LU-5393 osd-ldiskfs: read i_size once to protect against race

[fs/lustre-release.git] / lustre / osd-ldiskfs / osd_io.c

diff --git a/lustre/osd-ldiskfs/osd_io.c b/lustre/osd-ldiskfs/osd_io.c
index eae5c00..45822ff 100644 (file)
--- a/lustre/osd-ldiskfs/osd_io.c
+++ b/lustre/osd-ldiskfs/osd_io.c
@@ -27,7 +27,7 @@
  * Copyright (c) 2010, Oracle and/or its affiliates. All rights reserved.
  * Use is subject to license terms.
  *
- * Copyright (c) 2012, 2013, Intel Corporation.
+ * Copyright (c) 2012, 2014, Intel Corporation.
  */
 /*
  * This file is part of Lustre, http://www.lustre.org/
@@ -790,10 +790,17 @@ static int osd_ldiskfs_map_bm_inode_pages(struct inode *inode,
                                          unsigned long *blocks, int create)
 {
        int blocks_per_page = PAGE_CACHE_SIZE >> inode->i_blkbits;
+       pgoff_t bitmap_max_page_index;
        unsigned long *b;
        int rc = 0, i;
 
+       bitmap_max_page_index = LDISKFS_SB(inode->i_sb)->s_bitmap_maxbytes >>
+                               PAGE_SHIFT;
        for (i = 0, b = blocks; i < pages; i++, page++) {
+               if ((*page)->index + 1 >= bitmap_max_page_index) {
+                       rc = -EFBIG;
+                       break;
+               }
                rc = ldiskfs_map_inode_page(inode, *page, b, create);
                if (rc) {
                        CERROR("ino %lu, blk %lu create %d: rc %d\n",
@@ -814,6 +821,9 @@ static int osd_ldiskfs_map_ext_inode_pages(struct inode *inode,
        int rc = 0, i = 0;
        struct page *fp = NULL;
        int clen = 0;
+       pgoff_t extent_max_page_index;
+
+       extent_max_page_index = inode->i_sb->s_maxbytes >> PAGE_SHIFT;
 
        CDEBUG(D_OTHER, "inode %lu: map %d pages from %lu\n",
                inode->i_ino, pages, (*page)->index);
@@ -835,6 +845,9 @@ static int osd_ldiskfs_map_ext_inode_pages(struct inode *inode,
                        continue;
                }
 
+               if (fp->index + i >= extent_max_page_index)
+                       GOTO(cleanup, rc = -EFBIG);
+
                /* process found extent */
                rc = osd_ldiskfs_map_nblocks(inode, fp->index * blocks_per_page,
                                             clen * blocks_per_page, blocks,
@@ -879,6 +892,9 @@ static int osd_ldiskfs_map_inode_pages(struct inode *inode, struct page **page,
        int rc = 0, i = 0;
        struct page *fp = NULL;
        int clen = 0;
+       pgoff_t max_page_index;
+
+       max_page_index = inode->i_sb->s_maxbytes >> PAGE_SHIFT;
 
        CDEBUG(D_OTHER, "inode %lu: map %d pages from %lu\n",
                inode->i_ino, pages, (*page)->index);
@@ -902,6 +918,8 @@ static int osd_ldiskfs_map_inode_pages(struct inode *inode, struct page **page,
                        if (++i != pages)
                                continue;
                }
+               if (fp->index + i >= max_page_index)
+                       GOTO(cleanup, rc = -EFBIG);
                /* process found extent */
                map.m_lblk = fp->index * blocks_per_page;
                map.m_len = blen = clen * blocks_per_page;
@@ -1031,7 +1049,7 @@ static int osd_write_prep(const struct lu_env *env, struct dt_object *dt,
 }
 
 /* Check if a block is allocated or not */
-static int osd_is_mapped(struct inode *inode, obd_size offset)
+static int osd_is_mapped(struct inode *inode, u64 offset)
 {
        sector_t (*fs_bmap)(struct address_space *, sector_t);
 
@@ -1267,7 +1285,8 @@ static int osd_read_prep(const struct lu_env *env, struct dt_object *dt,
         struct osd_device *osd = osd_obj2dev(osd_dt_obj(dt));
         struct timeval start, end;
         unsigned long timediff;
-       int rc = 0, i, m = 0, cache = 0, cache_hits = 0, cache_misses = 0;
+       int rc = 0, i, cache = 0, cache_hits = 0, cache_misses = 0;
+       loff_t isize;
 
         LASSERT(inode);
 
@@ -1275,26 +1294,25 @@ static int osd_read_prep(const struct lu_env *env, struct dt_object *dt,
        if (unlikely(rc != 0))
                RETURN(rc);
 
+       isize = i_size_read(inode);
+
        if (osd->od_read_cache)
                cache = 1;
-       if (i_size_read(inode) > osd->od_readcache_max_filesize)
+       if (isize > osd->od_readcache_max_filesize)
                cache = 0;
 
        do_gettimeofday(&start);
        for (i = 0; i < npages; i++) {
 
-               if (i_size_read(inode) <= lnb[i].lnb_file_offset)
+               if (isize <= lnb[i].lnb_file_offset)
                        /* If there's no more data, abort early.
                         * lnb->lnb_rc == 0, so it's easy to detect later. */
                        break;
 
-               if (i_size_read(inode) <
-                   lnb[i].lnb_file_offset + lnb[i].lnb_len - 1)
-                       lnb[i].lnb_rc = i_size_read(inode) -
-                               lnb[i].lnb_file_offset;
+               if (isize < lnb[i].lnb_file_offset + lnb[i].lnb_len - 1)
+                       lnb[i].lnb_rc = isize - lnb[i].lnb_file_offset;
                else
                        lnb[i].lnb_rc = lnb[i].lnb_len;
-               m += lnb[i].lnb_len;
 
                if (PageUptodate(lnb[i].lnb_page)) {
                        cache_hits++;
@@ -1640,7 +1658,7 @@ int osd_ldiskfs_write_record(struct inode *inode, void *buf, int bufsize,
                         "boffs %d size %d bh->b_size %lu\n",
                         boffs, size, (unsigned long)bh->b_size);
                 memcpy(bh->b_data + boffs, buf, size);
-                err = ldiskfs_journal_dirty_metadata(handle, bh);
+               err = ldiskfs_handle_dirty_metadata(handle, NULL, bh);
                 if (err)
                         break;