On Wed, Apr 13, 2016 at 6:07 PM, Carlos E. R.
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA256
On 2016-04-14 01:10, Lindsay Mathieson wrote:
On 14/04/2016 8:29 AM, Carlos E. R. wrote:
But it is not safe. It is more dangerous, in fact.
How so?
The proposal says "Having separate swap partitions is safer from the hibernation point of view," but it is not so. Booting system (B) while (A) is hibernated is terribly dangerous because the boot procedure of (B) will run fsck on the already mounted, but hibernated, filesystems that (A) mounted. When, in due time, system (A) thaws, it will try to use what he thinks are open files, on partitions that have run fsck on them meanwhile. The in-memory filesystem structures do not match the on-disk structures. This cycle causes terrible corruption, beyond repair.
If you don't believe me, try it, in a virtual environment with two Linuxes that you can afford to destroy.
Yes but that sequence involves multiple instances of sufficiently bad implementation that it's sabotage. Isn't it? - The DE needs permission and capability to make certain there's a bootloader and bootloader configuration that definitely boots the right OS, to ensure resume happens at next boot. That means tacit permission on BIOS systems to overwrite the existing bootloader installation. On UEFI there's already a shared location for bootloaders and their own configuration files, so all the DE needs to do is modify the bootloader configuration file and NVRAM boot next. - In an alternate universe, we only have bootloaders that can read standardized bootloader configuration formats, stored in a standard location which is shared among all distros; this would obviate the need for the DE to overwrite the bootloader each time hibernation is called on a multiboot system. Since neither of those things is apparently true with any certainty, the logical conclusion for any DE that cares about user data, is to not offer hibernation anywhere in the GUI by default. It's the domain of experts only, it's too complicated and fragile to extend this to regular users. So I think the DE's need to hide any reference to hibernation/suspend to disk by default. And the installer should not include resume= as a kernel boot parameter. It's better to drop the hibernation image entirely than for it to get picked up after the file system it applies to has even been mounted, let alone had fsck run on it. -- Chris Murphy -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-factory+owner@opensuse.org