From d7659ea8c1f3025bbeaca670aec7b2e584571750 Mon Sep 17 00:00:00 2001
From: Gregory Burd <greg@basho.com>
Date: Sun, 4 Dec 2011 14:45:30 -0500
Subject: [PATCH] Add BDB failchk support which requires thread_id and is_alive
 (implement thread_id_string while we're at it).

---
 c_src/bdberl_drv.c   | 26 ++++++++++++++++++++++++-
 c_src/bdberl_tpool.c | 46 ++++++++++++++++++++++++++++++++++++++++++++
 2 files changed, 71 insertions(+), 1 deletion(-)

diff --git a/c_src/bdberl_drv.c b/c_src/bdberl_drv.c
index aa16871..721f3e2 100644
--- a/c_src/bdberl_drv.c
+++ b/c_src/bdberl_drv.c
@@ -263,6 +263,9 @@ DRIVER_INIT(bdberl_drv)
         DB_RECOVER |            /* Enable support for recovering from failures */
         DB_CREATE |             /* Create files as necessary */
         DB_REGISTER |           /* Run recovery if needed */
+        DB_FAILCHK |            /* Release any database reads locks held by the
+                                   thread of control that exited and, if needed,
+                                   abort unresolved transaction. */
         DB_USE_ENVIRON |        /* Use DB_HOME environment variable */
         DB_THREAD;              /* Make the environment free-threaded */
 
@@ -285,9 +288,30 @@ DRIVER_INIT(bdberl_drv)
     }
     else
     {
+        // DB should use the safe allocation routines provided by the Erlang VM
+        DBG("G_DB_ENV->set_alloc(%p, ...)", &G_DB_ENV);
         G_DB_ENV_ERROR = G_DB_ENV->set_alloc(G_DB_ENV, driver_alloc, driver_realloc, driver_free);
-        DBG(" = %d", G_DB_ENV_ERROR);
+        DBG(" = %d\n", G_DB_ENV_ERROR);
 
+        // Inform DB of the number of threads that will be operating on the DB Environment
+        unsigned int nthreads = G_NUM_GENERAL_THREADS + G_NUM_TXN_THREADS;
+        DBG("G_DB_ENV->set_thread_count(%p, %d, ...)", &G_DB_ENV, nthreads);
+        G_DB_ENV_ERROR = G_DB_ENV->set_thread_count(G_DB_ENV, nthreads);
+        DBG(" = %d\n", G_DB_ENV_ERROR);
+
+        DBG("G_DB_ENV->set_thread_id(%p, ...)", &G_DB_ENV);
+        G_DB_ENV_ERROR = G_DB_ENV->set_thread_id(G_DB_ENV, &bdberl_tpool_thread_id);
+        DBG(" = %d\n", G_DB_ENV_ERROR);
+
+        DBG("G_DB_ENV->set_thread_id_string(%p, ...)", &G_DB_ENV);
+        G_DB_ENV_ERROR = G_DB_ENV->set_thread_id_string(G_DB_ENV, &bdberl_tpool_thread_id_string);
+        DBG(" = %d\n", G_DB_ENV_ERROR);
+
+        DBG("G_DB_ENV->set_is_alive(%p, ...)", &G_DB_ENV);
+        G_DB_ENV_ERROR = G_DB_ENV->set_isalive(G_DB_ENV, &bdberl_tpool_thread_is_alive);
+        DBG(" = %d\n", G_DB_ENV_ERROR);
+
+        // Open the DB Environment
         DBG("G_DB_ENV->open(%p, 0, %08X, 0)", &G_DB_ENV, flags);
         G_DB_ENV_ERROR = G_DB_ENV->open(G_DB_ENV, 0, flags, 0);
         DBG(" = %d\n", G_DB_ENV_ERROR);
diff --git a/c_src/bdberl_tpool.c b/c_src/bdberl_tpool.c
index 42129e7..e0210c7 100644
--- a/c_src/bdberl_tpool.c
+++ b/c_src/bdberl_tpool.c
@@ -25,11 +25,19 @@
  * THE SOFTWARE.
  *
  * ------------------------------------------------------------------- */
+
+#include <db.h>
+#include "bdberl_drv.h"
 #include "bdberl_tpool.h"
 
 #include <stdio.h>
 #include <stdlib.h>
 #include <string.h>
+#include <pthread.h>
+#include <signal.h>
+#include <errno.h>
+#include <sys/stat.h>
+#include <sys/types.h>
 
 static void* bdberl_tpool_main(void* tpool);
 static TPoolJob* next_job(TPool* tpool);
@@ -356,3 +364,41 @@ void bdberl_tpool_job_count(TPool* tpool, unsigned int *pending_count_ptr,
     *active_count_ptr = tpool->active_job_count;
     UNLOCK(tpool);
 }
+
+// Returns a unique identifier pair for the current thread of control
+void bdberl_tpool_thread_id(DB_ENV *env, pid_t *pid, db_threadid_t *tid)
+{
+    if (pid)
+        *pid = getpid();
+    if (tid)
+      *tid = (db_threadid_t)pthread_self();
+}
+
+char *bdberl_tpool_thread_id_string(DB_ENV *dbenv, pid_t pid, db_threadid_t tid, char *buf)
+{
+    snprintf(buf, DB_THREADID_STRLEN, "[pid:%08X/tid:%08X]", (unsigned int)pid, (unsigned int)tid);
+    return buf;
+}
+
+// Returns non-zero if the thread of control, identified by the pid and tid arguments,
+// is still running.
+// If DB_MUTEX_PROCESS_ONLY is set in flags then return only if the process (pid) is
+// alive, ignore the thread ID.
+int bdberl_tpool_thread_is_alive(DB_ENV *dbenv, pid_t pid, db_threadid_t tid, u_int32_t flags)
+{
+    static char path[200];
+    static struct stat sb;
+    int alive = 0;
+
+    snprintf(path, 200, "/dev/%d/status", pid);
+    if (stat(path, &sb))
+    {
+        if (flags & DB_MUTEX_PROCESS_ONLY)
+            alive = 1;
+        else
+            if (pthread_kill(tid, 0) != ESRCH)
+              alive = 1;
+    }
+    DBG("bdberl_tpool_thread_is_alive(%08X, %08X, %d) = %d\n", pid, tid, flags, alive);
+    return alive;
+}