#asahi · 2021-08-11 06:14
svenhm.. that *shouldn't* break anything#asahi · 2021-08-11 06:13
svenso this is not just the devel branch but devel branch + some of your patches?#asahi · 2021-08-11 06:13
svenit also could be a very subtle bug in the iommu code that's why i'd like to make sure#asahi · 2021-08-11 06:11
sventhere's a config options and also a boot argument for that#asahi · 2021-08-11 06:11
svenanother think you could try is to boot with the IOMMU in passthrough mode#asahi · 2021-08-11 06:11
svenwould be good to make sure it's really an issue somewhere in linux#asahi · 2021-08-11 06:07
svenit's also weird because even with 4k pagesize code that i know to be broken i haven't seen that severe corruption#asahi · 2021-08-11 06:04
svenpretty sure i've only read about that from amw#asahi · 2021-08-11 06:02
svenamw: and just to confirm: this specific usb drive works fine if you use it with another machine? (i'm asking because i have one broken usb drive that caused me headaches for a week or so until i realized it *also* failed in the same ways on macOS and another linux machine)#asahi · 2021-08-11 06:01
svenamw: and that didn't happen before? the code in nvme and devel is also essentially the same#asahi · 2021-08-10 14:33
svenamw: can you elaborate on that corruption? do you get any error messages? does this usb drive work with other machines? 'cause i've been using that branch for a while without issues so far#asahi · 2021-08-10 06:24
sventhat's 16k, so should be fine#asahi · 2021-08-10 06:23
sven(please don't let it be part of that framework. i really don't want to also implement multiple iommu per bus support in there)#asahi · 2021-08-10 06:22
svenbut GART may not be part of the iommu in the end#asahi · 2021-08-10 06:22
sventhere's no DART but probably the same "issues" apply to GART#asahi · 2021-08-10 06:22
svenah, good point#asahi · 2021-08-10 06:21
sven(and we'll make the broadcom wifi untrusted too)#asahi · 2021-08-10 06:20
sveni.e. thunderbolt#asahi · 2021-08-10 06:20
svenuntrusted in the kernel means "external facing device that can act as a DMA master"#asahi · 2021-08-10 06:20
svenAGx wouldn't be untrusted anyway#asahi · 2021-08-10 06:18
svenbut presumably it'll need some per-process mappings#asahi · 2021-08-10 06:18
svenAGX uses something called GART which we don't know much about yet#asahi · 2021-08-10 06:18
svenapple loves IOMMU *so* much that they have at least four different ones#asahi · 2021-08-10 06:16
svenamw: 4k pages or 16k pages? and is the iommu in DMA or in bypass mode?#asahi · 2021-08-10 06:13
svenj_ey: probably not, that was just a hack i tried#asahi · 2021-08-10 06:12
svenamw: uh.. so dart/devel wouldn't surprise me since that one has that 4k pagesize shenanigans #asahi · 2021-08-09 19:34
svenand then make it all seem to effortless like you two#asahi · 2021-08-09 19:34
svenyeah, i couldn't even imagine to code and stream at the same time#asahi · 2021-08-09 18:27
svenhttps://github.com/AsahiLinux/linux/pull/6 so that's essentially the mailbox thing that just exposes N endpoints and muxes them to the single FIFO + some (maybe) hack to the mailbox core to introduce a fast-path for submission#asahi · 2021-08-09 18:20
sven(as in, i haven't traced through all the code to know why it works. i just have a good guess)#asahi · 2021-08-09 18:20
sveni'm not sure it's correct yet, but yeah :D#asahi · 2021-08-09 18:19
svenyeah, it's like 1 fix for the aic-irqchip (if you disable irq N while you're in the handler for N you can never enable it again) + 300 lines of code or so#asahi · 2021-08-09 18:19
svenpipcet[m]: might make sense to open one with the "0 channels as well"#asahi · 2021-08-09 18:18
svenmaybe i should actually open one with the dumb mailbox with N channels things so that we have some actual code to discuss and see how things fit#asahi · 2021-08-09 18:17
svenpipcet[m]: anyways, feel free to open more PRs for now#asahi · 2021-08-09 18:17
svensounds good#asahi · 2021-08-09 18:16
svenit's literally just rc3 + dart or so#asahi · 2021-08-09 18:16
svenwe just created that branch yesterday, feel free to remove it again :D#asahi · 2021-08-09 18:15
sventhen i guess we'll keep the PRs open for now#asahi · 2021-08-09 18:15
svensounds good :)#asahi · 2021-08-09 18:15
svensure, but for now devel works.#asahi · 2021-08-09 18:14
svento catch common mistakes, etc.#asahi · 2021-08-09 18:14
svenwe should have PRs somewhere though for quicker internal review#asahi · 2021-08-09 18:14
svendunno, but i think that first one could actually be sent to the 'real' upstream at this point#asahi · 2021-08-09 18:10
svenyeah, right#asahi · 2021-08-09 18:10
svenliterally just "take this message, get the endpoint from the client priv data, stuff it to the fifo"#asahi · 2021-08-09 18:10
sventhe mailbox driver would be very dumb#asahi · 2021-08-09 18:09
sventhe endpoint map would be the responsibility of the rtkit lib#asahi · 2021-08-09 18:09
sven^-- that#asahi · 2021-08-09 18:09
svenuh, why do i need an endpoint to know how many endpoints there are?