Re: "vmalloc", friends and GFP flags

From: James Bottomley (James.Bottomley@steeleye.com)
Date: Wed Jan 01 2003 - 17:57:51 EST


> I've just been looking over the state of the ARM {dma,pci}_alloc_consis
> tent wrt interrupts etc in 2.5.53, and decided to have a look around
> this area for other users. I stumbled across the following unsafe
> areas.

> a) parisc. Looking at pa11_dma_alloc_consistent:

I agree with the flow analysis. However, there is a condition attached to
getting from pte_alloc_kernel to pte_alloc_one_kernel which makes the
GFP_KERNEL allocation. That condition looks to be that the pmd is not
present. I don't believe that's possible for an area of memory we just got
from __get_free_pages (otherwise the kernel could fault when accessing that
region which isn't allowed).

If the above analysis is correct (which I'll defer to better opinion) then we
can in practice never get down the problem path, and dma_ allocation from
interrupt is safe on parisc.

James

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/



This archive was generated by hypermail 2b29 : Tue Jan 07 2003 - 22:00:15 EST