On Thu 03-12-15 11:39:33, Sergey Senozhatsky wrote: > On (12/03/15 10:11), Sergey Senozhatsky wrote: > > On (12/02/15 15:57), akpm@xxxxxxxxxxxxxxxxxxxx wrote: > > [..] > > > @console_may_schedule tracks whether console_sem was acquired through lock > > > or trylock. If the former, we're inside a sleepable context and > > > console_conditional_schedule() performs cond_resched(). This allows > > > console drivers which use console_lock for synchronization to yield while > > > performing time-consuming operations such as scrolling. > > > > > > However, the actual console outputting is performed while holding irq-safe > > > logbuf_lock, so console_unlock() clears @console_may_schedule before > > > starting outputting lines. Also, only a few drivers call > > > console_conditional_schedule() to begin with. This means that when a lot > > > of lines need to be output by console_unlock(), for example on a console > > > registration, the task doing console_unlock() may not yield for a long > > > time on a non-preemptible kernel. > > > > > > If this happens with a slow console devices, for example a serial console, > > > the outputting task may occupy the cpu for a very long time. Long enough > > > to trigger softlockup and/or RCU stall warnings, which in turn pile more > > > messages, sometimes enough to trigger the next cycle of warnings > > > incapacitating the system. > > > > > > Fix it by making console_unlock() insert cond_resched() between lines if > > > @console_may_schedule. > > > > CPU2 still can cause lots of troubles. consider > > > > CPU0 CPU1 CPU2 > > printk > > ... printk_deferred > > printk wake_up_klogd > > wake_up_klogd_work_func > > console_trylock > > console_unlock > > > > printk_deferred() may be issued by scheduler, for example. > > IOW, may be we can start limiting the number of bytes printed in console_unlock() > from irq contexts. Which is quite ugly, yes. We basically don't know how much time > we spend in call_console_drivers(); some of the consoles can do 'internal' spin_lock > loops in ->write() handlers, etc. So something like this (below) probably will not > really help, but still it's not always OK to do `while (1)' loop in console_unlock() > for irqs. What we really want is pushing the printing into async context (unless forced by debug option or oops in progress). Because what you do here fixes only a small fraction of the problem space. I have patches which fix more of it (https://lkml.org/lkml/2015/10/26/16) but they are still not enough because on large machines e.g. udev times out because printing messages about inserted hardware over serial console just takes too long. Honza > > --- > > kernel/printk/printk.c | 6 +++++- > 1 file changed, 5 insertions(+), 1 deletion(-) > > diff --git a/kernel/printk/printk.c b/kernel/printk/printk.c > index 9da39e7..221a230 100644 > --- a/kernel/printk/printk.c > +++ b/kernel/printk/printk.c > @@ -2235,6 +2235,7 @@ void console_unlock(void) > unsigned long flags; > bool wake_klogd = false; > bool do_cond_resched, retry; > + int printed, irq_count = irq_count(); > > if (console_suspended) { > up_console_sem(); > @@ -2257,6 +2258,7 @@ void console_unlock(void) > /* flush buffered message fragment immediately to console */ > console_cont_flush(text, sizeof(text)); > again: > + printed = 0; > for (;;) { > struct printk_log *msg; > size_t ext_len = 0; > @@ -2326,6 +2328,8 @@ skip: > > if (do_cond_resched) > cond_resched(); > + if (irq_count && printed > LOG_LINE_MAX) > + break; > } > console_locked = 0; > > @@ -2344,7 +2348,7 @@ skip: > * flush, no worries. > */ > raw_spin_lock(&logbuf_lock); > - retry = console_seq != log_next_seq; > + retry = (console_seq != log_next_seq) && !!irq_count; > raw_spin_unlock_irqrestore(&logbuf_lock, flags); > > if (retry && console_trylock()) > > -- Jan Kara <jack@xxxxxxxx> SUSE Labs, CR -- To unsubscribe from this list: send the line "unsubscribe stable" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html