diff options
Diffstat (limited to 'kernel')
-rw-r--r-- | kernel/futex.c | 176 |
1 files changed, 6 insertions, 170 deletions
diff --git a/kernel/futex.c b/kernel/futex.c index 98092c9817f4..449def8074fe 100644 --- a/kernel/futex.c +++ b/kernel/futex.c | |||
@@ -104,10 +104,6 @@ struct futex_q { | |||
104 | /* Key which the futex is hashed on: */ | 104 | /* Key which the futex is hashed on: */ |
105 | union futex_key key; | 105 | union futex_key key; |
106 | 106 | ||
107 | /* For fd, sigio sent using these: */ | ||
108 | int fd; | ||
109 | struct file *filp; | ||
110 | |||
111 | /* Optional priority inheritance state: */ | 107 | /* Optional priority inheritance state: */ |
112 | struct futex_pi_state *pi_state; | 108 | struct futex_pi_state *pi_state; |
113 | struct task_struct *task; | 109 | struct task_struct *task; |
@@ -126,9 +122,6 @@ struct futex_hash_bucket { | |||
126 | 122 | ||
127 | static struct futex_hash_bucket futex_queues[1<<FUTEX_HASHBITS]; | 123 | static struct futex_hash_bucket futex_queues[1<<FUTEX_HASHBITS]; |
128 | 124 | ||
129 | /* Futex-fs vfsmount entry: */ | ||
130 | static struct vfsmount *futex_mnt; | ||
131 | |||
132 | /* | 125 | /* |
133 | * Take mm->mmap_sem, when futex is shared | 126 | * Take mm->mmap_sem, when futex is shared |
134 | */ | 127 | */ |
@@ -610,8 +603,6 @@ lookup_pi_state(u32 uval, struct futex_hash_bucket *hb, | |||
610 | static void wake_futex(struct futex_q *q) | 603 | static void wake_futex(struct futex_q *q) |
611 | { | 604 | { |
612 | plist_del(&q->list, &q->list.plist); | 605 | plist_del(&q->list, &q->list.plist); |
613 | if (q->filp) | ||
614 | send_sigio(&q->filp->f_owner, q->fd, POLL_IN); | ||
615 | /* | 606 | /* |
616 | * The lock in wake_up_all() is a crucial memory barrier after the | 607 | * The lock in wake_up_all() is a crucial memory barrier after the |
617 | * plist_del() and also before assigning to q->lock_ptr. | 608 | * plist_del() and also before assigning to q->lock_ptr. |
@@ -988,14 +979,10 @@ out: | |||
988 | } | 979 | } |
989 | 980 | ||
990 | /* The key must be already stored in q->key. */ | 981 | /* The key must be already stored in q->key. */ |
991 | static inline struct futex_hash_bucket * | 982 | static inline struct futex_hash_bucket *queue_lock(struct futex_q *q) |
992 | queue_lock(struct futex_q *q, int fd, struct file *filp) | ||
993 | { | 983 | { |
994 | struct futex_hash_bucket *hb; | 984 | struct futex_hash_bucket *hb; |
995 | 985 | ||
996 | q->fd = fd; | ||
997 | q->filp = filp; | ||
998 | |||
999 | init_waitqueue_head(&q->waiters); | 986 | init_waitqueue_head(&q->waiters); |
1000 | 987 | ||
1001 | get_futex_key_refs(&q->key); | 988 | get_futex_key_refs(&q->key); |
@@ -1006,7 +993,7 @@ queue_lock(struct futex_q *q, int fd, struct file *filp) | |||
1006 | return hb; | 993 | return hb; |
1007 | } | 994 | } |
1008 | 995 | ||
1009 | static inline void __queue_me(struct futex_q *q, struct futex_hash_bucket *hb) | 996 | static inline void queue_me(struct futex_q *q, struct futex_hash_bucket *hb) |
1010 | { | 997 | { |
1011 | int prio; | 998 | int prio; |
1012 | 999 | ||
@@ -1041,15 +1028,6 @@ queue_unlock(struct futex_q *q, struct futex_hash_bucket *hb) | |||
1041 | * exactly once. They are called with the hashed spinlock held. | 1028 | * exactly once. They are called with the hashed spinlock held. |
1042 | */ | 1029 | */ |
1043 | 1030 | ||
1044 | /* The key must be already stored in q->key. */ | ||
1045 | static void queue_me(struct futex_q *q, int fd, struct file *filp) | ||
1046 | { | ||
1047 | struct futex_hash_bucket *hb; | ||
1048 | |||
1049 | hb = queue_lock(q, fd, filp); | ||
1050 | __queue_me(q, hb); | ||
1051 | } | ||
1052 | |||
1053 | /* Return 1 if we were still queued (ie. 0 means we were woken) */ | 1031 | /* Return 1 if we were still queued (ie. 0 means we were woken) */ |
1054 | static int unqueue_me(struct futex_q *q) | 1032 | static int unqueue_me(struct futex_q *q) |
1055 | { | 1033 | { |
@@ -1194,7 +1172,7 @@ static int futex_wait(u32 __user *uaddr, struct rw_semaphore *fshared, | |||
1194 | if (unlikely(ret != 0)) | 1172 | if (unlikely(ret != 0)) |
1195 | goto out_release_sem; | 1173 | goto out_release_sem; |
1196 | 1174 | ||
1197 | hb = queue_lock(&q, -1, NULL); | 1175 | hb = queue_lock(&q); |
1198 | 1176 | ||
1199 | /* | 1177 | /* |
1200 | * Access the page AFTER the futex is queued. | 1178 | * Access the page AFTER the futex is queued. |
@@ -1238,7 +1216,7 @@ static int futex_wait(u32 __user *uaddr, struct rw_semaphore *fshared, | |||
1238 | goto out_unlock_release_sem; | 1216 | goto out_unlock_release_sem; |
1239 | 1217 | ||
1240 | /* Only actually queue if *uaddr contained val. */ | 1218 | /* Only actually queue if *uaddr contained val. */ |
1241 | __queue_me(&q, hb); | 1219 | queue_me(&q, hb); |
1242 | 1220 | ||
1243 | /* | 1221 | /* |
1244 | * Now the futex is queued and we have checked the data, we | 1222 | * Now the futex is queued and we have checked the data, we |
@@ -1386,7 +1364,7 @@ static int futex_lock_pi(u32 __user *uaddr, struct rw_semaphore *fshared, | |||
1386 | goto out_release_sem; | 1364 | goto out_release_sem; |
1387 | 1365 | ||
1388 | retry_unlocked: | 1366 | retry_unlocked: |
1389 | hb = queue_lock(&q, -1, NULL); | 1367 | hb = queue_lock(&q); |
1390 | 1368 | ||
1391 | retry_locked: | 1369 | retry_locked: |
1392 | ret = lock_taken = 0; | 1370 | ret = lock_taken = 0; |
@@ -1499,7 +1477,7 @@ static int futex_lock_pi(u32 __user *uaddr, struct rw_semaphore *fshared, | |||
1499 | /* | 1477 | /* |
1500 | * Only actually queue now that the atomic ops are done: | 1478 | * Only actually queue now that the atomic ops are done: |
1501 | */ | 1479 | */ |
1502 | __queue_me(&q, hb); | 1480 | queue_me(&q, hb); |
1503 | 1481 | ||
1504 | /* | 1482 | /* |
1505 | * Now the futex is queued and we have checked the data, we | 1483 | * Now the futex is queued and we have checked the data, we |
@@ -1746,121 +1724,6 @@ pi_faulted: | |||
1746 | return ret; | 1724 | return ret; |
1747 | } | 1725 | } |
1748 | 1726 | ||
1749 | static int futex_close(struct inode *inode, struct file *filp) | ||
1750 | { | ||
1751 | struct futex_q *q = filp->private_data; | ||
1752 | |||
1753 | unqueue_me(q); | ||
1754 | kfree(q); | ||
1755 | |||
1756 | return 0; | ||
1757 | } | ||
1758 | |||
1759 | /* This is one-shot: once it's gone off you need a new fd */ | ||
1760 | static unsigned int futex_poll(struct file *filp, | ||
1761 | struct poll_table_struct *wait) | ||
1762 | { | ||
1763 | struct futex_q *q = filp->private_data; | ||
1764 | int ret = 0; | ||
1765 | |||
1766 | poll_wait(filp, &q->waiters, wait); | ||
1767 | |||
1768 | /* | ||
1769 | * plist_node_empty() is safe here without any lock. | ||
1770 | * q->lock_ptr != 0 is not safe, because of ordering against wakeup. | ||
1771 | */ | ||
1772 | if (plist_node_empty(&q->list)) | ||
1773 | ret = POLLIN | POLLRDNORM; | ||
1774 | |||
1775 | return ret; | ||
1776 | } | ||
1777 | |||
1778 | static const struct file_operations futex_fops = { | ||
1779 | .release = futex_close, | ||
1780 | .poll = futex_poll, | ||
1781 | }; | ||
1782 | |||
1783 | /* | ||
1784 | * Signal allows caller to avoid the race which would occur if they | ||
1785 | * set the sigio stuff up afterwards. | ||
1786 | */ | ||
1787 | static int futex_fd(u32 __user *uaddr, int signal) | ||
1788 | { | ||
1789 | struct futex_q *q; | ||
1790 | struct file *filp; | ||
1791 | int ret, err; | ||
1792 | struct rw_semaphore *fshared; | ||
1793 | static unsigned long printk_interval; | ||
1794 | |||
1795 | if (printk_timed_ratelimit(&printk_interval, 60 * 60 * 1000)) { | ||
1796 | printk(KERN_WARNING "Process `%s' used FUTEX_FD, which " | ||
1797 | "will be removed from the kernel in June 2007\n", | ||
1798 | current->comm); | ||
1799 | } | ||
1800 | |||
1801 | ret = -EINVAL; | ||
1802 | if (!valid_signal(signal)) | ||
1803 | goto out; | ||
1804 | |||
1805 | ret = get_unused_fd(); | ||
1806 | if (ret < 0) | ||
1807 | goto out; | ||
1808 | filp = get_empty_filp(); | ||
1809 | if (!filp) { | ||
1810 | put_unused_fd(ret); | ||
1811 | ret = -ENFILE; | ||
1812 | goto out; | ||
1813 | } | ||
1814 | filp->f_op = &futex_fops; | ||
1815 | filp->f_path.mnt = mntget(futex_mnt); | ||
1816 | filp->f_path.dentry = dget(futex_mnt->mnt_root); | ||
1817 | filp->f_mapping = filp->f_path.dentry->d_inode->i_mapping; | ||
1818 | |||
1819 | if (signal) { | ||
1820 | err = __f_setown(filp, task_pid(current), PIDTYPE_PID, 1); | ||
1821 | if (err < 0) { | ||
1822 | goto error; | ||
1823 | } | ||
1824 | filp->f_owner.signum = signal; | ||
1825 | } | ||
1826 | |||
1827 | q = kmalloc(sizeof(*q), GFP_KERNEL); | ||
1828 | if (!q) { | ||
1829 | err = -ENOMEM; | ||
1830 | goto error; | ||
1831 | } | ||
1832 | q->pi_state = NULL; | ||
1833 | |||
1834 | fshared = ¤t->mm->mmap_sem; | ||
1835 | down_read(fshared); | ||
1836 | err = get_futex_key(uaddr, fshared, &q->key); | ||
1837 | |||
1838 | if (unlikely(err != 0)) { | ||
1839 | up_read(fshared); | ||
1840 | kfree(q); | ||
1841 | goto error; | ||
1842 | } | ||
1843 | |||
1844 | /* | ||
1845 | * queue_me() must be called before releasing mmap_sem, because | ||
1846 | * key->shared.inode needs to be referenced while holding it. | ||
1847 | */ | ||
1848 | filp->private_data = q; | ||
1849 | |||
1850 | queue_me(q, ret, filp); | ||
1851 | up_read(fshared); | ||
1852 | |||
1853 | /* Now we map fd to filp, so userspace can access it */ | ||
1854 | fd_install(ret, filp); | ||
1855 | out: | ||
1856 | return ret; | ||
1857 | error: | ||
1858 | put_unused_fd(ret); | ||
1859 | put_filp(filp); | ||
1860 | ret = err; | ||
1861 | goto out; | ||
1862 | } | ||
1863 | |||
1864 | /* | 1727 | /* |
1865 | * Support for robust futexes: the kernel cleans up held futexes at | 1728 | * Support for robust futexes: the kernel cleans up held futexes at |
1866 | * thread exit time. | 1729 | * thread exit time. |
@@ -2092,10 +1955,6 @@ long do_futex(u32 __user *uaddr, int op, u32 val, ktime_t *timeout, | |||
2092 | case FUTEX_WAKE_BITSET: | 1955 | case FUTEX_WAKE_BITSET: |
2093 | ret = futex_wake(uaddr, fshared, val, val3); | 1956 | ret = futex_wake(uaddr, fshared, val, val3); |
2094 | break; | 1957 | break; |
2095 | case FUTEX_FD: | ||
2096 | /* non-zero val means F_SETOWN(getpid()) & F_SETSIG(val) */ | ||
2097 | ret = futex_fd(uaddr, val); | ||
2098 | break; | ||
2099 | case FUTEX_REQUEUE: | 1958 | case FUTEX_REQUEUE: |
2100 | ret = futex_requeue(uaddr, fshared, uaddr2, val, val2, NULL); | 1959 | ret = futex_requeue(uaddr, fshared, uaddr2, val, val2, NULL); |
2101 | break; | 1960 | break; |
@@ -2156,19 +2015,6 @@ asmlinkage long sys_futex(u32 __user *uaddr, int op, u32 val, | |||
2156 | return do_futex(uaddr, op, val, tp, uaddr2, val2, val3); | 2015 | return do_futex(uaddr, op, val, tp, uaddr2, val2, val3); |
2157 | } | 2016 | } |
2158 | 2017 | ||
2159 | static int futexfs_get_sb(struct file_system_type *fs_type, | ||
2160 | int flags, const char *dev_name, void *data, | ||
2161 | struct vfsmount *mnt) | ||
2162 | { | ||
2163 | return get_sb_pseudo(fs_type, "futex", NULL, FUTEXFS_SUPER_MAGIC, mnt); | ||
2164 | } | ||
2165 | |||
2166 | static struct file_system_type futex_fs_type = { | ||
2167 | .name = "futexfs", | ||
2168 | .get_sb = futexfs_get_sb, | ||
2169 | .kill_sb = kill_anon_super, | ||
2170 | }; | ||
2171 | |||
2172 | static int __init futex_init(void) | 2018 | static int __init futex_init(void) |
2173 | { | 2019 | { |
2174 | u32 curval; | 2020 | u32 curval; |
@@ -2193,16 +2039,6 @@ static int __init futex_init(void) | |||
2193 | spin_lock_init(&futex_queues[i].lock); | 2039 | spin_lock_init(&futex_queues[i].lock); |
2194 | } | 2040 | } |
2195 | 2041 | ||
2196 | i = register_filesystem(&futex_fs_type); | ||
2197 | if (i) | ||
2198 | return i; | ||
2199 | |||
2200 | futex_mnt = kern_mount(&futex_fs_type); | ||
2201 | if (IS_ERR(futex_mnt)) { | ||
2202 | unregister_filesystem(&futex_fs_type); | ||
2203 | return PTR_ERR(futex_mnt); | ||
2204 | } | ||
2205 | |||
2206 | return 0; | 2042 | return 0; |
2207 | } | 2043 | } |
2208 | __initcall(futex_init); | 2044 | __initcall(futex_init); |