On 11/18/2016 04:03 PM, Christoph Hellwig wrote:
> We can't handle vfree itself from atomic context, but callers
> can explicitly use vfree_atomic instead, which defers the actual
> vfree to a workqueue. Unfortunately in_atomic does not work
> on non-preemptible kernels, so we can't just do the right thing
> by default.
>
> Signed-off-by: Christoph Hellwig <hch@xxxxxx>
> ---
> mm/vmalloc.c | 1 +
> 1 file changed, 1 insertion(+)
>
> diff --git a/mm/vmalloc.c b/mm/vmalloc.c
> index 80f3fae..e2030b4 100644
> --- a/mm/vmalloc.c
> +++ b/mm/vmalloc.c
> @@ -1530,6 +1530,7 @@ void vfree_atomic(const void *addr)
> void vfree(const void *addr)
> {
> BUG_ON(in_nmi());
> + WARN_ON_ONCE(in_atomic());
This one is wrong. We still can call vfree() from interrupt context.
So WARN_ON_ONCE(in_atomic() && !in_interrupt()) would be correct,
but also redundant. DEBUG_ATOMIC_SLEEP=y should catch illegal vfree() calls.
Let's just drop this patch, ok?
> kmemleak_free(addr);
>
>
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@xxxxxxxxx. For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@xxxxxxxxx";> email@xxxxxxxxx </a>