Re: [Ext4 punch hole 4/5 v6] Ext4 Punch Hole Support: Map blocks punch hole flag

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



On 5/3/2011 5:12 PM, Mingming Cao wrote:
On Tue, 2011-05-03 at 09:48 -0700, Allison Henderson wrote:
This patch adds a new flag to map blocks that specifies
the given range of blocks should be punched out.  Extents
are first converted to uninitialized extents before they are
punched out. Because punching a hole may require that the
extent be split, it is possible that the splitting may need
more blocks than are available.  To deal with this, reserved
blocks are temporarily borrowed to allow the split to proceed.
The borrowed blocks are then returned after the hole punch has
released more blocks.

Just want to make sure that in the case of ENOSPC, the reserved root
blocks are used/consumed to allow extent being splitted and then punch
hole being proceeded. The reserved blocks, if I understand right, is
being used, not returned back. No?

Mingming

Hi there,

That's correct, the last line of the description should have probably been removed to avoid confusion. To clarify for others, in an earlier version of the set before we had this flag, we had code to convert root blocks to user blocks when there was not enough blocks to split the extents, and then the blocks would be converted back after the punch hole had free'd up more user blocks. But there was concern that this would cause a race condition with other processes also trying to allocate user blocks, so we took that code out and added this flag instead. The flag only allows consumption of the root blocks, there is no logic for converting the blocks back anymore. The blocks are just returned to the pool normally when the remove space routines release them.

Allison Henderson



The routine then returns the number of blocks successfully
punched out.

Signed-off-by: Allison Henderson<achender@xxxxxxxxxx>
---
:100644 100644 0afb746... 7f0fcc1... M	fs/ext4/extents.c
  fs/ext4/extents.c |   98 +++++++++++++++++++++++++++++++++++++++++++++++------
  1 files changed, 87 insertions(+), 11 deletions(-)

diff --git a/fs/ext4/extents.c b/fs/ext4/extents.c
index 0afb746..7f0fcc1 100644
--- a/fs/ext4/extents.c
+++ b/fs/ext4/extents.c
@@ -3290,15 +3290,19 @@ int ext4_ext_map_blocks(handle_t *handle, struct inode *inode,
  	ext4_fsblk_t newblock = 0;
  	int err = 0, depth, ret;
  	unsigned int allocated = 0;
+	unsigned int punched_out = 0;
+	unsigned int result = 0;
  	struct ext4_allocation_request ar;
  	ext4_io_end_t *io = EXT4_I(inode)->cur_aio_dio;
+	struct ext4_map_blocks punch_map;

  	ext_debug("blocks %u/%u requested for inode %lu\n",
  		  map->m_lblk, map->m_len, inode->i_ino);
  	trace_ext4_ext_map_blocks_enter(inode, map->m_lblk, map->m_len, flags);

  	/* check in cache */
-	if (ext4_ext_in_cache(inode, map->m_lblk,&newex)) {
+	if (ext4_ext_in_cache(inode, map->m_lblk,&newex)&&
+		((flags&  EXT4_GET_BLOCKS_PUNCH_OUT_EXT) == 0)) {
  		if (!newex.ee_start_lo&&  !newex.ee_start_hi) {
  			if ((flags&  EXT4_GET_BLOCKS_CREATE) == 0) {
  				/*
@@ -3363,16 +3367,84 @@ int ext4_ext_map_blocks(handle_t *handle, struct inode *inode,
  			ext_debug("%u fit into %u:%d ->  %llu\n", map->m_lblk,
  				  ee_block, ee_len, newblock);

-			/* Do not put uninitialized extent in the cache */
-			if (!ext4_ext_is_uninitialized(ex)) {
-				ext4_ext_put_in_cache(inode, ee_block,
-							ee_len, ee_start);
-				goto out;
+			if ((flags&  EXT4_GET_BLOCKS_PUNCH_OUT_EXT) == 0) {
+				/*
+				 * Do not put uninitialized extent
+				 * in the cache
+				 */
+				if (!ext4_ext_is_uninitialized(ex)) {
+					ext4_ext_put_in_cache(inode, ee_block,
+						ee_len, ee_start);
+					goto out;
+				}
+				ret = ext4_ext_handle_uninitialized_extents(
+					handle, inode, map, path, flags,
+					allocated, newblock);
+				return ret;
  			}
-			ret = ext4_ext_handle_uninitialized_extents(handle,
-					inode, map, path, flags, allocated,
-					newblock);
-			return ret;
+
+			/*
+			 * Punch out the map length, but only to the
+			 * end of the extent
+			 */
+			punched_out = allocated<  map->m_len ?
+				allocated : map->m_len;
+
+			/*
+			 * Sense extents need to be converted to
+			 * uninitialized, they must fit in an
+			 * uninitialized extent
+			 */
+			if (punched_out>  EXT_UNINIT_MAX_LEN)
+				punched_out = EXT_UNINIT_MAX_LEN;
+
+			punch_map.m_lblk = map->m_lblk;
+			punch_map.m_pblk = newblock;
+			punch_map.m_len = punched_out;
+			punch_map.m_flags = 0;
+
+			/* Check to see if the extent needs to be split */
+			if (punch_map.m_len != ee_len ||
+				punch_map.m_lblk != ee_block) {
+
+				ret = ext4_split_extent(handle, inode,
+				path,&punch_map, 0,
+				EXT4_GET_BLOCKS_PUNCH_OUT_EXT |
+				EXT4_GET_BLOCKS_PRE_IO);
+
+				if (ret<  0) {
+					err = ret;
+					goto out2;
+				}
+				/*
+				 * find extent for the block at
+				 * the start of the hole
+				 */
+				ext4_ext_drop_refs(path);
+				kfree(path);
+
+				path = ext4_ext_find_extent(inode,
+				map->m_lblk, NULL);
+				if (IS_ERR(path)) {
+					err = PTR_ERR(path);
+					path = NULL;
+					goto out2;
+				}
+
+				depth = ext_depth(inode);
+				ex = path[depth].p_ext;
+				ee_len = ext4_ext_get_actual_len(ex);
+				ee_block = le32_to_cpu(ex->ee_block);
+				ee_start = ext4_ext_pblock(ex);
+
+			}
+
+			ext4_ext_mark_uninitialized(ex);
+
+			err = ext4_ext_remove_space(inode, map->m_lblk,
+				map->m_lblk + punched_out);
+
+			goto out2;
  		}
  	}

@@ -3517,7 +3589,11 @@ out2:
  	}
  	trace_ext4_ext_map_blocks_exit(inode, map->m_lblk,
  		newblock, map->m_len, err ? err : allocated);
-	return err ? err : allocated;
+
+	result = (flags&  EXT4_GET_BLOCKS_PUNCH_OUT_EXT) ?
+			punched_out : allocated;
+
+	return err ? err : result;
  }

  void ext4_ext_truncate(struct inode *inode)



--
To unsubscribe from this list: send the line "unsubscribe linux-ext4" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html


[Index of Archives]     [Reiser Filesystem Development]     [Ceph FS]     [Kernel Newbies]     [Security]     [Netfilter]     [Bugtraq]     [Linux FS]     [Yosemite National Park]     [MIPS Linux]     [ARM Linux]     [Linux Security]     [Linux RAID]     [Samba]     [Device Mapper]     [Linux Media]

  Powered by Linux