Skip to contentSkip to conversation

panic: fix panic_force_cpu= redirect races and NMI bypass

PATCH In review Bradley Morgan 3 participants 14 messages Aug 18 – Aug 25

Subsystems (3)
lore
Share & tools
lore
#0 lore raw

v6 cover panic: fix panic_force_cpu= redirect races and NMI bypass

The panic_force_cpu= parameter redirects a panic to a specific CPU so the crash kernel runs there. The redirect code in panic_try_force_cpu() had two races and an NMI bypass, all found by Sashiko. This series closes them and kills one more hack that the rework turned up.

Patch 1: fix the redirect CPU race.

The redirect is gated by an atomic cmpxchg on panic_redirect_cpu, so only one CPU sends the redirect IPI. The cmpxchg loser used to return false and fall through into vpanic(), where it could win panic_try_start() and run crash_kexec on the wrong CPU before the target ever received the IPI. The loser has to stop. It cannot just return true, though, because panic_try_force_cpu() can be called twice on the same CPU (a nested NMI during the message formatting, before the IPI is sent), and a blind stop on that second call would abandon the panic with no IPI sent. The loser now returns true to stop, unless it is reentering on the same CPU (old_cpu == this_cpu), in which case it returns false and falls through.

Patch 1 also fixes the panic_in_progress() guard and moves the panic_try_force_cpu() doc comment update for the new return values here, as requested by Petr.

The race, two non target CPUs A and B (target is C):

cpu A cpu B
---------- ----------
panic_try_force_cpu() panic_try_force_cpu()
cmpxchg wins cmpxchg fails
IPI -> C return false <- old BUG
return true panic_try_start() wins
panic_smp_self_stop() __crash_kexec() on B
(A stops) (target C bypassed)

Patch 2: flatten nmi_panic control flow.

A behavior preserving cleanup. panic() is noreturn, so the else after it is dropped and the body flattened, ready for patch 5 to add the redirect step without piling more onto the if else chain.

Patch 3: fix va_list reuse in panic_try_force_cpu().

vsnprintf() consumes the caller's va_list. When the redirect fails, vpanic() reuses it for the panic message, which is undefined behavior. Fixed with va_copy(). Already reviewed by Petr, unchanged since v5.

Patch 4: restore variable arguments to nmi_panic().

nmi_panic() used to take variable arguments until commit ebc41f20d77f ("panic: change nmi_panic from macro to function") flattened it to a final message string. Bring them back and format with vpanic() directly, so the next patch can hand the arguments to both panic_try_force_cpu() and vpanic() without any pointer passing. Every existing caller passes a plain string literal, so nothing changes for them.

Patch 5: allow force_cpu redirect from an NMI.

A panic from an NMI used to bypass the redirect entirely. nmi_panic() called panic_try_start() first, which claims panic_cpu, so by the time panic() reached panic_try_force_cpu() the panic_in_progress() check saw panic_cpu set, returned false, and never sent the redirect IPI. The crash kernel ran on the CPU that took the NMI instead of the requested one.

nmi_panic() now calls panic_try_force_cpu() before claiming panic_cpu, as suggested by Petr. The requested CPU claims panic_cpu itself when it runs panic(), so panic_cpu does not need to be handed off.

Patch 6: kill the "buffer unavailable" redirect fallback.

The redirect buffer is kmalloc'ed in a late_initcall with no error check. Until then the redirect hands the target CPU the message "Redirected panic (buffer unavailable)", so the crash kernel knows that it panicked but not why. The whole boot is that window. Make it a static 1KB buffer and kill the initcall.

Bradley Morgan (6):
panic: fix redirect CPU race in panic_try_force_cpu()
panic: flatten nmi_panic control flow
panic: fix va_list reuse in panic_try_force_cpu()
panic: restore variable arguments to nmi_panic()
panic: allow force_cpu redirect from an NMI
panic: kill the "buffer unavailable" redirect fallback

 include/linux/panic.h |  3 +-
 kernel/panic.c        | 79 ++++++++++++++++++++++++-------------------
 2 files changed, 47 insertions(+), 35 deletions(-)

--
2.47.3

---
Changes since v5 (all from Petr's review of 4/4):
- Dropped the va_list pointer and the NULL means final message convention. panic_try_force_cpu() takes a plain va_list again.
- Dropped the va_copy() in vpanic(). panic_try_force_cpu() already copies before formatting, so the caller's arguments are reusable.
- New patch 4 restores the variable arguments of nmi_panic() that ebc41f20d77f removed (solution B), as its own patch.
- The doc comment change for the new return values moved into patch 1.
- Patch 5 is now just the redirect block and the comment update, as sketched in Petr's review.
- New patch 6 kills the redirect buffer hack that the rework turned up.

#1 lore raw

PATCH v6 1/6 panic: fix redirect CPU race in panic_try_force_cpu()

The cmpxchg() in panic_try_force_cpu() makes sure that only one CPU tries to redirect panic() to the requested CPU. It is similar to the cmpxchg() in panic_try_start() which makes sure that only one CPU does the panic(). In both situations, only the winner of cmpxchg() should proceed further. Other CPUs should go offline.

There is a bug because the cmpxchg loser returns false and falls through into vpanic(). Two non-target CPUs A and B panic, the requested CPU is C:

cpu A cpu B
---------- ----------
panic() panic()
vpanic() vpanic()
panic_try_force_cpu() panic_try_force_cpu()
cmpxchg wins cmpxchg fails
redirect = A old_cpu = A
IPI -> C return false <- BUG
return true panic_try_start() wins
panic_smp_self_stop() __crash_kexec() on B
(A stops) (target C bypassed)

The loser must stop, not fall through. It cannot just return true, though. A CPU that already won the redirect cmpxchg can reenter panic_try_force_cpu() on the same CPU, for example a nested NMI during the message formatting, before the IPI is sent:

             cpu A (1st)                  cpu A (nested)
          ----------                     ----------
    panic()
    vpanic()
    panic_try_force_cpu()
        cmpxchg wins (redirect = A)
        vsnprintf(msg) ...
            <-- NMI, nested panic -->
                                     panic()
                                     vpanic()
                                     panic_try_force_cpu()
                                         cmpxchg fails
                                         old_cpu == A (this CPU)
                                         return true   <- would halt
                                     panic_smp_self_stop()
                                     (IPI never sent, panic abandoned)

Check old_cpu against this_cpu so a second call from the same CPU returns false and falls through to panic_try_start() instead.

Also fix the panic_in_progress() check. We must not redirect when panic_cpu is already assigned. Return true to stop when the panic is on another CPU, false to proceed when it is this one.

Update the panic_try_force_cpu() doc comment for the new return value semantics.

Fixes 2e171ab29f91 ("panic: add panic_force_cpu= parameter to redirect panic to a specific CPU")
---
 kernel/panic.c | 19 ++++++++++++-------
 1 file changed, 12 insertions(+), 7 deletions(-)
kernel/panic.c +12 -7
index 213725b612aa..010b331658b6 100644
@@ -371,8 +371,9 @@ int __weak panic_smp_redirect_cpu(int target_cpu, void *msg)
  * for the crash kernel to function correctly. This function redirects
  * panic handling to the CPU specified via the panic_force_cpu= boot parameter.
  *
- * Returns false if panic should proceed on current CPU.
- * Returns true if panic was redirected.
+ * Returns true when this CPU must stop: the panic was redirected or is
+ * already running on another CPU.
+ * Returns false when panic() should proceed on this CPU.
  */
 __printf(1, 0)
 static bool panic_try_force_cpu(const char *fmt, va_list args)
@@ -396,16 +397,20 @@ static bool panic_try_force_cpu(const char *fmt, va_list args)
 		return false;
 	}
 
-	/* Another panic already in progress */
+	/*
+	 * Don't redirect when a panic is already in progress. Stop this
+	 * CPU when it's another one, proceed when it's this one.
+	 */
 	if (panic_in_progress())
-		return false;
+		return panic_on_other_cpu();
 
 	/*
-	 * Only one CPU can do the redirect. Use atomic cmpxchg to ensure
-	 * we don't race with another CPU also trying to redirect.
+	 * Only one CPU can do the redirection. Others should go offline.
+	 * Continue with panic() when we already tried the redirection
+	 * from this CPU before, for example via nmi_panic().
 	 */
 	if (!atomic_try_cmpxchg(&panic_redirect_cpu, &old_cpu, this_cpu))
-		return false;
+		return old_cpu != this_cpu;
 
 	/*
 	 * Use dynamically allocated buffer if available, otherwise
Signature · 1 line
2.47.3
#2 lore raw

Re: [PATCH v6 1/6] panic: fix redirect CPU race in panic_try_force_cpu()

Quoted 71 lines · Bradley Morgan

The cmpxchg() in panic_try_force_cpu() makes sure that only one CPU tries to redirect panic() to the requested CPU. It is similar to the cmpxchg() in panic_try_start() which makes sure that only one CPU does the panic(). In both situations, only the winner of cmpxchg() should proceed further. Other CPUs should go offline.

There is a bug because the cmpxchg loser returns false and falls through into vpanic(). Two non-target CPUs A and B panic, the requested CPU is C:

cpu A cpu B
---------- ----------
panic() panic()
vpanic() vpanic()
panic_try_force_cpu() panic_try_force_cpu()
cmpxchg wins cmpxchg fails
redirect = A old_cpu = A
IPI -> C return false <- BUG
return true panic_try_start() wins
panic_smp_self_stop() __crash_kexec() on B
(A stops) (target C bypassed)

The loser must stop, not fall through. It cannot just return true, though. A CPU that already won the redirect cmpxchg can reenter panic_try_force_cpu() on the same CPU, for example a nested NMI during the message formatting, before the IPI is sent:

             cpu A (1st)                  cpu A (nested)
          ----------                     ----------
    panic()
    vpanic()
    panic_try_force_cpu()
        cmpxchg wins (redirect = A)
        vsnprintf(msg) ...
            <-- NMI, nested panic -->
                                     panic()
                                     vpanic()
                                     panic_try_force_cpu()
                                         cmpxchg fails
                                         old_cpu == A (this CPU)
                                         return true   <- would halt
                                     panic_smp_self_stop()
                                     (IPI never sent, panic abandoned)

Check old_cpu against this_cpu so a second call from the same CPU returns false and falls through to panic_try_start() instead.

Also fix the panic_in_progress() check. We must not redirect when panic_cpu is already assigned. Return true to stop when the panic is on another CPU, false to proceed when it is this one.

Update the panic_try_force_cpu() doc comment for the new return value semantics.

kernel/panic.c +5 -4
@@ -396,16 +397,20 @@ static bool panic_try_force_cpu(const char *fmt, va_list args)
 		return false;
 	}
 
-	/* Another panic already in progress */
+	/*
+	 * Don't redirect when a panic is already in progress. Stop this
+	 * CPU when it's another one, proceed when it's this one.
+	 */
 	if (panic_in_progress())
-		return false;
+		return panic_on_other_cpu();
 
 	/*
-	 * Only one CPU can do the redirect. Use atomic cmpxchg to ensure
-	 * we don't race with another CPU also trying to redirect.

kernel/panic.c static bool panic_try_force_cpu(const char *fmt, va_list args) · line 416

Quoted 10 lines
kernel/panic.c +4 -1
+	 * Only one CPU can do the redirection. Others should go offline.
+	 * Continue with panic() when we already tried the redirection
+	 * from this CPU before, for example via nmi_panic().
 	 */
 	if (!atomic_try_cmpxchg(&panic_redirect_cpu, &old_cpu, this_cpu))
-		return false;
+		return old_cpu != this_cpu;
 
 	/*
 	 * Use dynamically allocated buffer if available, otherwise

Just for completeness. Sashiko AI points out that the vsnprintf() in panic_try_force_cpu() uses the "args" and they might later be used again when the redirection fails, see
https://sashiko.dev/#/patchset/20260818163806.17460-1-include%40grrlz.net

This problem is fixed in by the 3rd patch in this patchset, see
https://lore.kernel.org/all/[email protected]/

So, we are on the safe side.

Best Regards,
Petr

#3 lore raw

PATCH v6 2/6 panic: flatten nmi_panic control flow

In reply to Bradley Morgan

panic() is __noreturn, so the else after panic_try_start() is dead. Drop it so the force_cpu path can be added cleanly on top.

---
 kernel/panic.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)
kernel/panic.c +2 -1
index 010b331658b6..e1b443150ba0 100644
@@ -517,7 +517,8 @@ void nmi_panic(struct pt_regs *regs, const char *msg)
 {
 	if (panic_try_start())
 		panic("%s", msg);
-	else if (panic_on_other_cpu())
+
+	if (panic_on_other_cpu())
 		nmi_panic_self_stop(regs);
 }
 EXPORT_SYMBOL(nmi_panic);
Signature · 1 line
2.47.3
#4 lore raw

PATCH v6 3/6 panic: fix va_list reuse in panic_try_force_cpu()

In reply to Bradley Morgan

vsnprintf() consumes the caller's va_list. When the redirect fails, vpanic() reuses it for the panic message, which is undefined behavior. Use va_copy().

Fixes 2e171ab29f91 ("panic: add panic_force_cpu= parameter to redirect panic to a specific CPU")
---
 kernel/panic.c | 7 ++++++-
 1 file changed, 6 insertions(+), 1 deletion(-)
kernel/panic.c +6 -1
index e1b443150ba0..6b5728c3c9ce 100644
@@ -417,7 +417,12 @@ static bool panic_try_force_cpu(const char *fmt, va_list args)
 	 * fall back to static message for early boot panics or allocation failure.
 	 */
 	if (panic_force_buf) {
-		vsnprintf(panic_force_buf, PANIC_MSG_BUFSZ, fmt, args);
+		va_list ap;
+
+		/* Do not consume args, the caller reuses it if we fail */
+		va_copy(ap, args);
+		vsnprintf(panic_force_buf, PANIC_MSG_BUFSZ, fmt, ap);
+		va_end(ap);
 		msg = panic_force_buf;
 	} else {
 		msg = "Redirected panic (buffer unavailable)";
Signature · 1 line
2.47.3
#5 lore raw

PATCH v6 4/6 panic: restore variable arguments to nmi_panic()

In reply to Bradley Morgan

nmi_panic() used to accept variable arguments until commit ebc41f20d77f ("panic: change nmi_panic from macro to function") flattened it to a final message string. vpanic() did not exist back then, so the function had to format through panic("%s", msg).

Bring the variable arguments back and format with vpanic() directly. The next patch makes nmi_panic() try the panic_force_cpu= redirect before claiming panic_cpu, which needs the arguments twice: once to format the message for the redirected CPU and once for vpanic() when no redirect happens. Passing a final string would lose that.

Every existing caller passes a plain string literal with no format specifiers, so nothing changes for them.

---
 include/linux/panic.h |  3 ++-
 kernel/panic.c        | 11 +++++++++--
 2 files changed, 11 insertions(+), 3 deletions(-)
include/linux/panic.h +2 -1
index f1dd417e54b2..a9128bf3c168 100644
@@ -13,7 +13,8 @@ __printf(1, 2)
 void panic(const char *fmt, ...) __noreturn __cold;
 __printf(1, 0)
 void vpanic(const char *fmt, va_list args) __noreturn __cold;
-void nmi_panic(struct pt_regs *regs, const char *msg);
+__printf(2, 3)
+void nmi_panic(struct pt_regs *regs, const char *fmt, ...);
 void check_panic_on_warn(const char *origin);
 extern void oops_enter(void);
 extern void oops_exit(void);
kernel/panic.c +9 -2
index 6b5728c3c9ce..bc142485faa4 100644
@@ -518,13 +518,20 @@ EXPORT_SYMBOL(panic_on_other_cpu);
  * nmi_panic_self_stop() which can provide architecture dependent code such
  * as saving register state for crash dump.
  */
-void nmi_panic(struct pt_regs *regs, const char *msg)
+__printf(2, 3)
+void nmi_panic(struct pt_regs *regs, const char *fmt, ...)
 {
+	va_list args;
+
+	va_start(args, fmt);
+
 	if (panic_try_start())
-		panic("%s", msg);
+		vpanic(fmt, args);
 
 	if (panic_on_other_cpu())
 		nmi_panic_self_stop(regs);
+
+	va_end(args);
 }
 EXPORT_SYMBOL(nmi_panic);
 
Signature · 1 line
2.47.3
#6 lore raw

Re: [PATCH v6 4/6] panic: restore variable arguments to nmi_panic()

Quoted 5 lines · Bradley Morgan

nmi_panic() used to accept variable arguments until commit ebc41f20d77f ("panic: change nmi_panic from macro to function") flattened it to a final message string. vpanic() did not exist back then, so the function had to format through panic("%s", msg).

Quoted 8 lines

Bring the variable arguments back and format with vpanic() directly. The next patch makes nmi_panic() try the panic_force_cpu= redirect before claiming panic_cpu, which needs the arguments twice: once to format the message for the redirected CPU and once for vpanic() when no redirect happens. Passing a final string would lose that.

Every existing caller passes a plain string literal with no format specifiers, so nothing changes for them.

Sashiko AI complains, see
https://sashiko.dev/#/patchset/20260818163806.17460-1-include%40grrlz.net

| Is this assertion accurate? Looking at hpwdt_pretimeout() in
| drivers/watchdog/hpwdt.c, it constructs a dynamic string before passing it:
| 
| drivers/watchdog/hpwdt.c:hpwdt_pretimeout() {
| 	...
| 	hex_byte_pack(panic_msg, nmistat);
| 	nmi_panic(regs, panic_msg);
| 	...
| }

It is true that @panic_msg is a pointer to a string. But there are only two variants and both are plain strings with no format specifiers.

Well, we will update the commit message anyway, see below.

include/linux/panic.h __printf(1, 2) · line 17

Quoted 12 lines
include/linux/panic.h +2 -1
@@ -13,7 +13,8 @@ __printf(1, 2)
 void panic(const char *fmt, ...) __noreturn __cold;
 __printf(1, 0)
 void vpanic(const char *fmt, va_list args) __noreturn __cold;
-void nmi_panic(struct pt_regs *regs, const char *msg);
+__printf(2, 3)
+void nmi_panic(struct pt_regs *regs, const char *fmt, ...);

Here Sashiko says:

| Will this __printf() annotation cause a -Wformat-security build failure in
| hpwdt_pretimeout() when compiled with CONFIG_HPWDT_NMI_DECODING, since
| panic_msg is passed directly as the format argument without a "%s"
| specifier?

And it is right. I have reproduced it. I have explictitely added -Wformat-security and got:

# CC drivers/watchdog/hpwdt.o
drivers/watchdog/hpwdt.c: In function ‘hpwdt_pretimeout’:
drivers/watchdog/hpwdt.c:202:9: warning: format not a string literal and no format arguments [-Wformat-security]

  202 |         nmi_panic(regs, panic_msg);
      |         ^~~~~~~~~

So, we should add the %s format to be on the safe side. The following works:

drivers/watchdog/hpwdt.c +1 -1
@@ -199,7 +199,7 @@ static int hpwdt_pretimeout(unsigned int ulReason, struct pt_regs *regs)
 	}
 
 	hex_byte_pack(panic_msg, nmistat);
-	nmi_panic(regs, panic_msg);
+	nmi_panic(regs, "%s", panic_msg);
 
 	return NMI_HANDLED;
 }

We should do this change in this patch and mention it in the commit message which should prevent the earlier complaint.

kernel/panic.c EXPORT_SYMBOL(panic_on_other_cpu); · line 521

Quoted 13 lines
include/linux/panic.h +0 -0
 void check_panic_on_warn(const char *origin);
 extern void oops_enter(void);
 extern void oops_exit(void);
kernel/panic.c +1 -1
index 6b5728c3c9ce..bc142485faa4 100644
@@ -518,13 +518,20 @@ EXPORT_SYMBOL(panic_on_other_cpu);
  * nmi_panic_self_stop() which can provide architecture dependent code such
  * as saving register state for crash dump.
  */
-void nmi_panic(struct pt_regs *regs, const char *msg)
+__printf(2, 3)

This is not needed. It is enough to declare __printf() in the header file.

Quoted 6 lines
kernel/panic.c +5 -0
+void nmi_panic(struct pt_regs *regs, const char *fmt, ...)
 {
+	va_list args;
+
+	va_start(args, fmt);
+

kernel/panic.c EXPORT_SYMBOL(panic_on_other_cpu); · line 536

Quoted 10 lines
kernel/panic.c +3 -1
 	if (panic_try_start())
-		panic("%s", msg);
+		vpanic(fmt, args);
 
 	if (panic_on_other_cpu())
 		nmi_panic_self_stop(regs);
+
+	va_end(args);
 }
 EXPORT_SYMBOL(nmi_panic);

Otherwise, it looks good to me.

Best Regards,
Petr

#7 lore raw

PATCH v6 5/6 panic: allow force_cpu redirect from an NMI

In reply to Bradley Morgan

nmi_panic() claims panic_cpu via panic_try_start() before calling panic(). When the panic later reaches panic_try_force_cpu(), the panic_in_progress() check sees panic_cpu set and refuses to redirect. The crash kernel runs on the CPU that took the NMI instead of the CPU requested with panic_force_cpu=:

nmi_panic()
panic_try_start() wins, panic_cpu = X
panic("%s", msg)
vpanic()
panic_try_force_cpu()
panic_in_progress() true, panic_cpu is X
return false redirect bypassed
panic_try_start() already won
__crash_kexec() on X, not the requested CPU

Try the redirect before claiming panic_cpu instead, as suggested by Petr Mladek. nmi_panic() now calls panic_try_force_cpu() first and claims panic_cpu only when no redirect happened. The requested CPU claims panic_cpu itself when it runs panic(), so panic_cpu does not need to be handed off. panic_try_force_cpu() copies the arguments before formatting (patch 3), so nmi_panic() can pass them to vpanic() again when no redirect happens.

The redirect IPI is sent with smp_call_function_single_async(), which is not guaranteed to work from NMI context. Treat it as best effort. It is worth the risk because the redirection is only used when the crash kernel would not work on the panicking CPU anyway.

Keep returning when the panic is already running on this CPU. A nested NMI, for example with unknown_nmi_panic while this CPU is inside panic(), must return and let the interrupted panic() continue instead of parking the CPU in nmi_panic_self_stop().

Mark the redirecting CPU offline before stopping it, like vpanic() does, so that panic_other_cpus_shutdown() on the target CPU does not wait for it.

Fixes 2e171ab29f91 ("panic: add panic_force_cpu= parameter to redirect panic to a specific CPU")
---
 kernel/panic.c | 19 +++++++++++++++----
 1 file changed, 15 insertions(+), 4 deletions(-)
kernel/panic.c +15 -4
index bc142485faa4..356c03f2361c 100644
@@ -513,10 +513,11 @@ bool panic_on_other_cpu(void)
 EXPORT_SYMBOL(panic_on_other_cpu);
 
 /*
- * A variant of panic() called from NMI context. We return if we've already
- * panicked on this CPU. If another CPU already panicked, loop in
- * nmi_panic_self_stop() which can provide architecture dependent code such
- * as saving register state for crash dump.
+ * A variant of panic() called from NMI context. The panic is first
+ * redirected to the CPU requested via panic_force_cpu=, when configured.
+ * We return if we've already panicked on this CPU. If another CPU already
+ * panicked, loop in nmi_panic_self_stop() which can provide architecture
+ * dependent code such as saving register state for crash dump.
  */
 __printf(2, 3)
 void nmi_panic(struct pt_regs *regs, const char *fmt, ...)
@@ -525,6 +526,16 @@ void nmi_panic(struct pt_regs *regs, const char *fmt, ...)
 
 	va_start(args, fmt);
 
+	/* Try to redirect to the requested CPU before claiming panic_cpu. */
+	if (panic_try_force_cpu(fmt, args)) {
+		/*
+		 * Mark ourselves offline so panic_other_cpus_shutdown() won't
+		 * wait for us on architectures that check num_online_cpus().
+		 */
+		set_cpu_online(raw_smp_processor_id(), false);
+		nmi_panic_self_stop(regs);
+	}
+
 	if (panic_try_start())
 		vpanic(fmt, args);
 
Signature · 1 line
2.47.3
Reviewed-by on v6 5/6
#8 lore raw

Re: [PATCH v6 5/6] panic: allow force_cpu redirect from an NMI

Quoted 38 lines · Bradley Morgan

nmi_panic() claims panic_cpu via panic_try_start() before calling panic(). When the panic later reaches panic_try_force_cpu(), the panic_in_progress() check sees panic_cpu set and refuses to redirect. The crash kernel runs on the CPU that took the NMI instead of the CPU requested with panic_force_cpu=:

nmi_panic()
panic_try_start() wins, panic_cpu = X
panic("%s", msg)
vpanic()
panic_try_force_cpu()
panic_in_progress() true, panic_cpu is X
return false redirect bypassed
panic_try_start() already won
__crash_kexec() on X, not the requested CPU

Try the redirect before claiming panic_cpu instead, as suggested by Petr Mladek. nmi_panic() now calls panic_try_force_cpu() first and claims panic_cpu only when no redirect happened. The requested CPU claims panic_cpu itself when it runs panic(), so panic_cpu does not need to be handed off. panic_try_force_cpu() copies the arguments before formatting (patch 3), so nmi_panic() can pass them to vpanic() again when no redirect happens.

The redirect IPI is sent with smp_call_function_single_async(), which is not guaranteed to work from NMI context. Treat it as best effort. It is worth the risk because the redirection is only used when the crash kernel would not work on the panicking CPU anyway.

Keep returning when the panic is already running on this CPU. A nested NMI, for example with unknown_nmi_panic while this CPU is inside panic(), must return and let the interrupted panic() continue instead of parking the CPU in nmi_panic_self_stop().

Mark the redirecting CPU offline before stopping it, like vpanic() does, so that panic_other_cpus_shutdown() on the target CPU does not wait for it.

Quoted 5 lines
Fixes 2e171ab29f91 ("panic: add panic_force_cpu= parameter to redirect panic to a specific CPU")

The patch looks good to me:

See some comments below.

Quoted 19 lines
kernel/panic.c +5 -4
@@ -513,10 +513,11 @@ bool panic_on_other_cpu(void)
 EXPORT_SYMBOL(panic_on_other_cpu);
 
 /*
- * A variant of panic() called from NMI context. We return if we've already
- * panicked on this CPU. If another CPU already panicked, loop in
- * nmi_panic_self_stop() which can provide architecture dependent code such
- * as saving register state for crash dump.
+ * A variant of panic() called from NMI context. The panic is first
+ * redirected to the CPU requested via panic_force_cpu=, when configured.
+ * We return if we've already panicked on this CPU. If another CPU already
+ * panicked, loop in nmi_panic_self_stop() which can provide architecture
+ * dependent code such as saving register state for crash dump.
  */
 __printf(2, 3)
 void nmi_panic(struct pt_regs *regs, const char *fmt, ...)

kernel/panic.c void nmi_panic(struct pt_regs *regs, const char *fmt, ...) · line 535

Quoted 10 lines
kernel/panic.c +7 -0
@@ -525,6 +526,16 @@ void nmi_panic(struct pt_regs *regs, const char *fmt, ...)
 
 	va_start(args, fmt);
 
+	/* Try to redirect to the requested CPU before claiming panic_cpu. */
+	if (panic_try_force_cpu(fmt, args)) {
+		/*
+		 * Mark ourselves offline so panic_other_cpus_shutdown() won't
+		 * wait for us on architectures that check num_online_cpus().
+		 */
+		set_cpu_online(raw_smp_processor_id(), false);

It is a bit strange that we set this CPU offline in this code path but not when the below panic_try_start() fails. nmi_panic_self_stop(regs) is called in both situations.

I guess that we should mark the CPU offline in the other case as well. But it is an indepent change. We could fix this later.

Feel free to keep this patch as is.

kernel/panic.c void nmi_panic(struct pt_regs *regs, const char *fmt, ...) · line 540

Quoted 5 lines
kernel/panic.c +3 -0
+		nmi_panic_self_stop(regs);
+	}
+
 	if (panic_try_start())
 		vpanic(fmt, args);

Sashiko AI complains here:

| Does this sequence result in printing the 'target CPU is offline' warning
| twice?
| 
| If panic_force_cpu is set to an offline CPU, the newly added call to
| panic_try_force_cpu() evaluates this logic:
| 
| 	/* Target CPU is offline, can't redirect */
| 	if (!cpu_online(panic_force_cpu)) {
| 		pr_warn("panic: target CPU %d is offline, continuing on CPU %d\n",
| 			panic_force_cpu, this_cpu);
| 		return false;
| 	}
| 
| Because the redirect fails, nmi_panic() continues and calls vpanic(). Since
| vpanic() unconditionally calls panic_try_force_cpu() as well, the same
| offline check will execute and the warning might be printed a second time
| before the panic_in_progress() check can abort the redundant redirect attempt.

It is right. We might fix this by checking that panic_try_force_cpu() has been called twice, for example, by checking the panic_redirect_cpu value at the beginning.

But I would personally ignore this problem. It is a corner case. The fix would make the code more hairy. IMHO, it is not worth it.

Best Regards,
Petr

#9 lore raw

PATCH v6 6/6 panic: kill the "buffer unavailable" redirect fallback

In reply to Bradley Morgan

The redirect buffer is a disgusting terrible hack. panic_force_buf is kmalloc'ed in a late_initcall, the return value is not even checked, and until then the redirect delivers this as the panic message:

Redirected panic (buffer unavailable)

The whole point of the redirect is to hand the panic message to the target CPU, so the crash kernel boots knowing it panicked and not why. And that window is the entire boot, from the early_param to the late_initcall, which is exactly when you most want the message. A failed kmalloc just keeps it broken forever, silently.

Make it a static 1KB buffer and kill the initcall. The cost is 1KB of .bss in SMP crash dump builds, and it is only ever touched when panic_force_cpu= is set anyway.

Fixes 2e171ab29f91 ("panic: add panic_force_cpu= parameter to redirect panic to a specific CPU")
---
 kernel/panic.c | 32 +++++++-------------------------
 1 file changed, 7 insertions(+), 25 deletions(-)
kernel/panic.c +7 -25
index 356c03f2361c..b41a7efc0931 100644
@@ -307,7 +307,7 @@ atomic_t panic_cpu = ATOMIC_INIT(PANIC_CPU_INVALID);
 atomic_t panic_redirect_cpu = ATOMIC_INIT(PANIC_CPU_INVALID);
 
 #if defined(CONFIG_SMP) && defined(CONFIG_CRASH_DUMP)
-static char *panic_force_buf;
+static char panic_force_buf[PANIC_MSG_BUFSZ];
 
 static int __init panic_force_cpu_setup(char *str)
 {
@@ -326,17 +326,6 @@ static int __init panic_force_cpu_setup(char *str)
 }
 early_param("panic_force_cpu", panic_force_cpu_setup);
 
-static int __init panic_force_cpu_late_init(void)
-{
-	if (panic_force_cpu < 0)
-		return 0;
-
-	panic_force_buf = kmalloc(PANIC_MSG_BUFSZ, GFP_KERNEL);
-
-	return 0;
-}
-late_initcall(panic_force_cpu_late_init);
-
 static void do_panic_on_target_cpu(void *info)
 {
 	panic("%s", (char *)info);
@@ -381,6 +370,7 @@ static bool panic_try_force_cpu(const char *fmt, va_list args)
 	int this_cpu = raw_smp_processor_id();
 	int old_cpu = PANIC_CPU_INVALID;
 	const char *msg;
+	va_list ap;
 
 	/* Feature not enabled via boot parameter */
 	if (panic_force_cpu < 0)
@@ -413,20 +403,12 @@ static bool panic_try_force_cpu(const char *fmt, va_list args)
 		return old_cpu != this_cpu;
 
 	/*
-	 * Use dynamically allocated buffer if available, otherwise
-	 * fall back to static message for early boot panics or allocation failure.
+	 * Do not consume args, the caller reuses them if we fail.
 	 */
-	if (panic_force_buf) {
-		va_list ap;
-
-		/* Do not consume args, the caller reuses it if we fail */
-		va_copy(ap, args);
-		vsnprintf(panic_force_buf, PANIC_MSG_BUFSZ, fmt, ap);
-		va_end(ap);
-		msg = panic_force_buf;
-	} else {
-		msg = "Redirected panic (buffer unavailable)";
-	}
+	va_copy(ap, args);
+	vsnprintf(panic_force_buf, PANIC_MSG_BUFSZ, fmt, ap);
+	va_end(ap);
+	msg = panic_force_buf;
 
 	console_verbose();
 	bust_spinlocks(1);
Signature · 1 line
2.47.3
#10 lore raw

Re: [PATCH v6 6/6] panic: kill the "buffer unavailable" redirect fallback

Quoted 2 lines · Bradley Morgan

The redirect buffer is a disgusting terrible hack. panic_force_buf is kmalloc'ed in a late_initcall, the return value is not even checked,

It is not checked intentionally. kmalloc() prints a debug message on its own when it is not able to allocate the memory. And it returns NULL in this case. It is enough in this case.

Please, remove this sentence from the commit message.

Quoted 29 lines

and until then the redirect delivers this as the panic message:

Redirected panic (buffer unavailable)

The whole point of the redirect is to hand the panic message to the target CPU, so the crash kernel boots knowing it panicked and not why. And that window is the entire boot, from the early_param to the late_initcall, which is exactly when you most want the message. A failed kmalloc just keeps it broken forever, silently.

Make it a static 1KB buffer and kill the initcall. The cost is 1KB of .bss in SMP crash dump builds, and it is only ever touched when panic_force_cpu= is set anyway.

kernel/panic.c +1 -7
@@ -413,20 +403,12 @@ static bool panic_try_force_cpu(const char *fmt, va_list args)
 		return old_cpu != this_cpu;
 
 	/*
-	 * Use dynamically allocated buffer if available, otherwise
-	 * fall back to static message for early boot panics or allocation failure.
+	 * Do not consume args, the caller reuses them if we fail.
 	 */
-	if (panic_force_buf) {
-		va_list ap;
-
-		/* Do not consume args, the caller reuses it if we fail */
-		va_copy(ap, args);

kernel/panic.c static bool panic_try_force_cpu(const char *fmt, va_list args) · line 411

Quoted 10 lines
kernel/panic.c +4 -6
-		vsnprintf(panic_force_buf, PANIC_MSG_BUFSZ, fmt, ap);
-		va_end(ap);
-		msg = panic_force_buf;
-	} else {
-		msg = "Redirected panic (buffer unavailable)";
-	}
+	va_copy(ap, args);
+	vsnprintf(panic_force_buf, PANIC_MSG_BUFSZ, fmt, ap);
+	va_end(ap);
+	msg = panic_force_buf;

Nit: The "msg" variable is no longer needed. It will always be set to
panic_force_buf. The later code could use:

panic_smp_redirect_cpu(panic_force_cpu, (void *)panic_force_buf)

Otherwise, the change looks good from my POV.

Best Regards,
Petr

#11 lore raw
In reply to Bradley Morgan
Quoted 4 lines · On Tue, 18 Aug 2026 16:38:00 +0000 Bradley Morgan <[email protected]> wrote:

The panic_force_cpu= parameter redirects a panic to a specific CPU so the crash kernel runs there. The redirect code in panic_try_force_cpu() had two races and an NMI bypass, all found by Sashiko. This series closes them and kills one more hack that the rework turned up.

Thanks. Sashiko had more to say:

	https://sashiko.dev/#/patchset/[email protected]

In the first one it suggests va_copy(), but perhaps this would be better addressed with va_end/va_start.

#12 lore raw
Quoted 8 lines · Andrew Morton
Quoted 4 lines · On Tue, 18 Aug 2026 16:38:00 +0000 Bradley Morgan <[email protected]> wrote:

The panic_force_cpu= parameter redirects a panic to a specific CPU so the crash kernel runs there. The redirect code in panic_try_force_cpu() had two races and an NMI bypass, all found by Sashiko. This series closes them and kills one more hack that the rework turned up.

Quoted 6 lines

Thanks. Sashiko had more to say:

	https://sashiko.dev/#/patchset/[email protected]

In the first one it suggests va_copy(), but perhaps this would be better addressed with va_end/va_start.

Sashiko said there was ANOTHER bug, I might just screammm, at this point, some sort of superhero needs to come by and suggest the fix that sashiko likes :sob:

Thanks!

#13 lore raw
Quoted 13 lines · Bradley Morgan
Quoted 11 lines · Andrew Morton
Quoted 4 lines · On Tue, 18 Aug 2026 16:38:00 +0000 Bradley Morgan <[email protected]> wrote:

The panic_force_cpu= parameter redirects a panic to a specific CPU so the crash kernel runs there. The redirect code in panic_try_force_cpu() had two races and an NMI bypass, all found by Sashiko. This series closes them and kills one more hack that the rework turned up.

Thanks. Sashiko had more to say:

	https://sashiko.dev/#/patchset/[email protected]
Quoted 7 lines
Quoted 3 lines

In the first one it suggests va_copy(), but perhaps this would be better addressed with va_end/va_start.

Sashiko said there was ANOTHER bug, I might just screammm, at this point, some sort of superhero needs to come by and suggest the fix that sashiko likes :sob:

We are all still learning how to deal with Sashiko. It is quite good at finding problems. We always have to decide/agree on what is worth fixing and what is not worth it.

I described my opinion in replies to the particular patches.

Anyway, we are getting close. I believe that v7 will be final ;-)

Best Regards,
Petr