{"id":2224052,"url":"http://patchwork.ozlabs.org/api/covers/2224052/?format=json","web_url":"http://patchwork.ozlabs.org/project/linux-pci/cover/20260416180107.777065-1-kwilczynski@kernel.org/","project":{"id":28,"url":"http://patchwork.ozlabs.org/api/projects/28/?format=json","name":"Linux PCI development","link_name":"linux-pci","list_id":"linux-pci.vger.kernel.org","list_email":"linux-pci@vger.kernel.org","web_url":null,"scm_url":null,"webscm_url":null,"list_archive_url":"","list_archive_url_format":"","commit_url_format":""},"msgid":"<20260416180107.777065-1-kwilczynski@kernel.org>","list_archive_url":null,"date":"2026-04-16T18:00:44","name":"[v5,00/23] PCI: Convert all dynamic sysfs attributes to static","submitter":{"id":86709,"url":"http://patchwork.ozlabs.org/api/people/86709/?format=json","name":"Krzysztof Wilczyński","email":"kwilczynski@kernel.org"},"mbox":"http://patchwork.ozlabs.org/project/linux-pci/cover/20260416180107.777065-1-kwilczynski@kernel.org/mbox/","series":[{"id":500193,"url":"http://patchwork.ozlabs.org/api/series/500193/?format=json","web_url":"http://patchwork.ozlabs.org/project/linux-pci/list/?series=500193","date":"2026-04-16T18:00:44","name":"PCI: Convert all dynamic sysfs attributes to static","version":5,"mbox":"http://patchwork.ozlabs.org/series/500193/mbox/"}],"comments":"http://patchwork.ozlabs.org/api/covers/2224052/comments/","headers":{"Return-Path":"\n ","X-Original-To":["incoming@patchwork.ozlabs.org","linux-pci@vger.kernel.org"],"Delivered-To":"patchwork-incoming@legolas.ozlabs.org","Authentication-Results":["legolas.ozlabs.org;\n\tdkim=pass (2048-bit key;\n unprotected) header.d=kernel.org header.i=@kernel.org header.a=rsa-sha256\n header.s=k20201202 header.b=IN0G++eL;\n\tdkim-atps=neutral","legolas.ozlabs.org;\n spf=pass (sender SPF authorized) smtp.mailfrom=vger.kernel.org\n (client-ip=2600:3c04:e001:36c::12fc:5321; helo=tor.lore.kernel.org;\n envelope-from=linux-pci+bounces-52624-incoming=patchwork.ozlabs.org@vger.kernel.org;\n receiver=patchwork.ozlabs.org)","smtp.subspace.kernel.org;\n\tdkim=pass (2048-bit key) header.d=kernel.org header.i=@kernel.org\n header.b=\"IN0G++eL\"","smtp.subspace.kernel.org;\n arc=none smtp.client-ip=10.30.226.201"],"Received":["from tor.lore.kernel.org (tor.lore.kernel.org\n [IPv6:2600:3c04:e001:36c::12fc:5321])\n\t(using TLSv1.3 with cipher TLS_AES_256_GCM_SHA384 (256/256 bits)\n\t key-exchange x25519 server-signature ECDSA (secp384r1) server-digest SHA384)\n\t(No client certificate requested)\n\tby legolas.ozlabs.org (Postfix) with ESMTPS id 4fxQms1KCRz1yD3\n\tfor ; Fri, 17 Apr 2026 04:01:17 +1000 (AEST)","from smtp.subspace.kernel.org (conduit.subspace.kernel.org\n [100.90.174.1])\n\tby tor.lore.kernel.org (Postfix) with ESMTP id ACAC930082AD\n\tfor ; Thu, 16 Apr 2026 18:01:13 +0000 (UTC)","from localhost.localdomain (localhost.localdomain [127.0.0.1])\n\tby smtp.subspace.kernel.org (Postfix) with ESMTP id BB13533262A;\n\tThu, 16 Apr 2026 18:01:12 +0000 (UTC)","from smtp.kernel.org (aws-us-west-2-korg-mail-1.web.codeaurora.org\n [10.30.226.201])\n\t(using TLSv1.2 with cipher ECDHE-RSA-AES256-GCM-SHA384 (256/256 bits))\n\t(No client certificate requested)\n\tby smtp.subspace.kernel.org (Postfix) with ESMTPS id 8DF78331A4C;\n\tThu, 16 Apr 2026 18:01:12 +0000 (UTC)","by smtp.kernel.org (Postfix) with ESMTPSA id CB3DBC2BCAF;\n\tThu, 16 Apr 2026 18:01:08 +0000 (UTC)"],"ARC-Seal":"i=1; a=rsa-sha256; d=subspace.kernel.org; s=arc-20240116;\n\tt=1776362472; cv=none;\n b=RYo648TNdh896cpQ5yzG6+Vb/q7hvXRWtq60nn4FsyuRjXs/eSf1ET4oIWLdnkxhhvtoj7zgV3fwE6e9rXDh/ZLQR1tLAfLOgsc3m6q41gJddmMbLn2n6tf8qQTIIzZb/OgibqFCcyOdplIoaJQ/HmDhbWDHrB/zV8eVVix7iBo=","ARC-Message-Signature":"i=1; a=rsa-sha256; d=subspace.kernel.org;\n\ts=arc-20240116; t=1776362472; c=relaxed/simple;\n\tbh=0gwbU28QnN9fAESh13GoaD14hEGTKIIjtJA8T2NO+Uw=;\n\th=From:To:Cc:Subject:Date:Message-ID:MIME-Version:Content-Type;\n b=OoUzx23SWvpm95USa6e+IaeFJswryzWNjV6gYLFHHpNg67lHhWzwnWYdwfAyAPm0zQiNttM3vFsqi3XdifXZVtjbhx1UFvssNjDOS/873d2Cl/gnhPCTlWJIKLQgq0NujzjybQCGcqqjGDYvte4TDmL0Qb/y1XusvUXhWsPeV5A=","ARC-Authentication-Results":"i=1; smtp.subspace.kernel.org;\n dkim=pass (2048-bit key) header.d=kernel.org header.i=@kernel.org\n header.b=IN0G++eL; arc=none smtp.client-ip=10.30.226.201","DKIM-Signature":"v=1; a=rsa-sha256; c=relaxed/simple; d=kernel.org;\n\ts=k20201202; t=1776362472;\n\tbh=0gwbU28QnN9fAESh13GoaD14hEGTKIIjtJA8T2NO+Uw=;\n\th=From:To:Cc:Subject:Date:From;\n\tb=IN0G++eLHaqph6Kdkoprfm9BaOBsVO4IXrABjvmTLcu7EshpBE/81pDBZ1HDt2jD5\n\t kjGtnIIxLMB5mTbkeGBWsq2AxR+WXfNF97N1MBLknGCxunaV5Y18N3Gbdk8Xty737n\n\t xjQupnIl2esXrUf5abrzQl+0xMrE/YARvWE7UNG2eR+/W9tH1zylHYN1E9tMUmbT0/\n\t D9ytm5XHFENFMuvyhtSvWb573/suSQlOMRCvxGUjRqVloXBnuvB8OZEdOkkJokFeD6\n\t LO9L044tbnp9YkY899nxjM/p551NJ2GYPYkPB181hDqk6MmZpwILoMUqy78rciGYuj\n\t /fzjVlIdWjxag==","From":"=?utf-8?q?Krzysztof_Wilczy=C5=84ski?= ","To":"Bjorn Helgaas ","Cc":"Bjorn Helgaas ,\n Manivannan Sadhasivam ,\n Lorenzo Pieralisi ,\n Magnus Lindholm , Matt Turner ,\n Richard Henderson ,\n Christophe Leroy ,\n Madhavan Srinivasan ,\n Michael Ellerman , Nicholas Piggin ,\n Dexuan Cui ,\n =?utf-8?q?Krzysztof_Ha=C5=82asa?= ,\n Lukas Wunner , \"Oliver O'Halloran\" ,\n Saurabh Singh Sengar ,\n Shuan He , Srivatsa Bhat ,\n\t=?utf-8?q?Ilpo_J=C3=A4rvinen?= ,\n linux-pci@vger.kernel.org, linux-alpha@vger.kernel.org,\n linuxppc-dev@lists.ozlabs.org","Subject":"[PATCH v5 00/23] PCI: Convert all dynamic sysfs attributes to static","Date":"Thu, 16 Apr 2026 18:00:44 +0000","Message-ID":"<20260416180107.777065-1-kwilczynski@kernel.org>","X-Mailer":"git-send-email 2.53.0","Precedence":"bulk","X-Mailing-List":"linux-pci@vger.kernel.org","List-Id":"","List-Subscribe":"","List-Unsubscribe":"","MIME-Version":"1.0","Content-Type":"text/plain; charset=UTF-8","Content-Transfer-Encoding":"8bit"},"content":"Hello,\n\nThis series converts every dynamically allocated PCI sysfs attribute to\na static const definition. After the full series, pci_sysfs_init() and\nsysfs_initialized are gone, and every sysfs file is created by the\ndriver model at device_add() time.\n\nCurrently, the PCI resource files (resourceN, resourceN_wc) and the\nlegacy bus files (legacy_io, legacy_mem) are created dynamically\nfrom two unsynchronised paths:\n\nPath A: late_initcall\n\n pci_sysfs_init (late_initcall)\n sysfs_initialized = 1\n for_each_pci_dev\n pci_create_sysfs_dev_files\n sysfs_create_bin_file (resourceN, resourceN_wc)\n pci_find_next_bus\n pci_create_legacy_files\n sysfs_create_bin_file (legacy_io, legacy_mem)\n\nPath B: device registration / hotplug\n\n pci_bus_add_devices\n pci_bus_add_device\n pci_create_sysfs_dev_files\n if (!sysfs_initialized) return <- only guard\n sysfs_create_bin_file (resourceN, resourceN_wc)\n\nOn most ACPI systems this does not race because PCI enumeration\ncompletes at subsys_initcall time, before pci_sysfs_init() runs:\n\n subsys_initcall (level 4):\n acpi_pci_root_add\n pci_bus_add_device\n pci_create_sysfs_dev_files\n if (!sysfs_initialized) <- not yet set\n return -EACCES\n\n late_initcall (level 7):\n pci_sysfs_init\n sysfs_initialized = 1\n for_each_pci_dev\n pci_create_sysfs_dev_files <- creates the files, no race\n\nOn Devicetree platforms the host controller is a platform driver that\nprobes via the driver model, often on a workqueue, and overlaps with the\nlate_initcall:\n\n CPU 0 (late_initcall) CPU 1 (driver probe)\n --------------------------- ----------------------------\n pci_sysfs_init()\n sysfs_initialized = 1\n for_each_pci_dev(pdev) pci_bus_add_device(pdev)\n pci_create_sysfs_dev_files() pci_create_sysfs_dev_files()\n sysfs_create_bin_file() sysfs_create_bin_file()\n -> \"duplicate filename\"\n\nThe same happens on ACPI when probing is asynchronous (hv_pci on\nAzure, RISC-V with ACPI).\n\nThe duplicate causes sysfs_create_bin_file() to fail with -EEXIST.\npci_create_resource_files() then calls pci_remove_resource_files() in\nits error unwind, tearing down files the other thread created and\nstill references through pdev->res_attr[]. This has caused kernel\npanics on i.MX6 and boot failures on other platforms.\n\nSeveral different fixes have been proposed over the years: reordering\nthe sysfs_initialized assignment, adding locks, checking\npci_dev_is_added(), setting pdev->res_attr[] to NULL after kfree\n(which only prevents a double-free on the teardown path, not the\nerror unwind removing the other thread's files). None would address the\nroot cause.\n\nThis has been reported a few times:\n\n - https://lore.kernel.org/linux-pci/20250702155112.40124-1-heshuan@bytedance.com/\n - https://lore.kernel.org/linux-pci/b51519d6-ce45-4b6d-8135-c70169bd110e@h-partners.com/\n - https://lore.kernel.org/linux-pci/1702093576-30405-1-git-send-email-ssengar@linux.microsoft.com/\n - https://lore.kernel.org/linux-pci/SY0P300MB04687548090B73E40AF97D8897B82@SY0P300MB0468.AUSP300.PROD.OUTLOOK.COM/\n - https://lore.kernel.org/linux-pci/20230105174736.GA1154719@bhelgaas/\n - https://lore.kernel.org/linux-pci/m3eebg9puj.fsf@t19.piap.pl/\n - https://lore.kernel.org/linux-pci/20200716110423.xtfyb3n6tn5ixedh@pali/\n - https://lore.kernel.org/linux-pci/1366196798-15929-1-git-send-email-artem.savkov@gmail.com/\n - https://bugzilla.kernel.org/show_bug.cgi?id=215515\n - https://bugzilla.kernel.org/show_bug.cgi?id=216888\n\nWith static attributes the driver model creates sysfs entries once per\ndevice at device_add() time, under the device lock, eliminating the\nlate_initcall iteration and the race along with it.\n\n\tKrzysztof\n\n---\nChanges in v5:\n https://lore.kernel.org/linux-pci/20260411080148.471335-1-kwilczynski@kernel.org/\n \n - Added new Tested-by, Reviewed-by, and Acked-by tags.\n - Used the existing _io function names in the static macro\n definitions, deferring the rename to the conversion commit\n where it belongs, to avoid a forward reference across\n commits. This was reported by Sashiko, see:\n https://sashiko.dev/#/patchset/20260411080148.471335-1-kwilczynski%40kernel.org?part=6\n - Folded the __resource_resize_store() conversion into the\n main static attributes commit so the resize path is never\n broken between commits. This was reported by Sashiko, see:\n https://sashiko.dev/#/patchset/20260410055040.39233-1-kwilczynski%40kernel.org?part=6\n https://sashiko.dev/#/patchset/20260411080148.471335-1-kwilczynski%40kernel.org?part=7\n - Dropped the unnecessary parentheses cleanup from the Alpha\n BAR index commit, as the line is replaced two commits later\n anyway, as per Ilpo Järvinen's feedback.\n - Squashed the Alpha accessor macro and cleanup commits into\n one, using pci_resource_is_mem() directly instead of the\n intermediate pci_resource_flags() step, as per Ilpo\n Järvinen's feedback.\n - Moved the raw literal conversion in pci_create_legacy_files()\n into the macro definition commit, so the macros and their\n usage are introduced together, as per Ilpo Järvinen's\n feedback.\n - Removed unnecessary backslash line continuation from the\n ternary in pci_mmap_legacy_page_range().\n - Kept pci_resource_len() for visibility checks instead of\n resource_assigned(). The static is_visible() callback\n runs at device_add() time during the PCI enumeration,\n before the pci_assign_unassigned_bus_resources() populates\n res->parent, as such, resource_assigned() returned false\n for every BAR, hiding all resource files. This is related\n to review feedback from Ilpo Järvinen.\n\nChanges in v4:\n https://lore.kernel.org/linux-pci/20260410055040.39233-1-kwilczynski@kernel.org/\n\n - Added new Reviewed-by tags.\n - Added pci_resource_is_io() and pci_resource_is_mem() helpers\n for resource type checks, replacing the open-coded bitwise\n flag tests in pci_mmap_resource(), pci_resource_io(), and\n Alpha's pci_mmap_resource(), as per Ilpo Järvinen's\n suggestion.\n - Split the __pci_mmap_fits() cleanup into two patches. An\n overflow fix for zero-length BARs, which now includes a\n Fixes: tag referencing the original Alpha PCI sysfs commit,\n and the WARN macro removal is a separate cleanup as per Ilpo\n Järvinen's suggestion.\n - Added a missing Fixes: tag to the Alpha lockdown check,\n referencing the commit that added the check to the generic\n path but missed Alpha's implementation.\n - Added PCI_LEGACY_IO_SIZE and PCI_LEGACY_MEM_SIZE macros to\n replace the raw literals used for legacy address space sizes.\n These are used in both Alpha's pci_mmap_legacy_page_range()\n and the static legacy attribute definitions, as per Ilpo\n Järvinen's suggestion.\n - Replaced sysfs_update_groups() in the BAR resize path with\n sysfs_remove_groups() before the resize and sysfs_create_groups()\n after, restoring the original teardown before BAR resize\n ordering. This was reported by Sashiko, see:\n https://sashiko.dev/#/patchset/20260410055040.39233-1-kwilczynski%40kernel.org?part=7\n - Defined pci_dev_resource_attr_groups as a NULL macro when\n HAVE_PCI_MMAP and ARCH_GENERIC_PCI_MMAP_RESOURCE are both\n absent, so the resize path compiles unconditionally without\n #ifdef guards in the function body. This was reported by\n Sashiko, see:\n https://sashiko.dev/#/patchset/20260410055040.39233-1-kwilczynski%40kernel.org?part=7\n - Moved the pci_legacy_has_sparse() prototype into the patch\n that introduces the function, alongside the existing\n pci_adjust_legacy_attr() declaration, to fix a bisection\n issue where Alpha would warn on -Wmissing-prototypes.\n This was reported by Sashiko, see:\n https://sashiko.dev/#/patchset/20260410055040.39233-1-kwilczynski%40kernel.org?part=18\n\nChanges in v3:\n https://lore.kernel.org/linux-pci/20210910202623.2293708-1-kw@linux.com/\n\n - Updated for modern kernel releases and expanded scope. The\n v2 only covered the generic resource files. This version\n also converts Alpha's sparse/dense resource files and the\n legacy bus attributes, removing pci_sysfs_init() entirely.\n - Split the single macro definition into three distinct ones\n (per I/O, UC, and WC), to make sure that each carries only\n the callbacks its resource type needs.\n - Updated to use the new .bin_size callback, as the attributes\n are const, to replace using a->size directly, which was not\n ideal. This required changes to pci_llseek_resource(), to\n ensure that it would work for device and bus-level attributes.\n - Updated the __resource_resize_store() to include CAP_SYS_ADMIN\n capabilities check.\n - Added the security_locked_down() check to Alpha's\n pci_mmap_resource(), to align with other architectures.\n\nChanges in v2:\n https://lore.kernel.org/linux-pci/20210825212255.878043-1-kw@linux.com/\n\n - Refactored code so that the macros, helpers and internal\n functions can be used to correctly leverage the read(),\n write() and mmap() callbacks rather than to use the\n .is_bin_visible() callback to set up sysfs objects\n internals as this is not supported.\n - Refactored some if-statements to check for a resource\n flag first, and then call either arch_can_pci_mmap_io()\n or arch_can_pci_mmap_wc(), plus store result of testing\n for IORESOURCE_MEM and IORESOURCE_PREFETCH flags into\n a boolean variable, as per Bjorn Helgaas' suggestion.\n - Renamed pci_read_resource_io() and pci_write_resource_io()\n callbacks so that these are not specifically tied to I/O\n BARs read() and write() operations also as per Bjorn\n Helgaas' suggestion.\n - Updated style for code handling bitwise operations to\n match the style that is preferred as per Bjorn Helgaas'\n suggestion.\n - Updated commit messages adding more details about the\n implementation as requested by Bjorn Helgaas.\n\nKrzysztof Wilczyński (23):\n PCI/sysfs: Use PCI resource accessor macros\n PCI: Add pci_resource_is_io() and pci_resource_is_mem() helpers\n PCI/sysfs: Only allow supported resource types in I/O and MMIO helpers\n PCI/sysfs: Use BAR length in pci_llseek_resource() when attr->size is\n zero\n PCI/sysfs: Add CAP_SYS_ADMIN check to __resource_resize_store()\n PCI/sysfs: Add static PCI resource attribute macros\n PCI/sysfs: Convert PCI resource files to static attributes\n PCI/sysfs: Warn about BAR resize failure in __resource_resize_store()\n PCI/sysfs: Add stubs for pci_{create,remove}_sysfs_dev_files()\n PCI/sysfs: Limit pci_sysfs_init() late_initcall compile scope\n alpha/PCI: Add security_locked_down() check to pci_mmap_resource()\n alpha/PCI: Use BAR index in sysfs attr->private instead of resource\n pointer\n alpha/PCI: Use PCI resource accessor macros\n alpha/PCI: Fix __pci_mmap_fits() overflow for zero-length BARs\n alpha/PCI: Remove WARN from __pci_mmap_fits()\n alpha/PCI: Add static PCI resource attribute macros\n alpha/PCI: Convert resource files to static attributes\n PCI/sysfs: Remove pci_{create,remove}_sysfs_dev_files()\n PCI: Add macros for legacy I/O and memory address space sizes\n alpha/PCI: Compute legacy size in pci_mmap_legacy_page_range()\n PCI/sysfs: Add __weak pci_legacy_has_sparse() helper\n PCI/sysfs: Convert legacy I/O and memory attributes to static\n definitions\n PCI/sysfs: Remove pci_create_legacy_files() and pci_sysfs_init()\n\n arch/alpha/include/asm/pci.h | 13 +-\n arch/alpha/kernel/pci-sysfs.c | 373 +++++++++++----------\n arch/powerpc/include/asm/pci.h | 2 -\n drivers/pci/bus.c | 1 -\n drivers/pci/pci-sysfs.c | 575 +++++++++++++++++++--------------\n drivers/pci/pci.h | 16 +-\n drivers/pci/probe.c | 6 -\n drivers/pci/remove.c | 3 -\n include/linux/pci.h | 39 ++-\n 9 files changed, 578 insertions(+), 450 deletions(-)"}