Re: [PATCH] fix floppy.c to store correct ro/rw status in underlying gendisk

[Jon Masters]

On 10/31/05, Evgeny Stambulchik <Evgeny.Stambulchik@xxxxxxxxxxxxxx> wrote:
> Jon Masters wrote:

> > Let me know if this fixes it for you - should bomb out now if you try.
> > The error isn't the cleanest (blame mount), but it does fail.

> This works fine, thanks!

Good. This raises the general problem of devices knowing about their
underlying media status.

> For what it worth, though, mount -o remount,rw
> says remounting read-only yet still returns success. (Opposite to
> busybox, which now says "Permission denied" - rather misleading, but at
> least it fails).

Mount needs some fixing :-) It just tries calling sys_mount and
retries ro if that fails, then returns success because the last call
worked ok. Or so it would seem from the strace output.

> My question is, shouldn't it be implemented at a more generic level?

Yes. Right now, I think the gendisk policy is the best place to store
this - but having some generic "check the media status" VFS callback
might be a good thing.

> Floppy is just one example. Others are all kind of USB storage, ZIP/Jazz
> drives, and even normal SCSI disks (which have a RO jumper - at least
> some of them do).

I'll check, but they probably do something like this in their implementation.

> I got an ancient USB disk on key with a write-protection switch. When I
> plug it in in the RO mode, everything goes exactly as it was (before
> your patch) with floppy. Now something interesting:
> 1. The thingy is plugged in RW and mounted
> 2. While connected, I switch it to RO
>    dmesg says:
>    -> SCSI device sda: 129024 512-byte hdwr sectors (66 MB)
>    -> sda: Write Protect is on
>    -> sda: Mode Sense: 03 00 80 00
>    -> sda: assuming drive cache: write through
> 3.
> # mount -o remount,rw  /mnt
> mount: block device /dev/sda1 is write-protected, mounting read-only
> # echo $?
> 0
>
> So it seems there is some layer in bd which does know about RO status
> (and furthermore it's set by hot events)?

There's a callback for remount requests, perhaps they use that. Bear
in mind that USB devices tend to be more communicative than ancient
floppy disk drives. I'll have a look into this though - anyone else
reading who thinks we need a generic way of checking media read/write
status in addition to media change detect?

Jon.
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/