resources: GPUFS: ROCm 7.0 and MI350 support

Author: abmerop

src/x86-ubuntu-gpu-ml/build.sh

@@ -14,5 +14,6 @@ fi
 # Install the needed plugins
 ./packer init x86-ubuntu-gpu-ml.pkr.hcl
 
-# Build the image
-./packer build x86-ubuntu-gpu-ml.pkr.hcl
+# Build the image - Pass command line options from this script to build
+# command. This can be used to set variable such as qemu path.
+./packer build "$@" x86-ubuntu-gpu-ml.pkr.hcl

src/x86-ubuntu-gpu-ml/files/load_amdgpu.sh

@@ -24,6 +24,7 @@ insmod /lib/modules/`uname -r`/updates/dkms/amdkcl.ko.zst
 insmod /lib/modules/`uname -r`/updates/dkms/amd-sched.ko.zst
 insmod /lib/modules/`uname -r`/updates/dkms/amdxcp.ko.zst
 insmod /lib/modules/`uname -r`/updates/dkms/amddrm_buddy.ko.zst
+insmod /lib/modules/`uname -r`/updates/dkms/amddrm_exec.ko.zst
 insmod /lib/modules/`uname -r`/updates/dkms/amdttm.ko.zst
 insmod /lib/modules/`uname -r`/updates/dkms/amddrm_ttm_helper.ko.zst
 

src/x86-ubuntu-gpu-ml/files/mi350_discovery

@@ -60,10 +60,10 @@ sudo mkdir --parents --mode=0755 /etc/apt/keyrings
 wget https://repo.radeon.com/rocm/rocm.gpg.key -O - | \
         gpg --dearmor | sudo tee /etc/apt/keyrings/rocm.gpg > /dev/null
 
-echo "deb [arch=amd64 signed-by=/etc/apt/keyrings/rocm.gpg] https://repo.radeon.com/amdgpu/6.4/ubuntu noble main" \
+echo "deb [arch=amd64 signed-by=/etc/apt/keyrings/rocm.gpg] https://repo.radeon.com/amdgpu/7.0/ubuntu noble main" \
         | sudo tee /etc/apt/sources.list.d/amdgpu.list
 
-echo "deb [arch=amd64 signed-by=/etc/apt/keyrings/rocm.gpg] https://repo.radeon.com/rocm/apt/6.4 noble main" \
+echo "deb [arch=amd64 signed-by=/etc/apt/keyrings/rocm.gpg] https://repo.radeon.com/rocm/apt/7.0 noble main" \
         | sudo tee --append /etc/apt/sources.list.d/rocm.list
 echo -e 'Package: *\nPin: release o=repo.radeon.com\nPin-Priority: 600' \
         | sudo tee /etc/apt/preferences.d/rocm-pin-600
@@ -86,8 +86,8 @@ sudo chmod 777 /usr/lib/firmware/amdgpu/ip_discovery.bin
 
 
 # Install a known-working version of Linux as this might change after stable
-# release. Installl this after DKMS so they are rebuilt.
-KERNEL=6.8.0-60-generic
+# release. Install this after DKMS so they are rebuilt.
+KERNEL=6.8.0-79-generic
 
 sudo apt -y install "linux-image-${KERNEL}"
 sudo apt -y install "linux-headers-${KERNEL}" "linux-modules-extra-${KERNEL}"
@@ -105,6 +105,13 @@ if [ ! -f ./gem5_wmi.ko ]; then
 fi
 popd
 
+# Make the discovery files writeable by packer
+touch /usr/lib/firmware/amdgpu/mi300_discovery
+touch /usr/lib/firmware/amdgpu/mi350_discovery
+
+chmod 777 /usr/lib/firmware/amdgpu/mi300_discovery
+chmod 777 /usr/lib/firmware/amdgpu/mi350_discovery
+
 # Note about pip: This disk is created for the express purpose of being run in
 # gem5 and is therefore effectively sandboxed enough that we can use the pip
 # option --break-system-packages. If you plan to modify this disk image with
@@ -121,11 +128,9 @@ pip3 install --break-system-packages torch torchvision torchaudio --index-url ht
 
 # For a newer version uncomment one below and remove the above install:
 # Warning: Absurdly slow compared to ROCm 6.0 *in simulation*:
-#pip3 install --break-system-packages torch torchvision torchaudio --index-url https://download.pytorch.org/whl/rocm6.2.4
-# Warning: Missing python module torch.sparse.......:
-#pip3 install --break-system-packages torch torchvision torchaudio --index-url https://download.pytorch.org/whl/rocm6.3
+#pip3 install --break-system-packages torch torchvision torchaudio --index-url https://download.pytorch.org/whl/rocm6.4
 # Warning: nightly build, may not work depending on day. Use at your own risk:
-#pip3 install --pre torch torchvision torchaudio --index-url https://download.pytorch.org/whl/nightly/rocm6.4/ --break-system-packages
+#pip3 install --pre torch torchvision --index-url https://download.pytorch.org/whl/nightly/rocm7.0 --break-system-packages
 
 # Setup gem5 auto login.
 mv /home/gem5/serial-getty@.service /lib/systemd/system/

src/x86-ubuntu-gpu-ml/scripts/rocm-install.sh

@@ -26,6 +26,11 @@ variable "ssh_username" {
   default = "gem5"
 }
 
+variable "qemu_path" {
+  type    = string
+  default = "/usr/bin/qemu-system-x86_64"
+}
+
 source "qemu" "initialize" {
   accelerator      = "kvm"
   boot_command     = ["e<wait>",
@@ -43,7 +48,7 @@ source "qemu" "initialize" {
   iso_urls         = ["https://releases.ubuntu.com/24.04.2/ubuntu-24.04.2-live-server-amd64.iso"]
   memory           = "8192"
   output_directory = "disk-image"
-  qemu_binary      = "/usr/bin/qemu-system-x86_64"
+  qemu_binary      = "${var.qemu_path}"
   qemuargs         = [["-cpu", "host"], ["-display", "none"]]
   shutdown_command = "echo '${var.ssh_password}'|sudo -S shutdown -P now"
   ssh_password     = "${var.ssh_password}"
@@ -97,10 +102,15 @@ build {
   }
 
   provisioner "file" {
-    destination = "/usr/lib/firmware/amdgpu/ip_discovery.bin"
+    destination = "/usr/lib/firmware/amdgpu/mi300_discovery"
     source      = "files/mi300_discovery"
   }
 
+  provisioner "file" {
+    destination = "/usr/lib/firmware/amdgpu/mi350_discovery"
+    source      = "files/mi350_discovery"
+  }
+
   provisioner "file" {
     source      = "/home/gem5/vmlinux-gpu-ml"
     destination = "vmlinux-gpu-ml"