Add more agressive memory allocation calls

If we hit a situation where we can't allocate a page of memory,
this will force a coalesce (defrag) a few times and then
eventually trigger other memory reclamation actions.

Also tweaked a few spots in the kernel to enhance debug:
- add more stops to look at errors (HB_BREAK_ON_ERROR)
- add more backtrace calls
- add a new debug flag to count the extra coalesce calls

Change-Id: Ibac7079a44a12dc61e41304de4c4ae518c206d13
Reviewed-on: http://rchgit01.rchland.ibm.com/gerrit1/71653
Tested-by: Jenkins Server <pfd-jenkins+hostboot@us.ibm.com>
Tested-by: Jenkins OP Build CI <op-jenkins+hostboot@us.ibm.com>
Reviewed-by: Matt Derksen <mderkse1@us.ibm.com>
Reviewed-by: Nicholas E. Bofferding <bofferdn@us.ibm.com>
Tested-by: Jenkins OP HW <op-hw-jenkins+hostboot@us.ibm.com>
Tested-by: FSP CI Jenkins <fsp-CI-jenkins+hostboot@us.ibm.com>
Reviewed-by: Daniel M. Crowell <dcrowell@us.ibm.com>

7 files changed, 43 insertions, 8 deletions

diff --git a/src/build/debug/Hostboot/MemStats.pm b/src/build/debug/Hostboot/MemStats.pm
index 9156d8267..723b76aae 100755
--- a/src/build/debug/Hostboot/MemStats.pm
+++ b/src/build/debug/Hostboot/MemStats.pm
@@ -5,7 +5,7 @@
 #
 # OpenPOWER HostBoot Project
 #
-# Contributors Listed Below - COPYRIGHT 2011,2018
+# Contributors Listed Below - COPYRIGHT 2011,2019
 # [+] International Business Machines Corp.
 #
 #
@@ -75,6 +75,10 @@ sub main
       ::findPointer("PAGEMCNT",
                     "PageManager::cv_coalesce_count");
 
+    my $page_alloc_coal = ::read64
+      ::findPointer("PAGEMACC",
+                    "PageManager::cv_alloc_coalesce_count");
+
     my $big_heap_pages_used = ::read32
       ::findPointer("HEAPMLPC",
                     "HeapManager::cv_largeheap_page_count");
@@ -118,6 +122,7 @@ sub main
     ::userDisplay "    Free pages:              $free_pages\n";
     ::userDisplay "    Free pages Low mark:     $free_min\n";
     ::userDisplay "    Page chunks coalesced:   $page_coal\n";
+    ::userDisplay "    Page alloc coalesces:    $page_alloc_coal\n";
     ::userDisplay "\nHeap:\n";
     ::userDisplay "    Pages used by heap:      $heap_total  ".
                   "(B:$big_heap_pages_used,S:$small_heap_pages_used)\n";
diff --git a/src/include/kernel/pagemgr.H b/src/include/kernel/pagemgr.H
index 2431c613d..7b5deddb4 100644
--- a/src/include/kernel/pagemgr.H
+++ b/src/include/kernel/pagemgr.H
@@ -5,7 +5,7 @@
 /*                                                                        */
 /* OpenPOWER HostBoot Project                                             */
 /*                                                                        */
-/* Contributors Listed Below - COPYRIGHT 2010,2018                        */
+/* Contributors Listed Below - COPYRIGHT 2010,2019                        */
 /* [+] International Business Machines Corp.                              */
 /*                                                                        */
 /*                                                                        */
@@ -200,6 +200,7 @@ class PageManager
 
         static size_t cv_coalesce_count;        //!< running coalesced counter
         static size_t cv_low_page_count;        //!< lowest page count
+        static size_t cv_alloc_coalesce_count;  //!< number of repeat allocs
 
     protected:
 
diff --git a/src/include/usr/debugpointers.H b/src/include/usr/debugpointers.H
index 01de1a0d8..5815761e8 100644
--- a/src/include/usr/debugpointers.H
+++ b/src/include/usr/debugpointers.H
@@ -95,6 +95,7 @@ constexpr uint64_t HEAPMANAGERFREECHUNKS     = 0x4845415043484e4b; //'HEAPCHNK'
 constexpr uint64_t PAGEMANAGER               = 0x504147454d475220; //'PAGEMGR '
 constexpr uint64_t PAGEMANAGERCOALESCECOUNT  = 0x504147454d434e54; //'PAGEMCNT'
 constexpr uint64_t PAGEMANAGERLOWPAGECOUNT   = 0x504147454d4c5043; //'PAGEMLPC'
+constexpr uint64_t PAGEMANAGERALLOCCOUNT     = 0x504147454d414343; //'PAGEMACC'
 constexpr uint64_t SEGMENTMANAGER            = 0x53474d4e544d4752; //'SGMNTMGR'
 constexpr uint64_t BLOCKREADONLYEVICT        = 0x424c4f434b524f45; //'BLOCKROE'
 constexpr uint64_t BLOCKREADWRITEEVICT       = 0x424c4f434b525745; //'BLOCKRWE'
diff --git a/src/kernel/cpumgr.C b/src/kernel/cpumgr.C
index 425cc2d28..87c5251c7 100644
--- a/src/kernel/cpumgr.C
+++ b/src/kernel/cpumgr.C
@@ -553,7 +553,7 @@ void CpuManager::critAssert(uint64_t i_failAddr)
         {
             // print status to the console.
             printk("TI initiated on all threads (crit_assert)\n");
-
+            MAGIC_INSTRUCTION(MAGIC_BREAK_ON_ERROR);
         }
 
         void activeMainWork()
diff --git a/src/kernel/exception.C b/src/kernel/exception.C
index d3a1dbe6b..938d57515 100644
--- a/src/kernel/exception.C
+++ b/src/kernel/exception.C
@@ -5,7 +5,7 @@
 /*                                                                        */
 /* OpenPOWER HostBoot Project                                             */
 /*                                                                        */
-/* Contributors Listed Below - COPYRIGHT 2010,2018                        */
+/* Contributors Listed Below - COPYRIGHT 2010,2019                        */
 /* [+] International Business Machines Corp.                              */
 /*                                                                        */
 /*                                                                        */
@@ -150,6 +150,7 @@ void kernel_execute_inst_storage()
     {
         printk("Inst Storage exception on %d: %lx, %lx\n",
                t->tid, getSRR0(), getSRR1());
+        KernelMisc::printkBacktrace(t);
         MAGIC_INSTRUCTION(MAGIC_BREAK_ON_ERROR);
         TaskManager::endTask(t, NULL, TASK_STATUS_CRASHED);
     }
diff --git a/src/kernel/msghandler.C b/src/kernel/msghandler.C
index 1f22a7a70..8e7244d2b 100644
--- a/src/kernel/msghandler.C
+++ b/src/kernel/msghandler.C
@@ -5,7 +5,7 @@
 /*                                                                        */
 /* OpenPOWER HostBoot Project                                             */
 /*                                                                        */
-/* Contributors Listed Below - COPYRIGHT 2011,2017                        */
+/* Contributors Listed Below - COPYRIGHT 2011,2019                        */
 /* [+] International Business Machines Corp.                              */
 /*                                                                        */
 /*                                                                        */
@@ -32,6 +32,8 @@
 #include <kernel/taskmgr.H>
 #include <kernel/console.H>
 #include <kernel/doorbell.H>
+#include <kernel/misc.H>
+
 
 void MessageHandler::sendMessage(msg_sys_types_t i_type, void* i_key,
                                  void* i_data, task_t* i_task)
@@ -181,6 +183,8 @@ int MessageHandler::recvMessage(msg_t* i_msg)
             printk("Unhandled msg rc %d (%s) for key %p on task %d @ %p\n",
                     msg_rc, ErrnoToString(msg_rc), key, deferred_task->tid,
                     deferred_task->context.nip);
+            KernelMisc::printkBacktrace(deferred_task);
+            MAGIC_INSTRUCTION(MAGIC_BREAK_ON_ERROR);
             endTaskList.insert(deferred_task);
         }
         else if (CONTINUE_DEFER == rc)
diff --git a/src/kernel/pagemgr.C b/src/kernel/pagemgr.C
index bfc1bbf7a..ee3a3ff39 100644
--- a/src/kernel/pagemgr.C
+++ b/src/kernel/pagemgr.C
@@ -5,7 +5,7 @@
 /*                                                                        */
 /* OpenPOWER HostBoot Project                                             */
 /*                                                                        */
-/* Contributors Listed Below - COPYRIGHT 2010,2018                        */
+/* Contributors Listed Below - COPYRIGHT 2010,2019                        */
 /* [+] International Business Machines Corp.                              */
 /*                                                                        */
 /*                                                                        */
@@ -39,11 +39,12 @@
 #include <kernel/bltohbdatamgr.H>
 #include <kernel/misc.H>
 #include <usr/debugpointers.H>
+#include <kernel/cpumgr.H>
 
 
 size_t PageManager::cv_coalesce_count = 0;
 size_t PageManager::cv_low_page_count = -1;
-
+size_t PageManager::cv_alloc_coalesce_count = 0;
 
 void PageManagerCore::addMemory( size_t i_addr, size_t i_pageCount )
 {
@@ -165,10 +166,11 @@ void* PageManager::allocatePage(size_t n, bool userspace)
 
             // Didn't successfully allocate, so yield in hopes that memory
             // will eventually free up (ex. VMM flushes).
+            constexpr size_t MAX_ATTEMPTS = 10000;
             if (NULL == page)
             {
                 l_attempts++;
-                if( l_attempts == 10000 )
+                if( l_attempts == MAX_ATTEMPTS )
                 {
                     printk( "Cannot allocate %ld pages to %d!\n",
                             n, task_gettid() );
@@ -176,7 +178,25 @@ void* PageManager::allocatePage(size_t n, bool userspace)
                     KernelMisc::printkBacktrace(nullptr);
                     task_crash();
                 }
+
                 task_yield();
+
+                // Force a defrag of the memory 10 times
+                if( l_attempts % (MAX_ATTEMPTS/10) == 0 )
+                {
+                    printkd( "Forcing coalesce to allocate %ld pages to %d!\n",
+                             n, task_gettid() );
+                    coalesce();
+                    ++PageManager::cv_alloc_coalesce_count;
+                }
+
+                // Try to evict some pages once
+                if( l_attempts == MAX_ATTEMPTS/2 )
+                {
+                    printkd( "Forcing periodics to allocate %ld pages to %d!\n",
+                             n, task_gettid() );
+                    CpuManager::forceMemoryPeriodic();
+                }
             }
         }
     }
@@ -496,4 +516,7 @@ void PageManager::_addDebugPointers()
     DEBUG::add_debug_pointer(DEBUG::PAGEMANAGERCOALESCECOUNT,
                              &PageManager::cv_coalesce_count,
                              sizeof(PageManager::cv_coalesce_count));
+    DEBUG::add_debug_pointer(DEBUG::PAGEMANAGERALLOCCOUNT,
+                             &PageManager::cv_alloc_coalesce_count,
+                             sizeof(PageManager::cv_alloc_coalesce_count));
 }