/*
* mdadm - manage Linux "md" devices aka RAID arrays.
*
- * Copyright (C) 2001-2006 Neil Brown <neilb@suse.de>
+ * Copyright (C) 2001-2009 Neil Brown <neilb@suse.de>
*
*
* This program is free software; you can redistribute it and/or modify
* Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
*
* Author: Neil Brown
- * Email: <neilb@cse.unsw.edu.au>
- * Paper: Neil Brown
- * School of Computer Science and Engineering
- * The University of New South Wales
- * Sydney, 2052
- * Australia
+ * Email: <neilb@suse.de>
*/
#include "mdadm.h"
__u64 utime; /* 40 bits second, 24 btes microseconds */
__u64 events; /* incremented when superblock updated */
__u64 resync_offset; /* data before this offset (from data_offset) known to be in sync */
- __u32 sb_csum; /* checksum upto devs[max_dev] */
- __u32 max_dev; /* size of devs[] array to consider */
+ __u32 sb_csum; /* checksum upto dev_roles[max_dev] */
+ __u32 max_dev; /* size of dev_roles[] array to consider */
__u8 pad3[64-32]; /* set to 0 when writing */
/* device state information. Indexed by dev_number.
return __cpu_to_le32(csum);
}
+static char abuf[4096+4096];
+static int aread(int fd, void *buf, int len)
+{
+ /* aligned read.
+ * On devices with a 4K sector size, we need to read
+ * the full sector and copy relevant bits into
+ * the buffer
+ */
+ int bsize;
+ char *b;
+ int n;
+ if (ioctl(fd, BLKSSZGET, &bsize) != 0 ||
+ bsize <= len)
+ return read(fd, buf, len);
+ if (bsize > 4096)
+ return -1;
+ b = (char*)(((long)(abuf+4096))&~4095UL);
+
+ n = read(fd, b, bsize);
+ if (n <= 0)
+ return n;
+ lseek(fd, len - n, 1);
+ if (n > len)
+ n = len;
+ memcpy(buf, b, n);
+ return n;
+}
+
+static int awrite(int fd, void *buf, int len)
+{
+ /* aligned write.
+ * On devices with a 4K sector size, we need to write
+ * the full sector. We pre-read if the sector is larger
+ * than the write.
+ * The address must be sector-aligned.
+ */
+ int bsize;
+ char *b;
+ int n;
+ if (ioctl(fd, BLKSSZGET, &bsize) != 0 ||
+ bsize <= len)
+ return write(fd, buf, len);
+ if (bsize > 4096)
+ return -1;
+ b = (char*)(((long)(abuf+4096))&~4095UL);
+
+ n = read(fd, b, bsize);
+ if (n <= 0)
+ return n;
+ lseek(fd, -n, 1);
+ memcpy(b, buf, len);
+ n = write(fd, b, bsize);
+ if (n <= 0)
+ return n;
+ lseek(fd, len - n, 1);
+ return len;
+}
+
#ifndef MDASSEMBLE
static void examine_super1(struct supertype *st, char *homehost)
{
struct mdp_superblock_1 *sb = st->sb;
time_t atime;
int d;
- int faulty;
+ int role;
+ int delta_extra = 0;
int i;
char *c;
int l = homehost ? strlen(homehost) : 0;
human_size(__le64_to_cpu(sb->reshape_position)<<9));
if (__le32_to_cpu(sb->delta_disks)) {
printf(" Delta Devices : %d", __le32_to_cpu(sb->delta_disks));
- if (__le32_to_cpu(sb->delta_disks))
- printf(" (%d->%d)\n",
- __le32_to_cpu(sb->raid_disks)-__le32_to_cpu(sb->delta_disks),
- __le32_to_cpu(sb->raid_disks));
- else
- printf(" (%d->%d)\n", __le32_to_cpu(sb->raid_disks),
- __le32_to_cpu(sb->raid_disks)+__le32_to_cpu(sb->delta_disks));
+ printf(" (%d->%d)\n",
+ __le32_to_cpu(sb->raid_disks)-__le32_to_cpu(sb->delta_disks),
+ __le32_to_cpu(sb->raid_disks));
+ if ((int)__le32_to_cpu(sb->delta_disks) < 0)
+ delta_extra = -__le32_to_cpu(sb->delta_disks);
}
if (__le32_to_cpu(sb->new_level) != __le32_to_cpu(sb->level)) {
c = map_num(pers, __le32_to_cpu(sb->new_level));
c = map_num(r5layout, __le32_to_cpu(sb->new_layout));
printf(" New Layout : %s\n", c?c:"-unknown-");
}
+ if (__le32_to_cpu(sb->level) == 6) {
+ c = map_num(r6layout, __le32_to_cpu(sb->new_layout));
+ printf(" New Layout : %s\n", c?c:"-unknown-");
+ }
if (__le32_to_cpu(sb->level) == 10) {
printf(" New Layout :");
print_r10_layout(__le32_to_cpu(sb->new_layout));
c = map_num(r5layout, __le32_to_cpu(sb->layout));
printf(" Layout : %s\n", c?c:"-unknown-");
}
+ if (__le32_to_cpu(sb->level) == 6) {
+ c = map_num(r6layout, __le32_to_cpu(sb->layout));
+ printf(" Layout : %s\n", c?c:"-unknown-");
+ }
if (__le32_to_cpu(sb->level) == 10) {
int lo = __le32_to_cpu(sb->layout);
printf(" Layout :");
default: break;
}
printf("\n");
+#if 0
+ /* This turns out to just be confusing */
printf(" Array Slot : %d (", __le32_to_cpu(sb->dev_number));
for (i= __le32_to_cpu(sb->max_dev); i> 0 ; i--)
if (__le16_to_cpu(sb->dev_roles[i-1]) != 0xffff)
else printf("%d", role);
}
printf(")\n");
+#endif
+ printf(" Device Role : ");
+ d = __le32_to_cpu(sb->dev_number);
+ if (d < __le32_to_cpu(sb->max_dev))
+ role = __le16_to_cpu(sb->dev_roles[d]);
+ else
+ role = 0xFFFF;
+ if (role >= 0xFFFE)
+ printf("spare\n");
+ else
+ printf("Active device %d\n", role);
+
printf(" Array State : ");
- for (d=0; d<__le32_to_cpu(sb->raid_disks); d++) {
+ for (d=0; d<__le32_to_cpu(sb->raid_disks) + delta_extra; d++) {
int cnt = 0;
int me = 0;
int i;
}
}
if (cnt > 1) printf("?");
- else if (cnt == 1 && me) printf("U");
- else if (cnt == 1) printf("u");
- else printf ("_");
+ else if (cnt == 1) printf("A");
+ else printf (".");
}
+#if 0
+ /* This is confusing too */
faulty = 0;
for (i=0; i< __le32_to_cpu(sb->max_dev); i++) {
int role = __le16_to_cpu(sb->dev_roles[i]);
faulty++;
}
if (faulty) printf(" %d failed", faulty);
+#endif
+ printf(" ('A' == active, '.' == missing)");
printf("\n");
}
-static void brief_examine_super1(struct supertype *st)
+static void brief_examine_super1(struct supertype *st, int verbose)
{
struct mdp_superblock_1 *sb = st->sb;
int i;
else if (sb->set_name[0])
nm = sb->set_name;
else
- nm = "??";
+ nm = NULL;
- printf("ARRAY /dev/md/%s level=%s ", nm, c?c:"-unknown-");
+ printf("ARRAY%s%s", nm ? " /dev/md/":"", nm);
+ if (verbose && c)
+ printf(" level=%s", c);
sb_offset = __le64_to_cpu(sb->super_offset);
if (sb_offset <= 4)
- printf("metadata=1.1 ");
+ printf(" metadata=1.1 ");
else if (sb_offset <= 8)
- printf("metadata=1.2 ");
+ printf(" metadata=1.2 ");
else
- printf("metadata=1.0 ");
- printf("num-devices=%d UUID=", __le32_to_cpu(sb->raid_disks));
+ printf(" metadata=1.0 ");
+ if (verbose)
+ printf("num-devices=%d ", __le32_to_cpu(sb->raid_disks));
+ printf("UUID=");
for (i=0; i<16; i++) {
if ((i&3)==0 && i != 0) printf(":");
printf("%02x", sb->set_uuid[i]);
info->disk.raid_disk = -1;
switch(role) {
case 0xFFFF:
- info->disk.state = 2; /* spare: ACTIVE, not sync, not faulty */
+ info->disk.state = 0; /* spare: not active, not sync, not faulty */
break;
case 0xFFFE:
info->disk.state = 1; /* faulty */
strncpy(info->name, sb->set_name, 32);
info->name[32] = 0;
+ if (sb->feature_map & __le32_to_cpu(MD_FEATURE_RECOVERY_OFFSET))
+ info->recovery_start = __le32_to_cpu(sb->recovery_offset);
+ else
+ info->recovery_start = MaxSector;
+
if (sb->feature_map & __le32_to_cpu(MD_FEATURE_RESHAPE_ACTIVE)) {
info->reshape_active = 1;
info->reshape_progress = __le64_to_cpu(sb->reshape_position);
info->delta_disks = __le32_to_cpu(sb->delta_disks);
info->new_layout = __le32_to_cpu(sb->new_layout);
info->new_chunk = __le32_to_cpu(sb->new_chunk)<<9;
+ if (info->delta_disks < 0)
+ info->array.raid_disks -= info->delta_disks;
} else
info->reshape_active = 0;
switch(__le32_to_cpu(sb->level)) {
case 5: case 4: case 6:
/* need to force clean */
- if (sb->resync_offset != ~0ULL)
+ if (sb->resync_offset != MaxSector)
rv = 1;
- sb->resync_offset = ~0ULL;
+ sb->resync_offset = MaxSector;
}
}
if (strcmp(update, "assemble")==0) {
if ((rfd = open("/dev/urandom", O_RDONLY)) < 0 ||
read(rfd, sb->device_uuid, 16) != 16) {
- *(__u32*)(sb->device_uuid) = random();
- *(__u32*)(sb->device_uuid+4) = random();
- *(__u32*)(sb->device_uuid+8) = random();
- *(__u32*)(sb->device_uuid+12) = random();
+ __u32 r[4] = {random(), random(), random(), random()};
+ memcpy(sb->device_uuid, r, 16);
}
+ if (rfd >= 0)
+ close(rfd);
sb->dev_roles[i] =
__cpu_to_le16(info->disk.raid_disk);
else {
if ((rfd = open("/dev/urandom", O_RDONLY)) < 0 ||
read(rfd, sb->set_uuid, 16) != 16) {
- *(__u32*)(sb->set_uuid) = random();
- *(__u32*)(sb->set_uuid+4) = random();
- *(__u32*)(sb->set_uuid+8) = random();
- *(__u32*)(sb->set_uuid+12) = random();
+ __u32 r[4] = {random(), random(), random(), random()};
+ memcpy(sb->set_uuid, r, 16);
}
if (rfd >= 0) close(rfd);
}
sb->utime = sb->ctime;
sb->events = __cpu_to_le64(1);
if (info->state & (1<<MD_SB_CLEAN))
- sb->resync_offset = ~0ULL;
+ sb->resync_offset = MaxSector;
else
sb->resync_offset = 0;
sb->max_dev = __cpu_to_le32((1024- sizeof(struct mdp_superblock_1))/
};
#ifndef MDASSEMBLE
/* Add a device to the superblock being created */
-static void add_to_super1(struct supertype *st, mdu_disk_info_t *dk,
+static int add_to_super1(struct supertype *st, mdu_disk_info_t *dk,
int fd, char *devname)
{
struct mdp_superblock_1 *sb = st->sb;
else
*rp = 0xfffe;
+ if (dk->number >= __le32_to_cpu(sb->max_dev) &&
+ __le32_to_cpu(sb->max_dev) < 384)
+ sb->max_dev = __cpu_to_le32(dk->number+1);
+
sb->dev_number = __cpu_to_le32(dk->number);
sb->sb_csum = calc_sb_1_csum(sb);
di->disk = *dk;
di->next = NULL;
*dip = di;
+
+ return 0;
}
#endif
sbsize = sizeof(*sb) + 2 * __le32_to_cpu(sb->max_dev);
sbsize = (sbsize+511)&(~511UL);
- if (write(fd, sb, sbsize) != sbsize)
+ if (awrite(fd, sb, sbsize) != sbsize)
return 4;
if (sb->feature_map & __cpu_to_le32(MD_FEATURE_BITMAP_OFFSET)) {
(((char*)sb)+1024);
if (__le32_to_cpu(bm->magic) == BITMAP_MAGIC) {
locate_bitmap1(st, fd);
- if (write(fd, bm, ROUND_UP(sizeof(*bm),512)) !=
- ROUND_UP(sizeof(*bm),512))
+ if (awrite(fd, bm, sizeof(*bm)) !=
+ sizeof(*bm))
return 5;
}
}
{
/* if the device is bigger than 8Gig, save 64k for bitmap usage,
* if bigger than 200Gig, save 128k
+ * NOTE: result must be multiple of 4K else bad things happen
+ * on 4K-sector devices.
*/
if (devsize < 64*2) return 0;
if (devsize - 64*2 >= 200*1024*1024*2)
int rfd;
int rv = 0;
int bm_space;
+ unsigned long long reserved;
struct devinfo *di;
unsigned long long dsize, array_size;
long long sb_offset;
if (di->fd < 0)
continue;
- Kill(di->devname, 0, 1, 1);
- Kill(di->devname, 0, 1, 1);
+ while (Kill(di->devname, NULL, 0, 1, 1) == 0)
+ ;
sb->dev_number = __cpu_to_le32(di->disk.number);
if (di->disk.state & (1<<MD_DISK_WRITEMOSTLY))
if ((rfd = open("/dev/urandom", O_RDONLY)) < 0 ||
read(rfd, sb->device_uuid, 16) != 16) {
- *(__u32*)(sb->device_uuid) = random();
- *(__u32*)(sb->device_uuid+4) = random();
- *(__u32*)(sb->device_uuid+8) = random();
- *(__u32*)(sb->device_uuid+12) = random();
+ __u32 r[4] = {random(), random(), random(), random()};
+ memcpy(sb->device_uuid, r, 16);
}
- if (rfd >= 0) close(rfd);
+ if (rfd >= 0)
+ close(rfd);
+
sb->events = 0;
refst =*st;
sb_offset &= ~(4*2-1);
sb->super_offset = __cpu_to_le64(sb_offset);
sb->data_offset = __cpu_to_le64(0);
- if (sb_offset - bm_space < array_size)
- bm_space = sb_offset - array_size;
+ if (sb_offset - bm_space < array_size)
+ bm_space = sb_offset - array_size;
sb->data_size = __cpu_to_le64(sb_offset - bm_space);
break;
case 1:
sb->super_offset = __cpu_to_le64(0);
- if (4*2 + bm_space + __le64_to_cpu(sb->size) > dsize)
- bm_space = dsize - __le64_to_cpu(sb->size) -4*2;
- sb->data_offset = __cpu_to_le64(bm_space + 4*2);
- sb->data_size = __cpu_to_le64(dsize - bm_space - 4*2);
+ reserved = bm_space + 4*2;
+ /* Try for multiple of 1Meg so it is nicely aligned */
+ #define ONE_MEG (2*1024)
+ reserved = ((reserved + ONE_MEG-1)/ONE_MEG) * ONE_MEG;
+ if (reserved + __le64_to_cpu(sb->size) > dsize)
+ reserved = dsize - __le64_to_cpu(sb->size);
+ /* force 4K alignment */
+ reserved &= ~7ULL;
+
+ sb->data_offset = __cpu_to_le64(reserved);
+ sb->data_size = __cpu_to_le64(dsize - reserved);
break;
case 2:
sb_offset = 4*2;
> dsize)
bm_space = dsize - __le64_to_cpu(sb->size)
- 4*2 - 4*2;
- sb->data_offset = __cpu_to_le64(4*2 + 4*2 + bm_space);
- sb->data_size = __cpu_to_le64(dsize - 4*2 - 4*2
- - bm_space );
+
+ reserved = bm_space + 4*2 + 4*2;
+ /* Try for multiple of 1Meg so it is nicely aligned */
+ #define ONE_MEG (2*1024)
+ reserved = ((reserved + ONE_MEG-1)/ONE_MEG) * ONE_MEG;
+ if (reserved + __le64_to_cpu(sb->size) > dsize)
+ reserved = dsize - __le64_to_cpu(sb->size);
+ /* force 4K alignment */
+ reserved &= ~7ULL;
+
+ sb->data_offset = __cpu_to_le64(reserved);
+ sb->data_size = __cpu_to_le64(dsize - reserved);
break;
default:
return -EINVAL;
return 1;
}
- if (read(fd, super, 1024) != 1024) {
+ if (aread(fd, super, 1024) != 1024) {
if (devname)
fprintf(stderr, Name ": Cannot read superblock on %s\n",
devname);
* should get that written out.
*/
locate_bitmap1(st, fd);
- if (read(fd, ((char*)super)+1024, 512)
+ if (aread(fd, ((char*)super)+1024, 512)
!= 512)
goto no_bitmap;
return st;
}
if (strcmp(arg, "1.1") == 0 ||
- strcmp(arg, "1.01") == 0) {
+ strcmp(arg, "1.01") == 0 ||
+ strcmp(arg, "") == 0 /* no metadata */
+ ) {
st->minor_version = 1;
return st;
}
if (strcmp(arg, "1.2") == 0 ||
+ strcmp(arg, "default") == 0 ||
strcmp(arg, "1.02") == 0) {
st->minor_version = 2;
return st;
}
#endif
+ if (st->minor_version < 0)
+ /* not specified, so time to set default */
+ st->minor_version = 2;
+ if (super == NULL && st->minor_version > 0) {
+ /* haven't committed to a size yet, so allow some
+ * slack for alignment of data_offset.
+ * We haven't access to device details so allow
+ * 1 Meg if bigger than 1Gig
+ */
+ if (devsize > 1024*1024*2)
+ devsize -= 1024*2;
+ }
switch(st->minor_version) {
- case -1: /* no specified. Now time to set default */
- st->minor_version = 0;
- /* FALL THROUGH */
case 0:
/* at end */
return ((devsize - 8*2 ) & ~(4*2-1));
switch(st->minor_version) {
case 0:
- /* either 3K after the superblock, or some amount of space
- * before.
+ /* either 3K after the superblock (when hot-add),
+ * or some amount of space before.
*/
if (may_change) {
/* We are creating array, so we *know* how much room has
*/
offset = 0;
room = choose_bm_space(__le64_to_cpu(sb->size));
- if (room == 4*2) {
- /* make it 3K after the superblock */
- room = 3*2;
- offset = 2;
- }
} else {
room = __le64_to_cpu(sb->super_offset)
- __le64_to_cpu(sb->data_offset)
min_chunk *= 2;
bits = (bits+1)/2;
}
- if (chunk == UnSet)
+ if (chunk == UnSet) {
+ /* For practical purpose, 64Meg is a good
+ * default chunk size for internal bitmaps.
+ */
chunk = min_chunk;
- else if (chunk < min_chunk)
+ if (chunk < 64*1024*1024)
+ chunk = 64*1024*1024;
+ } else if (chunk < min_chunk)
return 0; /* chunk size too small */
if (chunk == 0) /* rounding problem */
return 0;
if (offset == 0) {
+ /* start bitmap on a 4K boundary with enough space for
+ * the bitmap
+ */
bits = (size*512) / chunk + 1;
- room = ((bits+7)/8 + sizeof(bitmap_super_t) +511)/512;
+ room = ((bits+7)/8 + sizeof(bitmap_super_t) +4095)/4096;
+ room *= 8; /* convert 4K blocks to sectors */
offset = -room;
}
int rv = 0;
int towrite, n;
- char abuf[4096+512];
- char *buf = (char*)(((long)(abuf+512))&~511UL);
+ char *buf = (char*)(((long)(abuf+4096))&~4095UL);
locate_bitmap1(st, fd);
unsigned long long ldsize;
int fd;
- if (level == LEVEL_CONTAINER)
+ if (level == LEVEL_CONTAINER) {
+ if (verbose)
+ fprintf(stderr, Name ": 1.x metadata does not support containers\n");
return 0;
+ }
if (!subdev)
return 1;
#else
.swapuuid = 1,
#endif
+ .name = "1.x",
};
projects / thirdparty / mdadm.git / blobdiff