#dri-devel on 2021-10-22 — irc logs at oftc.irclog.whitequark.org

2021-07-26 22:56 ChanServ changed the topic of #dri-devel to: <ajax> nothing involved with X should ever be unable to find a bar

00:00 <Kayden> though I'm not sure what fds you actually want...there's dedup across fstat-equivalent /dev/dri/* files, and dedup across os_same_file_description() (same open() of the same file, i.e. just dup()'d)

00:00 <jekstrand> We don't for the EXT (it requires buffer-owned VA) but we do for the KHR

00:00 <jekstrand> bnieuwenhuizen: util/vma.h is your friend. :D

00:00 <bnieuwenhuizen> Kayden: it dedups across same major/minor

00:01 <bnieuwenhuizen> which means you actually need the right fd to have your manually create syncobj be valid

00:03 <Kayden> ah, right, amdgpu_device does that...more or less... (actually drmGetDeviceNameFromFd equivalence, which is...a bit more strange than that...)

00:03 <Kayden> radv_amdgpu_winsys_create should have the actual fd though

00:03 <Kayden> so you could just store it somewhere

00:03 <bnieuwenhuizen> Kayden: if you store it the handles might not be compatible with the internal fd in libdrm_amdgpu

00:06 danvet has quit [Ping timeout: 480 seconds]

00:16 Company has quit [Read error: Connection reset by peer]

00:25 ybogdano has quit [Ping timeout: 480 seconds]

00:38 Lucretia has quit []

00:44 <imirkin> zmike: well, that's definitely the path of least resistance :) not really gallium, but GL (and thus st/mesa) expect MSAA to work equally on any renderable format.

00:47 <imirkin> zmike: in practice, some of the DX10 chips support less MSAA on 128-bit formats than on other ones. i think we've sorta normalized that in a way that doesn't completely upset everything ... iirc we just make the FB incomplete or something

00:47 <imirkin> zmike: but perhaps i'm not completely understanding your problem

00:47 <imirkin> an illustrated example might help?

00:52 columbarius has joined #dri-devel

00:53 <zmike> imirkin: there's one format that most vk drivers can't render to as ms, but GL expects it, and returning unsupported fails the test instead of skipping it

00:53 co1umbarius has quit [Ping timeout: 480 seconds]

00:54 <imirkin> zmike: which format?

00:54 <zmike> the shared exponent one

00:54 <imirkin> rgb9e5?

00:54 <zmike> yep

00:55 <imirkin> annoying

00:55 <zmike> yep

00:55 <imirkin> slightly surprised that hw lets you render in the first place

00:55 <zmike> I pass the tests now but only because detection is broken

00:55 <zmike> and the test skips

00:56 <zmike> very stupid problem

00:56 <imirkin> nvidia doesn't support rendering to rgb9e5

00:56 <imirkin> tbh i'd just say don't report it renderable

00:56 <imirkin> (maybe the latest boards do, dunno, but certainly not even like kepler generation did)

00:56 <zmike> I have to or the tests fail

00:57 <zmike> it's bizarre

00:57 <imirkin> report rgb9e5 renderable??

00:57 <zmike> yep

00:57 <imirkin> i don't think so. if they're failing, it's for some other reason

00:57 <imirkin> like i said, nouveau only exposes rgb9e5 for texturing

00:58 <zmike> I still need to look at it more, it's hard to figure out the rube goldbergness

00:58 <zmike> but that's the tl;dr

01:01 <imirkin> yeah, really annoying to track this stuff down

01:02 <zmike> indeed

01:02 <imirkin> even when you're not dealing with layering

01:03 <imirkin> zmike: fyi, this was my latest fix for the more general situation: 561f9ae74b2b7da06bb4830aaca8d017a3dd2746

01:04 <zmike> will have to check tomorrow, I'm on mobile for the night

01:05 <imirkin> cool. lmk if you have more questions tomorrow (i tend not to be on irc during the day, but email will reach me)

01:06 <zmike> it's not that urgent

01:06 <zmike> slow and steady

01:07 <zmike> just like the ribs I made tonight

01:07 camus1 has joined #dri-devel

01:08 camus has quit [Ping timeout: 480 seconds]

02:17 boistordu has quit [Ping timeout: 480 seconds]

02:21 camus has joined #dri-devel

02:24 nchery has quit [Remote host closed the connection]

02:24 camus1 has quit [Ping timeout: 480 seconds]

02:29 rg3igalia_ has joined #dri-devel

02:29 ezequielg_ has joined #dri-devel

02:29 cwabbott has quit [Ping timeout: 480 seconds]

02:29 ezequielg has quit [Ping timeout: 480 seconds]

02:29 steev_ has joined #dri-devel

02:30 steev has quit [Ping timeout: 480 seconds]

02:30 steev_ is now known as steev

02:30 cwabbott has joined #dri-devel

02:31 rg3igalia has quit [Ping timeout: 480 seconds]

02:44 vivijim has quit [Ping timeout: 480 seconds]

02:56 aswar002 has quit [Quit: No Ping reply in 180 seconds.]

02:56 mbrost has joined #dri-devel

02:57 aswar002 has joined #dri-devel

03:01 camus1 has joined #dri-devel

03:02 camus has quit [Ping timeout: 480 seconds]

03:24 nchery has joined #dri-devel

03:26 camus has joined #dri-devel

03:30 camus1 has quit [Ping timeout: 480 seconds]

03:36 Sachiel has quit [Quit: WeeChat 3.2]

03:37 Sachiel has joined #dri-devel

03:42 shashank1202_ has joined #dri-devel

03:42 lemonzest has joined #dri-devel

03:47 mbrost has quit [Remote host closed the connection]

04:26 Daaanct12 has quit [Remote host closed the connection]

04:27 Daaanct12 has joined #dri-devel

04:34 pnowack has joined #dri-devel

04:39 camus has quit [Ping timeout: 480 seconds]

04:43 rpigott has quit [Read error: Connection reset by peer]

04:44 rpigott has joined #dri-devel

04:44 camus has joined #dri-devel

04:55 aravind has joined #dri-devel

05:01 Duke`` has joined #dri-devel

05:01 sdutt has quit [Remote host closed the connection]

05:07 i-garrison has quit [Ping timeout: 480 seconds]

05:29 itoral has joined #dri-devel

05:32 nchery has quit [Ping timeout: 482 seconds]

05:38 tarceri_ has joined #dri-devel

05:39 tarceri_ has quit [Read error: Connection reset by peer]

05:39 i-garrison has joined #dri-devel

05:39 tarceri_ has joined #dri-devel

05:44 tarceri has quit [Ping timeout: 480 seconds]

06:00 thellstrom has joined #dri-devel

06:05 thellstrom1 has quit [Ping timeout: 480 seconds]

06:12 mlankhorst has joined #dri-devel

06:24 thellstrom has quit [Quit: thellstrom]

06:24 thellstrom has joined #dri-devel

06:26 camus1 has joined #dri-devel

06:31 camus1 has quit [Remote host closed the connection]

06:31 alanc has quit [Remote host closed the connection]

06:32 camus has quit [Ping timeout: 480 seconds]

06:32 alanc has joined #dri-devel

06:37 mlankhorst has quit [Remote host closed the connection]

06:37 camus has joined #dri-devel

06:37 mlankhorst has joined #dri-devel

06:46 danvet has joined #dri-devel

06:48 <dj-death> jekstrand: yeah, I was pretty sure that for binary semaphore, it has to have been submitted all the way

06:52 shashank1202_ has quit [Quit: Connection closed for inactivity]

07:03 jkrzyszt has joined #dri-devel

07:18 tzimmermann has joined #dri-devel

07:18 <pq> mdnavare, in case it wasn't quite clear yet: Wayland has nothing to do with actual displays. Whether a compositor needs a display device or not is totally up to the particular compositor. There is nothing generic about that. If you use VKMS, it's possible that some compositors will then not support hw accel even if you had a render-only GPU.

07:19 <pq> mdnavare, Weston in its current would not support hw accel on VKMS, but it has a separate headless-backend that supports hw accel without any display (DRM KMS device).

07:23 <tzimmermann> danvet, mripard_, mlankhorst, i need to revert a patch. could one of you please rubber-stamp an a-b to https://lore.kernel.org/dri-devel/20211015072546.3705-1-tzimmermann@suse.de/

07:29 <javierm> tzimmermann: hi, I'm pushing to disable the remaining fbdev drivers in Fedora and just use simpledrm instead. A user gave some feedback that I think is interesting

07:29 <javierm> tzimmermann: he asked if there could be added a kernel cmdline to prevent the real DRM drivers to take over

07:30 <javierm> something like nomodeset but to keep using simpledrm, let's say that there's a bug in the platform DRM driver but the firmware frame buffer is enough for basic video output

07:30 <tzimmermann> javierm, there's sometimes a per-driver module parameter

07:31 <javierm> tzimmermann: yes, you could also use module.blacklist=foobar but is an interesting idea I think. That way distros could have a "basic graphics" or something entry that does this

07:31 <javierm> for troubleshooting would be useful

07:32 <tzimmermann> javierm, but driver loading is performed by the linux kernel internally. there's nothing that drm could do

07:32 <tzimmermann> at least not until the device get's created

07:33 <tzimmermann> we could make drm_aperture_remove_framebuffer() fail, but that's it's somewhat hackish

07:33 elongbug has joined #dri-devel

07:34 <javierm> tzimmermann: device -> driver matching is done by the driver model, yes but I thought about maybe making drm_dev_register() to fail

07:34 <javierm> I thought about setting a DRIVER_SIMPLE or something in struct drm_driver .driver_features, but being an u32 there's no much room there to add another flag

07:35 <javierm> tzimmermann: maybe is a silly idea but I thought that what he asked made sense

07:35 adjtm has quit [Remote host closed the connection]

07:35 <javierm> tzimmermann: another option is to make this a udev rule. Since simpledrm would have to be built-in, we can prevent any drm module to be loaded

07:35 <tzimmermann> javierm, as i said, native drivers have to call drm_aperture_remove_framebuffer() very early in their probe callback. if this fails, the native driver should fail as well.

07:36 adjtm has joined #dri-devel

07:36 <javierm> tzimmermann: yeah, but I agree that this feels hackish

07:36 <tzimmermann> a module parameter could be added to drm_aperture.c and drm_aperture_remove_framebuffer() would always fail if it's set

07:37 <javierm> tzimmermann: Ok, if you think that could be a reasonable option I'll look at writing a patch

07:39 camus1 has joined #dri-devel

07:39 <tzimmermann> javierm, yes please post a patch and let's see what other have to say

07:39 <javierm> tzimmermann: I will. Thanks a lot for your help!

07:39 camus has quit [Ping timeout: 480 seconds]

07:40 <tzimmermann> i can see the usecase at least, and the aperture helpers is probably the cleanest way of adding this

07:40 <tzimmermann> javierm, but there's still a bit of work to do for simpledrm. is it ready for fedora?

07:40 <javierm> tzimmermann: yeah, the usecase made a lot of sense to me. And I've seen cases where video output is not working with the DRM drivers but for example efifb works well

07:41 <javierm> tzimmermann: that's what we want to figure out, if there are still gaps and what these are

07:41 <javierm> for Fedora you make a change proposal and if accepted, try to do the changes and if doesn't work there's a contingency plan, i.e: https://fedoraproject.org/wiki/Changes/ReplaceFbdevDrivers

07:43 <tzimmermann> javierm, there's no flicker-free booting yet. this requries to read the display state backinto drm. i worked on drm helpers to do that, but it's not yet done

07:43 dongwonk has quit [Remote host closed the connection]

07:44 <tzimmermann> javierm, and xorg doesn't autoconfigure platform drivers. if you do startx with simpledrm, you have to add a configure file

07:45 <tzimmermann> that's the main things IIRC.

07:45 <tzimmermann> for suse, the xorg thing is still a showstopper. i work on mitigating that

07:46 adjtm has quit [Remote host closed the connection]

07:46 <javierm> tzimmermann: I see. Fedora has been using wayland by default for a while so the xorg may not be a showstopper

07:46 adjtm has joined #dri-devel

07:46 <javierm> in fact, simpledrm is an improvement because mutter doesn't have a fbdev backend, hence not being usable with efifb but should work with simpledrm

07:46 <javierm> the flicker-free booting may be an issue though. Thanks for the info

07:47 <tzimmermann> javierm, yeah simpledrm is an improvement in most ways. but some of our clients cannot replace xorg easily.

07:48 <javierm> tzimmermann: this is for F36 and the beta freeze is in 4 months so I believe there are still time to fix the remaining issues. And as mentioned in worst case the change could be reverted

07:49 <tzimmermann> javierm, i have similar plans for tumbleweed (and later suse's enterprise distro). that timeframe sounds about right

07:49 JohnnyonFlame has quit [Read error: Connection reset by peer]

07:50 <javierm> tzimmermann: cool

07:51 <tzimmermann> javierm, now that nvidia starts offering gbm/wayland support, using wayland+xwayland might be an option.

07:51 <tzimmermann> it's really just the xorg backend code that sucks

07:51 <javierm> tzimmermann: yes, that's what Fedora does. And yeah, nvidia used to be the only reason for people using xorg but as you said that's improving

07:52 <tzimmermann> great! i'm happy to see things moving forward :)

07:53 <javierm> tzimmermann: indeed, and having simpledrm is awesome for this :)

07:54 <javierm> now we "just" need to replace fbcon so we could disable fbdev and the vt console support completely

07:55 <javierm> tzimmermann: anyway, I'll let you know how the Fedora experiment goes

07:55 <tzimmermann> sure

07:57 <tzimmermann> javierm, wrt fbcon. i'd like to revive kmscon. i simply don't have the time to do so.

07:59 <javierm> tzimmermann: yeah, same. I also have forward ported david's drm_log patch but didn't have time to properly test it yet to post to the list

07:59 pcercuei has joined #dri-devel

07:59 <tzimmermann> nice :)

08:01 rasterman has joined #dri-devel

08:01 <pq> good plans!

08:02 pcercuei has quit []

08:03 pcercuei has joined #dri-devel

08:18 Ahuj has joined #dri-devel

08:40 gawin has joined #dri-devel

08:42 <danvet> tzimmermann, ack

08:42 <tzimmermann> danvet, thanks

08:43 <tzimmermann> danvet: well, ack for the patch or for the discussion?

08:43 <danvet> tzimmermann, ack on the revert

08:43 <danvet> I didn't see a ping on the discussion

08:44 <tzimmermann> davnet, so thanks again. there's certainly a deeper issue that needs to be addressed. but i'm on pto next week. no time to look into it today

08:45 <danvet> tzimmermann, yeah the patch looks very funny, but if we don't understand what's going on then revert gets an ack

08:45 <danvet> tzimmermann, while you're here, quick ack on this fix of mine? https://lore.kernel.org/dri-devel/20211021202048.2638668-1-daniel.vetter@ffwll.ch/

08:50 camus1 has quit [Ping timeout: 480 seconds]

08:52 camus has joined #dri-devel

09:06 <tzimmermann> danvet, a-b: me

09:07 <danvet> thx

09:09 Lucretia has joined #dri-devel

09:10 aravind has quit [Ping timeout: 480 seconds]

09:55 camus1 has joined #dri-devel

10:00 camus has quit [Ping timeout: 480 seconds]

10:05 <airlied> tzimmermann: i applied kims revert patch to fixes

10:05 <airlied> today for Linus, seemed the safest option

10:05 <tzimmermann> airlied, ok. i wanted to push this in a few minutes

10:07 <airlied> kim seemwd to indicate the connector detect patch was the causr

10:07 <tzimmermann> airlied, your revert should implicitly revert 572994bf18ff ("drm/ast: Zero is missing in detect function") as well

10:07 <tzimmermann> there was a report about this too. but i think it's the same root cause

10:08 <tzimmermann> i just don't have the time this week to investigate and won't be around next week

10:08 adjtm has quit [Quit: Leaving]

10:09 <tzimmermann> airlied, as you applied kim's patch, i consider this done for now

10:09 <tzimmermann> thanks for taking care of it

10:10 <airlied> tzimmermann: Sounda good to me

10:21 Daaanct12 is now known as Danct12

10:22 Company has joined #dri-devel

10:39 X-Scale` has joined #dri-devel

10:45 X-Scale has quit [Ping timeout: 480 seconds]

10:54 camus has joined #dri-devel

10:56 flacks has quit [Quit: Quitter]

10:59 camus1 has quit [Ping timeout: 480 seconds]

10:59 flacks has joined #dri-devel

11:09 X-Scale has joined #dri-devel

11:13 X-Scale` has quit [Ping timeout: 480 seconds]

11:16 tursulin has joined #dri-devel

11:44 f11f13 has joined #dri-devel

11:47 f11f12 has quit [Ping timeout: 480 seconds]

11:50 camus1 has joined #dri-devel

11:54 camus has quit [Ping timeout: 480 seconds]

12:15 itoral has quit []

12:20 camus1 has quit [Ping timeout: 480 seconds]

12:25 camus has joined #dri-devel

12:35 vivijim has joined #dri-devel

12:41 mareko has quit [Remote host closed the connection]

12:41 mareko has joined #dri-devel

12:41 dri-logger has quit [Remote host closed the connection]

12:41 glisse has quit [Remote host closed the connection]

12:41 glisse has joined #dri-devel

12:41 mslusarz has quit [Remote host closed the connection]

12:46 dri-logger has joined #dri-devel

12:46 mslusarz has joined #dri-devel

12:50 Rakko has joined #dri-devel

12:50 Rakko has left #dri-devel [#dri-devel]

12:51 omakk has joined #dri-devel

12:53 oakk has joined #dri-devel

12:53 omakk has quit []

13:01 sdutt has joined #dri-devel

13:04 iive has joined #dri-devel

13:16 unsolo has quit [Ping timeout: 480 seconds]

13:16 ced117_ has quit [Ping timeout: 480 seconds]

13:19 flto has quit [Remote host closed the connection]

13:20 flto has joined #dri-devel

13:24 unsolo has joined #dri-devel

13:44 sdutt has quit []

13:44 sdutt has joined #dri-devel

13:46 fxkamd has joined #dri-devel

14:07 camus1 has joined #dri-devel

14:10 gawin has quit [Ping timeout: 480 seconds]

14:10 camus has quit [Ping timeout: 480 seconds]

14:11 pekkari has joined #dri-devel

14:25 jkrzyszt has quit [Ping timeout: 480 seconds]

14:28 Ahuj has quit [Ping timeout: 480 seconds]

14:35 shashanks has joined #dri-devel

14:36 hansg has joined #dri-devel

14:42 camus1 has quit [Ping timeout: 480 seconds]

14:44 elongbug has quit [Ping timeout: 480 seconds]

14:44 nchery has joined #dri-devel

14:49 kenjigashu has joined #dri-devel

15:18 tobiasjakobi has joined #dri-devel

15:18 tobiasjakobi has quit [Remote host closed the connection]

15:19 pekkari has quit [Quit: Konversation terminated!]

15:31 kenjigashu has quit []

15:31 boistordu has joined #dri-devel

15:43 shashanks has quit [Remote host closed the connection]

15:44 camus has joined #dri-devel

15:44 ced117 has joined #dri-devel

15:45 shashanks has joined #dri-devel

15:54 rgallaispou has quit []

15:54 tursulin has quit [Read error: Connection reset by peer]

15:54 dongwonk has joined #dri-devel

16:05 kenjigashu has joined #dri-devel

16:05 kenjigashu has quit [Remote host closed the connection]

16:25 heat has joined #dri-devel

16:39 mbrost has joined #dri-devel

16:43 ybogdano has joined #dri-devel

17:00 f11f13 has quit [Remote host closed the connection]

17:00 f11f13 has joined #dri-devel

17:04 gouchi has joined #dri-devel

17:25 gawin has joined #dri-devel

17:44 gawin has quit [Ping timeout: 480 seconds]

17:51 nsneck has joined #dri-devel

17:52 ybogdano has quit [Quit: Leaving]

17:54 ybogdano has joined #dri-devel

18:02 kenjigashu has joined #dri-devel

18:10 kenjigashu has quit []

18:13 pendingchaos has quit [Ping timeout: 480 seconds]

18:20 pendingchaos has joined #dri-devel

18:22 mlankhorst has quit [Ping timeout: 480 seconds]

18:26 kenjigashu has joined #dri-devel

18:29 shashank_sharma has joined #dri-devel

18:31 shashank_sharma has quit []

18:37 pendingchaos has quit [Quit: Quit]

18:37 pendingchaos has joined #dri-devel

18:50 kenjigashu has quit []

18:51 gouchi_ has joined #dri-devel

18:51 gouchi has quit [Read error: Connection reset by peer]

18:56 gouchi_ has quit [Remote host closed the connection]

19:09 tzimmermann has quit [Quit: Leaving]

19:29 kenjigashu has joined #dri-devel

19:29 kenjigashu has quit [Remote host closed the connection]

19:32 nchery has quit [Quit: Leaving]

19:34 kenjigashu has joined #dri-devel

19:37 heat has quit [Ping timeout: 480 seconds]

19:46 hansg has quit [Quit: Leaving]

19:54 thellstrom1 has joined #dri-devel

19:57 danvet has quit [Ping timeout: 480 seconds]

20:00 thellstrom has quit [Ping timeout: 480 seconds]

20:07 nchery has joined #dri-devel

20:08 nchery has quit []

20:08 nchery has joined #dri-devel

20:09 ppascher has quit [Ping timeout: 480 seconds]

20:09 nchery is now known as Guest3804

20:09 nchery has joined #dri-devel

20:09 Guest3804 has quit []

20:10 vivijim has quit [Ping timeout: 480 seconds]

20:20 danvet has joined #dri-devel

20:29 rasterman has quit [Quit: Gettin' stinky!]

20:34 pnowack has quit [Quit: pnowack]

20:52 vivijim has joined #dri-devel

20:56 kenjigashu has quit []

21:04 kenjigashu has joined #dri-devel

21:05 kenjigashu has quit [Remote host closed the connection]

21:05 nchery has quit [Quit: Leaving]

21:05 nchery has joined #dri-devel

21:11 DPA- has quit [Ping timeout: 480 seconds]

21:21 danvet has quit [Ping timeout: 480 seconds]

22:01 thellstrom has joined #dri-devel

22:01 thellstrom1 has quit [Remote host closed the connection]

22:06 Duke`` has quit [Ping timeout: 480 seconds]

22:18 DPA has joined #dri-devel

22:50 fxkamd has quit [Remote host closed the connection]

22:56 X-Scale` has joined #dri-devel

23:00 X-Scale has quit [Ping timeout: 480 seconds]

23:04 anholt__ has joined #dri-devel

23:10 anholt_ has quit [Ping timeout: 480 seconds]

23:13 alyssa has joined #dri-devel

23:17 iive has quit []

23:26 pcercuei has quit [Quit: dodo]

23:27 <alyssa> anholt__: imirkin: should "mesa: Require MRT support for GL3/ES3" be cc stable?

23:28 <alyssa> on one hand - it is an unambiguous bug fix

23:28 <alyssa> on the other hand - if users are accidentally relying on the es3/gl3 we were accidentally exposing on t720 (and anyone else..?), that's breakage and that seems mean to backport

23:29 <alyssa> Maybe should be sent to 21.3 since that's still in pre-release, but specifically not backported to 21.2?

23:35 anholt__ is now known as anholt

23:35 <anholt> alyssa: that seems like a good compromise to me

23:35 <alyssa> ack

23:53 ybogdano has quit [Ping timeout: 480 seconds]

23:59 * jekstrand thinks his new Vulkan sync code might be working....