Re: [PATCH 2.6.15-rc6] block: Make CDROMEJECT more robust

[Jens Axboe <axboe@suse.de>]

On Mon, Dec 19 2005, Ben Collins wrote:
> > - Read request: direction bit not set, non-zero data length
> > - Write request: direction bit set, non-zero data length
> > 
> > How can you read or write anything, when there's nothing to read or
> > write?
> 
> Exactly, but the block layer (and I'd have to trace back through the
> code to find the exact failure point) has a check somewhere in there
> that checks the direction, and if it's a WRITE, checks data_len, and
> fails if it is 0.

I'd like to see that, if it's true it wants fixing of course.

> Note, my patch does not fix the case for the bug if I leave it as WRITE.
> It only works when it is READ.

We need to get that fixed as well, it should not make a difference.

> > So the medium removal command does require write permission on the
> > deviec, but it doesn't require root. If they need to rw to the device
> > fs, surely they need write permissions on the device in the first place?
> 
> bcollins@colorless:~$ id -a
> uid=1000(bcollins) gid=1000(bcollins)
> groups=4(adm),20(dialout),24(cdrom),25(floppy),29(audio),30(dip),44(video),46(plugdev),104(lpadmin),105(scanner),106(admin),1000(bcollins)
> bcollins@colorless:~$ ls -l /dev/hdc
> brw-rw---- 1 root plugdev 22, 0 Dec 19  2005 /dev/hdc
> bcollins@colorless:~$ eject -s /dev/hdc
> eject: unable to eject, last error: Operation not permitted
> bcollins@colorless:~$ eject -r /dev/hdc

Does eject open the device O_RDWR?

-- 
Jens Axboe