On 10/07/15 18:49, Cormac O'Brien wrote:
Previous versions of QEMU use a 2-byte header for CUDA ADB packets where it should have used a 3-byte one, so this commit allows cuda_adb_req() to differentiate between the two formats and act accordingly.
Signed-off-by: Cormac O'Brien cormac@c-obrien.org
drivers/cuda.c | 14 ++++++++++++-- 1 file changed, 12 insertions(+), 2 deletions(-)
diff --git a/drivers/cuda.c b/drivers/cuda.c index 9555dea..deceacf 100644 --- a/drivers/cuda.c +++ b/drivers/cuda.c @@ -144,8 +144,18 @@ static int cuda_adb_req (void *host, const uint8_t *snd_buf, int len, // CUDA_DPRINTF("len: %d %02x\n", len, snd_buf[0]); len = cuda_request(host, ADB_PACKET, snd_buf, len, buffer); if (len > 1 && buffer[0] == ADB_PACKET) {
pos = buffer + 2;len -= 2;
/* QEMU's previous model used a 2-byte header where it should use a* 3-byte one, so we check to see what kind of header we ought to use.*/if (len > 2 && buffer[2] == snd_buf[0]) {/* Correct 3-byte header */pos = buffer + 3;len -= 3;} else {/* Old 2-byte header */pos = buffer + 2;len -= 2;}
Technically the patch looks good to me - due to the nature of the change, can you confirm that this patch doesn't cause regressions in any other OS images during testing? Even better, is there a reference to a 3-byte header in any of the open OS sources to confirm that this is definitely the correct solution?
The only change I would like to see is a reference to specific QEMU versions in the comments and commit log, so instead of referencing "QEMU's previous model" it would be better to state "QEMU versions <= 2.4" instead. This just makes it a bit easier to understand to implications of removing the hack at a later date.
ATB,
Mark.