ChanServ changed the topic of #linux-msm to:
davidinux is now known as Guest6731
davidinux has joined #linux-msm
Guest6731 has quit [Ping timeout: 480 seconds]
davidinux has quit [Quit: WeeChat 3.5]
junari_ has joined #linux-msm
junari_ has quit [Read error: Connection reset by peer]
jhovold has joined #linux-msm
<Degdag_Med[m]> <aka_[m]> "6.5 is so pain...." <- > <@aka_:matrix.org> 6.5 is so pain.... (full message at <https://matrix.org/_matrix/media/v3/download/matrix.org/GEzYbYaZzdtoSbBDqcjiNakn>)
<aka_[m]> Degdag_Med[m]: > <@degdag:matrix.org> I had the same issue on 6.5-rc1... (full message at <https://matrix.org/_matrix/media/v3/download/matrix.org/JwRwXLruesmQqsFLRJjjybIC>)
<aka_[m]> til, defining dmas on BLSP while having them disabled leaves kernel probe defer not only blsps but also usb if you use extcon
<aka_[m]> Tooniis: do you get usb up on latest rc?
<Tooniis[m]> haven't tried it
<aka_[m]> it feels for me like bam-dma don't even probe anymore
<aka_[m]> ok im dumb
junari has joined #linux-msm
<aka_[m]> any idea why i get configfs-gadget showing me it configure usb yet i get no detection on pc?
<aka_[m]> msm_hsusb 78db000.usb: Failed to create device link (0x180) with 1-0047
<aka_[m]> seems its working now somehow
<aka_[m]> lol
<aka_[m]> im dumb fuck once again ngl
<aka_[m]> once i enable gpu tsens start shuting device off
<aka_[m]> lol
<aka_[m]> [ 52.486459] thermal thermal_zone6: cpu6-thermal: critical temperature reached, shutting down
<aka_[m]> thermal_zone6 4 7 8
<aka_[m]> Affe Null: everything dies
AffeNull[m] has joined #linux-msm
<AffeNull[m]> what dies? when?
<aka_[m]> when i try to enable adsp
<aka_[m]> with q6afecc patches
<AffeNull[m]> maybe the clocks don't get enabled
<aka_[m]> when i tried to use api v2 before it spammed me log
<aka_[m]> atleast
<aka_[m]> now its nothing
<aka_[m]> Affe Null: wait
<aka_[m]> i need to read patches
<aka_[m]> do you do mclk enable too?
<AffeNull[m]> which SoC is this?
<aka_[m]> Affe Null: mclk fails to enable it seems
<aka_[m]> 8976
<aka_[m]> [ +1.023871] qcom-q6afe aprsvc:service:4:4: AFE set params failed -110
<aka_[m]> [ +0.000032] q6afe-dai c200000.remoteproc:smd-edge:apr:service@4:dais: ASoC: error at snd_soc_dai_set_sysclk on PRI_MI2S_RX: -110
<aka_[m]> [ +0.000021] qcom-apq8016-sbc c051000.sound: Failed to disable LPAIF bit clk: -110
<aka_[m]> for sure its wrong
<aka_[m]> i seen this error for year long
<AffeNull[m]> did you apply the device tree changes I made for 8916?
<aka_[m]> not on 8916 obv but on mine its there
<aka_[m]> i believe we might get into some race
<aka_[m]> when card probe before wcd enabled mclk
<aka_[m]> i guess api does not return status on enable/disable
<aka_[m]> imma just start printing what it found as API
<aka_[m]> inside card
<aka_[m]> so we can see if it even detects it properly
<aka_[m]> havent imported apq changes
<aka_[m]> lol
<AffeNull[m]> <aka_[m]> "[ +1.023871] qcom-q6afe aprsvc..." <- > <@aka_:matrix.org> ```... (full message at <https://matrix.org/_matrix/media/v3/download/matrix.org/iTcKrHsAHylUHSaSRQgDDaDD>)
<aka_[m]> with these 3 patches
<aka_[m]> shouldn't wcd call q6afecc to set mclk?
<aka_[m]> before i did it from apq8016 itself
<aka_[m]> ah wait you do voting from q6afe itself?
<aka_[m]> so whatever it takes from apq it will convert based on fetched clock
<AffeNull[m]> these errors don't seem related to mclk
<aka_[m]> they are
<aka_[m]> thats what you get when you don;t set mclk
<aka_[m]> ah wait we are in irc
<AffeNull[m]> so the apq8016_sbc driver is trying to set the bit clk and fails because the mclk is not set?
<aka_[m]> yea
<aka_[m]> now how we do take care of ensuring digital_codec_clk is set
<AffeNull[m]> add it to the sound card's assigned-clocks?
<aka_[m]> its already under
<aka_[m]> lpass_codec: audio-codec@c0f0000
<aka_[m]> ?
<aka_[m]> probe order should be lpass->q6afe->wcd->card
<aka_[m]> trying with it added to qdsp6 card tho im not sure if its going to fly
<aka_[m]> still hangs badly and reboot
<krzk> aka_[m]: sound card should have dai links to the codec, so it won't probe earlier
<krzk> if you suspect something probes in wrong order, you probably miss some properties / clocks / dai links etc
<aka_[m]> in this case obv i know its mclk
<aka_[m]> because we are trying to get q6afe to be smarter about supported afe clk api
<aka_[m]> now i wonder why it does not set mclk
<aka_[m]> i could print when it execute specific set clock command i think
<AffeNull[m]> maybe my patch is wrong and doesn't send the right command to set the mclk
<aka_[m]> or 8976 is avs 2.7 and it also uses v1
<aka_[m]> i need to verify
<aka_[m]> what it reads as fw version
<aka_[m]> and what it does
<AffeNull[m]> then it is 2.6 and not 2.7
<aka_[m]> Affe Null: i think i know what happens
<aka_[m]> for digital codec clock you always send packet in context of port
<aka_[m]> for mclk you can just use primary one
<aka_[m]> and on yours it sets with NULL
<AffeNull[m]> yes, I know
<AffeNull[m]> I thought this would work, but maybe it doesn't
<aka_[m]> cannot you just copy LPAIF_DIG_CLK behavior?
<AffeNull[m]> Where do I get the primary port?
<aka_[m]> ¯\_(ツ)_/¯
<AffeNull[m]> actually we still need to call q6afe_port_put somewhere if we do this
<aka_[m]> yup its crying
<AffeNull[m]> doesn't work?
junari_ has joined #linux-msm
<aka_[m]> error: passing argument 1 of 'q6afe_port_get_from_id' from incompatible pointer type
<aka_[m]> afe->dev ?
<AffeNull[m]> yes
junari has quit [Ping timeout: 480 seconds]
<aka_[m]> weird
<aka_[m]> its still wrong
<aka_[m]> but does not reset anymore
<aka_[m]> q6afe-dai c200000.remoteproc:smd-edge:apr:service@4:dais: AFE Port already open
<aka_[m]> [ +0.000010] MultiMedia3: ASoC: error at dpcm_fe_dai_prepare on MultiMedia3: -22
<aka_[m]> Affe Null: apq card lacking quin support
<AffeNull[m]> "AFE Port already open" is expected because `q6afe_port_put` wasn't called
<aka_[m]> still doesnt want to work somehow
<aka_[m]> seems like i have dai disabled in dts
<aka_[m]> Affe Null:
<AffeNull[m]> looks like it's working now
<AffeNull[m]> right?
<aka_[m]> "depends"
<aka_[m]> i get noise on speaker test
<aka_[m]> ima try running mp3
<aka_[m]> maybe some pinctrls?
<aka_[m]> Affe Null: default sound for speaker test it seems
<aka_[m]> 🚢 ?
<aka_[m]> besides this we need also to add quin to apq-sbc
<aka_[m]> would be nice to get it for 6.6 so we can wire 8976/8953
<aka_[m]> and you can go for 8917
<aka_[m]> 👍️
<aka_[m]> im also going to clean my trees
junari_ has quit [Ping timeout: 480 seconds]
<aka_[m]> have you check if cpufreq also going really fast on your devices?
<aka_[m]> it switches freqs so often
<AffeNull[m]> I haven't checked anything cpufreq-related yet
<aka_[m]> i should probably dump all these freqs
<aka_[m]> leave just 3 or so to have cpr bins assigned to them
<aka_[m]> now i wonder if i can drop assigned clocks i set before
<aka_[m]> i wonder why OS just don't turn off A72 cores instead of keeping switching them
<aka_[m]> Affe Null: hmm
<aka_[m]> what about compatibles, do we really need new ones?
<aka_[m]> and device specific
<aka_[m]> what about msm89XX one
<aka_[m]> now 8916 one can be used universaly
<AffeNull[m]> no wildcard compatibles
<aka_[m]> uh
<aka_[m]> cannot we just do "qcom,msm8953-qdsp6-sndcard", "qcom,msm8916-qdsp6-sndcard" with first one being just documented?
<AffeNull[m]> yes
<AffeNull[m]> we don't need to add compatibles to the driver
<aka_[m]> i believe main reason to have device specific compats is so dts can be used by other kernels/oses with their own drivers
<aka_[m]> and be unique enuff
<aka_[m]> krzk: mind showing "da wae"
jhovold has quit [Ping timeout: 480 seconds]
junari has joined #linux-msm
<aka_[m]> i will probably resend after getting r-b on rest
<aka_[m]> oh huh, turned out i lost some fixes flags
<aka_[m]> lol
<aka_[m]> have to resend anywat
<aka_[m]> and tags locations f me
* aka_[m] posted a file: v2-0004-dt-bindings-clock-qcom-hfpll-Document-MSM8976-com.patch (1KiB) < https://matrix.org/_matrix/media/v3/download/matrix.org/vLZFuCxZpTxTIWxwWvkxzUhI >
<aka_[m]> is that ok?
<aka_[m]> aaaaaaaaa
<aka_[m]> i feel so stressed now
<aka_[m]> ...
junari has quit [Ping timeout: 480 seconds]
<aka_[m]> Marijn: indeed there is more IPC to be shifted, will send v3 tomorrow
<Marijn[m]> No rush, give other reviewers some time...
<aka_[m]> yea would like to get things reviewed before v3
<Marijn[m]> aka_: do you still need me to send MDSS for 8976?
<aka_[m]> unless iommu drops it doesnt matter much
<aka_[m]> we cannot utilize mdss without iommu or atleast its not happy there
<aka_[m]> i will go for rprocs once Otto patches get pushed
<Marijn[m]> True that... I think Konrad resent it once
<aka_[m]> CPR maybe but it doesn't work to well with current shedulers
<aka_[m]> *too
<Marijn[m]> However, we could have the DTS and just not status=okay it on our boards until the iommu changes land...
<aka_[m]> i get what you mean
<aka_[m]> like msm8974 was
<Marijn[m]> And posting it as a dependency may put some more pressure...
<Marijn[m]> I was not able to look into the remaining iommu review and changes myself though, it has been too long and DPU needs way too much attention.
<Marijn[m]> Fortunately Konrad did most of that :)
<aka_[m]> don't force yourself summer is bad for working
<aka_[m]> hmm i wonder if this small mismatch of ipcs is reason why wcnss keeps failing
<aka_[m]> these are intc of remoteprocs
<aka_[m]> so prob matters a lot
<aka_[m]> before i had my own base dts so it was working
<aka_[m]> albeit crashing on idle power save mode
<Marijn[m]> ungeskriptet: you have a non-DSC dual-DSI cmdmode panel on sm8250 right? I'd love to play some with that before bringing DSC into the mix for my own devices... Did you share a kernel branch (device tree + panel drivers + other patches) before?
<Marijn[m]> aka_: it's not that hot here... And "summer holiday" isn't exactly a thing anymore
<ungeskriptet> Yes, it's dual DSI cmdmode panel without dsc
<aka_[m]> this week is okay but like last one was constant 35C
<ungeskriptet> Marijn: I have my testing branch here: https://github.com/sm8250-mainline/linux/tree/sm8250/dsi-experiments
Danct12 has quit [Remote host closed the connection]
Danct12 has joined #linux-msm
<Marijn[m]> ungeskriptet: oh nice, that seems to contain my latest changes to DPU, and the panel and dts look good (if you un-comment dsi1)
<Marijn[m]> ungeskriptet: did you get it to work on dsi0 only, by also changing `has_dual_dsi=false` in the panel driver?
<ungeskriptet> Marijn[m]: No, I did not try that yet
<Marijn[m]> Since you commented out dsi1 in dts... you'll also need to change your driver to cope with that
<Marijn[m]> FWIW I've used the optionality of the port node for that in my driver. If of_graph_get_remote_node() returns NULL there's no second DSI node and I only initialize on dsi0
<Marijn[m]> (That should probably be an error later as the panel really requires dual-dsi, but helpful for back-to-back testing single vs dual DSI)
<Marijn[m]> We're not entirely confident that a dual-DSI panel should work on only one half when DSC is involved though... But it should without DSC
<ungeskriptet> I can try messing with the panel tomorrow
<JIaxyga[m]> Marijn: Dsc should work on next branch?
<ungeskriptet> I did however try using the lmdpdg generated driver since it only generates single dsi panel drivers
<ungeskriptet> But that attempt was unsuccessful too
<Marijn[m]> JIaxyga: DSC should work perfectly fine after the last round of reviews (as long as you hack the transfer time in the panel driver by pretending they're porches...), but it was developed with a limited scope in mind and not designed for a dual-DSI (or dual-INTF topology for that matter) at all
<Marijn[m]> I've done multiple passes at implementing that but have yet been unsuccessful
<Marijn[m]> ungeskriptet: if you generate a half-panel driver (you can also reuse the current one...) be sure to use half-width in the mode
<JIaxyga[m]> Marijn[m]: It's just that I can test it on surya sm7150 as soon as I'm done with dpu
<JIaxyga[m]> It uses single dsi as far as I understand
<JIaxyga[m]> So there shouldn't be a problem
<Marijn[m]> If it does you should be in luck
<Marijn[m]> Which DPU revision does it use, is it close to sm8150 (so DPU v5.x)?
<JIaxyga[m]> Marijn[m]: dpu 5.2
<JIaxyga[m]> A little mess
<Marijn[m]> Okay so DSC 1.1
<Marijn[m]> Clean it and send it, the earlier the better ;)
<JIaxyga[m]> Dpu already worked on v6.2, just need to migrate to the latest version
<Marijn[m]> Sure, it's just that there are more and more improvements/refactors planned, just wanting to save you time :)