MINOR: activity/cli: optionally support sorting by address on "show profiling"

"show profiling" by default sorts by usage/counts, which is suitable for
occasional use. But when called from scripts to monitor/search variations,
this is not very convenient. Let's add a new "byaddr" option to support
sorting the output by address. It also eases matching alloc/free calls
from within a same library, or reading grouped tasks costs by library.

diff --git a/doc/management.txt b/doc/management.txt
index a1c5b4c9050fcea3cc2edda9dfabf6f4b7811cf9..2307d72e80b1b37a40b460d04ee55fdc37cbe501 100644 (file)
--- a/doc/management.txt
+++ b/doc/management.txt
@@ -2485,7 +2485,7 @@ show pools
   as the SIGQUIT when running in foreground except that it does not flush
   the pools.
 
-show profiling [{all | status | tasks | memory}] [max_lines]
+show profiling [{all | status | tasks | memory}] [byaddr] [<max_lines>]
   Dumps the current profiling settings, one per line, as well as the command
   needed to change them. When tasks profiling is enabled, some per-function
   statistics collected by the scheduler will also be emitted, with a summary
@@ -2495,10 +2495,11 @@ show profiling [{all | status | tasks | memory}] [max_lines]
   limit the dump to only the profiling status, the tasks, or the memory
   profiling by specifying the respective keywords; by default all profiling
   information are dumped. It is also possible to limit the number of lines
-  of output of each category by specifying a numeric limit. Please note that
-  profiling is essentially aimed at developers since it gives hints about
-  where CPU cycles or memory are wasted in the code. There is nothing useful
-  to monitor there.
+  of output of each category by specifying a numeric limit. If is possible to
+  request that the output is sorted by address instead of usage, e.g. to ease
+  comparisons between subsequent calls. Please note that profiling is
+  essentially aimed at developers since it gives hints about where CPU cycles
+  or memory are wasted in the code. There is nothing useful to monitor there.
 
 show resolvers [<resolvers section id>]
   Dump statistics for the given resolvers section, or all resolvers sections

diff --git a/src/activity.c b/src/activity.c
index 9d697e717a90d3f601151da2060be4a4d27ef221..5994e98d98a72a6635a583d7dbb21fbbfea553c1 100644 (file)
--- a/src/activity.c
+++ b/src/activity.c
@@ -466,7 +466,7 @@ static int cli_parse_set_profiling(char **args, char *payload, struct appctx *ap
        return 1;
 }
 
-static int cmp_sched_activity(const void *a, const void *b)
+static int cmp_sched_activity_calls(const void *a, const void *b)
 {
        const struct sched_activity *l = (const struct sched_activity *)a;
        const struct sched_activity *r = (const struct sched_activity *)b;
@@ -479,6 +479,19 @@ static int cmp_sched_activity(const void *a, const void *b)
                return 0;
 }
 
+static int cmp_sched_activity_addr(const void *a, const void *b)
+{
+       const struct sched_activity *l = (const struct sched_activity *)a;
+       const struct sched_activity *r = (const struct sched_activity *)b;
+
+       if (l->func > r->func)
+               return -1;
+       else if (l->func < r->func)
+               return 1;
+       else
+               return 0;
+}
+
 #if USE_MEMORY_PROFILING
 /* used by qsort below */
 static int cmp_memprof_stats(const void *a, const void *b)
@@ -493,6 +506,19 @@ static int cmp_memprof_stats(const void *a, const void *b)
        else
                return 0;
 }
+
+static int cmp_memprof_addr(const void *a, const void *b)
+{
+       const struct memprof_stats *l = (const struct memprof_stats *)a;
+       const struct memprof_stats *r = (const struct memprof_stats *)b;
+
+       if (l->caller > r->caller)
+               return -1;
+       else if (l->caller < r->caller)
+               return 1;
+       else
+               return 0;
+}
 #endif // USE_MEMORY_PROFILING
 
 /* This function dumps all profiling settings. It returns 0 if the output
@@ -506,6 +532,9 @@ static int cmp_memprof_stats(const void *a, const void *b)
  *       restart line for each step (starts at zero)
  *    ctx.cli.o0:
  *       may contain a configured max line count for each step (0=not set)
+ *    ctx.cli.o1:
+ *       0: sort by usage
+ *       1: sort by address
  */
 static int cli_io_handler_show_profiling(struct appctx *appctx)
 {
@@ -556,7 +585,10 @@ static int cli_io_handler_show_profiling(struct appctx *appctx)
                goto skip_tasks;
 
        memcpy(tmp_activity, sched_activity, sizeof(tmp_activity));
-       qsort(tmp_activity, 256, sizeof(tmp_activity[0]), cmp_sched_activity);
+       if (appctx->ctx.cli.o1)
+               qsort(tmp_activity, 256, sizeof(tmp_activity[0]), cmp_sched_activity_addr);
+       else
+               qsort(tmp_activity, 256, sizeof(tmp_activity[0]), cmp_sched_activity_calls);
 
        if (!appctx->ctx.cli.i1)
                chunk_appendf(&trash, "Tasks activity:\n"
@@ -612,7 +644,10 @@ static int cli_io_handler_show_profiling(struct appctx *appctx)
                goto skip_mem;
 
        memcpy(tmp_memstats, memprof_stats, sizeof(tmp_memstats));
-       qsort(tmp_memstats, MEMPROF_HASH_BUCKETS+1, sizeof(tmp_memstats[0]), cmp_memprof_stats);
+       if (appctx->ctx.cli.o1)
+               qsort(tmp_memstats, MEMPROF_HASH_BUCKETS+1, sizeof(tmp_memstats[0]), cmp_memprof_addr);
+       else
+               qsort(tmp_memstats, MEMPROF_HASH_BUCKETS+1, sizeof(tmp_memstats[0]), cmp_memprof_stats);
 
        if (!appctx->ctx.cli.i1)
                chunk_appendf(&trash,
@@ -685,36 +720,39 @@ static int cli_io_handler_show_profiling(struct appctx *appctx)
        return 1;
 }
 
-/* parse a "show profiling" command. It returns 1 on failure, 0 if it starts to dump. */
+/* parse a "show profiling" command. It returns 1 on failure, 0 if it starts to dump.
+ *  - cli.i0 is set to the first state (0=all, 4=status, 5=tasks, 6=memory)
+ *  - cli.o1 is set to 1 if the output must be sorted by addr instead of usage
+ *  - cli.o0 is set to the number of lines of output
+ */
 static int cli_parse_show_profiling(char **args, char *payload, struct appctx *appctx, void *private)
 {
+       int arg;
+
        if (!cli_has_level(appctx, ACCESS_LVL_ADMIN))
                return 1;
 
-       if (strcmp(args[2], "all") == 0) {
-               appctx->ctx.cli.i0 = 0; // will cycle through 0,1,2; default
-               args++;
-       }
-       else if (strcmp(args[2], "status") == 0) {
-               appctx->ctx.cli.i0 = 4; // will visit status only
-               args++;
-       }
-       else if (strcmp(args[2], "tasks") == 0) {
-               appctx->ctx.cli.i0 = 5; // will visit tasks only
-               args++;
-       }
-       else if (strcmp(args[2], "memory") == 0) {
-               appctx->ctx.cli.i0 = 6; // will visit memory only
-               args++;
-       }
-       else if (*args[2] && !isdigit((unsigned char)*args[2]))
-               return cli_err(appctx, "Expects either 'all', 'status', 'tasks' or 'memory'.\n");
-
-       if (*args[2]) {
-               /* Second arg may set a limit to number of entries to dump; default is
-                * not set and means no limit.
-                */
-               appctx->ctx.cli.o0 = atoi(args[2]);
+       for (arg = 2; *args[arg]; arg++) {
+               if (strcmp(args[arg], "all") == 0) {
+                       appctx->ctx.cli.i0 = 0; // will cycle through 0,1,2; default
+               }
+               else if (strcmp(args[arg], "status") == 0) {
+                       appctx->ctx.cli.i0 = 4; // will visit status only
+               }
+               else if (strcmp(args[arg], "tasks") == 0) {
+                       appctx->ctx.cli.i0 = 5; // will visit tasks only
+               }
+               else if (strcmp(args[arg], "memory") == 0) {
+                       appctx->ctx.cli.i0 = 6; // will visit memory only
+               }
+               else if (strcmp(args[arg], "byaddr") == 0) {
+                       appctx->ctx.cli.o1 = 1; // sort output by address instead of usage
+               }
+               else if (isdigit((unsigned char)*args[arg])) {
+                       appctx->ctx.cli.o0 = atoi(args[arg]); // number of entries to dump
+               }
+               else
+                       return cli_err(appctx, "Expects either 'all', 'status', 'tasks', 'memory', 'byaddr' or a max number of output lines.\n");
        }
        return 0;
 }
@@ -825,7 +863,7 @@ static int cli_io_handler_show_tasks(struct appctx *appctx)
        for (i = 0; i < 256; i++)
                tot_calls += tmp_activity[i].calls;
 
-       qsort(tmp_activity, 256, sizeof(tmp_activity[0]), cmp_sched_activity);
+       qsort(tmp_activity, 256, sizeof(tmp_activity[0]), cmp_sched_activity_calls);
 
        chunk_appendf(&trash, "Running tasks: %d (%d threads)\n"
                      "  function                     places     %%    lat_tot   lat_avg\n",
@@ -875,7 +913,7 @@ INITCALL1(STG_REGISTER, cfg_register_keywords, &cfg_kws);
 /* register cli keywords */
 static struct cli_kw_list cli_kws = {{ },{
        { { "set",  "profiling", NULL }, "set profiling <what> {auto|on|off}      : enable/disable resource profiling (tasks,memory)", cli_parse_set_profiling,  NULL },
-       { { "show", "profiling", NULL }, "show profiling [<what>] [<max_lines>]   : show profiling state (all,status,tasks,memory)",   cli_parse_show_profiling, cli_io_handler_show_profiling, NULL },
+       { { "show", "profiling", NULL }, "show profiling [<what>|<#lines>|byaddr]*: show profiling state (all,status,tasks,memory)",   cli_parse_show_profiling, cli_io_handler_show_profiling, NULL },
        { { "show", "tasks", NULL },     "show tasks                              : show running tasks",                               NULL, cli_io_handler_show_tasks,     NULL },
        {{},}
 }};