drivers: i2c_nrfx_twi: Make use of NRFX_TWI_FLAG_SUSPEND #17801
Conversation
anangl
added
area: I2C
bug
|
@pabigot please, take a look at this bug-fix PR |
There was a problem hiding this comment.
We'd discussed this back in the time of #11612, and the group consensus was the API is poorly defined and needs to be fixed at that level. Support for multi-element I2C transactions and the precise behavioral implication of flags like I2C_MSG_RESTART are unclear, so even if this does work for Nordic there's no reason to believe the underlying construct will function on other systems.
I don't know what nrfx does with those flags (in my work I control the peripheral directly) and don't have time to investigate or experiment so can't comment on whether the solution fixes this for Nordic. As such I can't explicitly approve it, but I have no objections to this being merged if it helps people get things done and it doesn't break any existing use.
Make use of the new flag introduced in nrfx v1.7.2 that suspends transfers, in order to properly handle scattered I2C transactions, especially to perform i2c_burst_write() in the way expected by display drivers, i.e. as a continuous transaction on the bus, without the repeated START between its two parts. Signed-off-by: Andrzej Głąbek <[email protected]>
anangl
force-pushed
the
nrfx_twi_i2c_burst_write
branch
from
9f81381 to
aca8b70
Compare
5 participants
Make use of the new flag introduced in nrfx v1.7.2 that suspends
transfers, in order to properly handle scattered I2C transactions,
especially to perform i2c_burst_write() in the way expected by
display drivers, i.e. as a continuous transaction on the bus,
without the repeated START between its two parts.
Fixes #15116 (although the issue mentions the TWIM peripheral, on which the hardware limitation prevents the issue from being solved, on all SoCs of nRF52 Series TWI can be used instead of TWIM).