Call for help about issue with gbm_gralloc

[Mauro Rossi]

Hello,

r-x86 (Android R) release and s-x86 (Android S) release

will not use drm_gralloc due to its lack of support for binderization/dma_bufs

We are currently testing gbm_gralloc and minigbm as candidates for replacing drm_gralloc as default gralloc HAL

gbm_gralloc advantage is that it should work Out Of The Box with mesa drivers and does not require implementation of per driver backend; the only gap is that it is affected by a bug which causes stride/alignement artifacts in SW_WRITE_OFTEN buffers, the problem is particularly visible in s-x86 (Android S) where Skia is more used and many common Window message panes show the artifacts and Windows panes are not readable.

Due to complexity (at least for my skills point of view) I was not able to clearly identify the problem, but I believe the root cause is in the area of stride value setting at buffer allocation, or an issue in retrieving the correct stride at the Virtual Address mmapping required to write onto the buffers from another process.

minigbm does not show this problem, at first examination of the code paths stride is always aligned to 64/128/256 and this may prevent the artifacts from happening.

If the problem is only to align the stride mod 64/128/256 it is very strange that the issue was not noticed nor fixed before in gbm_gralloc

I wanted to try to ask to android-x86 community if someone could help to shed some light, provide suggestions or contribute to solving this problem, in order to have r-x86 and s-x86 releases not affected by annoying artifacts

Thanks for any info/suggestions

Mauro

 

[fguy]

Hello Mario

I periodically continue to collect intel celadon 11 and 12. I did not delve into the intricacies of the implementation of graphics, but both versions work stably and do not have noticeable artifacts when displaying applications and the android interface. Maybe it makes sense to adapt celadon 11 to load as android x86 from partition images - this is a stable assembly with codecs and correct video display on web pages in chrome. Houdini from ChromeOS or WSA is also available for it. Unfortunately, my experience is not enough to implement such a loading method.

Android 12 in celadon is also stable, but it needs to be built with kernel 5.4 - since 5.10 there are problems in the video with picture and sound lags. Unfortunately, there is no "donor" houdini for it, and in SDK there is only ndk for arm64. Android Sv2 also appeared with a bias for tablets - its interface would be more acceptable for android x86.

MS WSA in win11 is also an interesting solution with android 11 and houdini, but to use it you need to have at least 8 GB of RAM and a modern processor.

I am glad that you continue to work for the android x86 project - lately I got the impression that the project is rather dead than alive.

пятница, 29 октября 2021 г. в 22:16:39 UTC+3, issor...@gmail.com:

[Huy Minh Bùi]

Celadon use minigbm and we can already make it work on Android-x86, now we need to make gbm_gralloc less buggy like what Mauro said (his name is not Mario btw)

Vào lúc 22:06:48 UTC+7 ngày Thứ Bảy, 30 tháng 10, 2021, fguy đã viết:

[fguy]

sorry, google translator beguiled, but there is no way to correct the text of the message

суббота, 30 октября 2021 г. в 18:38:38 UTC+3, buin...@gmail.com:

[Huy Minh Bùi (HMTheBoy154)]

Update on the situation : the problem is being fixed, both gbm_gralloc and minigbm with gbm_gralloc as a backend (minigbm_gbm_mesa). 

https://github.com/BlissRoms-x86/gbm_gralloc/commit/e70acfc3f5acc64168a172987f34d141e55f6b54

The fix was made by aleasto from WayDroid team, a better fix for minigbm_gbm_mesa was made by aleasto and rsglobal (Roman Stratiienko) 

https://github.com/rsglobal/sandbox/tree/gbm_mesa/gbm_mesa_driver

Both me, Mauro and some people over BlissOS community confirmed that the fix work. So if anyone is trying to build r-x86, please notice this.

Vào lúc 02:16:39 UTC+7 ngày Thứ Bảy, 30 tháng 10, 2021, issor...@gmail.com đã viết:

[nownownownownownownownownownownownownownownownownownownownownownownownownownownownownownownownownownownownownownownownownownownownownownownownownownownownownownownownownownownownownownownownownownownow]

android only works on phones and tablets, might as well quit
developement as android 10 is long overdue and you are lacking security
fixes and accessibility features, curnal coconuts told me everything

[Huy Minh Bùi (HMTheBoy154)]

1. If you said that "android only works on phones and tablets," then you are living under a rock or you are trying to be stupid in this group.

2. This is not the only place to track Android-x86 updates and development, you don't even know about the development status and then just tell everyone to quit is very disrespect. 

Vào lúc 13:20:32 UTC+7 ngày Thứ Ba, 25 tháng 10, 2022, nownownownownownownownownownownownownownownownownownownownownownownownownownownownownownownownownownownownownownownownownownownownownownownownownownownownownownownownownownownownownownownownownownownow đã viết:

[macintosh]

its easy. use a real phone or tablet

--
You received this message because you are subscribed to the Google Groups "Android-x86" group.
To unsubscribe from this group and stop receiving emails from it, send an email to android-x86...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/android-x86/074790a4-0a25-4473-ae12-fd1bcfd43163n%40googlegroups.com.

[RoyalGraphX]

dude you’re in the fkn android for x86 thread, what’s your issue Jordan?

On Tue, Oct 25, 2022 at 10:37 AM ########################## <jordanli...@hotmail.com> wrote:

not my issue, its yours, android is ment for phones!

On 25/10/2022 10:44, nownownownownownownownownownownownownownownownownownownownownownownownownownownownownownownownownownownownownownownownownownownownownownownownownownownownownownownownownownownownownownownownownownownow wrote:

nownownownownownownownownownownownownownownownownownownownownownownownownownownownownownownownownownownownownownownownownownownownownownownownownownownownownownownownownow

On 25/10/2022 10:39, Huy Minh Bùi (HMTheBoy154) wrote:

--

You received this message because you are subscribed to the Google Groups "Android-x86" group.
To unsubscribe from this group and stop receiving emails from it, send an email to android-x86...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/android-x86/074790a4-0a25-4473-ae12-fd1bcfd43163n%40googlegroups.com.

--
You received this message because you are subscribed to the Google Groups "Android-x86" group.
To unsubscribe from this group and stop receiving emails from it, send an email to android-x86...@googlegroups.com.

To view this discussion on the web visit https://groups.google.com/d/msgid/android-x86/AS8P195MB19305DAFC76AA00B0857351FA1319%40AS8P195MB1930.EURP195.PROD.OUTLOOK.COM.

--
You received this message because you are subscribed to the Google Groups "Android-x86" group.
To unsubscribe from this group and stop receiving emails from it, send an email to android-x86...@googlegroups.com.

To view this discussion on the web visit https://groups.google.com/d/msgid/android-x86/AS8P195MB19307573AF847CB967DB5CEDA1319%40AS8P195MB1930.EURP195.PROD.OUTLOOK.COM.

[Chih-Wei Huang]

I've banned the spammer.
Sorry for letting him in.