| xnox | cjwatson: ubuntu1 plugin create a gnome-keyring and copies it into user's home dir in target. That works without ecryptfs homedir encryption. When ecryptfs homedir encryption is enabled, adduser unmounts ecrypted user dir after having copied in /etc/skel. What should u1 plugin do? | 13:11 |
|---|---|---|
| xnox | (a) mount user-dir again - that means caching the password for longer. | 13:11 |
| xnox | (b) modified adduser, to allow keeping userdir mounted | 13:11 |
| cjwatson | temporarily create an adduser.local hook? | 13:12 |
| xnox | (c) pre-populate (target)/etc/skel with keyring, and wipe it after user-setup is finished | 13:12 |
| xnox | cjwatson: there are hooks?! =) | 13:12 |
| * xnox looks | 13:12 | |
| cjwatson | it's a rather basic hook mechanism (one script) but it should be usable for this | 13:12 |
| cjwatson | populating target/etc/skel doesn't seem entirely unreasonable either if you're careful | 13:13 |
| xnox | cjwatson: i think I will go with populating target/etc/skel, as the adduser.local is called after create_homedir() exited and unmounted ecryptfs homedir. | 13:17 |
| cjwatson | OK | 13:19 |
| xnox | danjared: cjwatson: so typical good SSD from 2012 has sequential read speed of above 500MB/s (based on Tom's hardware charts) this means for a full 8GB dump a 16s read time. Which is more than cold-boot to default desktop time on ssd laptops (8-12s ?!) | 17:25 |
| xnox | obviously we win, if it's not full 8GB. Similarly 8GB limit, sounds like the "rapidness" threshold based on current maximum SSD technology read speed. | 17:26 |
| danjared | yes, it's still faster than cold boot (or conventional Linux hibernate) even if you're writing and reading 8GB of data | 17:55 |
| danjared | (because you appear to bypass post) | 17:56 |
| xnox | ok. true. | 17:56 |
| danjared | it's not entirely clear if your system is 100% off when you do an iRST hibernate | 17:56 |
| danjared | the various information I've seen from Intel is not terribly clear on that | 17:57 |
| xnox | should we talk to microsoft then instead?! =) | 17:57 |
| danjared | in one place (can't remember where), I saw mention of "up to 30 days hibernate", which is consistent with the appearance of going straight to dumping back to memory at "power on". but the latest Intel documentation I've seen says that the system goes to S4 | 17:58 |
| danjared | maybe it's S4 with firmware magic | 17:59 |
| danjared | xnox: this gives some vague information on how it deals with full-disk encryption: http://downloadmirror.intel.com/22647/eng/Intel%20Rapid%20Start%20Technology%20Deployment%20Guide%20v1.0.pdf | 18:06 |
| xnox | danjared: thanks, downloaded. | 18:09 |
| danjared | I'll a link to that once I finish updating the blueprint | 18:12 |
| danjared | cjwatson: ooh, interesting, section 2.5.2 of the document I mention above talks about an optional feature to allow iRST hibernate when the IFFS partition is smaller than the amount of memory | 18:24 |
| xnox | danjared: interesting =) i have >>8GB but if <<8GB is mapped I'd love to rapid start it. | 18:26 |
| cjwatson | what does it do with the stuff that doesn't fit? :-) | 18:26 |
| danjared | if it doesn't all fit, you stay in S3 | 18:27 |
| infinity | cjwatson: Discard odd-numbered pages, write the rest to disk, and watch the pretty colours on resume, of course. | 18:27 |
| xnox | sensible thing would be to keep laptop in sleep mode and keep the RAM powered =) | 18:27 |
| xnox | but infinity way is more fun =) | 18:27 |
| infinity | Alternately, write half your system RAM to a framebuffer in VRAM, apply jpeg compression, read it back out. | 18:28 |
| infinity | WHAT COULD POSSIBLY GO WRONG. | 18:28 |
| danjared | ha | 18:29 |
| xnox | infinity: funny you should say that, mac os x does take screenshot, blurries it and store it, thus on resume the instantly show the desktop & a progress bar of unfreezing. | 18:30 |
| xnox | s/the instantly/they instantly/ | 18:30 |
| danjared | wat | 18:31 |
| infinity | xnox: That's quite the information disclosure. | 18:31 |
| infinity | Unless it's always a screenshot of a lock screen. :P | 18:31 |
| danjared | hopefully you're on the right websites | 18:31 |
| xnox | infinity: danjared: there is no lock screen by default. | 18:33 |
| infinity | xnox: Right. I assume that if you enable screen locking, it also disabled the screenshot-on-resume feature. :) | 18:33 |
| infinity | (Cute feature, but seems somewhat useless, since you can't INTERACT WITH IT) | 18:33 |
| * xnox may have blasted out load some "resumed" video in a library before when opening laptop and walking away to pick up books from shelves. | 18:33 | |
| xnox | s/load/loud/ | 18:33 |
| === kentb is now known as kentb-out | ||
| xnox | cjwatson: i confused Intel Rapid Start Technology, with Intel Smart Response Technology. The later uses SSD as a cache for HDD. | 23:38 |
| cjwatson | Ah | 23:41 |
| cjwatson | Intel naming FTL | 23:41 |
| xnox | cjwatson: i have distilled the "Intel Rapid Start Technology Deployment Guide" into 6 quotes on the blueprint. (to save 29page read) | 23:42 |
| xnox | cjwatson: the "api" to enable flushing pages, is a windows registry key that intel windows daemon listens to - and "flushes active pages or not" | 23:43 |
| xnox | (daemon is called "service" in windows?!) | 23:43 |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!