Commit d5dcc336 authored by Pawel Laszczak's avatar Pawel Laszczak Committed by Greg Kroah-Hartman

usb: cdns3: fix incorrect handling TRB_SMM flag for ISOC transfer

The TRB_SMM flag indicates that DMA has completed the TD service with
this TRB. Usually it’s a last TRB in TD. In case of ISOC transfer for
bInterval > 1 each ISOC transfer contains more than one TD associated
with usb request (one TD per ITP). In such case the TRB_SMM flag will
be set in every TD and driver will recognize the end of transfer after
processing the first TD with TRB_SMM. In result driver stops updating
request->actual and returns incorrect actual length.
To fix this issue driver additionally must check TRB_CHAIN which is not
used for isochronous transfers.

Fixes: 249f0a25 ("usb: cdns3: gadget: handle sg list use case at completion correctly")
cc: <stable@vger.kernel.org>
Acked-by: default avatarPeter Chen <peter.chen@kernel.org>
Signed-off-by: default avatarPawel Laszczak <pawell@cadence.com>
Link: https://lore.kernel.org/r/20220825062207.5824-1-pawell@cadence.comSigned-off-by: default avatarGreg Kroah-Hartman <gregkh@linuxfoundation.org>
parent 9d4dc16e
...@@ -1530,7 +1530,8 @@ static void cdns3_transfer_completed(struct cdns3_device *priv_dev, ...@@ -1530,7 +1530,8 @@ static void cdns3_transfer_completed(struct cdns3_device *priv_dev,
TRB_LEN(le32_to_cpu(trb->length)); TRB_LEN(le32_to_cpu(trb->length));
if (priv_req->num_of_trb > 1 && if (priv_req->num_of_trb > 1 &&
le32_to_cpu(trb->control) & TRB_SMM) le32_to_cpu(trb->control) & TRB_SMM &&
le32_to_cpu(trb->control) & TRB_CHAIN)
transfer_end = true; transfer_end = true;
cdns3_ep_inc_deq(priv_ep); cdns3_ep_inc_deq(priv_ep);
......
Markdown is supported
0%
or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment