Skip to contentSkip to conversation

gspca main: reorganize loop

PATCH In review Németh Márton 3 participants 5 messages Nov 29, 2009 Media (V4L/DVB)

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 gspca main: reorganize loop

From: Márton Németh <[email protected]>

Eliminate redundant code by reorganizing the loop.

---

linux/drivers/media/video/gspca/gspca.c +5 -11
diff -r 064a82aa2daa linux/drivers/media/video/gspca/gspca.c
@@ -623,12 +623,12 @@
 		if (ret < 0)
 			goto out;
 	}
-	ep = get_ep(gspca_dev);
-	if (ep == NULL) {
-		ret = -EIO;
-		goto out;
-	}
 	for (;;) {
+		ep = get_ep(gspca_dev);
+		if (ep == NULL) {
+			ret = -EIO;
+			goto out;
+		}
 		PDEBUG(D_STREAM, "init transfer alt %d", gspca_dev->alt);
 		ret = create_urbs(gspca_dev, ep);
 		if (ret < 0)
@@ -677,12 +677,6 @@
 			ret = gspca_dev->sd_desc->isoc_nego(gspca_dev);
 			if (ret < 0)
 				goto out;
-		} else {
-			ep = get_ep(gspca_dev);
-			if (ep == NULL) {
-				ret = -EIO;
-				goto out;
-			}
 		}
 	}
 out:
#1 lore raw
Quoted 30 lines · On Sun, 29 Nov 2009 11:24:31 +0100 Németh Márton <[email protected]> wrote:

From: Márton Németh <[email protected]>

Eliminate redundant code by reorganizing the loop.

---
diff -r 064a82aa2daa linux/drivers/media/video/gspca/gspca.c
--- a/linux/drivers/media/video/gspca/gspca.c Thu Nov 26
19:36:40 2009 +0100 +++
b/linux/drivers/media/video/gspca/gspca.c Sun Nov 29 11:09:33
2009 +0100 @@ -623,12 +623,12 @@ if (ret < 0)

linux/drivers/media/video/gspca/gspca.c +5 -5
 			goto out;
 	}
-	ep = get_ep(gspca_dev);
-	if (ep == NULL) {
-		ret = -EIO;
-		goto out;
-	}
 	for (;;) {
+		ep = get_ep(gspca_dev);
+		if (ep == NULL) {
+			ret = -EIO;
+			goto out;
+		}
 		PDEBUG(D_STREAM, "init transfer alt %d",

gspca_dev->alt); ret = create_urbs(gspca_dev, ep);

linux/drivers/media/video/gspca/gspca.c +0 -0
 		if (ret < 0)
@@ -677,12 +677,6 @@
 			ret =
 gspca_dev->sd_desc->isoc_nego(gspca_dev); if (ret < 0)

linux/drivers/media/video/gspca/gspca.c line 682

Quoted 10 lines
linux/drivers/media/video/gspca/gspca.c +0 -6
 				goto out;
-		} else {
-			ep = get_ep(gspca_dev);
-			if (ep == NULL) {
-				ret = -EIO;
-				goto out;
-			}
 		}
 	}
 out:

Hello Márton,

As you may see, in the loop, get_ep() is called only when isoc_nego() is not called. So, your patch does not work.

Regards.

Signature · 2 lines
Ken ar c'hentañ	|	      ** Breizh ha Linux atav! **
Jef		|		http://moinejf.free.fr/
#2 lore raw
Quoted 46 lines · Jean-Francois Moine
Quoted 30 lines · On Sun, 29 Nov 2009 11:24:31 +0100 Németh Márton <[email protected]> wrote:

From: Márton Németh <[email protected]>

Eliminate redundant code by reorganizing the loop.

---
diff -r 064a82aa2daa linux/drivers/media/video/gspca/gspca.c
--- a/linux/drivers/media/video/gspca/gspca.c Thu Nov 26
19:36:40 2009 +0100 +++
b/linux/drivers/media/video/gspca/gspca.c Sun Nov 29 11:09:33
2009 +0100 @@ -623,12 +623,12 @@ if (ret < 0)

linux/drivers/media/video/gspca/gspca.c +5 -5
 			goto out;
 	}
-	ep = get_ep(gspca_dev);
-	if (ep == NULL) {
-		ret = -EIO;
-		goto out;
-	}
 	for (;;) {
+		ep = get_ep(gspca_dev);
+		if (ep == NULL) {
+			ret = -EIO;
+			goto out;
+		}
 		PDEBUG(D_STREAM, "init transfer alt %d",

gspca_dev->alt); ret = create_urbs(gspca_dev, ep);

linux/drivers/media/video/gspca/gspca.c +0 -0
 		if (ret < 0)
@@ -677,12 +677,6 @@
 			ret =
 gspca_dev->sd_desc->isoc_nego(gspca_dev); if (ret < 0)

linux/drivers/media/video/gspca/gspca.c line 682

Quoted 10 lines
linux/drivers/media/video/gspca/gspca.c +0 -6
 				goto out;
-		} else {
-			ep = get_ep(gspca_dev);
-			if (ep == NULL) {
-				ret = -EIO;
-				goto out;
-			}
 		}
 	}
 out:

Hello Márton,

Quoted 2 lines

As you may see, in the loop, get_ep() is called only when isoc_nego() is not called. So, your patch does not work.

You are right, I overseen that.

Is there any subdriver where the isoc_nego() is implemented? I couldn't find one. What would be the task of the isoc_nego() function? Should it set the interface by calling usb_set_interface() as the get_ep() does? Should it create URBs for the endpoint?

Although I found the patch where the isoc_nego() was introduced
( http://linuxtv.org/hg/v4l-dvb/rev/5a5b23605bdb56aec86c9a89de8ca8b8ae9cb925 ) it is not clear how the "ep" pointer is updated when not the isoc_nego() is called instead of get_ep() in the current implementation.

Regards,

Márton Németh

#3 lore raw
Quoted 10 lines · On Sun, 29 Nov 2009 12:19:46 +0100 Németh Márton <[email protected]> wrote:

Is there any subdriver where the isoc_nego() is implemented? I couldn't find one. What would be the task of the isoc_nego() function? Should it set the interface by calling usb_set_interface() as the get_ep() does? Should it create URBs for the endpoint?

Although I found the patch where the isoc_nego() was introduced
( http://linuxtv.org/hg/v4l-dvb/rev/5a5b23605bdb56aec86c9a89de8ca8b8ae9cb925 ) it is not clear how the "ep" pointer is updated when not the isoc_nego() is called instead of get_ep() in the current implementation.

Hello Hans,

This function (isoc_nego) was added to fix the Mauro's problem with the st6422. Was this problem solved in some other way, or is the fix still waiting to be pulled?

Best regards.

Signature · 2 lines
Ken ar c'hentañ	|	      ** Breizh ha Linux atav! **
Jef		|		http://moinejf.free.fr/
#4 lore raw

Hi,

Quoted 14 lines · Jean-Francois Moine
Quoted 10 lines · On Sun, 29 Nov 2009 12:19:46 +0100 Németh Márton<[email protected]> wrote:

Is there any subdriver where the isoc_nego() is implemented? I couldn't find one. What would be the task of the isoc_nego() function? Should it set the interface by calling usb_set_interface() as the get_ep() does? Should it create URBs for the endpoint?

Although I found the patch where the isoc_nego() was introduced
( http://linuxtv.org/hg/v4l-dvb/rev/5a5b23605bdb56aec86c9a89de8ca8b8ae9cb925 ) it is not clear how the "ep" pointer is updated when not the isoc_nego() is called instead of get_ep() in the current implementation.

Quoted 6 lines

Hello Hans,

This function (isoc_nego) was added to fix the Mauro's problem with the st6422. Was this problem solved in some other way, or is the fix still waiting to be pulled?

The fix is still waiting to be pulled, or actually to be made working :|

First a quick summary of the need for the isoc_nego() function (and to only call get_ep() once). Some cams, at least those based on stv06xx bridges, have only 1 alt setting, but they have a register which allows one to tell it to send isoc frames which are never bigger then the value set in the register. I've tested this and it works as advertised, when you create isoc urbs with a size < wMaxPacketSize, then normally you will get -EMSGSIZE errors (iirc) as the camera sends isoc frames, larger then the buffers in the isoc urbs created, but if you then write the size you created the isoc urbs with to the register in question, things will work. So this works as expected.

The problem is that the usb "scheduler" which checks the bandwidth constrains will always use wMaxPacketSize to check if there is enough bandwidth instead of the actual packet size of the isoc packets. So although I have a patch implementing the isoc_neg call for stv06xx cams, it does not actually help as although it is scaling back the bandwidth needed, the usb core does not see this and keeps returning -ENOSPC.

After finding out about this I ran out of time. This reminds me that I need to send a message about this to the usb mailing list. I will do so now, and then we will see what will come from this.

Regards,

Hans