queue-6.8/dump_stack-do-not-get-cpu_sync-for-panic-cpu.patch

   1 From ffdb28902359d97bbf24eaa25d7b1585a0f91142 Mon Sep 17 00:00:00 2001
   2 From: Sasha Levin <sashal@kernel.org>
   3 Date: Wed, 7 Feb 2024 14:47:03 +0106
   4 Subject: dump_stack: Do not get cpu_sync for panic CPU
   5
   6 From: John Ogness <john.ogness@linutronix.de>
   7
   8 [ Upstream commit 7412dc6d55eed6b76180e40ac3601412ebde29bd ]
   9
  10 dump_stack() is called in panic(). If for some reason another CPU
  11 is holding the printk_cpu_sync and is unable to release it, the
  12 panic CPU will be unable to continue and print the stacktrace.
  13
  14 Since non-panic CPUs are not allowed to store new printk messages
  15 anyway, there is no need to synchronize the stacktrace output in
  16 a panic situation.
  17
  18 For the panic CPU, do not get the printk_cpu_sync because it is
  19 not needed and avoids a potential deadlock scenario in panic().
  20
  21 Link: https://lore.kernel.org/lkml/ZcIGKU8sxti38Kok@alley
  22 Signed-off-by: John Ogness <john.ogness@linutronix.de>
  23 Reviewed-by: Petr Mladek <pmladek@suse.com>
  24 Link: https://lore.kernel.org/r/20240207134103.1357162-15-john.ogness@linutronix.de
  25 Signed-off-by: Petr Mladek <pmladek@suse.com>
  26 Signed-off-by: Sasha Levin <sashal@kernel.org>
  27 ---
  28  include/linux/printk.h   |  2 ++
  29  kernel/printk/internal.h |  1 -
  30  lib/dump_stack.c         | 16 +++++++++++++---
  31  3 files changed, 15 insertions(+), 4 deletions(-)
  32
  33 diff --git a/include/linux/printk.h b/include/linux/printk.h
  34 index 8ef499ab3c1ed..955e31860095e 100644
  35 --- a/include/linux/printk.h
  36 +++ b/include/linux/printk.h
  37 @@ -273,6 +273,8 @@ static inline void printk_trigger_flush(void)
  38  }
  39  #endif
  40
  41 +bool this_cpu_in_panic(void);
  42 +
  43  #ifdef CONFIG_SMP
  44  extern int __printk_cpu_sync_try_get(void);
  45  extern void __printk_cpu_sync_wait(void);
  46 diff --git a/kernel/printk/internal.h b/kernel/printk/internal.h
  47 index ac2d9750e5f81..6c2afee5ef620 100644
  48 --- a/kernel/printk/internal.h
  49 +++ b/kernel/printk/internal.h
  50 @@ -130,7 +130,6 @@ struct printk_message {
  51  };
  52
  53  bool other_cpu_in_panic(void);
  54 -bool this_cpu_in_panic(void);
  55  bool printk_get_next_message(struct printk_message *pmsg, u64 seq,
  56                              bool is_extended, bool may_supress);
  57
  58 diff --git a/lib/dump_stack.c b/lib/dump_stack.c
  59 index 83471e81501a7..222c6d6c8281a 100644
  60 --- a/lib/dump_stack.c
  61 +++ b/lib/dump_stack.c
  62 @@ -96,15 +96,25 @@ static void __dump_stack(const char *log_lvl)
  63   */
  64  asmlinkage __visible void dump_stack_lvl(const char *log_lvl)
  65  {
  66 +       bool in_panic = this_cpu_in_panic();
  67         unsigned long flags;
  68
  69         /*
  70          * Permit this cpu to perform nested stack dumps while serialising
  71 -        * against other CPUs
  72 +        * against other CPUs, unless this CPU is in panic.
  73 +        *
  74 +        * When in panic, non-panic CPUs are not permitted to store new
  75 +        * printk messages so there is no need to synchronize the output.
  76 +        * This avoids potential deadlock in panic() if another CPU is
  77 +        * holding and unable to release the printk_cpu_sync.
  78          */
  79 -       printk_cpu_sync_get_irqsave(flags);
  80 +       if (!in_panic)
  81 +               printk_cpu_sync_get_irqsave(flags);
  82 +
  83         __dump_stack(log_lvl);
  84 -       printk_cpu_sync_put_irqrestore(flags);
  85 +
  86 +       if (!in_panic)
  87 +               printk_cpu_sync_put_irqrestore(flags);
  88  }
  89  EXPORT_SYMBOL(dump_stack_lvl);
  90
  91 --
  92 2.43.0
  93