Update the MCTP socket timeout

On the miniBMC side when the IOC is being reset, the IOC
config page read can take longer.  Currently the total timeout
for the config page read is set to 6 seconds. This config page
read happens every time miniBMC receives redfish requests for
SATA drive resources. The redfish response timeout on the BMC
is set to 1 second. So in any case, if IOC goes of reset, the
config page read can easily take more than 2-3 seconds and
therefore BMC can timeout. Eventually miniBMC and BMC
communication can go out of sync.

This can happen when miniBMC is receiving SATA drive resource requests
while Host is booting up. During host bootup IOC resets 3 times.

So update the BMC MCTP socket timeout to a safe value.

Tested:
Ran repeated redfish queries for SATA drive resources for
over 8 hours on a machine where the Host is bootlooping and on every
boot it is resetting IOC 3 times. No issue of miniBMC and BMC going
out of sync is observed.

Google-Bug-Id: 300946789
Google-Bug-Id: 310061372
Google-Bug-Id: 308729272
Change-Id: Ie2913e8f3c9e13736b3b593b581334b0643cf8b1
Signed-off-by: Nikhil Namjoshi <nikhilnamjoshi@google.com>

rded.cpp

diff --git a/rded.cpp b/rded.cpp
index 1272643..75b78d3 100644
--- a/rded.cpp
+++ b/rded.cpp
@@ -9,6 +9,8 @@
 #include <cstddef>
 #include <string>
 
+constexpr int SOCKET_TIMEOUT_SECONDS = 8;
+
 void setSocketTimeout(int fd, int seconds, int milliseconds)
 {
     struct timeval tv;
@@ -29,7 +31,8 @@
 
     stdplus::print(stderr, "Socket Created with ID: {}. Setting timeouts...\n",
                    fd);
-    setSocketTimeout(fd, /*seconds=*/1, /*milliseconds=*/0);
+    setSocketTimeout(fd, /*seconds=*/SOCKET_TIMEOUT_SECONDS,
+                     /*milliseconds=*/0);
 
     socklen_t optLen;
     int currentSendBuffSize;