#asahi-dev · 2022-01-23 13:48
svenits still treating “owned by rtkit” and “is iomem” as the same thing#asahi-dev · 2022-01-23 13:47
svenhrm, no#asahi-dev · 2022-01-23 13:47
svenbut it’s just the is_iomem name that’s misleading now#asahi-dev · 2022-01-23 13:46
svenhrm, that still has the if (iova) is_iomem = true; logic, doesn’t it?#asahi-dev · 2022-01-23 13:04
svenI’d be surprised if that’s one isn’t just normal RAM#asahi-dev · 2022-01-23 12:53
svenisn’t the frame buffer also mapped through the disp0 dart instead?#asahi-dev · 2022-01-23 12:53
svenjannau: I’d have to double check but the frame buffer is the only thing that’s special iirc#asahi-dev · 2022-01-23 12:53
sven50:50 chance I chose the wrong mode ;)#asahi-dev · 2022-01-23 12:46
svenand for SMC the same as MMIO (nGnRE or whatever non-pcie uses) #asahi-dev · 2022-01-23 12:45
svenkettenis: for the DCP ones the same as RAM#asahi-dev · 2022-01-23 12:45
svenjannau: true #asahi-dev · 2022-01-23 12:16
svenhm, true#asahi-dev · 2022-01-23 12:01
svenshe also has to map carve out memory somewhre#asahi-dev · 2022-01-23 12:00
svenshe also has to#asahi-dev · 2022-01-23 12:00
svenah, right. take a look at alyssa’s locked dart hacks #asahi-dev · 2022-01-23 11:59
svenDCP has the crashlog buffer as owned by rtkit but the rest owned by Linux but both are normal RAM#asahi-dev · 2022-01-23 11:59
svenit cares if the buffer is iomem (MMIO) or normal RAM#asahi-dev · 2022-01-23 11:59
svenit doesn’t care who owns the buffers #asahi-dev · 2022-01-23 11:50
svenrtkit.c itself doesn’t need to care who owns the buffer I think#asahi-dev · 2022-01-23 11:49
svenIt’s just to check if we need to use memcpy_fromio instead of memcpy#asahi-dev · 2022-01-23 11:49
svenIt’s different from who owns the buffers#asahi-dev · 2022-01-23 11:49
svenlooks good, is_iomem should be set from the driver though. DCP never needs it but SMC always needs#asahi-dev · 2022-01-23 10:45
svenI’d also get rid of flags competely and just have some is_iomem variable as part of the shmem struct#asahi-dev · 2022-01-23 10:44
svenno bug_on please. Just returns a error code#asahi-dev · 2022-01-21 17:24
svenah, right. Didn’t even think about bootargs#asahi-dev · 2022-01-21 17:16
sven🤨🤨🤨#asahi-dev · 2022-01-21 17:06
svenAnd just with a single “v 1.3.whatever” commit instead of full commit history ofc#asahi-dev · 2022-01-21 16:57
svenyeah. I kinda hope we can make that as separate as possible eventually because I’m sure people will want to audit that#asahi-dev · 2022-01-21 16:55
svenI mainly thought about “possible attack surface” and less about absolute file size fwiw#asahi-dev · 2022-01-21 16:54
svenThe fonts are quite large iirc#asahi-dev · 2022-01-21 16:51
svenwell... i don't want a random person to be able to reboot my machine, connect a usb cable and get full access for once ;)#asahi-dev · 2022-01-21 16:50
svenno need to ship the hv and proxyclient to everyone i guess#asahi-dev · 2022-01-21 16:50
svenonce this all gets more end user friendly / stable / production-ready we'll probably want to add a config.h option (or somthing similar) to strip down m1n1 anyway #asahi-dev · 2022-01-21 16:48
svenand i'm still amazed how simple nvme is compared to the horror that is usb#asahi-dev · 2022-01-21 16:47
sveni mainly need it to investigate this weird random write + flush bug#asahi-dev · 2022-01-21 16:11
svenah! now it makes sense#asahi-dev · 2022-01-21 16:06
svenwhich isn't a big deal because those conflicts are trivial but we can easily avoid them *shrug*#asahi-dev · 2022-01-21 16:06
svenotherwise we have multiple PRs from multiple subsystems which all conflict in the same section of that MAINTAINERS file#asahi-dev · 2022-01-21 16:05
svenit doesn't need to but it apparently makes life of upstream easier during PR time#asahi-dev · 2022-01-21 16:05
svenpovik: okay, Acked-by: Sven Peter <sven@svenpeter.dev> :-P#asahi-dev · 2022-01-21 16:04
sveni think it only needs an ack if it's to-be-merged through a different tree