Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
Date:   Fri, 1 Apr 2022 13:15:47 +0200
From:   Sebastian Andrzej Siewior <bigeasy@linutronix.de>
To:     "Zhang, Qiang1" <qiang1.zhang@intel.com>
Cc:     "ryabinin.a.a@gmail.com" <ryabinin.a.a@gmail.com>,
        "glider@google.com" <glider@google.com>,
        "andreyknvl@gmail.com" <andreyknvl@gmail.com>,
        "dvyukov@google.com" <dvyukov@google.com>,
        "akpm@linux-foundation.org" <akpm@linux-foundation.org>,
        "kasan-dev@googlegroups.com" <kasan-dev@googlegroups.com>,
        "linux-kernel@vger.kernel.org" <linux-kernel@vger.kernel.org>,
        "linux-rt-users@vger.kernel.org" <linux-rt-users@vger.kernel.org>
Subject: Re: [PATCH] kasan: Fix sleeping function called from invalid context
 in PREEMPT_RT
Message-ID: <Ykbe46hLAfJ8TsnW@linutronix.de>
References: <20220401091006.2100058-1-qiang1.zhang@intel.com>
 <YkbFhgN1jZPTMfnS@linutronix.de>
 <PH0PR11MB58800917A1BF8D1A76BEF84EDAE09@PH0PR11MB5880.namprd11.prod.outlook.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=utf-8
Content-Disposition: inline
In-Reply-To: <PH0PR11MB58800917A1BF8D1A76BEF84EDAE09@PH0PR11MB5880.namprd11.prod.outlook.com>
Precedence: bulk

On 2022-04-01 10:10:38 [+0000], Zhang, Qiang1 wrote:
> >Could we fix in a way that we don't involve freeing memory from in-IRQ?
> >This could trigger a lockdep splat if the local-lock in SLUB is acquired from in-IRQ context on !PREEMPT_RT.
> 
> Hi, I  will move qlist_free_all() from IPI context to task context,
> This operation and the next release  members
> in the quarantine pool operate similarly
> 
> I don't know the phenomenon you described. Can you explain it in detail?

If you mean by phenomenon my second sentence then the kernel option
CONFIG_PROVE_RAW_LOCK_NESTING will trigger on !PREEMPT_RT in a code
sequence like
	raw_spin_lock()
	spin_lock();

which is wrong on PREEMPT_RT. So we have a warning on both
configurations.
The call chain in your case will probably not lead to a warning since
there is no raw_spinlock_t involved within the IPI call. We worked on
avoiding memory allocation and freeing from in-IRQ context therefore I
would prefer to have something that works for both and not just ifdef
around the RT-case.

> Thanks
> Zqiang

Sebastian