From mboxrd@z Thu Jan 1 00:00:00 1970 Received: from mail-wm2-f13.google.com (mail-wm2-f13.google.com [74.125.225.141]) (using TLSv1.2 with cipher ECDHE-RSA-AES128-GCM-SHA256 (128/128 bits)) (No client certificate requested) by smtp.subspace.kernel.org (Postfix) with ESMTPS id 2DC3F2874E1 for ; Sun, 27 Sep 2026 17:52:10 +0000 (UTC) Authentication-Results: smtp.subspace.kernel.org; arc=none smtp.client-ip=74.125.225.141 ARC-Seal:i=1; a=rsa-sha256; d=subspace.kernel.org; s=arc-20240116; t=1790531532; cv=none; b=N3MiT9gd4AdCoUBFNBPLD/1zKHd/SquW/tUe//oaPoQQuHurMzsd8ssjxxTwFbFLSeHvJmLhvUOl9nScAwZoRT7bRmJ2aJBdPykBjqH7V0NKKkzg5HR06Yc7iSJogxv54AzlrSXooGVIoy32pZXrFVxnHre0pQMq6vUM7GBMCAg= ARC-Message-Signature:i=1; a=rsa-sha256; d=subspace.kernel.org; s=arc-20240116; t=1790531532; c=relaxed/simple; bh=o7ZFXih1aRtJWFxQcya4mju/4ru4hLafjVLjye66/g4=; h=From:To:Cc:Subject:Date:Message-ID:MIME-Version; b=VTqTmzIrHqCI4n7nLOed3J9QmO90X1IDkLwMi/braMid/14o2GsH0Zss1tfYgqgX8dAVmWQXpaxYkYuAJNcARnj0U3PmtDz/LyVd30h96FJnualuI0AEZ3jtrSqYEGNELANAL6JR2Wax7YaqKr4f/F7gTCMRn3HvvCQiGyg1C2E= ARC-Authentication-Results:i=1; smtp.subspace.kernel.org; dmarc=pass (p=none dis=none) header.from=gmail.com; spf=pass smtp.mailfrom=gmail.com; dkim=pass (2048-bit key) header.d=gmail.com header.i=@gmail.com header.b=PXHyjPG9; arc=none smtp.client-ip=74.125.225.141 Authentication-Results: smtp.subspace.kernel.org; dmarc=pass (p=none dis=none) header.from=gmail.com Authentication-Results: smtp.subspace.kernel.org; spf=pass smtp.mailfrom=gmail.com Authentication-Results: smtp.subspace.kernel.org; dkim=pass (2048-bit key) header.d=gmail.com header.i=@gmail.com header.b="PXHyjPG9" Received: by mail-wm2-f13.google.com with SMTP id 5b1f17b1804b1-49e69b9e16aso25660335e9.1 for ; Sun, 27 Sep 2026 10:52:10 -0700 (PDT) DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=gmail.com; s=20251104; t=1790531529; x=1791136329; darn=vger.kernel.org; h=content-transfer-encoding:mime-version:message-id:date:subject:cc :to:from:from:to:cc:subject:date:message-id:reply-to:content-type; bh=A69PJQ7Nmi5xcTUI4BotLtqEyRQMpUL1rHDBHvcBg3M=; b=PXHyjPG91LAMJjuzCBxAoF96JlOHS84QxI51PtXP3yDh9yNYkuWc8iq0XqdkwFKe5T xnHc7vxDpKbPSg5t6JfZ6rbyFHypmLr1b0ntR5dIsyDZad6F6a8D2c/FLLjLF67/qHeW YGYuxaoiY2WniUa5NDE07WK9i+SrJRv8UJWHo0YQz3Dg1flHjZZDna8w+Jk48EkepBZk BFQXUGhDbW8hFqa63AXNl3AqqfAJEd+y/OA9pllc4djJnw9nsVPu4eAcEq5DXBaZdB5r KS9l0fbBGlacdDbL3SJ8w0AUyDa6jbpwjdoMuQ4NiqxrAQ3Pvhm7XVA/WHi+ddvcjL78 6K8g== X-Google-DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=1e100.net; s=20260707; t=1790531529; x=1791136329; h=content-transfer-encoding:mime-version:message-id:date:subject:cc :to:from:x-gm-gg:x-gm-message-state:from:to:cc:subject:date :message-id:reply-to:content-type; bh=A69PJQ7Nmi5xcTUI4BotLtqEyRQMpUL1rHDBHvcBg3M=; b=BkaUiHxn/6hBPjO+tLNls2+bRQ/+MZy8IfYSUi9lZ1F78UpQen9XEEcB6wLZFQJoLZ /xdbfFgC4Yii8DxdYYRBbWvIMSXxrpkmbykTHYddkTtX//HbBr6lV3CuT/+ojWUIRiDa 57B3w9QTNvmCrsM89EAN+2IUFp8sijyLkEB/XFSOj5nEjYzsRhHREx2cgsGQtIghBs+l mDcT7gG6FCzhCt2YOVlPaTfMo3iMQy6p6kA2AX+GTUQPlE70uucAZ/BbNqoT10253UGd Tcuoa2H4qe0sqtxFvnQFE3Xij7Gy20OMPb3J0w8rCXBJ7KjktFXxQLXtNzSwReUSwheW Fuew== X-Forwarded-Encrypted: i=1; AKwUvBwiuRT7ddIqlyvIRNZMgJ3NK5hwBgnFV6pXp7KR8EuXyUm2o4FdNMg0MVdUxIjDkgkFbIty827I7Qloe+Q=@vger.kernel.org X-Gm-Message-State: AFuF++mst2sumxMz2i/gQ8p4DRzh8YRc13F9v/cNvT0NqwJcv4aun50L kE1xu2nM1sFclAQXn5PA10YM9hrgBsoYU9aOKCNjiS+dpBGAFQ1nksSV X-Gm-Gg: AYBFou1FQsVXM3iCrn7BvBtGfpq3s3kVebUgsWzMJ0k/YFYY9+8goJtXn3irRI43J5X PNJOT3+bbsyQDOaipyp7N4kU9HLPGuljp+p+opx4AsOpxk9RyIPQ3ZwV6NFekmjN4UwqzyoFGrq ukcAic9T9vjBeqlmXxus938ALcpX15ymcLtJ7Xcmt0/cPJOquzq5ZjKOMLTTkf1Y2jghHbZR5Jn g0gfGTPuNF5CvYg2D/oWBCOq/wQtrAxHslQQahmDqBtk23z/xolUkvxAFh5o2uhvWD600ab5rtQ wBgFn9XxUYiiOMxnK1WZjdcngCv6zawx7uz0CBYcBQ7Ii7Iu6Zvy0yWvBfOKUlHw6SOw3DLPJXR UWmKdxmHTfwyV8Xcg1UZGbmSTE6yzdwvrGLRYuAOv2EhxfVagLRCAY3CQ58jbkthJZdqwQiMn3W cldG+Ui/a/veTsTR+cQtbtItatBixXY8g8XxMGqB7uU9pBLGLwfcb8z3S8+Js5uyTy2ckp347P4 /Qj4gEwTlQFq5FB8sYQoyaWoYr1hNgSR19hPLvDjYceX0QWVznXUDE3Pbw7Ua/D41w= X-Received: by 2002:a05:600c:19c9:b0:49c:d019:70c5 with SMTP id 5b1f17b1804b1-49fe7ae84a9mr187353025e9.0.1790531529031; Sun, 27 Sep 2026 10:52:09 -0700 (PDT) Received: from f3a6eae2255e.fritz.box (dynamic-002-214-014-217.2.214.pool.telefonica.de. [2.214.14.217]) by smtp.gmail.com with ESMTPSA id ffacd0b85a97d-4887a30bcc4sm22563923f8f.1.2026.09.27.10.52.07 (version=TLS1_3 cipher=TLS_AES_256_GCM_SHA384 bits=256/256); Sun, 27 Sep 2026 10:52:08 -0700 (PDT) From: Abhin Parekadan Jose To: Bjorn Helgaas , Lukas Wunner , "Michael S. Tsirkin" Cc: =?UTF-8?q?Ilpo=20J=C3=A4rvinen?= , Shuai Xue , Kees Cook , Mahesh J Salgaonkar , Oliver O'Halloran , linux-pci@vger.kernel.org, linuxppc-dev@lists.ozlabs.org, linux-kernel@vger.kernel.org, Abhin Parekadan Jose Subject: [PATCH RFC v3 0/5] PCI: pciehp: Report surprise removal during safe removal Date: Sun, 27 Sep 2026 17:51:57 +0000 Message-ID: <20260927175203.928270-1-abhinjoses@gmail.com> X-Mailer: git-send-email 2.51.1 Precedence: bulk X-Mailing-List: linux-kernel@vger.kernel.org List-Id: List-Subscribe: List-Unsubscribe: MIME-Version: 1.0 Content-Transfer-Encoding: 8bit If a safe removal is already in progress when the device is surprise removed, pciehp cannot report the disconnect [1]. The removal blocks waiting on a device interrupt or status read, and pciehp's IRQ thread is single-threaded and is itself executing that removal, so it never runs again to report the device gone. The removal hangs indefinitely. pciehp_isr() does run while the IRQ thread is blocked, but pciehp_ist() must ignore link and presence changes caused by Secondary Bus Reset or DPC, and telling those apart takes seconds which cannot be spent in hardirq. Patch 4 does not do that work in hardirq. pciehp_isr() only checks whether the card is present or the link is active, and defers the rest to a work item in process context. The work item does not need to know why the link changed, only whether the card is gone: it waits for DPC recovery or a Secondary Bus Reset in progress to complete, then checks again under reset_lock, and if the card is absent and the link is down marks the devices below disconnected, which schedules the driver's disconnect work from patch 1. pciehp_ist() is unchanged and remains the only consumer of the one-shot flags PCI_DPC_RECOVERED and PCI_LINK_CHANGED, which tell it whether a link change can be ignored. Patches: 1/5 Michael's "PCI: Report surprise removal event" from his RFC v5, which adds disconnect_work_enable and pdev->disconnect_work. Changed to serialize disconnect_work_enable with a per-device spinlock. 2/5 Add pci_hp_wait_link_change(), which awaits a Secondary Bus Reset in progress without consuming PCI_LINK_CHANGED. 3/5 Add pci_dpc_wait_recovery(), which awaits DPC recovery without consuming PCI_DPC_RECOVERED. 4/5 The pciehp change. Adds disconnect_work to struct controller, scheduled from pciehp_isr() on PDC or DLLSC when neither Presence Detect State nor Data Link Layer Link Active indicates a card. 5/5 A POC driver for the QEMU edu device that blocks in remove() waiting for an interrupt, standing in for del_gendisk() stuck in blk_mq_freeze_queue_wait(). Not for merge -- included so the hang can be reproduced. Changes since RFC v2 [4]: - 1/5: Protect disconnect_work_enable with a per-device spinlock, held while testing it and scheduling the work, instead of lockless accesses and barriers. pci_dev_set_disconnected() could otherwise test the flag, get preempted, and queue the work while a newly bound driver re-initializes it. With the lock, cancel_work_sync() is sufficient again, so go back to it from disable_work_sync(). (Sashiko) - 2/5: Test PCI_LINK_CHANGING with test_bit_acquire(), so that the caller's subsequent accesses are ordered after the end of the code section even if wait_event() returns without sleeping. (Sashiko) - 3/5, 4/5, 5/5: No changes. Changes since RFC v1 [2]: - 1/5: Use disable_work_sync() instead of cancel_work_sync() in pci_clear_disconnect_work(), so that a racing schedule_work() cannot queue the work after remove() has returned. (Sashiko) - 2/5, 3/5: New. - 4/5: Drop schedule_notification_work(); pci_dev_set_disconnected() schedules the driver's disconnect work for all callers again. (Michael) - 4/5: Don't consume PCI_DPC_RECOVERED and PCI_LINK_CHANGED in the work item. In v1 it ran the same spurious link change test as pciehp_ist(), and whichever ran first took the flags, so pciehp_ist() could tear down a device that was only reset. (Sashiko) - 4/5: Check presence with pciehp_card_present_or_link_active() in both pciehp_isr() and the work item, as pciehp_ist() does, so that a port with Presence Detect State hardwired to zero is not mistaken for an empty slot. - 4/5: Return early from the work item if pciehp_ist() has already taken the pending events, and treat a read error of the presence check as "not present". (Sashiko) - 4/5: In pciehp_isr(), check presence before dropping the runtime PM reference on the port's parent. In the work item, take a runtime PM reference and check presence under reset_lock, since a slot reset may make Presence Detect State and Link Active flap. - 5/5: Build only with CONFIG_EDU_SRPOC, which depends on PCI; don't claim the interrupt when the device reads all ones; clear bus mastering on teardown. (Sashiko) Testing Reproducing this needs QEMU changes, since neither device_del nor the attention button produces a true surprise removal. A branch with both is here [3]: - a delayed-IRQ register on the edu device (BAR0 0x30, write N ms) - a pcie_surprise_del monitor command that drops the device and generates PDC=1, DLLSC=1, PDS=0 Both tests use: ./qemu-system-aarch64 -machine virt,gic-version=3 -cpu cortex-a57 \ -m 512 -smp 2 -kernel Image -initrd initramfs.cpio.gz \ -device pcie-root-port,id=rp1,chassis=1,slot=1 \ -device edu,bus=rp1,id=edu0 -append "console=ttyAMA0 rdinit=/init" \ -nographic -monitor unix:/tmp/qemu-mon.sock,server,nowait and need a guest kernel with CONFIG_EDU_SRPOC=y. Test 1 (Hang in remove() on a user thread, then surprise removal): guest# echo 1 > /sys/bus/pci/devices/0000:01:00.0/remove & host$ echo "pcie_surprise_del edu0" | socat - unix-connect:/tmp/qemu-mon.sock This is the case patch 1 solves on its own: pciehp's IRQ thread is free and handles the removal. Test 2 (Hang in remove() on pciehp's IRQ thread, then surprise removal): guest# echo 0 > /sys/bus/pci/slots/1/power & host$ echo "pcie_surprise_del edu0" | socat - unix-connect:/tmp/qemu-mon.sock Without patch 4 the safe removal does not return until the delayed interrupt fires 600 s later. With it, pciehp_isr() schedules ctrl->disconnect_work, which marks the device disconnected; the POC driver's disconnect work completes the wait and remove() proceeds. Open questions - Is this a viable approach? [1] https://lore.kernel.org/all/aHlZE18kPuHuDtTT@wunner.de/ [2] https://lore.kernel.org/all/20260905183905.997833-1-abhinjoses@gmail.com/ [3] https://gitlab.com/abhinkop/qemu/-/commits/suprise-removal [4] https://lore.kernel.org/all/20260927165459.829900-1-abhinjoses@gmail.com/ Assisted-by: LLM Abhin Parekadan Jose (4): PCI: pciehp: Add pci_hp_wait_link_change() PCI/DPC: Add pci_dpc_wait_recovery() PCI: pciehp: Report surprise removal from pciehp_isr() misc: Add edu_srpoc surprise removal POC driver Michael S. Tsirkin (1): PCI: Report surprise removal event drivers/misc/Kconfig | 11 ++ drivers/misc/Makefile | 1 + drivers/misc/edu_srpoc.c | 171 +++++++++++++++++++++++++ drivers/pci/hotplug/pci_hotplug_core.c | 21 ++- drivers/pci/hotplug/pciehp.h | 1 + drivers/pci/hotplug/pciehp_hpc.c | 67 ++++++++++ drivers/pci/pci.h | 10 ++ drivers/pci/pcie/dpc.c | 54 ++++++-- drivers/pci/probe.c | 1 + include/linux/pci.h | 57 +++++++++ 10 files changed, 380 insertions(+), 14 deletions(-) create mode 100644 drivers/misc/edu_srpoc.c base-commit: fd179f8a05be3ccae366b9b96e176b51fbe54aab -- 2.51.1