2024-01-02 07:42:37

by alexs

Subject: [PATCH] x86/dumpstack: uniform die messages

From: Alex Shi <[email protected]>

panic() has a uniform preempt, like "Kernel panic - not syncing:". But
die() doesn't have.

We died thousands machine with very different reasons. With current die
messages in dmesg, it's hard to write scripts to collect and anaylis the
die reasons. So we'd better give die() a uniform preempt to make the
life of system administrators a bit easier.

Signed-off-by: Alex Shi <[email protected]>
To: [email protected]
To: "H. Peter Anvin" <[email protected]>
To: [email protected]
To: Dave Hansen <[email protected]>
To: Borislav Petkov <[email protected]>
To: Ingo Molnar <[email protected]>
To: Thomas Gleixner <[email protected]>
To: Peter Zijlstra <[email protected]>
To: Josh Poimboeuf <[email protected]>
---
arch/x86/kernel/dumpstack.c | 6 ++++--
1 file changed, 4 insertions(+), 2 deletions(-)

diff --git a/arch/x86/kernel/dumpstack.c b/arch/x86/kernel/dumpstack.c
index f18ca44c904b..00766e7ac077 100644
--- a/arch/x86/kernel/dumpstack.c
+++ b/arch/x86/kernel/dumpstack.c
@@ -393,6 +393,8 @@ void oops_end(unsigned long flags, struct pt_regs *regs, int signr)
}
NOKPROBE_SYMBOL(oops_end);

+#define DIE_PREEMPT "DIE:\n"
+
static void __die_header(const char *str, struct pt_regs *regs, long err)
{
const char *pr = "";
@@ -405,8 +407,8 @@ static void __die_header(const char *str, struct pt_regs *regs, long err)
pr = IS_ENABLED(CONFIG_PREEMPT_RT) ? " PREEMPT_RT" : " PREEMPT";

printk(KERN_DEFAULT
- "%s: %04lx [#%d]%s%s%s%s%s\n", str, err & 0xffff, ++die_counter,
- pr,
+ DIE_PREEMPT "%s: %04lx [#%d]%s%s%s%s%s\n", str, err & 0xffff,
+ ++die_counter, pr,
IS_ENABLED(CONFIG_SMP) ? " SMP" : "",
debug_pagealloc_enabled() ? " DEBUG_PAGEALLOC" : "",
IS_ENABLED(CONFIG_KASAN) ? " KASAN" : "",
--
2.43.0

2024-01-03 12:45:33

by Alex Shi

Subject: Re: [PATCH] x86/dumpstack: uniform die messages

On Tue, Jan 2, 2024 at 3:42 PM <[email protected]> wrote:
>
> From: Alex Shi <[email protected]>
>
> panic() has a uniform preempt, like "Kernel panic - not syncing:". But
> die() doesn't have.
>
> We died thousands machine with very different reasons. With current die
> messages in dmesg, it's hard to write scripts to collect and anaylis the
> die reasons. So we'd better give die() a uniform preempt to make the
> life of system administrators a bit easier.

Sorry for the typo, s/preempt/prompt/,
And any comments are appreciated!

>
> Signed-off-by: Alex Shi <[email protected]>
> To: [email protected]
> To: "H. Peter Anvin" <[email protected]>
> To: [email protected]
> To: Dave Hansen <[email protected]>
> To: Borislav Petkov <[email protected]>
> To: Ingo Molnar <[email protected]>
> To: Thomas Gleixner <[email protected]>
> To: Peter Zijlstra <[email protected]>
> To: Josh Poimboeuf <[email protected]>
> ---
> arch/x86/kernel/dumpstack.c | 6 ++++--
> 1 file changed, 4 insertions(+), 2 deletions(-)
>
> diff --git a/arch/x86/kernel/dumpstack.c b/arch/x86/kernel/dumpstack.c
> index f18ca44c904b..00766e7ac077 100644
> --- a/arch/x86/kernel/dumpstack.c
> +++ b/arch/x86/kernel/dumpstack.c
> @@ -393,6 +393,8 @@ void oops_end(unsigned long flags, struct pt_regs *regs, int signr)
> }
> NOKPROBE_SYMBOL(oops_end);
>
> +#define DIE_PREEMPT "DIE:\n"
> +
> static void __die_header(const char *str, struct pt_regs *regs, long err)
> {
> const char *pr = "";
> @@ -405,8 +407,8 @@ static void __die_header(const char *str, struct pt_regs *regs, long err)
> pr = IS_ENABLED(CONFIG_PREEMPT_RT) ? " PREEMPT_RT" : " PREEMPT";
>
> printk(KERN_DEFAULT
> - "%s: %04lx [#%d]%s%s%s%s%s\n", str, err & 0xffff, ++die_counter,
> - pr,
> + DIE_PREEMPT "%s: %04lx [#%d]%s%s%s%s%s\n", str, err & 0xffff,
> + ++die_counter, pr,
> IS_ENABLED(CONFIG_SMP) ? " SMP" : "",
> debug_pagealloc_enabled() ? " DEBUG_PAGEALLOC" : "",
> IS_ENABLED(CONFIG_KASAN) ? " KASAN" : "",
> --
> 2.43.0
>