From: Qian Cai <cai@lca.pw>
To: Michal Hocko <mhocko@kernel.org>
Cc: Christian Borntraeger <borntraeger@de.ibm.com>,
Petr Mladek <pmladek@suse.com>,
Peter Oberparleiter <oberpar@linux.ibm.com>,
akpm@linux-foundation.org, sergey.senozhatsky.work@gmail.com,
rostedt@goodmis.org, peterz@infradead.org, linux-mm@kvack.org,
john.ogness@linutronix.de, david@redhat.com,
linux-kernel@vger.kernel.org,
Heiko Carstens <heiko.carstens@de.ibm.com>,
Vasily Gorbik <gor@linux.ibm.com>
Subject: Re: [PATCH v2] mm/page_isolation: fix a deadlock with printk()
Date: Tue, 08 Oct 2019 15:06:13 -0400 [thread overview]
Message-ID: <1570561573.5576.307.camel@lca.pw> (raw)
In-Reply-To: <20191008183525.GQ6681@dhcp22.suse.cz>
On Tue, 2019-10-08 at 20:35 +0200, Michal Hocko wrote:
> On Tue 08-10-19 12:08:37, Qian Cai wrote:
> > On Tue, 2019-10-08 at 14:56 +0200, Christian Borntraeger wrote:
> > > Adding Peter Oberparleiter.
> > > Peter, can you have a look?
> > >
> > > On 08.10.19 10:27, Michal Hocko wrote:
> > > > On Tue 08-10-19 09:43:57, Petr Mladek wrote:
> > > > > On Mon 2019-10-07 16:49:37, Michal Hocko wrote:
> > > > > > [Cc s390 maintainers - the lockdep is http://lkml.kernel.org/r/1570228005-24979-1-git-send-email-cai@lca.pw
> > > > > > Petr has explained it is a false positive
> > > > > > http://lkml.kernel.org/r/20191007143002.l37bt2lzqtnqjqxu@pathway.suse.cz]
> > > > > > On Mon 07-10-19 16:30:02, Petr Mladek wrote:
> > > > > > [...]
> > > > > > > I believe that it cannot really happen because:
> > > > > > >
> > > > > > > static int __init
> > > > > > > sclp_console_init(void)
> > > > > > > {
> > > > > > > [...]
> > > > > > > rc = sclp_rw_init();
> > > > > > > [...]
> > > > > > > register_console(&sclp_console);
> > > > > > > return 0;
> > > > > > > }
> > > > > > >
> > > > > > > sclp_rw_init() is called before register_console(). And
> > > > > > > console_unlock() will never call sclp_console_write() before
> > > > > > > the console is registered.
> > > > > > >
> > > > > > > AFAIK, lockdep only compares existing chain of locks. It does
> > > > > > > not know about console registration that would make some
> > > > > > > code paths mutually exclusive.
> > > > > > >
> > > > > > > I believe that it is a false positive. I do not know how to
> > > > > > > avoid this lockdep report. I hope that it will disappear
> > > > > > > by deferring all printk() calls rather soon.
> > > > > >
> > > > > > Thanks a lot for looking into this Petr. I have also checked the code
> > > > > > and I really fail to see why the allocation has to be done under the
> > > > > > lock in the first place. sclp_read_sccb and sclp_init_sccb are global
> > > > > > variables but I strongly suspect that they need a synchronization during
> > > > > > early init, callbacks are registered only later IIUC:
> > > > >
> > > > > Good idea. It would work when the init function is called only once.
> > > > > But see below.
> > > > >
> > > > > > diff --git a/drivers/s390/char/sclp.c b/drivers/s390/char/sclp.c
> > > > > > index d2ab3f07c008..4b1c033e3255 100644
> > > > > > --- a/drivers/s390/char/sclp.c
> > > > > > +++ b/drivers/s390/char/sclp.c
> > > > > > @@ -1169,13 +1169,13 @@ sclp_init(void)
> > > > > > unsigned long flags;
> > > > > > int rc = 0;
> > > > > >
> > > > > > + sclp_read_sccb = (void *) __get_free_page(GFP_ATOMIC | GFP_DMA);
> > > > > > + sclp_init_sccb = (void *) __get_free_page(GFP_ATOMIC | GFP_DMA);
> > > > > > spin_lock_irqsave(&sclp_lock, flags);
> > > > > > /* Check for previous or running initialization */
> > > > > > if (sclp_init_state != sclp_init_state_uninitialized)
> > > > > > goto fail_unlock;
> > > > >
> > > > > It seems that sclp_init() could be called several times in parallel.
> > > > > I see it called from sclp_register() and sclp_initcall().
> > > >
> > > > Interesting. Something for s390 people to answer I guess.
> > > > Anyway, this should be quite trivial to workaround by a cmpxch or alike.
> > > >
> >
> > The above fix is simply insufficient,
>
> Isn't this yet another init time lockdep false possitive?
Again, this is not 100% false positive for sure yet.
>
> > 00: [ 3.654337] -> #3 (console_owner){....}:
> > 00: [ 3.654343] lock_acquire+0x21a/0x468
> > 00: [ 3.654345] console_unlock+0x3a6/0xa30
> > 00: [ 3.654346] vprintk_emit+0x184/0x3c8
> > 00: [ 3.654348] vprintk_default+0x44/0x50
> > 00: [ 3.654349] printk+0xa8/0xc0
> > 00: [ 3.654351] get_random_u64+0x40/0x108
> > 00: [ 3.654360] add_to_free_area_random+0x188/0x1c0
> > 00: [ 3.654364] free_one_page+0x72/0x128
> > 00: [ 3.654366] __free_pages_ok+0x51c/0xca0
> > 00: [ 3.654368] memblock_free_all+0x30a/0x3b0
> > 00: [ 3.654370] mem_init+0x84/0x200
> > 00: [ 3.654371] start_kernel+0x384/0x6a0
> > 00: [ 3.654373] startup_continue+0x70/0xd0
>
> This one is actually a nice example why trying to get printk out of the
> zone->lock is simply not viable. This one is likely a printk to warn
> that the random pool is not fully intiailized. Just because the
> allocator tries to randomize the initial free memory pool. You are not
> going to remove that printk, right?
Well, Sergey had a patch to convert that one to printk_deferred(), but even with
his patch, it will still trigger the lockdep splat here because the lock
dependency between zone->lock --> console_owner is still there from memory
offline.
>
> I fully agree that this class of lockdep splats are annoying especially
> when they make the lockdep unusable but please discuss this with lockdep
> maintainers and try to find some solution rather than go and try to
> workaround the problem all over the place. If there are places that
> would result in a cleaner code then go for it but please do not make the
> code worse just because of a non existent problem flagged by a false
> positive.
It makes me wonder what make you think it is a false positive for sure.
next prev parent reply other threads:[~2019-10-08 19:06 UTC|newest]
Thread overview: 73+ messages / expand[flat|nested] mbox.gz Atom feed top
2019-10-04 22:26 [PATCH v2] mm/page_isolation: fix a deadlock with printk() Qian Cai
2019-10-07 8:07 ` Michal Hocko
2019-10-07 9:05 ` Petr Mladek
2019-10-07 11:33 ` Michal Hocko
2019-10-07 12:34 ` Qian Cai
[not found] ` <FB72D947-A0F9-43E7-80D9-D7ACE33849C7@lca.pw>
2019-10-07 11:37 ` Michal Hocko
2019-10-07 12:11 ` Qian Cai
2019-10-07 12:43 ` Michal Hocko
2019-10-07 13:07 ` Qian Cai
2019-10-07 14:10 ` Petr Mladek
2019-10-07 14:30 ` Petr Mladek
2019-10-07 14:49 ` Michal Hocko
2019-10-08 7:43 ` Petr Mladek
2019-10-08 8:27 ` Michal Hocko
2019-10-08 12:56 ` Christian Borntraeger
2019-10-08 16:08 ` Qian Cai
2019-10-08 18:35 ` Michal Hocko
2019-10-08 19:06 ` Qian Cai [this message]
2019-10-08 19:17 ` Michal Hocko
2019-10-08 19:35 ` Qian Cai
2019-10-09 11:49 ` Petr Mladek
2019-10-09 13:06 ` Qian Cai
2019-10-09 13:27 ` Michal Hocko
2019-10-09 13:43 ` Qian Cai
2019-10-09 13:51 ` Michal Hocko
2019-10-09 14:19 ` Qian Cai
2019-10-09 14:34 ` Michal Hocko
2019-10-09 15:08 ` Qian Cai
2019-10-09 16:23 ` Michal Hocko
2019-10-10 9:01 ` Qian Cai
2019-10-10 10:59 ` Michal Hocko
2019-10-10 13:11 ` Qian Cai
2019-10-10 14:18 ` Michal Hocko
2019-10-10 14:47 ` Qian Cai
2019-10-10 17:30 ` Michal Hocko
2019-10-10 17:48 ` Qian Cai
2019-10-10 18:06 ` Michal Hocko
2019-10-10 18:59 ` David Hildenbrand
2019-10-09 14:24 ` Petr Mladek
2019-10-09 14:46 ` Qian Cai
2019-10-10 7:57 ` Petr Mladek
2019-10-09 11:39 ` Petr Mladek
2019-10-09 13:56 ` Peter Oberparleiter
2019-10-09 14:26 ` Michal Hocko
2019-10-10 5:12 ` Sergey Senozhatsky
2019-10-10 7:40 ` Michal Hocko
2019-10-10 8:16 ` Sergey Senozhatsky
2019-10-10 8:37 ` Michal Hocko
2019-10-10 8:21 ` Petr Mladek
2019-10-10 8:39 ` Sergey Senozhatsky
2019-10-10 11:11 ` Petr Mladek
2019-10-09 15:25 ` Qian Cai
2019-10-07 14:59 ` Qian Cai
2019-10-07 15:12 ` Michal Hocko
2019-10-07 15:33 ` Qian Cai
2019-10-08 8:15 ` Petr Mladek
2019-10-08 9:32 ` Qian Cai
2019-10-08 13:13 ` Steven Rostedt
2019-10-08 13:23 ` Qian Cai
2019-10-08 13:33 ` Steven Rostedt
2019-10-08 13:42 ` Petr Mladek
2019-10-08 13:48 ` Michal Hocko
2019-10-08 14:03 ` Qian Cai
2019-10-08 14:08 ` Michal Hocko
2019-10-08 8:40 ` Michal Hocko
2019-10-08 10:04 ` Qian Cai
2019-10-08 10:39 ` Michal Hocko
2019-10-08 12:00 ` Qian Cai
2019-10-08 12:39 ` Michal Hocko
2019-10-08 13:06 ` Qian Cai
2019-10-08 13:37 ` Michal Hocko
2019-10-08 13:08 ` Petr Mladek
2019-10-08 13:33 ` Qian Cai
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=1570561573.5576.307.camel@lca.pw \
--to=cai@lca.pw \
--cc=akpm@linux-foundation.org \
--cc=borntraeger@de.ibm.com \
--cc=david@redhat.com \
--cc=gor@linux.ibm.com \
--cc=heiko.carstens@de.ibm.com \
--cc=john.ogness@linutronix.de \
--cc=linux-kernel@vger.kernel.org \
--cc=linux-mm@kvack.org \
--cc=mhocko@kernel.org \
--cc=oberpar@linux.ibm.com \
--cc=peterz@infradead.org \
--cc=pmladek@suse.com \
--cc=rostedt@goodmis.org \
--cc=sergey.senozhatsky.work@gmail.com \
/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;
as well as URLs for NNTP newsgroup(s).