From a34832f11e5c2a629bac094eccdd2b7fcde6a129 Mon Sep 17 00:00:00 2001
From: Kent Russell <kent.russell@amd.com>
Date: Thu, 8 Aug 2019 09:00:22 -0400
Subject: [PATCH] Fix RAS change

RAS formatting changed, so get it to handle both types of sysfs output
until it's normalized
Change-Id: I56f2a2495af8ff4d01011bc614283376afb9ad0a
---
 include/rocm_smi/rocm_smi.h        |  3 ++-
 src/rocm_smi.cc                    | 10 +++++++---
 tests/rocm_smi_test/test_common.cc |  6 +++++-
 3 files changed, 14 insertions(+), 5 deletions(-)

diff --git a/include/rocm_smi/rocm_smi.h b/include/rocm_smi/rocm_smi.h
index 4434e93f39..a32b7e0809 100755
--- a/include/rocm_smi/rocm_smi.h
+++ b/include/rocm_smi/rocm_smi.h
@@ -383,8 +383,9 @@ typedef enum {
   RSMI_RAS_ERR_STATE_MULT_UC,    //!< Multiple uncorrectable errors
   RSMI_RAS_ERR_STATE_POISON,     //!< Firmware detected error and isolated
                                  //!< page. Treat as uncorrectable.
+  RSMI_RAS_ERR_STATE_ENABLED,    //!< ECC is enabled
 
-  RSMI_RAS_ERR_STATE_LAST = RSMI_RAS_ERR_STATE_POISON,
+  RSMI_RAS_ERR_STATE_LAST = RSMI_RAS_ERR_STATE_ENABLED,
   RSMI_RAS_ERR_STATE_INVALID = 0xFFFFFFFF
 } rsmi_ras_err_state_t;
 
diff --git a/src/rocm_smi.cc b/src/rocm_smi.cc
index 2b3953af80..bf8fb81404 100755
--- a/src/rocm_smi.cc
+++ b/src/rocm_smi.cc
@@ -532,8 +532,10 @@ static const std::map<std::string, rsmi_ras_err_state_t> kRocmSMIStateMap = {
     {"single_correctable", RSMI_RAS_ERR_STATE_SING_C},
     {"multi_uncorrectable", RSMI_RAS_ERR_STATE_MULT_UC},
     {"poison", RSMI_RAS_ERR_STATE_POISON},
+    {"off", RSMI_RAS_ERR_STATE_DISABLED},
+    {"on", RSMI_RAS_ERR_STATE_ENABLED},
 };
-static_assert(RSMI_RAS_ERR_STATE_LAST == RSMI_RAS_ERR_STATE_POISON,
+static_assert(RSMI_RAS_ERR_STATE_LAST == RSMI_RAS_ERR_STATE_ENABLED,
                  "rsmi_gpu_block_t and/or above name map need to be updated"
                                                      " and then this assert");
 
@@ -562,6 +564,7 @@ rsmi_status_t rsmi_dev_ecc_status_get(uint32_t dv_ind, rsmi_gpu_block_t block,
 
   std::string blk_line;
   std::string search_str = kRocmSMIBlockMap.at(block);
+  std::string sysfs_junk = " ras feature mask:";
   std::string state_str;
 
   search_str += ":";
@@ -570,8 +573,9 @@ rsmi_status_t rsmi_dev_ecc_status_get(uint32_t dv_ind, rsmi_gpu_block_t block,
     std::istringstream fs1(val_vec[i]);
 
     fs1 >> blk_line;
-
-    if (blk_line == search_str) {
+    if (blk_line == search_str || blk_line == kRocmSMIBlockMap.at(block)) {
+      if (blk_line.back() != ':')
+        fs1.ignore(sysfs_junk.length(), ':');
       fs1 >> state_str;
       assert(kRocmSMIStateMap.count(state_str));
       *state = kRocmSMIStateMap.at(state_str);
diff --git a/tests/rocm_smi_test/test_common.cc b/tests/rocm_smi_test/test_common.cc
index 50684065b8..6a7dfd112b 100755
--- a/tests/rocm_smi_test/test_common.cc
+++ b/tests/rocm_smi_test/test_common.cc
@@ -70,6 +70,8 @@ static const char * kRasErrStateStrings[] = {
     "Single, Correctable",           // RSMI_RAS_ERR_STATE_SING_C
     "Multiple, Uncorrectable",       // RSMI_RAS_ERR_STATE_MULT_UC
     "Poison"                         // RSMI_RAS_ERR_STATE_POISON
+    "off",                           // RSMI_RAS_ERR_STATE_DISABLED
+    "on",                            // RSMI_RAS_ERR_STATE_ENABLED
 };
 static_assert(
   sizeof(kRasErrStateStrings)/sizeof(char *) == (RSMI_RAS_ERR_STATE_LAST + 1),
@@ -89,8 +91,10 @@ static const std::map<rsmi_ras_err_state_t, const char *> kErrStateNameMap = {
                             kRasErrStateStrings[RSMI_RAS_ERR_STATE_MULT_UC]},
     {RSMI_RAS_ERR_STATE_POISON,
                             kRasErrStateStrings[RSMI_RAS_ERR_STATE_POISON]},
+    {RSMI_RAS_ERR_STATE_ENABLED,
+                            kRasErrStateStrings[RSMI_RAS_ERR_STATE_ENABLED]},
 };
-static_assert(RSMI_RAS_ERR_STATE_LAST == RSMI_RAS_ERR_STATE_POISON,
+static_assert(RSMI_RAS_ERR_STATE_LAST == RSMI_RAS_ERR_STATE_ENABLED,
                                       "kErrStateNameMap needs to be updated");
 
 static const struct option long_options[] = {