Skip to contentSkip to conversation

Fix __ucmpdi2 in v4l2_norm_to_name()

PATCH In review Stelian Pop 5 participants 10 messages Jan 4, 2007 – Jan 15, 2007 Media (V4L/DVB)

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 Fix __ucmpdi2 in v4l2_norm_to_name()

Hi,

This patch replaces a switch statement using 64 bit values with the if/else equivalent in order to prevent a call __ucmpdi2 generated by some versions of gcc (verified with gcc-4.1.2 20060928):

	drivers/built-in.o: In function `v4l2_norm_to_name':
	(.text+0x71100): undefined reference to `__ucmpdi2'

---

 drivers/media/video/v4l2-common.c |  126 ++++++++++++++++++-------------------
 1 files changed, 62 insertions(+), 64 deletions(-)
drivers/media/video/v4l2-common.c +62 -64
index 752c82c..0c3c2f6 100644
@@ -91,70 +91,68 @@ char *v4l2_norm_to_name(v4l2_std_id id)
 {
 	char *name;
 
-	switch (id) {
-	case V4L2_STD_PAL:
-		name="PAL";		break;
-	case V4L2_STD_PAL_BG:
-		name="PAL-BG";		break;
-	case V4L2_STD_PAL_DK:
-		name="PAL-DK";		break;
-	case V4L2_STD_PAL_B:
-		name="PAL-B";		break;
-	case V4L2_STD_PAL_B1:
-		name="PAL-B1";		break;
-	case V4L2_STD_PAL_G:
-		name="PAL-G";		break;
-	case V4L2_STD_PAL_H:
-		name="PAL-H";		break;
-	case V4L2_STD_PAL_I:
-		name="PAL-I";		break;
-	case V4L2_STD_PAL_D:
-		name="PAL-D";		break;
-	case V4L2_STD_PAL_D1:
-		name="PAL-D1";		break;
-	case V4L2_STD_PAL_K:
-		name="PAL-K";		break;
-	case V4L2_STD_PAL_M:
-		name="PAL-M";		break;
-	case V4L2_STD_PAL_N:
-		name="PAL-N";		break;
-	case V4L2_STD_PAL_Nc:
-		name="PAL-Nc";		break;
-	case V4L2_STD_PAL_60:
-		name="PAL-60";		break;
-	case V4L2_STD_NTSC:
-		name="NTSC";		break;
-	case V4L2_STD_NTSC_M:
-		name="NTSC-M";		break;
-	case V4L2_STD_NTSC_M_JP:
-		name="NTSC-M-JP";	break;
-	case V4L2_STD_NTSC_443:
-		name="NTSC-443";	break;
-	case V4L2_STD_NTSC_M_KR:
-		name="NTSC-M-KR";	break;
-	case V4L2_STD_SECAM:
-		name="SECAM";		break;
-	case V4L2_STD_SECAM_DK:
-		name="SECAM-DK";	break;
-	case V4L2_STD_SECAM_B:
-		name="SECAM-B";		break;
-	case V4L2_STD_SECAM_D:
-		name="SECAM-D";		break;
-	case V4L2_STD_SECAM_G:
-		name="SECAM-G";		break;
-	case V4L2_STD_SECAM_H:
-		name="SECAM-H";		break;
-	case V4L2_STD_SECAM_K:
-		name="SECAM-K";		break;
-	case V4L2_STD_SECAM_K1:
-		name="SECAM-K1";	break;
-	case V4L2_STD_SECAM_L:
-		name="SECAM-L";		break;
-	case V4L2_STD_SECAM_LC:
-		name="SECAM-LC";	break;
-	default:
-		name="Unknown";		break;
-	}
+	if (id == V4L2_STD_PAL)
+		name = "PAL";
+	else if (id == V4L2_STD_PAL_BG)
+		name = "PAL-BG";
+	else if (id == V4L2_STD_PAL_DK)
+		name = "PAL-DK";
+	else if (id == V4L2_STD_PAL_B)
+		name = "PAL-B";
+	else if (id == V4L2_STD_PAL_B1)
+		name = "PAL-B1";
+	else if (id == V4L2_STD_PAL_G)
+		name = "PAL-G";
+	else if (id == V4L2_STD_PAL_H)
+		name = "PAL-H";
+	else if (id == V4L2_STD_PAL_I)
+		name = "PAL-I";
+	else if (id == V4L2_STD_PAL_D)
+		name = "PAL-D";
+	else if (id == V4L2_STD_PAL_D1)
+		name = "PAL-D1";
+	else if (id == V4L2_STD_PAL_K)
+		name = "PAL-K";
+	else if (id == V4L2_STD_PAL_M)
+		name = "PAL-M";
+	else if (id == V4L2_STD_PAL_N)
+		name = "PAL-N";
+	else if (id == V4L2_STD_PAL_Nc)
+		name = "PAL-Nc";
+	else if (id == V4L2_STD_PAL_60)
+		name = "PAL-60";
+	else if (id == V4L2_STD_NTSC)
+		name = "NTSC";
+	else if (id == V4L2_STD_NTSC_M)
+		name = "NTSC-M";
+	else if (id == V4L2_STD_NTSC_M_JP)
+		name = "NTSC-M-JP";
+	else if (id == V4L2_STD_NTSC_443)
+		name = "NTSC-443";
+	else if (id == V4L2_STD_NTSC_M_KR)
+		name = "NTSC-M-KR";
+	else if (id == V4L2_STD_SECAM)
+		name = "SECAM";
+	else if (id == V4L2_STD_SECAM_DK)
+		name = "SECAM-DK";
+	else if (id == V4L2_STD_SECAM_B)
+		name = "SECAM-B";
+	else if (id == V4L2_STD_SECAM_D)
+		name = "SECAM-D";
+	else if (id == V4L2_STD_SECAM_G)
+		name = "SECAM-G";
+	else if (id == V4L2_STD_SECAM_H)
+		name = "SECAM-H";
+	else if (id == V4L2_STD_SECAM_K)
+		name = "SECAM-K";
+	else if (id == V4L2_STD_SECAM_K1)
+		name = "SECAM-K1";
+	else if (id == V4L2_STD_SECAM_L)
+		name = "SECAM-L";
+	else if (id == V4L2_STD_SECAM_LC)
+		name = "SECAM-LC";
+	else
+		name = "Unknown";
 
 	return name;
 }
Signature · 1 line
Stelian Pop <[email protected]>
#1 lore raw
Quoted 8 lines · Stelian Pop

This patch replaces a switch statement using 64 bit values with the if/else equivalent in order to prevent a call __ucmpdi2 generated by some versions of gcc (verified with gcc-4.1.2 20060928):

	drivers/built-in.o: In function `v4l2_norm_to_name':
	(.text+0x71100): undefined reference to `__ucmpdi2'

It looks like there is a much better way to handle this problem here:
http://lkml.org/lkml/2006/12/17/46

Basically:
A. Fix gcc so it doesn't call in __ucmpdi2 (already fixed?) B. Link in __ucmpdi2 from libgcc
C. Write arch specific code to provide __ucmpdi2

#2 lore raw
Quoted 10 lines · Trent Piepho
Quoted 8 lines · Stelian Pop

This patch replaces a switch statement using 64 bit values with the if/else equivalent in order to prevent a call __ucmpdi2 generated by some versions of gcc (verified with gcc-4.1.2 20060928):

	drivers/built-in.o: In function `v4l2_norm_to_name':
	(.text+0x71100): undefined reference to `__ucmpdi2'
Quoted 7 lines

It looks like there is a much better way to handle this problem here:
http://lkml.org/lkml/2006/12/17/46

Basically:
A. Fix gcc so it doesn't call in __ucmpdi2 (already fixed?) B. Link in __ucmpdi2 from libgcc
C. Write arch specific code to provide __ucmpdi2

Indeed, I didn't see that thread.

Linus, just make sure you apply a fix for this problem before 2.4.20 goes final...

Thanks and sorry for the noise.

Stelian.

Signature · 1 line
Stelian Pop <[email protected]>
#3 lore raw
In reply to Stelian Pop
Quoted 21 lines · On Thu, 04 Jan 2007 12:10:14 +0100 Stelian Pop <[email protected]> wrote:

Hi,

This patch replaces a switch statement using 64 bit values with the if/else equivalent in order to prevent a call __ucmpdi2 generated by some versions of gcc (verified with gcc-4.1.2 20060928):

	drivers/built-in.o: In function `v4l2_norm_to_name':
	(.text+0x71100): undefined reference to `__ucmpdi2'

---

 drivers/media/video/v4l2-common.c |  126 ++++++++++++++++++-------------------
 1 files changed, 62 insertions(+), 64 deletions(-)
drivers/media/video/v4l2-common.c +0 -0
index 752c82c..0c3c2f6 100644

drivers/media/video/v4l2-common.c char *v4l2_norm_to_name(v4l2_std_id id) · line 94

Quoted 4 lines
drivers/media/video/v4l2-common.c +0 -1
@@ -91,70 +91,68 @@ char *v4l2_norm_to_name(v4l2_std_id id)
 {
 	char *name;
 
-	switch (id) {

It'd be simpler to just do

	switch ((unsigned int)id) {

here (with a suitable comment).

The largest value we use here is 0x02000000. Perhaps v4l2_std_id shouldn't be 64-bit?

#4 lore raw

Em Qui, 2007-01-04 às 14:48 -0800, Andrew Morton escreveu:

Quoted 31 lines
Quoted 21 lines · On Thu, 04 Jan 2007 12:10:14 +0100 Stelian Pop <[email protected]> wrote:

Hi,

This patch replaces a switch statement using 64 bit values with the if/else equivalent in order to prevent a call __ucmpdi2 generated by some versions of gcc (verified with gcc-4.1.2 20060928):

	drivers/built-in.o: In function `v4l2_norm_to_name':
	(.text+0x71100): undefined reference to `__ucmpdi2'

---

 drivers/media/video/v4l2-common.c |  126 ++++++++++++++++++-------------------
 1 files changed, 62 insertions(+), 64 deletions(-)
drivers/media/video/v4l2-common.c +0 -0
index 752c82c..0c3c2f6 100644

drivers/media/video/v4l2-common.c char *v4l2_norm_to_name(v4l2_std_id id) · line 94

Quoted 4 lines
drivers/media/video/v4l2-common.c +0 -1
@@ -91,70 +91,68 @@ char *v4l2_norm_to_name(v4l2_std_id id)
 {
 	char *name;
 
-	switch (id) {

It'd be simpler to just do

Quoted 6 lines
	switch ((unsigned int)id) {

here (with a suitable comment).

The largest value we use here is 0x02000000. Perhaps v4l2_std_id shouldn't be 64-bit?

Too late to change it to 32 bits. It is at V4L2 userspace API since kernel 2.6.0. We can, however use this approach as a workaround, with the proper documentation. I'll handle it after I return from vacations next week.

Quoted 1 line

Cheers,
Mauro.

#5 lore raw
Quoted 4 lines · On Thu, 04 Jan 2007 20:59:08 -0200 Mauro Carvalho Chehab…
Quoted 2 lines

The largest value we use here is 0x02000000. Perhaps v4l2_std_id shouldn't be 64-bit?

Too late to change it to 32 bits. It is at V4L2 userspace API since kernel 2.6.0.

You could perhaps make it 32-bit internally, and still 64-bit on the
kernel<->userspace boundary. 64-bit quantities are expensive..

Quoted 3 lines

We can, however use this approach as a workaround, with
the proper documentation. I'll handle it after I return from vacations next week.

Thanks.

#6 lore raw

Em Qui, 2007-01-04 às 15:18 -0800, Andrew Morton escreveu:

Quoted 10 lines
Quoted 4 lines · On Thu, 04 Jan 2007 20:59:08 -0200 Mauro Carvalho Chehab…
Quoted 2 lines

The largest value we use here is 0x02000000. Perhaps v4l2_std_id shouldn't be 64-bit?

Too late to change it to 32 bits. It is at V4L2 userspace API since kernel 2.6.0.

You could perhaps make it 32-bit internally, and still 64-bit on the
kernel<->userspace boundary. 64-bit quantities are expensive..

Hmm... there are some discussions currently on v4l ML about the need to add some standards to support some digital streams, like those used on webcams. Depending on the result of those discussions, we can need to use more bits. So, I think it is not worth right now to replace video std on every place it occurs.

I'm to just do the fix at v4l2-common.c.

Quoted 4 lines
Quoted 4 lines

We can, however use this approach as a workaround, with
the proper documentation. I'll handle it after I return from vacations next week.

Ok, I've wrote such patch. I should send today or tomorrow to Linus, together with other patches.

Quoted 2 lines

Thanks.

Cheers,
Mauro.

#7 lore raw

Le dimanche 07 janvier 2007 à 09:44 -0200, Mauro Carvalho Chehab a écrit :

Quoted 5 lines
Quoted 3 lines
Quoted 3 lines

We can, however use this approach as a workaround, with
the proper documentation. I'll handle it after I return from vacations next week.

Ok, I've wrote such patch. I should send today or tomorrow to Linus, together with other patches.

Hi Mauro,

The __ucmpdi2() problem is still present in 2.6.20-rc5, please (re)send your patch to Linus before 2.6.20 goes final...

Thanks.

Signature · 1 line
Stelian Pop <[email protected]>
#8 lore raw
In reply to Mauro Carvalho Chehab
Quoted 5 lines · Mauro Carvalho Chehab
Quoted 2 lines

The largest value we use here is 0x02000000. Perhaps v4l2_std_id shouldn't be 64-bit?

Too late to change it to 32 bits. It is at V4L2 userspace API since kernel 2.6.0. We can, however use this approach as a workaround, with the proper documentation.

Maybe with a BUG_ON(id > UINT_MAX) ?

Linus, do you want a patch for this or will you take the _ucmppdi2 ppc patch as per http://lkml.org/lkml/2006/12/17/46 instead ?

Thanks,

Stelian.

Signature · 1 line
Stelian Pop <[email protected]>
#9 lore raw
Quoted 7 lines
Quoted 5 lines
Quoted 2 lines

The largest value we use here is 0x02000000. Perhaps v4l2_std_id shouldn't be 64-bit?

Too late to change it to 32 bits. It is at V4L2 userspace API since kernel 2.6.0. We can, however use this approach as a workaround, with the proper documentation.

Maybe with a BUG_ON(id > UINT_MAX) ?

If the code code is like

	u64 user_id;
	u32 kernel_id = user_id;

(or different types, just showing the difference in word lengths here) it's easiest and safest to do

	BUG_ON(kernel_id != user_id);

i.e. cast back up to 64-bit, see if it's identical to the original. You won't have to worry about sign extensions or similar that way, the BUG_ON() condition expresses the actual requirement directly.

Segher