2005-04-17 02:20:36 +04:00
|
|
|
/*
|
|
|
|
* linux/ipc/util.c
|
|
|
|
* Copyright (C) 1992 Krishna Balasubramanian
|
|
|
|
*
|
|
|
|
* Sep 1997 - Call suser() last after "normal" permission checks so we
|
|
|
|
* get BSD style process accounting right.
|
|
|
|
* Occurs in several places in the IPC code.
|
|
|
|
* Chris Evans, <chris@ferret.lmh.ox.ac.uk>
|
|
|
|
* Nov 1999 - ipc helper functions, unified SMP locking
|
2006-01-15 04:43:54 +03:00
|
|
|
* Manfred Spraul <manfred@colorfullife.com>
|
2005-04-17 02:20:36 +04:00
|
|
|
* Oct 2002 - One lock per IPC id. RCU ipc_free for lock-free grow_ary().
|
|
|
|
* Mingming Cao <cmm@us.ibm.com>
|
2006-04-03 01:07:33 +04:00
|
|
|
* Mar 2006 - support for audit of ipc object properties
|
|
|
|
* Dustin Kirkland <dustin.kirkland@us.ibm.com>
|
2006-10-02 13:18:20 +04:00
|
|
|
* Jun 2006 - namespaces ssupport
|
|
|
|
* OpenVZ, SWsoft Inc.
|
|
|
|
* Pavel Emelianov <xemul@openvz.org>
|
2013-09-12 01:26:26 +04:00
|
|
|
*
|
|
|
|
* General sysv ipc locking scheme:
|
2013-10-17 00:46:45 +04:00
|
|
|
* rcu_read_lock()
|
|
|
|
* obtain the ipc object (kern_ipc_perm) by looking up the id in an idr
|
|
|
|
* tree.
|
|
|
|
* - perform initial checks (capabilities, auditing and permission,
|
|
|
|
* etc).
|
|
|
|
* - perform read-only operations, such as STAT, INFO commands.
|
|
|
|
* acquire the ipc lock (kern_ipc_perm.lock) through
|
|
|
|
* ipc_lock_object()
|
|
|
|
* - perform data updates, such as SET, RMID commands and
|
|
|
|
* mechanism-specific operations (semop/semtimedop,
|
|
|
|
* msgsnd/msgrcv, shmat/shmdt).
|
|
|
|
* drop the ipc lock, through ipc_unlock_object().
|
|
|
|
* rcu_read_unlock()
|
|
|
|
*
|
|
|
|
* The ids->rwsem must be taken when:
|
|
|
|
* - creating, removing and iterating the existing entries in ipc
|
|
|
|
* identifier sets.
|
|
|
|
* - iterating through files under /proc/sysvipc/
|
|
|
|
*
|
|
|
|
* Note that sems have a special fast path that avoids kern_ipc_perm.lock -
|
|
|
|
* see sem_lock().
|
2005-04-17 02:20:36 +04:00
|
|
|
*/
|
|
|
|
|
|
|
|
#include <linux/mm.h>
|
|
|
|
#include <linux/shm.h>
|
|
|
|
#include <linux/init.h>
|
|
|
|
#include <linux/msg.h>
|
|
|
|
#include <linux/vmalloc.h>
|
|
|
|
#include <linux/slab.h>
|
2013-04-30 02:08:05 +04:00
|
|
|
#include <linux/notifier.h>
|
2006-01-11 23:17:46 +03:00
|
|
|
#include <linux/capability.h>
|
2005-04-17 02:20:36 +04:00
|
|
|
#include <linux/highuid.h>
|
|
|
|
#include <linux/security.h>
|
|
|
|
#include <linux/rcupdate.h>
|
|
|
|
#include <linux/workqueue.h>
|
2005-09-07 02:17:09 +04:00
|
|
|
#include <linux/seq_file.h>
|
|
|
|
#include <linux/proc_fs.h>
|
2006-04-03 01:07:33 +04:00
|
|
|
#include <linux/audit.h>
|
2006-10-02 13:18:20 +04:00
|
|
|
#include <linux/nsproxy.h>
|
2007-10-19 10:40:54 +04:00
|
|
|
#include <linux/rwsem.h>
|
2008-04-29 12:00:42 +04:00
|
|
|
#include <linux/memory.h>
|
2008-02-08 15:18:22 +03:00
|
|
|
#include <linux/ipc_namespace.h>
|
2005-04-17 02:20:36 +04:00
|
|
|
|
|
|
|
#include <asm/unistd.h>
|
|
|
|
|
|
|
|
#include "util.h"
|
|
|
|
|
2005-09-07 02:17:09 +04:00
|
|
|
struct ipc_proc_iface {
|
|
|
|
const char *path;
|
|
|
|
const char *header;
|
2006-10-02 13:18:20 +04:00
|
|
|
int ids;
|
2005-09-07 02:17:09 +04:00
|
|
|
int (*show)(struct seq_file *, void *);
|
|
|
|
};
|
|
|
|
|
2005-04-17 02:20:36 +04:00
|
|
|
/**
|
2014-01-28 05:07:05 +04:00
|
|
|
* ipc_init - initialise ipc subsystem
|
2005-04-17 02:20:36 +04:00
|
|
|
*
|
2014-01-28 05:07:05 +04:00
|
|
|
* The various sysv ipc resources (semaphores, messages and shared
|
|
|
|
* memory) are initialised.
|
|
|
|
*
|
|
|
|
* A callback routine is registered into the memory hotplug notifier
|
|
|
|
* chain: since msgmni scales to lowmem this callback routine will be
|
|
|
|
* called upon successful memory add / remove to recompute msmgni.
|
2005-04-17 02:20:36 +04:00
|
|
|
*/
|
|
|
|
static int __init ipc_init(void)
|
|
|
|
{
|
ipc: optimize semget/shmget/msgget for lots of keys
ipc_findkey() used to scan all objects to look for the wanted key. This
is slow when using a high number of keys. This change adds an rhashtable
of kern_ipc_perm objects in ipc_ids, so that one lookup cease to be O(n).
This change gives a 865% improvement of benchmark reaim.jobs_per_min on a
56 threads Intel(R) Xeon(R) CPU E5-2695 v3 @ 2.30GHz with 256G memory [1]
Other (more micro) benchmark results, by the author: On an i5 laptop, the
following loop executed right after a reboot took, without and with this
change:
for (int i = 0, k=0x424242; i < KEYS; ++i)
semget(k++, 1, IPC_CREAT | 0600);
total total max single max single
KEYS without with call without call with
1 3.5 4.9 µs 3.5 4.9
10 7.6 8.6 µs 3.7 4.7
32 16.2 15.9 µs 4.3 5.3
100 72.9 41.8 µs 3.7 4.7
1000 5,630.0 502.0 µs * *
10000 1,340,000.0 7,240.0 µs * *
31900 17,600,000.0 22,200.0 µs * *
*: unreliable measure: high variance
The duration for a lookup-only usage was obtained by the same loop once
the keys are present:
total total max single max single
KEYS without with call without call with
1 2.1 2.5 µs 2.1 2.5
10 4.5 4.8 µs 2.2 2.3
32 13.0 10.8 µs 2.3 2.8
100 82.9 25.1 µs * 2.3
1000 5,780.0 217.0 µs * *
10000 1,470,000.0 2,520.0 µs * *
31900 17,400,000.0 7,810.0 µs * *
Finally, executing each semget() in a new process gave, when still
summing only the durations of these syscalls:
creation:
total total
KEYS without with
1 3.7 5.0 µs
10 32.9 36.7 µs
32 125.0 109.0 µs
100 523.0 353.0 µs
1000 20,300.0 3,280.0 µs
10000 2,470,000.0 46,700.0 µs
31900 27,800,000.0 219,000.0 µs
lookup-only:
total total
KEYS without with
1 2.5 2.7 µs
10 25.4 24.4 µs
32 106.0 72.6 µs
100 591.0 352.0 µs
1000 22,400.0 2,250.0 µs
10000 2,510,000.0 25,700.0 µs
31900 28,200,000.0 115,000.0 µs
[1] http://lkml.kernel.org/r/20170814060507.GE23258@yexl-desktop
Link: http://lkml.kernel.org/r/20170815194954.ck32ta2z35yuzpwp@debix
Signed-off-by: Guillaume Knispel <guillaume.knispel@supersonicimagine.com>
Reviewed-by: Marc Pardo <marc.pardo@supersonicimagine.com>
Cc: Davidlohr Bueso <dave@stgolabs.net>
Cc: Kees Cook <keescook@chromium.org>
Cc: Manfred Spraul <manfred@colorfullife.com>
Cc: Alexey Dobriyan <adobriyan@gmail.com>
Cc: "Eric W. Biederman" <ebiederm@xmission.com>
Cc: "Peter Zijlstra (Intel)" <peterz@infradead.org>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Cc: Serge Hallyn <serge@hallyn.com>
Cc: Andrey Vagin <avagin@openvz.org>
Cc: Guillaume Knispel <guillaume.knispel@supersonicimagine.com>
Cc: Marc Pardo <marc.pardo@supersonicimagine.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-09-09 02:17:55 +03:00
|
|
|
int err_sem, err_msg;
|
|
|
|
|
|
|
|
err_sem = sem_init();
|
|
|
|
WARN(err_sem, "ipc: sysv sem_init failed: %d\n", err_sem);
|
|
|
|
err_msg = msg_init();
|
|
|
|
WARN(err_msg, "ipc: sysv msg_init failed: %d\n", err_msg);
|
2005-04-17 02:20:36 +04:00
|
|
|
shm_init();
|
ipc: optimize semget/shmget/msgget for lots of keys
ipc_findkey() used to scan all objects to look for the wanted key. This
is slow when using a high number of keys. This change adds an rhashtable
of kern_ipc_perm objects in ipc_ids, so that one lookup cease to be O(n).
This change gives a 865% improvement of benchmark reaim.jobs_per_min on a
56 threads Intel(R) Xeon(R) CPU E5-2695 v3 @ 2.30GHz with 256G memory [1]
Other (more micro) benchmark results, by the author: On an i5 laptop, the
following loop executed right after a reboot took, without and with this
change:
for (int i = 0, k=0x424242; i < KEYS; ++i)
semget(k++, 1, IPC_CREAT | 0600);
total total max single max single
KEYS without with call without call with
1 3.5 4.9 µs 3.5 4.9
10 7.6 8.6 µs 3.7 4.7
32 16.2 15.9 µs 4.3 5.3
100 72.9 41.8 µs 3.7 4.7
1000 5,630.0 502.0 µs * *
10000 1,340,000.0 7,240.0 µs * *
31900 17,600,000.0 22,200.0 µs * *
*: unreliable measure: high variance
The duration for a lookup-only usage was obtained by the same loop once
the keys are present:
total total max single max single
KEYS without with call without call with
1 2.1 2.5 µs 2.1 2.5
10 4.5 4.8 µs 2.2 2.3
32 13.0 10.8 µs 2.3 2.8
100 82.9 25.1 µs * 2.3
1000 5,780.0 217.0 µs * *
10000 1,470,000.0 2,520.0 µs * *
31900 17,400,000.0 7,810.0 µs * *
Finally, executing each semget() in a new process gave, when still
summing only the durations of these syscalls:
creation:
total total
KEYS without with
1 3.7 5.0 µs
10 32.9 36.7 µs
32 125.0 109.0 µs
100 523.0 353.0 µs
1000 20,300.0 3,280.0 µs
10000 2,470,000.0 46,700.0 µs
31900 27,800,000.0 219,000.0 µs
lookup-only:
total total
KEYS without with
1 2.5 2.7 µs
10 25.4 24.4 µs
32 106.0 72.6 µs
100 591.0 352.0 µs
1000 22,400.0 2,250.0 µs
10000 2,510,000.0 25,700.0 µs
31900 28,200,000.0 115,000.0 µs
[1] http://lkml.kernel.org/r/20170814060507.GE23258@yexl-desktop
Link: http://lkml.kernel.org/r/20170815194954.ck32ta2z35yuzpwp@debix
Signed-off-by: Guillaume Knispel <guillaume.knispel@supersonicimagine.com>
Reviewed-by: Marc Pardo <marc.pardo@supersonicimagine.com>
Cc: Davidlohr Bueso <dave@stgolabs.net>
Cc: Kees Cook <keescook@chromium.org>
Cc: Manfred Spraul <manfred@colorfullife.com>
Cc: Alexey Dobriyan <adobriyan@gmail.com>
Cc: "Eric W. Biederman" <ebiederm@xmission.com>
Cc: "Peter Zijlstra (Intel)" <peterz@infradead.org>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Cc: Serge Hallyn <serge@hallyn.com>
Cc: Andrey Vagin <avagin@openvz.org>
Cc: Guillaume Knispel <guillaume.knispel@supersonicimagine.com>
Cc: Marc Pardo <marc.pardo@supersonicimagine.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-09-09 02:17:55 +03:00
|
|
|
|
|
|
|
return err_msg ? err_msg : err_sem;
|
2005-04-17 02:20:36 +04:00
|
|
|
}
|
2014-04-08 02:39:18 +04:00
|
|
|
device_initcall(ipc_init);
|
2005-04-17 02:20:36 +04:00
|
|
|
|
ipc: optimize semget/shmget/msgget for lots of keys
ipc_findkey() used to scan all objects to look for the wanted key. This
is slow when using a high number of keys. This change adds an rhashtable
of kern_ipc_perm objects in ipc_ids, so that one lookup cease to be O(n).
This change gives a 865% improvement of benchmark reaim.jobs_per_min on a
56 threads Intel(R) Xeon(R) CPU E5-2695 v3 @ 2.30GHz with 256G memory [1]
Other (more micro) benchmark results, by the author: On an i5 laptop, the
following loop executed right after a reboot took, without and with this
change:
for (int i = 0, k=0x424242; i < KEYS; ++i)
semget(k++, 1, IPC_CREAT | 0600);
total total max single max single
KEYS without with call without call with
1 3.5 4.9 µs 3.5 4.9
10 7.6 8.6 µs 3.7 4.7
32 16.2 15.9 µs 4.3 5.3
100 72.9 41.8 µs 3.7 4.7
1000 5,630.0 502.0 µs * *
10000 1,340,000.0 7,240.0 µs * *
31900 17,600,000.0 22,200.0 µs * *
*: unreliable measure: high variance
The duration for a lookup-only usage was obtained by the same loop once
the keys are present:
total total max single max single
KEYS without with call without call with
1 2.1 2.5 µs 2.1 2.5
10 4.5 4.8 µs 2.2 2.3
32 13.0 10.8 µs 2.3 2.8
100 82.9 25.1 µs * 2.3
1000 5,780.0 217.0 µs * *
10000 1,470,000.0 2,520.0 µs * *
31900 17,400,000.0 7,810.0 µs * *
Finally, executing each semget() in a new process gave, when still
summing only the durations of these syscalls:
creation:
total total
KEYS without with
1 3.7 5.0 µs
10 32.9 36.7 µs
32 125.0 109.0 µs
100 523.0 353.0 µs
1000 20,300.0 3,280.0 µs
10000 2,470,000.0 46,700.0 µs
31900 27,800,000.0 219,000.0 µs
lookup-only:
total total
KEYS without with
1 2.5 2.7 µs
10 25.4 24.4 µs
32 106.0 72.6 µs
100 591.0 352.0 µs
1000 22,400.0 2,250.0 µs
10000 2,510,000.0 25,700.0 µs
31900 28,200,000.0 115,000.0 µs
[1] http://lkml.kernel.org/r/20170814060507.GE23258@yexl-desktop
Link: http://lkml.kernel.org/r/20170815194954.ck32ta2z35yuzpwp@debix
Signed-off-by: Guillaume Knispel <guillaume.knispel@supersonicimagine.com>
Reviewed-by: Marc Pardo <marc.pardo@supersonicimagine.com>
Cc: Davidlohr Bueso <dave@stgolabs.net>
Cc: Kees Cook <keescook@chromium.org>
Cc: Manfred Spraul <manfred@colorfullife.com>
Cc: Alexey Dobriyan <adobriyan@gmail.com>
Cc: "Eric W. Biederman" <ebiederm@xmission.com>
Cc: "Peter Zijlstra (Intel)" <peterz@infradead.org>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Cc: Serge Hallyn <serge@hallyn.com>
Cc: Andrey Vagin <avagin@openvz.org>
Cc: Guillaume Knispel <guillaume.knispel@supersonicimagine.com>
Cc: Marc Pardo <marc.pardo@supersonicimagine.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-09-09 02:17:55 +03:00
|
|
|
static const struct rhashtable_params ipc_kht_params = {
|
|
|
|
.head_offset = offsetof(struct kern_ipc_perm, khtnode),
|
|
|
|
.key_offset = offsetof(struct kern_ipc_perm, key),
|
|
|
|
.key_len = FIELD_SIZEOF(struct kern_ipc_perm, key),
|
|
|
|
.locks_mul = 1,
|
|
|
|
.automatic_shrinking = true,
|
|
|
|
};
|
|
|
|
|
2005-04-17 02:20:36 +04:00
|
|
|
/**
|
2014-01-28 05:07:05 +04:00
|
|
|
* ipc_init_ids - initialise ipc identifiers
|
|
|
|
* @ids: ipc identifier set
|
2005-04-17 02:20:36 +04:00
|
|
|
*
|
2014-01-28 05:07:05 +04:00
|
|
|
* Set up the sequence range to use for the ipc identifier range (limited
|
ipc: optimize semget/shmget/msgget for lots of keys
ipc_findkey() used to scan all objects to look for the wanted key. This
is slow when using a high number of keys. This change adds an rhashtable
of kern_ipc_perm objects in ipc_ids, so that one lookup cease to be O(n).
This change gives a 865% improvement of benchmark reaim.jobs_per_min on a
56 threads Intel(R) Xeon(R) CPU E5-2695 v3 @ 2.30GHz with 256G memory [1]
Other (more micro) benchmark results, by the author: On an i5 laptop, the
following loop executed right after a reboot took, without and with this
change:
for (int i = 0, k=0x424242; i < KEYS; ++i)
semget(k++, 1, IPC_CREAT | 0600);
total total max single max single
KEYS without with call without call with
1 3.5 4.9 µs 3.5 4.9
10 7.6 8.6 µs 3.7 4.7
32 16.2 15.9 µs 4.3 5.3
100 72.9 41.8 µs 3.7 4.7
1000 5,630.0 502.0 µs * *
10000 1,340,000.0 7,240.0 µs * *
31900 17,600,000.0 22,200.0 µs * *
*: unreliable measure: high variance
The duration for a lookup-only usage was obtained by the same loop once
the keys are present:
total total max single max single
KEYS without with call without call with
1 2.1 2.5 µs 2.1 2.5
10 4.5 4.8 µs 2.2 2.3
32 13.0 10.8 µs 2.3 2.8
100 82.9 25.1 µs * 2.3
1000 5,780.0 217.0 µs * *
10000 1,470,000.0 2,520.0 µs * *
31900 17,400,000.0 7,810.0 µs * *
Finally, executing each semget() in a new process gave, when still
summing only the durations of these syscalls:
creation:
total total
KEYS without with
1 3.7 5.0 µs
10 32.9 36.7 µs
32 125.0 109.0 µs
100 523.0 353.0 µs
1000 20,300.0 3,280.0 µs
10000 2,470,000.0 46,700.0 µs
31900 27,800,000.0 219,000.0 µs
lookup-only:
total total
KEYS without with
1 2.5 2.7 µs
10 25.4 24.4 µs
32 106.0 72.6 µs
100 591.0 352.0 µs
1000 22,400.0 2,250.0 µs
10000 2,510,000.0 25,700.0 µs
31900 28,200,000.0 115,000.0 µs
[1] http://lkml.kernel.org/r/20170814060507.GE23258@yexl-desktop
Link: http://lkml.kernel.org/r/20170815194954.ck32ta2z35yuzpwp@debix
Signed-off-by: Guillaume Knispel <guillaume.knispel@supersonicimagine.com>
Reviewed-by: Marc Pardo <marc.pardo@supersonicimagine.com>
Cc: Davidlohr Bueso <dave@stgolabs.net>
Cc: Kees Cook <keescook@chromium.org>
Cc: Manfred Spraul <manfred@colorfullife.com>
Cc: Alexey Dobriyan <adobriyan@gmail.com>
Cc: "Eric W. Biederman" <ebiederm@xmission.com>
Cc: "Peter Zijlstra (Intel)" <peterz@infradead.org>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Cc: Serge Hallyn <serge@hallyn.com>
Cc: Andrey Vagin <avagin@openvz.org>
Cc: Guillaume Knispel <guillaume.knispel@supersonicimagine.com>
Cc: Marc Pardo <marc.pardo@supersonicimagine.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-09-09 02:17:55 +03:00
|
|
|
* below IPCMNI) then initialise the keys hashtable and ids idr.
|
2005-04-17 02:20:36 +04:00
|
|
|
*/
|
ipc: optimize semget/shmget/msgget for lots of keys
ipc_findkey() used to scan all objects to look for the wanted key. This
is slow when using a high number of keys. This change adds an rhashtable
of kern_ipc_perm objects in ipc_ids, so that one lookup cease to be O(n).
This change gives a 865% improvement of benchmark reaim.jobs_per_min on a
56 threads Intel(R) Xeon(R) CPU E5-2695 v3 @ 2.30GHz with 256G memory [1]
Other (more micro) benchmark results, by the author: On an i5 laptop, the
following loop executed right after a reboot took, without and with this
change:
for (int i = 0, k=0x424242; i < KEYS; ++i)
semget(k++, 1, IPC_CREAT | 0600);
total total max single max single
KEYS without with call without call with
1 3.5 4.9 µs 3.5 4.9
10 7.6 8.6 µs 3.7 4.7
32 16.2 15.9 µs 4.3 5.3
100 72.9 41.8 µs 3.7 4.7
1000 5,630.0 502.0 µs * *
10000 1,340,000.0 7,240.0 µs * *
31900 17,600,000.0 22,200.0 µs * *
*: unreliable measure: high variance
The duration for a lookup-only usage was obtained by the same loop once
the keys are present:
total total max single max single
KEYS without with call without call with
1 2.1 2.5 µs 2.1 2.5
10 4.5 4.8 µs 2.2 2.3
32 13.0 10.8 µs 2.3 2.8
100 82.9 25.1 µs * 2.3
1000 5,780.0 217.0 µs * *
10000 1,470,000.0 2,520.0 µs * *
31900 17,400,000.0 7,810.0 µs * *
Finally, executing each semget() in a new process gave, when still
summing only the durations of these syscalls:
creation:
total total
KEYS without with
1 3.7 5.0 µs
10 32.9 36.7 µs
32 125.0 109.0 µs
100 523.0 353.0 µs
1000 20,300.0 3,280.0 µs
10000 2,470,000.0 46,700.0 µs
31900 27,800,000.0 219,000.0 µs
lookup-only:
total total
KEYS without with
1 2.5 2.7 µs
10 25.4 24.4 µs
32 106.0 72.6 µs
100 591.0 352.0 µs
1000 22,400.0 2,250.0 µs
10000 2,510,000.0 25,700.0 µs
31900 28,200,000.0 115,000.0 µs
[1] http://lkml.kernel.org/r/20170814060507.GE23258@yexl-desktop
Link: http://lkml.kernel.org/r/20170815194954.ck32ta2z35yuzpwp@debix
Signed-off-by: Guillaume Knispel <guillaume.knispel@supersonicimagine.com>
Reviewed-by: Marc Pardo <marc.pardo@supersonicimagine.com>
Cc: Davidlohr Bueso <dave@stgolabs.net>
Cc: Kees Cook <keescook@chromium.org>
Cc: Manfred Spraul <manfred@colorfullife.com>
Cc: Alexey Dobriyan <adobriyan@gmail.com>
Cc: "Eric W. Biederman" <ebiederm@xmission.com>
Cc: "Peter Zijlstra (Intel)" <peterz@infradead.org>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Cc: Serge Hallyn <serge@hallyn.com>
Cc: Andrey Vagin <avagin@openvz.org>
Cc: Guillaume Knispel <guillaume.knispel@supersonicimagine.com>
Cc: Marc Pardo <marc.pardo@supersonicimagine.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-09-09 02:17:55 +03:00
|
|
|
int ipc_init_ids(struct ipc_ids *ids)
|
2005-04-17 02:20:36 +04:00
|
|
|
{
|
ipc: optimize semget/shmget/msgget for lots of keys
ipc_findkey() used to scan all objects to look for the wanted key. This
is slow when using a high number of keys. This change adds an rhashtable
of kern_ipc_perm objects in ipc_ids, so that one lookup cease to be O(n).
This change gives a 865% improvement of benchmark reaim.jobs_per_min on a
56 threads Intel(R) Xeon(R) CPU E5-2695 v3 @ 2.30GHz with 256G memory [1]
Other (more micro) benchmark results, by the author: On an i5 laptop, the
following loop executed right after a reboot took, without and with this
change:
for (int i = 0, k=0x424242; i < KEYS; ++i)
semget(k++, 1, IPC_CREAT | 0600);
total total max single max single
KEYS without with call without call with
1 3.5 4.9 µs 3.5 4.9
10 7.6 8.6 µs 3.7 4.7
32 16.2 15.9 µs 4.3 5.3
100 72.9 41.8 µs 3.7 4.7
1000 5,630.0 502.0 µs * *
10000 1,340,000.0 7,240.0 µs * *
31900 17,600,000.0 22,200.0 µs * *
*: unreliable measure: high variance
The duration for a lookup-only usage was obtained by the same loop once
the keys are present:
total total max single max single
KEYS without with call without call with
1 2.1 2.5 µs 2.1 2.5
10 4.5 4.8 µs 2.2 2.3
32 13.0 10.8 µs 2.3 2.8
100 82.9 25.1 µs * 2.3
1000 5,780.0 217.0 µs * *
10000 1,470,000.0 2,520.0 µs * *
31900 17,400,000.0 7,810.0 µs * *
Finally, executing each semget() in a new process gave, when still
summing only the durations of these syscalls:
creation:
total total
KEYS without with
1 3.7 5.0 µs
10 32.9 36.7 µs
32 125.0 109.0 µs
100 523.0 353.0 µs
1000 20,300.0 3,280.0 µs
10000 2,470,000.0 46,700.0 µs
31900 27,800,000.0 219,000.0 µs
lookup-only:
total total
KEYS without with
1 2.5 2.7 µs
10 25.4 24.4 µs
32 106.0 72.6 µs
100 591.0 352.0 µs
1000 22,400.0 2,250.0 µs
10000 2,510,000.0 25,700.0 µs
31900 28,200,000.0 115,000.0 µs
[1] http://lkml.kernel.org/r/20170814060507.GE23258@yexl-desktop
Link: http://lkml.kernel.org/r/20170815194954.ck32ta2z35yuzpwp@debix
Signed-off-by: Guillaume Knispel <guillaume.knispel@supersonicimagine.com>
Reviewed-by: Marc Pardo <marc.pardo@supersonicimagine.com>
Cc: Davidlohr Bueso <dave@stgolabs.net>
Cc: Kees Cook <keescook@chromium.org>
Cc: Manfred Spraul <manfred@colorfullife.com>
Cc: Alexey Dobriyan <adobriyan@gmail.com>
Cc: "Eric W. Biederman" <ebiederm@xmission.com>
Cc: "Peter Zijlstra (Intel)" <peterz@infradead.org>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Cc: Serge Hallyn <serge@hallyn.com>
Cc: Andrey Vagin <avagin@openvz.org>
Cc: Guillaume Knispel <guillaume.knispel@supersonicimagine.com>
Cc: Marc Pardo <marc.pardo@supersonicimagine.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-09-09 02:17:55 +03:00
|
|
|
int err;
|
2005-04-17 02:20:36 +04:00
|
|
|
ids->in_use = 0;
|
|
|
|
ids->seq = 0;
|
2013-01-05 03:34:50 +04:00
|
|
|
ids->next_id = -1;
|
2014-01-28 05:07:09 +04:00
|
|
|
init_rwsem(&ids->rwsem);
|
ipc: optimize semget/shmget/msgget for lots of keys
ipc_findkey() used to scan all objects to look for the wanted key. This
is slow when using a high number of keys. This change adds an rhashtable
of kern_ipc_perm objects in ipc_ids, so that one lookup cease to be O(n).
This change gives a 865% improvement of benchmark reaim.jobs_per_min on a
56 threads Intel(R) Xeon(R) CPU E5-2695 v3 @ 2.30GHz with 256G memory [1]
Other (more micro) benchmark results, by the author: On an i5 laptop, the
following loop executed right after a reboot took, without and with this
change:
for (int i = 0, k=0x424242; i < KEYS; ++i)
semget(k++, 1, IPC_CREAT | 0600);
total total max single max single
KEYS without with call without call with
1 3.5 4.9 µs 3.5 4.9
10 7.6 8.6 µs 3.7 4.7
32 16.2 15.9 µs 4.3 5.3
100 72.9 41.8 µs 3.7 4.7
1000 5,630.0 502.0 µs * *
10000 1,340,000.0 7,240.0 µs * *
31900 17,600,000.0 22,200.0 µs * *
*: unreliable measure: high variance
The duration for a lookup-only usage was obtained by the same loop once
the keys are present:
total total max single max single
KEYS without with call without call with
1 2.1 2.5 µs 2.1 2.5
10 4.5 4.8 µs 2.2 2.3
32 13.0 10.8 µs 2.3 2.8
100 82.9 25.1 µs * 2.3
1000 5,780.0 217.0 µs * *
10000 1,470,000.0 2,520.0 µs * *
31900 17,400,000.0 7,810.0 µs * *
Finally, executing each semget() in a new process gave, when still
summing only the durations of these syscalls:
creation:
total total
KEYS without with
1 3.7 5.0 µs
10 32.9 36.7 µs
32 125.0 109.0 µs
100 523.0 353.0 µs
1000 20,300.0 3,280.0 µs
10000 2,470,000.0 46,700.0 µs
31900 27,800,000.0 219,000.0 µs
lookup-only:
total total
KEYS without with
1 2.5 2.7 µs
10 25.4 24.4 µs
32 106.0 72.6 µs
100 591.0 352.0 µs
1000 22,400.0 2,250.0 µs
10000 2,510,000.0 25,700.0 µs
31900 28,200,000.0 115,000.0 µs
[1] http://lkml.kernel.org/r/20170814060507.GE23258@yexl-desktop
Link: http://lkml.kernel.org/r/20170815194954.ck32ta2z35yuzpwp@debix
Signed-off-by: Guillaume Knispel <guillaume.knispel@supersonicimagine.com>
Reviewed-by: Marc Pardo <marc.pardo@supersonicimagine.com>
Cc: Davidlohr Bueso <dave@stgolabs.net>
Cc: Kees Cook <keescook@chromium.org>
Cc: Manfred Spraul <manfred@colorfullife.com>
Cc: Alexey Dobriyan <adobriyan@gmail.com>
Cc: "Eric W. Biederman" <ebiederm@xmission.com>
Cc: "Peter Zijlstra (Intel)" <peterz@infradead.org>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Cc: Serge Hallyn <serge@hallyn.com>
Cc: Andrey Vagin <avagin@openvz.org>
Cc: Guillaume Knispel <guillaume.knispel@supersonicimagine.com>
Cc: Marc Pardo <marc.pardo@supersonicimagine.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-09-09 02:17:55 +03:00
|
|
|
err = rhashtable_init(&ids->key_ht, &ipc_kht_params);
|
|
|
|
if (err)
|
|
|
|
return err;
|
2007-10-19 10:40:48 +04:00
|
|
|
idr_init(&ids->ipcs_idr);
|
ipc: optimize semget/shmget/msgget for lots of keys
ipc_findkey() used to scan all objects to look for the wanted key. This
is slow when using a high number of keys. This change adds an rhashtable
of kern_ipc_perm objects in ipc_ids, so that one lookup cease to be O(n).
This change gives a 865% improvement of benchmark reaim.jobs_per_min on a
56 threads Intel(R) Xeon(R) CPU E5-2695 v3 @ 2.30GHz with 256G memory [1]
Other (more micro) benchmark results, by the author: On an i5 laptop, the
following loop executed right after a reboot took, without and with this
change:
for (int i = 0, k=0x424242; i < KEYS; ++i)
semget(k++, 1, IPC_CREAT | 0600);
total total max single max single
KEYS without with call without call with
1 3.5 4.9 µs 3.5 4.9
10 7.6 8.6 µs 3.7 4.7
32 16.2 15.9 µs 4.3 5.3
100 72.9 41.8 µs 3.7 4.7
1000 5,630.0 502.0 µs * *
10000 1,340,000.0 7,240.0 µs * *
31900 17,600,000.0 22,200.0 µs * *
*: unreliable measure: high variance
The duration for a lookup-only usage was obtained by the same loop once
the keys are present:
total total max single max single
KEYS without with call without call with
1 2.1 2.5 µs 2.1 2.5
10 4.5 4.8 µs 2.2 2.3
32 13.0 10.8 µs 2.3 2.8
100 82.9 25.1 µs * 2.3
1000 5,780.0 217.0 µs * *
10000 1,470,000.0 2,520.0 µs * *
31900 17,400,000.0 7,810.0 µs * *
Finally, executing each semget() in a new process gave, when still
summing only the durations of these syscalls:
creation:
total total
KEYS without with
1 3.7 5.0 µs
10 32.9 36.7 µs
32 125.0 109.0 µs
100 523.0 353.0 µs
1000 20,300.0 3,280.0 µs
10000 2,470,000.0 46,700.0 µs
31900 27,800,000.0 219,000.0 µs
lookup-only:
total total
KEYS without with
1 2.5 2.7 µs
10 25.4 24.4 µs
32 106.0 72.6 µs
100 591.0 352.0 µs
1000 22,400.0 2,250.0 µs
10000 2,510,000.0 25,700.0 µs
31900 28,200,000.0 115,000.0 µs
[1] http://lkml.kernel.org/r/20170814060507.GE23258@yexl-desktop
Link: http://lkml.kernel.org/r/20170815194954.ck32ta2z35yuzpwp@debix
Signed-off-by: Guillaume Knispel <guillaume.knispel@supersonicimagine.com>
Reviewed-by: Marc Pardo <marc.pardo@supersonicimagine.com>
Cc: Davidlohr Bueso <dave@stgolabs.net>
Cc: Kees Cook <keescook@chromium.org>
Cc: Manfred Spraul <manfred@colorfullife.com>
Cc: Alexey Dobriyan <adobriyan@gmail.com>
Cc: "Eric W. Biederman" <ebiederm@xmission.com>
Cc: "Peter Zijlstra (Intel)" <peterz@infradead.org>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Cc: Serge Hallyn <serge@hallyn.com>
Cc: Andrey Vagin <avagin@openvz.org>
Cc: Guillaume Knispel <guillaume.knispel@supersonicimagine.com>
Cc: Marc Pardo <marc.pardo@supersonicimagine.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-09-09 02:17:55 +03:00
|
|
|
ids->tables_initialized = true;
|
|
|
|
return 0;
|
2005-04-17 02:20:36 +04:00
|
|
|
}
|
|
|
|
|
2005-09-07 02:17:09 +04:00
|
|
|
#ifdef CONFIG_PROC_FS
|
2007-02-12 11:55:35 +03:00
|
|
|
static const struct file_operations sysvipc_proc_fops;
|
2005-09-07 02:17:09 +04:00
|
|
|
/**
|
2014-01-28 05:07:05 +04:00
|
|
|
* ipc_init_proc_interface - create a proc interface for sysipc types using a seq_file interface.
|
|
|
|
* @path: Path in procfs
|
|
|
|
* @header: Banner to be printed at the beginning of the file.
|
|
|
|
* @ids: ipc id table to iterate.
|
|
|
|
* @show: show routine.
|
2005-09-07 02:17:09 +04:00
|
|
|
*/
|
|
|
|
void __init ipc_init_proc_interface(const char *path, const char *header,
|
2006-10-02 13:18:20 +04:00
|
|
|
int ids, int (*show)(struct seq_file *, void *))
|
2005-09-07 02:17:09 +04:00
|
|
|
{
|
|
|
|
struct proc_dir_entry *pde;
|
|
|
|
struct ipc_proc_iface *iface;
|
|
|
|
|
|
|
|
iface = kmalloc(sizeof(*iface), GFP_KERNEL);
|
|
|
|
if (!iface)
|
|
|
|
return;
|
|
|
|
iface->path = path;
|
|
|
|
iface->header = header;
|
|
|
|
iface->ids = ids;
|
|
|
|
iface->show = show;
|
|
|
|
|
2008-04-29 12:02:12 +04:00
|
|
|
pde = proc_create_data(path,
|
|
|
|
S_IRUGO, /* world readable */
|
|
|
|
NULL, /* parent dir */
|
|
|
|
&sysvipc_proc_fops,
|
|
|
|
iface);
|
2014-01-28 05:07:06 +04:00
|
|
|
if (!pde)
|
2005-09-07 02:17:09 +04:00
|
|
|
kfree(iface);
|
|
|
|
}
|
|
|
|
#endif
|
|
|
|
|
2005-04-17 02:20:36 +04:00
|
|
|
/**
|
2014-01-28 05:07:05 +04:00
|
|
|
* ipc_findkey - find a key in an ipc identifier set
|
|
|
|
* @ids: ipc identifier set
|
|
|
|
* @key: key to find
|
2014-06-07 01:37:37 +04:00
|
|
|
*
|
2014-01-28 05:07:05 +04:00
|
|
|
* Returns the locked pointer to the ipc structure if found or NULL
|
|
|
|
* otherwise. If key is found ipc points to the owning ipc structure
|
|
|
|
*
|
ipc: optimize semget/shmget/msgget for lots of keys
ipc_findkey() used to scan all objects to look for the wanted key. This
is slow when using a high number of keys. This change adds an rhashtable
of kern_ipc_perm objects in ipc_ids, so that one lookup cease to be O(n).
This change gives a 865% improvement of benchmark reaim.jobs_per_min on a
56 threads Intel(R) Xeon(R) CPU E5-2695 v3 @ 2.30GHz with 256G memory [1]
Other (more micro) benchmark results, by the author: On an i5 laptop, the
following loop executed right after a reboot took, without and with this
change:
for (int i = 0, k=0x424242; i < KEYS; ++i)
semget(k++, 1, IPC_CREAT | 0600);
total total max single max single
KEYS without with call without call with
1 3.5 4.9 µs 3.5 4.9
10 7.6 8.6 µs 3.7 4.7
32 16.2 15.9 µs 4.3 5.3
100 72.9 41.8 µs 3.7 4.7
1000 5,630.0 502.0 µs * *
10000 1,340,000.0 7,240.0 µs * *
31900 17,600,000.0 22,200.0 µs * *
*: unreliable measure: high variance
The duration for a lookup-only usage was obtained by the same loop once
the keys are present:
total total max single max single
KEYS without with call without call with
1 2.1 2.5 µs 2.1 2.5
10 4.5 4.8 µs 2.2 2.3
32 13.0 10.8 µs 2.3 2.8
100 82.9 25.1 µs * 2.3
1000 5,780.0 217.0 µs * *
10000 1,470,000.0 2,520.0 µs * *
31900 17,400,000.0 7,810.0 µs * *
Finally, executing each semget() in a new process gave, when still
summing only the durations of these syscalls:
creation:
total total
KEYS without with
1 3.7 5.0 µs
10 32.9 36.7 µs
32 125.0 109.0 µs
100 523.0 353.0 µs
1000 20,300.0 3,280.0 µs
10000 2,470,000.0 46,700.0 µs
31900 27,800,000.0 219,000.0 µs
lookup-only:
total total
KEYS without with
1 2.5 2.7 µs
10 25.4 24.4 µs
32 106.0 72.6 µs
100 591.0 352.0 µs
1000 22,400.0 2,250.0 µs
10000 2,510,000.0 25,700.0 µs
31900 28,200,000.0 115,000.0 µs
[1] http://lkml.kernel.org/r/20170814060507.GE23258@yexl-desktop
Link: http://lkml.kernel.org/r/20170815194954.ck32ta2z35yuzpwp@debix
Signed-off-by: Guillaume Knispel <guillaume.knispel@supersonicimagine.com>
Reviewed-by: Marc Pardo <marc.pardo@supersonicimagine.com>
Cc: Davidlohr Bueso <dave@stgolabs.net>
Cc: Kees Cook <keescook@chromium.org>
Cc: Manfred Spraul <manfred@colorfullife.com>
Cc: Alexey Dobriyan <adobriyan@gmail.com>
Cc: "Eric W. Biederman" <ebiederm@xmission.com>
Cc: "Peter Zijlstra (Intel)" <peterz@infradead.org>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Cc: Serge Hallyn <serge@hallyn.com>
Cc: Andrey Vagin <avagin@openvz.org>
Cc: Guillaume Knispel <guillaume.knispel@supersonicimagine.com>
Cc: Marc Pardo <marc.pardo@supersonicimagine.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-09-09 02:17:55 +03:00
|
|
|
* Called with writer ipc_ids.rwsem held.
|
2005-04-17 02:20:36 +04:00
|
|
|
*/
|
2007-10-19 10:40:49 +04:00
|
|
|
static struct kern_ipc_perm *ipc_findkey(struct ipc_ids *ids, key_t key)
|
2005-04-17 02:20:36 +04:00
|
|
|
{
|
ipc: optimize semget/shmget/msgget for lots of keys
ipc_findkey() used to scan all objects to look for the wanted key. This
is slow when using a high number of keys. This change adds an rhashtable
of kern_ipc_perm objects in ipc_ids, so that one lookup cease to be O(n).
This change gives a 865% improvement of benchmark reaim.jobs_per_min on a
56 threads Intel(R) Xeon(R) CPU E5-2695 v3 @ 2.30GHz with 256G memory [1]
Other (more micro) benchmark results, by the author: On an i5 laptop, the
following loop executed right after a reboot took, without and with this
change:
for (int i = 0, k=0x424242; i < KEYS; ++i)
semget(k++, 1, IPC_CREAT | 0600);
total total max single max single
KEYS without with call without call with
1 3.5 4.9 µs 3.5 4.9
10 7.6 8.6 µs 3.7 4.7
32 16.2 15.9 µs 4.3 5.3
100 72.9 41.8 µs 3.7 4.7
1000 5,630.0 502.0 µs * *
10000 1,340,000.0 7,240.0 µs * *
31900 17,600,000.0 22,200.0 µs * *
*: unreliable measure: high variance
The duration for a lookup-only usage was obtained by the same loop once
the keys are present:
total total max single max single
KEYS without with call without call with
1 2.1 2.5 µs 2.1 2.5
10 4.5 4.8 µs 2.2 2.3
32 13.0 10.8 µs 2.3 2.8
100 82.9 25.1 µs * 2.3
1000 5,780.0 217.0 µs * *
10000 1,470,000.0 2,520.0 µs * *
31900 17,400,000.0 7,810.0 µs * *
Finally, executing each semget() in a new process gave, when still
summing only the durations of these syscalls:
creation:
total total
KEYS without with
1 3.7 5.0 µs
10 32.9 36.7 µs
32 125.0 109.0 µs
100 523.0 353.0 µs
1000 20,300.0 3,280.0 µs
10000 2,470,000.0 46,700.0 µs
31900 27,800,000.0 219,000.0 µs
lookup-only:
total total
KEYS without with
1 2.5 2.7 µs
10 25.4 24.4 µs
32 106.0 72.6 µs
100 591.0 352.0 µs
1000 22,400.0 2,250.0 µs
10000 2,510,000.0 25,700.0 µs
31900 28,200,000.0 115,000.0 µs
[1] http://lkml.kernel.org/r/20170814060507.GE23258@yexl-desktop
Link: http://lkml.kernel.org/r/20170815194954.ck32ta2z35yuzpwp@debix
Signed-off-by: Guillaume Knispel <guillaume.knispel@supersonicimagine.com>
Reviewed-by: Marc Pardo <marc.pardo@supersonicimagine.com>
Cc: Davidlohr Bueso <dave@stgolabs.net>
Cc: Kees Cook <keescook@chromium.org>
Cc: Manfred Spraul <manfred@colorfullife.com>
Cc: Alexey Dobriyan <adobriyan@gmail.com>
Cc: "Eric W. Biederman" <ebiederm@xmission.com>
Cc: "Peter Zijlstra (Intel)" <peterz@infradead.org>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Cc: Serge Hallyn <serge@hallyn.com>
Cc: Andrey Vagin <avagin@openvz.org>
Cc: Guillaume Knispel <guillaume.knispel@supersonicimagine.com>
Cc: Marc Pardo <marc.pardo@supersonicimagine.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-09-09 02:17:55 +03:00
|
|
|
struct kern_ipc_perm *ipcp = NULL;
|
2007-10-19 10:40:48 +04:00
|
|
|
|
ipc: optimize semget/shmget/msgget for lots of keys
ipc_findkey() used to scan all objects to look for the wanted key. This
is slow when using a high number of keys. This change adds an rhashtable
of kern_ipc_perm objects in ipc_ids, so that one lookup cease to be O(n).
This change gives a 865% improvement of benchmark reaim.jobs_per_min on a
56 threads Intel(R) Xeon(R) CPU E5-2695 v3 @ 2.30GHz with 256G memory [1]
Other (more micro) benchmark results, by the author: On an i5 laptop, the
following loop executed right after a reboot took, without and with this
change:
for (int i = 0, k=0x424242; i < KEYS; ++i)
semget(k++, 1, IPC_CREAT | 0600);
total total max single max single
KEYS without with call without call with
1 3.5 4.9 µs 3.5 4.9
10 7.6 8.6 µs 3.7 4.7
32 16.2 15.9 µs 4.3 5.3
100 72.9 41.8 µs 3.7 4.7
1000 5,630.0 502.0 µs * *
10000 1,340,000.0 7,240.0 µs * *
31900 17,600,000.0 22,200.0 µs * *
*: unreliable measure: high variance
The duration for a lookup-only usage was obtained by the same loop once
the keys are present:
total total max single max single
KEYS without with call without call with
1 2.1 2.5 µs 2.1 2.5
10 4.5 4.8 µs 2.2 2.3
32 13.0 10.8 µs 2.3 2.8
100 82.9 25.1 µs * 2.3
1000 5,780.0 217.0 µs * *
10000 1,470,000.0 2,520.0 µs * *
31900 17,400,000.0 7,810.0 µs * *
Finally, executing each semget() in a new process gave, when still
summing only the durations of these syscalls:
creation:
total total
KEYS without with
1 3.7 5.0 µs
10 32.9 36.7 µs
32 125.0 109.0 µs
100 523.0 353.0 µs
1000 20,300.0 3,280.0 µs
10000 2,470,000.0 46,700.0 µs
31900 27,800,000.0 219,000.0 µs
lookup-only:
total total
KEYS without with
1 2.5 2.7 µs
10 25.4 24.4 µs
32 106.0 72.6 µs
100 591.0 352.0 µs
1000 22,400.0 2,250.0 µs
10000 2,510,000.0 25,700.0 µs
31900 28,200,000.0 115,000.0 µs
[1] http://lkml.kernel.org/r/20170814060507.GE23258@yexl-desktop
Link: http://lkml.kernel.org/r/20170815194954.ck32ta2z35yuzpwp@debix
Signed-off-by: Guillaume Knispel <guillaume.knispel@supersonicimagine.com>
Reviewed-by: Marc Pardo <marc.pardo@supersonicimagine.com>
Cc: Davidlohr Bueso <dave@stgolabs.net>
Cc: Kees Cook <keescook@chromium.org>
Cc: Manfred Spraul <manfred@colorfullife.com>
Cc: Alexey Dobriyan <adobriyan@gmail.com>
Cc: "Eric W. Biederman" <ebiederm@xmission.com>
Cc: "Peter Zijlstra (Intel)" <peterz@infradead.org>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Cc: Serge Hallyn <serge@hallyn.com>
Cc: Andrey Vagin <avagin@openvz.org>
Cc: Guillaume Knispel <guillaume.knispel@supersonicimagine.com>
Cc: Marc Pardo <marc.pardo@supersonicimagine.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-09-09 02:17:55 +03:00
|
|
|
if (likely(ids->tables_initialized))
|
|
|
|
ipcp = rhashtable_lookup_fast(&ids->key_ht, &key,
|
|
|
|
ipc_kht_params);
|
2007-10-19 10:40:48 +04:00
|
|
|
|
ipc: optimize semget/shmget/msgget for lots of keys
ipc_findkey() used to scan all objects to look for the wanted key. This
is slow when using a high number of keys. This change adds an rhashtable
of kern_ipc_perm objects in ipc_ids, so that one lookup cease to be O(n).
This change gives a 865% improvement of benchmark reaim.jobs_per_min on a
56 threads Intel(R) Xeon(R) CPU E5-2695 v3 @ 2.30GHz with 256G memory [1]
Other (more micro) benchmark results, by the author: On an i5 laptop, the
following loop executed right after a reboot took, without and with this
change:
for (int i = 0, k=0x424242; i < KEYS; ++i)
semget(k++, 1, IPC_CREAT | 0600);
total total max single max single
KEYS without with call without call with
1 3.5 4.9 µs 3.5 4.9
10 7.6 8.6 µs 3.7 4.7
32 16.2 15.9 µs 4.3 5.3
100 72.9 41.8 µs 3.7 4.7
1000 5,630.0 502.0 µs * *
10000 1,340,000.0 7,240.0 µs * *
31900 17,600,000.0 22,200.0 µs * *
*: unreliable measure: high variance
The duration for a lookup-only usage was obtained by the same loop once
the keys are present:
total total max single max single
KEYS without with call without call with
1 2.1 2.5 µs 2.1 2.5
10 4.5 4.8 µs 2.2 2.3
32 13.0 10.8 µs 2.3 2.8
100 82.9 25.1 µs * 2.3
1000 5,780.0 217.0 µs * *
10000 1,470,000.0 2,520.0 µs * *
31900 17,400,000.0 7,810.0 µs * *
Finally, executing each semget() in a new process gave, when still
summing only the durations of these syscalls:
creation:
total total
KEYS without with
1 3.7 5.0 µs
10 32.9 36.7 µs
32 125.0 109.0 µs
100 523.0 353.0 µs
1000 20,300.0 3,280.0 µs
10000 2,470,000.0 46,700.0 µs
31900 27,800,000.0 219,000.0 µs
lookup-only:
total total
KEYS without with
1 2.5 2.7 µs
10 25.4 24.4 µs
32 106.0 72.6 µs
100 591.0 352.0 µs
1000 22,400.0 2,250.0 µs
10000 2,510,000.0 25,700.0 µs
31900 28,200,000.0 115,000.0 µs
[1] http://lkml.kernel.org/r/20170814060507.GE23258@yexl-desktop
Link: http://lkml.kernel.org/r/20170815194954.ck32ta2z35yuzpwp@debix
Signed-off-by: Guillaume Knispel <guillaume.knispel@supersonicimagine.com>
Reviewed-by: Marc Pardo <marc.pardo@supersonicimagine.com>
Cc: Davidlohr Bueso <dave@stgolabs.net>
Cc: Kees Cook <keescook@chromium.org>
Cc: Manfred Spraul <manfred@colorfullife.com>
Cc: Alexey Dobriyan <adobriyan@gmail.com>
Cc: "Eric W. Biederman" <ebiederm@xmission.com>
Cc: "Peter Zijlstra (Intel)" <peterz@infradead.org>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Cc: Serge Hallyn <serge@hallyn.com>
Cc: Andrey Vagin <avagin@openvz.org>
Cc: Guillaume Knispel <guillaume.knispel@supersonicimagine.com>
Cc: Marc Pardo <marc.pardo@supersonicimagine.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-09-09 02:17:55 +03:00
|
|
|
if (ipcp) {
|
2013-09-12 01:26:29 +04:00
|
|
|
rcu_read_lock();
|
ipc: optimize semget/shmget/msgget for lots of keys
ipc_findkey() used to scan all objects to look for the wanted key. This
is slow when using a high number of keys. This change adds an rhashtable
of kern_ipc_perm objects in ipc_ids, so that one lookup cease to be O(n).
This change gives a 865% improvement of benchmark reaim.jobs_per_min on a
56 threads Intel(R) Xeon(R) CPU E5-2695 v3 @ 2.30GHz with 256G memory [1]
Other (more micro) benchmark results, by the author: On an i5 laptop, the
following loop executed right after a reboot took, without and with this
change:
for (int i = 0, k=0x424242; i < KEYS; ++i)
semget(k++, 1, IPC_CREAT | 0600);
total total max single max single
KEYS without with call without call with
1 3.5 4.9 µs 3.5 4.9
10 7.6 8.6 µs 3.7 4.7
32 16.2 15.9 µs 4.3 5.3
100 72.9 41.8 µs 3.7 4.7
1000 5,630.0 502.0 µs * *
10000 1,340,000.0 7,240.0 µs * *
31900 17,600,000.0 22,200.0 µs * *
*: unreliable measure: high variance
The duration for a lookup-only usage was obtained by the same loop once
the keys are present:
total total max single max single
KEYS without with call without call with
1 2.1 2.5 µs 2.1 2.5
10 4.5 4.8 µs 2.2 2.3
32 13.0 10.8 µs 2.3 2.8
100 82.9 25.1 µs * 2.3
1000 5,780.0 217.0 µs * *
10000 1,470,000.0 2,520.0 µs * *
31900 17,400,000.0 7,810.0 µs * *
Finally, executing each semget() in a new process gave, when still
summing only the durations of these syscalls:
creation:
total total
KEYS without with
1 3.7 5.0 µs
10 32.9 36.7 µs
32 125.0 109.0 µs
100 523.0 353.0 µs
1000 20,300.0 3,280.0 µs
10000 2,470,000.0 46,700.0 µs
31900 27,800,000.0 219,000.0 µs
lookup-only:
total total
KEYS without with
1 2.5 2.7 µs
10 25.4 24.4 µs
32 106.0 72.6 µs
100 591.0 352.0 µs
1000 22,400.0 2,250.0 µs
10000 2,510,000.0 25,700.0 µs
31900 28,200,000.0 115,000.0 µs
[1] http://lkml.kernel.org/r/20170814060507.GE23258@yexl-desktop
Link: http://lkml.kernel.org/r/20170815194954.ck32ta2z35yuzpwp@debix
Signed-off-by: Guillaume Knispel <guillaume.knispel@supersonicimagine.com>
Reviewed-by: Marc Pardo <marc.pardo@supersonicimagine.com>
Cc: Davidlohr Bueso <dave@stgolabs.net>
Cc: Kees Cook <keescook@chromium.org>
Cc: Manfred Spraul <manfred@colorfullife.com>
Cc: Alexey Dobriyan <adobriyan@gmail.com>
Cc: "Eric W. Biederman" <ebiederm@xmission.com>
Cc: "Peter Zijlstra (Intel)" <peterz@infradead.org>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Cc: Serge Hallyn <serge@hallyn.com>
Cc: Andrey Vagin <avagin@openvz.org>
Cc: Guillaume Knispel <guillaume.knispel@supersonicimagine.com>
Cc: Marc Pardo <marc.pardo@supersonicimagine.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-09-09 02:17:55 +03:00
|
|
|
ipc_lock_object(ipcp);
|
|
|
|
return ipcp;
|
2005-04-17 02:20:36 +04:00
|
|
|
}
|
2007-10-19 10:40:48 +04:00
|
|
|
|
|
|
|
return NULL;
|
2005-04-17 02:20:36 +04:00
|
|
|
}
|
|
|
|
|
2007-10-19 10:40:48 +04:00
|
|
|
/**
|
2014-01-28 05:07:05 +04:00
|
|
|
* ipc_get_maxid - get the last assigned id
|
|
|
|
* @ids: ipc identifier set
|
2007-10-19 10:40:48 +04:00
|
|
|
*
|
2014-01-28 05:07:05 +04:00
|
|
|
* Called with ipc_ids.rwsem held.
|
2005-04-17 02:20:36 +04:00
|
|
|
*/
|
2007-10-19 10:40:48 +04:00
|
|
|
int ipc_get_maxid(struct ipc_ids *ids)
|
|
|
|
{
|
|
|
|
struct kern_ipc_perm *ipc;
|
|
|
|
int max_id = -1;
|
|
|
|
int total, id;
|
|
|
|
|
|
|
|
if (ids->in_use == 0)
|
|
|
|
return -1;
|
2005-04-17 02:20:36 +04:00
|
|
|
|
2007-10-19 10:40:48 +04:00
|
|
|
if (ids->in_use == IPCMNI)
|
|
|
|
return IPCMNI - 1;
|
|
|
|
|
|
|
|
/* Look for the last assigned id */
|
|
|
|
total = 0;
|
|
|
|
for (id = 0; id < IPCMNI && total < ids->in_use; id++) {
|
|
|
|
ipc = idr_find(&ids->ipcs_idr, id);
|
|
|
|
if (ipc != NULL) {
|
|
|
|
max_id = id;
|
|
|
|
total++;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
return max_id;
|
2005-04-17 02:20:36 +04:00
|
|
|
}
|
|
|
|
|
|
|
|
/**
|
2014-01-28 05:07:05 +04:00
|
|
|
* ipc_addid - add an ipc identifier
|
|
|
|
* @ids: ipc identifier set
|
|
|
|
* @new: new ipc permission set
|
|
|
|
* @size: limit for the number of used ids
|
2005-04-17 02:20:36 +04:00
|
|
|
*
|
2014-01-28 05:07:05 +04:00
|
|
|
* Add an entry 'new' to the ipc ids idr. The permissions object is
|
|
|
|
* initialised and the first free entry is set up and the id assigned
|
|
|
|
* is returned. The 'new' entry is returned in a locked state on success.
|
|
|
|
* On failure the entry is not locked and a negative err-code is returned.
|
2005-04-17 02:20:36 +04:00
|
|
|
*
|
2014-01-28 05:07:05 +04:00
|
|
|
* Called with writer ipc_ids.rwsem held.
|
2005-04-17 02:20:36 +04:00
|
|
|
*/
|
2014-01-28 05:07:04 +04:00
|
|
|
int ipc_addid(struct ipc_ids *ids, struct kern_ipc_perm *new, int size)
|
2005-04-17 02:20:36 +04:00
|
|
|
{
|
2012-02-08 04:54:11 +04:00
|
|
|
kuid_t euid;
|
|
|
|
kgid_t egid;
|
ipc: optimize semget/shmget/msgget for lots of keys
ipc_findkey() used to scan all objects to look for the wanted key. This
is slow when using a high number of keys. This change adds an rhashtable
of kern_ipc_perm objects in ipc_ids, so that one lookup cease to be O(n).
This change gives a 865% improvement of benchmark reaim.jobs_per_min on a
56 threads Intel(R) Xeon(R) CPU E5-2695 v3 @ 2.30GHz with 256G memory [1]
Other (more micro) benchmark results, by the author: On an i5 laptop, the
following loop executed right after a reboot took, without and with this
change:
for (int i = 0, k=0x424242; i < KEYS; ++i)
semget(k++, 1, IPC_CREAT | 0600);
total total max single max single
KEYS without with call without call with
1 3.5 4.9 µs 3.5 4.9
10 7.6 8.6 µs 3.7 4.7
32 16.2 15.9 µs 4.3 5.3
100 72.9 41.8 µs 3.7 4.7
1000 5,630.0 502.0 µs * *
10000 1,340,000.0 7,240.0 µs * *
31900 17,600,000.0 22,200.0 µs * *
*: unreliable measure: high variance
The duration for a lookup-only usage was obtained by the same loop once
the keys are present:
total total max single max single
KEYS without with call without call with
1 2.1 2.5 µs 2.1 2.5
10 4.5 4.8 µs 2.2 2.3
32 13.0 10.8 µs 2.3 2.8
100 82.9 25.1 µs * 2.3
1000 5,780.0 217.0 µs * *
10000 1,470,000.0 2,520.0 µs * *
31900 17,400,000.0 7,810.0 µs * *
Finally, executing each semget() in a new process gave, when still
summing only the durations of these syscalls:
creation:
total total
KEYS without with
1 3.7 5.0 µs
10 32.9 36.7 µs
32 125.0 109.0 µs
100 523.0 353.0 µs
1000 20,300.0 3,280.0 µs
10000 2,470,000.0 46,700.0 µs
31900 27,800,000.0 219,000.0 µs
lookup-only:
total total
KEYS without with
1 2.5 2.7 µs
10 25.4 24.4 µs
32 106.0 72.6 µs
100 591.0 352.0 µs
1000 22,400.0 2,250.0 µs
10000 2,510,000.0 25,700.0 µs
31900 28,200,000.0 115,000.0 µs
[1] http://lkml.kernel.org/r/20170814060507.GE23258@yexl-desktop
Link: http://lkml.kernel.org/r/20170815194954.ck32ta2z35yuzpwp@debix
Signed-off-by: Guillaume Knispel <guillaume.knispel@supersonicimagine.com>
Reviewed-by: Marc Pardo <marc.pardo@supersonicimagine.com>
Cc: Davidlohr Bueso <dave@stgolabs.net>
Cc: Kees Cook <keescook@chromium.org>
Cc: Manfred Spraul <manfred@colorfullife.com>
Cc: Alexey Dobriyan <adobriyan@gmail.com>
Cc: "Eric W. Biederman" <ebiederm@xmission.com>
Cc: "Peter Zijlstra (Intel)" <peterz@infradead.org>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Cc: Serge Hallyn <serge@hallyn.com>
Cc: Andrey Vagin <avagin@openvz.org>
Cc: Guillaume Knispel <guillaume.knispel@supersonicimagine.com>
Cc: Marc Pardo <marc.pardo@supersonicimagine.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-09-09 02:17:55 +03:00
|
|
|
int id, err;
|
2013-01-05 03:34:50 +04:00
|
|
|
int next_id = ids->next_id;
|
2005-04-17 02:20:36 +04:00
|
|
|
|
2007-10-19 10:40:48 +04:00
|
|
|
if (size > IPCMNI)
|
|
|
|
size = IPCMNI;
|
|
|
|
|
ipc: optimize semget/shmget/msgget for lots of keys
ipc_findkey() used to scan all objects to look for the wanted key. This
is slow when using a high number of keys. This change adds an rhashtable
of kern_ipc_perm objects in ipc_ids, so that one lookup cease to be O(n).
This change gives a 865% improvement of benchmark reaim.jobs_per_min on a
56 threads Intel(R) Xeon(R) CPU E5-2695 v3 @ 2.30GHz with 256G memory [1]
Other (more micro) benchmark results, by the author: On an i5 laptop, the
following loop executed right after a reboot took, without and with this
change:
for (int i = 0, k=0x424242; i < KEYS; ++i)
semget(k++, 1, IPC_CREAT | 0600);
total total max single max single
KEYS without with call without call with
1 3.5 4.9 µs 3.5 4.9
10 7.6 8.6 µs 3.7 4.7
32 16.2 15.9 µs 4.3 5.3
100 72.9 41.8 µs 3.7 4.7
1000 5,630.0 502.0 µs * *
10000 1,340,000.0 7,240.0 µs * *
31900 17,600,000.0 22,200.0 µs * *
*: unreliable measure: high variance
The duration for a lookup-only usage was obtained by the same loop once
the keys are present:
total total max single max single
KEYS without with call without call with
1 2.1 2.5 µs 2.1 2.5
10 4.5 4.8 µs 2.2 2.3
32 13.0 10.8 µs 2.3 2.8
100 82.9 25.1 µs * 2.3
1000 5,780.0 217.0 µs * *
10000 1,470,000.0 2,520.0 µs * *
31900 17,400,000.0 7,810.0 µs * *
Finally, executing each semget() in a new process gave, when still
summing only the durations of these syscalls:
creation:
total total
KEYS without with
1 3.7 5.0 µs
10 32.9 36.7 µs
32 125.0 109.0 µs
100 523.0 353.0 µs
1000 20,300.0 3,280.0 µs
10000 2,470,000.0 46,700.0 µs
31900 27,800,000.0 219,000.0 µs
lookup-only:
total total
KEYS without with
1 2.5 2.7 µs
10 25.4 24.4 µs
32 106.0 72.6 µs
100 591.0 352.0 µs
1000 22,400.0 2,250.0 µs
10000 2,510,000.0 25,700.0 µs
31900 28,200,000.0 115,000.0 µs
[1] http://lkml.kernel.org/r/20170814060507.GE23258@yexl-desktop
Link: http://lkml.kernel.org/r/20170815194954.ck32ta2z35yuzpwp@debix
Signed-off-by: Guillaume Knispel <guillaume.knispel@supersonicimagine.com>
Reviewed-by: Marc Pardo <marc.pardo@supersonicimagine.com>
Cc: Davidlohr Bueso <dave@stgolabs.net>
Cc: Kees Cook <keescook@chromium.org>
Cc: Manfred Spraul <manfred@colorfullife.com>
Cc: Alexey Dobriyan <adobriyan@gmail.com>
Cc: "Eric W. Biederman" <ebiederm@xmission.com>
Cc: "Peter Zijlstra (Intel)" <peterz@infradead.org>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Cc: Serge Hallyn <serge@hallyn.com>
Cc: Andrey Vagin <avagin@openvz.org>
Cc: Guillaume Knispel <guillaume.knispel@supersonicimagine.com>
Cc: Marc Pardo <marc.pardo@supersonicimagine.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-09-09 02:17:55 +03:00
|
|
|
if (!ids->tables_initialized || ids->in_use >= size)
|
2007-10-19 10:40:57 +04:00
|
|
|
return -ENOSPC;
|
2007-10-19 10:40:48 +04:00
|
|
|
|
2013-02-28 05:04:53 +04:00
|
|
|
idr_preload(GFP_KERNEL);
|
|
|
|
|
2017-09-09 02:17:45 +03:00
|
|
|
refcount_set(&new->refcount, 1);
|
2008-11-20 02:36:08 +03:00
|
|
|
spin_lock_init(&new->lock);
|
2014-01-28 05:07:02 +04:00
|
|
|
new->deleted = false;
|
2008-11-20 02:36:08 +03:00
|
|
|
rcu_read_lock();
|
|
|
|
spin_lock(&new->lock);
|
|
|
|
|
2015-09-30 19:48:40 +03:00
|
|
|
current_euid_egid(&euid, &egid);
|
|
|
|
new->cuid = new->uid = euid;
|
|
|
|
new->gid = new->cgid = egid;
|
|
|
|
|
2013-02-28 05:04:53 +04:00
|
|
|
id = idr_alloc(&ids->ipcs_idr, new,
|
|
|
|
(next_id < 0) ? 0 : ipcid_to_idx(next_id), 0,
|
|
|
|
GFP_NOWAIT);
|
|
|
|
idr_preload_end();
|
ipc: optimize semget/shmget/msgget for lots of keys
ipc_findkey() used to scan all objects to look for the wanted key. This
is slow when using a high number of keys. This change adds an rhashtable
of kern_ipc_perm objects in ipc_ids, so that one lookup cease to be O(n).
This change gives a 865% improvement of benchmark reaim.jobs_per_min on a
56 threads Intel(R) Xeon(R) CPU E5-2695 v3 @ 2.30GHz with 256G memory [1]
Other (more micro) benchmark results, by the author: On an i5 laptop, the
following loop executed right after a reboot took, without and with this
change:
for (int i = 0, k=0x424242; i < KEYS; ++i)
semget(k++, 1, IPC_CREAT | 0600);
total total max single max single
KEYS without with call without call with
1 3.5 4.9 µs 3.5 4.9
10 7.6 8.6 µs 3.7 4.7
32 16.2 15.9 µs 4.3 5.3
100 72.9 41.8 µs 3.7 4.7
1000 5,630.0 502.0 µs * *
10000 1,340,000.0 7,240.0 µs * *
31900 17,600,000.0 22,200.0 µs * *
*: unreliable measure: high variance
The duration for a lookup-only usage was obtained by the same loop once
the keys are present:
total total max single max single
KEYS without with call without call with
1 2.1 2.5 µs 2.1 2.5
10 4.5 4.8 µs 2.2 2.3
32 13.0 10.8 µs 2.3 2.8
100 82.9 25.1 µs * 2.3
1000 5,780.0 217.0 µs * *
10000 1,470,000.0 2,520.0 µs * *
31900 17,400,000.0 7,810.0 µs * *
Finally, executing each semget() in a new process gave, when still
summing only the durations of these syscalls:
creation:
total total
KEYS without with
1 3.7 5.0 µs
10 32.9 36.7 µs
32 125.0 109.0 µs
100 523.0 353.0 µs
1000 20,300.0 3,280.0 µs
10000 2,470,000.0 46,700.0 µs
31900 27,800,000.0 219,000.0 µs
lookup-only:
total total
KEYS without with
1 2.5 2.7 µs
10 25.4 24.4 µs
32 106.0 72.6 µs
100 591.0 352.0 µs
1000 22,400.0 2,250.0 µs
10000 2,510,000.0 25,700.0 µs
31900 28,200,000.0 115,000.0 µs
[1] http://lkml.kernel.org/r/20170814060507.GE23258@yexl-desktop
Link: http://lkml.kernel.org/r/20170815194954.ck32ta2z35yuzpwp@debix
Signed-off-by: Guillaume Knispel <guillaume.knispel@supersonicimagine.com>
Reviewed-by: Marc Pardo <marc.pardo@supersonicimagine.com>
Cc: Davidlohr Bueso <dave@stgolabs.net>
Cc: Kees Cook <keescook@chromium.org>
Cc: Manfred Spraul <manfred@colorfullife.com>
Cc: Alexey Dobriyan <adobriyan@gmail.com>
Cc: "Eric W. Biederman" <ebiederm@xmission.com>
Cc: "Peter Zijlstra (Intel)" <peterz@infradead.org>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Cc: Serge Hallyn <serge@hallyn.com>
Cc: Andrey Vagin <avagin@openvz.org>
Cc: Guillaume Knispel <guillaume.knispel@supersonicimagine.com>
Cc: Marc Pardo <marc.pardo@supersonicimagine.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-09-09 02:17:55 +03:00
|
|
|
|
|
|
|
if (id >= 0 && new->key != IPC_PRIVATE) {
|
|
|
|
err = rhashtable_insert_fast(&ids->key_ht, &new->khtnode,
|
|
|
|
ipc_kht_params);
|
|
|
|
if (err < 0) {
|
|
|
|
idr_remove(&ids->ipcs_idr, id);
|
|
|
|
id = err;
|
|
|
|
}
|
|
|
|
}
|
2013-02-28 05:04:53 +04:00
|
|
|
if (id < 0) {
|
2008-11-20 02:36:08 +03:00
|
|
|
spin_unlock(&new->lock);
|
|
|
|
rcu_read_unlock();
|
2013-02-28 05:04:53 +04:00
|
|
|
return id;
|
2008-11-20 02:36:08 +03:00
|
|
|
}
|
2007-10-19 10:40:48 +04:00
|
|
|
|
2005-04-17 02:20:36 +04:00
|
|
|
ids->in_use++;
|
|
|
|
|
2013-01-05 03:34:50 +04:00
|
|
|
if (next_id < 0) {
|
|
|
|
new->seq = ids->seq++;
|
2014-01-28 05:07:09 +04:00
|
|
|
if (ids->seq > IPCID_SEQ_MAX)
|
2013-01-05 03:34:50 +04:00
|
|
|
ids->seq = 0;
|
|
|
|
} else {
|
|
|
|
new->seq = ipcid_to_seqx(next_id);
|
|
|
|
ids->next_id = -1;
|
|
|
|
}
|
2005-04-17 02:20:36 +04:00
|
|
|
|
2008-04-29 12:00:35 +04:00
|
|
|
new->id = ipc_buildid(id, new->seq);
|
2005-04-17 02:20:36 +04:00
|
|
|
return id;
|
|
|
|
}
|
|
|
|
|
2007-10-19 10:40:49 +04:00
|
|
|
/**
|
2014-01-28 05:07:05 +04:00
|
|
|
* ipcget_new - create a new ipc object
|
|
|
|
* @ns: ipc namespace
|
2014-09-08 20:27:23 +04:00
|
|
|
* @ids: ipc identifier set
|
2014-01-28 05:07:05 +04:00
|
|
|
* @ops: the actual creation routine to call
|
|
|
|
* @params: its parameters
|
|
|
|
*
|
|
|
|
* This routine is called by sys_msgget, sys_semget() and sys_shmget()
|
|
|
|
* when the key is IPC_PRIVATE.
|
2007-10-19 10:40:49 +04:00
|
|
|
*/
|
2008-02-08 15:18:54 +03:00
|
|
|
static int ipcget_new(struct ipc_namespace *ns, struct ipc_ids *ids,
|
2014-06-07 01:37:36 +04:00
|
|
|
const struct ipc_ops *ops, struct ipc_params *params)
|
2007-10-19 10:40:49 +04:00
|
|
|
{
|
|
|
|
int err;
|
|
|
|
|
2013-09-12 01:26:24 +04:00
|
|
|
down_write(&ids->rwsem);
|
2007-10-19 10:40:49 +04:00
|
|
|
err = ops->getnew(ns, params);
|
2013-09-12 01:26:24 +04:00
|
|
|
up_write(&ids->rwsem);
|
2007-10-19 10:40:49 +04:00
|
|
|
return err;
|
|
|
|
}
|
|
|
|
|
|
|
|
/**
|
2014-01-28 05:07:05 +04:00
|
|
|
* ipc_check_perms - check security and permissions for an ipc object
|
|
|
|
* @ns: ipc namespace
|
|
|
|
* @ipcp: ipc permission set
|
|
|
|
* @ops: the actual security routine to call
|
|
|
|
* @params: its parameters
|
2007-10-19 10:40:53 +04:00
|
|
|
*
|
2014-01-28 05:07:05 +04:00
|
|
|
* This routine is called by sys_msgget(), sys_semget() and sys_shmget()
|
|
|
|
* when the key is not IPC_PRIVATE and that key already exists in the
|
|
|
|
* ds IDR.
|
2007-10-19 10:40:53 +04:00
|
|
|
*
|
2014-01-28 05:07:05 +04:00
|
|
|
* On success, the ipc id is returned.
|
2007-10-19 10:40:53 +04:00
|
|
|
*
|
2014-01-28 05:07:05 +04:00
|
|
|
* It is called with ipc_ids.rwsem and ipcp->lock held.
|
2007-10-19 10:40:49 +04:00
|
|
|
*/
|
2011-03-24 02:43:24 +03:00
|
|
|
static int ipc_check_perms(struct ipc_namespace *ns,
|
|
|
|
struct kern_ipc_perm *ipcp,
|
2014-06-07 01:37:36 +04:00
|
|
|
const struct ipc_ops *ops,
|
2011-03-24 02:43:24 +03:00
|
|
|
struct ipc_params *params)
|
2007-10-19 10:40:49 +04:00
|
|
|
{
|
|
|
|
int err;
|
|
|
|
|
2011-03-24 02:43:24 +03:00
|
|
|
if (ipcperms(ns, ipcp, params->flg))
|
2007-10-19 10:40:49 +04:00
|
|
|
err = -EACCES;
|
|
|
|
else {
|
|
|
|
err = ops->associate(ipcp, params->flg);
|
|
|
|
if (!err)
|
|
|
|
err = ipcp->id;
|
|
|
|
}
|
|
|
|
|
|
|
|
return err;
|
|
|
|
}
|
|
|
|
|
|
|
|
/**
|
2014-01-28 05:07:05 +04:00
|
|
|
* ipcget_public - get an ipc object or create a new one
|
|
|
|
* @ns: ipc namespace
|
2014-09-08 20:27:23 +04:00
|
|
|
* @ids: ipc identifier set
|
2014-01-28 05:07:05 +04:00
|
|
|
* @ops: the actual creation routine to call
|
|
|
|
* @params: its parameters
|
|
|
|
*
|
|
|
|
* This routine is called by sys_msgget, sys_semget() and sys_shmget()
|
|
|
|
* when the key is not IPC_PRIVATE.
|
|
|
|
* It adds a new entry if the key is not found and does some permission
|
|
|
|
* / security checkings if the key is found.
|
|
|
|
*
|
|
|
|
* On success, the ipc id is returned.
|
2007-10-19 10:40:49 +04:00
|
|
|
*/
|
2008-02-08 15:18:54 +03:00
|
|
|
static int ipcget_public(struct ipc_namespace *ns, struct ipc_ids *ids,
|
2014-06-07 01:37:36 +04:00
|
|
|
const struct ipc_ops *ops, struct ipc_params *params)
|
2007-10-19 10:40:49 +04:00
|
|
|
{
|
|
|
|
struct kern_ipc_perm *ipcp;
|
|
|
|
int flg = params->flg;
|
|
|
|
int err;
|
|
|
|
|
2007-10-19 10:40:54 +04:00
|
|
|
/*
|
|
|
|
* Take the lock as a writer since we are potentially going to add
|
|
|
|
* a new entry + read locks are not "upgradable"
|
|
|
|
*/
|
2013-09-12 01:26:24 +04:00
|
|
|
down_write(&ids->rwsem);
|
2007-10-19 10:40:49 +04:00
|
|
|
ipcp = ipc_findkey(ids, params->key);
|
|
|
|
if (ipcp == NULL) {
|
|
|
|
/* key not used */
|
|
|
|
if (!(flg & IPC_CREAT))
|
|
|
|
err = -ENOENT;
|
|
|
|
else
|
|
|
|
err = ops->getnew(ns, params);
|
|
|
|
} else {
|
|
|
|
/* ipc object has been locked by ipc_findkey() */
|
|
|
|
|
|
|
|
if (flg & IPC_CREAT && flg & IPC_EXCL)
|
|
|
|
err = -EEXIST;
|
|
|
|
else {
|
|
|
|
err = 0;
|
|
|
|
if (ops->more_checks)
|
|
|
|
err = ops->more_checks(ipcp, params);
|
|
|
|
if (!err)
|
2007-10-19 10:40:53 +04:00
|
|
|
/*
|
|
|
|
* ipc_check_perms returns the IPC id on
|
|
|
|
* success
|
|
|
|
*/
|
2011-03-24 02:43:24 +03:00
|
|
|
err = ipc_check_perms(ns, ipcp, ops, params);
|
2007-10-19 10:40:49 +04:00
|
|
|
}
|
|
|
|
ipc_unlock(ipcp);
|
|
|
|
}
|
2013-09-12 01:26:24 +04:00
|
|
|
up_write(&ids->rwsem);
|
2007-10-19 10:40:49 +04:00
|
|
|
|
|
|
|
return err;
|
|
|
|
}
|
|
|
|
|
ipc: optimize semget/shmget/msgget for lots of keys
ipc_findkey() used to scan all objects to look for the wanted key. This
is slow when using a high number of keys. This change adds an rhashtable
of kern_ipc_perm objects in ipc_ids, so that one lookup cease to be O(n).
This change gives a 865% improvement of benchmark reaim.jobs_per_min on a
56 threads Intel(R) Xeon(R) CPU E5-2695 v3 @ 2.30GHz with 256G memory [1]
Other (more micro) benchmark results, by the author: On an i5 laptop, the
following loop executed right after a reboot took, without and with this
change:
for (int i = 0, k=0x424242; i < KEYS; ++i)
semget(k++, 1, IPC_CREAT | 0600);
total total max single max single
KEYS without with call without call with
1 3.5 4.9 µs 3.5 4.9
10 7.6 8.6 µs 3.7 4.7
32 16.2 15.9 µs 4.3 5.3
100 72.9 41.8 µs 3.7 4.7
1000 5,630.0 502.0 µs * *
10000 1,340,000.0 7,240.0 µs * *
31900 17,600,000.0 22,200.0 µs * *
*: unreliable measure: high variance
The duration for a lookup-only usage was obtained by the same loop once
the keys are present:
total total max single max single
KEYS without with call without call with
1 2.1 2.5 µs 2.1 2.5
10 4.5 4.8 µs 2.2 2.3
32 13.0 10.8 µs 2.3 2.8
100 82.9 25.1 µs * 2.3
1000 5,780.0 217.0 µs * *
10000 1,470,000.0 2,520.0 µs * *
31900 17,400,000.0 7,810.0 µs * *
Finally, executing each semget() in a new process gave, when still
summing only the durations of these syscalls:
creation:
total total
KEYS without with
1 3.7 5.0 µs
10 32.9 36.7 µs
32 125.0 109.0 µs
100 523.0 353.0 µs
1000 20,300.0 3,280.0 µs
10000 2,470,000.0 46,700.0 µs
31900 27,800,000.0 219,000.0 µs
lookup-only:
total total
KEYS without with
1 2.5 2.7 µs
10 25.4 24.4 µs
32 106.0 72.6 µs
100 591.0 352.0 µs
1000 22,400.0 2,250.0 µs
10000 2,510,000.0 25,700.0 µs
31900 28,200,000.0 115,000.0 µs
[1] http://lkml.kernel.org/r/20170814060507.GE23258@yexl-desktop
Link: http://lkml.kernel.org/r/20170815194954.ck32ta2z35yuzpwp@debix
Signed-off-by: Guillaume Knispel <guillaume.knispel@supersonicimagine.com>
Reviewed-by: Marc Pardo <marc.pardo@supersonicimagine.com>
Cc: Davidlohr Bueso <dave@stgolabs.net>
Cc: Kees Cook <keescook@chromium.org>
Cc: Manfred Spraul <manfred@colorfullife.com>
Cc: Alexey Dobriyan <adobriyan@gmail.com>
Cc: "Eric W. Biederman" <ebiederm@xmission.com>
Cc: "Peter Zijlstra (Intel)" <peterz@infradead.org>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Cc: Serge Hallyn <serge@hallyn.com>
Cc: Andrey Vagin <avagin@openvz.org>
Cc: Guillaume Knispel <guillaume.knispel@supersonicimagine.com>
Cc: Marc Pardo <marc.pardo@supersonicimagine.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-09-09 02:17:55 +03:00
|
|
|
/**
|
|
|
|
* ipc_kht_remove - remove an ipc from the key hashtable
|
|
|
|
* @ids: ipc identifier set
|
|
|
|
* @ipcp: ipc perm structure containing the key to remove
|
|
|
|
*
|
|
|
|
* ipc_ids.rwsem (as a writer) and the spinlock for this ID are held
|
|
|
|
* before this function is called, and remain locked on the exit.
|
|
|
|
*/
|
|
|
|
static void ipc_kht_remove(struct ipc_ids *ids, struct kern_ipc_perm *ipcp)
|
|
|
|
{
|
|
|
|
if (ipcp->key != IPC_PRIVATE)
|
|
|
|
rhashtable_remove_fast(&ids->key_ht, &ipcp->khtnode,
|
|
|
|
ipc_kht_params);
|
|
|
|
}
|
2007-10-19 10:40:49 +04:00
|
|
|
|
2005-04-17 02:20:36 +04:00
|
|
|
/**
|
2014-01-28 05:07:05 +04:00
|
|
|
* ipc_rmid - remove an ipc identifier
|
|
|
|
* @ids: ipc identifier set
|
|
|
|
* @ipcp: ipc perm structure containing the identifier to remove
|
2005-04-17 02:20:36 +04:00
|
|
|
*
|
2014-01-28 05:07:05 +04:00
|
|
|
* ipc_ids.rwsem (as a writer) and the spinlock for this ID are held
|
|
|
|
* before this function is called, and remain locked on the exit.
|
2005-04-17 02:20:36 +04:00
|
|
|
*/
|
2007-10-19 10:40:48 +04:00
|
|
|
void ipc_rmid(struct ipc_ids *ids, struct kern_ipc_perm *ipcp)
|
2005-04-17 02:20:36 +04:00
|
|
|
{
|
2007-10-19 10:40:52 +04:00
|
|
|
int lid = ipcid_to_idx(ipcp->id);
|
2007-10-19 10:40:48 +04:00
|
|
|
|
|
|
|
idr_remove(&ids->ipcs_idr, lid);
|
ipc: optimize semget/shmget/msgget for lots of keys
ipc_findkey() used to scan all objects to look for the wanted key. This
is slow when using a high number of keys. This change adds an rhashtable
of kern_ipc_perm objects in ipc_ids, so that one lookup cease to be O(n).
This change gives a 865% improvement of benchmark reaim.jobs_per_min on a
56 threads Intel(R) Xeon(R) CPU E5-2695 v3 @ 2.30GHz with 256G memory [1]
Other (more micro) benchmark results, by the author: On an i5 laptop, the
following loop executed right after a reboot took, without and with this
change:
for (int i = 0, k=0x424242; i < KEYS; ++i)
semget(k++, 1, IPC_CREAT | 0600);
total total max single max single
KEYS without with call without call with
1 3.5 4.9 µs 3.5 4.9
10 7.6 8.6 µs 3.7 4.7
32 16.2 15.9 µs 4.3 5.3
100 72.9 41.8 µs 3.7 4.7
1000 5,630.0 502.0 µs * *
10000 1,340,000.0 7,240.0 µs * *
31900 17,600,000.0 22,200.0 µs * *
*: unreliable measure: high variance
The duration for a lookup-only usage was obtained by the same loop once
the keys are present:
total total max single max single
KEYS without with call without call with
1 2.1 2.5 µs 2.1 2.5
10 4.5 4.8 µs 2.2 2.3
32 13.0 10.8 µs 2.3 2.8
100 82.9 25.1 µs * 2.3
1000 5,780.0 217.0 µs * *
10000 1,470,000.0 2,520.0 µs * *
31900 17,400,000.0 7,810.0 µs * *
Finally, executing each semget() in a new process gave, when still
summing only the durations of these syscalls:
creation:
total total
KEYS without with
1 3.7 5.0 µs
10 32.9 36.7 µs
32 125.0 109.0 µs
100 523.0 353.0 µs
1000 20,300.0 3,280.0 µs
10000 2,470,000.0 46,700.0 µs
31900 27,800,000.0 219,000.0 µs
lookup-only:
total total
KEYS without with
1 2.5 2.7 µs
10 25.4 24.4 µs
32 106.0 72.6 µs
100 591.0 352.0 µs
1000 22,400.0 2,250.0 µs
10000 2,510,000.0 25,700.0 µs
31900 28,200,000.0 115,000.0 µs
[1] http://lkml.kernel.org/r/20170814060507.GE23258@yexl-desktop
Link: http://lkml.kernel.org/r/20170815194954.ck32ta2z35yuzpwp@debix
Signed-off-by: Guillaume Knispel <guillaume.knispel@supersonicimagine.com>
Reviewed-by: Marc Pardo <marc.pardo@supersonicimagine.com>
Cc: Davidlohr Bueso <dave@stgolabs.net>
Cc: Kees Cook <keescook@chromium.org>
Cc: Manfred Spraul <manfred@colorfullife.com>
Cc: Alexey Dobriyan <adobriyan@gmail.com>
Cc: "Eric W. Biederman" <ebiederm@xmission.com>
Cc: "Peter Zijlstra (Intel)" <peterz@infradead.org>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Cc: Serge Hallyn <serge@hallyn.com>
Cc: Andrey Vagin <avagin@openvz.org>
Cc: Guillaume Knispel <guillaume.knispel@supersonicimagine.com>
Cc: Marc Pardo <marc.pardo@supersonicimagine.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-09-09 02:17:55 +03:00
|
|
|
ipc_kht_remove(ids, ipcp);
|
2005-04-17 02:20:36 +04:00
|
|
|
ids->in_use--;
|
2014-01-28 05:07:02 +04:00
|
|
|
ipcp->deleted = true;
|
2005-04-17 02:20:36 +04:00
|
|
|
}
|
|
|
|
|
ipc: optimize semget/shmget/msgget for lots of keys
ipc_findkey() used to scan all objects to look for the wanted key. This
is slow when using a high number of keys. This change adds an rhashtable
of kern_ipc_perm objects in ipc_ids, so that one lookup cease to be O(n).
This change gives a 865% improvement of benchmark reaim.jobs_per_min on a
56 threads Intel(R) Xeon(R) CPU E5-2695 v3 @ 2.30GHz with 256G memory [1]
Other (more micro) benchmark results, by the author: On an i5 laptop, the
following loop executed right after a reboot took, without and with this
change:
for (int i = 0, k=0x424242; i < KEYS; ++i)
semget(k++, 1, IPC_CREAT | 0600);
total total max single max single
KEYS without with call without call with
1 3.5 4.9 µs 3.5 4.9
10 7.6 8.6 µs 3.7 4.7
32 16.2 15.9 µs 4.3 5.3
100 72.9 41.8 µs 3.7 4.7
1000 5,630.0 502.0 µs * *
10000 1,340,000.0 7,240.0 µs * *
31900 17,600,000.0 22,200.0 µs * *
*: unreliable measure: high variance
The duration for a lookup-only usage was obtained by the same loop once
the keys are present:
total total max single max single
KEYS without with call without call with
1 2.1 2.5 µs 2.1 2.5
10 4.5 4.8 µs 2.2 2.3
32 13.0 10.8 µs 2.3 2.8
100 82.9 25.1 µs * 2.3
1000 5,780.0 217.0 µs * *
10000 1,470,000.0 2,520.0 µs * *
31900 17,400,000.0 7,810.0 µs * *
Finally, executing each semget() in a new process gave, when still
summing only the durations of these syscalls:
creation:
total total
KEYS without with
1 3.7 5.0 µs
10 32.9 36.7 µs
32 125.0 109.0 µs
100 523.0 353.0 µs
1000 20,300.0 3,280.0 µs
10000 2,470,000.0 46,700.0 µs
31900 27,800,000.0 219,000.0 µs
lookup-only:
total total
KEYS without with
1 2.5 2.7 µs
10 25.4 24.4 µs
32 106.0 72.6 µs
100 591.0 352.0 µs
1000 22,400.0 2,250.0 µs
10000 2,510,000.0 25,700.0 µs
31900 28,200,000.0 115,000.0 µs
[1] http://lkml.kernel.org/r/20170814060507.GE23258@yexl-desktop
Link: http://lkml.kernel.org/r/20170815194954.ck32ta2z35yuzpwp@debix
Signed-off-by: Guillaume Knispel <guillaume.knispel@supersonicimagine.com>
Reviewed-by: Marc Pardo <marc.pardo@supersonicimagine.com>
Cc: Davidlohr Bueso <dave@stgolabs.net>
Cc: Kees Cook <keescook@chromium.org>
Cc: Manfred Spraul <manfred@colorfullife.com>
Cc: Alexey Dobriyan <adobriyan@gmail.com>
Cc: "Eric W. Biederman" <ebiederm@xmission.com>
Cc: "Peter Zijlstra (Intel)" <peterz@infradead.org>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Cc: Serge Hallyn <serge@hallyn.com>
Cc: Andrey Vagin <avagin@openvz.org>
Cc: Guillaume Knispel <guillaume.knispel@supersonicimagine.com>
Cc: Marc Pardo <marc.pardo@supersonicimagine.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-09-09 02:17:55 +03:00
|
|
|
/**
|
|
|
|
* ipc_set_key_private - switch the key of an existing ipc to IPC_PRIVATE
|
|
|
|
* @ids: ipc identifier set
|
|
|
|
* @ipcp: ipc perm structure containing the key to modify
|
|
|
|
*
|
|
|
|
* ipc_ids.rwsem (as a writer) and the spinlock for this ID are held
|
|
|
|
* before this function is called, and remain locked on the exit.
|
|
|
|
*/
|
|
|
|
void ipc_set_key_private(struct ipc_ids *ids, struct kern_ipc_perm *ipcp)
|
|
|
|
{
|
|
|
|
ipc_kht_remove(ids, ipcp);
|
|
|
|
ipcp->key = IPC_PRIVATE;
|
|
|
|
}
|
|
|
|
|
2017-07-13 00:34:41 +03:00
|
|
|
int ipc_rcu_getref(struct kern_ipc_perm *ptr)
|
2005-04-17 02:20:36 +04:00
|
|
|
{
|
2017-09-09 02:17:45 +03:00
|
|
|
return refcount_inc_not_zero(&ptr->refcount);
|
2006-11-22 17:55:48 +03:00
|
|
|
}
|
|
|
|
|
2017-07-13 00:34:41 +03:00
|
|
|
void ipc_rcu_putref(struct kern_ipc_perm *ptr,
|
|
|
|
void (*func)(struct rcu_head *head))
|
2005-04-17 02:20:36 +04:00
|
|
|
{
|
2017-09-09 02:17:45 +03:00
|
|
|
if (!refcount_dec_and_test(&ptr->refcount))
|
2005-04-17 02:20:36 +04:00
|
|
|
return;
|
|
|
|
|
2017-07-13 00:34:41 +03:00
|
|
|
call_rcu(&ptr->rcu, func);
|
ipc: fix race with LSMs
Currently, IPC mechanisms do security and auditing related checks under
RCU. However, since security modules can free the security structure,
for example, through selinux_[sem,msg_queue,shm]_free_security(), we can
race if the structure is freed before other tasks are done with it,
creating a use-after-free condition. Manfred illustrates this nicely,
for instance with shared mem and selinux:
-> do_shmat calls rcu_read_lock()
-> do_shmat calls shm_object_check().
Checks that the object is still valid - but doesn't acquire any locks.
Then it returns.
-> do_shmat calls security_shm_shmat (e.g. selinux_shm_shmat)
-> selinux_shm_shmat calls ipc_has_perm()
-> ipc_has_perm accesses ipc_perms->security
shm_close()
-> shm_close acquires rw_mutex & shm_lock
-> shm_close calls shm_destroy
-> shm_destroy calls security_shm_free (e.g. selinux_shm_free_security)
-> selinux_shm_free_security calls ipc_free_security(&shp->shm_perm)
-> ipc_free_security calls kfree(ipc_perms->security)
This patch delays the freeing of the security structures after all RCU
readers are done. Furthermore it aligns the security life cycle with
that of the rest of IPC - freeing them based on the reference counter.
For situations where we need not free security, the current behavior is
kept. Linus states:
"... the old behavior was suspect for another reason too: having the
security blob go away from under a user sounds like it could cause
various other problems anyway, so I think the old code was at least
_prone_ to bugs even if it didn't have catastrophic behavior."
I have tested this patch with IPC testcases from LTP on both my
quad-core laptop and on a 64 core NUMA server. In both cases selinux is
enabled, and tests pass for both voluntary and forced preemption models.
While the mentioned races are theoretical (at least no one as reported
them), I wanted to make sure that this new logic doesn't break anything
we weren't aware of.
Suggested-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Davidlohr Bueso <davidlohr@hp.com>
Acked-by: Manfred Spraul <manfred@colorfullife.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2013-09-24 04:04:45 +04:00
|
|
|
}
|
|
|
|
|
2005-04-17 02:20:36 +04:00
|
|
|
/**
|
2014-01-28 05:07:05 +04:00
|
|
|
* ipcperms - check ipc permissions
|
|
|
|
* @ns: ipc namespace
|
|
|
|
* @ipcp: ipc permission set
|
|
|
|
* @flag: desired permission set
|
2005-04-17 02:20:36 +04:00
|
|
|
*
|
2014-01-28 05:07:05 +04:00
|
|
|
* Check user, group, other permissions for access
|
|
|
|
* to ipc resources. return 0 if allowed
|
2011-03-24 02:43:24 +03:00
|
|
|
*
|
2017-03-30 23:11:36 +03:00
|
|
|
* @flag will most probably be 0 or ``S_...UGO`` from <linux/stat.h>
|
2005-04-17 02:20:36 +04:00
|
|
|
*/
|
2011-03-24 02:43:24 +03:00
|
|
|
int ipcperms(struct ipc_namespace *ns, struct kern_ipc_perm *ipcp, short flag)
|
|
|
|
{
|
2012-02-08 04:54:11 +04:00
|
|
|
kuid_t euid = current_euid();
|
2008-12-10 11:40:06 +03:00
|
|
|
int requested_mode, granted_mode;
|
2005-04-17 02:20:36 +04:00
|
|
|
|
2008-12-10 11:40:06 +03:00
|
|
|
audit_ipc_obj(ipcp);
|
2005-04-17 02:20:36 +04:00
|
|
|
requested_mode = (flag >> 6) | (flag >> 3) | flag;
|
|
|
|
granted_mode = ipcp->mode;
|
2012-02-08 04:54:11 +04:00
|
|
|
if (uid_eq(euid, ipcp->cuid) ||
|
|
|
|
uid_eq(euid, ipcp->uid))
|
2005-04-17 02:20:36 +04:00
|
|
|
granted_mode >>= 6;
|
|
|
|
else if (in_group_p(ipcp->cgid) || in_group_p(ipcp->gid))
|
|
|
|
granted_mode >>= 3;
|
|
|
|
/* is there some bit set in requested_mode but not in granted_mode? */
|
2014-06-07 01:37:37 +04:00
|
|
|
if ((requested_mode & ~granted_mode & 0007) &&
|
2011-03-24 02:43:24 +03:00
|
|
|
!ns_capable(ns->user_ns, CAP_IPC_OWNER))
|
2005-04-17 02:20:36 +04:00
|
|
|
return -1;
|
|
|
|
|
|
|
|
return security_ipc_permission(ipcp, flag);
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Functions to convert between the kern_ipc_perm structure and the
|
|
|
|
* old/new ipc_perm structures
|
|
|
|
*/
|
|
|
|
|
|
|
|
/**
|
2014-01-28 05:07:05 +04:00
|
|
|
* kernel_to_ipc64_perm - convert kernel ipc permissions to user
|
|
|
|
* @in: kernel permissions
|
|
|
|
* @out: new style ipc permissions
|
2005-04-17 02:20:36 +04:00
|
|
|
*
|
2014-01-28 05:07:05 +04:00
|
|
|
* Turn the kernel object @in into a set of permissions descriptions
|
|
|
|
* for returning to userspace (@out).
|
2005-04-17 02:20:36 +04:00
|
|
|
*/
|
2014-01-28 05:07:04 +04:00
|
|
|
void kernel_to_ipc64_perm(struct kern_ipc_perm *in, struct ipc64_perm *out)
|
2005-04-17 02:20:36 +04:00
|
|
|
{
|
|
|
|
out->key = in->key;
|
2012-02-08 04:54:11 +04:00
|
|
|
out->uid = from_kuid_munged(current_user_ns(), in->uid);
|
|
|
|
out->gid = from_kgid_munged(current_user_ns(), in->gid);
|
|
|
|
out->cuid = from_kuid_munged(current_user_ns(), in->cuid);
|
|
|
|
out->cgid = from_kgid_munged(current_user_ns(), in->cgid);
|
2005-04-17 02:20:36 +04:00
|
|
|
out->mode = in->mode;
|
|
|
|
out->seq = in->seq;
|
|
|
|
}
|
|
|
|
|
|
|
|
/**
|
2014-01-28 05:07:05 +04:00
|
|
|
* ipc64_perm_to_ipc_perm - convert new ipc permissions to old
|
|
|
|
* @in: new style ipc permissions
|
|
|
|
* @out: old style ipc permissions
|
2005-04-17 02:20:36 +04:00
|
|
|
*
|
2014-01-28 05:07:05 +04:00
|
|
|
* Turn the new style permissions object @in into a compatibility
|
|
|
|
* object and store it into the @out pointer.
|
2005-04-17 02:20:36 +04:00
|
|
|
*/
|
2014-01-28 05:07:04 +04:00
|
|
|
void ipc64_perm_to_ipc_perm(struct ipc64_perm *in, struct ipc_perm *out)
|
2005-04-17 02:20:36 +04:00
|
|
|
{
|
|
|
|
out->key = in->key;
|
|
|
|
SET_UID(out->uid, in->uid);
|
|
|
|
SET_GID(out->gid, in->gid);
|
|
|
|
SET_UID(out->cuid, in->cuid);
|
|
|
|
SET_GID(out->cgid, in->cgid);
|
|
|
|
out->mode = in->mode;
|
|
|
|
out->seq = in->seq;
|
|
|
|
}
|
|
|
|
|
2013-05-01 06:15:19 +04:00
|
|
|
/**
|
ipc: optimize semget/shmget/msgget for lots of keys
ipc_findkey() used to scan all objects to look for the wanted key. This
is slow when using a high number of keys. This change adds an rhashtable
of kern_ipc_perm objects in ipc_ids, so that one lookup cease to be O(n).
This change gives a 865% improvement of benchmark reaim.jobs_per_min on a
56 threads Intel(R) Xeon(R) CPU E5-2695 v3 @ 2.30GHz with 256G memory [1]
Other (more micro) benchmark results, by the author: On an i5 laptop, the
following loop executed right after a reboot took, without and with this
change:
for (int i = 0, k=0x424242; i < KEYS; ++i)
semget(k++, 1, IPC_CREAT | 0600);
total total max single max single
KEYS without with call without call with
1 3.5 4.9 µs 3.5 4.9
10 7.6 8.6 µs 3.7 4.7
32 16.2 15.9 µs 4.3 5.3
100 72.9 41.8 µs 3.7 4.7
1000 5,630.0 502.0 µs * *
10000 1,340,000.0 7,240.0 µs * *
31900 17,600,000.0 22,200.0 µs * *
*: unreliable measure: high variance
The duration for a lookup-only usage was obtained by the same loop once
the keys are present:
total total max single max single
KEYS without with call without call with
1 2.1 2.5 µs 2.1 2.5
10 4.5 4.8 µs 2.2 2.3
32 13.0 10.8 µs 2.3 2.8
100 82.9 25.1 µs * 2.3
1000 5,780.0 217.0 µs * *
10000 1,470,000.0 2,520.0 µs * *
31900 17,400,000.0 7,810.0 µs * *
Finally, executing each semget() in a new process gave, when still
summing only the durations of these syscalls:
creation:
total total
KEYS without with
1 3.7 5.0 µs
10 32.9 36.7 µs
32 125.0 109.0 µs
100 523.0 353.0 µs
1000 20,300.0 3,280.0 µs
10000 2,470,000.0 46,700.0 µs
31900 27,800,000.0 219,000.0 µs
lookup-only:
total total
KEYS without with
1 2.5 2.7 µs
10 25.4 24.4 µs
32 106.0 72.6 µs
100 591.0 352.0 µs
1000 22,400.0 2,250.0 µs
10000 2,510,000.0 25,700.0 µs
31900 28,200,000.0 115,000.0 µs
[1] http://lkml.kernel.org/r/20170814060507.GE23258@yexl-desktop
Link: http://lkml.kernel.org/r/20170815194954.ck32ta2z35yuzpwp@debix
Signed-off-by: Guillaume Knispel <guillaume.knispel@supersonicimagine.com>
Reviewed-by: Marc Pardo <marc.pardo@supersonicimagine.com>
Cc: Davidlohr Bueso <dave@stgolabs.net>
Cc: Kees Cook <keescook@chromium.org>
Cc: Manfred Spraul <manfred@colorfullife.com>
Cc: Alexey Dobriyan <adobriyan@gmail.com>
Cc: "Eric W. Biederman" <ebiederm@xmission.com>
Cc: "Peter Zijlstra (Intel)" <peterz@infradead.org>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Cc: Serge Hallyn <serge@hallyn.com>
Cc: Andrey Vagin <avagin@openvz.org>
Cc: Guillaume Knispel <guillaume.knispel@supersonicimagine.com>
Cc: Marc Pardo <marc.pardo@supersonicimagine.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-09-09 02:17:55 +03:00
|
|
|
* ipc_obtain_object_idr
|
2013-05-01 06:15:19 +04:00
|
|
|
* @ids: ipc identifier set
|
|
|
|
* @id: ipc id to look for
|
|
|
|
*
|
|
|
|
* Look for an id in the ipc ids idr and return associated ipc object.
|
|
|
|
*
|
|
|
|
* Call inside the RCU critical section.
|
|
|
|
* The ipc object is *not* locked on exit.
|
|
|
|
*/
|
2015-07-01 00:58:42 +03:00
|
|
|
struct kern_ipc_perm *ipc_obtain_object_idr(struct ipc_ids *ids, int id)
|
2013-05-01 06:15:19 +04:00
|
|
|
{
|
|
|
|
struct kern_ipc_perm *out;
|
|
|
|
int lid = ipcid_to_idx(id);
|
|
|
|
|
ipc: optimize semget/shmget/msgget for lots of keys
ipc_findkey() used to scan all objects to look for the wanted key. This
is slow when using a high number of keys. This change adds an rhashtable
of kern_ipc_perm objects in ipc_ids, so that one lookup cease to be O(n).
This change gives a 865% improvement of benchmark reaim.jobs_per_min on a
56 threads Intel(R) Xeon(R) CPU E5-2695 v3 @ 2.30GHz with 256G memory [1]
Other (more micro) benchmark results, by the author: On an i5 laptop, the
following loop executed right after a reboot took, without and with this
change:
for (int i = 0, k=0x424242; i < KEYS; ++i)
semget(k++, 1, IPC_CREAT | 0600);
total total max single max single
KEYS without with call without call with
1 3.5 4.9 µs 3.5 4.9
10 7.6 8.6 µs 3.7 4.7
32 16.2 15.9 µs 4.3 5.3
100 72.9 41.8 µs 3.7 4.7
1000 5,630.0 502.0 µs * *
10000 1,340,000.0 7,240.0 µs * *
31900 17,600,000.0 22,200.0 µs * *
*: unreliable measure: high variance
The duration for a lookup-only usage was obtained by the same loop once
the keys are present:
total total max single max single
KEYS without with call without call with
1 2.1 2.5 µs 2.1 2.5
10 4.5 4.8 µs 2.2 2.3
32 13.0 10.8 µs 2.3 2.8
100 82.9 25.1 µs * 2.3
1000 5,780.0 217.0 µs * *
10000 1,470,000.0 2,520.0 µs * *
31900 17,400,000.0 7,810.0 µs * *
Finally, executing each semget() in a new process gave, when still
summing only the durations of these syscalls:
creation:
total total
KEYS without with
1 3.7 5.0 µs
10 32.9 36.7 µs
32 125.0 109.0 µs
100 523.0 353.0 µs
1000 20,300.0 3,280.0 µs
10000 2,470,000.0 46,700.0 µs
31900 27,800,000.0 219,000.0 µs
lookup-only:
total total
KEYS without with
1 2.5 2.7 µs
10 25.4 24.4 µs
32 106.0 72.6 µs
100 591.0 352.0 µs
1000 22,400.0 2,250.0 µs
10000 2,510,000.0 25,700.0 µs
31900 28,200,000.0 115,000.0 µs
[1] http://lkml.kernel.org/r/20170814060507.GE23258@yexl-desktop
Link: http://lkml.kernel.org/r/20170815194954.ck32ta2z35yuzpwp@debix
Signed-off-by: Guillaume Knispel <guillaume.knispel@supersonicimagine.com>
Reviewed-by: Marc Pardo <marc.pardo@supersonicimagine.com>
Cc: Davidlohr Bueso <dave@stgolabs.net>
Cc: Kees Cook <keescook@chromium.org>
Cc: Manfred Spraul <manfred@colorfullife.com>
Cc: Alexey Dobriyan <adobriyan@gmail.com>
Cc: "Eric W. Biederman" <ebiederm@xmission.com>
Cc: "Peter Zijlstra (Intel)" <peterz@infradead.org>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Cc: Serge Hallyn <serge@hallyn.com>
Cc: Andrey Vagin <avagin@openvz.org>
Cc: Guillaume Knispel <guillaume.knispel@supersonicimagine.com>
Cc: Marc Pardo <marc.pardo@supersonicimagine.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-09-09 02:17:55 +03:00
|
|
|
if (unlikely(!ids->tables_initialized))
|
|
|
|
return ERR_PTR(-EINVAL);
|
|
|
|
|
2013-05-01 06:15:19 +04:00
|
|
|
out = idr_find(&ids->ipcs_idr, lid);
|
|
|
|
if (!out)
|
|
|
|
return ERR_PTR(-EINVAL);
|
|
|
|
|
|
|
|
return out;
|
|
|
|
}
|
|
|
|
|
2007-10-19 10:40:53 +04:00
|
|
|
/**
|
2014-01-28 05:07:05 +04:00
|
|
|
* ipc_lock - lock an ipc structure without rwsem held
|
|
|
|
* @ids: ipc identifier set
|
2007-10-19 10:40:53 +04:00
|
|
|
* @id: ipc id to look for
|
|
|
|
*
|
|
|
|
* Look for an id in the ipc ids idr and lock the associated ipc object.
|
|
|
|
*
|
2013-05-01 06:15:19 +04:00
|
|
|
* The ipc object is locked on successful exit.
|
2007-10-19 10:40:53 +04:00
|
|
|
*/
|
2007-10-19 10:40:48 +04:00
|
|
|
struct kern_ipc_perm *ipc_lock(struct ipc_ids *ids, int id)
|
2005-04-17 02:20:36 +04:00
|
|
|
{
|
2007-10-19 10:40:48 +04:00
|
|
|
struct kern_ipc_perm *out;
|
2005-04-17 02:20:36 +04:00
|
|
|
|
|
|
|
rcu_read_lock();
|
2015-07-01 00:58:42 +03:00
|
|
|
out = ipc_obtain_object_idr(ids, id);
|
2013-05-01 06:15:19 +04:00
|
|
|
if (IS_ERR(out))
|
ipc,sysv: make return -EIDRM when racing with RMID consistent
The ipc_lock helper is used by all forms of sysv ipc to acquire the ipc
object's spinlock. Upon error (bogus identifier), we always return
-EINVAL, whether the problem be in the idr path or because we raced with a
task performing RMID. For the later, however, all ipc related manpages,
state the that for:
EIDRM <ID> points to a removed identifier.
And return:
EINVAL Invalid <ID> value, or unaligned, etc.
Which (EINVAL) should only return once the ipc resource is deleted. For
all types of ipc this is done immediately upon a RMID command. However,
shared memory behaves slightly different as it can merely mark a segment
for deletion, and delay the actual freeing until there are no more active
consumers. Per shmctl(IPC_RMID) manpage:
""
Mark the segment to be destroyed. The segment will only actually
be destroyed after the last process detaches it (i.e., when the
shm_nattch member of the associated structure shmid_ds is zero).
""
Unlike ipc_lock, paths that behave "correctly", at least per the manpage,
involve controlling the ipc resource via *ctl(), doing the exact same
validity check as ipc_lock after right acquiring the spinlock:
if (!ipc_valid_object()) {
err = -EIDRM;
goto out_unlock;
}
Thus make ipc_lock consistent with the rest of ipc code and return -EIDRM
in ipc_lock when !ipc_valid_object().
Signed-off-by: Davidlohr Bueso <dbueso@suse.de>
Cc: Manfred Spraul <manfred@colorfullife.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-07-01 00:58:45 +03:00
|
|
|
goto err;
|
2007-10-19 10:40:48 +04:00
|
|
|
|
2005-04-17 02:20:36 +04:00
|
|
|
spin_lock(&out->lock);
|
2013-05-01 06:15:19 +04:00
|
|
|
|
ipc,sysv: make return -EIDRM when racing with RMID consistent
The ipc_lock helper is used by all forms of sysv ipc to acquire the ipc
object's spinlock. Upon error (bogus identifier), we always return
-EINVAL, whether the problem be in the idr path or because we raced with a
task performing RMID. For the later, however, all ipc related manpages,
state the that for:
EIDRM <ID> points to a removed identifier.
And return:
EINVAL Invalid <ID> value, or unaligned, etc.
Which (EINVAL) should only return once the ipc resource is deleted. For
all types of ipc this is done immediately upon a RMID command. However,
shared memory behaves slightly different as it can merely mark a segment
for deletion, and delay the actual freeing until there are no more active
consumers. Per shmctl(IPC_RMID) manpage:
""
Mark the segment to be destroyed. The segment will only actually
be destroyed after the last process detaches it (i.e., when the
shm_nattch member of the associated structure shmid_ds is zero).
""
Unlike ipc_lock, paths that behave "correctly", at least per the manpage,
involve controlling the ipc resource via *ctl(), doing the exact same
validity check as ipc_lock after right acquiring the spinlock:
if (!ipc_valid_object()) {
err = -EIDRM;
goto out_unlock;
}
Thus make ipc_lock consistent with the rest of ipc code and return -EIDRM
in ipc_lock when !ipc_valid_object().
Signed-off-by: Davidlohr Bueso <dbueso@suse.de>
Cc: Manfred Spraul <manfred@colorfullife.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-07-01 00:58:45 +03:00
|
|
|
/*
|
|
|
|
* ipc_rmid() may have already freed the ID while ipc_lock()
|
|
|
|
* was spinning: here verify that the structure is still valid.
|
|
|
|
* Upon races with RMID, return -EIDRM, thus indicating that
|
|
|
|
* the ID points to a removed identifier.
|
2005-04-17 02:20:36 +04:00
|
|
|
*/
|
2014-01-28 05:07:02 +04:00
|
|
|
if (ipc_valid_object(out))
|
2013-05-01 06:15:19 +04:00
|
|
|
return out;
|
|
|
|
|
|
|
|
spin_unlock(&out->lock);
|
ipc,sysv: make return -EIDRM when racing with RMID consistent
The ipc_lock helper is used by all forms of sysv ipc to acquire the ipc
object's spinlock. Upon error (bogus identifier), we always return
-EINVAL, whether the problem be in the idr path or because we raced with a
task performing RMID. For the later, however, all ipc related manpages,
state the that for:
EIDRM <ID> points to a removed identifier.
And return:
EINVAL Invalid <ID> value, or unaligned, etc.
Which (EINVAL) should only return once the ipc resource is deleted. For
all types of ipc this is done immediately upon a RMID command. However,
shared memory behaves slightly different as it can merely mark a segment
for deletion, and delay the actual freeing until there are no more active
consumers. Per shmctl(IPC_RMID) manpage:
""
Mark the segment to be destroyed. The segment will only actually
be destroyed after the last process detaches it (i.e., when the
shm_nattch member of the associated structure shmid_ds is zero).
""
Unlike ipc_lock, paths that behave "correctly", at least per the manpage,
involve controlling the ipc resource via *ctl(), doing the exact same
validity check as ipc_lock after right acquiring the spinlock:
if (!ipc_valid_object()) {
err = -EIDRM;
goto out_unlock;
}
Thus make ipc_lock consistent with the rest of ipc code and return -EIDRM
in ipc_lock when !ipc_valid_object().
Signed-off-by: Davidlohr Bueso <dbueso@suse.de>
Cc: Manfred Spraul <manfred@colorfullife.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-07-01 00:58:45 +03:00
|
|
|
out = ERR_PTR(-EIDRM);
|
|
|
|
err:
|
2013-05-01 06:15:19 +04:00
|
|
|
rcu_read_unlock();
|
|
|
|
return out;
|
|
|
|
}
|
2007-10-19 10:40:48 +04:00
|
|
|
|
2013-05-01 06:15:19 +04:00
|
|
|
/**
|
|
|
|
* ipc_obtain_object_check
|
|
|
|
* @ids: ipc identifier set
|
|
|
|
* @id: ipc id to look for
|
|
|
|
*
|
2015-07-01 00:58:42 +03:00
|
|
|
* Similar to ipc_obtain_object_idr() but also checks
|
2013-05-01 06:15:19 +04:00
|
|
|
* the ipc object reference counter.
|
|
|
|
*
|
|
|
|
* Call inside the RCU critical section.
|
|
|
|
* The ipc object is *not* locked on exit.
|
|
|
|
*/
|
|
|
|
struct kern_ipc_perm *ipc_obtain_object_check(struct ipc_ids *ids, int id)
|
|
|
|
{
|
2015-07-01 00:58:42 +03:00
|
|
|
struct kern_ipc_perm *out = ipc_obtain_object_idr(ids, id);
|
2013-05-01 06:15:19 +04:00
|
|
|
|
|
|
|
if (IS_ERR(out))
|
|
|
|
goto out;
|
|
|
|
|
|
|
|
if (ipc_checkid(out, id))
|
2015-07-01 00:58:48 +03:00
|
|
|
return ERR_PTR(-EINVAL);
|
2013-05-01 06:15:19 +04:00
|
|
|
out:
|
2005-04-17 02:20:36 +04:00
|
|
|
return out;
|
|
|
|
}
|
|
|
|
|
2008-02-08 15:18:54 +03:00
|
|
|
/**
|
|
|
|
* ipcget - Common sys_*get() code
|
2014-09-08 20:27:23 +04:00
|
|
|
* @ns: namespace
|
2014-01-28 05:07:05 +04:00
|
|
|
* @ids: ipc identifier set
|
|
|
|
* @ops: operations to be called on ipc object creation, permission checks
|
|
|
|
* and further checks
|
|
|
|
* @params: the parameters needed by the previous operations.
|
2008-02-08 15:18:54 +03:00
|
|
|
*
|
|
|
|
* Common routine called by sys_msgget(), sys_semget() and sys_shmget().
|
|
|
|
*/
|
|
|
|
int ipcget(struct ipc_namespace *ns, struct ipc_ids *ids,
|
2014-06-07 01:37:36 +04:00
|
|
|
const struct ipc_ops *ops, struct ipc_params *params)
|
2008-02-08 15:18:54 +03:00
|
|
|
{
|
|
|
|
if (params->key == IPC_PRIVATE)
|
|
|
|
return ipcget_new(ns, ids, ops, params);
|
|
|
|
else
|
|
|
|
return ipcget_public(ns, ids, ops, params);
|
|
|
|
}
|
|
|
|
|
2008-04-29 12:00:51 +04:00
|
|
|
/**
|
2014-01-28 05:07:05 +04:00
|
|
|
* ipc_update_perm - update the permissions of an ipc object
|
2008-04-29 12:00:51 +04:00
|
|
|
* @in: the permission given as input.
|
|
|
|
* @out: the permission of the ipc to set.
|
|
|
|
*/
|
2012-02-08 04:54:11 +04:00
|
|
|
int ipc_update_perm(struct ipc64_perm *in, struct kern_ipc_perm *out)
|
2008-04-29 12:00:51 +04:00
|
|
|
{
|
2012-02-08 04:54:11 +04:00
|
|
|
kuid_t uid = make_kuid(current_user_ns(), in->uid);
|
|
|
|
kgid_t gid = make_kgid(current_user_ns(), in->gid);
|
|
|
|
if (!uid_valid(uid) || !gid_valid(gid))
|
|
|
|
return -EINVAL;
|
|
|
|
|
|
|
|
out->uid = uid;
|
|
|
|
out->gid = gid;
|
2008-04-29 12:00:51 +04:00
|
|
|
out->mode = (out->mode & ~S_IRWXUGO)
|
|
|
|
| (in->mode & S_IRWXUGO);
|
2012-02-08 04:54:11 +04:00
|
|
|
|
|
|
|
return 0;
|
2008-04-29 12:00:51 +04:00
|
|
|
}
|
|
|
|
|
2008-04-29 12:00:54 +04:00
|
|
|
/**
|
2013-09-12 01:26:17 +04:00
|
|
|
* ipcctl_pre_down_nolock - retrieve an ipc and check permissions for some IPC_XXX cmd
|
2014-01-28 05:07:05 +04:00
|
|
|
* @ns: ipc namespace
|
2008-04-29 12:00:54 +04:00
|
|
|
* @ids: the table of ids where to look for the ipc
|
|
|
|
* @id: the id of the ipc to retrieve
|
|
|
|
* @cmd: the cmd to check
|
|
|
|
* @perm: the permission to set
|
|
|
|
* @extra_perm: one extra permission parameter used by msq
|
|
|
|
*
|
|
|
|
* This function does some common audit and permissions check for some IPC_XXX
|
|
|
|
* cmd and is called from semctl_down, shmctl_down and msgctl_down.
|
2017-03-30 23:11:36 +03:00
|
|
|
* It must be called without any lock held and:
|
|
|
|
*
|
|
|
|
* - retrieves the ipc with the given id in the given table.
|
|
|
|
* - performs some audit and permission check, depending on the given cmd
|
|
|
|
* - returns a pointer to the ipc object or otherwise, the corresponding
|
|
|
|
* error.
|
2013-07-09 03:01:12 +04:00
|
|
|
*
|
2013-09-12 01:26:24 +04:00
|
|
|
* Call holding the both the rwsem and the rcu read lock.
|
2008-04-29 12:00:54 +04:00
|
|
|
*/
|
2013-05-01 06:15:24 +04:00
|
|
|
struct kern_ipc_perm *ipcctl_pre_down_nolock(struct ipc_namespace *ns,
|
2013-09-12 01:26:17 +04:00
|
|
|
struct ipc_ids *ids, int id, int cmd,
|
|
|
|
struct ipc64_perm *perm, int extra_perm)
|
2013-05-01 06:15:24 +04:00
|
|
|
{
|
2012-02-08 04:54:11 +04:00
|
|
|
kuid_t euid;
|
2013-05-01 06:15:24 +04:00
|
|
|
int err = -EPERM;
|
|
|
|
struct kern_ipc_perm *ipcp;
|
2008-04-29 12:00:54 +04:00
|
|
|
|
2013-05-01 06:15:24 +04:00
|
|
|
ipcp = ipc_obtain_object_check(ids, id);
|
2008-04-29 12:00:54 +04:00
|
|
|
if (IS_ERR(ipcp)) {
|
|
|
|
err = PTR_ERR(ipcp);
|
2013-07-09 03:01:12 +04:00
|
|
|
goto err;
|
2008-04-29 12:00:54 +04:00
|
|
|
}
|
|
|
|
|
2008-12-10 11:40:06 +03:00
|
|
|
audit_ipc_obj(ipcp);
|
2008-12-10 11:47:15 +03:00
|
|
|
if (cmd == IPC_SET)
|
|
|
|
audit_ipc_set_perm(extra_perm, perm->uid,
|
2013-05-01 06:15:24 +04:00
|
|
|
perm->gid, perm->mode);
|
2008-11-14 02:39:06 +03:00
|
|
|
|
|
|
|
euid = current_euid();
|
2012-02-08 04:54:11 +04:00
|
|
|
if (uid_eq(euid, ipcp->cuid) || uid_eq(euid, ipcp->uid) ||
|
2011-03-24 02:43:24 +03:00
|
|
|
ns_capable(ns->user_ns, CAP_SYS_ADMIN))
|
2013-07-09 03:01:12 +04:00
|
|
|
return ipcp; /* successful lookup */
|
|
|
|
err:
|
2008-04-29 12:00:54 +04:00
|
|
|
return ERR_PTR(err);
|
|
|
|
}
|
|
|
|
|
2012-07-31 01:42:46 +04:00
|
|
|
#ifdef CONFIG_ARCH_WANT_IPC_PARSE_VERSION
|
2005-04-17 02:20:36 +04:00
|
|
|
|
|
|
|
|
|
|
|
/**
|
2014-01-28 05:07:05 +04:00
|
|
|
* ipc_parse_version - ipc call version
|
|
|
|
* @cmd: pointer to command
|
2005-04-17 02:20:36 +04:00
|
|
|
*
|
2014-01-28 05:07:05 +04:00
|
|
|
* Return IPC_64 for new style IPC and IPC_OLD for old style IPC.
|
|
|
|
* The @cmd value is turned from an encoding command and version into
|
|
|
|
* just the command code.
|
2005-04-17 02:20:36 +04:00
|
|
|
*/
|
2014-01-28 05:07:04 +04:00
|
|
|
int ipc_parse_version(int *cmd)
|
2005-04-17 02:20:36 +04:00
|
|
|
{
|
|
|
|
if (*cmd & IPC_64) {
|
|
|
|
*cmd ^= IPC_64;
|
|
|
|
return IPC_64;
|
|
|
|
} else {
|
|
|
|
return IPC_OLD;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2012-07-31 01:42:46 +04:00
|
|
|
#endif /* CONFIG_ARCH_WANT_IPC_PARSE_VERSION */
|
2005-09-07 02:17:09 +04:00
|
|
|
|
|
|
|
#ifdef CONFIG_PROC_FS
|
2007-02-12 11:52:10 +03:00
|
|
|
struct ipc_proc_iter {
|
|
|
|
struct ipc_namespace *ns;
|
|
|
|
struct ipc_proc_iface *iface;
|
|
|
|
};
|
|
|
|
|
2007-10-19 10:40:48 +04:00
|
|
|
/*
|
|
|
|
* This routine locks the ipc structure found at least at position pos.
|
|
|
|
*/
|
2008-02-06 12:36:28 +03:00
|
|
|
static struct kern_ipc_perm *sysvipc_find_ipc(struct ipc_ids *ids, loff_t pos,
|
|
|
|
loff_t *new_pos)
|
2005-09-07 02:17:09 +04:00
|
|
|
{
|
2007-10-19 10:40:48 +04:00
|
|
|
struct kern_ipc_perm *ipc;
|
|
|
|
int total, id;
|
2006-10-02 13:18:20 +04:00
|
|
|
|
2007-10-19 10:40:48 +04:00
|
|
|
total = 0;
|
|
|
|
for (id = 0; id < pos && total < ids->in_use; id++) {
|
|
|
|
ipc = idr_find(&ids->ipcs_idr, id);
|
|
|
|
if (ipc != NULL)
|
|
|
|
total++;
|
|
|
|
}
|
2005-09-07 02:17:09 +04:00
|
|
|
|
2007-10-19 10:40:48 +04:00
|
|
|
if (total >= ids->in_use)
|
|
|
|
return NULL;
|
2005-09-07 02:17:09 +04:00
|
|
|
|
2014-01-28 05:07:04 +04:00
|
|
|
for (; pos < IPCMNI; pos++) {
|
2007-10-19 10:40:48 +04:00
|
|
|
ipc = idr_find(&ids->ipcs_idr, pos);
|
|
|
|
if (ipc != NULL) {
|
|
|
|
*new_pos = pos + 1;
|
2013-09-12 01:26:29 +04:00
|
|
|
rcu_read_lock();
|
|
|
|
ipc_lock_object(ipc);
|
2005-09-07 02:17:09 +04:00
|
|
|
return ipc;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
/* Out of range - return NULL to terminate iteration */
|
|
|
|
return NULL;
|
|
|
|
}
|
|
|
|
|
2007-10-19 10:40:48 +04:00
|
|
|
static void *sysvipc_proc_next(struct seq_file *s, void *it, loff_t *pos)
|
|
|
|
{
|
|
|
|
struct ipc_proc_iter *iter = s->private;
|
|
|
|
struct ipc_proc_iface *iface = iter->iface;
|
|
|
|
struct kern_ipc_perm *ipc = it;
|
|
|
|
|
|
|
|
/* If we had an ipc id locked before, unlock it */
|
|
|
|
if (ipc && ipc != SEQ_START_TOKEN)
|
|
|
|
ipc_unlock(ipc);
|
|
|
|
|
2008-02-08 15:18:57 +03:00
|
|
|
return sysvipc_find_ipc(&iter->ns->ids[iface->ids], *pos, pos);
|
2007-10-19 10:40:48 +04:00
|
|
|
}
|
|
|
|
|
2005-09-07 02:17:09 +04:00
|
|
|
/*
|
2007-10-19 10:40:53 +04:00
|
|
|
* File positions: pos 0 -> header, pos n -> ipc id = n - 1.
|
|
|
|
* SeqFile iterator: iterator value locked ipc pointer or SEQ_TOKEN_START.
|
2005-09-07 02:17:09 +04:00
|
|
|
*/
|
|
|
|
static void *sysvipc_proc_start(struct seq_file *s, loff_t *pos)
|
|
|
|
{
|
2007-02-12 11:52:10 +03:00
|
|
|
struct ipc_proc_iter *iter = s->private;
|
|
|
|
struct ipc_proc_iface *iface = iter->iface;
|
2006-10-02 13:18:20 +04:00
|
|
|
struct ipc_ids *ids;
|
|
|
|
|
2008-02-08 15:18:57 +03:00
|
|
|
ids = &iter->ns->ids[iface->ids];
|
2005-09-07 02:17:09 +04:00
|
|
|
|
|
|
|
/*
|
|
|
|
* Take the lock - this will be released by the corresponding
|
|
|
|
* call to stop().
|
|
|
|
*/
|
2013-09-12 01:26:24 +04:00
|
|
|
down_read(&ids->rwsem);
|
2005-09-07 02:17:09 +04:00
|
|
|
|
|
|
|
/* pos < 0 is invalid */
|
|
|
|
if (*pos < 0)
|
|
|
|
return NULL;
|
|
|
|
|
|
|
|
/* pos == 0 means header */
|
|
|
|
if (*pos == 0)
|
|
|
|
return SEQ_START_TOKEN;
|
|
|
|
|
|
|
|
/* Find the (pos-1)th ipc */
|
2007-10-19 10:40:48 +04:00
|
|
|
return sysvipc_find_ipc(ids, *pos - 1, pos);
|
2005-09-07 02:17:09 +04:00
|
|
|
}
|
|
|
|
|
|
|
|
static void sysvipc_proc_stop(struct seq_file *s, void *it)
|
|
|
|
{
|
|
|
|
struct kern_ipc_perm *ipc = it;
|
2007-02-12 11:52:10 +03:00
|
|
|
struct ipc_proc_iter *iter = s->private;
|
|
|
|
struct ipc_proc_iface *iface = iter->iface;
|
2006-10-02 13:18:20 +04:00
|
|
|
struct ipc_ids *ids;
|
2005-09-07 02:17:09 +04:00
|
|
|
|
2007-10-19 10:40:53 +04:00
|
|
|
/* If we had a locked structure, release it */
|
2005-09-07 02:17:09 +04:00
|
|
|
if (ipc && ipc != SEQ_START_TOKEN)
|
|
|
|
ipc_unlock(ipc);
|
|
|
|
|
2008-02-08 15:18:57 +03:00
|
|
|
ids = &iter->ns->ids[iface->ids];
|
2005-09-07 02:17:09 +04:00
|
|
|
/* Release the lock we took in start() */
|
2013-09-12 01:26:24 +04:00
|
|
|
up_read(&ids->rwsem);
|
2005-09-07 02:17:09 +04:00
|
|
|
}
|
|
|
|
|
|
|
|
static int sysvipc_proc_show(struct seq_file *s, void *it)
|
|
|
|
{
|
2007-02-12 11:52:10 +03:00
|
|
|
struct ipc_proc_iter *iter = s->private;
|
|
|
|
struct ipc_proc_iface *iface = iter->iface;
|
2005-09-07 02:17:09 +04:00
|
|
|
|
2015-04-16 02:17:54 +03:00
|
|
|
if (it == SEQ_START_TOKEN) {
|
|
|
|
seq_puts(s, iface->header);
|
|
|
|
return 0;
|
|
|
|
}
|
2005-09-07 02:17:09 +04:00
|
|
|
|
|
|
|
return iface->show(s, it);
|
|
|
|
}
|
|
|
|
|
2009-09-23 03:43:43 +04:00
|
|
|
static const struct seq_operations sysvipc_proc_seqops = {
|
2005-09-07 02:17:09 +04:00
|
|
|
.start = sysvipc_proc_start,
|
|
|
|
.stop = sysvipc_proc_stop,
|
|
|
|
.next = sysvipc_proc_next,
|
|
|
|
.show = sysvipc_proc_show,
|
|
|
|
};
|
|
|
|
|
2007-02-12 11:52:10 +03:00
|
|
|
static int sysvipc_proc_open(struct inode *inode, struct file *file)
|
|
|
|
{
|
|
|
|
struct ipc_proc_iter *iter;
|
|
|
|
|
2014-10-14 02:54:14 +04:00
|
|
|
iter = __seq_open_private(file, &sysvipc_proc_seqops, sizeof(*iter));
|
2007-02-12 11:52:10 +03:00
|
|
|
if (!iter)
|
2014-10-14 02:54:14 +04:00
|
|
|
return -ENOMEM;
|
2007-02-12 11:52:10 +03:00
|
|
|
|
2013-04-01 02:16:14 +04:00
|
|
|
iter->iface = PDE_DATA(inode);
|
2007-02-12 11:52:10 +03:00
|
|
|
iter->ns = get_ipc_ns(current->nsproxy->ipc_ns);
|
2014-10-14 02:54:14 +04:00
|
|
|
|
|
|
|
return 0;
|
2007-02-12 11:52:10 +03:00
|
|
|
}
|
|
|
|
|
|
|
|
static int sysvipc_proc_release(struct inode *inode, struct file *file)
|
|
|
|
{
|
|
|
|
struct seq_file *seq = file->private_data;
|
|
|
|
struct ipc_proc_iter *iter = seq->private;
|
|
|
|
put_ipc_ns(iter->ns);
|
|
|
|
return seq_release_private(inode, file);
|
2005-09-07 02:17:09 +04:00
|
|
|
}
|
|
|
|
|
2007-02-12 11:55:35 +03:00
|
|
|
static const struct file_operations sysvipc_proc_fops = {
|
2005-09-07 02:17:09 +04:00
|
|
|
.open = sysvipc_proc_open,
|
|
|
|
.read = seq_read,
|
|
|
|
.llseek = seq_lseek,
|
2007-02-12 11:52:10 +03:00
|
|
|
.release = sysvipc_proc_release,
|
2005-09-07 02:17:09 +04:00
|
|
|
};
|
|
|
|
#endif /* CONFIG_PROC_FS */
|