#
✕ nick:sven
#asahi · 2022-08-05 12:55
svenyeah, nvm. So they did make it in time
#asahi · 2022-08-05 12:55
svener, wait
#asahi · 2022-08-05 12:54
svenah, for 6.0 though
#asahi · 2022-08-05 12:52
svenohhh… because the PR was sent too late?
#asahi · 2022-08-05 12:52
svenhuh
#asahi · 2022-08-05 12:52
svenyou probably also want to take individual bits/ branches instead of manually picking patches
#asahi · 2022-08-05 12:51
svenusb will work as well as long as you have the dtb nodes
#asahi · 2022-08-05 12:51
svennvme is upstream and will even work with 0 patches
#asahi · 2022-08-05 12:24
svenyeah, would’ve been a bit easier to argue that brcm,board-type is a good idea if it was already upstream for WiFi
#asahi · 2022-08-05 12:06
svenand it’ll mostly work. usually up until the point you try to unplug the device and use the port for another one :D
#asahi · 2022-08-05 11:39
svenyou can reset it, that's good enough!
#asahi · 2022-08-05 11:38
sven^--
#asahi · 2022-08-05 11:38
sventhe mini is quite useable with the upstream kernel fwiw. (for my definition of usable at least)
#asahi · 2022-08-05 08:34
svenso, back to the original question, does pcie even come back up from s2idle? i remember that the power domain is set to always-on now but is that enough?
#asahi · 2022-08-05 08:33
svenyeah, that too
#asahi · 2022-08-05 08:33
sven*sigh*
#asahi · 2022-08-05 08:33
sven<mps> no from mem suspend
#asahi · 2022-08-05 08:33
sven<sven> from s2idle?
#asahi · 2022-08-05 08:33
sven** s2idle **
#asahi · 2022-08-05 08:32
svenmem is just an alias for that
#asahi · 2022-08-05 08:32
svenwhat's inside /sys/power/mem_sleep?
#asahi · 2022-08-05 08:32
svenyeah, i doubt that
#asahi · 2022-08-05 08:32
sventhat sounds like suspend-to-ram which i'd be very surprised about
#asahi · 2022-08-05 08:31
svenwhat is mem suspend?
#asahi · 2022-08-05 08:30
svenfrom s2idle?
#asahi · 2022-08-05 08:26
svenj`ey: would be great! though I dunno if pcie even comes back up after suspend
#asahi · 2022-08-05 08:24
svenbluetooth suspend seems to work on the t2 chips fwiw
#asahi · 2022-08-05 08:24
svenhuh, i thought suspend just doesn't work at all on the asahi branch?!
#asahi · 2022-08-05 08:06
svenI’m gonna use this as my daily driver eventually and I really want it to work well
#asahi · 2022-08-05 08:06
svenand even if there are weird issues we’ll make sure to fix them :)
#asahi · 2022-08-05 07:50
sventhere are others that have older/different chips that didn't expose bluetooth using that "Apple Converged IPC" but as UART instead
#asahi · 2022-08-05 07:49
svenat least some of them
#asahi · 2022-08-05 07:49
svenyes
#asahi · 2022-08-05 07:48
svenshould be BCM4378B3 for M2
#asahi · 2022-08-05 07:47
svenit's either broadcom 4378 or 4387 fwiw
#asahi · 2022-08-05 07:18
svenchadmed: yeah, imho this entire project is fun because we can very easily stay away from grey areas and just focus on making linux work. i kinda burned out on the whole console hacking things years ago because we always had to deal with entitled users who just wanted to pirates games
#asahi · 2022-08-04 16:41
sveniirc joerg picks up new code between rc4 and 6(?). Still some time left ;)
#asahi · 2022-08-04 16:29
svenbut the code is smart enough to use direct mappings for anything dma_alloc_coherent and anything that happens to be page aligned
#asahi · 2022-08-04 16:28
svenyup, for dma_map_single at least
#asahi · 2022-08-04 16:28
svenyeah, just need to pass through the page offset/size all the way to the iommu driver
#asahi · 2022-08-04 16:27
svenyeah… it seems to use dma_map_single at least in one place to map < PAGE_SIZE buffers so it’ll already expose something adjacent
#asahi · 2022-08-04 16:24
svenbut if brcmfmac uses non-coherent mappings anywhere it’ll still expose adjacent memory
#asahi · 2022-08-04 16:23
svenit doesn’t work at all with that patch set because those iommus don’t do pass through
#asahi · 2022-08-04 16:20
svenhm, that reminds me: didn’t you want to look into making WiFi untrusted as well because Broadcom, marcan?
#asahi · 2022-08-04 16:15
svenbut that reminds me that I have to add an api to allocate unmanaged domains with a page size mismatch once I eventually get back to that 4K patch
#asahi · 2022-08-04 16:14
svenjannau: yeah, that’s also a fairly big change unfortunately:(
#asahi · 2022-08-04 16:09
svenI don’t think drm uses it in general, just some drivers iirc
#asahi · 2022-08-04 16:08
svendcp clearly abuses that api
#asahi · 2022-08-04 16:08
svenit depends
#asahi · 2022-08-04 16:08
svenyeah, but DCP just needs to stop abusing the dma api

plus ›