Merge tag 'trace-v7.3-rc2' of git://git.kernel.org/pub/scm/linux/kernel/git/trace/linux-tracegrafted

Pull tracing fixes from Steven Rostedt:

 - Don't destroy user event fields when removal fails

   User event fields are destroyed before the event is removed from
   visibility. But that can fail leaving the still visible event with no
   fields. Move the destroying of the fields to after the event is
   successfully removed from visibility.

 - Initialize function graph state is fork before calling
   copy_exec_state()

   For non-CLONE_VM forks, copy_exec_state() allocates a new
   task_exec_state. If that allocation fails, ftrace_graph_exit_task()
   will free the tasks ret_stack pointer. Since that pointer is still
   using the parent's ret_stack, it mistakenly frees the parent's
   pointer too.

   Call ftrace_graph_init() on the task first which will NULL out the
   new tasks's ret_stack and if the copy fails, it will not free
   anything.

 - Remove FGRAPH_MAX_INDEX

   The macro FGRAPH_MAX_INDEX was added but never used. Remove it.

 - Save ent_size in function graph printing of nested functions

   The function graph tracer needs to look at the next event to see if
   the next event is the return of the current function entry. If it is,
   it prints a single line:

	ktime_get();

   Otherwise it prints it like a nested function:

	tick_nohz_irq_exit() {
	    ktime_get();
	    kcpustat_irq_exit();
	}

   In order to look at the next event, it must save the current event so
   that it has the information to print from it. It saves the event in
   the iterator descriptor called "ent". What it doesn't save is the
   ent_size of the event which is now used to know if the function graph
   arguments are to be printed. The peek doesn't save the size so the
   size used happens to be that of the size of the last event that was
   seen.

   Save the entry event size in the iterator descriptor so that the
   correct size is used.

 - Fix several errors with freeing data in the histogram code

   The histogram code had a lot of leaked or or incorrect accounting
   when failures happen. Correct them.

 - Fix histogram regression of .percent and .graph modifiers

   Up until 6.3 histogram values could have "percent" or "graph"
   modifiers that changed how they were printed. But a change that added
   restricting histograms values from being strings, stack traces and
   other modifiers inadvertently prevented them from using the percent
   and graph modifiers, which were legal use cases for values.

   Put back the percent and graph modifiers.

 - Fix various typos in the comments

 - Set the trace_clock before initializing a histogram with clock
   argument

   The histogram API allows the user to specific which trace clock to
   use via a "clock=" string. The histogram is set up first before the
   clock is checked. If the passed in clock is not valid, it exits
   without fully fixing up the histogram leaving it on the list and a
   use-after-free can trigger.

   Update the clock argument first and if it fails then exit gracefully
   before the histogram trigger is placed on any lists.

 - Restore :mod: trailer after parsing in ftrace_set_clr_event

   The function ftrace_set_clr_event() modifies the parse string and
   needs to put it back to what was passed in. It searches for ":mod:"
   via a strsep() but fails to put back the first ':' in the string.

   Add back the ':' in the passed in string.

 - Take trace_array reference when opening a tracer options file

   The options files are dynamically created and some tracers add their
   own options. When a tracer adds their own list of options, the
   trace_array holding them has an array to hold the list of options for
   each tracer. This array increases in size via a krealloc(), and the
   new entry gets a newly allocated array to hold the options of the new
   tracer being added.

   The element in each entry of the tracer's option array holds a
   pointer back to the trace_array, a pointer to the tracer it is
   associated to, a pointer to the flags of the option.

   The issue is that these arrays are freed when the trace_array is
   freed when its instance it represents is removed from the instances
   directory. There's a race that an open of one of these options files
   can happen when the instance is being removed.

   Add a new helper function to be called by the open function of the
   options file to iterate all existing trace_arrays under a lock and
   find the one that has the given option element in one of it's tracer
   arrays. If found, then update the associated trace_array's reference
   counter to keep it from being freed. If not found, have the open call
   return -ENODEV.

 - Disable interrupts when acquiring the lock in rb_wake_up_waiters()

   The function rb_wake_up_waiters() assumes it will be called in
   interrupt context and does not disable irqs when taking
   cpu_buffer->reader_lock, which can be called in hard interrupt
   context. The issue is in PREEMPT_RT, this function is called in
   thread context leaving this lock open to a deadlock.

   Take the lock with interrupts disabled.

 - Use rcu_assign_pointer() for tmp_ops filter hash

   The tmp_ops used in update_ftrace_direct_mod() assigns its
   filter_hash field directly, but that field is annotated as __rcu and
   sparse complains. Assign it with rcu_assign_pointer()

 - Fix use-after-free in enable_trigger_private_data_free()

   The trace_event_call is accessed through the event_trigger_data's
   trace_event_file pointer to put the trace_event_call on freeing. The
   issue is that the trace_event_file data may have been freed already
   causing a use-after-free. Add a field to the event_trigger_data that
   points directly to the trace_event_call so that it can decrement its
   reference directly without needing to go through the
   trace_event_file.

 - Fix accounting of buffer data remote headers

   trace_buffer_desc_size() and trace_remote_alloc_buffer() undercount
   the number of pages is needed for the asked for size as it doesn't
   take into account the meta data on each page. Add a helper function
   to do the calculation properly and use that in these functions.

 - Catch nr_page_va overflow in ring_buffer_desc sizing

   The number of pages per remote ring buffer is capped by
   ring_buffer_desc::nr_page_va (32 bits). A buffer_size large enough to
   overflow that field would silently allocate a descriptor smaller than
   what was asked for.

 - Do not resize the subbuf order if any per_cpu buffer is disabled

   The mmapping of ring buffers disables resizing the subbuffers, but it
   is done per-cpu whereas the subbuf size change is done for all the
   per_cpu buffers under the buffer->mutex. It could change the size of
   some while the mapping is happening on others. Have the resize of the
   subbuf order check all the per_cpu buffers under the lock to see if
   any of them is disabled before starting and causing an inconsistency
   between buffers that are being mapped.

* tag 'trace-v7.3-rc2' of git://git.kernel.org/pub/scm/linux/kernel/git/trace/linux-trace: (25 commits)
  ring-buffer: Check resize_disabled before publishing the new subbuf order
  tracing/remotes: Catch nr_page_va overflow in ring_buffer_desc sizing
  tracing/remotes: Account for ring buffer page header in size calculation
  tracing: Don't dereference trace_event_file in deferred trigger free
  ftrace: Use rcu_assign_pointer() for tmp_ops filter hash
  ring-buffer: Acquire the lock with irqsave in rb_wake_up_waiters()
  tracing: Take trace_array reference when opening a tracer options file
  tracing: Fix ring_buffer_read_page_size() kernel-doc
  tracing: Restore :mod: trailer after parsing in ftrace_set_clr_event()
  tracing: Fix memory corruption from a "STACKTRACE" histogram key
  tracing: Fix memory corruption from the histogram stacktrace modifier
  tracing: Undo the registration when enabling the histogram trigger fails
  tracing: Take the reference before publishing the named histogram trigger
  tracing: Set the trace clock before registering the histogram trigger
  tracing: Fix typo "preceeded" in comment
  tracing: Fix typo "availabe" in comment
  tracing: Let histogram values keep the percent and graph modifiers
  tracing: Keep the entry count when the histogram stats allocation fails
  tracing: Free histogram the field rejected for a bad modifier
  tracing: Free histogram the var ref when its initialization fails
  ...

7 files changed, 750 insertions, 0 deletions

diff --git a/tools/testing/selftests/acct/.gitignore b/tools/testing/selftests/acct/.gitignore
new file mode 100644
index 000000000..fe0896f54
--- /dev/null
+++ b/tools/testing/selftests/acct/.gitignore
@@ -0,0 +1,5 @@
+acct_syscall
+taskstats_fill_stats_tgid
+cgroupstats
+config
+process_log
diff --git a/tools/testing/selftests/acct/Makefile b/tools/testing/selftests/acct/Makefile
new file mode 100644
index 000000000..93a11a28a
--- /dev/null
+++ b/tools/testing/selftests/acct/Makefile
@@ -0,0 +1,19 @@
+# SPDX-License-Identifier: GPL-2.0
+TEST_GEN_PROGS := acct_syscall
+TEST_GEN_PROGS += taskstats_fill_stats_tgid
+TEST_GEN_PROGS += cgroupstats
+
+NETLINK_HELPER_PROGS := cgroupstats taskstats_fill_stats_tgid
+
+CFLAGS += -Wall
+LDLIBS += -lpthread
+
+include ../lib.mk
+
+$(NETLINK_HELPER_PROGS): %: %.c netlink_helper.c netlink_helper.h
+	$(call msg,CC,,$@)
+	$(Q)$(LINK.c) $< netlink_helper.c $(LDLIBS) -o $@
+
+$(addprefix $(OUTPUT)/,$(NETLINK_HELPER_PROGS)): $(OUTPUT)/%: %.c netlink_helper.c netlink_helper.h
+	$(call msg,CC,,$@)
+	$(Q)$(LINK.c) $< netlink_helper.c $(LDLIBS) -o $@
diff --git a/tools/testing/selftests/acct/acct_syscall.c b/tools/testing/selftests/acct/acct_syscall.c
new file mode 100644
index 000000000..421adbdc2
--- /dev/null
+++ b/tools/testing/selftests/acct/acct_syscall.c
@@ -0,0 +1,78 @@
+// SPDX-License-Identifier: GPL-2.0
+
+/* kselftest for acct() system call
+ *  The acct() system call enables or disables process accounting.
+ */
+
+#include <stdio.h>
+#include <errno.h>
+#include <string.h>
+#include <sys/wait.h>
+
+#include "kselftest.h"
+
+int main(void)
+{
+	char filename[] = "process_log";
+	FILE *fp;
+	pid_t child_pid;
+	int sz;
+
+	// Setting up kselftest framework
+	ksft_print_header();
+	ksft_set_plan(1);
+
+	// Check if test is run a root
+	if (geteuid()) {
+		ksft_exit_skip("This test needs root to run!\n");
+		return 1;
+	}
+
+	// Create file to log closed processes
+	fp = fopen(filename, "w");
+
+	if (!fp) {
+		ksft_test_result_error("%s.\n", strerror(errno));
+		ksft_finished();
+		return 1;
+	}
+
+	acct(filename);
+
+	// Handle error conditions
+	if (errno) {
+		ksft_test_result_error("%s.\n", strerror(errno));
+		fclose(fp);
+		ksft_finished();
+		return 1;
+	}
+
+	// Create child process and wait for it to terminate.
+
+	child_pid = fork();
+
+	if (child_pid < 0) {
+		ksft_test_result_error("Creating a child process to log failed\n");
+		acct(NULL);
+		return 1;
+	} else if (child_pid > 0) {
+		wait(NULL);
+		fseek(fp, 0L, SEEK_END);
+		sz = ftell(fp);
+
+		acct(NULL);
+
+		if (sz <= 0) {
+			ksft_test_result_fail("Terminated child process not logged\n");
+			ksft_exit_fail();
+			return 1;
+		}
+
+		ksft_test_result_pass("Successfully logged terminated process.\n");
+		fclose(fp);
+		ksft_exit_pass();
+		return 0;
+	}
+
+	return 1;
+}
diff --git a/tools/testing/selftests/acct/cgroupstats.c b/tools/testing/selftests/acct/cgroupstats.c
new file mode 100644
index 000000000..0b421a4ca
--- /dev/null
+++ b/tools/testing/selftests/acct/cgroupstats.c
@@ -0,0 +1,231 @@
+// SPDX-License-Identifier: GPL-2.0
+#define _GNU_SOURCE
+
+#include <errno.h>
+#include <fcntl.h>
+#include <linux/cgroupstats.h>
+#include <linux/genetlink.h>
+#include <linux/netlink.h>
+#include <stdbool.h>
+#include <stdint.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+#include <sys/mount.h>
+#include <sys/socket.h>
+#include <sys/types.h>
+#include <unistd.h>
+
+#include "netlink_helper.h"
+#include "kselftest.h"
+
+static int send_cgroupstats_cmd(int fd, int family_id, uint32_t cgroup_fd,
+				int flags)
+{
+	struct {
+		struct nlmsghdr nlh;
+		struct genlmsghdr genl;
+		char buf[256];
+	} req = { 0 };
+	struct nlattr *na;
+
+	req.nlh.nlmsg_len = NLMSG_LENGTH(GENL_HDRLEN);
+	req.nlh.nlmsg_type = family_id;
+	req.nlh.nlmsg_flags = NLM_F_REQUEST | flags;
+	req.nlh.nlmsg_seq = 2;
+	req.nlh.nlmsg_pid = getpid();
+
+	req.genl.cmd = CGROUPSTATS_CMD_GET;
+	req.genl.version = 1;
+
+	na = (struct nlattr *)((char *)&req + NLMSG_ALIGN(req.nlh.nlmsg_len));
+	na->nla_type = CGROUPSTATS_CMD_ATTR_FD;
+	na->nla_len = NLA_HDRLEN + sizeof(cgroup_fd);
+	memcpy(nla_data(na), &cgroup_fd, sizeof(cgroup_fd));
+	req.nlh.nlmsg_len = NLMSG_ALIGN(req.nlh.nlmsg_len) + NLA_ALIGN(na->nla_len);
+
+	return send_request(fd, &req, req.nlh.nlmsg_len);
+}
+
+/*
+ * Receive and decode a cgroupstats response.
+ *
+ * Returns:
+ *   0             — success, stats filled from CGROUPSTATS_CMD_NEW reply
+ *   <0            — NLMSG_ERROR errno (e.g. -EBADF, -EINVAL)
+ */
+static int recv_cgroupstats_response(int fd, struct cgroupstats *stats)
+{
+	char resp[8192];
+	struct nlmsghdr *nlh;
+	struct genlmsghdr *genl;
+	struct nlattr *na;
+	int len;
+	int rem;
+
+	memset(stats, 0, sizeof(*stats));
+
+	len = recv(fd, resp, sizeof(resp), 0);
+	if (len < 0)
+		return -errno;
+
+	for (nlh = (struct nlmsghdr *)resp; NLMSG_OK(nlh, len);
+	     nlh = NLMSG_NEXT(nlh, len)) {
+		if (nlh->nlmsg_type == NLMSG_ERROR) {
+			struct nlmsgerr *err = NLMSG_DATA(nlh);
+
+			return err->error;
+		}
+
+		genl = (struct genlmsghdr *)NLMSG_DATA(nlh);
+		if (genl->cmd != CGROUPSTATS_CMD_NEW)
+			continue;
+
+		rem = nlh->nlmsg_len - NLMSG_HDRLEN - GENL_HDRLEN;
+		na = (struct nlattr *)((char *)genl + GENL_HDRLEN);
+		while (nla_ok(na, rem)) {
+			if (na->nla_type == CGROUPSTATS_TYPE_CGROUP_STATS) {
+				memcpy(stats, nla_data(na), sizeof(*stats));
+				return 0;
+			}
+			na = nla_next(na, &rem);
+		}
+	}
+
+	return -EIO;
+}
+
+/* mkdtemp() modifies the template in place, so this cannot be const. */
+static char cg_mountpoint[32];
+static bool cg_mounted;
+
+static int setup_cgroup_v1(void)
+{
+	strcpy(cg_mountpoint, "/tmp/cgstats_test_XXXXXX");
+
+	if (!mkdtemp(cg_mountpoint))
+		return -errno;
+
+	if (mount("cgstats_test", cg_mountpoint, "cgroup", 0,
+		  "none,name=cgstats_test") < 0) {
+		int ret = -errno;
+
+		rmdir(cg_mountpoint);
+		return ret;
+	}
+
+	cg_mounted = true;
+	return 0;
+}
+
+static void cleanup_cgroup_v1(void)
+{
+	if (!cg_mounted)
+		return;
+	umount2(cg_mountpoint, MNT_DETACH);
+	rmdir(cg_mountpoint);
+	cg_mounted = false;
+}
+
+int main(void)
+{
+	struct cgroupstats stats;
+	uint64_t total_tasks;
+	int family_id;
+	int nl_fd;
+	int cg_fd;
+	int ret;
+
+	ksft_print_header();
+
+	nl_fd = netlink_open();
+	if (nl_fd < 0)
+		ksft_exit_skip("failed to open generic netlink socket: %s\n",
+			       strerror(-nl_fd));
+
+	family_id = get_family_id(nl_fd, TASKSTATS_GENL_NAME);
+	if (family_id < 0)
+		ksft_exit_skip("taskstats generic netlink family unavailable: %s\n",
+			       strerror(-family_id));
+
+	ksft_set_plan(3);
+
+	/*
+	 * Test 1: mount a private cgroup v1 hierarchy, query it, and
+	 * verify the response contains sane task counts. If the test
+	 * environment cannot create a private cgroup v1 mount, skip this
+	 * case and continue with the unprivileged regression checks below.
+	 */
+	ret = setup_cgroup_v1();
+	if (ret) {
+		ksft_test_result_skip("cgroupstats query: cannot mount cgroup v1: %s\n",
+				      strerror(-ret));
+	} else {
+		cg_fd = open(cg_mountpoint, O_RDONLY | O_DIRECTORY);
+		if (cg_fd < 0) {
+			ksft_test_result_fail("cgroupstats query: open mountpoint: %s\n",
+					      strerror(errno));
+		} else {
+			ret = send_cgroupstats_cmd(nl_fd, family_id,
+						   (uint32_t)cg_fd, 0);
+			if (ret) {
+				ksft_test_result_fail("cgroupstats query: send: %s\n",
+						      strerror(-ret));
+			} else {
+				ret = recv_cgroupstats_response(nl_fd, &stats);
+				if (ret < 0) {
+					ksft_test_result_fail("cgroupstats query: %s\n",
+							      strerror(-ret));
+				} else {
+					total_tasks = (uint64_t)stats.nr_sleeping +
+						      (uint64_t)stats.nr_running +
+						      (uint64_t)stats.nr_stopped +
+						      (uint64_t)stats.nr_uninterruptible +
+						      (uint64_t)stats.nr_io_wait;
+
+					ksft_print_msg("cgroupstats query: total_tasks=%llu\n",
+						       (unsigned long long)total_tasks);
+
+					ksft_test_result(total_tasks > 0,
+							 "cgroupstats query returns valid stats\n");
+				}
+			}
+			close(cg_fd);
+		}
+	}
+	cleanup_cgroup_v1();
+
+	/*
+	 * Test 2: invalid fd without NLM_F_ACK.  The kernel should
+	 * return -EBADF via NLMSG_ERROR regardless of whether the
+	 * client requested an explicit ACK.
+	 */
+	ret = send_cgroupstats_cmd(nl_fd, family_id, 0xFFFFFFFF, 0);
+	if (ret)
+		ksft_exit_fail_msg("send test 2 failed: %s\n", strerror(-ret));
+
+	ret = recv_cgroupstats_response(nl_fd, &stats);
+	ksft_print_msg("bad fd (no ACK): response=%d (%s)\n",
+		       ret, ret < 0 ? strerror(-ret) : "unexpected success");
+	ksft_test_result(ret == -EBADF,
+			 "cgroupstats rejects bad fd without NLM_F_ACK\n");
+
+	/*
+	 * Test 3: invalid fd with NLM_F_ACK.  Same expectation as
+	 * test 2, but exercised through a different netlink flag
+	 * path in the kernel's ack/error handling.
+	 */
+	ret = send_cgroupstats_cmd(nl_fd, family_id, 0xFFFFFFFF, NLM_F_ACK);
+	if (ret)
+		ksft_exit_fail_msg("send test 3 failed: %s\n", strerror(-ret));
+
+	ret = recv_cgroupstats_response(nl_fd, &stats);
+	ksft_print_msg("bad fd (with ACK): response=%d (%s)\n",
+		       ret, ret < 0 ? strerror(-ret) : "unexpected success");
+	ksft_test_result(ret == -EBADF,
+			 "cgroupstats rejects bad fd with NLM_F_ACK\n");
+
+	close(nl_fd);
+	ksft_finished();
+	return ksft_get_fail_cnt() ? KSFT_FAIL : KSFT_PASS;
+}
diff --git a/tools/testing/selftests/acct/netlink_helper.c b/tools/testing/selftests/acct/netlink_helper.c
new file mode 100644
index 000000000..3ed834f0e
--- /dev/null
+++ b/tools/testing/selftests/acct/netlink_helper.c
@@ -0,0 +1,116 @@
+// SPDX-License-Identifier: GPL-2.0
+
+#include <errno.h>
+#include <stdint.h>
+#include <string.h>
+#include <sys/socket.h>
+#include <sys/time.h>
+#include <unistd.h>
+#include <linux/genetlink.h>
+
+#include "netlink_helper.h"
+
+int netlink_open(void)
+{
+	struct timeval tv = { .tv_sec = ACCT_RCV_TIMEOUT_SEC };
+	struct sockaddr_nl addr = {
+		.nl_family = AF_NETLINK,
+		.nl_pid = getpid(),
+	};
+	int fd;
+
+	fd = socket(AF_NETLINK, SOCK_RAW, NETLINK_GENERIC);
+	if (fd < 0)
+		return -errno;
+
+	if (setsockopt(fd, SOL_SOCKET, SO_RCVTIMEO, &tv, sizeof(tv)) < 0) {
+		int err = -errno;
+
+		close(fd);
+		return err;
+	}
+
+	if (bind(fd, (struct sockaddr *)&addr, sizeof(addr)) < 0) {
+		int err = -errno;
+
+		close(fd);
+		return err;
+	}
+
+	return fd;
+}
+
+int send_request(int fd, void *buf, size_t len)
+{
+	struct sockaddr_nl addr = {
+		.nl_family = AF_NETLINK,
+	};
+
+	if (sendto(fd, buf, len, 0, (struct sockaddr *)&addr, sizeof(addr)) < 0)
+		return -errno;
+
+	return 0;
+}
+
+/*
+ * Resolve the generic netlink family ID for @name.
+ * Returns the family ID (>= 0) on success, negative errno on failure.
+ */
+int get_family_id(int fd, const char *name)
+{
+	struct {
+		struct nlmsghdr nlh;
+		struct genlmsghdr genl;
+		char buf[256];
+	} req = { 0 };
+	char resp[8192];
+	struct nlmsghdr *nlh;
+	struct genlmsghdr *genl;
+	struct nlattr *na;
+	int len;
+	int rem;
+	int ret;
+
+	req.nlh.nlmsg_len = NLMSG_LENGTH(GENL_HDRLEN);
+	req.nlh.nlmsg_type = GENL_ID_CTRL;
+	req.nlh.nlmsg_flags = NLM_F_REQUEST;
+	req.nlh.nlmsg_seq = 1;
+	req.nlh.nlmsg_pid = getpid();
+
+	req.genl.cmd = CTRL_CMD_GETFAMILY;
+	req.genl.version = 1;
+
+	na = (struct nlattr *)((char *)&req + NLMSG_ALIGN(req.nlh.nlmsg_len));
+	na->nla_type = CTRL_ATTR_FAMILY_NAME;
+	na->nla_len = NLA_HDRLEN + strlen(name) + 1;
+	memcpy(nla_data(na), name, strlen(name) + 1);
+	req.nlh.nlmsg_len = NLMSG_ALIGN(req.nlh.nlmsg_len) + NLA_ALIGN(na->nla_len);
+
+	ret = send_request(fd, &req, req.nlh.nlmsg_len);
+	if (ret)
+		return ret;
+
+	len = recv(fd, resp, sizeof(resp), 0);
+	if (len < 0)
+		return -errno;
+
+	for (nlh = (struct nlmsghdr *)resp; NLMSG_OK(nlh, len);
+	     nlh = NLMSG_NEXT(nlh, len)) {
+		if (nlh->nlmsg_type == NLMSG_ERROR) {
+			struct nlmsgerr *err = NLMSG_DATA(nlh);
+
+			return err->error ? err->error : -ENOENT;
+		}
+
+		genl = (struct genlmsghdr *)NLMSG_DATA(nlh);
+		rem = nlh->nlmsg_len - NLMSG_HDRLEN - GENL_HDRLEN;
+		na = (struct nlattr *)((char *)genl + GENL_HDRLEN);
+		while (nla_ok(na, rem)) {
+			if (na->nla_type == CTRL_ATTR_FAMILY_ID)
+				return *(uint16_t *)nla_data(na);
+			na = nla_next(na, &rem);
+		}
+	}
+
+	return -ENOENT;
+}
diff --git a/tools/testing/selftests/acct/netlink_helper.h b/tools/testing/selftests/acct/netlink_helper.h
new file mode 100644
index 000000000..0320729c4
--- /dev/null
+++ b/tools/testing/selftests/acct/netlink_helper.h
@@ -0,0 +1,44 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+/*
+ * Shared generic netlink helpers for the acct selftests.
+ */
+#ifndef ACSELFTESTS_ACCT_NETLINK_HELPER_H
+#define ACSELFTESTS_ACCT_NETLINK_HELPER_H
+
+#include <stdbool.h>
+#include <linux/netlink.h>
+
+#ifndef NLA_ALIGNTO
+#define NLA_ALIGNTO 4
+#define NLA_ALIGN(len) (((len) + NLA_ALIGNTO - 1) & ~(NLA_ALIGNTO - 1))
+#define NLA_HDRLEN ((int)NLA_ALIGN(sizeof(struct nlattr)))
+#endif
+
+/* Fail an individual test case instead of hanging the whole binary. */
+#define ACCT_RCV_TIMEOUT_SEC 2
+
+static inline void *nla_data(const struct nlattr *na)
+{
+	return (void *)((char *)na + NLA_HDRLEN);
+}
+
+static inline bool nla_ok(const struct nlattr *na, int remaining)
+{
+	return remaining >= (int)sizeof(*na) &&
+	       na->nla_len >= sizeof(*na) &&
+	       na->nla_len <= remaining;
+}
+
+static inline struct nlattr *nla_next(const struct nlattr *na, int *remaining)
+{
+	int aligned_len = NLA_ALIGN(na->nla_len);
+
+	*remaining -= aligned_len;
+	return (struct nlattr *)((char *)na + aligned_len);
+}
+
+int netlink_open(void);
+int send_request(int fd, void *buf, size_t len);
+int get_family_id(int fd, const char *name);
+
+#endif /* ACSELFTESTS_ACCT_NETLINK_HELPER_H */
diff --git a/tools/testing/selftests/acct/taskstats_fill_stats_tgid.c b/tools/testing/selftests/acct/taskstats_fill_stats_tgid.c
new file mode 100644
index 000000000..9a4c1554d
--- /dev/null
+++ b/tools/testing/selftests/acct/taskstats_fill_stats_tgid.c
@@ -0,0 +1,257 @@
+// SPDX-License-Identifier: GPL-2.0
+#define _GNU_SOURCE
+
+#include <errno.h>
+#include <linux/genetlink.h>
+#include <linux/netlink.h>
+#include <linux/taskstats.h>
+#include <pthread.h>
+#include <stdbool.h>
+#include <stdint.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+#include <sys/socket.h>
+#include <sys/types.h>
+#include <time.h>
+#include <unistd.h>
+
+#include "netlink_helper.h"
+#include "kselftest.h"
+
+#define BUSY_NS (200ULL * 1000 * 1000)
+
+struct worker_ctx {
+	pthread_mutex_t lock;
+	pthread_cond_t cond;
+	bool ready;
+	bool release;
+};
+
+static unsigned long busy_sink;
+
+static uint64_t timespec_diff_ns(const struct timespec *start,
+				 const struct timespec *end)
+{
+	return (uint64_t)(end->tv_sec - start->tv_sec) * 1000000000ULL +
+	       (uint64_t)(end->tv_nsec - start->tv_nsec);
+}
+
+static void burn_cpu_for_ns(uint64_t runtime_ns)
+{
+	struct timespec start, now;
+	unsigned long acc = 0;
+
+	if (clock_gettime(CLOCK_MONOTONIC, &start)) {
+		perror("clock_gettime");
+		exit(EXIT_FAILURE);
+	}
+
+	do {
+		for (int i = 0; i < 100000; i++)
+			acc += i;
+		if (clock_gettime(CLOCK_MONOTONIC, &now)) {
+			perror("clock_gettime");
+			exit(EXIT_FAILURE);
+		}
+	} while (timespec_diff_ns(&start, &now) < runtime_ns);
+
+	busy_sink = acc;
+}
+
+static int get_taskstats(int fd, int family_id, uint16_t attr_type, uint32_t id,
+			 struct taskstats *stats)
+{
+	struct {
+		struct nlmsghdr nlh;
+		struct genlmsghdr genl;
+		char buf[256];
+	} req = { 0 };
+	char resp[16384];
+	struct nlmsghdr *nlh;
+	struct genlmsghdr *genl;
+	struct nlattr *na;
+	struct nlattr *nested;
+	int len;
+	int rem;
+	int nrem;
+	int ret;
+
+	memset(stats, 0, sizeof(*stats));
+
+	req.nlh.nlmsg_len = NLMSG_LENGTH(GENL_HDRLEN);
+	req.nlh.nlmsg_type = family_id;
+	req.nlh.nlmsg_flags = NLM_F_REQUEST;
+	req.nlh.nlmsg_seq = 2;
+	req.nlh.nlmsg_pid = getpid();
+
+	req.genl.cmd = TASKSTATS_CMD_GET;
+	req.genl.version = 1;
+
+	na = (struct nlattr *)((char *)&req + NLMSG_ALIGN(req.nlh.nlmsg_len));
+	na->nla_type = attr_type;
+	na->nla_len = NLA_HDRLEN + sizeof(id);
+	memcpy(nla_data(na), &id, sizeof(id));
+	req.nlh.nlmsg_len = NLMSG_ALIGN(req.nlh.nlmsg_len) + NLA_ALIGN(na->nla_len);
+
+	ret = send_request(fd, &req, req.nlh.nlmsg_len);
+	if (ret)
+		return ret;
+
+	len = recv(fd, resp, sizeof(resp), 0);
+	if (len < 0)
+		return -errno;
+
+	for (nlh = (struct nlmsghdr *)resp; NLMSG_OK(nlh, len);
+	     nlh = NLMSG_NEXT(nlh, len)) {
+		if (nlh->nlmsg_type == NLMSG_ERROR) {
+			struct nlmsgerr *err = NLMSG_DATA(nlh);
+
+			return err->error ? err->error : -ENOENT;
+		}
+
+		genl = (struct genlmsghdr *)NLMSG_DATA(nlh);
+		rem = nlh->nlmsg_len - NLMSG_HDRLEN - GENL_HDRLEN;
+		na = (struct nlattr *)((char *)genl + GENL_HDRLEN);
+		while (nla_ok(na, rem)) {
+			if (na->nla_type == TASKSTATS_TYPE_AGGR_PID ||
+			    na->nla_type == TASKSTATS_TYPE_AGGR_TGID) {
+				nested = (struct nlattr *)nla_data(na);
+				nrem = na->nla_len - NLA_HDRLEN;
+				while (nla_ok(nested, nrem)) {
+					if (nested->nla_type == TASKSTATS_TYPE_STATS) {
+						memcpy(stats, nla_data(nested),
+						       sizeof(*stats));
+						return 0;
+					}
+					nested = nla_next(nested, &nrem);
+				}
+			}
+			na = nla_next(na, &rem);
+		}
+	}
+
+	return -ENOENT;
+}
+
+static uint64_t cpu_total(const struct taskstats *stats)
+{
+	return (uint64_t)stats->ac_utime + (uint64_t)stats->ac_stime;
+}
+
+static void print_stats(const char *label, const struct taskstats *stats)
+{
+	ksft_print_msg("%s: cpu_total=%llu nvcsw=%llu nivcsw=%llu\n",
+		       label, (unsigned long long)cpu_total(stats),
+		       (unsigned long long)stats->nvcsw,
+		       (unsigned long long)stats->nivcsw);
+}
+
+static void *worker_thread(void *arg)
+{
+	struct worker_ctx *ctx = arg;
+
+	burn_cpu_for_ns(BUSY_NS);
+
+	pthread_mutex_lock(&ctx->lock);
+	ctx->ready = true;
+	pthread_cond_broadcast(&ctx->cond);
+	while (!ctx->release)
+		pthread_cond_wait(&ctx->cond, &ctx->lock);
+	pthread_mutex_unlock(&ctx->lock);
+
+	return NULL;
+}
+
+int main(void)
+{
+	struct worker_ctx ctx = {
+		.lock = PTHREAD_MUTEX_INITIALIZER,
+		.cond = PTHREAD_COND_INITIALIZER,
+	};
+	struct taskstats before, after;
+	pthread_t thread;
+	pid_t tgid = getpid();
+	int family_id;
+	int fd;
+	int ret;
+
+	ksft_print_header();
+	ksft_set_plan(1);
+
+	if (geteuid())
+		ksft_exit_skip("taskstats_fill_stats_tgid needs root\n");
+
+	fd = netlink_open();
+	if (fd < 0)
+		ksft_exit_skip("failed to open generic netlink socket: %s\n",
+			       strerror(-fd));
+
+	family_id = get_family_id(fd, TASKSTATS_GENL_NAME);
+	if (family_id < 0)
+		ksft_exit_skip("taskstats generic netlink family unavailable: %s\n",
+			       strerror(-family_id));
+
+	/* Create worker thread that burns 200ms of CPU */
+	if (pthread_create(&thread, NULL, worker_thread, &ctx) != 0)
+		ksft_exit_fail_msg("pthread_create failed: %s\n", strerror(errno));
+
+	/* Wait for worker to finish generating activity */
+	pthread_mutex_lock(&ctx.lock);
+	while (!ctx.ready)
+		pthread_cond_wait(&ctx.cond, &ctx.lock);
+	pthread_mutex_unlock(&ctx.lock);
+
+	/*
+	 * Snapshot A: TGID stats while worker is alive and sleeping.
+	 * Contains main thread + worker contributions.
+	 */
+	ret = get_taskstats(fd, family_id, TASKSTATS_CMD_ATTR_TGID, tgid, &before);
+	if (ret)
+		ksft_exit_fail_msg("TGID query before exit failed: %s\n",
+				   strerror(-ret));
+
+	/* Release worker so it can exit, then join (deterministic wait).
+	 *
+	 * Kernel exit path ordering guarantees:
+	 *   do_exit()
+	 *     taskstats_exit() -> fill_tgid_exit()  (accumulates worker into signal->stats)
+	 *     exit_notify()                         (releases the thread)
+	 *     do_task_dead() -> __schedule()        (wakes joiner)
+	 *
+	 * So pthread_join() returns only after fill_tgid_exit() has completed.
+	 */
+	pthread_mutex_lock(&ctx.lock);
+	ctx.release = true;
+	pthread_cond_broadcast(&ctx.cond);
+	pthread_mutex_unlock(&ctx.lock);
+
+	pthread_join(thread, NULL);
+
+	/*
+	 * Snapshot B: TGID stats after worker has exited.
+	 * fill_stats_for_tgid() does:
+	 *   memcpy(signal->stats)   <- includes fill_tgid_exit accumulation
+	 *   + scan live threads      <- only main thread now
+	 */
+	ret = get_taskstats(fd, family_id, TASKSTATS_CMD_ATTR_TGID, tgid, &after);
+	if (ret)
+		ksft_exit_fail_msg("TGID query after exit failed: %s\n",
+				   strerror(-ret));
+
+	print_stats("TGID before worker exit", &before);
+	print_stats("TGID after  worker exit", &after);
+
+	/*
+	 * The worker burned 200ms of CPU before the first snapshot.
+	 * If the kernel correctly retained its contribution via
+	 * fill_tgid_exit(), then the TGID CPU total after exit must be at
+	 * least as large as the TGID CPU total before exit.
+	 */
+	ksft_test_result(cpu_total(&after) >= cpu_total(&before),
+			 "TGID CPU stats should not regress after thread exit\n");
+
+	close(fd);
+	ksft_finished();
+	return ksft_get_fail_cnt() ? KSFT_FAIL : KSFT_PASS;
+}