kernel/drivers/misc/mei/main.c, branch linux-rolling-stable

mei: Fix error handling in mei_register

2026-01-02T11:57:15Z

commit a6dab2f61d23c1eb32f1d08fa7b4919a2478950b upstream. mei_register() fails to release the device reference in error paths after device_initialize(). During normal device registration, the reference is properly handled through mei_deregister() which calls device_destroy(). However, in error handling paths (such as cdev_alloc failure, cdev_add failure, etc.), missing put_device() calls cause reference count leaks, preventing the device's release function (mei_device_release) from being called and resulting in memory leaks of mei_device. Found by code review. Cc: stable Fixes: 7704e6be4ed2 ("mei: hook mei_device on class device") Signed-off-by: Ma Ke Acked-by: Alexander Usyskin Link: https://patch.msgid.link/20251104020133.5017-1-make24@iscas.ac.cn Signed-off-by: Greg Kroah-Hartman

Merge patch series "mei: connect to card in D3cold"

2025-09-18T16:29:34Z

Alexander Usyskin says: When discrete graphic card enters D3cold th CSC engine is powered down. On wakeup from the D3cold full HECI link reset is required. The driver should detect that firmware requests link reset and initiate the link reset flow. In the usual flow the connect IOCTL will trigger the wake from D3cold and corresponding link reset. The MEI driver invalidates all open handles on link reset including the one that triggered the wake rendering this connection unusable. To break this loop make connect detect that it is interrupted by link reset and retry connect attempt after reset was completed. Link: https://lore.kernel.org/r/20250918130435.3327400-1-alexander.usyskin@intel.com Signed-off-by: Greg Kroah-Hartman

mei: retry connect if interrupted by link reset

2025-09-18T16:29:33Z

When device is in D3cold the connect message will wake device and cause link reset. Link reset flow cleans all queues and wakes all waiters. Retry the connect flow if connect is failed and link reset is detected. Signed-off-by: Alexander Usyskin Signed-off-by: Greg Kroah-Hartman Link: https://lore.kernel.org/r/20250918130435.3327400-4-alexander.usyskin@intel.com

mei: make a local copy of client uuid in connect

2025-09-18T16:29:33Z

Connect ioctl has the same memory for in and out parameters. Copy in parameter (client uuid) to the local stack to avoid it be overwritten by out parameters fill. Signed-off-by: Alexander Usyskin Signed-off-by: Greg Kroah-Hartman Link: https://lore.kernel.org/r/20250918130435.3327400-3-alexander.usyskin@intel.com

mei: gsc: fix remove operations order

2025-09-18T16:27:40Z

The mei disconnect should be the last operation in remove flow. Otherwise the device is used after destruction. Fix minor free flow that happens after device destruction too. The fault leads to the following oops in Intel Gfx CI: <4>[ 267.871331] Oops: general protection fault, probably for non-canonical address 0x6b6b6b6b6b6b6bcb: 0000 [#1] SMP NOPTI ... <4>[ 267.871410] RIP: 0010:mei_gsc_remove+0x44/0x90 [mei_gsc] ... <4>[ 267.871555] Call Trace: <4>[ 267.871562] <4>[ 267.871570] auxiliary_bus_remove+0x1b/0x30 <4>[ 267.871589] device_remove+0x43/0x80 <4>[ 267.871604] device_release_driver_internal+0x215/0x280 <4>[ 267.871619] device_release_driver+0x12/0x20 <4>[ 267.871630] bus_remove_device+0xdc/0x150 <4>[ 267.871645] device_del+0x15f/0x3b0 <4>[ 267.871656] ? bus_unregister_notifier+0x37/0x50 <4>[ 267.871672] gsc_destroy_one.isra.0+0x44/0x210 [i915] <4>[ 267.872295] intel_gsc_fini+0x28/0x50 [i915] <4>[ 267.872860] intel_gt_driver_unregister+0x2c/0x80 [i915] <4>[ 267.873300] i915_driver_remove+0x6e/0x150 [i915] <4>[ 267.873694] i915_pci_remove+0x1e/0x40 [i915] <4>[ 267.874095] pci_device_remove+0x3e/0xb0 <4>[ 267.874111] device_remove+0x43/0x80 <4>[ 267.874126] device_release_driver_internal+0x215/0x280 <4>[ 267.874137] ? bus_find_device+0xa5/0xe0 <4>[ 267.874153] device_driver_detach+0x14/0x20 <4>[ 267.874164] unbind_store+0xac/0xc0 <4>[ 267.874178] drv_attr_store+0x21/0x50 <4>[ 267.874190] sysfs_kf_write+0x4a/0x80 <4>[ 267.874204] kernfs_fop_write_iter+0x188/0x240 <4>[ 267.874222] vfs_write+0x283/0x540 <4>[ 267.874241] ksys_write+0x6f/0xf0 <4>[ 267.874253] __x64_sys_write+0x19/0x30 <4>[ 267.874264] x64_sys_call+0x79/0x26a0 <4>[ 267.874277] do_syscall_64+0x93/0xd50 <4>[ 267.874291] ? do_syscall_64+0x1a2/0xd50 <4>[ 267.874301] ? do_syscall_64+0x1a2/0xd50 <4>[ 267.874313] ? do_syscall_64+0x1a2/0xd50 <4>[ 267.874324] ? clear_bhb_loop+0x30/0x80 <4>[ 267.874336] ? clear_bhb_loop+0x30/0x80 <4>[ 267.874349] entry_SYSCALL_64_after_hwframe+0x76/0x7e Fixes: 7704e6be4ed2 ("mei: hook mei_device on class device") Signed-off-by: Alexander Usyskin Link: https://lore.kernel.org/r/20250915124554.2263330-1-alexander.usyskin@intel.com Signed-off-by: Greg Kroah-Hartman

mei: hook mei_device on class device

2025-09-06T17:50:54Z

mei_device lifetime was managed by devm procedure of parent device. But such memory is freed on device_del. Mei_device object is used by client object that may be alive after parent device is removed. It may lead to use-after-free if discrete graphics driver unloads mei_gsc auxiliary device while user-space holds open handle to mei character device. Connect mei_device structure lifteme to mei class device lifetime by adding mei_device free to class device remove callback. Move exising parent device pointer to separate field in mei_device to avoid misuse. Allocate character device dynamically and allow to control its own lifetime as it may outlive mei_device structure while character device closes after parent device is removed from the system. Leave power management on parent device as we overwrite pci runtime pm procedure and user-space is expecting it there. Closes: https://gitlab.freedesktop.org/drm/i915/kernel/-/issues/14201 Signed-off-by: Alexander Usyskin Link: https://lore.kernel.org/r/20250826125617.1166546-1-alexander.usyskin@intel.com Signed-off-by: Greg Kroah-Hartman

mei: more prints with client prefix

2025-07-19T07:57:55Z

Use client-aware print macro instead of usual device print in more places to expand debug-ability. The client-aware print macro prefixes the usual device print with current connection endpoints. Signed-off-by: Alexander Usyskin Link: https://lore.kernel.org/r/20250717141112.1696482-3-alexander.usyskin@intel.com Signed-off-by: Greg Kroah-Hartman

[tree-wide] finally take no_llseek out

2024-09-27T15:18:43Z

no_llseek had been defined to NULL two years ago, in commit 868941b14441 ("fs: remove no_llseek") To quote that commit, At -rc1 we'll need do a mechanical removal of no_llseek - git grep -l -w no_llseek | grep -v porting.rst | while read i; do sed -i '/\/d' $i done would do it. Unfortunately, that hadn't been done. Linus, could you do that now, so that we could finally put that thing to rest? All instances are of the form .llseek = no_llseek, so it's obviously safe. Signed-off-by: Al Viro Signed-off-by: Linus Torvalds

mei: demote client disconnect warning on suspend to debug

2024-06-04T15:11:44Z

Change level for the "not connected" client message in the write callback from error to debug. The MEI driver currently disconnects all clients upon system suspend. This behavior is by design and user-space applications with open connections before the suspend are expected to handle errors upon resume, by reopening their handles, reconnecting, and retrying their operations. However, the current driver implementation logs an error message every time a write operation is attempted on a disconnected client. Since this is a normal and expected flow after system resume logging this as an error can be misleading. Signed-off-by: Alexander Usyskin Signed-off-by: Tomas Winkler Link: https://lore.kernel.org/r/20240530091415.725247-1-tomas.winkler@intel.com Signed-off-by: Greg Kroah-Hartman

misc: mei: main.c: fix kernel-doc warnings

2023-10-18T08:01:34Z

Fix kernel-doc warnings in main.c: main.c:465: warning: contents before sections main.c:590: warning: missing initial short description on line: * mei_ioctl_client_notify_request - Signed-off-by: Randy Dunlap Cc: Tomas Winkler Cc: Arnd Bergmann Cc: Greg Kroah-Hartman Link: https://lore.kernel.org/r/20231012024845.29169-8-rdunlap@infradead.org Signed-off-by: Greg Kroah-Hartman