2019-06-20 06:53:21

by jinho lim

Subject: [PATCH v2] arm64: rename dump_instr as dump_kernel_instr

[v2]
dump_kernel_instr does not work for user mode.
rename dump_instr function and remove __dump_instr.

Signed-off-by: jinho lim <[email protected]>
---

Thanks for review, I rename dump_instr function and merge __dump_instr in it.

arch/arm64/kernel/traps.c | 29 ++++++++++++++---------------
1 file changed, 14 insertions(+), 15 deletions(-)

diff --git a/arch/arm64/kernel/traps.c b/arch/arm64/kernel/traps.c
index ccc13b45d9b1..7053165cb31a 100644
--- a/arch/arm64/kernel/traps.c
+++ b/arch/arm64/kernel/traps.c
@@ -66,11 +66,20 @@ static void dump_backtrace_entry(unsigned long where)
printk(" %pS\n", (void *)where);
}

-static void __dump_instr(const char *lvl, struct pt_regs *regs)
+static void dump_kernel_instr(const char *lvl, struct pt_regs *regs)
{
- unsigned long addr = instruction_pointer(regs);
+ unsigned long addr;
char str[sizeof("00000000 ") * 5 + 2 + 1], *p = str;
int i;
+ mm_segment_t fs;
+
+ if (user_mode(regs))
+ return;
+
+ addr = instruction_pointer(regs);
+
+ fs = get_fs();
+ set_fs(KERNEL_DS);

for (i = -4; i < 1; i++) {
unsigned int val, bad;
@@ -84,19 +93,10 @@ static void __dump_instr(const char *lvl, struct pt_regs *regs)
break;
}
}
+
printk("%sCode: %s\n", lvl, str);
-}

-static void dump_instr(const char *lvl, struct pt_regs *regs)
-{
- if (!user_mode(regs)) {
- mm_segment_t fs = get_fs();
- set_fs(KERNEL_DS);
- __dump_instr(lvl, regs);
- set_fs(fs);
- } else {
- __dump_instr(lvl, regs);
- }
+ set_fs(fs);
}

void dump_backtrace(struct pt_regs *regs, struct task_struct *tsk)
@@ -182,8 +182,7 @@ static int __die(const char *str, int err, struct pt_regs *regs)
print_modules();
show_regs(regs);

- if (!user_mode(regs))
- dump_instr(KERN_EMERG, regs);
+ dump_kernel_instr(KERN_EMERG, regs);

return ret;
}
--
2.17.1

2019-06-24 14:15:38

by Will Deacon

Subject: Re: [PATCH v2] arm64: rename dump_instr as dump_kernel_instr

On Thu, Jun 20, 2019 at 03:52:49PM +0900, jinho lim wrote:
> [v2]

The version information is not usually part of the commit message. Please
drop that...

> dump_kernel_instr does not work for user mode.
> rename dump_instr function and remove __dump_instr.

... and rewrite this so it explains the problem that you're solving.

> Signed-off-by: jinho lim <[email protected]>
> ---
>
> Thanks for review, I rename dump_instr function and merge __dump_instr in it.
>
> arch/arm64/kernel/traps.c | 29 ++++++++++++++---------------
> 1 file changed, 14 insertions(+), 15 deletions(-)
>
> diff --git a/arch/arm64/kernel/traps.c b/arch/arm64/kernel/traps.c
> index ccc13b45d9b1..7053165cb31a 100644
> --- a/arch/arm64/kernel/traps.c
> +++ b/arch/arm64/kernel/traps.c
> @@ -66,11 +66,20 @@ static void dump_backtrace_entry(unsigned long where)
> printk(" %pS\n", (void *)where);
> }
>
> -static void __dump_instr(const char *lvl, struct pt_regs *regs)
> +static void dump_kernel_instr(const char *lvl, struct pt_regs *regs)
> {
> - unsigned long addr = instruction_pointer(regs);
> + unsigned long addr;
> char str[sizeof("00000000 ") * 5 + 2 + 1], *p = str;
> int i;
> + mm_segment_t fs;
> +
> + if (user_mode(regs))
> + return;
> +
> + addr = instruction_pointer(regs);
> +
> + fs = get_fs();
> + set_fs(KERNEL_DS);

Actually, if we use aarch64_insn_read() instead of get_user() then we can
avoid having to mess directly with the fs and we'll also get endianness
correction for free when running a big-endian kernel.

Will