Skip to contentSkip to conversation

usb/gadget/pxa27x_udc: test ep0state == IN_DATA_STAGE rather than non-bool USB_DIR_IN

PATCH In review Roel Kluin 2 participants 2 messages May 14, 2008 USB

lore
Share & tools
lore

Roel Kluin

author
#0 lore raw

PATCH 1/1 usb/gadget/pxa27x_udc: test ep0state == IN_DATA_STAGE rather than non-bool USB_DIR_IN

The patch below was not tested!
#------------------------------------------------------------------------------#
drivers/usb/gadget/pxa27x_udc.c uses USB_DIR_IN as if boolean in functions inc_ep_stats_{reqs,bytes}, lines 763, 795, 933, 985, 1037 and 1074, but USB_DIR_IN is defined 0x80.

handle_ep0_ctrl_req() tests USB_DIR_IN bitwise and pxa_udc's ep0_state is set IN_DATA_STAGE or OUT_DATA_STAGE. so ep->dev->ep0state == IN_DATA_STAGE should be tested rather than USB_DIR_IN

I am uncertain whether this fix is correct and sufficient, feedback appreciated.
- Maybe more tests are required: is ep->dev == NULL or similar
- Maybe I should test ep->dev->ep0state == OUT_DATA_STAGE rather than ep->dev->ep0state != IN_DATA_STAGE (instead of !USB_DIR_IN)

---
USB_DIR_IN is not boolean, test ep0state == IN_DATA_STAGE instead.

Signed-off-by Roel Kluin <[email protected]>

---

drivers/usb/gadget/pxa27x_udc.c +8 -5
index 75eba20..ea18995 100644
@@ -760,7 +760,7 @@ static void req_done(struct pxa_ep *ep, struct pxa27x_request *req, int status)
  */
 static void ep_end_out_req(struct pxa_ep *ep, struct pxa27x_request *req)
 {
-	inc_ep_stats_reqs(ep, !USB_DIR_IN);
+	inc_ep_stats_reqs(ep, ep->dev->ep0state != IN_DATA_STAGE);
 	req_done(ep, req, 0);
 }
 

@@ -792,7 +792,7 @@ static void ep0_end_out_req(struct pxa_ep *ep, struct pxa27x_request *req)
  */
 static void ep_end_in_req(struct pxa_ep *ep, struct pxa27x_request *req)
 {
-	inc_ep_stats_reqs(ep, USB_DIR_IN);
+	inc_ep_stats_reqs(ep, ep->dev->ep0state == IN_DATA_STAGE);
 	req_done(ep, req, 0);
 }
 

@@ -930,7 +930,8 @@ static int read_fifo(struct pxa_ep *ep, struct pxa27x_request *req)
 
 	while (epout_has_pkt(ep)) {
 		count = read_packet(ep, req);
-		inc_ep_stats_bytes(ep, count, !USB_DIR_IN);
+		inc_ep_stats_bytes(ep, count,
+				ep->dev->ep0state != IN_DATA_STAGE);
 
 		is_short = (count < ep->fifo_size);
 		ep_dbg(ep, "read udccsr:%03x, count:%d bytes%s req %p %d/%d\n",

@@ -982,7 +983,8 @@ static int write_fifo(struct pxa_ep *ep, struct pxa27x_request *req)
 		}
 
 		count = write_packet(ep, req, max);
-		inc_ep_stats_bytes(ep, count, USB_DIR_IN);
+		inc_ep_stats_bytes(ep, count,
+				ep->dev->ep0state == IN_DATA_STAGE);
 		totcount += count;
 
 		/* last packet is usually short (or a zlp) */

@@ -1034,7 +1036,8 @@ static int read_ep0_fifo(struct pxa_ep *ep, struct pxa27x_request *req)
 	while (epout_has_pkt(ep)) {
 		count = read_packet(ep, req);
 		udc_ep_writel(ep, UDCCSR, UDCCSR0_OPC);
-		inc_ep_stats_bytes(ep, count, !USB_DIR_IN);
+		inc_ep_stats_bytes(ep, count,
+				ep->dev->ep0state != IN_DATA_STAGE);
 
 		is_short = (count < ep->fifo_size);
 		ep_dbg(ep, "read udccsr:%03x, count:%d bytes%s req %p %d/%d\n",
#1 lore raw
Quoted 3 lines · Roel Kluin

drivers/usb/gadget/pxa27x_udc.c uses USB_DIR_IN as if boolean in functions inc_ep_stats_{reqs,bytes}, lines 763, 795, 933, 985, 1037 and 1074, but USB_DIR_IN is defined 0x80.

Yes, USB_DIR_IN could be considered as a boolean (0x80 => true, 0x0 => false). I'm pretty sure it's correct.

Quoted 3 lines

handle_ep0_ctrl_req() tests USB_DIR_IN bitwise and pxa_udc's ep0_state is set IN_DATA_STAGE or OUT_DATA_STAGE. so ep->dev->ep0state == IN_DATA_STAGE should be tested rather than USB_DIR_IN

No, I don't agree. Have you checked the USB 2.0 spec, especially on control endpoint handling and control requests ?

Sorry Roel, I can't ack this patch.

--
Robert