Re: [PATCH] target: Allow for target_submit_cmd() returning errors

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

On 07/18/2012 02:05 AM, Nicholas A. Bellinger wrote:
index 02ace18..9ddf315 100644
--- a/drivers/usb/gadget/tcm_usb_gadget.c
+++ b/drivers/usb/gadget/tcm_usb_gadget.c
@@ -1060,7 +1060,10 @@ static void usbg_cmd_work(struct work_struct *work)
  	tpg = cmd->fu->tpg;
  	tv_nexus = tpg->tpg_nexus;
  	dir = get_cmd_dir(cmd->cmd_buf);
-	if (dir<  0) {
+	if (dir<  0 ||
+	    target_submit_cmd(se_cmd, tv_nexus->tvn_se_sess,
+			      cmd->cmd_buf, cmd->sense_iu.sense, cmd->unpacked_lun,
+			      0, cmd->prio_attr, dir, TARGET_SCF_UNKNOWN_SIZE)) {
  				tv_nexus->tvn_se_sess, cmd->data_len, DMA_NONE,
@@ -1069,12 +1072,7 @@ static void usbg_cmd_work(struct work_struct *work)
-		return;
-	target_submit_cmd(se_cmd, tv_nexus->tvn_se_sess,
-			cmd->cmd_buf, cmd->sense_iu.sense, cmd->unpacked_lun,
-			0, cmd->prio_attr, dir, TARGET_SCF_UNKNOWN_SIZE);

  static int usbg_submit_command(struct f_uas *fu,

Looking at this part again target_submit_cmd() has been moved ahead of
target_init_se_cmd() in the exception path, which ends up getting called
twice during a target_get_sess_cmd() failure..

It looks like usbg_cmd_work() + bot_cmd_work() will need some work if
they intends to use a non zero failure to signal exception here, without
invoking a CHECK_CONDITION and sense.  Actually, I'm not even sure
sending a CHECK_CONDITION here after the target_submit_cmd() is going to
work for other fabrics drivers, but it appears to work with usb-gadget.

For UAS the status/ sense URB is sent right away (without any data) and
this worked the last time I tested.
For BOT things are a little complicated. What I do is to send empty
data (to fill the data pipe) and send a bad status so the other side
learns that the transfer failed somehow. The sense information is lost.
What I should do is to queue this sense data for the next MODE_SENSE
request which will be coming soon. Right now WinXP repeats a failed
command thrice if MODE_SENSE returns no error. "This" is on my to-fix

So for the moment, lets fix the double se_cmd init and make things a
little easier to read..   Sebastian, please have a look and double check
that the (dir<  0) + target_submit_cmd() failures cases are both going
to work as expected whilst sending a CHECK_CONDITION response.

The sense code should only be sent once and
transport_send_check_condition_and_sense() sets
SCF_SENT_CHECK_CONDITION and checks for it so doing it twice should not
do any harm right? But grep does not say when this flag gets removed.



To unsubscribe from this list: send the line "unsubscribe linux-usb" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at

B and H Foto and Electronics Corp.

[Linux Media]     [Video for Linux]     [Linux Input]     [Linux Audio Users]     [Photo]     [Yosemite News]    [Yosemite Photos]    [Free Online Dating]     [Linux Kernel]     [Linux SCSI]     [Old Linux USB Devel Archive]     [More Archives]

Add to Google Powered by Linux