wlan: Fix for WDI Timeout issue.
Fix for WDI Timeout issue due to race condition where if MC thread gets
stuck while waiting for response from HAL and in between response comes
from HAL and also new request comes from UMAC and timer also expires
before MC thread can resume and process response. So timerevent would be
in SYS Q but WDI will get timer event call back after sending new UMAC
request (after processing response to first message) and WDI will assume
that timer event is for the new UMAC Req it just sent out to HAL.
Change-Id: I506c3bdecf0e9abe37f223801890c4e7fb75ea7d
CR-Fixed: 433295
(cherry picked from commit a5e3edec1c41c8819b8172ce93460772cc0517cd)
diff --git a/CORE/WDI/CP/src/wlan_qct_wdi.c b/CORE/WDI/CP/src/wlan_qct_wdi.c
index 0d23541..a6705b0 100644
--- a/CORE/WDI/CP/src/wlan_qct_wdi.c
+++ b/CORE/WDI/CP/src/wlan_qct_wdi.c
@@ -83,6 +83,7 @@
/*----------------------------------------------------------------------------
* Include Files
* -------------------------------------------------------------------------*/
+#include <vos_trace.h>
#include "wlan_qct_wdi.h"
#include "wlan_qct_wdi_i.h"
#include "wlan_qct_wdi_sta.h"
@@ -19399,6 +19400,7 @@
)
{
WDI_Status wdiStatus = WDI_STATUS_SUCCESS;
+ wpt_uint32 ret;
/*- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -*/
/*------------------------------------------------------------------------
@@ -19413,13 +19415,30 @@
- notify transport failure
Note: CTS is reponsible for freeing the message buffer.
-----------------------------------------------------------------------*/
- if ( 0 != WCTS_SendMessage( pWDICtx->wctsHandle, (void*)pSendBuffer, usSendSize ))
+ ret = WCTS_SendMessage(pWDICtx->wctsHandle, (void*)pSendBuffer, usSendSize);
+ if ((eWLAN_PAL_STATUS_SUCCESS != ret) &&
+ (eWLAN_PAL_STATUS_E_RESOURCES != ret))
{
WPAL_TRACE(eWLAN_MODULE_DAL_CTRL, eWLAN_PAL_TRACE_LEVEL_FATAL,
"Failed to send message over the bus - catastrophic failure");
wdiStatus = WDI_STATUS_E_FAILURE;
}
+ else
+ {
+ /* even when message was placed in CTS deferred Q, we will treat it
+ success but log this info
+ */
+ if (eWLAN_PAL_STATUS_E_RESOURCES == ret)
+ {
+ WPAL_TRACE(eWLAN_MODULE_DAL_CTRL, eWLAN_PAL_TRACE_LEVEL_FATAL,
+ "WDI_SendMsg: message placed in CTS deferred Q, expected "
+ "response %s (%d)",
+ WDI_getRespMsgString(pWDICtx->wdiExpectedResponse),
+ pWDICtx->wdiExpectedResponse);
+ VOS_ASSERT(0);
+ }
+ }
/*Check if originator provided a request status callback*/
if ( NULL != pWDICtx->wdiReqStatusCB )
@@ -19444,6 +19463,9 @@
{
/*Start timer for the expected response */
wpalTimerStart(&pWDICtx->wptResponseTimer, WDI_RESPONSE_TIMEOUT);
+
+ /*cache current timestamp for debugging */
+ pWDICtx->uTimeStampRspTmrStart = vos_timer_get_system_time();
}
else
{
@@ -19493,14 +19515,18 @@
WPAL_TRACE(eWLAN_MODULE_DAL_CTRL, eWLAN_PAL_TRACE_LEVEL_INFO,
"Send indication status : %d", uStatus);
- pWDICtx->wdiReqStatusCB( (uStatus != 0 ) ? WDI_STATUS_E_FAILURE:
- WDI_STATUS_SUCCESS,
+ /* even if CTS placed indication into its deferred Q, we treat it
+ * as success and let CTS drain its queue as per smd interrupt to CTS
+ */
+ pWDICtx->wdiReqStatusCB( ((uStatus != eWLAN_PAL_STATUS_SUCCESS) && (uStatus != eWLAN_PAL_STATUS_E_RESOURCES)) ? WDI_STATUS_E_FAILURE: WDI_STATUS_SUCCESS,
pWDICtx->pReqStatusUserData);
}
/*If sending of the message failed - it is considered catastrophic and
indicates an error with the device*/
- if ( 0 != uStatus)
+ if (( eWLAN_PAL_STATUS_SUCCESS != uStatus) &&
+ ( eWLAN_PAL_STATUS_E_RESOURCES != uStatus))
+
{
WPAL_TRACE(eWLAN_MODULE_DAL_CTRL, eWLAN_PAL_TRACE_LEVEL_FATAL,
"Failed to send indication over the bus - catastrophic failure");
@@ -19598,14 +19624,32 @@
return;
}
+ /*cache current timestamp for debugging */
+ pWDICtx->uTimeStampRspTmrExp = vos_timer_get_system_time();
+
+ /* If response timer is running at this time that means this timer
+ * event is not for the last request but rather last-to-last request and
+ * this timer event has come after we recevied respone for last-to-last
+ * message
+ */
+ if (VOS_TIMER_STATE_RUNNING == wpalTimerGetCurStatus(&pWDICtx->wptResponseTimer))
+ {
+ WPAL_TRACE(eWLAN_MODULE_DAL_CTRL, eWLAN_PAL_TRACE_LEVEL_FATAL,
+ "WDI_ResponseTimerCB: timer in running state on timer event, "
+ "ignore tmr event, timeStampTmrStart: %ld, timeStampTmrExp: %ld",
+ pWDICtx->uTimeStampRspTmrStart, pWDICtx->uTimeStampRspTmrExp);
+ return;
+ }
+
if ( WDI_MAX_RESP != pWDICtx->wdiExpectedResponse )
{
WPAL_TRACE(eWLAN_MODULE_DAL_CTRL, eWLAN_PAL_TRACE_LEVEL_FATAL,
"Timeout occurred while waiting for %s (%d) message from device "
- " - catastrophic failure",
+ " - catastrophic failure, timeStampTmrStart: %ld, timeStampTmrExp: %ld",
WDI_getRespMsgString(pWDICtx->wdiExpectedResponse),
- pWDICtx->wdiExpectedResponse);
+ pWDICtx->wdiExpectedResponse, pWDICtx->uTimeStampRspTmrStart,
+ pWDICtx->uTimeStampRspTmrExp);
/* WDI timeout means Riva is not responding or SMD communication to Riva
* is not happening. The only possible way to recover from this error
* is to initiate SSR from APPS
@@ -19624,8 +19668,10 @@
else
{
WPAL_TRACE( eWLAN_MODULE_DAL_CTRL, eWLAN_PAL_TRACE_LEVEL_ERROR,
- "Timeout occurred but not waiting for any response %d",
- pWDICtx->wdiExpectedResponse);
+ "Timeout occurred but not waiting for any response %d "
+ "timeStampTmrStart: %ld, timeStampTmrExp: %ld",
+ pWDICtx->wdiExpectedResponse, pWDICtx->uTimeStampRspTmrStart,
+ pWDICtx->uTimeStampRspTmrExp);
}
return;