Kernel Panic when using MTP transfers with an Android phone

[fatigatti]

Samsung Galaxy A03 core (Android 11), Android File Transfer application.

Steps to reproduce: 1) Connect the phone in MTP mode and open Android File Transfer from the terminal. It may work a couple of times. Keep plugging/unplugging and reopening the app until it hangs with the message "device found, opening session" and nothing happens. 2) Kill the app using CTRL+C on the terminal. 3) Open the app again and it will crash the system (picture attached, sorry about the quality), in this case not allowing to continue and thus not registering to the syslog.

[waddlesplash]

Any chance you have any USB3 ports, or any devices with USB3 ports, and can test connecting to those instead of a USB2 port and see if the behavior differs?

[fatigatti]

I do. I couldn't reproduce this particular panic on USB3. However, when I switched the connection mode to tethering just to try things out I got #17797, which I reported yesterday. They are different and while I can get out of #17797 and resume working, this one needs a hard reset (I can type exit or co, but the it just prints the stack trace again).

There's also another bug: under certain circunstances that I need to investigate more but possibly include plugging and unplugging a few times the entire USB stack freezes. Ports work as charge only and after I disconnect all devices listusb keeps displaying them as present, and obviously new devices are not recognised.

[waddlesplash]

Yes, so it's a problem with our USB2 driver itself, it seems. It's not surprising that USB3 isn't affected, the equivalent routines there have been "battle-tested" against stuff like this at this point :)

The problem in #17797 is unrelated to USB, it's a problem with the network stack itself somehow, probably related to device removal.

[waddlesplash]

And, the other panic is an assertion failure which indeed can be gotten past (depending on severity, some indicate things so broken that continuing may make things worse.) This panic is a memory fault, much more serious and indeed can't be worked around.

[waddlesplash]

Please test after hrev56192 and see if it makes any difference.

[fatigatti]

Just updated to hrev56192, makes no difference. Still getting the page fault.

[diver]

Please attach a new photo anyway.

[fatigatti]

Same page fault in EHCI

[fatigatti]

A different page fault

[fatigatti]

I re-tested for the picture and managed to get the same EHCI page fault and a different one (admittedly plugging, opening the app, changing the phone's mode, unplugging, doing nasty things).