diff options
| -rw-r--r-- | STATE-CURRENT.md | 59 |
1 files changed, 30 insertions, 29 deletions
diff --git a/STATE-CURRENT.md b/STATE-CURRENT.md index 237eba6..5481313 100644 --- a/STATE-CURRENT.md +++ b/STATE-CURRENT.md @@ -1,6 +1,6 @@ # Open-IOV Current State -> **Snapshot: 2026-09-30 06:45.** This is a dated public status page, not live +> **Snapshot: 2026-09-30 08:58.** This is a dated public status page, not live > state; it is refreshed on every progress check. Live working notes are > maintained outside this repository; the public known-knowns/unknowns document > is authoritative here: @@ -29,8 +29,8 @@ meant to execute on the physical GPU. | `nvidia-smi -L` enumerates the emulated TU104 | ✅ | | CUDA userspace: `cuInit`, `cuDeviceGetCount`, `cuDeviceGet` | ✅ | | `cuCtxCreate` returns | ✅ (the whole smoke test now completes) | -| `cuMemAlloc` + `cuMemcpy` roundtrip (`ALL OK` in the smoke test) | 🟡 all calls succeed and the test prints `ALL OK`; the 64-byte data roundtrip is not yet correct | -| Full `nvidia-smi` output | ❌ (faults on the CUDA path) | +| `cuMemAlloc` + `cuMemcpy` roundtrip (`ALL OK` in the smoke test) | ✅ payload roundtrip verified 64/64 in the guest | +| Full `nvidia-smi` output | ❌ next: run real CUDA workloads in the guest | | CUDA kernel execution (`vectorAdd`) | ❌ not started (forwarding layer absent) | | Graphics (3D), video | ❌ | | Faults, multi-version, Windows, migration, multi-tenancy, performance | ❌ | @@ -43,20 +43,20 @@ chain succeeds. ## Right now (frontier) -- **The smoke test now completes end-to-end**: `cuInit -> cuDeviceGet -> - cuCtxCreate -> cuMemAlloc -> cuMemcpyHtoD -> cuMemcpyDtoH -> cuMemFree -> - cuCtxDestroy` all succeed and the test prints `ALL OK`. The context-creation - error that stalled this for days is root-caused and fixed: the emulated GR - information table left the shader-pipe count and related clock fields zero, - which the client's clock-table validation rejected. -- **The payload's host-to-device path is found**: the copy is a - compute-class inline-data DMA (the payload rides inside the push itself), - not a copy-engine transfer. The decoder now recognises the method set and - attempts the write. The remaining blocker is the destination's - GPU-virtual-address translation: the emulated page-table walk has never - resolved entries (every translation so far came from pins or the discovery - map), so the write cannot land yet. The live page-directory writes are - being audited to fix the walk. +- **The full CUDA driver-API smoke test passes with correct data**: the client + completes `cuCtxCreate -> cuMemAlloc -> cuMemcpyHtoD -> cuMemcpyDtoH -> + cuMemFree -> cuCtxDestroy` and prints `ALL OK`, with the 64-byte payload + roundtripping exactly (64/64 verified in the guest). +- **The last two walls**: the GPU page-table walk's address decode (child + pointers and leaves were 16x too low; fixed with the real entry encoding - + walks now succeed at scale) and the device-to-host copy (a copy-engine + class transfer bound on a compute channel whose work was denied by the + engine-type gate; trusted host-injected pushes may now execute copy-engine + work as well). +- **Next: run a real CUDA workload in the guest** (`nvidia-smi`, a + `vectorAdd`-style kernel launch) and iterate on whatever the runtime path + asks for - this moves the frontier toward kernel execution, which still + requires the execution backend. - Recap of the fixes behind this point: all 20/20 compute-channel releases land (three decode paths routed to the exact counter slot); registration delivers the incrementing tracking release for every consumed submission; @@ -64,9 +64,8 @@ chain succeeds. the copy-engine decoder consults pinned translations before the scanned map, so every push-data operation executes (`memset=19473`, `ringops=10551` live). -- Active work: audit the live page-directory writes to fix the page-table - walk, so the inline write (and then the device-to-host copy) resolves - client device addresses; iterate with a fresh boot per run. +- Active work: run a real CUDA workload in the guest (`nvidia-smi`, a kernel + launch) and iterate on the runtime path; iterate with a fresh boot per run. - Observability: a size-keyed copy watch, a VRAM shadow dump/search, and an operation log are now module parameters, and the test's output survives guest-agent stalls via the serial console. @@ -119,18 +118,19 @@ Status tags: `[done]`, `[frontier]` (actively worked), `[partial]`, | GR + CE channels; push buffers | semaphore pool; completion counters v - [partial] ---------------------------------------------> cuMemAlloc + cuMemcpy roundtrip - | (all calls succeed; payload data path pending) - [not yet] ---------------------------------------------> full nvidia-smi output + [done] ------------------------------------------------> cuMemAlloc + cuMemcpy roundtrip + | (passes end-to-end; payload 64/64 correct) + [not yet] ---------------------------------------------> full nvidia-smi output / kernel launch ``` Work execution (host service vs client synchronization): ``` - [partial] host decodes guest pushes (method streams): memsets, copies, releases - - CE pushes execute at scale (memset=19473, ringops=10551; misses at zero) - - the 64-byte test payload's HtoD path is not decoded yet - [partial] GPU VA -> guest PA translation and pinning; server-owned page tables in progress + [done] host decodes guest pushes (method streams): memsets, copies, releases + - CE pushes execute at scale; the compute-class inline-DMA HtoD and + the copy-engine-on-compute DtoH both execute; payload 64/64 + [partial] GPU VA -> guest PA translation: page-table walk fixed and succeeding; + server-owned page-table population still in progress [not yet] real work forwarded to the physical GPU (host-RM proxy or direct submission) [not yet] correct results (vectorAdd) ``` @@ -153,8 +153,9 @@ Recent chain of walls removed (2026-09-28 to 2026-09-30): -> deterministic chain solved -> pin-drop and release-push fixes -> copy-engine data path executes -> compute-window pinning -> registration release delivery -> 20/20 counter slots -> GR-info clock fix - -> cuCtxCreate OK and the full smoke test ALL OK; payload data path remains - -> payload found (compute-class inline DMA) -> page-table walk audit + -> cuCtxCreate OK; payload found (compute-class inline DMA) + -> GMMU walk decode fixed (child/leaf encoding) -> inline HtoD lands + -> DtoH copy-engine-on-compute allowed -> payload roundtrip 64/64 ``` ## Tools, code paths and fixes behind this progress |