Christian Mauderer commented on a discussion on bsps/aarch64/raspberrypi/i2c/raspberrypi-i2c.c: https://gitlab.rtems.org/rtems/rtos/rtems/-/merge_requests/682#note_129455 > bool read_transfer; > } raspberrypi_i2c_bus; > > -static int i2c_polling_read( raspberrypi_i2c_bus *bus ) > +static int rpi_i2c_bus_transfer( raspberrypi_i2c_bus *bus ) > { > - while ( !( S_REG( bus ) & S_DONE ) && ( bus->remaining_bytes > 0 ) ) { > - while ( ( S_REG( bus ) & S_RXD ) && ( bus->remaining_bytes > 0 ) ) { > + while ( bus->remaining_bytes > 0 ) { > + if ( bus->read_transfer ) { > + while ( ( S_REG( bus ) & S_RXD ) == 0 ) { > + } Do you know how the clock stretching feature works? Basically, a slave is allowed to just keep the clock line on low if the master is too fast. That can happen during a transfer. So it's quite possible that instead of a S_RXD, you get a S_CLKT. In that case, you will never receive the S_RXD and your code will hang here. So instead of the `while ( ( S_REG( bus ) & S_RXD ) == 0 )` you need something like a `while ( ( S_REG( bus ) & (S_RXD | S_CLKT) ) == 0 )` and then you have to check whether the status is S_RXD or S_CLKT to decide whether you continue or abort the transfer. -- View it on GitLab: https://gitlab.rtems.org/rtems/rtos/rtems/-/merge_requests/682#note_129455 You're receiving this email because of your account on gitlab.rtems.org.
_______________________________________________ bugs mailing list [email protected] http://lists.rtems.org/mailman/listinfo/bugs
