proc: faster /proc/*/status

top(1) opens the following files for every PID:

	/proc/*/stat
	/proc/*/statm
	/proc/*/status

This patch switches /proc/*/status away from seq_printf().
The result is 13.5% speedup.

Benchmark is open("/proc/self/status")+read+close 1.000.000 million times.

				BEFORE
$ perf stat -r 10 taskset -c 3 ./proc-self-status

 Performance counter stats for 'taskset -c 3 ./proc-self-status' (10 runs):

      10748.474301      task-clock (msec)         #    0.954 CPUs utilized            ( +-  0.91% )
                12      context-switches          #    0.001 K/sec                    ( +-  1.09% )
                 1      cpu-migrations            #    0.000 K/sec
               104      page-faults               #    0.010 K/sec                    ( +-  0.45% )
    37,424,127,876      cycles                    #    3.482 GHz                      ( +-  0.04% )
     8,453,010,029      stalled-cycles-frontend   #   22.59% frontend cycles idle     ( +-  0.12% )
     3,747,609,427      stalled-cycles-backend    #  10.01% backend cycles idle       ( +-  0.68% )
    65,632,764,147      instructions              #    1.75  insn per cycle
                                                  #    0.13  stalled cycles per insn  ( +-  0.00% )
    13,981,324,775      branches                  # 1300.773 M/sec                    ( +-  0.00% )
       138,967,110      branch-misses             #    0.99% of all branches          ( +-  0.18% )

      11.263885428 seconds time elapsed                                          ( +-  0.04% )
      ^^^^^^^^^^^^

				AFTER
$ perf stat -r 10 taskset -c 3 ./proc-self-status

 Performance counter stats for 'taskset -c 3 ./proc-self-status' (10 runs):

       9010.521776      task-clock (msec)         #    0.925 CPUs utilized            ( +-  1.54% )
                11      context-switches          #    0.001 K/sec                    ( +-  1.54% )
                 1      cpu-migrations            #    0.000 K/sec                    ( +- 11.11% )
               103      page-faults               #    0.011 K/sec                    ( +-  0.60% )
    32,352,310,603      cycles                    #    3.591 GHz                      ( +-  0.07% )
     7,849,199,578      stalled-cycles-frontend   #   24.26% frontend cycles idle     ( +-  0.27% )
     3,269,738,842      stalled-cycles-backend    #  10.11% backend cycles idle       ( +-  0.73% )
    56,012,163,567      instructions              #    1.73  insn per cycle
                                                  #    0.14  stalled cycles per insn  ( +-  0.00% )
    11,735,778,795      branches                  # 1302.453 M/sec                    ( +-  0.00% )
        98,084,459      branch-misses             #    0.84% of all branches          ( +-  0.28% )

       9.741247736 seconds time elapsed                                          ( +-  0.07% )
       ^^^^^^^^^^^

Link: http://lkml.kernel.org/r/20160806125608.GB1187@p183.telecom.by
Signed-off-by: Alexey Dobriyan <adobriyan@gmail.com>
Cc: Joe Perches <joe@perches.com>
Cc: Andi Kleen <andi@firstfloor.org>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

1 files changed, 47 insertions, 40 deletions

diff --git a/fs/proc/array.c b/fs/proc/array.c
index 88c7de1..5e7d252 100644
--- a/fs/proc/array.c
+++ b/fs/proc/array.c
@@ -186,51 +186,52 @@ static inline void task_state(struct seq_file *m, struct pid_namespace *ns,
 	task_unlock(p);
 	rcu_read_unlock();
 
-	seq_printf(m,
-		"State:\t%s\n"
-		"Tgid:\t%d\n"
-		"Ngid:\t%d\n"
-		"Pid:\t%d\n"
-		"PPid:\t%d\n"
-		"TracerPid:\t%d\n"
-		"Uid:\t%d\t%d\t%d\t%d\n"
-		"Gid:\t%d\t%d\t%d\t%d\n"
-		"FDSize:\t%d\nGroups:\t",
-		get_task_state(p),
-		tgid, ngid, pid_nr_ns(pid, ns), ppid, tpid,
-		from_kuid_munged(user_ns, cred->uid),
-		from_kuid_munged(user_ns, cred->euid),
-		from_kuid_munged(user_ns, cred->suid),
-		from_kuid_munged(user_ns, cred->fsuid),
-		from_kgid_munged(user_ns, cred->gid),
-		from_kgid_munged(user_ns, cred->egid),
-		from_kgid_munged(user_ns, cred->sgid),
-		from_kgid_munged(user_ns, cred->fsgid),
-		max_fds);
-
+	seq_printf(m, "State:\t%s", get_task_state(p));
+
+	seq_puts(m, "\nTgid:\t");
+	seq_put_decimal_ull(m, 0, tgid);
+	seq_puts(m, "\nNgid:\t");
+	seq_put_decimal_ull(m, 0, ngid);
+	seq_puts(m, "\nPid:\t");
+	seq_put_decimal_ull(m, 0, pid_nr_ns(pid, ns));
+	seq_puts(m, "\nPPid:\t");
+	seq_put_decimal_ull(m, 0, ppid);
+	seq_puts(m, "\nTracerPid:\t");
+	seq_put_decimal_ull(m, 0, tpid);
+	seq_puts(m, "\nUid:");
+	seq_put_decimal_ull(m, '\t', from_kuid_munged(user_ns, cred->uid));
+	seq_put_decimal_ull(m, '\t', from_kuid_munged(user_ns, cred->euid));
+	seq_put_decimal_ull(m, '\t', from_kuid_munged(user_ns, cred->suid));
+	seq_put_decimal_ull(m, '\t', from_kuid_munged(user_ns, cred->fsuid));
+	seq_puts(m, "\nGid:");
+	seq_put_decimal_ull(m, '\t', from_kgid_munged(user_ns, cred->gid));
+	seq_put_decimal_ull(m, '\t', from_kgid_munged(user_ns, cred->egid));
+	seq_put_decimal_ull(m, '\t', from_kgid_munged(user_ns, cred->sgid));
+	seq_put_decimal_ull(m, '\t', from_kgid_munged(user_ns, cred->fsgid));
+	seq_puts(m, "\nFDSize:\t");
+	seq_put_decimal_ull(m, 0, max_fds);
+
+	seq_puts(m, "\nGroups:\t");
 	group_info = cred->group_info;
 	for (g = 0; g < group_info->ngroups; g++)
-		seq_printf(m, "%d ",
-			   from_kgid_munged(user_ns, GROUP_AT(group_info, g)));
+		seq_put_decimal_ull(m, g ? ' ' : 0, from_kgid_munged(user_ns, GROUP_AT(group_info, g)));
 	put_cred(cred);
+	/* Trailing space shouldn't have been added in the first place. */
+	seq_putc(m, ' ');
 
 #ifdef CONFIG_PID_NS
 	seq_puts(m, "\nNStgid:");
 	for (g = ns->level; g <= pid->level; g++)
-		seq_printf(m, "\t%d",
-			task_tgid_nr_ns(p, pid->numbers[g].ns));
+		seq_put_decimal_ull(m, '\t', task_tgid_nr_ns(p, pid->numbers[g].ns));
 	seq_puts(m, "\nNSpid:");
 	for (g = ns->level; g <= pid->level; g++)
-		seq_printf(m, "\t%d",
-			task_pid_nr_ns(p, pid->numbers[g].ns));
+		seq_put_decimal_ull(m, '\t', task_pid_nr_ns(p, pid->numbers[g].ns));
 	seq_puts(m, "\nNSpgid:");
 	for (g = ns->level; g <= pid->level; g++)
-		seq_printf(m, "\t%d",
-			task_pgrp_nr_ns(p, pid->numbers[g].ns));
+		seq_put_decimal_ull(m, '\t', task_pgrp_nr_ns(p, pid->numbers[g].ns));
 	seq_puts(m, "\nNSsid:");
 	for (g = ns->level; g <= pid->level; g++)
-		seq_printf(m, "\t%d",
-			task_session_nr_ns(p, pid->numbers[g].ns));
+		seq_put_decimal_ull(m, '\t', task_session_nr_ns(p, pid->numbers[g].ns));
 #endif
 	seq_putc(m, '\n');
 }
@@ -299,11 +300,14 @@ static inline void task_sig(struct seq_file *m, struct task_struct *p)
 		unlock_task_sighand(p, &flags);
 	}
 
-	seq_printf(m, "Threads:\t%d\n", num_threads);
-	seq_printf(m, "SigQ:\t%lu/%lu\n", qsize, qlim);
+	seq_puts(m, "Threads:\t");
+	seq_put_decimal_ull(m, 0, num_threads);
+	seq_puts(m, "\nSigQ:\t");
+	seq_put_decimal_ull(m, 0, qsize);
+	seq_put_decimal_ull(m, '/', qlim);
 
 	/* render them all */
-	render_sigset_t(m, "SigPnd:\t", &pending);
+	render_sigset_t(m, "\nSigPnd:\t", &pending);
 	render_sigset_t(m, "ShdPnd:\t", &shpending);
 	render_sigset_t(m, "SigBlk:\t", &blocked);
 	render_sigset_t(m, "SigIgn:\t", &ignored);
@@ -348,17 +352,20 @@ static inline void task_cap(struct seq_file *m, struct task_struct *p)
 static inline void task_seccomp(struct seq_file *m, struct task_struct *p)
 {
 #ifdef CONFIG_SECCOMP
-	seq_printf(m, "Seccomp:\t%d\n", p->seccomp.mode);
+	seq_puts(m, "Seccomp:\t");
+	seq_put_decimal_ull(m, 0, p->seccomp.mode);
+	seq_putc(m, '\n');
 #endif
 }
 
 static inline void task_context_switch_counts(struct seq_file *m,
 						struct task_struct *p)
 {
-	seq_printf(m,	"voluntary_ctxt_switches:\t%lu\n"
-			"nonvoluntary_ctxt_switches:\t%lu\n",
-			p->nvcsw,
-			p->nivcsw);
+	seq_puts(m, "voluntary_ctxt_switches:\t");
+	seq_put_decimal_ull(m, 0, p->nvcsw);
+	seq_puts(m, "\nnonvoluntary_ctxt_switches:\t");
+	seq_put_decimal_ull(m, 0, p->nivcsw);
+	seq_putc(m, '\n');
 }
 
 static void task_cpus_allowed(struct seq_file *m, struct task_struct *task)