Re: [PATCH v4 3/8] ext4: introduce new callback for IOMAP_REPORT operations

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

 





On 10/9/19 5:38 PM, Matthew Bobrowski wrote:
On Wed, Oct 09, 2019 at 11:30:21AM +0530, Ritesh Harjani wrote:
+static u16 ext4_iomap_check_delalloc(struct inode *inode,
+				     struct ext4_map_blocks *map)
+{
+	struct extent_status es;
+	ext4_lblk_t end = map->m_lblk + map->m_len - 1;
+
+	ext4_es_find_extent_range(inode, &ext4_es_is_delayed, map->m_lblk,
+				  end, &es);
+
+	/* Entire range is a hole */
+	if (!es.es_len || es.es_lblk > end)
+		return IOMAP_HOLE;
+	if (es.es_lblk <= map->m_lblk) {
+		ext4_lblk_t offset = 0;
+
+		if (es.es_lblk < map->m_lblk)
+			offset = map->m_lblk - es.es_lblk;
+		map->m_lblk = es.es_lblk + offset;
This looks redundant no? map->m_lblk never changes actually.
So this is not needed here.

Well, it depends if map->m_lblk == es.es_lblk + offset prior to the
assignment? If that's always true, then sure, it'd be redundant. But
honestly, I don't know what the downstream effect would be if this was
removed. I'd have to look at the code, perform some tests, and figure
it out.

<snip>
3334         if (es.es_lblk <= map->m_lblk) {
3335                 ext4_lblk_t offset = 0;
3336
3337                 if (es.es_lblk < map->m_lblk)
3338                         offset = map->m_lblk - es.es_lblk;
3339                 map->m_lblk = es.es_lblk + offset;
3340                 map->m_len = es.es_len - offset;
3341                 return IOMAP_DELALLOC;
3342         }

I saw it this way-

In condition "if (es.es_lblk <= map->m_lblk)" there are 2 cases.

Case 1: es.es_lblk is equal to map->m_lblk (equality)
   For this case, "offset" will remain 0.
   So map->lblk = es.es_lblk + 0 (but since es.es_lblk is same as
map->m_lblk in equality case, so it is redundant).


Case 2: es.es_lblk < map->m_lblk (less than)
In this case "offset = map->m_lblk - es.es_lblk"
Now replacing this val of offset in "map->m_lblk = es.es_lblk + offset"
map->m_lblk = es.es_lblk + map->m_lblk - es.es_lblk
which again is map->m_lblk = map->m_lblk - again redundant.


What did I miss?
But sure feel free to test as per your convenience.



+	map.m_lblk = first_block;
+	map.m_len = last_block = first_block + 1;
+	ret = ext4_map_blocks(NULL, inode, &map, 0);
+	if (ret < 0)
+		return ret;
+	if (ret == 0)
+		type = ext4_iomap_check_delalloc(inode, &map);
+	return ext4_set_iomap(inode, iomap, type, first_block, &map);
We don't need to send first_block here. Since map->m_lblk
is same as first_block.
Also with Jan comment, we don't even need 'type' parameter.
Then we should be able to rename the function
ext4_set_iomap ==> ext4_map_to_iomap. This better reflects what it is
doing. Thoughts?

Depends on what we conclude in 1/8. :)

I'm for removing 'first_block', but still not convinced removing
'type' is heading down the right track if I were to forward think a
little.

Only once you are convinced that map->m_lblk will not change even in
function ext4_iomap_check_delalloc(), then only you should
drop "first_block" argument from ext4_set_iomap.

Please check above comments once.

-ritesh




[Index of Archives]     [Linux Ext4 Filesystem]     [Union Filesystem]     [Filesystem Testing]     [Ceph Users]     [Ecryptfs]     [AutoFS]     [Kernel Newbies]     [Share Photos]     [Security]     [Netfilter]     [Bugtraq]     [Yosemite News]     [MIPS Linux]     [ARM Linux]     [Linux Security]     [Linux Cachefs]     [Reiser Filesystem]     [Linux RAID]     [Samba]     [Device Mapper]     [CEPH Development]

  Powered by Linux