Attachment 195450 Details for Bug 214338 – Proposed patch (since 473551 revision)

[patch] Proposed patch (since 473551 revision)

glib20.diff (text/plain), 40.83 KB, created by lightside on 2018-07-25 16:32:35 UTC

(hide)

Description:

Filename:

MIME Type:

Creator: lightside

Created: 2018-07-25 16:32:35 UTC

Size: 40.83 KB

patch

obsolete

>Index: devel/glib20/Makefile
>===================================================================
>--- devel/glib20/Makefile	(revision 473551)
>+++ devel/glib20/Makefile	(working copy)
>@@ -48,7 +48,7 @@
> 		glib-compile-resources.1 gresource.1 gdbus-codegen.1
> gobject_MAN=	glib-genmarshal.1 glib-mkenums.1 gobject-query.1
> 
>-OPTIONS_DEFINE=	COLLATION_FIX DEBUG NLS
>+OPTIONS_DEFINE=	COLLATION_FIX DEBUG FAM_ALTBACKEND NLS
> OPTIONS_SUB=	yes
> # libc collation was fixed by https://svnweb.freebsd.org/changeset/base/290494
> COLLATION_FIX_DESC=	Use ICU for UTF-8 string collation (if libc is broken)
>@@ -65,6 +65,8 @@
> # http://www.freebsd.org/cgi/query-pr.cgi?pr=175930 see comment by ed@
> #COLLATION_FIX_CONFIGURE_ENV+=	CFLAGS="-D__STDC_ISO_10646__"
> 
>+FAM_ALTBACKEND_DESC=		Alternate file monitor backend
>+
> .include <bsd.port.pre.mk>
> 
> # doesn't build yet
>@@ -98,6 +100,13 @@
> 		s|-Werror|| ; \
> 		s|#define HAVE_SYS_INOTIFY_H 1||' ${WRKSRC}/configure
> 
>+post-patch-FAM_ALTBACKEND-on:
>+	@${REINPLACE_CMD} -e '/^libkqueue_la_OBJECTS =/s|$$(am_libkqueue_la_OBJECTS)|libkqueue_la-gkqueuefilemonitor.lo libkqueue_la-kqueue-helper.lo|' \
>+		${WRKSRC}/gio/kqueue/Makefile.in
>+	@${CP} -f ${FILESDIR}/gkqueuefilemonitor.c ${WRKSRC}/gio/kqueue/gkqueuefilemonitor.c
>+	@${CP} -f ${FILESDIR}/kqueue_fnm.c ${WRKSRC}/gio/kqueue/kqueue-helper.c
>+	@${CP} ${FILESDIR}/kqueue_fnm.h ${WRKSRC}/gio/kqueue/kqueue_fnm.h
>+
> post-install:
> 	@${MKDIR} ${STAGEDIR}${PREFIX}/share/GConf/gsettings
> 	@${MKDIR} ${STAGEDIR}${PREFIX}/lib/gio/modules
>Index: devel/glib20/files/gkqueuefilemonitor.c
>===================================================================
>--- devel/glib20/files/gkqueuefilemonitor.c	(nonexistent)
>+++ devel/glib20/files/gkqueuefilemonitor.c	(working copy)
>@@ -0,0 +1,217 @@
>+/*-
>+ * Copyright (c) 2016 - 2018 Rozhuk Ivan <rozhuk.im@gmail.com>
>+ * All rights reserved.
>+ *
>+ * Redistribution and use in source and binary forms, with or without
>+ * modification, are permitted provided that the following conditions
>+ * are met:
>+ * 1. Redistributions of source code must retain the above copyright
>+ *    notice, this list of conditions and the following disclaimer.
>+ * 2. Redistributions in binary form must reproduce the above copyright
>+ *    notice, this list of conditions and the following disclaimer in the
>+ *    documentation and/or other materials provided with the distribution.
>+ *
>+ * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
>+ * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
>+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
>+ * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
>+ * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
>+ * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
>+ * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
>+ * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
>+ * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
>+ * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
>+ * SUCH DAMAGE.
>+ *
>+ * Author: Rozhuk Ivan <rozhuk.im@gmail.com>
>+ *
>+ */
>+
>+#include "config.h"
>+
>+#include <glib-object.h>
>+#include <string.h>
>+#include <gio/gfilemonitor.h>
>+#include <gio/glocalfilemonitor.h>
>+#include <gio/giomodule.h>
>+#include "glib-private.h"
>+#include <glib-unix.h>
>+#include "kqueue_fnm.h"
>+
>+/* Defaults. */
>+#ifndef KQUEUE_MON_RATE_LIMIT_TIME_INIT
>+#	define KQUEUE_MON_RATE_LIMIT_TIME_INIT		1000
>+#endif
>+#ifndef KQUEUE_MON_RATE_LIMIT_TIME_MAX
>+#	define KQUEUE_MON_RATE_LIMIT_TIME_MAX		8000
>+#endif
>+#ifndef KQUEUE_MON_RATE_LIMIT_TIME_MUL
>+#	define KQUEUE_MON_RATE_LIMIT_TIME_MUL		2
>+#endif
>+#ifndef KQUEUE_MON_MAX_DIR_FILES
>+#	define KQUEUE_MON_MAX_DIR_FILES			128
>+#endif
>+#ifndef KQUEUE_MON_LOCAL_SUBFILES
>+#	define KQUEUE_MON_LOCAL_SUBFILES		1
>+#endif
>+#ifndef KQUEUE_MON_LOCAL_SUBDIRS
>+#	define KQUEUE_MON_LOCAL_SUBDIRS			0
>+#endif
>+
>+
>+static GMutex			kqueue_lock;
>+static volatile kq_fnm_p	kqueue_fnm = NULL;
>+/* Exclude from file changes monitoring, watch only for dirs. */
>+static const char *non_local_fs[] = {
>+	"fusefs.sshfs",
>+	NULL
>+};
>+
>+#define G_TYPE_KQUEUE_FILE_MONITOR      (g_kqueue_file_monitor_get_type())
>+#define G_KQUEUE_FILE_MONITOR(inst)     (G_TYPE_CHECK_INSTANCE_CAST((inst), \
>+					 G_TYPE_KQUEUE_FILE_MONITOR, GKqueueFileMonitor))
>+
>+typedef GLocalFileMonitorClass	GKqueueFileMonitorClass;
>+
>+typedef struct {
>+	GLocalFileMonitor	parent_instance;
>+	kq_fnmo_p		fnmo;
>+} GKqueueFileMonitor;
>+
>+GType g_kqueue_file_monitor_get_type(void);
>+G_DEFINE_TYPE_WITH_CODE (GKqueueFileMonitor, g_kqueue_file_monitor, G_TYPE_LOCAL_FILE_MONITOR,
>+       g_io_extension_point_implement(G_LOCAL_FILE_MONITOR_EXTENSION_POINT_NAME,
>+               g_define_type_id,
>+               "kqueue",
>+               10))
>+
>+
>+static void
>+kqueue_event_handler(kq_fnm_p kfnm,
>+    kq_fnmo_p fnmo, void *udata, uint32_t event,
>+    const char *base, const char *filename, const char *new_filename) {
>+	static const uint32_t kfnm_to_glib_map[] = {
>+		0,				/* KF_EVENT_NOT_CHANGED */
>+		G_FILE_MONITOR_EVENT_CREATED,	/* KF_EVENT_CREATED */
>+		G_FILE_MONITOR_EVENT_DELETED,	/* KF_EVENT_DELETED */
>+		G_FILE_MONITOR_EVENT_RENAMED,	/* KF_EVENT_RENAMED */
>+		G_FILE_MONITOR_EVENT_CHANGED	/* KF_EVENT_CHANGED */
>+	};
>+
>+	if (NULL == kfnm || NULL == filename ||
>+	    KF_EVENT_CREATED > event ||
>+	    KF_EVENT_CHANGED < event)
>+		return;
>+	g_file_monitor_source_handle_event(udata,
>+	    kfnm_to_glib_map[event],
>+	    filename, new_filename, NULL,
>+	    g_get_monotonic_time());
>+}
>+
>+static gboolean
>+g_kqueue_file_monitor_is_supported(void) {
>+	kq_file_mon_settings_t kfms;
>+
>+	if (NULL != kqueue_fnm)
>+		return (TRUE);
>+	/* Init only once. */
>+	g_mutex_lock(&kqueue_lock);
>+	if (NULL != kqueue_fnm) {
>+		g_mutex_unlock(&kqueue_lock);
>+		return (TRUE); /* Initialized while wait lock. */
>+	}
>+
>+	memset(&kfms, 0x00, sizeof(kq_file_mon_settings_t));
>+	kfms.rate_limit_time_init = KQUEUE_MON_RATE_LIMIT_TIME_INIT;
>+	kfms.rate_limit_time_max = KQUEUE_MON_RATE_LIMIT_TIME_MAX;
>+	kfms.rate_limit_time_mul = KQUEUE_MON_RATE_LIMIT_TIME_MUL;
>+	kfms.max_dir_files = KQUEUE_MON_MAX_DIR_FILES;
>+	kfms.mon_local_subfiles = KQUEUE_MON_LOCAL_SUBFILES;
>+	kfms.mon_local_subdirs = KQUEUE_MON_LOCAL_SUBDIRS;
>+	kfms.local_fs = NULL;
>+	kfms.non_local_fs = non_local_fs;
>+
>+	kqueue_fnm = kq_fnm_create(&kfms, kqueue_event_handler);
>+	if (NULL == kqueue_fnm) {
>+		g_mutex_unlock(&kqueue_lock);
>+		return (FALSE); /* Init fail. */
>+	}
>+	g_mutex_unlock(&kqueue_lock);
>+
>+	return (TRUE);
>+}
>+
>+static gboolean
>+g_kqueue_file_monitor_cancel(GFileMonitor *monitor) {
>+	GKqueueFileMonitor *gffm = G_KQUEUE_FILE_MONITOR(monitor);
>+
>+	kq_fnm_del(kqueue_fnm, gffm->fnmo);
>+	gffm->fnmo = NULL;
>+
>+	return (TRUE);
>+}
>+
>+static void
>+g_kqueue_file_monitor_finalize(GObject *object) {
>+	//GKqueueFileMonitor *gffm = G_KQUEUE_FILE_MONITOR(object);
>+
>+	//g_mutex_lock(&kqueue_lock);
>+	//kq_fnm_free(kqueue_fnm);
>+	//kqueue_fnm = NULL;
>+	//g_mutex_unlock(&kqueue_lock);
>+	//G_OBJECT_CLASS(g_kqueue_file_monitor_parent_class)->finalize(object);
>+}
>+
>+static void
>+g_kqueue_file_monitor_start(GLocalFileMonitor *local_monitor,
>+    const gchar *dirname, const gchar *basename,
>+    const gchar *filename, GFileMonitorSource *source) {
>+	GKqueueFileMonitor *gffm = G_KQUEUE_FILE_MONITOR(local_monitor);
>+
>+	g_assert(NULL != kqueue_fnm);
>+	//g_source_ref((GSource*)source);
>+
>+	if (NULL == filename) {
>+		filename = dirname;
>+	}
>+	gffm->fnmo = kq_fnm_add(kqueue_fnm, filename, source);
>+}
>+
>+static void
>+g_kqueue_file_monitor_init(GKqueueFileMonitor *monitor) {
>+
>+}
>+
>+static void
>+g_kqueue_file_monitor_class_init(GKqueueFileMonitorClass *class) {
>+	GObjectClass *gobject_class = G_OBJECT_CLASS(class);
>+	GFileMonitorClass *file_monitor_class = G_FILE_MONITOR_CLASS(class);
>+
>+	class->is_supported = g_kqueue_file_monitor_is_supported;
>+	class->start = g_kqueue_file_monitor_start;
>+	class->mount_notify = TRUE; /* TODO: ??? */
>+	file_monitor_class->cancel = g_kqueue_file_monitor_cancel;
>+	gobject_class->finalize = g_kqueue_file_monitor_finalize;
>+}
>+
>+static void
>+g_kqueue_file_monitor_class_finalize(GKqueueFileMonitorClass *class) {
>+
>+}
>+
>+void
>+g_io_module_load(GIOModule *module) {
>+
>+	g_type_module_use(G_TYPE_MODULE(module));
>+
>+	g_io_extension_point_implement(G_LOCAL_FILE_MONITOR_EXTENSION_POINT_NAME,
>+	    G_TYPE_KQUEUE_FILE_MONITOR, "kqueue", 10);
>+	g_io_extension_point_implement(G_NFS_FILE_MONITOR_EXTENSION_POINT_NAME,
>+	    G_TYPE_KQUEUE_FILE_MONITOR, "kqueue", 10);
>+}
>+
>+void
>+g_io_module_unload(GIOModule *module) {
>+
>+	g_assert_not_reached();
>+}
>Index: devel/glib20/files/kqueue_fnm.c
>===================================================================
>--- devel/glib20/files/kqueue_fnm.c	(nonexistent)
>+++ devel/glib20/files/kqueue_fnm.c	(working copy)
>@@ -0,0 +1,1086 @@
>+/*-
>+ * Copyright (c) 2016 - 2018 Rozhuk Ivan <rozhuk.im@gmail.com>
>+ * All rights reserved.
>+ *
>+ * Redistribution and use in source and binary forms, with or without
>+ * modification, are permitted provided that the following conditions
>+ * are met:
>+ * 1. Redistributions of source code must retain the above copyright
>+ *    notice, this list of conditions and the following disclaimer.
>+ * 2. Redistributions in binary form must reproduce the above copyright
>+ *    notice, this list of conditions and the following disclaimer in the
>+ *    documentation and/or other materials provided with the distribution.
>+ *
>+ * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
>+ * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
>+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
>+ * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
>+ * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
>+ * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
>+ * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
>+ * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
>+ * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
>+ * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
>+ * SUCH DAMAGE.
>+ *
>+ * Author: Rozhuk Ivan <rozhuk.im@gmail.com>
>+ *
>+ */
>+
>+#include <sys/param.h>
>+#include <sys/types.h>
>+#include <sys/event.h>
>+#include <sys/time.h>
>+#include <sys/stat.h>
>+#include <sys/mount.h>
>+#include <sys/fcntl.h> /* open, fcntl */
>+
>+#include <inttypes.h>
>+#include <stdlib.h> /* malloc, exit */
>+#include <unistd.h> /* close, write, sysconf */
>+#include <string.h> /* bcopy, bzero, memcpy, memmove, memset, strerror... */
>+#include <dirent.h> /* opendir, readdir */
>+#include <errno.h>
>+#include <pthread.h>
>+
>+#include "kqueue_fnm.h"
>+
>+
>+/* Preallocate items count. */
>+#ifndef FILES_ALLOC_BLK_SIZE
>+#	define FILES_ALLOC_BLK_SIZE	32
>+#endif
>+
>+
>+typedef struct readdir_context_s {
>+	int		fd;
>+	uint8_t		*buf;
>+	size_t		buf_size;
>+	size_t		buf_used;
>+	size_t		buf_pos;
>+} readdir_ctx_t, *readdir_ctx_p;
>+
>+
>+typedef struct file_info_s { /* Directory file. */
>+	int		fd;		/* For notify kqueue(). */
>+	struct dirent 	de;		/* d_reclen used for action. */
>+	struct stat	sb;
>+} file_info_t, *file_info_p;
>+
>+
>+typedef struct kq_file_nmon_obj_s {
>+	int		fd;		/* For notify kqueue(). */
>+	int		is_dir;
>+	int		is_local;	/* Is file system local. */
>+	struct stat	sb;
>+	char		path[(PATH_MAX + 2)];
>+	size_t		path_size;
>+	size_t		name_offset;	/* Parent path size. */
>+	uint32_t	rate_lim_cur_interval;	/* From rate_limit_time_init to rate_limit_time_max. 0 disabled. */
>+	size_t		rate_lim_ev_cnt;	/* Events count then rate_lim_cur_interval != 0 since last report. */
>+	sbintime_t	rate_lim_ev_last;	/* Last event time. */
>+	void		*udata;
>+	kq_fnm_p	kfnm;
>+	/* For dir. */
>+	file_info_p	files;
>+	volatile size_t	files_count;
>+	size_t		files_allocated;
>+} kq_fnmo_t;
>+
>+
>+typedef struct kq_file_nonify_monitor_s {
>+	int		fd;		/* kqueue() fd. */
>+	int		pfd[2];		/* pipe queue specific. */
>+	kfnm_event_handler_cb cb_func;	/* Callback on dir/file change. */
>+	kq_file_mon_settings_t s;
>+	sbintime_t	rate_lim_time_init; /* rate_limit_time_init */
>+	pthread_t	tid;
>+} kq_fnm_t;
>+
>+
>+typedef void (*kq_msg_cb)(void *arg);
>+
>+typedef struct kq_file_mon_msg_pkt_s {
>+	size_t		magic;
>+	kq_msg_cb	msg_cb;
>+	void		*arg;
>+	size_t		chk_sum;
>+} kq_fnm_msg_pkt_t, *kq_fnm_msg_pkt_p;
>+
>+#define KF_MSG_PKT_MAGIC	0xffddaa00
>+
>+
>+#ifndef O_NOATIME
>+#	define O_NOATIME	0
>+#endif
>+#ifndef O_EVTONLY
>+#	define O_EVTONLY	O_RDONLY
>+#endif
>+#define OPEN_FILE_FLAGS		(O_EVTONLY | O_NONBLOCK | O_NOFOLLOW | O_NOATIME | O_CLOEXEC)
>+
>+#define EVFILT_VNODE_SUB_FLAGS	(NOTE_WRITE |				\
>+				NOTE_EXTEND |				\
>+				NOTE_ATTRIB |				\
>+				NOTE_LINK |				\
>+				NOTE_CLOSE_WRITE)
>+
>+#define EVFILT_VNODE_FLAGS_ALL	(NOTE_DELETE |				\
>+				EVFILT_VNODE_SUB_FLAGS |		\
>+				NOTE_RENAME |				\
>+				NOTE_REVOKE)
>+
>+#ifndef _GENERIC_DIRSIZ
>+#	define _GENERIC_DIRSIZ(__de)	MIN((__de)->d_reclen, sizeof(struct dirent))
>+#endif
>+
>+#define IS_NAME_DOTS(__name)	('.' == (__name)[0] &&			\
>+				 ('\0' == (__name)[1] || 		\
>+				  ('.' == (__name)[1] && '\0' == (__name)[2])))
>+#define IS_DE_NAME_EQ(__de1, __de2)  (0 == mem_cmpn((__de1)->d_name,	\
>+						    (__de1)->d_namlen,	\
>+						    (__de2)->d_name,	\
>+						    (__de2)->d_namlen))
>+#define zalloc(__size)		calloc(1, (__size))
>+
>+#if (!defined(HAVE_REALLOCARRAY) && (!defined(__FreeBSD_version) || __FreeBSD_version < 1100000))
>+#	define reallocarray(__mem, __size, __count)	realloc((__mem), ((__size) * (__count)))
>+#endif
>+
>+#ifndef CLOCK_MONOTONIC_FAST
>+#	define CLOCK_MONOTONIC_FAST	CLOCK_MONOTONIC
>+#endif
>+
>+
>+#if defined(__FreeBSD_version) && (__FreeBSD_version < 1200040)
>+/* https://svnweb.freebsd.org/base?view=revision&revision=321686
>+ * https://svnweb.freebsd.org/base/head/sys/sys/time.h?view=markup&pathrev=321686#l198
>+ * https://svnweb.freebsd.org/base/head/sys/sys/param.h?revision=321688&view=markup#l61
>+ */
>+static __inline sbintime_t
>+mstosbt(int64_t _ms)
>+{
>+	return ((_ms * (((uint64_t)1 << 63) / 500)) >> 32);
>+}
>+#endif
>+
>+void 	*kq_fnm_proccess_events_proc(void *data);
>+
>+static inline int
>+mem_cmpn(const void *buf1, const size_t buf1_size,
>+    const void *buf2, const size_t buf2_size) {
>+
>+	if (buf1_size != buf2_size)
>+		return (((buf1_size > buf2_size) ? 127 : -127));
>+	if (0 == buf1_size || buf1 == buf2)
>+		return (0);
>+	if (NULL == buf1)
>+		return (-127);
>+	if (NULL == buf2)
>+		return (127);
>+	return (memcmp(buf1, buf2, buf1_size));
>+}
>+
>+static int
>+realloc_items(void **items, const size_t item_size,
>+    size_t *allocated, const size_t alloc_blk_cnt, const size_t count) {
>+	size_t allocated_prev, allocated_new;
>+	uint8_t *items_new;
>+
>+	if (NULL == items || 0 == item_size || NULL == allocated ||
>+	    0 == alloc_blk_cnt)
>+		return (EINVAL);
>+	allocated_prev = (*allocated);
>+	if (NULL != (*items) &&
>+	    allocated_prev > count &&
>+	    allocated_prev <= (count + alloc_blk_cnt))
>+		return (0);
>+	allocated_new = (((count / alloc_blk_cnt) + 1) * alloc_blk_cnt);
>+	items_new = (uint8_t*)reallocarray((*items), item_size, allocated_new);
>+	if (NULL == items_new) /* Realloc fail! */
>+		return (ENOMEM);
>+
>+	if (allocated_new > allocated_prev) { /* Init new mem. */
>+		memset((items_new + (allocated_prev * item_size)), 0x00,
>+		    ((allocated_new - allocated_prev) * item_size));
>+	}
>+	(*items) = items_new;
>+	(*allocated) = allocated_new;
>+
>+	return (0);
>+}
>+
>+
>+static int
>+readdir_start(int fd, struct stat *sb, size_t exp_count, readdir_ctx_p rdd) {
>+	size_t buf_size;
>+
>+	if (-1 == fd || NULL == sb || NULL == rdd)
>+		return (EINVAL);
>+	if (-1 == lseek(fd, 0, SEEK_SET))
>+		return (errno);
>+	/* Calculate buf size for getdents(). */
>+	buf_size = MAX((size_t)sb->st_size, (exp_count * sizeof(struct dirent)));
>+	if (0 == buf_size) {
>+		buf_size = (16 * PAGE_SIZE);
>+	}
>+	/* Make buf size well aligned. */
>+	if (0 != sb->st_blksize) {
>+		if (powerof2(sb->st_blksize)) {
>+			buf_size = roundup2(buf_size, sb->st_blksize);
>+		} else {
>+			buf_size = roundup(buf_size, sb->st_blksize);
>+		}
>+	} else {
>+		buf_size = round_page(buf_size);
>+	}
>+	/* Init. */
>+	memset(rdd, 0x00, sizeof(readdir_ctx_t));
>+	rdd->buf = malloc(buf_size);
>+	if (NULL == rdd->buf)
>+		return (ENOMEM);
>+	rdd->buf_size = buf_size;
>+	rdd->fd = fd;
>+
>+	return (0);
>+}
>+
>+static void
>+readdir_free(readdir_ctx_p rdd) {
>+
>+	if (NULL == rdd || NULL == rdd->buf)
>+		return;
>+	free(rdd->buf);
>+	memset(rdd, 0x00, sizeof(readdir_ctx_t));
>+}
>+
>+static int
>+readdir_next(readdir_ctx_p rdd, struct dirent *de) {
>+	int error = 0, ios;
>+	uint8_t *ptr;
>+
>+	if (NULL == rdd || NULL == rdd->buf || NULL == de)
>+		return (EINVAL);
>+
>+	for (;;) {
>+		if (rdd->buf_used <= rdd->buf_pos) {
>+			/* Called once if buf size calculated ok. */
>+			ios = getdents(rdd->fd, (char*)rdd->buf, (int)rdd->buf_size);
>+			if (-1 == ios) {
>+				error = errno;
>+				break;
>+			}
>+			if (0 == ios) {
>+				error = ESPIPE; /* EOF. */
>+				break;
>+			}
>+			rdd->buf_used = (size_t)ios;
>+			rdd->buf_pos = 0;
>+		}
>+		/* Keep data aligned. */
>+		ptr = (rdd->buf + rdd->buf_pos);
>+		memcpy(de, ptr, (sizeof(struct dirent) - sizeof(de->d_name)));
>+		if (0 == de->d_reclen) {
>+			error = ESPIPE; /* EOF. */
>+			break;
>+		}
>+		rdd->buf_pos += de->d_reclen;
>+#ifdef DT_WHT
>+		if (DT_WHT == de->d_type)
>+			continue;
>+#endif
>+		memcpy(de, ptr, _GENERIC_DIRSIZ(de));
>+		if (!IS_NAME_DOTS(de->d_name))
>+			return (0); /* OK. */
>+	}
>+
>+	/* Err or no more files. */
>+	readdir_free(rdd);
>+
>+	return (error);
>+}
>+
>+
>+static int
>+file_info_find_ni(file_info_p files, size_t files_count,
>+    file_info_p fi, size_t *idx) {
>+	size_t i;
>+	mode_t st_ftype;
>+
>+	if (NULL == files || NULL == fi || NULL == idx)
>+		return (0);
>+	st_ftype = (S_IFMT & fi->sb.st_mode);
>+	for (i = 0; i < files_count; i ++) {
>+		if ((S_IFMT & files[i].sb.st_mode) != st_ftype)
>+			continue;
>+		if ((fi->sb.st_ino != files[i].sb.st_ino ||
>+		     fi->de.d_fileno != files[i].de.d_fileno) &&
>+		    0 == IS_DE_NAME_EQ(&fi->de, &files[i].de))
>+			continue;
>+		(*idx) = i;
>+		return (1);
>+	}
>+	(*idx) = files_count;
>+	return (0);
>+}
>+
>+static int
>+file_info_find_ino(file_info_p files, size_t files_count,
>+    file_info_p fi, size_t *idx) {
>+	size_t i;
>+	mode_t st_ftype;
>+
>+	if (NULL == files || NULL == fi || NULL == idx)
>+		return (0);
>+	st_ftype = (S_IFMT & fi->sb.st_mode);
>+	for (i = 0; i < files_count; i ++) {
>+		if ((S_IFMT & files[i].sb.st_mode) != st_ftype ||
>+		    fi->sb.st_ino != files[i].sb.st_ino ||
>+		    fi->de.d_fileno != files[i].de.d_fileno)
>+			continue;
>+		(*idx) = i;
>+		return (1);
>+	}
>+	(*idx) = files_count;
>+	return (0);
>+}
>+
>+static int
>+file_info_find_name(file_info_p files, size_t files_count,
>+    file_info_p fi, size_t *idx) {
>+	size_t i;
>+	mode_t st_ftype;
>+
>+	if (NULL == files || NULL == fi || NULL == idx)
>+		return (0);
>+	st_ftype = (S_IFMT & fi->sb.st_mode);
>+	for (i = 0; i < files_count; i ++) {
>+		if ((S_IFMT & files[i].sb.st_mode) != st_ftype ||
>+		    0 == IS_DE_NAME_EQ(&fi->de, &files[i].de))
>+			continue;
>+		(*idx) = i;
>+		return (1);
>+	}
>+	(*idx) = files_count;
>+	return (0);
>+}
>+
>+static void
>+file_info_fd_close(file_info_p files, size_t files_count) {
>+	size_t i;
>+
>+	if (NULL == files || 0 == files_count)
>+		return;
>+	for (i = 0; i < files_count; i ++) {
>+		if (-1 == files[i].fd)
>+			continue;
>+		close(files[i].fd);
>+		files[i].fd = -1;
>+	}
>+}
>+
>+
>+static int
>+is_fs_local(struct statfs *stfs, const char **local_fs, const char **non_local_fs) {
>+	size_t i;
>+
>+	if (NULL == stfs)
>+		return (0);
>+	if (NULL != local_fs) {
>+		for (i = 0; NULL != local_fs[i]; i ++) {
>+			if (0 == strncmp(stfs->f_fstypename, local_fs[i],
>+			    sizeof(stfs->f_fstypename)))
>+				return (1);
>+		}
>+	}
>+	if (0 == (MNT_LOCAL & stfs->f_flags))
>+		return (0);
>+	if (NULL != non_local_fs) {
>+		for (i = 0; NULL != non_local_fs[i]; i ++) {
>+			if (0 == strncmp(stfs->f_fstypename, non_local_fs[i],
>+			    sizeof(stfs->f_fstypename)))
>+				return (0);
>+		}
>+	}
>+	return (1);
>+}
>+
>+
>+static void
>+kq_fnmo_rate_lim_stop(kq_fnmo_p fnmo) {
>+	struct kevent kev;
>+
>+	if (NULL == fnmo || -1 == fnmo->fd || 0 == fnmo->rate_lim_cur_interval)
>+		return;
>+	fnmo->rate_lim_cur_interval = 0;
>+	fnmo->rate_lim_ev_cnt = 0;
>+	EV_SET(&kev, fnmo->fd, EVFILT_TIMER, EV_DELETE, 0, 0, NULL);
>+	kevent(fnmo->kfnm->fd, &kev, 1, NULL, 0, NULL);
>+}
>+
>+static int
>+kq_fnmo_rate_lim_shedule_next(kq_fnmo_p fnmo) {
>+	u_short flags = (EV_ADD | EV_CLEAR | EV_ONESHOT);
>+	struct kevent kev;
>+
>+	if (NULL == fnmo || -1 == fnmo->fd || 0 == fnmo->kfnm->s.rate_limit_time_init)
>+		return (EINVAL);
>+	if (0 == fnmo->rate_lim_cur_interval) { /* First call. */
>+		fnmo->rate_lim_cur_interval = fnmo->kfnm->s.rate_limit_time_init;
>+	} else {
>+		if (fnmo->rate_lim_cur_interval == fnmo->kfnm->s.rate_limit_time_max)
>+			return (0); /* No need to modify timer. */
>+		/* Increase rate limit interval. */
>+		fnmo->rate_lim_cur_interval *= fnmo->kfnm->s.rate_limit_time_mul;
>+	}
>+	if (fnmo->rate_lim_cur_interval >= fnmo->kfnm->s.rate_limit_time_max) {
>+		/* Check upper limit and shedule periodic timer with upper rate limit time. */
>+		flags &= ~EV_ONESHOT;
>+		fnmo->rate_lim_cur_interval = fnmo->kfnm->s.rate_limit_time_max;
>+	}
>+	EV_SET(&kev, fnmo->fd, EVFILT_TIMER, flags,
>+	    NOTE_MSECONDS, fnmo->rate_lim_cur_interval, fnmo);
>+	if (-1 == kevent(fnmo->kfnm->fd, &kev, 1, NULL, 0, NULL)) {
>+		fnmo->rate_lim_cur_interval = 0;
>+		return (errno);
>+	}
>+	return (0);
>+}
>+
>+/* Return:
>+ * 0 for events that not handled
>+ * 1 for handled = rate limited
>+ * -1 on error.
>+ */
>+static int
>+kq_fnmo_rate_lim_check(kq_fnmo_p fnmo) {
>+	sbintime_t sbt, sbt_now;
>+	struct timespec ts;
>+
>+	if (NULL == fnmo)
>+		return (-1);
>+	if (-1 == fnmo->fd ||
>+	    0 == fnmo->kfnm->s.rate_limit_time_init)
>+		return (0);
>+	if (0 != fnmo->rate_lim_cur_interval) {
>+		fnmo->rate_lim_ev_cnt ++; /* Count event, timer is active. */
>+		return (1);
>+	}
>+
>+	/* Do we need to enable rate limit? */
>+	if (0 != clock_gettime(CLOCK_MONOTONIC_FAST, &ts))
>+		return (-1);
>+	sbt_now = tstosbt(ts);
>+	sbt = (fnmo->rate_lim_ev_last + fnmo->kfnm->rate_lim_time_init);
>+	fnmo->rate_lim_ev_last = sbt_now;
>+	if (sbt < sbt_now) /* Events rate to low. */
>+		return (0);
>+	/* Try to enable rate limit. */
>+	if (0 != kq_fnmo_rate_lim_shedule_next(fnmo))
>+		return (-1);
>+	/* Ok. */
>+	fnmo->rate_lim_ev_cnt ++;
>+
>+	return (1);
>+}
>+
>+static void
>+kq_fnmo_clean(kq_fnmo_p fnmo) {
>+
>+	if (NULL == fnmo)
>+		return;
>+	if (-1 != fnmo->fd) {
>+		kq_fnmo_rate_lim_stop(fnmo);
>+		close(fnmo->fd);
>+		fnmo->fd = -1;
>+	}
>+	if (0 != fnmo->is_local) { /* Stop monitoring files/dirs. */
>+		file_info_fd_close(fnmo->files, fnmo->files_count);
>+	}
>+	free(fnmo->files);
>+	fnmo->files = NULL;
>+	fnmo->files_count = 0;
>+	fnmo->files_allocated = 0;
>+}
>+
>+static void
>+kq_fnmo_free(void *arg) {
>+	kq_fnmo_p fnmo = arg;
>+
>+	if (NULL == fnmo)
>+		return;
>+	kq_fnmo_clean(fnmo);
>+	free(fnmo);
>+}
>+
>+static kq_fnmo_p
>+kq_fnmo_alloc(kq_fnm_p kfnm, const char *path, void *udata) {
>+	kq_fnmo_p fnmo;
>+
>+	if (NULL == kfnm || NULL == path)
>+		return (NULL);
>+	fnmo = zalloc(sizeof(kq_fnmo_t));
>+	if (NULL == fnmo)
>+		return (NULL);
>+	/* Remember args. */
>+	fnmo->path_size = strlcpy(fnmo->path, path, PATH_MAX);
>+	fnmo->name_offset = fnmo->path_size;
>+	fnmo->udata = udata;
>+	fnmo->kfnm = kfnm;
>+
>+	return (fnmo);
>+}
>+
>+static int
>+kq_fnmo_readdir(kq_fnmo_p fnmo, size_t exp_count) {
>+	int error;
>+	struct dirent *de;
>+	file_info_p tmfi;
>+	readdir_ctx_t rdd;
>+
>+	if (NULL == fnmo || 0 == fnmo->is_dir)
>+		return (EINVAL);
>+
>+	free(fnmo->files);
>+	fnmo->files = NULL;
>+	fnmo->files_count = 0;
>+	fnmo->files_allocated = 0;
>+	/* Pre allocate. */
>+	if (0 != realloc_items((void**)&fnmo->files,
>+	    sizeof(file_info_t), &fnmo->files_allocated,
>+	    FILES_ALLOC_BLK_SIZE, (exp_count + 1)))
>+		return (ENOMEM);
>+
>+	error = readdir_start(fnmo->fd, &fnmo->sb, exp_count, &rdd);
>+	if (0 != error)
>+		return (error);
>+	for (;;) {
>+		if (0 != realloc_items((void**)&fnmo->files,
>+		    sizeof(file_info_t), &fnmo->files_allocated,
>+		    FILES_ALLOC_BLK_SIZE, fnmo->files_count)) {
>+			free(fnmo->files);
>+			fnmo->files = NULL;
>+			fnmo->files_count = 0;
>+			fnmo->files_allocated = 0;
>+			readdir_free(&rdd);
>+			return (ENOMEM);
>+		}
>+		de = &fnmo->files[fnmo->files_count].de; /* Use short name. */
>+		/* Get file name from folder. */
>+		if (0 != readdir_next(&rdd, de))
>+			break;
>+		/* Get file attrs. */
>+		if (0 != fstatat(fnmo->fd, de->d_name,
>+		    &fnmo->files[fnmo->files_count].sb,
>+		    AT_SYMLINK_NOFOLLOW)) {
>+			memset(&fnmo->files[fnmo->files_count].sb, 0x00,
>+			    sizeof(struct stat));
>+		}
>+		fnmo->files[fnmo->files_count].fd = -1;
>+		fnmo->files_count ++;
>+	}
>+	/* Mem compact. */
>+	tmfi = reallocarray(fnmo->files, sizeof(file_info_t), (fnmo->files_count + 1));
>+	if (NULL != tmfi) { /* realloc ok. */
>+		fnmo->files = tmfi;
>+		fnmo->files_allocated = (fnmo->files_count + 1);
>+	}
>+
>+	readdir_free(&rdd);
>+
>+	return (0); /* OK. */
>+}
>+
>+
>+static void
>+kq_fnmo_fi_start(kq_fnmo_p fnmo, file_info_p fi) {
>+	struct kevent kev;
>+
>+	if (NULL == fnmo || NULL == fi)
>+		return;
>+	fi->fd = openat(fnmo->fd, fi->de.d_name, OPEN_FILE_FLAGS);
>+	if (-1 == fi->fd)
>+		return;
>+	EV_SET(&kev, fi->fd, EVFILT_VNODE,
>+	    (EV_ADD | EV_CLEAR),
>+	    EVFILT_VNODE_SUB_FLAGS, 0, fnmo);
>+	kevent(fnmo->kfnm->fd, &kev, 1, NULL, 0, NULL);
>+}
>+
>+static int
>+kq_fnmo_is_fi_monitored(kq_fnmo_p fnmo, file_info_p fi) {
>+
>+	if (NULL == fnmo)
>+		return (0);
>+	if (0 == fnmo->is_local ||
>+	    (0 != fnmo->kfnm->s.max_dir_files &&
>+	     fnmo->kfnm->s.max_dir_files < fnmo->files_count))
>+		return (0);
>+	if (NULL != fi &&
>+	    0 == fnmo->kfnm->s.mon_local_subdirs &&
>+	    S_ISDIR(fi->sb.st_mode))
>+		return (0);
>+	return (1);
>+}
>+
>+static void
>+kq_fnmo_init(void *arg) {
>+	kq_fnmo_p fnmo = arg;
>+	size_t i;
>+	struct statfs stfs;
>+	struct kevent kev;
>+
>+	if (NULL == fnmo)
>+		return;
>+	fnmo->fd = open(fnmo->path, OPEN_FILE_FLAGS);
>+	if (-1 == fnmo->fd)
>+		return;
>+	if (0 != fstat(fnmo->fd, &fnmo->sb))
>+		goto err_out;
>+	
>+	/* Get parent folder name. */
>+	if (S_ISDIR(fnmo->sb.st_mode)) {
>+		fnmo->is_dir = 1;
>+		/* Be sure that folder contain trailing '/'. */
>+		if ('/' != fnmo->path[(fnmo->path_size - 1)]) {
>+			fnmo->path[fnmo->path_size] = '/';
>+			fnmo->path_size ++;
>+			fnmo->path[fnmo->path_size] = 0;
>+		}
>+		/* Skip last '/' for parent dir search. */
>+		fnmo->name_offset = (fnmo->path_size - 1);
>+	}
>+
>+	/* Is file system local? */
>+	if (0 != fnmo->is_dir &&
>+	    0 != fnmo->kfnm->s.mon_local_subfiles &&
>+	    0 == fstatfs(fnmo->fd, &stfs)) {
>+		fnmo->is_local = is_fs_local(&stfs, fnmo->kfnm->s.local_fs,
>+		    fnmo->kfnm->s.non_local_fs);
>+	}
>+
>+	/* Find parent dir path size. */
>+	while (0 < fnmo->name_offset && '/' != fnmo->path[(fnmo->name_offset - 1)]) {
>+		fnmo->name_offset --;
>+	}
>+
>+	/* Dir special processing. */
>+	if (0 != fnmo->is_dir) {
>+		/* Read and remember dir content. */
>+		if (0 != kq_fnmo_readdir(fnmo, 0))
>+			goto err_out;
>+	}
>+	/* Add to kqueue. */
>+	EV_SET(&kev, fnmo->fd, EVFILT_VNODE, (EV_ADD | EV_CLEAR),
>+	    EVFILT_VNODE_FLAGS_ALL, 0, fnmo);
>+	if (-1 == kevent(fnmo->kfnm->fd, &kev, 1, NULL, 0, NULL))
>+		goto err_out;
>+	/* Add monitor sub files/dirs, ignory errors. */
>+	/* Check twice for performance reason. */
>+	if (0 != kq_fnmo_is_fi_monitored(fnmo, NULL)) {
>+		for (i = 0; i < fnmo->files_count; i ++) {
>+			if (0 != kq_fnmo_is_fi_monitored(fnmo, &fnmo->files[i])) {
>+				kq_fnmo_fi_start(fnmo, &fnmo->files[i]);
>+			}
>+		}
>+	}
>+
>+	return; /* OK. */
>+
>+err_out:
>+	kq_fnmo_clean(fnmo);
>+}
>+
>+static void
>+kq_handle_changes(kq_fnm_p kfnm, kq_fnmo_p fnmo) {
>+	size_t i, k, files_count;
>+	file_info_p files;
>+
>+	if (NULL == kfnm || NULL == fnmo)
>+		return;
>+	if (0 != fstat(fnmo->fd, &fnmo->sb) ||
>+	    0 == fnmo->sb.st_nlink) {
>+		kq_fnmo_clean(fnmo);
>+		kfnm->cb_func(kfnm, fnmo, fnmo->udata, KF_EVENT_DELETED,
>+		    fnmo->path, "", NULL);
>+		return;
>+	}
>+	if (0 == fnmo->is_dir) {
>+		fnmo->path[fnmo->name_offset] = 0;
>+		kfnm->cb_func(kfnm, fnmo, fnmo->udata, KF_EVENT_CHANGED,
>+		    fnmo->path, (fnmo->path + fnmo->name_offset), NULL);
>+		fnmo->path[fnmo->name_offset] = '/';
>+		return;
>+	}
>+
>+	/* Dir processing. */
>+
>+	/* Save prev. */
>+	files = fnmo->files;
>+	files_count = fnmo->files_count;
>+	fnmo->files = NULL;
>+	fnmo->files_count = 0;
>+	/* Update dir. */
>+	if (0 != kq_fnmo_readdir(fnmo, files_count)) {
>+		/* Restore prev state on fail. */
>+		fnmo->files = files;
>+		fnmo->files_count = files_count;
>+		return;
>+	}
>+	/* Notify removed first. */
>+	for (i = 0; i < files_count; i ++) {
>+		if (0 != file_info_find_ni(fnmo->files, fnmo->files_count, &files[i], &k)) /* Deleted? */
>+			continue;
>+		if (-1 != files[i].fd) {
>+			close(files[i].fd);
>+			files[i].fd = -1;
>+		}
>+		kfnm->cb_func(kfnm, fnmo, fnmo->udata, KF_EVENT_DELETED,
>+		    fnmo->path, files[i].de.d_name, NULL);
>+	}
>+	/* Notify. */
>+	for (i = 0; i < fnmo->files_count; i ++) {
>+		/* Is new file/folder? */
>+		if (0 == file_info_find_ino(files, files_count, &fnmo->files[i], &k) &&
>+		    0 == file_info_find_name(files, files_count, &fnmo->files[i], &k)) { /* Add new. */
>+			/* Add monitor sub files/dirs, ignory errors. */
>+			if (0 != kq_fnmo_is_fi_monitored(fnmo, &fnmo->files[i])) {
>+				kq_fnmo_fi_start(fnmo, &fnmo->files[i]);
>+			}
>+			kfnm->cb_func(kfnm, fnmo, fnmo->udata, KF_EVENT_CREATED,
>+			    fnmo->path, fnmo->files[i].de.d_name, NULL);
>+			continue;
>+		}
>+		/* Keep file fd. */
>+		fnmo->files[i].fd = files[k].fd;
>+		files[k].fd = -1;
>+		/* Is renamed? */
>+		if (0 == IS_DE_NAME_EQ(&files[k].de, &fnmo->files[i].de)) {
>+			kfnm->cb_func(kfnm, fnmo, fnmo->udata, KF_EVENT_RENAMED,
>+			    fnmo->path, files[k].de.d_name, fnmo->files[i].de.d_name);
>+			continue;
>+		}
>+		/* Is modified? */
>+		if (0 != memcmp(&fnmo->files[i].sb, &files[k].sb, sizeof(struct stat))) {
>+			kfnm->cb_func(kfnm, fnmo, fnmo->udata, KF_EVENT_CHANGED,
>+			    fnmo->path, fnmo->files[i].de.d_name, NULL);
>+			continue;
>+		}
>+		/* Not changed. */
>+	}
>+
>+	/* Prevent FD leak die to race conditions.
>+	 * All fd must be -1, check this while debuging.
>+	 */
>+	file_info_fd_close(files, files_count);
>+	free(files);
>+}
>+
>+static void
>+kq_handle_rename(kq_fnm_p kfnm, kq_fnmo_p fnmo) {
>+	int up_dir_fd, found = 0;
>+	readdir_ctx_t rdd;
>+	struct dirent de;
>+	struct stat sb;
>+	char old_filename[(MAXNAMLEN + 2)];
>+	size_t old_filename_size;
>+
>+	if (NULL == kfnm || NULL == fnmo)
>+		return;
>+	if (0 != fstat(fnmo->fd, &fnmo->sb) ||
>+	    0 == fnmo->sb.st_nlink) {
>+notify_removed:
>+		kq_fnmo_clean(fnmo);
>+		kfnm->cb_func(kfnm, fnmo, fnmo->udata, KF_EVENT_DELETED,
>+		    fnmo->path, "", NULL);
>+		return;
>+	}
>+	/* Save old file name. */
>+	old_filename_size = (fnmo->path_size - fnmo->name_offset - (size_t)fnmo->is_dir);
>+	memcpy(old_filename,
>+	    (fnmo->path + fnmo->name_offset),
>+	    old_filename_size);
>+	old_filename[old_filename_size] = 0;
>+
>+	/* Get parent folder name. */
>+	fnmo->path[fnmo->name_offset] = 0;
>+	/* Try to open. */
>+	up_dir_fd = open(fnmo->path, (OPEN_FILE_FLAGS | O_DIRECTORY));
>+	/* Restore '/' after parent folder. */
>+	fnmo->path[fnmo->name_offset] = '/';
>+	if (-1 == up_dir_fd ||
>+	    0 != fstat(up_dir_fd, &sb) ||
>+	    0 != readdir_start(up_dir_fd, &sb, 0, &rdd)) {
>+		close(up_dir_fd);
>+		return;
>+	}
>+	/* Find new name by inode. */
>+	while (0 == readdir_next(&rdd, &de)) {
>+		if (0 == fstatat(up_dir_fd, de.d_name, &sb, AT_SYMLINK_NOFOLLOW) &&
>+		    0 == memcmp(&fnmo->sb, &sb, sizeof(struct stat))) {
>+			found ++;
>+			break;
>+		}
>+	}
>+	close(up_dir_fd);
>+	if (0 == found)
>+		goto notify_removed; /* Not found. */
>+	/* Update name. */
>+	if (PATH_MAX <= (fnmo->name_offset + de.d_namlen))
>+		return; /* Too long. */
>+	memcpy((fnmo->path + fnmo->name_offset), de.d_name, de.d_namlen);
>+	fnmo->path_size = (fnmo->name_offset + de.d_namlen);
>+	/* Add last '/' for dir. */
>+	fnmo->path[fnmo->path_size] = '/';
>+	fnmo->path_size += (size_t)fnmo->is_dir;
>+	fnmo->path[fnmo->path_size] = 0;
>+	/* Notify. */
>+	kfnm->cb_func(kfnm, fnmo, fnmo->udata, KF_EVENT_RENAMED,
>+	    fnmo->path, old_filename, de.d_name);
>+}
>+
>+
>+static void
>+kq_fnm_delay_call_process(kq_fnm_p kfnm, kq_msg_cb forced_msg_cb) {
>+	ssize_t ios;
>+	kq_fnm_msg_pkt_t msg;
>+
>+	for (;;) {
>+		ios = read(kfnm->pfd[0], &msg, sizeof(msg));
>+		if (0 >= ios)
>+			return;
>+		if (sizeof(msg) != ios ||
>+		    KF_MSG_PKT_MAGIC != msg.magic ||
>+		    (((size_t)msg.msg_cb) ^ ((size_t)msg.arg)) != msg.chk_sum)
>+			continue;
>+		if (NULL != forced_msg_cb) {
>+			forced_msg_cb(msg.arg);
>+			continue;
>+		}
>+		if (NULL == msg.msg_cb)
>+			continue;
>+		msg.msg_cb(msg.arg);
>+	}
>+}
>+
>+static int
>+kq_fnm_delay_call(kq_fnm_p kfnm, kq_msg_cb msg_cb,
>+    void *arg) {
>+	kq_fnm_msg_pkt_t msg;
>+
>+	if (NULL == kfnm || NULL == arg)
>+		return (EINVAL);
>+	msg.magic = KF_MSG_PKT_MAGIC;
>+	msg.msg_cb = msg_cb;
>+	msg.arg = arg;
>+	msg.chk_sum = (((size_t)msg.msg_cb) ^ ((size_t)msg.arg));
>+	if (sizeof(msg) == write(kfnm->pfd[1], &msg, sizeof(msg)))
>+		return (0);
>+	return (errno);
>+}
>+
>+
>+static void
>+kq_fnm_free_cb(void *arg) {
>+	kq_fnm_p kfnm = arg;
>+
>+	if (NULL == kfnm)
>+		return;
>+	close(kfnm->fd);
>+	kfnm->fd = -1;
>+}
>+void
>+kq_fnm_free(kq_fnm_p kfnm) {
>+
>+	if (NULL == kfnm)
>+		return;
>+	kq_fnm_delay_call(kfnm, kq_fnm_free_cb, kfnm);
>+	pthread_join(kfnm->tid, NULL);
>+	/* Free all in delay calls queue. */
>+	close(kfnm->pfd[1]);
>+	kq_fnm_delay_call_process(kfnm, kq_fnmo_free);
>+	close(kfnm->pfd[0]);
>+	free(kfnm);
>+}
>+
>+kq_fnm_p
>+kq_fnm_create(kq_file_mon_settings_p s, kfnm_event_handler_cb cb_func) {
>+	kq_fnm_p kfnm;
>+	struct kevent kev;
>+
>+	if (NULL == s || NULL == cb_func)
>+		return (NULL);
>+	kfnm = zalloc(sizeof(kq_fnm_t));
>+	if (NULL == kfnm)
>+		return (NULL);
>+	kfnm->fd = kqueue();
>+	if (-1 == kfnm->fd)
>+		goto err_out;
>+	if (-1 == pipe2(kfnm->pfd, O_NONBLOCK))
>+		goto err_out;
>+	kfnm->cb_func = cb_func;
>+	memcpy(&kfnm->s, s, sizeof(kq_file_mon_settings_t));
>+	if (kfnm->s.rate_limit_time_init >= kfnm->s.rate_limit_time_max) {
>+		kfnm->s.rate_limit_time_max = kfnm->s.rate_limit_time_init;
>+	}
>+	if (0 == kfnm->s.rate_limit_time_mul) {
>+		kfnm->s.rate_limit_time_mul ++;
>+	}
>+	kfnm->rate_lim_time_init = mstosbt(kfnm->s.rate_limit_time_init);
>+
>+	EV_SET(&kev, kfnm->pfd[0], EVFILT_READ, EV_ADD, 0, 0, NULL);
>+	if (-1 == kevent(kfnm->fd, &kev, 1, NULL, 0, NULL))
>+		goto err_out;
>+	if (0 != pthread_create(&kfnm->tid, NULL,
>+	    kq_fnm_proccess_events_proc, kfnm))
>+		goto err_out;
>+
>+	return (kfnm);
>+
>+err_out:
>+	kq_fnm_free(kfnm);
>+	return (NULL);
>+}
>+
>+kq_fnmo_p
>+kq_fnm_add(kq_fnm_p kfnm, const char *path, void *udata) {
>+	int error;
>+	kq_fnmo_p fnmo;
>+
>+	if (NULL == kfnm || NULL == path)
>+		return (NULL);
>+	fnmo = kq_fnmo_alloc(kfnm, path, udata);
>+	if (NULL == fnmo)
>+		return (NULL);
>+	/* Shedule delay call to init. */
>+	error = kq_fnm_delay_call(kfnm, kq_fnmo_init, fnmo);
>+	if (0 != error) { /* Error, do no directly init to avoid freezes. */
>+		kq_fnmo_free(fnmo);
>+		return (NULL);
>+	}
>+	return (fnmo);
>+}
>+
>+void
>+kq_fnm_del(kq_fnm_p kfnm, kq_fnmo_p fnmo) {
>+	int error;
>+
>+	if (NULL == kfnm || NULL == fnmo)
>+		return;
>+	/* Cancel notifications. */
>+	kq_fnmo_rate_lim_stop(fnmo);
>+	close(fnmo->fd);
>+	fnmo->fd = -1;
>+	/* Shedule delay call to free. */
>+	error = kq_fnm_delay_call(kfnm, kq_fnmo_free, fnmo);
>+	if (0 == error)
>+		return;
>+	/* Error, free directly. */
>+	kq_fnmo_free(fnmo);
>+}
>+
>+
>+static void
>+kq_fnm_proccess_event(kq_fnm_p kfnm, struct kevent *kev) {
>+	kq_fnmo_p fnmo;
>+	file_info_p fi;
>+	size_t i;
>+	int is_rate_lim_checked = 0;
>+	struct stat sb;
>+
>+	if (NULL == kfnm || NULL == kev)
>+		return;
>+
>+	/* Handle delay calls. */
>+	if (kev->ident == (uintptr_t)kfnm->pfd[0]) {
>+		if (kev->filter == EVFILT_READ) {
>+			kq_fnm_delay_call_process(kfnm, NULL);
>+		}
>+		return;
>+	}
>+
>+	if (0 == kev->udata)
>+		return; /* No associated data, skip. */
>+	fnmo = (kq_fnmo_p)kev->udata;
>+
>+	/* FS delayed notifications. */
>+	if (EVFILT_TIMER == kev->filter) {
>+		if (0 == fnmo->rate_lim_ev_cnt) {
>+			/* No delayed events, disable rate limit polling. */
>+			kq_fnmo_rate_lim_stop(fnmo);
>+			return;
>+		}
>+		fnmo->rate_lim_ev_cnt = 0; /* Reset counter. */
>+		kq_fnmo_rate_lim_shedule_next(fnmo);
>+		kq_handle_changes(kfnm, fnmo);
>+		return;
>+	}
>+
>+	/* FS notifications. */
>+	if (EVFILT_VNODE != kev->filter)
>+		return; /* Unknown event, skip. */
>+	/* Subdir/file */
>+	if (kev->ident != (uintptr_t)fnmo->fd) {
>+		/* Is files changes rate limited? */
>+		if (1 == kq_fnmo_rate_lim_check(fnmo))
>+			return;
>+		is_rate_lim_checked ++;
>+		/* Try to find file and check it, without call kq_handle_changes(). */
>+		fi = NULL;
>+		for (i = 0; i < fnmo->files_count; i ++) {
>+			if (kev->ident != (uintptr_t)fnmo->files[i].fd)
>+				continue;
>+			fi = &fnmo->files[i];
>+			break;
>+		}
>+		if (NULL != fi) {
>+			/* Get file attrs. */
>+			if (0 != fstat(fi->fd, &sb)) {
>+				memset(&sb, 0x00, sizeof(struct stat));
>+			}
>+			/* Is modified? */
>+			if (0 != memcmp(&fi->sb, &sb, sizeof(struct stat))) {
>+				memcpy(&fi->sb, &sb, sizeof(struct stat));
>+				kfnm->cb_func(kfnm, fnmo, fnmo->udata, KF_EVENT_CHANGED,
>+				    fnmo->path, fi->de.d_name, NULL);
>+				return;
>+			}
>+		}
>+		/* fd not found or changes not found, rescan dir. */
>+		kev->fflags = NOTE_WRITE;
>+	}
>+	/* Monitored object. */
>+	/* All flags from EVFILT_VNODE_FLAGS_ALL must be handled here. */
>+	if (EV_ERROR & kev->flags) {
>+		kev->fflags |= NOTE_REVOKE; /* Treat error as unmount. */
>+	}
>+	if (NOTE_RENAME & kev->fflags) {
>+		kq_handle_rename(kfnm, fnmo);
>+	}
>+	if ((NOTE_WRITE | NOTE_EXTEND | NOTE_ATTRIB | NOTE_LINK | NOTE_CLOSE_WRITE) & kev->fflags) {
>+		/* Only count changes, do not prevent NOTE_DELETE event handling. */
>+		if (0 == is_rate_lim_checked &&
>+		    1 != kq_fnmo_rate_lim_check(fnmo)) {
>+			kq_handle_changes(kfnm, fnmo);
>+		}
>+	}
>+	if ((NOTE_DELETE | NOTE_REVOKE) & kev->fflags) {
>+		/* No report about childs. */
>+		kq_fnmo_clean(fnmo);
>+		kfnm->cb_func(kfnm, fnmo, fnmo->udata, KF_EVENT_DELETED,
>+		    fnmo->path, "", NULL);
>+	}
>+}
>+
>+void *
>+kq_fnm_proccess_events_proc(void *data) {
>+	struct kevent kev;
>+	kq_fnm_p kfnm = data;
>+
>+	if (NULL == kfnm)
>+		return (NULL);
>+	/* Get and proccess events, no wait. */
>+	while (0 < kevent(kfnm->fd, NULL, 0, &kev, 1, NULL)) {
>+		kq_fnm_proccess_event(kfnm, &kev);
>+	}
>+	return (NULL);
>+}
>Index: devel/glib20/files/kqueue_fnm.h
>===================================================================
>--- devel/glib20/files/kqueue_fnm.h	(nonexistent)
>+++ devel/glib20/files/kqueue_fnm.h	(working copy)
>@@ -0,0 +1,74 @@
>+/*-
>+ * Copyright (c) 2016 - 2018 Rozhuk Ivan <rozhuk.im@gmail.com>
>+ * All rights reserved.
>+ *
>+ * Redistribution and use in source and binary forms, with or without
>+ * modification, are permitted provided that the following conditions
>+ * are met:
>+ * 1. Redistributions of source code must retain the above copyright
>+ *    notice, this list of conditions and the following disclaimer.
>+ * 2. Redistributions in binary form must reproduce the above copyright
>+ *    notice, this list of conditions and the following disclaimer in the
>+ *    documentation and/or other materials provided with the distribution.
>+ *
>+ * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
>+ * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
>+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
>+ * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
>+ * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
>+ * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
>+ * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
>+ * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
>+ * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
>+ * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
>+ * SUCH DAMAGE.
>+ *
>+ * Author: Rozhuk Ivan <rozhuk.im@gmail.com>
>+ *
>+ */
>+
>+#ifndef __KQUEUE_FILE_NOTIFY_MONITOR_H__
>+#define __KQUEUE_FILE_NOTIFY_MONITOR_H__
>+
>+#include <sys/param.h>
>+#include <sys/types.h>
>+#include <inttypes.h>
>+
>+
>+typedef struct kq_file_nonify_monitor_s		*kq_fnm_p;
>+typedef struct kq_file_nmon_obj_s		*kq_fnmo_p;
>+
>+typedef void (*kfnm_event_handler_cb)(kq_fnm_p kfnm,
>+					kq_fnmo_p fnmo, void *udata,
>+					uint32_t event,
>+					const char *base,
>+					const char *filename,
>+					const char *new_filename);
>+#define KF_EVENT_NOT_CHANGED	0 /* Internal use. */
>+#define KF_EVENT_CREATED	1
>+#define KF_EVENT_DELETED	2
>+#define KF_EVENT_RENAMED	3
>+#define KF_EVENT_CHANGED	4
>+
>+
>+typedef struct kq_file_nonify_mon_settings_s {
>+	uint32_t	rate_limit_time_init;	/* Fire events for dir min interval, mseconds. */
>+	uint32_t	rate_limit_time_max;	/* Fire events for dir max interval, mseconds. */
>+	uint32_t	rate_limit_time_mul;	/* Fire events time increment, mseconds. */
>+	size_t		max_dir_files;		/* If dir contain more than n files - do not mon files changes. */
>+	int		mon_local_subfiles;	/* Enable monitoring files changes on local file systems. */
>+	int		mon_local_subdirs;	/* Also mon for subdirs changes . */
>+	const char 	**local_fs;		/* NULL terminated fs names list that threat as local. Keep utill kq_fnm_free() return. */
>+	const char	**non_local_fs;		/* NULL terminated fs names list that threat as not local. Keep utill kq_fnm_free() return. */
>+} kq_file_mon_settings_t, *kq_file_mon_settings_p;
>+
>+kq_fnm_p	kq_fnm_create(kq_file_mon_settings_p s,
>+		    kfnm_event_handler_cb cb_func);
>+void		kq_fnm_free(kq_fnm_p kfnm);
>+
>+kq_fnmo_p	kq_fnm_add(kq_fnm_p kfnm,
>+			    const char *path, void *udata);
>+void		kq_fnm_del(kq_fnm_p kfnm, kq_fnmo_p fnmo);
>+
>+
>+#endif /* __KQUEUE_FILE_NOTIFY_MONITOR_H__ */

Actions: View | Diff

Attachments on bug 214338: 176801 | 178938 | 181557 | 181816 | 182423 | 182749 | 184001 | 184287 | 184290 | 184301 | 184456 | 188088 | 188134 | 188144 | 188261 | 188262 | 190711 | 190712 | 191448 | 195450 | 195570 | 195597 | 195598 | 195618 | 195619 | 195649 | 195650 | 198889 | 198900 | 199221 | 199229 | 202832 | 202839 | 202935 | 202948 | 204279 | 204331 | 218219 | 222442