From: Petr Mladek <pmladek@suse.com>
To: John Ogness <john.ogness@linutronix.de>
Cc: Sergey Senozhatsky <senozhatsky@chromium.org>,
Steven Rostedt <rostedt@goodmis.org>,
Thomas Gleixner <tglx@linutronix.de>,
linux-kernel@vger.kernel.org,
Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Subject: Re: [PATCH printk 06/18] printk: Protect [un]register_console() with a mutex
Date: Fri, 30 Sep 2022 20:04:44 +0200 [thread overview]
Message-ID: <YzcvvPargLYDHhgq@alley> (raw)
In-Reply-To: <87leq1uev5.fsf@jogness.linutronix.de>
On Fri 2022-09-30 16:22:30, John Ogness wrote:
> On 2022-09-30, Petr Mladek <pmladek@suse.com> wrote:
> > You know, the more locks we have, the bigger is the risk of
> > deadlocks, and the more hacks would be needed in
> > console_flush_on_panic(). And I am afraid
> > that console_lock() will be with us for many years and
> > maybe forever.
>
> Sure. Removing console_lock() will be a long battle involving many
> drivers. I am not trying to fight that battle right now. I just want
> console_lock() out of the way of NOBKL consoles.
There is some misunderstanding. I am going to think more about your
arguments over the weekend.
But maybe, the above is important. You want to get cosole_lock() out
of the way of NOBLK consoles. What does it exactly mean, please?
What code paths are important to achieve this?
From my POV, the most important code path is the kthread. But it
should use SRCU. I mean that the kthread will take neither
cosnole_lock() nor console_list_lock().
Is there any other code path where console_list_lock() will help
you to get console_lock() out of the way?
From my POV, the proposed code does:
register_console()
{
console_list_lock();
console_lock();
/* manipulate struct console and the console_list */
console_unlock();
console_list_unlock();
}
register_console()
{
console_list_lock();
console_lock();
/* manipulate struct console and the console_list */
console_unlock();
console_list_unlock();
}
printk_kthread()
{
while() {
srcu_read_lock();
if (read_flags_srcu())
/* print line */
srcu_read_unlock();
}
}
vprintk_emit()
{
/* store message */
if (do_not_allow_sync_mode)
return;
if (console_trylock()) {
console_flush_all();
__console_unlock();
}
}
some_other_func()
{
console_list_lock();
/* do something with all registered consoles */
console_list_unlock();
}
console_flush_all()
{
do_something_with_all_consoles();
do_something_else_with_all_consoles();
}
What if?
do_something_with_all_consoles()
{
console_list_lock();
/* do something */
console_list_unlock();
}
Wait, there is a possible ABBA deadlock because
do_something_with_all_consoles() takes console_list_lock()
under console_lock(). And register_console() does it
the other way around.
But it is less obvious because these are different locks.
From my POV, both locks serialize the same things
(console_list manipulation). SRCU walk should be
enough for most iterations over the list.
And I do not see which code path would really benefit from
having the new console_list_lock() instead of console_lock().
Best Regards,
Petr
next prev parent reply other threads:[~2022-09-30 18:05 UTC|newest]
Thread overview: 75+ messages / expand[flat|nested] mbox.gz Atom feed top
2022-09-24 0:04 [PATCH printk 00/18] preparation for threaded/atomic printing John Ogness
2022-09-24 0:04 ` [PATCH printk 01/18] printk: Make pr_flush() static John Ogness
2022-09-26 14:12 ` Petr Mladek
2022-09-24 0:04 ` [PATCH printk 02/18] printk: Declare log_wait properly John Ogness
2022-09-26 14:22 ` Petr Mladek
2022-09-24 0:04 ` [PATCH printk 03/18] printk: Remove write only variable nr_ext_console_drivers John Ogness
2022-09-26 14:25 ` Petr Mladek
2022-09-24 0:04 ` [PATCH printk 04/18] printk: Remove bogus comment vs. boot consoles John Ogness
2022-09-26 14:26 ` Petr Mladek
2022-09-24 0:04 ` [PATCH printk 05/18] printk: Mark __printk percpu data ready __ro_after_init John Ogness
2022-09-26 14:27 ` Petr Mladek
2022-09-24 0:04 ` [PATCH printk 06/18] printk: Protect [un]register_console() with a mutex John Ogness
2022-09-24 9:31 ` Sergey Senozhatsky
2022-09-27 15:16 ` Petr Mladek
2022-09-28 9:46 ` Sergey Senozhatsky
2022-09-28 23:42 ` John Ogness
2022-09-29 15:43 ` Petr Mladek
2022-09-30 9:24 ` Petr Mladek
2022-09-30 14:16 ` John Ogness
2022-09-30 18:04 ` Petr Mladek [this message]
2022-09-30 20:26 ` John Ogness
2022-10-03 14:37 ` Petr Mladek
2022-10-03 19:35 ` John Ogness
2022-10-04 2:06 ` Sergey Senozhatsky
2022-10-04 7:28 ` Petr Mladek
2022-09-30 13:30 ` John Ogness
2022-09-24 0:04 ` [PATCH printk 07/18] printk: Convert console list walks for readers to list lock John Ogness
2022-09-27 14:07 ` Petr Mladek
2022-09-24 0:04 ` [PATCH printk 08/18] parisc: Put console abuse into one place John Ogness
2022-09-24 0:20 ` Steven Rostedt
2022-09-30 7:54 ` Petr Mladek
2022-09-24 0:04 ` [PATCH printk 09/18] serial: kgdboc: Lock console list in probe function John Ogness
2022-09-28 23:32 ` Doug Anderson
2022-09-30 8:07 ` Petr Mladek
2022-09-24 0:04 ` [PATCH printk 10/18] kgbd: Pretend that console list walk is safe John Ogness
2022-09-26 9:33 ` Aaron Tomlin
2022-09-28 23:32 ` Doug Anderson
2022-09-30 8:39 ` Petr Mladek
2022-09-30 13:44 ` John Ogness
2022-09-30 17:27 ` Petr Mladek
2022-09-24 0:04 ` [PATCH printk 11/18] printk: Convert console_drivers list to hlist John Ogness
2022-09-24 10:53 ` Sergey Senozhatsky
2022-09-24 17:20 ` Helge Deller
2022-09-25 0:43 ` Sergey Senozhatsky
2022-09-24 17:27 ` Helge Deller
2022-09-30 14:20 ` Petr Mladek
2022-09-30 16:53 ` Helge Deller
2022-09-30 19:46 ` John Ogness
2022-09-30 22:41 ` Helge Deller
2022-09-24 0:04 ` [PATCH printk 12/18] printk: Prepare for SCRU console list protection John Ogness
2022-09-24 10:58 ` Sergey Senozhatsky
2022-09-24 0:04 ` [PATCH printk 13/18] printk: Move buffer size defines John Ogness
2022-09-24 11:01 ` Sergey Senozhatsky
2022-10-07 9:11 ` Petr Mladek
2022-09-24 0:04 ` [PATCH printk 14/18] printk: Document struct console John Ogness
2022-09-24 11:08 ` Sergey Senozhatsky
2022-10-07 11:57 ` Petr Mladek
2022-09-24 0:04 ` [PATCH printk 15/18] printk: Add struct cons_text_buf John Ogness
2022-09-24 11:09 ` Sergey Senozhatsky
2022-10-07 15:15 ` Petr Mladek
2022-09-24 0:04 ` [PATCH printk 16/18] printk: Use " John Ogness
2022-09-24 11:34 ` Sergey Senozhatsky
2022-10-10 10:11 ` Petr Mladek
2022-09-24 0:04 ` [PATCH printk 17/18] printk: Use an output descriptor struct for emit John Ogness
2022-10-10 15:40 ` Petr Mladek
2022-09-24 0:04 ` [PATCH printk 18/18] printk: Handle dropped message smarter John Ogness
2022-09-26 4:19 ` Sergey Senozhatsky
2022-09-26 7:54 ` John Ogness
2022-09-26 9:18 ` Sergey Senozhatsky
2022-10-10 16:07 ` Petr Mladek
2022-09-26 9:22 ` Sergey Senozhatsky
2022-09-24 6:44 ` [PATCH printk 00/18] preparation for threaded/atomic printing Greg Kroah-Hartman
2022-09-25 15:23 ` John Ogness
2022-09-24 9:47 ` Sergey Senozhatsky
2022-09-29 16:33 ` Petr Mladek
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=YzcvvPargLYDHhgq@alley \
--to=pmladek@suse.com \
--cc=gregkh@linuxfoundation.org \
--cc=john.ogness@linutronix.de \
--cc=linux-kernel@vger.kernel.org \
--cc=rostedt@goodmis.org \
--cc=senozhatsky@chromium.org \
--cc=tglx@linutronix.de \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox
all inboxes | Powered by JetHome®