Re: [RFC PATCH 17/25] hw/virtio: Use correct type sizes

[Philippe Mathieu-Daudé]

On 5/20/21 6:27 PM, Richard Henderson wrote:
> On 5/18/21 1:36 PM, Philippe Mathieu-Daudé wrote:
>> -static inline int virtio_lduw_p(VirtIODevice *vdev, const void *ptr)
>> +static inline uint16_t virtio_lduw_p(VirtIODevice *vdev, const void
>> *ptr)
> 
> While this one looks obviously correct,
> 
>> -static inline int virtio_ldl_p(VirtIODevice *vdev, const void *ptr)
>> +static inline uint32_t virtio_ldl_p(VirtIODevice *vdev, const void *ptr)
>>   {
>>       if (virtio_access_is_big_endian(vdev)) {
>>           return ldl_be_p(ptr);
> 
> this one isn't so obvious.
> 
> Are all of the users unsigned?

All except this one for which I'm not sure:

hw/block/virtio-blk.c:137:            int p =
virtio_ldl_p(VIRTIO_DEVICE(s), &req->out.type);
hw/block/virtio-blk.c-138-            bool is_read = !(p &
VIRTIO_BLK_T_OUT);

--

hw/block/virtio-blk.c:183:    bool is_write_zeroes =
(virtio_ldl_p(VIRTIO_DEVICE(s), &req->out.type) &
hw/block/virtio-blk.c-184-
~VIRTIO_BLK_T_BARRIER) == VIRTIO_BLK_T_WRITE_ZEROES;
hw/block/virtio-blk.c-185-

/* Barrier before this op. */
#define VIRTIO_BLK_T_BARRIER    0x80000000

>  If so, should we rename it ldul?

OK.