Hello,
We are developing a tool called Spinner to detect locking violations
in the bpf subsytem. Spinner reported the raw_spin_lock() in
percpu_counter_add_batch(). This function is used by
htab_percpu_map_update_elem() and htab_percpu_map_delete_elem(), both
of which can be called from NMI. A deadlock can happen if a bpf
program holding the lock is interrupted by another program in NMI that
tries to acquire the same lock. The report was generated for kernel
version 6.6-rc4, however, we believe this should still exist in the
latest kernel.
We tried to validate the report on v6.10 by running a PoC. Below is
the lockdep splat. The PoC is attached at the end.
While executing the PoC, lockdep makes another report for the
spin_lock used in htab_lock_bucket() which we believe to be a false
positive.
Thanks,
Priya
[ 113.618910] ================================
[ 113.618912] WARNING: inconsistent lock state
[ 113.618914] 6.10.0-rc7+ #35 Not tainted
[ 113.618916] --------------------------------
[ 113.618917] inconsistent {INITIAL USE} -> {IN-NMI} usage.
[ 113.618919] percpu_perf/1140 [HC1[1]:SC0[0]:HE0:SE1] takes:
[ 113.618923] ffff888113274460 (key#18){....}-{2:2}, at:
percpu_counter_add_batch+0xd4/0x180
[ 113.618939] {INITIAL USE} state was registered at:
[ 113.618941] lock_acquire+0x1be/0x560
[ 113.618946] _raw_spin_lock+0x3b/0x80
[ 113.618952] percpu_counter_add_batch+0xd4/0x180
[ 113.618955] alloc_htab_elem+0x452/0x820
[ 113.618959] __htab_percpu_map_update_elem+0x2bb/0x430
[ 113.618963] htab_percpu_map_update_elem+0x15/0x20
[ 113.618966] bpf_prog_67104ba9eab51b63_callback+0x88/0xf9
[ 113.618970] bpf_prog_18ed4d80ded39420_tp+0x82/0xa3
[ 113.618972] trace_call_bpf+0x24d/0x810
[ 113.618976] kprobe_perf_func+0x108/0x8c0
[ 113.618979] kprobe_dispatcher+0xbc/0x160
[ 113.618981] kprobe_ftrace_handler+0x2f3/0x4d0
[ 113.618985] e1000_init_module+0xe9/0xff0 [e1000]
[ 113.618990] do_nanosleep+0x1/0x470
[ 113.618993] common_nsleep+0x81/0xc0
[ 113.618997] __x64_sys_clock_nanosleep+0x2c5/0x4b0
[ 113.619000] x64_sys_call+0xf8d/0x1f20
[ 113.619004] do_syscall_64+0x8b/0x140
[ 113.619007] entry_SYSCALL_64_after_hwframe+0x76/0x7e
[ 113.619013] irq event stamp: 30488
[ 113.619014] hardirqs last enabled at (30487): [<ffffffff841ce30f>]
irqentry_exit+0x6f/0xa0
[ 113.619019] hardirqs last disabled at (30488): [<ffffffff841c9a46>]
exc_nmi+0x106/0x2a0
[ 113.619023] softirqs last enabled at (30486): [<ffffffff8124b209>]
__irq_exit_rcu+0xa9/0x120
[ 113.619027] softirqs last disabled at (30479): [<ffffffff8124b209>]
__irq_exit_rcu+0xa9/0x120
[ 113.619030]
other info that might help us debug this:
[ 113.619032] Possible unsafe locking scenario:
[ 113.619033] CPU0
[ 113.619034] ----
[ 113.619034] lock(key#18);
[ 113.619037] <Interrupt>
[ 113.619038] lock(key#18);
[ 113.619040]
*** DEADLOCK ***
[ 113.619041] no locks held by percpu_perf/1140.
[ 113.619043]
stack backtrace:
[ 113.619045] CPU: 2 PID: 1140 Comm: percpu_perf Not tainted 6.10.0-rc7+ #35
[ 113.619048] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996),
BIOS 1.13.0-1ubuntu1.1 04/01/2014
[ 113.619057] Call Trace:
[ 113.619059] <TASK>
[ 113.619062] dump_stack_lvl+0x9f/0xf0
[ 113.619067] dump_stack+0x14/0x20
[ 113.619069] print_usage_bug.part.0+0x408/0x690
[ 113.619074] lock_acquire+0x3ab/0x560
[ 113.619077] ? llist_add_batch+0xd6/0x160
[ 113.619081] ? __pfx_lock_acquire+0x10/0x10
[ 113.619084] ? percpu_counter_add_batch+0xd4/0x180
[ 113.619088] ? lock_acquire+0x48c/0x560
[ 113.619090] ? rcu_is_watching+0x17/0xd0
[ 113.619096] ? __pfx_lock_acquire+0x10/0x10
[ 113.619100] _raw_spin_lock+0x3b/0x80
[ 113.619102] ? percpu_counter_add_batch+0xd4/0x180
[ 113.619106] percpu_counter_add_batch+0xd4/0x180
[ 113.619111] alloc_htab_elem+0x452/0x820
[ 113.619117] __htab_percpu_map_update_elem+0x2bb/0x430
[ 113.619122] ? __pfx___htab_percpu_map_update_elem+0x10/0x10
[ 113.619128] htab_percpu_map_update_elem+0x15/0x20
[ 113.619132] bpf_prog_67104ba9eab51b63_callback+0x88/0xf9
[ 113.619136] bpf_prog_72e9553d3d58c8e4_percpu_perf+0x82/0xa3
[ 113.619140] __perf_event_overflow+0x232/0xe40
[ 113.619144] ? x86_perf_event_set_period+0x268/0x560
[ 113.619148] ? __pfx___perf_event_overflow+0x10/0x10
[ 113.619155] perf_event_overflow+0x1d/0x30
[ 113.619158] handle_pmi_common+0x38a/0x9f0
[ 113.619165] ? __pfx_handle_pmi_common+0x10/0x10
[ 113.619177] ? __this_cpu_preempt_check+0x17/0x20
[ 113.619182] ? __pfx_intel_bts_interrupt+0x10/0x10
[ 113.619186] ? debug_smp_processor_id+0x1b/0x30
[ 113.619190] intel_pmu_handle_irq+0x24a/0xb20
[ 113.619195] ? __this_cpu_preempt_check+0x17/0x20
[ 113.619200] perf_event_nmi_handler+0x40/0x60
[ 113.619204] nmi_handle+0x168/0x4a0
[ 113.619209] ? rcu_is_watching+0x17/0xd0
[ 113.619214] default_do_nmi+0x6e/0x180
[ 113.619218] exc_nmi+0x1cd/0x2a0
[ 113.619222] asm_exc_nmi+0xbc/0x105
[ 113.619225] RIP: 0033:0x5580d124f606
[ 113.619238] Code: ff ff ff ff 48 8b 05 59 9a 04 00 48 89 c1 ba 2a
00 00 00 be 01 00 00 00 48 8d 05 e5 1b 03 00 48 89 c7 e8 dd f6 ff ff
eb 10 90 <0f> b6 05 3c 9a 04 00 83 f0 01 84 c0 75 f2 90 48 83 bd 08 fe
ff ff
[ 113.619241] RSP: 002b:00007fff1b3d6ff0 EFLAGS: 00000202
[ 113.619244] RAX: 0000000000000001 RBX: 00005580feb35920 RCX: 0000000000000000
[ 113.619246] RDX: 0000000000000000 RSI: 00005580feb35010 RDI: 0000000000000007
[ 113.619248] RBP: 00007fff1b3d7220 R08: 00005580feb35860 R09: 00000004feb359f0
[ 113.619250] R10: 0000000000000000 R11: a43a236fd00a0e7c R12: 00007fff1b3d7338
[ 113.619252] R13: 00005580d124f084 R14: 00005580d12969d8 R15: 00007f26cec79040
[ 113.619259] </TASK>
The lockdep warning can be triggered using the following bpf and user programs.
============================================================
#include "vmlinux.h"
#include <linux/version.h>
#include <bpf/bpf_helpers.h>
#include <bpf/bpf_tracing.h>
#include <bpf/bpf_core_read.h>
struct {
__uint(type, BPF_MAP_TYPE_PERCPU_HASH);
__type(key, int);
__type(value,int);
__uint(max_entries, 2048);
__uint(map_flags, BPF_F_NO_PREALLOC);
} pb SEC(".maps");
struct callback_ctx {
int output;
};
__u32 stop_index = -1;
int i = 0;
int j = 0;
static int callback(__u32 index, void *data)
{
struct callback_ctx *ctx = data;
int value = 2;
bpf_map_update_elem(&pb, &i, &value, BPF_ANY);
i+=1;
if (index >= stop_index)
return 1;
ctx->output += index;
return 0;
}
static int callback_tp(__u32 index, void *data)
{
struct callback_ctx *ctx = data;
int value = 2;
bpf_map_update_elem(&pb, &i, &value, BPF_ANY);
i+=1;
if (index >= stop_index)
return 1;
ctx->output += index;
return 0;
}
SEC("perf_event")
int percpu_perf(void *ctx)
{
bpf_printk("perf event");
struct callback_ctx data = {};
int nr_loops = bpf_loop(10, callback, &data, 0);
return 0;
}
SEC("kprobe/do_nanosleep")
int tp(void *ctx)
{
int i;
bpf_printk("tp");
struct callback_ctx data = {};
int nr_loops = bpf_loop(10, callback, &data, 0);
return 0;
}
char _license[] SEC("license") = "GPL";
============================================================
#include <unistd.h>
#include <sys/syscall.h>
#include <linux/perf_event.h>
#include <bpf/libbpf.h>
#include <bpf/bpf.h>
#include <sys/resource.h>
#include <signal.h>
#include "percpu_perf.skel.h"
static volatile bool exiting = false;
static void sig_handler(int sig)
{
exiting = true;
return;
}
extern int parse_cpu_mask_file(const char *fcpu, bool **mask, int *mask_sz);
static long perf_event_open(struct perf_event_attr *hw_event, pid_t
pid, int cpu, int group_fd,
unsigned long flags)
{
int ret;
ret = syscall(__NR_perf_event_open, hw_event, pid, cpu,
group_fd, flags);
return ret;
}
void bump_memlock_rlimit(void)
{
struct rlimit rlim_new = {
.rlim_cur = RLIM_INFINITY,
.rlim_max = RLIM_INFINITY,
};
if (setrlimit(RLIMIT_MEMLOCK, &rlim_new)) {
fprintf(stderr, "Failed to increase RLIMIT_MEMLOCK limit!\n");
exit(1);
}
return;
}
int main(int argc, char *const argv[])
{
const char *online_cpus_file = "/sys/devices/system/cpu/online";
int cpu;
struct percpu_perf_bpf *skel = NULL;
struct perf_event_attr attr;
struct bpf_link **links = NULL;
int num_cpus, num_online_cpus;
int *pefds = NULL, pefd;
int i, err = 0;
bool *online_mask = NULL;
struct bpf_program *prog;
struct bpf_object *obj;
struct bpf_map *map;
char filename[256];
bump_memlock_rlimit();
signal(SIGINT, sig_handler);
signal(SIGTERM, sig_handler);
err = parse_cpu_mask_file(online_cpus_file, &online_mask,
&num_online_cpus);
if (err) {
fprintf(stderr, "Fail to get online CPU numbers: %d\n", err);
goto cleanup;
}
num_cpus = libbpf_num_possible_cpus();
if (num_cpus <= 0) {
fprintf(stderr, "Fail to get the number of processors\n");
err = -1;
goto cleanup;
}
snprintf(filename, sizeof(filename), ".output/percpu_perf.bpf.o");
obj = bpf_object__open_file(filename, NULL);
if (libbpf_get_error(obj)) {
fprintf(stderr, "ERROR: opening BPF object file failed\n");
goto cleanup;
}
map = bpf_object__find_map_by_name(obj, "pb");
if (libbpf_get_error(map)) {
fprintf(stderr, "ERROR: finding a map in obj file failed\n");
goto cleanup;
}
if (bpf_object__load(obj)) {
fprintf(stderr, "ERROR: loading BPF object file failed\n");
goto cleanup;
}
pefds = malloc(num_cpus * sizeof(int));
for (i = 0; i < 1; i++) {
pefds[i] = -1;
}
links = calloc(num_cpus, sizeof(struct bpf_link *));
memset(&attr, 0, sizeof(attr));
attr.type = PERF_TYPE_HARDWARE;
attr.config = PERF_COUNT_HW_CPU_CYCLES;
attr.sample_freq = 10;
attr.inherit = 1;
attr.freq = 1;
for (cpu = 0; cpu < 1; cpu++) {
//skip offline/not present CPUs
if (cpu >= num_online_cpus || !online_mask[cpu])
continue;
// Set up performance monitoring on a CPU/Core
pefd = perf_event_open(&attr, 0, -1, -1, 0);
if (pefd < 0) {
fprintf(stderr, "Fail to set up performance
monitor on a CPU/Core\n");
err = -1;
goto cleanup;
}
pefds[cpu] = pefd;
prog = bpf_object__find_program_by_name(obj, "percpu_perf");
if (!prog) {
fprintf(stderr, "ERROR: finding a prog in obj
file failed\n");
goto cleanup;
}
links[cpu] = bpf_program__attach_perf_event(prog, pefds[cpu]);
if (!links[cpu]) {
err = -1;
fprintf(stderr, "ERROR: bpf_program__attach failed\n");
goto cleanup;
}
}
struct bpf_program *prog2;
struct bpf_link *link_tp;
prog2 = bpf_object__find_program_by_name(obj, "tp");
if (!prog2) {
fprintf(stderr, "ERROR: finding a prog in obj file failed\n");
goto cleanup;
}
link_tp = bpf_program__attach(prog2);
if (!link_tp) {
err = -1;
fprintf(stderr, "ERROR: bpf_program__attach failed\n");
goto cleanup;
}
while(!exiting){
}
cleanup:
if (links) {
for (cpu = 0; cpu < num_cpus; cpu++)
bpf_link__destroy(links[cpu]);
free(links);
}
if (pefds) {
for (i = 0; i < num_cpus; i++) {
if (pefds[i] >= 0)
close(pefds[i]);
}
free(pefds);
}
percpu_perf_bpf__destroy(skel);
free(online_mask);
return -err;
}
