Skip to contentSkip to conversation

arm64: clean up trampoline vector loads

PATCH Applied Rémi Denis-Courmont 3 participants 6 messages Mar 16, 2020 – Mar 18, 2020 arm64

v1 of 3 Applied Latest is v3 →

lore
Share & tools
lore
#0 lore raw

PATCH 1/3 arm64: clean up trampoline vector loads

From: Rémi Denis-Courmont <[email protected]>

This switches from custom instruction patterns to the regular large memory model sequence with ADRP and LDR. In doing so, the ADD instruction can be eliminated in the SDEI handler, and the code no longer assumes that the trampoline vectors and the vectors address both start on a page boundary.

---
 arch/arm64/kernel/entry.S | 9 ++++-----
 1 file changed, 4 insertions(+), 5 deletions(-)
arch/arm64/kernel/entry.S +4 -5
index e5d4e30ee242..24f828739696 100644
@@ -805,9 +805,9 @@ alternative_else_nop_endif
 2:
 	tramp_map_kernel	x30
 #ifdef CONFIG_RANDOMIZE_BASE
-	adr	x30, tramp_vectors + PAGE_SIZE
+	adrp	x30, tramp_vectors + PAGE_SIZE
 alternative_insn isb, nop, ARM64_WORKAROUND_QCOM_FALKOR_E1003
-	ldr	x30, [x30]
+	ldr	x30, [x30, #:lo12:__entry_tramp_data_start]
 #else
 	ldr	x30, =vectors
 #endif
@@ -953,9 +953,8 @@ SYM_CODE_START(__sdei_asm_entry_trampoline)
 1:	str	x4, [x1, #(SDEI_EVENT_INTREGS + S_ORIG_ADDR_LIMIT)]
 
 #ifdef CONFIG_RANDOMIZE_BASE
-	adr	x4, tramp_vectors + PAGE_SIZE
-	add	x4, x4, #:lo12:__sdei_asm_trampoline_next_handler
-	ldr	x4, [x4]
+	adrp	x4, tramp_vectors + PAGE_SIZE
+	ldr	x4, [x4, #:lo12:__sdei_asm_trampoline_next_handler]
 #else
 	ldr	x4, =__sdei_asm_handler
 #endif
Signature · 1 line
2.25.1
maintainer Acked-by on v1 1/3
#1 lore raw
Quoted 31 lines · Rémi Denis-Courmont

From: Rémi Denis-Courmont <[email protected]>

This switches from custom instruction patterns to the regular large memory model sequence with ADRP and LDR. In doing so, the ADD instruction can be eliminated in the SDEI handler, and the code no longer assumes that the trampoline vectors and the vectors address both start on a page boundary.

---
 arch/arm64/kernel/entry.S | 9 ++++-----
 1 file changed, 4 insertions(+), 5 deletions(-)
arch/arm64/kernel/entry.S +2 -2
index e5d4e30ee242..24f828739696 100644
@@ -805,9 +805,9 @@ alternative_else_nop_endif
 2:
 	tramp_map_kernel	x30
 #ifdef CONFIG_RANDOMIZE_BASE
-	adr	x30, tramp_vectors + PAGE_SIZE
+	adrp	x30, tramp_vectors + PAGE_SIZE
 alternative_insn isb, nop, ARM64_WORKAROUND_QCOM_FALKOR_E1003
-	ldr	x30, [x30]
+	ldr	x30, [x30, #:lo12:__entry_tramp_data_start]
 #else
 	ldr	x30, =vectors
 #endif
@@ -953,9 +953,8 @@ SYM_CODE_START(__sdei_asm_entry_trampoline)
 1:	str	x4, [x1, #(SDEI_EVENT_INTREGS + S_ORIG_ADDR_LIMIT)]

arch/arm64/kernel/entry.S SYM_CODE_START(__sdei_asm_entry_trampoline) · line 960

Quoted 10 lines
arch/arm64/kernel/entry.S +2 -3
 
 #ifdef CONFIG_RANDOMIZE_BASE
-	adr	x4, tramp_vectors + PAGE_SIZE
-	add	x4, x4, #:lo12:__sdei_asm_trampoline_next_handler
-	ldr	x4, [x4]
+	adrp	x4, tramp_vectors + PAGE_SIZE
+	ldr	x4, [x4, #:lo12:__sdei_asm_trampoline_next_handler]
 #else
 	ldr	x4, =__sdei_asm_handler
 #endif

Will

maintainer Applied
#2 lore raw
In reply to Rémi Denis-Courmont
Quoted 9 lines · Rémi Denis-Courmont

From: Rémi Denis-Courmont <[email protected]>

This switches from custom instruction patterns to the regular large memory model sequence with ADRP and LDR. In doing so, the ADD instruction can be eliminated in the SDEI handler, and the code no longer assumes that the trampoline vectors and the vectors address both start on a page boundary.

I queued the 3 trampoline patches for 5.7. Thanks.

Signature · 1 line
Catalin
maintainer
#3 lore raw
Quoted 9 lines · Catalin Marinas
Quoted 8 lines · Rémi Denis-Courmont

From: Rémi Denis-Courmont <[email protected]>

This switches from custom instruction patterns to the regular large memory model sequence with ADRP and LDR. In doing so, the ADD instruction can be eliminated in the SDEI handler, and the code no longer assumes that the trampoline vectors and the vectors address both start on a page boundary.

Quoted 3 lines
Quoted 1 line

I queued the 3 trampoline patches for 5.7. Thanks.

... and removed. I applied them on top of arm64 for-next/asm-annotations and with defconfig I get:

LD .tmp_vmlinux1
arch/arm64/kernel/entry.o: in function `tramp_vectors':
arch/arm64/kernel/entry.S:838:(.entry.tramp.text+0x43c): relocation truncated to fit: R_AARCH64_LDST64_ABS_LO12_NC against symbol `__entry_tramp_data_start' defined in .rodata section in arch/arm64/kernel/entry.o
ld: arch/arm64/kernel/entry.S:838: warning: one possible cause of this error is that the symbol is being referenced in the indicated code as if it had a larger alignment than was declared where it was defined arch/arm64/kernel/entry.S:839:(.entry.tramp.text+0x4bc): relocation truncated to fit: R_AARCH64_LDST64_ABS_LO12_NC against symbol `__entry_tramp_data_start' defined in .rodata section in arch/arm64/kernel/entry.o
ld: arch/arm64/kernel/entry.S:839: warning: one possible cause of this error is that the symbol is being referenced in the indicated code as if it had a larger alignment than was declared where it was defined arch/arm64/kernel/entry.S:840:(.entry.tramp.text+0x53c): relocation truncated to fit: R_AARCH64_LDST64_ABS_LO12_NC against symbol `__entry_tramp_data_start' defined in .rodata section in arch/arm64/kernel/entry.o
ld: arch/arm64/kernel/entry.S:840: warning: one possible cause of this error is that the symbol is being referenced in the indicated code as if it had a larger alignment than was declared where it was defined arch/arm64/kernel/entry.S:841:(.entry.tramp.text+0x5bc): relocation truncated to fit: R_AARCH64_LDST64_ABS_LO12_NC against symbol `__entry_tramp_data_start' defined in .rodata section in arch/arm64/kernel/entry.o
ld: arch/arm64/kernel/entry.S:841: warning: one possible cause of this error is that the symbol is being referenced in the indicated code as if it had a larger alignment than was declared where it was defined arch/arm64/kernel/entry.S:843:(.entry.tramp.text+0x638): relocation truncated to fit: R_AARCH64_LDST64_ABS_LO12_NC against symbol `__entry_tramp_data_start' defined in .rodata section in arch/arm64/kernel/entry.o
ld: arch/arm64/kernel/entry.S:843: warning: one possible cause of this error is that the symbol is being referenced in the indicated code as if it had a larger alignment than was declared where it was defined arch/arm64/kernel/entry.S:844:(.entry.tramp.text+0x6b8): relocation truncated to fit: R_AARCH64_LDST64_ABS_LO12_NC against symbol `__entry_tramp_data_start' defined in .rodata section in arch/arm64/kernel/entry.o
ld: arch/arm64/kernel/entry.S:844: warning: one possible cause of this error is that the symbol is being referenced in the indicated code as if it had a larger alignment than was declared where it was defined arch/arm64/kernel/entry.S:845:(.entry.tramp.text+0x738): relocation truncated to fit: R_AARCH64_LDST64_ABS_LO12_NC against symbol `__entry_tramp_data_start' defined in .rodata section in arch/arm64/kernel/entry.o
ld: arch/arm64/kernel/entry.S:845: warning: one possible cause of this error is that the symbol is being referenced in the indicated code as if it had a larger alignment than was declared where it was defined arch/arm64/kernel/entry.S:846:(.entry.tramp.text+0x7b8): relocation truncated to fit: R_AARCH64_LDST64_ABS_LO12_NC against symbol `__entry_tramp_data_start' defined in .rodata section in arch/arm64/kernel/entry.o
ld: arch/arm64/kernel/entry.S:846: warning: one possible cause of this error is that the symbol is being referenced in the indicated code as if it had a larger alignment than was declared where it was defined make[1]: *** [Makefile:1077: vmlinux] Error 1

I haven't bisected to see which patch caused this issue.

$ gcc --version
gcc (Debian 9.2.1-30) 9.2.1 20200224
$ ld --version
GNU ld (GNU Binutils for Debian) 2.34
Signature · 1 line
Catalin
#4 lore raw
Quoted 17 lines · Le keskiviikkona 18. maaliskuuta 2020, 20.06.30 EET Catalin Marinas a écrit :
Quoted 9 lines · Catalin Marinas
Quoted 8 lines · Rémi Denis-Courmont

From: Rémi Denis-Courmont <[email protected]>

This switches from custom instruction patterns to the regular large memory model sequence with ADRP and LDR. In doing so, the ADD instruction can be eliminated in the SDEI handler, and the code no longer assumes that the trampoline vectors and the vectors address both start on a page boundary.

Quoted 3 lines
Quoted 1 line

I queued the 3 trampoline patches for 5.7. Thanks.

... and removed. I applied them on top of arm64 for-next/asm-annotations and with defconfig I get:

Quoted 7 lines

LD .tmp_vmlinux1
arch/arm64/kernel/entry.o: in function `tramp_vectors':
arch/arm64/kernel/entry.S:838:(.entry.tramp.text+0x43c): relocation truncated to fit: R_AARCH64_LDST64_ABS_LO12_NC against symbol `__entry_tramp_data_start' defined in .rodata section in

I haven't bisected to see which patch caused this issue.

Uho, right :-( It only builds with SDEI enabled :-$

I'll check further.

Signature · 2 lines
Rémi Denis-Courmont
http://www.remlab.net/
#5 lore raw

Le 2020-03-18 20:29, Rémi Denis-Courmont a écrit :

Quoted 18 lines · Le keskiviikkona 18. maaliskuuta 2020, 20.06.30 EET Catalin Marinas a écrit :
Quoted 17 lines · Catalin Marinas
Quoted 9 lines · Rémi Denis-Courmont
Quoted 8 lines

From: Rémi Denis-Courmont <[email protected]>

This switches from custom instruction patterns to the regular large memory model sequence with ADRP and LDR. In doing so, the ADD instruction can be eliminated in the SDEI handler, and the code no longer assumes that the trampoline vectors and the vectors address both start on a page boundary.

Quoted 3 lines
Quoted 1 line

I queued the 3 trampoline patches for 5.7. Thanks.

... and removed. I applied them on top of arm64 for-next/asm-annotations and with defconfig I get:

Quoted 7 lines
Quoted 7 lines

LD .tmp_vmlinux1
arch/arm64/kernel/entry.o: in function `tramp_vectors':
arch/arm64/kernel/entry.S:838:(.entry.tramp.text+0x43c): relocation truncated to fit: R_AARCH64_LDST64_ABS_LO12_NC against symbol `__entry_tramp_data_start' defined in .rodata section in

I haven't bisected to see which patch caused this issue.

It's the third patch.

Quoted 3 lines

Uho, right :-( It only builds with SDEI enabled :-$

I'll check further.

It seems that the SYM_DATA_START macro does not align the data on its natural boundary. I guess that is all fine on x86 where data needs not be aligned, but it leads to this kind of mischief on arm64. Though even then, the address is of course actually aligned correctly on an 8-bytes boundary, so I suppose binutils is just being pointlessly pedantic here?

Signature · 1 line
Rémi Denis-Courmont