org/~tbm/dma/dma-patch to the attached > > appears to have fixed the problem for me, giving me a stable DMA-enabled > > squeeze kernel. > Thanks a lot for investigating and fixing this issue! I'm building an > updated kernel for lenny with this fix now. Sorry, it turns out this didn't fix the problem after all. :/ Putting NFS load on the system caused the panic to reappear again, and even downgrading to the 2.6.26 dma kernel doesn't fix it; so it seems it was just coincidence that it first appeared after I upgraded, and this patch is a wash. Vince Sanders suggests this is a misuse-of-the-DMA-engine problem generally, perhaps related to alignment issues. -- Steve Langasek Give me a lever long enough and a Free OS Debian Developer to set it on, and I can move the world. Ubuntu Developer http://www.debian.org/ slangasek@ubuntu.com vorlon@debian.org
Attachment:
signature.asc
Description: Digital signature