Subject : "Fix ATAPI transfer lengths" causes CD writing regression Submitter : Daniel Drake <dsd@gentoo.org> References : http://lkml.org/lkml/2007/10/30/242 Handled-By : Alan Cox <alan@lxorguk.ukuu.org.uk> Jens Axboe <jens.axboe@oracle.com> Jeff Garzik <jeff@garzik.org> Tejun Heo <htejun@gmail.com>
Caused by: commit 2db78dd302d26d242d3e8e5c4c5024b6c3ea93c2 Author: Alan Cox <alan@lxorguk.ukuu.org.uk> Date: Tue Oct 2 13:53:04 2007 -0700 libata_scsi: Fix ATAPI transfer lengths http://git.kernel.org/gitweb.cgi?p=linux/kernel/git/torvalds/linux-2.6.git;a=commit;h=2db78dd302d26d242d3e8e5c4c5024b6c3ea93c2
This is a bug caused by a combination of a drive firmware bug and an application trying to be too clever. The fact it now fails is a feature not a bug and good news as the alternatives could be silent lossage. This is not therefore properly characterised as a regression but a bugfix.
Tejun Heo discovered this bug was in the userspace software sending the I/O request - it missed a byte, likely due to confusing layout in the scsi specs, causing the length to be wrong. Sorry for the noise!