Bug 198543 - intel-ish-hid: Does not recover after kexec
Summary: intel-ish-hid: Does not recover after kexec
Status: NEW
Alias: None
Product: Drivers
Classification: Unclassified
Component: Other (show other bugs)
Hardware: Intel Linux
: P1 low
Assignee: drivers_other
URL:
Keywords:
Depends on:
Blocks:
 
Reported: 2018-01-21 11:29 UTC by Sam Edwards
Modified: 2022-12-01 04:55 UTC (History)
3 users (show)

See Also:
Kernel Version: 4.14.14
Subsystem:
Regression: No
Bisected commit-id:


Attachments
photo from the kernel panic after recovery from suspend to disk (167.38 KB, image/jpeg)
2022-04-14 09:43 UTC, Christian
Details

Description Sam Edwards 2018-01-21 11:29:03 UTC
The intel-ish-hid driver does not properly reinitialize the device after kexec. Presumably, the driver relies on a full hardware reset in order to put the ISH into a known state. If performing a cold boot or full system reboot, everything works fine. However, during a kexec, the kernel stalls for 10 seconds, the sensor hub is NOT available during that boot, and the following appears in dmesg:

[   10.719218] intel_ish_ipc 0000:00:13.0: [ishtp-ish]: Timed out waiting for FW-initiated reset
[   10.719365] intel_ish_ipc 0000:00:13.0: ISH: hw start failed.
[   10.719999] intel_ish_ipc 0000:00:13.0: ISH: PCI driver initialization failed.

This is, naturally, fixed upon the next full reboot. However, it is a minor annoyance to users who use kexec in their workflow and could be indicative of a missing HW reset in the driver initialization or cleanup.
Comment 1 Christian 2022-04-14 09:41:43 UTC
I am not sure if this is the same problem but I see same error messages after a suspend-to-disk recovery. I can only recover with a hard reset.

This is the case for several kernel releases now using latest mainline kernels in AlmaLinux 8 with a Dell Latitude 7410 Laptop.
Comment 2 Christian 2022-04-14 09:43:22 UTC
Created attachment 300755 [details]
photo from the kernel panic after recovery from suspend to disk

Note You need to log in before you can comment on or make changes to this bug.