On Wed 2026-01-28 06:17:39, Breno Leitao wrote:

Convert netconsole from the legacy console API to the NBCON framework.
NBCON provides threaded printing which unblocks printk()s and flushes in
a thread, decoupling network TX from printk() when netconsole is
in use.

Since netconsole relies on the network stack which cannot safely operate
from all atomic contexts, mark both consoles with
CON_NBCON_ATOMIC_UNSAFE. (See discussion in [1])

CON_NBCON_ATOMIC_UNSAFE restricts write_atomic() usage to emergency
scenarios (panic) where regular messages are sent in threaded mode.

Implementation changes:
- Unify write_ext_msg() and write_msg() into netconsole_write()
- Add device_lock/device_unlock callbacks to manage target_list_lock
- Use nbcon_enter_unsafe()/nbcon_exit_unsafe() around network
  operations.
  - If nbcon_enter_unsafe() fails, just return given netconsole lost
    the ownership of the console.

I was just curious and scratched my head around this a bit.
If I get it correctly then it might fail only in a single situation.
It actually should never happen on systems which stop CPUs by NMI.

My thiking is:

  1. nbcon->write_thread() is called only from the dedicated kthread
     so that there is always only one instance.

     It might actually be called also by the legacy loop when a boot
     console is registed. But the kthread is blocked in this case.

     Anyway, the callback is serialized also using netcon->device_lock().


  2. nbcon->write_atomic() is called only by the final
     nbcon_atomic_flush_unsafe() because of CON_NBCON_ATOMIC_UNSAFE.

     nbcon_enter_unsafe() always succeeds here because the _usafe_
     takeover is allowed.


  3. No other lock is synchronized with nbcon context.

     It is acceptable because nbcon->write_atomic() is called
     only by nbcon_atomic_flush_unsafe() where the nbcon context
     synchronization is ignored anyway.

So, nbcon_enter_unsafe() might fail only in netcon->write_thread()
when the kthread is still running on another CPU while
nbcon_atomic_flush_unsafe() is called on the panic CPU.

And it should never happen when the non-panic CPUs are stopped
by NMI.

By other words, the nbcon context synchronization does not have
much value when CON_NBCON_ATOMIC_UNSAFE is used.

It is acceptable from my POV. I write this just to make
the expectations clear. I wish, I had time to write a more
comprehensible documentation about the printk design...

- Set write_thread and write_atomic callbacks (both use same function)

Link: https://lore.kernel.org/all/b2qps3uywhmjaym4mht2wpxul4yqtuuayeoq4iv4k3zf5wdgh3@tocu6c7mj4lt/ (local) [1]
Signed-off-by: Breno Leitao <leitao@debian.org>

Feel free to use:

Reviewed-by: Petr Mladek <pmladek@suse.com>

Best Regards,
Petr