#asahi-dev · 2026-09-24 19:21
jannauyuka: strict dt-style-check from next improves the results for t603*/t8122*#asahi-dev · 2026-09-24 18:37
jannau`ioreg -l | grep chip-id` from a M5 Max / Ultra should clear things up#asahi-dev · 2026-09-24 18:35
jannaum5 ultra has "clpc,t6050d", dispext has "dispext0,t605x"#asahi-dev · 2026-09-24 18:33
jannauI would expect the combination of cpu and base tile to have each their own chip-id, 27.0 Devicetrees from the ipsw have 0 as chip-id#asahi-dev · 2026-09-24 18:21
jannaum5 ultra has the same chip_id as max? this sounds unexpected. I would have also expected that M5 Pro and Max have different chip_ids#asahi-dev · 2026-09-24 18:18
jannau m1 max has only 'compatible = [pmgr1,t6001]'. it's based on the chip_id#asahi-dev · 2026-09-24 14:40
jannau80 char line length moved to --mode stricter in next#asahi-dev · 2026-09-24 14:36
jannauwas the not exactly aligned strict check dropped? it sounds like that would conflict with mixed-indent-chars #asahi-dev · 2026-09-24 10:32
jannauprobably in dri-devel#asahi-dev · 2026-09-24 10:32
jannauI think there were recently discussions to use custom create buffer IOCTLs for kms only devices to support video / compressed buffers#asahi-dev · 2026-09-24 10:12
jannauafaik DRM_IOCTL_MODE_CREATE_DUMB is not supposed to be work/used with modifiers#asahi-dev · 2026-09-24 09:54
jannausven: indeed, took a while to consider that the apple one is the only one which needs dt-bindings#asahi-dev · 2026-09-24 09:53
jannausofus: then gstreamer is broken and implicitly relies on either having just a single plane, no offset for the chroma plane(s) or "knowing" the right offset for specific hardware #asahi-dev · 2026-09-24 09:44
jannaudoes anyone know if the "NVM EXPRESS DRIVER" MAINTAINERS entry intentionally does not cover Documentation/devicetree/bindings/nvme/#asahi-dev · 2026-09-24 09:36
jannaudata_offset#asahi-dev · 2026-09-24 09:36
jannausofus: data offset in v4l2_plane: https://www.kernel.org/doc/html/v4.9/media/uapi/v4l/buffer.html#c.v4l2_plane#asahi-dev · 2026-09-24 08:44
jannauhttps://gist.github.com/jannau/3ccc65814c73a8549c04f82e2262730f#asahi-dev · 2026-09-21 15:33
jannauvideos should be published. I can ask for slides#asahi-dev · 2026-09-21 15:22
jannauexcept that we can use iommu merging via the dart driver#asahi-dev · 2026-09-21 15:18
jannaupresentation just arrived there#asahi-dev · 2026-09-21 15:14
jannaumight be even easier for us since each avd has its own iommu/dart#asahi-dev · 2026-09-21 15:13
jannauyes, it looks like we can just do the same#asahi-dev · 2026-09-21 15:13
jannaucordination also depends on policy. it's probably more power-efficient to decode two playback stream on one device but for maximal performance both shall be used#asahi-dev · 2026-09-21 15:10
jannaudifferent userspace programs can't coordinate which v4l2 device to use for decoding#asahi-dev · 2026-09-21 14:56
jannausofus: the ultra chips have 2 avd instances#asahi-dev · 2026-09-21 13:20
jannausofus: is there anything to ask/discuss related to this talk: https://kernel-recipes.org/en/2026/schedule/dual-core-video-decoding-on-rk3588-upstreaming-vdpu381-and-the-road-to-multi-core/ ?#asahi-dev · 2026-09-20 19:01
jannauplumbers is in October, Miguel and Danilo are at kernel recipes (at least on wednesday)#asahi-dev · 2026-09-20 18:52
jannaumight has been discussed last week at kangrejos. that might explain why the r4l zulip isn't that responsive right now#asahi-dev · 2026-09-20 18:47
jannauchaos_princess: please remind to discuss your rust/rustavailable patch in person with Miguel on friday afternoon (this is probably reminder enough for me)#asahi-dev · 2026-09-20 18:40
jannauefuses driver and bindings are upstream since v5.19-rc1 but unused (even downstream)#asahi-dev · 2026-09-20 18:34
jannauv2 nvmem chage applied after 12 minutes, more than a year after v1#asahi-dev · 2026-09-20 07:36
jannauI have the slight recollection that for dcp the initial mappings conflict in iova space. so even if we want we couldn't use combined mappings#asahi-dev · 2026-09-20 07:32
jannauI haven't really looked at avd but for dcp piodma is used to map the actual display controller MMIO into dcp's address space#asahi-dev · 2026-09-19 18:57
jannauof course most dts changes also depend (trivially but still annoyingly) on each other#asahi-dev · 2026-09-19 18:55
jannaudevicetree changes itself are also a little annoying since I made everyone's life harder fixing most `dt-check-style --mode strict` issues#asahi-dev · 2026-09-19 18:54
jannauI have M3* nvme, pcie and smc queued for upstream but the dependency situation is annoying. nvme is unfortunately first and it depends on the m4 dt-bindings changes which is just in nvme-7.3 but not in linux-next#asahi-dev · 2026-09-18 10:55
jannauand imacs, all base m1 devices. I would assume it's not needed as cs42l84 has '.symmetric_sample_bits = 1,'#asahi-dev · 2026-09-17 08:57
jannauhttps://git.infradead.org/?p=nvme.git;a=shortlog;h=refs/heads/nvme-7.3#asahi-dev · 2026-09-17 08:52
jannaulooks like the nvme tree isn't merged directly into next but via block and the changes haven't made it yet into block's for-next branch. I wouldn't expect the changes to land in 7.3. nvme-7.3 is either a mistaken branch or how that tree operates#asahi-dev · 2026-09-14 08:32
jannausofus: nothing, I have that queued#asahi-dev · 2026-09-11 11:51
jannautagged asahi-7.1.13-2 with the m3 pro jack hpd irq fix#asahi-dev · 2026-09-11 10:19
jannauwhich matches disp_reserved_regions_t602x#asahi-dev · 2026-09-11 10:18
jannauI have following segments (numbers are region id): 49, 50, os_log, 57, 94, 95, 14/vram, 157#asahi-dev · 2026-09-11 10:06
jannaumy segment-ranges has for some reason one additional segment#asahi-dev · 2026-09-11 10:04
jannauthat's not mapped either#asahi-dev · 2026-09-11 10:03
jannauthat is the third segment / __OS_LOG?#asahi-dev · 2026-09-11 09:53
jannauchaos_princess: where do you see 0x100045c4000? I only see phys ram from the top of memory#asahi-dev · 2026-09-11 09:03
jannauprobably unavoidable for disp0 although one of the regions is vram#asahi-dev · 2026-09-11 09:01
jannauthey don't match. I'd recommend using segment-ranges dart-dcp* and use region-id tables only if unavoidable#asahi-dev · 2026-09-11 08:52
jannaumatch the phys ram from the dart against the addresses in the ADT