zed additional features

This commit adds two features to zed, that macOS desires. The first
is that when you unload the kernel module, zed would enter into a
cpubusy loop calling zfs_events_next() repeatedly. We now look for
ENODEV, returned by kernel, so zed can exit gracefully.

Second feature is -I (idle) (alas -P persist was taken) is for the
deamon to;

1; if started without ZFS kernel module, stick around waiting for it.
2; if kernel module is unloaded, go back to 1.

This is due to daemons in macOS is started by launchctl, and is
expected to stick around.

Currently, the busy loop only exists when errno is ENODEV. This is
to ensure that functionality that upstream expects is not changed.
It did not care about errors before, and it still does not. (with the
exception of ENODEV).

However, it is probably better that all errors
(ERESTART notwithstanding) exits the loop, and the issues complaining
about zed taking all CPU will go away.

Reviewed-by: Brian Behlendorf <[email protected]>
Signed-off-by: Jorgen Lundman <[email protected]>
Closes #10476 

6 files changed, 62 insertions, 12 deletions

diff --git a/cmd/zed/zed.c b/cmd/zed/zed.c
index bba8b8f64..0784e3834 100644
--- a/cmd/zed/zed.c
+++ b/cmd/zed/zed.c
@@ -263,18 +263,43 @@ main(int argc, char *argv[])
 	if (zed_conf_read_state(zcp, &saved_eid, saved_etime) < 0)
 		exit(EXIT_FAILURE);
 
-	zed_event_init(zcp);
+idle:
+	/*
+	 * If -I is specified, attempt to open /dev/zfs repeatedly until
+	 * successful.
+	 */
+	do {
+		if (!zed_event_init(zcp))
+			break;
+		/* Wait for some time and try again. tunable? */
+		sleep(30);
+	} while (!_got_exit && zcp->do_idle);
+
+	if (_got_exit)
+		goto out;
+
 	zed_event_seek(zcp, saved_eid, saved_etime);
 
 	while (!_got_exit) {
+		int rv;
 		if (_got_hup) {
 			_got_hup = 0;
 			(void) zed_conf_scan_dir(zcp);
 		}
-		zed_event_service(zcp);
+		rv = zed_event_service(zcp);
+
+		/* ENODEV: When kernel module is unloaded (osx) */
+		if (rv == ENODEV)
+			break;
 	}
+
 	zed_log_msg(LOG_NOTICE, "Exiting");
 	zed_event_fini(zcp);
+
+	if (zcp->do_idle && !_got_exit)
+		goto idle;
+
+out:
 	zed_conf_destroy(zcp);
 	zed_log_fini();
 	exit(EXIT_SUCCESS);
diff --git a/cmd/zed/zed_conf.c b/cmd/zed/zed_conf.c
index 86671369c..52370eb87 100644
--- a/cmd/zed/zed_conf.c
+++ b/cmd/zed/zed_conf.c
@@ -153,6 +153,8 @@ _zed_conf_display_help(const char *prog, int got_err)
 	    "Force daemon to run.");
 	fprintf(fp, "%*c%*s %s\n", w1, 0x20, -w2, "-F",
 	    "Run daemon in the foreground.");
+	fprintf(fp, "%*c%*s %s\n", w1, 0x20, -w2, "-I",
+	    "Idle daemon until kernel module is (re)loaded.");
 	fprintf(fp, "%*c%*s %s\n", w1, 0x20, -w2, "-M",
 	    "Lock all pages in memory.");
 	fprintf(fp, "%*c%*s %s\n", w1, 0x20, -w2, "-P",
@@ -249,7 +251,7 @@ _zed_conf_parse_path(char **resultp, const char *path)
 void
 zed_conf_parse_opts(struct zed_conf *zcp, int argc, char **argv)
 {
-	const char * const opts = ":hLVc:d:p:P:s:vfFMZ";
+	const char * const opts = ":hLVc:d:p:P:s:vfFMZI";
 	int opt;
 
 	if (!zcp || !argv || !argv[0])
@@ -274,6 +276,9 @@ zed_conf_parse_opts(struct zed_conf *zcp, int argc, char **argv)
 		case 'd':
 			_zed_conf_parse_path(&zcp->zedlet_dir, optarg);
 			break;
+		case 'I':
+			zcp->do_idle = 1;
+			break;
 		case 'p':
 			_zed_conf_parse_path(&zcp->pid_file, optarg);
 			break;
diff --git a/cmd/zed/zed_conf.h b/cmd/zed/zed_conf.h
index 7d6b63b1d..424cb2c01 100644
--- a/cmd/zed/zed_conf.h
+++ b/cmd/zed/zed_conf.h
@@ -25,6 +25,7 @@ struct zed_conf {
 	unsigned	do_memlock:1;		/* true if locking memory */
 	unsigned	do_verbose:1;		/* true if verbosity enabled */
 	unsigned	do_zero:1;		/* true if zeroing state */
+	unsigned	do_idle:1;		/* true if idle enabled */
 	int		syslog_facility;	/* syslog facility value */
 	int		min_events;		/* RESERVED FOR FUTURE USE */
 	int		max_events;		/* RESERVED FOR FUTURE USE */
diff --git a/cmd/zed/zed_event.c b/cmd/zed/zed_event.c
index 197c25386..1c5d00e29 100644
--- a/cmd/zed/zed_event.c
+++ b/cmd/zed/zed_event.c
@@ -41,25 +41,36 @@
 /*
  * Open the libzfs interface.
  */
-void
+int
 zed_event_init(struct zed_conf *zcp)
 {
 	if (!zcp)
 		zed_log_die("Failed zed_event_init: %s", strerror(EINVAL));
 
 	zcp->zfs_hdl = libzfs_init();
-	if (!zcp->zfs_hdl)
+	if (!zcp->zfs_hdl) {
+		if (zcp->do_idle)
+			return (-1);
 		zed_log_die("Failed to initialize libzfs");
+	}
 
 	zcp->zevent_fd = open(ZFS_DEV, O_RDWR);
-	if (zcp->zevent_fd < 0)
+	if (zcp->zevent_fd < 0) {
+		if (zcp->do_idle)
+			return (-1);
 		zed_log_die("Failed to open \"%s\": %s",
 		    ZFS_DEV, strerror(errno));
+	}
 
 	zfs_agent_init(zcp->zfs_hdl);
 
-	if (zed_disk_event_init() != 0)
+	if (zed_disk_event_init() != 0) {
+		if (zcp->do_idle)
+			return (-1);
 		zed_log_die("Failed to initialize disk events");
+	}
+
+	return (0);
 }
 
 /*
@@ -873,7 +884,7 @@ _zed_event_add_time_strings(uint64_t eid, zed_strings_t *zsp, int64_t etime[])
 /*
  * Service the next zevent, blocking until one is available.
  */
-void
+int
 zed_event_service(struct zed_conf *zcp)
 {
 	nvlist_t *nvl;
@@ -891,13 +902,13 @@ zed_event_service(struct zed_conf *zcp)
 		errno = EINVAL;
 		zed_log_msg(LOG_ERR, "Failed to service zevent: %s",
 		    strerror(errno));
-		return;
+		return (EINVAL);
 	}
 	rv = zpool_events_next(zcp->zfs_hdl, &nvl, &n_dropped, ZEVENT_NONE,
 	    zcp->zevent_fd);
 
 	if ((rv != 0) || !nvl)
-		return;
+		return (errno);
 
 	if (n_dropped > 0) {
 		zed_log_msg(LOG_WARNING, "Missed %d events", n_dropped);
@@ -950,4 +961,5 @@ zed_event_service(struct zed_conf *zcp)
 		zed_strings_destroy(zsp);
 	}
 	nvlist_free(nvl);
+	return (0);
 }
diff --git a/cmd/zed/zed_event.h b/cmd/zed/zed_event.h
index 9f37b80fe..c1455c3a0 100644
--- a/cmd/zed/zed_event.h
+++ b/cmd/zed/zed_event.h
@@ -17,13 +17,13 @@
 
 #include <stdint.h>
 
-void zed_event_init(struct zed_conf *zcp);
+int zed_event_init(struct zed_conf *zcp);
 
 void zed_event_fini(struct zed_conf *zcp);
 
 int zed_event_seek(struct zed_conf *zcp, uint64_t saved_eid,
     int64_t saved_etime[]);
 
-void zed_event_service(struct zed_conf *zcp);
+int zed_event_service(struct zed_conf *zcp);
 
 #endif	/* !ZED_EVENT_H */
diff --git a/man/man8/zed.8.in b/man/man8/zed.8.in
index 2db223a68..2ca393572 100644
--- a/man/man8/zed.8.in
+++ b/man/man8/zed.8.in
@@ -24,6 +24,7 @@ ZED \- ZFS Event Daemon
 [\fB\-f\fR]
 [\fB\-F\fR]
 [\fB\-h\fR]
+[\fB\-I\fR]
 [\fB\-L\fR]
 [\fB\-M\fR]
 [\fB\-p\fR \fIpidfile\fR]
@@ -66,6 +67,12 @@ Lock all current and future pages in the virtual memory address space.
 This may help the daemon remain responsive when the system is under heavy
 memory pressure.
 .TP
+.BI \-I
+Request that the daemon idle rather than exit when the kernel modules are
+not loaded. Processing of events will start, or resume, when the kernel
+modules are (re)loaded. Under Linux the kernel modules cannot be unloaded
+while the daemon is running.
+.TP
 .BI \-Z
 Zero the daemon's state, thereby allowing zevents still within the kernel
 to be reprocessed.