libbpf: auto-bump RLIMIT_MEMLOCK if kernel needs it for BPF

The need to increase RLIMIT_MEMLOCK to do anything useful with BPF is
one of the first extremely frustrating gotchas that all new BPF users go
through and in some cases have to learn it a very hard way.

Luckily, starting with upstream Linux kernel version 5.11, BPF subsystem
dropped the dependency on memlock and uses memcg-based memory accounting
instead. Unfortunately, detecting memcg-based BPF memory accounting is
far from trivial (as can be evidenced by this patch), so in practice
most BPF applications still do unconditional RLIMIT_MEMLOCK increase.

As we move towards libbpf 1.0, it would be good to allow users to forget
about RLIMIT_MEMLOCK vs memcg and let libbpf do the sensible adjustment
automatically. This patch paves the way forward in this matter. Libbpf
will do feature detection of memcg-based accounting, and if detected,
will do nothing. But if the kernel is too old, just like BCC, libbpf
will automatically increase RLIMIT_MEMLOCK on behalf of user
application ([0]).

As this is technically a breaking change, during the transition period
applications have to opt into libbpf 1.0 mode by setting
LIBBPF_STRICT_AUTO_RLIMIT_MEMLOCK bit when calling
libbpf_set_strict_mode().

Libbpf allows to control the exact amount of set RLIMIT_MEMLOCK limit
with libbpf_set_memlock_rlim_max() API. Passing 0 will make libbpf do
nothing with RLIMIT_MEMLOCK. libbpf_set_memlock_rlim_max() has to be
called before the first bpf_prog_load(), bpf_btf_load(), or
bpf_object__load() call, otherwise it has no effect and will return
-EBUSY.

  [0] Closes: libbpf/libbpf#369

Signed-off-by: Andrii Nakryiko <andrii@kernel.org>

tools/lib/bpf/bpf.c

@@ -29,6 +29,7 @@
 #include <errno.h>
 #include <linux/bpf.h>
 #include <limits.h>
+#include <sys/resource.h>
 #include "bpf.h"
 #include "libbpf.h"
 #include "libbpf_internal.h"
@@ -94,6 +95,106 @@ static inline int sys_bpf_prog_load(union bpf_attr *attr, unsigned int size, int
 	return fd;
 }
 
+/* Probe whether kernel switched from memlock-based (RLIMIT_MEMLOCK) to
+ * memcg-based memory accounting for BPF maps and progs. This was done in [0].
+ * We use the difference in reporting memlock value in BPF map's fdinfo before
+ * and after [0] to detect whether memcg accounting is done for BPF subsystem
+ * or not.
+ *
+ * Before the change, memlock value for ARRAY map would be calculated as:
+ *
+ *   memlock = sizeof(struct bpf_array) + round_up(value_size, 8) * max_entries;
+ *   memlock = round_up(memlock, PAGE_SIZE);
+ *
+ *
+ * After, memlock is approximated as:
+ *
+ *   memlock = round_up(key_size + value_size, 8) * max_entries;
+ *   memlock = round_up(memlock, PAGE_SIZE);
+ *
+ * In this check we use the fact that sizeof(struct bpf_array) is about 300
+ * bytes, so if we use value_size = (PAGE_SIZE - 100), before memcg
+ * approximation memlock would be rounded up to 2 * PAGE_SIZE, while with
+ * memcg approximation it will stay at single PAGE_SIZE (key_size is 4 for
+ * array and doesn't make much difference given 100 byte decrement we use for
+ * value_size).
+ *
+ *   [0] https://lore.kernel.org/bpf/20201201215900.3569844-1-guro@fb.com/
+ */
+int probe_memcg_account(void)
+{
+	const size_t map_create_attr_sz = offsetofend(union bpf_attr, map_extra);
+	long page_sz = sysconf(_SC_PAGESIZE), memlock_sz;
+	char buf[128];
+	union bpf_attr attr;
+	int map_fd;
+	FILE *f;
+
+	memset(&attr, 0, map_create_attr_sz);
+	attr.map_type = BPF_MAP_TYPE_ARRAY;
+	attr.key_size = 4;
+	attr.value_size = page_sz - 100;
+	attr.max_entries = 1;
+	map_fd = sys_bpf_fd(BPF_MAP_CREATE, &attr, map_create_attr_sz);
+	if (map_fd < 0)
+		return -errno;
+
+	sprintf(buf, "/proc/self/fdinfo/%d", map_fd);
+	f = fopen(buf, "r");
+	while (f && !feof(f) && fgets(buf, sizeof(buf), f)) {
+		if (fscanf(f, "memlock: %ld\n", &memlock_sz) == 1) {
+			fclose(f);
+			close(map_fd);
+			return memlock_sz == page_sz ? 1 : 0;
+		}
+	}
+
+	/* proc FS is disabled or we failed to parse fdinfo properly, assume
+	 * we need setrlimit
+	 */
+	if (f)
+		fclose(f);
+	close(map_fd);
+	return 0;
+}
+
+static bool memlock_bumped;
+static rlim_t memlock_rlim = RLIM_INFINITY;
+
+int libbpf_set_memlock_rlim(size_t memlock_bytes)
+{
+	if (memlock_bumped)
+		return libbpf_err(-EBUSY);
+
+	memlock_rlim = memlock_bytes;
+	return 0;
+}
+
+int bump_rlimit_memlock(void)
+{
+	struct rlimit rlim;
+
+	/* this the default in libbpf 1.0, but for now user has to opt-in explicitly */
+	if (!(libbpf_mode & LIBBPF_STRICT_AUTO_RLIMIT_MEMLOCK))
+		return 0;
+
+	/* if kernel supports memcg-based accounting, skip bumping RLIMIT_MEMLOCK */
+	if (memlock_bumped || kernel_supports(NULL, FEAT_MEMCG_ACCOUNT))
+		return 0;
+
+	memlock_bumped = true;
+
+	/* zero memlock_rlim_max disables auto-bumping RLIMIT_MEMLOCK */
+	if (memlock_rlim == 0)
+		return 0;
+
+	rlim.rlim_cur = rlim.rlim_max = memlock_rlim;
+	if (setrlimit(RLIMIT_MEMLOCK, &rlim))
+		return -errno;
+
+	return 0;
+}
+
 int bpf_map_create(enum bpf_map_type map_type,
 		   const char *map_name,
 		   __u32 key_size,
@@ -105,6 +206,8 @@ int bpf_map_create(enum bpf_map_type map_type,
 	union bpf_attr attr;
 	int fd;
 
+	bump_rlimit_memlock();
+
 	memset(&attr, 0, attr_sz);
 
 	if (!OPTS_VALID(opts, bpf_map_create_opts))
@@ -251,6 +354,8 @@ int bpf_prog_load_v0_6_0(enum bpf_prog_type prog_type,
 	union bpf_attr attr;
 	char *log_buf;
 
+	bump_rlimit_memlock();
+
 	if (!OPTS_VALID(opts, bpf_prog_load_opts))
 		return libbpf_err(-EINVAL);
 
@@ -456,6 +561,8 @@ int bpf_verify_program(enum bpf_prog_type type, const struct bpf_insn *insns,
 	union bpf_attr attr;
 	int fd;
 
+	bump_rlimit_memlock();
+
 	memset(&attr, 0, sizeof(attr));
 	attr.prog_type = type;
 	attr.insn_cnt = (__u32)insns_cnt;
@@ -1056,6 +1163,8 @@ int bpf_btf_load(const void *btf_data, size_t btf_size, const struct bpf_btf_loa
 	__u32 log_level;
 	int fd;
 
+	bump_rlimit_memlock();
+
 	memset(&attr, 0, attr_sz);
 
 	if (!OPTS_VALID(opts, bpf_btf_load_opts))

tools/lib/bpf/bpf.h

@@ -35,6 +35,8 @@
 extern "C" {
 #endif
 
+int libbpf_set_memlock_rlim(size_t memlock_bytes);
+
 struct bpf_map_create_opts {
 	size_t sz; /* size of this struct for forward/backward compatibility */
 

tools/lib/bpf/libbpf.c

@@ -187,42 +187,6 @@ const char *libbpf_version_string(void)
 #undef __S
 }
 
-enum kern_feature_id {
-	/* v4.14: kernel support for program & map names. */
-	FEAT_PROG_NAME,
-	/* v5.2: kernel support for global data sections. */
-	FEAT_GLOBAL_DATA,
-	/* BTF support */
-	FEAT_BTF,
-	/* BTF_KIND_FUNC and BTF_KIND_FUNC_PROTO support */
-	FEAT_BTF_FUNC,
-	/* BTF_KIND_VAR and BTF_KIND_DATASEC support */
-	FEAT_BTF_DATASEC,
-	/* BTF_FUNC_GLOBAL is supported */
-	FEAT_BTF_GLOBAL_FUNC,
-	/* BPF_F_MMAPABLE is supported for arrays */
-	FEAT_ARRAY_MMAP,
-	/* kernel support for expected_attach_type in BPF_PROG_LOAD */
-	FEAT_EXP_ATTACH_TYPE,
-	/* bpf_probe_read_{kernel,user}[_str] helpers */
-	FEAT_PROBE_READ_KERN,
-	/* BPF_PROG_BIND_MAP is supported */
-	FEAT_PROG_BIND_MAP,
-	/* Kernel support for module BTFs */
-	FEAT_MODULE_BTF,
-	/* BTF_KIND_FLOAT support */
-	FEAT_BTF_FLOAT,
-	/* BPF perf link support */
-	FEAT_PERF_LINK,
-	/* BTF_KIND_DECL_TAG support */
-	FEAT_BTF_DECL_TAG,
-	/* BTF_KIND_TYPE_TAG support */
-	FEAT_BTF_TYPE_TAG,
-	__FEAT_CNT,
-};
-
-static bool kernel_supports(const struct bpf_object *obj, enum kern_feature_id feat_id);
-
 enum reloc_type {
 	RELO_LD64,
 	RELO_CALL,
@@ -4354,6 +4318,10 @@ bpf_object__probe_loading(struct bpf_object *obj)
 	if (obj->gen_loader)
 		return 0;
 
+	ret = bump_rlimit_memlock();
+	if (ret)
+		pr_warn("Failed to bump RLIMIT_MEMLOCK (err = %d), you might need to do it explicitly!\n", ret);
+
 	/* make sure basic loading works */
 	ret = bpf_prog_load(BPF_PROG_TYPE_SOCKET_FILTER, NULL, "GPL", insns, insn_cnt, NULL);
 	if (ret < 0)
@@ -4720,14 +4688,17 @@ static struct kern_feature_desc {
 	[FEAT_BTF_TYPE_TAG] = {
 		"BTF_KIND_TYPE_TAG support", probe_kern_btf_type_tag,
 	},
+	[FEAT_MEMCG_ACCOUNT] = {
+		"memcg-based memory accounting", probe_memcg_account,
+	},
 };
 
-static bool kernel_supports(const struct bpf_object *obj, enum kern_feature_id feat_id)
+bool kernel_supports(const struct bpf_object *obj, enum kern_feature_id feat_id)
 {
 	struct kern_feature_desc *feat = &feature_probes[feat_id];
 	int ret;
 
-	if (obj->gen_loader)
+	if (obj && obj->gen_loader)
 		/* To generate loader program assume the latest kernel
 		 * to avoid doing extra prog_load, map_create syscalls.
 		 */

tools/lib/bpf/libbpf.map

@@ -427,4 +427,5 @@ LIBBPF_0.7.0 {
 		bpf_program__log_level;
 		bpf_program__set_log_buf;
 		bpf_program__set_log_level;
+		libbpf_set_memlock_rlim_max;
 };

tools/lib/bpf/libbpf_internal.h

@@ -272,6 +272,45 @@ static inline bool libbpf_validate_opts(const char *opts,
 					(opts)->sz - __off);		      \
 })
 
+enum kern_feature_id {
+	/* v4.14: kernel support for program & map names. */
+	FEAT_PROG_NAME,
+	/* v5.2: kernel support for global data sections. */
+	FEAT_GLOBAL_DATA,
+	/* BTF support */
+	FEAT_BTF,
+	/* BTF_KIND_FUNC and BTF_KIND_FUNC_PROTO support */
+	FEAT_BTF_FUNC,
+	/* BTF_KIND_VAR and BTF_KIND_DATASEC support */
+	FEAT_BTF_DATASEC,
+	/* BTF_FUNC_GLOBAL is supported */
+	FEAT_BTF_GLOBAL_FUNC,
+	/* BPF_F_MMAPABLE is supported for arrays */
+	FEAT_ARRAY_MMAP,
+	/* kernel support for expected_attach_type in BPF_PROG_LOAD */
+	FEAT_EXP_ATTACH_TYPE,
+	/* bpf_probe_read_{kernel,user}[_str] helpers */
+	FEAT_PROBE_READ_KERN,
+	/* BPF_PROG_BIND_MAP is supported */
+	FEAT_PROG_BIND_MAP,
+	/* Kernel support for module BTFs */
+	FEAT_MODULE_BTF,
+	/* BTF_KIND_FLOAT support */
+	FEAT_BTF_FLOAT,
+	/* BPF perf link support */
+	FEAT_PERF_LINK,
+	/* BTF_KIND_DECL_TAG support */
+	FEAT_BTF_DECL_TAG,
+	/* BTF_KIND_TYPE_TAG support */
+	FEAT_BTF_TYPE_TAG,
+	/* memcg-based accounting for BPF maps and progs */
+	FEAT_MEMCG_ACCOUNT,
+	__FEAT_CNT,
+};
+
+int probe_memcg_account(void);
+bool kernel_supports(const struct bpf_object *obj, enum kern_feature_id feat_id);
+int bump_rlimit_memlock(void);
 
 int parse_cpu_mask_str(const char *s, bool **mask, int *mask_sz);
 int parse_cpu_mask_file(const char *fcpu, bool **mask, int *mask_sz);

tools/lib/bpf/libbpf_legacy.h

@@ -45,7 +45,6 @@ enum libbpf_strict_mode {
 	 * (positive) error code.
 	 */
 	LIBBPF_STRICT_DIRECT_ERRS = 0x02,
-
 	/*
 	 * Enforce strict BPF program section (SEC()) names.
 	 * E.g., while prefiously SEC("xdp_whatever") or SEC("perf_event_blah") were
@@ -63,6 +62,17 @@ enum libbpf_strict_mode {
 	 * Clients can maintain it on their own if it is valuable for them.
 	 */
 	LIBBPF_STRICT_NO_OBJECT_LIST = 0x08,
+	/*
+	 * Automatically bump RLIMIT_MEMLOCK using setrlimit() before the
+	 * first BPF program or map creation operation. This is done only if
+	 * kernel is too old to support memcg-based memory accounting for BPF
+	 * subsystem. By default, RLIMIT_MEMLOCK limit is set to RLIM_INFINITY,
+	 * but it can be overriden with libbpf_set_memlock_rlim_max() API.
+	 * Note that libbpf_set_memlock_rlim_max() needs to be called before
+	 * the very first bpf_prog_load(), bpf_map_create() or bpf_object__load()
+	 * operation.
+	 */
+	LIBBPF_STRICT_AUTO_RLIMIT_MEMLOCK = 0x10,
 
 	__LIBBPF_STRICT_LAST,
 };