#
✕ nick:phire
#asahi-re · 2022-02-04 15:41
phirehowever... the cpu didn't write into it
#asahi-re · 2022-02-04 15:40
phirenope, it's clear at boot
#asahi-re · 2022-02-04 15:39
phireum, let me just check this isn't just stale data from iboot
#asahi-re · 2022-02-04 15:36
phireimplies something has direct access to physical memory
#asahi-re · 2022-02-04 15:35
phirethe fact that I'm seeing physical pointers in this gfx-shared-region is interesting
#asahi-re · 2022-02-04 15:34
phirenot sure I want to go much beyond looking at strings RE wise
#asahi-re · 2022-02-04 15:32
phireand doesn't look virtual
#asahi-re · 2022-02-04 15:32
phireit's outside of physical memory
#asahi-re · 2022-02-04 15:24
phireit's only bits 16-23 that change
#asahi-re · 2022-02-04 15:23
phirefew hundred 64bit values with slight increments between them
#asahi-re · 2022-02-04 15:23
phiresecond points to a table with data like: 0x00600004061e0447, 0x00600004061e4447, 0x00600004061e8447...
#asahi-re · 2022-02-04 15:20
phirefirst pointer points to nulls
#asahi-re · 2022-02-04 15:20
phirestuff that doesn't seem to be that interesting at a first glance
#asahi-re · 2022-02-04 15:15
phireor 0x1
#asahi-re · 2022-02-04 15:15
phireyeah, it's always 0x3
#asahi-re · 2022-02-04 14:59
phirethey are also all unaligned
#asahi-re · 2022-02-04 14:56
phireoh hang on, the pointers witten to gfx-shared-region-base actually point at real physical memory
#asahi-re · 2022-02-04 14:51
phireThere is very little data sent via the ASC interface. It basically looks like little more than a keepalive
#asahi-re · 2022-02-04 14:44
phireThat's where I'm getting these "pointers" that I don't know where they point to
#asahi-re · 2022-02-04 14:43
phireyes
#asahi-re · 2022-02-04 14:40
phireno, that only has the expected asc traffic
#asahi-re · 2022-02-04 14:31
phiremaybe it's on the agx-asc node's registers
#asahi-re · 2022-02-04 14:04
phireyeah, I think it does
#asahi-re · 2022-02-04 14:04
phiredepends on if trace_device("/arm-io/sgx", True) automatically picks them up
#asahi-re · 2022-02-04 14:03
phirebut it's passing in these pointers that we are assuming are GART pointers
#asahi-re · 2022-02-04 14:02
phireI suspect work is being passed in via gfx-handoff
#asahi-re · 2022-02-04 14:02
phireThen there are a few shared memory regions, gpu-region-base, gfx-handoff-base and gfx-shared-region-base
#asahi-re · 2022-02-04 14:01
phirea few of the registers seem to count up when work is executed
#asahi-re · 2022-02-04 14:01
phire0x11008-0x11014 seem to be some perf counters
#asahi-re · 2022-02-04 14:00
phire0x4000-0x401c seem to be id/version registers
#asahi-re · 2022-02-04 13:59
phireI've never seen those registers written to. only read
#asahi-re · 2022-02-04 13:58
phirethe sgx node?
#asahi-re · 2022-02-04 13:57
phireSo, assuming it is registers, they seem to be missing from ADT
#asahi-re · 2022-02-04 13:56
phireIt could require interaction on both sides?
#asahi-re · 2022-02-04 13:55
phirethough I'm with alyssa. I'm expecting registers controlled by the cpu
#asahi-re · 2022-02-04 13:55
phirewhich susgest the firmware is somewhat involved in the process of mapping
#asahi-re · 2022-02-04 13:54
phirewell the strings include: "AGXSecureGart::unmap Timed out waiting for FW to process secure unmap after %llu ns" @%s:%d
#asahi-re · 2022-02-04 13:54
phirepoint