Skip to contentSkip to conversation

armv8: fix #if around spin-table code in start.S

PATCH In review Oded Gabbay 3 participants 6 messages Dec 27, 2016 – Jan 17, 2017 Arm (32-bit)

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 armv8: fix #if around spin-table code in start.S

Using CONFIG_IS_ENABLED() doesn't work in SPL. This patch replaces the only occurrence of CONFIG_IS_ENABLED() in start.S to a regular #if defined(). It also adds "&& !defined(CONFIG_SPL_BUILD)" to that #if statement because the spin-table code can't currently work in SPL, and the spin-table file isn't even compiled in SPL.

---
 arch/arm/cpu/armv8/start.S | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
arch/arm/cpu/armv8/start.S +1 -1
index 4f5f6d8..2f975a0 100644
@@ -99,7 +99,7 @@ save_boot_params_ret:
 	/* Processor specific initialization */
 	bl	lowlevel_init
 
-#if CONFIG_IS_ENABLED(ARMV8_SPIN_TABLE)
+#if defined(CONFIG_ARMV8_SPIN_TABLE) && !defined(CONFIG_SPL_BUILD)
 	branch_if_master x0, x1, master_cpu
 	b	spin_table_secondary_jump
 	/* never return */
Signature · 1 line
2.7.4
#1 lore raw

PATCH 1/1 armv8: fix #if around spin-table code in start.S

Quoted 16 lines · Oded Gabbay

Using CONFIG_IS_ENABLED() doesn't work in SPL. This patch replaces the only occurrence of CONFIG_IS_ENABLED() in start.S to a regular #if defined(). It also adds "&& !defined(CONFIG_SPL_BUILD)" to that #if statement because the spin-table code can't currently work in SPL, and the spin-table file isn't even compiled in SPL.

---
 arch/arm/cpu/armv8/start.S | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
arch/arm/cpu/armv8/start.S +0 -0
index 4f5f6d8..2f975a0 100644

arch/arm/cpu/armv8/start.S save_boot_params_ret: · line 105

Quoted 8 lines
arch/arm/cpu/armv8/start.S +1 -1
@@ -99,7 +99,7 @@ save_boot_params_ret:
 	/* Processor specific initialization */
 	bl	lowlevel_init
 
-#if CONFIG_IS_ENABLED(ARMV8_SPIN_TABLE)
+#if defined(CONFIG_ARMV8_SPIN_TABLE) && !defined(CONFIG_SPL_BUILD)
 	branch_if_master x0, x1, master_cpu
 	b	spin_table_secondary_jump
 	/* never return */

I don't get it, sorry. Looking at include/linux/kconfig.h:
CONFIG_IS_ENABLED(ARMV8_SPIN_TABLE) is 1 if CONFIG_ARMV8_SPIN_TABLE and CONFIG_SPL_BUILD is undefined. CONFIG_IS_ENABLED(ARMV8_SPIN_TABLE) is always 0 since CONFIG_SPL_ARMV8_SPIN_TABLE is not a symbol. And since we don't link the spin table objects in outside of SPL, we wouldn't want to try and call those functions.

So, what's the bug exactly? Just a lack-of-clarity by using CONFIG_IS_ENABLED(ARMV8_SPIN_TABLE) vs 'defined(CONFIG_ARMV8_SPIN_TABLE) && !defined(CONFIG_SPL_BUILD)' ? Thanks!

Signature · 8 lines
Tom
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 819 bytes
Desc: Digital signature
URL: <http://lists.denx.de/pipermail/u-boot/attachments/20170110/6c24dfb7/attachment.sig>
#2 lore raw

PATCH 1/1 armv8: fix #if around spin-table code in start.S

Quoted 22 lines · Tom Rini
Quoted 19 lines · Oded Gabbay

Using CONFIG_IS_ENABLED() doesn't work in SPL. This patch replaces the only occurrence of CONFIG_IS_ENABLED() in start.S to a regular #if defined(). It also adds "&& !defined(CONFIG_SPL_BUILD)" to that #if statement because the spin-table code can't currently work in SPL, and the spin-table file isn't even compiled in SPL.

---
 arch/arm/cpu/armv8/start.S | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
arch/arm/cpu/armv8/start.S +0 -0
index 4f5f6d8..2f975a0 100644
@@ -99,7 +99,7 @@ save_boot_params_ret:
      /* Processor specific initialization */
      bl      lowlevel_init
 
Quoted 13 lines

arch/arm/cpu/armv8/start.S save_boot_params_ret: · line 105

Quoted 5 lines
arch/arm/cpu/armv8/start.S +1 -1
-#if CONFIG_IS_ENABLED(ARMV8_SPIN_TABLE)
+#if defined(CONFIG_ARMV8_SPIN_TABLE) && !defined(CONFIG_SPL_BUILD)
      branch_if_master x0, x1, master_cpu
      b       spin_table_secondary_jump
      /* never return */

I don't get it, sorry. Looking at include/linux/kconfig.h:
CONFIG_IS_ENABLED(ARMV8_SPIN_TABLE) is 1 if CONFIG_ARMV8_SPIN_TABLE and CONFIG_SPL_BUILD is undefined. CONFIG_IS_ENABLED(ARMV8_SPIN_TABLE) is always 0 since CONFIG_SPL_ARMV8_SPIN_TABLE is not a symbol. And since we don't link the spin table objects in outside of SPL, we wouldn't want to try and call those functions.

hmm, I'm really confused. For some reason, I can't recreate the original problem that caused me to create this patch. I also agree with your analysis. I must conclude that my original problem was due to a different reason.
Sorry for the waste of time.

Quoted 4 lines

So, what's the bug exactly? Just a lack-of-clarity by using CONFIG_IS_ENABLED(ARMV8_SPIN_TABLE) vs 'defined(CONFIG_ARMV8_SPIN_TABLE) && !defined(CONFIG_SPL_BUILD)' ? Thanks!

Although my original purpose was to fix an actual bug (which doesn't exists), I do think there is a clarity issue here, which this patch addresses. You need to go to kconfig.h, read the comments there to understand how CONFIG_IS_ENABLED is working with SPL, which is more tiresome than just doing straight #ifdef. It is definitely more confusing for a newbee.

In addition, this patch makes the code more consistent, because all other configuration checks in start.S use a straight #ifdef and not CONFIG_IS_ENABLED.

But since this patch doesn't fix an actual bug, I guess you can drop it if you don't want it.
Thanks,
Oded

Quoted 3 lines

--
Tom

#3 lore raw
In reply to Oded Gabbay
Quoted 7 lines · Oded Gabbay

Using CONFIG_IS_ENABLED() doesn't work in SPL. This patch replaces the only occurrence of CONFIG_IS_ENABLED() in start.S to a regular #if defined(). It also adds "&& !defined(CONFIG_SPL_BUILD)" to that #if statement because the spin-table code can't currently work in SPL, and the spin-table file isn't even compiled in SPL.

Applied to u-boot/master, thanks!

Signature · 8 lines
Tom
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 819 bytes
Desc: Digital signature
URL: <http://lists.denx.de/pipermail/u-boot/attachments/20170115/760bc578/attachment.sig>
#4 lore raw

Hi Tom.

2017-01-16 3:29 GMT+09:00 Tom Rini <[email protected]>:

Quoted 8 lines · Oded Gabbay
Quoted 7 lines

Using CONFIG_IS_ENABLED() doesn't work in SPL. This patch replaces the only occurrence of CONFIG_IS_ENABLED() in start.S to a regular #if defined(). It also adds "&& !defined(CONFIG_SPL_BUILD)" to that #if statement because the spin-table code can't currently work in SPL, and the spin-table file isn't even compiled in SPL.

Quoted 7 lines
Quoted 1 line

Applied to u-boot/master, thanks!

--
Tom

I had not noticed this patch until it was applied.

At least, the statement in the git-log
"Using CONFIG_IS_ENABLED() doesn't work in SPL" is wrong. So, when I saw the git history today, I wondered what was going on. Then, I found this discussion in the ML.

It does not matter to either apply or discard this patch because it is a matter of taste.

If you decide to apply it,
the git-log should have been replaced with Oded's comment:

--------
You need to go to kconfig.h, read the comments there to understand how CONFIG_IS_ENABLED is working with SPL, which is more tiresome than just doing straight #ifdef. It is definitely more confusing for a newbee.

In addition, this patch makes the code more consistent, because all other configuration checks in start.S use a straight #ifdef and not CONFIG_IS_ENABLED.
----------

It is too late this time, but please take care of it next time.

Signature · 2 lines
Best Regards
Masahiro Yamada
#5 lore raw
Quoted 41 lines · Masahiro Yamada

Hi Tom.

2017-01-16 3:29 GMT+09:00 Tom Rini <[email protected]>:

Quoted 8 lines · Oded Gabbay
Quoted 7 lines

Using CONFIG_IS_ENABLED() doesn't work in SPL. This patch replaces the only occurrence of CONFIG_IS_ENABLED() in start.S to a regular #if defined(). It also adds "&& !defined(CONFIG_SPL_BUILD)" to that #if statement because the spin-table code can't currently work in SPL, and the spin-table file isn't even compiled in SPL.

Quoted 7 lines
Quoted 1 line

Applied to u-boot/master, thanks!

--
Tom

I had not noticed this patch until it was applied.

At least, the statement in the git-log
"Using CONFIG_IS_ENABLED() doesn't work in SPL" is wrong. So, when I saw the git history today, I wondered what was going on. Then, I found this discussion in the ML.

It does not matter to either apply or discard this patch because it is a matter of taste.

If you decide to apply it,
the git-log should have been replaced with Oded's comment:

--------
You need to go to kconfig.h, read the comments there to understand how CONFIG_IS_ENABLED is working with SPL, which is more tiresome than just doing straight #ifdef. It is definitely more confusing for a newbee.

Quoted 7 lines

In addition, this patch makes the code more consistent, because all other configuration checks in start.S use a straight #ifdef and not CONFIG_IS_ENABLED.
----------

It is too late this time, but please take care of it next time.

Yeah, I should have asked for a v2 with a different commit message.

Signature · 8 lines
Tom
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 819 bytes
Desc: Digital signature
URL: <http://lists.denx.de/pipermail/u-boot/attachments/20170116/db6284f1/attachment.sig>