From patchwork Mon Nov 5 23:25:00 2018 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Marek Vasut X-Patchwork-Id: 10669335 X-Patchwork-Delegate: bhelgaas@google.com Return-Path: Received: from mail.wl.linuxfoundation.org (pdx-wl-mail.web.codeaurora.org [172.30.200.125]) by pdx-korg-patchwork-2.web.codeaurora.org (Postfix) with ESMTP id C66FC15E9 for ; Mon, 5 Nov 2018 23:25:10 +0000 (UTC) Received: from mail.wl.linuxfoundation.org (localhost [127.0.0.1]) by mail.wl.linuxfoundation.org (Postfix) with ESMTP id BE86529FF4 for ; Mon, 5 Nov 2018 23:25:10 +0000 (UTC) Received: by mail.wl.linuxfoundation.org (Postfix, from userid 486) id 9FCE029FED; Mon, 5 Nov 2018 23:25:10 +0000 (UTC) X-Spam-Checker-Version: SpamAssassin 3.3.1 (2010-03-16) on pdx-wl-mail.web.codeaurora.org X-Spam-Level: X-Spam-Status: No, score=-8.0 required=2.0 tests=BAYES_00,DKIM_SIGNED, DKIM_VALID,DKIM_VALID_AU,FREEMAIL_FROM,MAILING_LIST_MULTI,RCVD_IN_DNSWL_HI autolearn=unavailable version=3.3.1 Received: from vger.kernel.org (vger.kernel.org [209.132.180.67]) by mail.wl.linuxfoundation.org (Postfix) with ESMTP id 3A7B629FEE for ; Mon, 5 Nov 2018 23:25:10 +0000 (UTC) Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1725839AbeKFIrS (ORCPT ); Tue, 6 Nov 2018 03:47:18 -0500 Received: from mail-wm1-f68.google.com ([209.85.128.68]:33877 "EHLO mail-wm1-f68.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1725838AbeKFIrR (ORCPT ); Tue, 6 Nov 2018 03:47:17 -0500 Received: by mail-wm1-f68.google.com with SMTP id f1-v6so7482054wmg.1; Mon, 05 Nov 2018 15:25:07 -0800 (PST) DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=gmail.com; s=20161025; h=from:to:cc:subject:date:message-id; bh=3XCO7P5FWCHRNFc55rpR1NNXWKHmhwb46/s8mgDKOYI=; b=RX6C8qejlmRklqjnBTgag2EmeRn1s/Gfh2ny5lVxTxabVRezSjSg8EYZcEpL6z0c8u tPAfTSg9Uj3Q7v1XLiagtWwJMENheqhFeh+7/goCjbuIflB0qiE2AnbPz/9j49okHyQo Qp7nUfQGnk0NS3NAUDiv919T45g+d9NwAwCNrYMytPyz7CgZJOhod52zg9ZvMciSSWGc 5WcvoNM8y3PILfRK4rgl8ybiOAuiIlaMFjvnL6KdLDADDsxOihlgfqRIFRwLIVtXR1IM Oit6PRbOXdNnrXqKVyS8b3HqOvxoidWtj/dI2OBV4zTKXK9tlPxM4XjOHyNHPZNtv9aI qCzQ== X-Google-DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=1e100.net; s=20161025; h=x-gm-message-state:from:to:cc:subject:date:message-id; bh=3XCO7P5FWCHRNFc55rpR1NNXWKHmhwb46/s8mgDKOYI=; b=CwKYjAs5hrSczTW9Gi0ZEfxWrvQ7AhTzXy1dzVQyaOld4nT856e/LaXOSYZTDpx5A3 mMSSUWZu5RZVNEvDrbrDb7gx8G9lgcAuhBmcGHR1fh8V0jNSVpWLL/wW1foSJX4eIJRI YEV4DGTCJfDUSd3WacZD3sAnvaNlDTGKH/vVUXvRzduKZAYsCy/yjp6QO7sd+zoQhw/N OfRv/KMnIn/EdR+mqgLdR7hLDkJzMTRpqit9kivq3DkxCD7J61dvyqWv1Vmqo1/mXeff wmrpYDOABGfYeSJFH7HaCAdHHrOvD2y+qn+7Bn8HbEeiFLrYYF0vsy4TA/OBbueBY1nK iwqg== X-Gm-Message-State: AGRZ1gLYER771VJoXkiBoRBOCncj6PMVW9xEYqyAB7vPy1De75M/o1NB rIRm6JQmrz5RFLTD0uTrS8C78oWr X-Google-Smtp-Source: AJdET5cQwuZclwsV+c+Tpf3XfK9YRbpaddgTut9+MQutWyocRla6pP9XXL57/onpa5Vxqsrsmulcbg== X-Received: by 2002:a1c:f313:: with SMTP id q19-v6mr102079wmq.87.1541460306316; Mon, 05 Nov 2018 15:25:06 -0800 (PST) Received: from kurokawa.lan (ip-86-49-110-70.net.upcbroadband.cz. [86.49.110.70]) by smtp.gmail.com with ESMTPSA id o18-v6sm6827309wrp.31.2018.11.05.15.25.05 (version=TLS1_2 cipher=ECDHE-RSA-AES128-GCM-SHA256 bits=128/128); Mon, 05 Nov 2018 15:25:05 -0800 (PST) From: Marek Vasut X-Google-Original-From: Marek Vasut To: linux-pci@vger.kernel.org Cc: linux-renesas-soc@vger.kernel.org, Tho Vu Subject: [RFC][PATCH] PCI: Avoid PCI device removing/rescanning through sysfs triggers a deadlock Date: Tue, 6 Nov 2018 00:25:00 +0100 Message-Id: <20181105232500.19146-1-marek.vasut+renesas@gmail.com> X-Mailer: git-send-email 2.18.0 Sender: linux-pci-owner@vger.kernel.org Precedence: bulk List-ID: X-Mailing-List: linux-pci@vger.kernel.org X-Virus-Scanned: ClamAV using ClamSMTP From: Tho Vu This patch fixes deadlock warning in removing/rescanning through sysfs when CONFIG_PROVE_LOCKING is enabled. The issue can be reproduced by these steps: 1. Enable CONFIG_PROVE_LOCKING via defconfig or menuconfig 2. Insert Ethernet card into PCIe CH0 and start up. After kernel starting up, execute the following command. echo 1 > /sys/class/pci_bus/0000\:00/device/0000\:00\:00.0/remove 3. Rescan PCI device by this command echo 1 > /sys/class/pci_bus/0000\:00/rescan The deadlock warnings will occur. ============================================ WARNING: possible recursive locking detected 4.14.70-ltsi-yocto-standard #27 Not tainted -------------------------------------------- sh/3402 is trying to acquire lock: (kn->count#78){++++}, at: kernfs_remove_by_name_ns+0x50/0xa8 but task is already holding lock: (kn->count#78){++++}, at: kernfs_remove_self+0xe0/0x130 other info that might help us debug this: Possible unsafe locking scenario: CPU0 ---- lock(kn->count#78); lock(kn->count#78); *** DEADLOCK *** May be due to missing lock nesting notation 4 locks held by sh/3402: #0: (sb_writers#4){.+.+}, at: vfs_write+0x198/0x1b0 #1: (&of->mutex){+.+.}, at: kernfs_fop_write+0x108/0x210 #2: (kn->count#78){++++}, at: kernfs_remove_self+0xe0/0x130 #3: (pci_rescan_remove_lock){+.+.}, at: pci_lock_rescan_remove+0x1c/0x28 stack backtrace: CPU: 3 PID: 3402 Comm: sh Not tainted 4.14.70-ltsi-yocto-standard #27 Hardware name: Renesas Salvator-X 2nd version board based on r8a7795 ES3.0+ with 8GiB (4 x 2 GiB) (DT) Call trace: dump_backtrace+0x0/0x3d8 show_stack+0x14/0x20 dump_stack+0xbc/0xf4 __lock_acquire+0x930/0x18a8 lock_acquire+0x48/0x68 __kernfs_remove+0x280/0x2f8 kernfs_remove_by_name_ns+0x50/0xa8 remove_files.isra.0+0x38/0x78 sysfs_remove_group+0x4c/0xa0 sysfs_remove_groups+0x38/0x60 device_remove_attrs+0x54/0x78 device_del+0x1ac/0x308 pci_remove_bus_device+0x78/0xf8 pci_remove_bus_device+0x34/0xf8 pci_stop_and_remove_bus_device_locked+0x24/0x38 remove_store+0x6c/0x78 dev_attr_store+0x18/0x28 sysfs_kf_write+0x4c/0x78 kernfs_fop_write+0x138/0x210 __vfs_write+0x18/0x118 vfs_write+0xa4/0x1b0 SyS_write+0x48/0xb0 This warning occurs due to a self-deletion attribute using in the sysfs PCI device directory. This kind of attribute is really tricky, it does not allow pci framework drop this attribute until all active .show() and .store() callbacks have finished unless sysfs_break_active_protection() is called. Hence this patch avoids writing into this attribute triggers a deadlock. Referrence commit 5b55b24cec4c ("scsi: core: Avoid that SCSI device removal through sysfs triggers a deadlock") of scsi driver Signed-off-by: Tho Vu --- drivers/pci/pci-sysfs.c | 20 +++++++++++++++++++- 1 file changed, 19 insertions(+), 1 deletion(-) diff --git a/drivers/pci/pci-sysfs.c b/drivers/pci/pci-sysfs.c index 9ecfe13157c0..d522bd8368d9 100644 --- a/drivers/pci/pci-sysfs.c +++ b/drivers/pci/pci-sysfs.c @@ -470,12 +470,22 @@ static ssize_t remove_store(struct device *dev, struct device_attribute *attr, const char *buf, size_t count) { unsigned long val; + struct kernfs_node *kn; + + kn = sysfs_break_active_protection(&dev->kobj, &attr->attr); + WARN_ON_ONCE(!kn); if (kstrtoul(buf, 0, &val) < 0) return -EINVAL; - if (val && device_remove_file_self(dev, attr)) + if (val) { + device_remove_file(dev, attr); pci_stop_and_remove_bus_device_locked(to_pci_dev(dev)); + } + + if (kn) + sysfs_unbreak_active_protection(kn); + return count; } static struct device_attribute dev_remove_attr = __ATTR(remove, @@ -487,11 +497,15 @@ static ssize_t dev_bus_rescan_store(struct device *dev, const char *buf, size_t count) { unsigned long val; + struct kernfs_node *kn; struct pci_bus *bus = to_pci_bus(dev); if (kstrtoul(buf, 0, &val) < 0) return -EINVAL; + kn = sysfs_break_active_protection(&dev->kobj, &attr->attr); + WARN_ON_ONCE(!kn); + if (val) { pci_lock_rescan_remove(); if (!pci_is_root_bus(bus) && list_empty(&bus->devices)) @@ -500,6 +514,10 @@ static ssize_t dev_bus_rescan_store(struct device *dev, pci_rescan_bus(bus); pci_unlock_rescan_remove(); } + + if (kn) + sysfs_unbreak_active_protection(kn); + return count; } static DEVICE_ATTR(rescan, (S_IWUSR|S_IWGRP), NULL, dev_bus_rescan_store);