"Also sprach ptb:"
> mbio = bio_clone(bio, GFP_NOIO);
> ...
> mbio->bi_sector = r1_bio->sector + conf->mirrors[i].rdev->data_offset;
OK - the GFP_NOIO includes wait flags, so it seems to wait at least for
a bit.
include/linux/gfp.h:#define GFP_NOIO (__GFP_WAIT)
But can't the allocation still fail sometimes? And if not won't we just
stall forever in a deadlock if for example the only available memory is
in buffers that are aimed at this same raid device?
> but I don't think bio_clone is GUARANTEED to produce a non-null result,
> is it? It uses bio_alloc to get a bio, and bio_alloc does ...
>
> bio = mempool_alloc(bio_pool, gfp_mask);
> if (unlikely(!bio))
> goto out;
>
> out:
> return bio;
> }
>
> mempool_free(bio, bio_pool);
> bio = NULL;
> goto out;
>
> So it looks perfectly capable of producing a NULL.
Peter
-
To unsubscribe from this list: send the line "unsubscribe linux-raid" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
