On Fri, 27 May 2011 11:50:03 +0800
Bob Liu <lliubbo@xxxxxxxxx> wrote:
> remap_pfn_range() does not update vma->end on no mmu arch which will
> cause munmap() fail because it can't match the vma.
>
> eg. fb_mmap() in fbmem.c will call io_remap_pfn_range() which is
> remap_pfn_range() on nommu arch, if an address is not page aligned vma->start
> will be changed in remap_pfn_range(), but neither size nor vma->end will be
> updated. Then munmap(start, len) can't find the vma to free, because it need to
> compare (start + len) with vma->end.
>
> Signed-off-by: Bob Liu <lliubbo@xxxxxxxxx>
> ---
> mm/nommu.c | 1 +
> 1 files changed, 1 insertions(+), 0 deletions(-)
>
> diff --git a/mm/nommu.c b/mm/nommu.c
> index 1fd0c51..829848a 100644
> --- a/mm/nommu.c
> +++ b/mm/nommu.c
> @@ -1817,6 +1817,7 @@ int remap_pfn_range(struct vm_area_struct *vma, unsigned long from,
> unsigned long to, unsigned long size, pgprot_t prot)
> {
> vma->vm_start = vma->vm_pgoff << PAGE_SHIFT;
> + vma->vm_end = vma->vm_start + size;
> return 0;
> }
> EXPORT_SYMBOL(remap_pfn_range);
hm.
The MMU version of remap_pfn_range() doesn't do this. Seems that it
just leaves the omitted parts of the vma unmapped. Obviously nommu
can't do that, but the divergence is always a concern.
Thsi implementation could lead to overlapping vmas. Should we be
checking that it fits?
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@xxxxxxxxxx For more info on Linux MM,
see: http://www.linux-mm.org/ .
Fight unfair telecom internet charges in Canada: sign http://stopthemeter.ca/
Don't email: <a href=mailto:"dont@xxxxxxxxx";> email@xxxxxxxxx </a>