Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]

Groups > linux.kernel > #1373539 > unrolled thread

[PATCH v11 0/3] printk: Make printk() completely async

Back to article view | Back to linux.kernel

Contents

  [PATCH v11 0/3] printk: Make printk() completely async Sergey Senozhatsky <sergey.senozhatsky@gmail.com> - 2016-04-07 18:40 +0200
    [PATCH v11 2/3] printk: Make wake_up_klogd_work_func() async Sergey Senozhatsky <sergey.senozhatsky@gmail.com> - 2016-04-07 18:40 +0200
    Re: [PATCH v11 0/3] printk: Make printk() completely async Sergey Senozhatsky <sergey.senozhatsky.work@gmail.com> - 2016-04-16 05:00 +0200
      Re: [PATCH v11 0/3] printk: Make printk() completely async Joe Perches <joe@perches.com> - 2016-04-16 07:50 +0200
        Re: [PATCH v11 0/3] printk: Make printk() completely async Sergey Senozhatsky <sergey.senozhatsky.work@gmail.com> - 2016-04-21 04:20 +0200
          Re: [PATCH v11 0/3] printk: Make printk() completely async Joe Perches <joe@perches.com> - 2016-04-21 04:20 +0200
            Re: [PATCH v11 0/3] printk: Make printk() completely async Jan Kara <jack@suse.cz> - 2016-04-21 15:20 +0200

#1373539 — [PATCH v11 0/3] printk: Make printk() completely async

Hello,

This patch set makes printk() completely asynchronous: new messages
are getting upended to the kernel printk buffer, but instead of 'direct'
printing the actual print job is performed by a dedicated kthread.
This has the advantage that printing always happens from a schedulable
context and thus we don't lockup any particular CPU or even interrupts.

against next-20160407

v11:
-- switch default to sync printk
-- make `synchronous' param RW (Andrew, Jan)
-- set RT priority to printk kthread (Andrew)
-- correct comments (Andrew)

v10:
-- simplify printk_kthread_need_flush_console (Jan, Petr)

v9:
-- move need_flush_console assignment down in vprintk_emit (Jan)
-- simplify need_flush_console assignment rules (Petr)
-- clear need_flush_console in printing function (Petr)
-- rename need_flush_console (Petr)

v8:
-- rename kthread printing function (Petr)
-- clear need_flush_console in console_unlock() under logbuf (Petr)

v7:
-- do not set global printk_sync in panic in vrintk_emit() (Petr)
-- simplify vprintk_emit(). drop some of local variables (Petr)
-- move handling of LOGLEVEL_SCHED messages back to printk_deferred()
   so we wake_up_process()/console_trylock() in vprintk_emit() only
   for !in_sched messages

v6:
-- move wake_up_process out of logbuf lock (Jan, Byungchul)
-- do not disable async printk in recursion handling code.
-- rebase against next-20160321 (w/NMI patches)

v5:
-- make printk.synchronous RO (Petr)
-- make printing_func() correct and do not use wait_queue (Petr)
-- do not panic() when can't allocate printing thread (Petr)
-- do not wake_up_process() only in IRQ, prefer vprintk_emit() (Jan)
-- move wake_up_klogd_work_func() to a separate patch (Petr)
-- move wake_up_process() under logbuf lock so printk recursion logic can
   help us out
-- switch to sync_print mode if printk recursion occured
-- drop "printk: Skip messages on oops" patch

v4:
-- do not directly wake_up() the printing kthread from vprintk_emit(), need
   to go via IRQ->wake_up() to avoid sched deadlocks (Jan)

v3:
-- use a dedicated kthread for printing instead of using wq (Jan, Tetsuo, Tejun)

v2:
- use dedicated printk workqueue with WQ_MEM_RECLAIM bit
- fallback to system-wide workqueue only if allocation of printk_wq has
  failed
- do not use system_wq as a fallback wq. both console_lock() and onsole_unlock()
  can spend a significant amount of time; so we need to use system_long_wq.
- rework sync/!sync detection logic
  a) we can have deferred (in_sched) messages before we allocate printk_wq,
     so the only way to handle those messages is via IRQ context
  b) even in printk.synchronous mode, deferred messages must not be printed
     directly, and should go via IRQ context
  c) even if we allocated printk_wq and have !sync_printk mode, we must route
     deferred messages via IRQ context
- so this adds additional bool flags to vprint_emit() and introduces a new
  pending bit to `printk_pending'
- fix build on !PRINTK configs


Jan Kara (2):
  printk: Make printk() completely async
  printk: Make wake_up_klogd_work_func() async

Sergey Senozhatsky (1):
  printk: make printk.synchronous param rw

 Documentation/kernel-parameters.txt |  12 +++
 kernel/printk/printk.c              | 155 +++++++++++++++++++++++++++++++++---
 2 files changed, 157 insertions(+), 10 deletions(-)

-- 
2.8.0

[toc] | [next] | [standalone]

#1373542 — [PATCH v11 2/3] printk: Make wake_up_klogd_work_func() async

From: Jan Kara <jack@suse.cz>

Offload printing of scheduler deferred messages from IRQ context
to a schedulable printing kthread, when possible (the same way we
do it in vprintk_emit()). Otherwise, the CPU can spend unbounded
amount of time doing printing in console_unlock() from IRQ.

Signed-off-by: Jan Kara <jack@suse.cz>
Signed-off-by: Sergey Senozhatsky <sergey.senozhatsky@gmail.com>
Reviewed-by: Petr Mladek <pmladek@suse.com>
---
 kernel/printk/printk.c | 13 ++++++++++---
 1 file changed, 10 insertions(+), 3 deletions(-)

diff --git a/kernel/printk/printk.c b/kernel/printk/printk.c
index 64a98ea..89f5441 100644
--- a/kernel/printk/printk.c
+++ b/kernel/printk/printk.c
@@ -2820,9 +2820,16 @@ static void wake_up_klogd_work_func(struct irq_work *irq_work)
 	int pending = __this_cpu_xchg(printk_pending, 0);
 
 	if (pending & PRINTK_PENDING_OUTPUT) {
-		/* If trylock fails, someone else is doing the printing */
-		if (console_trylock())
-			console_unlock();
+		if (printk_kthread) {
+			wake_up_process(printk_kthread);
+		} else {
+			/*
+			 * If trylock fails, someone else is doing
+			 * the printing
+			 */
+			if (console_trylock())
+				console_unlock();
+		}
 	}
 
 	if (pending & PRINTK_PENDING_WAKEUP)
-- 
2.8.0

[toc] | [prev] | [next] | [standalone]

#1380471

On (04/08/16 02:31), Sergey Senozhatsky wrote:
> Hello,
> 
> This patch set makes printk() completely asynchronous: new messages
> are getting upended to the kernel printk buffer, but instead of 'direct'
> printing the actual print job is performed by a dedicated kthread.
> This has the advantage that printing always happens from a schedulable
> context and thus we don't lockup any particular CPU or even interrupts.

Hello,

Sir, is there anything else you want me to improve in this patch set?

	-ss

> against next-20160407
> 
> v11:
> -- switch default to sync printk
> -- make `synchronous' param RW (Andrew, Jan)
> -- set RT priority to printk kthread (Andrew)
> -- correct comments (Andrew)
> 
> v10:
> -- simplify printk_kthread_need_flush_console (Jan, Petr)
> 
> v9:
> -- move need_flush_console assignment down in vprintk_emit (Jan)
> -- simplify need_flush_console assignment rules (Petr)
> -- clear need_flush_console in printing function (Petr)
> -- rename need_flush_console (Petr)
> 
> v8:
> -- rename kthread printing function (Petr)
> -- clear need_flush_console in console_unlock() under logbuf (Petr)
> 
> v7:
> -- do not set global printk_sync in panic in vrintk_emit() (Petr)
> -- simplify vprintk_emit(). drop some of local variables (Petr)
> -- move handling of LOGLEVEL_SCHED messages back to printk_deferred()
>    so we wake_up_process()/console_trylock() in vprintk_emit() only
>    for !in_sched messages
> 
> v6:
> -- move wake_up_process out of logbuf lock (Jan, Byungchul)
> -- do not disable async printk in recursion handling code.
> -- rebase against next-20160321 (w/NMI patches)
> 
> v5:
> -- make printk.synchronous RO (Petr)
> -- make printing_func() correct and do not use wait_queue (Petr)
> -- do not panic() when can't allocate printing thread (Petr)
> -- do not wake_up_process() only in IRQ, prefer vprintk_emit() (Jan)
> -- move wake_up_klogd_work_func() to a separate patch (Petr)
> -- move wake_up_process() under logbuf lock so printk recursion logic can
>    help us out
> -- switch to sync_print mode if printk recursion occured
> -- drop "printk: Skip messages on oops" patch
> 
> v4:
> -- do not directly wake_up() the printing kthread from vprintk_emit(), need
>    to go via IRQ->wake_up() to avoid sched deadlocks (Jan)
> 
> v3:
> -- use a dedicated kthread for printing instead of using wq (Jan, Tetsuo, Tejun)
> 
> v2:
> - use dedicated printk workqueue with WQ_MEM_RECLAIM bit
> - fallback to system-wide workqueue only if allocation of printk_wq has
>   failed
> - do not use system_wq as a fallback wq. both console_lock() and onsole_unlock()
>   can spend a significant amount of time; so we need to use system_long_wq.
> - rework sync/!sync detection logic
>   a) we can have deferred (in_sched) messages before we allocate printk_wq,
>      so the only way to handle those messages is via IRQ context
>   b) even in printk.synchronous mode, deferred messages must not be printed
>      directly, and should go via IRQ context
>   c) even if we allocated printk_wq and have !sync_printk mode, we must route
>      deferred messages via IRQ context
> - so this adds additional bool flags to vprint_emit() and introduces a new
>   pending bit to `printk_pending'
> - fix build on !PRINTK configs
> 
> 
> Jan Kara (2):
>   printk: Make printk() completely async
>   printk: Make wake_up_klogd_work_func() async
> 
> Sergey Senozhatsky (1):
>   printk: make printk.synchronous param rw
> 
>  Documentation/kernel-parameters.txt |  12 +++
>  kernel/printk/printk.c              | 155 +++++++++++++++++++++++++++++++++---
>  2 files changed, 157 insertions(+), 10 deletions(-)
> 
> -- 
> 2.8.0
> 

[toc] | [prev] | [next] | [standalone]

#1380483

On Sat, 2016-04-16 at 11:55 +0900, Sergey Senozhatsky wrote:
> On (04/08/16 02:31), Sergey Senozhatsky wrote:
> > 
> > Hello,
> > 
> > This patch set makes printk() completely asynchronous: new messages
> > are getting upended to the kernel printk buffer, but instead of 'direct'
> > printing the actual print job is performed by a dedicated kthread.
> > This has the advantage that printing always happens from a schedulable
> > context and thus we don't lockup any particular CPU or even interrupts.
> Hello,
> 
> Sir, is there anything else you want me to improve in this patch set?

I'm not sir, but my preference would be to move as much of the
async/thread functionality as possible into a separate file.

[toc] | [prev] | [next] | [standalone]

#1383835

On (04/15/16 22:44), Joe Perches wrote:
[..]
> > Sir, is there anything else you want me to improve in this patch set?
> 
> I'm not sir, but my preference would be to move as much of the
> async/thread functionality as possible into a separate file.

hm, we are talking about some 50-60 lines of code in total (seems
that the patch set adds more comments than code), but your point
is interesting. let's say, if there will be more opinions that it
better land in async_printk.{c,h} files, then I'll take a look on
it. how does it sound?

	-ss

[toc] | [prev] | [next] | [standalone]

#1383836

On Thu, 2016-04-21 at 11:14 +0900, Sergey Senozhatsky wrote:
> On (04/15/16 22:44), Joe Perches wrote:
> [..]
> > > Sir, is there anything else you want me to improve in this patch
> > > set?
> > I'm not sir, but my preference would be to move as much of the
> > async/thread functionality as possible into a separate file.
> hm, we are talking about some 50-60 lines of code in total (seems
> that the patch set adds more comments than code), but your point
> is interesting. let's say, if there will be more opinions that it
> better land in async_printk.{c,h} files, then I'll take a look on
> it. how does it sound?

I think printk.c is pretty large, complicated and should
be broken up into several bits.

I did that once, but it's a real development timing issue 
https://lkml.org/lkml/2012/10/17/41

cheers, Joe

[toc] | [prev] | [next] | [standalone]

#1384206

On Wed 20-04-16 19:17:13, Joe Perches wrote:
> On Thu, 2016-04-21 at 11:14 +0900, Sergey Senozhatsky wrote:
> > On (04/15/16 22:44), Joe Perches wrote:
> > [..]
> > > > Sir, is there anything else you want me to improve in this patch
> > > > set?
> > > I'm not sir, but my preference would be to move as much of the
> > > async/thread functionality as possible into a separate file.
> > hm, we are talking about some 50-60 lines of code in total (seems
> > that the patch set adds more comments than code), but your point
> > is interesting. let's say, if there will be more opinions that it
> > better land in async_printk.{c,h} files, then I'll take a look on
> > it. how does it sound?
> 
> I think printk.c is pretty large, complicated and should
> be broken up into several bits.
> 
> I did that once, but it's a real development timing issue 
> https://lkml.org/lkml/2012/10/17/41

As much as that may be a useful excercise, I don't think it really belongs
in this rather small patch set.

								Honza
-- 
Jan Kara <jack@suse.com>
SUSE Labs, CR

[toc] | [prev] | [standalone]

Back to top | Article view | linux.kernel

csiph-web