3 files changed, 77 insertions, 35 deletions

diff --git a/fs/ext4/ext4.h b/fs/ext4/ext4.h
index af7b62699ea9..874d169a193e 100644
--- a/fs/ext4/ext4.h
+++ b/fs/ext4/ext4.h
@@ -361,14 +361,11 @@ struct ext4_new_group_data {
            so set the magic i_delalloc_reserve_flag after taking the 
            inode allocation semaphore for */
 #define EXT4_GET_BLOCKS_DELALLOC_RESERVE        0x0004
-        /* Call ext4_da_update_reserve_space() after successfully 
-           allocating the blocks */
-#define EXT4_GET_BLOCKS_UPDATE_RESERVE_SPACE    0x0008
         /* caller is from the direct IO path, request to creation of an
         unitialized extents if not allocated, split the uninitialized
         extent if blocks has been preallocated already*/
-#define EXT4_GET_BLOCKS_DIO                     0x0010
-#define EXT4_GET_BLOCKS_CONVERT                 0x0020
+#define EXT4_GET_BLOCKS_DIO                     0x0008
+#define EXT4_GET_BLOCKS_CONVERT                 0x0010
 #define EXT4_GET_BLOCKS_DIO_CREATE_EXT          (EXT4_GET_BLOCKS_DIO|\
                                          EXT4_GET_BLOCKS_CREATE_UNINIT_EXT)
         /* Convert extent to initialized after direct IO complete */
@@ -1443,6 +1440,8 @@ extern int ext4_block_truncate_page(handle_t *handle,
 extern int ext4_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf);
 extern qsize_t *ext4_get_reserved_space(struct inode *inode);
 extern int flush_aio_dio_completed_IO(struct inode *inode);
+extern void ext4_da_update_reserve_space(struct inode *inode,
+                                        int used, int quota_claim);
 /* ioctl.c */
 extern long ext4_ioctl(struct file *, unsigned int, unsigned long);
 extern long ext4_compat_ioctl(struct file *, unsigned int, unsigned long);
diff --git a/fs/ext4/extents.c b/fs/ext4/extents.c
index 7d7b74e94687..765a4826b118 100644
--- a/fs/ext4/extents.c
+++ b/fs/ext4/extents.c
@@ -3132,7 +3132,19 @@ out:
                 unmap_underlying_metadata_blocks(inode->i_sb->s_bdev,
                                         newblock + max_blocks,
                                         allocated - max_blocks);
+                allocated = max_blocks;
         }
+
+        /*
+         * If we have done fallocate with the offset that is already
+         * delayed allocated, we would have block reservation
+         * and quota reservation done in the delayed write path.
+         * But fallocate would have already updated quota and block
+         * count for this offset. So cancel these reservation
+         */
+        if (flags & EXT4_GET_BLOCKS_DELALLOC_RESERVE)
+                ext4_da_update_reserve_space(inode, allocated, 0);
+
 map_out:
         set_buffer_mapped(bh_result);
 out1:
@@ -3368,9 +3380,18 @@ int ext4_ext_get_blocks(handle_t *handle, struct inode *inode,
         /* previous routine could use block we allocated */
         newblock = ext_pblock(&newex);
         allocated = ext4_ext_get_actual_len(&newex);
+        if (allocated > max_blocks)
+                allocated = max_blocks;
         set_buffer_new(bh_result);
 
         /*
+         * Update reserved blocks/metadata blocks after successful
+         * block allocation which had been deferred till now.
+         */
+        if (flags & EXT4_GET_BLOCKS_DELALLOC_RESERVE)
+                ext4_da_update_reserve_space(inode, allocated, 1);
+
+        /*
          * Cache the extent and update transaction to commit on fdatasync only
          * when it is _not_ an uninitialized extent.
          */
diff --git a/fs/ext4/inode.c b/fs/ext4/inode.c
index c818972c8302..e11952404e02 100644
--- a/fs/ext4/inode.c
+++ b/fs/ext4/inode.c
@@ -1053,11 +1053,12 @@ static int ext4_calc_metadata_amount(struct inode *inode, sector_t lblock)
  * Called with i_data_sem down, which is important since we can call
  * ext4_discard_preallocations() from here.
  */
-static void ext4_da_update_reserve_space(struct inode *inode, int used)
+void ext4_da_update_reserve_space(struct inode *inode,
+                                        int used, int quota_claim)
 {
         struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb);
         struct ext4_inode_info *ei = EXT4_I(inode);
-        int mdb_free = 0;
+        int mdb_free = 0, allocated_meta_blocks = 0;
 
         spin_lock(&ei->i_block_reservation_lock);
         if (unlikely(used > ei->i_reserved_data_blocks)) {
@@ -1073,6 +1074,7 @@ static void ext4_da_update_reserve_space(struct inode *inode, int used)
         ei->i_reserved_data_blocks -= used;
         used += ei->i_allocated_meta_blocks;
         ei->i_reserved_meta_blocks -= ei->i_allocated_meta_blocks;
+        allocated_meta_blocks = ei->i_allocated_meta_blocks;
         ei->i_allocated_meta_blocks = 0;
         percpu_counter_sub(&sbi->s_dirtyblocks_counter, used);
 
@@ -1090,9 +1092,23 @@ static void ext4_da_update_reserve_space(struct inode *inode, int used)
         spin_unlock(&EXT4_I(inode)->i_block_reservation_lock);
 
         /* Update quota subsystem */
-        vfs_dq_claim_block(inode, used);
-        if (mdb_free)
-                vfs_dq_release_reservation_block(inode, mdb_free);
+        if (quota_claim) {
+                vfs_dq_claim_block(inode, used);
+                if (mdb_free)
+                        vfs_dq_release_reservation_block(inode, mdb_free);
+        } else {
+                /*
+                 * We did fallocate with an offset that is already delayed
+                 * allocated. So on delayed allocated writeback we should
+                 * not update the quota for allocated blocks. But then
+                 * converting an fallocate region to initialized region would
+                 * have caused a metadata allocation. So claim quota for
+                 * that
+                 */
+                if (allocated_meta_blocks)
+                        vfs_dq_claim_block(inode, allocated_meta_blocks);
+                vfs_dq_release_reservation_block(inode, mdb_free + used);
+        }
 
         /*
          * If we have done all the pending block allocations and if
@@ -1292,18 +1308,20 @@ int ext4_get_blocks(handle_t *handle, struct inode *inode, sector_t block,
                          */
                         EXT4_I(inode)->i_state &= ~EXT4_STATE_EXT_MIGRATE;
                 }
-        }
 
+                /*
+                 * Update reserved blocks/metadata blocks after successful
+                 * block allocation which had been deferred till now. We don't
+                 * support fallocate for non extent files. So we can update
+                 * reserve space here.
+                 */
+                if ((retval > 0) &&
+                        (flags & EXT4_GET_BLOCKS_DELALLOC_RESERVE))
+                        ext4_da_update_reserve_space(inode, retval, 1);
+        }
         if (flags & EXT4_GET_BLOCKS_DELALLOC_RESERVE)
                 EXT4_I(inode)->i_delalloc_reserved_flag = 0;
 
-        /*
-         * Update reserved blocks/metadata blocks after successful
-         * block allocation which had been deferred till now.
-         */
-        if ((retval > 0) && (flags & EXT4_GET_BLOCKS_UPDATE_RESERVE_SPACE))
-                ext4_da_update_reserve_space(inode, retval);
-
         up_write((&EXT4_I(inode)->i_data_sem));
         if (retval > 0 && buffer_mapped(bh)) {
                 int ret = check_block_validity(inode, "file system "
@@ -1835,24 +1853,12 @@ repeat:
          * later. Real quota accounting is done at pages writeout
          * time.
          */
-        if (vfs_dq_reserve_block(inode, md_needed + 1)) {
-                /* 
-                 * We tend to badly over-estimate the amount of
-                 * metadata blocks which are needed, so if we have
-                 * reserved any metadata blocks, try to force out the
-                 * inode and see if we have any better luck.
-                 */
-                if (md_reserved && retries++ <= 3)
-                        goto retry;
+        if (vfs_dq_reserve_block(inode, md_needed + 1))
                 return -EDQUOT;
-        }
 
         if (ext4_claim_free_blocks(sbi, md_needed + 1)) {
                 vfs_dq_release_reservation_block(inode, md_needed + 1);
                 if (ext4_should_retry_alloc(inode->i_sb, &retries)) {
-                retry:
-                        if (md_reserved)
-                                write_inode_now(inode, (retries == 3));
                         yield();
                         goto repeat;
                 }
@@ -2213,10 +2219,10 @@ static int mpage_da_map_blocks(struct mpage_da_data *mpd)
          * variables are updated after the blocks have been allocated.
          */
         new.b_state = 0;
-        get_blocks_flags = (EXT4_GET_BLOCKS_CREATE |
-                            EXT4_GET_BLOCKS_DELALLOC_RESERVE);
+        get_blocks_flags = EXT4_GET_BLOCKS_CREATE;
         if (mpd->b_state & (1 << BH_Delay))
-                get_blocks_flags |= EXT4_GET_BLOCKS_UPDATE_RESERVE_SPACE;
+                get_blocks_flags |= EXT4_GET_BLOCKS_DELALLOC_RESERVE;
+
         blks = ext4_get_blocks(handle, mpd->inode, next, max_blocks,
                                &new, get_blocks_flags);
         if (blks < 0) {
@@ -3032,7 +3038,7 @@ static int ext4_da_write_begin(struct file *file, struct address_space *mapping,
                                loff_t pos, unsigned len, unsigned flags,
                                struct page **pagep, void **fsdata)
 {
-        int ret, retries = 0;
+        int ret, retries = 0, quota_retries = 0;
         struct page *page;
         pgoff_t index;
         unsigned from, to;
@@ -3091,6 +3097,22 @@ retry:
 
         if (ret == -ENOSPC && ext4_should_retry_alloc(inode->i_sb, &retries))
                 goto retry;
+
+        if ((ret == -EDQUOT) &&
+            EXT4_I(inode)->i_reserved_meta_blocks &&
+            (quota_retries++ < 3)) {
+                /*
+                 * Since we often over-estimate the number of meta
+                 * data blocks required, we may sometimes get a
+                 * spurios out of quota error even though there would
+                 * be enough space once we write the data blocks and
+                 * find out how many meta data blocks were _really_
+                 * required.  So try forcing the inode write to see if
+                 * that helps.
+                 */
+                write_inode_now(inode, (quota_retries == 3));
+                goto retry;
+        }
 out:
         return ret;
 }