Modify IPMI Timeout Value

  -- Currently the Timeout value for an IPMI message is assigned
     when the message is placed in the queue. The timeout value
     is the same whether we have 0 or 100 messages in the queue
     which leads to the occasional IPMI timeout when the BMC
     is sent a lot of messages in a short period of time.

  -- This fix is to implement the timeout as a per message timeout.
     I.E. if the message being put on the queue is the 3rd message
     on the queue it would get a timeout of 3x the timeout value.

Change-Id: Ieb328d96144f2b1fe9dc39f57191e86cf4821c2c
CQ: SW461931
Reviewed-on: http://rchgit01.rchland.ibm.com/gerrit1/75538
Tested-by: Jenkins Server <pfd-jenkins+hostboot@us.ibm.com>
Tested-by: Jenkins OP Build CI <op-jenkins+hostboot@us.ibm.com>
Reviewed-by: Ilya Smirnov <ismirno@us.ibm.com>
Tested-by: FSP CI Jenkins <fsp-CI-jenkins+hostboot@us.ibm.com>
Tested-by: Jenkins OP HW <op-hw-jenkins+hostboot@us.ibm.com>
Reviewed-by: Christian R. Geddes <crgeddes@us.ibm.com>
Reviewed-by: Daniel M. Crowell <dcrowell@us.ibm.com>

1 files changed, 7 insertions, 5 deletions

diff --git a/src/usr/ipmibase/ipmirp.C b/src/usr/ipmibase/ipmirp.C
index 2def821bb..0bb785949 100644
--- a/src/usr/ipmibase/ipmirp.C
+++ b/src/usr/ipmibase/ipmirp.C
@@ -935,8 +935,12 @@ void IpmiRP::queueForResponse(IPMI::Message& i_msg)
     mutex_lock(&iv_mutex);
 
     // BMC request-to-response times are always seconds, 1 - 30.
-    // And I don't think we care about roll over here.
-    i_msg.iv_timeout.tv_sec += iv_bmc_timeout;
+    // And I don't think we care about roll over here. Enforce the
+    // timeout as a timeout per-message. Meaning if there are 2
+    // messages on the timeout queue, make the timeout of this message
+    // iv_bmc_timeout + iv_bmc_timeout*2
+    i_msg.iv_timeout.tv_sec +=
+                  (iv_bmc_timeout + (iv_bmc_timeout*iv_timeoutq.size()) );
 
     // Put this message on the response queue so we can find it later
     // for a response and on the timeout queue so if it times out
@@ -944,9 +948,7 @@ void IpmiRP::queueForResponse(IPMI::Message& i_msg)
     // timeout - mostly. Every message sent before the BMC tells us
     // the timeout (at least one message) will have the shortest possible
     // timeout. The BMC might lengthen the timeout, but can not shorten
-    // it. All messages after that will have the same timeout. So the
-    // timeout queue is "sorted."
-
+    // it.
     iv_respondq[i_msg.iv_seq] = i_msg.iv_msg;
     iv_timeoutq.push_back(i_msg.iv_msg);