On Fri, Feb 21, 2020 at 02:43:05PM +0000, David Howells wrote:

Darrick J. Wong [off-list ref] wrote:

quoted

quoted

+	memcpy(ctx->buffer, es->s_volume_name, sizeof(es->s_volume_name));

Shouldn't this be checking that ctx->buffer is large enough to hold
s_volume_name?

Well, the buffer is guaranteed to be 4KiB in size.

Ah, ok.

quoted

quoted

+	return strlen(ctx->buffer);

s_volume_name is /not/ a null-terminated string if the label is 16
characters long.

And the buffer is precleared, so it's automatically NULL terminated.

<nod>

quoted

quoted

+#define FSINFO_ATTR_EXT4_TIMESTAMPS	0x400	/* Ext4 superblock timestamps */

I guess each filesystem gets ... 256 different attrs, and the third
nibble determines the namespace?

No.  Think of it as allocating namespace in 256-number blocks.  That means
there are 16 million of them.  If a filesystem uses up an entire block, it can
always allocate another one.  I don't think it likely that we'll get
sufficient filesystems to eat them all.

Ah.  In that case I declare that we would like to reserve 0x5800-0x58FF
for XFS. :)

--D

David