[for,v5.5,0/2] input/rmi4 regression fix

Message

Hans Verkuil Jan. 15, 2020, 12:48 p.m. UTC

Hi Dmitry,

Timo Kaufmann reported that the "don't increment rmiaddr" patch broke
his Thinkpad T450 (the cursor started drifting after a while).

Some more digging revealed that that patch was bad, and that the
real solution for the original problem (function F54 reported bad
data) had to be done in rmi_f54.c.

So the first patch reverts the old fix (and so will break F54 again),
and this should be merged for v5.5 asap. The second patch that fixes
F54 can be postponed to v5.6: F54 is rarely used and I don't think it
is important enough for v5.5.

I have only tested the F54 fix for smbus, I don't know if it is OK
for i2c/spi since I have no hardware to test that. It seems sane :-)

Regards,

	Hans

Hans Verkuil (2):
  Revert "Input: synaptics-rmi4 - don't increment rmiaddr for SMBus
    transfers"
  Input: rmi_f54: read from FIFO in 32 byte blocks

 drivers/input/rmi4/rmi_f54.c   | 43 +++++++++++++++++++++-------------
 drivers/input/rmi4/rmi_smbus.c |  2 ++
 2 files changed, 29 insertions(+), 16 deletions(-)