Merge branch 'iovisor:master' into yf_dev

debian/changelog

@@ -1,3 +1,15 @@
+bcc (0.32.0-1) unstable; urgency=low
+
+  * Support for kernel up to 6.11.
+  * bcc tool update: wakeuptime, readahead, shmsnoop, offcputime, cachestat, cachetop, hardirqs
+  * libbpf tool update: futexctn, profile, readhead, softirqs, hardirqs
+  * Multiple enhancements for memleak, better error path checking, adding mremap uprobe
+  * Support get pid/tgid in pid namespaces (cpudist, profile)
+  * multiple pid filtering support: profile, offcputime
+  * detect whether elf binary is PIE even if the binary is marked as DYN
+  * Fix several compilation issues with llvm20
+  * doc update, other bug fixes and tools improvement.
+
 bcc (0.31.0-1) unstable; urgency=low
 
   * Support for kernel up to 6.9.

examples/cpp/KModRetExample.cc

@@ -20,7 +20,6 @@
 #include <iomanip>
 #include <string>
 
-#include <error.h>
 #include <sys/types.h>
 #include <sys/stat.h>
 #include <fcntl.h>

examples/networking/sockmap.py

@@ -39,8 +39,9 @@
 BPF_SOCKHASH(sock_hash, struct sock_key, MAX_SOCK_OPS_MAP_ENTRIES);
 
 static __always_inline void bpf_sock_ops_ipv4(struct bpf_sock_ops *skops) {
+    volatile __u32 remote_ip4 = skops->remote_ip4;
     struct sock_key skk = {
-        .remote_ip4 = skops->remote_ip4,
+        .remote_ip4 = remote_ip4,
         .local_ip4  = skops->local_ip4,
         .local_port = skops->local_port,
         .remote_port  = bpf_ntohl(skops->remote_port),

libbpf-tools/.gitignore

@@ -71,3 +71,5 @@
 /wakeuptime
 /xfsdist
 /xfsslower
+/zfsdist
+/zfsslower

libbpf-tools/Makefile

@@ -11,6 +11,7 @@ LIBBLAZESYM_SRC := $(abspath blazesym/target/release/libblazesym.a)
 INCLUDES := -I$(OUTPUT) -I../src/cc/libbpf/include/uapi
 CFLAGS := -g -O2 -Wall -Wmissing-field-initializers -Werror -Werror=undef
 BPFCFLAGS := -g -O2 -Wall -Werror=undef
+BPFCFLAGS_softirqs := $(BPFCFLAGS) -mcpu=v3
 INSTALL ?= install
 prefix ?= /usr/local
 bindir := $(prefix)/bin
@@ -100,8 +101,8 @@ APPS = \
 # export variables that are used in Makefile.btfgen as well.
 export OUTPUT BPFTOOL ARCH BTFHUB_ARCHIVE APPS
 
-FSDIST_ALIASES = btrfsdist ext4dist nfsdist xfsdist f2fsdist bcachefsdist
-FSSLOWER_ALIASES = btrfsslower ext4slower nfsslower xfsslower f2fsslower bcachefsslower
+FSDIST_ALIASES = btrfsdist ext4dist nfsdist xfsdist f2fsdist bcachefsdist zfsdist
+FSSLOWER_ALIASES = btrfsslower ext4slower nfsslower xfsslower f2fsslower bcachefsslower zfsslower
 SIGSNOOP_ALIAS = killsnoop
 APP_ALIASES = $(FSDIST_ALIASES) $(FSSLOWER_ALIASES) ${SIGSNOOP_ALIAS}
 
@@ -200,6 +201,8 @@ $(OUTPUT)/%.skel.h: $(OUTPUT)/%.bpf.o | $(OUTPUT) $(BPFTOOL)
 	$(call msg,GEN-SKEL,$@)
 	$(Q)$(BPFTOOL) gen skeleton $< > $@
 
+$(OUTPUT)/softirqs.bpf.o: BPFCFLAGS = $(BPFCFLAGS_softirqs)
+
 $(OUTPUT)/%.bpf.o: %.bpf.c $(LIBBPF_OBJ) $(wildcard %.h) $(ARCH)/vmlinux.h | $(OUTPUT)
 	$(call msg,BPF,$@)
 	$(Q)$(CLANG) $(BPFCFLAGS) -target bpf -D__TARGET_ARCH_$(ARCH)	      \

libbpf-tools/fsdist.c

@@ -38,6 +38,7 @@ enum fs_type {
 	XFS,
 	F2FS,
 	BCACHEFS,
+	ZFS,
 };
 
 static struct fs_config {
@@ -86,6 +87,13 @@ static struct fs_config {
 		[F_FSYNC] = "bch2_fsync",
 		[F_GETATTR] = "bch2_getattr",
 	}},
+	[ZFS] = { "zfs", {
+		[F_READ] = "zpl_iter_read",
+		[F_WRITE] = "zpl_iter_write",
+		[F_OPEN] = "zpl_open",
+		[F_FSYNC] = "zpl_fsync",
+		[F_GETATTR] = NULL, /* not supported */
+	}},
 };
 
 static char *file_op_names[] = {
@@ -126,7 +134,7 @@ static const struct argp_option opts[] = {
 	{ "timestamp", 'T', NULL, 0, "Print timestamp", 0 },
 	{ "milliseconds", 'm', NULL, 0, "Millisecond histogram", 0 },
 	{ "pid", 'p', "PID", 0, "Process ID to trace", 0 },
-	{ "type", 't', "Filesystem", 0, "Which filesystem to trace, [btrfs/ext4/nfs/xfs/f2fs/bcachefs]", 0 },
+	{ "type", 't', "Filesystem", 0, "Which filesystem to trace, [btrfs/ext4/nfs/xfs/f2fs/bcachefs/zfs]", 0 },
 	{ "verbose", 'v', NULL, 0, "Verbose debug output", 0 },
 	{ NULL, 'h', NULL, OPTION_HIDDEN, "Show the full help", 0 },
 	{},
@@ -159,6 +167,8 @@ static error_t parse_arg(int key, char *arg, struct argp_state *state)
 			fs_type = F2FS;
 		} else if (!strcmp(arg, "bcachefs")) {
 			fs_type = BCACHEFS;
+		} else if (!strcmp(arg, "zfs")) {
+			fs_type = ZFS;
 		} else {
 			warn("invalid filesystem\n");
 			argp_usage(state);
@@ -217,6 +227,8 @@ static void alias_parse(char *prog)
 		fs_type = F2FS;
 	} else if (strstr(name, "bcachefsdist")){
 		fs_type = BCACHEFS;
+	} else if (strstr(name, "zfsdist")) {
+		fs_type = ZFS;
 	}
 }
 

libbpf-tools/fsslower.c

@@ -41,6 +41,7 @@ enum fs_type {
 	XFS,
 	F2FS,
 	BCACHEFS,
+	ZFS,
 };
 
 static struct fs_config {
@@ -83,6 +84,12 @@ static struct fs_config {
 		[F_OPEN] = "bch2_open",
 		[F_FSYNC] = "bch2_fsync",
 	}},
+	[ZFS] = { "zfs", {
+		[F_READ] = "zpl_iter_read",
+		[F_WRITE] = "zpl_iter_write",
+		[F_OPEN] = "zpl_open",
+		[F_FSYNC] = "zpl_fsync",
+	}},
 };
 
 static char file_op[] = {
@@ -120,7 +127,7 @@ static const struct argp_option opts[] = {
 	{ "duration", 'd', "DURATION", 0, "Total duration of trace in seconds", 0 },
 	{ "pid", 'p', "PID", 0, "Process ID to trace", 0 },
 	{ "min", 'm', "MIN", 0, "Min latency to trace, in ms (default 10)", 0 },
-	{ "type", 't', "Filesystem", 0, "Which filesystem to trace, [btrfs/ext4/nfs/xfs/f2fs/bcachefs]", 0 },
+	{ "type", 't', "Filesystem", 0, "Which filesystem to trace, [btrfs/ext4/nfs/xfs/f2fs/bcachefs/zfs]", 0 },
 	{ "verbose", 'v', NULL, 0, "Verbose debug output", 0 },
 	{ NULL, 'h', NULL, OPTION_HIDDEN, "Show the full help", 0 },
 	{},
@@ -163,6 +170,8 @@ static error_t parse_arg(int key, char *arg, struct argp_state *state)
 			fs_type = F2FS;
 		} else if (!strcmp(arg, "bcachefs")) {
 			fs_type = BCACHEFS;
+		} else if (!strcmp(arg, "zfs")) {
+			fs_type = ZFS;
 		} else {
 			warn("invalid filesystem\n");
 			argp_usage(state);
@@ -201,6 +210,8 @@ static void alias_parse(char *prog)
 		fs_type = F2FS;
 	} else if (strstr(name, "bcachefsslower")){
 		fs_type = BCACHEFS;
+	} else if (!strcmp(name, "zfsslower")) {
+		fs_type = ZFS;
 	}
 }
 

libbpf-tools/gethostlatency.c

@@ -124,12 +124,11 @@ static void handle_lost_events(void *ctx, int cpu, __u64 lost_cnt)
 
 static int get_libc_path(char *path)
 {
-	FILE *f;
+	char proc_path[PATH_MAX + 32] = {};
 	char buf[PATH_MAX] = {};
-	char map_fname[PATH_MAX] = {};
-	char proc_path[PATH_MAX] = {};
 	char *filename;
 	float version;
+	FILE *f;
 
 	if (libc_path) {
 		memcpy(path, libc_path, strlen(libc_path));
@@ -139,8 +138,8 @@ static int get_libc_path(char *path)
 	if (target_pid == 0) {
 		f = fopen("/proc/self/maps", "r");
 	} else {
-		snprintf(map_fname, sizeof(map_fname), "/proc/%d/maps", target_pid);
-		f = fopen(map_fname, "r");
+		snprintf(buf, sizeof(buf), "/proc/%d/maps", target_pid);
+		f = fopen(buf, "r");
 	}
 	if (!f)
 		return -errno;

libbpf-tools/hardirqs.bpf.c

@@ -14,6 +14,7 @@ const volatile bool filter_cg = false;
 const volatile bool targ_dist = false;
 const volatile bool targ_ns = false;
 const volatile bool do_count = false;
+const volatile int targ_cpu = -1;
 
 struct {
 	__uint(type, BPF_MAP_TYPE_CGROUP_ARRAY);
@@ -38,10 +39,19 @@ struct {
 
 static struct info zero;
 
+static __always_inline bool is_target_cpu() {
+	if (targ_cpu < 0)
+		return true;
+
+	return targ_cpu == bpf_get_smp_processor_id();
+}
+
 static int handle_entry(int irq, struct irqaction *action)
 {
 	if (filter_cg && !bpf_current_task_under_cgroup(&cgroup_map, 0))
 		return 0;
+	if (!is_target_cpu())
+		return 0;
 
 	if (do_count) {
 		struct irq_key key = {};
@@ -76,6 +86,9 @@ static int handle_exit(int irq, struct irqaction *action)
 	if (filter_cg && !bpf_current_task_under_cgroup(&cgroup_map, 0))
 		return 0;
 
+	if (!is_target_cpu())
+		return 0;
+
 	tsp = bpf_map_lookup_elem(&start, &key);
 	if (!tsp)
 		return 0;

libbpf-tools/hardirqs.c

@@ -27,9 +27,11 @@ struct env {
 	bool verbose;
 	char *cgroupspath;
 	bool cg;
+	int targ_cpu;
 } env = {
 	.interval = 99999999,
 	.times = 99999999,
+	.targ_cpu = -1,
 };
 
 static volatile bool exiting;
@@ -47,12 +49,14 @@ const char argp_program_doc[] =
 "    hardirqs -d         # show hard irq event time as histograms\n"
 "    hardirqs 1 10       # print 1 second summaries, 10 times\n"
 "    hardirqs -c CG      # Trace process under cgroupsPath CG\n"
+"    hardirqs --cpu 1    # only stat irq on cpu 1\n"
 "    hardirqs -NT 1      # 1s summaries, nanoseconds, and timestamps\n";
 
 static const struct argp_option opts[] = {
 	{ "count", 'C', NULL, 0, "Show event counts instead of timing", 0 },
 	{ "distributed", 'd', NULL, 0, "Show distributions as histograms", 0 },
 	{ "cgroup", 'c', "/sys/fs/cgroup/unified", 0, "Trace process in cgroup path", 0 },
+	{ "cpu", 's', "CPU", 0, "Only stat irq on selected cpu", 0 },
 	{ "timestamp", 'T', NULL, 0, "Include timestamp on output", 0 },
 	{ "nanoseconds", 'N', NULL, 0, "Output in nanoseconds", 0 },
 	{ "verbose", 'v', NULL, 0, "Verbose debug output", 0 },
@@ -77,6 +81,14 @@ static error_t parse_arg(int key, char *arg, struct argp_state *state)
 	case 'C':
 		env.count = true;
 		break;
+	case 's':
+		errno = 0;
+		env.targ_cpu = atoi(arg);
+		if (errno || env.targ_cpu < 0) {
+			fprintf(stderr, "invalid cpu: %s\n", arg);
+			argp_usage(state);
+		}
+		break;
 	case 'c':
 		env.cgroupspath = arg;
 		env.cg = true;
@@ -218,6 +230,7 @@ int main(int argc, char **argv)
 
 	obj->rodata->filter_cg = env.cg;
 	obj->rodata->do_count = env.count;
+	obj->rodata->targ_cpu = env.targ_cpu;
 
 	/* initialize global data (filtering options) */
 	if (!env.count) {

libbpf-tools/javagc.c

@@ -125,25 +125,39 @@ static int get_jvmso_path(char *path)
 	size_t seg_start, seg_end, seg_off;
 	FILE *f;
 	int i = 0;
+	bool found = false;
+
+	if (env.pid == -1) {
+		fprintf(stderr, "not specify pid, see --pid.\n");
+		return -1;
+	}
 
 	sprintf(buf, "/proc/%d/maps", env.pid);
 	f = fopen(buf, "r");
-	if (!f)
+	if (!f) {
+		fprintf(stderr, "open %s failed: %m\n", buf);
 		return -1;
+	}
 
 	while (fscanf(f, "%zx-%zx %s %zx %*s %*d%[^\n]\n",
 			&seg_start, &seg_end, mode, &seg_off, line) == 5) {
 		i = 0;
 		while (isblank(line[i]))
 			i++;
 		if (strstr(line + i, "libjvm.so")) {
+			found = true;
+			strcpy(path, line + i);
 			break;
 		}
 	}
 
-	strcpy(path, line + i);
 	fclose(f);
 
+	if (!found) {
+		fprintf(stderr, "Not found libjvm.so.\n");
+		return -ENOENT;
+	}
+
 	return 0;
 }
 

libbpf-tools/memleak.bpf.c

@@ -119,7 +119,7 @@ static int gen_alloc_exit2(void *ctx, u64 address)
 	info.size = *size;
 	bpf_map_delete_elem(&sizes, &tid);
 
-	if (address != 0) {
+	if (address != 0 && address != MAP_FAILED) {
 		info.timestamp_ns = bpf_ktime_get_ns();
 
 		info.stack_id = bpf_get_stackid(ctx, &stack_traces, stack_flags);
@@ -223,6 +223,20 @@ int BPF_UPROBE(munmap_enter, void *address)
 	return gen_free_enter(address);
 }
 
+SEC("uprobe")
+int BPF_UPROBE(mremap_enter, void *old_address, size_t old_size, size_t new_size, int flags)
+{
+	gen_free_enter(old_address);
+
+	return gen_alloc_enter(new_size);
+}
+
+SEC("uretprobe")
+int BPF_URETPROBE(mremap_exit)
+{
+	return gen_alloc_exit(ctx);
+}
+
 SEC("uprobe")
 int BPF_UPROBE(posix_memalign_enter, void **memptr, size_t alignment, size_t size)
 {