Skip to contentSkip to conversation

Resend: PATCH: Stop 2.6.12rc rmmod from being able to destroy CD hardware

PATCH Alan Cox 4 participants 4 messages May 27, 2005

lore
Share & tools
lore

Alan Cox

author
#0 lore raw

On an rmmod the cdrom layer when used with ide-cd issues a cache flush atapi command to devices including those that do not support it. According to Jens earlier discussion this isn't merely a minor glitch but can destroy some CD hardware due to firmware bugs in the drive (as per the Mandrake incident)

The IDE CD layer uses a mask of unsupported features, this means that because ide-cd doesn't know about MRW writables it doesn't set the relevant bit for non writables and harm can occur.

The simple fix is attached, making the driver start from ~0 and mask bits the other direction would longer term be safer.

drivers/ide/ide-cd.c +3 -0
diff -u --new-file --recursive --exclude-from /usr/src/exclude linux.vanilla-2.6.12rc3/drivers/ide/ide-cd.c linux-2.6.12rc3-minoride/drivers/ide/ide-cd.c
@@ -2860,6 +2922,9 @@
 		devinfo->mask |= CDC_CLOSE_TRAY;
 	if (!CDROM_CONFIG_FLAGS(drive)->mo_drive)
 		devinfo->mask |= CDC_MO_DRIVE;
+		
+	/* We must have this masked unless a drive definitely handles it */
+	devinfo->mask |= CDC_MRW_W;
 
 	devinfo->disk = info->disk;
 	return register_cdrom(devinfo);
#1 lore raw

I believe that this was fixed differently almost moth ago by Jens:
http://kernel.org/git/?p=linux/kernel/git/torvalds/linux-2.6.git;a=commitdiff;h=7da21a02b3587157bd43910ea6d4c76661228ebb;hp=76530da1a9e8ce05963b1f49a098eddc6ec6c534

It would be nice to have features mask in ide-cd.c done right but this sounds like 2.6.13 material.

Quoted 26 lines · Alan Cox

On an rmmod the cdrom layer when used with ide-cd issues a cache flush atapi command to devices including those that do not support it. According to Jens earlier discussion this isn't merely a minor glitch but can destroy some CD hardware due to firmware bugs in the drive (as per the Mandrake incident)

The IDE CD layer uses a mask of unsupported features, this means that because ide-cd doesn't know about MRW writables it doesn't set the relevant bit for non writables and harm can occur.

The simple fix is attached, making the driver start from ~0 and mask bits the other direction would longer term be safer.

diff -u --new-file --recursive --exclude-from /usr/src/exclude linux.vanilla-2.6.12rc3/drivers/ide/ide-cd.c linux-2.6.12rc3-minoride/drivers/ide/ide-cd.c

drivers/ide/ide-cd.c 2005-05-01 14:09:35.000000000 +0100 renamed from drivers/ide/ide-cd.c 2005-04-27 16:01:29.000000000 +0100 +3 -0
@@ -2860,6 +2922,9 @@
                devinfo->mask |= CDC_CLOSE_TRAY;
        if (!CDROM_CONFIG_FLAGS(drive)->mo_drive)
                devinfo->mask |= CDC_MO_DRIVE;
+
+       /* We must have this masked unless a drive definitely handles it */
+       devinfo->mask |= CDC_MRW_W;
 
        devinfo->disk = info->disk;
        return register_cdrom(devinfo);
#2 lore raw
In reply to Alan Cox
Quoted 3 lines · Alan Cox

The simple fix is attached, making the driver start from ~0 and mask bits the other direction would longer term be safer.

As per Bartlomiej, I've not applied this, since the "don't do it unless you've written to it" fix from Jens went in. I assume CD-RW's know how to cache flush..

That said, it does sound like the capabilities bitmap should be changed around to be the other way (either by starting at ~0, or, if the changes aren't too invasive, just change the semantics of it to be the sane way around ("drive can do this" rather than "drive can _not_ do this").

Linus

#3 lore raw
Quoted 7 lines · Linus Torvalds
Quoted 3 lines · Alan Cox

The simple fix is attached, making the driver start from ~0 and mask bits the other direction would longer term be safer.

Quoted 8 lines

As per Bartlomiej, I've not applied this, since the "don't do it unless you've written to it" fix from Jens went in. I assume CD-RW's know how to cache flush..

That said, it does sound like the capabilities bitmap should be changed around to be the other way (either by starting at ~0, or, if the changes aren't too invasive, just change the semantics of it to be the sane way around ("drive can do this" rather than "drive can _not_ do this").

They should, it's always annoyed me because it's counter intuitive.

Signature · 1 line
Jens Axboe