2005-04-17 06:20:36 +08:00
|
|
|
/*
|
|
|
|
* linux/ipc/sem.c
|
|
|
|
* Copyright (C) 1992 Krishna Balasubramanian
|
|
|
|
* Copyright (C) 1995 Eric Schenk, Bruno Haible
|
|
|
|
*
|
|
|
|
* /proc/sysvipc/sem support (c) 1999 Dragos Acostachioaie <dragos@iname.com>
|
|
|
|
*
|
|
|
|
* SMP-threaded, sysctl's added
|
2006-01-15 09:43:54 +08:00
|
|
|
* (c) 1999 Manfred Spraul <manfred@colorfullife.com>
|
2005-04-17 06:20:36 +08:00
|
|
|
* Enforced range limit on SEM_UNDO
|
2009-01-05 22:06:29 +08:00
|
|
|
* (c) 2001 Red Hat Inc
|
2005-04-17 06:20:36 +08:00
|
|
|
* Lockless wakeup
|
|
|
|
* (c) 2003 Manfred Spraul <manfred@colorfullife.com>
|
2010-05-27 05:43:43 +08:00
|
|
|
* Further wakeup optimizations, documentation
|
|
|
|
* (c) 2010 Manfred Spraul <manfred@colorfullife.com>
|
2006-04-03 05:07:33 +08:00
|
|
|
*
|
|
|
|
* support for audit of ipc object properties and permission changes
|
|
|
|
* Dustin Kirkland <dustin.kirkland@us.ibm.com>
|
2006-10-02 17:18:22 +08:00
|
|
|
*
|
|
|
|
* namespaces support
|
|
|
|
* OpenVZ, SWsoft Inc.
|
|
|
|
* Pavel Emelianov <xemul@openvz.org>
|
2010-05-27 05:43:43 +08:00
|
|
|
*
|
|
|
|
* Implementation notes: (May 2010)
|
|
|
|
* This file implements System V semaphores.
|
|
|
|
*
|
|
|
|
* User space visible behavior:
|
|
|
|
* - FIFO ordering for semop() operations (just FIFO, not starvation
|
|
|
|
* protection)
|
|
|
|
* - multiple semaphore operations that alter the same semaphore in
|
|
|
|
* one semop() are handled.
|
|
|
|
* - sem_ctime (time of last semctl()) is updated in the IPC_SET, SETVAL and
|
|
|
|
* SETALL calls.
|
|
|
|
* - two Linux specific semctl() commands: SEM_STAT, SEM_INFO.
|
|
|
|
* - undo adjustments at process exit are limited to 0..SEMVMX.
|
|
|
|
* - namespace are supported.
|
|
|
|
* - SEMMSL, SEMMNS, SEMOPM and SEMMNI can be configured at runtine by writing
|
|
|
|
* to /proc/sys/kernel/sem.
|
|
|
|
* - statistics about the usage are reported in /proc/sysvipc/sem.
|
|
|
|
*
|
|
|
|
* Internals:
|
|
|
|
* - scalability:
|
|
|
|
* - all global variables are read-mostly.
|
|
|
|
* - semop() calls and semctl(RMID) are synchronized by RCU.
|
|
|
|
* - most operations do write operations (actually: spin_lock calls) to
|
|
|
|
* the per-semaphore array structure.
|
|
|
|
* Thus: Perfect SMP scaling between independent semaphore arrays.
|
|
|
|
* If multiple semaphores in one array are used, then cache line
|
|
|
|
* trashing on the semaphore array spinlock will limit the scaling.
|
|
|
|
* - semncnt and semzcnt are calculated on demand in count_semncnt() and
|
|
|
|
* count_semzcnt()
|
|
|
|
* - the task that performs a successful semop() scans the list of all
|
|
|
|
* sleeping tasks and completes any pending operations that can be fulfilled.
|
|
|
|
* Semaphores are actively given to waiting tasks (necessary for FIFO).
|
|
|
|
* (see update_queue())
|
|
|
|
* - To improve the scalability, the actual wake-up calls are performed after
|
|
|
|
* dropping all locks. (see wake_up_sem_queue_prepare(),
|
|
|
|
* wake_up_sem_queue_do())
|
|
|
|
* - All work is done by the waker, the woken up task does not have to do
|
|
|
|
* anything - not even acquiring a lock or dropping a refcount.
|
|
|
|
* - A woken up task may not even touch the semaphore array anymore, it may
|
|
|
|
* have been destroyed already by a semctl(RMID).
|
|
|
|
* - The synchronizations between wake-ups due to a timeout/signal and a
|
|
|
|
* wake-up due to a completed semaphore operation is achieved by using an
|
|
|
|
* intermediate state (IN_WAKEUP).
|
|
|
|
* - UNDO values are stored in an array (one per process and per
|
|
|
|
* semaphore array, lazily allocated). For backwards compatibility, multiple
|
|
|
|
* modes for the UNDO variables are supported (per process, per thread)
|
|
|
|
* (see copy_semundo, CLONE_SYSVSEM)
|
|
|
|
* - There are two lists of the pending operations: a per-array list
|
|
|
|
* and per-semaphore list (stored in the array). This allows to achieve FIFO
|
|
|
|
* ordering without always scanning all pending operations.
|
|
|
|
* The worst-case behavior is nevertheless O(N^2) for N wakeups.
|
2005-04-17 06:20:36 +08:00
|
|
|
*/
|
|
|
|
|
|
|
|
#include <linux/slab.h>
|
|
|
|
#include <linux/spinlock.h>
|
|
|
|
#include <linux/init.h>
|
|
|
|
#include <linux/proc_fs.h>
|
|
|
|
#include <linux/time.h>
|
|
|
|
#include <linux/security.h>
|
|
|
|
#include <linux/syscalls.h>
|
|
|
|
#include <linux/audit.h>
|
2006-01-12 04:17:46 +08:00
|
|
|
#include <linux/capability.h>
|
2005-09-07 06:17:10 +08:00
|
|
|
#include <linux/seq_file.h>
|
2007-10-19 14:40:54 +08:00
|
|
|
#include <linux/rwsem.h>
|
2006-10-02 17:18:22 +08:00
|
|
|
#include <linux/nsproxy.h>
|
2008-02-08 20:18:22 +08:00
|
|
|
#include <linux/ipc_namespace.h>
|
2006-03-26 17:37:17 +08:00
|
|
|
|
2005-04-17 06:20:36 +08:00
|
|
|
#include <asm/uaccess.h>
|
|
|
|
#include "util.h"
|
|
|
|
|
2011-11-03 04:38:54 +08:00
|
|
|
/* One semaphore structure for each semaphore in the system. */
|
|
|
|
struct sem {
|
|
|
|
int semval; /* current value */
|
|
|
|
int sempid; /* pid of last operation */
|
2013-05-01 10:15:44 +08:00
|
|
|
spinlock_t lock; /* spinlock for fine-grained semtimedop */
|
2011-11-03 04:38:54 +08:00
|
|
|
struct list_head sem_pending; /* pending single-sop operations */
|
|
|
|
};
|
|
|
|
|
|
|
|
/* One queue for each sleeping process in the system. */
|
|
|
|
struct sem_queue {
|
|
|
|
struct list_head list; /* queue of pending operations */
|
|
|
|
struct task_struct *sleeper; /* this process */
|
|
|
|
struct sem_undo *undo; /* undo structure */
|
|
|
|
int pid; /* process id of requesting process */
|
|
|
|
int status; /* completion status of operation */
|
|
|
|
struct sembuf *sops; /* array of pending operations */
|
|
|
|
int nsops; /* number of operations */
|
|
|
|
int alter; /* does *sops alter the array? */
|
|
|
|
};
|
|
|
|
|
|
|
|
/* Each task has a list of undo requests. They are executed automatically
|
|
|
|
* when the process exits.
|
|
|
|
*/
|
|
|
|
struct sem_undo {
|
|
|
|
struct list_head list_proc; /* per-process list: *
|
|
|
|
* all undos from one process
|
|
|
|
* rcu protected */
|
|
|
|
struct rcu_head rcu; /* rcu struct for sem_undo */
|
|
|
|
struct sem_undo_list *ulp; /* back ptr to sem_undo_list */
|
|
|
|
struct list_head list_id; /* per semaphore array list:
|
|
|
|
* all undos for one array */
|
|
|
|
int semid; /* semaphore set identifier */
|
|
|
|
short *semadj; /* array of adjustments */
|
|
|
|
/* one per semaphore */
|
|
|
|
};
|
|
|
|
|
|
|
|
/* sem_undo_list controls shared access to the list of sem_undo structures
|
|
|
|
* that may be shared among all a CLONE_SYSVSEM task group.
|
|
|
|
*/
|
|
|
|
struct sem_undo_list {
|
|
|
|
atomic_t refcnt;
|
|
|
|
spinlock_t lock;
|
|
|
|
struct list_head list_proc;
|
|
|
|
};
|
|
|
|
|
|
|
|
|
2008-02-08 20:18:57 +08:00
|
|
|
#define sem_ids(ns) ((ns)->ids[IPC_SEM_IDS])
|
2006-10-02 17:18:22 +08:00
|
|
|
|
2007-10-19 14:40:55 +08:00
|
|
|
#define sem_checkid(sma, semid) ipc_checkid(&sma->sem_perm, semid)
|
2005-04-17 06:20:36 +08:00
|
|
|
|
2007-10-19 14:40:49 +08:00
|
|
|
static int newary(struct ipc_namespace *, struct ipc_params *);
|
2008-02-08 20:18:57 +08:00
|
|
|
static void freeary(struct ipc_namespace *, struct kern_ipc_perm *);
|
2005-04-17 06:20:36 +08:00
|
|
|
#ifdef CONFIG_PROC_FS
|
2005-09-07 06:17:10 +08:00
|
|
|
static int sysvipc_sem_proc_show(struct seq_file *s, void *it);
|
2005-04-17 06:20:36 +08:00
|
|
|
#endif
|
|
|
|
|
|
|
|
#define SEMMSL_FAST 256 /* 512 bytes on stack */
|
|
|
|
#define SEMOPM_FAST 64 /* ~ 372 bytes on stack */
|
|
|
|
|
|
|
|
/*
|
|
|
|
* linked list protection:
|
|
|
|
* sem_undo.id_next,
|
|
|
|
* sem_array.sem_pending{,last},
|
|
|
|
* sem_array.sem_undo: sem_lock() for read/write
|
|
|
|
* sem_undo.proc_next: only "current" is allowed to read/write that field.
|
|
|
|
*
|
|
|
|
*/
|
|
|
|
|
2006-10-02 17:18:22 +08:00
|
|
|
#define sc_semmsl sem_ctls[0]
|
|
|
|
#define sc_semmns sem_ctls[1]
|
|
|
|
#define sc_semopm sem_ctls[2]
|
|
|
|
#define sc_semmni sem_ctls[3]
|
|
|
|
|
2008-02-08 20:18:57 +08:00
|
|
|
void sem_init_ns(struct ipc_namespace *ns)
|
2006-10-02 17:18:22 +08:00
|
|
|
{
|
|
|
|
ns->sc_semmsl = SEMMSL;
|
|
|
|
ns->sc_semmns = SEMMNS;
|
|
|
|
ns->sc_semopm = SEMOPM;
|
|
|
|
ns->sc_semmni = SEMMNI;
|
|
|
|
ns->used_sems = 0;
|
2008-02-08 20:18:57 +08:00
|
|
|
ipc_init_ids(&ns->ids[IPC_SEM_IDS]);
|
2006-10-02 17:18:22 +08:00
|
|
|
}
|
|
|
|
|
2008-02-08 20:18:22 +08:00
|
|
|
#ifdef CONFIG_IPC_NS
|
2006-10-02 17:18:22 +08:00
|
|
|
void sem_exit_ns(struct ipc_namespace *ns)
|
|
|
|
{
|
2008-02-08 20:18:57 +08:00
|
|
|
free_ipcs(ns, &sem_ids(ns), freeary);
|
2009-12-16 08:47:27 +08:00
|
|
|
idr_destroy(&ns->ids[IPC_SEM_IDS].ipcs_idr);
|
2006-10-02 17:18:22 +08:00
|
|
|
}
|
2008-02-08 20:18:22 +08:00
|
|
|
#endif
|
2005-04-17 06:20:36 +08:00
|
|
|
|
|
|
|
void __init sem_init (void)
|
|
|
|
{
|
2008-02-08 20:18:57 +08:00
|
|
|
sem_init_ns(&init_ipc_ns);
|
2005-09-07 06:17:10 +08:00
|
|
|
ipc_init_proc_interface("sysvipc/sem",
|
|
|
|
" key semid perms nsems uid gid cuid cgid otime ctime\n",
|
2006-10-02 17:18:22 +08:00
|
|
|
IPC_SEM_IDS, sysvipc_sem_proc_show);
|
2005-04-17 06:20:36 +08:00
|
|
|
}
|
|
|
|
|
2013-05-01 10:15:44 +08:00
|
|
|
/*
|
|
|
|
* If the request contains only one semaphore operation, and there are
|
|
|
|
* no complex transactions pending, lock only the semaphore involved.
|
|
|
|
* Otherwise, lock the entire semaphore array, since we either have
|
|
|
|
* multiple semaphores in our own semops, or we need to look at
|
|
|
|
* semaphores from other pending complex operations.
|
|
|
|
*
|
|
|
|
* Carefully guard against sma->complex_count changing between zero
|
|
|
|
* and non-zero while we are spinning for the lock. The value of
|
|
|
|
* sma->complex_count cannot change while we are holding the lock,
|
|
|
|
* so sem_unlock should be fine.
|
|
|
|
*
|
|
|
|
* The global lock path checks that all the local locks have been released,
|
|
|
|
* checking each local lock once. This means that the local lock paths
|
|
|
|
* cannot start their critical sections while the global lock is held.
|
|
|
|
*/
|
|
|
|
static inline int sem_lock(struct sem_array *sma, struct sembuf *sops,
|
|
|
|
int nsops)
|
|
|
|
{
|
|
|
|
int locknum;
|
|
|
|
again:
|
|
|
|
if (nsops == 1 && !sma->complex_count) {
|
|
|
|
struct sem *sem = sma->sem_base + sops->sem_num;
|
|
|
|
|
|
|
|
/* Lock just the semaphore we are interested in. */
|
|
|
|
spin_lock(&sem->lock);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* If sma->complex_count was set while we were spinning,
|
|
|
|
* we may need to look at things we did not lock here.
|
|
|
|
*/
|
|
|
|
if (unlikely(sma->complex_count)) {
|
|
|
|
spin_unlock(&sem->lock);
|
|
|
|
goto lock_array;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Another process is holding the global lock on the
|
|
|
|
* sem_array; we cannot enter our critical section,
|
|
|
|
* but have to wait for the global lock to be released.
|
|
|
|
*/
|
|
|
|
if (unlikely(spin_is_locked(&sma->sem_perm.lock))) {
|
|
|
|
spin_unlock(&sem->lock);
|
|
|
|
spin_unlock_wait(&sma->sem_perm.lock);
|
|
|
|
goto again;
|
|
|
|
}
|
|
|
|
|
|
|
|
locknum = sops->sem_num;
|
|
|
|
} else {
|
|
|
|
int i;
|
|
|
|
/*
|
|
|
|
* Lock the semaphore array, and wait for all of the
|
|
|
|
* individual semaphore locks to go away. The code
|
|
|
|
* above ensures no new single-lock holders will enter
|
|
|
|
* their critical section while the array lock is held.
|
|
|
|
*/
|
|
|
|
lock_array:
|
|
|
|
spin_lock(&sma->sem_perm.lock);
|
|
|
|
for (i = 0; i < sma->sem_nsems; i++) {
|
|
|
|
struct sem *sem = sma->sem_base + i;
|
|
|
|
spin_unlock_wait(&sem->lock);
|
|
|
|
}
|
|
|
|
locknum = -1;
|
|
|
|
}
|
|
|
|
return locknum;
|
|
|
|
}
|
|
|
|
|
|
|
|
static inline void sem_unlock(struct sem_array *sma, int locknum)
|
|
|
|
{
|
|
|
|
if (locknum == -1) {
|
|
|
|
spin_unlock(&sma->sem_perm.lock);
|
|
|
|
} else {
|
|
|
|
struct sem *sem = sma->sem_base + locknum;
|
|
|
|
spin_unlock(&sem->lock);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2007-10-19 14:40:54 +08:00
|
|
|
/*
|
|
|
|
* sem_lock_(check_) routines are called in the paths where the rw_mutex
|
|
|
|
* is not held.
|
2013-05-05 01:47:57 +08:00
|
|
|
*
|
|
|
|
* The caller holds the RCU read lock.
|
2007-10-19 14:40:54 +08:00
|
|
|
*/
|
2013-05-01 10:15:44 +08:00
|
|
|
static inline struct sem_array *sem_obtain_lock(struct ipc_namespace *ns,
|
|
|
|
int id, struct sembuf *sops, int nsops, int *locknum)
|
2007-10-19 14:40:51 +08:00
|
|
|
{
|
2013-05-01 10:15:35 +08:00
|
|
|
struct kern_ipc_perm *ipcp;
|
|
|
|
struct sem_array *sma;
|
2007-10-19 14:40:51 +08:00
|
|
|
|
2013-05-01 10:15:35 +08:00
|
|
|
ipcp = ipc_obtain_object(&sem_ids(ns), id);
|
2013-05-05 01:47:57 +08:00
|
|
|
if (IS_ERR(ipcp))
|
|
|
|
return ERR_CAST(ipcp);
|
IPC: fix error check in all new xxx_lock() and xxx_exit_ns() functions
In the new implementation of the [sem|shm|msg]_lock[_check]() routines, we
use the return value of ipc_lock() in container_of() without any check.
But ipc_lock may return a errcode. The use of this errcode in
container_of() may alter this errcode, and we don't want this.
And in xxx_exit_ns, the pointer return by idr_find is of type 'struct
kern_ipc_per'...
Today, the code will work as is because the member used in these
container_of() is the first member of its container (offset == 0), the
errcode isn't changed then. But in the general case, we can't count on
this assumption and this may lead later to a real bug if we don't correct
this.
Again, the proposed solution is simple and correct. But, as pointed by
Nadia, with this solution, the same check will be done several times (in
all sub-callers...), what is not very funny/optimal...
Signed-off-by: Pierre Peiffer <pierre.peiffer@bull.net>
Cc: Nadia Derbey <Nadia.Derbey@bull.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-02-06 17:36:23 +08:00
|
|
|
|
2013-05-01 10:15:44 +08:00
|
|
|
sma = container_of(ipcp, struct sem_array, sem_perm);
|
|
|
|
*locknum = sem_lock(sma, sops, nsops);
|
2013-05-01 10:15:35 +08:00
|
|
|
|
|
|
|
/* ipc_rmid() may have already freed the ID while sem_lock
|
|
|
|
* was spinning: verify that the structure is still valid
|
|
|
|
*/
|
|
|
|
if (!ipcp->deleted)
|
|
|
|
return container_of(ipcp, struct sem_array, sem_perm);
|
|
|
|
|
2013-05-01 10:15:44 +08:00
|
|
|
sem_unlock(sma, *locknum);
|
2013-05-05 01:47:57 +08:00
|
|
|
return ERR_PTR(-EINVAL);
|
2007-10-19 14:40:51 +08:00
|
|
|
}
|
|
|
|
|
ipc,sem: do not hold ipc lock more than necessary
Instead of holding the ipc lock for permissions and security checks, among
others, only acquire it when necessary.
Some numbers....
1) With Rik's semop-multi.c microbenchmark we can see the following
results:
Baseline (3.9-rc1):
cpus 4, threads: 256, semaphores: 128, test duration: 30 secs
total operations: 151452270, ops/sec 5048409
+ 59.40% a.out [kernel.kallsyms] [k] _raw_spin_lock
+ 6.14% a.out [kernel.kallsyms] [k] sys_semtimedop
+ 3.84% a.out [kernel.kallsyms] [k] avc_has_perm_flags
+ 3.64% a.out [kernel.kallsyms] [k] __audit_syscall_exit
+ 2.06% a.out [kernel.kallsyms] [k] copy_user_enhanced_fast_string
+ 1.86% a.out [kernel.kallsyms] [k] ipc_lock
With this patchset:
cpus 4, threads: 256, semaphores: 128, test duration: 30 secs
total operations: 273156400, ops/sec 9105213
+ 18.54% a.out [kernel.kallsyms] [k] _raw_spin_lock
+ 11.72% a.out [kernel.kallsyms] [k] sys_semtimedop
+ 7.70% a.out [kernel.kallsyms] [k] ipc_has_perm.isra.21
+ 6.58% a.out [kernel.kallsyms] [k] avc_has_perm_flags
+ 6.54% a.out [kernel.kallsyms] [k] __audit_syscall_exit
+ 4.71% a.out [kernel.kallsyms] [k] ipc_obtain_object_check
2) While on an Oracle swingbench DSS (data mining) workload the
improvements are not as exciting as with Rik's benchmark, we can see
some positive numbers. For an 8 socket machine the following are the
percentages of %sys time incurred in the ipc lock:
Baseline (3.9-rc1):
100 swingbench users: 8,74%
400 swingbench users: 21,86%
800 swingbench users: 84,35%
With this patchset:
100 swingbench users: 8,11%
400 swingbench users: 19,93%
800 swingbench users: 77,69%
[riel@redhat.com: fix two locking bugs]
[sasha.levin@oracle.com: prevent releasing RCU read lock twice in semctl_main]
[akpm@linux-foundation.org: coding-style fixes]
Signed-off-by: Davidlohr Bueso <davidlohr.bueso@hp.com>
Signed-off-by: Rik van Riel <riel@redhat.com>
Reviewed-by: Chegu Vinod <chegu_vinod@hp.com>
Acked-by: Michel Lespinasse <walken@google.com>
Cc: Rik van Riel <riel@redhat.com>
Cc: Jason Low <jason.low2@hp.com>
Cc: Emmanuel Benisty <benisty.e@gmail.com>
Cc: Peter Hurley <peter@hurleysoftware.com>
Cc: Stanislav Kinsbursky <skinsbursky@parallels.com>
Tested-by: Sedat Dilek <sedat.dilek@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2013-05-01 10:15:29 +08:00
|
|
|
static inline struct sem_array *sem_obtain_object(struct ipc_namespace *ns, int id)
|
|
|
|
{
|
|
|
|
struct kern_ipc_perm *ipcp = ipc_obtain_object(&sem_ids(ns), id);
|
|
|
|
|
|
|
|
if (IS_ERR(ipcp))
|
|
|
|
return ERR_CAST(ipcp);
|
|
|
|
|
|
|
|
return container_of(ipcp, struct sem_array, sem_perm);
|
|
|
|
}
|
|
|
|
|
|
|
|
static inline struct sem_array *sem_obtain_object_check(struct ipc_namespace *ns,
|
|
|
|
int id)
|
|
|
|
{
|
|
|
|
struct kern_ipc_perm *ipcp = ipc_obtain_object_check(&sem_ids(ns), id);
|
|
|
|
|
|
|
|
if (IS_ERR(ipcp))
|
|
|
|
return ERR_CAST(ipcp);
|
IPC: fix error check in all new xxx_lock() and xxx_exit_ns() functions
In the new implementation of the [sem|shm|msg]_lock[_check]() routines, we
use the return value of ipc_lock() in container_of() without any check.
But ipc_lock may return a errcode. The use of this errcode in
container_of() may alter this errcode, and we don't want this.
And in xxx_exit_ns, the pointer return by idr_find is of type 'struct
kern_ipc_per'...
Today, the code will work as is because the member used in these
container_of() is the first member of its container (offset == 0), the
errcode isn't changed then. But in the general case, we can't count on
this assumption and this may lead later to a real bug if we don't correct
this.
Again, the proposed solution is simple and correct. But, as pointed by
Nadia, with this solution, the same check will be done several times (in
all sub-callers...), what is not very funny/optimal...
Signed-off-by: Pierre Peiffer <pierre.peiffer@bull.net>
Cc: Nadia Derbey <Nadia.Derbey@bull.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-02-06 17:36:23 +08:00
|
|
|
|
2007-10-19 14:40:51 +08:00
|
|
|
return container_of(ipcp, struct sem_array, sem_perm);
|
2007-10-19 14:40:51 +08:00
|
|
|
}
|
|
|
|
|
2008-04-29 16:00:46 +08:00
|
|
|
static inline void sem_lock_and_putref(struct sem_array *sma)
|
|
|
|
{
|
2013-05-01 10:15:44 +08:00
|
|
|
sem_lock(sma, NULL, -1);
|
2008-04-29 16:00:46 +08:00
|
|
|
ipc_rcu_putref(sma);
|
|
|
|
}
|
|
|
|
|
|
|
|
static inline void sem_putref(struct sem_array *sma)
|
|
|
|
{
|
2013-05-04 06:22:00 +08:00
|
|
|
ipc_rcu_putref(sma);
|
2008-04-29 16:00:46 +08:00
|
|
|
}
|
|
|
|
|
2007-10-19 14:40:48 +08:00
|
|
|
static inline void sem_rmid(struct ipc_namespace *ns, struct sem_array *s)
|
|
|
|
{
|
|
|
|
ipc_rmid(&sem_ids(ns), &s->sem_perm);
|
|
|
|
}
|
|
|
|
|
2005-04-17 06:20:36 +08:00
|
|
|
/*
|
|
|
|
* Lockless wakeup algorithm:
|
|
|
|
* Without the check/retry algorithm a lockless wakeup is possible:
|
|
|
|
* - queue.status is initialized to -EINTR before blocking.
|
|
|
|
* - wakeup is performed by
|
|
|
|
* * unlinking the queue entry from sma->sem_pending
|
|
|
|
* * setting queue.status to IN_WAKEUP
|
|
|
|
* This is the notification for the blocked thread that a
|
|
|
|
* result value is imminent.
|
|
|
|
* * call wake_up_process
|
|
|
|
* * set queue.status to the final value.
|
|
|
|
* - the previously blocked thread checks queue.status:
|
|
|
|
* * if it's IN_WAKEUP, then it must wait until the value changes
|
|
|
|
* * if it's not -EINTR, then the operation was completed by
|
|
|
|
* update_queue. semtimedop can return queue.status without
|
2006-03-26 17:37:17 +08:00
|
|
|
* performing any operation on the sem array.
|
2005-04-17 06:20:36 +08:00
|
|
|
* * otherwise it must acquire the spinlock and check what's up.
|
|
|
|
*
|
|
|
|
* The two-stage algorithm is necessary to protect against the following
|
|
|
|
* races:
|
|
|
|
* - if queue.status is set after wake_up_process, then the woken up idle
|
|
|
|
* thread could race forward and try (and fail) to acquire sma->lock
|
|
|
|
* before update_queue had a chance to set queue.status
|
|
|
|
* - if queue.status is written before wake_up_process and if the
|
|
|
|
* blocked process is woken up by a signal between writing
|
|
|
|
* queue.status and the wake_up_process, then the woken up
|
|
|
|
* process could return from semtimedop and die by calling
|
|
|
|
* sys_exit before wake_up_process is called. Then wake_up_process
|
|
|
|
* will oops, because the task structure is already invalid.
|
|
|
|
* (yes, this happened on s390 with sysv msg).
|
|
|
|
*
|
|
|
|
*/
|
|
|
|
#define IN_WAKEUP 1
|
|
|
|
|
2007-10-19 14:40:53 +08:00
|
|
|
/**
|
|
|
|
* newary - Create a new semaphore set
|
|
|
|
* @ns: namespace
|
|
|
|
* @params: ptr to the structure that contains key, semflg and nsems
|
|
|
|
*
|
2007-10-19 14:40:54 +08:00
|
|
|
* Called with sem_ids.rw_mutex held (as a writer)
|
2007-10-19 14:40:53 +08:00
|
|
|
*/
|
|
|
|
|
2007-10-19 14:40:49 +08:00
|
|
|
static int newary(struct ipc_namespace *ns, struct ipc_params *params)
|
2005-04-17 06:20:36 +08:00
|
|
|
{
|
|
|
|
int id;
|
|
|
|
int retval;
|
|
|
|
struct sem_array *sma;
|
|
|
|
int size;
|
2007-10-19 14:40:49 +08:00
|
|
|
key_t key = params->key;
|
|
|
|
int nsems = params->u.nsems;
|
|
|
|
int semflg = params->flg;
|
2009-12-16 08:47:32 +08:00
|
|
|
int i;
|
2005-04-17 06:20:36 +08:00
|
|
|
|
|
|
|
if (!nsems)
|
|
|
|
return -EINVAL;
|
2006-10-02 17:18:22 +08:00
|
|
|
if (ns->used_sems + nsems > ns->sc_semmns)
|
2005-04-17 06:20:36 +08:00
|
|
|
return -ENOSPC;
|
|
|
|
|
|
|
|
size = sizeof (*sma) + nsems * sizeof (struct sem);
|
|
|
|
sma = ipc_rcu_alloc(size);
|
|
|
|
if (!sma) {
|
|
|
|
return -ENOMEM;
|
|
|
|
}
|
|
|
|
memset (sma, 0, size);
|
|
|
|
|
|
|
|
sma->sem_perm.mode = (semflg & S_IRWXUGO);
|
|
|
|
sma->sem_perm.key = key;
|
|
|
|
|
|
|
|
sma->sem_perm.security = NULL;
|
|
|
|
retval = security_sem_alloc(sma);
|
|
|
|
if (retval) {
|
|
|
|
ipc_rcu_putref(sma);
|
|
|
|
return retval;
|
|
|
|
}
|
|
|
|
|
2006-10-02 17:18:22 +08:00
|
|
|
id = ipc_addid(&sem_ids(ns), &sma->sem_perm, ns->sc_semmni);
|
2007-10-19 14:40:57 +08:00
|
|
|
if (id < 0) {
|
2005-04-17 06:20:36 +08:00
|
|
|
security_sem_free(sma);
|
|
|
|
ipc_rcu_putref(sma);
|
2007-10-19 14:40:57 +08:00
|
|
|
return id;
|
2005-04-17 06:20:36 +08:00
|
|
|
}
|
2006-10-02 17:18:22 +08:00
|
|
|
ns->used_sems += nsems;
|
2005-04-17 06:20:36 +08:00
|
|
|
|
|
|
|
sma->sem_base = (struct sem *) &sma[1];
|
2009-12-16 08:47:32 +08:00
|
|
|
|
2013-05-01 10:15:44 +08:00
|
|
|
for (i = 0; i < nsems; i++) {
|
2009-12-16 08:47:32 +08:00
|
|
|
INIT_LIST_HEAD(&sma->sem_base[i].sem_pending);
|
2013-05-01 10:15:44 +08:00
|
|
|
spin_lock_init(&sma->sem_base[i].lock);
|
|
|
|
}
|
2009-12-16 08:47:32 +08:00
|
|
|
|
|
|
|
sma->complex_count = 0;
|
2008-07-25 16:48:06 +08:00
|
|
|
INIT_LIST_HEAD(&sma->sem_pending);
|
2008-07-25 16:48:04 +08:00
|
|
|
INIT_LIST_HEAD(&sma->list_id);
|
2005-04-17 06:20:36 +08:00
|
|
|
sma->sem_nsems = nsems;
|
|
|
|
sma->sem_ctime = get_seconds();
|
2013-05-01 10:15:44 +08:00
|
|
|
sem_unlock(sma, -1);
|
ipc: move rcu_read_unlock() out of sem_unlock() and into callers
The IPC locking is a mess, and sem_unlock() unlocks not only the
semaphore spinlock, it also drops the rcu read lock. Unlike sem_lock(),
which just gets the spin-lock, and expects the caller to get the rcu
read lock.
This all makes things very hard to follow, and it's very confusing when
you take the rcu read lock in one function, and then release it in
another. And it has caused actual bugs: the sem_obtain_lock() function
ended up dropping the RCU read lock twice in one error path, because it
first did the sem_unlock(), and then did a rcu_read_unlock() to match
the rcu_read_lock() it had done.
This is just a totally mindless "remove rcu_read_unlock() from
sem_unlock() and add it immediately after each caller" (except for the
aforementioned bug where we did too many rcu_read_unlock(), and in
find_alloc_undo() where we just got the rcu_read_lock() to correct for
the fact that sem_unlock would immediately drop it again).
We can (and should) clean things up further, but this fixes the bug with
the minimal amount of subtlety.
Reviewed-by: Davidlohr Bueso <davidlohr.bueso@hp.com>
Cc: Rik van Riel <riel@redhat.com>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2013-05-04 06:04:40 +08:00
|
|
|
rcu_read_unlock();
|
2005-04-17 06:20:36 +08:00
|
|
|
|
2007-10-19 14:40:48 +08:00
|
|
|
return sma->sem_perm.id;
|
2005-04-17 06:20:36 +08:00
|
|
|
}
|
|
|
|
|
2007-10-19 14:40:49 +08:00
|
|
|
|
2007-10-19 14:40:53 +08:00
|
|
|
/*
|
2007-10-19 14:40:54 +08:00
|
|
|
* Called with sem_ids.rw_mutex and ipcp locked.
|
2007-10-19 14:40:53 +08:00
|
|
|
*/
|
2007-10-19 14:40:51 +08:00
|
|
|
static inline int sem_security(struct kern_ipc_perm *ipcp, int semflg)
|
2007-10-19 14:40:49 +08:00
|
|
|
{
|
2007-10-19 14:40:51 +08:00
|
|
|
struct sem_array *sma;
|
|
|
|
|
|
|
|
sma = container_of(ipcp, struct sem_array, sem_perm);
|
|
|
|
return security_sem_associate(sma, semflg);
|
2007-10-19 14:40:49 +08:00
|
|
|
}
|
|
|
|
|
2007-10-19 14:40:53 +08:00
|
|
|
/*
|
2007-10-19 14:40:54 +08:00
|
|
|
* Called with sem_ids.rw_mutex and ipcp locked.
|
2007-10-19 14:40:53 +08:00
|
|
|
*/
|
2007-10-19 14:40:51 +08:00
|
|
|
static inline int sem_more_checks(struct kern_ipc_perm *ipcp,
|
|
|
|
struct ipc_params *params)
|
2007-10-19 14:40:49 +08:00
|
|
|
{
|
2007-10-19 14:40:51 +08:00
|
|
|
struct sem_array *sma;
|
|
|
|
|
|
|
|
sma = container_of(ipcp, struct sem_array, sem_perm);
|
|
|
|
if (params->u.nsems > sma->sem_nsems)
|
2007-10-19 14:40:49 +08:00
|
|
|
return -EINVAL;
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2009-01-14 21:14:27 +08:00
|
|
|
SYSCALL_DEFINE3(semget, key_t, key, int, nsems, int, semflg)
|
2005-04-17 06:20:36 +08:00
|
|
|
{
|
2006-10-02 17:18:22 +08:00
|
|
|
struct ipc_namespace *ns;
|
2007-10-19 14:40:49 +08:00
|
|
|
struct ipc_ops sem_ops;
|
|
|
|
struct ipc_params sem_params;
|
2006-10-02 17:18:22 +08:00
|
|
|
|
|
|
|
ns = current->nsproxy->ipc_ns;
|
2005-04-17 06:20:36 +08:00
|
|
|
|
2006-10-02 17:18:22 +08:00
|
|
|
if (nsems < 0 || nsems > ns->sc_semmsl)
|
2005-04-17 06:20:36 +08:00
|
|
|
return -EINVAL;
|
2007-10-19 14:40:48 +08:00
|
|
|
|
2007-10-19 14:40:49 +08:00
|
|
|
sem_ops.getnew = newary;
|
|
|
|
sem_ops.associate = sem_security;
|
|
|
|
sem_ops.more_checks = sem_more_checks;
|
|
|
|
|
|
|
|
sem_params.key = key;
|
|
|
|
sem_params.flg = semflg;
|
|
|
|
sem_params.u.nsems = nsems;
|
2005-04-17 06:20:36 +08:00
|
|
|
|
2007-10-19 14:40:49 +08:00
|
|
|
return ipcget(ns, &sem_ids(ns), &sem_ops, &sem_params);
|
2005-04-17 06:20:36 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Determine whether a sequence of semaphore operations would succeed
|
|
|
|
* all at once. Return 0 if yes, 1 if need to sleep, else return error code.
|
|
|
|
*/
|
|
|
|
|
|
|
|
static int try_atomic_semop (struct sem_array * sma, struct sembuf * sops,
|
|
|
|
int nsops, struct sem_undo *un, int pid)
|
|
|
|
{
|
|
|
|
int result, sem_op;
|
|
|
|
struct sembuf *sop;
|
|
|
|
struct sem * curr;
|
|
|
|
|
|
|
|
for (sop = sops; sop < sops + nsops; sop++) {
|
|
|
|
curr = sma->sem_base + sop->sem_num;
|
|
|
|
sem_op = sop->sem_op;
|
|
|
|
result = curr->semval;
|
|
|
|
|
|
|
|
if (!sem_op && result)
|
|
|
|
goto would_block;
|
|
|
|
|
|
|
|
result += sem_op;
|
|
|
|
if (result < 0)
|
|
|
|
goto would_block;
|
|
|
|
if (result > SEMVMX)
|
|
|
|
goto out_of_range;
|
|
|
|
if (sop->sem_flg & SEM_UNDO) {
|
|
|
|
int undo = un->semadj[sop->sem_num] - sem_op;
|
|
|
|
/*
|
|
|
|
* Exceeding the undo range is an error.
|
|
|
|
*/
|
|
|
|
if (undo < (-SEMAEM - 1) || undo > SEMAEM)
|
|
|
|
goto out_of_range;
|
|
|
|
}
|
|
|
|
curr->semval = result;
|
|
|
|
}
|
|
|
|
|
|
|
|
sop--;
|
|
|
|
while (sop >= sops) {
|
|
|
|
sma->sem_base[sop->sem_num].sempid = pid;
|
|
|
|
if (sop->sem_flg & SEM_UNDO)
|
|
|
|
un->semadj[sop->sem_num] -= sop->sem_op;
|
|
|
|
sop--;
|
|
|
|
}
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
|
|
|
|
out_of_range:
|
|
|
|
result = -ERANGE;
|
|
|
|
goto undo;
|
|
|
|
|
|
|
|
would_block:
|
|
|
|
if (sop->sem_flg & IPC_NOWAIT)
|
|
|
|
result = -EAGAIN;
|
|
|
|
else
|
|
|
|
result = 1;
|
|
|
|
|
|
|
|
undo:
|
|
|
|
sop--;
|
|
|
|
while (sop >= sops) {
|
|
|
|
sma->sem_base[sop->sem_num].semval -= sop->sem_op;
|
|
|
|
sop--;
|
|
|
|
}
|
|
|
|
|
|
|
|
return result;
|
|
|
|
}
|
|
|
|
|
2010-05-27 05:43:41 +08:00
|
|
|
/** wake_up_sem_queue_prepare(q, error): Prepare wake-up
|
|
|
|
* @q: queue entry that must be signaled
|
|
|
|
* @error: Error value for the signal
|
|
|
|
*
|
|
|
|
* Prepare the wake-up of the queue entry q.
|
2009-12-16 08:47:30 +08:00
|
|
|
*/
|
2010-05-27 05:43:41 +08:00
|
|
|
static void wake_up_sem_queue_prepare(struct list_head *pt,
|
|
|
|
struct sem_queue *q, int error)
|
2009-12-16 08:47:30 +08:00
|
|
|
{
|
2010-05-27 05:43:41 +08:00
|
|
|
if (list_empty(pt)) {
|
|
|
|
/*
|
|
|
|
* Hold preempt off so that we don't get preempted and have the
|
|
|
|
* wakee busy-wait until we're scheduled back on.
|
|
|
|
*/
|
|
|
|
preempt_disable();
|
|
|
|
}
|
2009-12-16 08:47:30 +08:00
|
|
|
q->status = IN_WAKEUP;
|
2010-05-27 05:43:41 +08:00
|
|
|
q->pid = error;
|
|
|
|
|
2013-05-01 10:15:39 +08:00
|
|
|
list_add_tail(&q->list, pt);
|
2010-05-27 05:43:41 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
/**
|
|
|
|
* wake_up_sem_queue_do(pt) - do the actual wake-up
|
|
|
|
* @pt: list of tasks to be woken up
|
|
|
|
*
|
|
|
|
* Do the actual wake-up.
|
|
|
|
* The function is called without any locks held, thus the semaphore array
|
|
|
|
* could be destroyed already and the tasks can disappear as soon as the
|
|
|
|
* status is set to the actual return code.
|
|
|
|
*/
|
|
|
|
static void wake_up_sem_queue_do(struct list_head *pt)
|
|
|
|
{
|
|
|
|
struct sem_queue *q, *t;
|
|
|
|
int did_something;
|
|
|
|
|
|
|
|
did_something = !list_empty(pt);
|
2013-05-01 10:15:39 +08:00
|
|
|
list_for_each_entry_safe(q, t, pt, list) {
|
2010-05-27 05:43:41 +08:00
|
|
|
wake_up_process(q->sleeper);
|
|
|
|
/* q can disappear immediately after writing q->status. */
|
|
|
|
smp_wmb();
|
|
|
|
q->status = q->pid;
|
|
|
|
}
|
|
|
|
if (did_something)
|
|
|
|
preempt_enable();
|
2009-12-16 08:47:30 +08:00
|
|
|
}
|
|
|
|
|
2009-12-16 08:47:32 +08:00
|
|
|
static void unlink_queue(struct sem_array *sma, struct sem_queue *q)
|
|
|
|
{
|
|
|
|
list_del(&q->list);
|
2013-05-01 10:15:39 +08:00
|
|
|
if (q->nsops > 1)
|
2009-12-16 08:47:32 +08:00
|
|
|
sma->complex_count--;
|
|
|
|
}
|
|
|
|
|
2010-05-27 05:43:40 +08:00
|
|
|
/** check_restart(sma, q)
|
|
|
|
* @sma: semaphore array
|
|
|
|
* @q: the operation that just completed
|
|
|
|
*
|
|
|
|
* update_queue is O(N^2) when it restarts scanning the whole queue of
|
|
|
|
* waiting operations. Therefore this function checks if the restart is
|
|
|
|
* really necessary. It is called after a previously waiting operation
|
|
|
|
* was completed.
|
|
|
|
*/
|
|
|
|
static int check_restart(struct sem_array *sma, struct sem_queue *q)
|
|
|
|
{
|
|
|
|
struct sem *curr;
|
|
|
|
struct sem_queue *h;
|
|
|
|
|
|
|
|
/* if the operation didn't modify the array, then no restart */
|
|
|
|
if (q->alter == 0)
|
|
|
|
return 0;
|
|
|
|
|
|
|
|
/* pending complex operations are too difficult to analyse */
|
|
|
|
if (sma->complex_count)
|
|
|
|
return 1;
|
|
|
|
|
|
|
|
/* we were a sleeping complex operation. Too difficult */
|
|
|
|
if (q->nsops > 1)
|
|
|
|
return 1;
|
|
|
|
|
|
|
|
curr = sma->sem_base + q->sops[0].sem_num;
|
|
|
|
|
|
|
|
/* No-one waits on this queue */
|
|
|
|
if (list_empty(&curr->sem_pending))
|
|
|
|
return 0;
|
|
|
|
|
|
|
|
/* the new semaphore value */
|
|
|
|
if (curr->semval) {
|
|
|
|
/* It is impossible that someone waits for the new value:
|
|
|
|
* - q is a previously sleeping simple operation that
|
|
|
|
* altered the array. It must be a decrement, because
|
|
|
|
* simple increments never sleep.
|
|
|
|
* - The value is not 0, thus wait-for-zero won't proceed.
|
|
|
|
* - If there are older (higher priority) decrements
|
|
|
|
* in the queue, then they have observed the original
|
|
|
|
* semval value and couldn't proceed. The operation
|
|
|
|
* decremented to value - thus they won't proceed either.
|
|
|
|
*/
|
|
|
|
BUG_ON(q->sops[0].sem_op >= 0);
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
/*
|
|
|
|
* semval is 0. Check if there are wait-for-zero semops.
|
2013-05-01 10:15:39 +08:00
|
|
|
* They must be the first entries in the per-semaphore queue
|
2010-05-27 05:43:40 +08:00
|
|
|
*/
|
2013-05-01 10:15:39 +08:00
|
|
|
h = list_first_entry(&curr->sem_pending, struct sem_queue, list);
|
2010-05-27 05:43:40 +08:00
|
|
|
BUG_ON(h->nsops != 1);
|
|
|
|
BUG_ON(h->sops[0].sem_num != q->sops[0].sem_num);
|
|
|
|
|
|
|
|
/* Yes, there is a wait-for-zero semop. Restart */
|
|
|
|
if (h->sops[0].sem_op == 0)
|
|
|
|
return 1;
|
|
|
|
|
|
|
|
/* Again - no-one is waiting for the new value. */
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2009-12-16 08:47:33 +08:00
|
|
|
|
|
|
|
/**
|
|
|
|
* update_queue(sma, semnum): Look for tasks that can be completed.
|
|
|
|
* @sma: semaphore array.
|
|
|
|
* @semnum: semaphore that was modified.
|
2010-05-27 05:43:41 +08:00
|
|
|
* @pt: list head for the tasks that must be woken up.
|
2009-12-16 08:47:33 +08:00
|
|
|
*
|
|
|
|
* update_queue must be called after a semaphore in a semaphore array
|
2013-05-01 10:15:39 +08:00
|
|
|
* was modified. If multiple semaphores were modified, update_queue must
|
|
|
|
* be called with semnum = -1, as well as with the number of each modified
|
|
|
|
* semaphore.
|
2010-05-27 05:43:41 +08:00
|
|
|
* The tasks that must be woken up are added to @pt. The return code
|
|
|
|
* is stored in q->pid.
|
|
|
|
* The function return 1 if at least one semop was completed successfully.
|
2005-04-17 06:20:36 +08:00
|
|
|
*/
|
2010-05-27 05:43:41 +08:00
|
|
|
static int update_queue(struct sem_array *sma, int semnum, struct list_head *pt)
|
2005-04-17 06:20:36 +08:00
|
|
|
{
|
2009-12-16 08:47:33 +08:00
|
|
|
struct sem_queue *q;
|
|
|
|
struct list_head *walk;
|
|
|
|
struct list_head *pending_list;
|
2010-05-27 05:43:41 +08:00
|
|
|
int semop_completed = 0;
|
2009-12-16 08:47:33 +08:00
|
|
|
|
2013-05-01 10:15:39 +08:00
|
|
|
if (semnum == -1)
|
2009-12-16 08:47:33 +08:00
|
|
|
pending_list = &sma->sem_pending;
|
2013-05-01 10:15:39 +08:00
|
|
|
else
|
2009-12-16 08:47:33 +08:00
|
|
|
pending_list = &sma->sem_base[semnum].sem_pending;
|
2009-12-16 08:47:29 +08:00
|
|
|
|
|
|
|
again:
|
2009-12-16 08:47:33 +08:00
|
|
|
walk = pending_list->next;
|
|
|
|
while (walk != pending_list) {
|
2010-05-27 05:43:40 +08:00
|
|
|
int error, restart;
|
2009-12-16 08:47:33 +08:00
|
|
|
|
2013-05-01 10:15:39 +08:00
|
|
|
q = container_of(walk, struct sem_queue, list);
|
2009-12-16 08:47:33 +08:00
|
|
|
walk = walk->next;
|
2005-04-17 06:20:36 +08:00
|
|
|
|
2009-12-16 08:47:34 +08:00
|
|
|
/* If we are scanning the single sop, per-semaphore list of
|
|
|
|
* one semaphore and that semaphore is 0, then it is not
|
|
|
|
* necessary to scan the "alter" entries: simple increments
|
|
|
|
* that affect only one entry succeed immediately and cannot
|
|
|
|
* be in the per semaphore pending queue, and decrements
|
|
|
|
* cannot be successful if the value is already 0.
|
|
|
|
*/
|
|
|
|
if (semnum != -1 && sma->sem_base[semnum].semval == 0 &&
|
|
|
|
q->alter)
|
|
|
|
break;
|
|
|
|
|
2005-04-17 06:20:36 +08:00
|
|
|
error = try_atomic_semop(sma, q->sops, q->nsops,
|
|
|
|
q->undo, q->pid);
|
|
|
|
|
|
|
|
/* Does q->sleeper still need to sleep? */
|
2009-12-16 08:47:29 +08:00
|
|
|
if (error > 0)
|
|
|
|
continue;
|
|
|
|
|
2009-12-16 08:47:32 +08:00
|
|
|
unlink_queue(sma, q);
|
2009-12-16 08:47:29 +08:00
|
|
|
|
2010-05-27 05:43:41 +08:00
|
|
|
if (error) {
|
2010-05-27 05:43:40 +08:00
|
|
|
restart = 0;
|
2010-05-27 05:43:41 +08:00
|
|
|
} else {
|
|
|
|
semop_completed = 1;
|
2010-05-27 05:43:40 +08:00
|
|
|
restart = check_restart(sma, q);
|
2010-05-27 05:43:41 +08:00
|
|
|
}
|
2010-05-27 05:43:40 +08:00
|
|
|
|
2010-05-27 05:43:41 +08:00
|
|
|
wake_up_sem_queue_prepare(pt, q, error);
|
2010-05-27 05:43:40 +08:00
|
|
|
if (restart)
|
2009-12-16 08:47:29 +08:00
|
|
|
goto again;
|
2005-04-17 06:20:36 +08:00
|
|
|
}
|
2010-05-27 05:43:41 +08:00
|
|
|
return semop_completed;
|
2005-04-17 06:20:36 +08:00
|
|
|
}
|
|
|
|
|
2010-05-27 05:43:41 +08:00
|
|
|
/**
|
|
|
|
* do_smart_update(sma, sops, nsops, otime, pt) - optimized update_queue
|
2010-05-27 05:43:40 +08:00
|
|
|
* @sma: semaphore array
|
|
|
|
* @sops: operations that were performed
|
|
|
|
* @nsops: number of operations
|
2010-05-27 05:43:41 +08:00
|
|
|
* @otime: force setting otime
|
|
|
|
* @pt: list head of the tasks that must be woken up.
|
2010-05-27 05:43:40 +08:00
|
|
|
*
|
|
|
|
* do_smart_update() does the required called to update_queue, based on the
|
|
|
|
* actual changes that were performed on the semaphore array.
|
2010-05-27 05:43:41 +08:00
|
|
|
* Note that the function does not do the actual wake-up: the caller is
|
|
|
|
* responsible for calling wake_up_sem_queue_do(@pt).
|
|
|
|
* It is safe to perform this call after dropping all locks.
|
2010-05-27 05:43:40 +08:00
|
|
|
*/
|
2010-05-27 05:43:41 +08:00
|
|
|
static void do_smart_update(struct sem_array *sma, struct sembuf *sops, int nsops,
|
|
|
|
int otime, struct list_head *pt)
|
2010-05-27 05:43:40 +08:00
|
|
|
{
|
|
|
|
int i;
|
|
|
|
|
|
|
|
if (sma->complex_count || sops == NULL) {
|
2010-05-27 05:43:41 +08:00
|
|
|
if (update_queue(sma, -1, pt))
|
|
|
|
otime = 1;
|
2013-05-01 10:15:39 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
if (!sops) {
|
|
|
|
/* No semops; something special is going on. */
|
|
|
|
for (i = 0; i < sma->sem_nsems; i++) {
|
|
|
|
if (update_queue(sma, i, pt))
|
|
|
|
otime = 1;
|
|
|
|
}
|
2010-05-27 05:43:41 +08:00
|
|
|
goto done;
|
2010-05-27 05:43:40 +08:00
|
|
|
}
|
|
|
|
|
2013-05-01 10:15:39 +08:00
|
|
|
/* Check the semaphores that were modified. */
|
2010-05-27 05:43:40 +08:00
|
|
|
for (i = 0; i < nsops; i++) {
|
|
|
|
if (sops[i].sem_op > 0 ||
|
|
|
|
(sops[i].sem_op < 0 &&
|
|
|
|
sma->sem_base[sops[i].sem_num].semval == 0))
|
2010-05-27 05:43:41 +08:00
|
|
|
if (update_queue(sma, sops[i].sem_num, pt))
|
|
|
|
otime = 1;
|
2010-05-27 05:43:40 +08:00
|
|
|
}
|
2010-05-27 05:43:41 +08:00
|
|
|
done:
|
|
|
|
if (otime)
|
|
|
|
sma->sem_otime = get_seconds();
|
2010-05-27 05:43:40 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
|
2005-04-17 06:20:36 +08:00
|
|
|
/* The following counts are associated to each semaphore:
|
|
|
|
* semncnt number of tasks waiting on semval being nonzero
|
|
|
|
* semzcnt number of tasks waiting on semval being zero
|
|
|
|
* This model assumes that a task waits on exactly one semaphore.
|
|
|
|
* Since semaphore operations are to be performed atomically, tasks actually
|
|
|
|
* wait on a whole sequence of semaphores simultaneously.
|
|
|
|
* The counts we return here are a rough approximation, but still
|
|
|
|
* warrant that semncnt+semzcnt>0 if the task is on the pending queue.
|
|
|
|
*/
|
|
|
|
static int count_semncnt (struct sem_array * sma, ushort semnum)
|
|
|
|
{
|
|
|
|
int semncnt;
|
|
|
|
struct sem_queue * q;
|
|
|
|
|
|
|
|
semncnt = 0;
|
2008-07-25 16:48:06 +08:00
|
|
|
list_for_each_entry(q, &sma->sem_pending, list) {
|
2005-04-17 06:20:36 +08:00
|
|
|
struct sembuf * sops = q->sops;
|
|
|
|
int nsops = q->nsops;
|
|
|
|
int i;
|
|
|
|
for (i = 0; i < nsops; i++)
|
|
|
|
if (sops[i].sem_num == semnum
|
|
|
|
&& (sops[i].sem_op < 0)
|
|
|
|
&& !(sops[i].sem_flg & IPC_NOWAIT))
|
|
|
|
semncnt++;
|
|
|
|
}
|
|
|
|
return semncnt;
|
|
|
|
}
|
2008-07-25 16:48:06 +08:00
|
|
|
|
2005-04-17 06:20:36 +08:00
|
|
|
static int count_semzcnt (struct sem_array * sma, ushort semnum)
|
|
|
|
{
|
|
|
|
int semzcnt;
|
|
|
|
struct sem_queue * q;
|
|
|
|
|
|
|
|
semzcnt = 0;
|
2008-07-25 16:48:06 +08:00
|
|
|
list_for_each_entry(q, &sma->sem_pending, list) {
|
2005-04-17 06:20:36 +08:00
|
|
|
struct sembuf * sops = q->sops;
|
|
|
|
int nsops = q->nsops;
|
|
|
|
int i;
|
|
|
|
for (i = 0; i < nsops; i++)
|
|
|
|
if (sops[i].sem_num == semnum
|
|
|
|
&& (sops[i].sem_op == 0)
|
|
|
|
&& !(sops[i].sem_flg & IPC_NOWAIT))
|
|
|
|
semzcnt++;
|
|
|
|
}
|
|
|
|
return semzcnt;
|
|
|
|
}
|
|
|
|
|
2007-10-19 14:40:54 +08:00
|
|
|
/* Free a semaphore set. freeary() is called with sem_ids.rw_mutex locked
|
|
|
|
* as a writer and the spinlock for this semaphore set hold. sem_ids.rw_mutex
|
|
|
|
* remains locked on exit.
|
2005-04-17 06:20:36 +08:00
|
|
|
*/
|
2008-02-08 20:18:57 +08:00
|
|
|
static void freeary(struct ipc_namespace *ns, struct kern_ipc_perm *ipcp)
|
2005-04-17 06:20:36 +08:00
|
|
|
{
|
2008-07-25 16:48:06 +08:00
|
|
|
struct sem_undo *un, *tu;
|
|
|
|
struct sem_queue *q, *tq;
|
2008-02-08 20:18:57 +08:00
|
|
|
struct sem_array *sma = container_of(ipcp, struct sem_array, sem_perm);
|
2010-05-27 05:43:41 +08:00
|
|
|
struct list_head tasks;
|
2013-05-01 10:15:39 +08:00
|
|
|
int i;
|
2005-04-17 06:20:36 +08:00
|
|
|
|
2008-07-25 16:48:06 +08:00
|
|
|
/* Free the existing undo structures for this semaphore set. */
|
2008-07-25 16:48:04 +08:00
|
|
|
assert_spin_locked(&sma->sem_perm.lock);
|
2008-07-25 16:48:06 +08:00
|
|
|
list_for_each_entry_safe(un, tu, &sma->list_id, list_id) {
|
|
|
|
list_del(&un->list_id);
|
|
|
|
spin_lock(&un->ulp->lock);
|
2005-04-17 06:20:36 +08:00
|
|
|
un->semid = -1;
|
2008-07-25 16:48:06 +08:00
|
|
|
list_del_rcu(&un->list_proc);
|
|
|
|
spin_unlock(&un->ulp->lock);
|
2011-03-18 12:09:35 +08:00
|
|
|
kfree_rcu(un, rcu);
|
2008-07-25 16:48:06 +08:00
|
|
|
}
|
2005-04-17 06:20:36 +08:00
|
|
|
|
|
|
|
/* Wake up all pending processes and let them fail with EIDRM. */
|
2010-05-27 05:43:41 +08:00
|
|
|
INIT_LIST_HEAD(&tasks);
|
2008-07-25 16:48:06 +08:00
|
|
|
list_for_each_entry_safe(q, tq, &sma->sem_pending, list) {
|
2009-12-16 08:47:32 +08:00
|
|
|
unlink_queue(sma, q);
|
2010-05-27 05:43:41 +08:00
|
|
|
wake_up_sem_queue_prepare(&tasks, q, -EIDRM);
|
2005-04-17 06:20:36 +08:00
|
|
|
}
|
2013-05-01 10:15:39 +08:00
|
|
|
for (i = 0; i < sma->sem_nsems; i++) {
|
|
|
|
struct sem *sem = sma->sem_base + i;
|
|
|
|
list_for_each_entry_safe(q, tq, &sem->sem_pending, list) {
|
|
|
|
unlink_queue(sma, q);
|
|
|
|
wake_up_sem_queue_prepare(&tasks, q, -EIDRM);
|
|
|
|
}
|
|
|
|
}
|
2005-04-17 06:20:36 +08:00
|
|
|
|
2007-10-19 14:40:48 +08:00
|
|
|
/* Remove the semaphore set from the IDR */
|
|
|
|
sem_rmid(ns, sma);
|
2013-05-01 10:15:44 +08:00
|
|
|
sem_unlock(sma, -1);
|
ipc: move rcu_read_unlock() out of sem_unlock() and into callers
The IPC locking is a mess, and sem_unlock() unlocks not only the
semaphore spinlock, it also drops the rcu read lock. Unlike sem_lock(),
which just gets the spin-lock, and expects the caller to get the rcu
read lock.
This all makes things very hard to follow, and it's very confusing when
you take the rcu read lock in one function, and then release it in
another. And it has caused actual bugs: the sem_obtain_lock() function
ended up dropping the RCU read lock twice in one error path, because it
first did the sem_unlock(), and then did a rcu_read_unlock() to match
the rcu_read_lock() it had done.
This is just a totally mindless "remove rcu_read_unlock() from
sem_unlock() and add it immediately after each caller" (except for the
aforementioned bug where we did too many rcu_read_unlock(), and in
find_alloc_undo() where we just got the rcu_read_lock() to correct for
the fact that sem_unlock would immediately drop it again).
We can (and should) clean things up further, but this fixes the bug with
the minimal amount of subtlety.
Reviewed-by: Davidlohr Bueso <davidlohr.bueso@hp.com>
Cc: Rik van Riel <riel@redhat.com>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2013-05-04 06:04:40 +08:00
|
|
|
rcu_read_unlock();
|
2005-04-17 06:20:36 +08:00
|
|
|
|
2010-05-27 05:43:41 +08:00
|
|
|
wake_up_sem_queue_do(&tasks);
|
2006-10-02 17:18:22 +08:00
|
|
|
ns->used_sems -= sma->sem_nsems;
|
2005-04-17 06:20:36 +08:00
|
|
|
security_sem_free(sma);
|
|
|
|
ipc_rcu_putref(sma);
|
|
|
|
}
|
|
|
|
|
|
|
|
static unsigned long copy_semid_to_user(void __user *buf, struct semid64_ds *in, int version)
|
|
|
|
{
|
|
|
|
switch(version) {
|
|
|
|
case IPC_64:
|
|
|
|
return copy_to_user(buf, in, sizeof(*in));
|
|
|
|
case IPC_OLD:
|
|
|
|
{
|
|
|
|
struct semid_ds out;
|
|
|
|
|
2010-10-01 06:15:31 +08:00
|
|
|
memset(&out, 0, sizeof(out));
|
|
|
|
|
2005-04-17 06:20:36 +08:00
|
|
|
ipc64_perm_to_ipc_perm(&in->sem_perm, &out.sem_perm);
|
|
|
|
|
|
|
|
out.sem_otime = in->sem_otime;
|
|
|
|
out.sem_ctime = in->sem_ctime;
|
|
|
|
out.sem_nsems = in->sem_nsems;
|
|
|
|
|
|
|
|
return copy_to_user(buf, &out, sizeof(out));
|
|
|
|
}
|
|
|
|
default:
|
|
|
|
return -EINVAL;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2008-02-08 20:18:56 +08:00
|
|
|
static int semctl_nolock(struct ipc_namespace *ns, int semid,
|
2013-03-06 04:04:55 +08:00
|
|
|
int cmd, int version, void __user *p)
|
2005-04-17 06:20:36 +08:00
|
|
|
{
|
2009-12-16 08:47:35 +08:00
|
|
|
int err;
|
2005-04-17 06:20:36 +08:00
|
|
|
struct sem_array *sma;
|
|
|
|
|
|
|
|
switch(cmd) {
|
|
|
|
case IPC_INFO:
|
|
|
|
case SEM_INFO:
|
|
|
|
{
|
|
|
|
struct seminfo seminfo;
|
|
|
|
int max_id;
|
|
|
|
|
|
|
|
err = security_sem_semctl(NULL, cmd);
|
|
|
|
if (err)
|
|
|
|
return err;
|
|
|
|
|
|
|
|
memset(&seminfo,0,sizeof(seminfo));
|
2006-10-02 17:18:22 +08:00
|
|
|
seminfo.semmni = ns->sc_semmni;
|
|
|
|
seminfo.semmns = ns->sc_semmns;
|
|
|
|
seminfo.semmsl = ns->sc_semmsl;
|
|
|
|
seminfo.semopm = ns->sc_semopm;
|
2005-04-17 06:20:36 +08:00
|
|
|
seminfo.semvmx = SEMVMX;
|
|
|
|
seminfo.semmnu = SEMMNU;
|
|
|
|
seminfo.semmap = SEMMAP;
|
|
|
|
seminfo.semume = SEMUME;
|
2007-10-19 14:40:54 +08:00
|
|
|
down_read(&sem_ids(ns).rw_mutex);
|
2005-04-17 06:20:36 +08:00
|
|
|
if (cmd == SEM_INFO) {
|
2006-10-02 17:18:22 +08:00
|
|
|
seminfo.semusz = sem_ids(ns).in_use;
|
|
|
|
seminfo.semaem = ns->used_sems;
|
2005-04-17 06:20:36 +08:00
|
|
|
} else {
|
|
|
|
seminfo.semusz = SEMUSZ;
|
|
|
|
seminfo.semaem = SEMAEM;
|
|
|
|
}
|
2007-10-19 14:40:48 +08:00
|
|
|
max_id = ipc_get_maxid(&sem_ids(ns));
|
2007-10-19 14:40:54 +08:00
|
|
|
up_read(&sem_ids(ns).rw_mutex);
|
2013-03-06 04:04:55 +08:00
|
|
|
if (copy_to_user(p, &seminfo, sizeof(struct seminfo)))
|
2005-04-17 06:20:36 +08:00
|
|
|
return -EFAULT;
|
|
|
|
return (max_id < 0) ? 0: max_id;
|
|
|
|
}
|
2008-02-08 20:18:56 +08:00
|
|
|
case IPC_STAT:
|
2005-04-17 06:20:36 +08:00
|
|
|
case SEM_STAT:
|
|
|
|
{
|
|
|
|
struct semid64_ds tbuf;
|
ipc,sem: do not hold ipc lock more than necessary
Instead of holding the ipc lock for permissions and security checks, among
others, only acquire it when necessary.
Some numbers....
1) With Rik's semop-multi.c microbenchmark we can see the following
results:
Baseline (3.9-rc1):
cpus 4, threads: 256, semaphores: 128, test duration: 30 secs
total operations: 151452270, ops/sec 5048409
+ 59.40% a.out [kernel.kallsyms] [k] _raw_spin_lock
+ 6.14% a.out [kernel.kallsyms] [k] sys_semtimedop
+ 3.84% a.out [kernel.kallsyms] [k] avc_has_perm_flags
+ 3.64% a.out [kernel.kallsyms] [k] __audit_syscall_exit
+ 2.06% a.out [kernel.kallsyms] [k] copy_user_enhanced_fast_string
+ 1.86% a.out [kernel.kallsyms] [k] ipc_lock
With this patchset:
cpus 4, threads: 256, semaphores: 128, test duration: 30 secs
total operations: 273156400, ops/sec 9105213
+ 18.54% a.out [kernel.kallsyms] [k] _raw_spin_lock
+ 11.72% a.out [kernel.kallsyms] [k] sys_semtimedop
+ 7.70% a.out [kernel.kallsyms] [k] ipc_has_perm.isra.21
+ 6.58% a.out [kernel.kallsyms] [k] avc_has_perm_flags
+ 6.54% a.out [kernel.kallsyms] [k] __audit_syscall_exit
+ 4.71% a.out [kernel.kallsyms] [k] ipc_obtain_object_check
2) While on an Oracle swingbench DSS (data mining) workload the
improvements are not as exciting as with Rik's benchmark, we can see
some positive numbers. For an 8 socket machine the following are the
percentages of %sys time incurred in the ipc lock:
Baseline (3.9-rc1):
100 swingbench users: 8,74%
400 swingbench users: 21,86%
800 swingbench users: 84,35%
With this patchset:
100 swingbench users: 8,11%
400 swingbench users: 19,93%
800 swingbench users: 77,69%
[riel@redhat.com: fix two locking bugs]
[sasha.levin@oracle.com: prevent releasing RCU read lock twice in semctl_main]
[akpm@linux-foundation.org: coding-style fixes]
Signed-off-by: Davidlohr Bueso <davidlohr.bueso@hp.com>
Signed-off-by: Rik van Riel <riel@redhat.com>
Reviewed-by: Chegu Vinod <chegu_vinod@hp.com>
Acked-by: Michel Lespinasse <walken@google.com>
Cc: Rik van Riel <riel@redhat.com>
Cc: Jason Low <jason.low2@hp.com>
Cc: Emmanuel Benisty <benisty.e@gmail.com>
Cc: Peter Hurley <peter@hurleysoftware.com>
Cc: Stanislav Kinsbursky <skinsbursky@parallels.com>
Tested-by: Sedat Dilek <sedat.dilek@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2013-05-01 10:15:29 +08:00
|
|
|
int id = 0;
|
|
|
|
|
|
|
|
memset(&tbuf, 0, sizeof(tbuf));
|
2005-04-17 06:20:36 +08:00
|
|
|
|
2008-02-08 20:18:56 +08:00
|
|
|
if (cmd == SEM_STAT) {
|
ipc,sem: do not hold ipc lock more than necessary
Instead of holding the ipc lock for permissions and security checks, among
others, only acquire it when necessary.
Some numbers....
1) With Rik's semop-multi.c microbenchmark we can see the following
results:
Baseline (3.9-rc1):
cpus 4, threads: 256, semaphores: 128, test duration: 30 secs
total operations: 151452270, ops/sec 5048409
+ 59.40% a.out [kernel.kallsyms] [k] _raw_spin_lock
+ 6.14% a.out [kernel.kallsyms] [k] sys_semtimedop
+ 3.84% a.out [kernel.kallsyms] [k] avc_has_perm_flags
+ 3.64% a.out [kernel.kallsyms] [k] __audit_syscall_exit
+ 2.06% a.out [kernel.kallsyms] [k] copy_user_enhanced_fast_string
+ 1.86% a.out [kernel.kallsyms] [k] ipc_lock
With this patchset:
cpus 4, threads: 256, semaphores: 128, test duration: 30 secs
total operations: 273156400, ops/sec 9105213
+ 18.54% a.out [kernel.kallsyms] [k] _raw_spin_lock
+ 11.72% a.out [kernel.kallsyms] [k] sys_semtimedop
+ 7.70% a.out [kernel.kallsyms] [k] ipc_has_perm.isra.21
+ 6.58% a.out [kernel.kallsyms] [k] avc_has_perm_flags
+ 6.54% a.out [kernel.kallsyms] [k] __audit_syscall_exit
+ 4.71% a.out [kernel.kallsyms] [k] ipc_obtain_object_check
2) While on an Oracle swingbench DSS (data mining) workload the
improvements are not as exciting as with Rik's benchmark, we can see
some positive numbers. For an 8 socket machine the following are the
percentages of %sys time incurred in the ipc lock:
Baseline (3.9-rc1):
100 swingbench users: 8,74%
400 swingbench users: 21,86%
800 swingbench users: 84,35%
With this patchset:
100 swingbench users: 8,11%
400 swingbench users: 19,93%
800 swingbench users: 77,69%
[riel@redhat.com: fix two locking bugs]
[sasha.levin@oracle.com: prevent releasing RCU read lock twice in semctl_main]
[akpm@linux-foundation.org: coding-style fixes]
Signed-off-by: Davidlohr Bueso <davidlohr.bueso@hp.com>
Signed-off-by: Rik van Riel <riel@redhat.com>
Reviewed-by: Chegu Vinod <chegu_vinod@hp.com>
Acked-by: Michel Lespinasse <walken@google.com>
Cc: Rik van Riel <riel@redhat.com>
Cc: Jason Low <jason.low2@hp.com>
Cc: Emmanuel Benisty <benisty.e@gmail.com>
Cc: Peter Hurley <peter@hurleysoftware.com>
Cc: Stanislav Kinsbursky <skinsbursky@parallels.com>
Tested-by: Sedat Dilek <sedat.dilek@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2013-05-01 10:15:29 +08:00
|
|
|
rcu_read_lock();
|
|
|
|
sma = sem_obtain_object(ns, semid);
|
|
|
|
if (IS_ERR(sma)) {
|
|
|
|
err = PTR_ERR(sma);
|
|
|
|
goto out_unlock;
|
|
|
|
}
|
2008-02-08 20:18:56 +08:00
|
|
|
id = sma->sem_perm.id;
|
|
|
|
} else {
|
ipc,sem: do not hold ipc lock more than necessary
Instead of holding the ipc lock for permissions and security checks, among
others, only acquire it when necessary.
Some numbers....
1) With Rik's semop-multi.c microbenchmark we can see the following
results:
Baseline (3.9-rc1):
cpus 4, threads: 256, semaphores: 128, test duration: 30 secs
total operations: 151452270, ops/sec 5048409
+ 59.40% a.out [kernel.kallsyms] [k] _raw_spin_lock
+ 6.14% a.out [kernel.kallsyms] [k] sys_semtimedop
+ 3.84% a.out [kernel.kallsyms] [k] avc_has_perm_flags
+ 3.64% a.out [kernel.kallsyms] [k] __audit_syscall_exit
+ 2.06% a.out [kernel.kallsyms] [k] copy_user_enhanced_fast_string
+ 1.86% a.out [kernel.kallsyms] [k] ipc_lock
With this patchset:
cpus 4, threads: 256, semaphores: 128, test duration: 30 secs
total operations: 273156400, ops/sec 9105213
+ 18.54% a.out [kernel.kallsyms] [k] _raw_spin_lock
+ 11.72% a.out [kernel.kallsyms] [k] sys_semtimedop
+ 7.70% a.out [kernel.kallsyms] [k] ipc_has_perm.isra.21
+ 6.58% a.out [kernel.kallsyms] [k] avc_has_perm_flags
+ 6.54% a.out [kernel.kallsyms] [k] __audit_syscall_exit
+ 4.71% a.out [kernel.kallsyms] [k] ipc_obtain_object_check
2) While on an Oracle swingbench DSS (data mining) workload the
improvements are not as exciting as with Rik's benchmark, we can see
some positive numbers. For an 8 socket machine the following are the
percentages of %sys time incurred in the ipc lock:
Baseline (3.9-rc1):
100 swingbench users: 8,74%
400 swingbench users: 21,86%
800 swingbench users: 84,35%
With this patchset:
100 swingbench users: 8,11%
400 swingbench users: 19,93%
800 swingbench users: 77,69%
[riel@redhat.com: fix two locking bugs]
[sasha.levin@oracle.com: prevent releasing RCU read lock twice in semctl_main]
[akpm@linux-foundation.org: coding-style fixes]
Signed-off-by: Davidlohr Bueso <davidlohr.bueso@hp.com>
Signed-off-by: Rik van Riel <riel@redhat.com>
Reviewed-by: Chegu Vinod <chegu_vinod@hp.com>
Acked-by: Michel Lespinasse <walken@google.com>
Cc: Rik van Riel <riel@redhat.com>
Cc: Jason Low <jason.low2@hp.com>
Cc: Emmanuel Benisty <benisty.e@gmail.com>
Cc: Peter Hurley <peter@hurleysoftware.com>
Cc: Stanislav Kinsbursky <skinsbursky@parallels.com>
Tested-by: Sedat Dilek <sedat.dilek@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2013-05-01 10:15:29 +08:00
|
|
|
rcu_read_lock();
|
|
|
|
sma = sem_obtain_object_check(ns, semid);
|
|
|
|
if (IS_ERR(sma)) {
|
|
|
|
err = PTR_ERR(sma);
|
|
|
|
goto out_unlock;
|
|
|
|
}
|
2008-02-08 20:18:56 +08:00
|
|
|
}
|
2005-04-17 06:20:36 +08:00
|
|
|
|
|
|
|
err = -EACCES;
|
2011-03-24 07:43:24 +08:00
|
|
|
if (ipcperms(ns, &sma->sem_perm, S_IRUGO))
|
2005-04-17 06:20:36 +08:00
|
|
|
goto out_unlock;
|
|
|
|
|
|
|
|
err = security_sem_semctl(sma, cmd);
|
|
|
|
if (err)
|
|
|
|
goto out_unlock;
|
|
|
|
|
|
|
|
kernel_to_ipc64_perm(&sma->sem_perm, &tbuf.sem_perm);
|
|
|
|
tbuf.sem_otime = sma->sem_otime;
|
|
|
|
tbuf.sem_ctime = sma->sem_ctime;
|
|
|
|
tbuf.sem_nsems = sma->sem_nsems;
|
ipc,sem: do not hold ipc lock more than necessary
Instead of holding the ipc lock for permissions and security checks, among
others, only acquire it when necessary.
Some numbers....
1) With Rik's semop-multi.c microbenchmark we can see the following
results:
Baseline (3.9-rc1):
cpus 4, threads: 256, semaphores: 128, test duration: 30 secs
total operations: 151452270, ops/sec 5048409
+ 59.40% a.out [kernel.kallsyms] [k] _raw_spin_lock
+ 6.14% a.out [kernel.kallsyms] [k] sys_semtimedop
+ 3.84% a.out [kernel.kallsyms] [k] avc_has_perm_flags
+ 3.64% a.out [kernel.kallsyms] [k] __audit_syscall_exit
+ 2.06% a.out [kernel.kallsyms] [k] copy_user_enhanced_fast_string
+ 1.86% a.out [kernel.kallsyms] [k] ipc_lock
With this patchset:
cpus 4, threads: 256, semaphores: 128, test duration: 30 secs
total operations: 273156400, ops/sec 9105213
+ 18.54% a.out [kernel.kallsyms] [k] _raw_spin_lock
+ 11.72% a.out [kernel.kallsyms] [k] sys_semtimedop
+ 7.70% a.out [kernel.kallsyms] [k] ipc_has_perm.isra.21
+ 6.58% a.out [kernel.kallsyms] [k] avc_has_perm_flags
+ 6.54% a.out [kernel.kallsyms] [k] __audit_syscall_exit
+ 4.71% a.out [kernel.kallsyms] [k] ipc_obtain_object_check
2) While on an Oracle swingbench DSS (data mining) workload the
improvements are not as exciting as with Rik's benchmark, we can see
some positive numbers. For an 8 socket machine the following are the
percentages of %sys time incurred in the ipc lock:
Baseline (3.9-rc1):
100 swingbench users: 8,74%
400 swingbench users: 21,86%
800 swingbench users: 84,35%
With this patchset:
100 swingbench users: 8,11%
400 swingbench users: 19,93%
800 swingbench users: 77,69%
[riel@redhat.com: fix two locking bugs]
[sasha.levin@oracle.com: prevent releasing RCU read lock twice in semctl_main]
[akpm@linux-foundation.org: coding-style fixes]
Signed-off-by: Davidlohr Bueso <davidlohr.bueso@hp.com>
Signed-off-by: Rik van Riel <riel@redhat.com>
Reviewed-by: Chegu Vinod <chegu_vinod@hp.com>
Acked-by: Michel Lespinasse <walken@google.com>
Cc: Rik van Riel <riel@redhat.com>
Cc: Jason Low <jason.low2@hp.com>
Cc: Emmanuel Benisty <benisty.e@gmail.com>
Cc: Peter Hurley <peter@hurleysoftware.com>
Cc: Stanislav Kinsbursky <skinsbursky@parallels.com>
Tested-by: Sedat Dilek <sedat.dilek@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2013-05-01 10:15:29 +08:00
|
|
|
rcu_read_unlock();
|
2013-03-06 04:04:55 +08:00
|
|
|
if (copy_semid_to_user(p, &tbuf, version))
|
2005-04-17 06:20:36 +08:00
|
|
|
return -EFAULT;
|
|
|
|
return id;
|
|
|
|
}
|
|
|
|
default:
|
|
|
|
return -EINVAL;
|
|
|
|
}
|
|
|
|
out_unlock:
|
ipc,sem: do not hold ipc lock more than necessary
Instead of holding the ipc lock for permissions and security checks, among
others, only acquire it when necessary.
Some numbers....
1) With Rik's semop-multi.c microbenchmark we can see the following
results:
Baseline (3.9-rc1):
cpus 4, threads: 256, semaphores: 128, test duration: 30 secs
total operations: 151452270, ops/sec 5048409
+ 59.40% a.out [kernel.kallsyms] [k] _raw_spin_lock
+ 6.14% a.out [kernel.kallsyms] [k] sys_semtimedop
+ 3.84% a.out [kernel.kallsyms] [k] avc_has_perm_flags
+ 3.64% a.out [kernel.kallsyms] [k] __audit_syscall_exit
+ 2.06% a.out [kernel.kallsyms] [k] copy_user_enhanced_fast_string
+ 1.86% a.out [kernel.kallsyms] [k] ipc_lock
With this patchset:
cpus 4, threads: 256, semaphores: 128, test duration: 30 secs
total operations: 273156400, ops/sec 9105213
+ 18.54% a.out [kernel.kallsyms] [k] _raw_spin_lock
+ 11.72% a.out [kernel.kallsyms] [k] sys_semtimedop
+ 7.70% a.out [kernel.kallsyms] [k] ipc_has_perm.isra.21
+ 6.58% a.out [kernel.kallsyms] [k] avc_has_perm_flags
+ 6.54% a.out [kernel.kallsyms] [k] __audit_syscall_exit
+ 4.71% a.out [kernel.kallsyms] [k] ipc_obtain_object_check
2) While on an Oracle swingbench DSS (data mining) workload the
improvements are not as exciting as with Rik's benchmark, we can see
some positive numbers. For an 8 socket machine the following are the
percentages of %sys time incurred in the ipc lock:
Baseline (3.9-rc1):
100 swingbench users: 8,74%
400 swingbench users: 21,86%
800 swingbench users: 84,35%
With this patchset:
100 swingbench users: 8,11%
400 swingbench users: 19,93%
800 swingbench users: 77,69%
[riel@redhat.com: fix two locking bugs]
[sasha.levin@oracle.com: prevent releasing RCU read lock twice in semctl_main]
[akpm@linux-foundation.org: coding-style fixes]
Signed-off-by: Davidlohr Bueso <davidlohr.bueso@hp.com>
Signed-off-by: Rik van Riel <riel@redhat.com>
Reviewed-by: Chegu Vinod <chegu_vinod@hp.com>
Acked-by: Michel Lespinasse <walken@google.com>
Cc: Rik van Riel <riel@redhat.com>
Cc: Jason Low <jason.low2@hp.com>
Cc: Emmanuel Benisty <benisty.e@gmail.com>
Cc: Peter Hurley <peter@hurleysoftware.com>
Cc: Stanislav Kinsbursky <skinsbursky@parallels.com>
Tested-by: Sedat Dilek <sedat.dilek@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2013-05-01 10:15:29 +08:00
|
|
|
rcu_read_unlock();
|
2005-04-17 06:20:36 +08:00
|
|
|
return err;
|
|
|
|
}
|
|
|
|
|
2013-03-06 04:04:55 +08:00
|
|
|
static int semctl_setval(struct ipc_namespace *ns, int semid, int semnum,
|
|
|
|
unsigned long arg)
|
|
|
|
{
|
|
|
|
struct sem_undo *un;
|
|
|
|
struct sem_array *sma;
|
|
|
|
struct sem* curr;
|
|
|
|
int err;
|
|
|
|
struct list_head tasks;
|
|
|
|
int val;
|
|
|
|
#if defined(CONFIG_64BIT) && defined(__BIG_ENDIAN)
|
|
|
|
/* big-endian 64bit */
|
|
|
|
val = arg >> 32;
|
|
|
|
#else
|
|
|
|
/* 32bit or little-endian 64bit */
|
|
|
|
val = arg;
|
|
|
|
#endif
|
|
|
|
|
2013-05-01 10:15:44 +08:00
|
|
|
if (val > SEMVMX || val < 0)
|
|
|
|
return -ERANGE;
|
2013-03-06 04:04:55 +08:00
|
|
|
|
|
|
|
INIT_LIST_HEAD(&tasks);
|
|
|
|
|
2013-05-01 10:15:44 +08:00
|
|
|
rcu_read_lock();
|
|
|
|
sma = sem_obtain_object_check(ns, semid);
|
|
|
|
if (IS_ERR(sma)) {
|
|
|
|
rcu_read_unlock();
|
|
|
|
return PTR_ERR(sma);
|
|
|
|
}
|
|
|
|
|
|
|
|
if (semnum < 0 || semnum >= sma->sem_nsems) {
|
|
|
|
rcu_read_unlock();
|
|
|
|
return -EINVAL;
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
if (ipcperms(ns, &sma->sem_perm, S_IWUGO)) {
|
|
|
|
rcu_read_unlock();
|
|
|
|
return -EACCES;
|
|
|
|
}
|
2013-03-06 04:04:55 +08:00
|
|
|
|
|
|
|
err = security_sem_semctl(sma, SETVAL);
|
2013-05-01 10:15:44 +08:00
|
|
|
if (err) {
|
|
|
|
rcu_read_unlock();
|
|
|
|
return -EACCES;
|
|
|
|
}
|
2013-03-06 04:04:55 +08:00
|
|
|
|
2013-05-01 10:15:44 +08:00
|
|
|
sem_lock(sma, NULL, -1);
|
2013-03-06 04:04:55 +08:00
|
|
|
|
|
|
|
curr = &sma->sem_base[semnum];
|
|
|
|
|
|
|
|
assert_spin_locked(&sma->sem_perm.lock);
|
|
|
|
list_for_each_entry(un, &sma->list_id, list_id)
|
|
|
|
un->semadj[semnum] = 0;
|
|
|
|
|
|
|
|
curr->semval = val;
|
|
|
|
curr->sempid = task_tgid_vnr(current);
|
|
|
|
sma->sem_ctime = get_seconds();
|
|
|
|
/* maybe some queued-up processes were waiting for this */
|
|
|
|
do_smart_update(sma, NULL, 0, 0, &tasks);
|
2013-05-01 10:15:44 +08:00
|
|
|
sem_unlock(sma, -1);
|
ipc: move rcu_read_unlock() out of sem_unlock() and into callers
The IPC locking is a mess, and sem_unlock() unlocks not only the
semaphore spinlock, it also drops the rcu read lock. Unlike sem_lock(),
which just gets the spin-lock, and expects the caller to get the rcu
read lock.
This all makes things very hard to follow, and it's very confusing when
you take the rcu read lock in one function, and then release it in
another. And it has caused actual bugs: the sem_obtain_lock() function
ended up dropping the RCU read lock twice in one error path, because it
first did the sem_unlock(), and then did a rcu_read_unlock() to match
the rcu_read_lock() it had done.
This is just a totally mindless "remove rcu_read_unlock() from
sem_unlock() and add it immediately after each caller" (except for the
aforementioned bug where we did too many rcu_read_unlock(), and in
find_alloc_undo() where we just got the rcu_read_lock() to correct for
the fact that sem_unlock would immediately drop it again).
We can (and should) clean things up further, but this fixes the bug with
the minimal amount of subtlety.
Reviewed-by: Davidlohr Bueso <davidlohr.bueso@hp.com>
Cc: Rik van Riel <riel@redhat.com>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2013-05-04 06:04:40 +08:00
|
|
|
rcu_read_unlock();
|
2013-03-06 04:04:55 +08:00
|
|
|
wake_up_sem_queue_do(&tasks);
|
2013-05-01 10:15:44 +08:00
|
|
|
return 0;
|
2013-03-06 04:04:55 +08:00
|
|
|
}
|
|
|
|
|
2006-10-02 17:18:22 +08:00
|
|
|
static int semctl_main(struct ipc_namespace *ns, int semid, int semnum,
|
2013-03-06 04:04:55 +08:00
|
|
|
int cmd, void __user *p)
|
2005-04-17 06:20:36 +08:00
|
|
|
{
|
|
|
|
struct sem_array *sma;
|
|
|
|
struct sem* curr;
|
ipc,sem: do not hold ipc lock more than necessary
Instead of holding the ipc lock for permissions and security checks, among
others, only acquire it when necessary.
Some numbers....
1) With Rik's semop-multi.c microbenchmark we can see the following
results:
Baseline (3.9-rc1):
cpus 4, threads: 256, semaphores: 128, test duration: 30 secs
total operations: 151452270, ops/sec 5048409
+ 59.40% a.out [kernel.kallsyms] [k] _raw_spin_lock
+ 6.14% a.out [kernel.kallsyms] [k] sys_semtimedop
+ 3.84% a.out [kernel.kallsyms] [k] avc_has_perm_flags
+ 3.64% a.out [kernel.kallsyms] [k] __audit_syscall_exit
+ 2.06% a.out [kernel.kallsyms] [k] copy_user_enhanced_fast_string
+ 1.86% a.out [kernel.kallsyms] [k] ipc_lock
With this patchset:
cpus 4, threads: 256, semaphores: 128, test duration: 30 secs
total operations: 273156400, ops/sec 9105213
+ 18.54% a.out [kernel.kallsyms] [k] _raw_spin_lock
+ 11.72% a.out [kernel.kallsyms] [k] sys_semtimedop
+ 7.70% a.out [kernel.kallsyms] [k] ipc_has_perm.isra.21
+ 6.58% a.out [kernel.kallsyms] [k] avc_has_perm_flags
+ 6.54% a.out [kernel.kallsyms] [k] __audit_syscall_exit
+ 4.71% a.out [kernel.kallsyms] [k] ipc_obtain_object_check
2) While on an Oracle swingbench DSS (data mining) workload the
improvements are not as exciting as with Rik's benchmark, we can see
some positive numbers. For an 8 socket machine the following are the
percentages of %sys time incurred in the ipc lock:
Baseline (3.9-rc1):
100 swingbench users: 8,74%
400 swingbench users: 21,86%
800 swingbench users: 84,35%
With this patchset:
100 swingbench users: 8,11%
400 swingbench users: 19,93%
800 swingbench users: 77,69%
[riel@redhat.com: fix two locking bugs]
[sasha.levin@oracle.com: prevent releasing RCU read lock twice in semctl_main]
[akpm@linux-foundation.org: coding-style fixes]
Signed-off-by: Davidlohr Bueso <davidlohr.bueso@hp.com>
Signed-off-by: Rik van Riel <riel@redhat.com>
Reviewed-by: Chegu Vinod <chegu_vinod@hp.com>
Acked-by: Michel Lespinasse <walken@google.com>
Cc: Rik van Riel <riel@redhat.com>
Cc: Jason Low <jason.low2@hp.com>
Cc: Emmanuel Benisty <benisty.e@gmail.com>
Cc: Peter Hurley <peter@hurleysoftware.com>
Cc: Stanislav Kinsbursky <skinsbursky@parallels.com>
Tested-by: Sedat Dilek <sedat.dilek@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2013-05-01 10:15:29 +08:00
|
|
|
int err, nsems;
|
2005-04-17 06:20:36 +08:00
|
|
|
ushort fast_sem_io[SEMMSL_FAST];
|
|
|
|
ushort* sem_io = fast_sem_io;
|
2010-05-27 05:43:41 +08:00
|
|
|
struct list_head tasks;
|
2005-04-17 06:20:36 +08:00
|
|
|
|
ipc,sem: do not hold ipc lock more than necessary
Instead of holding the ipc lock for permissions and security checks, among
others, only acquire it when necessary.
Some numbers....
1) With Rik's semop-multi.c microbenchmark we can see the following
results:
Baseline (3.9-rc1):
cpus 4, threads: 256, semaphores: 128, test duration: 30 secs
total operations: 151452270, ops/sec 5048409
+ 59.40% a.out [kernel.kallsyms] [k] _raw_spin_lock
+ 6.14% a.out [kernel.kallsyms] [k] sys_semtimedop
+ 3.84% a.out [kernel.kallsyms] [k] avc_has_perm_flags
+ 3.64% a.out [kernel.kallsyms] [k] __audit_syscall_exit
+ 2.06% a.out [kernel.kallsyms] [k] copy_user_enhanced_fast_string
+ 1.86% a.out [kernel.kallsyms] [k] ipc_lock
With this patchset:
cpus 4, threads: 256, semaphores: 128, test duration: 30 secs
total operations: 273156400, ops/sec 9105213
+ 18.54% a.out [kernel.kallsyms] [k] _raw_spin_lock
+ 11.72% a.out [kernel.kallsyms] [k] sys_semtimedop
+ 7.70% a.out [kernel.kallsyms] [k] ipc_has_perm.isra.21
+ 6.58% a.out [kernel.kallsyms] [k] avc_has_perm_flags
+ 6.54% a.out [kernel.kallsyms] [k] __audit_syscall_exit
+ 4.71% a.out [kernel.kallsyms] [k] ipc_obtain_object_check
2) While on an Oracle swingbench DSS (data mining) workload the
improvements are not as exciting as with Rik's benchmark, we can see
some positive numbers. For an 8 socket machine the following are the
percentages of %sys time incurred in the ipc lock:
Baseline (3.9-rc1):
100 swingbench users: 8,74%
400 swingbench users: 21,86%
800 swingbench users: 84,35%
With this patchset:
100 swingbench users: 8,11%
400 swingbench users: 19,93%
800 swingbench users: 77,69%
[riel@redhat.com: fix two locking bugs]
[sasha.levin@oracle.com: prevent releasing RCU read lock twice in semctl_main]
[akpm@linux-foundation.org: coding-style fixes]
Signed-off-by: Davidlohr Bueso <davidlohr.bueso@hp.com>
Signed-off-by: Rik van Riel <riel@redhat.com>
Reviewed-by: Chegu Vinod <chegu_vinod@hp.com>
Acked-by: Michel Lespinasse <walken@google.com>
Cc: Rik van Riel <riel@redhat.com>
Cc: Jason Low <jason.low2@hp.com>
Cc: Emmanuel Benisty <benisty.e@gmail.com>
Cc: Peter Hurley <peter@hurleysoftware.com>
Cc: Stanislav Kinsbursky <skinsbursky@parallels.com>
Tested-by: Sedat Dilek <sedat.dilek@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2013-05-01 10:15:29 +08:00
|
|
|
INIT_LIST_HEAD(&tasks);
|
|
|
|
|
|
|
|
rcu_read_lock();
|
|
|
|
sma = sem_obtain_object_check(ns, semid);
|
|
|
|
if (IS_ERR(sma)) {
|
|
|
|
rcu_read_unlock();
|
2007-10-19 14:40:51 +08:00
|
|
|
return PTR_ERR(sma);
|
ipc,sem: do not hold ipc lock more than necessary
Instead of holding the ipc lock for permissions and security checks, among
others, only acquire it when necessary.
Some numbers....
1) With Rik's semop-multi.c microbenchmark we can see the following
results:
Baseline (3.9-rc1):
cpus 4, threads: 256, semaphores: 128, test duration: 30 secs
total operations: 151452270, ops/sec 5048409
+ 59.40% a.out [kernel.kallsyms] [k] _raw_spin_lock
+ 6.14% a.out [kernel.kallsyms] [k] sys_semtimedop
+ 3.84% a.out [kernel.kallsyms] [k] avc_has_perm_flags
+ 3.64% a.out [kernel.kallsyms] [k] __audit_syscall_exit
+ 2.06% a.out [kernel.kallsyms] [k] copy_user_enhanced_fast_string
+ 1.86% a.out [kernel.kallsyms] [k] ipc_lock
With this patchset:
cpus 4, threads: 256, semaphores: 128, test duration: 30 secs
total operations: 273156400, ops/sec 9105213
+ 18.54% a.out [kernel.kallsyms] [k] _raw_spin_lock
+ 11.72% a.out [kernel.kallsyms] [k] sys_semtimedop
+ 7.70% a.out [kernel.kallsyms] [k] ipc_has_perm.isra.21
+ 6.58% a.out [kernel.kallsyms] [k] avc_has_perm_flags
+ 6.54% a.out [kernel.kallsyms] [k] __audit_syscall_exit
+ 4.71% a.out [kernel.kallsyms] [k] ipc_obtain_object_check
2) While on an Oracle swingbench DSS (data mining) workload the
improvements are not as exciting as with Rik's benchmark, we can see
some positive numbers. For an 8 socket machine the following are the
percentages of %sys time incurred in the ipc lock:
Baseline (3.9-rc1):
100 swingbench users: 8,74%
400 swingbench users: 21,86%
800 swingbench users: 84,35%
With this patchset:
100 swingbench users: 8,11%
400 swingbench users: 19,93%
800 swingbench users: 77,69%
[riel@redhat.com: fix two locking bugs]
[sasha.levin@oracle.com: prevent releasing RCU read lock twice in semctl_main]
[akpm@linux-foundation.org: coding-style fixes]
Signed-off-by: Davidlohr Bueso <davidlohr.bueso@hp.com>
Signed-off-by: Rik van Riel <riel@redhat.com>
Reviewed-by: Chegu Vinod <chegu_vinod@hp.com>
Acked-by: Michel Lespinasse <walken@google.com>
Cc: Rik van Riel <riel@redhat.com>
Cc: Jason Low <jason.low2@hp.com>
Cc: Emmanuel Benisty <benisty.e@gmail.com>
Cc: Peter Hurley <peter@hurleysoftware.com>
Cc: Stanislav Kinsbursky <skinsbursky@parallels.com>
Tested-by: Sedat Dilek <sedat.dilek@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2013-05-01 10:15:29 +08:00
|
|
|
}
|
2005-04-17 06:20:36 +08:00
|
|
|
|
|
|
|
nsems = sma->sem_nsems;
|
|
|
|
|
|
|
|
err = -EACCES;
|
2011-03-24 07:43:24 +08:00
|
|
|
if (ipcperms(ns, &sma->sem_perm,
|
ipc,sem: do not hold ipc lock more than necessary
Instead of holding the ipc lock for permissions and security checks, among
others, only acquire it when necessary.
Some numbers....
1) With Rik's semop-multi.c microbenchmark we can see the following
results:
Baseline (3.9-rc1):
cpus 4, threads: 256, semaphores: 128, test duration: 30 secs
total operations: 151452270, ops/sec 5048409
+ 59.40% a.out [kernel.kallsyms] [k] _raw_spin_lock
+ 6.14% a.out [kernel.kallsyms] [k] sys_semtimedop
+ 3.84% a.out [kernel.kallsyms] [k] avc_has_perm_flags
+ 3.64% a.out [kernel.kallsyms] [k] __audit_syscall_exit
+ 2.06% a.out [kernel.kallsyms] [k] copy_user_enhanced_fast_string
+ 1.86% a.out [kernel.kallsyms] [k] ipc_lock
With this patchset:
cpus 4, threads: 256, semaphores: 128, test duration: 30 secs
total operations: 273156400, ops/sec 9105213
+ 18.54% a.out [kernel.kallsyms] [k] _raw_spin_lock
+ 11.72% a.out [kernel.kallsyms] [k] sys_semtimedop
+ 7.70% a.out [kernel.kallsyms] [k] ipc_has_perm.isra.21
+ 6.58% a.out [kernel.kallsyms] [k] avc_has_perm_flags
+ 6.54% a.out [kernel.kallsyms] [k] __audit_syscall_exit
+ 4.71% a.out [kernel.kallsyms] [k] ipc_obtain_object_check
2) While on an Oracle swingbench DSS (data mining) workload the
improvements are not as exciting as with Rik's benchmark, we can see
some positive numbers. For an 8 socket machine the following are the
percentages of %sys time incurred in the ipc lock:
Baseline (3.9-rc1):
100 swingbench users: 8,74%
400 swingbench users: 21,86%
800 swingbench users: 84,35%
With this patchset:
100 swingbench users: 8,11%
400 swingbench users: 19,93%
800 swingbench users: 77,69%
[riel@redhat.com: fix two locking bugs]
[sasha.levin@oracle.com: prevent releasing RCU read lock twice in semctl_main]
[akpm@linux-foundation.org: coding-style fixes]
Signed-off-by: Davidlohr Bueso <davidlohr.bueso@hp.com>
Signed-off-by: Rik van Riel <riel@redhat.com>
Reviewed-by: Chegu Vinod <chegu_vinod@hp.com>
Acked-by: Michel Lespinasse <walken@google.com>
Cc: Rik van Riel <riel@redhat.com>
Cc: Jason Low <jason.low2@hp.com>
Cc: Emmanuel Benisty <benisty.e@gmail.com>
Cc: Peter Hurley <peter@hurleysoftware.com>
Cc: Stanislav Kinsbursky <skinsbursky@parallels.com>
Tested-by: Sedat Dilek <sedat.dilek@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2013-05-01 10:15:29 +08:00
|
|
|
cmd == SETALL ? S_IWUGO : S_IRUGO)) {
|
|
|
|
rcu_read_unlock();
|
|
|
|
goto out_wakeup;
|
|
|
|
}
|
2005-04-17 06:20:36 +08:00
|
|
|
|
|
|
|
err = security_sem_semctl(sma, cmd);
|
ipc,sem: do not hold ipc lock more than necessary
Instead of holding the ipc lock for permissions and security checks, among
others, only acquire it when necessary.
Some numbers....
1) With Rik's semop-multi.c microbenchmark we can see the following
results:
Baseline (3.9-rc1):
cpus 4, threads: 256, semaphores: 128, test duration: 30 secs
total operations: 151452270, ops/sec 5048409
+ 59.40% a.out [kernel.kallsyms] [k] _raw_spin_lock
+ 6.14% a.out [kernel.kallsyms] [k] sys_semtimedop
+ 3.84% a.out [kernel.kallsyms] [k] avc_has_perm_flags
+ 3.64% a.out [kernel.kallsyms] [k] __audit_syscall_exit
+ 2.06% a.out [kernel.kallsyms] [k] copy_user_enhanced_fast_string
+ 1.86% a.out [kernel.kallsyms] [k] ipc_lock
With this patchset:
cpus 4, threads: 256, semaphores: 128, test duration: 30 secs
total operations: 273156400, ops/sec 9105213
+ 18.54% a.out [kernel.kallsyms] [k] _raw_spin_lock
+ 11.72% a.out [kernel.kallsyms] [k] sys_semtimedop
+ 7.70% a.out [kernel.kallsyms] [k] ipc_has_perm.isra.21
+ 6.58% a.out [kernel.kallsyms] [k] avc_has_perm_flags
+ 6.54% a.out [kernel.kallsyms] [k] __audit_syscall_exit
+ 4.71% a.out [kernel.kallsyms] [k] ipc_obtain_object_check
2) While on an Oracle swingbench DSS (data mining) workload the
improvements are not as exciting as with Rik's benchmark, we can see
some positive numbers. For an 8 socket machine the following are the
percentages of %sys time incurred in the ipc lock:
Baseline (3.9-rc1):
100 swingbench users: 8,74%
400 swingbench users: 21,86%
800 swingbench users: 84,35%
With this patchset:
100 swingbench users: 8,11%
400 swingbench users: 19,93%
800 swingbench users: 77,69%
[riel@redhat.com: fix two locking bugs]
[sasha.levin@oracle.com: prevent releasing RCU read lock twice in semctl_main]
[akpm@linux-foundation.org: coding-style fixes]
Signed-off-by: Davidlohr Bueso <davidlohr.bueso@hp.com>
Signed-off-by: Rik van Riel <riel@redhat.com>
Reviewed-by: Chegu Vinod <chegu_vinod@hp.com>
Acked-by: Michel Lespinasse <walken@google.com>
Cc: Rik van Riel <riel@redhat.com>
Cc: Jason Low <jason.low2@hp.com>
Cc: Emmanuel Benisty <benisty.e@gmail.com>
Cc: Peter Hurley <peter@hurleysoftware.com>
Cc: Stanislav Kinsbursky <skinsbursky@parallels.com>
Tested-by: Sedat Dilek <sedat.dilek@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2013-05-01 10:15:29 +08:00
|
|
|
if (err) {
|
|
|
|
rcu_read_unlock();
|
|
|
|
goto out_wakeup;
|
|
|
|
}
|
2005-04-17 06:20:36 +08:00
|
|
|
|
|
|
|
err = -EACCES;
|
|
|
|
switch (cmd) {
|
|
|
|
case GETALL:
|
|
|
|
{
|
2013-03-06 04:04:55 +08:00
|
|
|
ushort __user *array = p;
|
2005-04-17 06:20:36 +08:00
|
|
|
int i;
|
|
|
|
|
2013-05-03 07:30:49 +08:00
|
|
|
sem_lock(sma, NULL, -1);
|
2005-04-17 06:20:36 +08:00
|
|
|
if(nsems > SEMMSL_FAST) {
|
2013-05-03 07:30:49 +08:00
|
|
|
if (!ipc_rcu_getref(sma)) {
|
|
|
|
sem_unlock(sma, -1);
|
ipc: move rcu_read_unlock() out of sem_unlock() and into callers
The IPC locking is a mess, and sem_unlock() unlocks not only the
semaphore spinlock, it also drops the rcu read lock. Unlike sem_lock(),
which just gets the spin-lock, and expects the caller to get the rcu
read lock.
This all makes things very hard to follow, and it's very confusing when
you take the rcu read lock in one function, and then release it in
another. And it has caused actual bugs: the sem_obtain_lock() function
ended up dropping the RCU read lock twice in one error path, because it
first did the sem_unlock(), and then did a rcu_read_unlock() to match
the rcu_read_lock() it had done.
This is just a totally mindless "remove rcu_read_unlock() from
sem_unlock() and add it immediately after each caller" (except for the
aforementioned bug where we did too many rcu_read_unlock(), and in
find_alloc_undo() where we just got the rcu_read_lock() to correct for
the fact that sem_unlock would immediately drop it again).
We can (and should) clean things up further, but this fixes the bug with
the minimal amount of subtlety.
Reviewed-by: Davidlohr Bueso <davidlohr.bueso@hp.com>
Cc: Rik van Riel <riel@redhat.com>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2013-05-04 06:04:40 +08:00
|
|
|
rcu_read_unlock();
|
2013-05-03 07:30:49 +08:00
|
|
|
err = -EIDRM;
|
|
|
|
goto out_free;
|
|
|
|
}
|
|
|
|
sem_unlock(sma, -1);
|
ipc: move rcu_read_unlock() out of sem_unlock() and into callers
The IPC locking is a mess, and sem_unlock() unlocks not only the
semaphore spinlock, it also drops the rcu read lock. Unlike sem_lock(),
which just gets the spin-lock, and expects the caller to get the rcu
read lock.
This all makes things very hard to follow, and it's very confusing when
you take the rcu read lock in one function, and then release it in
another. And it has caused actual bugs: the sem_obtain_lock() function
ended up dropping the RCU read lock twice in one error path, because it
first did the sem_unlock(), and then did a rcu_read_unlock() to match
the rcu_read_lock() it had done.
This is just a totally mindless "remove rcu_read_unlock() from
sem_unlock() and add it immediately after each caller" (except for the
aforementioned bug where we did too many rcu_read_unlock(), and in
find_alloc_undo() where we just got the rcu_read_lock() to correct for
the fact that sem_unlock would immediately drop it again).
We can (and should) clean things up further, but this fixes the bug with
the minimal amount of subtlety.
Reviewed-by: Davidlohr Bueso <davidlohr.bueso@hp.com>
Cc: Rik van Riel <riel@redhat.com>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2013-05-04 06:04:40 +08:00
|
|
|
rcu_read_unlock();
|
2005-04-17 06:20:36 +08:00
|
|
|
sem_io = ipc_alloc(sizeof(ushort)*nsems);
|
|
|
|
if(sem_io == NULL) {
|
2008-04-29 16:00:46 +08:00
|
|
|
sem_putref(sma);
|
2005-04-17 06:20:36 +08:00
|
|
|
return -ENOMEM;
|
|
|
|
}
|
|
|
|
|
2013-05-05 01:13:40 +08:00
|
|
|
rcu_read_lock();
|
2008-04-29 16:00:46 +08:00
|
|
|
sem_lock_and_putref(sma);
|
2005-04-17 06:20:36 +08:00
|
|
|
if (sma->sem_perm.deleted) {
|
2013-05-01 10:15:44 +08:00
|
|
|
sem_unlock(sma, -1);
|
ipc: move rcu_read_unlock() out of sem_unlock() and into callers
The IPC locking is a mess, and sem_unlock() unlocks not only the
semaphore spinlock, it also drops the rcu read lock. Unlike sem_lock(),
which just gets the spin-lock, and expects the caller to get the rcu
read lock.
This all makes things very hard to follow, and it's very confusing when
you take the rcu read lock in one function, and then release it in
another. And it has caused actual bugs: the sem_obtain_lock() function
ended up dropping the RCU read lock twice in one error path, because it
first did the sem_unlock(), and then did a rcu_read_unlock() to match
the rcu_read_lock() it had done.
This is just a totally mindless "remove rcu_read_unlock() from
sem_unlock() and add it immediately after each caller" (except for the
aforementioned bug where we did too many rcu_read_unlock(), and in
find_alloc_undo() where we just got the rcu_read_lock() to correct for
the fact that sem_unlock would immediately drop it again).
We can (and should) clean things up further, but this fixes the bug with
the minimal amount of subtlety.
Reviewed-by: Davidlohr Bueso <davidlohr.bueso@hp.com>
Cc: Rik van Riel <riel@redhat.com>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2013-05-04 06:04:40 +08:00
|
|
|
rcu_read_unlock();
|
2005-04-17 06:20:36 +08:00
|
|
|
err = -EIDRM;
|
|
|
|
goto out_free;
|
|
|
|
}
|
2013-05-03 07:30:49 +08:00
|
|
|
}
|
2005-04-17 06:20:36 +08:00
|
|
|
for (i = 0; i < sma->sem_nsems; i++)
|
|
|
|
sem_io[i] = sma->sem_base[i].semval;
|
2013-05-01 10:15:44 +08:00
|
|
|
sem_unlock(sma, -1);
|
ipc: move rcu_read_unlock() out of sem_unlock() and into callers
The IPC locking is a mess, and sem_unlock() unlocks not only the
semaphore spinlock, it also drops the rcu read lock. Unlike sem_lock(),
which just gets the spin-lock, and expects the caller to get the rcu
read lock.
This all makes things very hard to follow, and it's very confusing when
you take the rcu read lock in one function, and then release it in
another. And it has caused actual bugs: the sem_obtain_lock() function
ended up dropping the RCU read lock twice in one error path, because it
first did the sem_unlock(), and then did a rcu_read_unlock() to match
the rcu_read_lock() it had done.
This is just a totally mindless "remove rcu_read_unlock() from
sem_unlock() and add it immediately after each caller" (except for the
aforementioned bug where we did too many rcu_read_unlock(), and in
find_alloc_undo() where we just got the rcu_read_lock() to correct for
the fact that sem_unlock would immediately drop it again).
We can (and should) clean things up further, but this fixes the bug with
the minimal amount of subtlety.
Reviewed-by: Davidlohr Bueso <davidlohr.bueso@hp.com>
Cc: Rik van Riel <riel@redhat.com>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2013-05-04 06:04:40 +08:00
|
|
|
rcu_read_unlock();
|
2005-04-17 06:20:36 +08:00
|
|
|
err = 0;
|
|
|
|
if(copy_to_user(array, sem_io, nsems*sizeof(ushort)))
|
|
|
|
err = -EFAULT;
|
|
|
|
goto out_free;
|
|
|
|
}
|
|
|
|
case SETALL:
|
|
|
|
{
|
|
|
|
int i;
|
|
|
|
struct sem_undo *un;
|
|
|
|
|
2013-05-01 10:15:44 +08:00
|
|
|
if (!ipc_rcu_getref(sma)) {
|
|
|
|
rcu_read_unlock();
|
|
|
|
return -EIDRM;
|
|
|
|
}
|
ipc,sem: do not hold ipc lock more than necessary
Instead of holding the ipc lock for permissions and security checks, among
others, only acquire it when necessary.
Some numbers....
1) With Rik's semop-multi.c microbenchmark we can see the following
results:
Baseline (3.9-rc1):
cpus 4, threads: 256, semaphores: 128, test duration: 30 secs
total operations: 151452270, ops/sec 5048409
+ 59.40% a.out [kernel.kallsyms] [k] _raw_spin_lock
+ 6.14% a.out [kernel.kallsyms] [k] sys_semtimedop
+ 3.84% a.out [kernel.kallsyms] [k] avc_has_perm_flags
+ 3.64% a.out [kernel.kallsyms] [k] __audit_syscall_exit
+ 2.06% a.out [kernel.kallsyms] [k] copy_user_enhanced_fast_string
+ 1.86% a.out [kernel.kallsyms] [k] ipc_lock
With this patchset:
cpus 4, threads: 256, semaphores: 128, test duration: 30 secs
total operations: 273156400, ops/sec 9105213
+ 18.54% a.out [kernel.kallsyms] [k] _raw_spin_lock
+ 11.72% a.out [kernel.kallsyms] [k] sys_semtimedop
+ 7.70% a.out [kernel.kallsyms] [k] ipc_has_perm.isra.21
+ 6.58% a.out [kernel.kallsyms] [k] avc_has_perm_flags
+ 6.54% a.out [kernel.kallsyms] [k] __audit_syscall_exit
+ 4.71% a.out [kernel.kallsyms] [k] ipc_obtain_object_check
2) While on an Oracle swingbench DSS (data mining) workload the
improvements are not as exciting as with Rik's benchmark, we can see
some positive numbers. For an 8 socket machine the following are the
percentages of %sys time incurred in the ipc lock:
Baseline (3.9-rc1):
100 swingbench users: 8,74%
400 swingbench users: 21,86%
800 swingbench users: 84,35%
With this patchset:
100 swingbench users: 8,11%
400 swingbench users: 19,93%
800 swingbench users: 77,69%
[riel@redhat.com: fix two locking bugs]
[sasha.levin@oracle.com: prevent releasing RCU read lock twice in semctl_main]
[akpm@linux-foundation.org: coding-style fixes]
Signed-off-by: Davidlohr Bueso <davidlohr.bueso@hp.com>
Signed-off-by: Rik van Riel <riel@redhat.com>
Reviewed-by: Chegu Vinod <chegu_vinod@hp.com>
Acked-by: Michel Lespinasse <walken@google.com>
Cc: Rik van Riel <riel@redhat.com>
Cc: Jason Low <jason.low2@hp.com>
Cc: Emmanuel Benisty <benisty.e@gmail.com>
Cc: Peter Hurley <peter@hurleysoftware.com>
Cc: Stanislav Kinsbursky <skinsbursky@parallels.com>
Tested-by: Sedat Dilek <sedat.dilek@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2013-05-01 10:15:29 +08:00
|
|
|
rcu_read_unlock();
|
2005-04-17 06:20:36 +08:00
|
|
|
|
|
|
|
if(nsems > SEMMSL_FAST) {
|
|
|
|
sem_io = ipc_alloc(sizeof(ushort)*nsems);
|
|
|
|
if(sem_io == NULL) {
|
2008-04-29 16:00:46 +08:00
|
|
|
sem_putref(sma);
|
2005-04-17 06:20:36 +08:00
|
|
|
return -ENOMEM;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2013-03-06 04:04:55 +08:00
|
|
|
if (copy_from_user (sem_io, p, nsems*sizeof(ushort))) {
|
2008-04-29 16:00:46 +08:00
|
|
|
sem_putref(sma);
|
2005-04-17 06:20:36 +08:00
|
|
|
err = -EFAULT;
|
|
|
|
goto out_free;
|
|
|
|
}
|
|
|
|
|
|
|
|
for (i = 0; i < nsems; i++) {
|
|
|
|
if (sem_io[i] > SEMVMX) {
|
2008-04-29 16:00:46 +08:00
|
|
|
sem_putref(sma);
|
2005-04-17 06:20:36 +08:00
|
|
|
err = -ERANGE;
|
|
|
|
goto out_free;
|
|
|
|
}
|
|
|
|
}
|
2013-05-05 01:13:40 +08:00
|
|
|
rcu_read_lock();
|
2008-04-29 16:00:46 +08:00
|
|
|
sem_lock_and_putref(sma);
|
2005-04-17 06:20:36 +08:00
|
|
|
if (sma->sem_perm.deleted) {
|
2013-05-01 10:15:44 +08:00
|
|
|
sem_unlock(sma, -1);
|
ipc: move rcu_read_unlock() out of sem_unlock() and into callers
The IPC locking is a mess, and sem_unlock() unlocks not only the
semaphore spinlock, it also drops the rcu read lock. Unlike sem_lock(),
which just gets the spin-lock, and expects the caller to get the rcu
read lock.
This all makes things very hard to follow, and it's very confusing when
you take the rcu read lock in one function, and then release it in
another. And it has caused actual bugs: the sem_obtain_lock() function
ended up dropping the RCU read lock twice in one error path, because it
first did the sem_unlock(), and then did a rcu_read_unlock() to match
the rcu_read_lock() it had done.
This is just a totally mindless "remove rcu_read_unlock() from
sem_unlock() and add it immediately after each caller" (except for the
aforementioned bug where we did too many rcu_read_unlock(), and in
find_alloc_undo() where we just got the rcu_read_lock() to correct for
the fact that sem_unlock would immediately drop it again).
We can (and should) clean things up further, but this fixes the bug with
the minimal amount of subtlety.
Reviewed-by: Davidlohr Bueso <davidlohr.bueso@hp.com>
Cc: Rik van Riel <riel@redhat.com>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2013-05-04 06:04:40 +08:00
|
|
|
rcu_read_unlock();
|
2005-04-17 06:20:36 +08:00
|
|
|
err = -EIDRM;
|
|
|
|
goto out_free;
|
|
|
|
}
|
|
|
|
|
|
|
|
for (i = 0; i < nsems; i++)
|
|
|
|
sma->sem_base[i].semval = sem_io[i];
|
2008-07-25 16:48:04 +08:00
|
|
|
|
|
|
|
assert_spin_locked(&sma->sem_perm.lock);
|
|
|
|
list_for_each_entry(un, &sma->list_id, list_id) {
|
2005-04-17 06:20:36 +08:00
|
|
|
for (i = 0; i < nsems; i++)
|
|
|
|
un->semadj[i] = 0;
|
2008-07-25 16:48:04 +08:00
|
|
|
}
|
2005-04-17 06:20:36 +08:00
|
|
|
sma->sem_ctime = get_seconds();
|
|
|
|
/* maybe some queued-up processes were waiting for this */
|
2010-05-27 05:43:41 +08:00
|
|
|
do_smart_update(sma, NULL, 0, 0, &tasks);
|
2005-04-17 06:20:36 +08:00
|
|
|
err = 0;
|
|
|
|
goto out_unlock;
|
|
|
|
}
|
2013-03-06 04:04:55 +08:00
|
|
|
/* GETVAL, GETPID, GETNCTN, GETZCNT: fall-through */
|
2005-04-17 06:20:36 +08:00
|
|
|
}
|
|
|
|
err = -EINVAL;
|
ipc,sem: do not hold ipc lock more than necessary
Instead of holding the ipc lock for permissions and security checks, among
others, only acquire it when necessary.
Some numbers....
1) With Rik's semop-multi.c microbenchmark we can see the following
results:
Baseline (3.9-rc1):
cpus 4, threads: 256, semaphores: 128, test duration: 30 secs
total operations: 151452270, ops/sec 5048409
+ 59.40% a.out [kernel.kallsyms] [k] _raw_spin_lock
+ 6.14% a.out [kernel.kallsyms] [k] sys_semtimedop
+ 3.84% a.out [kernel.kallsyms] [k] avc_has_perm_flags
+ 3.64% a.out [kernel.kallsyms] [k] __audit_syscall_exit
+ 2.06% a.out [kernel.kallsyms] [k] copy_user_enhanced_fast_string
+ 1.86% a.out [kernel.kallsyms] [k] ipc_lock
With this patchset:
cpus 4, threads: 256, semaphores: 128, test duration: 30 secs
total operations: 273156400, ops/sec 9105213
+ 18.54% a.out [kernel.kallsyms] [k] _raw_spin_lock
+ 11.72% a.out [kernel.kallsyms] [k] sys_semtimedop
+ 7.70% a.out [kernel.kallsyms] [k] ipc_has_perm.isra.21
+ 6.58% a.out [kernel.kallsyms] [k] avc_has_perm_flags
+ 6.54% a.out [kernel.kallsyms] [k] __audit_syscall_exit
+ 4.71% a.out [kernel.kallsyms] [k] ipc_obtain_object_check
2) While on an Oracle swingbench DSS (data mining) workload the
improvements are not as exciting as with Rik's benchmark, we can see
some positive numbers. For an 8 socket machine the following are the
percentages of %sys time incurred in the ipc lock:
Baseline (3.9-rc1):
100 swingbench users: 8,74%
400 swingbench users: 21,86%
800 swingbench users: 84,35%
With this patchset:
100 swingbench users: 8,11%
400 swingbench users: 19,93%
800 swingbench users: 77,69%
[riel@redhat.com: fix two locking bugs]
[sasha.levin@oracle.com: prevent releasing RCU read lock twice in semctl_main]
[akpm@linux-foundation.org: coding-style fixes]
Signed-off-by: Davidlohr Bueso <davidlohr.bueso@hp.com>
Signed-off-by: Rik van Riel <riel@redhat.com>
Reviewed-by: Chegu Vinod <chegu_vinod@hp.com>
Acked-by: Michel Lespinasse <walken@google.com>
Cc: Rik van Riel <riel@redhat.com>
Cc: Jason Low <jason.low2@hp.com>
Cc: Emmanuel Benisty <benisty.e@gmail.com>
Cc: Peter Hurley <peter@hurleysoftware.com>
Cc: Stanislav Kinsbursky <skinsbursky@parallels.com>
Tested-by: Sedat Dilek <sedat.dilek@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2013-05-01 10:15:29 +08:00
|
|
|
if (semnum < 0 || semnum >= nsems) {
|
|
|
|
rcu_read_unlock();
|
|
|
|
goto out_wakeup;
|
|
|
|
}
|
2005-04-17 06:20:36 +08:00
|
|
|
|
2013-05-01 10:15:44 +08:00
|
|
|
sem_lock(sma, NULL, -1);
|
2005-04-17 06:20:36 +08:00
|
|
|
curr = &sma->sem_base[semnum];
|
|
|
|
|
|
|
|
switch (cmd) {
|
|
|
|
case GETVAL:
|
|
|
|
err = curr->semval;
|
|
|
|
goto out_unlock;
|
|
|
|
case GETPID:
|
|
|
|
err = curr->sempid;
|
|
|
|
goto out_unlock;
|
|
|
|
case GETNCNT:
|
|
|
|
err = count_semncnt(sma,semnum);
|
|
|
|
goto out_unlock;
|
|
|
|
case GETZCNT:
|
|
|
|
err = count_semzcnt(sma,semnum);
|
|
|
|
goto out_unlock;
|
|
|
|
}
|
ipc,sem: do not hold ipc lock more than necessary
Instead of holding the ipc lock for permissions and security checks, among
others, only acquire it when necessary.
Some numbers....
1) With Rik's semop-multi.c microbenchmark we can see the following
results:
Baseline (3.9-rc1):
cpus 4, threads: 256, semaphores: 128, test duration: 30 secs
total operations: 151452270, ops/sec 5048409
+ 59.40% a.out [kernel.kallsyms] [k] _raw_spin_lock
+ 6.14% a.out [kernel.kallsyms] [k] sys_semtimedop
+ 3.84% a.out [kernel.kallsyms] [k] avc_has_perm_flags
+ 3.64% a.out [kernel.kallsyms] [k] __audit_syscall_exit
+ 2.06% a.out [kernel.kallsyms] [k] copy_user_enhanced_fast_string
+ 1.86% a.out [kernel.kallsyms] [k] ipc_lock
With this patchset:
cpus 4, threads: 256, semaphores: 128, test duration: 30 secs
total operations: 273156400, ops/sec 9105213
+ 18.54% a.out [kernel.kallsyms] [k] _raw_spin_lock
+ 11.72% a.out [kernel.kallsyms] [k] sys_semtimedop
+ 7.70% a.out [kernel.kallsyms] [k] ipc_has_perm.isra.21
+ 6.58% a.out [kernel.kallsyms] [k] avc_has_perm_flags
+ 6.54% a.out [kernel.kallsyms] [k] __audit_syscall_exit
+ 4.71% a.out [kernel.kallsyms] [k] ipc_obtain_object_check
2) While on an Oracle swingbench DSS (data mining) workload the
improvements are not as exciting as with Rik's benchmark, we can see
some positive numbers. For an 8 socket machine the following are the
percentages of %sys time incurred in the ipc lock:
Baseline (3.9-rc1):
100 swingbench users: 8,74%
400 swingbench users: 21,86%
800 swingbench users: 84,35%
With this patchset:
100 swingbench users: 8,11%
400 swingbench users: 19,93%
800 swingbench users: 77,69%
[riel@redhat.com: fix two locking bugs]
[sasha.levin@oracle.com: prevent releasing RCU read lock twice in semctl_main]
[akpm@linux-foundation.org: coding-style fixes]
Signed-off-by: Davidlohr Bueso <davidlohr.bueso@hp.com>
Signed-off-by: Rik van Riel <riel@redhat.com>
Reviewed-by: Chegu Vinod <chegu_vinod@hp.com>
Acked-by: Michel Lespinasse <walken@google.com>
Cc: Rik van Riel <riel@redhat.com>
Cc: Jason Low <jason.low2@hp.com>
Cc: Emmanuel Benisty <benisty.e@gmail.com>
Cc: Peter Hurley <peter@hurleysoftware.com>
Cc: Stanislav Kinsbursky <skinsbursky@parallels.com>
Tested-by: Sedat Dilek <sedat.dilek@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2013-05-01 10:15:29 +08:00
|
|
|
|
2005-04-17 06:20:36 +08:00
|
|
|
out_unlock:
|
2013-05-01 10:15:44 +08:00
|
|
|
sem_unlock(sma, -1);
|
ipc: move rcu_read_unlock() out of sem_unlock() and into callers
The IPC locking is a mess, and sem_unlock() unlocks not only the
semaphore spinlock, it also drops the rcu read lock. Unlike sem_lock(),
which just gets the spin-lock, and expects the caller to get the rcu
read lock.
This all makes things very hard to follow, and it's very confusing when
you take the rcu read lock in one function, and then release it in
another. And it has caused actual bugs: the sem_obtain_lock() function
ended up dropping the RCU read lock twice in one error path, because it
first did the sem_unlock(), and then did a rcu_read_unlock() to match
the rcu_read_lock() it had done.
This is just a totally mindless "remove rcu_read_unlock() from
sem_unlock() and add it immediately after each caller" (except for the
aforementioned bug where we did too many rcu_read_unlock(), and in
find_alloc_undo() where we just got the rcu_read_lock() to correct for
the fact that sem_unlock would immediately drop it again).
We can (and should) clean things up further, but this fixes the bug with
the minimal amount of subtlety.
Reviewed-by: Davidlohr Bueso <davidlohr.bueso@hp.com>
Cc: Rik van Riel <riel@redhat.com>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2013-05-04 06:04:40 +08:00
|
|
|
rcu_read_unlock();
|
ipc,sem: do not hold ipc lock more than necessary
Instead of holding the ipc lock for permissions and security checks, among
others, only acquire it when necessary.
Some numbers....
1) With Rik's semop-multi.c microbenchmark we can see the following
results:
Baseline (3.9-rc1):
cpus 4, threads: 256, semaphores: 128, test duration: 30 secs
total operations: 151452270, ops/sec 5048409
+ 59.40% a.out [kernel.kallsyms] [k] _raw_spin_lock
+ 6.14% a.out [kernel.kallsyms] [k] sys_semtimedop
+ 3.84% a.out [kernel.kallsyms] [k] avc_has_perm_flags
+ 3.64% a.out [kernel.kallsyms] [k] __audit_syscall_exit
+ 2.06% a.out [kernel.kallsyms] [k] copy_user_enhanced_fast_string
+ 1.86% a.out [kernel.kallsyms] [k] ipc_lock
With this patchset:
cpus 4, threads: 256, semaphores: 128, test duration: 30 secs
total operations: 273156400, ops/sec 9105213
+ 18.54% a.out [kernel.kallsyms] [k] _raw_spin_lock
+ 11.72% a.out [kernel.kallsyms] [k] sys_semtimedop
+ 7.70% a.out [kernel.kallsyms] [k] ipc_has_perm.isra.21
+ 6.58% a.out [kernel.kallsyms] [k] avc_has_perm_flags
+ 6.54% a.out [kernel.kallsyms] [k] __audit_syscall_exit
+ 4.71% a.out [kernel.kallsyms] [k] ipc_obtain_object_check
2) While on an Oracle swingbench DSS (data mining) workload the
improvements are not as exciting as with Rik's benchmark, we can see
some positive numbers. For an 8 socket machine the following are the
percentages of %sys time incurred in the ipc lock:
Baseline (3.9-rc1):
100 swingbench users: 8,74%
400 swingbench users: 21,86%
800 swingbench users: 84,35%
With this patchset:
100 swingbench users: 8,11%
400 swingbench users: 19,93%
800 swingbench users: 77,69%
[riel@redhat.com: fix two locking bugs]
[sasha.levin@oracle.com: prevent releasing RCU read lock twice in semctl_main]
[akpm@linux-foundation.org: coding-style fixes]
Signed-off-by: Davidlohr Bueso <davidlohr.bueso@hp.com>
Signed-off-by: Rik van Riel <riel@redhat.com>
Reviewed-by: Chegu Vinod <chegu_vinod@hp.com>
Acked-by: Michel Lespinasse <walken@google.com>
Cc: Rik van Riel <riel@redhat.com>
Cc: Jason Low <jason.low2@hp.com>
Cc: Emmanuel Benisty <benisty.e@gmail.com>
Cc: Peter Hurley <peter@hurleysoftware.com>
Cc: Stanislav Kinsbursky <skinsbursky@parallels.com>
Tested-by: Sedat Dilek <sedat.dilek@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2013-05-01 10:15:29 +08:00
|
|
|
out_wakeup:
|
2010-05-27 05:43:41 +08:00
|
|
|
wake_up_sem_queue_do(&tasks);
|
2005-04-17 06:20:36 +08:00
|
|
|
out_free:
|
|
|
|
if(sem_io != fast_sem_io)
|
|
|
|
ipc_free(sem_io, sizeof(ushort)*nsems);
|
|
|
|
return err;
|
|
|
|
}
|
|
|
|
|
2008-04-29 16:00:50 +08:00
|
|
|
static inline unsigned long
|
|
|
|
copy_semid_from_user(struct semid64_ds *out, void __user *buf, int version)
|
2005-04-17 06:20:36 +08:00
|
|
|
{
|
|
|
|
switch(version) {
|
|
|
|
case IPC_64:
|
2008-04-29 16:00:50 +08:00
|
|
|
if (copy_from_user(out, buf, sizeof(*out)))
|
2005-04-17 06:20:36 +08:00
|
|
|
return -EFAULT;
|
|
|
|
return 0;
|
|
|
|
case IPC_OLD:
|
|
|
|
{
|
|
|
|
struct semid_ds tbuf_old;
|
|
|
|
|
|
|
|
if(copy_from_user(&tbuf_old, buf, sizeof(tbuf_old)))
|
|
|
|
return -EFAULT;
|
|
|
|
|
2008-04-29 16:00:50 +08:00
|
|
|
out->sem_perm.uid = tbuf_old.sem_perm.uid;
|
|
|
|
out->sem_perm.gid = tbuf_old.sem_perm.gid;
|
|
|
|
out->sem_perm.mode = tbuf_old.sem_perm.mode;
|
2005-04-17 06:20:36 +08:00
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
default:
|
|
|
|
return -EINVAL;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2008-04-29 16:00:49 +08:00
|
|
|
/*
|
|
|
|
* This function handles some semctl commands which require the rw_mutex
|
|
|
|
* to be held in write mode.
|
|
|
|
* NOTE: no locks must be held, the rw_mutex is taken inside this function.
|
|
|
|
*/
|
2008-04-29 16:00:49 +08:00
|
|
|
static int semctl_down(struct ipc_namespace *ns, int semid,
|
2013-03-06 04:04:55 +08:00
|
|
|
int cmd, int version, void __user *p)
|
2005-04-17 06:20:36 +08:00
|
|
|
{
|
|
|
|
struct sem_array *sma;
|
|
|
|
int err;
|
2008-04-29 16:00:50 +08:00
|
|
|
struct semid64_ds semid64;
|
2005-04-17 06:20:36 +08:00
|
|
|
struct kern_ipc_perm *ipcp;
|
|
|
|
|
|
|
|
if(cmd == IPC_SET) {
|
2013-03-06 04:04:55 +08:00
|
|
|
if (copy_semid_from_user(&semid64, p, version))
|
2005-04-17 06:20:36 +08:00
|
|
|
return -EFAULT;
|
|
|
|
}
|
2006-04-03 05:07:33 +08:00
|
|
|
|
ipc,sem: do not hold ipc lock more than necessary
Instead of holding the ipc lock for permissions and security checks, among
others, only acquire it when necessary.
Some numbers....
1) With Rik's semop-multi.c microbenchmark we can see the following
results:
Baseline (3.9-rc1):
cpus 4, threads: 256, semaphores: 128, test duration: 30 secs
total operations: 151452270, ops/sec 5048409
+ 59.40% a.out [kernel.kallsyms] [k] _raw_spin_lock
+ 6.14% a.out [kernel.kallsyms] [k] sys_semtimedop
+ 3.84% a.out [kernel.kallsyms] [k] avc_has_perm_flags
+ 3.64% a.out [kernel.kallsyms] [k] __audit_syscall_exit
+ 2.06% a.out [kernel.kallsyms] [k] copy_user_enhanced_fast_string
+ 1.86% a.out [kernel.kallsyms] [k] ipc_lock
With this patchset:
cpus 4, threads: 256, semaphores: 128, test duration: 30 secs
total operations: 273156400, ops/sec 9105213
+ 18.54% a.out [kernel.kallsyms] [k] _raw_spin_lock
+ 11.72% a.out [kernel.kallsyms] [k] sys_semtimedop
+ 7.70% a.out [kernel.kallsyms] [k] ipc_has_perm.isra.21
+ 6.58% a.out [kernel.kallsyms] [k] avc_has_perm_flags
+ 6.54% a.out [kernel.kallsyms] [k] __audit_syscall_exit
+ 4.71% a.out [kernel.kallsyms] [k] ipc_obtain_object_check
2) While on an Oracle swingbench DSS (data mining) workload the
improvements are not as exciting as with Rik's benchmark, we can see
some positive numbers. For an 8 socket machine the following are the
percentages of %sys time incurred in the ipc lock:
Baseline (3.9-rc1):
100 swingbench users: 8,74%
400 swingbench users: 21,86%
800 swingbench users: 84,35%
With this patchset:
100 swingbench users: 8,11%
400 swingbench users: 19,93%
800 swingbench users: 77,69%
[riel@redhat.com: fix two locking bugs]
[sasha.levin@oracle.com: prevent releasing RCU read lock twice in semctl_main]
[akpm@linux-foundation.org: coding-style fixes]
Signed-off-by: Davidlohr Bueso <davidlohr.bueso@hp.com>
Signed-off-by: Rik van Riel <riel@redhat.com>
Reviewed-by: Chegu Vinod <chegu_vinod@hp.com>
Acked-by: Michel Lespinasse <walken@google.com>
Cc: Rik van Riel <riel@redhat.com>
Cc: Jason Low <jason.low2@hp.com>
Cc: Emmanuel Benisty <benisty.e@gmail.com>
Cc: Peter Hurley <peter@hurleysoftware.com>
Cc: Stanislav Kinsbursky <skinsbursky@parallels.com>
Tested-by: Sedat Dilek <sedat.dilek@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2013-05-01 10:15:29 +08:00
|
|
|
ipcp = ipcctl_pre_down_nolock(ns, &sem_ids(ns), semid, cmd,
|
|
|
|
&semid64.sem_perm, 0);
|
2008-04-29 16:00:54 +08:00
|
|
|
if (IS_ERR(ipcp))
|
|
|
|
return PTR_ERR(ipcp);
|
2006-04-03 05:07:33 +08:00
|
|
|
|
2008-04-29 16:00:54 +08:00
|
|
|
sma = container_of(ipcp, struct sem_array, sem_perm);
|
2005-04-17 06:20:36 +08:00
|
|
|
|
|
|
|
err = security_sem_semctl(sma, cmd);
|
ipc,sem: do not hold ipc lock more than necessary
Instead of holding the ipc lock for permissions and security checks, among
others, only acquire it when necessary.
Some numbers....
1) With Rik's semop-multi.c microbenchmark we can see the following
results:
Baseline (3.9-rc1):
cpus 4, threads: 256, semaphores: 128, test duration: 30 secs
total operations: 151452270, ops/sec 5048409
+ 59.40% a.out [kernel.kallsyms] [k] _raw_spin_lock
+ 6.14% a.out [kernel.kallsyms] [k] sys_semtimedop
+ 3.84% a.out [kernel.kallsyms] [k] avc_has_perm_flags
+ 3.64% a.out [kernel.kallsyms] [k] __audit_syscall_exit
+ 2.06% a.out [kernel.kallsyms] [k] copy_user_enhanced_fast_string
+ 1.86% a.out [kernel.kallsyms] [k] ipc_lock
With this patchset:
cpus 4, threads: 256, semaphores: 128, test duration: 30 secs
total operations: 273156400, ops/sec 9105213
+ 18.54% a.out [kernel.kallsyms] [k] _raw_spin_lock
+ 11.72% a.out [kernel.kallsyms] [k] sys_semtimedop
+ 7.70% a.out [kernel.kallsyms] [k] ipc_has_perm.isra.21
+ 6.58% a.out [kernel.kallsyms] [k] avc_has_perm_flags
+ 6.54% a.out [kernel.kallsyms] [k] __audit_syscall_exit
+ 4.71% a.out [kernel.kallsyms] [k] ipc_obtain_object_check
2) While on an Oracle swingbench DSS (data mining) workload the
improvements are not as exciting as with Rik's benchmark, we can see
some positive numbers. For an 8 socket machine the following are the
percentages of %sys time incurred in the ipc lock:
Baseline (3.9-rc1):
100 swingbench users: 8,74%
400 swingbench users: 21,86%
800 swingbench users: 84,35%
With this patchset:
100 swingbench users: 8,11%
400 swingbench users: 19,93%
800 swingbench users: 77,69%
[riel@redhat.com: fix two locking bugs]
[sasha.levin@oracle.com: prevent releasing RCU read lock twice in semctl_main]
[akpm@linux-foundation.org: coding-style fixes]
Signed-off-by: Davidlohr Bueso <davidlohr.bueso@hp.com>
Signed-off-by: Rik van Riel <riel@redhat.com>
Reviewed-by: Chegu Vinod <chegu_vinod@hp.com>
Acked-by: Michel Lespinasse <walken@google.com>
Cc: Rik van Riel <riel@redhat.com>
Cc: Jason Low <jason.low2@hp.com>
Cc: Emmanuel Benisty <benisty.e@gmail.com>
Cc: Peter Hurley <peter@hurleysoftware.com>
Cc: Stanislav Kinsbursky <skinsbursky@parallels.com>
Tested-by: Sedat Dilek <sedat.dilek@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2013-05-01 10:15:29 +08:00
|
|
|
if (err) {
|
|
|
|
rcu_read_unlock();
|
2013-05-05 01:25:11 +08:00
|
|
|
goto out_up;
|
ipc,sem: do not hold ipc lock more than necessary
Instead of holding the ipc lock for permissions and security checks, among
others, only acquire it when necessary.
Some numbers....
1) With Rik's semop-multi.c microbenchmark we can see the following
results:
Baseline (3.9-rc1):
cpus 4, threads: 256, semaphores: 128, test duration: 30 secs
total operations: 151452270, ops/sec 5048409
+ 59.40% a.out [kernel.kallsyms] [k] _raw_spin_lock
+ 6.14% a.out [kernel.kallsyms] [k] sys_semtimedop
+ 3.84% a.out [kernel.kallsyms] [k] avc_has_perm_flags
+ 3.64% a.out [kernel.kallsyms] [k] __audit_syscall_exit
+ 2.06% a.out [kernel.kallsyms] [k] copy_user_enhanced_fast_string
+ 1.86% a.out [kernel.kallsyms] [k] ipc_lock
With this patchset:
cpus 4, threads: 256, semaphores: 128, test duration: 30 secs
total operations: 273156400, ops/sec 9105213
+ 18.54% a.out [kernel.kallsyms] [k] _raw_spin_lock
+ 11.72% a.out [kernel.kallsyms] [k] sys_semtimedop
+ 7.70% a.out [kernel.kallsyms] [k] ipc_has_perm.isra.21
+ 6.58% a.out [kernel.kallsyms] [k] avc_has_perm_flags
+ 6.54% a.out [kernel.kallsyms] [k] __audit_syscall_exit
+ 4.71% a.out [kernel.kallsyms] [k] ipc_obtain_object_check
2) While on an Oracle swingbench DSS (data mining) workload the
improvements are not as exciting as with Rik's benchmark, we can see
some positive numbers. For an 8 socket machine the following are the
percentages of %sys time incurred in the ipc lock:
Baseline (3.9-rc1):
100 swingbench users: 8,74%
400 swingbench users: 21,86%
800 swingbench users: 84,35%
With this patchset:
100 swingbench users: 8,11%
400 swingbench users: 19,93%
800 swingbench users: 77,69%
[riel@redhat.com: fix two locking bugs]
[sasha.levin@oracle.com: prevent releasing RCU read lock twice in semctl_main]
[akpm@linux-foundation.org: coding-style fixes]
Signed-off-by: Davidlohr Bueso <davidlohr.bueso@hp.com>
Signed-off-by: Rik van Riel <riel@redhat.com>
Reviewed-by: Chegu Vinod <chegu_vinod@hp.com>
Acked-by: Michel Lespinasse <walken@google.com>
Cc: Rik van Riel <riel@redhat.com>
Cc: Jason Low <jason.low2@hp.com>
Cc: Emmanuel Benisty <benisty.e@gmail.com>
Cc: Peter Hurley <peter@hurleysoftware.com>
Cc: Stanislav Kinsbursky <skinsbursky@parallels.com>
Tested-by: Sedat Dilek <sedat.dilek@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2013-05-01 10:15:29 +08:00
|
|
|
}
|
2005-04-17 06:20:36 +08:00
|
|
|
|
|
|
|
switch(cmd){
|
|
|
|
case IPC_RMID:
|
2013-05-01 10:15:44 +08:00
|
|
|
sem_lock(sma, NULL, -1);
|
2008-02-08 20:18:57 +08:00
|
|
|
freeary(ns, ipcp);
|
2008-04-29 16:00:49 +08:00
|
|
|
goto out_up;
|
2005-04-17 06:20:36 +08:00
|
|
|
case IPC_SET:
|
2013-05-01 10:15:44 +08:00
|
|
|
sem_lock(sma, NULL, -1);
|
2012-02-08 08:54:11 +08:00
|
|
|
err = ipc_update_perm(&semid64.sem_perm, ipcp);
|
|
|
|
if (err)
|
|
|
|
goto out_unlock;
|
2005-04-17 06:20:36 +08:00
|
|
|
sma->sem_ctime = get_seconds();
|
|
|
|
break;
|
|
|
|
default:
|
ipc,sem: do not hold ipc lock more than necessary
Instead of holding the ipc lock for permissions and security checks, among
others, only acquire it when necessary.
Some numbers....
1) With Rik's semop-multi.c microbenchmark we can see the following
results:
Baseline (3.9-rc1):
cpus 4, threads: 256, semaphores: 128, test duration: 30 secs
total operations: 151452270, ops/sec 5048409
+ 59.40% a.out [kernel.kallsyms] [k] _raw_spin_lock
+ 6.14% a.out [kernel.kallsyms] [k] sys_semtimedop
+ 3.84% a.out [kernel.kallsyms] [k] avc_has_perm_flags
+ 3.64% a.out [kernel.kallsyms] [k] __audit_syscall_exit
+ 2.06% a.out [kernel.kallsyms] [k] copy_user_enhanced_fast_string
+ 1.86% a.out [kernel.kallsyms] [k] ipc_lock
With this patchset:
cpus 4, threads: 256, semaphores: 128, test duration: 30 secs
total operations: 273156400, ops/sec 9105213
+ 18.54% a.out [kernel.kallsyms] [k] _raw_spin_lock
+ 11.72% a.out [kernel.kallsyms] [k] sys_semtimedop
+ 7.70% a.out [kernel.kallsyms] [k] ipc_has_perm.isra.21
+ 6.58% a.out [kernel.kallsyms] [k] avc_has_perm_flags
+ 6.54% a.out [kernel.kallsyms] [k] __audit_syscall_exit
+ 4.71% a.out [kernel.kallsyms] [k] ipc_obtain_object_check
2) While on an Oracle swingbench DSS (data mining) workload the
improvements are not as exciting as with Rik's benchmark, we can see
some positive numbers. For an 8 socket machine the following are the
percentages of %sys time incurred in the ipc lock:
Baseline (3.9-rc1):
100 swingbench users: 8,74%
400 swingbench users: 21,86%
800 swingbench users: 84,35%
With this patchset:
100 swingbench users: 8,11%
400 swingbench users: 19,93%
800 swingbench users: 77,69%
[riel@redhat.com: fix two locking bugs]
[sasha.levin@oracle.com: prevent releasing RCU read lock twice in semctl_main]
[akpm@linux-foundation.org: coding-style fixes]
Signed-off-by: Davidlohr Bueso <davidlohr.bueso@hp.com>
Signed-off-by: Rik van Riel <riel@redhat.com>
Reviewed-by: Chegu Vinod <chegu_vinod@hp.com>
Acked-by: Michel Lespinasse <walken@google.com>
Cc: Rik van Riel <riel@redhat.com>
Cc: Jason Low <jason.low2@hp.com>
Cc: Emmanuel Benisty <benisty.e@gmail.com>
Cc: Peter Hurley <peter@hurleysoftware.com>
Cc: Stanislav Kinsbursky <skinsbursky@parallels.com>
Tested-by: Sedat Dilek <sedat.dilek@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2013-05-01 10:15:29 +08:00
|
|
|
rcu_read_unlock();
|
2005-04-17 06:20:36 +08:00
|
|
|
err = -EINVAL;
|
ipc,sem: do not hold ipc lock more than necessary
Instead of holding the ipc lock for permissions and security checks, among
others, only acquire it when necessary.
Some numbers....
1) With Rik's semop-multi.c microbenchmark we can see the following
results:
Baseline (3.9-rc1):
cpus 4, threads: 256, semaphores: 128, test duration: 30 secs
total operations: 151452270, ops/sec 5048409
+ 59.40% a.out [kernel.kallsyms] [k] _raw_spin_lock
+ 6.14% a.out [kernel.kallsyms] [k] sys_semtimedop
+ 3.84% a.out [kernel.kallsyms] [k] avc_has_perm_flags
+ 3.64% a.out [kernel.kallsyms] [k] __audit_syscall_exit
+ 2.06% a.out [kernel.kallsyms] [k] copy_user_enhanced_fast_string
+ 1.86% a.out [kernel.kallsyms] [k] ipc_lock
With this patchset:
cpus 4, threads: 256, semaphores: 128, test duration: 30 secs
total operations: 273156400, ops/sec 9105213
+ 18.54% a.out [kernel.kallsyms] [k] _raw_spin_lock
+ 11.72% a.out [kernel.kallsyms] [k] sys_semtimedop
+ 7.70% a.out [kernel.kallsyms] [k] ipc_has_perm.isra.21
+ 6.58% a.out [kernel.kallsyms] [k] avc_has_perm_flags
+ 6.54% a.out [kernel.kallsyms] [k] __audit_syscall_exit
+ 4.71% a.out [kernel.kallsyms] [k] ipc_obtain_object_check
2) While on an Oracle swingbench DSS (data mining) workload the
improvements are not as exciting as with Rik's benchmark, we can see
some positive numbers. For an 8 socket machine the following are the
percentages of %sys time incurred in the ipc lock:
Baseline (3.9-rc1):
100 swingbench users: 8,74%
400 swingbench users: 21,86%
800 swingbench users: 84,35%
With this patchset:
100 swingbench users: 8,11%
400 swingbench users: 19,93%
800 swingbench users: 77,69%
[riel@redhat.com: fix two locking bugs]
[sasha.levin@oracle.com: prevent releasing RCU read lock twice in semctl_main]
[akpm@linux-foundation.org: coding-style fixes]
Signed-off-by: Davidlohr Bueso <davidlohr.bueso@hp.com>
Signed-off-by: Rik van Riel <riel@redhat.com>
Reviewed-by: Chegu Vinod <chegu_vinod@hp.com>
Acked-by: Michel Lespinasse <walken@google.com>
Cc: Rik van Riel <riel@redhat.com>
Cc: Jason Low <jason.low2@hp.com>
Cc: Emmanuel Benisty <benisty.e@gmail.com>
Cc: Peter Hurley <peter@hurleysoftware.com>
Cc: Stanislav Kinsbursky <skinsbursky@parallels.com>
Tested-by: Sedat Dilek <sedat.dilek@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2013-05-01 10:15:29 +08:00
|
|
|
goto out_up;
|
2005-04-17 06:20:36 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
out_unlock:
|
2013-05-01 10:15:44 +08:00
|
|
|
sem_unlock(sma, -1);
|
ipc: move rcu_read_unlock() out of sem_unlock() and into callers
The IPC locking is a mess, and sem_unlock() unlocks not only the
semaphore spinlock, it also drops the rcu read lock. Unlike sem_lock(),
which just gets the spin-lock, and expects the caller to get the rcu
read lock.
This all makes things very hard to follow, and it's very confusing when
you take the rcu read lock in one function, and then release it in
another. And it has caused actual bugs: the sem_obtain_lock() function
ended up dropping the RCU read lock twice in one error path, because it
first did the sem_unlock(), and then did a rcu_read_unlock() to match
the rcu_read_lock() it had done.
This is just a totally mindless "remove rcu_read_unlock() from
sem_unlock() and add it immediately after each caller" (except for the
aforementioned bug where we did too many rcu_read_unlock(), and in
find_alloc_undo() where we just got the rcu_read_lock() to correct for
the fact that sem_unlock would immediately drop it again).
We can (and should) clean things up further, but this fixes the bug with
the minimal amount of subtlety.
Reviewed-by: Davidlohr Bueso <davidlohr.bueso@hp.com>
Cc: Rik van Riel <riel@redhat.com>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2013-05-04 06:04:40 +08:00
|
|
|
rcu_read_unlock();
|
2008-04-29 16:00:49 +08:00
|
|
|
out_up:
|
|
|
|
up_write(&sem_ids(ns).rw_mutex);
|
2005-04-17 06:20:36 +08:00
|
|
|
return err;
|
|
|
|
}
|
|
|
|
|
2013-03-06 04:04:55 +08:00
|
|
|
SYSCALL_DEFINE4(semctl, int, semid, int, semnum, int, cmd, unsigned long, arg)
|
2005-04-17 06:20:36 +08:00
|
|
|
{
|
|
|
|
int version;
|
2006-10-02 17:18:22 +08:00
|
|
|
struct ipc_namespace *ns;
|
2013-03-06 04:04:55 +08:00
|
|
|
void __user *p = (void __user *)arg;
|
2005-04-17 06:20:36 +08:00
|
|
|
|
|
|
|
if (semid < 0)
|
|
|
|
return -EINVAL;
|
|
|
|
|
|
|
|
version = ipc_parse_version(&cmd);
|
2006-10-02 17:18:22 +08:00
|
|
|
ns = current->nsproxy->ipc_ns;
|
2005-04-17 06:20:36 +08:00
|
|
|
|
|
|
|
switch(cmd) {
|
|
|
|
case IPC_INFO:
|
|
|
|
case SEM_INFO:
|
2008-02-08 20:18:56 +08:00
|
|
|
case IPC_STAT:
|
2005-04-17 06:20:36 +08:00
|
|
|
case SEM_STAT:
|
2013-03-06 04:04:55 +08:00
|
|
|
return semctl_nolock(ns, semid, cmd, version, p);
|
2005-04-17 06:20:36 +08:00
|
|
|
case GETALL:
|
|
|
|
case GETVAL:
|
|
|
|
case GETPID:
|
|
|
|
case GETNCNT:
|
|
|
|
case GETZCNT:
|
|
|
|
case SETALL:
|
2013-03-06 04:04:55 +08:00
|
|
|
return semctl_main(ns, semid, semnum, cmd, p);
|
|
|
|
case SETVAL:
|
|
|
|
return semctl_setval(ns, semid, semnum, arg);
|
2005-04-17 06:20:36 +08:00
|
|
|
case IPC_RMID:
|
|
|
|
case IPC_SET:
|
2013-03-06 04:04:55 +08:00
|
|
|
return semctl_down(ns, semid, cmd, version, p);
|
2005-04-17 06:20:36 +08:00
|
|
|
default:
|
|
|
|
return -EINVAL;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
/* If the task doesn't already have a undo_list, then allocate one
|
|
|
|
* here. We guarantee there is only one thread using this undo list,
|
|
|
|
* and current is THE ONE
|
|
|
|
*
|
|
|
|
* If this allocation and assignment succeeds, but later
|
|
|
|
* portions of this code fail, there is no need to free the sem_undo_list.
|
|
|
|
* Just let it stay associated with the task, and it'll be freed later
|
|
|
|
* at exit time.
|
|
|
|
*
|
|
|
|
* This can block, so callers must hold no locks.
|
|
|
|
*/
|
|
|
|
static inline int get_undo_list(struct sem_undo_list **undo_listp)
|
|
|
|
{
|
|
|
|
struct sem_undo_list *undo_list;
|
|
|
|
|
|
|
|
undo_list = current->sysvsem.undo_list;
|
|
|
|
if (!undo_list) {
|
2006-10-02 17:18:25 +08:00
|
|
|
undo_list = kzalloc(sizeof(*undo_list), GFP_KERNEL);
|
2005-04-17 06:20:36 +08:00
|
|
|
if (undo_list == NULL)
|
|
|
|
return -ENOMEM;
|
2005-08-06 05:05:27 +08:00
|
|
|
spin_lock_init(&undo_list->lock);
|
2005-04-17 06:20:36 +08:00
|
|
|
atomic_set(&undo_list->refcnt, 1);
|
2008-07-25 16:48:04 +08:00
|
|
|
INIT_LIST_HEAD(&undo_list->list_proc);
|
|
|
|
|
2005-04-17 06:20:36 +08:00
|
|
|
current->sysvsem.undo_list = undo_list;
|
|
|
|
}
|
|
|
|
*undo_listp = undo_list;
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2009-12-16 08:47:28 +08:00
|
|
|
static struct sem_undo *__lookup_undo(struct sem_undo_list *ulp, int semid)
|
2005-04-17 06:20:36 +08:00
|
|
|
{
|
2009-12-16 08:47:28 +08:00
|
|
|
struct sem_undo *un;
|
2008-07-25 16:48:04 +08:00
|
|
|
|
2009-12-16 08:47:28 +08:00
|
|
|
list_for_each_entry_rcu(un, &ulp->list_proc, list_proc) {
|
|
|
|
if (un->semid == semid)
|
|
|
|
return un;
|
2005-04-17 06:20:36 +08:00
|
|
|
}
|
2008-07-25 16:48:04 +08:00
|
|
|
return NULL;
|
2005-04-17 06:20:36 +08:00
|
|
|
}
|
|
|
|
|
2009-12-16 08:47:28 +08:00
|
|
|
static struct sem_undo *lookup_undo(struct sem_undo_list *ulp, int semid)
|
|
|
|
{
|
|
|
|
struct sem_undo *un;
|
|
|
|
|
|
|
|
assert_spin_locked(&ulp->lock);
|
|
|
|
|
|
|
|
un = __lookup_undo(ulp, semid);
|
|
|
|
if (un) {
|
|
|
|
list_del_rcu(&un->list_proc);
|
|
|
|
list_add_rcu(&un->list_proc, &ulp->list_proc);
|
|
|
|
}
|
|
|
|
return un;
|
|
|
|
}
|
|
|
|
|
2008-07-25 16:48:04 +08:00
|
|
|
/**
|
|
|
|
* find_alloc_undo - Lookup (and if not present create) undo array
|
|
|
|
* @ns: namespace
|
|
|
|
* @semid: semaphore array id
|
|
|
|
*
|
|
|
|
* The function looks up (and if not present creates) the undo structure.
|
|
|
|
* The size of the undo structure depends on the size of the semaphore
|
|
|
|
* array, thus the alloc path is not that straightforward.
|
2008-07-25 16:48:06 +08:00
|
|
|
* Lifetime-rules: sem_undo is rcu-protected, on success, the function
|
|
|
|
* performs a rcu_read_lock().
|
2008-07-25 16:48:04 +08:00
|
|
|
*/
|
|
|
|
static struct sem_undo *find_alloc_undo(struct ipc_namespace *ns, int semid)
|
2005-04-17 06:20:36 +08:00
|
|
|
{
|
|
|
|
struct sem_array *sma;
|
|
|
|
struct sem_undo_list *ulp;
|
|
|
|
struct sem_undo *un, *new;
|
2013-05-01 10:15:44 +08:00
|
|
|
int nsems, error;
|
2005-04-17 06:20:36 +08:00
|
|
|
|
|
|
|
error = get_undo_list(&ulp);
|
|
|
|
if (error)
|
|
|
|
return ERR_PTR(error);
|
|
|
|
|
2008-07-25 16:48:06 +08:00
|
|
|
rcu_read_lock();
|
2007-10-19 14:40:55 +08:00
|
|
|
spin_lock(&ulp->lock);
|
2005-04-17 06:20:36 +08:00
|
|
|
un = lookup_undo(ulp, semid);
|
2007-10-19 14:40:55 +08:00
|
|
|
spin_unlock(&ulp->lock);
|
2005-04-17 06:20:36 +08:00
|
|
|
if (likely(un!=NULL))
|
|
|
|
goto out;
|
|
|
|
|
|
|
|
/* no undo structure around - allocate one. */
|
2008-07-25 16:48:04 +08:00
|
|
|
/* step 1: figure out the size of the semaphore array */
|
ipc,sem: do not hold ipc lock more than necessary
Instead of holding the ipc lock for permissions and security checks, among
others, only acquire it when necessary.
Some numbers....
1) With Rik's semop-multi.c microbenchmark we can see the following
results:
Baseline (3.9-rc1):
cpus 4, threads: 256, semaphores: 128, test duration: 30 secs
total operations: 151452270, ops/sec 5048409
+ 59.40% a.out [kernel.kallsyms] [k] _raw_spin_lock
+ 6.14% a.out [kernel.kallsyms] [k] sys_semtimedop
+ 3.84% a.out [kernel.kallsyms] [k] avc_has_perm_flags
+ 3.64% a.out [kernel.kallsyms] [k] __audit_syscall_exit
+ 2.06% a.out [kernel.kallsyms] [k] copy_user_enhanced_fast_string
+ 1.86% a.out [kernel.kallsyms] [k] ipc_lock
With this patchset:
cpus 4, threads: 256, semaphores: 128, test duration: 30 secs
total operations: 273156400, ops/sec 9105213
+ 18.54% a.out [kernel.kallsyms] [k] _raw_spin_lock
+ 11.72% a.out [kernel.kallsyms] [k] sys_semtimedop
+ 7.70% a.out [kernel.kallsyms] [k] ipc_has_perm.isra.21
+ 6.58% a.out [kernel.kallsyms] [k] avc_has_perm_flags
+ 6.54% a.out [kernel.kallsyms] [k] __audit_syscall_exit
+ 4.71% a.out [kernel.kallsyms] [k] ipc_obtain_object_check
2) While on an Oracle swingbench DSS (data mining) workload the
improvements are not as exciting as with Rik's benchmark, we can see
some positive numbers. For an 8 socket machine the following are the
percentages of %sys time incurred in the ipc lock:
Baseline (3.9-rc1):
100 swingbench users: 8,74%
400 swingbench users: 21,86%
800 swingbench users: 84,35%
With this patchset:
100 swingbench users: 8,11%
400 swingbench users: 19,93%
800 swingbench users: 77,69%
[riel@redhat.com: fix two locking bugs]
[sasha.levin@oracle.com: prevent releasing RCU read lock twice in semctl_main]
[akpm@linux-foundation.org: coding-style fixes]
Signed-off-by: Davidlohr Bueso <davidlohr.bueso@hp.com>
Signed-off-by: Rik van Riel <riel@redhat.com>
Reviewed-by: Chegu Vinod <chegu_vinod@hp.com>
Acked-by: Michel Lespinasse <walken@google.com>
Cc: Rik van Riel <riel@redhat.com>
Cc: Jason Low <jason.low2@hp.com>
Cc: Emmanuel Benisty <benisty.e@gmail.com>
Cc: Peter Hurley <peter@hurleysoftware.com>
Cc: Stanislav Kinsbursky <skinsbursky@parallels.com>
Tested-by: Sedat Dilek <sedat.dilek@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2013-05-01 10:15:29 +08:00
|
|
|
sma = sem_obtain_object_check(ns, semid);
|
|
|
|
if (IS_ERR(sma)) {
|
|
|
|
rcu_read_unlock();
|
2010-05-27 05:43:44 +08:00
|
|
|
return ERR_CAST(sma);
|
ipc,sem: do not hold ipc lock more than necessary
Instead of holding the ipc lock for permissions and security checks, among
others, only acquire it when necessary.
Some numbers....
1) With Rik's semop-multi.c microbenchmark we can see the following
results:
Baseline (3.9-rc1):
cpus 4, threads: 256, semaphores: 128, test duration: 30 secs
total operations: 151452270, ops/sec 5048409
+ 59.40% a.out [kernel.kallsyms] [k] _raw_spin_lock
+ 6.14% a.out [kernel.kallsyms] [k] sys_semtimedop
+ 3.84% a.out [kernel.kallsyms] [k] avc_has_perm_flags
+ 3.64% a.out [kernel.kallsyms] [k] __audit_syscall_exit
+ 2.06% a.out [kernel.kallsyms] [k] copy_user_enhanced_fast_string
+ 1.86% a.out [kernel.kallsyms] [k] ipc_lock
With this patchset:
cpus 4, threads: 256, semaphores: 128, test duration: 30 secs
total operations: 273156400, ops/sec 9105213
+ 18.54% a.out [kernel.kallsyms] [k] _raw_spin_lock
+ 11.72% a.out [kernel.kallsyms] [k] sys_semtimedop
+ 7.70% a.out [kernel.kallsyms] [k] ipc_has_perm.isra.21
+ 6.58% a.out [kernel.kallsyms] [k] avc_has_perm_flags
+ 6.54% a.out [kernel.kallsyms] [k] __audit_syscall_exit
+ 4.71% a.out [kernel.kallsyms] [k] ipc_obtain_object_check
2) While on an Oracle swingbench DSS (data mining) workload the
improvements are not as exciting as with Rik's benchmark, we can see
some positive numbers. For an 8 socket machine the following are the
percentages of %sys time incurred in the ipc lock:
Baseline (3.9-rc1):
100 swingbench users: 8,74%
400 swingbench users: 21,86%
800 swingbench users: 84,35%
With this patchset:
100 swingbench users: 8,11%
400 swingbench users: 19,93%
800 swingbench users: 77,69%
[riel@redhat.com: fix two locking bugs]
[sasha.levin@oracle.com: prevent releasing RCU read lock twice in semctl_main]
[akpm@linux-foundation.org: coding-style fixes]
Signed-off-by: Davidlohr Bueso <davidlohr.bueso@hp.com>
Signed-off-by: Rik van Riel <riel@redhat.com>
Reviewed-by: Chegu Vinod <chegu_vinod@hp.com>
Acked-by: Michel Lespinasse <walken@google.com>
Cc: Rik van Riel <riel@redhat.com>
Cc: Jason Low <jason.low2@hp.com>
Cc: Emmanuel Benisty <benisty.e@gmail.com>
Cc: Peter Hurley <peter@hurleysoftware.com>
Cc: Stanislav Kinsbursky <skinsbursky@parallels.com>
Tested-by: Sedat Dilek <sedat.dilek@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2013-05-01 10:15:29 +08:00
|
|
|
}
|
2007-10-19 14:40:51 +08:00
|
|
|
|
2005-04-17 06:20:36 +08:00
|
|
|
nsems = sma->sem_nsems;
|
2013-05-01 10:15:44 +08:00
|
|
|
if (!ipc_rcu_getref(sma)) {
|
|
|
|
rcu_read_unlock();
|
|
|
|
un = ERR_PTR(-EIDRM);
|
|
|
|
goto out;
|
|
|
|
}
|
ipc,sem: do not hold ipc lock more than necessary
Instead of holding the ipc lock for permissions and security checks, among
others, only acquire it when necessary.
Some numbers....
1) With Rik's semop-multi.c microbenchmark we can see the following
results:
Baseline (3.9-rc1):
cpus 4, threads: 256, semaphores: 128, test duration: 30 secs
total operations: 151452270, ops/sec 5048409
+ 59.40% a.out [kernel.kallsyms] [k] _raw_spin_lock
+ 6.14% a.out [kernel.kallsyms] [k] sys_semtimedop
+ 3.84% a.out [kernel.kallsyms] [k] avc_has_perm_flags
+ 3.64% a.out [kernel.kallsyms] [k] __audit_syscall_exit
+ 2.06% a.out [kernel.kallsyms] [k] copy_user_enhanced_fast_string
+ 1.86% a.out [kernel.kallsyms] [k] ipc_lock
With this patchset:
cpus 4, threads: 256, semaphores: 128, test duration: 30 secs
total operations: 273156400, ops/sec 9105213
+ 18.54% a.out [kernel.kallsyms] [k] _raw_spin_lock
+ 11.72% a.out [kernel.kallsyms] [k] sys_semtimedop
+ 7.70% a.out [kernel.kallsyms] [k] ipc_has_perm.isra.21
+ 6.58% a.out [kernel.kallsyms] [k] avc_has_perm_flags
+ 6.54% a.out [kernel.kallsyms] [k] __audit_syscall_exit
+ 4.71% a.out [kernel.kallsyms] [k] ipc_obtain_object_check
2) While on an Oracle swingbench DSS (data mining) workload the
improvements are not as exciting as with Rik's benchmark, we can see
some positive numbers. For an 8 socket machine the following are the
percentages of %sys time incurred in the ipc lock:
Baseline (3.9-rc1):
100 swingbench users: 8,74%
400 swingbench users: 21,86%
800 swingbench users: 84,35%
With this patchset:
100 swingbench users: 8,11%
400 swingbench users: 19,93%
800 swingbench users: 77,69%
[riel@redhat.com: fix two locking bugs]
[sasha.levin@oracle.com: prevent releasing RCU read lock twice in semctl_main]
[akpm@linux-foundation.org: coding-style fixes]
Signed-off-by: Davidlohr Bueso <davidlohr.bueso@hp.com>
Signed-off-by: Rik van Riel <riel@redhat.com>
Reviewed-by: Chegu Vinod <chegu_vinod@hp.com>
Acked-by: Michel Lespinasse <walken@google.com>
Cc: Rik van Riel <riel@redhat.com>
Cc: Jason Low <jason.low2@hp.com>
Cc: Emmanuel Benisty <benisty.e@gmail.com>
Cc: Peter Hurley <peter@hurleysoftware.com>
Cc: Stanislav Kinsbursky <skinsbursky@parallels.com>
Tested-by: Sedat Dilek <sedat.dilek@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2013-05-01 10:15:29 +08:00
|
|
|
rcu_read_unlock();
|
2005-04-17 06:20:36 +08:00
|
|
|
|
2008-07-25 16:48:04 +08:00
|
|
|
/* step 2: allocate new undo structure */
|
2006-12-07 12:38:51 +08:00
|
|
|
new = kzalloc(sizeof(struct sem_undo) + sizeof(short)*nsems, GFP_KERNEL);
|
2005-04-17 06:20:36 +08:00
|
|
|
if (!new) {
|
2008-04-29 16:00:46 +08:00
|
|
|
sem_putref(sma);
|
2005-04-17 06:20:36 +08:00
|
|
|
return ERR_PTR(-ENOMEM);
|
|
|
|
}
|
|
|
|
|
2008-07-25 16:48:06 +08:00
|
|
|
/* step 3: Acquire the lock on semaphore array */
|
2013-05-05 01:13:40 +08:00
|
|
|
rcu_read_lock();
|
2008-04-29 16:00:46 +08:00
|
|
|
sem_lock_and_putref(sma);
|
2005-04-17 06:20:36 +08:00
|
|
|
if (sma->sem_perm.deleted) {
|
2013-05-01 10:15:44 +08:00
|
|
|
sem_unlock(sma, -1);
|
ipc: move rcu_read_unlock() out of sem_unlock() and into callers
The IPC locking is a mess, and sem_unlock() unlocks not only the
semaphore spinlock, it also drops the rcu read lock. Unlike sem_lock(),
which just gets the spin-lock, and expects the caller to get the rcu
read lock.
This all makes things very hard to follow, and it's very confusing when
you take the rcu read lock in one function, and then release it in
another. And it has caused actual bugs: the sem_obtain_lock() function
ended up dropping the RCU read lock twice in one error path, because it
first did the sem_unlock(), and then did a rcu_read_unlock() to match
the rcu_read_lock() it had done.
This is just a totally mindless "remove rcu_read_unlock() from
sem_unlock() and add it immediately after each caller" (except for the
aforementioned bug where we did too many rcu_read_unlock(), and in
find_alloc_undo() where we just got the rcu_read_lock() to correct for
the fact that sem_unlock would immediately drop it again).
We can (and should) clean things up further, but this fixes the bug with
the minimal amount of subtlety.
Reviewed-by: Davidlohr Bueso <davidlohr.bueso@hp.com>
Cc: Rik van Riel <riel@redhat.com>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2013-05-04 06:04:40 +08:00
|
|
|
rcu_read_unlock();
|
2005-04-17 06:20:36 +08:00
|
|
|
kfree(new);
|
|
|
|
un = ERR_PTR(-EIDRM);
|
|
|
|
goto out;
|
|
|
|
}
|
2008-07-25 16:48:06 +08:00
|
|
|
spin_lock(&ulp->lock);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* step 4: check for races: did someone else allocate the undo struct?
|
|
|
|
*/
|
|
|
|
un = lookup_undo(ulp, semid);
|
|
|
|
if (un) {
|
|
|
|
kfree(new);
|
|
|
|
goto success;
|
|
|
|
}
|
2008-07-25 16:48:04 +08:00
|
|
|
/* step 5: initialize & link new undo structure */
|
|
|
|
new->semadj = (short *) &new[1];
|
2008-07-25 16:48:06 +08:00
|
|
|
new->ulp = ulp;
|
2008-07-25 16:48:04 +08:00
|
|
|
new->semid = semid;
|
|
|
|
assert_spin_locked(&ulp->lock);
|
2008-07-25 16:48:06 +08:00
|
|
|
list_add_rcu(&new->list_proc, &ulp->list_proc);
|
2008-07-25 16:48:04 +08:00
|
|
|
assert_spin_locked(&sma->sem_perm.lock);
|
|
|
|
list_add(&new->list_id, &sma->list_id);
|
2008-07-25 16:48:06 +08:00
|
|
|
un = new;
|
2008-07-25 16:48:04 +08:00
|
|
|
|
2008-07-25 16:48:06 +08:00
|
|
|
success:
|
2007-10-19 14:40:55 +08:00
|
|
|
spin_unlock(&ulp->lock);
|
2013-05-01 10:15:44 +08:00
|
|
|
sem_unlock(sma, -1);
|
2005-04-17 06:20:36 +08:00
|
|
|
out:
|
|
|
|
return un;
|
|
|
|
}
|
|
|
|
|
2010-07-21 04:24:23 +08:00
|
|
|
|
|
|
|
/**
|
|
|
|
* get_queue_result - Retrieve the result code from sem_queue
|
|
|
|
* @q: Pointer to queue structure
|
|
|
|
*
|
|
|
|
* Retrieve the return code from the pending queue. If IN_WAKEUP is found in
|
|
|
|
* q->status, then we must loop until the value is replaced with the final
|
|
|
|
* value: This may happen if a task is woken up by an unrelated event (e.g.
|
|
|
|
* signal) and in parallel the task is woken up by another task because it got
|
|
|
|
* the requested semaphores.
|
|
|
|
*
|
|
|
|
* The function can be called with or without holding the semaphore spinlock.
|
|
|
|
*/
|
|
|
|
static int get_queue_result(struct sem_queue *q)
|
|
|
|
{
|
|
|
|
int error;
|
|
|
|
|
|
|
|
error = q->status;
|
|
|
|
while (unlikely(error == IN_WAKEUP)) {
|
|
|
|
cpu_relax();
|
|
|
|
error = q->status;
|
|
|
|
}
|
|
|
|
|
|
|
|
return error;
|
|
|
|
}
|
|
|
|
|
|
|
|
|
2009-01-14 21:14:27 +08:00
|
|
|
SYSCALL_DEFINE4(semtimedop, int, semid, struct sembuf __user *, tsops,
|
|
|
|
unsigned, nsops, const struct timespec __user *, timeout)
|
2005-04-17 06:20:36 +08:00
|
|
|
{
|
|
|
|
int error = -EINVAL;
|
|
|
|
struct sem_array *sma;
|
|
|
|
struct sembuf fast_sops[SEMOPM_FAST];
|
|
|
|
struct sembuf* sops = fast_sops, *sop;
|
|
|
|
struct sem_undo *un;
|
2013-05-01 10:15:44 +08:00
|
|
|
int undos = 0, alter = 0, max, locknum;
|
2005-04-17 06:20:36 +08:00
|
|
|
struct sem_queue queue;
|
|
|
|
unsigned long jiffies_left = 0;
|
2006-10-02 17:18:22 +08:00
|
|
|
struct ipc_namespace *ns;
|
2010-05-27 05:43:41 +08:00
|
|
|
struct list_head tasks;
|
2006-10-02 17:18:22 +08:00
|
|
|
|
|
|
|
ns = current->nsproxy->ipc_ns;
|
2005-04-17 06:20:36 +08:00
|
|
|
|
|
|
|
if (nsops < 1 || semid < 0)
|
|
|
|
return -EINVAL;
|
2006-10-02 17:18:22 +08:00
|
|
|
if (nsops > ns->sc_semopm)
|
2005-04-17 06:20:36 +08:00
|
|
|
return -E2BIG;
|
|
|
|
if(nsops > SEMOPM_FAST) {
|
|
|
|
sops = kmalloc(sizeof(*sops)*nsops,GFP_KERNEL);
|
|
|
|
if(sops==NULL)
|
|
|
|
return -ENOMEM;
|
|
|
|
}
|
|
|
|
if (copy_from_user (sops, tsops, nsops * sizeof(*tsops))) {
|
|
|
|
error=-EFAULT;
|
|
|
|
goto out_free;
|
|
|
|
}
|
|
|
|
if (timeout) {
|
|
|
|
struct timespec _timeout;
|
|
|
|
if (copy_from_user(&_timeout, timeout, sizeof(*timeout))) {
|
|
|
|
error = -EFAULT;
|
|
|
|
goto out_free;
|
|
|
|
}
|
|
|
|
if (_timeout.tv_sec < 0 || _timeout.tv_nsec < 0 ||
|
|
|
|
_timeout.tv_nsec >= 1000000000L) {
|
|
|
|
error = -EINVAL;
|
|
|
|
goto out_free;
|
|
|
|
}
|
|
|
|
jiffies_left = timespec_to_jiffies(&_timeout);
|
|
|
|
}
|
|
|
|
max = 0;
|
|
|
|
for (sop = sops; sop < sops + nsops; sop++) {
|
|
|
|
if (sop->sem_num >= max)
|
|
|
|
max = sop->sem_num;
|
|
|
|
if (sop->sem_flg & SEM_UNDO)
|
2005-06-23 15:10:06 +08:00
|
|
|
undos = 1;
|
|
|
|
if (sop->sem_op != 0)
|
2005-04-17 06:20:36 +08:00
|
|
|
alter = 1;
|
|
|
|
}
|
|
|
|
|
2013-05-01 10:15:44 +08:00
|
|
|
INIT_LIST_HEAD(&tasks);
|
|
|
|
|
2005-04-17 06:20:36 +08:00
|
|
|
if (undos) {
|
2013-05-01 10:15:44 +08:00
|
|
|
/* On success, find_alloc_undo takes the rcu_read_lock */
|
2008-07-25 16:48:04 +08:00
|
|
|
un = find_alloc_undo(ns, semid);
|
2005-04-17 06:20:36 +08:00
|
|
|
if (IS_ERR(un)) {
|
|
|
|
error = PTR_ERR(un);
|
|
|
|
goto out_free;
|
|
|
|
}
|
2013-05-01 10:15:44 +08:00
|
|
|
} else {
|
2005-04-17 06:20:36 +08:00
|
|
|
un = NULL;
|
2013-05-01 10:15:44 +08:00
|
|
|
rcu_read_lock();
|
|
|
|
}
|
2005-04-17 06:20:36 +08:00
|
|
|
|
ipc,sem: do not hold ipc lock more than necessary
Instead of holding the ipc lock for permissions and security checks, among
others, only acquire it when necessary.
Some numbers....
1) With Rik's semop-multi.c microbenchmark we can see the following
results:
Baseline (3.9-rc1):
cpus 4, threads: 256, semaphores: 128, test duration: 30 secs
total operations: 151452270, ops/sec 5048409
+ 59.40% a.out [kernel.kallsyms] [k] _raw_spin_lock
+ 6.14% a.out [kernel.kallsyms] [k] sys_semtimedop
+ 3.84% a.out [kernel.kallsyms] [k] avc_has_perm_flags
+ 3.64% a.out [kernel.kallsyms] [k] __audit_syscall_exit
+ 2.06% a.out [kernel.kallsyms] [k] copy_user_enhanced_fast_string
+ 1.86% a.out [kernel.kallsyms] [k] ipc_lock
With this patchset:
cpus 4, threads: 256, semaphores: 128, test duration: 30 secs
total operations: 273156400, ops/sec 9105213
+ 18.54% a.out [kernel.kallsyms] [k] _raw_spin_lock
+ 11.72% a.out [kernel.kallsyms] [k] sys_semtimedop
+ 7.70% a.out [kernel.kallsyms] [k] ipc_has_perm.isra.21
+ 6.58% a.out [kernel.kallsyms] [k] avc_has_perm_flags
+ 6.54% a.out [kernel.kallsyms] [k] __audit_syscall_exit
+ 4.71% a.out [kernel.kallsyms] [k] ipc_obtain_object_check
2) While on an Oracle swingbench DSS (data mining) workload the
improvements are not as exciting as with Rik's benchmark, we can see
some positive numbers. For an 8 socket machine the following are the
percentages of %sys time incurred in the ipc lock:
Baseline (3.9-rc1):
100 swingbench users: 8,74%
400 swingbench users: 21,86%
800 swingbench users: 84,35%
With this patchset:
100 swingbench users: 8,11%
400 swingbench users: 19,93%
800 swingbench users: 77,69%
[riel@redhat.com: fix two locking bugs]
[sasha.levin@oracle.com: prevent releasing RCU read lock twice in semctl_main]
[akpm@linux-foundation.org: coding-style fixes]
Signed-off-by: Davidlohr Bueso <davidlohr.bueso@hp.com>
Signed-off-by: Rik van Riel <riel@redhat.com>
Reviewed-by: Chegu Vinod <chegu_vinod@hp.com>
Acked-by: Michel Lespinasse <walken@google.com>
Cc: Rik van Riel <riel@redhat.com>
Cc: Jason Low <jason.low2@hp.com>
Cc: Emmanuel Benisty <benisty.e@gmail.com>
Cc: Peter Hurley <peter@hurleysoftware.com>
Cc: Stanislav Kinsbursky <skinsbursky@parallels.com>
Tested-by: Sedat Dilek <sedat.dilek@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2013-05-01 10:15:29 +08:00
|
|
|
sma = sem_obtain_object_check(ns, semid);
|
2007-10-19 14:40:51 +08:00
|
|
|
if (IS_ERR(sma)) {
|
2013-05-01 10:15:44 +08:00
|
|
|
rcu_read_unlock();
|
2007-10-19 14:40:51 +08:00
|
|
|
error = PTR_ERR(sma);
|
2005-04-17 06:20:36 +08:00
|
|
|
goto out_free;
|
2007-10-19 14:40:51 +08:00
|
|
|
}
|
|
|
|
|
ipc,sem: do not hold ipc lock more than necessary
Instead of holding the ipc lock for permissions and security checks, among
others, only acquire it when necessary.
Some numbers....
1) With Rik's semop-multi.c microbenchmark we can see the following
results:
Baseline (3.9-rc1):
cpus 4, threads: 256, semaphores: 128, test duration: 30 secs
total operations: 151452270, ops/sec 5048409
+ 59.40% a.out [kernel.kallsyms] [k] _raw_spin_lock
+ 6.14% a.out [kernel.kallsyms] [k] sys_semtimedop
+ 3.84% a.out [kernel.kallsyms] [k] avc_has_perm_flags
+ 3.64% a.out [kernel.kallsyms] [k] __audit_syscall_exit
+ 2.06% a.out [kernel.kallsyms] [k] copy_user_enhanced_fast_string
+ 1.86% a.out [kernel.kallsyms] [k] ipc_lock
With this patchset:
cpus 4, threads: 256, semaphores: 128, test duration: 30 secs
total operations: 273156400, ops/sec 9105213
+ 18.54% a.out [kernel.kallsyms] [k] _raw_spin_lock
+ 11.72% a.out [kernel.kallsyms] [k] sys_semtimedop
+ 7.70% a.out [kernel.kallsyms] [k] ipc_has_perm.isra.21
+ 6.58% a.out [kernel.kallsyms] [k] avc_has_perm_flags
+ 6.54% a.out [kernel.kallsyms] [k] __audit_syscall_exit
+ 4.71% a.out [kernel.kallsyms] [k] ipc_obtain_object_check
2) While on an Oracle swingbench DSS (data mining) workload the
improvements are not as exciting as with Rik's benchmark, we can see
some positive numbers. For an 8 socket machine the following are the
percentages of %sys time incurred in the ipc lock:
Baseline (3.9-rc1):
100 swingbench users: 8,74%
400 swingbench users: 21,86%
800 swingbench users: 84,35%
With this patchset:
100 swingbench users: 8,11%
400 swingbench users: 19,93%
800 swingbench users: 77,69%
[riel@redhat.com: fix two locking bugs]
[sasha.levin@oracle.com: prevent releasing RCU read lock twice in semctl_main]
[akpm@linux-foundation.org: coding-style fixes]
Signed-off-by: Davidlohr Bueso <davidlohr.bueso@hp.com>
Signed-off-by: Rik van Riel <riel@redhat.com>
Reviewed-by: Chegu Vinod <chegu_vinod@hp.com>
Acked-by: Michel Lespinasse <walken@google.com>
Cc: Rik van Riel <riel@redhat.com>
Cc: Jason Low <jason.low2@hp.com>
Cc: Emmanuel Benisty <benisty.e@gmail.com>
Cc: Peter Hurley <peter@hurleysoftware.com>
Cc: Stanislav Kinsbursky <skinsbursky@parallels.com>
Tested-by: Sedat Dilek <sedat.dilek@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2013-05-01 10:15:29 +08:00
|
|
|
error = -EFBIG;
|
|
|
|
if (max >= sma->sem_nsems) {
|
|
|
|
rcu_read_unlock();
|
|
|
|
goto out_wakeup;
|
|
|
|
}
|
|
|
|
|
|
|
|
error = -EACCES;
|
|
|
|
if (ipcperms(ns, &sma->sem_perm, alter ? S_IWUGO : S_IRUGO)) {
|
|
|
|
rcu_read_unlock();
|
|
|
|
goto out_wakeup;
|
|
|
|
}
|
|
|
|
|
|
|
|
error = security_sem_semop(sma, sops, nsops, alter);
|
|
|
|
if (error) {
|
|
|
|
rcu_read_unlock();
|
|
|
|
goto out_wakeup;
|
|
|
|
}
|
|
|
|
|
2005-04-17 06:20:36 +08:00
|
|
|
/*
|
2008-07-25 16:48:04 +08:00
|
|
|
* semid identifiers are not unique - find_alloc_undo may have
|
2005-04-17 06:20:36 +08:00
|
|
|
* allocated an undo structure, it was invalidated by an RMID
|
2008-07-25 16:48:04 +08:00
|
|
|
* and now a new array with received the same id. Check and fail.
|
2011-03-31 09:57:33 +08:00
|
|
|
* This case can be detected checking un->semid. The existence of
|
2008-07-25 16:48:06 +08:00
|
|
|
* "un" itself is guaranteed by rcu.
|
2005-04-17 06:20:36 +08:00
|
|
|
*/
|
2008-07-25 16:48:04 +08:00
|
|
|
error = -EIDRM;
|
2013-05-01 10:15:44 +08:00
|
|
|
locknum = sem_lock(sma, sops, nsops);
|
|
|
|
if (un && un->semid == -1)
|
|
|
|
goto out_unlock_free;
|
2008-07-25 16:48:04 +08:00
|
|
|
|
2007-10-19 14:40:14 +08:00
|
|
|
error = try_atomic_semop (sma, sops, nsops, un, task_tgid_vnr(current));
|
2005-04-17 06:20:36 +08:00
|
|
|
if (error <= 0) {
|
|
|
|
if (alter && error == 0)
|
2010-05-27 05:43:41 +08:00
|
|
|
do_smart_update(sma, sops, nsops, 1, &tasks);
|
2009-12-16 08:47:33 +08:00
|
|
|
|
2005-04-17 06:20:36 +08:00
|
|
|
goto out_unlock_free;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* We need to sleep on this operation, so we put the current
|
|
|
|
* task into the pending queue and go to sleep.
|
|
|
|
*/
|
|
|
|
|
|
|
|
queue.sops = sops;
|
|
|
|
queue.nsops = nsops;
|
|
|
|
queue.undo = un;
|
2007-10-19 14:40:14 +08:00
|
|
|
queue.pid = task_tgid_vnr(current);
|
2005-04-17 06:20:36 +08:00
|
|
|
queue.alter = alter;
|
|
|
|
|
2009-12-16 08:47:32 +08:00
|
|
|
if (nsops == 1) {
|
|
|
|
struct sem *curr;
|
|
|
|
curr = &sma->sem_base[sops->sem_num];
|
|
|
|
|
|
|
|
if (alter)
|
2013-05-01 10:15:39 +08:00
|
|
|
list_add_tail(&queue.list, &curr->sem_pending);
|
2009-12-16 08:47:32 +08:00
|
|
|
else
|
2013-05-01 10:15:39 +08:00
|
|
|
list_add(&queue.list, &curr->sem_pending);
|
2009-12-16 08:47:32 +08:00
|
|
|
} else {
|
2013-05-01 10:15:39 +08:00
|
|
|
if (alter)
|
|
|
|
list_add_tail(&queue.list, &sma->sem_pending);
|
|
|
|
else
|
|
|
|
list_add(&queue.list, &sma->sem_pending);
|
2009-12-16 08:47:32 +08:00
|
|
|
sma->complex_count++;
|
|
|
|
}
|
|
|
|
|
2005-04-17 06:20:36 +08:00
|
|
|
queue.status = -EINTR;
|
|
|
|
queue.sleeper = current;
|
2011-11-03 04:38:52 +08:00
|
|
|
|
|
|
|
sleep_again:
|
2005-04-17 06:20:36 +08:00
|
|
|
current->state = TASK_INTERRUPTIBLE;
|
2013-05-01 10:15:44 +08:00
|
|
|
sem_unlock(sma, locknum);
|
ipc: move rcu_read_unlock() out of sem_unlock() and into callers
The IPC locking is a mess, and sem_unlock() unlocks not only the
semaphore spinlock, it also drops the rcu read lock. Unlike sem_lock(),
which just gets the spin-lock, and expects the caller to get the rcu
read lock.
This all makes things very hard to follow, and it's very confusing when
you take the rcu read lock in one function, and then release it in
another. And it has caused actual bugs: the sem_obtain_lock() function
ended up dropping the RCU read lock twice in one error path, because it
first did the sem_unlock(), and then did a rcu_read_unlock() to match
the rcu_read_lock() it had done.
This is just a totally mindless "remove rcu_read_unlock() from
sem_unlock() and add it immediately after each caller" (except for the
aforementioned bug where we did too many rcu_read_unlock(), and in
find_alloc_undo() where we just got the rcu_read_lock() to correct for
the fact that sem_unlock would immediately drop it again).
We can (and should) clean things up further, but this fixes the bug with
the minimal amount of subtlety.
Reviewed-by: Davidlohr Bueso <davidlohr.bueso@hp.com>
Cc: Rik van Riel <riel@redhat.com>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2013-05-04 06:04:40 +08:00
|
|
|
rcu_read_unlock();
|
2005-04-17 06:20:36 +08:00
|
|
|
|
|
|
|
if (timeout)
|
|
|
|
jiffies_left = schedule_timeout(jiffies_left);
|
|
|
|
else
|
|
|
|
schedule();
|
|
|
|
|
2010-07-21 04:24:23 +08:00
|
|
|
error = get_queue_result(&queue);
|
2005-04-17 06:20:36 +08:00
|
|
|
|
|
|
|
if (error != -EINTR) {
|
|
|
|
/* fast path: update_queue already obtained all requested
|
2010-07-21 04:24:23 +08:00
|
|
|
* resources.
|
|
|
|
* Perform a smp_mb(): User space could assume that semop()
|
|
|
|
* is a memory barrier: Without the mb(), the cpu could
|
|
|
|
* speculatively read in user space stale data that was
|
|
|
|
* overwritten by the previous owner of the semaphore.
|
|
|
|
*/
|
|
|
|
smp_mb();
|
|
|
|
|
2005-04-17 06:20:36 +08:00
|
|
|
goto out_free;
|
|
|
|
}
|
|
|
|
|
2013-05-05 01:47:57 +08:00
|
|
|
rcu_read_lock();
|
2013-05-01 10:15:44 +08:00
|
|
|
sma = sem_obtain_lock(ns, semid, sops, nsops, &locknum);
|
2011-07-26 08:11:47 +08:00
|
|
|
|
|
|
|
/*
|
|
|
|
* Wait until it's guaranteed that no wakeup_sem_queue_do() is ongoing.
|
|
|
|
*/
|
|
|
|
error = get_queue_result(&queue);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Array removed? If yes, leave without sem_unlock().
|
|
|
|
*/
|
2007-10-19 14:40:51 +08:00
|
|
|
if (IS_ERR(sma)) {
|
2013-05-05 01:47:57 +08:00
|
|
|
rcu_read_unlock();
|
2005-04-17 06:20:36 +08:00
|
|
|
goto out_free;
|
|
|
|
}
|
|
|
|
|
2010-07-21 04:24:23 +08:00
|
|
|
|
2005-04-17 06:20:36 +08:00
|
|
|
/*
|
2011-07-26 08:11:47 +08:00
|
|
|
* If queue.status != -EINTR we are woken up by another process.
|
|
|
|
* Leave without unlink_queue(), but with sem_unlock().
|
2005-04-17 06:20:36 +08:00
|
|
|
*/
|
2010-07-21 04:24:23 +08:00
|
|
|
|
2005-04-17 06:20:36 +08:00
|
|
|
if (error != -EINTR) {
|
|
|
|
goto out_unlock_free;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* If an interrupt occurred we have to clean up the queue
|
|
|
|
*/
|
|
|
|
if (timeout && jiffies_left == 0)
|
|
|
|
error = -EAGAIN;
|
2011-11-03 04:38:52 +08:00
|
|
|
|
|
|
|
/*
|
|
|
|
* If the wakeup was spurious, just retry
|
|
|
|
*/
|
|
|
|
if (error == -EINTR && !signal_pending(current))
|
|
|
|
goto sleep_again;
|
|
|
|
|
2009-12-16 08:47:32 +08:00
|
|
|
unlink_queue(sma, &queue);
|
2005-04-17 06:20:36 +08:00
|
|
|
|
|
|
|
out_unlock_free:
|
2013-05-01 10:15:44 +08:00
|
|
|
sem_unlock(sma, locknum);
|
ipc: move rcu_read_unlock() out of sem_unlock() and into callers
The IPC locking is a mess, and sem_unlock() unlocks not only the
semaphore spinlock, it also drops the rcu read lock. Unlike sem_lock(),
which just gets the spin-lock, and expects the caller to get the rcu
read lock.
This all makes things very hard to follow, and it's very confusing when
you take the rcu read lock in one function, and then release it in
another. And it has caused actual bugs: the sem_obtain_lock() function
ended up dropping the RCU read lock twice in one error path, because it
first did the sem_unlock(), and then did a rcu_read_unlock() to match
the rcu_read_lock() it had done.
This is just a totally mindless "remove rcu_read_unlock() from
sem_unlock() and add it immediately after each caller" (except for the
aforementioned bug where we did too many rcu_read_unlock(), and in
find_alloc_undo() where we just got the rcu_read_lock() to correct for
the fact that sem_unlock would immediately drop it again).
We can (and should) clean things up further, but this fixes the bug with
the minimal amount of subtlety.
Reviewed-by: Davidlohr Bueso <davidlohr.bueso@hp.com>
Cc: Rik van Riel <riel@redhat.com>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2013-05-04 06:04:40 +08:00
|
|
|
rcu_read_unlock();
|
ipc,sem: do not hold ipc lock more than necessary
Instead of holding the ipc lock for permissions and security checks, among
others, only acquire it when necessary.
Some numbers....
1) With Rik's semop-multi.c microbenchmark we can see the following
results:
Baseline (3.9-rc1):
cpus 4, threads: 256, semaphores: 128, test duration: 30 secs
total operations: 151452270, ops/sec 5048409
+ 59.40% a.out [kernel.kallsyms] [k] _raw_spin_lock
+ 6.14% a.out [kernel.kallsyms] [k] sys_semtimedop
+ 3.84% a.out [kernel.kallsyms] [k] avc_has_perm_flags
+ 3.64% a.out [kernel.kallsyms] [k] __audit_syscall_exit
+ 2.06% a.out [kernel.kallsyms] [k] copy_user_enhanced_fast_string
+ 1.86% a.out [kernel.kallsyms] [k] ipc_lock
With this patchset:
cpus 4, threads: 256, semaphores: 128, test duration: 30 secs
total operations: 273156400, ops/sec 9105213
+ 18.54% a.out [kernel.kallsyms] [k] _raw_spin_lock
+ 11.72% a.out [kernel.kallsyms] [k] sys_semtimedop
+ 7.70% a.out [kernel.kallsyms] [k] ipc_has_perm.isra.21
+ 6.58% a.out [kernel.kallsyms] [k] avc_has_perm_flags
+ 6.54% a.out [kernel.kallsyms] [k] __audit_syscall_exit
+ 4.71% a.out [kernel.kallsyms] [k] ipc_obtain_object_check
2) While on an Oracle swingbench DSS (data mining) workload the
improvements are not as exciting as with Rik's benchmark, we can see
some positive numbers. For an 8 socket machine the following are the
percentages of %sys time incurred in the ipc lock:
Baseline (3.9-rc1):
100 swingbench users: 8,74%
400 swingbench users: 21,86%
800 swingbench users: 84,35%
With this patchset:
100 swingbench users: 8,11%
400 swingbench users: 19,93%
800 swingbench users: 77,69%
[riel@redhat.com: fix two locking bugs]
[sasha.levin@oracle.com: prevent releasing RCU read lock twice in semctl_main]
[akpm@linux-foundation.org: coding-style fixes]
Signed-off-by: Davidlohr Bueso <davidlohr.bueso@hp.com>
Signed-off-by: Rik van Riel <riel@redhat.com>
Reviewed-by: Chegu Vinod <chegu_vinod@hp.com>
Acked-by: Michel Lespinasse <walken@google.com>
Cc: Rik van Riel <riel@redhat.com>
Cc: Jason Low <jason.low2@hp.com>
Cc: Emmanuel Benisty <benisty.e@gmail.com>
Cc: Peter Hurley <peter@hurleysoftware.com>
Cc: Stanislav Kinsbursky <skinsbursky@parallels.com>
Tested-by: Sedat Dilek <sedat.dilek@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2013-05-01 10:15:29 +08:00
|
|
|
out_wakeup:
|
2010-05-27 05:43:41 +08:00
|
|
|
wake_up_sem_queue_do(&tasks);
|
2005-04-17 06:20:36 +08:00
|
|
|
out_free:
|
|
|
|
if(sops != fast_sops)
|
|
|
|
kfree(sops);
|
|
|
|
return error;
|
|
|
|
}
|
|
|
|
|
2009-01-14 21:14:27 +08:00
|
|
|
SYSCALL_DEFINE3(semop, int, semid, struct sembuf __user *, tsops,
|
|
|
|
unsigned, nsops)
|
2005-04-17 06:20:36 +08:00
|
|
|
{
|
|
|
|
return sys_semtimedop(semid, tsops, nsops, NULL);
|
|
|
|
}
|
|
|
|
|
|
|
|
/* If CLONE_SYSVSEM is set, establish sharing of SEM_UNDO state between
|
|
|
|
* parent and child tasks.
|
|
|
|
*/
|
|
|
|
|
|
|
|
int copy_semundo(unsigned long clone_flags, struct task_struct *tsk)
|
|
|
|
{
|
|
|
|
struct sem_undo_list *undo_list;
|
|
|
|
int error;
|
|
|
|
|
|
|
|
if (clone_flags & CLONE_SYSVSEM) {
|
|
|
|
error = get_undo_list(&undo_list);
|
|
|
|
if (error)
|
|
|
|
return error;
|
|
|
|
atomic_inc(&undo_list->refcnt);
|
|
|
|
tsk->sysvsem.undo_list = undo_list;
|
|
|
|
} else
|
|
|
|
tsk->sysvsem.undo_list = NULL;
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* add semadj values to semaphores, free undo structures.
|
|
|
|
* undo structures are not freed when semaphore arrays are destroyed
|
|
|
|
* so some of them may be out of date.
|
|
|
|
* IMPLEMENTATION NOTE: There is some confusion over whether the
|
|
|
|
* set of adjustments that needs to be done should be done in an atomic
|
|
|
|
* manner or not. That is, if we are attempting to decrement the semval
|
|
|
|
* should we queue up and wait until we can do so legally?
|
|
|
|
* The original implementation attempted to do this (queue and wait).
|
|
|
|
* The current implementation does not do so. The POSIX standard
|
|
|
|
* and SVID should be consulted to determine what behavior is mandated.
|
|
|
|
*/
|
|
|
|
void exit_sem(struct task_struct *tsk)
|
|
|
|
{
|
2008-07-25 16:48:04 +08:00
|
|
|
struct sem_undo_list *ulp;
|
2005-04-17 06:20:36 +08:00
|
|
|
|
2008-07-25 16:48:04 +08:00
|
|
|
ulp = tsk->sysvsem.undo_list;
|
|
|
|
if (!ulp)
|
2005-04-17 06:20:36 +08:00
|
|
|
return;
|
2008-04-29 16:00:57 +08:00
|
|
|
tsk->sysvsem.undo_list = NULL;
|
2005-04-17 06:20:36 +08:00
|
|
|
|
2008-07-25 16:48:04 +08:00
|
|
|
if (!atomic_dec_and_test(&ulp->refcnt))
|
2005-04-17 06:20:36 +08:00
|
|
|
return;
|
|
|
|
|
2008-07-25 16:48:06 +08:00
|
|
|
for (;;) {
|
2005-04-17 06:20:36 +08:00
|
|
|
struct sem_array *sma;
|
2008-07-25 16:48:06 +08:00
|
|
|
struct sem_undo *un;
|
2010-05-27 05:43:41 +08:00
|
|
|
struct list_head tasks;
|
2013-05-01 10:15:44 +08:00
|
|
|
int semid, i;
|
2008-07-25 16:48:04 +08:00
|
|
|
|
2008-07-25 16:48:06 +08:00
|
|
|
rcu_read_lock();
|
2009-04-15 02:17:16 +08:00
|
|
|
un = list_entry_rcu(ulp->list_proc.next,
|
|
|
|
struct sem_undo, list_proc);
|
2008-07-25 16:48:06 +08:00
|
|
|
if (&un->list_proc == &ulp->list_proc)
|
|
|
|
semid = -1;
|
|
|
|
else
|
|
|
|
semid = un->semid;
|
2008-07-25 16:48:04 +08:00
|
|
|
|
2013-05-01 10:15:44 +08:00
|
|
|
if (semid == -1) {
|
|
|
|
rcu_read_unlock();
|
2008-07-25 16:48:06 +08:00
|
|
|
break;
|
2013-05-01 10:15:44 +08:00
|
|
|
}
|
2005-04-17 06:20:36 +08:00
|
|
|
|
2013-05-01 10:15:44 +08:00
|
|
|
sma = sem_obtain_object_check(tsk->nsproxy->ipc_ns, un->semid);
|
2008-07-25 16:48:06 +08:00
|
|
|
/* exit_sem raced with IPC_RMID, nothing to do */
|
2013-05-01 10:15:44 +08:00
|
|
|
if (IS_ERR(sma)) {
|
|
|
|
rcu_read_unlock();
|
2008-07-25 16:48:06 +08:00
|
|
|
continue;
|
2013-05-01 10:15:44 +08:00
|
|
|
}
|
2005-04-17 06:20:36 +08:00
|
|
|
|
2013-05-01 10:15:44 +08:00
|
|
|
sem_lock(sma, NULL, -1);
|
2009-12-16 08:47:28 +08:00
|
|
|
un = __lookup_undo(ulp, semid);
|
2008-07-25 16:48:06 +08:00
|
|
|
if (un == NULL) {
|
|
|
|
/* exit_sem raced with IPC_RMID+semget() that created
|
|
|
|
* exactly the same semid. Nothing to do.
|
|
|
|
*/
|
2013-05-01 10:15:44 +08:00
|
|
|
sem_unlock(sma, -1);
|
ipc: move rcu_read_unlock() out of sem_unlock() and into callers
The IPC locking is a mess, and sem_unlock() unlocks not only the
semaphore spinlock, it also drops the rcu read lock. Unlike sem_lock(),
which just gets the spin-lock, and expects the caller to get the rcu
read lock.
This all makes things very hard to follow, and it's very confusing when
you take the rcu read lock in one function, and then release it in
another. And it has caused actual bugs: the sem_obtain_lock() function
ended up dropping the RCU read lock twice in one error path, because it
first did the sem_unlock(), and then did a rcu_read_unlock() to match
the rcu_read_lock() it had done.
This is just a totally mindless "remove rcu_read_unlock() from
sem_unlock() and add it immediately after each caller" (except for the
aforementioned bug where we did too many rcu_read_unlock(), and in
find_alloc_undo() where we just got the rcu_read_lock() to correct for
the fact that sem_unlock would immediately drop it again).
We can (and should) clean things up further, but this fixes the bug with
the minimal amount of subtlety.
Reviewed-by: Davidlohr Bueso <davidlohr.bueso@hp.com>
Cc: Rik van Riel <riel@redhat.com>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2013-05-04 06:04:40 +08:00
|
|
|
rcu_read_unlock();
|
2008-07-25 16:48:06 +08:00
|
|
|
continue;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* remove un from the linked lists */
|
2008-07-25 16:48:04 +08:00
|
|
|
assert_spin_locked(&sma->sem_perm.lock);
|
|
|
|
list_del(&un->list_id);
|
|
|
|
|
2008-07-25 16:48:06 +08:00
|
|
|
spin_lock(&ulp->lock);
|
|
|
|
list_del_rcu(&un->list_proc);
|
|
|
|
spin_unlock(&ulp->lock);
|
|
|
|
|
2008-07-25 16:48:04 +08:00
|
|
|
/* perform adjustments registered in un */
|
|
|
|
for (i = 0; i < sma->sem_nsems; i++) {
|
2006-03-26 17:37:17 +08:00
|
|
|
struct sem * semaphore = &sma->sem_base[i];
|
2008-07-25 16:48:04 +08:00
|
|
|
if (un->semadj[i]) {
|
|
|
|
semaphore->semval += un->semadj[i];
|
2005-04-17 06:20:36 +08:00
|
|
|
/*
|
|
|
|
* Range checks of the new semaphore value,
|
|
|
|
* not defined by sus:
|
|
|
|
* - Some unices ignore the undo entirely
|
|
|
|
* (e.g. HP UX 11i 11.22, Tru64 V5.1)
|
|
|
|
* - some cap the value (e.g. FreeBSD caps
|
|
|
|
* at 0, but doesn't enforce SEMVMX)
|
|
|
|
*
|
|
|
|
* Linux caps the semaphore value, both at 0
|
|
|
|
* and at SEMVMX.
|
|
|
|
*
|
|
|
|
* Manfred <manfred@colorfullife.com>
|
|
|
|
*/
|
2006-03-26 17:37:17 +08:00
|
|
|
if (semaphore->semval < 0)
|
|
|
|
semaphore->semval = 0;
|
|
|
|
if (semaphore->semval > SEMVMX)
|
|
|
|
semaphore->semval = SEMVMX;
|
2007-10-19 14:40:14 +08:00
|
|
|
semaphore->sempid = task_tgid_vnr(current);
|
2005-04-17 06:20:36 +08:00
|
|
|
}
|
|
|
|
}
|
|
|
|
/* maybe some queued-up processes were waiting for this */
|
2010-05-27 05:43:41 +08:00
|
|
|
INIT_LIST_HEAD(&tasks);
|
|
|
|
do_smart_update(sma, NULL, 0, 1, &tasks);
|
2013-05-01 10:15:44 +08:00
|
|
|
sem_unlock(sma, -1);
|
ipc: move rcu_read_unlock() out of sem_unlock() and into callers
The IPC locking is a mess, and sem_unlock() unlocks not only the
semaphore spinlock, it also drops the rcu read lock. Unlike sem_lock(),
which just gets the spin-lock, and expects the caller to get the rcu
read lock.
This all makes things very hard to follow, and it's very confusing when
you take the rcu read lock in one function, and then release it in
another. And it has caused actual bugs: the sem_obtain_lock() function
ended up dropping the RCU read lock twice in one error path, because it
first did the sem_unlock(), and then did a rcu_read_unlock() to match
the rcu_read_lock() it had done.
This is just a totally mindless "remove rcu_read_unlock() from
sem_unlock() and add it immediately after each caller" (except for the
aforementioned bug where we did too many rcu_read_unlock(), and in
find_alloc_undo() where we just got the rcu_read_lock() to correct for
the fact that sem_unlock would immediately drop it again).
We can (and should) clean things up further, but this fixes the bug with
the minimal amount of subtlety.
Reviewed-by: Davidlohr Bueso <davidlohr.bueso@hp.com>
Cc: Rik van Riel <riel@redhat.com>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2013-05-04 06:04:40 +08:00
|
|
|
rcu_read_unlock();
|
2010-05-27 05:43:41 +08:00
|
|
|
wake_up_sem_queue_do(&tasks);
|
2008-07-25 16:48:06 +08:00
|
|
|
|
2011-03-18 12:09:35 +08:00
|
|
|
kfree_rcu(un, rcu);
|
2005-04-17 06:20:36 +08:00
|
|
|
}
|
2008-07-25 16:48:04 +08:00
|
|
|
kfree(ulp);
|
2005-04-17 06:20:36 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
#ifdef CONFIG_PROC_FS
|
2005-09-07 06:17:10 +08:00
|
|
|
static int sysvipc_sem_proc_show(struct seq_file *s, void *it)
|
2005-04-17 06:20:36 +08:00
|
|
|
{
|
2012-02-08 08:54:11 +08:00
|
|
|
struct user_namespace *user_ns = seq_user_ns(s);
|
2005-09-07 06:17:10 +08:00
|
|
|
struct sem_array *sma = it;
|
|
|
|
|
|
|
|
return seq_printf(s,
|
2009-12-16 08:47:32 +08:00
|
|
|
"%10d %10d %4o %10u %5u %5u %5u %5u %10lu %10lu\n",
|
2005-09-07 06:17:10 +08:00
|
|
|
sma->sem_perm.key,
|
2007-10-19 14:40:48 +08:00
|
|
|
sma->sem_perm.id,
|
2005-09-07 06:17:10 +08:00
|
|
|
sma->sem_perm.mode,
|
|
|
|
sma->sem_nsems,
|
2012-02-08 08:54:11 +08:00
|
|
|
from_kuid_munged(user_ns, sma->sem_perm.uid),
|
|
|
|
from_kgid_munged(user_ns, sma->sem_perm.gid),
|
|
|
|
from_kuid_munged(user_ns, sma->sem_perm.cuid),
|
|
|
|
from_kgid_munged(user_ns, sma->sem_perm.cgid),
|
2005-09-07 06:17:10 +08:00
|
|
|
sma->sem_otime,
|
|
|
|
sma->sem_ctime);
|
2005-04-17 06:20:36 +08:00
|
|
|
}
|
|
|
|
#endif
|