#asahi-dev · 2024-01-06 10:03
kettenis(this is causing me some minor grief in the OpenBSD port)#asahi-dev · 2024-01-06 10:02
kettenisshould it include that file directly?#asahi-dev · 2024-01-06 10:02
kettenisjanneg, sven: dptxep.c uses stuff from <linux/bitfields.h> and ends up including that header through some convoluted route in <asm/sysreg.h>#asahi-dev · 2024-01-04 11:58
kettenisoops, well spotted; thanks!#asahi-dev · 2023-12-30 16:22
kettenisI can probably figure out how to hook this up the way it is in the OpenBSD code#asahi-dev · 2023-12-30 16:21
kettenisso probably best to leave things as-is#asahi-dev · 2023-12-30 16:20
kettenisok, it seems that there is plenty of prior art in the linux tree to do what you do now#asahi-dev · 2023-12-30 16:07
kettenisand at resume time you poll the pin, which makes sense to me#asahi-dev · 2023-12-30 16:05
kettenisa level-triggered interrupt would continue to be triggered#asahi-dev · 2023-12-30 15:58
kettenisI'll explore that path#asahi-dev · 2023-12-30 15:57
kettenisit would make my life a bit easier, and it would decouple the interrupt from the detect pin in case apple comes up with a design where that is the case#asahi-dev · 2023-12-30 15:45
kettenisjanneg: is there a reason why the hdmi hpd interrupt isn't described by an "interrupts-extended" property?#asahi-dev · 2023-12-30 12:37
kettenisthis happens for DCP mappings that are also mapped into the disp0 DART#asahi-dev · 2023-12-30 12:36
kettenisit seems the linux driver accepts mapping iova's that don't fall into the range specified by that property#asahi-dev · 2023-12-30 12:35
kettenismarcan: I have some questions about the apple,dma-range property that was added for the DARTs some time ago#asahi-dev · 2023-12-29 11:43
kettenis\8/#asahi-dev · 2023-12-29 10:10
kettenisi509vcb: is that 3900x really much more powerful than an m1?#asahi-dev · 2023-12-28 21:24
kettenisprobably not the solution you'r looking for#asahi-dev · 2023-12-28 21:23
kettenisright, although in the rk3588 case PCIE and SATA are mutually exclusive#asahi-dev · 2023-12-28 21:15
kettenissee for example how the "rockchip,rk3588-naneng-combphy" is switched between PCIE and SATA mode for certain boards#asahi-dev · 2023-12-28 21:11
kettenissome SoCs use additional cells in the PHY specification to set the mode#asahi-dev · 2023-12-28 12:52
ketteniscreated a PR with a possible fix for the backlight resume issue#asahi-dev · 2023-12-28 12:51
kettenisthe DPMS issue I reported earlier is gone so I guess one of the commits that changed how modesets were triggered fixed that#asahi-dev · 2023-12-28 12:49
kettenisfully caught up with asahi-wip for the OpenBSD port of the dcp driver now#asahi-dev · 2023-12-27 22:34
kettenissince our OSes don't implement PPL, you don't really need to exploit this interesting hardware feature if you want to do something bad#asahi-dev · 2023-12-27 22:29
kettenishttps://securelist.com/operation-triangulation-the-last-hardware-mystery/111669/#asahi-dev · 2023-12-26 19:20
kettenisso older code with the latest dtb#asahi-dev · 2023-12-26 19:19
kettenisjanneg: ah, thanks, I'm still working on crossing the apple,dma-addr bump#asahi-dev · 2023-12-26 18:21
kettenisare we missing a register region on t600x?#asahi-dev · 2023-12-26 18:21
kettenisjaaneg: dcpep_cb_map_reg *WARNING* attempted to read invalid reg index 4#asahi-dev · 2023-12-26 14:50
kettenisbut there vm-size for dart-disp0 is 0xfffff0000 such that it includes the dcp range#asahi-dev · 2023-12-26 12:33
kettenisbut on t8103 for example, the apple,dma-range windows of those two DARTs do not overlap#asahi-dev · 2023-12-26 12:33
kettenisthe dcp driver enters some of the buffers into both the disp0 and dcp DARTs with the same iova#asahi-dev · 2023-12-26 12:32
kettenisI also noticed something else that's weird with the apple,dma-range properties for some of the SoCs#asahi-dev · 2023-12-26 12:30
kettenisjanneg: that is exactly what I mean#asahi-dev · 2023-12-25 15:54
kettenisI think that is wrong#asahi-dev · 2023-12-25 15:52
kettenishmm, the iommu-addresses properties under /reserved-memory have truncated IOVA's#asahi-dev · 2023-12-20 21:24
kettenisthe ABI differences cause problems, regardless of whether the driver is written in C or rust ;)#asahi-dev · 2023-12-20 14:21
kettenisnice to see that with dpms there are some significant power savings#asahi-dev · 2023-12-20 13:28
kettenislooks like I really should update to the latest code as soon as possible ;)#asahi-dev · 2023-12-20 13:26
kettenishmm, so maybe this is fixed by "drm: apple: Move modeset into drm_crtc's atomic_enable" that I don't have yet#asahi-dev · 2023-12-20 13:23
kettenisand for some reason dcp sometimes refuses to start; happens more often on the M1 MacBook than on the M1 mini#asahi-dev · 2023-12-20 13:21
kettenisbut that probably is a problem in the drm atomic modeset code#asahi-dev · 2023-12-20 13:21
kettenisthe other bit that I'm still struggling with is that drm_fb_helper_blank() doesn't seem to do the right thing#asahi-dev · 2023-12-20 13:18
kettenisok, I'll try that and if it works on OpenBSD I'll send you a diff to test on Linux ;)#asahi-dev · 2023-12-20 13:16
kettenisright, that was my idea; already has a check whether backlight.update is still set ;)#asahi-dev · 2023-12-20 13:08
kettenismaybe from some delayed scheduled work such that it can be avoided if a swap already happened?#asahi-dev · 2023-12-20 13:06
kettenisis it really that difficult to force a flush from the iomfbep_cb_enable_backlight_message_ap_gated() callback?#asahi-dev · 2023-12-20 13:04
kettenisjanneg: I still think this needs to be fixed; even if you believe Xorg should die there are other contexts where having a working dumb framebuffer is desirable#asahi-dev · 2023-12-20 11:04
kettenisbut I see the same thing on the 13" M2 MacBook Air with 13.5 fw