Skip to contentSkip to conversation

USB: misplaced parenthesis

PATCH In review Roel Kluin 3 participants 5 messages Feb 15, 2010 – Feb 17, 2010

Subsystems (2)
lore
Share & tools
lore

Roel Kluin

author
#0 lore raw

PATCH 1/1 USB: misplaced parenthesis

The parenthesis was misplaced

---
Is this maybe, as the comment states, why blanking a cdrw at speed 4 was unreliable?

drivers/usb/storage/shuttle_usbat.c +2 -2
index b62a288..b958db5 100644
@@ -1645,8 +1645,8 @@ static int usbat_hp8200e_transport(struct scsi_cmnd *srb, struct us_data *us)
 
 	if ((result = usbat_write_block(us,
 			USBAT_ATA, srb->cmnd, 12,
-				(srb->cmnd[0]==GPCMD_BLANK ? 75 : 10), 0) !=
-			     USB_STOR_TRANSPORT_GOOD)) {
+				(srb->cmnd[0]==GPCMD_BLANK ? 75 : 10), 0)) !=
+			     USB_STOR_TRANSPORT_GOOD) {
 		return result;
 	}
 
#1 lore raw
Quoted 12 lines · Roel Kluin

The parenthesis was misplaced

---
Is this maybe, as the comment states, why blanking a cdrw at speed 4 was unreliable?

drivers/usb/storage/shuttle_usbat.c +0 -0
index b62a288..b958db5 100644

drivers/usb/storage/shuttle_usbat.c static int usbat_hp8200e_transport(struct scsi_cmnd *srb, struct us_data *us) · line 1651

Quoted 9 lines
drivers/usb/storage/shuttle_usbat.c +2 -2
@@ -1645,8 +1645,8 @@ static int usbat_hp8200e_transport(struct scsi_cmnd *srb, struct us_data *us)
 
 	if ((result = usbat_write_block(us,
 			USBAT_ATA, srb->cmnd, 12,
-				(srb->cmnd[0]==GPCMD_BLANK ? 75 : 10), 0) !=
-			     USB_STOR_TRANSPORT_GOOD)) {
+				(srb->cmnd[0]==GPCMD_BLANK ? 75 : 10), 0)) !=
+			     USB_STOR_TRANSPORT_GOOD) {
 		return result;
 	}

I think it'd be better if you hoisted the set'n'test out of the if()

Isn't this the current logic?

	result = usbat_write_block(us, USBAT_ATA, srb->cmnd, 12,
				   srb->cmnd[0] == GPCMD_BLANK ? 75 : 10, 0);
	result = result != USB_STOR_TRANSPORT_GOOD;
	if (result)
		return result;

I wonder if it should be:

	result = usbat_write_block(us, USBAT_ATA, srb->cmnd, 12,
				   srb->cmnd[0] == GPCMD_BLANK ? 75 : 10, 0);
	if (result != USB_STOR_TRANSPORT_GOOD)
		return result;
author
#2 lore raw
Quoted 1 line

I think it'd be better if you hoisted the set'n'test out of the if()

ok, I agree.

Quoted 7 lines

Isn't this the current logic?

	result = usbat_write_block(us, USBAT_ATA, srb->cmnd, 12,
				   srb->cmnd[0] == GPCMD_BLANK ? 75 : 10, 0);
	result = result != USB_STOR_TRANSPORT_GOOD;
	if (result)
		return result;

Thanks for your comments, Yes that was the current logic, which I thought was wrong, but now I think it could also be obscurely written but right:

in drivers/usb/storage/transport.h line 100 note the definitions:

#define USB_STOR_TRANSPORT_GOOD    0   /* Transport good, command good     */
#define USB_STOR_TRANSPORT_FAILED  1   /* Transport good, command failed   */
#define USB_STOR_TRANSPORT_NO_SENSE 2  /* Command failed, no auto-sense    */
#define USB_STOR_TRANSPORT_ERROR   3   /* Transport bad (i.e. device dead) */

With the current logic usbat_hp8200e_transport() returns TRANSPORT_FAILED, even if usbat_write_block() returned TRANSPORT_NO_SENSE or TRANSPORT_ERROR.

This could be intended, but then the author chose a very obscure way to write:

	if (usbat_write_block(us, USBAT_ATA, srb->cmnd, 12,
			      srb->cmnd[0] == GPCMD_BLANK ? 75 : 10, 0) !=
			      USB_STOR_TRANSPORT_GOOD) 
		return USB_STOR_TRANSPORT_FAILED;

Or was the parenthesis misplaced and should it really be:

	result = usbat_write_block(us, USBAT_ATA, srb->cmnd, 12,
				   srb->cmnd[0] == GPCMD_BLANK ? 75 : 10, 0);

	if (result != USB_STOR_TRANSPORT_GOOD)

return result;

Maybe someone with the specs/more knowledge of this driver could look into this?

Thanks, Roel

maintainer
#3 lore raw
Quoted 34 lines · Roel Kluin
Quoted 1 line

I think it'd be better if you hoisted the set'n'test out of the if()

ok, I agree.

Quoted 7 lines

Isn't this the current logic?

	result = usbat_write_block(us, USBAT_ATA, srb->cmnd, 12,
				   srb->cmnd[0] == GPCMD_BLANK ? 75 : 10, 0);
	result = result != USB_STOR_TRANSPORT_GOOD;
	if (result)
		return result;

Thanks for your comments, Yes that was the current logic, which I thought was wrong, but now I think it could also be obscurely written but right:

in drivers/usb/storage/transport.h line 100 note the definitions:

#define USB_STOR_TRANSPORT_GOOD    0   /* Transport good, command good     */
#define USB_STOR_TRANSPORT_FAILED  1   /* Transport good, command failed   */
#define USB_STOR_TRANSPORT_NO_SENSE 2  /* Command failed, no auto-sense    */
#define USB_STOR_TRANSPORT_ERROR   3   /* Transport bad (i.e. device dead) */

With the current logic usbat_hp8200e_transport() returns TRANSPORT_FAILED, even if usbat_write_block() returned TRANSPORT_NO_SENSE or TRANSPORT_ERROR.

This could be intended, but then the author chose a very obscure way to write:

	if (usbat_write_block(us, USBAT_ATA, srb->cmnd, 12,
			      srb->cmnd[0] == GPCMD_BLANK ? 75 : 10, 0) !=
			      USB_STOR_TRANSPORT_GOOD) 
		return USB_STOR_TRANSPORT_FAILED;

Or was the parenthesis misplaced and should it really be:

Quoted 8 lines
	result = usbat_write_block(us, USBAT_ATA, srb->cmnd, 12,
				   srb->cmnd[0] == GPCMD_BLANK ? 75 : 10, 0);

	if (result != USB_STOR_TRANSPORT_GOOD)

return result;

Maybe someone with the specs/more knowledge of this driver could look into this?

It seems pretty clear that your patch was correct and the parens were misplaced. In usb-storage, transport routines like usbat_hp8200e_transport() are supposed to return one of the USB_STOR_TRANSPORT_* codes, not a Boolean value.

I do agree with Joe that it would be better form to separate the function call and the "if" into two statements, as in your second version above. Compare with the code a few lines higher:

        if ( (result = usbat_multiple_write(us, 
                        registers, data, 7)) != USB_STOR_TRANSPORT_GOOD) {
                return result;
        }

The meaning is clear, even though this also unnecessarily squeezes a function call and a test into one statement and includes unneeded {}'s.

Alan Stern

author
#4 lore raw

Due to a misplaced parenthesis the usbat_write_block() return value was not stored, but a boolean. USB_STOR_TRANSPORT_NO_SENSE and USB_STOR_TRANSPORT_ERROR were returned as USB_STOR_TRANSPORT_FAILED.

---

Quoted 7 lines
Quoted 6 lines

in drivers/usb/storage/transport.h line 100 note the definitions:

#define USB_STOR_TRANSPORT_GOOD    0   /* Transport good, command good     */
#define USB_STOR_TRANSPORT_FAILED  1   /* Transport good, command failed   */
#define USB_STOR_TRANSPORT_NO_SENSE 2  /* Command failed, no auto-sense    */
#define USB_STOR_TRANSPORT_ERROR   3   /* Transport bad (i.e. device dead) */
Quoted 8 lines

It seems pretty clear that your patch was correct and the parens were misplaced. In usb-storage, transport routines like usbat_hp8200e_transport() are supposed to return one of the USB_STOR_TRANSPORT_* codes, not a Boolean value.

I do agree with Joe that it would be better form to separate the function call and the "if" into two statements, as in your second version above.

Ok, thanks for comments, This should fix and separate the assignment as desired.

Roel

drivers/usb/storage/shuttle_usbat.c +7 -8
index b62a288..bd3f415 100644
@@ -1628,10 +1628,10 @@ static int usbat_hp8200e_transport(struct scsi_cmnd *srb, struct us_data *us)
 		return USB_STOR_TRANSPORT_ERROR;
 	}
 
-	if ( (result = usbat_multiple_write(us, 
-			registers, data, 7)) != USB_STOR_TRANSPORT_GOOD) {
+	result = usbat_multiple_write(us, registers, data, 7);
+
+	if (result != USB_STOR_TRANSPORT_GOOD)
 		return result;
-	}
 
 	/*
 	 * Write the 12-byte command header.
@@ -1643,12 +1643,11 @@ static int usbat_hp8200e_transport(struct scsi_cmnd *srb, struct us_data *us)
 	 * AT SPEED 4 IS UNRELIABLE!!!
 	 */
 
-	if ((result = usbat_write_block(us,
-			USBAT_ATA, srb->cmnd, 12,
-				(srb->cmnd[0]==GPCMD_BLANK ? 75 : 10), 0) !=
-			     USB_STOR_TRANSPORT_GOOD)) {
+	result = usbat_write_block(us, USBAT_ATA, srb->cmnd, 12,
+				   srb->cmnd[0] == GPCMD_BLANK ? 75 : 10, 0);
+
+	if (result != USB_STOR_TRANSPORT_GOOD)
 		return result;
-	}
 
 	/* If there is response data to be read in then do it here. */