linux-kernel - Re: commit dd4cf8c9e1f4 leads to failed boot

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives
Hash Suite: Windows password security audit tool. GUI, reports in PDF.
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <c5c7d5b8-f1a7-4485-a238-e7c523e742a6@paulmck-laptop>
Date: Thu, 24 Apr 2025 07:30:23 -0700
From: "Paul E. McKenney" <paulmck@...nel.org>
To: Bert Karwatzki <spasswolf@....de>
Cc: "Aithal, Srikanth" <sraithal@....com>,
	Andrew Morton <akpm@...ux-foundation.org>,
	Kuniyuki Iwashima <kuniyu@...zon.com>,
	Mateusz Guzik <mjguzik@...il.com>, Petr Mladek <pmladek@...e.com>,
	Steven Rostedt <rostedt@...dmis.org>,
	John Ogness <john.ogness@...utronix.de>,
	Sergey Senozhatsky <senozhatsky@...omium.org>,
	linux-kernel@...r.kernel.org,
	Linux-Next Mailing List <linux-next@...r.kernel.org>
Subject: Re: commit dd4cf8c9e1f4 leads to failed boot

On Thu, Apr 24, 2025 at 02:40:25PM +0200, Bert Karwatzki wrote:
> Am Mittwoch, dem 23.04.2025 um 12:56 -0700 schrieb Paul E. McKenney:
> > On Wed, Apr 23, 2025 at 09:19:56PM +0200, Bert Karwatzki wrote:
> > > Am Mittwoch, dem 23.04.2025 um 11:07 -0700 schrieb Paul E. McKenney:
> > > > On Wed, Apr 23, 2025 at 08:49:08PM +0530, Aithal, Srikanth wrote:
> > > > > On 4/23/2025 7:48 PM, Paul E. McKenney wrote:
> > > > > > On Wed, Apr 23, 2025 at 07:09:42PM +0530, Aithal, Srikanth wrote:
> > > > > > > On 4/23/2025 5:24 PM, Bert Karwatzki wrote:
> > > > > > > > Since linux next-20250422 booting fails on my MSI Alpha 15 Laptop runnning
> > > > > > > > debian sid. When booting kernel message appear on screen but no messages from
> > > > > > > > init (systemd). There are also no logs written even thought emergency sync
> > > > > > > > via magic sysrq works (a message is printed on screen), presumably because
> > > > > > > > / is not mounted. I bisected this (from 6.15-rc3 to next-20250422) and found
> > > > > > > > commit dd4cf8c9e1f4 as the first bad commit.
> > > > > > > > Reverting commit dd4cf8c9e1f4 in next-20250422 fixes the issue.
> > > > > > > 
> > > > > > > 
> > > > > > > Hello,
> > > > > > > 
> > > > > > > On AMD platform as well boot failed starting next-20250422, bisecting the
> > > > > > > issue led me to same commit dd4cf8c9e1f4. I have attached kernel config and
> > > > > > > logs.
> > > > > > 
> > > > > > Thank you all for the bisection and the report!
> > > > > > 
> > > > > > Please check out the predecessor of commit dd4cf8c9e1f4 ("ratelimit:
> > > > > > Force re-initialization when rate-limiting re-enabled"):
> > > > > > 
> > > > > > 13fa70e052dd ("ratelimit: Allow zero ->burst to disable ratelimiting")
> > > > > > 
> > > > > > Then please apply the patch shown below, and let me know what happens?
> > > > > > (Yes, I should have split that commit up...)
> > > > > > 
> > > > > > 							Thanx, Paul
> > > > > > 
> > > > > > ------------------------------------------------------------------------
> > > > > > 
> > > > > > diff --git a/lib/ratelimit.c b/lib/ratelimit.c
> > > > > > index 04f16b8e24575..13ed636642270 100644
> > > > > > --- a/lib/ratelimit.c
> > > > > > +++ b/lib/ratelimit.c
> > > > > > @@ -35,7 +35,7 @@ int ___ratelimit(struct ratelimit_state *rs, const char *func)
> > > > > >   	unsigned long flags;
> > > > > >   	int ret;
> > > > > > -	if (!interval || !burst)
> > > > > > +	if (interval <= 0 || burst <= 0)
> > > > > >   		return 1;
> > > > > >   	/*
> > > > > 
> > > > > 
> > > > > I applied above patch on top of 13fa70e052dd ("ratelimit: Allow zero ->burst
> > > > > to disable ratelimiting") [linux-20250423]. This is fixing the boot issue.
> > > > > 
> > > > > Tested-by: Srikanth Aithal <sraithal@....com>
> > > > 
> > > > Thank you both, and to Bert for intuiting the correct -next commit!
> > > > 
> > > > Could you please try the next increment, which is this patch, again
> > > > on top of 24ff89c63355 ("ratelimit: Allow zero ->burst to > disable
> > > > ratelimiting")?
> > > > 
> > > > In the meantime, I will expose the version you two just tested to
> > > > -next.
> > > > 
> > > > 							Thanx, Paul
> > > > 
> > > > ------------------------------------------------------------------------
> > > > 
> > > > diff --git a/lib/ratelimit.c b/lib/ratelimit.c
> > > > index 04f16b8e24575..8f6c54f719ef2 100644
> > > > --- a/lib/ratelimit.c
> > > > +++ b/lib/ratelimit.c
> > > > @@ -35,8 +35,10 @@ int ___ratelimit(struct ratelimit_state *rs, const char *func)
> > > >  	unsigned long flags;
> > > >  	int ret;
> > > >  
> > > > -	if (!interval || !burst)
> > > > +	if (interval <= 0 || burst <= 0) {
> > > > +		ret = burst > 0;
> > > >  		return 1;
> > > > +	}
> > > >  
> > > >  	/*
> > > >  	 * If we contend on this state's lock then just check if
> > > 
> > > If you set "ret = burst > 0", but "return 1" this will make no difference
> > > (except in the case of a major compiler bug, probably), as I wrote in my other
> > > email which overlapped yours, this fixes the issue in next-20250422:
> > > 
> > > diff --git a/lib/ratelimit.c b/lib/ratelimit.c
> > > index b5c727e976d2..fc28f6cf8269 100644
> > > --- a/lib/ratelimit.c
> > > +++ b/lib/ratelimit.c
> > > @@ -40,7 +40,7 @@ int ___ratelimit(struct ratelimit_state *rs, const char *func)
> > >          * interval says never limit.
> > >          */
> > >         if (interval <= 0 || burst <= 0) {
> > > -               ret = burst > 0;
> > > +               ret = 1;
> > >                 if (!(READ_ONCE(rs->flags) & RATELIMIT_INITIALIZED) ||
> > >                     !raw_spin_trylock_irqsave(&rs->lock, flags))
> > >                         return ret;
> > 
> > You are quite right, your patch does fix the issue that you three say.
> > Unfortunately, it prevents someone from completely suppressing output
> > by setting burst to zero.  Could you please try the patch below?
> > 
> > 							Thanx, Paul
> > 
> > ------------------------------------------------------------------------
> > 
> > diff --git a/lib/ratelimit.c b/lib/ratelimit.c
> > index 04f16b8e24575..d6531e5c6ec4e 100644
> > --- a/lib/ratelimit.c
> > +++ b/lib/ratelimit.c
> > @@ -35,8 +35,8 @@ int ___ratelimit(struct ratelimit_state *rs, const char *func)
> >  	unsigned long flags;
> >  	int ret;
> >  
> > -	if (!interval || !burst)
> > -		return 1;
> > +	if (interval <= 0 || burst <= 0)
> > +		return interval == 0 || burst > 0;
> >  
> >  	/*
> >  	 * If we contend on this state's lock then just check if
> 
> This patch will make no difference for me as I've monitored every call to
> ___ratelimit() with "printk(KERN_INFO "%s: interval = %d burst = %d\n",
> __func__, interval, burst)" and at least for me burst == 0 is always acompanied
> by intervall == 0.

You lost me on this one.

With my previous "burst > 0", if both are zero, it would set ret to
false, right?

With the new "interval == 0 || burst > 0", if both are zero, then interval
must be zero, so the result would instead be true, correct?

So what am I missing here?

							Thanx, Paul