Re: PATCH (for 2.6.3-rc1) for cdrom driver dvd_read_struct

[Jamie Lokier <jamie@shareable.org>]

Jens Axboe wrote:
> On Tue, Feb 10 2004, Stuart Hayes wrote:
> > The attached patch will make the "dvd_read_struct" function in cdrom.c 
> > check that the DVD drive can currently do the DVD read structure command 
> > before sending the command to the drive.  It does this by checking the 
> > "dvd read" feature using the "get configuration" command.
> > 
> > Currently, cdrom.c only checks that the drive is a DVD drive before 
> > allowing the dvd read structure command to go to the drive--it does not 
> > make sure that the DVD drive has a DVD in it.  Without this patch, if CD 
> > medium is in a DVD drive, and the DVD_READ_STRUCT ioctl is used, the drive
> > will spew an ugly "illegal request" error (magicdev does this).
> 
> I'm rather anxious about applying anything like this, in my experience
> it's much much safer to simply let the command fail. I don't see
> anything technically wrong with your approach, I'd just like it tested
> on 100 different dvd drives :)

Meaning that you don't trust the DVD media test to return true on all
drives when and only when there's DVD media currently in there?

If that's your logic, it follows that userspace programs shouldn't
trust the DVD media test either - they should always call
DVD_READ_STRUCT if they would like to treat DVDs differently to CDs.

So, can you add a flag to the request meaning "if this
DVD_READ_STRUCT request fails with illegal request, don't log it"?

-- Jamie