Skip to contentSkip to conversation

arm64: One fix and one improvement

PATCH Applied Andrew Jones 2 participants 6 messages Mar 25, 2021 KVM

lore
Share & tools
lore
#0 lore raw

cover arm64: One fix and one improvement

Fix the loading of argc. Nikos reported an issue with its alignment while testing target-efi (aligned to four bytes is OK, as long as we only load four bytes like we should). Also, take a patch developed while working on target-efi which can make debugging a bit more convenient (by doing some subtraction for the test developer).

Andrew Jones (2):
arm64: argc is an int
arm64: Output PC load offset on unhandled exceptions

 arm/cstart64.S        | 2 +-
 arm/flat.lds          | 1 +
 lib/arm64/processor.c | 7 +++++++
 3 files changed, 9 insertions(+), 1 deletion(-)
Signature · 1 line
2.26.3
#1 lore raw

PATCH 1/2 arm64: argc is an int

If argc isn't aligned to eight bytes then loading it as if it's eight bytes wide is a bad idea. It's only four bytes wide, so we should only load four bytes.

---
 arm/cstart64.S | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
arm/cstart64.S +1 -1
index 89321dad7aba..0a85338bcdae 100644
@@ -100,7 +100,7 @@ start:
 1:
 	/* run the test */
 	adrp	x0, __argc
-	ldr	x0, [x0, :lo12:__argc]
+	ldr	w0, [x0, :lo12:__argc]
 	adrp	x1, __argv
 	add	x1, x1, :lo12:__argv
 	adrp	x2, __environ
Signature · 1 line
2.26.3
Reviewed-by on v1 1/2
#2 lore raw

Re: [PATCH kvm-unit-tests 1/2] arm64: argc is an int

Quoted 6 lines · Andrew Jones

If argc isn't aligned to eight bytes then loading it as if it's eight bytes wide is a bad idea. It's only four bytes wide, so we should only load four bytes.

Thanks for fixing this.

Quoted 18 lines
---
  arm/cstart64.S | 2 +-
  1 file changed, 1 insertion(+), 1 deletion(-)
arm/cstart64.S +1 -1
index 89321dad7aba..0a85338bcdae 100644
@@ -100,7 +100,7 @@ start:
  1:
  	/* run the test */
  	adrp	x0, __argc
-	ldr	x0, [x0, :lo12:__argc]
+	ldr	w0, [x0, :lo12:__argc]
  	adrp	x1, __argv
  	add	x1, x1, :lo12:__argv
  	adrp	x2, __environ
#3 lore raw

PATCH 2/2 arm64: Output PC load offset on unhandled exceptions

In reply to Andrew Jones

Since for arm64 we can load the unit tests at different addresses, then let's make it easier to debug by calculating the PC offset for the user. The offset can then be directly used when looking at the disassembly of the test's elf file.

---
 arm/flat.lds          | 1 +
 lib/arm64/processor.c | 7 +++++++
 2 files changed, 8 insertions(+)
arm/flat.lds +1 -0
index 4d43cdfeab41..6ed377c0eaa0 100644
@@ -1,6 +1,7 @@
 
 SECTIONS
 {
+    PROVIDE(_text = .);
     .text : { *(.init) *(.text) *(.text.*) }
     . = ALIGN(64K);
     PROVIDE(etext = .);
lib/arm64/processor.c +7 -0
index ef558625e284..831207c16587 100644
@@ -99,12 +99,19 @@ bool get_far(unsigned int esr, unsigned long *far)
 	return false;
 }
 
+extern unsigned long _text;
+
 static void bad_exception(enum vector v, struct pt_regs *regs,
 			  unsigned int esr, bool esr_valid, bool bad_vector)
 {
 	unsigned long far;
 	bool far_valid = get_far(esr, &far);
 	unsigned int ec = esr >> ESR_EL1_EC_SHIFT;
+	uintptr_t text = (uintptr_t)&_text;
+
+	printf("Load address: %" PRIxPTR "\n", text);
+	printf("PC: %" PRIxPTR " PC offset: %" PRIxPTR "\n",
+	       (uintptr_t)regs->pc, (uintptr_t)regs->pc - text);
 
 	if (bad_vector) {
 		if (v < VECTOR_MAX)
Signature · 1 line
2.26.3
Reviewed-by on v1 2/2
#4 lore raw

Re: [PATCH kvm-unit-tests 2/2] arm64: Output PC load offset on unhandled exceptions

Quoted 6 lines · Andrew Jones

Since for arm64 we can load the unit tests at different addresses, then let's make it easier to debug by calculating the PC offset for the user. The offset can then be directly used when looking at the disassembly of the test's elf file.

Quoted 42 lines
---
  arm/flat.lds          | 1 +
  lib/arm64/processor.c | 7 +++++++
  2 files changed, 8 insertions(+)
arm/flat.lds +1 -0
index 4d43cdfeab41..6ed377c0eaa0 100644
@@ -1,6 +1,7 @@
 
  SECTIONS
  {
+    PROVIDE(_text = .);
      .text : { *(.init) *(.text) *(.text.*) }
      . = ALIGN(64K);
      PROVIDE(etext = .);
lib/arm64/processor.c +7 -0
index ef558625e284..831207c16587 100644
@@ -99,12 +99,19 @@ bool get_far(unsigned int esr, unsigned long *far)
  	return false;
  }
 
+extern unsigned long _text;
+
  static void bad_exception(enum vector v, struct pt_regs *regs,
  			  unsigned int esr, bool esr_valid, bool bad_vector)
  {
  	unsigned long far;
  	bool far_valid = get_far(esr, &far);
  	unsigned int ec = esr >> ESR_EL1_EC_SHIFT;
+	uintptr_t text = (uintptr_t)&_text;
+
+	printf("Load address: %" PRIxPTR "\n", text);
+	printf("PC: %" PRIxPTR " PC offset: %" PRIxPTR "\n",
+	       (uintptr_t)regs->pc, (uintptr_t)regs->pc - text);
 
  	if (bad_vector) {
  		if (v < VECTOR_MAX)
author Applied
#5 lore raw
In reply to Andrew Jones
Quoted 6 lines · Andrew Jones

Fix the loading of argc. Nikos reported an issue with its alignment while testing target-efi (aligned to four bytes is OK, as long as we only load four bytes like we should). Also, take a patch developed while working on target-efi which can make debugging a bit more convenient (by doing some subtraction for the test developer).

Quoted 12 lines

Andrew Jones (2):
arm64: argc is an int
arm64: Output PC load offset on unhandled exceptions

 arm/cstart64.S        | 2 +-
 arm/flat.lds          | 1 +
 lib/arm64/processor.c | 7 +++++++
 3 files changed, 9 insertions(+), 1 deletion(-)
Signature · 1 line
2.26.3

Thanks for the review, Nikos!

Applied to arm/queue

https://gitlab.com/rhdrjones/kvm-unit-tests/-/commits/arm/queue

Thanks,
drew