Skip to contentSkip to conversation

armv7: start.S: Fix relocation address caculation

PATCH New Jason Liu 4 participants 10 messages Dec 15, 2010 – Dec 16, 2010 Arm (32-bit)

lore
Share & tools
lore

Jason Liu

author
#0 lore raw

PATCH 1/1 armv7: start.S: Fix relocation address caculation

There will have issue if the _start not equal TEXT_BASE when enable relocation. The reason is as the followings,

The _dynsym_start_ofs and _rel_dyn_start_ofs is the offset from _start, so, need use _start address instead of _TEXT_BASE to caculate the rel dyn start and sym table address. This patch also correct the board_init_r function address caculation in relocation area. The addr of board_init_r after relocation is: _board_init_r_offs + relocation start address.

This patch also make code cleanup by removing some useless code

---
 arch/arm/cpu/armv7/start.S |    8 ++------
 1 files changed, 2 insertions(+), 6 deletions(-)
arch/arm/cpu/armv7/start.S +2 -6
index 684f2d2..08902b0 100644
@@ -171,7 +171,6 @@ stack_setup:
 	beq	clear_bss		/* skip relocation */
 #endif
 	mov	r1, r6			/* r1 <- scratch for copy_loop */
-	ldr	r2, _TEXT_BASE
 	ldr	r3, _bss_start_ofs
 	add	r2, r0, r3		/* r2 <- source end address	    */
 
@@ -185,7 +184,7 @@ copy_loop:
 	/*
 	 * fix .rel.dyn relocations
 	 */
-	ldr	r0, _TEXT_BASE		/* r0 <- Text base */
+	adr     r0, _start
 	sub	r9, r6, r0		/* r9 <- relocation offset */
 	ldr	r10, _dynsym_start_ofs	/* r10 <- sym table ofs */
 	add	r10, r10, r0		/* r10 <- sym table in FLASH */
@@ -224,7 +223,6 @@ fixnext:
 clear_bss:
 	ldr	r0, _bss_start_ofs
 	ldr	r1, _bss_end_ofs
-	ldr	r3, _TEXT_BASE		/* Text base */
 	mov	r4, r6			/* reloc addr */
 	add	r0, r0, r4
 	add	r1, r1, r4
@@ -242,9 +240,7 @@ clbss_l:str	r2, [r0]		/* clear loop...		    */
  */
 jump_2_ram:
 	ldr	r0, _board_init_r_ofs
-	adr	r1, _start
-	add	lr, r0, r1
-	add	lr, lr, r9
+	add	lr, r0, r4
 	/* setup parameters for board_init_r */
 	mov	r0, r5		/* gd_t */
 	mov	r1, r6		/* dest_addr */
Signature · 1 line
1.7.0.4
#1 lore raw

PATCH 1/1 armv7: start.S: Fix relocation address caculation

Hi Jason,

Le 15/12/2010 14:57, Jason Liu a ?crit :

Quoted 2 lines

There will have issue if the _start not equal TEXT_BASE when enable relocation.

In what case does this happen?

Amicalement,

Signature · 1 line
Albert.
#2 lore raw

PATCH 1/1 armv7: start.S: Fix relocation address caculation

Hi, Albert,

2010/12/16 Albert ARIBAUD <[email protected]>:

Quoted 7 lines

Hi Jason,

Le 15/12/2010 14:57, Jason Liu a ?crit :

Quoted 2 lines

There will have issue if the _start not equal TEXT_BASE when enable relocation.

In what case does this happen?

Some ARM SOC ROM need run the plug-in code first in IRAM and the plugin-in code need appear at the beginning of the u-boot. ROM will check the plugin-in header to do security check and run the plug-in code to init the DDR etc. In this case the _start will be not the same as TEXT_BASE.

Thanks for your comments. Cheers,

Quoted 9 lines

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

#3 lore raw

PATCH 1/1 armv7: start.S: Fix relocation address caculation

Le 16/12/2010 04:04, Jason Liu a ?crit :

Quoted 9 lines

Hi, Albert,

2010/12/16 Albert ARIBAUD<[email protected]>:

Quoted 6 lines

Hi Jason,

Le 15/12/2010 14:57, Jason Liu a ?crit :

Quoted 2 lines

There will have issue if the _start not equal TEXT_BASE when enable relocation.

Quoted 7 lines
Quoted 1 line

In what case does this happen?

Some ARM SOC ROM need run the plug-in code first in IRAM and the plugin-in code need appear at the beginning of the u-boot. ROM will check the plugin-in header to do security check and run the plug-in code to init the DDR etc. In this case the _start will be not the same as TEXT_BASE.

I still don't see why u-boot would not end up where specified.

The fact that there is a "plug-in" (I assume it's what I would call an IPL) does not change the fact that its payload (u-boot) can and will be loaded where specified, i.e. at TEXT_BASE -- and if it is loaded elsewhere, it is at a fixed address, so TEXT_BASE can be adjusted) All IPLs that I know of put their payload where specified.

Quoted 1 line

Thanks for your comments. Cheers,

Amicalement,

Signature · 1 line
Albert.
#4 lore raw

PATCH 1/1 armv7: start.S: Fix relocation address caculation

Hi, Albert,

2010/12/16 Albert ARIBAUD <[email protected]>:

Quoted 21 lines

Le 16/12/2010 04:04, Jason Liu a ?crit :

Quoted 12 lines

Hi, Albert,

2010/12/16 Albert ARIBAUD<[email protected]>:

Quoted 8 lines

Hi Jason,

Le 15/12/2010 14:57, Jason Liu a ?crit :

Quoted 3 lines

There will have issue if the _start not equal TEXT_BASE when enable relocation.

Quoted 7 lines
Quoted 1 line

In what case does this happen?

Some ARM SOC ROM need run the plug-in code first in IRAM and the plugin-in code need appear at the beginning of the u-boot. ROM will check the plugin-in header to do security check and run the plug-in code to init the DDR etc. In this case the _start will be not the same as TEXT_BASE.

Quoted 7 lines

I still don't see why u-boot would not end up where specified.

The fact that there is a "plug-in" (I assume it's what I would call an IPL) does not change the fact that its payload (u-boot) can and will be loaded where specified, i.e. at TEXT_BASE -- and if it is loaded elsewhere, it is at a fixed address, so TEXT_BASE can be adjusted) All IPLs that I know of put their payload where specified.

It's not an IPL. The layout is that as the following,

---- ----- TEXT_BASE
plug-in
---------- _start

---------- _end

No matter what you adjusted the TEXT_BASE, the _star is not equal to it.

The fix doe not affect the original functionality but just make it more flexible.

Quoted 7 lines
Quoted 2 lines

Thanks for your comments. Cheers,

Amicalement,
--
Albert.

#5 lore raw

PATCH 1/1 armv7: start.S: Fix relocation address caculation

Le 16/12/2010 10:18, Jason Liu a ?crit :

Quoted 38 lines

Hi, Albert,

2010/12/16 Albert ARIBAUD<[email protected]>:

Quoted 21 lines

Le 16/12/2010 04:04, Jason Liu a ?crit :

Quoted 12 lines

Hi, Albert,

2010/12/16 Albert ARIBAUD<[email protected]>:

Quoted 8 lines

Hi Jason,

Le 15/12/2010 14:57, Jason Liu a ?crit :

Quoted 3 lines

There will have issue if the _start not equal TEXT_BASE when enable relocation.

Quoted 7 lines
Quoted 1 line

In what case does this happen?

Some ARM SOC ROM need run the plug-in code first in IRAM and the plugin-in code need appear at the beginning of the u-boot. ROM will check the plugin-in header to do security check and run the plug-in code to init the DDR etc. In this case the _start will be not the same as TEXT_BASE.

Quoted 7 lines

I still don't see why u-boot would not end up where specified.

The fact that there is a "plug-in" (I assume it's what I would call an IPL) does not change the fact that its payload (u-boot) can and will be loaded where specified, i.e. at TEXT_BASE -- and if it is loaded elsewhere, it is at a fixed address, so TEXT_BASE can be adjusted) All IPLs that I know of put their payload where specified.

It's not an IPL. The layout is that as the following,

---- ----- TEXT_BASE
plug-in
---------- _start

Quoted 6 lines

---------- _end

No matter what you adjusted the TEXT_BASE, the _star is not equal to it.

The fix doe not affect the original functionality but just make it more flexible.

This layout is not that of u-boot for ARM; the fix thus corrects a fault not inherent to u-boot but introduced by inserting this "plug-in" where it should not be.

Why must you modify the original layout?

Also, what is this 'plug-in' if it is not an IPL?

Amicalement,

Signature · 1 line
Albert.
#6 lore raw

PATCH 1/1 armv7: start.S: Fix relocation address caculation

Dear Albert ARIBAUD,

Quoted 2 lines · In message <[email protected]> you wrote:

Why must you modify the original layout?

right!

Quoted 1 line

Also, what is this 'plug-in' if it is not an IPL?

Yes, and why must it be linked at TEXT_BASE?

Let's stop this here.

Let's wait until any code is submitted or proposed that actually needs such a feature. Then we can discuss the design and try to find a solution that makes sense.

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
History is only a confused heap of facts.
                                       -- Philip Earl of Chesterfield
#7 lore raw

PATCH 1/1 armv7: start.S: Fix relocation address caculation

Hi, Wolfgang,

2010/12/16 Wolfgang Denk <[email protected]>:

Quoted 10 lines

Dear Albert ARIBAUD,

Quoted 2 lines · In message <[email protected]> you wrote:

Why must you modify the original layout?

right!

Quoted 1 line

Also, what is this 'plug-in' if it is not an IPL?

Quoted 7 lines

Yes, and why must it be linked at TEXT_BASE?

Let's stop this here.

Let's wait until any code is submitted or proposed that actually needs such a feature. ?Then we can discuss the design and try to find a solution that makes sense.

In fact, this is indeed one bug in the code as I said in the comit log:

"the _dynsym_start_ofs and _rel_dyn_start_ofs is the offset from _start, so, need use _start address instead of _TEXT_BASE to caculate the rel dyn start and sym table address. This patch also correct the board_init_r function address caculation in relocation area. The addr of board_init_r after relocation is: _board_init_r_offs + relocation start address."

Is there any thing wrong with my patch?
The case I met just make the bug show up.

Quoted 12 lines

Best regards,

Wolfgang Denk

--
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 History is only a confused heap of facts.
? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? -- Philip Earl of Chesterfield

#8 lore raw

PATCH 1/1 armv7: start.S: Fix relocation address caculation

Dear Jason Liu,

Quoted 9 lines · In message <[email protected]> you…

In fact, this is indeed one bug in the code as I said in the comit log:

"the _dynsym_start_ofs and _rel_dyn_start_ofs is the offset from _start, so, need use _start address instead of _TEXT_BASE to caculate the rel dyn start and sym table address. This patch also correct the board_init_r function address caculation in relocation area. The addr of board_init_r after relocation is: _board_init_r_offs + relocation start address."

If there are really two different bugs, then please split your patch into to separate commits, and explain the respective problem in the commit messages.

Best regards,

Wolfgang Denk

Signature · 4 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
A modem is a baudy house.
#9 lore raw

PATCH 1/1 armv7: start.S: Fix relocation address caculation

In reply to Jason Liu

Dear Jason Liu,

Quoted 8 lines · In message <[email protected]> you…
Quoted 2 lines

In what case does this happen?

Some ARM SOC ROM need run the plug-in code first in IRAM and the plugin-in code need appear at the beginning of the u-boot. ROM will check the plugin-in header to do security check and run the plug-in code to init the DDR etc. In this case the _start will be not the same as TEXT_BASE.

So we are talking about aSoC that is not supported yet in U-Boot. Let's talk about this when such support gets added to U-Boot.

Best regards,

Wolfgang Denk

Signature · 4 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
Mistakes are often the stepping stones to utter failure.