Add kernel infrastructure to read RDT and assign RMIDs

.github/workflows/test-kernel-module.yaml

@@ -1,5 +1,17 @@
 name: Test Kernel Module
-on: workflow_dispatch  # Manual trigger for testing
+on: 
+  workflow_dispatch:  # Manual trigger for testing
+    inputs:
+      instance-type:
+        description: 'EC2 instance type to use'
+        required: false
+        default: 'm7i.metal-24xl'
+        type: string
+  push:
+    branches:
+      - main
+    paths:
+      - module/**
 
 permissions:
   id-token: write # Required for requesting the JWT
@@ -26,7 +38,7 @@ jobs:
           mode: start
           github-token: ${{ secrets.REPO_ADMIN_TOKEN }}
           ec2-image-id: ami-0884d2865dbe9de4b  # Ubuntu 22.04 LTS in us-east-2
-          ec2-instance-type: t3.large
+          ec2-instance-type: ${{ inputs.instance-type || 'm7i.metal-24xl' }}
           market-type: spot
           subnet-id: ${{ secrets.AWS_SUBNET_ID }}
           security-group-id: ${{ secrets.AWS_SECURITY_GROUP_ID }}
@@ -45,7 +57,7 @@ jobs:
   test-module:
     needs: start-runner
     runs-on: ${{ needs.start-runner.outputs.label }}
-    timeout-minutes: 10  # Add timeout in case system hangs
+    timeout-minutes: 2  # Add timeout in case system hangs
     steps:
       - name: Checkout code
         uses: actions/checkout@v4
@@ -105,31 +117,95 @@ jobs:
           # Now try the actual build
           make
 
-          ls -l build/memory_collector.ko
+          ls -l build/collector.ko
 
+      - name: Check RDT Capabilities
+        run: |
+          sudo mkdir -p /sys/fs/resctrl || true
+          sudo mount -t resctrl resctrl /sys/fs/resctrl || true
+
+          echo "Mounting resctrl filesystem"
+          mount | grep resctrl || true
+
+          echo "Checking RDT capabilities"
+          ls /sys/fs/resctrl/info || true
+
+          echo "Monitoring features:"
+          cat /sys/fs/resctrl/info/L3_MON/mon_features || true
+
+          echo "Number of available RMIDs:"
+          cat /sys/fs/resctrl/info/L3_MON/num_rmids || true
+
+          echo "Number of CAT classes:"
+          cat /sys/fs/resctrl/info/L3/num_closids || true
+
+          echo "head -n 35 /proc/cpuinfo:"
+          head -n 35 /proc/cpuinfo || true
+
+          echo "CPU RDT features (head):"
+          grep -E "cat_l3|cdp_l3|cqm_occup_llc|cqm_mbm_total|cqm_mbm_local" /proc/cpuinfo | head || true
+
+          # we do not unmount, maybe mounting affects the intel_cqm checks below
+          #sudo umount /sys/fs/resctrl || true           
+
       - name: Load and test module
+        id: load-and-test-module
+        continue-on-error: true
         working-directory: module
         run: |
+
+          # Check undefined symbols
+          sudo modinfo -F depends build/collector.ko
+          sudo objdump -d build/collector.ko | grep undefined || true
+
           # Load module
-          sudo insmod build/memory_collector.ko
+          echo "insmod build/collector.ko:"
+          sudo insmod build/collector.ko
 
           # Verify module is loaded
-          lsmod | grep memory_collector
+          echo "lsmod | grep collector:"
+          lsmod | grep collector
 
           # Check kernel logs for module initialization
-          dmesg | grep "Memory Collector" || true
+          echo "dmesg | grep 'Memory Collector':"
+          dmesg -c | grep "Memory Collector" || true
 
           # Unload module
-          sudo rmmod memory_collector
+          echo "rmmod collector:"
+          sudo rmmod collector
 
           # Verify module unloaded successfully
-          if lsmod | grep -q memory_collector; then
+          echo "lsmod | grep collector:"
+          ! lsmod | grep collector
+          if lsmod | grep -q collector; then
             echo "Error: Module still loaded"
             exit 1
           fi
 
           # Check kernel logs for cleanup message
-          dmesg | grep "Memory Collector" || true
+          echo "dmesg | grep 'Memory Collector':"
+          dmesg -c | grep "Memory Collector" || true
+
+      - name: Check dmesg on failure
+        if: steps.load-and-test-module.outcome == 'failure'
+        run: |
+          echo "load and test module failed, showing last kernel messages:"
+          sudo dmesg | tail -n 100
+          exit 1
+
+      - name: Install trace dependencies
+        run: |
+          sudo apt-get install -y trace-cmd
+
+      - name: Run module test script
+        working-directory: module
+        run: |
+          # run 10 times in quick succession to stress-test insmod/rmmod and collector
+          for i in {1..10}; do
+            echo "*** Run $i:"
+            ./test_module.sh
+          done
+
 
   stop-runner:
     name: Stop EC2 runner

Dockerfile.devcontainer

@@ -3,11 +3,27 @@ FROM ubuntu:latest
 # Avoid prompts during package installation
 ENV DEBIAN_FRONTEND=noninteractive
 
+# Add amd64 architecture
+RUN dpkg --add-architecture amd64
+
+# Set up repositories for both arm64 and amd64
+RUN echo "deb [arch=arm64] http://ports.ubuntu.com/ubuntu-ports noble main restricted universe multiverse\n\
+deb [arch=arm64] http://ports.ubuntu.com/ubuntu-ports noble-updates main restricted universe multiverse\n\
+deb [arch=arm64] http://ports.ubuntu.com/ubuntu-ports noble-backports main restricted universe multiverse\n\
+deb [arch=arm64] http://ports.ubuntu.com/ubuntu-ports noble-security main restricted universe multiverse\n\
+deb [arch=amd64] http://archive.ubuntu.com/ubuntu noble main restricted universe multiverse\n\
+deb [arch=amd64] http://archive.ubuntu.com/ubuntu noble-updates main restricted universe multiverse\n\
+deb [arch=amd64] http://archive.ubuntu.com/ubuntu noble-backports main restricted universe multiverse\n\
+deb [arch=amd64] http://security.ubuntu.com/ubuntu noble-security main restricted universe multiverse" > /etc/apt/sources.list
+
+RUN rm /etc/apt/sources.list.d/ubuntu.sources
+
 # Update and install essential build tools and kernel headers
 RUN apt-get update && apt-get install -y \
     build-essential \
-    linux-headers-6.8.0-52-generic \
-    linux-image-6.8.0-52-generic \
+    crossbuild-essential-amd64 \
+    linux-headers-6.8.0-52-generic:amd64 \
+    linux-image-6.8.0-52-generic:amd64 \
     git \
     vim \
     curl \
@@ -18,4 +34,4 @@ RUN apt-get update && apt-get install -y \
 WORKDIR /workspace
 
 # Keep container running
-CMD ["sleep", "infinity"] 
+CMD ["sleep", "infinity"]

module/.gitignore

@@ -0,0 +1 @@
+build/

module/Makefile

@@ -1,4 +1,11 @@
-obj-m += memory_collector.o
+EXTRA_CFLAGS += -I$(src)
+
+# Define the module name and its source files
+obj-m := collector.o
+collector-objs := memory_collector.o resctrl.o
+
+# Always set architecture to x86_64 (not just x86)
+ARCH := x86_64
 
 # Check if KVERSION is provided on command line
 ifdef KVERSION
@@ -19,15 +26,32 @@ endif
 KDIR := /lib/modules/$(KERNEL_VERSION)/build
 BUILD_DIR := $(PWD)/build
 
+# Set cross-compilation by default in container
+ifneq (,$(wildcard /.dockerenv))
+    CROSS_COMPILE := x86_64-linux-gnu-
+endif
+
+# Also set cross-compilation on ARM machines
+ifeq ($(shell uname -m),arm64)
+    CROSS_COMPILE := x86_64-linux-gnu-
+endif
+
 all: | $(BUILD_DIR)
 	@echo "Building for kernel version: $(KERNEL_VERSION)"
-	$(MAKE) -C $(KDIR) M=$(BUILD_DIR) src=$(PWD) modules
+	$(MAKE) -C $(KDIR) M=$(BUILD_DIR) src=$(PWD) \
+		ARCH=$(ARCH) \
+		CROSS_COMPILE=$(CROSS_COMPILE) \
+		modules
 
 $(BUILD_DIR):
 	mkdir -p $(BUILD_DIR)
 
 clean:
-	$(MAKE) -C $(KDIR) M=$(BUILD_DIR) src=$(PWD) clean
+	$(MAKE) -C $(KDIR) M=$(BUILD_DIR) src=$(PWD) \
+		ARCH=$(ARCH) \
+		CROSS_COMPILE=$(CROSS_COMPILE) \
+		clean
 	rm -rf $(BUILD_DIR)
+	rm -f Module.symvers Module.markers modules.order
 
 .PHONY: all clean