ocfs2: check/fix inode block for online file check
authorGang He <ghe@suse.com>
Tue, 22 Mar 2016 21:24:30 +0000 (14:24 -0700)
committerLinus Torvalds <torvalds@linux-foundation.org>
Tue, 22 Mar 2016 22:36:02 +0000 (15:36 -0700)
Implement online check or fix inode block during reading a inode block
to memory.

Signed-off-by: Gang He <ghe@suse.com>
Reviewed-by: Mark Fasheh <mfasheh@suse.de>
Cc: Joel Becker <jlbec@evilplan.org>
Cc: Junxiao Bi <junxiao.bi@oracle.com>
Cc: Joseph Qi <joseph.qi@huawei.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
fs/ocfs2/inode.c
fs/ocfs2/ocfs2_trace.h

index 36294446d9607ee9f0d07814552fd6f0bf97e6ab..ba495beff1c2b25dfccd4641db15de68ddb1d721 100644 (file)
@@ -53,6 +53,7 @@
 #include "xattr.h"
 #include "refcounttree.h"
 #include "ocfs2_trace.h"
+#include "filecheck.h"
 
 #include "buffer_head_io.h"
 
@@ -74,6 +75,14 @@ static int ocfs2_truncate_for_delete(struct ocfs2_super *osb,
                                    struct inode *inode,
                                    struct buffer_head *fe_bh);
 
+static int ocfs2_filecheck_read_inode_block_full(struct inode *inode,
+                                                struct buffer_head **bh,
+                                                int flags, int type);
+static int ocfs2_filecheck_validate_inode_block(struct super_block *sb,
+                                               struct buffer_head *bh);
+static int ocfs2_filecheck_repair_inode_block(struct super_block *sb,
+                                             struct buffer_head *bh);
+
 void ocfs2_set_inode_flags(struct inode *inode)
 {
        unsigned int flags = OCFS2_I(inode)->ip_attr;
@@ -127,6 +136,7 @@ struct inode *ocfs2_ilookup(struct super_block *sb, u64 blkno)
 struct inode *ocfs2_iget(struct ocfs2_super *osb, u64 blkno, unsigned flags,
                         int sysfile_type)
 {
+       int rc = 0;
        struct inode *inode = NULL;
        struct super_block *sb = osb->sb;
        struct ocfs2_find_inode_args args;
@@ -161,12 +171,17 @@ struct inode *ocfs2_iget(struct ocfs2_super *osb, u64 blkno, unsigned flags,
        }
        trace_ocfs2_iget5_locked(inode->i_state);
        if (inode->i_state & I_NEW) {
-               ocfs2_read_locked_inode(inode, &args);
+               rc = ocfs2_read_locked_inode(inode, &args);
                unlock_new_inode(inode);
        }
        if (is_bad_inode(inode)) {
                iput(inode);
-               inode = ERR_PTR(-ESTALE);
+               if ((flags & OCFS2_FI_FLAG_FILECHECK_CHK) ||
+                   (flags & OCFS2_FI_FLAG_FILECHECK_FIX))
+                       /* Return OCFS2_FILECHECK_ERR_XXX related errno */
+                       inode = ERR_PTR(rc);
+               else
+                       inode = ERR_PTR(-ESTALE);
                goto bail;
        }
 
@@ -410,7 +425,7 @@ static int ocfs2_read_locked_inode(struct inode *inode,
        struct ocfs2_super *osb;
        struct ocfs2_dinode *fe;
        struct buffer_head *bh = NULL;
-       int status, can_lock;
+       int status, can_lock, lock_level = 0;
        u32 generation = 0;
 
        status = -EINVAL;
@@ -478,7 +493,7 @@ static int ocfs2_read_locked_inode(struct inode *inode,
                        mlog_errno(status);
                        return status;
                }
-               status = ocfs2_inode_lock(inode, NULL, 0);
+               status = ocfs2_inode_lock(inode, NULL, lock_level);
                if (status) {
                        make_bad_inode(inode);
                        mlog_errno(status);
@@ -495,16 +510,32 @@ static int ocfs2_read_locked_inode(struct inode *inode,
        }
 
        if (can_lock) {
-               status = ocfs2_read_inode_block_full(inode, &bh,
-                                                    OCFS2_BH_IGNORE_CACHE);
+               if (args->fi_flags & OCFS2_FI_FLAG_FILECHECK_CHK)
+                       status = ocfs2_filecheck_read_inode_block_full(inode,
+                                               &bh, OCFS2_BH_IGNORE_CACHE, 0);
+               else if (args->fi_flags & OCFS2_FI_FLAG_FILECHECK_FIX)
+                       status = ocfs2_filecheck_read_inode_block_full(inode,
+                                               &bh, OCFS2_BH_IGNORE_CACHE, 1);
+               else
+                       status = ocfs2_read_inode_block_full(inode,
+                                               &bh, OCFS2_BH_IGNORE_CACHE);
        } else {
                status = ocfs2_read_blocks_sync(osb, args->fi_blkno, 1, &bh);
                /*
                 * If buffer is in jbd, then its checksum may not have been
                 * computed as yet.
                 */
-               if (!status && !buffer_jbd(bh))
-                       status = ocfs2_validate_inode_block(osb->sb, bh);
+               if (!status && !buffer_jbd(bh)) {
+                       if (args->fi_flags & OCFS2_FI_FLAG_FILECHECK_CHK)
+                               status = ocfs2_filecheck_validate_inode_block(
+                                                               osb->sb, bh);
+                       else if (args->fi_flags & OCFS2_FI_FLAG_FILECHECK_FIX)
+                               status = ocfs2_filecheck_repair_inode_block(
+                                                               osb->sb, bh);
+                       else
+                               status = ocfs2_validate_inode_block(
+                                                               osb->sb, bh);
+               }
        }
        if (status < 0) {
                mlog_errno(status);
@@ -532,11 +563,24 @@ static int ocfs2_read_locked_inode(struct inode *inode,
 
        BUG_ON(args->fi_blkno != le64_to_cpu(fe->i_blkno));
 
+       if (buffer_dirty(bh) && !buffer_jbd(bh)) {
+               if (can_lock) {
+                       ocfs2_inode_unlock(inode, lock_level);
+                       lock_level = 1;
+                       ocfs2_inode_lock(inode, NULL, lock_level);
+               }
+               status = ocfs2_write_block(osb, bh, INODE_CACHE(inode));
+               if (status < 0) {
+                       mlog_errno(status);
+                       goto bail;
+               }
+       }
+
        status = 0;
 
 bail:
        if (can_lock)
-               ocfs2_inode_unlock(inode, 0);
+               ocfs2_inode_unlock(inode, lock_level);
 
        if (status < 0)
                make_bad_inode(inode);
@@ -1397,6 +1441,169 @@ bail:
        return rc;
 }
 
+static int ocfs2_filecheck_validate_inode_block(struct super_block *sb,
+                                               struct buffer_head *bh)
+{
+       int rc = 0;
+       struct ocfs2_dinode *di = (struct ocfs2_dinode *)bh->b_data;
+
+       trace_ocfs2_filecheck_validate_inode_block(
+               (unsigned long long)bh->b_blocknr);
+
+       BUG_ON(!buffer_uptodate(bh));
+
+       /*
+        * Call ocfs2_validate_meta_ecc() first since it has ecc repair
+        * function, but we should not return error immediately when ecc
+        * validation fails, because the reason is quite likely the invalid
+        * inode number inputed.
+        */
+       rc = ocfs2_validate_meta_ecc(sb, bh->b_data, &di->i_check);
+       if (rc) {
+               mlog(ML_ERROR,
+                    "Filecheck: checksum failed for dinode %llu\n",
+                    (unsigned long long)bh->b_blocknr);
+               rc = -OCFS2_FILECHECK_ERR_BLOCKECC;
+       }
+
+       if (!OCFS2_IS_VALID_DINODE(di)) {
+               mlog(ML_ERROR,
+                    "Filecheck: invalid dinode #%llu: signature = %.*s\n",
+                    (unsigned long long)bh->b_blocknr, 7, di->i_signature);
+               rc = -OCFS2_FILECHECK_ERR_INVALIDINO;
+               goto bail;
+       } else if (rc)
+               goto bail;
+
+       if (le64_to_cpu(di->i_blkno) != bh->b_blocknr) {
+               mlog(ML_ERROR,
+                    "Filecheck: invalid dinode #%llu: i_blkno is %llu\n",
+                    (unsigned long long)bh->b_blocknr,
+                    (unsigned long long)le64_to_cpu(di->i_blkno));
+               rc = -OCFS2_FILECHECK_ERR_BLOCKNO;
+               goto bail;
+       }
+
+       if (!(di->i_flags & cpu_to_le32(OCFS2_VALID_FL))) {
+               mlog(ML_ERROR,
+                    "Filecheck: invalid dinode #%llu: OCFS2_VALID_FL "
+                    "not set\n",
+                    (unsigned long long)bh->b_blocknr);
+               rc = -OCFS2_FILECHECK_ERR_VALIDFLAG;
+               goto bail;
+       }
+
+       if (le32_to_cpu(di->i_fs_generation) !=
+           OCFS2_SB(sb)->fs_generation) {
+               mlog(ML_ERROR,
+                    "Filecheck: invalid dinode #%llu: fs_generation is %u\n",
+                    (unsigned long long)bh->b_blocknr,
+                    le32_to_cpu(di->i_fs_generation));
+               rc = -OCFS2_FILECHECK_ERR_GENERATION;
+               goto bail;
+       }
+
+bail:
+       return rc;
+}
+
+static int ocfs2_filecheck_repair_inode_block(struct super_block *sb,
+                                             struct buffer_head *bh)
+{
+       int changed = 0;
+       struct ocfs2_dinode *di = (struct ocfs2_dinode *)bh->b_data;
+
+       if (!ocfs2_filecheck_validate_inode_block(sb, bh))
+               return 0;
+
+       trace_ocfs2_filecheck_repair_inode_block(
+               (unsigned long long)bh->b_blocknr);
+
+       if (ocfs2_is_hard_readonly(OCFS2_SB(sb)) ||
+           ocfs2_is_soft_readonly(OCFS2_SB(sb))) {
+               mlog(ML_ERROR,
+                    "Filecheck: cannot repair dinode #%llu "
+                    "on readonly filesystem\n",
+                    (unsigned long long)bh->b_blocknr);
+               return -OCFS2_FILECHECK_ERR_READONLY;
+       }
+
+       if (buffer_jbd(bh)) {
+               mlog(ML_ERROR,
+                    "Filecheck: cannot repair dinode #%llu, "
+                    "its buffer is in jbd\n",
+                    (unsigned long long)bh->b_blocknr);
+               return -OCFS2_FILECHECK_ERR_INJBD;
+       }
+
+       if (!OCFS2_IS_VALID_DINODE(di)) {
+               /* Cannot fix invalid inode block */
+               return -OCFS2_FILECHECK_ERR_INVALIDINO;
+       }
+
+       if (!(di->i_flags & cpu_to_le32(OCFS2_VALID_FL))) {
+               /* Cannot just add VALID_FL flag back as a fix,
+                * need more things to check here.
+                */
+               return -OCFS2_FILECHECK_ERR_VALIDFLAG;
+       }
+
+       if (le64_to_cpu(di->i_blkno) != bh->b_blocknr) {
+               di->i_blkno = cpu_to_le64(bh->b_blocknr);
+               changed = 1;
+               mlog(ML_ERROR,
+                    "Filecheck: reset dinode #%llu: i_blkno to %llu\n",
+                    (unsigned long long)bh->b_blocknr,
+                    (unsigned long long)le64_to_cpu(di->i_blkno));
+       }
+
+       if (le32_to_cpu(di->i_fs_generation) !=
+           OCFS2_SB(sb)->fs_generation) {
+               di->i_fs_generation = cpu_to_le32(OCFS2_SB(sb)->fs_generation);
+               changed = 1;
+               mlog(ML_ERROR,
+                    "Filecheck: reset dinode #%llu: fs_generation to %u\n",
+                    (unsigned long long)bh->b_blocknr,
+                    le32_to_cpu(di->i_fs_generation));
+       }
+
+       if (changed || ocfs2_validate_meta_ecc(sb, bh->b_data, &di->i_check)) {
+               ocfs2_compute_meta_ecc(sb, bh->b_data, &di->i_check);
+               mark_buffer_dirty(bh);
+               mlog(ML_ERROR,
+                    "Filecheck: reset dinode #%llu: compute meta ecc\n",
+                    (unsigned long long)bh->b_blocknr);
+       }
+
+       return 0;
+}
+
+static int
+ocfs2_filecheck_read_inode_block_full(struct inode *inode,
+                                     struct buffer_head **bh,
+                                     int flags, int type)
+{
+       int rc;
+       struct buffer_head *tmp = *bh;
+
+       if (!type) /* Check inode block */
+               rc = ocfs2_read_blocks(INODE_CACHE(inode),
+                               OCFS2_I(inode)->ip_blkno,
+                               1, &tmp, flags,
+                               ocfs2_filecheck_validate_inode_block);
+       else /* Repair inode block */
+               rc = ocfs2_read_blocks(INODE_CACHE(inode),
+                               OCFS2_I(inode)->ip_blkno,
+                               1, &tmp, flags,
+                               ocfs2_filecheck_repair_inode_block);
+
+       /* If ocfs2_read_blocks() got us a new bh, pass it up. */
+       if (!rc && !*bh)
+               *bh = tmp;
+
+       return rc;
+}
+
 int ocfs2_read_inode_block_full(struct inode *inode, struct buffer_head **bh,
                                int flags)
 {
index a52a2dbc064e2d87919f934504fb1f156c73d176..24b7e7f591dccc91a16df3d98a56c34f4e062803 100644 (file)
@@ -1540,6 +1540,8 @@ DEFINE_OCFS2_ULL_INT_EVENT(ocfs2_read_locked_inode);
 DEFINE_OCFS2_INT_INT_EVENT(ocfs2_check_orphan_recovery_state);
 
 DEFINE_OCFS2_ULL_EVENT(ocfs2_validate_inode_block);
+DEFINE_OCFS2_ULL_EVENT(ocfs2_filecheck_validate_inode_block);
+DEFINE_OCFS2_ULL_EVENT(ocfs2_filecheck_repair_inode_block);
 
 TRACE_EVENT(ocfs2_inode_is_valid_to_delete,
        TP_PROTO(void *task, void *dc_task, unsigned long long ino,