aboutsummaryrefslogtreecommitdiffstats
path: root/drivers/md
diff options
context:
space:
mode:
Diffstat (limited to 'drivers/md')
-rw-r--r--drivers/md/md.c86
-rw-r--r--drivers/md/raid1.c5
-rw-r--r--drivers/md/raid10.c4
-rw-r--r--drivers/md/raid5.c84
4 files changed, 114 insertions, 65 deletions
diff --git a/drivers/md/md.c b/drivers/md/md.c
index e4e161372a3e..b6d16022a53e 100644
--- a/drivers/md/md.c
+++ b/drivers/md/md.c
@@ -110,7 +110,7 @@ static ctl_table raid_table[] = {
110 .procname = "speed_limit_min", 110 .procname = "speed_limit_min",
111 .data = &sysctl_speed_limit_min, 111 .data = &sysctl_speed_limit_min,
112 .maxlen = sizeof(int), 112 .maxlen = sizeof(int),
113 .mode = 0644, 113 .mode = S_IRUGO|S_IWUSR,
114 .proc_handler = &proc_dointvec, 114 .proc_handler = &proc_dointvec,
115 }, 115 },
116 { 116 {
@@ -118,7 +118,7 @@ static ctl_table raid_table[] = {
118 .procname = "speed_limit_max", 118 .procname = "speed_limit_max",
119 .data = &sysctl_speed_limit_max, 119 .data = &sysctl_speed_limit_max,
120 .maxlen = sizeof(int), 120 .maxlen = sizeof(int),
121 .mode = 0644, 121 .mode = S_IRUGO|S_IWUSR,
122 .proc_handler = &proc_dointvec, 122 .proc_handler = &proc_dointvec,
123 }, 123 },
124 { .ctl_name = 0 } 124 { .ctl_name = 0 }
@@ -129,7 +129,7 @@ static ctl_table raid_dir_table[] = {
129 .ctl_name = DEV_RAID, 129 .ctl_name = DEV_RAID,
130 .procname = "raid", 130 .procname = "raid",
131 .maxlen = 0, 131 .maxlen = 0,
132 .mode = 0555, 132 .mode = S_IRUGO|S_IXUGO,
133 .child = raid_table, 133 .child = raid_table,
134 }, 134 },
135 { .ctl_name = 0 } 135 { .ctl_name = 0 }
@@ -1062,6 +1062,11 @@ static int super_1_load(mdk_rdev_t *rdev, mdk_rdev_t *refdev, int minor_version)
1062 if (rdev->sb_size & bmask) 1062 if (rdev->sb_size & bmask)
1063 rdev-> sb_size = (rdev->sb_size | bmask)+1; 1063 rdev-> sb_size = (rdev->sb_size | bmask)+1;
1064 1064
1065 if (sb->level == cpu_to_le32(LEVEL_MULTIPATH))
1066 rdev->desc_nr = -1;
1067 else
1068 rdev->desc_nr = le32_to_cpu(sb->dev_number);
1069
1065 if (refdev == 0) 1070 if (refdev == 0)
1066 ret = 1; 1071 ret = 1;
1067 else { 1072 else {
@@ -1171,7 +1176,6 @@ static int super_1_validate(mddev_t *mddev, mdk_rdev_t *rdev)
1171 } 1176 }
1172 if (mddev->level != LEVEL_MULTIPATH) { 1177 if (mddev->level != LEVEL_MULTIPATH) {
1173 int role; 1178 int role;
1174 rdev->desc_nr = le32_to_cpu(sb->dev_number);
1175 role = le16_to_cpu(sb->dev_roles[rdev->desc_nr]); 1179 role = le16_to_cpu(sb->dev_roles[rdev->desc_nr]);
1176 switch(role) { 1180 switch(role) {
1177 case 0xffff: /* spare */ 1181 case 0xffff: /* spare */
@@ -1779,8 +1783,8 @@ state_store(mdk_rdev_t *rdev, const char *buf, size_t len)
1779 } 1783 }
1780 return err ? err : len; 1784 return err ? err : len;
1781} 1785}
1782static struct rdev_sysfs_entry 1786static struct rdev_sysfs_entry rdev_state =
1783rdev_state = __ATTR(state, 0644, state_show, state_store); 1787__ATTR(state, S_IRUGO|S_IWUSR, state_show, state_store);
1784 1788
1785static ssize_t 1789static ssize_t
1786super_show(mdk_rdev_t *rdev, char *page) 1790super_show(mdk_rdev_t *rdev, char *page)
@@ -1811,7 +1815,7 @@ errors_store(mdk_rdev_t *rdev, const char *buf, size_t len)
1811 return -EINVAL; 1815 return -EINVAL;
1812} 1816}
1813static struct rdev_sysfs_entry rdev_errors = 1817static struct rdev_sysfs_entry rdev_errors =
1814__ATTR(errors, 0644, errors_show, errors_store); 1818__ATTR(errors, S_IRUGO|S_IWUSR, errors_show, errors_store);
1815 1819
1816static ssize_t 1820static ssize_t
1817slot_show(mdk_rdev_t *rdev, char *page) 1821slot_show(mdk_rdev_t *rdev, char *page)
@@ -1845,7 +1849,7 @@ slot_store(mdk_rdev_t *rdev, const char *buf, size_t len)
1845 1849
1846 1850
1847static struct rdev_sysfs_entry rdev_slot = 1851static struct rdev_sysfs_entry rdev_slot =
1848__ATTR(slot, 0644, slot_show, slot_store); 1852__ATTR(slot, S_IRUGO|S_IWUSR, slot_show, slot_store);
1849 1853
1850static ssize_t 1854static ssize_t
1851offset_show(mdk_rdev_t *rdev, char *page) 1855offset_show(mdk_rdev_t *rdev, char *page)
@@ -1867,7 +1871,7 @@ offset_store(mdk_rdev_t *rdev, const char *buf, size_t len)
1867} 1871}
1868 1872
1869static struct rdev_sysfs_entry rdev_offset = 1873static struct rdev_sysfs_entry rdev_offset =
1870__ATTR(offset, 0644, offset_show, offset_store); 1874__ATTR(offset, S_IRUGO|S_IWUSR, offset_show, offset_store);
1871 1875
1872static ssize_t 1876static ssize_t
1873rdev_size_show(mdk_rdev_t *rdev, char *page) 1877rdev_size_show(mdk_rdev_t *rdev, char *page)
@@ -1891,7 +1895,7 @@ rdev_size_store(mdk_rdev_t *rdev, const char *buf, size_t len)
1891} 1895}
1892 1896
1893static struct rdev_sysfs_entry rdev_size = 1897static struct rdev_sysfs_entry rdev_size =
1894__ATTR(size, 0644, rdev_size_show, rdev_size_store); 1898__ATTR(size, S_IRUGO|S_IWUSR, rdev_size_show, rdev_size_store);
1895 1899
1896static struct attribute *rdev_default_attrs[] = { 1900static struct attribute *rdev_default_attrs[] = {
1897 &rdev_state.attr, 1901 &rdev_state.attr,
@@ -1922,6 +1926,8 @@ rdev_attr_store(struct kobject *kobj, struct attribute *attr,
1922 1926
1923 if (!entry->store) 1927 if (!entry->store)
1924 return -EIO; 1928 return -EIO;
1929 if (!capable(CAP_SYS_ADMIN))
1930 return -EACCES;
1925 return entry->store(rdev, page, length); 1931 return entry->store(rdev, page, length);
1926} 1932}
1927 1933
@@ -2128,7 +2134,7 @@ safe_delay_store(mddev_t *mddev, const char *cbuf, size_t len)
2128 return len; 2134 return len;
2129} 2135}
2130static struct md_sysfs_entry md_safe_delay = 2136static struct md_sysfs_entry md_safe_delay =
2131__ATTR(safe_mode_delay, 0644,safe_delay_show, safe_delay_store); 2137__ATTR(safe_mode_delay, S_IRUGO|S_IWUSR,safe_delay_show, safe_delay_store);
2132 2138
2133static ssize_t 2139static ssize_t
2134level_show(mddev_t *mddev, char *page) 2140level_show(mddev_t *mddev, char *page)
@@ -2163,7 +2169,7 @@ level_store(mddev_t *mddev, const char *buf, size_t len)
2163} 2169}
2164 2170
2165static struct md_sysfs_entry md_level = 2171static struct md_sysfs_entry md_level =
2166__ATTR(level, 0644, level_show, level_store); 2172__ATTR(level, S_IRUGO|S_IWUSR, level_show, level_store);
2167 2173
2168 2174
2169static ssize_t 2175static ssize_t
@@ -2188,7 +2194,7 @@ layout_store(mddev_t *mddev, const char *buf, size_t len)
2188 return len; 2194 return len;
2189} 2195}
2190static struct md_sysfs_entry md_layout = 2196static struct md_sysfs_entry md_layout =
2191__ATTR(layout, 0655, layout_show, layout_store); 2197__ATTR(layout, S_IRUGO|S_IWUSR, layout_show, layout_store);
2192 2198
2193 2199
2194static ssize_t 2200static ssize_t
@@ -2219,7 +2225,7 @@ raid_disks_store(mddev_t *mddev, const char *buf, size_t len)
2219 return rv ? rv : len; 2225 return rv ? rv : len;
2220} 2226}
2221static struct md_sysfs_entry md_raid_disks = 2227static struct md_sysfs_entry md_raid_disks =
2222__ATTR(raid_disks, 0644, raid_disks_show, raid_disks_store); 2228__ATTR(raid_disks, S_IRUGO|S_IWUSR, raid_disks_show, raid_disks_store);
2223 2229
2224static ssize_t 2230static ssize_t
2225chunk_size_show(mddev_t *mddev, char *page) 2231chunk_size_show(mddev_t *mddev, char *page)
@@ -2243,7 +2249,7 @@ chunk_size_store(mddev_t *mddev, const char *buf, size_t len)
2243 return len; 2249 return len;
2244} 2250}
2245static struct md_sysfs_entry md_chunk_size = 2251static struct md_sysfs_entry md_chunk_size =
2246__ATTR(chunk_size, 0644, chunk_size_show, chunk_size_store); 2252__ATTR(chunk_size, S_IRUGO|S_IWUSR, chunk_size_show, chunk_size_store);
2247 2253
2248static ssize_t 2254static ssize_t
2249resync_start_show(mddev_t *mddev, char *page) 2255resync_start_show(mddev_t *mddev, char *page)
@@ -2267,7 +2273,7 @@ resync_start_store(mddev_t *mddev, const char *buf, size_t len)
2267 return len; 2273 return len;
2268} 2274}
2269static struct md_sysfs_entry md_resync_start = 2275static struct md_sysfs_entry md_resync_start =
2270__ATTR(resync_start, 0644, resync_start_show, resync_start_store); 2276__ATTR(resync_start, S_IRUGO|S_IWUSR, resync_start_show, resync_start_store);
2271 2277
2272/* 2278/*
2273 * The array state can be: 2279 * The array state can be:
@@ -2437,7 +2443,8 @@ array_state_store(mddev_t *mddev, const char *buf, size_t len)
2437 else 2443 else
2438 return len; 2444 return len;
2439} 2445}
2440static struct md_sysfs_entry md_array_state = __ATTR(array_state, 0644, array_state_show, array_state_store); 2446static struct md_sysfs_entry md_array_state =
2447__ATTR(array_state, S_IRUGO|S_IWUSR, array_state_show, array_state_store);
2441 2448
2442static ssize_t 2449static ssize_t
2443null_show(mddev_t *mddev, char *page) 2450null_show(mddev_t *mddev, char *page)
@@ -2497,7 +2504,7 @@ new_dev_store(mddev_t *mddev, const char *buf, size_t len)
2497} 2504}
2498 2505
2499static struct md_sysfs_entry md_new_device = 2506static struct md_sysfs_entry md_new_device =
2500__ATTR(new_dev, 0200, null_show, new_dev_store); 2507__ATTR(new_dev, S_IWUSR, null_show, new_dev_store);
2501 2508
2502static ssize_t 2509static ssize_t
2503size_show(mddev_t *mddev, char *page) 2510size_show(mddev_t *mddev, char *page)
@@ -2535,7 +2542,7 @@ size_store(mddev_t *mddev, const char *buf, size_t len)
2535} 2542}
2536 2543
2537static struct md_sysfs_entry md_size = 2544static struct md_sysfs_entry md_size =
2538__ATTR(component_size, 0644, size_show, size_store); 2545__ATTR(component_size, S_IRUGO|S_IWUSR, size_show, size_store);
2539 2546
2540 2547
2541/* Metdata version. 2548/* Metdata version.
@@ -2583,7 +2590,7 @@ metadata_store(mddev_t *mddev, const char *buf, size_t len)
2583} 2590}
2584 2591
2585static struct md_sysfs_entry md_metadata = 2592static struct md_sysfs_entry md_metadata =
2586__ATTR(metadata_version, 0644, metadata_show, metadata_store); 2593__ATTR(metadata_version, S_IRUGO|S_IWUSR, metadata_show, metadata_store);
2587 2594
2588static ssize_t 2595static ssize_t
2589action_show(mddev_t *mddev, char *page) 2596action_show(mddev_t *mddev, char *page)
@@ -2651,12 +2658,11 @@ mismatch_cnt_show(mddev_t *mddev, char *page)
2651 (unsigned long long) mddev->resync_mismatches); 2658 (unsigned long long) mddev->resync_mismatches);
2652} 2659}
2653 2660
2654static struct md_sysfs_entry 2661static struct md_sysfs_entry md_scan_mode =
2655md_scan_mode = __ATTR(sync_action, S_IRUGO|S_IWUSR, action_show, action_store); 2662__ATTR(sync_action, S_IRUGO|S_IWUSR, action_show, action_store);
2656 2663
2657 2664
2658static struct md_sysfs_entry 2665static struct md_sysfs_entry md_mismatches = __ATTR_RO(mismatch_cnt);
2659md_mismatches = __ATTR_RO(mismatch_cnt);
2660 2666
2661static ssize_t 2667static ssize_t
2662sync_min_show(mddev_t *mddev, char *page) 2668sync_min_show(mddev_t *mddev, char *page)
@@ -2715,15 +2721,14 @@ static ssize_t
2715sync_speed_show(mddev_t *mddev, char *page) 2721sync_speed_show(mddev_t *mddev, char *page)
2716{ 2722{
2717 unsigned long resync, dt, db; 2723 unsigned long resync, dt, db;
2718 resync = (mddev->curr_resync - atomic_read(&mddev->recovery_active)); 2724 resync = (mddev->curr_mark_cnt - atomic_read(&mddev->recovery_active));
2719 dt = ((jiffies - mddev->resync_mark) / HZ); 2725 dt = ((jiffies - mddev->resync_mark) / HZ);
2720 if (!dt) dt++; 2726 if (!dt) dt++;
2721 db = resync - (mddev->resync_mark_cnt); 2727 db = resync - (mddev->resync_mark_cnt);
2722 return sprintf(page, "%ld\n", db/dt/2); /* K/sec */ 2728 return sprintf(page, "%ld\n", db/dt/2); /* K/sec */
2723} 2729}
2724 2730
2725static struct md_sysfs_entry 2731static struct md_sysfs_entry md_sync_speed = __ATTR_RO(sync_speed);
2726md_sync_speed = __ATTR_RO(sync_speed);
2727 2732
2728static ssize_t 2733static ssize_t
2729sync_completed_show(mddev_t *mddev, char *page) 2734sync_completed_show(mddev_t *mddev, char *page)
@@ -2739,8 +2744,7 @@ sync_completed_show(mddev_t *mddev, char *page)
2739 return sprintf(page, "%lu / %lu\n", resync, max_blocks); 2744 return sprintf(page, "%lu / %lu\n", resync, max_blocks);
2740} 2745}
2741 2746
2742static struct md_sysfs_entry 2747static struct md_sysfs_entry md_sync_completed = __ATTR_RO(sync_completed);
2743md_sync_completed = __ATTR_RO(sync_completed);
2744 2748
2745static ssize_t 2749static ssize_t
2746suspend_lo_show(mddev_t *mddev, char *page) 2750suspend_lo_show(mddev_t *mddev, char *page)
@@ -2857,6 +2861,8 @@ md_attr_store(struct kobject *kobj, struct attribute *attr,
2857 2861
2858 if (!entry->store) 2862 if (!entry->store)
2859 return -EIO; 2863 return -EIO;
2864 if (!capable(CAP_SYS_ADMIN))
2865 return -EACCES;
2860 rv = mddev_lock(mddev); 2866 rv = mddev_lock(mddev);
2861 if (!rv) { 2867 if (!rv) {
2862 rv = entry->store(mddev, page, length); 2868 rv = entry->store(mddev, page, length);
@@ -3091,7 +3097,6 @@ static int do_md_run(mddev_t * mddev)
3091 } 3097 }
3092 3098
3093 set_bit(MD_RECOVERY_NEEDED, &mddev->recovery); 3099 set_bit(MD_RECOVERY_NEEDED, &mddev->recovery);
3094 md_wakeup_thread(mddev->thread);
3095 3100
3096 if (mddev->sb_dirty) 3101 if (mddev->sb_dirty)
3097 md_update_sb(mddev); 3102 md_update_sb(mddev);
@@ -3112,7 +3117,7 @@ static int do_md_run(mddev_t * mddev)
3112 * start recovery here. If we leave it to md_check_recovery, 3117 * start recovery here. If we leave it to md_check_recovery,
3113 * it will remove the drives and not do the right thing 3118 * it will remove the drives and not do the right thing
3114 */ 3119 */
3115 if (mddev->degraded) { 3120 if (mddev->degraded && !mddev->sync_thread) {
3116 struct list_head *rtmp; 3121 struct list_head *rtmp;
3117 int spares = 0; 3122 int spares = 0;
3118 ITERATE_RDEV(mddev,rdev,rtmp) 3123 ITERATE_RDEV(mddev,rdev,rtmp)
@@ -3133,10 +3138,11 @@ static int do_md_run(mddev_t * mddev)
3133 mdname(mddev)); 3138 mdname(mddev));
3134 /* leave the spares where they are, it shouldn't hurt */ 3139 /* leave the spares where they are, it shouldn't hurt */
3135 mddev->recovery = 0; 3140 mddev->recovery = 0;
3136 } else 3141 }
3137 md_wakeup_thread(mddev->sync_thread);
3138 } 3142 }
3139 } 3143 }
3144 md_wakeup_thread(mddev->thread);
3145 md_wakeup_thread(mddev->sync_thread); /* possibly kick off a reshape */
3140 3146
3141 mddev->changed = 1; 3147 mddev->changed = 1;
3142 md_new_event(mddev); 3148 md_new_event(mddev);
@@ -4586,6 +4592,8 @@ void md_error(mddev_t *mddev, mdk_rdev_t *rdev)
4586 __builtin_return_address(0),__builtin_return_address(1), 4592 __builtin_return_address(0),__builtin_return_address(1),
4587 __builtin_return_address(2),__builtin_return_address(3)); 4593 __builtin_return_address(2),__builtin_return_address(3));
4588*/ 4594*/
4595 if (!mddev->pers)
4596 return;
4589 if (!mddev->pers->error_handler) 4597 if (!mddev->pers->error_handler)
4590 return; 4598 return;
4591 mddev->pers->error_handler(mddev,rdev); 4599 mddev->pers->error_handler(mddev,rdev);
@@ -4683,12 +4691,13 @@ static void status_resync(struct seq_file *seq, mddev_t * mddev)
4683 */ 4691 */
4684 dt = ((jiffies - mddev->resync_mark) / HZ); 4692 dt = ((jiffies - mddev->resync_mark) / HZ);
4685 if (!dt) dt++; 4693 if (!dt) dt++;
4686 db = resync - (mddev->resync_mark_cnt/2); 4694 db = (mddev->curr_mark_cnt - atomic_read(&mddev->recovery_active))
4687 rt = (dt * ((unsigned long)(max_blocks-resync) / (db/100+1)))/100; 4695 - mddev->resync_mark_cnt;
4696 rt = (dt * ((unsigned long)(max_blocks-resync) / (db/2/100+1)))/100;
4688 4697
4689 seq_printf(seq, " finish=%lu.%lumin", rt / 60, (rt % 60)/6); 4698 seq_printf(seq, " finish=%lu.%lumin", rt / 60, (rt % 60)/6);
4690 4699
4691 seq_printf(seq, " speed=%ldK/sec", db/dt); 4700 seq_printf(seq, " speed=%ldK/sec", db/2/dt);
4692} 4701}
4693 4702
4694static void *md_seq_start(struct seq_file *seq, loff_t *pos) 4703static void *md_seq_start(struct seq_file *seq, loff_t *pos)
@@ -5199,6 +5208,7 @@ void md_do_sync(mddev_t *mddev)
5199 5208
5200 j += sectors; 5209 j += sectors;
5201 if (j>1) mddev->curr_resync = j; 5210 if (j>1) mddev->curr_resync = j;
5211 mddev->curr_mark_cnt = io_sectors;
5202 if (last_check == 0) 5212 if (last_check == 0)
5203 /* this is the earliers that rebuilt will be 5213 /* this is the earliers that rebuilt will be
5204 * visible in /proc/mdstat 5214 * visible in /proc/mdstat
@@ -5645,8 +5655,8 @@ static int set_ro(const char *val, struct kernel_param *kp)
5645 return -EINVAL; 5655 return -EINVAL;
5646} 5656}
5647 5657
5648module_param_call(start_ro, set_ro, get_ro, NULL, 0600); 5658module_param_call(start_ro, set_ro, get_ro, NULL, S_IRUSR|S_IWUSR);
5649module_param(start_dirty_degraded, int, 0644); 5659module_param(start_dirty_degraded, int, S_IRUGO|S_IWUSR);
5650 5660
5651 5661
5652EXPORT_SYMBOL(register_md_personality); 5662EXPORT_SYMBOL(register_md_personality);
diff --git a/drivers/md/raid1.c b/drivers/md/raid1.c
index cead918578a7..1efe22a2d041 100644
--- a/drivers/md/raid1.c
+++ b/drivers/md/raid1.c
@@ -1145,7 +1145,7 @@ static int end_sync_write(struct bio *bio, unsigned int bytes_done, int error)
1145 long sectors_to_go = r1_bio->sectors; 1145 long sectors_to_go = r1_bio->sectors;
1146 /* make sure these bits doesn't get cleared. */ 1146 /* make sure these bits doesn't get cleared. */
1147 do { 1147 do {
1148 bitmap_end_sync(mddev->bitmap, r1_bio->sector, 1148 bitmap_end_sync(mddev->bitmap, s,
1149 &sync_blocks, 1); 1149 &sync_blocks, 1);
1150 s += sync_blocks; 1150 s += sync_blocks;
1151 sectors_to_go -= sync_blocks; 1151 sectors_to_go -= sync_blocks;
@@ -1509,6 +1509,9 @@ static void raid1d(mddev_t *mddev)
1509 s<<9, conf->tmppage, READ) == 0) 1509 s<<9, conf->tmppage, READ) == 0)
1510 /* Well, this device is dead */ 1510 /* Well, this device is dead */
1511 md_error(mddev, rdev); 1511 md_error(mddev, rdev);
1512 else
1513 printk(KERN_INFO "raid1:%s: read error corrected (%d sectors at %llu on %s)\n",
1514 mdname(mddev), s, (unsigned long long)(sect + rdev->data_offset), bdevname(rdev->bdev, b));
1512 } 1515 }
1513 } 1516 }
1514 } else { 1517 } else {
diff --git a/drivers/md/raid10.c b/drivers/md/raid10.c
index 7f636283a1ba..016ddb831c9b 100644
--- a/drivers/md/raid10.c
+++ b/drivers/md/raid10.c
@@ -1492,6 +1492,10 @@ static void raid10d(mddev_t *mddev)
1492 s<<9, conf->tmppage, READ) == 0) 1492 s<<9, conf->tmppage, READ) == 0)
1493 /* Well, this device is dead */ 1493 /* Well, this device is dead */
1494 md_error(mddev, rdev); 1494 md_error(mddev, rdev);
1495 else
1496 printk(KERN_INFO "raid10:%s: read error corrected (%d sectors at %llu on %s)\n",
1497 mdname(mddev), s, (unsigned long long)(sect+rdev->data_offset), bdevname(rdev->bdev, b));
1498
1495 rdev_dec_pending(rdev, mddev); 1499 rdev_dec_pending(rdev, mddev);
1496 rcu_read_lock(); 1500 rcu_read_lock();
1497 } 1501 }
diff --git a/drivers/md/raid5.c b/drivers/md/raid5.c
index 7433871f4b3a..450066007160 100644
--- a/drivers/md/raid5.c
+++ b/drivers/md/raid5.c
@@ -18,6 +18,30 @@
18 * Software Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA. 18 * Software Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
19 */ 19 */
20 20
21/*
22 * BITMAP UNPLUGGING:
23 *
24 * The sequencing for updating the bitmap reliably is a little
25 * subtle (and I got it wrong the first time) so it deserves some
26 * explanation.
27 *
28 * We group bitmap updates into batches. Each batch has a number.
29 * We may write out several batches at once, but that isn't very important.
30 * conf->bm_write is the number of the last batch successfully written.
31 * conf->bm_flush is the number of the last batch that was closed to
32 * new additions.
33 * When we discover that we will need to write to any block in a stripe
34 * (in add_stripe_bio) we update the in-memory bitmap and record in sh->bm_seq
35 * the number of the batch it will be in. This is bm_flush+1.
36 * When we are ready to do a write, if that batch hasn't been written yet,
37 * we plug the array and queue the stripe for later.
38 * When an unplug happens, we increment bm_flush, thus closing the current
39 * batch.
40 * When we notice that bm_flush > bm_write, we write out all pending updates
41 * to the bitmap, and advance bm_write to where bm_flush was.
42 * This may occasionally write a bit out twice, but is sure never to
43 * miss any bits.
44 */
21 45
22#include <linux/module.h> 46#include <linux/module.h>
23#include <linux/slab.h> 47#include <linux/slab.h>
@@ -88,12 +112,14 @@ static void __release_stripe(raid5_conf_t *conf, struct stripe_head *sh)
88 BUG_ON(!list_empty(&sh->lru)); 112 BUG_ON(!list_empty(&sh->lru));
89 BUG_ON(atomic_read(&conf->active_stripes)==0); 113 BUG_ON(atomic_read(&conf->active_stripes)==0);
90 if (test_bit(STRIPE_HANDLE, &sh->state)) { 114 if (test_bit(STRIPE_HANDLE, &sh->state)) {
91 if (test_bit(STRIPE_DELAYED, &sh->state)) 115 if (test_bit(STRIPE_DELAYED, &sh->state)) {
92 list_add_tail(&sh->lru, &conf->delayed_list); 116 list_add_tail(&sh->lru, &conf->delayed_list);
93 else if (test_bit(STRIPE_BIT_DELAY, &sh->state) && 117 blk_plug_device(conf->mddev->queue);
94 conf->seq_write == sh->bm_seq) 118 } else if (test_bit(STRIPE_BIT_DELAY, &sh->state) &&
119 sh->bm_seq - conf->seq_write > 0) {
95 list_add_tail(&sh->lru, &conf->bitmap_list); 120 list_add_tail(&sh->lru, &conf->bitmap_list);
96 else { 121 blk_plug_device(conf->mddev->queue);
122 } else {
97 clear_bit(STRIPE_BIT_DELAY, &sh->state); 123 clear_bit(STRIPE_BIT_DELAY, &sh->state);
98 list_add_tail(&sh->lru, &conf->handle_list); 124 list_add_tail(&sh->lru, &conf->handle_list);
99 } 125 }
@@ -270,7 +296,7 @@ static struct stripe_head *get_active_stripe(raid5_conf_t *conf, sector_t sector
270 < (conf->max_nr_stripes *3/4) 296 < (conf->max_nr_stripes *3/4)
271 || !conf->inactive_blocked), 297 || !conf->inactive_blocked),
272 conf->device_lock, 298 conf->device_lock,
273 unplug_slaves(conf->mddev) 299 raid5_unplug_device(conf->mddev->queue)
274 ); 300 );
275 conf->inactive_blocked = 0; 301 conf->inactive_blocked = 0;
276 } else 302 } else
@@ -281,7 +307,8 @@ static struct stripe_head *get_active_stripe(raid5_conf_t *conf, sector_t sector
281 } else { 307 } else {
282 if (!test_bit(STRIPE_HANDLE, &sh->state)) 308 if (!test_bit(STRIPE_HANDLE, &sh->state))
283 atomic_inc(&conf->active_stripes); 309 atomic_inc(&conf->active_stripes);
284 if (list_empty(&sh->lru)) 310 if (list_empty(&sh->lru) &&
311 !test_bit(STRIPE_EXPANDING, &sh->state))
285 BUG(); 312 BUG();
286 list_del_init(&sh->lru); 313 list_del_init(&sh->lru);
287 } 314 }
@@ -496,6 +523,8 @@ static int raid5_end_read_request(struct bio * bi, unsigned int bytes_done,
496 raid5_conf_t *conf = sh->raid_conf; 523 raid5_conf_t *conf = sh->raid_conf;
497 int disks = sh->disks, i; 524 int disks = sh->disks, i;
498 int uptodate = test_bit(BIO_UPTODATE, &bi->bi_flags); 525 int uptodate = test_bit(BIO_UPTODATE, &bi->bi_flags);
526 char b[BDEVNAME_SIZE];
527 mdk_rdev_t *rdev;
499 528
500 if (bi->bi_size) 529 if (bi->bi_size)
501 return 1; 530 return 1;
@@ -543,25 +572,39 @@ static int raid5_end_read_request(struct bio * bi, unsigned int bytes_done,
543 set_bit(R5_UPTODATE, &sh->dev[i].flags); 572 set_bit(R5_UPTODATE, &sh->dev[i].flags);
544#endif 573#endif
545 if (test_bit(R5_ReadError, &sh->dev[i].flags)) { 574 if (test_bit(R5_ReadError, &sh->dev[i].flags)) {
546 printk(KERN_INFO "raid5: read error corrected!!\n"); 575 rdev = conf->disks[i].rdev;
576 printk(KERN_INFO "raid5:%s: read error corrected (%lu sectors at %llu on %s)\n",
577 mdname(conf->mddev), STRIPE_SECTORS,
578 (unsigned long long)sh->sector + rdev->data_offset,
579 bdevname(rdev->bdev, b));
547 clear_bit(R5_ReadError, &sh->dev[i].flags); 580 clear_bit(R5_ReadError, &sh->dev[i].flags);
548 clear_bit(R5_ReWrite, &sh->dev[i].flags); 581 clear_bit(R5_ReWrite, &sh->dev[i].flags);
549 } 582 }
550 if (atomic_read(&conf->disks[i].rdev->read_errors)) 583 if (atomic_read(&conf->disks[i].rdev->read_errors))
551 atomic_set(&conf->disks[i].rdev->read_errors, 0); 584 atomic_set(&conf->disks[i].rdev->read_errors, 0);
552 } else { 585 } else {
586 const char *bdn = bdevname(conf->disks[i].rdev->bdev, b);
553 int retry = 0; 587 int retry = 0;
588 rdev = conf->disks[i].rdev;
589
554 clear_bit(R5_UPTODATE, &sh->dev[i].flags); 590 clear_bit(R5_UPTODATE, &sh->dev[i].flags);
555 atomic_inc(&conf->disks[i].rdev->read_errors); 591 atomic_inc(&rdev->read_errors);
556 if (conf->mddev->degraded) 592 if (conf->mddev->degraded)
557 printk(KERN_WARNING "raid5: read error not correctable.\n"); 593 printk(KERN_WARNING "raid5:%s: read error not correctable (sector %llu on %s).\n",
594 mdname(conf->mddev),
595 (unsigned long long)sh->sector + rdev->data_offset,
596 bdn);
558 else if (test_bit(R5_ReWrite, &sh->dev[i].flags)) 597 else if (test_bit(R5_ReWrite, &sh->dev[i].flags))
559 /* Oh, no!!! */ 598 /* Oh, no!!! */
560 printk(KERN_WARNING "raid5: read error NOT corrected!!\n"); 599 printk(KERN_WARNING "raid5:%s: read error NOT corrected!! (sector %llu on %s).\n",
561 else if (atomic_read(&conf->disks[i].rdev->read_errors) 600 mdname(conf->mddev),
601 (unsigned long long)sh->sector + rdev->data_offset,
602 bdn);
603 else if (atomic_read(&rdev->read_errors)
562 > conf->max_nr_stripes) 604 > conf->max_nr_stripes)
563 printk(KERN_WARNING 605 printk(KERN_WARNING
564 "raid5: Too many read errors, failing device.\n"); 606 "raid5:%s: Too many read errors, failing device %s.\n",
607 mdname(conf->mddev), bdn);
565 else 608 else
566 retry = 1; 609 retry = 1;
567 if (retry) 610 if (retry)
@@ -569,7 +612,7 @@ static int raid5_end_read_request(struct bio * bi, unsigned int bytes_done,
569 else { 612 else {
570 clear_bit(R5_ReadError, &sh->dev[i].flags); 613 clear_bit(R5_ReadError, &sh->dev[i].flags);
571 clear_bit(R5_ReWrite, &sh->dev[i].flags); 614 clear_bit(R5_ReWrite, &sh->dev[i].flags);
572 md_error(conf->mddev, conf->disks[i].rdev); 615 md_error(conf->mddev, rdev);
573 } 616 }
574 } 617 }
575 rdev_dec_pending(conf->disks[i].rdev, conf->mddev); 618 rdev_dec_pending(conf->disks[i].rdev, conf->mddev);
@@ -1270,9 +1313,9 @@ static int add_stripe_bio(struct stripe_head *sh, struct bio *bi, int dd_idx, in
1270 (unsigned long long)sh->sector, dd_idx); 1313 (unsigned long long)sh->sector, dd_idx);
1271 1314
1272 if (conf->mddev->bitmap && firstwrite) { 1315 if (conf->mddev->bitmap && firstwrite) {
1273 sh->bm_seq = conf->seq_write;
1274 bitmap_startwrite(conf->mddev->bitmap, sh->sector, 1316 bitmap_startwrite(conf->mddev->bitmap, sh->sector,
1275 STRIPE_SECTORS, 0); 1317 STRIPE_SECTORS, 0);
1318 sh->bm_seq = conf->seq_flush+1;
1276 set_bit(STRIPE_BIT_DELAY, &sh->state); 1319 set_bit(STRIPE_BIT_DELAY, &sh->state);
1277 } 1320 }
1278 1321
@@ -2554,13 +2597,6 @@ static int raid5_issue_flush(request_queue_t *q, struct gendisk *disk,
2554 return ret; 2597 return ret;
2555} 2598}
2556 2599
2557static inline void raid5_plug_device(raid5_conf_t *conf)
2558{
2559 spin_lock_irq(&conf->device_lock);
2560 blk_plug_device(conf->mddev->queue);
2561 spin_unlock_irq(&conf->device_lock);
2562}
2563
2564static int make_request(request_queue_t *q, struct bio * bi) 2600static int make_request(request_queue_t *q, struct bio * bi)
2565{ 2601{
2566 mddev_t *mddev = q->queuedata; 2602 mddev_t *mddev = q->queuedata;
@@ -2670,7 +2706,6 @@ static int make_request(request_queue_t *q, struct bio * bi)
2670 goto retry; 2706 goto retry;
2671 } 2707 }
2672 finish_wait(&conf->wait_for_overlap, &w); 2708 finish_wait(&conf->wait_for_overlap, &w);
2673 raid5_plug_device(conf);
2674 handle_stripe(sh, NULL); 2709 handle_stripe(sh, NULL);
2675 release_stripe(sh); 2710 release_stripe(sh);
2676 } else { 2711 } else {
@@ -2923,7 +2958,7 @@ static void raid5d (mddev_t *mddev)
2923 while (1) { 2958 while (1) {
2924 struct list_head *first; 2959 struct list_head *first;
2925 2960
2926 if (conf->seq_flush - conf->seq_write > 0) { 2961 if (conf->seq_flush != conf->seq_write) {
2927 int seq = conf->seq_flush; 2962 int seq = conf->seq_flush;
2928 spin_unlock_irq(&conf->device_lock); 2963 spin_unlock_irq(&conf->device_lock);
2929 bitmap_unplug(mddev->bitmap); 2964 bitmap_unplug(mddev->bitmap);
@@ -3246,9 +3281,6 @@ static int run(mddev_t *mddev)
3246 set_bit(MD_RECOVERY_RUNNING, &mddev->recovery); 3281 set_bit(MD_RECOVERY_RUNNING, &mddev->recovery);
3247 mddev->sync_thread = md_register_thread(md_do_sync, mddev, 3282 mddev->sync_thread = md_register_thread(md_do_sync, mddev,
3248 "%s_reshape"); 3283 "%s_reshape");
3249 /* FIXME if md_register_thread fails?? */
3250 md_wakeup_thread(mddev->sync_thread);
3251
3252 } 3284 }
3253 3285
3254 /* read-ahead size must cover two whole stripes, which is 3286 /* read-ahead size must cover two whole stripes, which is