On Fri, Jun 15, 2018 at 11:17 AM, Sekhar Nori [off-list ref] wrote:

Hi Rob,

On Wednesday 13 June 2018 02:36 AM, Rob Herring wrote:

quoted

On Tue, Jun 05, 2018 at 01:01:22AM -0500, Nishanth Menon wrote:

quoted

AM654 uses a UART controller that is compatible (partially) with
existing 8250 UART, however, has a few differences with respect to DMA
support and control paths. Introduce a base definition that allows us
to build up the differences in follow on patches.

Cc: Sekhar Nori <redacted>
Cc: Vignesh R <vigneshr@ti.com>
Signed-off-by: Nishanth Menon <nm@ti.com>
---
 Documentation/devicetree/bindings/serial/omap_serial.txt | 1 +
 drivers/tty/serial/8250/8250_omap.c                      | 1 +
 2 files changed, 2 insertions(+)

diff --git a/Documentation/devicetree/bindings/serial/omap_serial.txt b/Documentation/devicetree/bindings/serial/omap_serial.txt
index 4b0f05adb228..c35d5ece1156 100644
--- a/Documentation/devicetree/bindings/serial/omap_serial.txt
+++ b/Documentation/devicetree/bindings/serial/omap_serial.txt

@@ -1,6 +1,7 @@
 OMAP UART controller

 Required properties:
+- compatible : should be "ti,am654-uart" for AM654 controllers

Not compatible with any existing TI 8250 UARTs?

Curious on why you asked about this. Are you suggesting why not:

"ti,<new-soc>-uart", "ti,<old-soc>-uart"

Correct.

or you are asking why introduce "ti,<new-soc>-uart" unless there is
clear demonstrable need for using it in driver code.

In general, I think "ti,<new-soc>-uart", "ti,<old-soc>-uart" in
device-tree (and by extension in binding document) is better even in
there are no _known_ incompatibilities at the time of initial driver
submission. The reason is silicon integration and process differences
many times spill over into driver.

Yes, and chip designers can't be trusted. ;)

Of course, the idea is not to go postal and create a new compatible for
every pin-compatible part number that gets created, but probably a new
compatible should be created for a new silicon die.

Yes, that's the criteria I would use too. That's sometimes hard if
it's not the chip vendor doing the DT bindings.

We have just started introducing support for this SoC, and since it
reuses many IPs, this question is likely to come up again.

In this particular case though, Nishanth is perfectly right in not saying

compatible : should be "ti,am654-uart", "ti,omap4-uart"

Because we *know* UART DMA integration is different and a match against
omap4 would result in non-working UART once DMA is enabled by default.

Okay, makes sense. I'd suggest rewording the commit message to include
this. The "compatible to 8250 except for DMA" part I would have
applied to all TI UARTs rather than DMA differences with other TI
UARTs.

Rob