efr32 CSMA done twice? #2379
Description
It seems that both the radio and OpenThread perform CSMA.
In efr32 radio.c function otPlatRadioTransmit
RAIL_CsmaConfig_t csmaConfig = RAIL_CSMA_CONFIG_802_15_4_2003_2p4_GHz_OQPSK_CSMA;
...
railTxStatus = RAIL_TxStartWithOptions(aFrame->mChannel, &txOption, RAIL_CcaCsma, &csmaConfig);
RAIL_CSMA_CONFIG_802_15_4_2003_2p4_GHz_OQPSK_CSMA is
/**
* @def RAIL_CSMA_CONFIG_802_15_4_2003_2p4_GHz_OQPSK_CSMA
* @brief RAIL_CsmaConfig_t initializer configuring CSMA per 802.15.4-2003
* on 2.4 GHz OSPSK, commonly used by ZigBee.
*/
#define RAIL_CSMA_CONFIG_802_15_4_2003_2p4_GHz_OQPSK_CSMA { \
/* CSMA per 802.15.4-2003 on 2.4 GHz OSPSK, commonly used by ZigBee */ \
/* csmaMinBoExp */ 3, /* 2^3-1 for 0..7 backoffs on 1st try */ \
/* csmaMaxBoExp */ 5, /* 2^5-1 for 0..31 backoffs on 3rd+ tries */ \
/* csmaTries */ 5, /* 5 tries overall (4 re-tries) */ \
/* ccaThreshold */ -75, /* 10 dB above sensitivity */ \
/* ccaBackoff */ 320, /* 20 symbols at 16 us/symbol */ \
/* ccaDuration */ 128, /* 8 symbols at 16 us/symbol */ \
/* csmaTimeout */ 0, /* no timeout */ \
}
so SiLabs Rail is set to use CSMA.
Now if we look into MAC layer, we see that before transmitting it calls RadioSupportsCsmaBackoff() and if it returns false - performs its own CSMA. In efr32 radio.c function, otPlatRadioGetCaps doesn't set OT_RADIO_CAPS_TRANSMIT_RETRIES or OT_RADIO_CAPS_CSMA_BACKOFF bits so this is precisely what happens.