On Tue, 11 Jan 2011, Ted Ts'o wrote:

Argh.  In the future, I'd really appreciate if you explicitly label
patches with a version number, so I can more easily keep track of
which one is the latest.

It's also best if you send patches as free-standing separate e-mail
messages (one message per patch, using git format-patch and git
send-email), so they can more easily tracked using patchwork.  Sending
me patches which are included as a quoted reply (as you did here) means
I have to manually pick out the patch, or apply it the patch by hand.

Both of these would make my life much easier; and makes it more likely
I will apply your patches quickly.

Thanks,

						- Ted

I am sorry for causing you troubles. However I always send you patches
using git format-patch and git send-email, but it this case I just
replied to Jan's patch to let you know that my previous attempts to fix
this was not right. Jan sent it to you and to the ext4 list, so you should
have the original, no need to manually pick out this one.

However I did screw up with two different versions which both were wrong
and I am sorry about that...will try to do better next time and to
version my patches.

Thanks!
-Lukas

On Tue, Jan 11, 2011 at 11:40:49AM +0100, Lukas Czerner wrote:

quoted

On Mon, 10 Jan 2011, Jan Kara wrote:

quoted

When s_first_data_block is not zero (which happens e.g. when block size is 1KB)
and trim ioctl is called to start trimming from block 0, the math in
ext4_get_group_no_and_offset() overflows. The overall result is that ioctl
returns EINVAL which is kind of unexpected and we probably don't want
userspace tools to bother with internal details of filesystem structure.
So just silently increase starting offset (and shorten length) when starting
block is below s_first_data_block.

CC: Lukas Czerner <redacted>
Signed-off-by: Jan Kara <jack@suse.cz>
---
 fs/ext4/mballoc.c |    6 ++++++
 1 files changed, 6 insertions(+), 0 deletions(-)

diff --git a/fs/ext4/mballoc.c b/fs/ext4/mballoc.c
index 4c4766c..b9c2aad 100644
--- a/fs/ext4/mballoc.c
+++ b/fs/ext4/mballoc.c

@@ -4819,6 +4819,8 @@ int ext4_trim_fs(struct super_block *sb, struct fstrim_range *range)
 	ext4_group_t group, ngroups = ext4_get_groups_count(sb);
 	ext4_grpblk_t cnt = 0, first_block, last_block;
 	uint64_t start, len, minlen, trimmed;
+	ext4_fsblk_t first_data_blk =
+			le32_to_cpu(EXT4_SB(sb)->s_es->s_first_data_block);
 	int ret = 0;
 
 	start = range->start >> sb->s_blocksize_bits;

@@ -4828,6 +4830,10 @@ int ext4_trim_fs(struct super_block *sb, struct fstrim_range *range)
 
 	if (unlikely(minlen > EXT4_BLOCKS_PER_GROUP(sb)))
 		return -EINVAL;
+	if (start < first_data_blk) {
+		len -= first_data_blk - start;
+		start = first_data_blk;
+	}
 
 	/* Determine first and last group to examine based on start and len */
 	ext4_get_group_no_and_offset(sb, (ext4_fsblk_t) start,

Hi Ted,

forget my previous patch, this is the right one. Thanks Jan!

-Lukas

--