[take35 0/10] kevent: Generic event handling mechanism.

[take35 0/10] kevent: Generic event handling mechanism.

[Evgeniy Polyakov]

Generic event handling mechanism.

Kevent is a generic subsytem which allows to handle event notifications.
It supports both level and edge triggered events. It is similar to
poll/epoll in some cases, but it is more scalable, it is faster and
allows to work with essentially eny kind of events.

Events are provided into kernel through control syscall and can be read
back through ring buffer or using usual syscalls.
Kevent update (i.e. readiness switching) happens directly from internals
of the appropriate state machine of the underlying subsytem (like
network, filesystem, timer or any other).

Homepage:
http://tservice.net.ru/~s0mbre/old/?section=projects&item=kevent

Documentation page:
http://linux-net.osdl.org/index.php/Kevent

Consider for inclusion.

P.S. If you want to be removed from Cc: list just drop me a mail.

Changes from 'take34' patchset:
 * Ported to the 2.6.20-rc7 (9be5b038b1c9d1927c367bf91683458e10d5d4eb) tree.

Changes from 'take33' patchset:
 * Added optional header pointer and its size into aio_sendfile_path(), 
   which allows to send header and file in one syscall instead of
   send(header), open file, sendfile(file).

Changes from 'take32' patchset:
 * Updated documentation (aio_sendfile_path()).
 * Fixed typo in forward declaration.

Changes from 'take31' patchset:
 * Added aio_sendfile_path() - this syscall allows to asynchronosly transfer
   file specified by provided pathname to destination socket.
   Opened file descriptor is returned.
 * Added trivial scheduler which selects execution thread. It allows
   to specify given thread 'by-hands', but since kaio provides '-1' it uses
   round-robin to get processing thread. In theory it can be bound to
   scheduler statistics or gamma-ray receiver data.
 * Number of bug fixes in kevent based AIO mpage_readpages().
   
   Benchmark of the 100 1Mb files transfer (files are in VFS already) using 
   sync sendfile or this new version shows about 10Mb/sec performance win
   for aio_sendfile_path().

Changes from 'take30' patchset:
 * AIO state machine.
 * aio_sendfile() implementation.
 * moved kevent_user_get/kevent_user_put into header.
 * use *zalloc where needed.

Changes from 'take29' patchset:
 * new private userspace notifications - allows to queue any userspace private 
 	event and then mark it as ready using kevent_ctl(KEVENT_READY) command
 * KEVENT_REQ_READY flag - if set kevent will be marked as ready at enqueue time
 * port to 2.6.20-rc2 tree (54abb5fcdae74a811ed440ec6556cabc6b24f404 commit)
 * use struct kmem_cache instead of kmem_cache_t
 * added notificaion type into search key, this allows to have the same id for 
 	different types of notifications

Changes from 'take28' patchset:
 * optimized af_unix to use socket notifications
 * changed ALWAYS_QUEUE behaviour with poll/select notifications - previously
 	kevent was not queued into poll wait queue when ALWAYS_QUEUE flag
	is set
 * added KEVENT_POLL_POLLRDHUP definition into ukevent.h header
 * libevent-1.2 patch (Jamal, your request is completed, so I'm waiting two weeks
 	before starting final countdown :)
	All regression tests passed successfully except test_evbuffer(), which is
	crashed on my amd64 linux 2.6 test machine for all types of notifications,
	probably it was fixed in libevent-1.2a version, I did not check.
	Patch and README can be found at project homepage.

Changes from 'take27' patchset:
 * made kevent default yes in non embedded case.
 * added falgs to callback structures - currently used to check if kevent
 	can be requested from kernelspace only (posix timers) or 
	userspace (all others)

Changes from 'take26' patchset:
 * made kevent visible in config only in case of embedded setup.
 * added comment about KEVENT_MAX number.
 * spell fix.

Changes from 'take25' patchset:
 * use timespec as timeout parameter.
 * added high-resolution timer to handle absolute timeouts.
 * added flags to waiting and initialization syscalls.
 * kevent_commit() has new_uidx parameter.
 * kevent_wait() has old_uidx parameter, which, if not equal to u->uidx,
 	results in immediate wakeup (usefull for the case when entries
	are added asynchronously from kernel (not supported for now)).
 * added interface to mark any event as ready.
 * event POSIX timers support.
 * return -ENOSYS if there is no registered event type.
 * provided file descriptor must be checked for fifo type (spotted by Eric Dumazet).
 * signal notifications.
 * documentation update.
 * lighttpd patch updated (the latest benchmarks with lighttpd patch can be found in blog).

Changes from 'take24' patchset:
 * new (old (new)) ring buffer implementation with kernel and user indexes.
 * added initialization syscall instead of opening /dev/kevent
 * kevent_commit() syscall to commit ring buffer entries
 * changed KEVENT_REQ_WAKEUP_ONE flag to KEVENT_REQ_WAKEUP_ALL, kevent wakes
   only first thread always if that flag is not set
 * KEVENT_REQ_ALWAYS_QUEUE flag. If set, kevent will be queued into ready queue
   instead of copying back to userspace when kevent is ready immediately when
   it is added.
 * lighttpd patch (Hail! Although nothing really outstanding compared to epoll)

Changes from 'take23' patchset:
 * kevent PIPE notifications
 * KEVENT_REQ_LAST_CHECK flag, which allows to perform last check at dequeueing time
 * fixed poll/select notifications (were broken due to tree manipulations)
 * made Documentation/kevent.txt look nice in 80-col terminal
 * fix for copy_to_user() failure report for the first kevent (Andrew Morton)
 * minor function renames

Changes from 'take22' patchset:
 * new ring buffer implementation in process' memory
 * wakeup-one-thread flag
 * edge-triggered behaviour

Changes from 'take21' patchset:
 * minor cleanups (different return values, removed unneded variables, whitespaces and so on)
 * fixed bug in kevent removal in case when kevent being removed
   is the same as overflow_kevent (spotted by Eric Dumazet)

Changes from 'take20' patchset:
 * new ring buffer implementation
 * removed artificial limit on possible number of kevents

Changes from 'take19' patchset:
 * use __init instead of __devinit
 * removed 'default N' from config for user statistic
 * removed kevent_user_fini() since kevent can not be unloaded
 * use KERN_INFO for statistic output

Changes from 'take18' patchset:
 * use __init instead of __devinit
 * removed 'default N' from config for user statistic
 * removed kevent_user_fini() since kevent can not be unloaded
 * use KERN_INFO for statistic output

Changes from 'take17' patchset:
 * Use RB tree instead of hash table. 
	At least for a web sever, frequency of addition/deletion of new kevent 
	is comparable with number of search access, i.e. most of the time events 
	are added, accesed only couple of times and then removed, so it justifies 
	RB tree usage over AVL tree, since the latter does have much slower deletion 
	time (max O(log(N)) compared to 3 ops), 
	although faster search time (1.44*O(log(N)) vs. 2*O(log(N))). 
	So for kevents I use RB tree for now and later, when my AVL tree implementation 
	is ready, it will be possible to compare them.
 * Changed readiness check for socket notifications.

With both above changes it is possible to achieve more than 3380 req/second compared to 2200, 
sometimes 2500 req/second for epoll() for trivial web-server and httperf client on the same
hardware.
It is possible that above kevent limit is due to maximum allowed kevents in a time limit, which is
4096 events.

Changes from 'take16' patchset:
 * misc cleanups (__read_mostly, const ...)
 * created special macro which is used for mmap size (number of pages) calculation
 * export kevent_socket_notify(), since it is used in network protocols which can be 
	built as modules (IPv6 for example)

Changes from 'take15' patchset:
 * converted kevent_timer to high-resolution timers, this forces timer API update at
	http://linux-net.osdl.org/index.php/Kevent
 * use struct ukevent* instead of void * in syscalls (documentation has been updated)
 * added warning in kevent_add_ukevent() if ring has broken index (for testing)

Changes from 'take14' patchset:
 * added kevent_wait()
    This syscall waits until either timeout expires or at least one event
    becomes ready. It also commits that @num events from @start are processed
    by userspace and thus can be be removed or rearmed (depending on it's flags).
    It can be used for commit events read by userspace through mmap interface.
    Example userspace code (evtest.c) can be found on project's homepage.
 * added socket notifications (send/recv/accept)

Changes from 'take13' patchset:
 * do not get lock aroung user data check in __kevent_search()
 * fail early if there were no registered callbacks for given type of kevent
 * trailing whitespace cleanup

Changes from 'take12' patchset:
 * remove non-chardev interface for initialization
 * use pointer to kevent_mring instead of unsigned longs
 * use aligned 64bit type in raw user data (can be used by high-res timer if needed)
 * simplified enqueue/dequeue callbacks and kevent initialization
 * use nanoseconds for timeout
 * put number of milliseconds into timer's return data
 * move some definitions into user-visible header
 * removed filenames from comments

Changes from 'take11' patchset:
 * include missing headers into patchset
 * some trivial code cleanups (use goto instead of if/else games and so on)
 * some whitespace cleanups
 * check for ready_callback() callback before main loop which should save us some ticks

Changes from 'take10' patchset:
 * removed non-existent prototypes
 * added helper function for kevent_registered_callbacks
 * fixed 80 lines comments issues
 * added shared between userspace and kernelspace header instead of embedd them in one
 * core restructuring to remove forward declarations
 * s o m e w h i t e s p a c e c o d y n g s t y l e c l e a n u p
 * use vm_insert_page() instead of remap_pfn_range()

Changes from 'take9' patchset:
 * fixed ->nopage method

Changes from 'take8' patchset:
 * fixed mmap release bug
 * use module_init() instead of late_initcall()
 * use better structures for timer notifications

Changes from 'take7' patchset:
 * new mmap interface (not tested, waiting for other changes to be acked)
	- use nopage() method to dynamically substitue pages
	- allocate new page for events only when new added kevent requres it
	- do not use ugly index dereferencing, use structure instead
	- reduced amount of data in the ring (id and flags), 
		maximum 12 pages on x86 per kevent fd

Changes from 'take6' patchset:
 * a lot of comments!
 * do not use list poisoning for detection of the fact, that entry is in the list
 * return number of ready kevents even if copy*user() fails
 * strict check for number of kevents in syscall
 * use ARRAY_SIZE for array size calculation
 * changed superblock magic number
 * use SLAB_PANIC instead of direct panic() call
 * changed -E* return values
 * a lot of small cleanups and indent fixes

Changes from 'take5' patchset:
 * removed compilation warnings about unused wariables when lockdep is not turned on
 * do not use internal socket structures, use appropriate (exported) wrappers instead
 * removed default 1 second timeout
 * removed AIO stuff from patchset

Changes from 'take4' patchset:
 * use miscdevice instead of chardevice
 * comments fixes

Changes from 'take3' patchset:
 * removed serializing mutex from kevent_user_wait()
 * moved storage list processing to RCU
 * removed lockdep screaming - all storage locks are initialized in the same function, so it was
learned 
	to differentiate between various cases
 * remove kevent from storage if is marked as broken after callback
 * fixed a typo in mmaped buffer implementation which would end up in wrong index calcualtion 

Changes from 'take2' patchset:
 * split kevent_finish_user() to locked and unlocked variants
 * do not use KEVENT_STAT ifdefs, use inline functions instead
 * use array of callbacks of each type instead of each kevent callback initialization
 * changed name of ukevent guarding lock
 * use only one kevent lock in kevent_user for all hash buckets instead of per-bucket locks
 * do not use kevent_user_ctl structure instead provide needed arguments as syscall parameters
 * various indent cleanups
 * added optimisation, which is aimed to help when a lot of kevents are being copied from
userspace
 * mapped buffer (initial) implementation (no userspace yet)

Changes from 'take1' patchset:
 - rebased against 2.6.18-git tree
 - removed ioctl controlling
 - added new syscall kevent_get_events(int fd, unsigned int min_nr, unsigned int max_nr,
			unsigned int timeout, void __user *buf, unsigned flags)
 - use old syscall kevent_ctl for creation/removing, modification and initial kevent 
	initialization
 - use mutuxes instead of semaphores
 - added file descriptor check and return error if provided descriptor does not match
	kevent file operations
 - various indent fixes
 - removed aio_sendfile() declarations.

Thank you.

Signed-off-by: Evgeniy Polyakov <johnpol@2ka.mipt.ru>

[take35 1/10] kevent: Description.

[Evgeniy Polyakov]

Description.


diff --git a/Documentation/kevent.txt b/Documentation/kevent.txt
new file mode 100644
index 0000000..d6e126f
--- /dev/null
+++ b/Documentation/kevent.txt
@@ -0,0 +1,271 @@
+Description.
+
+int kevent_init(struct kevent_ring *ring, unsigned int ring_size, 
+	unsigned int flags);
+
+num - size of the ring buffer in events 
+ring - pointer to allocated ring buffer
+flags - various flags, see KEVENT_FLAGS_* definitions.
+
+Return value: kevent control file descriptor or negative error value.
+
+ struct kevent_ring
+ {
+   unsigned int ring_kidx, ring_over;
+   struct ukevent event[0];
+ }
+
+ring_kidx - index in the ring buffer where kernel will put new events 
+		when kevent_wait() or kevent_get_events() is called 
+ring_over - number of overflows of ring_uidx happend from the start.
+	Overflow counter is used to prevent situation when two threads 
+	are going to free the same events, but one of them was scheduled 
+	away for too long, so ring indexes were wrapped, so when that 
+	thread will be awakened, it will free not those events, which 
+	it suppose to free.
+
+Example userspace code (ring_buffer.c) can be found on project's homepage.
+
+Each kevent syscall can be so called cancellation point in glibc, i.e. when 
+thread has been cancelled in kevent syscall, thread can be safely removed 
+and no events will be lost, since each syscall (kevent_wait() or 
+kevent_get_events()) will copy event into special ring buffer, accessible 
+from other threads or even processes (if shared memory is used).
+
+When kevent is removed (not dequeued when it is ready, but just removed), 
+even if it was ready, it is not copied into ring buffer, since if it is 
+removed, no one cares about it (otherwise user would wait until it becomes 
+ready and got it through usual way using kevent_get_events() or kevent_wait()) 
+and thus no need to copy it to the ring buffer.
+
+-------------------------------------------------------------------------------
+
+
+int kevent_ctl(int fd, unsigned int cmd, unsigned int num, struct ukevent *arg);
+
+fd - is the file descriptor referring to the kevent queue to manipulate. 
+It is created by opening "/dev/kevent" char device, which is created with 
+dynamic minor number and major number assigned for misc devices. 
+
+cmd - is the requested operation. It can be one of the following:
+    KEVENT_CTL_ADD - add event notification 
+    KEVENT_CTL_REMOVE - remove event notification 
+    KEVENT_CTL_MODIFY - modify existing notification 
+    KEVENT_CTL_READY - mark existing events as ready, if number of events is zero,
+    	it just wakes up parked in syscall thread
+
+num - number of struct ukevent in the array pointed to by arg 
+arg - array of struct ukevent
+
+Return value: 
+ number of events processed or negative error value.
+
+When called, kevent_ctl will carry out the operation specified in the 
+cmd parameter.
+-------------------------------------------------------------------------------
+
+ int kevent_get_events(int ctl_fd, unsigned int min_nr, unsigned int max_nr, 
+ 		struct timespec timeout, struct ukevent *buf, unsigned flags);
+
+ctl_fd - file descriptor referring to the kevent queue 
+min_nr - minimum number of completed events that kevent_get_events will block 
+	 waiting for 
+max_nr - number of struct ukevent in buf 
+timeout - time to wait before returning less than min_nr 
+	  events. If this is -1, then wait forever. 
+buf - pointer to an array of struct ukevent. 
+flags - various flags, see KEVENT_FLAGS_* definitions.
+
+Return value:
+ number of events copied or negative error value.
+
+kevent_get_events will wait timeout milliseconds for at least min_nr completed 
+events, copying completed struct ukevents to buf and deleting any 
+KEVENT_REQ_ONESHOT event requests. In nonblocking mode it returns as many 
+events as possible, but not more than max_nr. In blocking mode it waits until 
+timeout or if at least min_nr events are ready.
+
+This function copies event into ring buffer if it was initialized, if ring buffer
+is full, KEVENT_RET_COPY_FAILED flag is set in ret_flags field.
+-------------------------------------------------------------------------------
+
+ int kevent_wait(int ctl_fd, unsigned int num, unsigned int old_uidx, 
+ 	struct timespec timeout, unsigned int flags);
+
+ctl_fd - file descriptor referring to the kevent queue 
+num - number of processed kevents 
+old_uidx - the last index user is aware of
+timeout - time to wait until there is free space in kevent queue
+flags - various flags, see KEVENT_FLAGS_* definitions.
+
+Return value:
+ number of events copied into ring buffer or negative error value.
+
+This syscall waits until either timeout expires or at least one event becomes 
+ready. It also copies events into special ring buffer. If ring buffer is full,
+it waits until there are ready events and then return.
+If kevent is one-shot kevent it is removed in this syscall.
+If kevent is edge-triggered (KEVENT_REQ_ET flag is set in 'req_flags') it is 
+requeued in this syscall for performance reasons.
+-------------------------------------------------------------------------------
+
+ int kevent_commit(int ctl_fd, unsigned int new_idx, unsigned int over);
+
+ctl_fd - file descriptor referring to the kevent queue 
+new_uidx - the last committed kevent
+over - overflow count for given $new_idx value
+
+Return value:
+ number of committed kevents or negative error value.
+
+This function commits, i.e. marks as empty, slots in the ring buffer, so
+they can be reused when userspace completes that entries processing.
+
+Overflow counter is used to prevent situation when two threads are going 
+to free the same events, but one of them was scheduled away for too long, 
+so ring indexes were wrapped, so when that thread will be awakened, it 
+will free not those events, which it suppose to free.
+
+It is possible that returned number of committed events will be smaller than
+requested number - it is possible when several threads try to commit the
+same events.
+-------------------------------------------------------------------------------
+
+long aio_sendfile(int kevent_fd, int sock_fd, int in_fd, off_t offset, size_t count);
+
+kevent_fd - file descriptor referring to the kevent queue
+sock_fd - destination socket file descriptor
+in_fd - source file descriptor
+offset - offset from the beginning of the source file
+count - number of bytes to transfer
+
+Async sendfile implementation.
+Returned coockie can be used to determine which entry has been returned by
+kevent_get_events() - it will be stored in event.ptr.
+event.ret_data will contain number of bytes actually transferred.
+-------------------------------------------------------------------------------
+
+long aio_sendfile_path(int kevent_fd, int sock_fd, 
+	void *header, size_t header_size, 
+	char *filename, off_t offset, size_t count);
+
+kevent_fd - file descriptor referring to the kevent queue
+sock_fd - destination socket file descriptor
+header - header data, which will be sent before file's data
+header_size - size of the header
+pathname - source filename
+offset - offset from the beginning of the source file
+count - number of bytes to transfer
+-------------------------------------------------------------------------------
+
+The bulk of the interface is entirely done through the ukevent struct. 
+It is used to add event requests, modify existing event requests, 
+specify which event requests to remove, and return completed events.
+
+struct ukevent contains the following members:
+
+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,
+
+    KEVENT_REQ_ONESHOT
+        event will be removed when it is ready 
+
+    KEVENT_REQ_WAKEUP_ALL
+        Kevent wakes up only first thread interested in given event, 
+	or all threads if this flag is set.
+
+    KEVENT_REQ_ET
+        Edge Triggered behaviour. It is an optimisation which allows to move 
+	ready and dequeued (i.e. copied to userspace) event to move into set 
+	of interest for given storage (socket, inode and so on) again. It is 
+	very usefull for cases when the same event should be used many times 
+	(like reading from pipe). It is similar to epoll()'s EPOLLET flag. 
+
+    KEVENT_REQ_LAST_CHECK
+        if set allows to perform the last check on kevent (call appropriate 
+	callback) when kevent is marked as ready and has been removed from 
+	ready queue. If it will be confirmed that kevent is ready 
+	(k->callbacks.callback(k) returns true) then kevent will be copied 
+	to userspace, otherwise it will be requeued back to storage. 
+	Second (checking) call is performed with this bit cleared, so callback 
+	can detect when it was called from kevent_storage_ready() - bit is set, 
+	or kevent_dequeue_ready() - bit is cleared. If kevent will be requeued, 
+	bit will be set again.
+
+   KEVENT_REQ_ALWAYS_QUEUE
+        If this flag is set kevent will be queued into ready queue if it is 
+	ready at enqueue time, otherwise it will be copied back to userspace
+	and will not be queued into the storage.
+
+   KEVENT_REQ_READY
+   	If this flag is set, kevent will be marked as ready immediately at enqueue
+	time.
+
+__u32 ret_flags
+    Per-event return flags
+
+    KEVENT_RET_BROKEN
+        Kevent is broken 
+
+    KEVENT_RET_DONE
+        Kevent processing was finished successfully 
+
+    KEVENT_RET_COPY_FAILED
+        Kevent was not copied into ring buffer due to some error conditions. 
+
+__u32 ret_data
+    Event return data. Event originator fills it with anything it likes 
+    (for example timer notifications put number of milliseconds when timer 
+    has fired 
+union { __u32 user[2]; void *ptr; }
+    User's data. It is not used, just copied to/from user. The whole structure 
+    is aligned to 8 bytes already, so the last union is aligned properly. 
+
+-------------------------------------------------------------------------------
+
+Kevent waiting syscall flags.
+
+KEVENT_FLAGS_ABSTIME - provided timespec parameter contains absolute time, 
+	for example Aug 27, 2194, or time(NULL) + 10.
+
+-------------------------------------------------------------------------------
+
+Usage
+
+For KEVENT_CTL_ADD, all fields relevant to the event type must be filled 
+(id, type, event, req_flags). 
+After kevent_ctl(..., KEVENT_CTL_ADD, ...) returns each struct's ret_flags 
+should be checked to see if the event is already broken or done.
+
+For KEVENT_CTL_MODIFY, the id, req_flags, and user and event fields must be 
+set and an existing kevent request must have matching id and user fields. If 
+match is found, req_flags and event are replaced with the newly supplied 
+values and requeueing is started, so modified kevent can be checked and 
+probably marked as ready immediately. If a match can't be found, the 
+passed in ukevent's ret_flags has KEVENT_RET_BROKEN set. KEVENT_RET_DONE is 
+always set.
+
+For KEVENT_CTL_REMOVE, the id and user fields must be set and an existing 
+kevent request must have matching id and user fields. If a match is found, 
+the kevent request is removed. If a match can't be found, the passed in 
+ukevent's ret_flags has KEVENT_RET_BROKEN set. KEVENT_RET_DONE is always set.
+
+For kevent_get_events, the entire structure is returned.
+
+-------------------------------------------------------------------------------
+
+Usage cases
+
+kevent_timer
+struct ukevent should contain following fields:
+    type - KEVENT_TIMER 
+    event - KEVENT_TIMER_FIRED 
+    req_flags - KEVENT_REQ_ONESHOT if you want to fire that timer only once 
+    id.raw[0] - number of seconds after commit when this timer shout expire 
+    id.raw[0] - additional to number of seconds number of nanoseconds 

[take35 3/10] kevent: poll/select() notifications.

[Evgeniy Polyakov]

poll/select() notifications.

This patch includes generic poll/select notifications.
kevent_poll works simialr to epoll and has the same issues (callback
is invoked not from internal state machine of the caller, but through
process awake, a lot of allocations and so on).

Signed-off-by: Evgeniy Polyakov <johnpol@2ka.mitp.ru>

diff --git a/fs/file_table.c b/fs/file_table.c
index 4c17a18..46f458c 100644
--- a/fs/file_table.c
+++ b/fs/file_table.c
@@ -20,6 +20,7 @@
 #include <linux/cdev.h>
 #include <linux/fsnotify.h>
 #include <linux/sysctl.h>
+#include <linux/kevent.h>
 #include <linux/percpu_counter.h>
 
 #include <asm/atomic.h>
@@ -119,6 +120,7 @@ struct file *get_empty_filp(void)
 	f->f_uid = tsk->fsuid;
 	f->f_gid = tsk->fsgid;
 	eventpoll_init_file(f);
+	kevent_init_file(f);
 	/* f->f_version: 0 */
 	return f;
 
@@ -164,6 +166,7 @@ void fastcall __fput(struct file *file)
 	 * in the file cleanup chain.
 	 */
 	eventpoll_release(file);
+	kevent_cleanup_file(file);
 	locks_remove_flock(file);
 
 	if (file->f_op && file->f_op->release)
diff --git a/include/linux/fs.h b/include/linux/fs.h
index 1410e53..24f605c 100644
--- a/include/linux/fs.h
+++ b/include/linux/fs.h
@@ -280,6 +280,7 @@ extern int dir_notify_enable;
 #include <linux/init.h>
 #include <linux/pid.h>
 #include <linux/mutex.h>
+#include <linux/kevent_storage.h>
 
 #include <asm/atomic.h>
 #include <asm/semaphore.h>
@@ -408,6 +409,8 @@ struct address_space_operations {
 
 	int (*readpages)(struct file *filp, struct address_space *mapping,
 			struct list_head *pages, unsigned nr_pages);
+	int (*aio_readpages)(struct file *filp, struct address_space *mapping,
+			struct list_head *pages, unsigned nr_pages, void *priv);
 
 	/*
 	 * ext3 requires that a successful prepare_write() call be followed
@@ -579,6 +582,10 @@ struct inode {
 	struct mutex		inotify_mutex;	/* protects the watches list */
 #endif
 
+#if defined CONFIG_KEVENT_SOCKET || defined CONFIG_KEVENT_PIPE
+	struct kevent_storage	st;
+#endif
+
 	unsigned long		i_state;
 	unsigned long		dirtied_when;	/* jiffies of first dirtying */
 
@@ -738,6 +745,9 @@ struct file {
 	struct list_head	f_ep_links;
 	spinlock_t		f_ep_lock;
 #endif /* #ifdef CONFIG_EPOLL */
+#ifdef CONFIG_KEVENT_POLL
+	struct kevent_storage	st;
+#endif
 	struct address_space	*f_mapping;
 };
 extern spinlock_t files_lock;
diff --git a/kernel/kevent/kevent_poll.c b/kernel/kevent/kevent_poll.c
new file mode 100644
index 0000000..58129fa
--- /dev/null
+++ b/kernel/kevent/kevent_poll.c
@@ -0,0 +1,234 @@
+/*
+ * 2006 Copyright (c) Evgeniy Polyakov <johnpol@2ka.mipt.ru>
+ * 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.
+ */
+
+#include <linux/kernel.h>
+#include <linux/types.h>
+#include <linux/list.h>
+#include <linux/slab.h>
+#include <linux/spinlock.h>
+#include <linux/timer.h>
+#include <linux/file.h>
+#include <linux/kevent.h>
+#include <linux/poll.h>
+#include <linux/fs.h>
+
+static struct kmem_cache *kevent_poll_container_cache;
+static struct kmem_cache *kevent_poll_priv_cache;
+
+struct kevent_poll_ctl
+{
+	struct poll_table_struct 	pt;
+	struct kevent			*k;
+};
+
+struct kevent_poll_wait_container
+{
+	struct list_head		container_entry;
+	wait_queue_head_t		*whead;
+	wait_queue_t			wait;
+	struct kevent			*k;
+};
+
+struct kevent_poll_private
+{
+	struct list_head		container_list;
+	spinlock_t			container_lock;
+};
+
+static int kevent_poll_enqueue(struct kevent *k);
+static int kevent_poll_dequeue(struct kevent *k);
+static int kevent_poll_callback(struct kevent *k);
+
+static int kevent_poll_wait_callback(wait_queue_t *wait,
+		unsigned mode, int sync, void *key)
+{
+	struct kevent_poll_wait_container *cont =
+		container_of(wait, struct kevent_poll_wait_container, wait);
+	struct kevent *k = cont->k;
+
+	kevent_storage_ready(k->st, NULL, KEVENT_MASK_ALL);
+	return 0;
+}
+
+static void kevent_poll_qproc(struct file *file, wait_queue_head_t *whead,
+		struct poll_table_struct *poll_table)
+{
+	struct kevent *k =
+		container_of(poll_table, struct kevent_poll_ctl, pt)->k;
+	struct kevent_poll_private *priv = k->priv;
+	struct kevent_poll_wait_container *cont;
+	unsigned long flags;
+
+	cont = kmem_cache_alloc(kevent_poll_container_cache, GFP_KERNEL);
+	if (!cont) {
+		kevent_break(k);
+		return;
+	}
+
+	cont->k = k;
+	init_waitqueue_func_entry(&cont->wait, kevent_poll_wait_callback);
+	cont->whead = whead;
+
+	spin_lock_irqsave(&priv->container_lock, flags);
+	list_add_tail(&cont->container_entry, &priv->container_list);
+	spin_unlock_irqrestore(&priv->container_lock, flags);
+
+	add_wait_queue(whead, &cont->wait);
+}
+
+static int kevent_poll_enqueue(struct kevent *k)
+{
+	struct file *file;
+	int err;
+	unsigned int revents;
+	unsigned long flags;
+	struct kevent_poll_ctl ctl;
+	struct kevent_poll_private *priv;
+
+	file = fget(k->event.id.raw[0]);
+	if (!file)
+		return -EBADF;
+	
+	err = -EINVAL;
+	if (!file->f_op || !file->f_op->poll)
+		goto err_out_fput;
+	
+	err = -ENOMEM;
+	priv = kmem_cache_alloc(kevent_poll_priv_cache, GFP_KERNEL);
+	if (!priv)
+		goto err_out_fput;
+
+	spin_lock_init(&priv->container_lock);
+	INIT_LIST_HEAD(&priv->container_list);
+
+	k->priv = priv;
+
+	ctl.k = k;
+	init_poll_funcptr(&ctl.pt, &kevent_poll_qproc);
+	
+	err = kevent_storage_enqueue(&file->st, k);
+	if (err)
+		goto err_out_free;
+
+	revents = file->f_op->poll(file, &ctl.pt);
+	if (k->event.req_flags & KEVENT_REQ_ALWAYS_QUEUE) {
+		kevent_requeue(k);
+	} else {
+		if (revents & k->event.event) {
+			err = 1;
+			goto out_dequeue;
+		}
+	}
+
+	spin_lock_irqsave(&k->ulock, flags);
+	k->event.req_flags |= KEVENT_REQ_LAST_CHECK;
+	spin_unlock_irqrestore(&k->ulock, flags);
+
+	return 0;
+
+out_dequeue:
+	kevent_storage_dequeue(k->st, k);
+err_out_free:
+	kmem_cache_free(kevent_poll_priv_cache, priv);
+err_out_fput:
+	fput(file);
+	return err;
+}
+
+static int kevent_poll_dequeue(struct kevent *k)
+{
+	struct file *file = k->st->origin;
+	struct kevent_poll_private *priv = k->priv;
+	struct kevent_poll_wait_container *w, *n;
+	unsigned long flags;
+
+	kevent_storage_dequeue(k->st, k);
+
+	spin_lock_irqsave(&priv->container_lock, flags);
+	list_for_each_entry_safe(w, n, &priv->container_list, container_entry) {
+		list_del(&w->container_entry);
+		remove_wait_queue(w->whead, &w->wait);
+		kmem_cache_free(kevent_poll_container_cache, w);
+	}
+	spin_unlock_irqrestore(&priv->container_lock, flags);
+
+	kmem_cache_free(kevent_poll_priv_cache, priv);
+	k->priv = NULL;
+
+	fput(file);
+
+	return 0;
+}
+
+static int kevent_poll_callback(struct kevent *k)
+{
+	if (k->event.req_flags & KEVENT_REQ_LAST_CHECK) {
+		return 1;
+	} else {
+		struct file *file = k->st->origin;
+		unsigned int revents = file->f_op->poll(file, NULL);
+
+		k->event.ret_data[0] = revents & k->event.event;
+
+		return (revents & k->event.event);
+	}
+}
+
+static int __init kevent_poll_sys_init(void)
+{
+	struct kevent_callbacks pc = {
+		.callback = &kevent_poll_callback,
+		.enqueue = &kevent_poll_enqueue,
+		.dequeue = &kevent_poll_dequeue,
+		.flags = 0,
+	};
+
+	kevent_poll_container_cache = kmem_cache_create("kevent_poll_container_cache",
+			sizeof(struct kevent_poll_wait_container), 0, 0, NULL, NULL);
+	if (!kevent_poll_container_cache) {
+		printk(KERN_ERR "Failed to create kevent poll container cache.\n");
+		return -ENOMEM;
+	}
+
+	kevent_poll_priv_cache = kmem_cache_create("kevent_poll_priv_cache",
+			sizeof(struct kevent_poll_private), 0, 0, NULL, NULL);
+	if (!kevent_poll_priv_cache) {
+		printk(KERN_ERR "Failed to create kevent poll private data cache.\n");
+		kmem_cache_destroy(kevent_poll_container_cache);
+		kevent_poll_container_cache = NULL;
+		return -ENOMEM;
+	}
+
+	kevent_add_callbacks(&pc, KEVENT_POLL);
+
+	printk(KERN_INFO "Kevent poll()/select() subsystem has been initialized.\n");
+	return 0;
+}
+
+static struct lock_class_key kevent_poll_key;
+
+void kevent_poll_reinit(struct file *file)
+{
+	lockdep_set_class(&file->st.lock, &kevent_poll_key);
+}
+
+static void __exit kevent_poll_sys_fini(void)
+{
+	kmem_cache_destroy(kevent_poll_priv_cache);
+	kmem_cache_destroy(kevent_poll_container_cache);
+}
+
+module_init(kevent_poll_sys_init);
+module_exit(kevent_poll_sys_fini);

[take35 2/10] kevent: Core files.

[Evgeniy Polyakov]

Core files.

This patch includes core kevent files:
 * userspace controlling
 * kernelspace interfaces
 * initialization
 * notification state machines

Some bits of documentation can be found on project's homepage (and links from there):
http://tservice.net.ru/~s0mbre/old/?section=projects&item=kevent

Signed-off-by: Evgeniy Polyakov <johnpol@2ka.mipt.ru>

diff --git a/arch/i386/kernel/syscall_table.S b/arch/i386/kernel/syscall_table.S
index 2697e92..9b4cbf1 100644
--- a/arch/i386/kernel/syscall_table.S
+++ b/arch/i386/kernel/syscall_table.S
@@ -319,3 +319,10 @@ ENTRY(sys_call_table)
 	.long sys_move_pages
 	.long sys_getcpu
 	.long sys_epoll_pwait
+	.long sys_kevent_get_events
+	.long sys_kevent_ctl		/* 320 */
+	.long sys_kevent_wait
+	.long sys_kevent_commit
+	.long sys_kevent_init
+	.long sys_aio_sendfile
+	.long sys_aio_sendfile_path
diff --git a/arch/x86_64/ia32/ia32entry.S b/arch/x86_64/ia32/ia32entry.S
index b4aa875..df87cde 100644
--- a/arch/x86_64/ia32/ia32entry.S
+++ b/arch/x86_64/ia32/ia32entry.S
@@ -714,8 +714,15 @@ ia32_sys_call_table:
 	.quad compat_sys_get_robust_list
 	.quad sys_splice
 	.quad sys_sync_file_range
-	.quad sys_tee
+	.quad sys_tee			/* 315 */
 	.quad compat_sys_vmsplice
 	.quad compat_sys_move_pages
 	.quad sys_getcpu
+	.quad sys_kevent_get_events
+	.quad sys_kevent_ctl		/* 320 */
+	.quad sys_kevent_wait
+	.quad sys_kevent_commit
+	.quad sys_kevent_init
+	.quad sys_aio_sendfile
+	.quad sys_aio_sendfile_path
 ia32_syscall_end:		
diff --git a/include/asm-i386/unistd.h b/include/asm-i386/unistd.h
index 833fa17..5800a2e 100644
--- a/include/asm-i386/unistd.h
+++ b/include/asm-i386/unistd.h
@@ -325,10 +325,17 @@
 #define __NR_move_pages		317
 #define __NR_getcpu		318
 #define __NR_epoll_pwait	319
+#define __NR_kevent_get_events	320
+#define __NR_kevent_ctl		321
+#define __NR_kevent_wait	322
+#define __NR_kevent_commit	323
+#define __NR_kevent_init	324
+#define __NR_aio_sendfile	325
+#define __NR_aio_sendfile_path	326
 
 #ifdef __KERNEL__
 
-#define NR_syscalls 320
+#define NR_syscalls 327
 
 #define __ARCH_WANT_IPC_PARSE_VERSION
 #define __ARCH_WANT_OLD_READDIR
diff --git a/include/asm-x86_64/unistd.h b/include/asm-x86_64/unistd.h
index c5f596e..6984f7c 100644
--- a/include/asm-x86_64/unistd.h
+++ b/include/asm-x86_64/unistd.h
@@ -619,8 +619,22 @@ __SYSCALL(__NR_sync_file_range, sys_sync_file_range)
 __SYSCALL(__NR_vmsplice, sys_vmsplice)
 #define __NR_move_pages		279
 __SYSCALL(__NR_move_pages, sys_move_pages)
-
-#define __NR_syscall_max __NR_move_pages
+#define __NR_kevent_get_events	280
+__SYSCALL(__NR_kevent_get_events, sys_kevent_get_events)
+#define __NR_kevent_ctl		281
+__SYSCALL(__NR_kevent_ctl, sys_kevent_ctl)
+#define __NR_kevent_wait	282
+__SYSCALL(__NR_kevent_wait, sys_kevent_wait)
+#define __NR_kevent_commit	283
+__SYSCALL(__NR_kevent_commit, sys_kevent_commit)
+#define __NR_kevent_init	284
+__SYSCALL(__NR_kevent_init, sys_kevent_init)
+#define __NR_aio_sendfile	285
+__SYSCALL(__NR_aio_sendfile, sys_aio_sendfile)
+#define __NR_aio_sendfile_path	286
+__SYSCALL(__NR_aio_sendfile_path, sys_aio_sendfile_path)
+
+#define __NR_syscall_max __NR_aio_sendfile_path
 
 #ifndef __NO_STUBS
 #define __ARCH_WANT_OLD_READDIR
diff --git a/include/linux/kevent.h b/include/linux/kevent.h
new file mode 100644
index 0000000..3040d01
--- /dev/null
+++ b/include/linux/kevent.h
@@ -0,0 +1,268 @@
+/*
+ * 2006 Copyright (c) Evgeniy Polyakov <johnpol@2ka.mipt.ru>
+ * 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
+#include <linux/types.h>
+#include <linux/list.h>
+#include <linux/rbtree.h>
+#include <linux/spinlock.h>
+#include <linux/mutex.h>
+#include <linux/wait.h>
+#include <linux/net.h>
+#include <linux/rcupdate.h>
+#include <linux/fs.h>
+#include <linux/sched.h>
+#include <linux/hrtimer.h>
+#include <linux/kevent_storage.h>
+#include <linux/ukevent.h>
+
+#define KEVENT_MIN_BUFFS_ALLOC	3
+
+struct kevent;
+struct kevent_storage;
+typedef int (* kevent_callback_t)(struct kevent *);
+
+/* @callback is called each time new event has been caught. */
+/* @enqueue is called each time new event is queued. */
+/* @dequeue is called each time event is dequeued. */
+/* @flags - flags for given set of callbacks. */
+
+struct kevent_callbacks {
+	kevent_callback_t	callback, enqueue, dequeue;
+	unsigned int		flags;
+};
+
+#define KEVENT_CALLBACKS_KERNELONLY	0x1
+
+int kevent_event_is_allowed(struct ukevent *e);
+
+#define KEVENT_READY		0x1
+#define KEVENT_STORAGE		0x2
+#define KEVENT_USER		0x4
+
+struct kevent
+{
+	/* Used for kevent freeing.*/
+	struct rcu_head		rcu_head;
+	struct ukevent		event;
+	/* This lock protects ukevent manipulations, e.g. ret_flags changes. */
+	spinlock_t		ulock;
+
+	/* Entry of user's tree. */
+	struct rb_node		kevent_node;
+	/* Entry of origin's queue. */
+	struct list_head	storage_entry;
+	/* Entry of user's ready. */
+	struct list_head	ready_entry;
+
+	u32			flags;
+
+	/* User who requested this kevent. */
+	struct kevent_user	*user;
+	/* Kevent container. */
+	struct kevent_storage	*st;
+
+	struct kevent_callbacks	callbacks;
+
+	/* Private data for different storages.
+	 * poll()/select storage has a list of wait_queue_t containers
+	 * for each ->poll() { poll_wait()' } here.
+	 */
+	void			*priv;
+};
+
+struct kevent_user
+{
+	struct rb_root		kevent_root;
+	spinlock_t		kevent_lock;
+	/* Number of queued kevents. */
+	unsigned int		kevent_num;
+
+	/* List of ready kevents. */
+	struct list_head	ready_list;
+	/* Number of ready kevents. */
+	unsigned int		ready_num;
+	/* Protects all manipulations with ready queue. */
+	spinlock_t 		ready_lock;
+
+	/* Protects against simultaneous kevent_user control manipulations. */
+	struct mutex		ctl_mutex;
+	/* Wait until some events are ready. */
+	wait_queue_head_t	wait;
+	/* Exit from syscall if someone wants us to do it */
+	int			need_exit;
+
+	/* Reference counter, increased for each new kevent. */
+	atomic_t		refcnt;
+
+	/* Mutex protecting userspace ring buffer. */
+	struct mutex		ring_lock;
+	/* Kernel index and size of the userspace ring buffer. */
+	unsigned int		kidx, uidx, ring_size, ring_over, full;
+	/* Pointer to userspace ring buffer. */
+	struct kevent_ring __user *pring;
+
+	/* Is used for absolute waiting times. */
+	struct hrtimer		timer;
+
+	/* Used for userspace private notifications. */
+	struct kevent_storage	st;
+
+#ifdef CONFIG_KEVENT_USER_STAT
+	unsigned long		im_num;
+	unsigned long		wait_num, ring_num;
+	unsigned long		total;
+#endif
+};
+
+int kevent_enqueue(struct kevent *k);
+int kevent_dequeue(struct kevent *k);
+int kevent_init(struct kevent *k);
+void kevent_requeue(struct kevent *k);
+int kevent_break(struct kevent *k);
+
+int kevent_add_callbacks(const struct kevent_callbacks *cb, int pos);
+
+void kevent_storage_ready(struct kevent_storage *st,
+		kevent_callback_t ready_callback, u32 event);
+int kevent_storage_init(void *origin, struct kevent_storage *st);
+void kevent_storage_fini(struct kevent_storage *st);
+int kevent_storage_enqueue(struct kevent_storage *st, struct kevent *k);
+void kevent_storage_dequeue(struct kevent_storage *st, struct kevent *k);
+
+void kevent_ready(struct kevent *k, int ret);
+
+int kevent_user_add_ukevent(struct ukevent *uk, struct kevent_user *u);
+
+#ifdef CONFIG_KEVENT_POLL
+void kevent_poll_reinit(struct file *file);
+#else
+static inline void kevent_poll_reinit(struct file *file)
+{
+}
+#endif
+
+#ifdef CONFIG_KEVENT_USER_STAT
+static inline void kevent_stat_init(struct kevent_user *u)
+{
+	u->wait_num = u->im_num = u->total = u->ring_num = 0;
+}
+static inline void kevent_stat_print(struct kevent_user *u)
+{
+	printk(KERN_INFO "%s: u: %p, wait: %lu, ring: %lu, immediately: %lu, total: %lu.\n",
+			__func__, u, u->wait_num, u->ring_num, u->im_num, u->total);
+}
+static inline void kevent_stat_im(struct kevent_user *u)
+{
+	u->im_num++;
+}
+static inline void kevent_stat_ring(struct kevent_user *u)
+{
+	u->ring_num++;
+}
+static inline void kevent_stat_wait(struct kevent_user *u)
+{
+	u->wait_num++;
+}
+static inline void kevent_stat_total(struct kevent_user *u)
+{
+	u->total++;
+}
+#else
+#define kevent_stat_print(u)		({ (void) u;})
+#define kevent_stat_init(u)		({ (void) u;})
+#define kevent_stat_im(u)		({ (void) u;})
+#define kevent_stat_wait(u)		({ (void) u;})
+#define kevent_stat_ring(u)		({ (void) u;})
+#define kevent_stat_total(u)		({ (void) u;})
+#endif
+
+void kevent_user_free(struct kevent_user *u);
+
+/*
+ * Kevent userspace control block reference counting.
+ * Set to 1 at creation time, when appropriate kevent file descriptor
+ * is closed, that reference counter is decreased.
+ * When counter hits zero block is freed.
+ */
+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)) {
+		kevent_user_free(u);
+
+	}
+}
+
+#ifdef CONFIG_LOCKDEP
+void kevent_socket_reinit(struct socket *sock);
+void kevent_sk_reinit(struct sock *sk);
+#else
+static inline void kevent_socket_reinit(struct socket *sock)
+{
+}
+static inline void kevent_sk_reinit(struct sock *sk)
+{
+}
+#endif
+#ifdef CONFIG_KEVENT_SOCKET
+void kevent_socket_notify(struct sock *sock, u32 event);
+int kevent_socket_dequeue(struct kevent *k);
+int kevent_socket_enqueue(struct kevent *k);
+#define sock_async(__sk) sock_flag(__sk, SOCK_ASYNC)
+#else
+static inline void kevent_socket_notify(struct sock *sock, u32 event)
+{
+}
+#define sock_async(__sk)	({ (void)__sk; 0; })
+#endif
+
+#ifdef CONFIG_KEVENT_POLL
+static inline void kevent_init_file(struct file *file)
+{
+	kevent_storage_init(file, &file->st);
+}
+
+static inline void kevent_cleanup_file(struct file *file)
+{
+	kevent_storage_fini(&file->st);
+}
+#else
+static inline void kevent_init_file(struct file *file) {}
+static inline void kevent_cleanup_file(struct file *file) {}
+#endif
+
+#ifdef CONFIG_KEVENT_PIPE
+extern void kevent_pipe_notify(struct inode *inode, u32 events);
+#else
+static inline void kevent_pipe_notify(struct inode *inode, u32 events) {}
+#endif
+
+#ifdef CONFIG_KEVENT_SIGNAL
+extern int kevent_signal_notify(struct task_struct *tsk, int sig);
+#else
+static inline int kevent_signal_notify(struct task_struct *tsk, int sig) {return 0;}
+#endif
+
+#endif /* __KEVENT_H */
diff --git a/include/linux/kevent_storage.h b/include/linux/kevent_storage.h
new file mode 100644
index 0000000..a38575d
--- /dev/null
+++ b/include/linux/kevent_storage.h
@@ -0,0 +1,11 @@
+#ifndef __KEVENT_STORAGE_H
+#define __KEVENT_STORAGE_H
+
+struct kevent_storage
+{
+	void			*origin;		/* Originator's pointer, e.g. struct sock or struct file. Can be NULL. */
+	struct list_head	list;			/* List of queued kevents. */
+	spinlock_t		lock;			/* Protects users queue. */
+};
+
+#endif /* __KEVENT_STORAGE_H */
diff --git a/include/linux/syscalls.h b/include/linux/syscalls.h
index 1912c6c..3dc47a3 100644
--- a/include/linux/syscalls.h
+++ b/include/linux/syscalls.h
@@ -54,6 +54,8 @@ struct compat_stat;
 struct compat_timeval;
 struct robust_list_head;
 struct getcpu_cache;
+struct ukevent;
+struct kevent_ring;
 
 #include <linux/types.h>
 #include <linux/aio_abi.h>
@@ -603,6 +605,19 @@ asmlinkage long sys_set_robust_list(struct robust_list_head __user *head,
 				    size_t len);
 asmlinkage long sys_getcpu(unsigned __user *cpu, unsigned __user *node, struct getcpu_cache __user *cache);
 
+asmlinkage long sys_kevent_get_events(int ctl_fd, unsigned int min, unsigned int max,
+		struct timespec timeout, struct ukevent __user *buf, unsigned flags);
+asmlinkage long sys_kevent_ctl(int ctl_fd, unsigned int cmd, unsigned int num, struct ukevent __user *buf);
+asmlinkage long sys_kevent_wait(int ctl_fd, unsigned int num, unsigned int old_uidx, 
+		struct timespec timeout, unsigned int flags);
+asmlinkage long sys_kevent_commit(int ctl_fd, unsigned int new_uidx, unsigned int over);
+asmlinkage long sys_kevent_init(struct kevent_ring __user *ring, unsigned int num, unsigned int flags);
+
 int kernel_execve(const char *filename, char *const argv[], char *const envp[]);
 
+asmlinkage long sys_aio_sendfile(int kevent_fd, int sock_fd, int in_fd, off_t offset, size_t count);
+asmlinkage long sys_aio_sendfile_path(int kevent_fd, int sock_fd, 
+		void __user *header, size_t header_size, 
+		char __user *filename, off_t offset, size_t count);
+
 #endif
diff --git a/include/linux/ukevent.h b/include/linux/ukevent.h
new file mode 100644
index 0000000..d975407
--- /dev/null
+++ b/include/linux/ukevent.h
@@ -0,0 +1,191 @@
+/*
+ * 2006 Copyright (c) Evgeniy Polyakov <johnpol@2ka.mipt.ru>
+ * 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 __UKEVENT_H
+#define __UKEVENT_H
+
+#include <linux/types.h>
+
+/*
+ * Kevent request flags.
+ */
+
+/* Process this event only once and then remove it. */
+#define KEVENT_REQ_ONESHOT	0x1
+/* Kevent wakes up only first thread interested in given event,
+ * or all threads if this flag is set.
+ */
+#define KEVENT_REQ_WAKEUP_ALL	0x2
+/* Edge Triggered behaviour. */
+#define KEVENT_REQ_ET		0x4
+/* Perform the last check on kevent (call appropriate callback) when
+ * kevent is marked as ready and has been removed from ready queue.
+ * If it will be confirmed that kevent is ready 
+ * (k->callbacks.callback(k) returns true) then kevent will be copied
+ * to userspace, otherwise it will be requeued back to storage. 
+ * Second (checking) call is performed with this bit _cleared_ so
+ * callback can detect when it was called from 
+ * kevent_storage_ready() - bit is set, or 
+ * kevent_dequeue_ready() - bit is cleared. 
+ * If kevent will be requeued, bit will be set again. */
+#define KEVENT_REQ_LAST_CHECK	0x8
+/*
+ * Always queue kevent even if it is immediately ready.
+ */
+#define KEVENT_REQ_ALWAYS_QUEUE	0x10
+/*
+ * Mark event as ready immediately on enqueueing time.
+ */
+#define KEVENT_REQ_READY	0x20
+
+/*
+ * Kevent return flags.
+ */
+/* Kevent is broken. */
+#define KEVENT_RET_BROKEN	0x1
+/* Kevent processing was finished successfully. */
+#define KEVENT_RET_DONE		0x2
+/* Kevent was not copied into ring buffer due to some error conditions. */
+#define KEVENT_RET_COPY_FAILED	0x4
+
+/*
+ * Kevent type set.
+ */
+#define KEVENT_SOCKET 		0
+#define KEVENT_INODE		1
+#define KEVENT_TIMER		2
+#define KEVENT_POLL		3
+#define KEVENT_NAIO		4
+#define KEVENT_AIO		5
+#define KEVENT_PIPE		6
+#define KEVENT_SIGNAL		7
+#define KEVENT_POSIX_TIMER	8
+#define KEVENT_UNOTIFY		9
+
+/* Used as array size, so must be equal to the last KEVENT type + 1 */
+#define	KEVENT_MAX		10
+
+/*
+ * Per-type event sets.
+ * Number of per-event sets should be exactly as number of kevent types.
+ */
+
+/*
+ * Timer events.
+ */
+#define	KEVENT_TIMER_FIRED	0x1
+
+/*
+ * Socket/network asynchronous IO and PIPE events.
+ */
+#define	KEVENT_SOCKET_RECV	0x1
+#define	KEVENT_SOCKET_ACCEPT	0x2
+#define	KEVENT_SOCKET_SEND	0x4
+
+/*
+ * Inode events.
+ */
+#define	KEVENT_INODE_CREATE	0x1
+#define	KEVENT_INODE_REMOVE	0x2
+
+/*
+ * Poll events.
+ */
+#define	KEVENT_POLL_POLLIN	0x0001
+#define	KEVENT_POLL_POLLPRI	0x0002
+#define	KEVENT_POLL_POLLOUT	0x0004
+#define	KEVENT_POLL_POLLERR	0x0008
+#define	KEVENT_POLL_POLLHUP	0x0010
+#define	KEVENT_POLL_POLLNVAL	0x0020
+
+#define	KEVENT_POLL_POLLRDNORM	0x0040
+#define	KEVENT_POLL_POLLRDBAND	0x0080
+#define	KEVENT_POLL_POLLWRNORM	0x0100
+#define	KEVENT_POLL_POLLWRBAND	0x0200
+#define	KEVENT_POLL_POLLMSG	0x0400
+#define	KEVENT_POLL_POLLREMOVE	0x1000
+#define KEVENT_POLL_POLLRDHUP	0x2000
+
+/*
+ * Asynchronous IO events.
+ */
+#define	KEVENT_AIO_BIO		0x1
+
+/*
+ * Signal events.
+ */
+#define KEVENT_SIGNAL_DELIVERY		0x1
+
+/* If set in raw64, then given signals will not be delivered
+ * in a usual way through sigmask update and signal callback 
+ * invocation. */
+#define KEVENT_SIGNAL_NOMASK	0x8000000000000000ULL
+
+/* Mask of all possible event values. */
+#define KEVENT_MASK_ALL		0xffffffff
+/* Empty mask of ready events. */
+#define KEVENT_MASK_EMPTY	0x0
+
+struct kevent_id
+{
+	union {
+		__u32		raw[2];
+		__u64		raw_u64 __attribute__((aligned(8)));
+	};
+};
+
+struct ukevent
+{
+	/* Id of this request, e.g. socket number, file descriptor and so on... */
+	struct kevent_id	id;
+	/* Event type, e.g. KEVENT_SOCK, KEVENT_INODE, KEVENT_TIMER and so on... */
+	__u32			type;
+	/* Event itself, e.g. SOCK_ACCEPT, INODE_CREATED, TIMER_FIRED... */
+	__u32			event;
+	/* Per-event request flags */
+	__u32			req_flags;
+	/* Per-event return flags */
+	__u32			ret_flags;
+	/* Event return data. Event originator fills it with anything it likes. */
+	__u32			ret_data[2];
+	/* User's data. It is not used, just copied to/from user.
+	 * The whole structure is aligned to 8 bytes already, so the last union
+	 * is aligned properly.
+	 */
+	union {
+		__u32		user[2];
+		void		*ptr;
+	};
+};
+
+struct kevent_ring
+{
+	unsigned int		ring_kidx, ring_over;
+	struct ukevent		event[0];
+};
+
+#define	KEVENT_CTL_ADD 		0
+#define	KEVENT_CTL_REMOVE	1
+#define	KEVENT_CTL_MODIFY	2
+#define	KEVENT_CTL_READY	3
+
+/* Provided timespec parameter uses absolute time, i.e. 'wait until Aug 27, 2194' */
+#define KEVENT_FLAGS_ABSTIME	1
+
+#endif /* __UKEVENT_H */
diff --git a/init/Kconfig b/init/Kconfig
index a3f83e2..561a0fa 100644
--- a/init/Kconfig
+++ b/init/Kconfig
@@ -219,6 +219,8 @@ config AUDITSYSCALL
 	  such as SELinux.  To use audit's filesystem watch feature, please
 	  ensure that INOTIFY is configured.
 
+source "kernel/kevent/Kconfig"
+
 config IKCONFIG
 	tristate "Kernel .config support"
 	---help---
diff --git a/kernel/Makefile b/kernel/Makefile
index 14f4d45..6bccdbb 100644
--- a/kernel/Makefile
+++ b/kernel/Makefile
@@ -46,6 +46,7 @@ obj-$(CONFIG_DETECT_SOFTLOCKUP) += softlockup.o
 obj-$(CONFIG_GENERIC_HARDIRQS) += irq/
 obj-$(CONFIG_SECCOMP) += seccomp.o
 obj-$(CONFIG_RCU_TORTURE_TEST) += rcutorture.o
+obj-$(CONFIG_KEVENT) += kevent/
 obj-$(CONFIG_RELAY) += relay.o
 obj-$(CONFIG_UTS_NS) += utsname.o
 obj-$(CONFIG_TASK_DELAY_ACCT) += delayacct.o
diff --git a/kernel/kevent/Kconfig b/kernel/kevent/Kconfig
new file mode 100644
index 0000000..a23e84f
--- /dev/null
+++ b/kernel/kevent/Kconfig
@@ -0,0 +1,82 @@
+config KEVENT
+	bool "Kernel event notification mechanism" if EMBEDDED
+	default y
+	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_USER_STAT
+	bool "Kevent user statistic"
+	depends on KEVENT
+	help
+	  This option will turn kevent_user statistic collection on.
+	  Statistic data includes total number of kevent, number of kevents
+	  which are ready immediately at insertion time and number of kevents
+	  which were removed through readiness completion.
+	  It will be printed each time control kevent descriptor is closed.
+
+config KEVENT_TIMER
+	bool "Kernel event notifications for timers"
+	depends on KEVENT
+	default y
+	help
+	  This option allows to use timers through KEVENT subsystem.
+
+config KEVENT_POLL
+	bool "Kernel event notifications for poll()/select()"
+	depends on KEVENT
+	default y
+	help
+	  This option allows to use kevent subsystem for poll()/select()
+	  notifications.
+
+config KEVENT_SOCKET
+	bool "Kernel event notifications for sockets"
+	depends on NET && KEVENT
+	default y
+	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_PIPE
+	bool "Kernel event notifications for pipes"
+	depends on KEVENT
+	default y
+	help
+	  This option enables notifications through KEVENT subsystem of 
+	  pipe read/write operations.
+
+config KEVENT_SIGNAL
+	bool "Kernel event notifications for signals"
+	depends on KEVENT
+	default y
+	help
+	  This option enables signal delivery through KEVENT subsystem.
+	  Signals which were requested to be delivered through kevent
+	  subsystem must be registered through usual signal() and others
+	  syscalls, this option allows alternative delivery.
+	  With KEVENT_SIGNAL_NOMASK flag being set in kevent for set of 
+	  signals, they will not be delivered in a usual way.
+	  Kevents for appropriate signals are not copied when process forks,
+	  new process must add new kevents after fork(). Mask of signals
+	  is copied as before.
+
+config KEVENT_UNOTIFY
+	bool "Private userspace notifications over kevent"
+	depends on KEVENT
+	default y
+	help
+	  This option enable possibility to insert private userspace events,
+	  which can be marked as ready on demand using kevent_ctl(KEVENT_CTL_READY)
+	  command.
+
+config KEVENT_AIO
+	bool "Kevent based AIO"
+	depends on KEVENT && NET
+	default y
+	help
+	  This option allows to work with kevent based AIO state machine.
+	  Among others this allows to implement aio_sendfile() syscall.
diff --git a/kernel/kevent/Makefile b/kernel/kevent/Makefile
new file mode 100644
index 0000000..a179bea
--- /dev/null
+++ b/kernel/kevent/Makefile
@@ -0,0 +1,8 @@
+obj-y := kevent.o kevent_user.o
+obj-$(CONFIG_KEVENT_TIMER) += kevent_timer.o
+obj-$(CONFIG_KEVENT_POLL) += kevent_poll.o epoll.o
+obj-$(CONFIG_KEVENT_SOCKET) += kevent_socket.o
+obj-$(CONFIG_KEVENT_PIPE) += kevent_pipe.o
+obj-$(CONFIG_KEVENT_SIGNAL) += kevent_signal.o
+obj-$(CONFIG_KEVENT_UNOTIFY) += kevent_unotify.o
+obj-$(CONFIG_KEVENT_AIO) += kevent_aio.o
diff --git a/kernel/kevent/kevent.c b/kernel/kevent/kevent.c
new file mode 100644
index 0000000..743cd0c
--- /dev/null
+++ b/kernel/kevent/kevent.c
@@ -0,0 +1,267 @@
+/*
+ * 2006 Copyright (c) Evgeniy Polyakov <johnpol@2ka.mipt.ru>
+ * 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 <linux/kernel.h>
+#include <linux/types.h>
+#include <linux/list.h>
+#include <linux/slab.h>
+#include <linux/spinlock.h>
+#include <linux/mempool.h>
+#include <linux/sched.h>
+#include <linux/wait.h>
+#include <linux/kevent.h>
+
+/*
+ * 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.
+ * ->enqueue() callback must increase origin's reference counter.
+ */
+int kevent_enqueue(struct kevent *k)
+{
+	return k->callbacks.enqueue(k);
+}
+
+/*
+ * Remove event from the appropriate queue.
+ * ->dequeue() callback must decrease origin's reference counter.
+ */
+int kevent_dequeue(struct kevent *k)
+{
+	return k->callbacks.dequeue(k);
+}
+
+/*
+ * Mark kevent as broken.
+ */
+int kevent_break(struct kevent *k)
+{
+	unsigned long flags;
+
+	spin_lock_irqsave(&k->ulock, flags);
+	k->event.ret_flags |= KEVENT_RET_BROKEN;
+	spin_unlock_irqrestore(&k->ulock, flags);
+	return -EINVAL;
+}
+
+static struct kevent_callbacks kevent_registered_callbacks[KEVENT_MAX] __read_mostly;
+
+int kevent_event_is_allowed(struct ukevent *e)
+{
+	if (unlikely(e->type >= KEVENT_MAX))
+		return 0;
+
+	if (!kevent_registered_callbacks[e->type].callback)
+		return 0;
+
+	if (unlikely(kevent_registered_callbacks[e->type].callback == kevent_break))
+		return 0;
+
+	if (kevent_registered_callbacks[e->type].flags & KEVENT_CALLBACKS_KERNELONLY)
+		return 0;
+	
+	return 1;
+}
+
+int kevent_add_callbacks(const struct kevent_callbacks *cb, int pos)
+{
+	struct kevent_callbacks *p;
+
+	if (pos >= KEVENT_MAX)
+		return -EINVAL;
+
+	p = &kevent_registered_callbacks[pos];
+
+	p->enqueue = (cb->enqueue) ? cb->enqueue : kevent_break;
+	p->dequeue = (cb->dequeue) ? cb->dequeue : kevent_break;
+	p->callback = (cb->callback) ? cb->callback : kevent_break;
+	p->flags = cb->flags;
+
+	printk(KERN_INFO "KEVENT: Added callbacks for type %d.\n", pos);
+	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_enqueue() 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)
+{
+	spin_lock_init(&k->ulock);
+	k->flags = 0;
+
+	if (unlikely(k->event.type >= KEVENT_MAX)) {
+		kevent_break(k);
+		return -ENOSYS;
+	}
+
+	if (!kevent_registered_callbacks[k->event.type].callback) {
+		kevent_break(k);
+		return -ENOSYS;
+	}
+
+	k->callbacks = kevent_registered_callbacks[k->event.type];
+	if (unlikely(k->callbacks.callback == kevent_break)) {
+		kevent_break(k);
+		return -ENOSYS;
+	}
+
+	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);
+	list_add_tail_rcu(&k->storage_entry, &st->list);
+	k->flags |= KEVENT_STORAGE;
+	if (k->event.req_flags & KEVENT_REQ_READY)
+		kevent_ready(k, 1);
+	spin_unlock_irqrestore(&st->lock, flags);
+	return 0;
+}
+
+/*
+ * Dequeue kevent from origin's queue.
+ * It does not decrease origin's reference counter in any way
+ * and must be called before it, so storage itself must be valid.
+ * 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);
+	if (k->flags & KEVENT_STORAGE) {
+		list_del_rcu(&k->storage_entry);
+		k->flags &= ~KEVENT_STORAGE;
+	}
+	spin_unlock_irqrestore(&st->lock, flags);
+}
+
+void kevent_ready(struct kevent *k, int ret)
+{
+	unsigned long flags;
+	int rem;
+
+	spin_lock_irqsave(&k->ulock, flags);
+	if (ret > 0)
+		k->event.ret_flags |= KEVENT_RET_DONE;
+	else if (ret < 0)
+		k->event.ret_flags |= (KEVENT_RET_BROKEN | KEVENT_RET_DONE);
+	else
+		ret = (k->event.ret_flags & (KEVENT_RET_BROKEN|KEVENT_RET_DONE));
+	rem = (k->event.req_flags & KEVENT_REQ_ONESHOT);
+	spin_unlock_irqrestore(&k->ulock, flags);
+
+	if (ret) {
+		if ((rem || ret < 0) && (k->flags & KEVENT_STORAGE)) {
+			list_del_rcu(&k->storage_entry);
+			k->flags &= ~KEVENT_STORAGE;
+		}
+
+		spin_lock_irqsave(&k->user->ready_lock, flags);
+		if (!(k->flags & KEVENT_READY)) {
+			list_add_tail(&k->ready_entry, &k->user->ready_list);
+			k->flags |= KEVENT_READY;
+			k->user->ready_num++;
+		}
+		spin_unlock_irqrestore(&k->user->ready_lock, flags);
+		wake_up(&k->user->wait);
+	}
+}
+
+/*
+ * Call kevent ready callback and queue it into ready queue if needed.
+ * If kevent is marked as one-shot, then remove it from storage queue.
+ */
+static int __kevent_requeue(struct kevent *k, u32 event)
+{
+	int ret;
+
+	ret = k->callbacks.callback(k);
+
+	kevent_ready(k, ret);
+
+	return ret;
+}
+
+/*
+ * Check if kevent is ready (by invoking it's callback) and requeue/remove
+ * if needed.
+ */
+void kevent_requeue(struct kevent *k)
+{
+	unsigned long flags;
+
+	spin_lock_irqsave(&k->st->lock, flags);
+	__kevent_requeue(k, 0);
+	spin_unlock_irqrestore(&k->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, u32 event)
+{
+	struct kevent *k;
+	int wake_num = 0;
+
+	rcu_read_lock();
+	if (unlikely(ready_callback))
+		list_for_each_entry_rcu(k, &st->list, storage_entry)
+			(*ready_callback)(k);
+
+	list_for_each_entry_rcu(k, &st->list, storage_entry) {
+		if (event & k->event.event)
+			if ((k->event.req_flags & KEVENT_REQ_WAKEUP_ALL) || wake_num == 0)
+				if (__kevent_requeue(k, event))
+					wake_num++;
+	}
+	rcu_read_unlock();
+}
+
+int kevent_storage_init(void *origin, struct kevent_storage *st)
+{
+	spin_lock_init(&st->lock);
+	st->origin = origin;
+	INIT_LIST_HEAD(&st->list);
+	return 0;
+}
+
+/*
+ * Mark all events as broken, that will remove them from storage,
+ * so storage origin (inode, socket and so on) can be safely removed.
+ * No new entries are allowed to be added into the storage at this point.
+ * (Socket is removed from file table at this point for example).
+ */
+void kevent_storage_fini(struct kevent_storage *st)
+{
+	kevent_storage_ready(st, kevent_break, KEVENT_MASK_ALL);
+}
diff --git a/kernel/kevent/kevent_user.c b/kernel/kevent/kevent_user.c
new file mode 100644
index 0000000..f25d696
--- /dev/null
+++ b/kernel/kevent/kevent_user.c
@@ -0,0 +1,1360 @@
+/*
+ * 2006 Copyright (c) Evgeniy Polyakov <johnpol@2ka.mipt.ru>
+ * 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 <linux/kernel.h>
+#include <linux/module.h>
+#include <linux/types.h>
+#include <linux/list.h>
+#include <linux/slab.h>
+#include <linux/spinlock.h>
+#include <linux/fs.h>
+#include <linux/file.h>
+#include <linux/mount.h>
+#include <linux/device.h>
+#include <linux/poll.h>
+#include <linux/kevent.h>
+#include <linux/miscdevice.h>
+#include <asm/io.h>
+
+static struct kmem_cache *kevent_cache __read_mostly;
+static struct kmem_cache *kevent_user_cache __read_mostly;
+
+static int kevent_debug_abstime;
+
+/*
+ * kevents are pollable, return POLLIN and POLLRDNORM
+ * when there is at least one ready kevent.
+ */
+static unsigned int kevent_user_poll(struct file *file, struct poll_table_struct *wait)
+{
+	struct kevent_user *u = file->private_data;
+	unsigned int mask;
+
+	poll_wait(file, &u->wait, wait);
+	mask = 0;
+
+	if (u->ready_num || u->need_exit)
+		mask |= POLLIN | POLLRDNORM;
+	u->need_exit = 0;
+
+	return mask;
+}
+
+static inline unsigned int kevent_ring_space(struct kevent_user *u)
+{
+	if (!u->pring)
+		return 1;
+
+	if (u->full)
+		return 0;
+
+	return (u->uidx > u->kidx)?
+		(u->uidx - u->kidx):
+		(u->ring_size - (u->kidx - u->uidx));
+}
+
+static inline int kevent_ring_index_inc(unsigned int *pidx, unsigned int size)
+{
+	unsigned int idx = *pidx;
+
+	if (++idx >= size)
+		idx = 0;
+	*pidx = idx;
+	return (idx == 0);
+}
+
+/*
+ * Copies kevent into userspace ring buffer if it was initialized.
+ * Returns 
+ *  0 on success or if ring buffer is not used
+ *  -EAGAIN if there were no place for that kevent
+ *  -EFAULT if copy_to_user() failed.
+ *
+ *  Must be called under kevent_user->ring_lock locked.
+ */
+static int kevent_copy_ring_buffer(struct kevent *k)
+{
+	struct kevent_ring __user *ring;
+	struct kevent_user *u = k->user;
+	unsigned long flags;
+	int err;
+
+	ring = u->pring;
+	if (!ring)
+		return 0;
+
+	if (!kevent_ring_space(u))
+		return -EAGAIN;
+
+	if (copy_to_user(&ring->event[u->kidx], &k->event, sizeof(struct ukevent))) {
+		err = -EFAULT;
+		goto err_out_exit;
+	}
+
+	kevent_ring_index_inc(&u->kidx, u->ring_size);
+
+	if (u->kidx == u->uidx)
+		u->full = 1;
+
+	if (put_user(u->kidx, &ring->ring_kidx)) {
+		err = -EFAULT;
+		goto err_out_exit;
+	}
+
+	return 0;
+
+err_out_exit:
+	spin_lock_irqsave(&k->ulock, flags);
+	k->event.ret_flags |= KEVENT_RET_COPY_FAILED;
+	spin_unlock_irqrestore(&k->ulock, flags);
+	return err;
+}
+
+static struct kevent_user *kevent_user_alloc(struct kevent_ring __user *ring, unsigned int num)
+{
+	struct kevent_user *u;
+
+	u = kmem_cache_zalloc(kevent_user_cache, GFP_KERNEL);
+	if (!u)
+		return NULL;
+
+	INIT_LIST_HEAD(&u->ready_list);
+	spin_lock_init(&u->ready_lock);
+	kevent_stat_init(u);
+	spin_lock_init(&u->kevent_lock);
+	u->kevent_root = RB_ROOT;
+
+	mutex_init(&u->ctl_mutex);
+	init_waitqueue_head(&u->wait);
+	u->need_exit = 0;
+
+	atomic_set(&u->refcnt, 1);
+
+	mutex_init(&u->ring_lock);
+	u->kidx = u->uidx = u->ring_over = u->full = 0;
+
+	u->pring = ring;
+	u->ring_size = num;
+
+	hrtimer_init(&u->timer, CLOCK_REALTIME, HRTIMER_ABS);
+
+	kevent_storage_init(u, &u->st);
+
+	return u;
+}
+
+void kevent_user_free(struct kevent_user *u)
+{
+	kevent_stat_print(u);
+	hrtimer_cancel(&u->timer);
+	kmem_cache_free(kevent_user_cache, u);
+}
+
+static inline int kevent_compare_id(struct kevent_id *left, struct kevent_id *right, 
+		__u32 type_left, __u32 type_right)
+{
+	if (left->raw_u64 > right->raw_u64)
+		return -1;
+
+	if (right->raw_u64 > left->raw_u64)
+		return 1;
+
+	if (type_left > type_right)
+		return -1;
+
+	if (type_left < type_right)
+		return 1;
+
+	return 0;
+}
+
+/*
+ * RCU protects storage list (kevent->storage_entry).
+ * Free entry in RCU callback, it is dequeued from all lists at
+ * this point.
+ */
+
+static void kevent_free_rcu(struct rcu_head *rcu)
+{
+	struct kevent *kevent = container_of(rcu, struct kevent, rcu_head);
+	kmem_cache_free(kevent_cache, kevent);
+}
+
+/*
+ * Must be called under u->ready_lock.
+ * This function unlinks kevent from ready queue.
+ */
+static inline void kevent_unlink_ready(struct kevent *k)
+{
+	list_del(&k->ready_entry);
+	k->flags &= ~KEVENT_READY;
+	k->user->ready_num--;
+}
+
+static void kevent_remove_ready(struct kevent *k)
+{
+	struct kevent_user *u = k->user;
+	unsigned long flags;
+
+	spin_lock_irqsave(&u->ready_lock, flags);
+	if (k->flags & KEVENT_READY)
+		kevent_unlink_ready(k);
+	spin_unlock_irqrestore(&u->ready_lock, flags);
+}
+
+/*
+ * Complete kevent removing - it dequeues kevent from storage list
+ * if it is requested, removes kevent from ready list, drops userspace
+ * control block reference counter and schedules kevent freeing through RCU.
+ */
+static void kevent_finish_user_complete(struct kevent *k, int deq)
+{
+	if (deq)
+		kevent_dequeue(k);
+
+	kevent_remove_ready(k);
+
+	kevent_user_put(k->user);
+	call_rcu(&k->rcu_head, kevent_free_rcu);
+}
+
+/*
+ * Remove from all lists and free kevent.
+ * Must be called under kevent_user->kevent_lock to protect
+ * kevent->kevent_entry removing.
+ */
+static void __kevent_finish_user(struct kevent *k, int deq)
+{
+	struct kevent_user *u = k->user;
+
+	rb_erase(&k->kevent_node, &u->kevent_root);
+	k->flags &= ~KEVENT_USER;
+	u->kevent_num--;
+	kevent_finish_user_complete(k, deq);
+}
+
+/*
+ * Remove kevent from user's list of all events,
+ * 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 deq)
+{
+	struct kevent_user *u = k->user;
+	unsigned long flags;
+
+	spin_lock_irqsave(&u->kevent_lock, flags);
+	rb_erase(&k->kevent_node, &u->kevent_root);
+	k->flags &= ~KEVENT_USER;
+	u->kevent_num--;
+	spin_unlock_irqrestore(&u->kevent_lock, flags);
+	kevent_finish_user_complete(k, deq);
+}
+
+static struct kevent *__kevent_dequeue_ready_one(struct kevent_user *u)
+{
+	unsigned long flags;
+	struct kevent *k = NULL;
+
+	if (u->ready_num) {
+		spin_lock_irqsave(&u->ready_lock, flags);
+		if (u->ready_num && !list_empty(&u->ready_list)) {
+			k = list_entry(u->ready_list.next, struct kevent, ready_entry);
+			kevent_unlink_ready(k);
+		}
+		spin_unlock_irqrestore(&u->ready_lock, flags);
+	}
+
+	return k;
+}
+
+static struct kevent *kevent_dequeue_ready_one(struct kevent_user *u)
+{
+	struct kevent *k = NULL;
+
+	while (u->ready_num && !k) {
+		k = __kevent_dequeue_ready_one(u);
+
+		if (k && (k->event.req_flags & KEVENT_REQ_LAST_CHECK)) {
+			unsigned long flags;
+
+			spin_lock_irqsave(&k->ulock, flags);
+			k->event.req_flags &= ~KEVENT_REQ_LAST_CHECK;
+			spin_unlock_irqrestore(&k->ulock, flags);
+
+			if (!k->callbacks.callback(k)) {
+				spin_lock_irqsave(&k->ulock, flags);
+				k->event.req_flags |= KEVENT_REQ_LAST_CHECK;
+				k->event.ret_flags = 0;
+				k->event.ret_data[0] = k->event.ret_data[1] = 0;
+				spin_unlock_irqrestore(&k->ulock, flags);
+				k = NULL;
+			}
+		} else
+			break;
+	}
+
+	return k;
+}
+
+static inline void kevent_copy_ring(struct kevent *k)
+{
+	unsigned long flags;
+
+	if (!k)
+		return;
+
+	if (kevent_copy_ring_buffer(k)) {
+		spin_lock_irqsave(&k->ulock, flags);
+		k->event.ret_flags |= KEVENT_RET_COPY_FAILED;
+		spin_unlock_irqrestore(&k->ulock, flags);
+	}
+}
+
+/*
+ * Dequeue one entry from user's ready queue.
+ */
+static struct kevent *kevent_dequeue_ready(struct kevent_user *u)
+{
+	struct kevent *k;
+
+	mutex_lock(&u->ring_lock);
+	k = kevent_dequeue_ready_one(u);
+	kevent_copy_ring(k);
+	mutex_unlock(&u->ring_lock);
+
+	return k;
+}
+
+/*
+ * Dequeue one entry from user's ready queue if there is space in ring buffer.
+ */
+static struct kevent *kevent_dequeue_ready_ring(struct kevent_user *u)
+{
+	struct kevent *k = NULL;
+
+	mutex_lock(&u->ring_lock);
+	if (kevent_ring_space(u)) {
+		k = kevent_dequeue_ready_one(u);
+		kevent_copy_ring(k);
+	}
+	mutex_unlock(&u->ring_lock);
+
+	return k;
+}
+
+static void kevent_complete_ready(struct kevent *k)
+{
+	if (k->event.req_flags & KEVENT_REQ_ONESHOT)
+		/*
+		 * If it is one-shot kevent, it has been removed already from
+		 * origin's queue, so we can easily free it here.
+		 */
+		kevent_finish_user(k, 1);
+	else if (k->event.req_flags & KEVENT_REQ_ET) {
+		unsigned long flags;
+
+		/*
+		 * Edge-triggered behaviour: mark event as clear new one.
+		 */
+
+		spin_lock_irqsave(&k->ulock, flags);
+		k->event.ret_flags = 0;
+		k->event.ret_data[0] = k->event.ret_data[1] = 0;
+		spin_unlock_irqrestore(&k->ulock, flags);
+	}
+}
+
+/*
+ * Search a kevent inside kevent tree for given ukevent.
+ */
+static struct kevent *__kevent_search(struct kevent_id *id, __u32 type, struct kevent_user *u)
+{
+	struct kevent *k, *ret = NULL;
+	struct rb_node *n = u->kevent_root.rb_node;
+	int cmp;
+
+	while (n) {
+		k = rb_entry(n, struct kevent, kevent_node);
+		cmp = kevent_compare_id(&k->event.id, id, k->event.type, type);
+
+		if (cmp > 0)
+			n = n->rb_right;
+		else if (cmp < 0)
+			n = n->rb_left;
+		else {
+			ret = k;
+			break;
+		}
+	}
+
+	return ret;
+}
+
+/*
+ * Search and modify kevent according to provided ukevent.
+ */
+static int kevent_modify(struct ukevent *uk, struct kevent_user *u)
+{
+	struct kevent *k;
+	int err = -ENODEV;
+	unsigned long flags;
+
+	spin_lock_irqsave(&u->kevent_lock, flags);
+	k = __kevent_search(&uk->id, uk->type, u);
+	if (k) {
+		spin_lock(&k->ulock);
+		k->event.event = uk->event;
+		k->event.req_flags = uk->req_flags;
+		k->event.ret_flags = 0;
+		spin_unlock(&k->ulock);
+		kevent_requeue(k);
+		err = 0;
+	}
+	spin_unlock_irqrestore(&u->kevent_lock, flags);
+
+	return err;
+}
+
+/*
+ * Remove kevent which matches provided ukevent.
+ */
+static int kevent_remove(struct ukevent *uk, struct kevent_user *u)
+{
+	int err = -ENODEV;
+	struct kevent *k;
+	unsigned long flags;
+
+	spin_lock_irqsave(&u->kevent_lock, flags);
+	k = __kevent_search(&uk->id, uk->type, u);
+	if (k) {
+		__kevent_finish_user(k, 1);
+		err = 0;
+	}
+	spin_unlock_irqrestore(&u->kevent_lock, flags);
+
+	return err;
+}
+
+/*
+ * Detaches userspace control block from file descriptor
+ * and decrease it's reference counter.
+ * No new kevents can be added or removed from any list at this point.
+ */
+static int kevent_user_release(struct inode *inode, struct file *file)
+{
+	struct kevent_user *u = file->private_data;
+	struct kevent *k;
+	struct rb_node *n;
+
+	for (n = rb_first(&u->kevent_root); n; n = rb_next(n)) {
+		k = rb_entry(n, struct kevent, kevent_node);
+		kevent_finish_user(k, 1);
+	}
+
+	kevent_user_put(u);
+	file->private_data = NULL;
+
+	return 0;
+}
+
+/*
+ * Read requested number of ukevents in one shot.
+ */
+static struct ukevent *kevent_get_user(unsigned int num, void __user *arg)
+{
+	struct ukevent *ukev;
+
+	ukev = kmalloc(sizeof(struct ukevent) * num, GFP_KERNEL);
+	if (!ukev)
+		return NULL;
+
+	if (copy_from_user(ukev, arg, sizeof(struct ukevent) * num)) {
+		kfree(ukev);
+		return NULL;
+	}
+
+	return ukev;
+}
+
+static int kevent_mark_ready(struct ukevent *uk, struct kevent_user *u)
+{
+	struct kevent *k;
+	int err = -ENODEV;
+	unsigned long flags;
+
+	spin_lock_irqsave(&u->kevent_lock, flags);
+	k = __kevent_search(&uk->id, uk->type, u);
+	if (k) {
+		spin_lock(&k->st->lock);
+		kevent_ready(k, 1);
+		spin_unlock(&k->st->lock);
+		err = 0;
+	}
+	spin_unlock_irqrestore(&u->kevent_lock, flags);
+
+	return err;
+}
+
+/*
+ * Mark appropriate kevents as ready.
+ * If number of events is zero just wake up one listener.
+ */
+static int kevent_user_ctl_ready(struct kevent_user *u, unsigned int num, void __user *arg)
+{
+	int err = -EINVAL, cerr = 0, rnum = 0, i;
+	void __user *orig = arg;
+	struct ukevent uk;
+
+	if (num > u->kevent_num)
+		return err;
+	
+	if (!num) {
+		u->need_exit = 1;
+		wake_up(&u->wait);
+		return 0;
+	}
+
+	mutex_lock(&u->ctl_mutex);
+
+	if (num > KEVENT_MIN_BUFFS_ALLOC) {
+		struct ukevent *ukev;
+
+		ukev = kevent_get_user(num, arg);
+		if (ukev) {
+			for (i = 0; i < num; ++i) {
+				err = kevent_mark_ready(&ukev[i], u);
+				if (err) {
+					if (i != rnum)
+						memcpy(&ukev[rnum], &ukev[i], sizeof(struct ukevent));
+					rnum++;
+				}
+			}
+			if (copy_to_user(orig, ukev, rnum*sizeof(struct ukevent)))
+				cerr = -EFAULT;
+			kfree(ukev);
+			goto out_setup;
+		}
+	}
+
+	for (i = 0; i < num; ++i) {
+		if (copy_from_user(&uk, arg, sizeof(struct ukevent))) {
+			cerr = -EFAULT;
+			break;
+		}
+		arg += sizeof(struct ukevent);
+
+		err = kevent_mark_ready(&uk, u);
+		if (err) {
+			if (copy_to_user(orig, &uk, sizeof(struct ukevent))) {
+				cerr = -EFAULT;
+				break;
+			}
+			orig += sizeof(struct ukevent);
+			rnum++;
+		}
+	}
+
+out_setup:
+	if (cerr < 0) {
+		err = cerr;
+		goto out_remove;
+	}
+
+	err = num - rnum;
+out_remove:
+	mutex_unlock(&u->ctl_mutex);
+
+	return err;
+}
+
+/*
+ * Read from userspace all ukevents and modify appropriate kevents.
+ * If provided number of ukevents is more that threshold, it is faster
+ * to allocate a room for them and copy in one shot instead of copy
+ * one-by-one and then process them.
+ */
+static int kevent_user_ctl_modify(struct kevent_user *u, unsigned int num, void __user *arg)
+{
+	int err = 0, i;
+	struct ukevent uk;
+
+	mutex_lock(&u->ctl_mutex);
+
+	if (num > u->kevent_num) {
+		err = -EINVAL;
+		goto out;
+	}
+
+	if (num > KEVENT_MIN_BUFFS_ALLOC) {
+		struct ukevent *ukev;
+
+		ukev = kevent_get_user(num, arg);
+		if (ukev) {
+			for (i = 0; i < num; ++i) {
+				if (kevent_modify(&ukev[i], u))
+					ukev[i].ret_flags |= KEVENT_RET_BROKEN;
+				ukev[i].ret_flags |= KEVENT_RET_DONE;
+			}
+			if (copy_to_user(arg, ukev, num*sizeof(struct ukevent)))
+				err = -EFAULT;
+			kfree(ukev);
+			goto out;
+		}
+	}
+
+	for (i = 0; i < num; ++i) {
+		if (copy_from_user(&uk, arg, sizeof(struct ukevent))) {
+			err = -EFAULT;
+			break;
+		}
+
+		if (kevent_modify(&uk, u))
+			uk.ret_flags |= KEVENT_RET_BROKEN;
+		uk.ret_flags |= KEVENT_RET_DONE;
+
+		if (copy_to_user(arg, &uk, sizeof(struct ukevent))) {
+			err = -EFAULT;
+			break;
+		}
+
+		arg += sizeof(struct ukevent);
+	}
+out:
+	mutex_unlock(&u->ctl_mutex);
+
+	return err;
+}
+
+/*
+ * Read from userspace all ukevents and remove appropriate kevents.
+ * If provided number of ukevents is more that threshold, it is faster
+ * to allocate a room for them and copy in one shot instead of copy
+ * one-by-one and then process them.
+ */
+static int kevent_user_ctl_remove(struct kevent_user *u, unsigned int num, void __user *arg)
+{
+	int err = 0, i;
+	struct ukevent uk;
+
+	mutex_lock(&u->ctl_mutex);
+
+	if (num > u->kevent_num) {
+		err = -EINVAL;
+		goto out;
+	}
+
+	if (num > KEVENT_MIN_BUFFS_ALLOC) {
+		struct ukevent *ukev;
+
+		ukev = kevent_get_user(num, arg);
+		if (ukev) {
+			for (i = 0; i < num; ++i) {
+				if (kevent_remove(&ukev[i], u))
+					ukev[i].ret_flags |= KEVENT_RET_BROKEN;
+				ukev[i].ret_flags |= KEVENT_RET_DONE;
+			}
+			if (copy_to_user(arg, ukev, num*sizeof(struct ukevent)))
+				err = -EFAULT;
+			kfree(ukev);
+			goto out;
+		}
+	}
+
+	for (i = 0; i < num; ++i) {
+		if (copy_from_user(&uk, arg, sizeof(struct ukevent))) {
+			err = -EFAULT;
+			break;
+		}
+
+		if (kevent_remove(&uk, u))
+			uk.ret_flags |= KEVENT_RET_BROKEN;
+
+		uk.ret_flags |= KEVENT_RET_DONE;
+
+		if (copy_to_user(arg, &uk, sizeof(struct ukevent))) {
+			err = -EFAULT;
+			break;
+		}
+
+		arg += sizeof(struct ukevent);
+	}
+out:
+	mutex_unlock(&u->ctl_mutex);
+
+	return err;
+}
+
+/*
+ * Queue kevent into userspace control block and increase
+ * it's reference counter.
+ */
+static int kevent_user_enqueue(struct kevent_user *u, struct kevent *new)
+{
+	unsigned long flags;
+	struct rb_node **p = &u->kevent_root.rb_node, *parent = NULL;
+	struct kevent *k;
+	int err = 0, cmp;
+
+	spin_lock_irqsave(&u->kevent_lock, flags);
+	while (*p) {
+		parent = *p;
+		k = rb_entry(parent, struct kevent, kevent_node);
+
+		cmp = kevent_compare_id(&k->event.id, &new->event.id,
+				k->event.type, new->event.type);
+		if (cmp > 0)
+			p = &parent->rb_right;
+		else if (cmp < 0)
+			p = &parent->rb_left;
+		else {
+			err = -EEXIST;
+			break;
+		}
+	}
+	if (likely(!err)) {
+		rb_link_node(&new->kevent_node, parent, p);
+		rb_insert_color(&new->kevent_node, &u->kevent_root);
+		new->flags |= KEVENT_USER;
+		u->kevent_num++;
+		kevent_user_get(u);
+	}
+	spin_unlock_irqrestore(&u->kevent_lock, flags);
+
+	return err;
+}
+
+/*
+ * Add kevent from both kernel and userspace users.
+ * This function allocates and queues kevent, returns negative value
+ * on error, positive if kevent is ready immediately and zero
+ * if kevent has been queued.
+ */
+int kevent_user_add_ukevent(struct ukevent *uk, struct kevent_user *u)
+{
+	struct kevent *k;
+	int err;
+
+	k = kmem_cache_alloc(kevent_cache, GFP_KERNEL);
+	if (!k) {
+		err = -ENOMEM;
+		goto err_out_exit;
+	}
+
+	memcpy(&k->event, uk, sizeof(struct ukevent));
+	INIT_RCU_HEAD(&k->rcu_head);
+
+	k->event.ret_flags = 0;
+
+	err = kevent_init(k);
+	if (err) {
+		kmem_cache_free(kevent_cache, k);
+		goto err_out_exit;
+	}
+	k->user = u;
+	kevent_stat_total(u);
+	err = kevent_user_enqueue(u, k);
+	if (err) {
+		kmem_cache_free(kevent_cache, k);
+		goto err_out_exit;
+	}
+
+	err = kevent_enqueue(k);
+	if (err) {
+		memcpy(uk, &k->event, sizeof(struct ukevent));
+		kevent_finish_user(k, 0);
+		goto err_out_exit;
+	}
+
+	return 0;
+
+err_out_exit:
+	if (err < 0) {
+		uk->ret_flags |= KEVENT_RET_BROKEN | KEVENT_RET_DONE;
+		uk->ret_data[1] = err;
+	} else if (err > 0)
+		uk->ret_flags |= KEVENT_RET_DONE;
+	return err;
+}
+
+
+/*
+ * Copy all ukevents from userspace, allocate kevent for each one
+ * and add them into appropriate kevent_storages,
+ * e.g. sockets, inodes and so on...
+ * Ready events will replace ones provided by used and number
+ * of ready events is returned.
+ * User must check ret_flags field of each ukevent structure
+ * to determine if it is fired or failed event.
+ */
+static int kevent_user_ctl_add(struct kevent_user *u, unsigned int num, void __user *arg)
+{
+	int err, cerr = 0, rnum = 0, i;
+	void __user *orig = arg;
+	struct ukevent uk;
+
+	mutex_lock(&u->ctl_mutex);
+
+	err = -EINVAL;
+	if (num > KEVENT_MIN_BUFFS_ALLOC) {
+		struct ukevent *ukev;
+
+		ukev = kevent_get_user(num, arg);
+		if (ukev) {
+			for (i = 0; i < num; ++i) {
+				if (!kevent_event_is_allowed(&ukev[i])) {
+					if (i != rnum)
+						memcpy(&ukev[rnum], 
+							&ukev[i], 
+							sizeof(struct ukevent));
+					rnum++;
+				} else {
+					err = kevent_user_add_ukevent(&ukev[i], u);
+					if (err) {
+						kevent_stat_im(u);
+						if (i != rnum)
+							memcpy(&ukev[rnum], 
+								&ukev[i], 
+								sizeof(struct ukevent));
+						rnum++;
+					}
+				}
+			}
+			if (copy_to_user(orig, ukev, rnum*sizeof(struct ukevent)))
+				cerr = -EFAULT;
+			kfree(ukev);
+			goto out_setup;
+		}
+	}
+
+	for (i = 0; i < num; ++i) {
+		if (copy_from_user(&uk, arg, sizeof(struct ukevent))) {
+			cerr = -EFAULT;
+			break;
+		}
+		arg += sizeof(struct ukevent);
+
+		if (!kevent_event_is_allowed(&uk)) {
+			err = 1;
+		} else {
+			err = kevent_user_add_ukevent(&uk, u);
+			if (err)
+				kevent_stat_im(u);
+		}
+		if (err) {
+			if (copy_to_user(orig, &uk, sizeof(struct ukevent))) {
+				cerr = -EFAULT;
+				break;
+			}
+			orig += sizeof(struct ukevent);
+			rnum++;
+		}
+	}
+
+out_setup:
+	if (cerr < 0) {
+		err = cerr;
+		goto out_remove;
+	}
+
+	err = rnum;
+out_remove:
+	mutex_unlock(&u->ctl_mutex);
+
+	return err;
+}
+
+/* Used to wakeup waiting syscalls in case high-resolution timer is used. */
+static int kevent_user_wake(struct hrtimer *timer)
+{
+	struct kevent_user *u = container_of(timer, struct kevent_user, timer);
+
+	u->need_exit = 1;
+	wake_up(&u->wait);
+
+	return HRTIMER_NORESTART;
+}
+
+
+/*
+ * In nonblocking mode it returns as many events as possible, but not more than @max_nr.
+ * In blocking mode it waits until timeout or if at least @min_nr events are ready.
+ */
+static int kevent_user_wait(struct file *file, struct kevent_user *u,
+		unsigned int min_nr, unsigned int max_nr, struct timespec timeout,
+		void __user *buf, unsigned int flags)
+{
+	struct kevent *k;
+	int num = 0;
+	long tm = MAX_SCHEDULE_TIMEOUT;
+
+	if (!(file->f_flags & O_NONBLOCK)) {
+		if (!timespec_valid(&timeout))
+			return -EINVAL;
+
+		if (flags & KEVENT_FLAGS_ABSTIME) {
+			hrtimer_cancel(&u->timer);
+			hrtimer_init(&u->timer, CLOCK_REALTIME, HRTIMER_ABS);
+			u->timer.expires = ktime_set(timeout.tv_sec, timeout.tv_nsec);
+			u->timer.function = &kevent_user_wake;
+			hrtimer_start(&u->timer, u->timer.expires, HRTIMER_ABS);
+			if (unlikely(kevent_debug_abstime == 0)) {
+				printk(KERN_INFO "kevent: author was wrong, "
+						"someone uses absolute time in %s, "
+						"please report to remove this warning.\n", __func__);
+				kevent_debug_abstime = 1;
+			}
+		} else {
+			tm = timespec_to_jiffies(&timeout);
+		}
+
+		wait_event_interruptible_timeout(u->wait,
+			((u->ready_num >= 1) && kevent_ring_space(u)) || u->need_exit, tm);
+	}
+	u->need_exit = 0;
+
+	while (num < max_nr && ((k = kevent_dequeue_ready(u)) != NULL)) {
+		if (copy_to_user(buf + num*sizeof(struct ukevent),
+					&k->event, sizeof(struct ukevent))) {
+			if (num == 0)
+				num = -EFAULT;
+			break;
+		}
+		kevent_complete_ready(k);
+		++num;
+		kevent_stat_wait(u);
+	}
+
+	return num;
+}
+
+struct file_operations kevent_user_fops = {
+	.release	= kevent_user_release,
+	.poll		= kevent_user_poll,
+	.owner		= THIS_MODULE,
+};
+
+static int kevent_ctl_process(struct file *file, unsigned int cmd, unsigned int num, void __user *arg)
+{
+	int err;
+	struct kevent_user *u = file->private_data;
+
+	switch (cmd) {
+	case KEVENT_CTL_ADD:
+		err = kevent_user_ctl_add(u, num, arg);
+		break;
+	case KEVENT_CTL_REMOVE:
+		err = kevent_user_ctl_remove(u, num, arg);
+		break;
+	case KEVENT_CTL_MODIFY:
+		err = kevent_user_ctl_modify(u, num, arg);
+		break;
+	case KEVENT_CTL_READY:
+		err = kevent_user_ctl_ready(u, num, arg);
+		break;
+	default:
+		err = -EINVAL;
+		break;
+	}
+
+	return err;
+}
+
+/*
+ * Used to get ready kevents from queue.
+ * @ctl_fd - kevent control descriptor which must be obtained through kevent_ctl(KEVENT_CTL_INIT).
+ * @min_nr - minimum number of ready kevents.
+ * @max_nr - maximum number of ready kevents.
+ * @timeout - time to wait until some events are ready.
+ * @buf - buffer to place ready events.
+ * @flags - various flags (see include/linux/ukevent.h KEVENT_FLAGS_*).
+ */
+asmlinkage long sys_kevent_get_events(int ctl_fd, unsigned int min_nr, unsigned int max_nr,
+		struct timespec timeout, struct ukevent __user *buf, unsigned flags)
+{
+	int err = -EINVAL;
+	struct file *file;
+	struct kevent_user *u;
+
+	file = fget(ctl_fd);
+	if (!file)
+		return -EBADF;
+
+	if (file->f_op != &kevent_user_fops)
+		goto out_fput;
+	u = file->private_data;
+
+	err = kevent_user_wait(file, u, min_nr, max_nr, timeout, buf, flags);
+out_fput:
+	fput(file);
+	return err;
+}
+
+static struct vfsmount *kevent_mnt __read_mostly;
+
+static int kevent_get_sb(struct file_system_type *fs_type, int flags,
+		   const char *dev_name, void *data, struct vfsmount *mnt)
+{
+	return get_sb_pseudo(fs_type, "kevent", NULL, 0xaabbccdd, mnt);
+}
+
+static struct file_system_type kevent_fs_type = {
+	.name		= "keventfs",
+	.get_sb		= kevent_get_sb,
+	.kill_sb	= kill_anon_super,
+};
+
+static int keventfs_delete_dentry(struct dentry *dentry)
+{
+	return 1;
+}
+
+static struct dentry_operations keventfs_dentry_operations = {
+	.d_delete	= keventfs_delete_dentry,
+};
+
+asmlinkage long sys_kevent_init(struct kevent_ring __user *ring, unsigned int num, unsigned int flags)
+{
+	struct qstr this;
+	char name[32];
+	struct dentry *dentry;
+	struct inode *inode;
+	struct file *file;
+	int err = -ENFILE, fd;
+	struct kevent_user *u;
+
+	if ((ring && !num) || (!ring && num) || (num == 1))
+		return -EINVAL;
+
+	file = get_empty_filp();
+	if (!file)
+		goto err_out_exit;
+
+	inode = new_inode(kevent_mnt->mnt_sb);
+	if (!inode)
+		goto err_out_fput;
+
+	inode->i_fop = &kevent_user_fops;
+
+	inode->i_state = I_DIRTY;
+	inode->i_mode = S_IRUSR | S_IWUSR;
+	inode->i_uid = current->fsuid;
+	inode->i_gid = current->fsgid;
+	inode->i_atime = inode->i_mtime = inode->i_ctime = CURRENT_TIME;
+
+	err = get_unused_fd();
+	if (err < 0)
+		goto err_out_iput;
+	fd = err;
+
+	err = -ENOMEM;
+	u = kevent_user_alloc(ring, num);
+	if (!u)
+		goto err_out_put_fd;
+
+	sprintf(name, "[%lu]", inode->i_ino);
+	this.name = name;
+	this.len = strlen(name);
+	this.hash = inode->i_ino;
+	dentry = d_alloc(kevent_mnt->mnt_sb->s_root, &this);
+	if (!dentry)
+		goto err_out_free;
+	dentry->d_op = &keventfs_dentry_operations;
+	d_add(dentry, inode);
+	file->f_vfsmnt = mntget(kevent_mnt);
+	file->f_dentry = dentry;
+	file->f_mapping = inode->i_mapping;
+	file->f_pos = 0;
+	file->f_flags = O_RDONLY;
+	file->f_op = &kevent_user_fops;
+	file->f_mode = FMODE_READ;
+	file->f_version = 0;
+	file->private_data = u;
+
+	fd_install(fd, file);
+
+	return fd;
+
+err_out_free:
+	kmem_cache_free(kevent_user_cache, u);
+err_out_put_fd:
+	put_unused_fd(fd);
+err_out_iput:
+	iput(inode);
+err_out_fput:
+	put_filp(file);
+err_out_exit:
+	return err;
+}
+
+/*
+ * Commits user's index (consumer index).
+ * Must be called under u->ring_lock mutex held.
+ */
+static int __kevent_user_commit(struct kevent_user *u, unsigned int new_uidx, unsigned int over)
+{
+	int err = -EOVERFLOW, comm = 0;
+	struct kevent_ring __user *ring = u->pring;
+
+	if (!ring) {
+		err = 0;
+		goto err_out_exit;
+	}
+
+	if (new_uidx >= u->ring_size) {
+		err = -EINVAL;
+		goto err_out_exit;
+	}
+
+	if ((over != u->ring_over - 1) && (over != u->ring_over))
+		goto err_out_exit;
+
+	if (u->uidx < u->kidx && new_uidx > u->kidx) {
+		err = -EINVAL;
+		goto err_out_exit;
+	}
+
+	if (new_uidx > u->uidx) {
+		if (over != u->ring_over)
+			goto err_out_exit;
+
+		comm = new_uidx - u->uidx;
+		u->uidx = new_uidx;
+		u->full = 0;
+	} else if (new_uidx < u->uidx) {
+		comm = u->ring_size - (u->uidx - new_uidx);
+		u->uidx = new_uidx;
+		u->full = 0;
+		u->ring_over++;
+
+		if (put_user(u->ring_over, &ring->ring_over)) {
+			err = -EFAULT;
+			goto err_out_exit;
+		}
+	}
+
+	return comm;
+
+err_out_exit:
+	return err;
+}
+
+/*
+ * This syscall is used to perform waiting until there is free space in the ring
+ * buffer, in that case some events will be copied there.
+ * Function returns number of actually copied ready events in ring buffer.
+ * After this function is completed userspace ring->ring_kidx will be updated.
+ *
+ * @ctl_fd - kevent file descriptor.
+ * @num - number of kevents to process.
+ * @old_uidx - the last index user is aware of.
+ * @timeout - time to wait until there is free space in kevent queue.
+ * @flags - various flags (see include/linux/ukevent.h KEVENT_FLAGS_*).
+ *
+ * When we need to commit @num events, it means we should just remove first @num
+ * kevents from ready queue and copy them into the buffer. 
+ * Kevents will be copied into ring buffer in order they were placed into ready queue.
+ * One-shot kevents will be removed here, since there is no way they can be reused.
+ * Edge-triggered events will be requeued here for better performance.
+ */
+asmlinkage long sys_kevent_wait(int ctl_fd, unsigned int num, unsigned int old_uidx, 
+		struct timespec timeout, unsigned int flags)
+{
+	int err = -EINVAL, copied = 0;
+	struct file *file;
+	struct kevent_user *u;
+	struct kevent *k;
+	struct kevent_ring __user *ring;
+	long tm = MAX_SCHEDULE_TIMEOUT;
+	unsigned int i;
+
+	file = fget(ctl_fd);
+	if (!file)
+		return -EBADF;
+
+	if (file->f_op != &kevent_user_fops)
+		goto out_fput;
+	u = file->private_data;
+
+	ring = u->pring;
+	if (!ring || num > u->ring_size)
+		goto out_fput;
+#if 0
+	/*
+	 * Allow to immediately update ring index, but it is not supported,
+	 * since syscall() has limited number of arguments which is actually
+	 * a good idea - use kevent_commit() instead.
+	 */
+	if ((u->uidx != new_uidx) && (new_uidx != 0xffffffff)) {
+		mutex_lock(&u->ring_lock);
+		__kevent_user_commit(u, new_uidx, over);
+		mutex_unlock(&u->ring_lock);
+	}
+#endif
+
+	if (!(file->f_flags & O_NONBLOCK)) {
+		if (!timespec_valid(&timeout))
+			goto out_fput;
+
+		if (flags & KEVENT_FLAGS_ABSTIME) {
+			hrtimer_cancel(&u->timer);
+			hrtimer_init(&u->timer, CLOCK_REALTIME, HRTIMER_ABS);
+			u->timer.expires = ktime_set(timeout.tv_sec, timeout.tv_nsec);
+			u->timer.function = &kevent_user_wake;
+			hrtimer_start(&u->timer, u->timer.expires, HRTIMER_ABS);
+			if (unlikely(kevent_debug_abstime == 0)) {
+				printk(KERN_INFO "kevent: author was wrong, "
+						"someone uses absolute time in %s, "
+						"please report to remove this warning.\n", __func__);
+				kevent_debug_abstime = 1;
+			}
+		} else {
+			tm = timespec_to_jiffies(&timeout);
+		}
+
+		wait_event_interruptible_timeout(u->wait,
+			((u->ready_num >= 1) && kevent_ring_space(u)) || 
+				u->need_exit || old_uidx != u->uidx,
+				tm);
+	}
+	u->need_exit = 0;
+
+	for (i=0; i<num; ++i) {
+		k = kevent_dequeue_ready_ring(u);
+		if (!k)
+			break;
+		kevent_complete_ready(k);
+
+		if (k->event.ret_flags & KEVENT_RET_COPY_FAILED)
+			break;
+		kevent_stat_ring(u);
+		copied++;
+	}
+
+	fput(file);
+
+	return copied;
+out_fput:
+	fput(file);
+	return err;
+}
+
+/*
+ * This syscall is used to commit events in ring buffer, i.e. mark appropriate
+ * entries as unused by userspace so subsequent kevent_wait() could overwrite them.
+ * This fucntion returns actual number of kevents which were committed.
+ * After this function is completed userspace ring->ring_over can be updated.
+ *
+ * @ctl_fd - kevent file descriptor.
+ * @new_uidx - the last committed kevent.
+ * @over - number of overflows given queue had.
+ */
+asmlinkage long sys_kevent_commit(int ctl_fd, unsigned int new_uidx, unsigned int over)
+{
+	int err = -EINVAL, comm = 0;
+	struct file *file;
+	struct kevent_user *u;
+
+	file = fget(ctl_fd);
+	if (!file)
+		return -EBADF;
+
+	if (file->f_op != &kevent_user_fops)
+		goto out_fput;
+	u = file->private_data;
+
+	mutex_lock(&u->ring_lock);
+	err = __kevent_user_commit(u, new_uidx, over);
+	if (err < 0)
+		goto err_out_unlock;
+	comm = err;
+	mutex_unlock(&u->ring_lock);
+
+	fput(file);
+
+	return comm;
+
+err_out_unlock:
+	mutex_unlock(&u->ring_lock);
+out_fput:
+	fput(file);
+	return err;
+}
+
+/*
+ * This syscall is used to perform various control operations
+ * on given kevent queue, which is obtained through kevent file descriptor @fd.
+ * @cmd - type of operation.
+ * @num - number of kevents to be processed.
+ * @arg - pointer to array of struct ukevent.
+ */
+asmlinkage long sys_kevent_ctl(int fd, unsigned int cmd, unsigned int num, struct ukevent __user *arg)
+{
+	int err = -EINVAL;
+	struct file *file;
+
+	file = fget(fd);
+	if (!file)
+		return -EBADF;
+
+	if (file->f_op != &kevent_user_fops)
+		goto out_fput;
+
+	err = kevent_ctl_process(file, cmd, num, arg);
+
+out_fput:
+	fput(file);
+	return err;
+}
+
+/*
+ * Kevent subsystem initialization - create caches and register
+ * filesystem to get control file descriptors from.
+ */
+static int __init kevent_user_init(void)
+{
+	int err = 0;
+
+	kevent_cache = kmem_cache_create("kevent_cache",
+			sizeof(struct kevent), 0, SLAB_PANIC, NULL, NULL);
+	
+	kevent_user_cache = kmem_cache_create("kevent_user_cache",
+			sizeof(struct kevent_user), 0, SLAB_PANIC, NULL, NULL);
+	
+	err = register_filesystem(&kevent_fs_type);
+	if (err)
+		goto err_out_exit;
+	
+	kevent_mnt = kern_mount(&kevent_fs_type);
+	err = PTR_ERR(kevent_mnt);
+	if (IS_ERR(kevent_mnt))
+		goto err_out_unreg;
+
+	printk(KERN_INFO "KEVENT subsystem has been successfully registered.\n");
+
+	return 0;
+
+err_out_unreg:
+	unregister_filesystem(&kevent_fs_type);
+err_out_exit:
+	kmem_cache_destroy(kevent_cache);
+	return err;
+}
+
+module_init(kevent_user_init);
diff --git a/kernel/sys_ni.c b/kernel/sys_ni.c
index d7306d0..22a775a 100644
--- a/kernel/sys_ni.c
+++ b/kernel/sys_ni.c
@@ -123,6 +123,15 @@ cond_syscall(ppc_rtas);
 cond_syscall(sys_spu_run);
 cond_syscall(sys_spu_create);
 
+cond_syscall(sys_kevent_get_events);
+cond_syscall(sys_kevent_ctl);
+cond_syscall(sys_kevent_wait);
+cond_syscall(sys_kevent_commit);
+cond_syscall(sys_kevent_init);
+
+cond_syscall(sys_aio_sendfile);
+cond_syscall(sys_aio_sendfile_path);
+
 /* mmu depending weak syscall entries */
 cond_syscall(sys_mprotect);
 cond_syscall(sys_msync);

[take35 4/10] kevent: Socket notifications.

[Evgeniy Polyakov]

Socket notifications.

This patch includes socket send/recv/accept notifications.

Signed-off-by: Evgeniy Polyakov <johnpol@2ka.mipt.ru>

diff --git a/fs/inode.c b/fs/inode.c
index bf21dc6..82817b1 100644
--- a/fs/inode.c
+++ b/fs/inode.c
@@ -21,6 +21,7 @@
 #include <linux/cdev.h>
 #include <linux/bootmem.h>
 #include <linux/inotify.h>
+#include <linux/kevent.h>
 #include <linux/mount.h>
 
 /*
@@ -164,12 +165,18 @@ static struct inode *alloc_inode(struct super_block *sb)
 		}
 		inode->i_private = NULL;
 		inode->i_mapping = mapping;
+#if defined CONFIG_KEVENT_SOCKET || defined CONFIG_KEVENT_PIPE
+		kevent_storage_init(inode, &inode->st);
+#endif
 	}
 	return inode;
 }
 
 void destroy_inode(struct inode *inode) 
 {
+#if defined CONFIG_KEVENT_SOCKET || defined CONFIG_KEVENT_PIPE
+	kevent_storage_fini(&inode->st);
+#endif
 	BUG_ON(inode_has_buffers(inode));
 	security_inode_free(inode);
 	if (inode->i_sb->s_op->destroy_inode)
diff --git a/include/net/sock.h b/include/net/sock.h
index 03684e7..d840399 100644
--- a/include/net/sock.h
+++ b/include/net/sock.h
@@ -49,6 +49,7 @@
 #include <linux/skbuff.h>	/* struct sk_buff */
 #include <linux/mm.h>
 #include <linux/security.h>
+#include <linux/kevent.h>
 
 #include <linux/filter.h>
 
@@ -451,6 +452,21 @@ static inline int sk_stream_memory_free(struct sock *sk)
 
 extern void sk_stream_rfree(struct sk_buff *skb);
 
+struct socket_alloc {
+	struct socket socket;
+	struct inode vfs_inode;
+};
+
+static inline struct socket *SOCKET_I(struct inode *inode)
+{
+	return &container_of(inode, struct socket_alloc, vfs_inode)->socket;
+}
+
+static inline struct inode *SOCK_INODE(struct socket *socket)
+{
+	return &container_of(socket, struct socket_alloc, socket)->vfs_inode;
+}
+
 static inline void sk_stream_set_owner_r(struct sk_buff *skb, struct sock *sk)
 {
 	skb->sk = sk;
@@ -478,6 +494,7 @@ static inline void sk_add_backlog(struct sock *sk, struct sk_buff *skb)
 		sk->sk_backlog.tail = skb;
 	}
 	skb->next = NULL;
+	kevent_socket_notify(sk, KEVENT_SOCKET_RECV);
 }
 
 #define sk_wait_event(__sk, __timeo, __condition)		\
@@ -679,21 +696,6 @@ static inline struct kiocb *siocb_to_kiocb(struct sock_iocb *si)
 	return si->kiocb;
 }
 
-struct socket_alloc {
-	struct socket socket;
-	struct inode vfs_inode;
-};
-
-static inline struct socket *SOCKET_I(struct inode *inode)
-{
-	return &container_of(inode, struct socket_alloc, vfs_inode)->socket;
-}
-
-static inline struct inode *SOCK_INODE(struct socket *socket)
-{
-	return &container_of(socket, struct socket_alloc, socket)->vfs_inode;
-}
-
 extern void __sk_stream_mem_reclaim(struct sock *sk);
 extern int sk_stream_mem_schedule(struct sock *sk, int size, int kind);
 
diff --git a/include/net/tcp.h b/include/net/tcp.h
index cd8fa0c..7344d61 100644
--- a/include/net/tcp.h
+++ b/include/net/tcp.h
@@ -864,6 +864,7 @@ static inline int tcp_prequeue(struct sock *sk, struct sk_buff *skb)
 			tp->ucopy.memory = 0;
 		} else if (skb_queue_len(&tp->ucopy.prequeue) == 1) {
 			wake_up_interruptible(sk->sk_sleep);
+			kevent_socket_notify(sk, KEVENT_SOCKET_RECV|KEVENT_SOCKET_SEND);
 			if (!inet_csk_ack_scheduled(sk))
 				inet_csk_reset_xmit_timer(sk, ICSK_TIME_DACK,
 						          (3 * TCP_RTO_MIN) / 4,
diff --git a/kernel/kevent/kevent_socket.c b/kernel/kevent/kevent_socket.c
new file mode 100644
index 0000000..d1a2701
--- /dev/null
+++ b/kernel/kevent/kevent_socket.c
@@ -0,0 +1,144 @@
+/*
+ * 	kevent_socket.c
+ * 
+ * 2006 Copyright (c) Evgeniy Polyakov <johnpol@2ka.mipt.ru>
+ * 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 <linux/kernel.h>
+#include <linux/types.h>
+#include <linux/list.h>
+#include <linux/slab.h>
+#include <linux/spinlock.h>
+#include <linux/timer.h>
+#include <linux/file.h>
+#include <linux/tcp.h>
+#include <linux/kevent.h>
+
+#include <net/sock.h>
+#include <net/request_sock.h>
+#include <net/inet_connection_sock.h>
+
+static int kevent_socket_callback(struct kevent *k)
+{
+	struct inode *inode = k->st->origin;
+	unsigned int events = SOCKET_I(inode)->ops->poll(SOCKET_I(inode)->file, SOCKET_I(inode), NULL);
+
+	if ((events & (POLLIN | POLLRDNORM)) && (k->event.event & (KEVENT_SOCKET_RECV | KEVENT_SOCKET_ACCEPT)))
+		return 1;
+	if ((events & (POLLOUT | POLLWRNORM)) && (k->event.event & KEVENT_SOCKET_SEND))
+		return 1;
+	if (events & (POLLERR | POLLHUP | POLLRDHUP | POLLREMOVE))
+		return -1;
+	return 0;
+}
+
+int kevent_socket_enqueue(struct kevent *k)
+{
+	struct inode *inode;
+	struct socket *sock;
+	int err = -EBADF;
+
+	sock = sockfd_lookup(k->event.id.raw[0], &err);
+	if (!sock)
+		goto err_out_exit;
+
+	inode = igrab(SOCK_INODE(sock));
+	if (!inode)
+		goto err_out_fput;
+
+	err = kevent_storage_enqueue(&inode->st, k);
+	if (err)
+		goto err_out_iput;
+
+	if (k->event.req_flags & KEVENT_REQ_ALWAYS_QUEUE) {
+		kevent_requeue(k);
+		err = 0;
+	} else {
+		err = k->callbacks.callback(k);
+		if (err)
+			goto err_out_dequeue;
+	}
+
+	return err;
+
+err_out_dequeue:
+	kevent_storage_dequeue(k->st, k);
+err_out_iput:
+	iput(inode);
+err_out_fput:
+	sockfd_put(sock);
+err_out_exit:
+	return err;
+}
+
+int kevent_socket_dequeue(struct kevent *k)
+{
+	struct inode *inode = k->st->origin;
+	struct socket *sock;
+
+	kevent_storage_dequeue(k->st, k);
+
+	sock = SOCKET_I(inode);
+	iput(inode);
+	sockfd_put(sock);
+
+	return 0;
+}
+
+void kevent_socket_notify(struct sock *sk, u32 event)
+{
+	if (sk->sk_socket)
+		kevent_storage_ready(&SOCK_INODE(sk->sk_socket)->st, NULL, event);
+}
+
+/*
+ * It is required for network protocols compiled as modules, like IPv6.
+ */
+EXPORT_SYMBOL_GPL(kevent_socket_notify);
+
+#ifdef CONFIG_LOCKDEP
+static struct lock_class_key kevent_sock_key;
+
+void kevent_socket_reinit(struct socket *sock)
+{
+	struct inode *inode = SOCK_INODE(sock);
+
+	lockdep_set_class(&inode->st.lock, &kevent_sock_key);
+}
+
+void kevent_sk_reinit(struct sock *sk)
+{
+	if (sk->sk_socket) {
+		struct inode *inode = SOCK_INODE(sk->sk_socket);
+
+		lockdep_set_class(&inode->st.lock, &kevent_sock_key);
+	}
+}
+#endif
+static int __init kevent_init_socket(void)
+{
+	struct kevent_callbacks sc = {
+		.callback = &kevent_socket_callback,
+		.enqueue = &kevent_socket_enqueue,
+		.dequeue = &kevent_socket_dequeue,
+		.flags = 0,
+	};
+
+	return kevent_add_callbacks(&sc, KEVENT_SOCKET);
+}
+module_init(kevent_init_socket);
diff --git a/net/core/sock.c b/net/core/sock.c
index 0ed5b4f..e687f54 100644
--- a/net/core/sock.c
+++ b/net/core/sock.c
@@ -1393,6 +1393,7 @@ static void sock_def_wakeup(struct sock *sk)
 	if (sk->sk_sleep && waitqueue_active(sk->sk_sleep))
 		wake_up_interruptible_all(sk->sk_sleep);
 	read_unlock(&sk->sk_callback_lock);
+	kevent_socket_notify(sk, KEVENT_SOCKET_RECV|KEVENT_SOCKET_SEND);
 }
 
 static void sock_def_error_report(struct sock *sk)
@@ -1402,6 +1403,7 @@ static void sock_def_error_report(struct sock *sk)
 		wake_up_interruptible(sk->sk_sleep);
 	sk_wake_async(sk,0,POLL_ERR); 
 	read_unlock(&sk->sk_callback_lock);
+	kevent_socket_notify(sk, KEVENT_SOCKET_RECV|KEVENT_SOCKET_SEND);
 }
 
 static void sock_def_readable(struct sock *sk, int len)
@@ -1411,6 +1413,7 @@ static void sock_def_readable(struct sock *sk, int len)
 		wake_up_interruptible(sk->sk_sleep);
 	sk_wake_async(sk,1,POLL_IN);
 	read_unlock(&sk->sk_callback_lock);
+	kevent_socket_notify(sk, KEVENT_SOCKET_RECV|KEVENT_SOCKET_SEND);
 }
 
 static void sock_def_write_space(struct sock *sk)
@@ -1430,6 +1433,7 @@ static void sock_def_write_space(struct sock *sk)
 	}
 
 	read_unlock(&sk->sk_callback_lock);
+	kevent_socket_notify(sk, KEVENT_SOCKET_SEND|KEVENT_SOCKET_RECV);
 }
 
 static void sock_def_destruct(struct sock *sk)
@@ -1480,6 +1484,8 @@ void sock_init_data(struct socket *sock, struct sock *sk)
 	sk->sk_state		=	TCP_CLOSE;
 	sk->sk_socket		=	sock;
 
+	kevent_sk_reinit(sk);
+
 	sock_set_flag(sk, SOCK_ZAPPED);
 
 	if(sock)
@@ -1546,8 +1552,10 @@ void fastcall release_sock(struct sock *sk)
 	if (sk->sk_backlog.tail)
 		__release_sock(sk);
 	sk->sk_lock.owner = NULL;
-	if (waitqueue_active(&sk->sk_lock.wq))
+	if (waitqueue_active(&sk->sk_lock.wq)) {
 		wake_up(&sk->sk_lock.wq);
+		kevent_socket_notify(sk, KEVENT_SOCKET_RECV|KEVENT_SOCKET_SEND);
+	}
 	spin_unlock_bh(&sk->sk_lock.slock);
 }
 EXPORT_SYMBOL(release_sock);
diff --git a/net/core/stream.c b/net/core/stream.c
index d1d7dec..2878c2a 100644
--- a/net/core/stream.c
+++ b/net/core/stream.c
@@ -36,6 +36,7 @@ void sk_stream_write_space(struct sock *sk)
 			wake_up_interruptible(sk->sk_sleep);
 		if (sock->fasync_list && !(sk->sk_shutdown & SEND_SHUTDOWN))
 			sock_wake_async(sock, 2, POLL_OUT);
+		kevent_socket_notify(sk, KEVENT_SOCKET_SEND|KEVENT_SOCKET_RECV);
 	}
 }
 
diff --git a/net/ipv4/tcp_input.c b/net/ipv4/tcp_input.c
index c26076f..8e56c71 100644
--- a/net/ipv4/tcp_input.c
+++ b/net/ipv4/tcp_input.c
@@ -3128,6 +3128,7 @@ static void tcp_ofo_queue(struct sock *sk)
 
 		__skb_unlink(skb, &tp->out_of_order_queue);
 		__skb_queue_tail(&sk->sk_receive_queue, skb);
+		kevent_socket_notify(sk, KEVENT_SOCKET_RECV);
 		tp->rcv_nxt = TCP_SKB_CB(skb)->end_seq;
 		if(skb->h.th->fin)
 			tcp_fin(skb, sk, skb->h.th);
diff --git a/net/ipv4/tcp_ipv4.c b/net/ipv4/tcp_ipv4.c
index 12de90a..fca282e 100644
--- a/net/ipv4/tcp_ipv4.c
+++ b/net/ipv4/tcp_ipv4.c
@@ -61,6 +61,7 @@
 #include <linux/jhash.h>
 #include <linux/init.h>
 #include <linux/times.h>
+#include <linux/kevent.h>
 
 #include <net/icmp.h>
 #include <net/inet_hashtables.h>
@@ -1391,6 +1392,7 @@ int tcp_v4_conn_request(struct sock *sk, struct sk_buff *skb)
 	   	reqsk_free(req);
 	} else {
 		inet_csk_reqsk_queue_hash_add(sk, req, TCP_TIMEOUT_INIT);
+		kevent_socket_notify(sk, KEVENT_SOCKET_ACCEPT);
 	}
 	return 0;
 
diff --git a/net/socket.c b/net/socket.c
index 4e39631..776dc2e 100644
--- a/net/socket.c
+++ b/net/socket.c
@@ -84,6 +84,7 @@
 #include <linux/kmod.h>
 #include <linux/audit.h>
 #include <linux/wireless.h>
+#include <linux/kevent.h>
 
 #include <asm/uaccess.h>
 #include <asm/unistd.h>
@@ -496,6 +497,8 @@ static struct socket *sock_alloc(void)
 	inode->i_uid = current->fsuid;
 	inode->i_gid = current->fsgid;
 
+	kevent_socket_reinit(sock);
+
 	get_cpu_var(sockets_in_use)++;
 	put_cpu_var(sockets_in_use);
 	return sock;
diff --git a/net/unix/af_unix.c b/net/unix/af_unix.c
index 2f208c7..835e20f 100644
--- a/net/unix/af_unix.c
+++ b/net/unix/af_unix.c
@@ -1563,8 +1563,10 @@ static int unix_dgram_recvmsg(struct kiocb *iocb, struct socket *sock,
 	struct scm_cookie tmp_scm;
 	struct sock *sk = sock->sk;
 	struct unix_sock *u = unix_sk(sk);
+	struct sock *other;
 	int noblock = flags & MSG_DONTWAIT;
 	struct sk_buff *skb;
+
 	int err;
 
 	err = -EOPNOTSUPP;
@@ -1580,6 +1582,12 @@ static int unix_dgram_recvmsg(struct kiocb *iocb, struct socket *sock,
 		goto out_unlock;
 
 	wake_up_interruptible(&u->peer_wait);
+	other =unix_peer_get(sk);
+	if (other) {
+		kevent_socket_notify(other, KEVENT_SOCKET_SEND);
+		sock_put(other);
+	} else
+		kevent_socket_notify(sk, KEVENT_SOCKET_RECV);
 
 	if (msg->msg_name)
 		unix_copy_addr(msg, skb->sk);
@@ -1674,7 +1682,7 @@ static int unix_stream_recvmsg(struct kiocb *iocb, struct socket *sock,
 {
 	struct sock_iocb *siocb = kiocb_to_siocb(iocb);
 	struct scm_cookie tmp_scm;
-	struct sock *sk = sock->sk;
+	struct sock *sk = sock->sk, *other;
 	struct unix_sock *u = unix_sk(sk);
 	struct sockaddr_un *sunaddr=msg->msg_name;
 	int copied = 0;
@@ -1803,6 +1811,14 @@ static int unix_stream_recvmsg(struct kiocb *iocb, struct socket *sock,
 		}
 	} while (size);
 
+	other =unix_peer_get(sk);
+	if (other) {
+		kevent_socket_notify(other, KEVENT_SOCKET_SEND);
+		sock_put(other);
+	}
+	if (sk->sk_shutdown & RCV_SHUTDOWN || !other)
+		kevent_socket_notify(sk, KEVENT_SOCKET_RECV);
+
 	mutex_unlock(&u->readlock);
 	scm_recv(sock, msg, siocb->scm, flags);
 out:
@@ -1824,6 +1840,10 @@ static int unix_shutdown(struct socket *sock, int mode)
 			sock_hold(other);
 		unix_state_wunlock(sk);
 		sk->sk_state_change(sk);
+		kevent_socket_notify(sk, KEVENT_SOCKET_SEND|KEVENT_SOCKET_RECV);
+		if (other)
+			kevent_socket_notify(other, KEVENT_SOCKET_SEND|KEVENT_SOCKET_RECV);
+		kevent_socket_notify(sk, KEVENT_SOCKET_SEND|KEVENT_SOCKET_RECV);
 
 		if (other &&
 			(sk->sk_type == SOCK_STREAM || sk->sk_type == SOCK_SEQPACKET)) {

[take35 5/10] kevent: Timer notifications.

[Evgeniy Polyakov]

Timer notifications.

Timer notifications can be used for fine grained per-process time 
management, since interval timers are very inconvenient to use, 
and they are limited.

This subsystem uses high-resolution timers.

id.raw[0] is used as number of seconds
id.raw[1] is used as number of nanoseconds

Signed-off-by: Evgeniy Polyakov <johnpol@2ka.mipt.ru>

diff --git a/kernel/kevent/kevent_timer.c b/kernel/kevent/kevent_timer.c
new file mode 100644
index 0000000..c21a155
--- /dev/null
+++ b/kernel/kevent/kevent_timer.c
@@ -0,0 +1,114 @@
+/*
+ * 2006 Copyright (c) Evgeniy Polyakov <johnpol@2ka.mipt.ru>
+ * 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 <linux/kernel.h>
+#include <linux/types.h>
+#include <linux/list.h>
+#include <linux/slab.h>
+#include <linux/spinlock.h>
+#include <linux/hrtimer.h>
+#include <linux/jiffies.h>
+#include <linux/kevent.h>
+
+struct kevent_timer
+{
+	struct hrtimer		ktimer;
+	struct kevent_storage	ktimer_storage;
+	struct kevent		*ktimer_event;
+};
+
+static int kevent_timer_func(struct hrtimer *timer)
+{
+	struct kevent_timer *t = container_of(timer, struct kevent_timer, ktimer);
+	struct kevent *k = t->ktimer_event;
+
+	kevent_storage_ready(&t->ktimer_storage, NULL, KEVENT_MASK_ALL);
+	hrtimer_forward(timer, timer->base->softirq_time,
+			ktime_set(k->event.id.raw[0], k->event.id.raw[1]));
+	return HRTIMER_RESTART;
+}
+
+static struct lock_class_key kevent_timer_key;
+
+static int kevent_timer_enqueue(struct kevent *k)
+{
+	int err;
+	struct kevent_timer *t;
+
+	t = kmalloc(sizeof(struct kevent_timer), GFP_KERNEL);
+	if (!t)
+		return -ENOMEM;
+
+	hrtimer_init(&t->ktimer, CLOCK_MONOTONIC, HRTIMER_REL);
+	t->ktimer.expires = ktime_set(k->event.id.raw[0], k->event.id.raw[1]);
+	t->ktimer.function = kevent_timer_func;
+	t->ktimer_event = k;
+
+	err = kevent_storage_init(&t->ktimer, &t->ktimer_storage);
+	if (err)
+		goto err_out_free;
+	lockdep_set_class(&t->ktimer_storage.lock, &kevent_timer_key);
+
+	err = kevent_storage_enqueue(&t->ktimer_storage, k);
+	if (err)
+		goto err_out_st_fini;
+
+	hrtimer_start(&t->ktimer, t->ktimer.expires, HRTIMER_REL);
+
+	return 0;
+
+err_out_st_fini:
+	kevent_storage_fini(&t->ktimer_storage);
+err_out_free:
+	kfree(t);
+
+	return err;
+}
+
+static int kevent_timer_dequeue(struct kevent *k)
+{
+	struct kevent_storage *st = k->st;
+	struct kevent_timer *t = container_of(st, struct kevent_timer, ktimer_storage);
+
+	hrtimer_cancel(&t->ktimer);
+	kevent_storage_dequeue(st, k);
+	kfree(t);
+
+	return 0;
+}
+
+static int kevent_timer_callback(struct kevent *k)
+{
+	k->event.ret_data[0] = jiffies_to_msecs(jiffies);
+	return 1;
+}
+
+static int __init kevent_init_timer(void)
+{
+	struct kevent_callbacks tc = {
+		.callback = &kevent_timer_callback,
+		.enqueue = &kevent_timer_enqueue,
+		.dequeue = &kevent_timer_dequeue,
+		.flags = 0,
+	};
+
+	return kevent_add_callbacks(&tc, KEVENT_TIMER);
+}
+module_init(kevent_init_timer);
+

[take35 6/10] kevent: Pipe notifications.

[Evgeniy Polyakov]

Pipe notifications.


diff --git a/fs/pipe.c b/fs/pipe.c
index 68090e8..0c75bf1 100644
--- a/fs/pipe.c
+++ b/fs/pipe.c
@@ -16,6 +16,7 @@
 #include <linux/uio.h>
 #include <linux/highmem.h>
 #include <linux/pagemap.h>
+#include <linux/kevent.h>
 
 #include <asm/uaccess.h>
 #include <asm/ioctls.h>
@@ -313,6 +314,7 @@ redo:
 			break;
 		}
 		if (do_wakeup) {
+			kevent_pipe_notify(inode, KEVENT_SOCKET_SEND);
 			wake_up_interruptible_sync(&pipe->wait);
  			kill_fasync(&pipe->fasync_writers, SIGIO, POLL_OUT);
 		}
@@ -322,6 +324,7 @@ redo:
 
 	/* Signal writers asynchronously that there is more room. */
 	if (do_wakeup) {
+		kevent_pipe_notify(inode, KEVENT_SOCKET_SEND);
 		wake_up_interruptible(&pipe->wait);
 		kill_fasync(&pipe->fasync_writers, SIGIO, POLL_OUT);
 	}
@@ -484,6 +487,7 @@ redo2:
 			break;
 		}
 		if (do_wakeup) {
+			kevent_pipe_notify(inode, KEVENT_SOCKET_RECV);
 			wake_up_interruptible_sync(&pipe->wait);
 			kill_fasync(&pipe->fasync_readers, SIGIO, POLL_IN);
 			do_wakeup = 0;
@@ -495,6 +499,7 @@ redo2:
 out:
 	mutex_unlock(&inode->i_mutex);
 	if (do_wakeup) {
+		kevent_pipe_notify(inode, KEVENT_SOCKET_RECV);
 		wake_up_interruptible(&pipe->wait);
 		kill_fasync(&pipe->fasync_readers, SIGIO, POLL_IN);
 	}
@@ -590,6 +595,7 @@ pipe_release(struct inode *inode, int decr, int decw)
 		free_pipe_info(inode);
 	} else {
 		wake_up_interruptible(&pipe->wait);
+		kevent_pipe_notify(inode, KEVENT_SOCKET_SEND|KEVENT_SOCKET_RECV);
 		kill_fasync(&pipe->fasync_readers, SIGIO, POLL_IN);
 		kill_fasync(&pipe->fasync_writers, SIGIO, POLL_OUT);
 	}
diff --git a/kernel/kevent/kevent_pipe.c b/kernel/kevent/kevent_pipe.c
new file mode 100644
index 0000000..91dc1eb
--- /dev/null
+++ b/kernel/kevent/kevent_pipe.c
@@ -0,0 +1,123 @@
+/*
+ * 	kevent_pipe.c
+ * 
+ * 2006 Copyright (c) Evgeniy Polyakov <johnpol@2ka.mipt.ru>
+ * 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 <linux/kernel.h>
+#include <linux/types.h>
+#include <linux/slab.h>
+#include <linux/spinlock.h>
+#include <linux/file.h>
+#include <linux/fs.h>
+#include <linux/kevent.h>
+#include <linux/pipe_fs_i.h>
+
+static int kevent_pipe_callback(struct kevent *k)
+{
+	struct inode *inode = k->st->origin;
+	struct pipe_inode_info *pipe = inode->i_pipe;
+	int nrbufs = pipe->nrbufs;
+
+	if (k->event.event & KEVENT_SOCKET_RECV && nrbufs > 0) {
+		if (!pipe->writers)
+			return -1;
+		return 1;
+	}
+	
+	if (k->event.event & KEVENT_SOCKET_SEND && nrbufs < PIPE_BUFFERS) {
+		if (!pipe->readers)
+			return -1;
+		return 1;
+	}
+
+	return 0;
+}
+
+int kevent_pipe_enqueue(struct kevent *k)
+{
+	struct file *pipe;
+	int err = -EBADF;
+	struct inode *inode;
+
+	pipe = fget(k->event.id.raw[0]);
+	if (!pipe)
+		goto err_out_exit;
+
+	inode = igrab(pipe->f_dentry->d_inode);
+	if (!inode)
+		goto err_out_fput;
+
+	err = -EINVAL;
+	if (!S_ISFIFO(inode->i_mode))
+		goto err_out_iput;
+
+	err = kevent_storage_enqueue(&inode->st, k);
+	if (err)
+		goto err_out_iput;
+
+	if (k->event.req_flags & KEVENT_REQ_ALWAYS_QUEUE) {
+		kevent_requeue(k);
+		err = 0;
+	} else {
+		err = k->callbacks.callback(k);
+		if (err)
+			goto err_out_dequeue;
+	}
+
+	fput(pipe);
+
+	return err;
+
+err_out_dequeue:
+	kevent_storage_dequeue(k->st, k);
+err_out_iput:
+	iput(inode);
+err_out_fput:
+	fput(pipe);
+err_out_exit:
+	return err;
+}
+
+int kevent_pipe_dequeue(struct kevent *k)
+{
+	struct inode *inode = k->st->origin;
+
+	kevent_storage_dequeue(k->st, k);
+	iput(inode);
+
+	return 0;
+}
+
+void kevent_pipe_notify(struct inode *inode, u32 event)
+{
+	kevent_storage_ready(&inode->st, NULL, event);
+}
+
+static int __init kevent_init_pipe(void)
+{
+	struct kevent_callbacks sc = {
+		.callback = &kevent_pipe_callback,
+		.enqueue = &kevent_pipe_enqueue,
+		.dequeue = &kevent_pipe_dequeue,
+		.flags = 0,
+	};
+
+	return kevent_add_callbacks(&sc, KEVENT_PIPE);
+}
+module_init(kevent_init_pipe);

[take35 7/10] kevent: Signal notifications.

[Evgeniy Polyakov]

Signal notifications.

This type of notifications allows to deliver signals through kevent queue.
One can find example application signal.c on project homepage.

If KEVENT_SIGNAL_NOMASK bit is set in raw_u64 id then signal will be
delivered only through queue, otherwise both delivery types are used - old
through update of mask of pending signals and through queue.

If signal is delivered only through kevent queue, mask of pending signals
is not updated at all, which is equal to putting signal into blocked mask,
but with delivery of that signal through kevent queue.

Signed-off-by: Evgeniy Polyakov <johnpol@2ka.mipt.ru>


diff --git a/include/linux/sched.h b/include/linux/sched.h
index 4463735..e7372f2 100644
--- a/include/linux/sched.h
+++ b/include/linux/sched.h
@@ -82,6 +82,7 @@ struct sched_param {
 #include <linux/resource.h>
 #include <linux/timer.h>
 #include <linux/hrtimer.h>
+#include <linux/kevent_storage.h>
 #include <linux/task_io_accounting.h>
 
 #include <asm/processor.h>
@@ -1048,6 +1049,10 @@ struct task_struct {
 #ifdef	CONFIG_TASK_DELAY_ACCT
 	struct task_delay_info *delays;
 #endif
+#ifdef CONFIG_KEVENT_SIGNAL
+	struct kevent_storage st;
+	u32 kevent_signals;
+#endif
 #ifdef CONFIG_FAULT_INJECTION
 	int make_it_fail;
 #endif
diff --git a/kernel/fork.c b/kernel/fork.c
index fc723e5..fd7c749 100644
--- a/kernel/fork.c
+++ b/kernel/fork.c
@@ -49,6 +49,7 @@
 #include <linux/delayacct.h>
 #include <linux/taskstats_kern.h>
 #include <linux/random.h>
+#include <linux/kevent.h>
 
 #include <asm/pgtable.h>
 #include <asm/pgalloc.h>
@@ -118,6 +119,9 @@ void __put_task_struct(struct task_struct *tsk)
 	WARN_ON(atomic_read(&tsk->usage));
 	WARN_ON(tsk == current);
 
+#ifdef CONFIG_KEVENT_SIGNAL
+	kevent_storage_fini(&tsk->st);
+#endif
 	security_task_free(tsk);
 	free_uid(tsk->user);
 	put_group_info(tsk->group_info);
@@ -1126,6 +1130,10 @@ static struct task_struct *copy_process(unsigned long clone_flags,
 	if (retval)
 		goto bad_fork_cleanup_namespaces;
 
+#ifdef CONFIG_KEVENT_SIGNAL
+	kevent_storage_init(p, &p->st);
+#endif
+
 	p->set_child_tid = (clone_flags & CLONE_CHILD_SETTID) ? child_tidptr : NULL;
 	/*
 	 * Clear TID on mm_release()?
diff --git a/kernel/kevent/kevent_signal.c b/kernel/kevent/kevent_signal.c
new file mode 100644
index 0000000..abe3972
--- /dev/null
+++ b/kernel/kevent/kevent_signal.c
@@ -0,0 +1,94 @@
+/*
+ * 	kevent_signal.c
+ * 
+ * 2006 Copyright (c) Evgeniy Polyakov <johnpol@2ka.mipt.ru>
+ * 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 <linux/kernel.h>
+#include <linux/types.h>
+#include <linux/slab.h>
+#include <linux/spinlock.h>
+#include <linux/file.h>
+#include <linux/fs.h>
+#include <linux/kevent.h>
+
+static int kevent_signal_callback(struct kevent *k)
+{
+	struct task_struct *tsk = k->st->origin;
+	int sig = k->event.id.raw[0];
+	int ret = 0;
+
+	if (sig == tsk->kevent_signals)
+		ret = 1;
+
+	if (ret && (k->event.id.raw_u64 & KEVENT_SIGNAL_NOMASK))
+		tsk->kevent_signals |= 0x80000000;
+
+	return ret;
+}
+
+int kevent_signal_enqueue(struct kevent *k)
+{
+	int err;
+
+	err = kevent_storage_enqueue(&current->st, k);
+	if (err)
+		goto err_out_exit;
+
+	if (k->event.req_flags & KEVENT_REQ_ALWAYS_QUEUE) {
+		kevent_requeue(k);
+		err = 0;
+	} else {
+		err = k->callbacks.callback(k);
+		if (err)
+			goto err_out_dequeue;
+	}
+
+	return err;
+
+err_out_dequeue:
+	kevent_storage_dequeue(k->st, k);
+err_out_exit:
+	return err;
+}
+
+int kevent_signal_dequeue(struct kevent *k)
+{
+	kevent_storage_dequeue(k->st, k);
+	return 0;
+}
+
+int kevent_signal_notify(struct task_struct *tsk, int sig)
+{
+	tsk->kevent_signals = sig;
+	kevent_storage_ready(&tsk->st, NULL, KEVENT_SIGNAL_DELIVERY);
+	return (tsk->kevent_signals & 0x80000000);
+}
+
+static int __init kevent_init_signal(void)
+{
+	struct kevent_callbacks sc = {
+		.callback = &kevent_signal_callback,
+		.enqueue = &kevent_signal_enqueue,
+		.dequeue = &kevent_signal_dequeue,
+		.flags = 0,
+	};
+
+	return kevent_add_callbacks(&sc, KEVENT_SIGNAL);
+}
+module_init(kevent_init_signal);
diff --git a/kernel/signal.c b/kernel/signal.c
index 5630255..f12ebc0 100644
--- a/kernel/signal.c
+++ b/kernel/signal.c
@@ -23,6 +23,7 @@
 #include <linux/ptrace.h>
 #include <linux/signal.h>
 #include <linux/capability.h>
+#include <linux/kevent.h>
 #include <linux/freezer.h>
 #include <linux/pid_namespace.h>
 #include <linux/nsproxy.h>
@@ -714,6 +715,9 @@ static int send_signal(int sig, struct siginfo *info, struct task_struct *t,
 {
 	struct sigqueue * q = NULL;
 	int ret = 0;
+	
+	if (kevent_signal_notify(t, sig))
+		return 1;
 
 	/*
 	 * fast-pathed signals for kernel-internal things like SIGSTOP
@@ -793,6 +797,17 @@ specific_send_sig_info(int sig, struct siginfo *info, struct task_struct *t)
 	ret = send_signal(sig, info, t, &t->pending);
 	if (!ret && !sigismember(&t->blocked, sig))
 		signal_wake_up(t, sig == SIGKILL);
+#ifdef CONFIG_KEVENT_SIGNAL
+	/*
+	 * Kevent allows to deliver signals through kevent queue, 
+	 * it is possible to setup kevent to not deliver
+	 * signal through the usual way, in that case send_signal()
+	 * returns 1 and signal is delivered only through kevent queue.
+	 * We simulate successfull delivery notification through this hack:
+	 */
+	if (ret == 1)
+		ret = 0;
+#endif
 out:
 	return ret;
 }
@@ -982,6 +997,17 @@ __group_send_sig_info(int sig, struct siginfo *info, struct task_struct *p)
 	 * to avoid several races.
 	 */
 	ret = send_signal(sig, info, p, &p->signal->shared_pending);
+#ifdef CONFIG_KEVENT_SIGNAL
+	/*
+	 * Kevent allows to deliver signals through kevent queue, 
+	 * it is possible to setup kevent to not deliver
+	 * signal through the usual way, in that case send_signal()
+	 * returns 1 and signal is delivered only through kevent queue.
+	 * We simulate successfull delivery notification through this hack:
+	 */
+	if (ret == 1)
+		ret = 0;
+#endif
 	if (unlikely(ret))
 		return ret;
 

[take35 9/10] kevent: Private userspace notifications.

[Evgeniy Polyakov]

Private userspace notifications.

Allows to register notifications of any private userspace
events over kevent. Events can be marked as ready using 
kevent_ctl(KEVENT_READY) command.

Signed-off-by: Evgeniy Polyakov <johnpol@2ka.mipt.ru>

diff --git a/kernel/kevent/kevent_unotify.c b/kernel/kevent/kevent_unotify.c
new file mode 100644
index 0000000..618c09c
--- /dev/null
+++ b/kernel/kevent/kevent_unotify.c
@@ -0,0 +1,62 @@
+/*
+ * 2006 Copyright (c) Evgeniy Polyakov <johnpol@2ka.mipt.ru>
+ * 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 <linux/kernel.h>
+#include <linux/kevent.h>
+
+static int kevent_unotify_callback(struct kevent *k)
+{
+	return 1;
+}
+
+int kevent_unotify_enqueue(struct kevent *k)
+{
+	int err;
+
+	err = kevent_storage_enqueue(&k->user->st, k);
+	if (err)
+		goto err_out_exit;
+
+	if (k->event.req_flags & KEVENT_REQ_ALWAYS_QUEUE)
+		kevent_requeue(k);
+
+	return 0;
+
+err_out_exit:
+	return err;
+}
+
+int kevent_unotify_dequeue(struct kevent *k)
+{
+	kevent_storage_dequeue(k->st, k);
+	return 0;
+}
+
+static int __init kevent_init_unotify(void)
+{
+	struct kevent_callbacks sc = {
+		.callback = &kevent_unotify_callback,
+		.enqueue = &kevent_unotify_enqueue,
+		.dequeue = &kevent_unotify_dequeue,
+		.flags = 0,
+	};
+
+	return kevent_add_callbacks(&sc, KEVENT_UNOTIFY);
+}
+module_init(kevent_init_unotify);

[take35 8/10] kevent: Kevent posix timer notifications.

[Evgeniy Polyakov]

Kevent posix timer notifications.

Simple extension to POSIX timers which allows
to deliver notification of the timer expiration
through kevent queue.

Example application posix_timer.c can be found
in archive on project homepage.

Signed-off-by: Evgeniy Polyakov <johnpol@2ka.mipt.ru>


diff --git a/include/asm-generic/siginfo.h b/include/asm-generic/siginfo.h
index 8786e01..3768746 100644
--- a/include/asm-generic/siginfo.h
+++ b/include/asm-generic/siginfo.h
@@ -235,6 +235,7 @@ typedef struct siginfo {
 #define SIGEV_NONE	1	/* other notification: meaningless */
 #define SIGEV_THREAD	2	/* deliver via thread creation */
 #define SIGEV_THREAD_ID 4	/* deliver to thread */
+#define SIGEV_KEVENT	8	/* deliver through kevent queue */
 
 /*
  * This works because the alignment is ok on all current architectures
@@ -260,6 +261,8 @@ typedef struct sigevent {
 			void (*_function)(sigval_t);
 			void *_attribute;	/* really pthread_attr_t */
 		} _sigev_thread;
+
+		int kevent_fd;
 	} _sigev_un;
 } sigevent_t;
 
diff --git a/include/linux/posix-timers.h b/include/linux/posix-timers.h
index a7dd38f..4b9deb4 100644
--- a/include/linux/posix-timers.h
+++ b/include/linux/posix-timers.h
@@ -4,6 +4,7 @@
 #include <linux/spinlock.h>
 #include <linux/list.h>
 #include <linux/sched.h>
+#include <linux/kevent_storage.h>
 
 union cpu_time_count {
 	cputime_t cpu;
@@ -49,6 +50,9 @@ struct k_itimer {
 	sigval_t it_sigev_value;	/* value word of sigevent struct */
 	struct task_struct *it_process;	/* process to send signal to */
 	struct sigqueue *sigq;		/* signal queue entry. */
+#ifdef CONFIG_KEVENT_TIMER
+	struct kevent_storage st;
+#endif
 	union {
 		struct {
 			struct hrtimer timer;
diff --git a/kernel/posix-timers.c b/kernel/posix-timers.c
index 5fe87de..5ec805e 100644
--- a/kernel/posix-timers.c
+++ b/kernel/posix-timers.c
@@ -48,6 +48,8 @@
 #include <linux/wait.h>
 #include <linux/workqueue.h>
 #include <linux/module.h>
+#include <linux/kevent.h>
+#include <linux/file.h>
 
 /*
  * Management arrays for POSIX timers.	 Timers are kept in slab memory
@@ -224,6 +226,100 @@ static int posix_ktime_get_ts(clockid_t which_clock, struct timespec *tp)
 	return 0;
 }
 
+#ifdef CONFIG_KEVENT_TIMER
+static int posix_kevent_enqueue(struct kevent *k)
+{
+	/*
+	 * It is not ugly - there is no pointer in the id field union, 
+	 * but its size is 64bits, which is ok for any known pointer size.
+	 */
+	struct k_itimer *tmr = (struct k_itimer *)(unsigned long)k->event.id.raw_u64;
+	return kevent_storage_enqueue(&tmr->st, k);
+}
+static int posix_kevent_dequeue(struct kevent *k)
+{
+	struct k_itimer *tmr = (struct k_itimer *)(unsigned long)k->event.id.raw_u64;
+	kevent_storage_dequeue(&tmr->st, k);
+	return 0;
+}
+static int posix_kevent_callback(struct kevent *k)
+{
+	return 1;
+}
+static int posix_kevent_init(void)
+{
+	struct kevent_callbacks tc = {
+		.callback = &posix_kevent_callback,
+		.enqueue = &posix_kevent_enqueue,
+		.dequeue = &posix_kevent_dequeue,
+		.flags = KEVENT_CALLBACKS_KERNELONLY};
+
+	return kevent_add_callbacks(&tc, KEVENT_POSIX_TIMER);
+}
+
+extern struct file_operations kevent_user_fops;
+
+static int posix_kevent_init_timer(struct k_itimer *tmr, int fd)
+{
+	struct ukevent uk;
+	struct file *file;
+	struct kevent_user *u;
+	int err;
+
+	file = fget(fd);
+	if (!file) {
+		err = -EBADF;
+		goto err_out;
+	}
+
+	if (file->f_op != &kevent_user_fops) {
+		err = -EINVAL;
+		goto err_out_fput;
+	}
+
+	u = file->private_data;
+
+	memset(&uk, 0, sizeof(struct ukevent));
+
+	uk.event = KEVENT_MASK_ALL;
+	uk.type = KEVENT_POSIX_TIMER;
+	uk.id.raw_u64 = (unsigned long)(tmr); /* Just cast to something unique */
+	uk.req_flags = KEVENT_REQ_ONESHOT | KEVENT_REQ_ALWAYS_QUEUE;
+	uk.ptr = tmr->it_sigev_value.sival_ptr;
+
+	err = kevent_user_add_ukevent(&uk, u);
+	if (err)
+		goto err_out_fput;
+
+	fput(file);
+
+	return 0;
+
+err_out_fput:
+	fput(file);
+err_out:
+	return err;
+}
+
+static void posix_kevent_fini_timer(struct k_itimer *tmr)
+{
+	kevent_storage_fini(&tmr->st);
+}
+#else
+static int posix_kevent_init_timer(struct k_itimer *tmr, int fd)
+{
+	return -ENOSYS;
+}
+static int posix_kevent_init(void)
+{
+	return 0;
+}
+static void posix_kevent_fini_timer(struct k_itimer *tmr)
+{
+}
+#endif
+
+
 /*
  * Initialize everything, well, just everything in Posix clocks/timers ;)
  */
@@ -241,6 +337,11 @@ static __init int init_posix_timers(void)
 	register_posix_clock(CLOCK_REALTIME, &clock_realtime);
 	register_posix_clock(CLOCK_MONOTONIC, &clock_monotonic);
 
+	if (posix_kevent_init()) {
+		printk(KERN_ERR "Failed to initialize kevent posix timers.\n");
+		BUG();
+	}
+
 	posix_timers_cache = kmem_cache_create("posix_timers_cache",
 					sizeof (struct k_itimer), 0, 0, NULL, NULL);
 	idr_init(&posix_timers_id);
@@ -343,23 +444,29 @@ static int posix_timer_fn(struct hrtimer *timer)
 
 	timr = container_of(timer, struct k_itimer, it.real.timer);
 	spin_lock_irqsave(&timr->it_lock, flags);
+	
+	if (timr->it_sigev_notify == SIGEV_KEVENT) {
+#ifdef CONFIG_KEVENT_TIMER
+		kevent_storage_ready(&timr->st, NULL, KEVENT_MASK_ALL);
+#endif
+	} else {
+		if (timr->it.real.interval.tv64 != 0)
+			si_private = ++timr->it_requeue_pending;
 
-	if (timr->it.real.interval.tv64 != 0)
-		si_private = ++timr->it_requeue_pending;
-
-	if (posix_timer_event(timr, si_private)) {
-		/*
-		 * signal was not sent because of sig_ignor
-		 * we will not get a call back to restart it AND
-		 * it should be restarted.
-		 */
-		if (timr->it.real.interval.tv64 != 0) {
-			timr->it_overrun +=
-				hrtimer_forward(timer,
-						timer->base->softirq_time,
-						timr->it.real.interval);
-			ret = HRTIMER_RESTART;
-			++timr->it_requeue_pending;
+		if (posix_timer_event(timr, si_private)) {
+			/*
+			 * signal was not sent because of sig_ignor
+			 * we will not get a call back to restart it AND
+			 * it should be restarted.
+			 */
+			if (timr->it.real.interval.tv64 != 0) {
+				timr->it_overrun +=
+					hrtimer_forward(timer,
+							timer->base->softirq_time,
+							timr->it.real.interval);
+				ret = HRTIMER_RESTART;
+				++timr->it_requeue_pending;
+			}
 		}
 	}
 
@@ -407,6 +514,9 @@ static struct k_itimer * alloc_posix_timer(void)
 		kmem_cache_free(posix_timers_cache, tmr);
 		tmr = NULL;
 	}
+#ifdef CONFIG_KEVENT_TIMER
+	kevent_storage_init(tmr, &tmr->st);
+#endif
 	return tmr;
 }
 
@@ -424,6 +534,7 @@ static void release_posix_timer(struct k_itimer *tmr, int it_id_set)
 	if (unlikely(tmr->it_process) &&
 	    tmr->it_sigev_notify == (SIGEV_SIGNAL|SIGEV_THREAD_ID))
 		put_task_struct(tmr->it_process);
+	posix_kevent_fini_timer(tmr);
 	kmem_cache_free(posix_timers_cache, tmr);
 }
 
@@ -496,40 +607,52 @@ sys_timer_create(const clockid_t which_clock,
 		new_timer->it_sigev_signo = event.sigev_signo;
 		new_timer->it_sigev_value = event.sigev_value;
 
-		read_lock(&tasklist_lock);
-		if ((process = good_sigevent(&event))) {
-			/*
-			 * We may be setting up this process for another
-			 * thread.  It may be exiting.  To catch this
-			 * case the we check the PF_EXITING flag.  If
-			 * the flag is not set, the siglock will catch
-			 * him before it is too late (in exit_itimers).
-			 *
-			 * The exec case is a bit more invloved but easy
-			 * to code.  If the process is in our thread
-			 * group (and it must be or we would not allow
-			 * it here) and is doing an exec, it will cause
-			 * us to be killed.  In this case it will wait
-			 * for us to die which means we can finish this
-			 * linkage with our last gasp. I.e. no code :)
-			 */
+		if (event.sigev_notify == SIGEV_KEVENT) {
+			error = posix_kevent_init_timer(new_timer, event._sigev_un.kevent_fd);
+			if (error)
+				goto out;
+
+			process = current->group_leader;
 			spin_lock_irqsave(&process->sighand->siglock, flags);
-			if (!(process->flags & PF_EXITING)) {
-				new_timer->it_process = process;
-				list_add(&new_timer->list,
-					 &process->signal->posix_timers);
-				spin_unlock_irqrestore(&process->sighand->siglock, flags);
-				if (new_timer->it_sigev_notify == (SIGEV_SIGNAL|SIGEV_THREAD_ID))
-					get_task_struct(process);
-			} else {
-				spin_unlock_irqrestore(&process->sighand->siglock, flags);
-				process = NULL;
+			new_timer->it_process = process;
+			list_add(&new_timer->list, &process->signal->posix_timers);
+			spin_unlock_irqrestore(&process->sighand->siglock, flags);
+		} else {
+			read_lock(&tasklist_lock);
+			if ((process = good_sigevent(&event))) {
+				/*
+				 * We may be setting up this process for another
+				 * thread.  It may be exiting.  To catch this
+				 * case the we check the PF_EXITING flag.  If
+				 * the flag is not set, the siglock will catch
+				 * him before it is too late (in exit_itimers).
+				 *
+				 * The exec case is a bit more invloved but easy
+				 * to code.  If the process is in our thread
+				 * group (and it must be or we would not allow
+				 * it here) and is doing an exec, it will cause
+				 * us to be killed.  In this case it will wait
+				 * for us to die which means we can finish this
+				 * linkage with our last gasp. I.e. no code :)
+				 */
+				spin_lock_irqsave(&process->sighand->siglock, flags);
+				if (!(process->flags & PF_EXITING)) {
+					new_timer->it_process = process;
+					list_add(&new_timer->list,
+						 &process->signal->posix_timers);
+					spin_unlock_irqrestore(&process->sighand->siglock, flags);
+					if (new_timer->it_sigev_notify == (SIGEV_SIGNAL|SIGEV_THREAD_ID))
+						get_task_struct(process);
+				} else {
+					spin_unlock_irqrestore(&process->sighand->siglock, flags);
+					process = NULL;
+				}
+			}
+			read_unlock(&tasklist_lock);
+			if (!process) {
+				error = -EINVAL;
+				goto out;
 			}
-		}
-		read_unlock(&tasklist_lock);
-		if (!process) {
-			error = -EINVAL;
-			goto out;
 		}
 	} else {
 		new_timer->it_sigev_notify = SIGEV_SIGNAL;

[take35 10/10] kevent: Kevent based AIO (aio_sendfile()/aio_sendfile_path()).

[Evgeniy Polyakov]

Kevent based AIO (aio_sendfile()/aio_sendfile_path()).

aio_sendfile()/aio_sendfile_path() contains of two major parts: AIO 
state machine and page processing code. 
The former is just a small subsystem, which allows to queue callback 
for theirs invocation in process' context on behalf of pool of kernel 
threads. It allows to queue caches of callbacks to the local thread 
or to any other specified. Each cache of callbacks is processed until 
there are callbacks in it, callbacks can requeue themselfs into the 
same cache.

Real work is being done in page processing code - code which populates 
pages into VFS cache and then sends pages to the destination socket 
via ->sendpage(). Unlike previous aio_sendfile() implementation, new 
one does not require low-level filesystem specific callbacks (->get_block())
at all, instead I extended struct address_space_operations to contain new 
member called ->aio_readpages(), which is exactly the same as ->readpage() 
(read: mpage_readpages()) except different BIO allocation and sumbission 
routines. I changed mpage_readpages() to provide mpage_alloc() and 
mpage_bio_submit() to the new function called __mpage_readpages(), which is 
exactly old mpage_readpages() with provided callback invocation instead of 
usage for old functions. mpage_readpages_aio() provides kevent specific 
callbacks, which calls old functions, but with different destructor callbacks,
which are essentially the same, except that they reschedule AIO processing.

aio_sendfile_path() is essentially aio_sendfile(), except that it takes
source filename as parameter, has a pointer to private header
and its size (which allows to send header and file's content in one syscall
instead of three (open, send, sendfile) and returns opened file descriptor.

Benchmark of the 100 1MB files transfer (files are in VFS already) using sync 
sendfile() against aio_sendfile_path() shows about 10MB/sec performance win 
(78 MB/s vs 66-72 MB/s over 1 Gb network, sendfile sending server is one-way 
AMD Athlong 64 3500+) for aio_sendfile_path().

AIO state machine is a base for network AIO (which becomes
quite trivial), but I will not start implementation until
roadback of kevent as a whole and AIO implementation become more clear.

Signed-off-by: Evgeniy Polyakov <johnpol@2ka.mipt.ru>

diff --git a/fs/bio.c b/fs/bio.c
index 7618bcb..291e7e8 100644
--- a/fs/bio.c
+++ b/fs/bio.c
@@ -120,7 +120,7 @@ void bio_free(struct bio *bio, struct bio_set *bio_set)
 /*
  * default destructor for a bio allocated with bio_alloc_bioset()
  */
-static void bio_fs_destructor(struct bio *bio)
+void bio_fs_destructor(struct bio *bio)
 {
 	bio_free(bio, fs_bio_set);
 }
diff --git a/fs/ext3/inode.c b/fs/ext3/inode.c
index beaf25f..f08c957 100644
--- a/fs/ext3/inode.c
+++ b/fs/ext3/inode.c
@@ -1650,6 +1650,13 @@ ext3_readpages(struct file *file, struct address_space *mapping,
 	return mpage_readpages(mapping, pages, nr_pages, ext3_get_block);
 }
 
+static int
+ext3_readpages_aio(struct file *file, struct address_space *mapping,
+		struct list_head *pages, unsigned nr_pages, void *priv)
+{
+	return mpage_readpages_aio(mapping, pages, nr_pages, ext3_get_block, priv);
+}
+
 static void ext3_invalidatepage(struct page *page, unsigned long offset)
 {
 	journal_t *journal = EXT3_JOURNAL(page->mapping->host);
@@ -1768,6 +1775,7 @@ static int ext3_journalled_set_page_dirty(struct page *page)
 }
 
 static const struct address_space_operations ext3_ordered_aops = {
+	.aio_readpages	= ext3_readpages_aio,
 	.readpage	= ext3_readpage,
 	.readpages	= ext3_readpages,
 	.writepage	= ext3_ordered_writepage,
diff --git a/fs/mpage.c b/fs/mpage.c
index 692a3e5..e5ba44b 100644
--- a/fs/mpage.c
+++ b/fs/mpage.c
@@ -102,7 +102,7 @@ static struct bio *mpage_bio_submit(int rw, struct bio *bio)
 static struct bio *
 mpage_alloc(struct block_device *bdev,
 		sector_t first_sector, int nr_vecs,
-		gfp_t gfp_flags)
+		gfp_t gfp_flags, void *priv)
 {
 	struct bio *bio;
 
@@ -116,6 +116,7 @@ mpage_alloc(struct block_device *bdev,
 	if (bio) {
 		bio->bi_bdev = bdev;
 		bio->bi_sector = first_sector;
+		bio->bi_private = priv;
 	}
 	return bio;
 }
@@ -175,7 +176,10 @@ map_buffer_to_page(struct page *page, struct buffer_head *bh, int page_block)
 static struct bio *
 do_mpage_readpage(struct bio *bio, struct page *page, unsigned nr_pages,
 		sector_t *last_block_in_bio, struct buffer_head *map_bh,
-		unsigned long *first_logical_block, get_block_t get_block)
+		unsigned long *first_logical_block, get_block_t get_block,
+		struct bio *(*alloc)(struct block_device *bdev, sector_t first_sector, 
+			int nr_vecs, gfp_t gfp_flags, void *priv),
+		struct bio *(*submit)(int rw, struct bio *bio), void *priv)
 {
 	struct inode *inode = page->mapping->host;
 	const unsigned blkbits = inode->i_blkbits;
@@ -302,25 +306,25 @@ do_mpage_readpage(struct bio *bio, struct page *page, unsigned nr_pages,
 	 * This page will go to BIO.  Do we need to send this BIO off first?
 	 */
 	if (bio && (*last_block_in_bio != blocks[0] - 1))
-		bio = mpage_bio_submit(READ, bio);
+		bio = submit(READ, bio);
 
 alloc_new:
 	if (bio == NULL) {
-		bio = mpage_alloc(bdev, blocks[0] << (blkbits - 9),
+		bio = alloc(bdev, blocks[0] << (blkbits - 9),
 			  	min_t(int, nr_pages, bio_get_nr_vecs(bdev)),
-				GFP_KERNEL);
+				GFP_KERNEL, priv);
 		if (bio == NULL)
 			goto confused;
 	}
 
 	length = first_hole << blkbits;
 	if (bio_add_page(bio, page, length, 0) < length) {
-		bio = mpage_bio_submit(READ, bio);
+		bio = submit(READ, bio);
 		goto alloc_new;
 	}
 
 	if (buffer_boundary(map_bh) || (first_hole != blocks_per_page))
-		bio = mpage_bio_submit(READ, bio);
+		bio = submit(READ, bio);
 	else
 		*last_block_in_bio = blocks[blocks_per_page - 1];
 out:
@@ -328,7 +332,7 @@ out:
 
 confused:
 	if (bio)
-		bio = mpage_bio_submit(READ, bio);
+		bio = submit(READ, bio);
 	if (!PageUptodate(page))
 	        block_read_full_page(page, get_block);
 	else
@@ -336,6 +340,48 @@ confused:
 	goto out;
 }
 
+int
+__mpage_readpages(struct address_space *mapping, struct list_head *pages,
+				unsigned nr_pages, get_block_t get_block,
+		struct bio *(*alloc)(struct block_device *bdev, sector_t first_sector, 
+			int nr_vecs, gfp_t gfp_flags, void *priv),
+		struct bio *(*submit)(int rw, struct bio *bio), 
+		void *priv)
+{
+	struct bio *bio = NULL;
+	unsigned page_idx;
+	sector_t last_block_in_bio = 0;
+	struct pagevec lru_pvec;
+	struct buffer_head map_bh;
+	unsigned long first_logical_block = 0;
+
+	clear_buffer_mapped(&map_bh);
+	pagevec_init(&lru_pvec, 0);
+	for (page_idx = 0; page_idx < nr_pages; page_idx++) {
+		struct page *page = list_entry(pages->prev, struct page, lru);
+
+		prefetchw(&page->flags);
+		list_del(&page->lru);
+		if (!add_to_page_cache(page, mapping,
+					page->index, GFP_KERNEL)) {
+			bio = do_mpage_readpage(bio, page,
+					nr_pages - page_idx,
+					&last_block_in_bio, &map_bh,
+					&first_logical_block,
+					get_block, alloc, submit, priv);
+			if (!pagevec_add(&lru_pvec, page))
+				__pagevec_lru_add(&lru_pvec);
+		} else {
+			page_cache_release(page);
+		}
+	}
+	pagevec_lru_add(&lru_pvec);
+	BUG_ON(!list_empty(pages));
+	if (bio)
+		submit(READ, bio);
+	return 0;
+}
+
 /**
  * mpage_readpages - populate an address space with some pages, and
  *                       start reads against them.
@@ -386,40 +432,28 @@ int
 mpage_readpages(struct address_space *mapping, struct list_head *pages,
 				unsigned nr_pages, get_block_t get_block)
 {
-	struct bio *bio = NULL;
-	unsigned page_idx;
-	sector_t last_block_in_bio = 0;
-	struct pagevec lru_pvec;
-	struct buffer_head map_bh;
-	unsigned long first_logical_block = 0;
+	return __mpage_readpages(mapping, pages, nr_pages, get_block,
+			mpage_alloc, mpage_bio_submit, NULL);
+}
+EXPORT_SYMBOL(mpage_readpages);
 
-	clear_buffer_mapped(&map_bh);
-	pagevec_init(&lru_pvec, 0);
-	for (page_idx = 0; page_idx < nr_pages; page_idx++) {
-		struct page *page = list_entry(pages->prev, struct page, lru);
+#ifdef CONFIG_KEVENT_AIO
+extern struct bio *kaio_mpage_alloc(struct block_device *bdev, sector_t first_sector, 
+		int nr_vecs, gfp_t gfp_flags, void *priv);
+extern struct bio *kaio_mpage_bio_submit(int rw, struct bio *bio);
+#else
+#define kaio_mpage_alloc	mpage_alloc
+#define kaio_mpage_bio_submit	mpage_bio_submit
+#endif
 
-		prefetchw(&page->flags);
-		list_del(&page->lru);
-		if (!add_to_page_cache(page, mapping,
-					page->index, GFP_KERNEL)) {
-			bio = do_mpage_readpage(bio, page,
-					nr_pages - page_idx,
-					&last_block_in_bio, &map_bh,
-					&first_logical_block,
-					get_block);
-			if (!pagevec_add(&lru_pvec, page))
-				__pagevec_lru_add(&lru_pvec);
-		} else {
-			page_cache_release(page);
-		}
-	}
-	pagevec_lru_add(&lru_pvec);
-	BUG_ON(!list_empty(pages));
-	if (bio)
-		mpage_bio_submit(READ, bio);
-	return 0;
+int
+mpage_readpages_aio(struct address_space *mapping, struct list_head *pages,
+				unsigned nr_pages, get_block_t get_block, void *priv)
+{
+	return __mpage_readpages(mapping, pages, nr_pages, get_block,
+			kaio_mpage_alloc, kaio_mpage_bio_submit, priv);
 }
-EXPORT_SYMBOL(mpage_readpages);
+EXPORT_SYMBOL(mpage_readpages_aio);
 
 /*
  * This isn't called much at all
@@ -433,7 +467,8 @@ int mpage_readpage(struct page *page, get_block_t get_block)
 
 	clear_buffer_mapped(&map_bh);
 	bio = do_mpage_readpage(bio, page, 1, &last_block_in_bio,
-			&map_bh, &first_logical_block, get_block);
+			&map_bh, &first_logical_block, get_block,
+			mpage_alloc, mpage_bio_submit, NULL);
 	if (bio)
 		mpage_bio_submit(READ, bio);
 	return 0;
@@ -595,7 +630,7 @@ page_is_mapped:
 alloc_new:
 	if (bio == NULL) {
 		bio = mpage_alloc(bdev, blocks[0] << (blkbits - 9),
-				bio_get_nr_vecs(bdev), GFP_NOFS|__GFP_HIGH);
+				bio_get_nr_vecs(bdev), GFP_NOFS|__GFP_HIGH, NULL);
 		if (bio == NULL)
 			goto confused;
 	}
diff --git a/include/linux/mpage.h b/include/linux/mpage.h
index cc5fb75..accdbdd 100644
--- a/include/linux/mpage.h
+++ b/include/linux/mpage.h
@@ -16,6 +16,8 @@ typedef int (writepage_t)(struct page *page, struct writeback_control *wbc);
 
 int mpage_readpages(struct address_space *mapping, struct list_head *pages,
 				unsigned nr_pages, get_block_t get_block);
+int mpage_readpages_aio(struct address_space *mapping, struct list_head *pages,
+				unsigned nr_pages, get_block_t get_block, void *priv);
 int mpage_readpage(struct page *page, get_block_t get_block);
 int mpage_writepages(struct address_space *mapping,
 		struct writeback_control *wbc, get_block_t get_block);
diff --git a/kernel/kevent/kevent_aio.c b/kernel/kevent/kevent_aio.c
new file mode 100644
index 0000000..9bb5e99
--- /dev/null
+++ b/kernel/kevent/kevent_aio.c
@@ -0,0 +1,939 @@
+/*
+ * 2006 Copyright (c) Evgeniy Polyakov <johnpol@2ka.mipt.ru>
+ * 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 <linux/kernel.h>
+#include <linux/module.h>
+#include <linux/types.h>
+#include <linux/spinlock.h>
+#include <linux/list.h>
+#include <linux/kthread.h>
+#include <linux/socket.h>
+#include <linux/slab.h>
+#include <linux/bio.h>
+#include <linux/pagemap.h>
+#include <linux/file.h>
+#include <linux/swap.h>
+#include <linux/kevent.h>
+
+#define KAIO_CALL_NUM		8
+#define KAIO_THREAD_NUM		8
+
+//#define KEVENT_AIO_DEBUG
+
+#ifdef KEVENT_AIO_DEBUG
+#define dprintk(f, a...) printk(f, ##a)
+#else
+#define dprintk(f, a...) do {} while (0)
+#endif
+
+struct kaio_thread
+{
+	struct list_head	req_list;
+	spinlock_t		req_lock;
+	struct task_struct	*thread;
+	int			refcnt;
+	wait_queue_head_t 	wait;
+};
+
+extern struct file_operations kevent_user_fops;
+static DEFINE_PER_CPU(u32, kaio_req_counter);
+static DEFINE_PER_CPU(int, kaio_req_cpu);
+
+/*
+ * Array of working threads.
+ * It can only be accessed under RCU protection, 
+ * so threads reference counters are not atomic.
+ */
+static struct kaio_thread *kaio_threads[KAIO_THREAD_NUM] __read_mostly;
+static struct kmem_cache *kaio_req_cache __read_mostly;
+static struct kmem_cache *kaio_priv_cache __read_mostly;
+
+struct kaio_req;
+typedef int (* kaio_callback)(struct kaio_req *req, int direct);
+
+#define KAIO_REQ_PENDING	0
+
+/*
+ * Cache of kaio request callbacks.
+ * It is not allowed to change the same cache entry
+ * simultaneously (for example it is forbidden to add entries
+ * in parallel).
+ *
+ * When cache entry is scheduled for execution in one of the threads,
+ * it is forbidden to access it, since it will be freed when
+ * all callbacks have been invoked.
+ *
+ * It is possible to add callbacks into this cache from callbacks itself.
+ */
+struct kaio_req
+{
+	struct list_head	req_entry;
+	kaio_callback		call[KAIO_CALL_NUM];
+	int			read_idx, add_idx;
+	int			cpu;
+	long			flags;
+	atomic_t		refcnt;
+	void			(*destructor)(struct kaio_req *);
+	void			*priv;
+};
+
+/*
+ * Returns pointer to thread entry for given index.
+ * Must be called under RCU protection.
+ */
+static struct kaio_thread *kaio_get_thread(int cpu)
+{
+	struct kaio_thread *th;
+
+	if (cpu == -1) {
+#if 1
+		int *cnt = &__get_cpu_var(kaio_req_cpu);
+		cpu = *cnt;
+
+		*cnt = *cnt + 1;
+		if (*cnt >= KAIO_THREAD_NUM)
+			*cnt = 0;
+#else
+		cpu = 0;
+#endif
+	}
+
+	if (unlikely(cpu >= KAIO_THREAD_NUM || !kaio_threads[cpu]))
+		return NULL;
+
+	th = kaio_threads[cpu];
+	th->refcnt++;
+
+	return th;
+}
+
+/*
+ * Drops reference counter for given thread.
+ * Must be called under RCU protection.
+ */
+static inline void kaio_put_thread(struct kaio_thread *th)
+{
+	th->refcnt--;
+}
+
+void kaio_schedule_req(struct kaio_req *req)
+{
+	struct kaio_thread *th;
+	unsigned long flags;
+
+	rcu_read_lock();
+	th = kaio_get_thread(req->cpu);
+	if (!th) {
+		req->cpu = -1;
+		th = kaio_get_thread(-1);
+		BUG_ON(!th);
+	}
+
+	if (!test_and_set_bit(KAIO_REQ_PENDING, &req->flags)) {
+		spin_lock_irqsave(&th->req_lock, flags);
+		list_add_tail(&req->req_entry, &th->req_list);
+		spin_unlock_irqrestore(&th->req_lock, flags);
+	}
+
+	wake_up(&th->wait);
+
+	kaio_put_thread(th);
+	rcu_read_unlock();
+}
+
+EXPORT_SYMBOL_GPL(kaio_schedule_req);
+
+static inline void kaio_req_get(struct kaio_req *req)
+{
+	atomic_inc(&req->refcnt);
+}
+
+static inline int kaio_req_put(struct kaio_req *req)
+{
+	if (atomic_dec_and_test(&req->refcnt)) {
+		dprintk("%s: freeing req: %p, priv: %p.\n", __func__, req, req->priv);
+		if (req->destructor)
+			req->destructor(req);
+		kmem_cache_free(kaio_req_cache, req);
+		return 1;
+	}
+
+	return 0;
+}
+
+/*
+ * Append a call request into cache.
+ * Returns -EOVERFLOW in case cache is full, and 0 otherwise.
+ */
+int kaio_append_call(struct kaio_req *req, kaio_callback call)
+{
+	if ((req->add_idx + 1 == req->read_idx) ||
+			((req->add_idx + 1 == KAIO_CALL_NUM) && req->read_idx == 0))
+		return -EOVERFLOW;
+
+	req->call[req->add_idx] = call;
+
+	dprintk("%s: req: %p, read_idx: %d, add_idx: %d, call: %p [%p].\n",
+			__func__, req, req->read_idx, req->add_idx, 
+			req->call[req->read_idx], req->call[req->add_idx]);
+	if (++req->add_idx == KAIO_CALL_NUM)
+		req->add_idx = 0;
+
+	kaio_req_get(req);
+	
+	return 0;
+}
+
+EXPORT_SYMBOL_GPL(kaio_append_call);
+
+/*
+ * Adds one call request into given cache.
+ * If cache is NULL or full, allocate new one.
+ */
+struct kaio_req *kaio_add_call(struct kaio_req *req, kaio_callback call, int cpu, gfp_t gflags)
+{
+	if (req && !kaio_append_call(req, call)) {
+		kaio_schedule_req(req);
+		return req;
+	}
+
+	req = kmem_cache_alloc(kaio_req_cache, gflags);
+	if (!req)
+		return NULL;
+
+	memset(req->call, 0, sizeof(req->call));
+
+	req->destructor = NULL;
+	req->cpu = cpu;
+	req->call[0] = call;
+	req->add_idx = 1;
+	req->read_idx = 0;
+	req->flags = 0;
+	atomic_set(&req->refcnt, 1);
+
+	dprintk("%s: req: %p, call: %p [%p].\n", __func__, req, call, req->call[0]);
+
+	return req;
+}
+
+EXPORT_SYMBOL_GPL(kaio_add_call);
+
+/*
+ * Call appropriate callbacks in cache.
+ * This can only be called by working threads, which means that cache
+ * is filled (probably partially) and are not even accessible from
+ * the originator of requests, which means that cache will be freed
+ * when all callbacks are invoked.
+ *
+ * Callback itself can reschedule new callback into the same cache.
+ *
+ * If callback returns negative value, the whole cache will be freed.
+ * If positive value is returned, then further processing is stopped, 
+ * so cache can be queued into the end of the processing FIFO by callback.
+ * If zero is returned, next callback will be invoked if any.
+ */
+static int kaio_call(struct kaio_req *req)
+{
+	int err = -EINVAL;
+
+	if (likely(req->add_idx != req->read_idx)) {
+		dprintk("%s: req: %p, read_idx: %d, add_idx: %d, call: %p [%p].\n",
+				__func__, req, req->read_idx, req->add_idx, 
+				req->call[req->read_idx], req->call[0]);
+		err = (*req->call[req->read_idx])(req, 0);
+		if (++req->read_idx == KAIO_CALL_NUM)
+			req->read_idx = 0;
+
+		if (kaio_req_put(req))
+			err = 0;
+	}
+	return err;
+}
+
+static int kaio_thread_process(void *data)
+{
+	struct kaio_thread *th = data;
+	unsigned long flags;
+	struct kaio_req *req, *first;
+	DECLARE_WAITQUEUE(wait, current);
+	int err;
+
+	add_wait_queue_exclusive(&th->wait, &wait);
+
+	while (!kthread_should_stop()) {
+		first = req = NULL;
+		do {
+			req = NULL;
+			spin_lock_irqsave(&th->req_lock, flags);
+			if (!list_empty(&th->req_list)) {
+				req = list_entry(th->req_list.prev, struct kaio_req, req_entry);
+				if (first != req)
+					list_del(&req->req_entry);
+			}
+			spin_unlock_irqrestore(&th->req_lock, flags);
+				
+			if (!first)
+				first = req;
+			else if (first == req)
+				break;
+
+			if (req) {
+				err = 0;
+				while ((req->read_idx != req->add_idx) && !kthread_should_stop()) {
+					dprintk("%s: req: %p, read_idx: %d, add_idx: %d, err: %d.\n",
+							__func__, req, req->read_idx, req->add_idx, err);
+					err = kaio_call(req);
+					if (err != 0)
+						break;
+				}
+
+				if (err > 0) {
+					spin_lock_irqsave(&th->req_lock, flags);
+					list_add_tail(&req->req_entry, &th->req_list);
+					spin_unlock_irqrestore(&th->req_lock, flags);
+				}
+			}
+		} while (req);
+		__set_current_state(TASK_INTERRUPTIBLE);
+		schedule_timeout(HZ);
+		__set_current_state(TASK_RUNNING);
+	}
+	
+	remove_wait_queue(&th->wait, &wait);
+
+	return 0;
+}
+
+struct kaio_private
+{
+	union {
+		void 		*sptr;
+		__u64		sdata;
+	};
+	union {
+		void 		*dptr;
+		__u64		ddata;
+	};
+	void			__user *header;
+	size_t			header_size;
+	__u64			offset, processed;
+	__u64			count, limit;
+	struct kevent_user	*kevent_user;
+};
+
+extern void bio_fs_destructor(struct bio *bio);
+
+static void kaio_bio_destructor(struct bio *bio)
+{
+	dprintk("%s: bio=%p, num=%u.\n", __func__, bio, bio->bi_vcnt);
+	bio_fs_destructor(bio);
+}
+
+static int kaio_read_send_pages(struct kaio_req *req, int direct);
+
+static int kaio_mpage_end_io_read(struct bio *bio, unsigned int bytes_done, int err)
+{
+	const int uptodate = test_bit(BIO_UPTODATE, &bio->bi_flags);
+	struct bio_vec *bvec = bio->bi_io_vec + bio->bi_vcnt - 1;
+	struct kaio_req *req = bio->bi_private;
+
+	if (bio->bi_size)
+		return 1;
+
+	do {
+		struct page *page = bvec->bv_page;
+
+		if (--bvec >= bio->bi_io_vec)
+			prefetchw(&bvec->bv_page->flags);
+
+		if (uptodate) {
+			SetPageUptodate(page);
+		} else {
+			ClearPageUptodate(page);
+			SetPageError(page);
+		}
+		unlock_page(page);
+	} while (bvec >= bio->bi_io_vec);
+
+	dprintk("%s: bio: %p, req: %p, pending: %d.\n", 
+			__func__, bio, req, test_bit(KAIO_REQ_PENDING, &req->flags));
+
+	kaio_append_call(req, kaio_read_send_pages);
+	kaio_req_put(req);
+	kaio_schedule_req(req);
+
+	bio_put(bio);
+	return 0;
+}
+
+struct bio *kaio_mpage_bio_submit(int rw, struct bio *bio)
+{
+	if (bio) {
+		bio->bi_end_io = kaio_mpage_end_io_read;
+		dprintk("%s: bio=%p, num=%u.\n", __func__, bio, bio->bi_vcnt);
+		submit_bio(READ, bio);
+	}
+	return NULL;
+}
+
+struct bio *kaio_mpage_alloc(struct block_device *bdev,
+		sector_t first_sector, int nr_vecs, gfp_t gfp_flags, void *priv)
+{
+	struct bio *bio;
+
+	bio = bio_alloc(gfp_flags, nr_vecs);
+
+	if (bio == NULL && (current->flags & PF_MEMALLOC)) {
+		while (!bio && (nr_vecs /= 2))
+			bio = bio_alloc(gfp_flags, nr_vecs);
+	}
+
+	if (bio) {
+		struct kaio_req *req = priv;
+
+		bio->bi_bdev = bdev;
+		bio->bi_sector = first_sector;
+		bio->bi_private = priv;
+		bio->bi_destructor = kaio_bio_destructor;
+		kaio_req_get(req);
+		dprintk("%s: bio: %p, req: %p, num: %d.\n", __func__, bio, priv, nr_vecs);
+	}
+	return bio;
+}
+
+static ssize_t kaio_vfs_read_actor(struct kaio_private *priv, struct page *page, size_t len)
+{
+	struct socket *sock = priv->dptr;
+	struct file *file = sock->file;
+	
+	return file->f_op->sendpage(file, page, 0, len, &file->f_pos, 1);
+}
+
+static int kaio_vfs_read(struct kaio_private *priv,
+		ssize_t (*actor)(struct kaio_private *, struct page *, size_t))
+{
+	struct address_space *mapping;
+	struct file *file = priv->sptr;
+	ssize_t actor_size;
+	loff_t isize;
+	int i = 0, pg_num;
+
+	mapping = file->f_mapping;
+	isize = i_size_read(file->f_dentry->d_inode);
+
+	if (priv->processed >= isize) {
+		priv->count = 0;
+		return 0;
+	}
+	priv->count = isize - priv->processed;
+	pg_num = ALIGN(min_t(u64, isize, priv->count), PAGE_SIZE) >> PAGE_SHIFT;
+	
+	dprintk("%s: start: priv: %p, ret: %d, num: %d, count: %Lu, offset: %Lu, processed: %Lu.\n", 
+			__func__, priv, i, pg_num, priv->count, priv->offset, priv->processed);
+
+	for (i=0; i<pg_num && priv->count; ++i) {
+		struct page *page;
+		size_t nr = PAGE_CACHE_SIZE;
+
+		page = find_get_page(mapping, priv->processed >> PAGE_CACHE_SHIFT);
+		if (unlikely(page == NULL))
+			break;
+		if (!PageUptodate(page)) {
+			dprintk("%s: %2d: page=%p, processed=%Lu, count=%Lu not uptodate.\n", 
+					__func__, i, page, priv->processed, priv->count);
+			page_cache_release(page);
+			break;
+		}
+
+		if (mapping_writably_mapped(mapping))
+			flush_dcache_page(page);
+
+		mark_page_accessed(page);
+
+		if (nr + priv->processed > isize)
+			nr = isize - priv->processed;
+		if (nr > priv->count)
+			nr = priv->count;
+
+		actor_size = actor(priv, page, nr);
+		if (actor_size < 0) {
+			page_cache_release(page);
+			i = (int)actor_size;
+			break;
+		}
+
+		page_cache_release(page);
+
+		priv->processed += actor_size;
+		priv->count -= actor_size;
+	}
+
+	if (!priv->count)
+		i = pg_num;
+
+	dprintk("%s: end: priv: %p, ret: %d, num: %d, count: %Lu, offset: %Lu, processed: %Lu.\n", 
+			__func__, priv, i, pg_num, priv->count, priv->offset, priv->processed);
+
+	return i;
+}
+
+/*
+ * SThis function sends a header if it is available.
+ * Zero is returned if it was sent correctly,
+ * in case of fatal error negative value is returned,
+ * and positive one, if -EAGAIN was returned by underlying socket layer.
+ */
+static int kaio_send_header(struct kaio_private *priv)
+{
+	int err;
+	struct socket *sock = priv->dptr;
+	struct iovec iov;
+	struct msghdr msg;
+
+	memset(&msg, 0, sizeof(struct msghdr));
+	iov.iov_base = priv->header;
+	iov.iov_len = priv->header_size;
+	msg.msg_iov = &iov;
+	msg.msg_iovlen = 1;
+
+	while (priv->header_size) {
+		iov.iov_base = priv->header;
+		iov.iov_len = priv->header_size;
+
+		err = sock_sendmsg(sock, &msg, priv->header_size);
+
+		if (err < 0)
+			return err;
+		if (err == 0)
+			return -1;
+
+		priv->header += err;
+		priv->header_size -= err;
+	}
+
+	return 0;
+}
+
+static int kaio_read_send_pages(struct kaio_req *req, int direct)
+{
+	struct kaio_private *priv = req->priv;
+	struct file *file = priv->sptr;
+	struct address_space *mapping = file->f_mapping;
+	struct page *page;
+	int err = 0, i, num;
+	u64 offset;
+	LIST_HEAD(page_pool);
+
+	if (priv->header_size) {
+		err = kaio_send_header(priv);
+		if (err < 0)
+			return err;
+	}
+
+	if (err > 0) {
+		kaio_append_call(req, kaio_read_send_pages);
+		return 0;
+	}
+
+	err = kaio_vfs_read(priv, &kaio_vfs_read_actor);
+	if (err < 0)
+		return err;
+
+	if (err == 0) {
+		priv->limit >>= 1;
+	} else {
+		if (priv->limit)
+			priv->limit <<= 1;
+		else
+			priv->limit = 8;
+	}
+
+	if (priv->offset < priv->processed)
+		priv->offset = priv->processed;
+
+	if (!priv->count) {
+		kevent_storage_ready(&priv->kevent_user->st, NULL, KEVENT_MASK_ALL);
+		return 0;
+	}
+
+	if (priv->offset >= priv->processed + priv->count) {
+		kaio_append_call(req, kaio_read_send_pages);
+		return 0;
+	}
+
+	num = min_t(int, max_sane_readahead(priv->limit), 
+			ALIGN(priv->count, PAGE_SIZE) >> PAGE_SHIFT);
+
+	offset = priv->offset;
+	for (i=0; i<num; ++i) {
+		page = page_cache_alloc_cold(mapping);
+		if (!page)
+			break;
+
+		page->index = priv->offset >> PAGE_CACHE_SHIFT;
+		list_add(&page->lru, &page_pool);
+
+		priv->offset += PAGE_CACHE_SIZE;
+	}
+
+	dprintk("%s: submit: req: %p, priv: %p, offset: %Lu, num: %d, limit: %Lu.\n",
+			__func__, req, priv, offset, i, priv->limit);
+
+	err = mapping->a_ops->aio_readpages(file, mapping, &page_pool, i, req);
+	if (err) {
+		dprintk("%s: kevent_mpage_readpages failed: err=%d, count=%Lu.\n",
+				__func__, err, priv->count);
+		kaio_schedule_req(req);
+		return err;
+	}
+
+	return 1;
+}
+
+static int kaio_add_kevent(int fd, struct kaio_req *req)
+{
+	struct ukevent uk;
+	struct file *file;
+	struct kevent_user *u;
+	int err, need_fput = 0;
+	u32 *cnt;
+
+	file = fget_light(fd, &need_fput);
+	if (!file) {
+		err = -EBADF;
+		goto err_out;
+	}
+
+	if (file->f_op != &kevent_user_fops) {
+		err = -EINVAL;
+		goto err_out_fput;
+	}
+
+	u = file->private_data;
+
+	memset(&uk, 0, sizeof(struct ukevent));
+
+	uk.event = KEVENT_MASK_ALL;
+	uk.type = KEVENT_AIO;
+
+	preempt_disable();
+	uk.id.raw[0] = smp_processor_id();
+	cnt = &__get_cpu_var(kaio_req_counter);
+	uk.id.raw[1] = *cnt;
+	*cnt = *cnt + 1;
+	preempt_enable();
+
+	uk.req_flags = KEVENT_REQ_ONESHOT | KEVENT_REQ_ALWAYS_QUEUE;
+	uk.ptr = req;
+
+	err = kevent_user_add_ukevent(&uk, u);
+	if (err)
+		goto err_out_fput;
+
+	kevent_user_get(u);
+
+	fput_light(file, need_fput);
+
+	return 0;
+
+err_out_fput:
+	fput_light(file, need_fput);
+err_out:
+	return err;
+}
+
+static void kaio_destructor(struct kaio_req *req)
+{
+	struct kaio_private *priv = req->priv;
+	struct socket *sock = priv->dptr;
+	struct file *file = priv->sptr;
+
+	fput(file);
+	sockfd_put(sock);
+
+	kevent_storage_ready(&priv->kevent_user->st, NULL, KEVENT_MASK_ALL);
+	kevent_user_put(priv->kevent_user);
+
+	kmem_cache_free(kaio_priv_cache, req->priv);
+}
+
+static struct kaio_req *kaio_sendfile(int kevent_fd, int sock_fd, 
+		void __user *header, size_t header_size, 
+		struct file *file, off_t offset, size_t count)
+{
+	struct kaio_req *req;
+	struct socket *sock;
+	struct kaio_private *priv;
+	int err;
+
+	sock = sockfd_lookup(sock_fd, &err);
+	if (!sock)
+		goto err_out_exit;
+
+	priv = kmem_cache_alloc(kaio_priv_cache, GFP_KERNEL);
+	if (!priv)
+		goto err_out_sput;
+
+	priv->sptr = file;
+	priv->dptr = sock;
+	priv->offset = offset;
+	priv->count = min_t(u64, i_size_read(file->f_dentry->d_inode), count);
+	priv->processed = offset;
+	priv->limit = 128;
+	priv->header = header;
+	priv->header_size = header_size;
+
+	req = kaio_add_call(NULL, &kaio_read_send_pages, -1, GFP_KERNEL);
+	if (!req)
+		goto err_out_free;
+
+	req->destructor = kaio_destructor;
+	req->priv = priv;
+
+	err = kaio_add_kevent(kevent_fd, req);
+
+	dprintk("%s: req: %p, priv: %p, call: %p [%p], offset: %Lu, processed: %Lu, count: %Lu, err: %d.\n",
+			__func__, req, priv, &kaio_read_send_pages, 
+			kaio_read_send_pages, priv->offset, priv->processed, priv->count, err);
+
+	if (err)
+		goto err_out_remove;
+
+	kaio_schedule_req(req);
+
+	return req;
+
+err_out_remove:
+	/* It is safe to just free the object since it is guaranteed that it was not
+	 * queued for processing.
+	 */
+	kmem_cache_free(kaio_req_cache, req);
+err_out_free:
+	kmem_cache_free(kaio_priv_cache, priv);
+err_out_sput:
+	sockfd_put(sock);
+err_out_exit:
+	return NULL;
+
+}
+
+asmlinkage long sys_aio_sendfile(int kevent_fd, int sock_fd, int in_fd, off_t offset, size_t count)
+{
+	struct kaio_req *req;
+	struct file *file;
+	int err;
+
+	file = fget(in_fd);
+	if (!file) {
+		err = -EBADF;
+		goto err_out_exit;
+	}
+
+	req = kaio_sendfile(kevent_fd, sock_fd, NULL, 0, file, offset, count);
+	if (!req) {
+		err = -EINVAL;
+		goto err_out_fput;
+	}
+
+	return (long)req;
+
+err_out_fput:
+	fput(file);
+err_out_exit:
+	return err;
+}
+
+asmlinkage long sys_aio_sendfile_path(int kevent_fd, int sock_fd, 
+		void __user *header, size_t header_size, 
+		char __user *filename, off_t offset, size_t count)
+{
+	char *tmp = getname(filename);
+	int fd = PTR_ERR(tmp);
+	int flags = O_RDONLY, err;
+	struct nameidata nd;
+	struct file *file;
+	struct kaio_req *req;
+
+	if (force_o_largefile())
+		flags = O_LARGEFILE;
+
+	if (IS_ERR(tmp)) {
+		err = fd;
+		goto err_out_exit;
+	}
+
+	fd = get_unused_fd();
+	if (fd < 0) {
+		err = fd;
+		goto err_out_put_name;
+	}
+
+	if ((flags+1) & O_ACCMODE)
+		flags++;
+
+	err = open_namei(AT_FDCWD, tmp, flags, 0400, &nd);
+	if (err)
+		goto err_out_fdput;
+
+	file = nameidata_to_filp(&nd, flags);
+	if (!file)
+		goto err_out_fdput;
+
+	/* One reference will be released in sys_close(), 
+	 * second one through req->destructor() 
+	 */
+	atomic_inc(&file->f_count);
+
+	req = kaio_sendfile(kevent_fd, sock_fd, header, header_size, 
+			file, offset, count);
+	if (!req) {
+		err = -EINVAL;
+		goto err_out_fput;
+	}
+
+	fd_install(fd, file);
+
+	return fd;
+
+err_out_fput:
+	fput(file);
+	fput(file);
+err_out_fdput:
+	put_unused_fd(fd);
+err_out_put_name:
+	putname(tmp);
+err_out_exit:
+	return err;
+}
+
+static int kevent_aio_callback(struct kevent *k)
+{
+	struct kaio_req *req = k->event.ptr;
+	struct kaio_private *priv = req->priv;
+
+	if (!priv->count) {
+		__u32 *processed = (__u32 *)&priv->processed;
+		k->event.ret_data[0] = processed[0];
+		k->event.ret_data[1] = processed[1];
+		return 1;
+	}
+
+	return 0;
+}
+
+int kevent_aio_enqueue(struct kevent *k)
+{
+	int err;
+	struct kaio_req *req = k->event.ptr;
+	struct kaio_private *priv = req->priv;
+
+	err = kevent_storage_enqueue(&k->user->st, k);
+	if (err)
+		goto err_out_exit;
+
+	priv->kevent_user = k->user;
+	if (k->event.req_flags & KEVENT_REQ_ALWAYS_QUEUE)
+		kevent_requeue(k);
+
+	return 0;
+
+err_out_exit:
+	return err;
+}
+
+int kevent_aio_dequeue(struct kevent *k)
+{
+	kevent_storage_dequeue(k->st, k);
+
+	return 0;
+}
+
+static void kaio_thread_stop(struct kaio_thread *th)
+{
+	kthread_stop(th->thread);
+	kfree(th);
+}
+
+static int kaio_thread_start(struct kaio_thread **thp, int num)
+{
+	struct kaio_thread *th;
+
+	th = kzalloc(sizeof(struct kaio_thread), GFP_KERNEL);
+	if (!th)
+		return -ENOMEM;
+
+	th->refcnt = 1;
+	spin_lock_init(&th->req_lock);
+	INIT_LIST_HEAD(&th->req_list);
+	init_waitqueue_head(&th->wait);
+
+	th->thread = kthread_run(kaio_thread_process, th, "kaio/%d", num);
+	if (IS_ERR(th->thread)) {
+		int err = PTR_ERR(th->thread);
+		kfree(th);
+		return err;
+	}
+
+	*thp = th;
+	wmb();
+
+	return 0;
+}
+
+static int __init kevent_init_aio(void)
+{
+	struct kevent_callbacks sc = {
+		.callback = &kevent_aio_callback,
+		.enqueue = &kevent_aio_enqueue,
+		.dequeue = &kevent_aio_dequeue,
+		.flags = 0,
+	};
+	int err, i;
+
+	kaio_req_cache = kmem_cache_create("kaio_req", sizeof(struct kaio_req), 
+			0, SLAB_PANIC, NULL, NULL);
+	kaio_priv_cache = kmem_cache_create("kaio_priv", sizeof(struct kaio_private), 
+			0, SLAB_PANIC, NULL, NULL);
+
+	memset(kaio_threads, 0, sizeof(kaio_threads));
+
+	for (i=0; i<KAIO_THREAD_NUM; ++i) {
+		err = kaio_thread_start(&kaio_threads[i], i);
+		if (err)
+			goto err_out_stop;
+	}
+
+	err = kevent_add_callbacks(&sc, KEVENT_AIO);
+	if (err)
+		goto err_out_stop;
+
+	return 0;
+
+err_out_stop:
+	while (--i >= 0) {
+		struct kaio_thread *th = kaio_threads[i];
+
+		kaio_threads[i] = NULL;
+		wmb();
+
+		kaio_thread_stop(th);
+	}
+	return err;
+}
+module_init(kevent_init_aio);

/async open+send/ Re: [take35 10/10] kevent: Kevent based AIO (aio_sendfile()/aio_sendfile_path()).

[Evgeniy Polyakov]

On Thu, Feb 01, 2007 at 01:12:34PM +0300, Evgeniy Polyakov (johnpol@2ka.mipt.ru) wrote:
> 
> Kevent based AIO (aio_sendfile()/aio_sendfile_path()).
> 
> aio_sendfile()/aio_sendfile_path() contains of two major parts: AIO 
> state machine and page processing code. 
> The former is just a small subsystem, which allows to queue callback 
> for theirs invocation in process' context on behalf of pool of kernel 
> threads. It allows to queue caches of callbacks to the local thread 
> or to any other specified. Each cache of callbacks is processed until 
> there are callbacks in it, callbacks can requeue themselfs into the 
> same cache.

Below patch, casted by discussion in lkml@ about micro-thread AIO.
Async sending and file closing are implemented already, I added 
kaio_open() into state machine as example of what it can do, it was sync
before. It is an example only - kevent was asked to have feature-freeze
time.

Btw, could you please Cc: linux-fsdevel@ on AIO patchsets too.
Thanks.

diff --git a/kernel/kevent/kevent_aio.c b/kernel/kevent/kevent_aio.c
index 9bb5e99..803e9ff 100644
--- a/kernel/kevent/kevent_aio.c
+++ b/kernel/kevent/kevent_aio.c
@@ -759,16 +759,15 @@ err_out_exit:
 	return err;
 }
 
-asmlinkage long sys_aio_sendfile_path(int kevent_fd, int sock_fd, 
-		void __user *header, size_t header_size, 
-		char __user *filename, off_t offset, size_t count)
+static int kaio_open(struct kaio_req *req, int direct)
 {
+	struct kaio_private *priv = req->priv;
+	char *filename = priv->sptr;
 	char *tmp = getname(filename);
 	int fd = PTR_ERR(tmp);
 	int flags = O_RDONLY, err;
 	struct nameidata nd;
 	struct file *file;
-	struct kaio_req *req;
 
 	if (force_o_largefile())
 		flags = O_LARGEFILE;
@@ -795,25 +794,12 @@ asmlinkage long sys_aio_sendfile_path(int kevent_fd, int sock_fd,
 	if (!file)
 		goto err_out_fdput;
 
-	/* One reference will be released in sys_close(), 
-	 * second one through req->destructor() 
-	 */
-	atomic_inc(&file->f_count);
-
-	req = kaio_sendfile(kevent_fd, sock_fd, header, header_size, 
-			file, offset, count);
-	if (!req) {
-		err = -EINVAL;
-		goto err_out_fput;
-	}
-
 	fd_install(fd, file);
 
-	return fd;
+	priv->sptr = file;
+
+	return 0;
 
-err_out_fput:
-	fput(file);
-	fput(file);
 err_out_fdput:
 	put_unused_fd(fd);
 err_out_put_name:
@@ -822,6 +808,67 @@ err_out_exit:
 	return err;
 }
 
+asmlinkage long sys_aio_sendfile_path(int kevent_fd, int sock_fd, 
+		void __user *header, size_t header_size, 
+		char __user *filename, off_t offset, size_t count)
+{
+	struct kaio_req *req;
+	struct socket *sock;
+	struct kaio_private *priv;
+	int err;
+
+	sock = sockfd_lookup(sock_fd, &err);
+	if (!sock)
+		goto err_out_exit;
+
+	priv = kmem_cache_alloc(kaio_priv_cache, GFP_KERNEL);
+	if (!priv)
+		goto err_out_sput;
+
+	priv->sptr = filename;
+	priv->dptr = sock;
+	priv->offset = offset;
+	priv->count = count;
+	priv->processed = offset;
+	priv->limit = 128;
+	priv->header = header;
+	priv->header_size = header_size;
+
+	req = kaio_add_call(NULL, &kaio_open, -1, GFP_KERNEL);
+	if (!req)
+		goto err_out_free;
+	
+	kaio_append_call(req, &kaio_read_send_pages);
+
+	req->destructor = kaio_destructor;
+	req->priv = priv;
+
+	err = kaio_add_kevent(kevent_fd, req);
+
+	dprintk("%s: req: %p, priv: %p, call: %p [%p], offset: %Lu, processed: %Lu, count: %Lu, err: %d.\n",
+			__func__, req, priv, &kaio_read_send_pages, 
+			kaio_read_send_pages, priv->offset, priv->processed, priv->count, err);
+
+	if (err)
+		goto err_out_remove;
+
+	kaio_schedule_req(req);
+
+	return (long)req;
+
+err_out_remove:
+	/* It is safe to just free the object since it is guaranteed that it was not
+	 * queued for processing.
+	 */
+	kmem_cache_free(kaio_req_cache, req);
+err_out_free:
+	kmem_cache_free(kaio_priv_cache, priv);
+err_out_sput:
+	sockfd_put(sock);
+err_out_exit:
+	return -1;
+}
+
 static int kevent_aio_callback(struct kevent *k)
 {
 	struct kaio_req *req = k->event.ptr;


-- 
	Evgeniy Polyakov

Re: [take35 0/10] kevent: Generic event handling mechanism.

[Evgeniy Polyakov]

On Thu, Feb 01, 2007 at 01:12:30PM +0300, Evgeniy Polyakov (johnpol@2ka.mipt.ru) wrote:
> 
> Generic event handling mechanism.
> 
> Kevent is a generic subsytem which allows to handle event notifications.
> It supports both level and edge triggered events. It is similar to
> poll/epoll in some cases, but it is more scalable, it is faster and
> allows to work with essentially eny kind of events.
> 
> Events are provided into kernel through control syscall and can be read
> back through ring buffer or using usual syscalls.
> Kevent update (i.e. readiness switching) happens directly from internals
> of the appropriate state machine of the underlying subsytem (like
> network, filesystem, timer or any other).
> 
> Homepage:
> http://tservice.net.ru/~s0mbre/old/?section=projects&item=kevent
> 
> Documentation page:
> http://linux-net.osdl.org/index.php/Kevent
> 
> Consider for inclusion.
> 
> P.S. If you want to be removed from Cc: list just drop me a mail.
> 
> Changes from 'take34' patchset:
>  * Ported to the 2.6.20-rc7 (9be5b038b1c9d1927c367bf91683458e10d5d4eb) tree.

Is there some progress?
Some things to be implemented, described or discussed?

Thank you.

-- 
	Evgeniy Polyakov

Re: [take35 0/10] kevent: Generic event handling mechanism.

[David M. Lloyd]

On Thu, 2007-02-01 at 13:12 +0300, Evgeniy Polyakov wrote:
> Generic event handling mechanism.

The patch applied cleanly to 2.6.20 final, but I got a build error:

  CC      kernel/kevent/kevent.o
  CC      kernel/kevent/kevent_user.o
  CC      kernel/kevent/kevent_timer.o
  CC      kernel/kevent/kevent_poll.o
make[2]: *** No rule to make target `kernel/kevent/epoll.o', needed by `kernel/kevent/built-in.o'.  Stop.
make[1]: *** [kernel/kevent] Error 2
make: *** [kernel] Error 2

- DML

Re: [take35 0/10] kevent: Generic event handling mechanism.

[Evgeniy Polyakov]

On Mon, Feb 05, 2007 at 11:28:05AM -0600, David M. Lloyd (dmlloyd@flurg.com) wrote:
> On Thu, 2007-02-01 at 13:12 +0300, Evgeniy Polyakov wrote:
> > Generic event handling mechanism.
> 
> The patch applied cleanly to 2.6.20 final, but I got a build error:
> 
>   CC      kernel/kevent/kevent.o
>   CC      kernel/kevent/kevent_user.o
>   CC      kernel/kevent/kevent_timer.o
>   CC      kernel/kevent/kevent_poll.o
> make[2]: *** No rule to make target `kernel/kevent/epoll.o', needed by `kernel/kevent/built-in.o'.  Stop.
> make[1]: *** [kernel/kevent] Error 2
> make: *** [kernel] Error 2

It looks like I generated wrong diff from my tree - it compiles just
fine. This patch should fix it. That hunk is about epoll rewrite over
keent I posted couple of weeks ago.

Thanks for testing.

> - DML

diff --git a/kernel/kevent/Makefile b/kernel/kevent/Makefile
index a179bea..dc7f8b2 100644
--- a/kernel/kevent/Makefile
+++ b/kernel/kevent/Makefile
@@ -1,6 +1,6 @@
 obj-y := kevent.o kevent_user.o
 obj-$(CONFIG_KEVENT_TIMER) += kevent_timer.o
-obj-$(CONFIG_KEVENT_POLL) += kevent_poll.o epoll.o
+obj-$(CONFIG_KEVENT_POLL) += kevent_poll.o
 obj-$(CONFIG_KEVENT_SOCKET) += kevent_socket.o
 obj-$(CONFIG_KEVENT_PIPE) += kevent_pipe.o
 obj-$(CONFIG_KEVENT_SIGNAL) += kevent_signal.o


-- 
	Evgeniy Polyakov

Re: [take35 0/10] kevent: Generic event handling mechanism.

[Evgeniy Polyakov]

On Mon, Feb 05, 2007 at 05:23:36PM +0300, Evgeniy Polyakov (johnpol@2ka.mipt.ru) wrote:
> On Thu, Feb 01, 2007 at 01:12:30PM +0300, Evgeniy Polyakov (johnpol@2ka.mipt.ru) wrote:
> > 
> > Generic event handling mechanism.
> > 
> > Kevent is a generic subsytem which allows to handle event notifications.
> > It supports both level and edge triggered events. It is similar to
> > poll/epoll in some cases, but it is more scalable, it is faster and
> > allows to work with essentially eny kind of events.
> > 
> > Events are provided into kernel through control syscall and can be read
> > back through ring buffer or using usual syscalls.
> > Kevent update (i.e. readiness switching) happens directly from internals
> > of the appropriate state machine of the underlying subsytem (like
> > network, filesystem, timer or any other).
> > 
> > Homepage:
> > http://tservice.net.ru/~s0mbre/old/?section=projects&item=kevent
> > 
> > Documentation page:
> > http://linux-net.osdl.org/index.php/Kevent
> > 
> > Consider for inclusion.
> > 
> > P.S. If you want to be removed from Cc: list just drop me a mail.
> > 
> > Changes from 'take34' patchset:
> >  * Ported to the 2.6.20-rc7 (9be5b038b1c9d1927c367bf91683458e10d5d4eb) tree.
> 
> Is there some progress?
> Some things to be implemented, described or discussed?

I think that mean that everybody is happy with APi, design and set of
features. 

Andrew, do you consider kevent for inclusion or declining?

-- 
	Evgeniy Polyakov

Re: [take35 0/10] kevent: Generic event handling mechanism.

[Ulrich Drepper]

[-- Attachment #1: Type: text/plain, Size: 335 bytes --]

Evgeniy Polyakov wrote:
> I think that mean that everybody is happy with APi, design and set of
> features. 

No comment means that I still have not been able to test anything since
regardless of what version I tried, it failed to build.

-- 
➧ Ulrich Drepper ➧ Red Hat, Inc. ➧ 444 Castro St ➧ Mountain View, CA ❖


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 251 bytes --]

Re: [take35 0/10] kevent: Generic event handling mechanism.

[Andrew Morton]

> On Mon, 12 Feb 2007 13:35:10 +0300 Evgeniy Polyakov <johnpol@2ka.mipt.ru> wrote:
> Andrew, do you consider kevent for inclusion or declining?

I haven't had time to think about it in the past month or two, sorry.

However we might as well get it back in there for review-and-test - please
send a new patchset once 2.6.21-rc1 is released and copy me on it?

Re: [take35 0/10] kevent: Generic event handling mechanism.

[Evgeniy Polyakov]

On Mon, Feb 12, 2007 at 06:59:17AM -0800, Ulrich Drepper (drepper@redhat.com) wrote:
> Evgeniy Polyakov wrote:
> > I think that mean that everybody is happy with APi, design and set of
> > features. 
> 
> No comment means that I still have not been able to test anything since
> regardless of what version I tried, it failed to build.

I think if you would provide error message you saw, I fixed that in a
couple of moments, doesn't it?

I will send new patchset next week when rc1 will be out and your
feedback is greatly appreciated.

> -- 
> ➧ Ulrich Drepper ➧ Red Hat, Inc. ➧ 444 Castro St ➧ Mountain View, CA ❖
> 



-- 
	Evgeniy Polyakov

Re: [take35 0/10] kevent: Generic event handling mechanism.

[Evgeniy Polyakov]

On Mon, Feb 12, 2007 at 09:13:35AM -0800, Andrew Morton (akpm@linux-foundation.org) wrote:
> > On Mon, 12 Feb 2007 13:35:10 +0300 Evgeniy Polyakov <johnpol@2ka.mipt.ru> wrote:
> > Andrew, do you consider kevent for inclusion or declining?
> 
> I haven't had time to think about it in the past month or two, sorry.
> 
> However we might as well get it back in there for review-and-test - please
> send a new patchset once 2.6.21-rc1 is released and copy me on it?

Ok, I will.
You are always in copy for kevent patchsets.

-- 
	Evgeniy Polyakov