ChanServ changed the topic of #linux-msm to:
MollySophia[m] has joined #linux-msm
<MollySophia[m]> <Marijn[m]> "ungeskriptet adomerle Molly..." <- Wow awesome!
<MollySophia[m]> May try that on pipa this week
minecrell has quit [Read error: Connection reset by peer]
minecrell has joined #linux-msm
Daanct12 has joined #linux-msm
Daanct12 has quit [Quit: WeeChat 4.2.1]
Daanct12 has joined #linux-msm
Daanct12 has quit [Quit: WeeChat 4.2.1]
Daanct12 has joined #linux-msm
marvin24 has joined #linux-msm
marvin24_ has quit [Ping timeout: 480 seconds]
aklimov has quit [Ping timeout: 480 seconds]
rmsilva- has quit [Remote host closed the connection]
rmsilva has joined #linux-msm
Daanct12 has quit [Quit: WeeChat 4.2.1]
Daanct12 has joined #linux-msm
<aka_[m]> i know its not related to msm but is there any irc for memetek?
<aka_[m]> i see linux-mediatek but i would rather call it dead
jhovold has joined #linux-msm
<Daanct12> perhaps mtk just doesnt get much love as qcom
<aka_[m]> I have alldocube g99 tablet on the side and vendor appears to not really give a flying duck about firnware
<aka_[m]> Would be cool to run gnome on it
jhovold has quit [Quit: WeeChat 4.2.1]
jhovold has joined #linux-msm
pespin has joined #linux-msm
<calebccff> lumag: regarding pixel3 display, would be interesting to know what happens if you send a get_compression_mode message, not sure if reading is supported
Daanct12 has quit [Quit: WeeChat 4.2.1]
Daanct12 has joined #linux-msm
mgonzalez has joined #linux-msm
<mgonzalez> bamse: jhugo: did you guys see my DT msm8998 proposed work around for ath10k?
Daanct12 has quit [Quit: WeeChat 4.2.1]
Daanct12 has joined #linux-msm
<mgonzalez> lumag: my apologies: I forgot to CC you on v2 on the ath10k work-around for msm8998
<aka_[m]> huh
<aka_[m]> what they are doing
<mgonzalez> konradybcio: are jamie and angelo still active on MSM stuff?
<lumag> mgonzalez, angelo - no. I haven't seem him sending MSM patches in the last year or two.
<aka_[m]> mgonzalez: Angelo is on memeteks now
<lumag> mgonzalez, regarding msm8998 - n/p
<mgonzalez> "n/p" means "no problem" ? :)
<lumag> yes
<aka_[m]> but you have Konrad
<aka_[m]> he did some 8998
<mgonzalez> aka_[m]: is somainline still active as an org?
<aka_[m]> konradybcio: pushes his branches while working at linaro and Marijn doing display stuff outside of repo
<aka_[m]> and Angelo no longer msm-ing so i would say somainline outside of being staging repo is dead
<mgonzalez> aka_[m]: since konradybcio is the maintainer for several msm8998 DTs, I CCed him on my proposed work-around ;)
<lumag> mgonzalez, if the msa issue is firmware-related, then which firmware is it? I thought that it already happens before loading of wlanmdsp.mbn ?
<mgonzalez> lumag: I'm not sure how to answer your question. I never claimed it is FW related myself.
<lumag> I see
<mgonzalez> I guess I did assume it might be in some FW blob, since qcom does so much in software on remote processors
<mgonzalez> Now that I have a functional system, I suppose I can live test some hypothesis, but I also need to mainline other msm8998 pieces
<mgonzalez> Goal being to have a 99% functional mainline-supported DTS for our board
<mgonzalez> (apq8098 not msm8998, but close enough)
<mgonzalez> aka_[m]: I actually mentioned the postmarketOS webpage for msm8998 in my commit message
<lumag> mgonzalez, do you have a pointer for your log when you tried pushing it to the firmware-N.bin ?
<mgonzalez> It just crashes
<lumag> I think that I saw that in some email...
<mgonzalez> Basically, the field is still NULL at this point
<mgonzalez> Yeah I did dump the logs
<lumag> found it
<mgonzalez> Yup, right before you suggested to use a DT property ;)
<lumag> Ok, so wlanmdsp.mbn was loaded and it is running.
<mgonzalez> OK. Why is that important? Are you considering another way to work around the issue?
<mgonzalez> [ 15.626925] remoteproc remoteproc0: Booting fw image mba.mbn, size 234152
<mgonzalez> No explicit logs for other mbn files
<lumag> mgonzalez, wlanmdsp is loaded via tqftp, so you can check for it in syslog / journalctl
<lumag> Well, krzk raised a valid question.
<lumag> But I also can't find a corresponding piece in msm-4.4 kernel
<mgonzalez> not easy to correlate kernel timestamps with syslog dates :(
<mgonzalez> I had a patch somewhere to change syslog timestamps
<lumag> I see that the driver seems to always send the MSA_READY and then wait for the result
<lumag> dmesg -T ?
<mgonzalez> oh boy, never knew about that (embarrassed look)
<mgonzalez> I have no logs from tqftp, only qrtr-ns & rmtfs
<lumag> jhugo, do you remember any details by chance? I don't see any kind of special MSA READY handling in msm-4.4
<mgonzalez> lumag: jeff said that the vendor ath10k driver did not wait for MSA_READY
<lumag> As I wrote, I don't see it.
<lumag> See wlfw_msa_ready_send_sync_msg() in msm-4.4 and the calling code
<lumag> calebccff, i didn't try sending that command. I'm not even sure that read commands are supported by the panel.
<mgonzalez> lumag: maybe jeff meant the Android driver.
<lumag> msm-4.4 is the Android kernel
<mgonzalez> I thought there was a "naked linux" version and an "android linux" version.
<lumag> That's LE.soemthing vs LA.something
<mgonzalez> yeah that!
<mgonzalez> lumag: I didn't catch the valid question from kk though...??
<Marijn[m]> mgonzalez aka_ the SoMainline repository is still actively being used. All display stuff that I work on is pushed to that repo, _not_ outside of it?
<Marijn[m]> But thanks for calling me/us dead, effectively
<mgonzalez> Marijn[m]: thanks for clarifying the situation.
<mgonzalez> Marijn[m]: somainline is an embedded linux-focused consultancy org, like baylibre?
<Marijn[m]> mgonzalez: no, we're a bunch of hobbyists/enthusiasts who all happened to work on Qcom-based Sony devices
<mgonzalez> Oh so you were not paid by sony or qcom for all the mainlining effort?
<lumag> mgonzalez, I'd note that 4.4 has been sending host_cap after the MSA
<mgonzalez> what does "MSA" stand for?
<lumag> No idea (yet?)
<mgonzalez> my patch only deals with "MSA_READY indicator"
<Marijn[m]> mgonzalez: I don't know if receiving a device every once in a while counts as payment¸ if it only amounts to yet another thing to "work" on :)
<Marijn[m]> mgonzalez: why all these questions by the way?
<lumag> ?
aklimov has joined #linux-msm
<mgonzalez> Marijn[m]: I'm trying to mainline support for another msm8998 device, and I guess I'll be sending a few somainline patches
<mgonzalez> Marijn[m]: at the moment, I was working on WiFi support, specifically a work-around for the missing MSA_READY indicator.
<Marijn[m]> Anything important that we haven't sent yet? I'm about to grab the 8998 device for a specific panel request
<mgonzalez> Sending is one thing, but some patches are not mainline, right?
<mgonzalez> (I'm on 6.9-rc1)
<lumag> mgonzalez, if you have downstream running, there should be a debugfs for the wlan. Then it has nice 'stats' file which has msa_read_foo counters
<Marijn[m]> mgonzalez: well, if they were on the list at some point (but never) merged you have to be careful to have the right vX tag and address previous review
<Marijn[m]> * mgonzalez: well, if they were on the list at some point (but never merged) you have to be careful to have the right vX tag and address previous review
<mgonzalez> (My TODO list includes venus decoder, HDMI, cpufreq -- off the top of my head)
<mgonzalez> Marijn[m]: if you run the msm8998 device, can you give WiFi a try? :)
<Daanct12> i recently got back to msm stuff and ready to send some patches
<Daanct12> but i decided to hold it because it's a weird platform
<Daanct12> are device bring up without wcnss and stuffs are accepted in lkml?
<mgonzalez> Daanct12: I agree, getting WiFi working is pretty hairy with lots of FW blobs to send to a remoteproc
<mgonzalez> I'll try to test venus today & tomorrow
aklimov has quit [Quit: Leaving]
<mgonzalez> I think phh got it working, but he had to disable low-power mode, and when I submitted a patch to do that in mainline, a qcom maintainer said "that's too intrusive, let's find a better way" then forgot about me :(
aklimov has joined #linux-msm
<Daanct12> mgonzalez: which platform are you working on?
<phh> mgonzalez: correct
<mgonzalez> Daanct12: msm8998 (actually apq8098)
<Daanct12> ah
<Daanct12> how's the gpu situation?
<mgonzalez> I defer to phh :p
<phh> freedreno worked without any issue ootb, though we didn't test much, not even performance
<Daanct12> nice
<Daanct12> as long it runs mainline that's already very good :D
<Marijn[m]> Danct12: wifi is definitely not a requirement to upstream a device!
<Marijn[m]> The requirements are and should be really lenient (not counting strictness in quality of course) so that you have something to iterate on
<Marijn[m]> In fact it's always nicer if things land in small isolated chunks/commits
<Daanct12> ah! so even if initial bring up is just i2c and uart.. that's all?
<krzk> Daanct12: the rule is release early, release often
<krzk> or rather concept, not rule
<aklimov> krzk: even if it's not ready?
Daanct12 has quit [Quit: WeeChat 4.2.1]
<krzk> aklimov: what does it mean "not ready"?
<krzk> Then submitting patches explains how to divide your work and send it...
<mgonzalez> lumag: I don't run the downstream kernel.
<mgonzalez> krzk: hello, I hope the changelog I provided for the msm8998 ath10k work-around clarified the situation?
Caterpillar has joined #linux-msm
<jhugo> lumag: I don't remember any MSA ready issue
<lumag> jhugo, :-D
<jhugo> lumag: I vaguely remember a QMI issue which I thought I upstreamed
<jhugo> I haven't actually used Wifi on the 8998 laptop in a long while. Its possible something broke
<mgonzalez> jhugo: it's the same thing. You even documented it on a web page.
<mgonzalez> (Or not, looks like it was Jamie K who investigated the issue)
<jhugo> mgonzalez: no, this is my documentation - https://github.com/jhugo/linux/blob/5.5rc2_wifi/README
<mgonzalez> Right, I got mixed up
<mgonzalez> I don't understand why there's so much resistance getting a simple work-around accepted :(
<jhugo> mgonzalez: Yeah. I've been skimming the threads.
<mgonzalez> And the double standards... makes baby Jesus cry.
<lumag> mgonzalez, please don't take it as a personal offence or bias. It's not.
<mgonzalez> Not talking about your contribution ;)
<mgonzalez> Although making me read qcom code is pretty cruel & unusual punishment :p
<mgonzalez> lumag: are you saying it is mandatory to compile a vendor kernel to analyze why they don't have the issue / how they work around it?
<aka_[m]> bryanodonoghue: pinctrl funcs on 8976 are different
<aka_[m]> And for fw names I bet I copied 8953
<lumag> mgonzalez, no, it is not mandatory. Sometimes it is a last resort.
<lumag> icnss driver is not that bad
<mgonzalez> mainline ath10k_qmi_driver_event_work looks very similar to downstream ath10k_snoc_driver_event_work
<lumag> msm-4.4 has the icnss driver, it was used with msm8998 as far as I understand
<mgonzalez> I'm looking at drivers/net/wireless/ath/ath10k/qmi.c
<lumag> Hmm. I see that it was only enabled for the mediabox. All other devices used drivers/soc/qcom/icnss.c
<mgonzalez> msm_ath10k_wlan: qcom,msm_ath10k_wlan { compatible = "qcom,wcn3990-wifi";
<mgonzalez> I'm not sure I can intelligently compare mainline & vendor : the user-space tools are not the same
<mgonzalez> From my perspective, the issue seems to be stuck
<mgonzalez> It was noted that the driver waits for an indicator that does not come
<mgonzalez> Not waiting for the indicator works around the issue
<mgonzalez> The same issue affects at least 3 different msm8998 boards. I don't know if they ran exactly the same FW blobs.
<mgonzalez> Downstream DT has 2 nodes using the same reg addresses, I thought that's not even valid...
<lumag> mgonzalez, the first node is disabled by default
<lumag> And it gets enabled only on one device.
<lumag> mgonzalez, I won't insist on jumping into downstream. It's just it well might be that the msm8998 expects slightly different sequence somewhere.
<lumag> strongtz[m], could you please check if this fixes the issue you had with kernel pd-mapper? https://lore.kernel.org/linux-arm-msm/20240402-pmic-glink-fix-clients-v1-0-885440b81c65@linaro.org/
telent has quit [Ping timeout: 480 seconds]
pespin has quit []
telent has joined #linux-msm
jhovold has quit [Ping timeout: 480 seconds]
<aka_[m]> konradybcio: have you verified your lenovo memory layout to mainline dts?
<aka_[m]> i find some differences even sm6115.dtsi vs bengal.dtsi
<aka_[m]> some stuff is even more weird
<aka_[m]> like cdsp ion-heap has memory region reserved and address is different in DS vs mainline and my device don't even have this in fdt at all
<aka_[m]> oh well mainline vs ds is same
<aka_[m]> konradybcio: also regarding sm6115 shouldn't we add vdd-io-bias-supply to sdhci driver as optional sauce?
<aka_[m]> i see its also on bengal younger brother
<aka_[m]> eg Khaje
<aka_[m]> 680