Experimental Linux Patches Allow Kernel Tracing To Work Past Reboots/Crashes

[phoronix]

Phoronix: Experimental Linux Patches Allow Kernel Tracing To Work Past Reboots/Crashes

Steven Rostedt this week posted some interesting albeit experimental patches for the Linux kernel to support persistent traces that work across a reboot or crash...

https://www.phoronix.com/news/Linux-Tracing-Post-Reboots

[avis]

Great, 15 years after Android got this feature.

[darkbasic]

That would be extremely useful, I'm tired of having to attach serial consoles to debug kernel crashes.

[spicfoo]

Great, 15 years after Android got this feature.

Citation missing. Android uses Linux kernel. If Linux didn't have this feature, how exactly does Android have it?

[dragonn]

Android uses Linux kernel

A custom patched Linux kernel, not upstream (at least not yet as far I know).

[plonoma]

Great feature. Seems very useful for debugging.

[spicfoo]

A custom patched Linux kernel, not upstream (at least not yet as far I know).

Please link to the patch or docs or something. There are plenty of non upstreamble features added as hacks in forked kernels that the vendors never submitted upstream, so when someone else does it, it makes the feature more available to more users and is worth praising anyway.

[byteabit]

A custom patched Linux kernel, not upstream

That's the catch, the Android Kernel is always custom Kernel to specific hardware and for a specific operating system (Android). The upstream Linux Kernel can be installed on almost any hardware for any type of operating system and random hardware (PC, Server, Single Board, gaming consoles and so on). Having it upstream that works reliably on random installations is one thing, having it in a controlled hardware and software environment is another.

[okias]

/me having Vietnam flashbacks to ramoops, pstore etc. In general very useful thing.

[muncrief]

Okay, now this is something I think most everyone would agree would be very useful. Kernel crashes are usually difficult to diagnose, but good god random kernel crashes are sometimes impossible even with external terminals. There are a few times over the decades where I've simply had to randomly try removing or changing the positions of various components to isolate the problem, and just when I thought I'd figured it out the kernel would randomly crash again. Needless to say this is very frustrating and time consuming, and simply having any kind of persistent record of what happened would be awesome.