ChanServ changed the topic of #linux-msm to:
pg12_ has joined #linux-msm
pg12 has quit [Remote host closed the connection]
pg12_ is now known as pg12
marvin24_ has joined #linux-msm
marvin24 has quit [Ping timeout: 480 seconds]
z_z3ntu_bouncer has quit [Quit: ZNC 1.8.2 - https://znc.in]
z_z3ntu_bouncer has joined #linux-msm
rmsilva- is now known as rmsilva
Fekz115 has joined #linux-msm
Fekz115 has quit []
Fekz115 has joined #linux-msm
<Fekz115> Hi guys. I'm asking about msm dsi driver. I'm trying to enable panel with 8 bit per pixel and 10 bit per component(8 bpp and 10 bpc). Here is commit added to interfere me: https://github.com/torvalds/linux/commit/d053fbc449c47517b1f6516dbce2f917f2a9f51d
<Fekz115> It says, that actually dsc implemented only for 8 bpp and 8 bpc, but next it calls function:
<Fekz115> That already has mapping for my case:
<Fekz115> Is it mistake or driver still not ready for my panel?
<Fekz115> lumag:
<Marijn[m]> Fekz115: it looks like you're in the same boat as Adrian, who asked the same question in this channel a few days ago
<Fekz115> I joined here just right now, so I don't see what they answered him
<Marijn[m]> I can tell you, because I answered ;)
<Fekz115> Please
<Fekz115> Marijn[m]:
<Marijn[m]> Maybe have some patience, while I write up a reply for you ;)
<konradybcio> fekz115, fwiw this channel is logged
<Marijn[m]> I'll summarize it either way because I don't we wandered every possible avenue yet that isn't "let me send a patch series with 10-bpc support"
<Marijn[m]> Fekz115: first up you're right that the constants for 10-bpc are available in the kernel; they're not only used by Qcom hardware (it's all shared!)
<Marijn[m]> We were trying to first resolve some topology-related issues on his hardware before diving into this issue. We also found that the DSC stream is still 8 bits per pixel, so the total data sent to the panel should be the same. Adrian found a hint about using an 8-bpc mode in his firmware, but didn't find a panel command to change his panel to this. I don't think that might be necessary as the bits_per_component parameter is encoded in the DSI
<Marijn[m]> DSC Picture Parameter Set, which your panel driver should send to the panel.
<Marijn[m]> If it does send that PPS, you might try just setting bpc to 8 and hope that it might just work (or at the very least show colorful corruption, indicating parts are turned on and working at least)
<Marijn[m]> Adrian: did you also try setting bpc to 8 after applying the 1:1:1 topology commit?
<Adrian[m]> Marijn[m]: I am using bpc=8 with 1:1:1 topology
<Adrian[m]> Otherwise it would error due to your patch that was sent here earlier
<Fekz115> Sorry guys, didn't want to annoy you
<Marijn[m]> Adrian: we might try removing all the safeguards and set 10, but I'm sure some code will have to be updated. Likewise, I think the 10-bit DSI format should be added?
<Marijn[m]> It's been a while since I last looked into this, since Sony only provides fake 8-bit panels with FRC
<Adrian[m]> Marijn[m]: I will try removing that forced error and see what 1:1:1 topology with bpc=10 results in. Currently I don't even see jank so I'm not even sure if my configuration is otherwise correct :/
<Marijn[m]> I had hoped you'd at least see corruption on the screen. Are the errors in dmesg exclusively timeouts, or also a few -22s?
<Adrian[m]> I don't see any panel errors until I write a value to that backlight node or until ~5 minutes after boot when the panel seems to think it wants to turn off
<Adrian[m]> This is what it looks like if I don't touch anything and wait for a couple of minutes
<Marijn[m]> Is anything trying to present (e.g. a graphics session)? I'd expect timeouts
<Marijn[m]> If the panel remains black, maybe a panel_on() call is missing?
<Adrian[m]> Marijn[m]: Yeah this rootfs contains phosh, yet there are no timeouts
<Marijn[m]> Adrian: did you ever link your mainline panel driver in chat history? I'd like to take a look
<Marijn[m]> That looks correct on first glance
<Marijn[m]> Last time you were seeing 30fps, so the tear command is definitely not working or the panel is not responding (30 fps is what MDP generates)
<Marijn[m]> But the panel is also not rejecting initial setup commands, only the poweroff or poking at brightness
<Marijn[m]> What I've done sometimes is send a brightness command directly before or after power_on() and that worked
<Adrian[m]> wouldn't that just force it to error early?
<Adrian[m]> <Adrian[m]> "I will try removing that..." <- 1:1:1 with bpc=10 seems to be the same as bpc=8. modtest still reports 30Hz
<Adrian[m]> s/modtest/modetest/
<Marijn[m]> <Adrian[m]> "wouldn't that just force it to..." <- The other dsi commands are sent correctly, or so it seems
<Marijn[m]> So it might be that the panel stops responding after some time/command?
<Marijn[m]> <Adrian[m]> "1:1:1 with bpc=10 seems to be..." <- 30hz comes from sync_cfg_height, if we disable that we'll see the real (lack of) panel vblanls
<Marijn[m]> * In reply to @gelbpunkt:matrix.org
<Marijn[m]> 1:1:1 with bpc=10 seems to be the same as bpc=8. modetest still reports 30Hz
<Marijn[m]> 30hz comes from sync_cfg_height, if we disable that we'll see the real (lack of) panel vblanks
<Adrian[m]> <Marijn[m]> "30hz comes from sync_cfg_height,..." <- > <@marijns95:matrix.org> In reply to @gelbpunkt:matrix.org... (full message at <https://matrix.org/_matrix/media/v3/download/matrix.org/UjPtADAeCOlbpCfFujJvAAaN>)
<lumag> konradybcio, do you by chance have a pending series that pings the MPM timer on idle?