17:12 <alyssa> has anyone managed to use ecryptfs (or full disk encryption) successfully on internal storage on an RK3288 chromebook? it seems I'm getting disk corruption on every boot to the point where I think I'll have to revert back to a cleartext system by the end of the day, before it'll've lasted even 24h...

17:16 <Ke> has nothing to do with th internal, disable rk3288 crypto accelerator or revert to linux-4.9.91

17:17 <Ke> I complained about this to no avail, crypto accel has been broken about since linux-4.11ish iirc

17:18 <Ke> at that time it was trivially broken, now it's just corrupting frequently

17:19 <Ke> unless you are hitting a different issue

17:19 <Ke> if rk3288 crypto accel works for you, please note, so I can retry

17:20 <Ke> anyway 4.9 works like a charm, so I have almost no interest in updating for a year

17:22 <Ke> and obviously the plan is to ditch arm for aarch64 and rk3399

17:27 <Ke> to be honest, rk3288 crypto unit is probably simple enough to fix even with my skills if push comes to shove

17:27 <Esmil> alyssa: hey, I read your last blog on reversing the mail GPU. Very cool stuff :)

17:28 <Ke> yeah, it awesome, not sure why you are not getting more on liberapay

17:28 <Esmil> Did I understand correctly that you're working on Mali T7xx (Midgard 3rd gen) and someone else is working on Bifrost?

17:29 <Esmil> ..and you mention that they're very similar

17:29 <Esmil> So eventually Mali T8xx (Midgard 4th gen) should also be relatively easy to add support for right?

17:30 <alyssa> Ke: hm ok. i was thinking since the eMMC is unstable as it is, and I'm getting kernel lockups (dwc2 related, I think)... there's just general disk corruption which is amplified by the encryption.

17:30 <alyssa> Esmil: thanks! :)

17:32 <alyssa> our project (Panfrost) is interested in both Midgard and Bifrost. the command stream part (i.e. not the shaders) are, afaik, identical; in some sense, Bifrost's command stream appears to be T9xx, a mere incremental revision. the shaders are totally different. for shaders, I work only on Midgard; others work on Bifrost though; both will end up supported, but Midgard first. I privately suspect that T8xx might work without changes o

17:32 <alyssa> n the current codebase, but I haven't tested that

17:32 <Ke> it was actually one of the times I was glad I used btrfs

17:33 <Esmil> alyssa: Nice. I should try out on my rk3399 chromebook then

17:34 <alyssa> it's not really in a state to be tried by people not interested in developing it atm... unless you're interested in working on T8xx support, in which case by all means! :D

17:34 <Ke> what sort of stack would this be then, libdrm, mesa, arm opensource, but not upstream, drm module?

17:35 <alyssa> Kernel will be a cleaned up, (eventually DRM) fork of the ChromeOS kernel driver. Userspace is mesa. Same strategy as freedreno/etnaviv/lima

17:36 <Ke> thanks

17:37 <alyssa> sure

17:38 <Ke> I am definitely hoping some company will hire you to work on this full time

17:40 <Ke> obviously you may want to study or something... not my call anyway

20:37 <alyssa> okay, follow up question... does anyone know the safe way to revert from using ecryptfs? I still have the unencrypted backup from the migration since I was afraid this might happen

20:47 <alyssa> undone. that was fun while it lasted.