arm64: Changes to register and stack dumps

We can decode the PSTATE easily enough, so pretty-print it in register dumps.

Signed-off-by Will Deacon <[email protected]>

---
 arch/arm64/kernel/process.c | 20 +++++++++++++++++++-
 1 file changed, 19 insertions(+), 1 deletion(-)

arch/arm64/kernel/process.c +6 -0

index c20896b8fb2d..be3cf9cd7dec 100644
@@ -170,6 +170,23 @@ void machine_restart(char *cmd)
 	while (1);
 }
 
+static void print_pstate(struct pt_regs *regs)
+{
+	u64 pstate = regs->pstate;
+	printk("pstate: %08llx (%c%c%c%c %c%c%c%c %cPAN %cUAO)\n",
+		pstate,
+		pstate & PSR_N_BIT ? 'N' : 'n',

arch/arm64/kernel/process.c +3 -1

+
 void __show_regs(struct pt_regs *regs)
 {
 	int i, top_reg;

@@ -186,9 +203,10 @@ void __show_regs(struct pt_regs *regs)
 	}
 
 	show_regs_print_info(KERN_DEFAULT);
+	print_pstate(regs);
 	print_symbol("pc : %s\n", regs->pc);
 	print_symbol("lr : %s\n", lr);
-	printk("sp : %016llx pstate : %08llx\n", sp, regs->pstate);
+	printk("sp : %016llx\n", sp);
 
 	i = top_reg;
 

Signature · 1 line

2.1.4

Quoted 22 lines · Will Deacon

We can decode the PSTATE easily enough, so pretty-print it in register dumps.

Signed-off-by Will Deacon <[email protected]>

---
 arch/arm64/kernel/process.c | 20 +++++++++++++++++++-
 1 file changed, 19 insertions(+), 1 deletion(-)

arch/arm64/kernel/process.c +6 -0

index c20896b8fb2d..be3cf9cd7dec 100644
@@ -170,6 +170,23 @@ void machine_restart(char *cmd)
 	while (1);
 }
 
+static void print_pstate(struct pt_regs *regs)
+{
+	u64 pstate = regs->pstate;
+	printk("pstate: %08llx (%c%c%c%c %c%c%c%c %cPAN %cUAO)\n",
+		pstate,
+		pstate & PSR_N_BIT ? 'N' : 'n',

arch/arm64/kernel/process.c void machine_restart(char *cmd) · line 188

Quoted 10 lines

arch/arm64/kernel/process.c +10 -0

+		pstate & PSR_Z_BIT ? 'Z' : 'z',
+		pstate & PSR_C_BIT ? 'C' : 'c',
+		pstate & PSR_V_BIT ? 'V' : 'v',
+		pstate & PSR_D_BIT ? 'D' : 'd',
+		pstate & PSR_A_BIT ? 'A' : 'a',
+		pstate & PSR_I_BIT ? 'I' : 'i',
+		pstate & PSR_F_BIT ? 'F' : 'f',
+		pstate & PSR_PAN_BIT ? '+' : '-',
+		pstate & PSR_UAO_BIT ? '+' : '-');
+}

Given we use __show_regs for user state, I think we need to check compat_user_mode() regs, and adjust the decoding accordingly.

e.g. AArch32 has an E bit where the D in DAIF is.

Quoted 37 lines · Mark Rutland

Quoted 22 lines · Will Deacon

We can decode the PSTATE easily enough, so pretty-print it in register dumps.

Signed-off-by Will Deacon <[email protected]>

---
 arch/arm64/kernel/process.c | 20 +++++++++++++++++++-
 1 file changed, 19 insertions(+), 1 deletion(-)

arch/arm64/kernel/process.c +6 -0

index c20896b8fb2d..be3cf9cd7dec 100644
@@ -170,6 +170,23 @@ void machine_restart(char *cmd)
 	while (1);
 }
 
+static void print_pstate(struct pt_regs *regs)
+{
+	u64 pstate = regs->pstate;
+	printk("pstate: %08llx (%c%c%c%c %c%c%c%c %cPAN %cUAO)\n",
+		pstate,
+		pstate & PSR_N_BIT ? 'N' : 'n',

arch/arm64/kernel/process.c void machine_restart(char *cmd) · line 188

Quoted 10 lines

arch/arm64/kernel/process.c +10 -0

+		pstate & PSR_Z_BIT ? 'Z' : 'z',
+		pstate & PSR_C_BIT ? 'C' : 'c',
+		pstate & PSR_V_BIT ? 'V' : 'v',
+		pstate & PSR_D_BIT ? 'D' : 'd',
+		pstate & PSR_A_BIT ? 'A' : 'a',
+		pstate & PSR_I_BIT ? 'I' : 'i',
+		pstate & PSR_F_BIT ? 'F' : 'f',
+		pstate & PSR_PAN_BIT ? '+' : '-',
+		pstate & PSR_UAO_BIT ? '+' : '-');
+}

Given we use __show_regs for user state, I think we need to check compat_user_mode() regs, and adjust the decoding accordingly.

Quoted 1 line

e.g. AArch32 has an E bit where the D in DAIF is.

Yes, you're right. There's lots of exciting bits in the AArch32 pstate too. I ended up with the diff below.

Will

--->8

(quoted diff) +4 -0

+static void print_pstate(struct pt_regs *regs)
+{
+	u64 pstate = regs->pstate;
+

Hi all,

This handful of patches makes some significant changes to how we print register/stack dumps and brings us inline with recent changes to x86. The main changes are that we no longer print out the 'Exception stack:' section and PC/LR values are resolved to a symbol+offset format where possible.

Feedback welcome.

Will

--->8

Mark Rutland (1):
arm64: consistently log ESR and page table

Will Deacon (3):
arm64: traps: Don't print stack or raw PC/LR values in backtraces
arm64: traps: Pretty-print pstate in register dumps
arm64: uapi: Remove PSR_Q_BIT

 arch/arm64/include/uapi/asm/ptrace.h |  1 -
 arch/arm64/kernel/process.c          | 24 ++++++++++---
 arch/arm64/kernel/traps.c            | 65 ++----------------------------------
 arch/arm64/mm/fault.c                |  7 ++--
 4 files changed, 28 insertions(+), 69 deletions(-)

Quoted 6 lines · Will Deacon

This handful of patches makes some significant changes to how we print register/stack dumps and brings us inline with recent changes to x86. The main changes are that we no longer print out the 'Exception stack:' section and PC/LR values are resolved to a symbol+offset format where possible.

Quoted 2 lines

Feedback welcome.

In very petty complaints, I can no longer just use the PC/LR with addr2line. Yes I can use gdb or one of the scripts but that's a few more steps when I'm doing debugging on my own. Like I said, very minor and overall this is good for both security and alignment.

I ran this through several of the LKDTM tests and the backtraces looked fine. You're welcome to add

Tested-by Laura Abbott <[email protected]>