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 <behlendorf1@llnl.gov>
Signed-off-by: Jorgen Lundman <lundman@lundman.net>
Closes #10476
This commit is contained in:
Jorgen Lundman 2020-06-23 01:53:34 +09:00 committed by GitHub
parent 42d8d1d66a
commit 68301ba20e
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
6 changed files with 62 additions and 12 deletions

View File

@ -263,18 +263,43 @@ main(int argc, char *argv[])
if (zed_conf_read_state(zcp, &saved_eid, saved_etime) < 0) if (zed_conf_read_state(zcp, &saved_eid, saved_etime) < 0)
exit(EXIT_FAILURE); 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); zed_event_seek(zcp, saved_eid, saved_etime);
while (!_got_exit) { while (!_got_exit) {
int rv;
if (_got_hup) { if (_got_hup) {
_got_hup = 0; _got_hup = 0;
(void) zed_conf_scan_dir(zcp); (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_log_msg(LOG_NOTICE, "Exiting");
zed_event_fini(zcp); zed_event_fini(zcp);
if (zcp->do_idle && !_got_exit)
goto idle;
out:
zed_conf_destroy(zcp); zed_conf_destroy(zcp);
zed_log_fini(); zed_log_fini();
exit(EXIT_SUCCESS); exit(EXIT_SUCCESS);

View File

@ -153,6 +153,8 @@ _zed_conf_display_help(const char *prog, int got_err)
"Force daemon to run."); "Force daemon to run.");
fprintf(fp, "%*c%*s %s\n", w1, 0x20, -w2, "-F", fprintf(fp, "%*c%*s %s\n", w1, 0x20, -w2, "-F",
"Run daemon in the foreground."); "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", fprintf(fp, "%*c%*s %s\n", w1, 0x20, -w2, "-M",
"Lock all pages in memory."); "Lock all pages in memory.");
fprintf(fp, "%*c%*s %s\n", w1, 0x20, -w2, "-P", fprintf(fp, "%*c%*s %s\n", w1, 0x20, -w2, "-P",
@ -249,7 +251,7 @@ _zed_conf_parse_path(char **resultp, const char *path)
void void
zed_conf_parse_opts(struct zed_conf *zcp, int argc, char **argv) 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; int opt;
if (!zcp || !argv || !argv[0]) if (!zcp || !argv || !argv[0])
@ -274,6 +276,9 @@ zed_conf_parse_opts(struct zed_conf *zcp, int argc, char **argv)
case 'd': case 'd':
_zed_conf_parse_path(&zcp->zedlet_dir, optarg); _zed_conf_parse_path(&zcp->zedlet_dir, optarg);
break; break;
case 'I':
zcp->do_idle = 1;
break;
case 'p': case 'p':
_zed_conf_parse_path(&zcp->pid_file, optarg); _zed_conf_parse_path(&zcp->pid_file, optarg);
break; break;

View File

@ -25,6 +25,7 @@ struct zed_conf {
unsigned do_memlock:1; /* true if locking memory */ unsigned do_memlock:1; /* true if locking memory */
unsigned do_verbose:1; /* true if verbosity enabled */ unsigned do_verbose:1; /* true if verbosity enabled */
unsigned do_zero:1; /* true if zeroing state */ unsigned do_zero:1; /* true if zeroing state */
unsigned do_idle:1; /* true if idle enabled */
int syslog_facility; /* syslog facility value */ int syslog_facility; /* syslog facility value */
int min_events; /* RESERVED FOR FUTURE USE */ int min_events; /* RESERVED FOR FUTURE USE */
int max_events; /* RESERVED FOR FUTURE USE */ int max_events; /* RESERVED FOR FUTURE USE */

View File

@ -41,25 +41,36 @@
/* /*
* Open the libzfs interface. * Open the libzfs interface.
*/ */
void int
zed_event_init(struct zed_conf *zcp) zed_event_init(struct zed_conf *zcp)
{ {
if (!zcp) if (!zcp)
zed_log_die("Failed zed_event_init: %s", strerror(EINVAL)); zed_log_die("Failed zed_event_init: %s", strerror(EINVAL));
zcp->zfs_hdl = libzfs_init(); 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"); zed_log_die("Failed to initialize libzfs");
}
zcp->zevent_fd = open(ZFS_DEV, O_RDWR); 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", zed_log_die("Failed to open \"%s\": %s",
ZFS_DEV, strerror(errno)); ZFS_DEV, strerror(errno));
}
zfs_agent_init(zcp->zfs_hdl); 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"); 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. * Service the next zevent, blocking until one is available.
*/ */
void int
zed_event_service(struct zed_conf *zcp) zed_event_service(struct zed_conf *zcp)
{ {
nvlist_t *nvl; nvlist_t *nvl;
@ -891,13 +902,13 @@ zed_event_service(struct zed_conf *zcp)
errno = EINVAL; errno = EINVAL;
zed_log_msg(LOG_ERR, "Failed to service zevent: %s", zed_log_msg(LOG_ERR, "Failed to service zevent: %s",
strerror(errno)); strerror(errno));
return; return (EINVAL);
} }
rv = zpool_events_next(zcp->zfs_hdl, &nvl, &n_dropped, ZEVENT_NONE, rv = zpool_events_next(zcp->zfs_hdl, &nvl, &n_dropped, ZEVENT_NONE,
zcp->zevent_fd); zcp->zevent_fd);
if ((rv != 0) || !nvl) if ((rv != 0) || !nvl)
return; return (errno);
if (n_dropped > 0) { if (n_dropped > 0) {
zed_log_msg(LOG_WARNING, "Missed %d events", n_dropped); 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); zed_strings_destroy(zsp);
} }
nvlist_free(nvl); nvlist_free(nvl);
return (0);
} }

View File

@ -17,13 +17,13 @@
#include <stdint.h> #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); void zed_event_fini(struct zed_conf *zcp);
int zed_event_seek(struct zed_conf *zcp, uint64_t saved_eid, int zed_event_seek(struct zed_conf *zcp, uint64_t saved_eid,
int64_t saved_etime[]); int64_t saved_etime[]);
void zed_event_service(struct zed_conf *zcp); int zed_event_service(struct zed_conf *zcp);
#endif /* !ZED_EVENT_H */ #endif /* !ZED_EVENT_H */

View File

@ -24,6 +24,7 @@ ZED \- ZFS Event Daemon
[\fB\-f\fR] [\fB\-f\fR]
[\fB\-F\fR] [\fB\-F\fR]
[\fB\-h\fR] [\fB\-h\fR]
[\fB\-I\fR]
[\fB\-L\fR] [\fB\-L\fR]
[\fB\-M\fR] [\fB\-M\fR]
[\fB\-p\fR \fIpidfile\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 This may help the daemon remain responsive when the system is under heavy
memory pressure. memory pressure.
.TP .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 .BI \-Z
Zero the daemon's state, thereby allowing zevents still within the kernel Zero the daemon's state, thereby allowing zevents still within the kernel
to be reprocessed. to be reprocessed.