#
✕ nick:yuka
#asahi-dev · 2026-07-11 08:42
yukaonce I write BIT(3) in the mask is sticky active in the flag
#asahi-dev · 2026-07-11 08:41
yukait should be BIT(1)
#asahi-dev · 2026-07-11 08:41
yukathe mailbox driver assumes IRQ_RX = BIT(3), but this does not work in my experiments
#asahi-dev · 2026-07-11 08:31
yukaor maybe the dockchannel-uart hardware works slightly differently
#asahi-dev · 2026-07-11 08:31
yukaexactly
#asahi-dev · 2026-07-11 08:28
yukaI can double check that
#asahi-dev · 2026-07-11 08:28
yukabut I think something that requests the mailbox channel would probably be sufficient to trigger it
#asahi-dev · 2026-07-11 08:26
yukathis is the code: https://cyberchaos.dev/yuka/linux/-/commit/8c9fb4ca834101d5dbf31c76bbeb980e52dcd5ea
#asahi-dev · 2026-07-11 08:25
yukaat the same time, irq/41-2a050c000.mailbox is using lots of cpu
#asahi-dev · 2026-07-11 08:25
yukathe uart is idle
#asahi-dev · 2026-07-11 08:25
yukaI'm also not sending anything
#asahi-dev · 2026-07-11 08:25
yukano
#asahi-dev · 2026-07-11 07:40
yukaI think it might be inherent to using the proposed mailbox code on the dockchannel uart
#asahi-dev · 2026-07-11 07:39
yukaI've seen the space heater / irq storm effect return with my own version of the dockchannel mailbox uart
#asahi-dev · 2026-07-10 09:53
yukaM4 was in a similar situation for a while I think
#asahi-dev · 2026-07-10 09:53
yukaI do see the pmgr regs change, so I'm confident the offset is correct
#asahi-dev · 2026-07-10 09:52
yukaIf the offset was wrong, I wounldn't see the cpus pmgr regs change
#asahi-dev · 2026-07-10 07:47
yukaI would still like to support at least 2 channels over dockchannel uart inside of m1n1, so that one of them can be used as vuart on the Neo
#asahi-dev · 2026-07-10 07:32
yukaso that we can redirect it to cdc-acm?
#asahi-dev · 2026-07-10 06:58
yukaor maybe there is some way to enable irqs for the other channels (but I just don't know how to find it)
#asahi-dev · 2026-07-10 06:57
yukaand then a proper guest could use channel 0 with irqs
#asahi-dev · 2026-07-10 06:56
yukamaybe m1n1 should always use channel 1 instead of 0, since it does polling anyways
#asahi-dev · 2026-07-10 06:54
yukaWhich makes them kinda pointless to expose as ttys
#asahi-dev · 2026-07-10 06:52
yukaSo at this point it seems to me like the other channels don't have IRQs
#asahi-dev · 2026-07-10 06:52
yukadata appears in data1
#asahi-dev · 2026-07-10 06:51
yukanone of the IRQ regs change
#asahi-dev · 2026-07-10 06:51
yukaWrote stuff to channel 1
#asahi-dev · 2026-07-10 06:51
yuka(and the data appears in data0)
#asahi-dev · 2026-07-10 06:51
yukaWrote stuff to channel 0 -> IRQ regs (1)+IRQ_FLAG gets set
#asahi-dev · 2026-07-10 06:51
yuka(checked the other IRQ regs + IRQ_MASK are all 0 as well)
#asahi-dev · 2026-07-10 06:50
yukaenabled irqs (by writing 0 to IRQ regs (1)+IRQ_MASK)
#asahi-dev · 2026-07-10 06:50
yukaI added a RegMonitor on all the areas described in proxyclient/m1n1/hw/dockchannel.py
#asahi-dev · 2026-07-10 06:49
yukawe haven't seen them used in macOS, so...
#asahi-dev · 2026-07-10 06:49
yukamaybe they're actually half broken
#asahi-dev · 2026-07-10 06:48
yukaand how they work with the different fifos
#asahi-dev · 2026-07-10 06:48
yukaI'm so confused about dockchannel irqs
#asahi-dev · 2026-07-09 19:48
yukaor something
#asahi-dev · 2026-07-09 19:48
yukaeverything is a file
#asahi-dev · 2026-07-09 19:45
yukalike you write a mailbox driver and get a uart/tty-including-console-support for free
#asahi-dev · 2026-07-09 19:45
yukaI wonder if the mailbox-uart/tty-including-console-support part could be so generic that it isn't even apple-dockchannel specific
#asahi-dev · 2026-07-09 19:41
yukaWhen there was nothing to be sent
#asahi-dev · 2026-07-09 19:40
yukathe previous thing I tested was busy looping somewhere
#asahi-dev · 2026-07-09 19:34
yukaand I hope this implementation doesn't draw an extra 10W
#asahi-dev · 2026-07-09 19:34
yukaSo the earlycon would still work if the mbox was entirely missing, and the normal uart driver would still work if the regs were missing
#asahi-dev · 2026-07-09 19:34
yukaTo make this not completely horrible, I duplicated the fifo reg on the mailbox and the uart node. Both earlycon and the normal uart driver use to the same dt node, but look at different properties (regs vs mboxes)
#asahi-dev · 2026-07-09 19:32
yukaI think the earlycon still writes things to the fifo after mailbox has been initialized
#asahi-dev · 2026-07-09 19:32
yukaexcept when I enable both at once, the handover goes wrong
#asahi-dev · 2026-07-09 19:31
yukaOkay I have dockchannel-uart earlycon (direct mmio, tx-only) and console= / ttyDC0 (mailbox, tx/rx) working :-)
#asahi-dev · 2026-07-09 17:17
yukayes, only the hub has a "function-reset"
#asahi-dev · 2026-07-09 17:17
yukaand what was the way to reset it again?

plus ›