Hi, On Mon, Feb 16, 2026 at 10:11 AM Douglas Anderson <[email protected]> wrote: > > As talked about in the first patch in this series, passing NULL as the > 'mssg' argument to mbox_send_message() ends up causing confusion and > quirky behavior inside the mailbox core. Despite this, there are a > number of drivers that pass NULL. > > It is plausible that some of the drivers passing NULL may have been > taking advantage of the quirks of the mailbox core. Specifically, they > may have been taking advantage that calling "tx_done" wasn't truly > necessary for NULL messages (it was a noop) or that NULL messages were > passed onto the mailbox controller right away without queuing. > > This series introduces a new API call: mbox_ring_doorbell(). The new > API call tries to mimic the specific quirks that were helpful in the > old behavior and it's expected to be a nearly drop-in replacement. > > There are some subtle differences between the new call and the old > behavior, but it's expected that all of these differences are only for > cases where the old behavior made little sense. The description of the > first patch details these differences. > > The series attempts to convert all in-tree users to stop passing NULL > for mssg. As per above, there are some slight differences in behavior. > If any of the patches are causing problems, they can safely be > reverted while debugging the problems. Eventually, all code should be > converted over to stop passing NULL mssg. > > Changes in v3: > - Suggest mbox_ring_doorbell in the warning message > - Updated patch description based on Cristian's response. > > Changes in v2: > - Instead of just documenting NULL, introduce a new function
In case any mailbox client API users want to comment, Jassi has proposed an alternative to this series: https://lore.kernel.org/r/[email protected]/ Feel free to join the conversation over there. :-) -Doug
