robclark changed the topic of #aarch64-laptops to: Linux support for AArch64 Laptops (Chrome OS Trogdor Devices - Asus NovaGo TP370QL - HP Envy x2 - Lenovo Mixx 630 - Lenovo Yoga C630 - Lenovo ThinkPad X13s - and various other snapdragon laptops) - https://oftc.irclog.whitequark.org/aarch64-laptops
<robclark> does wlan really need 2mb of contiguous memory? That is something that is going to be hard to find on a system that has been running for a while
ellyq has quit []
mcbridematt has joined #aarch64-laptops
ellyq_ has quit []
ellyq has joined #aarch64-laptops
<steev> that, i cannot answer
shawnguo has joined #aarch64-laptops
<steev> oh nice, welcome back shawn :D
martiert_work has quit [Remote host closed the connection]
martiert_work has joined #aarch64-laptops
iivanov has joined #aarch64-laptops
iivanov has quit []
iivanov has joined #aarch64-laptops
shoragan has joined #aarch64-laptops
jhovold has joined #aarch64-laptops
<steev> also, woo, 6.10 is out
<abby> any new config options to enable in 6.10?
<steev> not that i'm aware of
<steev> at least, not if you've been using johan's or my stuff already
<steev> but the defconfig itself is kept to a working state as well
srinik has joined #aarch64-laptops
Caterpillar has joined #aarch64-laptops
hightower2 has joined #aarch64-laptops
Caterpillar has quit [Quit: Konversation terminated!]
jelly has quit []
jelly has joined #aarch64-laptops
ektor5 has quit [Quit: WeeChat 4.3.4]
ektor5 has joined #aarch64-laptops
iivanov has quit [Quit: Leaving...]
<gabertron> robclark: if we don’t specify that kernel cmdline related to cma does that impact the wifi driver perf somehow that you’ve seen? I assume not but unclear why people are specifying that on the cmdline to have a larger cma i guess?
<konradybcio> gabertron we work around that by having a pre-reserved region for cma in dt on 8280 / x1..
<robclark> konradybcio, why does wlan need cma? Isn't there an iommu? Or does it not have enough tlb entries?
<konradybcio> the pcie smmu is opaque to the system unless the hyp is kicked
<konradybcio> on other platforms the pcie is routed through the accessible apps smmu
<robclark> hmm
<robclark> I assume it is smmu-v3 again? IIRC that one is programmed via ring buffer, which should be easy to virtualize
<abby> ooh nice, 6.10.0 fixed the deferred probe for something audio-related i was having
<bluerise> on x1e the SMMUv3 can't be touched, like 8280, unless you escape to EL2 (which works like on 8280)
<bluerise> Did anyone try ITS MSIs?
<bluerise> Wonder if it's an OpenBSD or SoC problem
<travmurav[m]> did anyone already manage to switch to el2 on x1e? iirc it was broken on the devkit a while back
<bluerise> I did! :)
<bluerise> But OpenBSD is only semi-happy, same with x13s
<travmurav[m]> bluerise: nice! with slbounce or the official thing?
<bluerise> oh this one is running gunyah, right?
<bluerise> I used slbounce
<bluerise> used the same diff I was using for x13s
<travmurav[m]> oh cool, neither slbounce nor the official thing worked on the devkit months(?) ago
<bluerise> When I booted up it instantly blew up because... well touching USB makes it reboot on both 8280 and x1e
<travmurav[m]> nice to hear slbounce "just works"
<bluerise> and then... ah last thing was missing interrupts I think
<bluerise> MSIs iirc, but need to check again
<travmurav[m]> bluerise: what device is that btw?
<bluerise> Lenovo Yoga Slim 7x
<travmurav[m]> cool
<travmurav[m]> bluerise: the original firmware is gunyah based yes, but as far as I looked into it, at least with slbounce, you still get proper full takeover
<bluerise> I mean, it doesn't work great, but it looks like I'm in EL2. I can configure/touch the SMMUv3 and it breaks basically the same as x13s
<robclark> hmm, sltest didn't appear to work on my yoga slim.. but I have this weird problem where kb doesn't work in efi shell / grub / etc, so was a bit difficult to poke around with. (Ie. maybe it did manage to draw the green bar, but default overlay position made it not visible, or something like that?)
<bluerise> I ported the code into OpenBSD's bootloader
<bluerise> and yes, the keyboard is polled in UEFI and shit to use
<travmurav[m]> robclark: fwiw sltest draws the green bar as the only owner of the cpu (after ebs)
<travmurav[m]> but also hm, if bluerise has ported the code, possibly there is some bug in mine that prevents it from working on x1e
<travmurav[m]> might as well be something funny with restoring the cpu state in el2, I think I already had some trouble with that before
<travmurav[m]> (though it shouldn't prevent sltest from working, since it's just bare asm code to draw the green bar on the framebuffer)
<travmurav[m]> but I guess the important part is that we have /a/ way to boot into el2, which means it can be fixed and used anywhere if one wants to :>
<robclark> indeed
<travmurav[m]> fwiw I think sltest was also semi-broken on the sc8180 where only single green pixel was drawn after which the code died, but I called that a win as well since /some/ user code was executed in el2
<bluerise> Though I do wonder if our VMM should simply make use of Gunyah APIs
<bluerise> the thing is, all this remote proc work one has to figure out to replace what their EL2 is doing...
<bluerise> and I still don't know why USB makes it crash
<bluerise> probably some EL2 vectors being called that don't exist anymore lol
<bluerise> maybe some EL2 state I didn't properly setup after leaving Gunyah
<travmurav[m]> maybe some missing iommu configurations?
<travmurav[m]> I think linux sets all the context banks that were configured before it as bypass
<bluerise> Oh right I should probably turn SMMUv2 off lol
<travmurav[m]> so when you boot after slbounce you get literally everything bypassed
<bluerise> yeah, we don't touch SMMUv2 because the bypass thingy and re-use didn't properly work :(
<travmurav[m]> (so all remoteprocs, usb, mmc, mdss...)
<travmurav[m]> ... because this is what hyp had configured before takeover
<travmurav[m]> I think I was considering to even do some page walks in software at some point to try and figure out which ids correspond to which IP blocks but ended up putting all that stuff off after realizing I can't seem to get any iommu faults on my 7c anyway
<bluerise> heh
<travmurav[m]> and not like adsp was fetching anything from ram anyway even though the boot fsm reported that it has "booted"
<travmurav[m]> since I think while linux never got any iommu faults, the device would still just reset if a fault has occured, and I wasn't getting any resets even when I booted adsp with the code base of tz protected memory xD
<konradybcio> bluerise the linux gunyah port is likely not using the same apis as gunyah on your laptop
<konradybcio> there was a change of calling convention at one point
<bluerise> if it's GPL I can't take it anyway :P
<steev> gabertron: which cma is specified on x1e for wifi?
srinik has quit [Ping timeout: 480 seconds]
<bluerise> btw, the Lenovo Slim 7x also has a calibration variant
<bluerise> not sure if that needs a new ath12k property
<bluerise> > qcom,ath11k-calibration-variant = "LES790";
<JensGlathe[m]> but why ath11k
<steev> because one doesn't exist for ath12k yet
<JensGlathe[m]> hmm
<steev> at least, it's not in the docs
<bluerise> I'll send a diff
<bluerise> got one prepared already
<bluerise> this should do, I hope
<steev> bluerise: need to update the bindings too (docs)
<steev> nevermind, i'm blind :D
<bluerise> :)
<bluerise> Yeah that was the first thing I did, grep where the docs are
<steev> my browser opened the thread weirdly
<bluerise> at the bottom; mine, too
<konradybcio> that's due to the link
<konradybcio> you can link to a specific message in the thread
srinik has joined #aarch64-laptops
<konradybcio> bluerise does the calibration variant name from the windows binaries?
<konradybcio> i.e. does windows encode the boardfile in the same way as upstream linux
<JensGlathe[m]> imo its the name you give in the dt for the variant, at least that's the case for Volterra / Blackrock
<konradybcio> i mean yeah so far it's been making stuff up as we go, but i'm interested in whether that changed
<colemickens> for those hacking on x1e, is there a checklist of non-kernel stuff needed/recommended? pd-mapper, etc?
<colemickens> also, I hate to be annoying, but still looking for someone to send me a ZIP of the Yoga7x firmware files or link me to a linux-firmware repo/fork that carries them. 🙏
<robclark> colemickens: did you already nuke your windows partition? If not, just copy all the files with .mbn extension.. plus IIRC adsp/cdsp needed dtb files with a different extension
<bluerise> why am I not surprised
<bluerise> konradybcio: I got mine from the ACPI tables
<bluerise> which match up with what I found in a recent board-2.bin
<robclark> yeah, adsp_dtbs.elf and cdsp_dtbs.elf
srinik has quit [Ping timeout: 480 seconds]
f_ has quit [Ping timeout: 480 seconds]
f_ has joined #aarch64-laptops
<steev> i usually make a backup of the driverstore/filerepository just to have everything off the system
<colemickens> <robclark> "colemickens: did you already..." <- I don't actually have a device yet, I'm being a bit stubborn and trying to prep an image to just work when I get it. I'm honestly considering going to a Best Buy here and grabbing them off and to just actually see the laptop physically to confirm it's right for me.
<colemickens> huh, I can google around but "adsp_dtbs.elf" and "cdsp_dtbs.elf" are surprising filenames, I assumed the DTBs would just be a kernel/bootloader concern. 🤔
<steev> who knows what they're thinking when they name the files
\[m] has joined #aarch64-laptops
<\[m]> damn it I'm actually tempted 😲
<\[m]> it's that time of the week again
<steev> do it, and get one for me too :P
<steev> unless it's the yoga
<\[m]> not available in EU still mm
<\[m]> hahaha I was like mm I still have to repay a debt to a friend so maaaybe not 😄
<\[m]> soon ™️
<\[m]> maybe next month 🙂
<steev> saved by the lack of availability!
<\[m]> my friend thanks you lenovo !
<\[m]> colemickens how's the fans behaving? did you run a bit on windows?
<\[m]> also if you want to get rid of yours 🙂
<\[m]> I wouldn't mind buying
sera[m] has joined #aarch64-laptops
<colemickens> .sera: you rock so much, thank you.
<colemickens> <\[m]> "colemickens how's the fans..." <- sorry, to clarify I don't have the device yet.
<colemickens> Though now with .sera 's blobs, I could theoretically put together a USB stick and go try and the Best Buy here. They have a demo unit. Assuming UEFI / USB Boot plays nice.
<konradybcio> check your emails :P
<HdkR> Minimum 80. Nice
<HdkR> Gotta buy enough to supply a whole company :P
<sera[m]> In stock now
<sera[m]> Oh, as I reloaded the page, it changed from min 1 in stock to minimum 80 also ..
<colemickens> just refreshed, says increments of 1 minimum 1 right this second. good luck ;)
<colemickens> Still says "Quote" though, rather than "Buy Now" or some such.
<konradybcio> >Limit of 5 units per customer
<HdkR> Seems like their server is getting hit by hundreds of us and choking :P
<colemickens> for x1e, are y'all building tip of mesa or a more fun branch?
<Jasper[m]> <konradybcio> "check your emails :P" <- nothing sadly
<Jasper[m]> <HdkR> "https://www.arrow.com/en/..."; <- Oh god wtf, I'm never going to be able to buy one
<Jasper[m]> Arrow is insane with shipping and they stock so late
<\[m]> oooh I thought you had a t14s
<\[m]> gaberton, how's the fand in the t14s, noise wise?
<Jasper[m]> \[m]: Nah just the X13s
<Jasper[m]> If they ever start shipping higher sku's for less exorbitant pricing then maybe
<\[m]> yoga should not have fans?
<robclark> colemickens: so windows itself doesn't use dtb but the fw running on the *dsp's do.. so it's really a dsp bootloader (ie. the kernel driver) issue
<robclark> (and yes, this kinda means windows is using devicetree :-P)
<robclark> anyways, I'd suggest just backing up the windows partition or at least the directory steev mentioned
<konradybcio> you probably want to run through windows update before erasing w
<robclark> yes.. that and ofc disable bitlocker so you can copy the important things
<robclark> \[m]: yoga has fans.. not too loud.. I can hear them if I build kernel with distro config which maxes things out for a while.. but the yoga is officially the fastest thing I own when it comes to compiling stuff (if you don't count my 64c/128t/shitloads-o-ram threadripper work desktop)
<colemickens> what else do I need in userspace besides pd-mapper? anything?
<robclark> I think abelvesa or jhovold's kernel branch has the in-kernel pd-mapper, maybe just use that because I think it will land before too long