diff options
| author | Arthur Hanson Rasmusson <arthur@vgpu.io> | 2026-10-14 11:39:30 -0400 |
|---|---|---|
| committer | Arthur Hanson Rasmusson <arthur@vgpu.io> | 2026-10-14 11:39:30 -0400 |
| commit | 624438bb0901680f416b879727b30a411dd3d78b (patch) | |
| tree | ec43b6c3f392c08621d9129f20acfcec44542fc5 /tools/cuda-ab | |
| parent | ea23d78e56a550ea86370363f1f7d1441c717376 (diff) | |
| download | open-iov-624438bb0901680f416b879727b30a411dd3d78b.tar.gz open-iov-624438bb0901680f416b879727b30a411dd3d78b.zip | |
vgsp-sema: author the PMA scrubber channel's progress (third-client scrub freeze)
The guest PMA scrubber (in-guest CeUtils CE channel, 4096 GPFIFO entries)
waits on its channel buffer's finish payload (offset 0x6c004) after the
third CUDA session's pages must be scrubbed; nothing in the emulation
authored it, so cudaMalloc froze in scrubCheckAndWaitForSize.
- channel_add: detect the 4096-entry scrubber channel and stash its USERD
PA/ring geometry (the channel itself is left untouched: no rank change,
no pump change - the colliding numeric-handle FREE may still unregister
it, the geometry survives).
- sema_scrubbuf_add debug param: register the finish/host-semaphore word
PAs the guest scanner resolves (vmallocinfo + kcore walk), because the
RM kernel VA space collides numerically with the client CE window in the
global VA map and cannot be resolved through it.
- svc_scrub_pass (service thread, per pass): from the channel's own USERD
GPPut, track submitted scrub blocks across wraps, then author USERD
GP_GET = put, host semaphore = put and finish payload = submitted.
Forward-only, idempotent, zero writes while the scrubber idles.
Live-validated mechanism (r137 VM): authoring the finish word at the kcore-
resolved PA let the frozen victory_matmul leave scrubCheckAndWaitForSize
and proceed to its launch.
Diffstat (limited to 'tools/cuda-ab')
| -rwxr-xr-x | tools/cuda-ab/push_battery.sh | 22 |
1 files changed, 22 insertions, 0 deletions
diff --git a/tools/cuda-ab/push_battery.sh b/tools/cuda-ab/push_battery.sh new file mode 100755 index 0000000..644beba --- /dev/null +++ b/tools/cuda-ab/push_battery.sh @@ -0,0 +1,22 @@ +#!/bin/bash +# push_battery.sh - push every CUDA battery binary from tools/cuda-ab/run/ +# to the running guest via scp over the SSH hostfwd (port 2222). +# Usage: push_battery.sh [guest_dir] (default /tmp) +set -u +HERE="$(cd "$(dirname "$0")" && pwd)" +RUN="$HERE/run" +GUEST_DIR="${1:-/tmp}" +CLIENTS="victory_mail smiley vecadd_loop streams2 memcpy_roundtrip multikernel victory_saxpy victory_reduce victory_matmul victory_strided victory_memset2d victory_atomics victory_streams_events victory_math vector_add" +export SSHPASS="${VGPU_GUEST_PASS:-ubuntu}" +FILES="" +for c in $CLIENTS; do [ -x "$RUN/$c" ] && FILES="$FILES $RUN/$c"; done +[ -n "$FILES" ] || { echo "no binaries in $RUN"; exit 1; } +# shellcheck disable=SC2086 +sshpass -e scp -P "${VGPU_GUEST_PORT:-2222}" \ + -o StrictHostKeyChecking=no -o UserKnownHostsFile=/dev/null \ + -o PubkeyAuthentication=no -o PreferredAuthentications=password \ + -o ConnectTimeout=15 $FILES "ubuntu@127.0.0.1:$GUEST_DIR/" && \ +sshpass -e ssh -p "${VGPU_GUEST_PORT:-2222}" \ + -o StrictHostKeyChecking=no -o UserKnownHostsFile=/dev/null \ + -o PubkeyAuthentication=no -o PreferredAuthentications=password \ + "ubuntu@127.0.0.1" "chmod +x $GUEST_DIR/* 2>/dev/null; ls $GUEST_DIR | wc -l; echo BATTERY_PUSHED" |
