Branch b1_6

b=16566
o=Jonathan Li(jli@cray.com)
i=shadow, bobijam

Description: Upcall on Lustre log has been dumped
Details    : Allow for a user mode script to be called once a Lustre log has
             been dumped. It passes the filename of the dumped log to the
             script, the location of the script can be specified via
             /proc/sys/lnet/debug_log_upcall.

diff --git a/lnet/libcfs/debug.c b/lnet/libcfs/debug.c
index fa55ae6..fb2fe5c 100644 (file)
--- a/lnet/libcfs/debug.c
+++ b/lnet/libcfs/debug.c
@@ -421,6 +421,9 @@ libcfs_debug_str2mask(int *mask, const char *str, int is_subsys)
         return 0;
 }
 
+/**
+ * Dump Lustre log to ::debug_file_path by calling tracefile_dump_all_pages()
+ */
 void libcfs_debug_dumplog_internal(void *arg)
 {
         CFS_DECL_JOURNAL_DATA;
@@ -434,6 +437,7 @@ void libcfs_debug_dumplog_internal(void *arg)
                 printk(KERN_ALERT "LustreError: dumping log to %s\n",
                        debug_file_name);
                 tracefile_dump_all_pages(debug_file_name);
+                libcfs_run_debug_log_upcall(debug_file_name);
         }
         CFS_POP_JOURNAL;
 }

diff --git a/lnet/libcfs/linux/linux-debug.c b/lnet/libcfs/linux/linux-debug.c
index 7d2592c..e0eade8 100644 (file)
--- a/lnet/libcfs/linux/linux-debug.c
+++ b/lnet/libcfs/linux/linux-debug.c
@@ -79,6 +79,42 @@
 #endif
 
 char lnet_upcall[1024] = "/usr/lib/lustre/lnet_upcall";
+char lnet_debug_log_upcall[1024] = "/usr/lib/lustre/lnet_debug_log_upcall";
+
+/**
+ * Upcall function once a Lustre log has been dumped.
+ *
+ * \param file  path of the dumped log
+ */
+void libcfs_run_debug_log_upcall(char *file)
+{
+        char *argv[3];
+        int   rc;
+        char *envp[] = {
+                "HOME=/",
+                "PATH=/sbin:/bin:/usr/sbin:/usr/bin",
+                NULL};
+        ENTRY;
+
+        argv[0] = lnet_debug_log_upcall;
+
+        LASSERTF(file != NULL, "called on a null filename\n");
+        argv[1] = file; //only need to pass the path of the file
+
+        argv[2] = NULL;
+
+        rc = USERMODEHELPER(argv[0], argv, envp);
+        if (rc < 0 && rc != -ENOENT) {
+                CERROR("Error %d invoking LNET debug log upcall %s %s; "
+                       "check /proc/sys/lnet/debug_log_upcall\n",
+                       rc, argv[0], argv[1]);
+        } else {
+                CDEBUG(D_HA, "Invoked LNET debug log upcall %s %s\n",
+                       argv[0], argv[1]);
+        }
+
+        EXIT;
+}
 
 void libcfs_run_upcall(char **argv)
 {

diff --git a/lnet/libcfs/linux/linux-proc.c b/lnet/libcfs/linux/linux-proc.c
index 1b75f9e..cbada7c 100644 (file)
--- a/lnet/libcfs/linux/linux-proc.c
+++ b/lnet/libcfs/linux/linux-proc.c
@@ -78,6 +78,10 @@
 
 static cfs_sysctl_table_header_t *lnet_table_header = NULL;
 extern char lnet_upcall[1024];
+/**
+ * The path of debug log dump upcall script.
+ */
+extern char lnet_debug_log_upcall[1024];
 
 #define PSDEV_LNET  (0x100)
 enum {
@@ -97,11 +101,12 @@ enum {
         PSDEV_LNET_DUMP_KERNEL,   /* snapshot kernel debug buffer to file */
         PSDEV_LNET_DAEMON_FILE,   /* spool kernel debug buffer to file */
         PSDEV_LNET_DEBUG_MB,      /* size of debug buffer */
+        PSDEV_LNET_DEBUG_LOG_UPCALL, /* debug log upcall script */
 };
 
-static int 
-proc_call_handler(void *data, int write, 
-                  loff_t *ppos, void *buffer, size_t *lenp, 
+static int
+proc_call_handler(void *data, int write,
+                  loff_t *ppos, void *buffer, size_t *lenp,
                   int (*handler)(void *data, int write,
                                  loff_t pos, void *buffer, int len))
 {
@@ -130,7 +135,7 @@ LL_PROC_PROTO(name)                                     \
                                  __##name);             \
 }
 
-static int __proc_dobitmasks(void *data, int write, 
+static int __proc_dobitmasks(void *data, int write,
                              loff_t pos, void *buffer, int nob)
 {
         const int     tmpstrlen = 512;
@@ -176,7 +181,7 @@ static int __proc_dump_kernel(void *data, int write,
 {
         if (!write)
                 return 0;
-        
+
         return trace_dump_debug_buffer_usrstr(buffer, nob);
 }
 
@@ -187,14 +192,14 @@ static int __proc_daemon_file(void *data, int write,
 {
         if (!write) {
                 int len = strlen(tracefile);
-                
+
                 if (pos >= len)
                         return 0;
-                
-                return trace_copyout_string(buffer, nob, 
+
+                return trace_copyout_string(buffer, nob,
                                             tracefile + pos, "\n");
         }
-        
+
         return trace_daemon_command_usrstr(buffer, nob);
 }
 
@@ -210,10 +215,10 @@ static int __proc_debug_mb(void *data, int write,
 
                 if (pos >= len)
                         return 0;
-                
+
                 return trace_copyout_string(buffer, nob, tmpstr + pos, "\n");
         }
-        
+
         return trace_set_debug_mb_usrstr(buffer, nob);
 }
 
@@ -366,7 +371,6 @@ static cfs_sysctl_table_t lnet_table[] = {
                 .mode     = 0644,
                 .proc_handler = &proc_console_backoff
         },
-
         {
                 .ctl_name = PSDEV_DEBUG_PATH,
                 .procname = "debug_path",
@@ -375,7 +379,6 @@ static cfs_sysctl_table_t lnet_table[] = {
                 .mode     = 0644,
                 .proc_handler = &proc_dostring,
         },
-
         {
                 .ctl_name = PSDEV_LNET_UPCALL,
                 .procname = "upcall",
@@ -385,6 +388,14 @@ static cfs_sysctl_table_t lnet_table[] = {
                 .proc_handler = &proc_dostring,
         },
         {
+                .ctl_name = PSDEV_LNET_DEBUG_LOG_UPCALL,
+                .procname = "debug_log_upcall",
+                .data     = lnet_debug_log_upcall,
+                .maxlen   = sizeof(lnet_debug_log_upcall),
+                .mode     = 0644,
+                .proc_handler = &proc_dostring,
+        },
+        {
                 .ctl_name = PSDEV_LNET_MEMUSED,
                 .procname = "memused",
                 .data     = (int *)&libcfs_kmemory.counter,

diff --git a/lnet/libcfs/tracefile.h b/lnet/libcfs/tracefile.h
index 0493063..dd25327 100644 (file)
--- a/lnet/libcfs/tracefile.h
+++ b/lnet/libcfs/tracefile.h
@@ -45,6 +45,8 @@
 extern char      tracefile[TRACEFILE_NAME_SIZE];
 extern long long tracefile_size;
 
+extern void libcfs_run_debug_log_upcall(char *file);
+
 int  tracefile_init_arch(void);
 void tracefile_fini_arch(void);
 
@@ -96,7 +98,7 @@ extern int  trace_max_debug_mb(void);
 
 #define TRACEFILE_SIZE (500 << 20)
 
-/* Size of a buffer for sprinting console messages if we can't get a page 
+/* Size of a buffer for sprinting console messages if we can't get a page
  * from system */
 #define TRACE_CONSOLE_BUFFER_SIZE   1024
 
@@ -125,7 +127,7 @@ union trace_data_union {
 
                /*
                 * Maximal number of pages allowed on ->tcd_pages and
-                * ->tcd_daemon_pages each. 
+                * ->tcd_daemon_pages each.
                 * Always TCD_MAX_PAGES * tcd_pages_factor / 100 in current
                 * implementation.
                 */
@@ -233,7 +235,7 @@ struct trace_page {
         */
        unsigned short   cpu;
        /*
-        * type(context) of this page 
+        * type(context) of this page
         */
        unsigned short   type;
 };