nimble/ll: Drop received PDU if it is handled after scan was stopped

When doing continuous scan and channel is changed close to scan stop
it is possible that PDU is received before interval timer is cancelled
and RX is disabled. This may lead to processing received PDU only after
scanner is already stopped. While this case was already handled in LL
task it was not correctly handled in ISR. If this happen simply mark
PDU to be ignored by LL and don't process it in ISR.

This is more likely to happen when small scan window is used.

Mostly affects initiator state as this may lead to assert when happen
after CONNECTION_IND was already sent.
This commit is contained in:
Szymon Janc
2023-09-19 15:16:12 +02:00
parent cb41532dc8
commit a35cb28ef0
3 changed files with 7 additions and 0 deletions
@@ -179,6 +179,7 @@ STATS_SECT_START(ble_ll_stats)
STATS_SECT_ENTRY(rx_connect_reqs)
STATS_SECT_ENTRY(rx_scan_ind)
STATS_SECT_ENTRY(rx_aux_connect_rsp)
STATS_SECT_ENTRY(rx_pdu_on_scan_disabled)
STATS_SECT_ENTRY(adv_txg)
STATS_SECT_ENTRY(adv_late_starts)
STATS_SECT_ENTRY(adv_resched_pdu_fail)
+1
View File
@@ -320,6 +320,7 @@ STATS_NAME_START(ble_ll_stats)
STATS_NAME(ble_ll_stats, rx_connect_reqs)
STATS_NAME(ble_ll_stats, rx_scan_ind)
STATS_NAME(ble_ll_stats, rx_aux_connect_rsp)
STATS_NAME(ble_ll_stats, rx_pdu_on_scan_disabled)
STATS_NAME(ble_ll_stats, adv_txg)
STATS_NAME(ble_ll_stats, adv_late_starts)
STATS_NAME(ble_ll_stats, adv_resched_pdu_fail)
+5
View File
@@ -1610,6 +1610,11 @@ ble_ll_scan_rx_isr_end(struct os_mbuf *rxpdu, uint8_t crcok)
goto scan_rx_isr_ignore;
}
if (!scansm->scan_enabled) {
STATS_INC(ble_ll_stats, rx_pdu_on_scan_disabled);
goto scan_rx_isr_ignore;
}
rxbuf = rxpdu->om_data;
pdu_type = rxbuf[0] & BLE_ADV_PDU_HDR_TYPE_MASK;