Patchwork [U-Boot] arm1136: timer: Replace bss variable by gd

login
register
mail settings
Submitter Heiko Schocher
Date Dec. 10, 2010, 8:01 a.m.
Message ID <1291968075-29004-1-git-send-email-hs@denx.de>
Download mbox | patch
Permalink /patch/75035/
State Accepted
Commit c44bf4e8ac2610d650bad441a1dc6a617ae98a4f
Delegated to: Albert ARIBAUD
Headers show

Comments

Heiko Schocher - Dec. 10, 2010, 8:01 a.m.
Reuse the gd->tbl value for timestamp and add gd->lastinc for lastinc bss
values in the arm1136 timer driver for mx31 and omap24xx

The usage of bss values in drivers before initialisation of bss is forbidden.
In that special case some data in .rel.dyn gets corrupted.

This patch is similiar to the patch Dirk Behme posted
for the armv7/omap-common/timer.c

Tested on the mx31 based qong board

Signed-off-by: Heiko Schocher <hs@denx.de>
cc: Albert ARIBAUD <albert.aribaud@free.fr>
---
 arch/arm/cpu/arm1136/mx31/timer.c     |   19 +++++++++----------
 arch/arm/cpu/arm1136/omap24xx/timer.c |   19 +++++++++----------
 2 files changed, 18 insertions(+), 20 deletions(-)
Albert ARIBAUD - Dec. 10, 2010, 8:09 a.m.
Le 10/12/2010 09:01, Heiko Schocher a écrit :
> Reuse the gd->tbl value for timestamp and add gd->lastinc for lastinc bss
> values in the arm1136 timer driver for mx31 and omap24xx
>
> The usage of bss values in drivers before initialisation of bss is forbidden.
> In that special case some data in .rel.dyn gets corrupted.
>
> This patch is similiar to the patch Dirk Behme posted
> for the armv7/omap-common/timer.c
>
> Tested on the mx31 based qong board
>
> Signed-off-by: Heiko Schocher<hs@denx.de>
> cc: Albert ARIBAUD<albert.aribaud@free.fr>

FWIW,

Acked-by: Albert ARIBAUD <albert.aribaud@free.fr>

Amicalement,
Wolfgang Denk - Dec. 17, 2010, 8:13 p.m.
Dear Albert & Heiko,

In message <4D01E031.3090100@free.fr> you wrote:
>
> > Reuse the gd->tbl value for timestamp and add gd->lastinc for lastinc bss
> > values in the arm1136 timer driver for mx31 and omap24xx
> >
> > The usage of bss values in drivers before initialisation of bss is forbid> den.
> > In that special case some data in .rel.dyn gets corrupted.
> >
> > This patch is similiar to the patch Dirk Behme posted
> > for the armv7/omap-common/timer.c
> >
> > Tested on the mx31 based qong board
> >
> > Signed-off-by: Heiko Schocher<hs@denx.de>
> > cc: Albert ARIBAUD<albert.aribaud@free.fr>
>
> FWIW,
>
> Acked-by: Albert ARIBAUD <albert.aribaud@free.fr>

Should I pull this into v2010.12, or should we rather wait for amore
general solution and fix it in the next release?

Best regards,

Wolfgang Denk
Heiko Schocher - Dec. 19, 2010, 8:45 a.m.
Hello Wolfgang,

Wolfgang Denk wrote:
> Dear Albert & Heiko,
> 
> In message <4D01E031.3090100@free.fr> you wrote:
>>> Reuse the gd->tbl value for timestamp and add gd->lastinc for lastinc bss
>>> values in the arm1136 timer driver for mx31 and omap24xx
>>>
>>> The usage of bss values in drivers before initialisation of bss is forbid> den.
>>> In that special case some data in .rel.dyn gets corrupted.
>>>
>>> This patch is similiar to the patch Dirk Behme posted
>>> for the armv7/omap-common/timer.c
>>>
>>> Tested on the mx31 based qong board
>>>
>>> Signed-off-by: Heiko Schocher<hs@denx.de>
>>> cc: Albert ARIBAUD<albert.aribaud@free.fr>
>> FWIW,
>>
>> Acked-by: Albert ARIBAUD <albert.aribaud@free.fr>
> 
> Should I pull this into v2010.12, or should we rather wait for amore
> general solution and fix it in the next release?

I vote for pulling this into the v2010.12 release, beause it is a
bugfix and we get a working code with elf relocation(conceding the
patch is not a perfekt solution, but working)

bye,
Heiko
Albert ARIBAUD - Jan. 20, 2011, 8:34 p.m.
Le 19/12/2010 09:45, Heiko Schocher a écrit :
> Hello Wolfgang,
>
> Wolfgang Denk wrote:
>> Dear Albert&  Heiko,
>>
>> In message<4D01E031.3090100@free.fr>  you wrote:
>>>> Reuse the gd->tbl value for timestamp and add gd->lastinc for lastinc bss
>>>> values in the arm1136 timer driver for mx31 and omap24xx
>>>>
>>>> The usage of bss values in drivers before initialisation of bss is forbid>  den.
>>>> In that special case some data in .rel.dyn gets corrupted.
>>>>
>>>> This patch is similiar to the patch Dirk Behme posted
>>>> for the armv7/omap-common/timer.c
>>>>
>>>> Tested on the mx31 based qong board
>>>>
>>>> Signed-off-by: Heiko Schocher<hs@denx.de>
>>>> cc: Albert ARIBAUD<albert.aribaud@free.fr>
>>> FWIW,
>>>
>>> Acked-by: Albert ARIBAUD<albert.aribaud@free.fr>
>>
>> Should I pull this into v2010.12, or should we rather wait for amore
>> general solution and fix it in the next release?
>
> I vote for pulling this into the v2010.12 release, beause it is a
> bugfix and we get a working code with elf relocation(conceding the
> patch is not a perfekt solution, but working)
>
> bye,
> Heiko

Applied to u-boot-arm, thanks and apologies for the late action.

Amicalement,

Patch

diff --git a/arch/arm/cpu/arm1136/mx31/timer.c b/arch/arm/cpu/arm1136/mx31/timer.c
index b8848c4..f6be3b9 100644
--- a/arch/arm/cpu/arm1136/mx31/timer.c
+++ b/arch/arm/cpu/arm1136/mx31/timer.c
@@ -39,8 +39,7 @@ 
 #define GPTCR_CLKSOURCE_32	(4 << 6)	/* Clock source		*/
 #define GPTCR_TEN		1		/* Timer enable		*/
 
-static ulong timestamp;
-static ulong lastinc;
+DECLARE_GLOBAL_DATA_PTR;
 
 /* "time" is measured in 1 / CONFIG_SYS_HZ seconds, "tick" is internal timer period */
 #ifdef CONFIG_MX31_TIMER_HIGH_PRECISION
@@ -108,8 +107,8 @@  int timer_init (void)
 void reset_timer_masked (void)
 {
 	/* reset time */
-	lastinc = GPTCNT; /* capture current incrementer value time */
-	timestamp = 0; /* start "advancing" time stamp from 0 */
+	gd->lastinc = GPTCNT; /* capture current incrementer value time */
+	gd->tbl = 0; /* start "advancing" time stamp from 0 */
 }
 
 void reset_timer(void)
@@ -121,13 +120,13 @@  unsigned long long get_ticks (void)
 {
 	ulong now = GPTCNT; /* current tick value */
 
-	if (now >= lastinc)	/* normal mode (non roll) */
+	if (now >= gd->lastinc)	/* normal mode (non roll) */
 		/* move stamp forward with absolut diff ticks */
-		timestamp += (now - lastinc);
+		gd->tbl += (now - gd->lastinc);
 	else			/* we have rollover of incrementer */
-		timestamp += (0xFFFFFFFF - lastinc) + now;
-	lastinc = now;
-	return timestamp;
+		gd->tbl += (0xFFFFFFFF - gd->lastinc) + now;
+	gd->lastinc = now;
+	return gd->tbl;
 }
 
 ulong get_timer_masked (void)
@@ -148,7 +147,7 @@  ulong get_timer (ulong base)
 
 void set_timer (ulong t)
 {
-	timestamp = time_to_tick(t);
+	gd->tbl = time_to_tick(t);
 }
 
 /* delay x useconds AND preserve advance timestamp value */
diff --git a/arch/arm/cpu/arm1136/omap24xx/timer.c b/arch/arm/cpu/arm1136/omap24xx/timer.c
index 68fe1b2..228ceba 100644
--- a/arch/arm/cpu/arm1136/omap24xx/timer.c
+++ b/arch/arm/cpu/arm1136/omap24xx/timer.c
@@ -39,8 +39,7 @@ 
 /* macro to read the 32 bit timer */
 #define READ_TIMER (*((volatile ulong *)(CONFIG_SYS_TIMERBASE+TCRR)))
 
-static ulong timestamp;
-static ulong lastinc;
+DECLARE_GLOBAL_DATA_PTR;
 
 int timer_init (void)
 {
@@ -70,7 +69,7 @@  ulong get_timer (ulong base)
 
 void set_timer (ulong t)
 {
-	timestamp = t;
+	gd->tbl	= t;
 }
 
 /* delay x useconds AND preserve advance timestamp value */
@@ -99,20 +98,20 @@  void __udelay (unsigned long usec)
 void reset_timer_masked (void)
 {
 	/* reset time */
-	lastinc = READ_TIMER;		/* capture current incrementer value time */
-	timestamp = 0;			/* start "advancing" time stamp from 0 */
+	gd->lastinc = READ_TIMER;	/* capture current incrementer value time */
+	gd->tbl = 0;			/* start "advancing" time stamp from 0 */
 }
 
 ulong get_timer_masked (void)
 {
 	ulong now = READ_TIMER;		/* current tick value */
 
-	if (now >= lastinc)		/* normal mode (non roll) */
-		timestamp += (now - lastinc); /* move stamp fordward with absoulte diff ticks */
+	if (now >= gd->lastinc)		/* normal mode (non roll) */
+		gd->tbl += (now - gd->lastinc); /* move stamp fordward with absoulte diff ticks */
 	else				/* we have rollover of incrementer */
-		timestamp += (0xFFFFFFFF - lastinc) + now;
-	lastinc = now;
-	return timestamp;
+		gd->tbl += (0xFFFFFFFF - gd->lastinc) + now;
+	gd->lastinc = now;
+	return gd->tbl;
 }
 
 /* waits specified delay value and resets timestamp */