As the 'incomplete_write_byte' fault injector has shown, we need to take care that pulses used for recovery will not be misinterpreted as a regular cycle. This series implements that. This approach makes sure that STOP signals put on the bus do not create additional pulses. We still do 9 pulses only, but generate a STOP signal in case SDA was released again. The first patch implements it in a minimal way, so it is suitable for stable. The second patch adds some sanity checking and documentation on top which is not suitable for stable. The third patch simplifies the logic because of the new behaviour. It has been tested with a Renesas Lager board (R-Car H2). I will test it tomorrow a bit more once I taught the GPIO driver bus recovery. A branch (with further improvements) can be found here: git://git.kernel.org/pub/scm/linux/kernel/git/wsa/linux.git renesas/i2c/recovery/write-byte-fix Looking forward to comments. Kind regards, Wolfram Changes since V1: * we create STOP now always when set_sda is available independent of get_sda being available or not. This is much safer and allows for a cleaner logic implemented in the new patch 3. * hopefully improved comments Wolfram Sang (3): i2c: recovery: if possible send STOP with recovery pulses i2c: recovery: require either get_sda or set_sda i2c: recovery: refactor recovery function drivers/i2c/i2c-core-base.c | 36 +++++++++++++++++++----------------- include/linux/i2c.h | 12 ++++++------ 2 files changed, 25 insertions(+), 23 deletions(-) -- 2.11.0
