kernel/linux.git/drivers/base/core.c, branch v6.1.175

driver core: Don't let a device probe until it's ready

2026-06-01T15:38:45+00:00

[ Upstream commit a2225b6e834a838ae3c93709760edc0a169eb2f2 ] The moment we link a "struct device" into the list of devices for the bus, it's possible probe can happen. This is because another thread can load the driver at any time and that can cause the device to probe. This has been seen in practice with a stack crawl that looks like this [1]: really_probe() __driver_probe_device() driver_probe_device() __driver_attach() bus_for_each_dev() driver_attach() bus_add_driver() driver_register() __platform_driver_register() init_module() [some module] do_one_initcall() do_init_module() load_module() __arm64_sys_finit_module() invoke_syscall() As a result of the above, it was seen that device_links_driver_bound() could be called for the device before "dev->fwnode->dev" was assigned. This prevented __fw_devlink_pickup_dangling_consumers() from being called which meant that other devices waiting on our driver's sub-nodes were stuck deferring forever. It's believed that this problem is showing up suddenly for two reasons: 1. Android has recently (last ~1 year) implemented an optimization to the order it loads modules [2]. When devices opt-in to this faster loading, modules are loaded one-after-the-other very quickly. This is unlike how other distributions do it. The reproduction of this problem has only been seen on devices that opt-in to Android's "parallel module loading". 2. Android devices typically opt-in to fw_devlink, and the most noticeable issue is the NULL "dev->fwnode->dev" in device_links_driver_bound(). fw_devlink is somewhat new code and also not in use by all Linux devices. Even though the specific symptom where "dev->fwnode->dev" wasn't assigned could be fixed by moving that assignment higher in device_add(), other parts of device_add() (like the call to device_pm_add()) are also important to run before probe. Only moving the "dev->fwnode->dev" assignment would likely fix the current symptoms but lead to difficult-to-debug problems in the future. Fix the problem by preventing probe until device_add() has run far enough that the device is ready to probe. If somehow we end up trying to probe before we're allowed, __driver_probe_device() will return -EPROBE_DEFER which will make certain the device is noticed. In the race condition that was seen with Android's faster module loading, we will temporarily add the device to the deferred list and then take it off immediately when device_add() probes the device. Instead of adding another flag to the bitfields already in "struct device", instead add a new "flags" field and use that. This allows us to freely change the bit from different thread without worrying about corrupting nearby bits (and means threads changing other bit won't corrupt us). [1] Captured on a machine running a downstream 6.6 kernel [2] https://cs.android.com/android/platform/superproject/main/+/main:system/core/libmodprobe/libmodprobe.cpp?q=LoadModulesParallel Cc: stable@vger.kernel.org Fixes: 2023c610dc54 ("Driver core: add new device to bus's list before probing") Reviewed-by: Alan Stern Reviewed-by: Rafael J. Wysocki (Intel) Reviewed-by: Danilo Krummrich Acked-by: Greg Kroah-Hartman Acked-by: Marek Szyprowski Signed-off-by: Douglas Anderson Link: https://patch.msgid.link/20260406162231.v5.1.Id750b0fbcc94f23ed04b7aecabcead688d0d8c17@changeid Signed-off-by: Danilo Krummrich Signed-off-by: Douglas Anderson Signed-off-by: Sasha Levin

device property: Make modifications of fwnode "flags" thread safe

2026-06-01T15:38:45+00:00

commit f72e77c33e4b5657af35125e75bab249256030f3 upstream. In various places in the kernel, we modify the fwnode "flags" member by doing either: fwnode->flags |= SOME_FLAG; fwnode->flags &= ~SOME_FLAG; This type of modification is not thread-safe. If two threads are both mucking with the flags at the same time then one can clobber the other. While flags are often modified while under the "fwnode_link_lock", this is not universally true. Create some accessor functions for setting, clearing, and testing the FWNODE flags and move all users to these accessor functions. New accessor functions use set_bit() and clear_bit(), which are thread-safe. Cc: stable@vger.kernel.org Fixes: c2c724c868c4 ("driver core: Add fw_devlink_parse_fwtree()") Reviewed-by: Andy Shevchenko Acked-by: Mark Brown Reviewed-by: Wolfram Sang Signed-off-by: Douglas Anderson Reviewed-by: Rafael J. Wysocki (Intel) Reviewed-by: Saravana Kannan Link: https://patch.msgid.link/20260317090112.v2.1.I0a4d03104ecd5103df3d76f66c8d21b1d15a2e38@changeid [ Fix fwnode_clear_flag() argument alignment, restore dropped blank line in fwnode_dev_initialized(), and remove unnecessary parentheses around fwnode_test_flag() calls. - Danilo ] Signed-off-by: Danilo Krummrich Signed-off-by: Douglas Anderson Signed-off-by: Greg Kroah-Hartman

drivers: core: fix device leak in __fw_devlink_relax_cycles()

2025-03-13T11:53:22+00:00

commit 78eb41f518f414378643ab022241df2a9dcd008b upstream. Commit bac3b10b78e5 ("driver core: fw_devlink: Stop trying to optimize cycle detection logic") introduced a new struct device *con_dev and a get_dev_from_fwnode() call to get it, but without adding a corresponding put_device(). Closes: https://lore.kernel.org/all/20241204124826.2e055091@booty/ Fixes: bac3b10b78e5 ("driver core: fw_devlink: Stop trying to optimize cycle detection logic") Cc: stable@vger.kernel.org Reviewed-by: Saravana Kannan Signed-off-by: Luca Ceresoli Link: https://lore.kernel.org/r/20250213-fix__fw_devlink_relax_cycles_missing_device_put-v2-1-8cd3b03e6a3f@bootlin.com Signed-off-by: Greg Kroah-Hartman

driver core: fw_devlink: Stop trying to optimize cycle detection logic

2024-12-14T18:54:26+00:00

[ Upstream commit bac3b10b78e54b7da3cede397258f75a2180609b ] In attempting to optimize fw_devlink runtime, I introduced numerous cycle detection bugs by foregoing cycle detection logic under specific conditions. Each fix has further narrowed the conditions for optimization. It's time to give up on these optimization attempts and just run the cycle detection logic every time fw_devlink tries to create a device link. The specific bug report that triggered this fix involved a supplier fwnode that never gets a device created for it. Instead, the supplier fwnode is represented by the device that corresponds to an ancestor fwnode. In this case, fw_devlink didn't do any cycle detection because the cycle detection logic is only run when a device link is created between the devices that correspond to the actual consumer and supplier fwnodes. With this change, fw_devlink will run cycle detection logic even when creating SYNC_STATE_ONLY proxy device links from a device that is an ancestor of a consumer fwnode. Reported-by: Tomi Valkeinen Closes: https://lore.kernel.org/all/1a1ab663-d068-40fb-8c94-f0715403d276@ideasonboard.com/ Fixes: 6442d79d880c ("driver core: fw_devlink: Improve detection of overlapping cycles") Cc: stable Tested-by: Tomi Valkeinen Signed-off-by: Saravana Kannan Link: https://lore.kernel.org/r/20241030171009.1853340-1-saravanak@google.com Signed-off-by: Greg Kroah-Hartman Signed-off-by: Sasha Levin

driver core: Add FWLINK_FLAG_IGNORE to completely ignore a fwnode link

2024-12-14T18:54:26+00:00

[ Upstream commit b7e1241d8f77ed64404a5e4450f43a319310fc91 ] A fwnode link between specific supplier-consumer fwnodes can be added multiple times for multiple reasons. If that dependency doesn't exist, deleting the fwnode link once doesn't guarantee that it won't get created again. So, add FWLINK_FLAG_IGNORE flag to mark a fwnode link as one that needs to be completely ignored. Since a fwnode link's flags is an OR of all the flags passed to all the fwnode_link_add() calls to create that specific fwnode link, the FWLINK_FLAG_IGNORE flag is preserved and can be used to mark a fwnode link as on that need to be completely ignored until it is deleted. Signed-off-by: Saravana Kannan Acked-by: "Rafael J. Wysocki" Reviewed-by: Rob Herring Link: https://lore.kernel.org/r/20240305050458.1400667-3-saravanak@google.com Signed-off-by: Greg Kroah-Hartman Stable-dep-of: bac3b10b78e5 ("driver core: fw_devlink: Stop trying to optimize cycle detection logic") Signed-off-by: Sasha Levin

driver core: fw_devlink: Improve logs for cycle detection

2024-12-14T18:54:26+00:00

[ Upstream commit 6e7ad1aebb4fc9fed0217dd50ef6e58a53f17d81 ] The links in a cycle are not all logged in a consistent manner or not logged at all. Make them consistent by adding a "cycle:" string and log all the link in the cycles (even the child ==> parent dependency) so that it's easier to debug cycle detection code. Also, mark the start and end of a cycle so it's easy to tell when multiple cycles are logged back to back. Signed-off-by: Saravana Kannan Tested-by: Xu Yang Link: https://lore.kernel.org/r/20240202095636.868578-4-saravanak@google.com Signed-off-by: Greg Kroah-Hartman Stable-dep-of: bac3b10b78e5 ("driver core: fw_devlink: Stop trying to optimize cycle detection logic") Signed-off-by: Sasha Levin

Revert "driver core: Fix uevent_show() vs driver detach race"

2024-11-08T15:26:44+00:00

commit 9a71892cbcdb9d1459c84f5a4c722b14354158a5 upstream. This reverts commit 15fffc6a5624b13b428bb1c6e9088e32a55eb82c. This commit causes a regression, so revert it for now until it can come back in a way that works for everyone. Link: https://lore.kernel.org/all/172790598832.1168608.4519484276671503678.stgit@dwillia2-xfh.jf.intel.com/ Fixes: 15fffc6a5624 ("driver core: Fix uevent_show() vs driver detach race") Cc: stable Cc: Ashish Sangwan Cc: Namjae Jeon Cc: Dirk Behme Cc: Greg Kroah-Hartman Cc: Rafael J. Wysocki Cc: Dan Williams Signed-off-by: Greg Kroah-Hartman

driver core: Fix uevent_show() vs driver detach race

2024-08-14T11:52:57+00:00

commit 15fffc6a5624b13b428bb1c6e9088e32a55eb82c upstream. uevent_show() wants to de-reference dev->driver->name. There is no clean way for a device attribute to de-reference dev->driver unless that attribute is defined via (struct device_driver).dev_groups. Instead, the anti-pattern of taking the device_lock() in the attribute handler risks deadlocks with code paths that remove device attributes while holding the lock. This deadlock is typically invisible to lockdep given the device_lock() is marked lockdep_set_novalidate_class(), but some subsystems allocate a local lockdep key for @dev->mutex to reveal reports of the form: ====================================================== WARNING: possible circular locking dependency detected 6.10.0-rc7+ #275 Tainted: G OE N ------------------------------------------------------ modprobe/2374 is trying to acquire lock: ffff8c2270070de0 (kn->active#6){++++}-{0:0}, at: __kernfs_remove+0xde/0x220 but task is already holding lock: ffff8c22016e88f8 (&cxl_root_key){+.+.}-{3:3}, at: device_release_driver_internal+0x39/0x210 which lock already depends on the new lock. the existing dependency chain (in reverse order) is: -> #1 (&cxl_root_key){+.+.}-{3:3}: __mutex_lock+0x99/0xc30 uevent_show+0xac/0x130 dev_attr_show+0x18/0x40 sysfs_kf_seq_show+0xac/0xf0 seq_read_iter+0x110/0x450 vfs_read+0x25b/0x340 ksys_read+0x67/0xf0 do_syscall_64+0x75/0x190 entry_SYSCALL_64_after_hwframe+0x76/0x7e -> #0 (kn->active#6){++++}-{0:0}: __lock_acquire+0x121a/0x1fa0 lock_acquire+0xd6/0x2e0 kernfs_drain+0x1e9/0x200 __kernfs_remove+0xde/0x220 kernfs_remove_by_name_ns+0x5e/0xa0 device_del+0x168/0x410 device_unregister+0x13/0x60 devres_release_all+0xb8/0x110 device_unbind_cleanup+0xe/0x70 device_release_driver_internal+0x1c7/0x210 driver_detach+0x47/0x90 bus_remove_driver+0x6c/0xf0 cxl_acpi_exit+0xc/0x11 [cxl_acpi] __do_sys_delete_module.isra.0+0x181/0x260 do_syscall_64+0x75/0x190 entry_SYSCALL_64_after_hwframe+0x76/0x7e The observation though is that driver objects are typically much longer lived than device objects. It is reasonable to perform lockless de-reference of a @driver pointer even if it is racing detach from a device. Given the infrequency of driver unregistration, use synchronize_rcu() in module_remove_driver() to close any potential races. It is potentially overkill to suffer synchronize_rcu() just to handle the rare module removal racing uevent_show() event. Thanks to Tetsuo Handa for the debug analysis of the syzbot report [1]. Fixes: c0a40097f0bc ("drivers: core: synchronize really_probe() and dev_uevent()") Reported-by: syzbot+4762dd74e32532cda5ff@syzkaller.appspotmail.com Reported-by: Tetsuo Handa Closes: http://lore.kernel.org/5aa5558f-90a4-4864-b1b1-5d6784c5607d@I-love.SAKURA.ne.jp [1] Link: http://lore.kernel.org/669073b8ea479_5fffa294c1@dwillia2-xfh.jf.intel.com.notmuch Cc: stable@vger.kernel.org Cc: Ashish Sangwan Cc: Namjae Jeon Cc: Dirk Behme Cc: Greg Kroah-Hartman Cc: Rafael J. Wysocki Signed-off-by: Dan Williams Link: https://lore.kernel.org/r/172081332794.577428.9738802016494057132.stgit@dwillia2-xfh.jf.intel.com Signed-off-by: Greg Kroah-Hartman

drivers: core: synchronize really_probe() and dev_uevent()

2024-06-21T12:35:54+00:00

commit c0a40097f0bc81deafc15f9195d1fb54595cd6d0 upstream. Synchronize the dev->driver usage in really_probe() and dev_uevent(). These can run in different threads, what can result in the following race condition for dev->driver uninitialization: Thread #1: ========== really_probe() { ... probe_failed: ... device_unbind_cleanup(dev) { ... dev->driver = NULL; // <= Failed probe sets dev->driver to NULL ... } ... } Thread #2: ========== dev_uevent() { ... if (dev->driver) // If dev->driver is NULLed from really_probe() from here on, // after above check, the system crashes add_uevent_var(env, "DRIVER=%s", dev->driver->name); ... } really_probe() holds the lock, already. So nothing needs to be done there. dev_uevent() is called with lock held, often, too. But not always. What implies that we can't add any locking in dev_uevent() itself. So fix this race by adding the lock to the non-protected path. This is the path where above race is observed: dev_uevent+0x235/0x380 uevent_show+0x10c/0x1f0 <= Add lock here dev_attr_show+0x3a/0xa0 sysfs_kf_seq_show+0x17c/0x250 kernfs_seq_show+0x7c/0x90 seq_read_iter+0x2d7/0x940 kernfs_fop_read_iter+0xc6/0x310 vfs_read+0x5bc/0x6b0 ksys_read+0xeb/0x1b0 __x64_sys_read+0x42/0x50 x64_sys_call+0x27ad/0x2d30 do_syscall_64+0xcd/0x1d0 entry_SYSCALL_64_after_hwframe+0x77/0x7f Similar cases are reported by syzkaller in https://syzkaller.appspot.com/bug?extid=ffa8143439596313a85a But these are regarding the *initialization* of dev->driver dev->driver = drv; As this switches dev->driver to non-NULL these reports can be considered to be false-positives (which should be "fixed" by this commit, as well, though). The same issue was reported and tried to be fixed back in 2015 in https://lore.kernel.org/lkml/1421259054-2574-1-git-send-email-a.sangwan@samsung.com/ already. Fixes: 239378f16aa1 ("Driver core: add uevent vars for devices of a class") Cc: stable Cc: syzbot+ffa8143439596313a85a@syzkaller.appspotmail.com Cc: Ashish Sangwan Cc: Namjae Jeon Signed-off-by: Dirk Behme Link: https://lore.kernel.org/r/20240513050634.3964461-1-dirk.behme@de.bosch.com Signed-off-by: Greg Kroah-Hartman

driver core: Introduce device_link_wait_removal()

2024-04-10T14:28:32+00:00

commit 0462c56c290a99a7f03e817ae5b843116dfb575c upstream. The commit 80dd33cf72d1 ("drivers: base: Fix device link removal") introduces a workqueue to release the consumer and supplier devices used in the devlink. In the job queued, devices are release and in turn, when all the references to these devices are dropped, the release function of the device itself is called. Nothing is present to provide some synchronisation with this workqueue in order to ensure that all ongoing releasing operations are done and so, some other operations can be started safely. For instance, in the following sequence: 1) of_platform_depopulate() 2) of_overlay_remove() During the step 1, devices are released and related devlinks are removed (jobs pushed in the workqueue). During the step 2, OF nodes are destroyed but, without any synchronisation with devlink removal jobs, of_overlay_remove() can raise warnings related to missing of_node_put(): ERROR: memory leak, expected refcount 1 instead of 2 Indeed, the missing of_node_put() call is going to be done, too late, from the workqueue job execution. Introduce device_link_wait_removal() to offer a way to synchronize operations waiting for the end of devlink removals (i.e. end of workqueue jobs). Also, as a flushing operation is done on the workqueue, the workqueue used is moved from a system-wide workqueue to a local one. Cc: stable@vger.kernel.org Signed-off-by: Herve Codina Tested-by: Luca Ceresoli Reviewed-by: Nuno Sa Reviewed-by: Saravana Kannan Acked-by: Greg Kroah-Hartman Link: https://lore.kernel.org/r/20240325152140.198219-2-herve.codina@bootlin.com Signed-off-by: Rob Herring Signed-off-by: Greg Kroah-Hartman