Rediscover I2C Targets after Host I2C Reset

It's possible that some targets that use I2C to detect if they are
present  might not be present after a system re-IPLs and performs a
FSI I2C Reset.  Therefore, later in the IPL, after a Host I2C Reset
sequence an attempt to rediscover any of these targets is made.
If any new targets are found, the system will re-IPL.  Currently
only supporting looking for DIMMs now on OpenPower systems with
securemode enabled.

Change-Id: I010135231f2f74869529a3dbc3344413b6d19dc9
RTC:178973
Backport:release-op920
CQ:SW427365
Reviewed-on: http://ralgit01.raleigh.ibm.com/gerrit1/57744
Tested-by: Jenkins Server <pfd-jenkins+hostboot@us.ibm.com>
Tested-by: FSP CI Jenkins <fsp-CI-jenkins+hostboot@us.ibm.com>
Reviewed-by: Nicholas E. Bofferding <bofferdn@us.ibm.com>
Reviewed-by: ILYA SMIRNOV <ismirno@us.ibm.com>
Tested-by: Jenkins OP Build CI <op-jenkins+hostboot@us.ibm.com>
Tested-by: Jenkins OP HW <op-hw-jenkins+hostboot@us.ibm.com>
Reviewed-by: Daniel M. Crowell <dcrowell@us.ibm.com>

3 files changed, 213 insertions, 8 deletions

diff --git a/src/include/usr/isteps/istep_reasoncodes.H b/src/include/usr/isteps/istep_reasoncodes.H
index c0f30679a..01aa61bca 100644
--- a/src/include/usr/isteps/istep_reasoncodes.H
+++ b/src/include/usr/isteps/istep_reasoncodes.H
@@ -60,6 +60,7 @@ namespace ISTEP
         MOD_SET_IPL_PARMS                                 = 0x1D,
         MOD_OCC_XSTOP_HANDLER                             = 0x1E,
         MOD_VERIFY_AND_MOVE_PAYLOAD                       = 0x1F,
+        MOD_REDISCOVER_I2C_TARGETS                        = 0x20,
     };
 
     /**
@@ -115,6 +116,7 @@ namespace ISTEP
         RC_PREVENT_REBOOT_IN_MFG_TERM_MODE       = ISTEP_COMP_ID | 0x36,
         RC_FAILED_WRITE_SPR                      = ISTEP_COMP_ID | 0x37,
         RC_FAILED_TO_BOOT_SBE                    = ISTEP_COMP_ID | 0x38,
+        RC_REDISCOVERED_TARGETS                  = ISTEP_COMP_ID | 0x39,
     };
 
 };
diff --git a/src/usr/i2c/i2c.C b/src/usr/i2c/i2c.C
index 950f7bbff..5749ba115 100755
--- a/src/usr/i2c/i2c.C
+++ b/src/usr/i2c/i2c.C
@@ -1120,10 +1120,11 @@ bool i2cPresence( TARGETING::Target * i_target,
                         uint64_t i_engine,
                         uint64_t i_devAddr )
 {
-    TRACUCOMP(g_trac_i2c, ENTER_MRK"i2cPresence(): tgt=0x%X: e/p/devAddr="
-              "%d/%d/0x%X", TARGETING::get_huid(i_target), i_engine,
+    TRACUCOMP(g_trac_i2c, ENTER_MRK"i2cPresence(): tgt=0x%X: e%d/p%d/"
+              "devAddr=0x%X", TARGETING::get_huid(i_target), i_engine,
               i_port, i_devAddr );
 
+
     errlHndl_t err = NULL;
     bool l_mutex_success = false;
     bool l_present = false;
@@ -1324,9 +1325,9 @@ bool i2cPresence( TARGETING::Target * i_target,
     if( err )
     {
         TRACFCOMP( g_trac_i2c,
-                   ERR_MRK"i2cPresence() Error!"
-                   "tgt=0x%X",
-                   TARGETING::get_huid(i_target));
+                   ERR_MRK"i2cPresence() Error! "
+                   "tgt=0x%X: e%d/p%d/devAddr=0x%X",
+                   TARGETING::get_huid(i_target), i_engine, i_port, i_devAddr);
         errlCommit(err,
                    I2C_COMP_ID);
 
@@ -1342,8 +1343,8 @@ bool i2cPresence( TARGETING::Target * i_target,
                    args.engine );
     }
 
-    TRACUCOMP(g_trac_i2c, EXIT_MRK"i2cPresence(): tgt=0x%X: e/p/devAddr="
-              "%d/%d/0x%X: l_present=%d",
+    TRACUCOMP(g_trac_i2c, EXIT_MRK"i2cPresence(): tgt=0x%X: e%d/p%d/"
+              "devAddr=0x%X: l_present=%d",
               TARGETING::get_huid(i_target), i_engine, i_port, i_devAddr,
               l_present );
 
diff --git a/src/usr/isteps/istep10/call_host_slave_sbe_update.C b/src/usr/isteps/istep10/call_host_slave_sbe_update.C
index 147749da1..e8d436709 100644
--- a/src/usr/isteps/istep10/call_host_slave_sbe_update.C
+++ b/src/usr/isteps/istep10/call_host_slave_sbe_update.C
@@ -5,7 +5,7 @@
 /*                                                                        */
 /* OpenPOWER HostBoot Project                                             */
 /*                                                                        */
-/* Contributors Listed Below - COPYRIGHT 2015,2016                        */
+/* Contributors Listed Below - COPYRIGHT 2015,2018                        */
 /* [+] International Business Machines Corp.                              */
 /*                                                                        */
 /*                                                                        */
@@ -26,7 +26,10 @@
 #include    <errl/errludtarget.H>
 #include    <errl/errlmanager.H>
 #include    <isteps/hwpisteperror.H>
+#include    <isteps/istep_reasoncodes.H>
 #include    <initservice/isteps_trace.H>
+#include    <initservice/istepdispatcherif.H>
+#include    <initservice/initserviceif.H>
 
 //  targeting support
 #include    <targeting/common/commontargeting.H>
@@ -36,7 +39,17 @@
 #include    <sbe/sbeif.H>
 #include    <pnor/pnorif.H>
 #include    <i2c/i2cif.H>
+#include    <console/consoleif.H>
+#include    <config.h>
+#ifdef CONFIG_BMC_IPMI
+#include    <ipmi/ipmisensor.H>
+#endif
+#include    <sys/time.h>
+#include    <hwas/common/hwasCommon.H>
 
+// Easy macro replace for unit testing
+//#define TRACUCOMP(args...)  TRACFCOMP(args)
+#define TRACUCOMP(args...)
 
 using   namespace   ISTEP_ERROR;
 using   namespace   ISTEP;
@@ -45,6 +58,187 @@ using   namespace   ERRORLOG;
 
 namespace ISTEP_10
 {
+
+errlHndl_t rediscoverI2CTargets(void)
+{
+    errlHndl_t err = nullptr;
+    size_t dimm_count_before = 0;
+    size_t dimm_count_after = 0;
+
+    TRACUCOMP( ISTEPS_TRACE::g_trac_isteps_trace,ENTER_MRK
+               "rediscoverI2CTargets" );
+    do
+    {
+        if ((SECUREBOOT::enabled() == false ) ||
+            (INITSERVICE::spBaseServicesEnabled() == true ))
+        {
+            TRACUCOMP( ISTEPS_TRACE::g_trac_isteps_trace,
+                       "rediscoverI2CTargets: skipping because either security "
+                       "(%d) is not enabled or FSP-based system (%d)",
+                       SECUREBOOT::enabled(),
+                       INITSERVICE::spBaseServicesEnabled());
+
+            break;
+        }
+
+        // find CLASS_SYS (the top level target)
+        Target* pSys = nullptr;
+        targetService().getTopLevelTarget(pSys);
+        assert(pSys, "rediscoverI2CTargets: pSys is nullptr");
+
+        // find list of all we need to call platPresenceDetect against
+        PredicateCTM predDimm(CLASS_LOGICAL_CARD, TYPE_DIMM);
+        PredicatePostfixExpr checkExpr;
+
+        // Look for the ones that are 'not present'
+        PredicateHwas notPresent;
+        notPresent.present(false);
+
+        checkExpr.push(&predDimm).push(&notPresent).And();
+
+        TargetHandleList pDimmList;
+        targetService().getAssociated( pDimmList, pSys,
+            TargetService::CHILD, TargetService::ALL, &checkExpr );
+
+        dimm_count_before = pDimmList.size();
+        if ( dimm_count_before == 0 )
+        {
+            TRACUCOMP( ISTEPS_TRACE::g_trac_isteps_trace,
+                       "rediscoverI2CTargets: pDimmList has size of %d meaning "
+                       "no DIMMs to discover",
+                       dimm_count_before);
+            break;
+        }
+
+        // Pass this list to the hwas platform-specific api to reevaluate if
+        // these targets are now present
+        TRACUCOMP( ISTEPS_TRACE::g_trac_isteps_trace,
+                   "rediscoverI2CTargets: pDimmList size before: %d",
+                   dimm_count_before);
+        err = HWAS::platPresenceDetect(pDimmList);
+        if (err)
+        {
+            TRACFCOMP( ISTEPS_TRACE::g_trac_isteps_trace,
+                       "rediscoverI2CTargets: HWAS::platPresenceDetect "
+                       "returned err plid=0x%X, rc=0x%X",
+                       err->plid(), err->reasonCode());
+            break;
+        }
+
+        // HWAS::platPresenceDetect only keeps present targets, so save that
+        // count here
+        dimm_count_after = pDimmList.size();
+        TRACUCOMP( ISTEPS_TRACE::g_trac_isteps_trace,
+                   "rediscoverI2CTargets: pDimmList size after: %d (was %d)",
+                   dimm_count_after, dimm_count_before);
+
+        if ( dimm_count_after > 0 )
+        {
+            TRACFCOMP( ISTEPS_TRACE::g_trac_isteps_trace,ERR_MRK
+                       "rediscoverI2CTargets: New DIMM(s) detected: pDimmList "
+                       "size before=%d, after=%d. Requesting reboot",
+                       dimm_count_before, dimm_count_after);
+
+            uint32_t huid0 = get_huid(pDimmList.at(0));
+            uint32_t huid1 = 0;
+            if ( dimm_count_after >= 2)
+            {
+                huid1 = get_huid(pDimmList.at(1));
+            }
+
+            // Create info log
+
+            /*@
+             * @errortype
+             * @moduleid         ISTEP::MOD_REDISCOVER_I2C_TARGETS
+             * @reasoncode       ISTEP::RC_REDISCOVERED_TARGETS
+             * @userdata1[0:31]  HUID of 1st rediscovered target
+             * @userdata1[31:63] HUID of 2nd rediscovered target, if applicable
+             * @userdata2[0:31]  Target Count Before Rediscover Attempt
+             * @userdata2[31:63] Target Count After Rediscover Attempt
+             * @devdesc          Targets detected via I2C were rediscovered
+             *                   after an I2C reset.  Reboot to add to config
+             * @custdesc         A problem occurred during the IPL of the
+             *                   system and the system will reboot.
+             */
+            err = new ErrlEntry(ERRL_SEV_INFORMATIONAL,
+                                ISTEP::MOD_REDISCOVER_I2C_TARGETS,
+                                ISTEP::RC_REDISCOVERED_TARGETS,
+                                TWO_UINT32_TO_UINT64(huid0,
+                                                     huid1),
+                                TWO_UINT32_TO_UINT64(dimm_count_before,
+                                                     dimm_count_after));
+            err->collectTrace(ISTEP_COMP_NAME);
+            err->collectTrace(HWAS_COMP_NAME);
+
+            // Add Rediscovered Targets To The Error Log
+            for ( auto l_tgt : pDimmList )
+            {
+                ErrlUserDetailsTarget(l_tgt).addToLog(err);
+            }
+
+            // Commit Error Log here since requesting reboot and not returning
+            // error log to the caller
+            errlCommit(err, ISTEP_COMP_ID);
+
+#ifdef CONFIG_CONSOLE
+            CONSOLE::displayf(ISTEP_COMP_NAME,"Requesting Reboot after DIMM(s) "
+                              "were rediscovered after I2C Reset");
+            CONSOLE::flush();
+#endif
+
+#ifdef CONFIG_BMC_IPMI
+            uint16_t count = SENSOR::DEFAULT_REBOOT_COUNT;
+            SENSOR::RebootCountSensor l_sensor;
+
+            // Set reboot count to default value
+            TRACFCOMP( ISTEPS_TRACE::g_trac_isteps_trace,INFO_MRK
+                       "rediscoverI2CTargets: Writing Reboot Sensor Count=%d",
+                       count);
+
+            auto new_err = l_sensor.setRebootCount( count );
+            if ( new_err )
+            {
+                TRACFCOMP( ISTEPS_TRACE::g_trac_isteps_trace,
+                           ERR_MRK"rediscoverI2CTargets: "
+                           "FAIL Writing Reboot Sensor Count to %d. "
+                           "Committing Error Log rc=0x%.4X eid=0x%.8X "
+                           "plid=0x%.8X, but continuing shutdown",
+                           count,
+                           new_err->reasonCode(),
+                           new_err->eid(),
+                           new_err->plid());
+                new_err->collectTrace(ISTEP_COMP_NAME);
+                errlCommit( new_err, ISTEP_COMP_ID );
+
+                // No Break - Still send chassis power cycle
+            }
+
+            // Initiate a graceful power cycle
+            TRACFCOMP( ISTEPS_TRACE::g_trac_isteps_trace,INFO_MRK
+                   "rediscoverI2CTargets: requesting power cycle");
+            INITSERVICE::requestReboot();
+
+            // sleep here to give IPMI code so that the istep doesn't
+            // continue with SBE Update
+            while (true)
+            {
+                nanosleep(60,0); // 60 seconds
+            }
+#endif
+
+
+       }
+
+    } while(0);
+
+    TRACUCOMP( ISTEPS_TRACE::g_trac_isteps_trace,EXIT_MRK
+               "rediscoverI2CTargets: err plid=0x%X rc=0x%X",
+               ERRL_GETPLID_SAFE(err), ERRL_GETRC_SAFE(err) );
+
+    return err;
+}
+
 void* call_host_slave_sbe_update (void *io_pArgs)
 {
     errlHndl_t  l_errl  =   NULL;
@@ -67,6 +261,14 @@ void* call_host_slave_sbe_update (void *io_pArgs)
             errlCommit( l_errl, HWPF_COMP_ID );
         }
 
+        // Attempt to rediscover any I2C devices after the above reset
+        l_errl = rediscoverI2CTargets();
+        if (l_errl)
+        {
+            // Commit error and keep going
+            errlCommit( l_errl, HWPF_COMP_ID );
+        }
+
         // Call to check state of Processor SBE SEEPROMs and
         // make any necessary updates
         l_errl = SBE::updateProcessorSbeSeeproms();