mirror of https://lore.kernel.org/lkml/
 help / color / mirror / Atom feed
From: Calvin Owens <calvinowens@fb.com>
To: Andrew Morton <akpm@linux-foundation.org>,
	Alexey Dobriyan <adobriyan@gmail.com>,
	Oleg Nesterov <oleg@redhat.com>,
	"Eric W. Biederman" <ebiederm@xmission.com>,
	Al Viro <viro@zeniv.linux.org.uk>,
	"Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>,
	Peter Feiner <pfeiner@google.com>,
	Grant Likely <grant.likely@secretlab.ca>
Cc: Siddhesh Poyarekar <siddhesh.poyarekar@gmail.com>,
	<linux-kernel@vger.kernel.org>, <kernel-team@fb.com>,
	<calvinowens@fb.com>
Subject: Re: [RFC][PATCH] procfs: Add /proc/<pid>/mapped_files
Date: Tue, 13 Jan 2015 16:23:41 -0800	[thread overview]
Message-ID: <20150114002341.GA6469@mail.thefacebook.com> (raw)
In-Reply-To: <1421194829-28696-1-git-send-email-calvinowens@fb.com>

Here's a simple program to trigger the issue with /proc/<pid>/maps.

Thanks,
Calvin

/* Simple program to reproduce O(N^2) behavior reading /proc/<pid>/maps
 *
 * Example on a random server:
 *
 * 	$ ./map_repro 0
 * 	Spawning 0 threads
 * 	Reading /proc/self/maps... read 2189 bytes in 1 syscalls in 33us!
 * 	$ ./map_repro 10
 * 	Spawning 10 threads
 * 	Reading /proc/self/maps... read 3539 bytes in 1 syscalls in 55us!
 * 	$ ./map_repro 100
 * 	Spawning 100 threads
 * 	Reading /proc/self/maps... read 15689 bytes in 4 syscalls in 373us!
 * 	$ ./map_repro 1000
 * 	Spawning 1000 threads
 * 	Reading /proc/self/maps... read 137189 bytes in 34 syscalls in 32376us!
 * 	$ ./map_repro 2000
 * 	Spawning 2000 threads
 * 	Reading /proc/self/maps... read 272189 bytes in 68 syscalls in 119980us!
 * 	$ ./map_repro 4000
 * 	Spawning 4000 threads
 * 	Reading /proc/self/maps... read 544912 bytes in 134 syscalls in 712200us!
 * 	$ ./map_repro 8000
 * 	Spawning 8000 threads
 * 	Reading /proc/self/maps... read 1090189 bytes in 268 syscalls in 3650718us!
 * 	$ ./map_repro 16000
 * 	Spawning 16000 threads
 * 	Reading /proc/self/maps... read 2178189 bytes in 534 syscalls in 42701311us!
 */

#include <stdlib.h>
#include <stdio.h>
#include <errno.h>
#include <string.h>
#include <limits.h>
#include <pthread.h>
#include <unistd.h>
#include <time.h>
#include <fcntl.h>

static char buf[65536] = {0};
static void time_maps_read(void)
{
	struct timespec then, now;
	long usec_elapsed;
	int ret, fd;
	int count = 0;
	int rd = 0;

	fd = open("/proc/self/maps", O_RDONLY);
	if (fd == -1) {
		printf("Couldn't open /proc/self/maps, bailing...\n");
		return;
	}

	printf("Reading /proc/self/maps... ");
	ret = clock_gettime(CLOCK_MONOTONIC, &then);

	while (1) {
		ret = read(fd, &buf, 65536);
		if (!ret || ret == -1)
			break;
		rd += ret;
		count++;
	}

	ret = clock_gettime(CLOCK_MONOTONIC, &now); 
	usec_elapsed = (now.tv_sec - then.tv_sec) * 1000000L;
	usec_elapsed += (now.tv_nsec - then.tv_nsec) / 1000L;

	printf("read %d bytes in %d syscalls in %ldus!\n", rd, count, usec_elapsed);
	close(fd);
}

static void *do_nothing_forever(void *unused)
{
	while (1)
		sleep(60);

	return NULL;
}

int main(int args, char **argv)
{
	int i, ret, threads_to_spawn = 0;	
	pthread_t tmp;

	if (args != 1) {
		threads_to_spawn = atoi(argv[1]);
		printf("Spawning %d threads\n", threads_to_spawn);
	}

	for (i = 0; i < threads_to_spawn; i++) {
		ret = pthread_create(&tmp, NULL, do_nothing_forever, NULL);
		if (ret)
			printf("Thread %d failed to spawn?\n", i);
	}

	time_maps_read();
	return 0;
}

  reply	other threads:[~2015-01-14  0:23 UTC|newest]

Thread overview: 66+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2015-01-14  0:20 Calvin Owens
2015-01-14  0:23 ` Calvin Owens [this message]
2015-01-14 14:13 ` Rasmus Villemoes
2015-01-14 14:37   ` Siddhesh Poyarekar
2015-01-14 14:53     ` Rasmus Villemoes
2015-01-14 21:03       ` Calvin Owens
2015-01-14 22:45         ` Andrew Morton
2015-01-14 23:51           ` Rasmus Villemoes
2015-01-16  1:15             ` Andrew Morton
2015-01-16 11:00               ` Kirill A. Shutemov
2015-01-14 15:25 ` Kirill A. Shutemov
2015-01-14 15:33   ` Cyrill Gorcunov
2015-01-14 20:46     ` Calvin Owens
2015-01-14 21:16       ` Cyrill Gorcunov
2015-01-22  2:45         ` [RFC][PATCH] procfs: Always expose /proc/<pid>/map_files/ and make it readable Calvin Owens
2015-01-22  7:16           ` Cyrill Gorcunov
2015-01-22 11:02           ` Kirill A. Shutemov
2015-01-22 21:00             ` Calvin Owens
2015-01-22 21:27               ` Kirill A. Shutemov
2015-01-23  5:52                 ` Calvin Owens
2015-01-24  3:15           ` [RFC][PATCH v2] " Calvin Owens
2015-01-26 12:47             ` Kirill A. Shutemov
2015-01-26 21:00               ` Cyrill Gorcunov
2015-01-26 23:43                 ` Andrew Morton
2015-01-27  0:15                   ` Kees Cook
2015-01-27  7:37                     ` Cyrill Gorcunov
2015-01-27 19:53                       ` Kees Cook
2015-01-27 21:35                         ` Cyrill Gorcunov
2015-01-27 21:46                         ` Pavel Emelyanov
2015-01-27  0:19                   ` Kirill A. Shutemov
2015-01-27  6:46                   ` Cyrill Gorcunov
2015-01-27  6:50                     ` Andrew Morton
2015-01-27  7:23                       ` Cyrill Gorcunov
2015-01-28  4:38                   ` Calvin Owens
2015-01-30  1:30                     ` Kees Cook
2015-01-31  1:58                       ` Calvin Owens
2015-02-02 14:01                         ` Austin S Hemmelgarn
2015-02-04  3:53                           ` Calvin Owens
2015-02-02 20:16                         ` Andy Lutomirski
2015-02-04  3:28                           ` Calvin Owens
2015-02-12  2:29             ` [RFC][PATCH v3] " Calvin Owens
2015-02-12  7:45               ` Cyrill Gorcunov
2015-02-14 20:40               ` [RFC][PATCH v4] " Calvin Owens
2015-03-10 22:17                 ` Cyrill Gorcunov
2015-04-28 22:23                   ` Calvin Owens
2015-04-29  7:32                     ` Cyrill Gorcunov
2015-05-19  3:10                 ` [PATCH v5] " Calvin Owens
2015-05-19  3:29                   ` Joe Perches
2015-05-19 18:04                   ` Andy Lutomirski
2015-05-21  1:52                     ` Calvin Owens
2015-05-21  2:10                       ` Andy Lutomirski
2015-06-09  3:39                   ` [PATCH v6] " Calvin Owens
2015-06-09 17:27                     ` Kees Cook
2015-06-09 17:47                       ` Andy Lutomirski
2015-06-09 18:15                         ` Cyrill Gorcunov
2015-06-09 21:13                     ` Andrew Morton
2015-06-10  1:39                       ` Calvin Owens
2015-06-10 20:58                         ` Andrew Morton
2015-06-11 11:10                           ` Alexey Dobriyan
2015-06-11 18:49                             ` Andrew Morton
2015-06-12  9:55                               ` Alexey Dobriyan
2015-06-19  2:32                     ` [PATCH v7] " Calvin Owens
2015-07-15 22:21                       ` Andrew Morton
2015-07-15 23:39                         ` Calvin Owens
2015-02-14 20:44               ` [PATCH] procfs: Return -ESRCH on /proc/N/fd/* when PID N doesn't exist Calvin Owens
2015-01-14 22:40 ` [RFC][PATCH] procfs: Add /proc/<pid>/mapped_files Kirill A. Shutemov

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20150114002341.GA6469@mail.thefacebook.com \
    --to=calvinowens@fb.com \
    --cc=adobriyan@gmail.com \
    --cc=akpm@linux-foundation.org \
    --cc=ebiederm@xmission.com \
    --cc=grant.likely@secretlab.ca \
    --cc=kernel-team@fb.com \
    --cc=kirill.shutemov@linux.intel.com \
    --cc=linux-kernel@vger.kernel.org \
    --cc=oleg@redhat.com \
    --cc=pfeiner@google.com \
    --cc=siddhesh.poyarekar@gmail.com \
    --cc=viro@zeniv.linux.org.uk \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox

Powered by JetHome