#asahi-dev · 2026-11-04 10:48
mazyup, exactly that. and there was consensus to do so by exposing an existing helper.#asahi-dev · 2026-11-04 10:47
jannauthe ask was to my understanding to turn on power for each port individually in pcie-apple because the driver does everything else on a per-port basis#asahi-dev · 2026-11-04 10:46
yukapcie-apple#asahi-dev · 2026-11-04 10:46
jannauwhich driver? I think that already caused confusion in the initial submission#asahi-dev · 2026-11-04 10:42
yukadoes not make sense to me#asahi-dev · 2026-11-04 10:41
yukayeah but the whole driver can not power on a single port#asahi-dev · 2026-11-04 10:41
jannausee "To be clear, I'm not suggesting delaying switching of the power. I'm merely suggesting managing it on a per port basis."#asahi-dev · 2026-11-04 10:41
yukaAh well of course you can copy the same code into the driver#asahi-dev · 2026-11-04 10:41
jannauI interpretet that as referring to hiding the loop over all ports in pci_pwrctrl_power_on_devices()#asahi-dev · 2026-11-04 10:36
yukaI did not really know how to argue with that so I dropped the ball#asahi-dev · 2026-11-04 10:36
yukaThis was nacked as "[not] consistent with the way the rest of the driver"#asahi-dev · 2026-11-04 10:29
yukaI think I proposed this and it was nacked? Would have to double check though#asahi-dev · 2026-11-04 10:08
jannauI think the main reason why it is not picked up is that it was entangled in the previous power GPIO submission: https://lore.kernel.org/linux-pci/20220502093832.32778-3-marcan@marcan.st/#asahi-dev · 2026-11-04 10:06
jannaucan't we just power on devices in a loop before we loop over apple_pcie_setup_port(). we have https://github.com/AsahiLinux/linux/commit/1a71d5c8ad462e8a6cd0a436c5e92b3c56bc5271 downstream which essentially does the same for the PERST# gpios.#asahi-dev · 2026-11-04 09:52
yuka(practically this won't happen if all ports are powered on via the same smc gpio device or so, but honestly still feels quite bad)#asahi-dev · 2026-11-04 09:51
yukait's probably possible to handle this just very complicated#asahi-dev · 2026-11-04 09:51
yukayou might start initializing some pcie ports 1 and 2 and on the third you run into a -EPROBE_DEFER#asahi-dev · 2026-11-04 09:50
yukabut see my mailing list posts from back then#asahi-dev · 2026-11-04 09:49
yukaI mean you have to keep in mind that the gpio device might not be bound at the point when this first runs#asahi-dev · 2026-11-04 09:49
yuka"per port"?#asahi-dev · 2026-11-04 09:48
jannauyuka: into what kind of trouble did you run with powering on devices per port?#asahi-dev · 2026-11-04 09:47
jannauafter looking at the code pci_pwrctrl_create_device / pci_pwrctrl_power_on_device would work for M2 Pro as well#asahi-dev · 2026-11-04 09:29
yukaworks without#asahi-dev · 2026-11-04 09:29
yukaM1 Pro suspend is broken with PMP#asahi-dev · 2026-11-04 08:48
mazif we really need something of that shape, so be it, but let's have the discussion on the list.#asahi-dev · 2026-11-04 08:47
mazjannau: it'd be good to make progress on the PCIe power front (the "post and drop" strategy doesn't really help)#asahi-dev · 2026-11-04 07:49
jannauYouwes09: I saw your message but haven't looked at it yet. I saw the rust changes and remember that they might cause issues but forgot all details. I have to look at code#asahi-dev · 2026-11-04 07:45
yukaas in, it requires more than just device tree changes#asahi-dev · 2026-11-04 07:45
yukaAhh the PRs opened by a-ramses probably address the issue I was having where nothing happened with HDMI on my M3 Max#asahi-dev · 2026-11-04 06:39
sobek[m]<jmax[m]> "So cool! And a changelog entry!" <- Sounds like that bug was big enough that they wanted to tell us when they fixed it.#asahi-dev · 2026-11-04 03:08
Youwes09I fixed it by changing the RTKit layer so the driver owns the connection and shuts it down on unbind, which also makes sure the firmware stops before the buffers go away. does that seem like the right approach? asked about this before but I don't think it got seen...#asahi-dev · 2026-11-04 03:06
Youwes09jannau: question about the 7.4 rebase. mainline changed how Rust drivers hold DMA memory, so a buffer can now only exist while its driver is bound to the device. the PMP driver keeps its DMA buffers in a shared object it hands to RTKit, and that stops building because the compiler can't prove the buffers are gone before the driver unbinds.#asahi-dev · 2026-11-04 03:02
Youwes09myria: saw your message about the M4 events, thanks. I'm in the middle of some hypervisor work on thermals right now, but I'll look at it soon and let you know!#asahi-dev · 2026-11-03 23:09
jannauI've tagged and pushed asahi-7.2.15-1#asahi-dev · 2026-11-03 23:06
jannauno, my PCIe dts looks different, did you see that maz did not like pci_pwrctrl_power_on_devices() when yuka send PCI pwrctrl support in July. I fear we need it for Mac Pro support with the large PCIe switch though#asahi-dev · 2026-11-03 23:01
jannauhuh, I missed that nvme was picked up#asahi-dev · 2026-11-03 22:59
MichaelReeves[m]Ack, thanks will do later today#asahi-dev · 2026-11-03 22:51
jannauArend did not address how the DMA direction conflict should be handled, might still be advisable to send a v2 based on wireless.git with the conflict resolved (and Arand's Acked-by: imported)#asahi-dev · 2026-11-03 22:43
MichaelReeves[m]It’s probably the same as yours anyway so I don’t really mind. Yeah I guess NVMe is needed for firmware loading so it technically won’t work without it, but I think sv.en picked that up yesterday. What would you like to do?#asahi-dev · 2026-11-03 22:42
MichaelReeves[m]I also have commits 14-21 here which is what I was mentioning https://github.com/IntegralPilot/linux/pull/1#asahi-dev · 2026-11-03 22:40
jannauMichaelReeves[m]: I think t602x upstream has it already, for M3 pcie support is missing as whole. I have a prepared branch for it ready to send. I haven't send it so far as it losely depends on m3 nvme#asahi-dev · 2026-11-03 22:34
MichaelReeves[m]is it OK to send DT bits now so they land together in time for 7.5?#asahi-dev · 2026-11-03 22:34
MichaelReeves[m]yay wifi series got all acked#asahi-dev · 2026-11-03 20:46
yukanice!#asahi-dev · 2026-11-03 20:46
jannauj714c#asahi-dev · 2026-11-03 20:44
yukajannau which machine?#asahi-dev · 2026-11-03 20:39
jannau"Starting CPU 17 (0:2:5)... Started."#asahi-dev · 2026-11-03 19:58
svenif they aren’t annoying ill just fix them and pick the series up #asahi-dev · 2026-11-03 19:58
svenI’ll take a look at the conflicts tomorrow once I’m at the airport back from my $work trip