[PATCH] ARM: OMAP: Modified omap_mux_init_signal() to take in const char *

[PATCH] ARM: OMAP: Modified omap_mux_init_signal() to take in const char *

[Tim Nordell]

If one does the following line twice with the old code:

omap_mux_init_signal("uart3_rts_sd.gpio_164", OMAP_PIN_INPUT);
omap_mux_init_signal("uart3_rts_sd.gpio_164", OMAP_PIN_INPUT);

the compiler optimizes the two const strings into one string
internally in the compiler.  The old code would modify the string
by inserting a NULL effectively making the second call become:

omap_mux_init_signal("uart3_rts_sd", OMAP_PIN_INPUT);

Since the code changes _all_ uart3_rts_sd signals over to this,
it'll cause unknown behavior, as well as making it more
difficult to track down the reason since the string
"uart3_rts_sd" by itself may not actually exist in your
normal mux initialization sequence.

Signed-off-by: Tim Nordell <tim.nordell@logicpd.com>
---
 arch/arm/mach-omap2/mux.c |   13 ++++++++-----
 arch/arm/mach-omap2/mux.h |    4 ++--
 2 files changed, 10 insertions(+), 7 deletions(-)

diff --git a/arch/arm/mach-omap2/mux.c b/arch/arm/mach-omap2/mux.c
index ab403b2..b685540 100644
--- a/arch/arm/mach-omap2/mux.c
+++ b/arch/arm/mach-omap2/mux.c
@@ -127,19 +127,21 @@ int __init omap_mux_init_gpio(int gpio, int val)
 	return 0;
 }
 
-int __init omap_mux_init_signal(char *muxname, int val)
+int __init omap_mux_init_signal(const char *muxname, int val)
 {
 	struct omap_mux_entry *e;
-	char *m0_name = NULL, *mode_name = NULL;
+	const char *m0_name = NULL, *mode_name = NULL;
 	int found = 0;
+	int m0_name_len;
 
 	mode_name = strchr(muxname, '.');
 	if (mode_name) {
-		*mode_name = '\0';
 		mode_name++;
 		m0_name = muxname;
+		m0_name_len = mode_name - muxname - 1;
 	} else {
 		mode_name = muxname;
+		m0_name_len = strlen(muxname);
 	}
 
 	list_for_each_entry(e, &muxmodes, node) {
@@ -147,7 +149,8 @@ int __init omap_mux_init_signal(char *muxname, int val)
 		char *m0_entry = m->muxnames[0];
 		int i;
 
-		if (m0_name && strcmp(m0_name, m0_entry))
+		if (m0_name && (strncmp(m0_name, m0_entry, m0_name_len)
+				|| strlen(m0_entry) != m0_name_len))
 			continue;
 
 		for (i = 0; i < OMAP_MUX_NR_MODES; i++) {
@@ -164,7 +167,7 @@ int __init omap_mux_init_signal(char *muxname, int val)
 				mux_mode = val | i;
 				printk(KERN_DEBUG "mux: Setting signal "
 					"%s.%s 0x%04x -> 0x%04x\n",
-					m0_entry, muxname, old_mode, mux_mode);
+					m0_entry, mode_cur, old_mode, mux_mode);
 				omap_mux_write(mux_mode, m->reg_offset);
 				found++;
 			}
diff --git a/arch/arm/mach-omap2/mux.h b/arch/arm/mach-omap2/mux.h
index a8e040c..55e3a21 100644
--- a/arch/arm/mach-omap2/mux.h
+++ b/arch/arm/mach-omap2/mux.h
@@ -120,7 +120,7 @@ int omap_mux_init_gpio(int gpio, int val);
  * @muxname:		Mux name in mode0_name.signal_name format
  * @val:		Options for the mux register value
  */
-int omap_mux_init_signal(char *muxname, int val);
+int omap_mux_init_signal(const char *muxname, int val);
 
 #else
 
@@ -128,7 +128,7 @@ static inline int omap_mux_init_gpio(int gpio, int val)
 {
 	return 0;
 }
-static inline int omap_mux_init_signal(char *muxname, int val)
+static inline int omap_mux_init_signal(const char *muxname, int val)
 {
 	return 0;
 }
-- 
1.7.1

[PATCH] ARM: OMAP: Modified omap_mux_init_signal() to take in const char *

[Tim Nordell]

 On 10/21/10 13:50, Tim Nordell wrote:
> If one does the following line twice with the old code:
>
> omap_mux_init_signal("uart3_rts_sd.gpio_164", OMAP_PIN_INPUT);
> omap_mux_init_signal("uart3_rts_sd.gpio_164", OMAP_PIN_INPUT);
>
> the compiler optimizes the two const strings into one string
> internally in the compiler.  The old code would modify the string
> by inserting a NULL effectively making the second call become:
>
> omap_mux_init_signal("uart3_rts_sd", OMAP_PIN_INPUT);
>
> Since the code changes _all_ uart3_rts_sd signals over to this,
> it'll cause unknown behavior, as well as making it more
> difficult to track down the reason since the string
> "uart3_rts_sd" by itself may not actually exist in your
> normal mux initialization sequence.
>

Gah.  I just realized there was a patch in the linux-omap tree to do the
same thing.

http://git.kernel.org/?p=linux/kernel/git/tmlind/linux-omap-2.6.git;a=commitdiff;h=5a3b2f7a5a79082dd3a5f2294cbd85fc3b173d98;hp=7ad0e386d46e9edff64705ab25337ad9130baf63

It looks like by inspection that there could be a couple of things wrong
with that patch however.  Namely, on the comparison of muxname to
m0_entry, if they have the same string up to mode0_len, such as
"dss_data1" and "dss_data12" it'd match there prematurely as it'd stop
comparing.  Also, a minor display issue on the printk of the new muxname.

- Tim

[PATCH] ARM: OMAP: Modified omap_mux_init_signal() to take in const char *

[Tony Lindgren]

* Tim Nordell <tim.nordell@logicpd.com> [101021 12:02]:
>  On 10/21/10 13:50, Tim Nordell wrote:
> 
> It looks like by inspection that there could be a couple of things wrong
> with that patch however.  Namely, on the comparison of muxname to
> m0_entry, if they have the same string up to mode0_len, such as
> "dss_data1" and "dss_data12" it'd match there prematurely as it'd stop
> comparing.  Also, a minor display issue on the printk of the new muxname.

Good catch, it would be nice to get those fixed too. Got any patches
for those issues?

Regards,

Tony