diff --git a/docs/install-any-distro.md b/docs/install-any-distro.md index 033c4a09a..5dc8f64d4 100644 --- a/docs/install-any-distro.md +++ b/docs/install-any-distro.md @@ -209,7 +209,7 @@ that stays on its distribution's kernel only needs the second and third: | `linux-image-` / `linux-headers-` (deb), `1bit-kernel` + `1bit-kernel-devel` (rpm), `linux-1bit` + `linux-1bit-headers` (Arch) | the kernel 1bit OS boots: Linux 7.4-rc1 with this fleet's config, signed with the 1bit MOK | only if you want OUR kernel; the drivers build against yours as well | | `amdxdna-1bit-dkms` | `amd/xdna-driver` at `526c6081` plus the engine's patch series (`third_party/patches/amdxdna/0001-0005`), as DKMS source | the NPU. **6.10+ kernels only** (measured, not inherited — see below), signed with **your** MOK | | `ec-su-axb35-dkms` | the AXB35-02 board's embedded-controller driver (fans, EC temperature, power mode), GPL-2.0, upstream `cmetz/ec-su_axb35-linux` at `f62c2c22`, plus a DMI gate so it attaches only on an AXB35 | the fans on a Bosgame AXB35; inert on any other board | -| `1bit-npu-config` | `limits.d/99-amdnpu-memlock.conf` (unlimited memlock), `udev/rules.d/90-1bit-npu-power.rules` (NPU keep-awake), `modules-load.d/ec_su_axb35.conf`, and the public MOK certificate at `/usr/share/1bit/mok/MOK-Kernel.der` | the NPU runtime's locked command buffers, and engine#1051's suspend/resume hang | +| `1bit-npu-config` | `limits.d/90-1bit-npu-memlock.conf` (unlimited memlock — a **higher-numbered** file you keep still wins, because limits.d applies files in lexical order; the package sat at `99-` at first and a machine that had written its own `99-amdnpu-memlock.conf` by hand made `dpkg` refuse the install and a purge then deleted that file, which is why it is a namespaced `90-` now), `udev/rules.d/90-1bit-npu-power.rules` (NPU keep-awake), `modules-load.d/ec_su_axb35.conf`, and the public MOK certificate at `/usr/share/1bit/mok/MOK-Kernel.der` | the NPU runtime's locked command buffers, and engine#1051's suspend/resume hang | The three driver/config packages are the same set in all three families, and that set alone is a complete install for a machine that keeps its distribution's kernel — proved as its own path @@ -278,7 +278,7 @@ Verify (no model needed): dkms status # amdxdna-1bit/0.18.0 and ec-su-axb35/0.1, installed for your kernel modinfo -n amdxdna # .../updates/dkms/amdxdna.ko (Debian, Arch) or .../extra/amdxdna.ko.xz (Fedora) modinfo -F srcversion amdxdna # EE1F36A2B9BA294CE11AD80 is the series this release ships -cat /etc/security/limits.d/99-amdnpu-memlock.conf +cat /etc/security/limits.d/90-1bit-npu-memlock.conf ``` Uninstall puts the in-tree driver back: @@ -413,6 +413,7 @@ registry model, since this Lemonade version has no local-file registration): | a weekly release tag works in all three families, and upgrades | `scripts/mk-kernel-packages.sh 2026.45.0 --skip-kernel`, each family's own tool, in `fedora:latest` / `ubuntu:26.04` / `archlinux:latest` | deb `0.18.0-1 < 0.18.0-2026.45.0` (dpkg), rpm `0.18.0-1 < 0.18.0-2026.45.0` (rpmdev-vercmp), arch `vercmp` `-1`; each installs over the previous package (`rc=0`) and reports `0.18.0-2026.45.0` afterwards, with one `/usr/src/amdxdna-1bit-0.18.0` left on deb — no stale duplicate source tree | | uninstalling really restores the in-tree driver, in all three families | `scripts/pkg-proof.sh`, each family in its container, after the purge | the in-tree module is gone from the tree while ours is installed (dkms archives it to `/var/lib/dkms//original_module/`), and after removal both `modprobe --show-depends amdxdna` and the file are back on the `kernel/` path: `amdxdna.ko.zst` on Ubuntu and Arch, `amdxdna.ko.xz` on Fedora. Same three containers as the install rows, `PROOF-RC=0` | | uninstalling the kernel package leaves the machine as it found it | `scripts/pkg-proof.sh`, each family in its container, after removing the kernel package | deb `dpkg -P linux-image` rc=0, rpm `rpm -e 1bit-kernel-devel 1bit-kernel` rc=0, arch `pacman -R linux-1bit-headers linux-1bit` rc=0; `/boot/vmlinuz-7.4.0-rc1-1bit` (or `/boot/vmlinuz-1bit` on Arch) is gone and the module tree under `modules/7.4.0-rc1-1bit` has **0 files** left. deb **29/29**, rpm **27/27**, arch **27/27** rows PASS, `PROOF-RC=0` | +| a hand-written limits file does not block the install | `scripts/pkg-proof.sh`, all three families | with a pre-existing `99-amdnpu-memlock.conf` (65536, the file this fleet's own box wrote by hand on 2026-10-12) the install succeeds and both files are present, and removing the package leaves the hand-written one intact. Measured on the shipped `90-` name; at the original `99-` name the same scenario **failed** — `dpkg` rc=1 on the conffile conflict, and a purge then deleted the user's file. Cleaned up after the row, since a `99-` file legitimately overrides our `90-` and would change what the memlock rows below measure | | the memlock file is honoured, not just present | `scripts/pkg-proof.sh`, all three families, running the container with `--cap-add=sys_resource` | a probe login with the file rewritten to `4096` came back with exactly `4096` (lowering always works, so that value can only come from the file), and with the shipped file back the same login reports `ulimit -l: unlimited`. Without `CAP_SYS_RESOURCE` — which a container's default set excludes — raising the hard limit fails and the session keeps the container's `8192`, which is the container, not the package | | the distribution-kernel path, on its own | clean `ubuntu:26.04` container, packages built with `--skip-kernel`, `scripts/pkg-proof.sh deb` | 22 rows PASS and one `INFO no kernel packages in this dir` (the kernel rows skip themselves instead of failing): install, config, memlock reaching a login session, dkms building both modules against `7.0.0-43-generic`, `amdxdna` resolving to `updates/dkms`, uninstall, and the distribution's in-tree `amdxdna` back in place. This is the path a user on their own kernel takes | | the engine package and this set install on the same machine | clean `ubuntu:26.04` container, `scripts/pkg-proof.sh coexist`, engine `1bit-engine_2026.11.02-1_amd64.deb` | all 26 rows PASS: the engine installs alone, this set installs **over it** with `dpkg` reporting no file conflict, `/usr/share/1bit` holds both (`lemonade-config-fragment.json` from the engine, `mok/MOK-Kernel.der` from `1bit-npu-config`) with each file attributed to its own package, the engine still runs (`1bit 0.0.1`), both dkms modules build, purging this set leaves the engine's files intact, and the engine purges last. Neither `scripts/pack-proof.sh` (the engine) nor the rows above (this set) install both, so this is the combination nobody had tried | diff --git a/scripts/mk-kernel-packages.sh b/scripts/mk-kernel-packages.sh index 55d50986d..9890a6c38 100755 --- a/scripts/mk-kernel-packages.sh +++ b/scripts/mk-kernel-packages.sh @@ -123,7 +123,7 @@ mkdir -p "$cfg/etc/security/limits.d" "$cfg/etc/udev/rules.d" "$cfg/etc/modules- "$cfg/usr/share/1bit/mok" "$cfg/usr/share/doc/1bit-npu-config" "$cfg/usr/share/licenses/1bit-npu-config" # The NPU runtime is compiled into 1bit and locks command buffers; without unlimited memlock a loaded # model's allocations fail. The same two lines the box runs. -cat > "$cfg/etc/security/limits.d/99-amdnpu-memlock.conf" <<'EOF' +cat > "$cfg/etc/security/limits.d/90-1bit-npu-memlock.conf" <<'EOF' * soft memlock unlimited * hard memlock unlimited EOF @@ -152,7 +152,7 @@ The runtime configuration the 1bit NPU packages need. Configuration only: no bin | file | why | |---|---| -| `/etc/security/limits.d/99-amdnpu-memlock.conf` | the NPU runtime is compiled into `1bit` and locks command buffers; without unlimited memlock a loaded model's allocations fail | +| `/etc/security/limits.d/90-1bit-npu-memlock.conf` | the NPU runtime is compiled into `1bit` and locks command buffers; without unlimited memlock a loaded model's allocations fail | | `/etc/udev/rules.d/90-1bit-npu-power.rules` | keeps the XDNA2 NPU (`1022:17f0`) out of runtime autosuspend, whose resume path loses the firmware context (engine#1051) | | `/etc/modules-load.d/ec_su_axb35.conf` | loads the AXB35 board's EC driver (fans, EC temperature, EC power mode); the driver attaches only on an AXB35, so it is inert on other boards | | `/usr/share/1bit/mok/MOK-Kernel.der` | the **public** certificate the 1bit kernel and its modules are signed with, for Secure Boot | diff --git a/scripts/pkg-proof.sh b/scripts/pkg-proof.sh index 20bcc5071..f3e8e0afd 100755 --- a/scripts/pkg-proof.sh +++ b/scripts/pkg-proof.sh @@ -118,6 +118,39 @@ if [ -n "$intree" ]; then row PASS "the in-tree amdxdna is present" "$intree" else row FAIL "the in-tree amdxdna is present" "install the kernel's module tree (linux-modules- / kernel-core / linux)"; fi depmod -a "$kv" 2>/dev/null || : +# A machine may already carry its own memlock file — this fleet's own box wrote +# /etc/security/limits.d/99-amdnpu-memlock.conf by hand on 2026-10-12, before this package existed. With +# our file at that exact path the install FAILED (dpkg rc=1, conffile conflict) and a purge then deleted +# the hand-written file. So: the hand-written file stays, ours is a namespaced neighbour, and both facts +# are asserted here rather than assumed. +if [ "$fam" != coexist ]; then + hand=/etc/security/limits.d/99-amdnpu-memlock.conf + printf '* soft memlock 65536\n* hard memlock 65536\n# hand-written before this package\n' > "$hand" + case $fam in + deb) dpkg -i "$pkgdir"/1bit-npu-config_*.deb >/tmp/re.log 2>&1 ;; + rpm) rpm -i --nosignature "$pkgdir"/1bit-npu-config-[0-9]*.rpm >/tmp/re.log 2>&1 ;; + arch) pacman -U --noconfirm "$pkgdir"/1bit-npu-config-[0-9]*.pkg.tar.zst >/tmp/re.log 2>&1 ;; + esac + if [ $? -eq 0 ] && grep -q 65536 "$hand" && [ -f /etc/security/limits.d/90-1bit-npu-memlock.conf ]; then + row PASS "a hand-written limits file does not block the install" "ours at 90-…, ours-and-yours both present" + else + row FAIL "a hand-written limits file does not block the install" "$(tail -2 /tmp/re.log)" + fi + case $fam in + deb) dpkg -P 1bit-npu-config >/tmp/rr.log 2>&1 ;; + rpm) rpm -e 1bit-npu-config >/tmp/rr.log 2>&1 ;; + arch) pacman -R --noconfirm 1bit-npu-config >/tmp/rr.log 2>&1 ;; + esac + if [ $? -eq 0 ] && grep -q 65536 "$hand"; then + row PASS "the hand-written file survives our removal" "still 65536" + else + row FAIL "the hand-written file survives our removal" "$( [ -e "$hand" ] && echo 'content changed' || echo 'deleted')" + fi + # clean up after the scenario: a 99- file sorts after our 90- and would legitimately win (later file + # wins in limits.d), which would make the memlock rows below measure the user's limit, not ours + rm -f "$hand" +fi + # ---- install the three packages -------------------------------------------------------------- case $fam in coexist) @@ -151,7 +184,7 @@ fi # ---- 1bit-npu-config ------------------------------------------------------------------------ check "memlock limits.d" "2 lines, unlimited" \ - grep -q 'memlock unlimited' /etc/security/limits.d/99-amdnpu-memlock.conf + grep -q 'memlock unlimited' /etc/security/limits.d/90-1bit-npu-memlock.conf check "NPU keep-awake udev rule syntax" "udevadm verify" \ udevadm verify /etc/udev/rules.d/90-1bit-npu-power.rules check "modules-load.d entry" "ec_su_axb35" \ @@ -160,14 +193,14 @@ check "public MOK certificate (kernel key)" \ "$(openssl x509 -inform DER -in /usr/share/1bit/mok/MOK-Kernel.der -noout -subject 2>/dev/null | sed 's/subject=//')" \ test -s /usr/share/1bit/mok/MOK-Kernel.der check "config files are 0644" "not group-writable" \ - bash -c '! find /etc/security/limits.d/99-amdnpu-memlock.conf /etc/udev/rules.d/90-1bit-npu-power.rules -perm /022 | grep -q .' + bash -c '! find /etc/security/limits.d/90-1bit-npu-memlock.conf /etc/udev/rules.d/90-1bit-npu-power.rules -perm /022 | grep -q .' # Is the memlock file actually READ by the login path, or only present? The NPU runtime locks command # buffers and fails without it, so a right-looking file at a path pam_limits does not read would be a # confusing user-facing failure. Proof: write a distinctive finite limit into the installed file, log in # as a fresh user, and see that value come back; then restore the shipped content and read it again. if id kpkgprobe >/dev/null 2>&1 || useradd -m -s /bin/bash kpkgprobe >/dev/null 2>&1; then - lim=/etc/security/limits.d/99-amdnpu-memlock.conf + lim=/etc/security/limits.d/90-1bit-npu-memlock.conf keep=$(mktemp); cp "$lim" "$keep" # 4096, not 8192: LOWERING a limit is always permitted, so the value can only come from the file. # (The first version of this row wrote 8192, which happened to equal the container's inherited