diff --git a/include/linux/kevent.h b/include/linux/kevent.h new file mode 100644 index 0000000..d5995cf --- /dev/null +++ b/include/linux/kevent.h @@ -0,0 +1,201 @@ +/* + * kevent.h + * + * 2006 Copyright (c) Evgeniy Polyakov + * All rights reserved. + * + * This program is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + * GNU General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA + */ + +#ifndef __KEVENT_H +#define __KEVENT_H + +/* + * Kevent request flags. + */ + +#define KEVENT_REQ_ONESHOT 0x1 /* Process this event only once and then dequeue. */ + +/* + * Kevent return flags. + */ +#define KEVENT_RET_BROKEN 0x1 /* Kevent is broken. */ +#define KEVENT_RET_DONE 0x2 /* Kevent processing was finished successfully. */ + +struct kevent_id +{ + __u32 raw[2]; +}; + +struct ukevent +{ + struct kevent_id id; /* Id of this request, e.g. socket number, file descriptor and so on... */ + __u32 type; /* Event type, e.g. KEVENT_SOCK, KEVENT_INODE, KEVENT_TIMER and so on... */ + __u32 event; /* Event itself, e.g. SOCK_ACCEPT, INODE_CREATED, TIMER_FIRED... */ + __u32 req_flags; /* Per-event request flags */ + __u32 ret_flags; /* Per-event return flags */ + __u32 ret_data[2]; /* Event return data. Event originator fills it with anything it likes. */ + __u32 user[2]; /* User's data. It is not used, just copied to/from user. */ +}; + +struct kevent_user_control +{ + unsigned int cmd; /* Control command, e.g. KEVENT_ADD, KEVENT_REMOVE... */ + unsigned int num; /* Number of ukevents this strucutre controls. */ + unsigned int timeout; /* Timeout in milliseconds waiting for "ready_num" events to become ready. */ + unsigned int ready_num; /* Minimum number of ready events to inform about. */ +}; + +#define KEVENT_USER_SYMBOL 'K' +#define KEVENT_USER_CTL _IOWR(KEVENT_USER_SYMBOL, 0, struct kevent_user_control) +#define KEVENT_USER_WAIT _IOWR(KEVENT_USER_SYMBOL, 1, struct kevent_user_control) + +#ifdef __KERNEL__ + +#include +#include +#include +#include + +struct kqueue +{ + struct list_head list; /* List of queued kevents. */ + unsigned int qlen; +}; + +/* + * Kevent type set. + */ +enum { + KEVENT_SOCKET = 0, + KEVENT_INODE, + KEVENT_TIMER, + + KEVENT_MAX, +}; + +/* + * Per-type event sets. + * Number of per-event sets should be exactly as number of kevent types. + */ + +/* + * Timer events. + */ +enum { + KEVENT_TIMER_FIRED = 0, + + KEVENT_TIMER_MAX, +}; + +/* + * Socket events. + */ +enum { + KEVENT_SOCK_RECV = 0, + KEVENT_SOCK_ACCEPT, + + KEVENT_SOCK_MAX, +}; + +/* + * Inode events. + */ +enum { + KEVENT_INODE_CREATED = 0, + + KEVENT_INODE_MAX, +}; + +struct kevent; +struct kevent_storage; +typedef int (* kevent_callback_t)(struct kevent *); + +struct kevent +{ + struct ukevent event; + spinlock_t lock; /* This lock protects ukevent manipulations, e.g. ret_flags changes. */ + + struct list_head kevent_entry; /* Entry of user's queue. */ + struct list_head entry; /* Entry of origin's queue or user's ready queue. */ + + struct kevent_user *user; /* User who requested this kevent. */ + struct kevent_storage *st; /* Kevent container. */ + + kevent_callback_t callback; /* Is called each time new event has been caught. */ + kevent_callback_t enqueue; /* Is called each time new event is queued. */ + kevent_callback_t dequeue; /* Is called each time new event is dequeued. */ +}; + +struct kevent_user +{ + struct kqueue queue; /* List of all kevents from this user. */ + spinlock_t lock; /* Protects all manipulations with queue of kevents. */ + + struct kqueue ready_queue; /* Kevent is dequeued from storage->queue and placed here when it is ready. */ + spinlock_t ready_lock; /* Protects all manipulations with ready queue. */ + + unsigned int max_ready_num; /* Requested number of kevents. */ + + struct semaphore ctl_mutex; /* Protects against simultaneous kevent_user control manipulations. */ + struct semaphore wait_mutex; /* Protects against simultaneous kevent_user waits. */ + wait_queue_head_t wait; /* Wait until some events are ready. */ + + atomic_t refcnt; /* Reference counter, increased for each new kevent. */ +}; + +struct kevent_storage +{ + __u32 type; /* Event type, e.g. KEVENT_SOCK, KEVENT_INODE, KEVENT_TIMER and so on... */ + __u32 event; /* Event itself, e.g. SOCK_ACCEPT, INODE_CREATED, TIMER_FIRED... */ + void *origin; /* Originator's pointer, e.g. struct sock or struct file. Can be NULL. */ + struct kqueue queue; /* Queue of registered users. */ + spinlock_t lock; /* Protects users queue. */ +}; + +#define KEVENT_MAX_REQUESTS PAGE_SIZE/sizeof(struct kevent) + +int kevent_enqueue(struct kevent *k); +int kevent_dequeue(struct kevent *k); +int kevent_requeue(struct kevent *k); +int kevent_init(struct kevent *k); +struct kevent *__kqueue_dequeue_one(struct kqueue *q); +void __kqueue_dequeue_kevent(struct kqueue *q, struct kevent *k); +void __kqueue_enqueue_kevent(struct kqueue *q, struct kevent *k); +int kqueue_init(struct kqueue *q); +int kevent_storage_enqueue(struct kevent_storage *st, struct kevent *k); +void kevent_storage_dequeue(struct kevent_storage *st, struct kevent *k); +void kevent_storage_ready(struct kevent_storage *st, kevent_callback_t ready_callback); +int kevent_storage_init(__u32 type, __u32 event, void *origin, struct kevent_storage *st); +void kevent_storage_fini(struct kevent_storage *st); + +#define list_for_each_entry_reverse_safe(pos, n, head, member) \ + for (pos = list_entry((head)->prev, typeof(*pos), member), \ + n = list_entry(pos->member.prev, typeof(*pos), member); \ + prefetch(pos->member.prev), &pos->member != (head); \ + pos = n, n = list_entry(pos->member.prev, typeof(*pos), member)) + +int kevent_break(struct kevent *k); +int kevent_init(struct kevent *k); +int kevent_init_socket(struct kevent *k); +int kevent_init_inode(struct kevent *k); +int kevent_init_timer(struct kevent *k); + +void kevent_storage_ready(struct kevent_storage *st, kevent_callback_t ready_callback); +int kevent_storage_init(__u32 type, __u32 event, void *origin, struct kevent_storage *st); +void kevent_storage_fini(struct kevent_storage *st); + +#endif /* __KERNEL__ */ +#endif /* __KEVENT_H */ diff --git a/init/Kconfig b/init/Kconfig index 9fc0759..4a4f26c 100644 --- a/init/Kconfig +++ b/init/Kconfig @@ -224,6 +224,8 @@ config KOBJECT_UEVENT Say Y, unless you are building a system requiring minimal memory consumption. +source "kernel/kevent/Kconfig" + config IKCONFIG bool "Kernel .config support" ---help--- diff --git a/kernel/Makefile b/kernel/Makefile index 4f5a145..7c5aa88 100644 --- a/kernel/Makefile +++ b/kernel/Makefile @@ -32,6 +32,7 @@ obj-$(CONFIG_GENERIC_HARDIRQS) += irq/ obj-$(CONFIG_CRASH_DUMP) += crash_dump.o obj-$(CONFIG_SECCOMP) += seccomp.o obj-$(CONFIG_RCU_TORTURE_TEST) += rcutorture.o +obj-$(CONFIG_KEVENT) += kevent/ ifneq ($(CONFIG_SCHED_NO_NO_OMIT_FRAME_POINTER),y) # According to Alan Modra , the -fno-omit-frame-pointer is diff --git a/kernel/kevent/Kconfig b/kernel/kevent/Kconfig new file mode 100644 index 0000000..675bfe1 --- /dev/null +++ b/kernel/kevent/Kconfig @@ -0,0 +1,20 @@ +config KEVENT + bool "Kernel event notification mechanism" + help + This option enables event queue mechanism. + It can be used as replacement for poll()/select(), AIO callback invocations, + advanced timer notifications and other kernel object status changes. + +config KEVENT_SOCKET + bool "Kernel event notifications for sockets" + depends on NET && KEVENT + help + This option enables notifications through KEVENT subsystem of + sockets operations, like new packet receiving conditions, ready for accept + conditions and so on. + +config KEVENT_TIMER + bool "Kernel event notifications for timers" + depends on KEVENT + help + This option allows to use timers through KEVENT subsystem. diff --git a/kernel/kevent/Makefile b/kernel/kevent/Makefile new file mode 100644 index 0000000..b7999e4 --- /dev/null +++ b/kernel/kevent/Makefile @@ -0,0 +1,4 @@ +obj-y := kevent.o kevent_user.o kevent_init.o +obj-$(CONFIG_KEVENT_SOCKET) += kevent_socket.o +obj-$(CONFIG_KEVENT_INODE) += kevent_inode.o +obj-$(CONFIG_KEVENT_TIMER) += kevent_timer.o diff --git a/kernel/kevent/kevent.c b/kernel/kevent/kevent.c new file mode 100644 index 0000000..347e389 --- /dev/null +++ b/kernel/kevent/kevent.c @@ -0,0 +1,252 @@ +/* + * kevent.c + * + * 2006 Copyright (c) Evgeniy Polyakov + * All rights reserved. + * + * This program is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + * GNU General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA + */ + +#include +#include +#include +#include +#include +#include +#include +#include +#include + +/* + * Attempts to add an event into appropriate origin's queue. + * Returns positive value if this event is ready immediately, + * negative value in case of error and zero if event has been queued. + * If event is ready it is not queued into origin's queue. + * ->enqueue() callback must increase origin's reference counter. + */ +int kevent_enqueue(struct kevent *k) +{ + if (k->event.type >= KEVENT_MAX) + return -E2BIG; + + if (!k->enqueue) { + kevent_break(k); + return -EINVAL; + } + + return k->enqueue(k); +} + +/* + * Remove event from the appropriate queue. + * ->dequeue() callback must decrease origin's reference counter. + */ +int kevent_dequeue(struct kevent *k) +{ + if (k->event.type >= KEVENT_MAX) + return -E2BIG; + + if (!k->dequeue) { + kevent_break(k); + return -EINVAL; + } + + return k->dequeue(k); +} + +/* + * Reference for storage is being held until we call kevent_dequeue() and + * thus ->dequeue() callback which will decrement origin's reference counter, + * so kevent->st pointer is still valid. + */ +int kevent_requeue(struct kevent *k) +{ + unsigned long flags; + + spin_lock_irqsave(&k->st->lock, flags); + spin_lock(&k->user->lock); + __kqueue_enqueue_kevent(&k->st->queue, k); + spin_unlock(&k->user->lock); + spin_unlock_irqrestore(&k->st->lock, flags); + return 0; +} + +/* + * Must be called before event is going to be added into some origin's queue. + * Initializes ->enqueue(), ->dequeue() and ->callback() callbacks. + * If failed, kevent should not be used or kevent_add() will fail to add + * this kevent into origin's queue with setting + * KEVENT_RET_BROKEN flag in kevent->event.ret_flags. + */ +int kevent_init(struct kevent *k) +{ + int err; + + spin_lock_init(&k->lock); + k->kevent_entry.next = LIST_POISON1; + k->entry.next = LIST_POISON1; + + if (k->event.type >= KEVENT_MAX) + return -E2BIG; + + switch (k->event.type) { + case KEVENT_SOCKET: + err = kevent_init_socket(k); + break; + case KEVENT_INODE: + err = kevent_init_inode(k); + break; + case KEVENT_TIMER: + err = kevent_init_timer(k); + break; + default: + err = -ENODEV; + } + + return err; +} + +/* + * Dequeue one entry from either origin's queue or user's ready queue. + */ +struct kevent *__kqueue_dequeue_one(struct kqueue *q) +{ + struct kevent *k = NULL; + + if (q->qlen) { + k = list_entry(q->list.next, struct kevent, entry); + list_del(&k->entry); + q->qlen--; + } + + return k; +} + +/* + * Dequeue kevent from either origin's queue or user's ready queue. + */ +void __kqueue_dequeue_kevent(struct kqueue *q, struct kevent *k) +{ + BUG_ON(k->entry.next == LIST_POISON1); + + list_del(&k->entry); + q->qlen--; +} + +/* + * Dequeue kevent into either origin's queue or user's ready queue. + */ +void __kqueue_enqueue_kevent(struct kqueue *q, struct kevent *k) +{ + BUG_ON(k->entry.next != LIST_POISON1); + + list_add_tail(&k->entry, &q->list); + q->qlen++; +} + +int kqueue_init(struct kqueue *q) +{ + INIT_LIST_HEAD(&q->list); + q->qlen = 0; + + return 0; +} + +/* + * Called from ->enqueue() callback when reference counter for given + * origin (socket, inode...) has been increased. + */ +int kevent_storage_enqueue(struct kevent_storage *st, struct kevent *k) +{ + unsigned long flags; + + k->st = st; + spin_lock_irqsave(&st->lock, flags); + __kqueue_enqueue_kevent(&st->queue, k); + spin_unlock_irqrestore(&st->lock, flags); + + return 0; +} + +/* + * Dequeue kevent from origin's queue. It does not decrese origin's reference counter in any way. + * It is called from ->dequeue() callback. + */ +void kevent_storage_dequeue(struct kevent_storage *st, struct kevent *k) +{ + unsigned long flags; + + spin_lock_irqsave(&st->lock, flags); + /* + * It looks a bit ugly, but there is notthing wrong with this assumption here. + * kevent_storage_dequeue() is only called from ->dequeue() callback, + * which in turn is called from kevent_dequeue() when we want to dequeue + * some kevent from either origin's or user's ready queue. + * But if kevent was in user's ready queue it can not be in storage queue, + * and this will trigger BUG_ON() check in __kqueue_dequeue_kevent(). + * Probably there should be some flag in kevent which states from which + * queue we dequeue this kevent, but I do not think it is too ugly + * to check here if kevent is still in the queue. + */ + if (k->entry.next != LIST_POISON1) + __kqueue_dequeue_kevent(&st->queue, k); + spin_unlock_irqrestore(&st->lock, flags); +} + +/* + * Called each time some activity in origin (socket, inode...) is noticed. + */ +void kevent_storage_ready(struct kevent_storage *st, kevent_callback_t ready_callback) +{ + unsigned long flags; + struct kevent *k, *n; + int err; + + spin_lock_irqsave(&st->lock, flags); + list_for_each_entry_safe(k, n, &st->queue.list, entry) { + if (ready_callback) + ready_callback(k); + + err = k->callback(k); + if (err < 0) + kevent_break(k); + if (err) { + spin_lock(&k->lock); + k->event.ret_flags |= KEVENT_RET_DONE; + spin_unlock(&k->lock); + + spin_lock(&k->user->ready_lock); + __kqueue_dequeue_kevent(&st->queue, k); + __kqueue_enqueue_kevent(&k->user->ready_queue, k); + wake_up(&k->user->wait); + spin_unlock(&k->user->ready_lock); + } + } + spin_unlock_irqrestore(&st->lock, flags); +} + +int kevent_storage_init(__u32 type, __u32 event, void *origin, struct kevent_storage *st) +{ + spin_lock_init(&st->lock); + st->type = type; + st->event = event; + st->origin = origin; + return kqueue_init(&st->queue); +} + +void kevent_storage_fini(struct kevent_storage *st) +{ + kevent_storage_ready(st, kevent_break); +} diff --git a/kernel/kevent/kevent_init.c b/kernel/kevent/kevent_init.c new file mode 100644 index 0000000..d97a323 --- /dev/null +++ b/kernel/kevent/kevent_init.c @@ -0,0 +1,62 @@ +/* + * kevent_init.c + * + * 2006 Copyright (c) Evgeniy Polyakov + * All rights reserved. + * + * This program is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + * GNU General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA + */ + +#include +#include +#include +#include +#include +#include + +int kevent_break(struct kevent *k) +{ + unsigned long flags; + + spin_lock_irqsave(&k->lock, flags); + k->event.ret_flags |= KEVENT_RET_BROKEN; + spin_unlock_irqrestore(&k->lock, flags); + printk("%s: k=%p.\n", __func__, k); + return 0; +} + +#ifndef CONFIG_KEVENT_SOCKET +int kevent_init_socket(struct kevent *k) +{ + kevent_break(k); + return -ENODEV; +} +#endif + +#ifndef CONFIG_KEVENT_INODE +int kevent_init_inode(struct kevent *k) +{ + kevent_break(k); + return -ENODEV; +} +#endif + +#ifndef CONFIG_KEVENT_TIMER +int kevent_init_timer(struct kevent *k) +{ + kevent_break(k); + return -ENODEV; +} +#endif diff --git a/kernel/kevent/kevent_timer.c b/kernel/kevent/kevent_timer.c new file mode 100644 index 0000000..402f5eb --- /dev/null +++ b/kernel/kevent/kevent_timer.c @@ -0,0 +1,112 @@ +/* + * kevent_timer.c + * + * 2006 Copyright (c) Evgeniy Polyakov + * All rights reserved. + * + * This program is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + * GNU General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA + */ + +#include +#include +#include +#include +#include +#include +#include +#include + +static void kevent_timer_func(unsigned long data) +{ + struct kevent *k = (struct kevent *)data; + struct timer_list *t = k->st->origin; + + kevent_storage_ready(k->st, NULL); + mod_timer(t, jiffies + msecs_to_jiffies(k->event.id.raw[0])); +} + +static int kevent_timer_enqueue(struct kevent *k) +{ + struct timer_list *t; + struct kevent_storage *st; + int err; + + t = kmalloc(sizeof(struct timer_list) + sizeof(struct kevent_storage), GFP_KERNEL); + if (!t) + return -ENOMEM; + + init_timer(t); + t->function = kevent_timer_func; + t->expires = jiffies + msecs_to_jiffies(k->event.id.raw[0]); + t->data = (unsigned long)k; + + st = (struct kevent_storage *)(t+1); + err = kevent_storage_init(k->event.type, k->event.event, t, st); + if (err) + goto err_out_free; + + err = kevent_storage_enqueue(st, k); + if (err) + goto err_out_st_fini; + + add_timer(t); + + return 0; + +err_out_st_fini: + kevent_storage_fini(st); +err_out_free: + kfree(t); + + return err; +} + +static int kevent_timer_dequeue(struct kevent *k) +{ + struct kevent_storage *st = k->st; + struct timer_list *t = st->origin; + + if (!t) + return -ENODEV; + + del_timer_sync(t); + + kevent_storage_dequeue(st, k); + + kfree(t); + + return 0; +} + +static int kevent_timer_callback(struct kevent *k) +{ + struct kevent_storage *st = k->st; + struct timer_list *t = st->origin; + + if (!t) + return -ENODEV; + + k->event.ret_data[0] = (__u32)jiffies; + k->event.ret_data[1] = (__u32)((unsigned long)t & 0xffffffff); + return 1; +} + +int kevent_init_timer(struct kevent *k) +{ + k->enqueue = &kevent_timer_enqueue; + k->dequeue = &kevent_timer_dequeue; + k->callback = &kevent_timer_callback; + return 0; +} diff --git a/kernel/kevent/kevent_user.c b/kernel/kevent/kevent_user.c new file mode 100644 index 0000000..559aa58 --- /dev/null +++ b/kernel/kevent/kevent_user.c @@ -0,0 +1,340 @@ +/* + * kevent_user.c + * + * 2006 Copyright (c) Evgeniy Polyakov + * All rights reserved. + * + * This program is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + * GNU General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA + */ + +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include + +static struct class *kevent_user_class; +static char kevent_name[] = "kevent"; +static int kevent_user_major; + +static int kevent_user_open(struct inode *, struct file *); +static int kevent_user_release(struct inode *, struct file *); +static int kevent_user_ioctl(struct inode *, struct file *, unsigned int, unsigned long); + +static struct file_operations kevent_user_fops = { + .open = kevent_user_open, + .release = kevent_user_release, + .ioctl = kevent_user_ioctl, + .owner = THIS_MODULE, +}; + +static int kevent_user_open(struct inode *inode, struct file *file) +{ + struct kevent_user *u; + + u = kzalloc(sizeof(struct kevent_user), GFP_KERNEL); + if (!u) + return -ENOMEM; + + kqueue_init(&u->queue); + spin_lock_init(&u->lock); + + kqueue_init(&u->ready_queue); + spin_lock_init(&u->ready_lock); + + init_MUTEX(&u->ctl_mutex); + init_MUTEX(&u->wait_mutex); + init_waitqueue_head(&u->wait); + u->max_ready_num = 0; + + atomic_set(&u->refcnt, 1); + + file->private_data = u; + + return 0; +} + +static inline void kevent_user_get(struct kevent_user *u) +{ + atomic_inc(&u->refcnt); +} + +static inline void kevent_user_put(struct kevent_user *u) +{ + if (atomic_dec_and_test(&u->refcnt)) + kfree(u); +} + +/* + * Remove kevent from user->queue, dequeue it from storage and decrease user's reference counter, + * since this kevent does not exist anymore. That is why it is freed here. + */ +static void kevent_finish_user(struct kevent *k, int lock) +{ + struct kevent_user *u = k->user; + unsigned long flags; + + if (lock) { + spin_lock_irqsave(&u->lock, flags); + list_del(&k->kevent_entry); + u->queue.qlen--; + spin_unlock_irqrestore(&u->lock, flags); + } else { + list_del(&k->kevent_entry); + u->queue.qlen--; + } + kevent_dequeue(k); + kevent_user_put(u); + kfree(k); +} + +static struct kevent *kqueue_dequeue_ready(struct kevent_user *u) +{ + unsigned long flags; + struct kevent *k; + + spin_lock_irqsave(&u->ready_lock, flags); + k = __kqueue_dequeue_one(&u->ready_queue); + spin_unlock_irqrestore(&u->ready_lock, flags); + + return k; +} + +/* + * No new entry can be added or removed from any list at this point. + * It is not permitted to call ->ioctl() and ->release() in parallel. + */ +static int kevent_user_release(struct inode *inode, struct file *file) +{ + struct kevent_user *u = file->private_data; + struct kevent *k, *n; + unsigned long flags; + + spin_lock_irqsave(&u->lock, flags); + list_for_each_entry_safe(k, n, &u->queue.list, kevent_entry) + kevent_finish_user(k, 0); + spin_unlock_irqrestore(&u->lock, flags); + + kevent_user_put(u); + file->private_data = NULL; + + return 0; +} + +/* + * Copy all ukevents from userspace, allocate kevent for each one and add them + * into appropriate kevent_storages, e.g. sockets, inodes and so on... + * If something goes wrong, all events will be dequeued and negative error will be returned. + * On success zero is returned and ctl->num will be a number of finished events, either completed + * or failed. Array of finished events (struct ukevent) will be placed behind kevent_user_control structure. + * User must run through that array and check ret_flags field of each ukevent structure + * to determine if it is fired or failed event. + */ +static int kevent_user_ctl(struct kevent_user *u, struct kevent_user_control *ctl, unsigned long arg) +{ + int err = 0, cerr = 0, num = 0, knum = 0, i; + struct kevent *k, *n; + void __user *orig, *ctl_addr; + unsigned long flags; + + if (down_interruptible(&u->ctl_mutex)) + return -ERESTARTSYS; + + orig = (void __user *)arg; + ctl_addr = (void __user *)(arg - sizeof(struct kevent_user_control)); + + for (i=0; inum; ++i) { + k = kzalloc(sizeof(struct kevent), GFP_KERNEL); + if (!k) { + err = -ENOMEM; + break; + } + + if (copy_from_user(&k->event, (const void __user *)arg, sizeof(struct ukevent))) { + kfree(k); + err = -EINVAL; + break; + } + + arg += sizeof(struct ukevent); + + err = kevent_init(k); + if (err) { + kfree(k); + break; + } + k->user = u; + + err = kevent_enqueue(k); + if (err) { + if (copy_to_user(orig, &k->event, sizeof(struct ukevent))) + cerr = -EINVAL; + + orig += sizeof(struct ukevent); + kfree(k); + num++; + } else { + spin_lock_irqsave(&u->lock, flags); + list_add_tail(&k->kevent_entry, &u->queue.list); + u->queue.qlen++; + kevent_user_get(u); + spin_unlock_irqrestore(&u->lock, flags); + knum++; + } + } + + if (err) + goto err_out_remove; + + ctl->num = num; + if (copy_to_user(ctl_addr, ctl, sizeof(struct kevent_user_control))) + cerr = -EINVAL; + + if (cerr) + err = cerr; + +err_out_remove: + if (err && knum) { + num = 0; + /* + * Dequeue the rest of successfully added kevents. + */ + spin_lock_irqsave(&u->lock, flags); + list_for_each_entry_reverse_safe(k, n, &u->queue.list, kevent_entry) { + kevent_finish_user(k, 0); + + if (++num == knum) + break; + } + spin_unlock_irqrestore(&u->lock, flags); + } + up(&u->ctl_mutex); + + return err; +} + +/* + * Waits until at least ctl->ready_num events are ready or timeout and returns + * number of ready events (in case of timeout) or number of requested events. + */ +static int kevent_user_wait(struct kevent_user *u, struct kevent_user_control *ctl, unsigned long arg) +{ + struct kevent *k; + int cerr = 0, num = 0; + void __user *ptr = (void __user *)(arg + sizeof(struct kevent_user_control)); + + if (down_interruptible(&u->wait_mutex)) + return -ERESTARTSYS; + + wait_event_interruptible_timeout(u->wait, + u->ready_queue.qlen >= ctl->ready_num, msecs_to_jiffies(ctl->timeout)); + while (num < ctl->num && (k = kqueue_dequeue_ready(u))) { + if (copy_to_user(ptr + num*sizeof(struct ukevent), &k->event, sizeof(struct ukevent))) + cerr = -EINVAL; + /* + * If it is one-shot kevent, it has been removed already from + * origin's queue, so we can easily free it here. + */ + if (k->event.req_flags & KEVENT_REQ_ONESHOT) + kevent_finish_user(k, 1); + else + kevent_requeue(k); + ++num; + } + + ctl->num = num; + if (copy_to_user((void __user *)arg, ctl, sizeof(struct kevent_user_control))) + cerr = -EINVAL; + + up(&u->wait_mutex); + + return (cerr)?cerr:num; +} + +static int kevent_user_ioctl(struct inode *inode, struct file *file, unsigned int cmd, unsigned long arg) +{ + int err = -ENODEV; + struct kevent_user_control ctl; + struct kevent_user *u = file->private_data; + + if (copy_from_user(&ctl, (const void __user *)arg, sizeof(struct kevent_user_control))) + return -EINVAL; + + switch (cmd) { + case KEVENT_USER_CTL: + err = kevent_user_ctl(u, &ctl, arg+sizeof(struct kevent_user_control)); + break; + case KEVENT_USER_WAIT: + err = kevent_user_wait(u, &ctl, arg); + default: + break; + } + + return err; +} + +static int __devinit kevent_user_init(void) +{ + struct class_device *dev; + int err = 0; + + kevent_user_major = register_chrdev(0, kevent_name, &kevent_user_fops); + if (kevent_user_major < 0) { + printk(KERN_ERR "Failed to register \"%s\" char device: err=%d.\n", kevent_name, kevent_user_major); + return -ENODEV; + } + + kevent_user_class = class_create(THIS_MODULE, "kevent"); + if (IS_ERR(kevent_user_class)) { + printk(KERN_ERR "Failed to register \"%s\" class: err=%ld.\n", kevent_name, PTR_ERR(kevent_user_class)); + err = PTR_ERR(kevent_user_class); + goto err_out_unregister; + } + + dev = class_device_create(kevent_user_class, NULL, MKDEV(kevent_user_major, 0), NULL, kevent_name); + if (IS_ERR(dev)) { + printk(KERN_ERR "Failed to create %d.%d class device in \"%s\" class: err=%ld.\n", + kevent_user_major, 0, kevent_name, PTR_ERR(dev)); + err = PTR_ERR(dev); + goto err_out_class_destroy; + } + + printk("KEVENT subsystem: chardev helper: major=%d.\n", kevent_user_major); + + return 0; + +err_out_class_destroy: + class_destroy(kevent_user_class); +err_out_unregister: + unregister_chrdev(kevent_user_major, kevent_name); + + return err; +} + +static void __devexit kevent_user_fini(void) +{ + class_device_destroy(kevent_user_class, MKDEV(kevent_user_major, 0)); + class_destroy(kevent_user_class); + unregister_chrdev(kevent_user_major, kevent_name); +} + +module_init(kevent_user_init); +module_exit(kevent_user_fini);