On Fri, May 12, 2006 at 01:50:46PM -0700, Linus Torvalds wrote: > On Fri, 12 May 2006, Russell King wrote: > > Great, I'm fucked by the SCSI folk again. > > No, you introduced a regression. This isn't the SCSI layer being evil, > this is the "regressions aren't acceptable". No - I introduced a correct fix. I think actually we're heading towards needing Linux V2 - the rewrite. It seems that fixing simple bugs cause other bugs, and that means we're heading into a maintainability nightmare. > > Can we revert the patch which broke the MMC/SD layer - the one which > > added the mount/unmount hotplug events as well then. > > > > That way we get back to a working MMC/SD layer as well as a working > > SCSI layer. > > That's certainly the logical fix - push the pain up the chain to the > person who introduced it. Which commit is that, do you know? No idea I'm afraid, and since I've had a _really_ extremely shitty couple of days I'm not about to start going to look for it. > Really, the added ref-count should be gotten by whoever holds on to the > thing, and it sounds like it's the hotplug event that caused this and > should have held on to its hotplug reference. ... which would be the genhd layer - it's keeping the driverfs_dev around so that the genhd layer can generate hotplug events using it at mount/umount time. Thanks for just re-confirming my original fix and that it's SCSI which is the real problem. 8) -- Russell King Linux kernel 2.6 ARM Linux - http://www.arm.linux.org.uk/ maintainer of: 2.6 Serial core - : send the line "unsubscribe linux-scsi" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html