#
✕ nick:jannau
#asahi-dev · 2022-01-23 12:00
jannauI currently use ioremap_wc for the dcp crashlog, mostly since I'm drowned in noise when looking how to map carveout memory
#asahi-dev · 2022-01-23 11:59
jannauah right
#asahi-dev · 2022-01-23 11:57
jannauit cares in apple_rtkit_memcpy()
#asahi-dev · 2022-01-23 11:47
jannauupdated tree is at https://github.com/jannau/linux/tree/bits/050-nvme
#asahi-dev · 2022-01-23 11:47
jannausven: ah, firmware might change behavior, BUG_ON not really appropiate then. I only considered the new RTKit using driver case
#asahi-dev · 2022-01-23 10:43
jannausven: rtkit buffer changes are in https://github.com/jannau/linux/tree/bits/050-nvme
#asahi-dev · 2022-01-23 10:42
jannauugh, I just looked at the carveouts at the top of the memory. the crashlog buffer is in a 3 megabyte carveout so that would be fine if the ADT would tell to which device nodes a carveout belongs
#asahi-dev · 2022-01-23 09:29
jannauthe dcp crashlog buffer is in one of the carveout-memory-map regions. I don't see a mapping of carveout regions to devices in the ADT
#asahi-dev · 2022-01-22 11:45
jannaumaz: you could have udev chainload the current m1n1 + u-boot when the on device m1n1 connects. a little fragile but you don't need to go through macos recovery for updating the default u-boot
#asahi-dev · 2022-01-20 20:36
jannauI thought so. we could decide based on that save the buffer's ownership in struct apple_rtkit_shmem
#asahi-dev · 2022-01-20 20:28
jannaudcp's crashlog has an iova buffer but on all other endpoints host controlled buffers
#asahi-dev · 2022-01-20 20:27
jannausven: any idea how to handle RTKits with mixed rtkit/linux allocated buffers?
#asahi-dev · 2022-01-20 20:24
jannaufinally a dcp crashlog, not that the crashlog info would be helpful in figuring out why dcp crashes on unplug
#asahi-dev · 2022-01-18 22:26
jannauyes, it only does swaps with valid surfaces, I haven't filtered sawps without surfaces completely out in the rebased tree
#asahi-dev · 2022-01-18 22:13
jannauparse_log.py prints the structs in the arguments
#asahi-dev · 2022-01-18 22:11
jannauwell that only affects parse_log, the strings in hv/trace_dcp.py are still 11.4
#asahi-dev · 2022-01-18 22:09
jannauupdated in de5b5d996c7f29e8a70cddbf487d44249eac7b55 on August 14th
#asahi-dev · 2022-01-18 22:07
jannau+on
#asahi-dev · 2022-01-18 22:07
jannauno, the august era tracer is based 12.0 beta 3 or so
#asahi-dev · 2022-01-18 22:06
jannaubut it looks like there are less swaps until console boot so I think there are probably still dodgy ones with newer drm
#asahi-dev · 2022-01-18 22:05
jannauof course I now have to downgrade the tracer to the 11.4 ABI to see details
#asahi-dev · 2022-01-18 22:00
jannauhotplug works with https://github.com/mu-one/linux/commits/20211022
#asahi-dev · 2022-01-18 21:34
jannaudrivers/iommu/apple-dart.c has "writel(DART_STREAM_ALL, dart->regs + DART_STREAMS_ENABLE);" as addition
#asahi-dev · 2022-01-18 21:33
jannaudrivers/gpu/drm/apple/ is identical
#asahi-dev · 2022-01-18 21:31
jannauyes, changes were ment neutral
#asahi-dev · 2022-01-18 21:26
jannauso drm changes are the most likely suspect for problems
#asahi-dev · 2022-01-18 21:25
jannauI assume the dcp code itself doesn't differ (much) from the previous working branch
#asahi-dev · 2022-01-18 21:21
jannaudcp code I have is based on https://github.com/mu-one/linux/commits/20211209 with all non dcp/dart changes removed and rebased onto asahi
#asahi-dev · 2022-01-18 21:16
jannauunrelated to my changes additional changes
#asahi-dev · 2022-01-18 21:14
jannauit crashes with macos 11 as well (with just the api change reverted)
#asahi-dev · 2022-01-18 21:10
jannaumacos 12 does a swap with ".swap_enabled/completed = 0x80000007" after the unplug
#asahi-dev · 2022-01-18 21:09
jannauI've tested with my dcp_flush changes so those are not responsible although I'm not convinced that they are a good idea
#asahi-dev · 2022-01-18 21:07
jannauI'll test with macos 11 with the API change commit reverted
#asahi-dev · 2022-01-18 21:05
jannauor it doesn't like that we don't listen on the other endpoints
#asahi-dev · 2022-01-18 21:05
jannauswap_complete_intent_gated
#asahi-dev · 2022-01-18 21:04
jannauthe first missing callback compared to macos is
#asahi-dev · 2022-01-18 21:04
jannaumy best guess is it is either something in the swap structure it doesn't like
#asahi-dev · 2022-01-18 21:03
jannaudcp crashes during a series of callback after plugin out
#asahi-dev · 2022-01-18 21:02
jannauI don't see anything in dcp traces (compared to macos 12) how I could have broken hotplug
#asahi-dev · 2022-01-17 20:10
jannaumaybe just a dying usb device
#asahi-dev · 2022-01-17 19:54
jannauerr, it's with the usb-a ports so I guess nevermind. if it's dart related it would be PCIe
#asahi-dev · 2022-01-17 19:51
jannausven: have you seen "xhci_hcd 0000:02:00.0: WARN Set TR Deq Ptr cmd failed due to incorrect slot or ep state." before? possibly related to the dart changes in the dcp branch. I see that with 12.0 but not with 12.1. dcp dt nodes are commented out
#asahi-dev · 2022-01-17 16:49
jannauI wonder what happens if someone connects a screen with iphone5 dimensions
#asahi-dev · 2022-01-17 16:46
jannauit works on 12.0 and timeouts after 2 seconds without connected display. so far dcp was ready after 400 ms with all displays so we should be able reduce the timeout
#asahi-dev · 2022-01-17 16:32
jannau11.x without connected display?
#asahi-dev · 2022-01-16 23:32
jannauyes, the display works, I was looking if any of the other endpoints reports the EDID data
#asahi-dev · 2022-01-16 23:15
jannauthe av endpoint reports parsed EDID information but no timing information
#asahi-dev · 2022-01-16 22:41
jannaubut it manages to display something on an old dell 4:3 display and list the native mode 1600x1200@60. macos 11 iboot didn't initialize that display
#asahi-dev · 2022-01-16 22:38
jannaufor both monitors I tried knowing the native resolution is the only method I see to select the correct mode
#asahi-dev · 2022-01-16 22:37
jannautrying to select 1920x1080@60 seems to be the least bad option with the limited information provided

plus ›