X-Git-Url: http://git.ipfire.org/?p=thirdparty%2Fmdadm.git;a=blobdiff_plain;f=sysfs.c;h=c3bbbe32b1fca1ff0a4a6a02a339aafa20cd3632;hp=58c84e26339e8014027bcea6d722bf87b73b4030;hb=2904b26f059c5d82d9d631c9987e92e3f9af498c;hpb=492350045c7e40741069caa7d017209439db665b

diff --git a/sysfs.c b/sysfs.c
index 58c84e26..c3bbbe32 100644
--- a/sysfs.c
+++ b/sysfs.c
@@ -2,7 +2,7 @@
  * 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
@@ -81,27 +81,23 @@ int sysfs_open(int devnum, char *devname, char *attr)
 
 void sysfs_init(struct mdinfo *mdi, int fd, int devnum)
 {
+	mdi->sys_name[0] = 0;
 	if (fd >= 0) {
-		struct stat stb;
 		mdu_version_t vers;
- 		if (fstat(fd, &stb))
-			return;
 		if (ioctl(fd, RAID_VERSION, &vers) != 0)
 			return;
-		if (major(stb.st_rdev)==9)
-			sprintf(mdi->sys_name, "md%d", (int)minor(stb.st_rdev));
-		else
-			sprintf(mdi->sys_name, "md_d%d",
-				(int)minor(stb.st_rdev)>>MdpMinorShift);
-	} else {
-		if (devnum >= 0)
-			sprintf(mdi->sys_name, "md%d", devnum);
-		else
-			sprintf(mdi->sys_name, "md_d%d",
-				-1-devnum);
+		devnum = fd2devnum(fd);
 	}
+	if (devnum == NoMdDev)
+		return;
+	if (devnum >= 0)
+		sprintf(mdi->sys_name, "md%d", devnum);
+	else
+		sprintf(mdi->sys_name, "md_d%d",
+			-1-devnum);
 }
 
+
 struct mdinfo *sysfs_read(int fd, int devnum, unsigned long options)
 {
 	/* Longest possible name in sysfs, mounted at /sys, is
@@ -123,6 +119,10 @@ struct mdinfo *sysfs_read(int fd, int devnum, unsigned long options)
 		return sra;
 	memset(sra, 0, sizeof(*sra));
 	sysfs_init(sra, fd, devnum);
+	if (sra->sys_name[0] == 0) {
+		free(sra);
+		return NULL;
+	}
 
 	sprintf(fname, "/sys/block/%s/md/", sra->sys_name);
 	base = fname + strlen(fname);
@@ -272,18 +272,34 @@ struct mdinfo *sysfs_read(int fd, int devnum, unsigned long options)
 			}
 			
 		}
-		dev->next = sra->devs;
-		sra->devs = dev;
-
 		strcpy(dev->sys_name, de->d_name);
 		dev->disk.raid_disk = strtoul(buf, &ep, 10);
 		if (*ep) dev->disk.raid_disk = -1;
 
 		strcpy(dbase, "block/dev");
-		if (load_sys(fname, buf))
-			goto abort;
+		if (load_sys(fname, buf)) {
+			free(dev);
+			if (options & SKIP_GONE_DEVS)
+				continue;
+			else
+				goto abort;
+		}
 		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;
+			}
+		}
+
+		/* finally add this disk to the array */
+		dev->next = sra->devs;
+		sra->devs = dev;
+
 		if (options & GET_OFFSET) {
 			strcpy(dbase, "offset");
 			if (load_sys(fname, buf))
@@ -363,7 +379,7 @@ unsigned long long get_component_size(int fd)
 	char fname[50];
 	int n;
 	if (fstat(fd, &stb)) return 0;
-	if (major(stb.st_rdev) == 9)
+	if (major(stb.st_rdev) != get_mdp_major())
 		sprintf(fname, "/sys/block/md%d/md/component_size",
 			(int)minor(stb.st_rdev));
 	else
@@ -410,21 +426,44 @@ int sysfs_set_num(struct mdinfo *sra, struct mdinfo *dev,
 	return sysfs_set_str(sra, dev, name, valstr);
 }
 
-int sysfs_get_ll(struct mdinfo *sra, struct mdinfo *dev,
-		       char *name, unsigned long long *val)
+int sysfs_uevent(struct mdinfo *sra, char *event)
 {
 	char fname[50];
-	char buf[50];
 	int n;
 	int fd;
-	char *ep;
+
+	sprintf(fname, "/sys/block/%s/uevent",
+		sra->sys_name);
+	fd = open(fname, O_WRONLY);
+	if (fd < 0)
+		return -1;
+	n = write(fd, event, strlen(event));
+	close(fd);
+	return 0;
+}	
+
+int sysfs_get_fd(struct mdinfo *sra, struct mdinfo *dev,
+		       char *name)
+{
+	char fname[50];
+	int fd;
+
 	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;
@@ -434,6 +473,46 @@ int sysfs_get_ll(struct mdinfo *sra, struct mdinfo *dev,
 	return 0;
 }
 
+int sysfs_get_ll(struct mdinfo *sra, struct mdinfo *dev,
+		       char *name, unsigned long long *val)
+{
+	int n;
+	int fd;
+
+	fd = sysfs_get_fd(sra, dev, name);
+	if (fd < 0)
+		return -1;
+	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;
@@ -474,12 +553,26 @@ int sysfs_set_array(struct mdinfo *info, int vers)
 	rv |= sysfs_set_num(info, NULL, "chunk_size", info->array.chunk_size);
 	rv |= sysfs_set_num(info, NULL, "layout", info->array.layout);
 	rv |= sysfs_set_num(info, NULL, "component_size", info->component_size/2);
+	if (info->custom_array_size) {
+		int rc;
+
+		rc = sysfs_set_num(info, NULL, "array_size",
+				   info->custom_array_size/2);
+		if (rc && errno == ENOENT) {
+			fprintf(stderr, Name ": This kernel does not "
+				"have the md/array_size attribute, "
+				"the array may be larger than expected\n");
+			rc = 0;
+		}
+		rv |= rc;
+	}
+
 	if (info->array.level > 0)
 		rv |= sysfs_set_num(info, NULL, "resync_start", info->resync_start);
 	return rv;
 }
 
-int sysfs_add_disk(struct mdinfo *sra, struct mdinfo *sd)
+int sysfs_add_disk(struct mdinfo *sra, struct mdinfo *sd, int resume)
 {
 	char dv[100];
 	char nm[100];
@@ -502,11 +595,24 @@ int sysfs_add_disk(struct mdinfo *sra, struct mdinfo *sd)
 	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 (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", "insync");
 		rv |= sysfs_set_num(sra, sd, "slot", sd->disk.raid_disk);
-//		rv |= sysfs_set_str(sra, sd, "state", "in_sync");
+		if (resume)
+			sysfs_set_num(sra, sd, "recovery_start", sd->recovery_start);
 	}
 	return rv;
 }
@@ -689,3 +795,109 @@ int sysfs_unique_holder(int devnum, long rdev)
 	else
 		return found;
 }
+
+#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;
+
+		/* give mdmon a chance to checkpoint resync */
+		sysfs_set_str(mdi, NULL, "sync_action", "idle");
+
+		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 */