Indeed, only 2.2 trusted the check media change information and left the
cache valid on top of the floppy across close/open of the blkdev.
Maybe it's because a read now fills in the whole page rather than 1k.
So if you write 1k aligned, and you read it back, the read will it the
disk and it will have to read another 3k before it can return such 1k
info that was just in memory. It's exactly the same on the files on a 1k
ext2 filesystem.
Alexander just proposed some month ago to support the partial reads in
the pagecache like we do for partial writes: by not marking the page
uptodate for partial reads and by going down to the buffer layer every
time to check if the partial memory we're interested about is just
uptodate or not (and marking the page uptodate eventually if we end
reading the whole page). So it's nothing unfixable, we just need to
understand if that is the problem and/or if mtools is doing something
stupid that sorted out to work ok when partial reads were supported.
Andrea
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/