On Fri, Oct 13, 2000 at 10:13:11AM +0900, Tom Holroyd wrote:
> Mine works OK, except for that invalidate on last close thing
> (how do I go back to the old behavior? Why was it changed?).
You go back to the old behaviour by commenting out the
invalidate_buffers call in
if (atomic_dec_and_test(&bdev->bd_openers))
invalidate_buffers(rdev);
in blkdev_put().
A rather silly but effective measure is to start
the command openflop:
#include <fcntl.h>
main(){
int fd = open("/dev/fd0", O_RDONLY);
pause();
}
each time after inserting a floppy, and to kill it
before taking the floppy out again.
(For me reading a floppy takes 72.5 seconds each time
without this command, and 0.1 sec the second time
with this command.)
Why was it changed? I think it was a very bad move.
The main reason was that Alexander Viro didnt see
how to retain guaranteed correctness without it.
Alan pseudo-defended it by saying that some devices
have a broken detection of media change.
So, presently we leave detection of media change
to the user who has to run openflop.
You may find the discussion in the archives.
[Subject: why invalidate_buffers() in blkdev_put()?]
Andries
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
Please read the FAQ at http://www.tux.org/lkml/
This archive was generated by hypermail 2b29 : Sun Oct 15 2000 - 21:00:24 EST