mke2fs: only try discarding a single block to test if discard works

Commit d2bfdc7ff15c ("Use punch hole as "discard" on regular files")
added a test to see if the storage device actually supports discard.
The intent was to try discarding the first block but since
io_channel_discard() interprets the offset and count arguments in
blocks, and not bytes, mke2fs was actually discarding the first 16
megabytes (when the block size is 4k).  This is normally not a
problem, since most file systems are larger than that, and requests to
discard beyond the end of the block device are ignored.

However, when creating a small file system as part of a image
containing multiple partitions, the initial test discard can end up
discarding data beyond the file system being created.

Addresses-Debian-Bug: #989630
Reported-by: Josh Triplett <josh@joshtriplett.org>
Fixes: d2bfdc7ff15c ("Use punch hole as "discard" on regular files")
Signed-off-by: Theodore Ts'o <tytso@mit.edu>

diff --git a/misc/mke2fs.c b/misc/mke2fs.c
index 9fa6eaa..5a35e9e 100644 (file)
--- a/misc/mke2fs.c
+++ b/misc/mke2fs.c
@@ -2794,7 +2794,7 @@ static int mke2fs_discard_device(ext2_filsys fs)
        struct ext2fs_numeric_progress_struct progress;
        blk64_t blocks = ext2fs_blocks_count(fs->super);
        blk64_t count = DISCARD_STEP_MB;
-       blk64_t cur;
+       blk64_t cur = 0;
        int retval = 0;
 
        /*
@@ -2802,10 +2802,9 @@ static int mke2fs_discard_device(ext2_filsys fs)
         * we do not print numeric progress resulting in failure
         * afterwards.
         */
-       retval = io_channel_discard(fs->io, 0, fs->blocksize);
+       retval = io_channel_discard(fs->io, 0, 1);
        if (retval)
                return retval;
-       cur = fs->blocksize;
 
        count *= (1024 * 1024);
        count /= fs->blocksize;