On 2019-03-04, Sergey Senozhatsky <sergey.senozhatsky.work@xxxxxxxxx> wrote:
>> + /* the printk kthread never exits */
>> + for (;;) {
>> + ret = prb_iter_wait_next(&iter, buf,
>> + PRINTK_RECORD_MAX, &master_seq);
>> + if (ret == -ERESTARTSYS) {
>> + continue;
>> + } else if (ret < 0) {
>> + /* iterator invalid, start over */
>> + prb_iter_init(&iter, &printk_rb, NULL);
>> + continue;
>> + }
>> +
>> + msg = (struct printk_log *)buf;
>> + format_text(msg, master_seq, ext_text, &ext_len, text,
>> + &len, printk_time);
>> +
>> + console_lock();
>> + if (len > 0 || ext_len > 0) {
>> + call_console_drivers(ext_text, ext_len, text, len);
>> + boot_delay_msec(msg->level);
>> + printk_delay();
>> + }
>> + console_unlock();
>> + }
>
> This, theoretically, creates a whole new world of possibilities for
> console drivers. Now they can do GFP_KERNEL allocations and stall
> printk_kthread during OOM; or they can explicitly reschedule from
> ->write() callback (via console_conditional_schedule()) because
> console_lock() sets console_may_schedule.
This was the intention. Although, as I mentioned in a previous
response[0], perhaps we should not loosen the requirements on write().
> It's one thing to do cond_resched() (or to let preemption to take
> over) after call_console_drivers() (when we are done printing a
> message to all console drivers) and another thing to let preemption to
> take over while we are printing a messages to the consoles. It
> probably would make sense to disable preemption around
> call_console_drivers().
I could see disabling preemption and interrupts for emergency messages
in the printk-kthread in order to synchronize against an irq_work
secondary printer as suggested in my response[0]. But I don't see an
advantage to disabling preemption in general for
call_console_drivers(). It is exactly that disable_preempt() that is so
harmful for realtime tasks.
John Ogness
[0] https://lkml.kernel.org/r/87lg1rggcz.fsf@xxxxxxxxxxxxx
