xfs: switch to NOFS allocation under i_lock in xfs_readlink_bmap
[safe/jmp/linux-2.6] / fs / ocfs2 / super.c
index 2eb657c..201b40a 100644 (file)
@@ -42,6 +42,7 @@
 #include <linux/mount.h>
 #include <linux/seq_file.h>
 #include <linux/quotaops.h>
+#include <linux/smp_lock.h>
 
 #define MLOG_MASK_PREFIX ML_SUPER
 #include <cluster/masklog.h>
@@ -52,6 +53,7 @@
 #include "ocfs1_fs_compat.h"
 
 #include "alloc.h"
+#include "blockcheck.h"
 #include "dlmglue.h"
 #include "export.h"
 #include "extent_map.h"
@@ -125,7 +127,6 @@ static int ocfs2_get_sector(struct super_block *sb,
                            struct buffer_head **bh,
                            int block,
                            int sect_size);
-static void ocfs2_write_super(struct super_block *sb);
 static struct inode *ocfs2_alloc_inode(struct super_block *sb);
 static void ocfs2_destroy_inode(struct inode *inode);
 static int ocfs2_susp_quotas(struct ocfs2_super *osb, int unsuspend);
@@ -140,7 +141,6 @@ static const struct super_operations ocfs2_sops = {
        .clear_inode    = ocfs2_clear_inode,
        .delete_inode   = ocfs2_delete_inode,
        .sync_fs        = ocfs2_sync_fs,
-       .write_super    = ocfs2_write_super,
        .put_super      = ocfs2_put_super,
        .remount_fs     = ocfs2_remount,
        .show_options   = ocfs2_show_options,
@@ -200,15 +200,169 @@ static const match_table_t tokens = {
        {Opt_err, NULL}
 };
 
-/*
- * write_super and sync_fs ripped right out of ext3.
- */
-static void ocfs2_write_super(struct super_block *sb)
+#ifdef CONFIG_DEBUG_FS
+static int ocfs2_osb_dump(struct ocfs2_super *osb, char *buf, int len)
+{
+       int out = 0;
+       int i;
+       struct ocfs2_cluster_connection *cconn = osb->cconn;
+       struct ocfs2_recovery_map *rm = osb->recovery_map;
+
+       out += snprintf(buf + out, len - out,
+                       "%10s => Id: %-s  Uuid: %-s  Gen: 0x%X  Label: %-s\n",
+                       "Device", osb->dev_str, osb->uuid_str,
+                       osb->fs_generation, osb->vol_label);
+
+       out += snprintf(buf + out, len - out,
+                       "%10s => State: %d  Flags: 0x%lX\n", "Volume",
+                       atomic_read(&osb->vol_state), osb->osb_flags);
+
+       out += snprintf(buf + out, len - out,
+                       "%10s => Block: %lu  Cluster: %d\n", "Sizes",
+                       osb->sb->s_blocksize, osb->s_clustersize);
+
+       out += snprintf(buf + out, len - out,
+                       "%10s => Compat: 0x%X  Incompat: 0x%X  "
+                       "ROcompat: 0x%X\n",
+                       "Features", osb->s_feature_compat,
+                       osb->s_feature_incompat, osb->s_feature_ro_compat);
+
+       out += snprintf(buf + out, len - out,
+                       "%10s => Opts: 0x%lX  AtimeQuanta: %u\n", "Mount",
+                       osb->s_mount_opt, osb->s_atime_quantum);
+
+       out += snprintf(buf + out, len - out,
+                       "%10s => Stack: %s  Name: %*s  Version: %d.%d\n",
+                       "Cluster",
+                       (*osb->osb_cluster_stack == '\0' ?
+                        "o2cb" : osb->osb_cluster_stack),
+                       cconn->cc_namelen, cconn->cc_name,
+                       cconn->cc_version.pv_major, cconn->cc_version.pv_minor);
+
+       spin_lock(&osb->dc_task_lock);
+       out += snprintf(buf + out, len - out,
+                       "%10s => Pid: %d  Count: %lu  WakeSeq: %lu  "
+                       "WorkSeq: %lu\n", "DownCnvt",
+                       task_pid_nr(osb->dc_task), osb->blocked_lock_count,
+                       osb->dc_wake_sequence, osb->dc_work_sequence);
+       spin_unlock(&osb->dc_task_lock);
+
+       spin_lock(&osb->osb_lock);
+       out += snprintf(buf + out, len - out, "%10s => Pid: %d  Nodes:",
+                       "Recovery",
+                       (osb->recovery_thread_task ?
+                        task_pid_nr(osb->recovery_thread_task) : -1));
+       if (rm->rm_used == 0)
+               out += snprintf(buf + out, len - out, " None\n");
+       else {
+               for (i = 0; i < rm->rm_used; i++)
+                       out += snprintf(buf + out, len - out, " %d",
+                                       rm->rm_entries[i]);
+               out += snprintf(buf + out, len - out, "\n");
+       }
+       spin_unlock(&osb->osb_lock);
+
+       out += snprintf(buf + out, len - out,
+                       "%10s => Pid: %d  Interval: %lu  Needs: %d\n", "Commit",
+                       task_pid_nr(osb->commit_task), osb->osb_commit_interval,
+                       atomic_read(&osb->needs_checkpoint));
+
+       out += snprintf(buf + out, len - out,
+                       "%10s => State: %d  NumTxns: %d  TxnId: %lu\n",
+                       "Journal", osb->journal->j_state,
+                       atomic_read(&osb->journal->j_num_trans),
+                       osb->journal->j_trans_id);
+
+       out += snprintf(buf + out, len - out,
+                       "%10s => GlobalAllocs: %d  LocalAllocs: %d  "
+                       "SubAllocs: %d  LAWinMoves: %d  SAExtends: %d\n",
+                       "Stats",
+                       atomic_read(&osb->alloc_stats.bitmap_data),
+                       atomic_read(&osb->alloc_stats.local_data),
+                       atomic_read(&osb->alloc_stats.bg_allocs),
+                       atomic_read(&osb->alloc_stats.moves),
+                       atomic_read(&osb->alloc_stats.bg_extends));
+
+       out += snprintf(buf + out, len - out,
+                       "%10s => State: %u  Descriptor: %llu  Size: %u bits  "
+                       "Default: %u bits\n",
+                       "LocalAlloc", osb->local_alloc_state,
+                       (unsigned long long)osb->la_last_gd,
+                       osb->local_alloc_bits, osb->local_alloc_default_bits);
+
+       spin_lock(&osb->osb_lock);
+       out += snprintf(buf + out, len - out,
+                       "%10s => Slot: %d  NumStolen: %d\n", "Steal",
+                       osb->s_inode_steal_slot,
+                       atomic_read(&osb->s_num_inodes_stolen));
+       spin_unlock(&osb->osb_lock);
+
+       out += snprintf(buf + out, len - out, "%10s => %3s  %10s\n",
+                       "Slots", "Num", "RecoGen");
+
+       for (i = 0; i < osb->max_slots; ++i) {
+               out += snprintf(buf + out, len - out,
+                               "%10s  %c %3d  %10d\n",
+                               " ",
+                               (i == osb->slot_num ? '*' : ' '),
+                               i, osb->slot_recovery_generations[i]);
+       }
+
+       return out;
+}
+
+static int ocfs2_osb_debug_open(struct inode *inode, struct file *file)
+{
+       struct ocfs2_super *osb = inode->i_private;
+       char *buf = NULL;
+
+       buf = kmalloc(PAGE_SIZE, GFP_KERNEL);
+       if (!buf)
+               goto bail;
+
+       i_size_write(inode, ocfs2_osb_dump(osb, buf, PAGE_SIZE));
+
+       file->private_data = buf;
+
+       return 0;
+bail:
+       return -ENOMEM;
+}
+
+static int ocfs2_debug_release(struct inode *inode, struct file *file)
+{
+       kfree(file->private_data);
+       return 0;
+}
+
+static ssize_t ocfs2_debug_read(struct file *file, char __user *buf,
+                               size_t nbytes, loff_t *ppos)
+{
+       return simple_read_from_buffer(buf, nbytes, ppos, file->private_data,
+                                      i_size_read(file->f_mapping->host));
+}
+#else
+static int ocfs2_osb_debug_open(struct inode *inode, struct file *file)
 {
-       if (mutex_trylock(&sb->s_lock) != 0)
-               BUG();
-       sb->s_dirt = 0;
+       return 0;
+}
+static int ocfs2_debug_release(struct inode *inode, struct file *file)
+{
+       return 0;
+}
+static ssize_t ocfs2_debug_read(struct file *file, char __user *buf,
+                               size_t nbytes, loff_t *ppos)
+{
+       return 0;
 }
+#endif /* CONFIG_DEBUG_FS */
+
+static struct file_operations ocfs2_osb_debug_fops = {
+       .open =         ocfs2_osb_debug_open,
+       .release =      ocfs2_debug_release,
+       .read =         ocfs2_debug_read,
+       .llseek =       generic_file_llseek,
+};
 
 static int ocfs2_sync_fs(struct super_block *sb, int wait)
 {
@@ -216,8 +370,6 @@ static int ocfs2_sync_fs(struct super_block *sb, int wait)
        tid_t target;
        struct ocfs2_super *osb = OCFS2_SB(sb);
 
-       sb->s_dirt = 0;
-
        if (ocfs2_is_hard_readonly(osb))
                return -EROFS;
 
@@ -430,6 +582,8 @@ static int ocfs2_remount(struct super_block *sb, int *flags, char *data)
        struct mount_options parsed_options;
        struct ocfs2_super *osb = OCFS2_SB(sb);
 
+       lock_kernel();
+
        if (!ocfs2_parse_options(sb, data, &parsed_options, 1)) {
                ret = -EINVAL;
                goto out;
@@ -533,6 +687,7 @@ unlock_osb:
                        ocfs2_set_journal_params(osb);
        }
 out:
+       unlock_kernel();
        return ret;
 }
 
@@ -548,7 +703,7 @@ static int ocfs2_sb_probe(struct super_block *sb,
        *bh = NULL;
 
        /* may be > 512 */
-       *sector_size = bdev_hardsect_size(sb->s_bdev);
+       *sector_size = bdev_logical_block_size(sb->s_bdev);
        if (*sector_size > OCFS2_MAX_BLOCKSIZE) {
                mlog(ML_ERROR, "Hardware sector size too large: %d (max=%d)\n",
                     *sector_size, OCFS2_MAX_BLOCKSIZE);
@@ -925,6 +1080,16 @@ static int ocfs2_fill_super(struct super_block *sb, void *data, int silent)
                goto read_super_error;
        }
 
+       osb->osb_ctxt = debugfs_create_file("fs_state", S_IFREG|S_IRUSR,
+                                           osb->osb_debug_root,
+                                           osb,
+                                           &ocfs2_osb_debug_fops);
+       if (!osb->osb_ctxt) {
+               status = -EINVAL;
+               mlog_errno(status);
+               goto read_super_error;
+       }
+
        status = ocfs2_mount_volume(sb);
        if (osb->root_inode)
                inode = igrab(osb->root_inode);
@@ -1375,9 +1540,13 @@ static void ocfs2_put_super(struct super_block *sb)
 {
        mlog_entry("(0x%p)\n", sb);
 
+       lock_kernel();
+
        ocfs2_sync_blockdev(sb);
        ocfs2_dismount_volume(sb, 0);
 
+       unlock_kernel();
+
        mlog_exit_void();
 }
 
@@ -1536,6 +1705,13 @@ static int ocfs2_get_sector(struct super_block *sb,
        unlock_buffer(*bh);
        ll_rw_block(READ, 1, bh);
        wait_on_buffer(*bh);
+       if (!buffer_uptodate(*bh)) {
+               mlog_errno(-EIO);
+               brelse(*bh);
+               *bh = NULL;
+               return -EIO;
+       }
+
        return 0;
 }
 
@@ -1612,6 +1788,8 @@ static void ocfs2_dismount_volume(struct super_block *sb, int mnt_err)
        osb = OCFS2_SB(sb);
        BUG_ON(!osb);
 
+       debugfs_remove(osb->osb_ctxt);
+
        ocfs2_disable_quotas(osb);
 
        ocfs2_shutdown_local_alloc(osb);
@@ -1734,6 +1912,12 @@ static int ocfs2_initialize_super(struct super_block *sb,
        bbits = le32_to_cpu(di->id2.i_super.s_blocksize_bits);
        sb->s_maxbytes = ocfs2_max_file_offset(bbits, cbits);
 
+       osb->osb_dx_mask = (1 << (cbits - bbits)) - 1;
+
+       for (i = 0; i < 3; i++)
+               osb->osb_dx_seed[i] = le32_to_cpu(di->id2.i_super.s_dx_seed[i]);
+       osb->osb_dx_seed[3] = le32_to_cpu(di->id2.i_super.s_uuid_hash);
+
        osb->sb = sb;
        /* Save off for ocfs2_rw_direct */
        osb->s_sectsize_bits = blksize_bits(sector_size);
@@ -1746,6 +1930,7 @@ static int ocfs2_initialize_super(struct super_block *sb,
        INIT_LIST_HEAD(&osb->blocked_lock_list);
        osb->blocked_lock_count = 0;
        spin_lock_init(&osb->osb_lock);
+       spin_lock_init(&osb->osb_xattr_lock);
        ocfs2_init_inode_steal_slot(osb);
 
        atomic_set(&osb->alloc_stats.moves, 0);
@@ -1886,6 +2071,9 @@ static int ocfs2_initialize_super(struct super_block *sb,
        INIT_WORK(&journal->j_recovery_work, ocfs2_complete_recovery);
        journal->j_state = OCFS2_JOURNAL_FREE;
 
+       INIT_WORK(&osb->dentry_lock_work, ocfs2_drop_dl_inodes);
+       osb->dentry_lock_list = NULL;
+
        /* get some pseudo constants for clustersize bits */
        osb->s_clustersize_bits =
                le32_to_cpu(di->id2.i_super.s_clustersize_bits);
@@ -1989,6 +2177,15 @@ static int ocfs2_verify_volume(struct ocfs2_dinode *di,
 
        if (memcmp(di->i_signature, OCFS2_SUPER_BLOCK_SIGNATURE,
                   strlen(OCFS2_SUPER_BLOCK_SIGNATURE)) == 0) {
+               /* We have to do a raw check of the feature here */
+               if (le32_to_cpu(di->id2.i_super.s_feature_incompat) &
+                   OCFS2_FEATURE_INCOMPAT_META_ECC) {
+                       status = ocfs2_block_check_validate(bh->b_data,
+                                                           bh->b_size,
+                                                           &di->i_check);
+                       if (status)
+                               goto out;
+               }
                status = -EINVAL;
                if ((1 << le32_to_cpu(di->id2.i_super.s_blocksize_bits)) != blksz) {
                        mlog(ML_ERROR, "found superblock with incorrect block "
@@ -2030,6 +2227,7 @@ static int ocfs2_verify_volume(struct ocfs2_dinode *di,
                }
        }
 
+out:
        mlog_exit(status);
        return status;
 }
@@ -2108,6 +2306,12 @@ static int ocfs2_check_volume(struct ocfs2_super *osb)
         * lock, and it's marked as dirty, set the bit in the recover
         * map and launch a recovery thread for it. */
        status = ocfs2_mark_dead_nodes(osb);
+       if (status < 0) {
+               mlog_errno(status);
+               goto finally;
+       }
+
+       status = ocfs2_compute_replay_slots(osb);
        if (status < 0)
                mlog_errno(status);