CVE-2025-21951: Linux kernel · bus: mhi: host: pci_generic: Use pci_try_reset_function() to avoid deadlock
Source GCVE Database
Not known to be exploited · CVSS 5.5 Medium.
Data as of 29 Oct 2026, 22:52 UTC
CVSS:3.1/AV:L/AC:L/PR:L/UI:N/S:U/C:N/I:N/A:HExploitation evidence
Disclosed: evidence found.
CVE published 1 Apr 2025
Detectable: no evidence on record.
No public nuclei template indexed
Weaponized: no evidence on record.
No public exploit code indexed in Metasploit or Exploit-DB
Exploited in the wildExploited: no evidence on record.
Not in CISA KEV and no source reports exploitation
Observed by OffSeq MirageSeen by Mirage: no evidence on record.
No attack traffic recorded by OffSeq Mirage sensors
A missing step is not evidence of safety. Sources: cve.org (opens in a new tab), the Metasploit, Exploit-DB and nuclei-templates index as of 29 Oct 2026.
Description
In the Linux kernel, the following vulnerability has been resolved: bus: mhi: host: pci_generic: Use pci_try_reset_function() to avoid deadlock There are multiple places from where the recovery work gets scheduled asynchronously. Also, there are multiple places where the caller waits synchronously for the recovery to be completed. One such place is during the PM shutdown() callback. If the device is not alive during recovery_work, it will try to reset the device using pci_reset_function(). This function internally will take the device_lock() first before resetting the device. By this time, if the lock has already been acquired, then recovery_work will get stalled while waiting for the lock. And if the lock was already acquired by the caller which waits for the recovery_work to be completed, it will lead to deadlock. This is what happened on the X1E80100 CRD device when the device died before shutdown() callback. Driver core calls the driver's shutdown() callback while holding the device_lock() leading to deadlock. And this deadlock scenario can occur on other paths as well, like during the PM suspend() callback, where the driver core would hold the device_lock() before calling driver's suspend() callback. And if the recovery_work was already started, it could lead to deadlock. This is also observed on the X1E80100 CRD. So to fix both issues, use pci_try_reset_function() in recovery_work. This function first checks for the availability of the device_lock() before trying to reset the device. If the lock is available, it will acquire it and reset the device. Otherwise, it will return -EAGAIN. If that happens, recovery_work will fail with the error message "Recovery failed" as not much could be done.
References
- nvd.nist.gov/vuln/detail/CVE-2025-21951 (opens in a new tab)
- git.kernel.org/stable/c/1f9eb7078bc6b5fb5cbfbcb37c4bc01685332b95 (opens in a new tab)
- git.kernel.org/stable/c/62505657475c245c9cd46e42ac01026d1e61f027 (opens in a new tab)
- git.kernel.org/stable/c/7746f3bb8917fccb4571a576f3837d80fc513054 (opens in a new tab)
- git.kernel.org/stable/c/7a5ffadd54fe2662f5c99cdccf30144d060376f7 (opens in a new tab)
- git.kernel.org/stable/c/985d3cf56d8745ca637deee273929e01df449f85 (opens in a new tab)
- git.kernel.org/stable/c/a321d163de3d8aa38a6449ab2becf4b1581aed96 (opens in a new tab)
- lists.debian.org/debian-lts-announce/2025/05/msg00045.html (opens in a new tab)