On Mon, Jan 19, 2026 at 11:33:26AM +0000, Jonathan Cameron wrote:

On Wed, 14 Jan 2026 11:46:07 +0000
Cristian Marussi [off-list ref] wrote:

quoted

Allow protocols themselves to register for their own notifications and

and provide their own notifier callbacks.

quoted

providing their own notifier callbacks. While at that, allow for a protocol
to register events with compilation-time unknown report/event sizes: such
events will use the maximum transport size.

I'm not keen on the 'while at that' part of the patch. In an ideal
world that's a separate patch.

Yes indeed...it was tempting to do it together with the rework since it
was the only usecase that triggered the 'while-at' change...

...but this series in general needs more splitting both at the protocol
level and at the FS level (once I get some feedback from FS guys) so I
will split this out too.

One other comment inline.

Jonathan
p.s. You get to my review victim whilst I run a particularly annoying
bisection on the other screen (completely unrelated!) :)

quoted

Signed-off-by: Cristian Marussi <cristian.marussi@arm.com>
---
v1-->v2
 - Fixed multiline comment format
---
 drivers/firmware/arm_scmi/common.h    |  4 ++++
 drivers/firmware/arm_scmi/driver.c    | 12 ++++++++++++
 drivers/firmware/arm_scmi/notify.c    | 28 ++++++++++++++++++++-------
 drivers/firmware/arm_scmi/notify.h    |  8 ++++++--
 drivers/firmware/arm_scmi/protocols.h |  6 ++++++
 5 files changed, 49 insertions(+), 9 deletions(-)

quoted

diff --git a/drivers/firmware/arm_scmi/notify.c b/drivers/firmware/arm_scmi/notify.c
index 78e9e27dc9ec..e84b4dbefe82 100644
--- a/drivers/firmware/arm_scmi/notify.c
+++ b/drivers/firmware/arm_scmi/notify.c

quoted

@@ -779,8 +787,13 @@ int scmi_register_protocol_events(const struct scmi_handle *handle, u8 proto_id,
 	}
 
 	evt = ee->evts;
-	for (i = 0; i < ee->num_events; i++)
+	for (i = 0; i < ee->num_events; i++) {
+		if (evt[i].max_payld_sz == 0) {
+			payld_sz = max_msg_sz;
+			break;
+		}
 		payld_sz = max_t(size_t, payld_sz, evt[i].max_payld_sz);

Everything here seems to already be a size_t.  It is rare that we actually need max_t over
max, and definitely not when all the types match.
		payld_sz = max(payl_sz, evt[i].max_payld_sz);

...indeed...I will fix.

Thanks,
Cristian