kernel: bump 5.15 to 5.15.38

[rsalvaterra]

Deleted (upstreamed):
mediatek/patches-5.15/120-10-v5.18-mtd-nand-fix-ecc-parameters-for-mt7622.patch [1]

Other patches automatically rebased.

Run-tested:
mediatek/mt7622 (Belkin RT3200)

[1] https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git/commit/?h=v5.15.38&id=3ec920d755ae69c201b358e8d8e96c32f51145d8

Signed-off-by: Rui Salvaterra <rsalvaterra@gmail.com>

[lucize]

starting from 3.15.35 dell wyse cx0, Via CPU lockup if ethernet adapter is connected (even usb), with this version terminal is printing (until freezes)

SQUASHFS error: Unable to read metadata cache entry [...]
SQUASHFS error: Unable to read directory block [...]

it has an ide flash, I'll try to boot from an usb stick and see if it's the same, tried on multiple same devices
LE:
booting from usb didn't help, but no more squashfs errors printed on terminal

[rsalvaterra]

starting from 3.15.35 dell wyse cx0, Via CPU lockup if ethernet adapter is connected (even usb), with this version terminal is printing (until freezes)

SQUASHFS error: Unable to read metadata cache entry [...]
SQUASHFS error: Unable to read directory block [...]

it has an ide flash, I'll try to boot from an usb stick and see if it's the same, tried on multiple same devices LE: booting from usb didn't help, but no more squashfs errors printed on terminal

In order to rule out any possible squashfs issues, can you please build/boot an ext4 image and see if there are any errors with it?

[lucize]

yes, I'll try, but I guess it's just something that locks the system, when network traffic starts, and squashfs was just a collateral "damage", with .35 freezes instantly, with .36, ping is working a few seconds, I'll try also as an AP

LE:
so any usb device I try, beside of the onboard velocity seems to freeze the device
connecting to the wifi network works, the device release and ip but can't access the device via wifi, the device doesn't freeze
guess better luck with the next release of 5.15

[hauke]

If you think something is broken with locking you can activate these debug options and check if they find anything:

CONFIG_KERNEL_DETECT_HUNG_TASK=y
CONFIG_KERNEL_PROVE_LOCKING=y
CONFIG_KERNEL_SOFTLOCKUP_DETECTOR=y
CONFIG_KERNEL_WQ_WATCHDOG=y

[lucize]

oddly enough, with those kernel options, the device doesn't freeze anymore at boot if the cable is connected, but if I disconnect and reconnect it freezes again but in different way, the keyboard is not working anymore (caps/num lock seems to work, but no text) and the network cable status is displayed (every time I disconnect and reconnect), also acpi power button doesn't work

LE: on subsequent reboot it seems to freeze after a few seconds

[jospezial]

5.15.37 released

[lucize]

I guess the only way is to do a git bisect, at least 5.10 is still working

[graysky2]

Build-system: x86_64
Build-tested: mt7622/RT3200, bcm2711/RPi4B*
Run-tested: mt7622/RT3200, bcm2711/RPi4B*

*After applying #9675 and enabling CONFIG_TESTING_KERNEL

[lucize]

seems that the regression was fixed!
Tested x86