GFS2: Fix writing to non-page aligned gfs2_quota structures

This is the upstream fix for this bug. This patch differs
from the RHEL5 fix (Red Hat bz #555754) which simply writes to the 8-byte
value field of the quota. In upstream quota code, we're
required to write the entire quota (88 bytes) which can be split
across a page boundary. We check for such quotas, and read/write
the two parts from/to the corresponding pages holding these parts.

With this patch, I don't see the bug anymore using the reproducer
in Red Hat bz 555754. I successfully ran a couple of simple tests/mounts/
umounts and it doesn't seem like this patch breaks anything else.

Signed-off-by: Abhi Das <adas@redhat.com>
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>

diff --git a/fs/gfs2/quota.c b/fs/gfs2/quota.c
index 6ca0967..d5f4661 100644 (file)
--- a/fs/gfs2/quota.c
+++ b/fs/gfs2/quota.c
@@ -637,15 +637,40 @@ static int gfs2_adjust_quota(struct gfs2_inode *ip, loff_t loc,
        unsigned blocksize, iblock, pos;
        struct buffer_head *bh, *dibh;
        struct page *page;
-       void *kaddr;
-       struct gfs2_quota *qp;
-       s64 value;
-       int err = -EIO;
+       void *kaddr, *ptr;
+       struct gfs2_quota q, *qp;
+       int err, nbytes;
        u64 size;
 
        if (gfs2_is_stuffed(ip))
                gfs2_unstuff_dinode(ip, NULL);
-       
+
+       memset(&q, 0, sizeof(struct gfs2_quota));
+       err = gfs2_internal_read(ip, NULL, (char *)&q, &loc, sizeof(q));
+       if (err < 0)
+               return err;
+
+       err = -EIO;
+       qp = &q;
+       qp->qu_value = be64_to_cpu(qp->qu_value);
+       qp->qu_value += change;
+       qp->qu_value = cpu_to_be64(qp->qu_value);
+       qd->qd_qb.qb_value = qp->qu_value;
+       if (fdq) {
+               if (fdq->d_fieldmask & FS_DQ_BSOFT) {
+                       qp->qu_warn = cpu_to_be64(fdq->d_blk_softlimit);
+                       qd->qd_qb.qb_warn = qp->qu_warn;
+               }
+               if (fdq->d_fieldmask & FS_DQ_BHARD) {
+                       qp->qu_limit = cpu_to_be64(fdq->d_blk_hardlimit);
+                       qd->qd_qb.qb_limit = qp->qu_limit;
+               }
+       }
+
+       /* Write the quota into the quota file on disk */
+       ptr = qp;
+       nbytes = sizeof(struct gfs2_quota);
+get_a_page:
        page = grab_cache_page(mapping, index);
        if (!page)
                return -ENOMEM;
@@ -667,7 +692,12 @@ static int gfs2_adjust_quota(struct gfs2_inode *ip, loff_t loc,
        if (!buffer_mapped(bh)) {
                gfs2_block_map(inode, iblock, bh, 1);
                if (!buffer_mapped(bh))
-                       goto unlock;
+                       goto unlock_out;
+               /* If it's a newly allocated disk block for quota, zero it */
+               if (buffer_new(bh)) {
+                       memset(bh->b_data, 0, bh->b_size);
+                       set_buffer_uptodate(bh);
+               }
        }
 
        if (PageUptodate(page))
@@ -677,32 +707,34 @@ static int gfs2_adjust_quota(struct gfs2_inode *ip, loff_t loc,
                ll_rw_block(READ_META, 1, &bh);
                wait_on_buffer(bh);
                if (!buffer_uptodate(bh))
-                       goto unlock;
+                       goto unlock_out;
        }
 
        gfs2_trans_add_bh(ip->i_gl, bh, 0);
 
        kaddr = kmap_atomic(page, KM_USER0);
-       qp = kaddr + offset;
-       value = (s64)be64_to_cpu(qp->qu_value) + change;
-       qp->qu_value = cpu_to_be64(value);
-       qd->qd_qb.qb_value = qp->qu_value;
-       if (fdq) {
-               if (fdq->d_fieldmask & FS_DQ_BSOFT) {
-                       qp->qu_warn = cpu_to_be64(fdq->d_blk_softlimit);
-                       qd->qd_qb.qb_warn = qp->qu_warn;
-               }
-               if (fdq->d_fieldmask & FS_DQ_BHARD) {
-                       qp->qu_limit = cpu_to_be64(fdq->d_blk_hardlimit);
-                       qd->qd_qb.qb_limit = qp->qu_limit;
-               }
-       }
+       if (offset + sizeof(struct gfs2_quota) > PAGE_CACHE_SIZE)
+               nbytes = PAGE_CACHE_SIZE - offset;
+       memcpy(kaddr + offset, ptr, nbytes);
        flush_dcache_page(page);
        kunmap_atomic(kaddr, KM_USER0);
+       unlock_page(page);
+       page_cache_release(page);
+
+       /* If quota straddles page boundary, we need to update the rest of the
+        * quota at the beginning of the next page */
+       if (offset != 0) { /* first page, offset is closer to PAGE_CACHE_SIZE */
+               ptr = ptr + nbytes;
+               nbytes = sizeof(struct gfs2_quota) - nbytes;
+               offset = 0;
+               index++;
+               goto get_a_page;
+       }
 
+       /* Update the disk inode timestamp and size (if extended) */
        err = gfs2_meta_inode_buffer(ip, &dibh);
        if (err)
-               goto unlock;
+               goto out;
 
        size = loc + sizeof(struct gfs2_quota);
        if (size > inode->i_size) {
@@ -715,7 +747,9 @@ static int gfs2_adjust_quota(struct gfs2_inode *ip, loff_t loc,
        brelse(dibh);
        mark_inode_dirty(inode);
 
-unlock:
+out:
+       return err;
+unlock_out:
        unlock_page(page);
        page_cache_release(page);
        return err;
@@ -779,8 +813,10 @@ static int do_sync(unsigned int num_qd, struct gfs2_quota_data **qda)
         * rgrp since it won't be allocated during the transaction
         */
        al->al_requested = 1;
-       /* +1 in the end for block requested above for unstuffing */
-       blocks = num_qd * data_blocks + RES_DINODE + num_qd + 1;
+       /* +3 in the end for unstuffing block, inode size update block
+        * and another block in case quota straddles page boundary and 
+        * two blocks need to be updated instead of 1 */
+       blocks = num_qd * data_blocks + RES_DINODE + num_qd + 3;
 
        if (nalloc)
                al->al_requested += nalloc * (data_blocks + ind_blocks);