bpf: use raw_spin_trylock() for pcpu_freelist_push/pop in NMI

.travis.yml

@@ -0,0 +1,33 @@
+sudo: required
+language: bash
+dist: bionic
+services:
+    - docker
+
+env:
+    global:
+        - PROJECT_NAME='libbpf'
+        - AUTHOR_EMAIL="$(git log -1 --pretty=\"%aE\")"
+        - REPO_ROOT="$TRAVIS_BUILD_DIR"
+        - CI_ROOT="$REPO_ROOT/travis-ci"
+        - VMTEST_ROOT="$CI_ROOT/vmtest"
+
+addons:
+    apt:
+        packages:
+            - qemu-kvm
+            - zstd
+            - binutils-dev
+            - elfutils
+            - libcap-dev
+            - libelf-dev
+            - libdw-dev
+            - python3-docutils
+
+jobs:
+    include:
+        - stage: Builds & Tests
+          name: Kernel LATEST + selftests
+          language: bash
+          env: KERNEL=LATEST
+          script:  $CI_ROOT/vmtest/run_vmtest.sh || travis_terminate 1

kernel/bpf/percpu_freelist.c

@@ -17,6 +17,8 @@ int pcpu_freelist_init(struct pcpu_freelist *s)
 		raw_spin_lock_init(&head->lock);
 		head->first = NULL;
 	}
+	raw_spin_lock_init(&s->extralist.lock);
+	s->extralist.first = NULL;
 	return 0;
 }
 
@@ -40,12 +42,50 @@ static inline void ___pcpu_freelist_push(struct pcpu_freelist_head *head,
 	raw_spin_unlock(&head->lock);
 }
 
+static inline bool pcpu_freelist_try_push_extra(struct pcpu_freelist *s,
+						struct pcpu_freelist_node *node)
+{
+	if (!raw_spin_trylock(&s->extralist.lock))
+		return false;
+
+	pcpu_freelist_push_node(&s->extralist, node);
+	raw_spin_unlock(&s->extralist.lock);
+	return true;
+}
+
+static inline void ___pcpu_freelist_push_nmi(struct pcpu_freelist *s,
+					     struct pcpu_freelist_node *node)
+{
+	int cpu, orig_cpu;
+
+	orig_cpu = cpu = raw_smp_processor_id();
+	while (1) {
+		struct pcpu_freelist_head *head;
+
+		head = per_cpu_ptr(s->freelist, cpu);
+		if (raw_spin_trylock(&head->lock)) {
+			pcpu_freelist_push_node(head, node);
+			raw_spin_unlock(&head->lock);
+			return;
+		}
+		cpu = cpumask_next(cpu, cpu_possible_mask);
+		if (cpu >= nr_cpu_ids)
+			cpu = 0;
+
+		/* cannot lock any per cpu lock, try extralist */
+		if (cpu == orig_cpu &&
+		    pcpu_freelist_try_push_extra(s, node))
+			return;
+	}
+}
+
 void __pcpu_freelist_push(struct pcpu_freelist *s,
 			struct pcpu_freelist_node *node)
 {
-	struct pcpu_freelist_head *head = this_cpu_ptr(s->freelist);
-
-	___pcpu_freelist_push(head, node);
+	if (in_nmi())
+		___pcpu_freelist_push_nmi(s, node);
+	else
+		___pcpu_freelist_push(this_cpu_ptr(s->freelist), node);
 }
 
 void pcpu_freelist_push(struct pcpu_freelist *s,
@@ -81,7 +121,7 @@ void pcpu_freelist_populate(struct pcpu_freelist *s, void *buf, u32 elem_size,
 	}
 }
 
-struct pcpu_freelist_node *__pcpu_freelist_pop(struct pcpu_freelist *s)
+static struct pcpu_freelist_node *___pcpu_freelist_pop(struct pcpu_freelist *s)
 {
 	struct pcpu_freelist_head *head;
 	struct pcpu_freelist_node *node;
@@ -102,8 +142,59 @@ struct pcpu_freelist_node *__pcpu_freelist_pop(struct pcpu_freelist *s)
 		if (cpu >= nr_cpu_ids)
 			cpu = 0;
 		if (cpu == orig_cpu)
-			return NULL;
+			break;
+	}
+
+	/* per cpu lists are all empty, try extralist */
+	raw_spin_lock(&s->extralist.lock);
+	node = s->extralist.first;
+	if (node)
+		s->extralist.first = node->next;
+	raw_spin_unlock(&s->extralist.lock);
+	return node;
+}
+
+static struct pcpu_freelist_node *
+___pcpu_freelist_pop_nmi(struct pcpu_freelist *s)
+{
+	struct pcpu_freelist_head *head;
+	struct pcpu_freelist_node *node;
+	int orig_cpu, cpu;
+
+	orig_cpu = cpu = raw_smp_processor_id();
+	while (1) {
+		head = per_cpu_ptr(s->freelist, cpu);
+		if (raw_spin_trylock(&head->lock)) {
+			node = head->first;
+			if (node) {
+				head->first = node->next;
+				raw_spin_unlock(&head->lock);
+				return node;
+			}
+			raw_spin_unlock(&head->lock);
+		}
+		cpu = cpumask_next(cpu, cpu_possible_mask);
+		if (cpu >= nr_cpu_ids)
+			cpu = 0;
+		if (cpu == orig_cpu)
+			break;
 	}
+
+	/* cannot pop from per cpu lists, try extralist */
+	if (!raw_spin_trylock(&s->extralist.lock))
+		return NULL;
+	node = s->extralist.first;
+	if (node)
+		s->extralist.first = node->next;
+	raw_spin_unlock(&s->extralist.lock);
+	return node;
+}
+
+struct pcpu_freelist_node *__pcpu_freelist_pop(struct pcpu_freelist *s)
+{
+	if (in_nmi())
+		return ___pcpu_freelist_pop_nmi(s);
+	return ___pcpu_freelist_pop(s);
 }
 
 struct pcpu_freelist_node *pcpu_freelist_pop(struct pcpu_freelist *s)

kernel/bpf/percpu_freelist.h

@@ -13,6 +13,7 @@ struct pcpu_freelist_head {
 
 struct pcpu_freelist {
 	struct pcpu_freelist_head __percpu *freelist;
+	struct pcpu_freelist_head extralist;
 };
 
 struct pcpu_freelist_node {

travis-ci/managers/debian.sh

@@ -0,0 +1,84 @@
+#!/bin/bash
+
+PHASES=(${@:-SETUP RUN RUN_ASAN CLEANUP})
+DEBIAN_RELEASE="${DEBIAN_RELEASE:-testing}"
+CONT_NAME="${CONT_NAME:-libbpf-debian-$DEBIAN_RELEASE}"
+ENV_VARS="${ENV_VARS:-}"
+DOCKER_RUN="${DOCKER_RUN:-docker run}"
+REPO_ROOT="${REPO_ROOT:-$PWD}"
+ADDITIONAL_DEPS=(clang pkg-config gcc-8)
+CFLAGS="-g -O2 -Werror -Wall"
+
+function info() {
+    echo -e "\033[33;1m$1\033[0m"
+}
+
+function error() {
+    echo -e "\033[31;1m$1\033[0m"
+}
+
+function docker_exec() {
+    docker exec $ENV_VARS -it $CONT_NAME "$@"
+}
+
+set -e
+
+source "$(dirname $0)/travis_wait.bash"
+
+for phase in "${PHASES[@]}"; do
+    case $phase in
+        SETUP)
+            info "Setup phase"
+            info "Using Debian $DEBIAN_RELEASE"
+
+            sudo apt-get -y -o Dpkg::Options::="--force-confnew" install docker-ce
+            docker --version
+
+            docker pull debian:$DEBIAN_RELEASE
+            info "Starting container $CONT_NAME"
+            $DOCKER_RUN -v $REPO_ROOT:/build:rw \
+                        -w /build --privileged=true --name $CONT_NAME \
+                        -dit --net=host debian:$DEBIAN_RELEASE /bin/bash
+            docker_exec bash -c "echo deb-src http://deb.debian.org/debian $DEBIAN_RELEASE main >>/etc/apt/sources.list"
+            docker_exec apt-get -y update
+            docker_exec apt-get -y build-dep libelf-dev
+            docker_exec apt-get -y install libelf-dev
+            docker_exec apt-get -y install "${ADDITIONAL_DEPS[@]}"
+            ;;
+        RUN|RUN_CLANG|RUN_GCC8|RUN_ASAN|RUN_CLANG_ASAN|RUN_GCC8_ASAN)
+            if [[ "$phase" = *"CLANG"* ]]; then
+                ENV_VARS="-e CC=clang -e CXX=clang++"
+                CC="clang"
+            elif [[ "$phase" = *"GCC8"* ]]; then
+                ENV_VARS="-e CC=gcc-8 -e CXX=g++-8"
+                CC="gcc-8"
+            else
+                CFLAGS="${CFLAGS} -Wno-stringop-truncation"
+            fi
+            if [[ "$phase" = *"ASAN"* ]]; then
+                CFLAGS="${CFLAGS} -fsanitize=address,undefined"
+            fi
+            docker_exec mkdir build install
+            docker_exec ${CC:-cc} --version
+            info "build"
+	    docker_exec make -j$((4*$(nproc))) CFLAGS="${CFLAGS}" -C ./src -B OBJDIR=../build
+            info "ldd build/libbpf.so:"
+            docker_exec ldd build/libbpf.so
+            if ! docker_exec ldd build/libbpf.so | grep -q libelf; then
+                error "No reference to libelf.so in libbpf.so!"
+                exit 1
+            fi
+            info "install"
+            docker_exec make -j$((4*$(nproc))) -C src OBJDIR=../build DESTDIR=../install install
+            docker_exec rm -rf build install
+            ;;
+        CLEANUP)
+            info "Cleanup phase"
+            docker stop $CONT_NAME
+            docker rm -f $CONT_NAME
+            ;;
+        *)
+            echo >&2 "Unknown phase '$phase'"
+            exit 1
+    esac
+done

travis-ci/managers/travis_wait.bash

@@ -0,0 +1,61 @@
+# This was borrowed from https://github.com/travis-ci/travis-build/tree/master/lib/travis/build/bash
+# to get around https://github.com/travis-ci/travis-ci/issues/9979. It should probably be removed
+# as soon as Travis CI has started to provide an easy way to export the functions to bash scripts.
+
+travis_jigger() {
+  local cmd_pid="${1}"
+  shift
+  local timeout="${1}"
+  shift
+  local count=0
+
+  echo -e "\\n"
+
+  while [[ "${count}" -lt "${timeout}" ]]; do
+    count="$((count + 1))"
+    echo -ne "Still running (${count} of ${timeout}): ${*}\\r"
+    sleep 60
+  done
+
+  echo -e "\\n${ANSI_RED}Timeout (${timeout} minutes) reached. Terminating \"${*}\"${ANSI_RESET}\\n"
+  kill -9 "${cmd_pid}"
+}
+
+travis_wait() {
+  local timeout="${1}"
+
+  if [[ "${timeout}" =~ ^[0-9]+$ ]]; then
+    shift
+  else
+    timeout=20
+  fi
+
+  local cmd=("${@}")
+  local log_file="travis_wait_${$}.log"
+
+  "${cmd[@]}" &>"${log_file}" &
+  local cmd_pid="${!}"
+
+  travis_jigger "${!}" "${timeout}" "${cmd[@]}" &
+  local jigger_pid="${!}"
+  local result
+
+  {
+    set +e
+    wait "${cmd_pid}" 2>/dev/null
+    result="${?}"
+    ps -p"${jigger_pid}" &>/dev/null && kill "${jigger_pid}"
+    set -e
+  }
+
+  if [[ "${result}" -eq 0 ]]; then
+    echo -e "\\n${ANSI_GREEN}The command ${cmd[*]} exited with ${result}.${ANSI_RESET}"
+  else
+    echo -e "\\n${ANSI_RED}The command ${cmd[*]} exited with ${result}.${ANSI_RESET}"
+  fi
+
+  echo -e "\\n${ANSI_GREEN}Log:${ANSI_RESET}\\n"
+  cat "${log_file}"
+
+  return "${result}"
+}

travis-ci/managers/ubuntu.sh

@@ -0,0 +1,27 @@
+#!/bin/bash
+set -e
+set -x
+
+RELEASE="bionic"
+
+echo "deb-src http://archive.ubuntu.com/ubuntu/ $RELEASE main restricted universe multiverse" >>/etc/apt/sources.list
+
+apt-get update
+apt-get -y build-dep libelf-dev
+apt-get install -y libelf-dev pkg-config
+
+source "$(dirname $0)/travis_wait.bash"
+
+cd $REPO_ROOT
+
+CFLAGS="-g -O2 -Werror -Wall -fsanitize=address,undefined"
+mkdir build install
+cc --version
+make -j$((4*$(nproc))) CFLAGS="${CFLAGS}" -C ./src -B OBJDIR=../build
+ldd build/libbpf.so
+if ! ldd build/libbpf.so | grep -q libelf; then
+    echo "FAIL: No reference to libelf.so in libbpf.so!"
+    exit 1
+fi
+make -j$((4*$(nproc))) -C src OBJDIR=../build DESTDIR=../install install
+rm -rf build install

travis-ci/vmtest/build_pahole.sh

@@ -0,0 +1,30 @@
+#!/bin/bash
+
+set -eu
+
+source $(cd $(dirname $0) && pwd)/helpers.sh
+
+travis_fold start build_pahole "Building pahole"
+
+CWD=$(pwd)
+REPO_PATH=$1
+PAHOLE_ORIGIN=https://git.kernel.org/pub/scm/devel/pahole/pahole.git
+
+mkdir -p ${REPO_PATH}
+cd ${REPO_PATH}
+git init
+git remote add origin ${PAHOLE_ORIGIN}
+git fetch origin
+git checkout master
+
+mkdir -p build
+cd build
+cmake -DCMAKE_BUILD_TYPE=RelWithDebInfo -D__LIB=lib ..
+make -j$((4*$(nproc))) all
+sudo make install
+
+export LD_LIBRARY_PATH=${LD_LIBRARY_PATH:-}:/usr/local/lib
+ldd $(which pahole)
+pahole --version
+
+travis_fold end build_pahole