On 05/31/2016 08:29 AM, Cédric Le Goater wrote:
> On 05/31/2016 04:26 PM, Eric Blake wrote:
>> On 05/31/2016 05:36 AM, Cédric Le Goater wrote:
>>> commit 243e6f69c129 ("m25p80: Switch to byte-based block access")
>>> replaced blk_read() calls with blk_pread() but return values are
>>> different.
>>
>> Shoot, I completely missed that when I made the conversions. Now I need
>> to re-audit that entire series to see if the same problem happened
>> anywhere else.
>
> I took a quick look and most of the calls to blk_pread() test with < 0.
> So we should be fine and I should have mention that.
>
> C.
>
>>>
>>> Signed-off-by: Cédric Le Goater <[email protected]>
>>> ---
>>> hw/block/m25p80.c | 2 +-
>>> 1 file changed, 1 insertion(+), 1 deletion(-)
>>
>> Reviewed-by: Eric Blake <[email protected]>
>>
>>>
>>> Index: qemu-ast2400-mainline.git/hw/block/m25p80.c
>>> ===================================================================
>>> --- qemu-ast2400-mainline.git.orig/hw/block/m25p80.c
>>> +++ qemu-ast2400-mainline.git/hw/block/m25p80.c
>>> @@ -900,7 +900,7 @@ static int m25p80_init(SSISlave *ss)
>>> s->storage = blk_blockalign(s->blk, s->size);
>>>
>>> /* FIXME: Move to late init */
>>> - if (blk_pread(s->blk, 0, s->storage, s->size)) {
>>> + if (blk_pread(s->blk, 0, s->storage, s->size) != s->size) {As it is, blk_pread() (and blk_pwrite()) _always_ returns negative or the input count value; it appears that it is incapable of reporting a short read amount. I guess that's intentional, but a bit odd, when compared to the standard read()/write(). -- Eric Blake eblake redhat com +1-919-301-3266 Libvirt virtualization library http://libvirt.org
signature.asc
Description: OpenPGP digital signature
