Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.31 as permitted sender) client-ip=23.128.96.31;
Message-ID: <62ca6115-cf0d-427d-92c5-3539a361b476@samsung.com>
Date:   Fri, 6 Oct 2023 10:36:27 +0200
MIME-Version: 1.0
User-Agent: Mozilla Thunderbird
From:   Marek Szyprowski <m.szyprowski@samsung.com>
Subject: Re: [PATCH] sched/fair: fix pick_eevdf to always find the correct
 se
To:     Biju Das <biju.das.jz@bp.renesas.com>,
        Peter Zijlstra <peterz@infradead.org>
Cc:     "bsegall@google.com" <bsegall@google.com>,
        "bristot@redhat.com" <bristot@redhat.com>,
        "chris.hyser@oracle.com" <chris.hyser@oracle.com>,
        "corbet@lwn.net" <corbet@lwn.net>,
        "dietmar.eggemann@arm.com" <dietmar.eggemann@arm.com>,
        "efault@gmx.de" <efault@gmx.de>,
        "gregkh@linuxfoundation.org" <gregkh@linuxfoundation.org>,
        "joel@joelfernandes.org" <joel@joelfernandes.org>,
        "joshdon@google.com" <joshdon@google.com>,
        "juri.lelli@redhat.com" <juri.lelli@redhat.com>,
        "kprateek.nayak@amd.com" <kprateek.nayak@amd.com>,
        "linux-kernel@vger.kernel.org" <linux-kernel@vger.kernel.org>,
        "mingo@kernel.org" <mingo@kernel.org>,
        "patrick.bellasi@matbug.net" <patrick.bellasi@matbug.net>,
        Pavel Machek <pavel@ucw.cz>, "pjt@google.com" <pjt@google.com>,
        "qperret@google.com" <qperret@google.com>,
        "qyousef@layalina.io" <qyousef@layalina.io>,
        "rostedt@goodmis.org" <rostedt@goodmis.org>,
        "tglx@linutronix.de" <tglx@linutronix.de>,
        "tim.c.chen@linux.intel.com" <tim.c.chen@linux.intel.com>,
        "timj@gnu.org" <timj@gnu.org>,
        "vincent.guittot@linaro.org" <vincent.guittot@linaro.org>,
        "youssefesmat@chromium.org" <youssefesmat@chromium.org>,
        "yu.c.chen@intel.com" <yu.c.chen@intel.com>,
        "mgorman@suse.de" <mgorman@suse.de>,
        "linux-renesas-soc@vger.kernel.org" 
        <linux-renesas-soc@vger.kernel.org>
Content-Language: en-US
In-Reply-To: <OS0PR01MB592295E06AD01CAEFBA83BF386CAA@OS0PR01MB5922.jpnprd01.prod.outlook.com>
Content-Transfer-Encoding: 7bit
Content-Type: text/plain; charset="utf-8"
CMS-TYPE: 201P
References: <OS0PR01MB59220AF3959BDC5FEFC0340F86CAA@OS0PR01MB5922.jpnprd01.prod.outlook.com>
        <20231005150258.GA36277@noisy.programming.kicks-ass.net>
        <CGME20231005150845eucas1p1ed3aae6b90c411b5c26a5dfadf7e0733@eucas1p1.samsung.com>
        <OS0PR01MB592295E06AD01CAEFBA83BF386CAA@OS0PR01MB5922.jpnprd01.prod.outlook.com>
Precedence: bulk

On 05.10.2023 17:08, Biju Das wrote:
> Hi Peter Zijlstra,
>
>> Subject: Re: [PATCH] sched/fair: fix pick_eevdf to always find the correct
>> se
>>
>> On Thu, Oct 05, 2023 at 07:31:34AM +0000, Biju Das wrote:
>>
>>> [   26.099203] EEVDF scheduling fail, picking leftmost
>> This, that the problem.. the rest is just noise because printk stinks.
>>
>> Weirdly have not seen that trigger, and I've been running with this patch
>> on for a few days now :/
> I agree Original issue is "EEVDF scheduling fail, picking leftmost"
> Which is triggering noisy lock warning messages during boot.
>
> 2 platforms are affected both ARM platforms(Renesas and Samsung)
> Maybe other platforms affected too.


Just to note, I've run into this issue on the QEmu's 'arm64/virt' 
platform, not on the Samsung specific hardware.


You can easily reproduce it with the following steps:

# make ARCH=arm64 CROSS_COMPILE=aarch64-linux-gnu- defconfig

# ./scripts/config -e PROVE_LOCKING -e DEBUG_ATOMIC_SLEEP -e PM_DEBUG -e 
PM_ADVANCED_DEBUG

# make ARCH=arm64 CROSS_COMPILE=aarch64-linux-gnu- olddefconfig Image.gz

# wget 
https://cloud.debian.org/images/cloud/buster/20230802-1460/debian-10-nocloud-arm64-20230802-1460.tar.xz

# tar xJfv debian-10-nocloud-arm64-20230802-1460.tar.xz


Then run QEmu a few times until you see the lockdep splat:

# qemu-system-aarch64 -serial stdio -kernel arch/arm64/boot/Image 
-append "console=ttyAMA0 root=/dev/vda1 rootwait" -M virt -cpu 
cortex-a57 -smp 2 -m 1024 -netdev user,id=user -device 
virtio-net-device,netdev=user -display none -device 
virtio-blk-device,drive=virtio-blk0 -drive 
file=disk.raw,id=virtio-blk0,if=none,format=raw


I have no idea if this is ARM64-specific anyhow, but this is how I 
reproduced it. I hope this guide helps fixing the bug!

Best regards
-- 
Marek Szyprowski, PhD
Samsung R&D Institute Poland