open write failure not predicted by blockdev getro

open write failure not predicted by blockdev getro

[Pat LaVarre]

> Subject: Re: [PATCH] fix cdrom mt rainier probe

Does the reverse inconsistency also interest you?

By inserting a read-only DVD disc into a USB drive, I get to see the
paradoxical contradiction:

$ uname -msr
Linux 2.6.8-rc1 i686
$
$ sudo blockdev --getro /dev/scd0
0
$ sudo dd of=/dev/scd0 bs=64K count=1 if=/dev/zero
dd: opening `/dev/scd0': Read-only file system
$

I trust you'll tell me if I can help further.

Pat LaVarre

Re: open write failure not predicted by blockdev getro

[Jens Axboe]

On Tue, Jul 13 2004, Pat LaVarre wrote:
> > Subject: Re: [PATCH] fix cdrom mt rainier probe
> 
> Does the reverse inconsistency also interest you?
> 
> By inserting a read-only DVD disc into a USB drive, I get to see the
> paradoxical contradiction:
> 
> $ uname -msr
> Linux 2.6.8-rc1 i686
> $
> $ sudo blockdev --getro /dev/scd0
> 0
> $ sudo dd of=/dev/scd0 bs=64K count=1 if=/dev/zero
> dd: opening `/dev/scd0': Read-only file system
> $
> 
> I trust you'll tell me if I can help further.

It does interest me, please see if you can track this problem down. I
have no time myself these days.

-- 
Jens Axboe

Re: open write failure not predicted by blockdev getro

[Pat LaVarre]

> > inserting a read-only DVD disc into a USB drive, ...
> > the paradoxical contradiction:
> > ...
> > $ sudo blockdev --getro /dev/scd0
> > 0
> > $ sudo dd of=/dev/scd0 bs=64K count=1 if=/dev/zero
> > dd: opening `/dev/scd0': Read-only file system
> > $
> ...
> does interest me, ...

Excellent, hi.

> please see if you can track this problem down.

printk tells me,

We call set_disk_ro of drivers/block/genhd.c only once per device plug
in, not once per disc insert.

Logically therefore, given a mixed collection of rewritable and
read-only discs, our guessing blockdev --getro 1 at plug in will prevent
writes of rewritable discs, and our guessing blockdev --getro 0 at plug
in will allow writes of read-only discs.

We necessarily lose both ways.

Do we care?

May we say that blockdev --setrw --setro --getro is a facility for
write-protecting DVD/ CD devices, not DVD/ CD discs, and take on the job
of guessing which devices don't want blanket write protection?

Or do we learn to refresh set_disk_ro once per disc insert?

We can decide in light of the drivers/scsi/sd.c precedent.  I see there
a comment telling me we call set_disk_ro once per disc insert:

--- http://lxr.linux.no/source/drivers/scsi/sd.c?v=2.6.5#L1109
...
* read write protect setting, if possible - called only in sd_revalidate_disk()
...
---

Pat LaVarre

P.S. Newbie, me I haven't yet found the lines of cdrom.ko code we run
only once per disc insert.

The following patch ain't it, at least not yet.  dmesg tells me this
source line executes indefinitely often while I have a USB DVD disc
(specifically an Iomega RRD disc) present.  Even with PATAPI alone
connected, still it ran more often than the count of discs present.

diff -urp linux-2.6.8-rc1/drivers/cdrom/cdrom.c linux-2.6.8-rc1-pel/drivers/cdrom/cdrom.c
--- linux-2.6.8-rc1/drivers/cdrom/cdrom.c	2004-07-13 08:26:02.000000000 -0600
+++ linux-2.6.8-rc1-pel/drivers/cdrom/cdrom.c	2004-07-14 17:46:24.013859464 -0600
@@ -1297,6 +1297,7 @@ int media_changed(struct cdrom_device_in
 	    return ret;
 	/* changed since last call? */
 	if (cdi->ops->media_changed(cdi, CDSL_CURRENT)) {
+printk("PEL: could refresh set_disk_ro\n");
 		cdi->mc_flags = 0x3;    /* set bit on both queues */
 		ret |= 1;
 	}