printk question:why release console_sem after logbuf_lock
in the patch 0b5e1c5255e7ee8670e077e8224e5c2281229a5b, it releases console_sem after logbuf_lock, the description of this patch is as below: Release console_sem after unlocking the logbuf_lock so that we don't generate wakeups while holding logbuf_lock. This avoids some lock inversion troubles once we remove the lockdep_off bits between logbuf_lock and rq->lock (prints while holding rq->lock vs doing wakeups while holding logbuf_lock).There's of course still an actual deadlock where the printk()s under rq->lock will issue a wakeup from the up() call, but lockdep won't warn about that since semaphores are not tracked. could you please give me a detail example about the issue it tries to fix? thanks.
Details u have to look at the source code, my guess is based on the following posting: http://kerneltrap.org/mailarchive/linux-kernel/2008/1/23/595569 https://lkml.org/lkml/2011/6/21/249 logic: a. u want to be able to do printk from anywhere. b. but every call to printk requires a console_sem lock to be acquired. after acquiring console_sem, printk actually serializes the output to a memory buffer. c. now problem arises when printk is happening very fast, and so this type of locks is ill-suited for printk(). d. later than this patch is another attempt: https://patchwork.kernel.org/patch/1760211/ https://lkml.org/lkml/2012/10/20/90 where lazy irq work is being used instead. read through the comments in the intro to the patch - it covers a lot more than i mentioned here. In Documentation/lockdep_design.txt discuss about using irq tracing to trace the lock dependencies. Lock inversion is a common computer science problem....look up wiki. On Sat, Feb 9, 2013 at 10:55 AM, buyitian <buyit@live.cn> wrote:
in the patch 0b5e1c5255e7ee8670e077e8224e5c2281229a5b, it releases console_sem after logbuf_lock, the description of this patch is as below:
Release console_sem after unlocking the logbuf_lock so that we don't generate wakeups while holding logbuf_lock. This avoids some lock inversion troubles once we remove the lockdep_off bits between logbuf_lock and rq->lock (prints while holding rq->lock vs doing wakeups while holding logbuf_lock). There's of course still an actual deadlock where the printk()s under rq->lock will issue a wakeup from the up() call, but lockdep won't warn about that since semaphores are not tracked.
could you please give me a detail example about the issue it tries to fix? thanks.
_______________________________________________ Kernelnewbies mailing list Kernelnewbies@kernelnewbies.org http://lists.kernelnewbies.org/mailman/listinfo/kernelnewbies
-- Regards, Peter Teoh
在 2013-2-10，11:43，"Peter Teoh" <htmldeveloper@gmail.com> 写道：
Details u have to look at the source code, my guess is based on the following posting:
http://kerneltrap.org/mailarchive/linux-kernel/2008/1/23/595569
sorry, I cannot imagine why we must release console_sem after logbuf_lock. I need an example, thanks.
logic:
a. u want to be able to do printk from anywhere. b. but every call to printk requires a console_sem lock to be acquired. after acquiring console_sem, printk actually serializes the output to a memory buffer. c. now problem arises when printk is happening very fast, and so this type of locks is ill-suited for printk(). d. later than this patch is another attempt:
https://patchwork.kernel.org/patch/1760211/ https://lkml.org/lkml/2012/10/20/90
where lazy irq work is being used instead.
read through the comments in the intro to the patch - it covers a lot more than i mentioned here.
In Documentation/lockdep_design.txt discuss about using irq tracing to trace the lock dependencies.
Lock inversion is a common computer science problem....look up wiki.
On Sat, Feb 9, 2013 at 10:55 AM, buyitian <buyit@live.cn> wrote:
in the patch 0b5e1c5255e7ee8670e077e8224e5c2281229a5b, it releases console_sem after logbuf_lock, the description of this patch is as below:
Release console_sem after unlocking the logbuf_lock so that we don't generate wakeups while holding logbuf_lock. This avoids some lock inversion troubles once we remove the lockdep_off bits between logbuf_lock and rq->lock (prints while holding rq->lock vs doing wakeups while holding logbuf_lock). There's of course still an actual deadlock where the printk()s under rq->lock will issue a wakeup from the up() call, but lockdep won't warn about that since semaphores are not tracked.
could you please give me a detail example about the issue it tries to fix? thanks.
_______________________________________________ Kernelnewbies mailing list Kernelnewbies@kernelnewbies.org http://lists.kernelnewbies.org/mailman/listinfo/kernelnewbies
-- Regards, Peter Teoh _______________________________________________ Kernelnewbies mailing list Kernelnewbies@kernelnewbies.org http://lists.kernelnewbies.org/mailman/listinfo/kernelnewbies
_______________________________________________ Kernelnewbies mailing list Kernelnewbies@kernelnewbies.org http://lists.kernelnewbies.org/mailman/listinfo/kernelnewbies
participants (2)
-
buyitian -
Peter Teoh