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;
Message-ID: <479acc85-2349-d0ac-851c-f57b1bf6aa9e@suse.com>
Date:   Mon, 8 May 2023 09:44:13 +0200
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:102.0) Gecko/20100101
 Thunderbird/102.10.1
Subject: Re: [PATCH 1/2] PCI: mediatek-gen3: Stop acquiring spinlocks in
 {suspend,resume}_noirq
To:     AngeloGioacchino Del Regno 
        <angelogioacchino.delregno@collabora.com>, ryder.lee@mediatek.com
Cc:     jianjun.wang@mediatek.com, lpieralisi@kernel.org, kw@linux.com,
        robh@kernel.org, bhelgaas@google.com, p.zabel@pengutronix.de,
        matthias.bgg@gmail.com, linux-pci@vger.kernel.org,
        linux-mediatek@lists.infradead.org, linux-kernel@vger.kernel.org,
        linux-arm-kernel@lists.infradead.org
References: <20230504113509.184633-1-angelogioacchino.delregno@collabora.com>
 <20230504113509.184633-2-angelogioacchino.delregno@collabora.com>
Content-Language: en-US
From:   Oliver Neukum <oneukum@suse.com>
In-Reply-To: <20230504113509.184633-2-angelogioacchino.delregno@collabora.com>
Content-Type: text/plain; charset=UTF-8; format=flowed
Content-Transfer-Encoding: 7bit
MIME-Version: 1.0
X-MS-Exchange-AntiSpam-MessageData-ChunkCount: 1
X-MS-Exchange-AntiSpam-MessageData-0: =?utf-8?B?WGttRXBkYXpCUWhKb3FDQnB6MEd5YlFtSnkveGszaEFpN0ExY1JPY0VtcEkv?=
 =?utf-8?B?bDhhVHZCOTArQXlHQ0w3Tmc5WXFBMUdRdXdlVUxiUmRKUzBxTGdYSGhqTWhv?=
 =?utf-8?B?Q2ZiSmxaQ09IVmF4SCs1SWhjWlU0YWFSbTFwdWd5VUppZDhWb1RRSTFWNHhn?=
 =?utf-8?B?QkhPTGpSTm1PTFJURWd1b2ljbzhXNFNja2V2ZDZrMkVad09ES2ZScGpuUzJU?=
 =?utf-8?B?cVpKZmMwSzM2aXhzNU9XWm1LbmFKNUVyN2pyYXcrTzJlWWpxS21IL1V6N1JL?=
 =?utf-8?B?cUR4L1dFRmJqK3hjbDZKcGdVQUR5ZGJYRDBQVENoWjdscUdzaVFsRWFrWnc2?=
 =?utf-8?B?ZnVxRDhuVnkxL0NJRnpENFp2a3ZnODRFdzRRSFMvOWpkaks4SXJRcHBUQndx?=
 =?utf-8?B?dzU1R0U0ZlIreWxkRW9PMEt4dzJQMHNXZzFrclBKbzljK3NWcTFsbm91c1kw?=
 =?utf-8?B?U25GODJOVFJwS1FCd1Rlc0ZDSzNCNk95aWNNa2U1YmFqSnRHS0EyT0ZuTVU1?=
 =?utf-8?B?NTJoSVlCT1l5UDhKK0w2QjV3NkFlcnZQWENJTHorMWRwbXVLRDFQSVh5bTIx?=
 =?utf-8?B?UVdwRWpaWDF5Nm5nTkxrWFc0dEtmdjVsUjh3Y1ZuZDcwMjhNN09IM2lQayt0?=
 =?utf-8?B?UVpBVEFTTHRPY1R5MjRSeVQwaTVaR29tdUU2NS9NcVhNb0trcXY3czVZenQ3?=
 =?utf-8?B?RlRGTmhxZWgrd0VrcEZjTjEwZk80dnJDdEMrNzdNcURic09PTHNBSUEwZHpy?=
 =?utf-8?B?V3R1TUUzT1R1S1F2QW9ERDJIcXZKOGNESnJoU2MxeGpCdGpDOFg4QU9Bbmdj?=
 =?utf-8?B?VXV1SSsrdUllLzdodHVzSXovc21VVWpCZkdSV0c5Ri91QlNNZnpXRW91MUxN?=
 =?utf-8?B?TXVHMUFtZU41RjJvZ0xxSzhvNzBmdTFqWmtmZlYvVnl6T1pxM1RqcGZ4RFpz?=
 =?utf-8?B?UVFjbzJKT1l3N0c1L0N5eHp0Q29QWFNNQldVaHdxcGN6aWt3dU5MUmFEMExv?=
 =?utf-8?B?L0tsOVV6Mjd6dE4yWlIzaEcxenk1NHhZOUtZbFZUOGN6S1I0VmZpbTdEelds?=
 =?utf-8?B?Q2ZkcmF2cit1RVB0UlJINmEyaVBYcG1MazErZnZaTm9zazVzRVhsVGwzYXMz?=
 =?utf-8?B?MklhU3ZMUjRUNlZPcEpnV3h0Wk5sc1VFbmtXdGRuUXFaamtPZnVLenltRnVM?=
 =?utf-8?B?UmpnUTU2ODlBU2NCUEMzNjJGb2dNNXQwVG5FNmh4N0czUVY3a1IyMTJQbXFJ?=
 =?utf-8?B?S29JUkJ2aVRFNEF3SmNWS2gwUEcxU0gra2R4dnhESm9ZbGdBOXRXNlIzVG9w?=
 =?utf-8?B?czRkTHRPc2tIdFFkYjdtRG42azhYR2lsSzdOaGh6b3ZyRGVYeVlPMk04MFdH?=
 =?utf-8?B?bDB6WGl2UjFUeUFYT2V0Q25nS1Yvc0F6Q1c2R0w0Qk5oSDgwY3BEVzZUazJl?=
 =?utf-8?B?UGpreUN5MzRJbDBqd3pGOFJ3WTJ1YUs3L2ptUG5iMW96YWpYR2V1alRWZHdr?=
 =?utf-8?B?ZGczMjlneEw4bXdkUlljNEtScjlhUW1RSS9DWWU5azhnSWRCaWE0SGRxV3Mx?=
 =?utf-8?B?UzgwQXcrOGJWY1VhK0p5RWx2MW9Va200WE1BQ09NTy9XOFBhYnRqTXdCTU9R?=
 =?utf-8?B?dHpBQ3RINmdmeGU3TXpjTDBtVjY5RjdLQndoMEY5TnVodTRJbkk3VzVJclRU?=
 =?utf-8?B?dVFWaXBTMDI5U3RhWXVoMk5CV0ZYSldzSyt3M0RXQUZBVnp5YVFPNm5JYkJX?=
 =?utf-8?B?cFgwbmJDS1VZS21UZWFKT1BJaXRJaWdTZFl0RDFpZm01U2FSMkJWejZyVUg2?=
 =?utf-8?B?ekJ5b3RNN21Tc1EycnUrMmloYUt2M08zSnp2b1lEeTk5dFRmSGxHdXdpZTVj?=
 =?utf-8?B?KzBSWVhLTEZKT2FFczhldXBsNmtjTzVLRnVEUE9WcWFSM2ovWGtMTXlSeXN1?=
 =?utf-8?B?b05jK1lHMSttcDZSVDN3VnFMVTQrM00vL213REJremZPWjFaYXBZbXM0cmZV?=
 =?utf-8?B?VGtzc0hMWTVHQ3BqUnNNWmRvSVEwc2pFM0pSZlF6eUNmdExzWUk2d3lONEM5?=
 =?utf-8?B?ck5xbEUyMWVzcVpSU0U0WFVCdDNCQ1puOFFrZ3dkamZjdCs5UVM1dVF1TktO?=
 =?utf-8?B?a00xYUxTVzFCV09Wamt6Vk1SKzR4UWRXTW9xam5zV1B4cVdSdTBHQlI4NktX?=
 =?utf-8?Q?avcAuC2n6cX7ZTrI/lblcRulox6dKeCc1KHx6Nu7PwOh?=
X-OriginatorOrg: suse.com
X-MS-Exchange-CrossTenant-Network-Message-Id: 8b5f57e7-5c23-4ac1-e3da-08db4f9806d0
X-MS-Exchange-CrossTenant-AuthSource: VI1PR04MB7104.eurprd04.prod.outlook.com
X-MS-Exchange-CrossTenant-AuthAs: Internal
X-MS-Exchange-CrossTenant-OriginalArrivalTime: 08 May 2023 07:44:17.4418
 (UTC)
X-MS-Exchange-CrossTenant-FromEntityHeader: Hosted
X-MS-Exchange-CrossTenant-Id: f7a17af6-1c5c-4a36-aa8b-f5be247aa4ba
X-MS-Exchange-CrossTenant-MailboxType: HOSTED
X-MS-Exchange-CrossTenant-UserPrincipalName: BJmfDuH8+sHJPzcl4u/4gIP+u/w2PXJroQfRmUrGxskrGK7iZRNu4xzb9ri+48pMRo5AOcEuqUCWJ641hehTmg==
X-MS-Exchange-Transport-CrossTenantHeadersStamped: VI1PR04MB6816
X-Spam-Status: No, score=-5.1 required=5.0 tests=BAYES_00,DKIM_SIGNED,
        DKIM_VALID,DKIM_VALID_AU,DKIM_VALID_EF,NICE_REPLY_A,RCVD_IN_DNSWL_NONE,
        RCVD_IN_MSPIKE_H2,SPF_HELO_NONE,SPF_PASS,T_SCC_BODY_TEXT_LINE
        autolearn=ham autolearn_force=no version=3.4.6
X-Spam-Checker-Version: SpamAssassin 3.4.6 (2021-04-09) on
        lindbergh.monkeyblade.net
Precedence: bulk
List-ID: <linux-kernel.vger.kernel.org>
X-Mailing-List: linux-kernel@vger.kernel.org

On 04.05.23 13:35, AngeloGioacchino Del Regno wrote:

Hi,

looking at your patch I am afraid there is an issue.

> In mtk_pcie_suspend_noirq() and mtk_pcie_resume_noirq() we are,
> respectively, disabling and enabling generation of interrupts and
> then saving and restoring the enabled interrupts register: since
> we're using noirq PM callbacks, that can be safely done without
> holding any spin lock.

Why? You can still race with another CPU in task context.
That is if you say that you do not need locking to touch
PCIE_INT_ENABLE_REG that is fine, but then why do you remove
it from one place only?
It is also touched in mtk_pcie_probe() at a minimum.

  
> That was noticed because of, and solves, the following issue:
> 
> <4>[   74.185982] ========================================================
> <4>[   74.192629] WARNING: possible irq lock inversion dependency detected
> <4>[   74.199276] 6.3.0-next-20230428+ #51 Tainted: G        W
> <4>[   74.205664] --------------------------------------------------------
> <4>[   74.212309] systemd-sleep/809 just changed the state of lock:
> <4>[   74.218347] ffff65a5c34c65a0 (&pcie->irq_lock){+...}-{2:2}, at: mtk_pcie_resume+0x50/0xa8
> <4>[   74.226870] but this lock was taken by another, HARDIRQ-safe lock in the past:
> <4>[   74.234389]  (&irq_desc_lock_class){-.-.}-{2:2}
> <4>[   74.234409]
> <4>[   74.234409]
> <4>[   74.234409] and interrupts could create inverse lock ordering between them.
> <4>[   74.234409]
> <4>[   74.251704]
> <4>[   74.251704] other info that might help us debug this:
> <4>[   74.258785]  Possible interrupt unsafe locking scenario:
> <4>[   74.258785]
> <4>[   74.266126]        CPU0                    CPU1
> <4>[   74.270942]        ----                    ----
> <4>[   74.275758]   lock(&pcie->irq_lock);

Lock A

> <4>[   74.279627]                                local_irq_disable();

strictly speaking irrelevant

> <4>[   74.285836]                                lock(&irq_desc_lock_class);

lock B

> <4>[   74.292667]                                lock(&pcie->irq_lock);

lock A

> <4>[   74.299061]   <Interrupt>

You do not need that interrupt.

> <4>[   74.301960]     lock(&irq_desc_lock_class);

lock B

> <4>[   74.306438]
> <4>[   74.306438]  *** DEADLOCK ***