nfsd4: move check_stateid_generation check
[safe/jmp/linux-2.6] / fs / ocfs2 / super.c
index 60f1d29..7ac83a8 100644 (file)
@@ -41,6 +41,7 @@
 #include <linux/debugfs.h>
 #include <linux/mount.h>
 #include <linux/seq_file.h>
+#include <linux/quotaops.h>
 
 #define MLOG_MASK_PREFIX ML_SUPER
 #include <cluster/masklog.h>
@@ -51,6 +52,7 @@
 #include "ocfs1_fs_compat.h"
 
 #include "alloc.h"
+#include "blockcheck.h"
 #include "dlmglue.h"
 #include "export.h"
 #include "extent_map.h"
@@ -127,6 +129,9 @@ static int ocfs2_get_sector(struct super_block *sb,
 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);
+static int ocfs2_enable_quotas(struct ocfs2_super *osb);
+static void ocfs2_disable_quotas(struct ocfs2_super *osb);
 
 static const struct super_operations ocfs2_sops = {
        .statfs         = ocfs2_statfs,
@@ -165,6 +170,8 @@ enum {
        Opt_inode64,
        Opt_acl,
        Opt_noacl,
+       Opt_usrquota,
+       Opt_grpquota,
        Opt_err,
 };
 
@@ -189,6 +196,8 @@ static const match_table_t tokens = {
        {Opt_inode64, "inode64"},
        {Opt_acl, "acl"},
        {Opt_noacl, "noacl"},
+       {Opt_usrquota, "usrquota"},
+       {Opt_grpquota, "grpquota"},
        {Opt_err, NULL}
 };
 
@@ -452,6 +461,12 @@ static int ocfs2_remount(struct super_block *sb, int *flags, char *data)
 
        /* We're going to/from readonly mode. */
        if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) {
+               /* Disable quota accounting before remounting RO */
+               if (*flags & MS_RDONLY) {
+                       ret = ocfs2_susp_quotas(osb, 0);
+                       if (ret < 0)
+                               goto out;
+               }
                /* Lock here so the check of HARD_RO and the potential
                 * setting of SOFT_RO is atomic. */
                spin_lock(&osb->osb_lock);
@@ -487,6 +502,21 @@ static int ocfs2_remount(struct super_block *sb, int *flags, char *data)
                }
 unlock_osb:
                spin_unlock(&osb->osb_lock);
+               /* Enable quota accounting after remounting RW */
+               if (!ret && !(*flags & MS_RDONLY)) {
+                       if (sb_any_quota_suspended(sb))
+                               ret = ocfs2_susp_quotas(osb, 1);
+                       else
+                               ret = ocfs2_enable_quotas(osb);
+                       if (ret < 0) {
+                               /* Return back changes... */
+                               spin_lock(&osb->osb_lock);
+                               sb->s_flags |= MS_RDONLY;
+                               osb->osb_flags |= OCFS2_OSB_SOFT_RO;
+                               spin_unlock(&osb->osb_lock);
+                               goto out;
+                       }
+               }
        }
 
        if (!ret) {
@@ -647,6 +677,131 @@ static int ocfs2_verify_userspace_stack(struct ocfs2_super *osb,
        return 0;
 }
 
+static int ocfs2_susp_quotas(struct ocfs2_super *osb, int unsuspend)
+{
+       int type;
+       struct super_block *sb = osb->sb;
+       unsigned int feature[MAXQUOTAS] = { OCFS2_FEATURE_RO_COMPAT_USRQUOTA,
+                                            OCFS2_FEATURE_RO_COMPAT_GRPQUOTA};
+       int status = 0;
+
+       for (type = 0; type < MAXQUOTAS; type++) {
+               if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb, feature[type]))
+                       continue;
+               if (unsuspend)
+                       status = vfs_quota_enable(
+                                       sb_dqopt(sb)->files[type],
+                                       type, QFMT_OCFS2,
+                                       DQUOT_SUSPENDED);
+               else
+                       status = vfs_quota_disable(sb, type,
+                                                  DQUOT_SUSPENDED);
+               if (status < 0)
+                       break;
+       }
+       if (status < 0)
+               mlog(ML_ERROR, "Failed to suspend/unsuspend quotas on "
+                    "remount (error = %d).\n", status);
+       return status;
+}
+
+static int ocfs2_enable_quotas(struct ocfs2_super *osb)
+{
+       struct inode *inode[MAXQUOTAS] = { NULL, NULL };
+       struct super_block *sb = osb->sb;
+       unsigned int feature[MAXQUOTAS] = { OCFS2_FEATURE_RO_COMPAT_USRQUOTA,
+                                            OCFS2_FEATURE_RO_COMPAT_GRPQUOTA};
+       unsigned int ino[MAXQUOTAS] = { LOCAL_USER_QUOTA_SYSTEM_INODE,
+                                       LOCAL_GROUP_QUOTA_SYSTEM_INODE };
+       int status;
+       int type;
+
+       sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE | DQUOT_NEGATIVE_USAGE;
+       for (type = 0; type < MAXQUOTAS; type++) {
+               if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb, feature[type]))
+                       continue;
+               inode[type] = ocfs2_get_system_file_inode(osb, ino[type],
+                                                       osb->slot_num);
+               if (!inode[type]) {
+                       status = -ENOENT;
+                       goto out_quota_off;
+               }
+               status = vfs_quota_enable(inode[type], type, QFMT_OCFS2,
+                                               DQUOT_USAGE_ENABLED);
+               if (status < 0)
+                       goto out_quota_off;
+       }
+
+       for (type = 0; type < MAXQUOTAS; type++)
+               iput(inode[type]);
+       return 0;
+out_quota_off:
+       ocfs2_disable_quotas(osb);
+       for (type = 0; type < MAXQUOTAS; type++)
+               iput(inode[type]);
+       mlog_errno(status);
+       return status;
+}
+
+static void ocfs2_disable_quotas(struct ocfs2_super *osb)
+{
+       int type;
+       struct inode *inode;
+       struct super_block *sb = osb->sb;
+
+       /* We mostly ignore errors in this function because there's not much
+        * we can do when we see them */
+       for (type = 0; type < MAXQUOTAS; type++) {
+               if (!sb_has_quota_loaded(sb, type))
+                       continue;
+               inode = igrab(sb->s_dquot.files[type]);
+               /* Turn off quotas. This will remove all dquot structures from
+                * memory and so they will be automatically synced to global
+                * quota files */
+               vfs_quota_disable(sb, type, DQUOT_USAGE_ENABLED |
+                                           DQUOT_LIMITS_ENABLED);
+               if (!inode)
+                       continue;
+               iput(inode);
+       }
+}
+
+/* Handle quota on quotactl */
+static int ocfs2_quota_on(struct super_block *sb, int type, int format_id,
+                         char *path, int remount)
+{
+       unsigned int feature[MAXQUOTAS] = { OCFS2_FEATURE_RO_COMPAT_USRQUOTA,
+                                            OCFS2_FEATURE_RO_COMPAT_GRPQUOTA};
+
+       if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb, feature[type]))
+               return -EINVAL;
+
+       if (remount)
+               return 0;       /* Just ignore it has been handled in
+                                * ocfs2_remount() */
+       return vfs_quota_enable(sb_dqopt(sb)->files[type], type,
+                                   format_id, DQUOT_LIMITS_ENABLED);
+}
+
+/* Handle quota off quotactl */
+static int ocfs2_quota_off(struct super_block *sb, int type, int remount)
+{
+       if (remount)
+               return 0;       /* Ignore now and handle later in
+                                * ocfs2_remount() */
+       return vfs_quota_disable(sb, type, DQUOT_LIMITS_ENABLED);
+}
+
+static struct quotactl_ops ocfs2_quotactl_ops = {
+       .quota_on       = ocfs2_quota_on,
+       .quota_off      = ocfs2_quota_off,
+       .quota_sync     = vfs_quota_sync,
+       .get_info       = vfs_get_dqinfo,
+       .set_info       = vfs_set_dqinfo,
+       .get_dqblk      = vfs_get_dqblk,
+       .set_dqblk      = vfs_set_dqblk,
+};
+
 static int ocfs2_fill_super(struct super_block *sb, void *data, int silent)
 {
        struct dentry *root;
@@ -689,6 +844,22 @@ static int ocfs2_fill_super(struct super_block *sb, void *data, int silent)
        osb->osb_commit_interval = parsed_options.commit_interval;
        osb->local_alloc_default_bits = ocfs2_megabytes_to_clusters(sb, parsed_options.localalloc_opt);
        osb->local_alloc_bits = osb->local_alloc_default_bits;
+       if (osb->s_mount_opt & OCFS2_MOUNT_USRQUOTA &&
+           !OCFS2_HAS_RO_COMPAT_FEATURE(sb,
+                                        OCFS2_FEATURE_RO_COMPAT_USRQUOTA)) {
+               status = -EINVAL;
+               mlog(ML_ERROR, "User quotas were requested, but this "
+                    "filesystem does not have the feature enabled.\n");
+               goto read_super_error;
+       }
+       if (osb->s_mount_opt & OCFS2_MOUNT_GRPQUOTA &&
+           !OCFS2_HAS_RO_COMPAT_FEATURE(sb,
+                                        OCFS2_FEATURE_RO_COMPAT_GRPQUOTA)) {
+               status = -EINVAL;
+               mlog(ML_ERROR, "Group quotas were requested, but this "
+                    "filesystem does not have the feature enabled.\n");
+               goto read_super_error;
+       }
 
        status = ocfs2_verify_userspace_stack(osb, &parsed_options);
        if (status)
@@ -793,6 +964,28 @@ static int ocfs2_fill_super(struct super_block *sb, void *data, int silent)
        atomic_set(&osb->vol_state, VOLUME_MOUNTED);
        wake_up(&osb->osb_mount_event);
 
+       /* Now we can initialize quotas because we can afford to wait
+        * for cluster locks recovery now. That also means that truncation
+        * log recovery can happen but that waits for proper quota setup */
+       if (!(sb->s_flags & MS_RDONLY)) {
+               status = ocfs2_enable_quotas(osb);
+               if (status < 0) {
+                       /* We have to err-out specially here because
+                        * s_root is already set */
+                       mlog_errno(status);
+                       atomic_set(&osb->vol_state, VOLUME_DISABLED);
+                       wake_up(&osb->osb_mount_event);
+                       mlog_exit(status);
+                       return status;
+               }
+       }
+
+       ocfs2_complete_quota_recovery(osb);
+
+       /* Now we wake up again for processes waiting for quotas */
+       atomic_set(&osb->vol_state, VOLUME_MOUNTED_QUOTAS);
+       wake_up(&osb->osb_mount_event);
+
        mlog_exit(status);
        return status;
 
@@ -980,6 +1173,28 @@ static int ocfs2_parse_options(struct super_block *sb,
                case Opt_inode64:
                        mopt->mount_opt |= OCFS2_MOUNT_INODE64;
                        break;
+               case Opt_usrquota:
+                       /* We check only on remount, otherwise features
+                        * aren't yet initialized. */
+                       if (is_remount && !OCFS2_HAS_RO_COMPAT_FEATURE(sb,
+                           OCFS2_FEATURE_RO_COMPAT_USRQUOTA)) {
+                               mlog(ML_ERROR, "User quota requested but "
+                                    "filesystem feature is not set\n");
+                               status = 0;
+                               goto bail;
+                       }
+                       mopt->mount_opt |= OCFS2_MOUNT_USRQUOTA;
+                       break;
+               case Opt_grpquota:
+                       if (is_remount && !OCFS2_HAS_RO_COMPAT_FEATURE(sb,
+                           OCFS2_FEATURE_RO_COMPAT_GRPQUOTA)) {
+                               mlog(ML_ERROR, "Group quota requested but "
+                                    "filesystem feature is not set\n");
+                               status = 0;
+                               goto bail;
+                       }
+                       mopt->mount_opt |= OCFS2_MOUNT_GRPQUOTA;
+                       break;
 #ifdef CONFIG_OCFS2_FS_POSIX_ACL
                case Opt_acl:
                        mopt->mount_opt |= OCFS2_MOUNT_POSIX_ACL;
@@ -1056,6 +1271,10 @@ static int ocfs2_show_options(struct seq_file *s, struct vfsmount *mnt)
        if (osb->osb_cluster_stack[0])
                seq_printf(s, ",cluster_stack=%.*s", OCFS2_STACK_LABEL_LEN,
                           osb->osb_cluster_stack);
+       if (opts & OCFS2_MOUNT_USRQUOTA)
+               seq_printf(s, ",usrquota");
+       if (opts & OCFS2_MOUNT_GRPQUOTA)
+               seq_printf(s, ",grpquota");
 
        if (opts & OCFS2_MOUNT_NOUSERXATTR)
                seq_printf(s, ",nouser_xattr");
@@ -1318,6 +1537,13 @@ 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;
 }
 
@@ -1394,6 +1620,8 @@ static void ocfs2_dismount_volume(struct super_block *sb, int mnt_err)
        osb = OCFS2_SB(sb);
        BUG_ON(!osb);
 
+       ocfs2_disable_quotas(osb);
+
        ocfs2_shutdown_local_alloc(osb);
 
        ocfs2_truncate_log_shutdown(osb);
@@ -1504,6 +1732,8 @@ static int ocfs2_initialize_super(struct super_block *sb,
        sb->s_fs_info = osb;
        sb->s_op = &ocfs2_sops;
        sb->s_export_op = &ocfs2_export_ops;
+       sb->s_qcop = &ocfs2_quotactl_ops;
+       sb->dq_op = &ocfs2_quota_operations;
        sb->s_xattr = ocfs2_xattr_handlers;
        sb->s_time_gran = 1;
        sb->s_flags |= MS_NOATIME;
@@ -1524,6 +1754,7 @@ 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);
@@ -1664,6 +1895,9 @@ 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);
@@ -1767,6 +2001,15 @@ 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 "
@@ -1808,6 +2051,7 @@ static int ocfs2_verify_volume(struct ocfs2_dinode *di,
                }
        }
 
+out:
        mlog_exit(status);
        return status;
 }