06:47tzimmermann: sima, could you please forward drm-next to -rc5 ?
06:48tzimmermann: bbrezillon ^
06:49sima: tzimmermann, what's the rationale?
06:51tzimmermann: sima, bbrezillon needs a gem_lru fix from upstream. IIRC there's a possible conflict resolution at https://lore.kernel.org/dri-devel/20260518165225.145175b1@fedora/
06:51tzimmermann: i'd than backmerge into drm-misc-next
06:51tzimmermann: s/than/then
06:58sima: tzimmermann, ah yes, on it
06:58tzimmermann: cool, thanks
06:59sima: I'll do that, ping you and then shut down my machine here because anything to not heat my apartment during this heatwave :-/
07:02tzimmermann: oh well, i don't look forward to summer
07:12sima: ah crap, exynos doesn't even compile in drm-next
07:12sima: airlied, broke in the merge you've just done, sad noises
07:13sima: tzimmermann, this'll take a bit longer I guess
07:14tzimmermann: sima, ok no hurries
07:20sima: tzimmermann, https://paste.debian.net/hidden/97fb3444 feel like an ack while you're here?
07:21tzimmermann: sima, a-b me
07:21sima: thx!
07:23sima: oh lol arnd has one already, I'll pick that one up
08:10sima: tzimmermann, airlied habemus backmerge
09:18tzimmermann: sima, thx
09:36bbrezillon: tzimmermann: https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/drivers/gpu/drm/drm_gem_shmem_helper.c?h=v7.1-rc5&id=cb2a2a5b37adb34ec46d39346b1c71e255827116
09:36bbrezillon: > Note that this wasn't a problem before commit 28e3918179aa
09:36bbrezillon: ("drm/gem-shmem: Track folio accessed/dirty status in mmap"), because
09:36bbrezillon: the pgprot were not lowered to read-only before this commit (see the
09:36bbrezillon: vma_wants_writenotify() in vma_set_page_prot()).
09:40bbrezillon: https://elixir.bootlin.com/linux/v7.1-rc5/source/arch/arm64/mm/mmap.c#L16 for why clearing VM_SHARED results in a read-only PTE/PMD
09:41bbrezillon: lowering happens in https://elixir.bootlin.com/linux/v7.1-rc5/source/mm/mmap.c#L81, which is called as part of the mmap() process
09:42bbrezillon: hope that helps clarify why write=false doesn't matter in that case
10:20tzimmermann: bbrezillon, ok. thanks for clarifying
10:20tzimmermann: BTW drm-misc-next now has the gem-lru fix
10:24tzimmermann: bbrezillon, but that really answer my question about a possible performance impact of the huge-page split on writes. not a blocker though
11:31bbrezillon: tzimmermann, sima: thanks for the backmerge
11:32bbrezillon: tzimmermann: pages won't be split on write, because you won't get a fault to trigger the split in the first place
11:34bbrezillon: because pages PMDs are writable from the start
11:48tzimmermann: bbrezillion. ok. and thanks for the timely review. if I don't hear anything else, the patch will be in tomorrow's -misc-fixes PR
13:46mlankhorst: tzimmermann: I see you backmerged drm-next, I want to send out a pull request for drm-misc-next soon is there anything you need?
13:47tzimmermann: mlankhorst, maybe bbrezillon does ^
13:48mlankhorst: I don't want to wait too long since it's already thursday
13:48dakr: mlankhorst: I have a few nouveau patches that were waiting for the backmerge too.
13:49dakr: I can dig in this evening, but if you don't want to wait I can also put them in a topic branch and send to Dave separately.
13:49mlankhorst: What this is 'this evening'?
13:49mlankhorst: time*
13:51dakr: I can probably get it done until 8pm CEST.
13:51mlankhorst: I'll prepare the summary and wait a bit longer then
13:51dakr: Ok, thanks!
14:15bbrezillon: mlankhorst: actually, I'm about to push something for panthor
14:15bbrezillon: that I'd like to have in for the next release cycle
14:16bbrezillon: I'm testing it as we speak, will let you know when it's done/pushed
14:17bbrezillon: sorry for the annoyance
14:49bbrezillon: mlankhorst: done
14:56mlankhorst: bbrezillon: thanks, now waiting on dakr
15:30mlankhorst: That's not an invitation to push 25 tegra patches. :-)
17:43dakr: mlankhorst: done, thanks!
18:50nexion: hi, on Venus page this recipe is provided without much of explanation
18:50nexion: -device virtio-gpu-gl,hostmem=4G,blob=true,venus=true
18:51nexion: where can I learn more what do hostmem and blob settings mean?
18:56nexion: is it host GPU VRAM or something else, how to measure its consumption and estimate the proper amount to not exceed the physical capacity when running several VMs
19:01mlankhorst: drm-misc-next-fixes is now open!
22:15airlied: karolherbst: looks like I'm going to need the shared memory optimiser :-)
22:15airlied: llama.cpp's coopmat2 shader definitely pushed the boundaries
22:20HdkR: Just get a GPU with more shared memory :P
22:22airlied: I'll just spill shared memory to global, what could possibly get slower
22:33airlied: okay the shared mem opt branch lets the shader run
22:33HdkR: Well now I'm disappointed that you're not spilling to global
23:36karolherbst: airlied: that one MR should just work (tm)
23:36karolherbst: airlied: https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/33914