aboutsummaryrefslogtreecommitdiffstats
diff options
context:
space:
mode:
authoruser <user@clank>2026-10-08 16:30:08 +0200
committeruser <user@clank>2026-10-08 16:30:08 +0200
commit4bf30ca9cf2fd2ad0f625ef665c7243fa0056701 (patch)
tree7533bf0241466a0a8e9d8b42573cc30b2edc9efd
parentre-evaluate GCP attack surface (post-retraction): capability-capped SYS_DRV r... (diff)
downloadamd-psp-re-4bf30ca9cf2fd2ad0f625ef665c7243fa0056701.tar.gz
amd-psp-re-4bf30ca9cf2fd2ad0f625ef665c7243fa0056701.tar.xz
amd-psp-re-4bf30ca9cf2fd2ad0f625ef665c7243fa0056701.zip
crypto output read-back: principle OK (create handle), but RNG 'identical output' was STALE GCP memory
Follow-up to the hijack retraction. Pointing a crypto op output at GCP memory clear of the ROP chain should let us read engine output back. Principle validated by op 0x01 create: handle 0x10 read back via the message copy-back = a confirmed crypto output read-back. RNG (op 0x35) data-buffer demo BLOCKED: output->0x410500 (len 4 and 32) gate-rejects (status 0xffff000f = dword_E1DF48 buffer-gate fail) -> RNG never wrote. Decisive control: a create baseline (writes nothing to 0x410500) dumped the SAME bytes bb18bb63... as the RNG runs across 3 boots. So the 'identical random-looking output' is STALE GCP memory at 0x410500, NOT RNG output -> the deterministic- seed hypothesis does not apply (RNG didn't run). 0x410500 is GCP-mapped (dump reads it) but SVC 0xB5 rejects it as an op-0x35 OUTPUT buffer; earlier single-op ptr_sweep accepted op0x35->0x410500 len4 at status 0, so the gate accept/reject is harness/len/state-sensitive, not yet pinned. Updated harness doc.
-rw-r--r--fw/HARNESS-GCP-CRYPTO-ROP.md24
1 files changed, 24 insertions, 0 deletions
diff --git a/fw/HARNESS-GCP-CRYPTO-ROP.md b/fw/HARNESS-GCP-CRYPTO-ROP.md
index 6a87adf..e58ce49 100644
--- a/fw/HARNESS-GCP-CRYPTO-ROP.md
+++ b/fw/HARNESS-GCP-CRYPTO-ROP.md
@@ -111,3 +111,27 @@ its result inline in the message, or (b) defeating the cross-domain buffer bounc
where the SYS_DRV handler actually dereferences it) -- neither available as-is. This is the accurate
boundary of the GCP->SYS_DRV crypto surface: inline-result ops work end-to-end; buffer-I/O ops
execute but their data does not cross back.
+
+================================================================================
+## Reading crypto OUTPUT back via GCP shared memory -- principle OK, RNG demo blocked
+
+Follow-up to the retraction: since the crypto buffer write lands in GCP's own shared memory, pointing
+an op's output at GCP memory CLEAR of the ROP chain should let us read the engine output back.
+
+* **Principle validated** by `TEE_AllocateOperation` (op 0x01): it returns the context handle `0x10`
+ inline, and we read it back from GCP memory via the message copy-back. That is a confirmed crypto
+ output read-back.
+* **RNG (op 0x35) data-buffer demo: blocked.** Pointed op 0x35 output at `0x410500` (clear of the
+ chain), len 4 and 32, dumped `0x410500`. Result: `status = 0xffff000f` (crypto dispatcher
+ buffer-gate reject, `dword_E1DF48`) -- the RNG never wrote. Decisive control: a `create` baseline
+ (writes nothing to `0x410500`) dumped the SAME bytes `bb18bb63 aa3069a5 ...` as the RNG runs, across
+ 3 boots.
+* **=> the "identical random-looking output" is STALE GCP memory at `0x410500`, not RNG output.** The
+ seeded/deterministic-RNG hypothesis does not apply -- the RNG did not run. (`0x410500` reads fine
+ from GCP context, so it is GCP-mapped; but `SVC 0xB5` rejects it as an op-0x35 OUTPUT buffer -- the
+ earlier single-op `ptr_sweep` accepted `0x410500` for op 0x35 len 4 at status 0, so the gate's
+ accept/reject is harness/len/state-sensitive and not yet pinned.)
+
+Open: to demo a data-crypto output read-back, resolve why `SVC 0xB5` rejects the op-0x35 output buffer
+here (window/direction/len), or use an op whose output buffer the gate accepts, then dump it and
+compare against a no-write baseline to prove it is engine output (not stale).