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
derzahl has joined #aarch64-laptops
<agl>
steev: Have you contacted clover[m]?
<steev>
why would i?
<agl>
steev: Because the EndeavourOS arm there worked all and after an upgrade (yay -Syyu) the grafical system does not work.
<steev>
i don't use endeavoros
<agl>
yes, you want bevor some days that I send some thinks form /lib/firmware (you look about the versions) and other directories and you find nothing. Then you say you would contakt clover[m] when he is back.
<steev>
no, i said you should, because i don't use endeavor and i have no idea where to look or why that would occur. i'm not going to do your work for you and just be a middle man
<agl>
ok, I understood that you would contact him. Ok, I see clover[m] is not in the user list ... then I have to wait if he is back ....
<abby>
probably the matrix bridge being bad
<agl>
ok I can try to use matix. I have until now not an Account there.
<agl>
steev: Sorry, that I mean you want to contact clover[m].
<agl>
steev: My english language is not so good.
<steev>
your english is better than my german :)
<agl>
ok, the german language is a little bit difficult to learn it for a english man ;-)
agl7[m] has joined #aarch64-laptops
clover[m] has joined #aarch64-laptops
<clover[m]>
i think the bridge is working again
<abby>
every time the bridge crashes it tries to reconnect every user who's ever connected, which just leads to it crashing again and again until it finally recovers within a week or two
<colemickens>
I think I am one linux-firmware repo away from having a test nixos iso with grub entries for each DTB. :D
tobhe_ has joined #aarch64-laptops
tobhe has quit [Ping timeout: 480 seconds]
phire has joined #aarch64-laptops
<agl>
clover[m]: In this moment the grafical system starts under EndeavourOS arm, but some times it starts not. I can not understand why sometimes the grafical system starts and why not at an another start.
<clover[m]>
Okay did you get any log info about it
<agl>
not yet it runs in this moment
<agl>
I look
<agl>
clover[m]: Under dmesg it says adreno is loaded and running
cachoeira has quit [Remote host closed the connection]
cachoeira has joined #aarch64-laptops
<strongtz[m]>
colemickens: not much actually, you can use Abel's branch or simply linux-next
phire_ has joined #aarch64-laptops
phire is now known as Guest646
phire_ is now known as phire
Guest646 has quit [Ping timeout: 480 seconds]
alfredo has joined #aarch64-laptops
alfredo1 has joined #aarch64-laptops
alfredo has quit [Ping timeout: 480 seconds]
alfredo1 is now known as alfredo
phire has quit [Ping timeout: 480 seconds]
phire has joined #aarch64-laptops
alfredo1 has joined #aarch64-laptops
alfredo has quit [Ping timeout: 480 seconds]
alfredo1 is now known as alfredo
alfredo1 has joined #aarch64-laptops
paddymahoney has quit [Remote host closed the connection]
<mani_s>
maz, I'm seeing a behavior with gic-v3-its driver where if odd number of MSIs are requested (ex. 3) using pci_alloc_irq_vectors_affinity(), then its_alloc_device_irq() allocates log2 of the MSI count (ex. 4). But when pci_free_irq_vectors() is called, it only frees the 3 MSIs
<mani_s>
leaving 1 MSI always allocated. When the driver is unloaded and loaded again, requesting 3 MSIs again is failing because of the previous 1 MSI already allocated (here max MSI is 4).
<mani_s>
I can see that this behavior is applicable to other irqchip drivers also using bitmap to manage irqs.
<mani_s>
'driver is unloaded' -> I'm referring to the PCI driver.
kettenis_ is now known as kettenis
<kettenis>
mani_s: Someone should tell Qualcomm that shipping hardware with PCI devices that uses multiple MSI vectors but doesn't support MSI-X is unacceptable in 2024
<mani_s>
kettenis, I'm don't agree with this statement. If the PCI device has no requirement to support > 32 MSI vectors per function, then there is no need for MSI-X
<mani_s>
Most of the chipsets that Qcom supports for PCI devices are for modems, WLAN, where 32 MSIs is more than enough
<mani_s>
That's one of the reason why PCIe spec hasn't deprecated MSI
<kettenis>
multi-vector MSI has some serious design flaws that make it interact poorly with many interrupt controller designs
<kettenis>
MSI-X fixes those flaws and every other vendor seems to be able to implement it even for devices that need significantly less than 32 vectors
<kettenis>
the issue you're seeing is entirely because of that design flaw
<kettenis>
I just implemented multi-vector MSI support for an OpenBSD PCIe controller driver that has exactly the same flaw ;)
agl has joined #aarch64-laptops
<gabertron>
Has anybody made a working / partially working dts for the t14s x1 laptop? Was taking a look at that but thought I’d check if there was one out there that I could start with
<mani_s>
kettenis, ah okay... I was checking if this is something already known or not. Sounds like it is ;)
agl has quit [Ping timeout: 480 seconds]
agl has joined #aarch64-laptops
<JensGlathe[m]>
<mani_s> "maz, I'm seeing a behavior..." <- Hmm looks loke some bit shifting only in one of the driver functions. Maybe a good idea to take a look at this driver
<mani_s>
JensGlathe[m], It is not *this* driver specific issue, but applies to other drivers as well.
<JensGlathe[m]>
mani_s: I figured after reading again. Fix would be to allocate log2 vectors only to ensure it doesn't bite
<robclark>
gabertron: most folks interested in t14 seem to be holding out for option to configure device
<mani_s>
JensGlathe[m], yes, that looks like the only possible solution to me. But currently, drivers just don't do that
<mani_s>
It is also tricky because, if a driver just wants odd number of vectors, why should it request more vectors? And what should it do with those? Just return IRQ_HANDLED? That sounds sub optimal
<JensGlathe[m]>
It would solve the problem. Wouldn't cost too much tbh, if the spec is flawed as kettenis suggests.
<clee>
gabertron: in the branch with the (mostly) working DTS/DTB for my Yoga Slim 7X, I see: x1e80100-{crd,qcp,surface,vivobook,yoga}.dts so I'm guessing not yet for the T14S
<clee>
(I wonder how different it is from the Yoga…)
<mani_s>
JensGlathe[m], in that case, the restriction should be imposed in the API level i.e, for all pci_alloc_irq_vectors*() API
<JensGlathe[m]>
It could even wire up the unused vectors with the "dummy" handler
<JensGlathe[m]>
Keeping track of them might be tricky...?
agl_ has joined #aarch64-laptops
agl has quit [Read error: Connection reset by peer]
<gabertron>
Yea I tried the crd dts but no luck so far but will try the yoga to see if that works
<konradybcio>
gabertron send me an acpi dump and ill send you a dts back..
<colemickens>
Anyone testing on the Surface Pro? I'm considering it over the Yoga but worried about the Flex keyboard being some kind of proprietary Bluetooth handshake with the tablet/windows.