03:29 zzoon[m]: airlied: could you please review https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/28591 ?
03:36 airlied: zzoon[m]: done
04:06 zzoon[m]: thanks!
10:44 jani: sima: airlied: there are a number of intel specific headers under include/drm, named i915_*, xe_* and intel*. what if we added an include/drm/intel subdirectory?
10:44 jani: demarchi: ^
10:47 airlied: jani: do they all need to be in there? but yeah we could add one
10:49 sima: jani, yeah seems reasonable
10:49 sima: airlied, I guess because a per-kmod include dir is too much to ask from kbuild?
10:49 sima: for kmod that are spread across directories
10:51 jani: airlied: at a quick glance, they're all shared between i915 and drivers/misc/mei, drivers/char/agp/intel*, drivers/platform/x86/intel_ips.c, arch/x86/kernel/early-quirks.c, or sound/x86/intel_hdmi_audio.c
10:51 sima: oh right those need to all be there
10:52 airlied: ah cool yeah if they are used across the kernel it makes sense they be there
10:52 jani: it's not like moving them under a new subdir is making things a whole lot prettier, just a bit of lipstick
10:53 jani: plus it would be a place for some stuff maybe shared betwen i915 and xe and intel display in the future
12:04 zmike: mareko: do you want to ack https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/28308
12:54 jfalempe: sima, tzimmermann: Do you think the v11 of drm_panic is good enough ? (https://patchwork.freedesktop.org/series/122244) I agree it's not that useful yet, but I want to have this first version merged, before doing more work on it.
12:55 tzimmermann: jfalempe, i'll take another look tomorrow
12:56 jfalempe: tzimmermann: ok thanks. I think you already approved patches 6-8 on the v7, but I've rewritten them in v8 and v9.
14:29 willy: i've got to the point in the folio project where i can't put off changing fb_defio any longer
14:31 willy: there's a few ways this could be done, but i think the best way is to divorce fb_defio from struct page altogether. no more relying on PageDirty, or page_mkclean()
14:33 willy: we'd get rid of fb_deferred_io_pageref altogether, and replace it with a bitmap
14:33 willy: should get us much better performance and less code
14:34 willy: the only downside is having to touch a dozen drivers, and i'm not really set up to test graphics
16:30 demarchi: jani: ack on the subdir, sounds good to me
17:31 co1umbarius: Only dmabufs with one plane can be allocated with the implicit modifier api, right? So YUY type buffers would have their planes at an offset in the same memory range accessible with gbm_map?