Re: writable mmc profiles actually are writable

[Jens Axboe <axboe@suse.de>]

On Mon, Oct 06 2003, Pat LaVarre wrote:
> > assuming this has to do with misdetected hardware?
> 
> Seemingly so, aye.
> 
> > hardware?
> 
> I'm plugging in a pre-production sample of a new device.
> 
> Seemingly I cannot mkudffs its disks except by patching the kernel.
> 
> That's normal for yet another new kind of dvd/cd, yes?

Completely.

> Via sg I see 35 GB of 64 KiB/block writes, 2 KiB/block reads, with
> writable profile x0002 removable_disk of the 1999 mmc 2 standard.  Yes
> pdt x05 dvd/cd, yes feature x0020 random_writable, no not the cd_r,
> cd_rw, dvd, dvd_r, and dvd_ram enumerated in the obsolete legacy 1997
> mmc 1 mode page x2A Capabilities.
> 
> Previously near here we discussed some as-yet-ineffective alternatives
> to my patch: the Andy Polyakov patch for mount -w of dvd+rw, the Jens
> Axboe patch for mount -w of cd-rw, and blockdev --setrw.

blockdev --setrw is the same bypass again. dvd+rw patch and cd-rw patch
are two different beasts. The latter should probably be allowed to die
given the hideousness of that hardware, I haven't even looked at the
dvd+rw patch to pass any judgement on that (but I presume dvd+rw design
is saner and doesn't include software supported read-modify-write cycles
and defect management).

> > The uniform layer uses CDC_DVD_RAM as meaning
> > randomly writable media,
> 
> When I try setting CDC_DVD_RAM for a device, I see
> /proc/sys/dev/cdrom/info erroneously reports:
> 
> Can write DVD-RAM:      1
> 
> I conclude we wish to express the discovery of feature x0020
> random_writable differently.  Wrong?

Correct.

> In weak confirmation, earlier I did see the usb-storage folk reject a
> patch to tweak mode page x2A to appear to have mode page x2A byte 3 mask
> x20 dvd_ram_write set.  Should I submit a patch like that for sr_mod.ko
> and ide-cd.ko?

I'd propose just adding better detection of new devices, and adding a
specific "randomly writable" flag instead of abusing CDC_DVD_RAM.

> > Calling read-only media failing to be opened
> > read-write as "uncooperative" is confusing,
> 
> Clear now?  I mean to say, sg_dd shows the real disk of the real device
> is randomly rewritable.  But cdrom.ko and sr_mod.ko refuse to pass
> writes thru.

Well yes, now that you explain what the issue actually is rather than
show up with an obscure "patch" :)

-- 
Jens Axboe