On Sun, 14 Jan 2007, [EMAIL PROTECTED] wrote: > problem. If ifdown isn't being called, I wonder if also sync is not > called? > It would explain file system corruption bugs etch / debian folk have written > of across reboots.
Actually, the lack of a shutdown bus operation on ide, scsi and libata, and the lack of an automatic cache flush (always) and disk head unload (only when undergoing system poweroff or a device coldplug unplug operation) when the bus is shutdown is a rather ugly little misfeature of the *Linux* kernel (and maybe the others, I wouldn't know). It has been doing dumb things like this for a long time now. It was a MAJOR stupid mistake to ever try to work around this on userspace (userspace simply does not know enough to get it right always, and it simply *cannot* do it in suspend-to-disk, and other corner cases), and it will take two or three years before we manage to fix it and remove the "do it in userspace by default" crap halt(8) has to carry around. As for WOL, it looks like we need to stop with the stupid black magic games, and get an userland-kernel interface that does whatever is right on sleep and shutdown depending on WOL state, *regardless* of what userspace might have done to the interface state (up/down). If we enable WOL, the KERNEL driver should do whatever it takes to get it to work on sleep and shutdown. If we disable WOL, the drive should do whatever it takes to make sure it will not wake the machine on sleep and shutdown. Especially since it is clearly device-specific. And all needs to be done on .shutdown, and when entering any ACPI sleep, of course. There is only so much we can do on userspace to fix kernel bugs or limitations. The same way that there is only so much the kernel can do to work around bugs and limitations on userland stuff... -- "One disk to rule them all, One disk to find them. One disk to bring them all and in the darkness grind them. In the Land of Redmond where the shadows lie." -- The Silicon Valley Tarot Henrique Holschuh -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
