#asahi-re · 2021-09-13 11:17
alyssa08:09 <@marcan> the dp/usb/phy/tcpm model is the same as what we have#asahi-re · 2021-09-13 11:17
alyssa08:09 <@marcan> sven: fwiw, rk3399 is a good model for this#asahi-re · 2021-09-12 15:34
alyssauh#asahi-re · 2021-09-12 15:01
alyssajust a question of whether the AP can bypass the DCP (directly or indirectly), and whether that's ever useful#asahi-re · 2021-09-12 14:59
alyssawell, "fake" .... there still an i2c bus that the DCP is using internally to get at display properties#asahi-re · 2021-09-12 14:59
alyssausually there's an i2c bus for the display and that does get routed into userspace but AFAIK there's no good reason to fake it in our case#asahi-re · 2021-09-12 14:58
alyssai mean#asahi-re · 2021-09-12 14:58
alyssano, I don't think so#asahi-re · 2021-09-12 14:58
alyssait's unclear to me if the other endpoints expose raw "send i2c, recv i2c" calls#asahi-re · 2021-09-12 14:58
alyssaif everything we need is accessible over dcp, we don't /need/ i2c from the ap#asahi-re · 2021-09-12 14:57
alyssasven: what makes you say that?#asahi-re · 2021-09-12 11:49
alyssalooks like we can get to i2c via dcp over one of the other endpoints (dpdev, mdcp29xx, ..)#asahi-re · 2021-09-04 12:17
alyssa :)#asahi-re · 2021-09-04 10:55
alyssasven: Are you saying this for driver purposes or exploitation purposes? :-p#asahi-re · 2021-09-04 10:54
alyssalovely#asahi-re · 2021-09-02 18:45
alyssadebating matching that in my code so the syslog makes more sense#asahi-re · 2021-09-02 18:44
alyssamarcan: btw, I think the DCP calls channels "streams"#asahi-re · 2021-08-31 01:09
alyssaso.... reg1 = scratch and reg2:bit = doorbell#asahi-re · 2021-08-31 01:08
alyssamarcan: Hmmmm that looks very familiar :')#asahi-re · 2021-08-31 01:07
alyssa[syslog] * [UPPipeDCP_H13P.cpp:9025]IOMFB Int RTBandwidth: using legacy. Doorbell 23bc3c000, bit 2#asahi-re · 2021-08-31 01:07
alyssa[syslog] * [UPPipeDCP_H13P.cpp:9005]IOMFB Int RTBandwidth: program_M3_rt_config: Using Legacy, scratch 23b738014#asahi-re · 2021-08-30 03:33
alyssabbl sleep :p#asahi-re · 2021-08-30 03:33
alyssa\o#asahi-re · 2021-08-30 03:30
alyssa....I feel better about my own code now#asahi-re · 2021-08-30 03:28
alyssa size = 0x0000000000800000, Container: #asahi-re · 2021-08-30 03:28
alyssa addr = 0x0000000031800000#asahi-re · 2021-08-30 03:28
alyssaoh er that was dispext0 i'm tired#asahi-re · 2021-08-30 03:27
alyssa size = 0x0000000000800000, Container: #asahi-re · 2021-08-30 03:27
alyssa addr = 0x0000000071800000#asahi-re · 2021-08-30 03:27
alyssaof the disp0 regs only suspicious one is#asahi-re · 2021-08-30 03:27
alyssa(so the regs are still in the DT, not hardcoded)#asahi-re · 2021-08-30 03:27
alyssa<reference> being a reference to a DT name/index/handle/whatever?#asahi-re · 2021-08-30 03:26
alyssaGrumble#asahi-re · 2021-08-30 03:25
alyssasince AFAIK we have no reason to touch those registers ourselves ever#asahi-re · 2021-08-30 03:25
alyssaI see no reason not to just include the regs in Apple order in the DT#asahi-re · 2021-08-30 03:25
alyssathat's what I assumed but am trying to check my assumptions off you since you complained I don't ask enough questions 😋#asahi-re · 2021-08-30 03:24
alyssa👍#asahi-re · 2021-08-30 03:23
alyssa(Obviously letting a copro map whatever memory it wants defeats the purpose of having an IOMMU)#asahi-re · 2021-08-30 03:22
alyssamap_physical, the DCP only calls it with disp0 regs, yeah?#asahi-re · 2021-08-30 03:22
alyssaoh meant to ask about this#asahi-re · 2021-08-30 03:22
alyssa+ /* XXX: need to validate the DCP is allowed to access */#asahi-re · 2021-08-30 03:22
alyssaSnrkt#asahi-re · 2021-08-30 03:21
alyssaalright, won't worry so much then#asahi-re · 2021-08-30 03:20
alyssaalthough then I needed deferred probe nonsense --- yeah#asahi-re · 2021-08-30 03:20
alyssaOkay, that's what I have now#asahi-re · 2021-08-30 03:20
alyssain the DT is that modeled as sibling nodes or as parent/child?#asahi-re · 2021-08-30 03:19
alyssaHum, alright. That seems sane#asahi-re · 2021-08-30 03:19
alyssaon the other hand, the DCP module needs to "own" the DCP dart so the rtkit crap can just do dma_alloc_coherent properly#asahi-re · 2021-08-30 03:18
alyssathe KMS driver needs to "own" disp0 stream 0, so that way framebuffer memory can be allocated by the DRM core and be automatically mapped with no driver twiddling#asahi-re · 2021-08-30 03:18
alyssasven linked me to your clock gate thing as an example..?