Disable bad MCAs based on CRP0:Lx keyword data

The CRP0 record has 8 records (L1-L8) that correspond to the 8 ports
(MCA targets) for the chip.  One of the fields inside the keyword indicates
if the port is disabled.  This field marks the MCA as non-functional, but
the MCA remains present.

Change-Id: Iccc943ad477a90a25aeaff67572b66126c759913
RTC:166354
Reviewed-on: http://ralgit01.raleigh.ibm.com/gerrit1/34885
Tested-by: Jenkins Server <pfd-jenkins+hostboot@us.ibm.com>
Tested-by: FSP CI Jenkins <fsp-CI-jenkins+hostboot@us.ibm.com>
Tested-by: Jenkins OP Build CI <op-jenkins+hostboot@us.ibm.com>
Reviewed-by: William G. Hoffa <wghoffa@us.ibm.com>
Reviewed-by: Dean Sanner <dsanner@us.ibm.com>
Reviewed-by: Christian R. Geddes <crgeddes@us.ibm.com>

7 files changed, 245 insertions, 4 deletions

diff --git a/src/include/usr/hwas/common/deconfigGard.H b/src/include/usr/hwas/common/deconfigGard.H
index 7efb854cd..0c133f427 100644
--- a/src/include/usr/hwas/common/deconfigGard.H
+++ b/src/include/usr/hwas/common/deconfigGard.H
@@ -5,7 +5,7 @@
 /*                                                                        */
 /* OpenPOWER HostBoot Project                                             */
 /*                                                                        */
-/* Contributors Listed Below - COPYRIGHT 2012,2016                        */
+/* Contributors Listed Below - COPYRIGHT 2012,2017                        */
 /* [+] International Business Machines Corp.                              */
 /*                                                                        */
 /*                                                                        */
@@ -163,6 +163,7 @@ public:
         DECONFIGURED_BY_PEC_DECONFIG,           // BASE | 0x13
         DECONFIGURED_BY_NO_CHILD_MCS,           // BASE | 0x14
         DECONFIGURED_BY_NO_PARENT_MCBIST,       // BASE | 0x15
+        DECONFIGURED_BY_DISABLED_PORT,          // BASE | 0x16
 
         // mask - these bits mean it's a PLID and not an enum
         DECONFIGURED_BY_PLID_MASK                = 0xFFFF0000,
diff --git a/src/include/usr/hwas/common/hwas.H b/src/include/usr/hwas/common/hwas.H
index 891af5e3d..138a51887 100644
--- a/src/include/usr/hwas/common/hwas.H
+++ b/src/include/usr/hwas/common/hwas.H
@@ -5,7 +5,7 @@
 /*                                                                        */
 /* OpenPOWER HostBoot Project                                             */
 /*                                                                        */
-/* Contributors Listed Below - COPYRIGHT 2012,2016                        */
+/* Contributors Listed Below - COPYRIGHT 2012,2017                        */
 /* [+] International Business Machines Corp.                              */
 /*                                                                        */
 /*                                                                        */
@@ -201,6 +201,15 @@ void setChipletGardsOnProc(TARGETING::Target * i_procTarget);
  */
 void calculateEffectiveEC();
 
+
+/**
+ * @brief Mark any MCA units that are present but have a disabled port as
+ *        non-functional
+ *
+ * @return error log handle
+ */
+errlHndl_t markDisabledMcas();
+
 };   // end namespace
 
 #endif
diff --git a/src/include/usr/hwas/common/hwasCommon.H b/src/include/usr/hwas/common/hwasCommon.H
index b3fc5e17b..1aa4f12b1 100644
--- a/src/include/usr/hwas/common/hwasCommon.H
+++ b/src/include/usr/hwas/common/hwasCommon.H
@@ -5,7 +5,7 @@
 /*                                                                        */
 /* OpenPOWER HostBoot Project                                             */
 /*                                                                        */
-/* Contributors Listed Below - COPYRIGHT 2012,2016                        */
+/* Contributors Listed Below - COPYRIGHT 2012,2017                        */
 /* [+] International Business Machines Corp.                              */
 /*                                                                        */
 /*                                                                        */
@@ -263,6 +263,37 @@ errlHndl_t platReadPR(const TARGETING::TargetHandle_t &i_target,
 // constants the platReadPR will use for looking at the VPD data
 const uint32_t VPD_VINI_PR_DATA_LENGTH      = 8; //@deprecrated
 
+
+/**
+ * @brief platform specific code to determine the Lx vector of the input
+ *  target. The platform specific code is responsible for returning the
+ *  vector. The caller is responsible for allocating and de-allocating the
+ *  space.
+ *
+ * @param[in]  i_proc      processor target to read the Lx keyword from
+ * @param[in]  i_mca       MCA target indicating which Lx keyword to read
+ * @param[out] o_lxData    pointer to area that will hold the Lx keyword
+ *                          read from VPD; must be malloc'ed by the caller,
+ *                          and must be VPD_CRP0_LX_DATA_LENGTH in size.
+ *
+ * @return errlHndl_t       valid errlHndl_t handle if there was an error
+ *                          NULL if no errors;
+ */
+errlHndl_t platReadLx(const TARGETING::TargetHandle_t &i_proc,
+                      const TARGETING::TargetHandle_t &i_mca,
+                      void *o_lxData);
+
+// constants the platReadLx will use for looking at the VPD data
+const uint32_t VPD_CRP0_LX_HDR_LENGTH       = 1;
+const uint32_t VPD_CRP0_LX_DATA_LENGTH      = 256;
+
+const uint32_t VPD_CRP0_LX_FREQ_INDEP_INDEX = 8;
+const uint32_t VPD_CRP0_LX_PORT_DISABLED    = 0;
+
+const uint8_t  VPD_CRP0_LX_MIN_X            = 1;
+const uint8_t  VPD_CRP0_LX_MAX_X            = 8;
+
+
 /**
  * @brief platform specific code to read the Field Core Override
  *
diff --git a/src/include/usr/hwas/hwasplatreasoncodes.H b/src/include/usr/hwas/hwasplatreasoncodes.H
index 3ffcb9abb..9e6ea9611 100644
--- a/src/include/usr/hwas/hwasplatreasoncodes.H
+++ b/src/include/usr/hwas/hwasplatreasoncodes.H
@@ -5,7 +5,7 @@
 /*                                                                        */
 /* OpenPOWER HostBoot Project                                             */
 /*                                                                        */
-/* Contributors Listed Below - COPYRIGHT 2014,2016                        */
+/* Contributors Listed Below - COPYRIGHT 2014,2017                        */
 /* [+] International Business Machines Corp.                              */
 /*                                                                        */
 /*                                                                        */
@@ -37,6 +37,7 @@ namespace HWAS
         MOD_HOST_DISCOVER_TARGETS = 0x80,
         MOD_PLAT_DECONFIG_GARD    = 0x81,
         MOD_PLAT_READIDEC         = 0x82,
+        MOD_PLAT_READLX           = 0x83,
     };
 
     enum HwasPlatReasonCode
@@ -47,6 +48,7 @@ namespace HWAS
         RC_TARGET_NOT_GARDABLE              = HWAS_COMP_ID | 0x81,
         RC_GARD_REPOSITORY_FULL             = HWAS_COMP_ID | 0x82,
         RC_BAD_CHIPID                       = HWAS_COMP_ID | 0x83,
+        RC_BAD_LX                           = HWAS_COMP_ID | 0x84,
     };
 };
 
diff --git a/src/usr/hwas/common/hwas.C b/src/usr/hwas/common/hwas.C
index fa0c16882..c9955b5fd 100644
--- a/src/usr/hwas/common/hwas.C
+++ b/src/usr/hwas/common/hwas.C
@@ -455,6 +455,15 @@ errlHndl_t discoverTargets()
             break;
         }
 
+        // Mark any MCA units that are present but have a disabled port
+        //  as non-functional
+        errl = markDisabledMcas();
+        if (errl)
+        {
+            HWAS_ERR("discoverTargets: markDisabledMcas failed");
+            break;
+        }
+
         // call invokePresentByAssoc() to obtain functional MCSs, MEMBUFs, and
         // DIMMs for non-direct memory or MCSs, MCAs, and DIMMs for direct
         // memory. Call algorithm function presentByAssoc() to determine
@@ -2449,4 +2458,78 @@ void calculateEffectiveEC()
 
 } //calculateEffectiveEC
 
+errlHndl_t markDisabledMcas()
+{
+    errlHndl_t l_errl = nullptr;
+    uint8_t lxData[HWAS::VPD_CRP0_LX_DATA_LENGTH];
+
+    HWAS_INF("markDisabledMcas entry");
+
+    do
+    {
+        //Get all functional chips
+        TARGETING::TargetHandleList l_procList;
+        getAllChips(l_procList, TYPE_PROC);
+
+        //Loop through all functional procs
+        for(auto l_proc : l_procList)
+        {
+            //Get the functional MCAs for this proc
+            TargetHandleList l_mcaList;
+            getChildChiplets(l_mcaList, l_proc, TYPE_MCA, true);
+
+            for (auto l_mca : l_mcaList)
+            {
+                // fill the Lx data buffer with zeros
+                memset(lxData, 0x00, VPD_CRP0_LX_DATA_LENGTH);
+
+#ifdef __HOSTBOOT_MODULE
+                //@TODO RTC:167294 Need to remove conditional after
+                //                 additional implementation
+                //Read Lx keyword for associated proc and MCA
+                l_errl = platReadLx(l_proc,
+                                    l_mca,
+                                    lxData);
+#endif
+
+                if (l_errl)
+                {
+                    // commit the error but keep going
+                    errlCommit(l_errl, HWAS_COMP_ID);
+                }
+
+                if (lxData[VPD_CRP0_LX_FREQ_INDEP_INDEX
+                                + VPD_CRP0_LX_PORT_DISABLED] != 0)
+                {
+                    // Since port is disabled, MCA is not functional, but
+                    // it's present.
+                    enableHwasState(l_mca,
+                                    true, // present
+                                    false, // not functional
+                                    DeconfigGard::DECONFIGURED_BY_DISABLED_PORT
+                                    );
+                    HWAS_DBG("MCA %.8X - marked present, not functional",
+                             l_mca->getAttr<ATTR_HUID>());
+
+                    TargetInfo l_TargetInfo;
+                    l_TargetInfo.affinityPath =
+                        l_mca->getAttr<ATTR_AFFINITY_PATH>();
+                    l_TargetInfo.pThisTarget = l_mca;
+                    l_TargetInfo.type = l_mca->getAttr<ATTR_TYPE>();
+                    l_TargetInfo.reason =
+                        DeconfigGard::DECONFIGURED_BY_DISABLED_PORT;
+
+                    // Deconfigure child targets for this MCA
+                    deconfigPresentByAssoc(l_TargetInfo);
+                }
+            }
+        }
+
+    }while(0);
+
+    HWAS_INF("markDisabledMcas exit");
+    return l_errl;
+
+} //markDisabledMcas
+
 };   // end namespace
diff --git a/src/usr/hwas/hwasPlat.C b/src/usr/hwas/hwasPlat.C
index 38618e3e7..57a2e9ef6 100644
--- a/src/usr/hwas/hwasPlat.C
+++ b/src/usr/hwas/hwasPlat.C
@@ -303,6 +303,82 @@ errlHndl_t platReadPR(const TargetHandle_t &i_target,
 } // platReadPR
 
 //******************************************************************************
+// platReadLx function
+//******************************************************************************
+errlHndl_t platReadLx(const TargetHandle_t &i_proc,
+                      const TargetHandle_t &i_mca,
+                      void *o_lxData)
+{
+    errlHndl_t errl = nullptr;
+    uint8_t l_chip_unit = i_mca->getAttr<TARGETING::ATTR_CHIP_UNIT>();
+    uint8_t l_x = VPD_CRP0_LX_MIN_X + l_chip_unit;
+
+    HWAS_DBG( "i_proc %.8X, i_mca %.8X, Lx = L%1d",
+              i_proc->getAttr<ATTR_HUID>(),
+              i_mca->getAttr<ATTR_HUID>(),
+              l_x);
+
+    //Look for an invalid x value
+    if( l_x > VPD_CRP0_LX_MAX_X)
+    {
+        HWAS_ERR("Invalid Lx with x=%1d for MCA %.8X on %.8X",
+                 l_x,
+                 i_mca->getAttr<ATTR_HUID>(),
+                 i_proc->getAttr<ATTR_HUID>());
+        /*@
+         * @errortype    ERRORLOG::ERRL_SEV_UNRECOVERABLE
+         * @moduleid     HWAS::MOD_PLAT_READLX
+         * @reasoncode   HWAS::RC_BAD_LX
+         * @userdata1[0:31]     Target proc HUID
+         * @userdata1[32:63]    Target MCA HUID
+         * @userdata2    Value of x for Lx keyword
+         * @devdesc      platReadLx> Invalid Lx keyword
+         */
+        errl = new ERRORLOG::ErrlEntry(
+                                       ERRORLOG::ERRL_SEV_UNRECOVERABLE,
+                                       HWAS::MOD_PLAT_READLX,
+                                       HWAS::RC_BAD_LX,
+                                       TWO_UINT32_TO_UINT64(
+                                           TARGETING::get_huid(i_proc),
+                                           TARGETING::get_huid(i_mca)),
+                                       l_x);
+
+        // make code the highest callout
+        errl->addProcedureCallout(HWAS::EPUB_PRC_HB_CODE,
+                                  HWAS::SRCI_PRIORITY_HIGH);
+    }
+
+    if (errl == nullptr)
+    {   // no error, so we got a valid chip unit value back
+        // call deviceRead() to find the Lx record
+        uint8_t lxRaw[VPD_CRP0_LX_HDR_LENGTH + VPD_CRP0_LX_DATA_LENGTH];
+        size_t lxSize = sizeof(lxRaw);
+
+        errl = deviceRead(i_proc, lxRaw, lxSize,
+                          DEVICE_MVPD_ADDRESS(MVPD::CRP0,
+                                              MVPD::L1 + l_chip_unit));
+
+        if (errl != nullptr)
+        {   // trace the error condition
+            HWAS_INF( "i_proc %.8X, i_mca %.8X - failed L%1d read",
+                      i_proc->getAttr<ATTR_HUID>(),
+                      i_mca->getAttr<ATTR_HUID>(),
+                      l_x);
+        }
+        else
+        {
+            // skip past the header
+            void *lxData = static_cast<void *>(&lxRaw[0]);
+            HWAS_DBG_BIN("Lx record", lxData, VPD_CRP0_LX_DATA_LENGTH);
+            // copy the data back into the caller's buffer
+            memcpy(o_lxData, lxData, VPD_CRP0_LX_DATA_LENGTH);
+        }
+    }
+
+    return errl;
+} // platReadLx
+
+//******************************************************************************
 // platGetFCO function
 //******************************************************************************
 errlHndl_t platGetFCO(const TargetHandle_t &i_node,
diff --git a/src/usr/hwas/test/hwas1test.H b/src/usr/hwas/test/hwas1test.H
index 7dd7b70ed..cf7461623 100644
--- a/src/usr/hwas/test/hwas1test.H
+++ b/src/usr/hwas/test/hwas1test.H
@@ -1627,6 +1627,45 @@ public:
 
         TS_INFO( "testHWASpervStates exit" );
     }
+
+    /**
+     * @brief   test platReadLx
+     */
+    void testHWASplatReadLx()
+    {
+        using namespace TARGETING;
+
+        TS_INFO( "testHWASplatReadLx entry" );
+
+        // call platReadLx with target that isn't in the VPD
+        errlHndl_t l_errl;
+
+        Target* pSys;
+        targetService().getTopLevelTarget(pSys);
+
+        TARGETING::TargetHandleList l_procList;
+        getAllChips(l_procList, TYPE_PROC);
+
+        TargetHandleList l_mcaList;
+        getChildChiplets(l_mcaList, *(l_procList.begin()), TYPE_MCA, true);
+
+        uint8_t lxData[HWAS::VPD_CRP0_LX_DATA_LENGTH];
+
+        l_errl = HWAS::platReadLx(pSys, *(l_mcaList.begin()), lxData);
+
+        if (l_errl)
+        {
+            // error log is expected case, delete it
+            delete l_errl;
+        }
+        else
+        {
+            TS_FAIL("testHWASplatReadLx>"
+                    "No error from platReadLx(pSys, *(l_mcaList.begin()).");
+        }
+
+        TS_INFO( "testHWASplatReadLx exit" );
+    }
 };
 
 #endif //