commit b08723d7ce4843c1604abaaa9d18ff4e0461e033
author Paul Fertser <fercerpav@gmail.com> Fri Dec 16 11:49:08 2022 +0000
committer Paul Fertser <fercerpav@gmail.com> Sat Dec 17 08:47:01 2022 +0000
tree 88a326d25b0084ee4d849fd85b46270c47f778f0
parent 541637caadc0469b251cd1d191b8e785e574f1f4

Try up to 15 times in case of timeouts

When the host OS is idle the CPUs can reach C6 state in which PECI is
unresponsive. It's often enough to try reading the same data more
times to catch a moment when some background daemon wakes the CPU up.

Tested on a Tioga Pass board with dual Xeon Scalable 1st gen CPUs
running Windows 10 pro 22H2 and Ubuntu 22.10 desktop.

Signed-off-by: Paul Fertser <fercerpav@gmail.com>
Change-Id: I63805303dae774a8c60925c2cd3df856c4197a19

src/peci_pcie.cpp

diff --git a/src/peci_pcie.cpp b/src/peci_pcie.cpp
index e65f33f..eabdf84 100644
--- a/src/peci_pcie.cpp
+++ b/src/peci_pcie.cpp
@@ -177,7 +177,7 @@
     std::array<uint8_t, pciReadSize> data;
     uint8_t cc;
     int ret = PECI_CC_TIMEOUT;
-    for (int index = 0; (index < 5) && (ret == PECI_CC_TIMEOUT); index++)
+    for (int index = 0; (index < 15) && (ret == PECI_CC_TIMEOUT); index++)
     {
 #ifdef USE_RDENDPOINTCFG
         ret = peci_RdEndPointConfigPci(clientAddr,  // CPU Address