Skip to contentSkip to conversation

ARM Cortex A8: Move OMAP3 specific reset handler to OMAP3 code

PATCH New Dirk Behme 4 participants 7 messages May 30, 2009 – Jun 2, 2009 Arm/actions semi architecture

lore
Share & tools
lore

Dirk Behme

author
#0 lore raw

PATCH 1/1 ARM Cortex A8: Move OMAP3 specific reset handler to OMAP3 code

Reset is SoC specific and not ARM Cortex A8 generic. Move it from generic code to OMAP3 SoC specific file.

CC "Kim, Heung Jun" <[email protected]>

---

This patches fixes the second issue found by riverful in

http://lists.denx.de/pipermail/u-boot/2009-May/053433.html

The first issue is fixed by

http://lists.denx.de/pipermail/u-boot/2009-May/053444.html

 cpu/arm_cortexa8/omap3/lowlevel_init.S |   12 ++++++++++++
 cpu/arm_cortexa8/start.S               |   14 --------------
 2 files changed, 12 insertions(+), 14 deletions(-)
cpu/arm_cortexa8/omap3/lowlevel_init.S +12 -0
Index: u-boot-arm/cpu/arm_cortexa8/omap3/lowlevel_init.S
===================================================================
@@ -181,6 +181,18 @@ lowlevel_init:
 	/* back to arch calling code */
 	mov	pc, lr
 
+.global reset_cpu
+reset_cpu:
+	ldr	r1, rstctl			@ get addr for global reset
+						@ reg
+	mov	r3, #0x2			@ full reset pll + mpu
+	str	r3, [r1]			@ force reset
+	mov	r0, r0
+_loop_forever:
+	b	_loop_forever
+rstctl:
+	.word	PRM_RSTCTRL
+
 	/* the literal pools origin */
 	.ltorg
 
cpu/arm_cortexa8/start.S +0 -14
Index: u-boot-arm/cpu/arm_cortexa8/start.S
===================================================================
@@ -500,17 +500,3 @@ finished_inval:
 						@ but we compile with armv5
 
 	ldmfd	r13!, {r0 - r5, r7, r9 - r12, pc}
-
-
-	.align	5
-.global reset_cpu
-reset_cpu:
-	ldr	r1, rstctl			@ get addr for global reset
-						@ reg
-	mov	r3, #0x2			@ full reset pll + mpu
-	str	r3, [r1]			@ force reset
-	mov	r0, r0
-_loop_forever:
-	b	_loop_forever
-rstctl:
-	.word	PRM_RSTCTRL
#1 lore raw

PATCH 1/1 ARM Cortex A8: Move OMAP3 specific reset handler to OMAP3 code

Quoted 30 lines · Dirk Behme

Reset is SoC specific and not ARM Cortex A8 generic. Move it from generic code to OMAP3 SoC specific file.

CC "Kim, Heung Jun" <[email protected]>

---

This patches fixes the second issue found by riverful in

http://lists.denx.de/pipermail/u-boot/2009-May/053433.html

The first issue is fixed by

http://lists.denx.de/pipermail/u-boot/2009-May/053444.html

 cpu/arm_cortexa8/omap3/lowlevel_init.S |   12 ++++++++++++
 cpu/arm_cortexa8/start.S               |   14 --------------
 2 files changed, 12 insertions(+), 14 deletions(-)

Index: u-boot-arm/cpu/arm_cortexa8/omap3/lowlevel_init.S
===================================================================

cpu/arm_cortexa8/omap3/lowlevel_init.S +2 -0
@@ -181,6 +181,18 @@ lowlevel_init:
 	/* back to arch calling code */
 	mov	pc, lr
 
+.global reset_cpu
+reset_cpu:

cpu/arm_cortexa8/omap3/lowlevel_init.S lowlevel_init: · line 195

Quoted 10 lines
cpu/arm_cortexa8/omap3/lowlevel_init.S +10 -0
+	ldr	r1, rstctl			@ get addr for global reset
+						@ reg
+	mov	r3, #0x2			@ full reset pll + mpu
+	str	r3, [r1]			@ force reset
+	mov	r0, r0
+_loop_forever:
+	b	_loop_forever
+rstctl:
+	.word	PRM_RSTCTRL
+

please move this to reset.S other wise fine

Best Regards,
J.

author
#2 lore raw

PATCH 1/1 ARM Cortex A8: Move OMAP3 specific reset handler to OMAP3 code

Dear Jean-Christophe,

Quoted 29 lines · Jean-Christophe PLAGNIOL-VILLARD
Quoted 28 lines · Dirk Behme

Reset is SoC specific and not ARM Cortex A8 generic. Move it from generic code to OMAP3 SoC specific file.

CC "Kim, Heung Jun" <[email protected]>

---

This patches fixes the second issue found by riverful in

http://lists.denx.de/pipermail/u-boot/2009-May/053433.html

The first issue is fixed by

http://lists.denx.de/pipermail/u-boot/2009-May/053444.html

 cpu/arm_cortexa8/omap3/lowlevel_init.S |   12 ++++++++++++
 cpu/arm_cortexa8/start.S               |   14 --------------
 2 files changed, 12 insertions(+), 14 deletions(-)

Index: u-boot-arm/cpu/arm_cortexa8/omap3/lowlevel_init.S
===================================================================

cpu/arm_cortexa8/omap3/lowlevel_init.S +0 -0
@@ -181,6 +181,18 @@ lowlevel_init:
 	/* back to arch calling code */
 	mov	pc, lr
 
Quoted 13 lines

cpu/arm_cortexa8/omap3/lowlevel_init.S lowlevel_init: · line 195

Quoted 12 lines
cpu/arm_cortexa8/omap3/lowlevel_init.S +12 -0
+.global reset_cpu
+reset_cpu:
+	ldr	r1, rstctl			@ get addr for global reset
+						@ reg
+	mov	r3, #0x2			@ full reset pll + mpu
+	str	r3, [r1]			@ force reset
+	mov	r0, r0
+_loop_forever:
+	b	_loop_forever
+rstctl:
+	.word	PRM_RSTCTRL
+

please move this to reset.S other wise fine

Most probably your idea is that each file should only contain functionality which fits 100% (120%?) what the file name implies (?). While from general point of view this is correct, it makes no sense to create new files again and again just to follow this rule. We already created a cache.c on your request, now you request a new file reset.S for ~5 assembly lines. This new file would contain more comments (e.g. GPL header) than useful code.

So while in general case having file names reflecting more or less the functionality in these files, in this case it doesn't make sense. It doesn't make sense to pollute the source tree with a new file containing ~5 assembly lines just to make your rules apply. For such small code, re-using existing file is the better way to go. So NACK in this case.

Best regards

Dirk

#3 lore raw

PATCH 1/1 ARM Cortex A8: Move OMAP3 specific reset handler to OMAP3 code

Dear Jean & Dirk,

Jean's opinion seems that file naming & func naming must match for soruce maintaining, & definitely I agree with that. Moreover it's a very first time to implement the new arm_cortexa8 code for u-boot source. So, the matching works is needed.

On the other hand, Dirk's opinion seems that it dosen't make sense in this case, if there is many files when the developer make the code. I think that too, especially when the code is a little simple like the boot code.
The u-boot's architecture & SoC directory tree is already brilliant, so it's alright we just look at this directories what arch & SoC ever we develop.

So, my opinions is re-using the files seems to be right in this case. the Dirk's patch code is not difficult to understand & use. After the code polluted cause of not-matching the file & func naming, I believe that the developers seem be able to fix this issue easily, like my case.

BTW, when am I able to re-update my new patch??
I wanna do that :)

Best Regards,
riverful

2009/6/1 Dirk Behme <[email protected]>:

Quoted 70 lines

Dear Jean-Christophe,

Quoted 29 lines · Jean-Christophe PLAGNIOL-VILLARD
Quoted 28 lines · Dirk Behme

Reset is SoC specific and not ARM Cortex A8 generic. Move it from generic code to OMAP3 SoC specific file.

CC "Kim, Heung Jun" <[email protected]>

---

This patches fixes the second issue found by riverful in

http://lists.denx.de/pipermail/u-boot/2009-May/053433.html

The first issue is fixed by

http://lists.denx.de/pipermail/u-boot/2009-May/053444.html

?cpu/arm_cortexa8/omap3/lowlevel_init.S | ? 12 ++++++++++++ ?cpu/arm_cortexa8/start.S ? ? ? ? ? ? ? | ? 14 -------------- ?2 files changed, 12 insertions(+), 14 deletions(-)

Index: u-boot-arm/cpu/arm_cortexa8/omap3/lowlevel_init.S
===================================================================

cpu/arm_cortexa8/omap3/lowlevel_init.S +0 -0
@@ -181,6 +181,18 @@ lowlevel_init:

? ? ?/* back to arch calling code */
? ? ?mov ? ? pc, lr

Quoted 13 lines

cpu/arm_cortexa8/omap3/lowlevel_init.S

Quoted 12 lines

+.global reset_cpu

cpu/arm_cortexa8/omap3/lowlevel_init.S +11 -0
+reset_cpu:
+ ? ?ldr ? ? r1, rstctl ? ? ? ? ? ? ? ? ? ? ?@ get addr for global reset
+ ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ?@ reg
+ ? ?mov ? ? r3, #0x2 ? ? ? ? ? ? ? ? ? ? ? ?@ full reset pll + mpu
+ ? ?str ? ? r3, [r1] ? ? ? ? ? ? ? ? ? ? ? ?@ force reset
+ ? ?mov ? ? r0, r0
+_loop_forever:
+ ? ?b ? ? ? _loop_forever
+rstctl:
+ ? ?.word ? PRM_RSTCTRL
+

please move this to reset.S other wise fine

Most probably your idea is that each file should only contain functionality which fits 100% (120%?) what the file name implies (?). While from general point of view this is correct, it makes no sense to create new files again and again just to follow this rule. We already created a cache.c on your request, now you request a new file reset.S for ~5 assembly lines. This new file would contain more comments (e.g. GPL header) than useful code.

So while in general case having file names reflecting more or less the functionality in these files, in this case it doesn't make sense. It doesn't make sense to pollute the source tree with a new file containing ~5 assembly lines just to make your rules apply. For such small code, re-using existing file is the better way to go. So NACK in this case.

Best regards

Dirk

_______________________________________________
U-Boot mailing list
U-Boot at lists.denx.de
http://lists.denx.de/mailman/listinfo/u-boot

#4 lore raw

PATCH 1/1 ARM Cortex A8: Move OMAP3 specific reset handler to OMAP3 code

Dear Riverful,

Quoted 3 lines · In message <[email protected]> you…

Jean's opinion seems that file naming & func naming must match for soruce maintaining, & definitely I agree with that.

I do not agree.

Yes, they _should_ match. That means we should try to acchieve this whenever it makes sense.

But there is no strict must that we need to create a new source file for each and every function. Maintaining tons of tiny files does not make much sense either.

In this case, I agree with Dirk that splitting of a new source file for a small (10 lines or so including comments) function is overkill.

Jean-Christiphe sent a note about this code, and Dirk probvided a reasonable explanation why the code was written as is.

I think that should be enough in this case. From my point of view, the code can go in as is.

Best regards,

Wolfgang Denk

Signature · 5 lines
DENX Software Engineering GmbH,     MD: Wolfgang Denk & Detlev Zundel
HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany
Phone: (+49)-8142-66989-10 Fax: (+49)-8142-66989-80 Email: wd at denx.de
"The algorithm to do that is extremely nasty. You might want  to  mug
someone with it."                   - M. Devine, Computer Science 340
#5 lore raw

PATCH 1/1 ARM Cortex A8: Move OMAP3 specific reset handler to OMAP3 code

In reply to Dirk Behme
Quoted 45 lines · Dirk Behme

Dear Jean-Christophe,

Quoted 33 lines · Jean-Christophe PLAGNIOL-VILLARD
Quoted 32 lines · Dirk Behme

Reset is SoC specific and not ARM Cortex A8 generic. Move it from generic code to OMAP3 SoC specific file.

CC "Kim, Heung Jun" <[email protected]>

---

This patches fixes the second issue found by riverful in

http://lists.denx.de/pipermail/u-boot/2009-May/053433.html

The first issue is fixed by

http://lists.denx.de/pipermail/u-boot/2009-May/053444.html

 cpu/arm_cortexa8/omap3/lowlevel_init.S |   12 ++++++++++++
 cpu/arm_cortexa8/start.S               |   14 --------------
 2 files changed, 12 insertions(+), 14 deletions(-)

Index: u-boot-arm/cpu/arm_cortexa8/omap3/lowlevel_init.S
===================================================================

cpu/arm_cortexa8/omap3/lowlevel_init.S +4 -0
@@ -181,6 +181,18 @@ lowlevel_init:
 	/* back to arch calling code */
 	mov	pc, lr
 +.global reset_cpu
+reset_cpu:
+	ldr	r1, rstctl			@ get addr for global reset
+						@ reg
+	mov	r3, #0x2			@ full reset pll + mpu
Quoted 8 lines

cpu/arm_cortexa8/omap3/lowlevel_init.S lowlevel_init: · line 195

Quoted 7 lines
cpu/arm_cortexa8/omap3/lowlevel_init.S +7 -0
+	str	r3, [r1]			@ force reset
+	mov	r0, r0
+_loop_forever:
+	b	_loop_forever
+rstctl:
+	.word	PRM_RSTCTRL
+

please move this to reset.S other wise fine

Quoted 7 lines

Most probably your idea is that each file should only contain functionality which fits 100% (120%?) what the file name implies (?). While from general point of view this is correct, it makes no sense to create new files again and again just to follow this rule. We already created a cache.c on your request, now you request a new file reset.S for ~5 assembly lines. This new file would contain more comments (e.g. GPL header) than useful code.

the idea is different here
I want to have only code in lowlevel_init.S that can be disable by CONFIG_SKIP_LOWLEVEL_INIT and do it via Makefile

Best Regards,
J.

author
#6 lore raw

PATCH 1/1 ARM Cortex A8: Move OMAP3 specific reset handler to OMAP3 code

Dear Jean-Christophe,

Quoted 52 lines · Jean-Christophe PLAGNIOL-VILLARD
Quoted 44 lines · Dirk Behme

Dear Jean-Christophe,

Quoted 33 lines · Jean-Christophe PLAGNIOL-VILLARD
Quoted 32 lines · Dirk Behme

Reset is SoC specific and not ARM Cortex A8 generic. Move it from generic code to OMAP3 SoC specific file.

CC "Kim, Heung Jun" <[email protected]>

---

This patches fixes the second issue found by riverful in

http://lists.denx.de/pipermail/u-boot/2009-May/053433.html

The first issue is fixed by

http://lists.denx.de/pipermail/u-boot/2009-May/053444.html

 cpu/arm_cortexa8/omap3/lowlevel_init.S |   12 ++++++++++++
 cpu/arm_cortexa8/start.S               |   14 --------------
 2 files changed, 12 insertions(+), 14 deletions(-)

Index: u-boot-arm/cpu/arm_cortexa8/omap3/lowlevel_init.S
===================================================================

cpu/arm_cortexa8/omap3/lowlevel_init.S +4 -0
@@ -181,6 +181,18 @@ lowlevel_init:
 	/* back to arch calling code */
 	mov	pc, lr
 +.global reset_cpu
+reset_cpu:
+	ldr	r1, rstctl			@ get addr for global reset
+						@ reg
+	mov	r3, #0x2			@ full reset pll + mpu
Quoted 8 lines

cpu/arm_cortexa8/omap3/lowlevel_init.S lowlevel_init: · line 195

Quoted 7 lines
cpu/arm_cortexa8/omap3/lowlevel_init.S +7 -0
+	str	r3, [r1]			@ force reset
+	mov	r0, r0
+_loop_forever:
+	b	_loop_forever
+rstctl:
+	.word	PRM_RSTCTRL
+

please move this to reset.S other wise fine

Quoted 7 lines

Most probably your idea is that each file should only contain functionality which fits 100% (120%?) what the file name implies (?). While from general point of view this is correct, it makes no sense to create new files again and again just to follow this rule. We already created a cache.c on your request, now you request a new file reset.S for ~5 assembly lines. This new file would contain more comments (e.g. GPL header) than useful code.

Quoted 3 lines

the idea is different here
I want to have only code in lowlevel_init.S that can be disable by CONFIG_SKIP_LOWLEVEL_INIT and do it via Makefile

Looking at recent OMAP3 lowlevel_init.S most probably some other stuff has to be moved to make this work, too. So for the moment, the cleanest way is to move above reset_cpu to low_levelinit.S. And then later, after thorough investigation and testing, move the stuff needed for your idea to an appropriate place. This move will be consistent then and will avoid polluting source tree with unnecessary files until then.

So let's do it in two steps:

a) Now, move reset_cpu to lowlevel_init.S so that Riverful can go on with his work

b) Later, move everything necessary in one consistent patch set while you implement your "CONFIG_SKIP_LOWLEVEL_INIT via Makefile" idea

Best regards

Dirk