On Wed 2022-04-27 19:49:19, Guilherme G. Piccoli wrote:

Currently the parameter "panic_print" relies in a function called
directly on panic path; one of the flags the users can set for
panic_print triggers a console replay mechanism, to show the
entire kernel log buffer (from the beginning) in a panic event.

Two problems with that: the dual nature of the panic_print
isn't really appropriate, the function was originally meant
to allow users dumping system information on panic events,
and was "overridden" to also force a console flush of the full
kernel log buffer. It also turns the code a bit more complex
and duplicate than it needs to be.

This patch proposes 2 changes: first, we decouple panic_print
from the console flushing mechanism, in the form of a new kernel
core parameter (panic_console_replay); we kept the functionality
on panic_print to avoid userspace breakage, although we comment
in both code and documentation that this panic_print usage is
deprecated.

We converted panic_print function to a panic notifier too, adding
it on the panic informational notifier list, executed as the final
callback. This allows a more clear code and makes sense, as
panic_print_sys_info() is really a panic-time only function.
We also moved its code to kernel/printk.c, it seems to make more
sense given it's related to printing stuff.

I really like both changes. Just please split it them into two
patchset. I mean one patch for the new "panic_console_replay"
parameter and 2nd that moves "printk_info" into the notifier.

Best Regards,
Petr