* sysfs - extract md related information from sysfs. Part of:
* mdadm - manage Linux "md" devices aka RAID arrays.
*
- * Copyright (C) 2006 Neil Brown <neilb@suse.de>
+ * Copyright (C) 2006-2009 Neil Brown <neilb@suse.de>
*
*
* This program is free software; you can redistribute it and/or modify
struct mdinfo *sysfs_read(int fd, int devnum, unsigned long options)
{
- /* Longest possible name in sysfs, mounted at /sys, is
- * /sys/block/md_dXXX/md/dev-XXXXX/block/dev
- * /sys/block/md_dXXX/md/metadata_version
- * which is about 41 characters. 50 should do for now
- */
- char fname[50];
- char buf[1024];
+ char fname[PATH_MAX];
+ char buf[PATH_MAX];
char *base;
char *dbase;
struct mdinfo *sra;
if (options & GET_SAFEMODE) {
int scale = 1;
int dot = 0;
- int i;
+ unsigned i;
unsigned long msec;
size_t len;
strcpy(dbase, "block/dev");
if (load_sys(fname, buf)) {
+ /* assume this is a stale reference to a hot
+ * removed device
+ */
free(dev);
- if (options & SKIP_GONE_DEVS)
- continue;
- else
- goto abort;
+ continue;
}
sscanf(buf, "%d:%d", &dev->disk.major, &dev->disk.minor);
/* special case check for block devices that can go 'offline' */
- if (options & SKIP_GONE_DEVS) {
- strcpy(dbase, "block/device/state");
- if (load_sys(fname, buf) == 0 &&
- strncmp(buf, "offline", 7) == 0) {
- free(dev);
- continue;
- }
+ strcpy(dbase, "block/device/state");
+ if (load_sys(fname, buf) == 0 &&
+ strncmp(buf, "offline", 7) == 0) {
+ free(dev);
+ continue;
}
/* finally add this disk to the array */
char fname[50];
int n;
if (fstat(fd, &stb)) return 0;
- if (major(stb.st_rdev) != get_mdp_major())
+ if (major(stb.st_rdev) != (unsigned)get_mdp_major())
sprintf(fname, "/sys/block/md%d/md/component_size",
(int)minor(stb.st_rdev));
else
char *name, char *val)
{
char fname[50];
- int n;
+ unsigned int n;
int fd;
sprintf(fname, "/sys/block/%s/md/%s/%s",
return 0;
}
-int sysfs_get_ll(struct mdinfo *sra, struct mdinfo *dev,
- char *name, unsigned long long *val)
+int sysfs_attribute_available(struct mdinfo *sra, struct mdinfo *dev, char *name)
+{
+ char fname[50];
+ struct stat st;
+
+ sprintf(fname, "/sys/block/%s/md/%s/%s",
+ sra->sys_name, dev?dev->sys_name:"", name);
+
+ return stat(fname, &st) == 0;
+}
+
+int sysfs_get_fd(struct mdinfo *sra, struct mdinfo *dev,
+ char *name)
{
char fname[50];
- char buf[50];
- int n;
int fd;
- char *ep;
+
sprintf(fname, "/sys/block/%s/md/%s/%s",
sra->sys_name, dev?dev->sys_name:"", name);
- fd = open(fname, O_RDONLY);
+ fd = open(fname, O_RDWR);
if (fd < 0)
- return -1;
+ fd = open(fname, O_RDONLY);
+ return fd;
+}
+
+int sysfs_fd_get_ll(int fd, unsigned long long *val)
+{
+ char buf[50];
+ int n;
+ char *ep;
+
+ lseek(fd, 0, 0);
n = read(fd, buf, sizeof(buf));
- close(fd);
if (n <= 0)
return -1;
buf[n] = 0;
return 0;
}
-int sysfs_get_str(struct mdinfo *sra, struct mdinfo *dev,
- char *name, char *val, int size)
+int sysfs_get_ll(struct mdinfo *sra, struct mdinfo *dev,
+ char *name, unsigned long long *val)
{
- char fname[50];
int n;
int fd;
- sprintf(fname, "/sys/block/%s/md/%s/%s",
- sra->sys_name, dev?dev->sys_name:"", name);
- fd = open(fname, O_RDONLY);
+
+ fd = sysfs_get_fd(sra, dev, name);
if (fd < 0)
return -1;
- n = read(fd, val, size);
+ n = sysfs_fd_get_ll(fd, val);
close(fd);
+ return n;
+}
+
+int sysfs_fd_get_str(int fd, char *val, int size)
+{
+ int n;
+
+ lseek(fd, 0, 0);
+ n = read(fd, val, size);
if (n <= 0)
return -1;
val[n] = 0;
return n;
}
+int sysfs_get_str(struct mdinfo *sra, struct mdinfo *dev,
+ char *name, char *val, int size)
+{
+ int n;
+ int fd;
+
+ fd = sysfs_get_fd(sra, dev, name);
+ if (fd < 0)
+ return -1;
+ n = sysfs_fd_get_str(fd, val, size);
+ close(fd);
+ return n;
+}
+
int sysfs_set_safemode(struct mdinfo *sra, unsigned long ms)
{
unsigned long sec;
return rv;
}
-int sysfs_add_disk(struct mdinfo *sra, struct mdinfo *sd, int in_sync)
+int sysfs_add_disk(struct mdinfo *sra, struct mdinfo *sd, int resume)
{
- char dv[100];
- char nm[100];
+ char dv[PATH_MAX];
+ char nm[PATH_MAX];
char *dname;
int rv;
strcpy(sd->sys_name, "dev-");
strcpy(sd->sys_name+4, dname);
+ /* test write to see if 'recovery_start' is available */
+ if (resume && sd->recovery_start < MaxSector &&
+ sysfs_set_num(sra, sd, "recovery_start", 0)) {
+ sysfs_set_str(sra, sd, "state", "remove");
+ return -1;
+ }
+
rv = sysfs_set_num(sra, sd, "offset", sd->data_offset);
rv |= sysfs_set_num(sra, sd, "size", (sd->component_size+1) / 2);
if (sra->array.level != LEVEL_CONTAINER) {
- if (in_sync)
+ if (sd->recovery_start == MaxSector)
/* This can correctly fail if array isn't started,
* yet, so just ignore status for now.
*/
- sysfs_set_str(sra, sd, "state", "in_sync");
+ sysfs_set_str(sra, sd, "state", "insync");
rv |= sysfs_set_num(sra, sd, "slot", sd->disk.raid_disk);
+ if (resume)
+ sysfs_set_num(sra, sd, "recovery_start", sd->recovery_start);
}
return rv;
}
else
return found;
}
+
+int sysfs_freeze_array(struct mdinfo *sra)
+{
+ /* Try to freeze resync/rebuild on this array/container.
+ * Return -1 if the array is busy,
+ * return -2 container cannot be frozen,
+ * return 0 if this kernel doesn't support 'frozen'
+ * return 1 if it worked.
+ */
+ char buf[20];
+
+ if (!sysfs_attribute_available(sra, NULL, "sync_action"))
+ return 1; /* no sync_action == frozen */
+ if (sysfs_get_str(sra, NULL, "sync_action", buf, 20) <= 0)
+ return 0;
+ if (strcmp(buf, "idle\n") != 0 &&
+ strcmp(buf, "frozen\n") != 0)
+ return -1;
+ if (sysfs_set_str(sra, NULL, "sync_action", "frozen") < 0)
+ return 0;
+ return 1;
+}
+
+#ifndef MDASSEMBLE
+
+static char *clean_states[] = {
+ "clear", "inactive", "readonly", "read-auto", "clean", NULL };
+
+int WaitClean(char *dev, int sock, int verbose)
+{
+ int fd;
+ struct mdinfo *mdi;
+ int rv = 1;
+ int devnum;
+
+ fd = open(dev, O_RDONLY);
+ if (fd < 0) {
+ if (verbose)
+ fprintf(stderr, Name ": Couldn't open %s: %s\n", dev, strerror(errno));
+ return 1;
+ }
+
+ devnum = fd2devnum(fd);
+ mdi = sysfs_read(fd, devnum, GET_VERSION|GET_LEVEL|GET_SAFEMODE);
+ if (!mdi) {
+ if (verbose)
+ fprintf(stderr, Name ": Failed to read sysfs attributes for "
+ "%s\n", dev);
+ close(fd);
+ return 0;
+ }
+
+ switch(mdi->array.level) {
+ case LEVEL_LINEAR:
+ case LEVEL_MULTIPATH:
+ case 0:
+ /* safemode delay is irrelevant for these levels */
+ rv = 0;
+
+ }
+
+ /* for internal metadata the kernel handles the final clean
+ * transition, containers can never be dirty
+ */
+ if (!is_subarray(mdi->text_version))
+ rv = 0;
+
+ /* safemode disabled ? */
+ if (mdi->safe_mode_delay == 0)
+ rv = 0;
+
+ if (rv) {
+ int state_fd = sysfs_open(fd2devnum(fd), NULL, "array_state");
+ char buf[20];
+ fd_set fds;
+ struct timeval tm;
+
+ /* minimize the safe_mode_delay and prepare to wait up to 5s
+ * for writes to quiesce
+ */
+ sysfs_set_safemode(mdi, 1);
+ tm.tv_sec = 5;
+ tm.tv_usec = 0;
+
+ FD_ZERO(&fds);
+
+ /* wait for array_state to be clean */
+ while (1) {
+ rv = read(state_fd, buf, sizeof(buf));
+ if (rv < 0)
+ break;
+ if (sysfs_match_word(buf, clean_states) <= 4)
+ break;
+ FD_SET(state_fd, &fds);
+ rv = select(state_fd + 1, NULL, NULL, &fds, &tm);
+ if (rv < 0 && errno != EINTR)
+ break;
+ lseek(state_fd, 0, SEEK_SET);
+ }
+ if (rv < 0)
+ rv = 1;
+ else if (fping_monitor(sock) == 0 ||
+ ping_monitor(mdi->text_version) == 0) {
+ /* we need to ping to close the window between array
+ * state transitioning to clean and the metadata being
+ * marked clean
+ */
+ rv = 0;
+ } else
+ rv = 1;
+ if (rv && verbose)
+ fprintf(stderr, Name ": Error waiting for %s to be clean\n",
+ dev);
+
+ /* restore the original safe_mode_delay */
+ sysfs_set_safemode(mdi, mdi->safe_mode_delay);
+ close(state_fd);
+ }
+
+ sysfs_free(mdi);
+ close(fd);
+
+ return rv;
+}
+#endif /* MDASSEMBLE */