On 06/26/2016 11:06 PM, Martin Blumenstingl wrote:

The meson-ir driver uses the wrong offset (at least according to
Amlogic's reference driver as well as the datasheets of the
Meson8b/S805 and GXBB/S905).
This means that we are getting incorrect durations (REG1_TIME_IV)
reported from the hardware.

Hi,

I'm quite sure the registers are good for meson6 actually, and
it seems reasonable Amlogic made the HW evolve for the Meson8 and GXBB platforms.

This problem was also noticed by some people trying to use this on an
ODROID-C1 and ODROID-C2 - the workaround there (probably because the
datasheets were not publicy available yet at that time) was to switch
to ir_raw_event_store_edge (which leaves it up to the kernel to measure
the duration of a pulse). See [0] and [1] for the corresponding
patches.

Since we are using devicetree, the correct way to achieve this fix is not
to drop support for meson6 (what you do) but add a logic to select the correct
register for meson8 and gxbb if their compatible string are encountered.

Please note that I was only able to test this on an GXBB/S905 based
device (due to lack of other hardware).

I made this fix already but lacked time to actually test it on HW :
https://github.com/torvalds/linux/compare/master...superna9999:amlogic/v4.7/ir

My patch is missing the meson8b support, and may need a supplementary compatible check or
a separate dt match table.

Neil

[0] https://github.com/erdoukki/linux-amlogic-1/commit/969b2e2242fb14a13cb651f9a1cf771b599c958b
[1] http://forum.odroid.com/viewtopic.php?f=135&t=20504

PS: BTW could you format the cover letter using the git format-patch --cover-letter instead and
add the v2 using the -subject-prefix like :
# git format-patch --cover-letter --signoff --subject-prefix "PATCH v2" -2

_______________________________________________
linux-amlogic mailing list
linux-amlogic@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-amlogic