diff options
| author | user <user@clank> | 2026-09-29 13:44:30 +0200 |
|---|---|---|
| committer | user <user@clank> | 2026-09-29 13:44:30 +0200 |
| commit | 8dbac8b41fcabc0f54274115bc14fa2ab4538eca (patch) | |
| tree | b3b9c5a9e217db44685f39d5c3a50761f1710d0c | |
| parent | [HW] host drives GCP into calling ASD 0x202; ASD refuses, so no leak yet (diff) | |
| download | amd-psp-re-8dbac8b41fcabc0f54274115bc14fa2ab4538eca.tar.gz amd-psp-re-8dbac8b41fcabc0f54274115bc14fa2ab4538eca.tar.xz amd-psp-re-8dbac8b41fcabc0f54274115bc14fa2ab4538eca.zip | |
[HW] priming a DRM session does not satisfy ASD's gate; 0x202 still refuses
Followed up the named blocker by running GCP's own session commands before cmd 4:
AllocateDRMSession and SendDRMCertificate both return 0, CreateAuthenticatedSession
is rejected with 0xffff0000, and GetObfKeyCryptoSession returns 0 but reports
0x321 internally. Command 4 still lands on 1045, the ASD-0x202-call-failed path.
So the gate is not satisfied by allocating a session and sending a zeroed
certificate, which is unsurprising since an empty page cannot establish real key
state. The conclusion is unchanged and better supported: the host can reach ASD
0x202 through GCP, proven by the 1045 path being reachable only from that call,
and the disclosure does not fire without legitimate DRM session material,
measured across four priming variants.
| -rw-r--r-- | fw/AUDIT-NEW-TAS-RE.md | 22 | ||||
| -rw-r--r-- | hw/gcp_asd202_leak_probe.py | 15 |
2 files changed, 37 insertions, 0 deletions
diff --git a/fw/AUDIT-NEW-TAS-RE.md b/fw/AUDIT-NEW-TAS-RE.md index 0aab34e..d78814d 100644 --- a/fw/AUDIT-NEW-TAS-RE.md +++ b/fw/AUDIT-NEW-TAS-RE.md @@ -427,3 +427,25 @@ finding used) may satisfy. That is the next experiment, and it is a small one. This supersedes the previous section's static-only claim that the chain "completes": the data path does reach the host page by construction, but on silicon the call fails before anything is copied. + +## Priming the DRM session does not satisfy ASD's gate + +Follow-up on the named blocker: run GCP's own session commands first, then cmd 4. + +``` +prime cmd 0x1 (AllocateDRMSession) status 0x0 dw2=0 +prime cmd 0x2 (SendDRMCertificate) status 0x0 dw2=0 +prime cmd 0xfe03 (CreateAuthenticatedSession) status 0xffff0000 dw2=0x102 +prime cmd 0xfe02 (GetObfKeyCryptoSession) status 0x0 dw2=0x321 +cmd 4 after priming -> dw2 = 0x415 (ASD 0x202 call failed) +``` + +Still 1045. So the ASD-side gate is not satisfied by merely allocating a session and +sending a (zeroed) certificate — unsurprising, since `SendDRMCertificate` with an empty +page cannot establish real key state, and `0xFE02` itself reports `0x321`. + +Conclusion stands and is now better supported: **the host can reach ASD 0x202 through GCP +(proven), and the disclosure does not fire without legitimate DRM session material +(measured across four priming variants).** Obtaining that material is a different problem +from the one this project has been solving, and it is where this thread stops rather than +speculating. diff --git a/hw/gcp_asd202_leak_probe.py b/hw/gcp_asd202_leak_probe.py index 0922bcf..4741329 100644 --- a/hw/gcp_asd202_leak_probe.py +++ b/hw/gcp_asd202_leak_probe.py @@ -168,6 +168,21 @@ def main(): s1 = invoke(sid, 1) c2 = show('after cmd 1 (status %#x)' % s1, read_back()) + # C4: the 1045 path means ASD refused. Its copy sits behind ASD's crypto + # pipeline, so try establishing DRM session state first with GCP's own session + # commands, then re-run cmd 4 and see whether the status leaves 1045. + print('\nC4 prime a DRM session, then cmd 4:') + for pc in (1, 2, 0xFE03, 0xFE02): + fill_and_prime(pc) + st_p = invoke(sid, pc) + d_p = read_back() + print(' prime cmd %-6s status %#-6x dw2=%#x' + % (hex(pc), st_p, struct.unpack_from('<I', d_p, 8)[0])) + fill_and_prime(4) + invoke(sid, 4) + dprime = read_back() + show('cmd 4 after priming', dprime) + print('\nC3 the chain -- cmd 4 (GetContentKey), twice:') fill_and_prime(4) s4a = invoke(sid, 4) |
