arm: mvebu: Do not define or overwrite vectors in SPL build
Subsystems (4)
PATCH 1/1 arm: mvebu: Do not define or overwrite vectors in SPL build
U-Boot SPL is executed by the BootROM. And BootROM expects that U-Boot SPL code returns back to the BootROM. Reset vectors during execution of U-Boot SPL should not be changed as BootROM does not expect it and uses its own reset vectors.
Add ifdefs which disable overwriting reset vectors for 32-bit mvebu platforms on which U-Boot SPL should returns back to the BootROM.
--- arch/arm/cpu/armv7/start.S | 7 ++++--- arch/arm/lib/vectors.S | 6 ++++++ 2 files changed, 10 insertions(+), 3 deletions(-)
arch/arm/cpu/armv7/start.S +4 -3
@@ -99,10 +99,11 @@ switch_to_hypervisor_ret: /* * Setup vector: - * (OMAP4 spl TEXT_BASE is not 32 byte aligned. - * Continue to use ROM code vector only in OMAP4 spl) + * OMAP4 spl TEXT_BASE is not 32 byte aligned. + * 32-bit mvebu spl returns execution back to BootROM and should not change vectors. + * Continue to use ROM code vector only in OMAP4 or 32-bit mvebu spl. */ -#if !(defined(CONFIG_OMAP44XX) && defined(CONFIG_SPL_BUILD)) +#if !((defined(CONFIG_OMAP44XX) || (defined(CONFIG_ARCH_MVEBU) && defined(CONFIG_ARMADA_32BIT))) && defined(CONFIG_SPL_BUILD)) /* Set V=0 in CP15 SCTLR register - for VBAR to point to vector */ mrc p15, 0, r0, c1, c0, 0 @ Read CP15 SCTLR Register bic r0, #CR_V @ V = 0
arch/arm/lib/vectors.S +6 -0
@@ -24,6 +24,7 @@ #else b reset #endif +#if !(defined(CONFIG_ARCH_MVEBU) && defined(CONFIG_ARMADA_32BIT) && defined(CONFIG_SPL_BUILD)) ldr pc, _undefined_instruction ldr pc, _software_interrupt ldr pc, _prefetch_abort
@@ -31,6 +32,7 @@ ldr pc, _not_used ldr pc, _irq ldr pc, _fiq +#endif .endm
@@ -87,6 +89,7 @@ _start: ARM_VECTORS #endif /* !defined(CONFIG_ENABLE_ARM_SOC_BOOT0_HOOK) */ +#if !(defined(CONFIG_ARCH_MVEBU) && defined(CONFIG_ARMADA_32BIT) && defined(CONFIG_SPL_BUILD)) /* ************************************************************************* *
@@ -118,6 +121,7 @@ _irq: .word irq _fiq: .word fiq .balignl 16,0xdeadbeef +#endif /* *************************************************************************
@@ -131,6 +135,7 @@ _fiq: .word fiq #ifdef CONFIG_SPL_BUILD +#if !(defined(CONFIG_ARCH_MVEBU) && defined(CONFIG_ARMADA_32BIT) && defined(CONFIG_SPL_BUILD)) .align 5 undefined_instruction: software_interrupt:
@@ -141,6 +146,7 @@ irq: fiq: 1: b 1b /* hang and never return */ +#endif #else /* !CONFIG_SPL_BUILD */
Signature · 1 line
2.20.1
Quoted 4 lines · Pali Rohár
U-Boot SPL is executed by the BootROM. And BootROM expects that U-Boot SPL code returns back to the BootROM. Reset vectors during execution of U-Boot SPL should not be changed as BootROM does not expect it and uses its own reset vectors.
Just curious: How did you detect this problem? What actually happens when the reset vectors are overwritten here?
Quoted 15 lines
Add ifdefs which disable overwriting reset vectors for 32-bit mvebu platforms on which U-Boot SPL should returns back to the BootROM.
Signed-off-by Pali Rohár <[email protected]>--- arch/arm/cpu/armv7/start.S | 7 ++++--- arch/arm/lib/vectors.S | 6 ++++++ 2 files changed, 10 insertions(+), 3 deletions(-)arch/arm/cpu/armv7/start.S +0
-0@@ -99,10 +99,11 @@ switch_to_hypervisor_ret:
arch/arm/cpu/armv7/start.S switch_to_hypervisor_ret: · line 106
Quoted 10 lines
arch/arm/cpu/armv7/start.S +4
-3/* * Setup vector: - * (OMAP4 spl TEXT_BASE is not 32 byte aligned. - * Continue to use ROM code vector only in OMAP4 spl) + * OMAP4 spl TEXT_BASE is not 32 byte aligned. + * 32-bit mvebu spl returns execution back to BootROM and should not change vectors. + * Continue to use ROM code vector only in OMAP4 or 32-bit mvebu spl. */ -#if !(defined(CONFIG_OMAP44XX) && defined(CONFIG_SPL_BUILD)) +#if !((defined(CONFIG_OMAP44XX) || (defined(CONFIG_ARCH_MVEBU) && defined(CONFIG_ARMADA_32BIT))) && defined(CONFIG_SPL_BUILD))
Will this file ever be compiled for (defined(CONFIG_ARCH_MVEBU) && !defined(CONFIG_ARMADA_32BIT))? Meaning cpu/armv7 pretty much means 32bit, or did I miss something? If this is correct, you can probably drop one of the checks making this line a bit easier to read (again).
Perhaps it's better to introduce a new Kconfig option for this instead and select it for the necessary platforms/targets?
Thanks,
Stefan
Quoted 48 lines
arch/arm/cpu/armv7/start.S +0
-0/* Set V=0 in CP15 SCTLR register - for VBAR to point to vector */ mrc p15, 0, r0, c1, c0, 0 @ Read CP15 SCTLR Register bic r0, #CR_V @ V = 0arch/arm/lib/vectors.S +5
-0@@ -24,6 +24,7 @@ #else b reset #endif +#if !(defined(CONFIG_ARCH_MVEBU) && defined(CONFIG_ARMADA_32BIT) && defined(CONFIG_SPL_BUILD)) ldr pc, _undefined_instruction ldr pc, _software_interrupt ldr pc, _prefetch_abort@@ -31,6 +32,7 @@ ldr pc, _not_used ldr pc, _irq ldr pc, _fiq +#endif .endm@@ -87,6 +89,7 @@ _start: ARM_VECTORS #endif /* !defined(CONFIG_ENABLE_ARM_SOC_BOOT0_HOOK) */ +#if !(defined(CONFIG_ARCH_MVEBU) && defined(CONFIG_ARMADA_32BIT) && defined(CONFIG_SPL_BUILD)) /* ************************************************************************* *@@ -118,6 +121,7 @@ _irq: .word irq _fiq: .word fiq .balignl 16,0xdeadbeef +#endif /* *************************************************************************@@ -131,6 +135,7 @@ _fiq: .word fiq #ifdef CONFIG_SPL_BUILD +#if !(defined(CONFIG_ARCH_MVEBU) && defined(CONFIG_ARMADA_32BIT) && defined(CONFIG_SPL_BUILD)) .align 5 undefined_instruction: software_interrupt:
arch/arm/lib/vectors.S irq: · line 152
Quoted 7 lines
arch/arm/lib/vectors.S +1
-0@@ -141,6 +146,7 @@ irq: fiq: 1: b 1b /* hang and never return */ +#endif #else /* !CONFIG_SPL_BUILD */
Viele Grüße,
Stefan Roese
Signature · 3 lines
DENX Software Engineering GmbH, Managing Director: Wolfgang Denk HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany Phone: (+49)-8142-66989-51 Fax: (+49)-8142-66989-80 Email: [email protected]
Quoted 8 lines · On Wednesday 16 February 2022 11:08:19 Stefan Roese wrote:
Quoted 4 lines · Pali Rohár
U-Boot SPL is executed by the BootROM. And BootROM expects that U-Boot SPL code returns back to the BootROM. Reset vectors during execution of U-Boot SPL should not be changed as BootROM does not expect it and uses its own reset vectors.
Just curious: How did you detect this problem? What actually happens when the reset vectors are overwritten here?
BootROM loose ability to detect / count issues with unbootable sources which crashes. Nothing important when SPI source is fully bootable without any issue. There are registers into which BootROM stores reset information, so during _early_ debugging it could be useful to have it filled.
Quoted 21 lines
Quoted 21 lines
Add ifdefs which disable overwriting reset vectors for 32-bit mvebu platforms on which U-Boot SPL should returns back to the BootROM.
Signed-off-by Pali Rohár <[email protected]>--- arch/arm/cpu/armv7/start.S | 7 ++++--- arch/arm/lib/vectors.S | 6 ++++++ 2 files changed, 10 insertions(+), 3 deletions(-)arch/arm/cpu/armv7/start.S +3
-2@@ -99,10 +99,11 @@ switch_to_hypervisor_ret: /* * Setup vector: - * (OMAP4 spl TEXT_BASE is not 32 byte aligned. - * Continue to use ROM code vector only in OMAP4 spl) + * OMAP4 spl TEXT_BASE is not 32 byte aligned. + * 32-bit mvebu spl returns execution back to BootROM and should not change vectors. + * Continue to use ROM code vector only in OMAP4 or 32-bit mvebu spl.
Quoted 8 lines
arch/arm/cpu/armv7/start.S switch_to_hypervisor_ret: · line 106
Quoted 3 lines
arch/arm/cpu/armv7/start.S +1
-1*/ -#if !(defined(CONFIG_OMAP44XX) && defined(CONFIG_SPL_BUILD)) +#if !((defined(CONFIG_OMAP44XX) || (defined(CONFIG_ARCH_MVEBU) && defined(CONFIG_ARMADA_32BIT))) && defined(CONFIG_SPL_BUILD))Will this file ever be compiled for (
defined(CONFIG_ARCH_MVEBU)&& !defined(CONFIG_ARMADA_32BIT))? Meaning cpu/armv7 pretty much means 32bit, or did I miss something? If this is correct, you can probably drop one of the checks making this line a bit easier to read (again).
First I added just CONFIG_ARCH_MVEBU and realized that it is enabled also for 64-bit mvebu platforms. So later I added CONFIG_ARMADA_32BIT and forgot about it. And you are right CONFIG_ARMADA_32BIT should be enough.
Quoted 2 lines
Perhaps it's better to introduce a new
Kconfigoption for this instead and select it for the necessary platforms/targets?
This should be possible, but I do not know how to do this config symbol magic correctly when it should be applied only for SPL builds...
Quoted 57 lines
Thanks,
StefanQuoted 41 lines
arch/arm/cpu/armv7/start.S +0
-0/* Set V=0 in CP15 SCTLR register - for VBAR to point to vector */ mrc p15, 0, r0, c1, c0, 0 @ Read CP15 SCTLR Register bic r0, #CR_V @ V = 0arch/arm/lib/vectors.S +5
-0@@ -24,6 +24,7 @@ #else b reset #endif +#if !(defined(CONFIG_ARCH_MVEBU) && defined(CONFIG_ARMADA_32BIT) && defined(CONFIG_SPL_BUILD)) ldr pc, _undefined_instruction ldr pc, _software_interrupt ldr pc, _prefetch_abort@@ -31,6 +32,7 @@ ldr pc, _not_used ldr pc, _irq ldr pc, _fiq +#endif .endm@@ -87,6 +89,7 @@ _start: ARM_VECTORS #endif /* !defined(CONFIG_ENABLE_ARM_SOC_BOOT0_HOOK) */ +#if !(defined(CONFIG_ARCH_MVEBU) && defined(CONFIG_ARMADA_32BIT) && defined(CONFIG_SPL_BUILD)) /* ************************************************************************* *@@ -118,6 +121,7 @@ _irq: .word irq _fiq: .word fiq .balignl 16,0xdeadbeef +#endif /* *************************************************************************@@ -131,6 +135,7 @@ _fiq: .word fiq #ifdef CONFIG_SPL_BUILD +#if !(defined(CONFIG_ARCH_MVEBU) && defined(CONFIG_ARMADA_32BIT) && defined(CONFIG_SPL_BUILD)) .align 5 undefined_instruction: software_interrupt:arch/arm/lib/vectors.S irq: · line 150
Quoted 5 lines
arch/arm/lib/vectors.S +1
-0@@ -141,6 +146,7 @@ irq: fiq: 1: b 1b /* hang and never return */ +#endif #else /* !CONFIG_SPL_BUILD */Viele Grüße,
Stefan RoeseSignature · 3 lines
DENX Software Engineering GmbH, Managing Director: Wolfgang Denk HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany Phone: (+49)-8142-66989-51 Fax: (+49)-8142-66989-80 Email: [email protected]
PATCH v2 2/3 arm: Do not compile vector table when SYS_NO_VECTOR_TABLE is enabled
In reply to Pali RohárVector table is not used when SYS_NO_VECTOR_TABLE is enabled. So do not compile it and reduce image size.
--- arch/arm/lib/vectors.S | 6 ++++++ 1 file changed, 6 insertions(+)
arch/arm/lib/vectors.S +6 -0
@@ -24,6 +24,7 @@ #else b reset #endif +#if !CONFIG_IS_ENABLED(SYS_NO_VECTOR_TABLE) ldr pc, _undefined_instruction ldr pc, _software_interrupt ldr pc, _prefetch_abort
@@ -31,6 +32,7 @@ ldr pc, _not_used ldr pc, _irq ldr pc, _fiq +#endif .endm
@@ -87,6 +89,7 @@ _start: ARM_VECTORS #endif /* !defined(CONFIG_ENABLE_ARM_SOC_BOOT0_HOOK) */ +#if !CONFIG_IS_ENABLED(SYS_NO_VECTOR_TABLE) /* ************************************************************************* *
@@ -118,6 +121,7 @@ _irq: .word irq _fiq: .word fiq .balignl 16,0xdeadbeef +#endif /* *************************************************************************
@@ -131,6 +135,7 @@ _fiq: .word fiq #ifdef CONFIG_SPL_BUILD +#if !CONFIG_IS_ENABLED(SYS_NO_VECTOR_TABLE) .align 5 undefined_instruction: software_interrupt:
@@ -141,6 +146,7 @@ irq: fiq: 1: b 1b /* hang and never return */ +#endif #else /* !CONFIG_SPL_BUILD */
Signature · 1 line
2.20.1
PATCH v2 1/3 arm: Introduce new CONFIG_SPL_SYS_NO_VECTOR_TABLE option
In reply to Pali RohárMove OMAP4 specific option for disabling overwriting vector table into config option CONFIG_SPL_SYS_NO_VECTOR_TABLE.
--- arch/arm/Kconfig | 4 ++++ arch/arm/cpu/armv7/start.S | 4 +--- arch/arm/mach-omap2/Kconfig | 1 + 3 files changed, 6 insertions(+), 3 deletions(-)
arch/arm/Kconfig +4 -0
@@ -57,6 +57,10 @@ config SYS_INIT_SP_BSS_OFFSET that the early malloc region, global data (gd), and early stack usage do not overlap any appended DTB. +config SPL_SYS_NO_VECTOR_TABLE + depends on SPL + bool + config LINUX_KERNEL_IMAGE_HEADER depends on ARM64 bool
arch/arm/cpu/armv7/start.S +1 -3
@@ -97,12 +97,10 @@ switch_to_hypervisor_ret: orr r0, r0, #0xc0 @ disable FIQ and IRQ msr cpsr,r0 +#if !CONFIG_IS_ENABLED(SYS_NO_VECTOR_TABLE) /* * Setup vector: - * (OMAP4 spl TEXT_BASE is not 32 byte aligned. - * Continue to use ROM code vector only in OMAP4 spl) */ -#if !(defined(CONFIG_OMAP44XX) && defined(CONFIG_SPL_BUILD)) /* Set V=0 in CP15 SCTLR register - for VBAR to point to vector */ mrc p15, 0, r0, c1, c0, 0 @ Read CP15 SCTLR Register bic r0, #CR_V @ V = 0
arch/arm/mach-omap2/Kconfig +1 -0
@@ -32,6 +32,7 @@ config OMAP34XX config OMAP44XX bool "OMAP44XX SoC" select SPL_USE_TINY_PRINTF + select SPL_SYS_NO_VECTOR_TABLE if SPL imply NAND_OMAP_ELM imply NAND_OMAP_GPMC imply SPL_DISPLAY_PRINT
Signature · 1 line
2.20.1
PATCH v2 3/3 arm: mvebu: Enable CONFIG_SPL_SYS_NO_VECTOR_TABLE for 32-bit mvebu
U-Boot SPL is on 32-bit mvebu executed by the BootROM. And BootROM expects that U-Boot SPL returns execution back to the BootROM. Vectors during execution of U-Boot SPL should not be changed as BootROM does not expect it and uses its own vectors. So do not overwrite vectors in SPL build.
--- arch/arm/mach-mvebu/Kconfig | 1 + 1 file changed, 1 insertion(+)
arch/arm/mach-mvebu/Kconfig +1 -0
@@ -16,6 +16,7 @@ config ARMADA_32BIT select SUPPORT_SPL select TRANSLATION_OFFSET select TOOLS_KWBIMAGE if SPL + select SPL_SYS_NO_VECTOR_TABLE if SPL config ARMADA_64BIT bool
Signature · 1 line
2.20.1
Re: [PATCH u-boot-marvell v2 1/3] arm: Introduce new CONFIG_SPL_SYS_NO_VECTOR_TABLE option
In reply to Pali RohárQuoted 4 lines · Pali Rohár
Move OMAP4 specific option for disabling overwriting vector table into config option
CONFIG_SPL_SYS_NO_VECTOR_TABLE.Signed-off-by Pali Rohár <[email protected]>
Applied to u-boot-marvell/master
Thanks,
Stefan
Quoted 44 lines
--- arch/arm/Kconfig | 4 ++++ arch/arm/cpu/armv7/start.S | 4 +--- arch/arm/mach-omap2/Kconfig | 1 + 3 files changed, 6 insertions(+), 3 deletions(-)arch/arm/Kconfig +4
-0@@ -57,6 +57,10 @@ config SYS_INIT_SP_BSS_OFFSET that the early malloc region, global data (gd), and early stack usage do not overlap any appended DTB. +config SPL_SYS_NO_VECTOR_TABLE + depends on SPL + bool + config LINUX_KERNEL_IMAGE_HEADER depends on ARM64 boolarch/arm/cpu/armv7/start.S +1
-3@@ -97,12 +97,10 @@ switch_to_hypervisor_ret: orr r0, r0, #0xc0 @ disable FIQ and IRQ msr cpsr,r0 +#if !CONFIG_IS_ENABLED(SYS_NO_VECTOR_TABLE) /* * Setup vector: - * (OMAP4 spl TEXT_BASE is not 32 byte aligned. - * Continue to use ROM code vector only in OMAP4 spl) */ -#if !(defined(CONFIG_OMAP44XX) && defined(CONFIG_SPL_BUILD)) /* Set V=0 in CP15 SCTLR register - for VBAR to point to vector */ mrc p15, 0, r0, c1, c0, 0 @ Read CP15 SCTLR Register bic r0, #CR_V @ V = 0arch/arm/mach-omap2/Kconfig +0
-0
arch/arm/mach-omap2/Kconfig config OMAP34XX · line 38
Quoted 7 lines
arch/arm/mach-omap2/Kconfig +1
-0@@ -32,6 +32,7 @@ config OMAP34XX config OMAP44XX bool "OMAP44XX SoC" select SPL_USE_TINY_PRINTF + select SPL_SYS_NO_VECTOR_TABLE if SPL imply NAND_OMAP_ELM imply NAND_OMAP_GPMC imply SPL_DISPLAY_PRINT
Viele Grüße,
Stefan Roese
Signature · 3 lines
DENX Software Engineering GmbH, Managing Director: Wolfgang Denk HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany Phone: (+49)-8142-66989-51 Fax: (+49)-8142-66989-80 Email: [email protected]
Re: [PATCH u-boot-marvell v2 2/3] arm: Do not compile vector table when SYS_NO_VECTOR_TABLE is enabled
In reply to Pali RohárQuoted 4 lines · Pali Rohár
Vector table is not used when
SYS_NO_VECTOR_TABLEis enabled. So do not compile it and reduce image size.Signed-off-by Pali Rohár <[email protected]>
Applied to u-boot-marvell/master
Thanks,
Stefan
Quoted 49 lines
--- arch/arm/lib/vectors.S | 6 ++++++ 1 file changed, 6 insertions(+)arch/arm/lib/vectors.S +5
-0@@ -24,6 +24,7 @@ #else b reset #endif +#if !CONFIG_IS_ENABLED(SYS_NO_VECTOR_TABLE) ldr pc, _undefined_instruction ldr pc, _software_interrupt ldr pc, _prefetch_abort@@ -31,6 +32,7 @@ ldr pc, _not_used ldr pc, _irq ldr pc, _fiq +#endif .endm@@ -87,6 +89,7 @@ _start: ARM_VECTORS #endif /* !defined(CONFIG_ENABLE_ARM_SOC_BOOT0_HOOK) */ +#if !CONFIG_IS_ENABLED(SYS_NO_VECTOR_TABLE) /* ************************************************************************* *@@ -118,6 +121,7 @@ _irq: .word irq _fiq: .word fiq .balignl 16,0xdeadbeef +#endif /* *************************************************************************@@ -131,6 +135,7 @@ _fiq: .word fiq #ifdef CONFIG_SPL_BUILD +#if !CONFIG_IS_ENABLED(SYS_NO_VECTOR_TABLE) .align 5 undefined_instruction: software_interrupt:
arch/arm/lib/vectors.S irq: · line 152
Quoted 7 lines
arch/arm/lib/vectors.S +1
-0@@ -141,6 +146,7 @@ irq: fiq: 1: b 1b /* hang and never return */ +#endif #else /* !CONFIG_SPL_BUILD */
Viele Grüße,
Stefan Roese
Signature · 3 lines
DENX Software Engineering GmbH, Managing Director: Wolfgang Denk HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany Phone: (+49)-8142-66989-51 Fax: (+49)-8142-66989-80 Email: [email protected]
Re: [PATCH u-boot-marvell v2 3/3] arm: mvebu: Enable CONFIG_SPL_SYS_NO_VECTOR_TABLE for 32-bit mvebu
In reply to Pali RohárQuoted 6 lines · Pali Rohár
U-Boot SPL is on 32-bit mvebu executed by the BootROM. And BootROM expects that U-Boot SPL returns execution back to the BootROM. Vectors during execution of U-Boot SPL should not be changed as BootROM does not expect it and uses its own vectors. So do not overwrite vectors in SPL build.
Signed-off-by Pali Rohár <[email protected]>
Applied to u-boot-marvell/master
Thanks,
Stefan
Quoted 9 lines
--- arch/arm/mach-mvebu/Kconfig | 1 + 1 file changed, 1 insertion(+)arch/arm/mach-mvebu/Kconfig +0
-0
arch/arm/mach-mvebu/Kconfig config ARMADA_32BIT · line 22
Quoted 7 lines
arch/arm/mach-mvebu/Kconfig +1
-0@@ -16,6 +16,7 @@ config ARMADA_32BIT select SUPPORT_SPL select TRANSLATION_OFFSET select TOOLS_KWBIMAGE if SPL + select SPL_SYS_NO_VECTOR_TABLE if SPL config ARMADA_64BIT bool
Viele Grüße,
Stefan Roese
Signature · 3 lines
DENX Software Engineering GmbH, Managing Director: Wolfgang Denk HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany Phone: (+49)-8142-66989-51 Fax: (+49)-8142-66989-80 Email: [email protected]
Re: [PATCH u-boot-marvell v2 1/3] arm: Introduce new CONFIG_SPL_SYS_NO_VECTOR_TABLE option
In reply to Pali RohárQuoted 15 lines · Pali Rohár
Move OMAP4 specific option for disabling overwriting vector table into config option
CONFIG_SPL_SYS_NO_VECTOR_TABLE.Signed-off-by Pali Rohár <[email protected]>--- arch/arm/Kconfig | 4 ++++ arch/arm/cpu/armv7/start.S | 4 +--- arch/arm/mach-omap2/Kconfig | 1 + 3 files changed, 6 insertions(+), 3 deletions(-)arch/arm/Kconfig +0
-0
arch/arm/Kconfig config SYS_INIT_SP_BSS_OFFSET · line 63
Quoted 7 lines
arch/arm/Kconfig +4
-0@@ -57,6 +57,10 @@ config SYS_INIT_SP_BSS_OFFSET that the early malloc region, global data (gd), and early stack usage do not overlap any appended DTB. +config SPL_SYS_NO_VECTOR_TABLE + depends on SPL + bool +
ARMv8 has symbol CONFIG_ARMV8_SPL_EXCEPTION_VECTORS.
Why do we need two separate setting which have the same usage (but with inverse logic)? Can't we consolidate the two?
Why should this symbol not be editable?
I guess it needs the same warning as for ARMv8 that if you set it there will be no crash dump in SPL.
Best regards
Heinrich
Quoted 33 lines
arch/arm/Kconfig +0
-0config LINUX_KERNEL_IMAGE_HEADER depends on ARM64 boolarch/arm/cpu/armv7/start.S +1
-3@@ -97,12 +97,10 @@ switch_to_hypervisor_ret: orr r0, r0, #0xc0 @ disable FIQ and IRQ msr cpsr,r0 +#if !CONFIG_IS_ENABLED(SYS_NO_VECTOR_TABLE) /* * Setup vector: - * (OMAP4 spl TEXT_BASE is not 32 byte aligned. - * Continue to use ROM code vector only in OMAP4 spl) */ -#if !(defined(CONFIG_OMAP44XX) && defined(CONFIG_SPL_BUILD)) /* Set V=0 in CP15 SCTLR register - for VBAR to point to vector */ mrc p15, 0, r0, c1, c0, 0 @ Read CP15 SCTLR Register bic r0, #CR_V @ V = 0arch/arm/mach-omap2/Kconfig +1
-0@@ -32,6 +32,7 @@ config OMAP34XX config OMAP44XX bool "OMAP44XX SoC" select SPL_USE_TINY_PRINTF + select SPL_SYS_NO_VECTOR_TABLE if SPL imply NAND_OMAP_ELM imply NAND_OMAP_GPMC imply SPL_DISPLAY_PRINT
Re: [PATCH u-boot-marvell v2 1/3] arm: Introduce new CONFIG_SPL_SYS_NO_VECTOR_TABLE option
Quoted 26 lines · On Wednesday 03 August 2022 02:02:22 Heinrich Schuchardt wrote:
Quoted 15 lines · Pali Rohár
Move OMAP4 specific option for disabling overwriting vector table into config option
CONFIG_SPL_SYS_NO_VECTOR_TABLE.Signed-off-by Pali Rohár <[email protected]>--- arch/arm/Kconfig | 4 ++++ arch/arm/cpu/armv7/start.S | 4 +--- arch/arm/mach-omap2/Kconfig | 1 + 3 files changed, 6 insertions(+), 3 deletions(-)arch/arm/Kconfig +0
-0arch/arm/Kconfig config SYS_INIT_SP_BSS_OFFSET · line 63
Quoted 7 lines
arch/arm/Kconfig +4
-0@@ -57,6 +57,10 @@ config SYS_INIT_SP_BSS_OFFSET that the early malloc region, global data (gd), and early stack usage do not overlap any appended DTB. +config SPL_SYS_NO_VECTOR_TABLE + depends on SPL + bool +ARMv8 has symbol
CONFIG_ARMV8_SPL_EXCEPTION_VECTORS.
Quoted 4 lines
Why do we need two separate setting which have the same usage (but with inverse logic)? Can't we consolidate the two?
Why should this symbol not be editable?
These two symbols are different. CONFIG_ARMV8_SPL_EXCEPTION_VECTORS describes that it is there for ARMv8 and its usage is if user wants to save space, to decrease U-Boot binary.
This new SPL_SYS_NO_VECTOR_TABLE is there for ARMv7 and as you pointed correctly it is not user editable. This symbol is there for platforms which _require_ that SPL does not modify exception tables. If platform sets it as needed then user must not be able to disable this symbol.
I hope that I explained it in commit message, SPL on 32-bit mvebu boards is started by BootROM "as a function". SPL must returns control to BootROM and BootROM expects that vector table is not modified.
Quoted 2 lines
I guess it needs the same warning as for ARMv8 that if you set it there will be no crash dump in SPL.
As this symbol is not user editable, it does not have description (requirement by Kconfig language). If crash happens then it is handled by BootROM itself.
But I think that description is not needed if symbol is not user editable. It is up to the maintainer of architecture to decide if this option must be enabled or disabled (default) based on other things.
Probably what could make sense is to "lift" CONFIG_ARMV8_SPL_EXCEPTION_VECTORS option to be user selectable and for both ARMv7 and ARMv8. To allow user to disable generating vectors on ARMv7 too (but not to enable them if architecture has to have them disabled).
Quoted 38 lines
Best regards
Heinrich
Quoted 33 lines
arch/arm/Kconfig +0
-0config LINUX_KERNEL_IMAGE_HEADER depends on ARM64 boolarch/arm/cpu/armv7/start.S +1
-3@@ -97,12 +97,10 @@ switch_to_hypervisor_ret: orr r0, r0, #0xc0 @ disable FIQ and IRQ msr cpsr,r0 +#if !CONFIG_IS_ENABLED(SYS_NO_VECTOR_TABLE) /* * Setup vector: - * (OMAP4 spl TEXT_BASE is not 32 byte aligned. - * Continue to use ROM code vector only in OMAP4 spl) */ -#if !(defined(CONFIG_OMAP44XX) && defined(CONFIG_SPL_BUILD)) /* Set V=0 in CP15 SCTLR register - for VBAR to point to vector */ mrc p15, 0, r0, c1, c0, 0 @ Read CP15 SCTLR Register bic r0, #CR_V @ V = 0arch/arm/mach-omap2/Kconfig +1
-0@@ -32,6 +32,7 @@ config OMAP34XX config OMAP44XX bool "OMAP44XX SoC" select SPL_USE_TINY_PRINTF + select SPL_SYS_NO_VECTOR_TABLE if SPL imply NAND_OMAP_ELM imply NAND_OMAP_GPMC imply SPL_DISPLAY_PRINT