printk: Defer legacy printing when holding printk_cpu_sync

The documentation of printk_cpu_sync_get() clearly states
that the owner must never perform any activities where it waits
for a CPU. For legacy printing there can be spinning on the
console_lock and on the port lock. Therefore legacy printing
must be deferred when holding the printk_cpu_sync.

Note that in the case of emergency states, atomic consoles
are not prevented from printing when printk is deferred. This
is appropriate because they do not spin-wait indefinitely for
other CPUs.

Reported-by: Rik van Riel <riel@surriel.com>
Closes: https://lore.kernel.org/r/20240715232052.73eb7fb1@imladris.surriel.com
Signed-off-by: John Ogness <john.ogness@linutronix.de>
Fixes: 55d6af1d66 ("lib/nmi_backtrace: explicitly serialize banner and regs")
Reviewed-by: Petr Mladek <pmladek@suse.com>
Link: https://lore.kernel.org/r/20241209111746.192559-3-john.ogness@linutronix.de
Signed-off-by: Petr Mladek <pmladek@suse.com>
This commit is contained in:
John Ogness 2024-12-09 12:23:46 +01:06 committed by Petr Mladek
parent f1c21cf470
commit 0161e2d695
3 changed files with 17 additions and 1 deletions

View File

@ -338,3 +338,9 @@ bool printk_get_next_message(struct printk_message *pmsg, u64 seq,
void console_prepend_dropped(struct printk_message *pmsg, unsigned long dropped);
void console_prepend_replay(struct printk_message *pmsg);
#endif
#ifdef CONFIG_SMP
bool is_printk_cpu_sync_owner(void);
#else
static inline bool is_printk_cpu_sync_owner(void) { return false; }
#endif

View File

@ -4922,6 +4922,11 @@ void console_try_replay_all(void)
static atomic_t printk_cpu_sync_owner = ATOMIC_INIT(-1);
static atomic_t printk_cpu_sync_nested = ATOMIC_INIT(0);
bool is_printk_cpu_sync_owner(void)
{
return (atomic_read(&printk_cpu_sync_owner) == raw_smp_processor_id());
}
/**
* __printk_cpu_sync_wait() - Busy wait until the printk cpu-reentrant
* spinning lock is not owned by any CPU.

View File

@ -61,10 +61,15 @@ bool is_printk_legacy_deferred(void)
/*
* The per-CPU variable @printk_context can be read safely in any
* context. CPU migration is always disabled when set.
*
* A context holding the printk_cpu_sync must not spin waiting for
* another CPU. For legacy printing, it could be the console_lock
* or the port lock.
*/
return (force_legacy_kthread() ||
this_cpu_read(printk_context) ||
in_nmi());
in_nmi() ||
is_printk_cpu_sync_owner());
}
asmlinkage int vprintk(const char *fmt, va_list args)