Skip to contentSkip to conversation

cgroup bpf auto-detachment

PATCH Applied Roman Gushchin 2 participants 6 messages May 25, 2019 – May 28, 2019

Subsystems (7)
lore
Share & tools
lore
#0 lore raw

v5 cover cgroup bpf auto-detachment

This patchset implements a cgroup bpf auto-detachment functionality:
bpf programs are detached as soon as possible after removal of the cgroup, without waiting for the release of all associated resources.

Patches 2 and 3 are required to implement a corresponding kselftest in patch 4.

v5:
1) rebase

v4:
1) release cgroup bpf data using a workqueue
2) add test_cgroup_attach to .gitignore

v3:
1) some minor changes and typo fixes

v2:
1) removed a bogus check in patch 4
2) moved buf[len] = 0 in patch 2

Roman Gushchin (4):
bpf: decouple the lifetime of cgroup_bpf from cgroup itself selftests/bpf: convert test_cgrp2_attach2 example into kselftest selftests/bpf: enable all available cgroup v2 controllers selftests/bpf: add auto-detach test

 include/linux/bpf-cgroup.h                    |  11 +-
 include/linux/cgroup.h                        |  18 +++
 kernel/bpf/cgroup.c                           |  41 ++++-
 kernel/cgroup/cgroup.c                        |  11 +-
 samples/bpf/Makefile                          |   2 -
 tools/testing/selftests/bpf/.gitignore        |   1 +
 tools/testing/selftests/bpf/Makefile          |   3 +-
 tools/testing/selftests/bpf/cgroup_helpers.c  |  57 +++++++
 .../selftests/bpf/test_cgroup_attach.c        | 146 ++++++++++++++++--
 9 files changed, 261 insertions(+), 29 deletions(-)
 rename samples/bpf/test_cgrp2_attach2.c => tools/testing/selftests/bpf/test_cgroup_attach.c (79%)
Signature · 1 line
2.21.0
#1 lore raw

PATCH v5 1/4 bpf: decouple the lifetime of cgroup_bpf from cgroup itself

Currently the lifetime of bpf programs attached to a cgroup is bound to the lifetime of the cgroup itself. It means that if a user forgets (or intentionally avoids) to detach a bpf program before removing the cgroup, it will stay attached up to the release of the cgroup. Since the cgroup can stay in the dying state (the state between being rmdir()'ed and being released) for a very long time, it leads to a waste of memory. Also, it blocks a possibility to implement the memcg-based memory accounting for bpf objects, because a circular reference dependency will occur. Charged memory pages are pinning the corresponding memory cgroup, and if the memory cgroup is pinning the attached bpf program, nothing will be ever released.

A dying cgroup can not contain any processes, so the only chance for an attached bpf program to be executed is a live socket associated with the cgroup. So in order to release all bpf data early, let's count associated sockets using a new percpu refcounter. On cgroup removal the counter is transitioned to the atomic mode, and as soon as it reaches 0, all bpf programs are detached.

Because cgroup_bpf_release() can block, it can't be called from the percpu ref counter callback directly, so instead an asynchronous work is scheduled.

The reference counter is not socket specific, and can be used for any other types of programs, which can be executed from a cgroup-bpf hook outside of the process context, had such a need arise in the future.

---
 include/linux/bpf-cgroup.h | 11 ++++++++--
 include/linux/cgroup.h     | 18 +++++++++++++++++
 kernel/bpf/cgroup.c        | 41 ++++++++++++++++++++++++++++++++++----
 kernel/cgroup/cgroup.c     | 11 +++++++---
 4 files changed, 72 insertions(+), 9 deletions(-)
include/linux/bpf-cgroup.h +9 -2
index cb3c6b3b89c8..9f100fc422c3 100644
@@ -6,6 +6,7 @@
 #include <linux/errno.h>
 #include <linux/jump_label.h>
 #include <linux/percpu.h>
+#include <linux/percpu-refcount.h>
 #include <linux/rbtree.h>
 #include <uapi/linux/bpf.h>
 
@@ -72,10 +73,16 @@ struct cgroup_bpf {
 
 	/* temp storage for effective prog array used by prog_attach/detach */
 	struct bpf_prog_array __rcu *inactive;
+
+	/* reference counter used to detach bpf programs after cgroup removal */
+	struct percpu_ref refcnt;
+
+	/* cgroup_bpf is released using a work queue */
+	struct work_struct release_work;
 };
 
-void cgroup_bpf_put(struct cgroup *cgrp);
 int cgroup_bpf_inherit(struct cgroup *cgrp);
+void cgroup_bpf_offline(struct cgroup *cgrp);
 
 int __cgroup_bpf_attach(struct cgroup *cgrp, struct bpf_prog *prog,
 			enum bpf_attach_type type, u32 flags);
@@ -283,8 +290,8 @@ int cgroup_bpf_prog_query(const union bpf_attr *attr,
 
 struct bpf_prog;
 struct cgroup_bpf {};
-static inline void cgroup_bpf_put(struct cgroup *cgrp) {}
 static inline int cgroup_bpf_inherit(struct cgroup *cgrp) { return 0; }
+static inline void cgroup_bpf_offline(struct cgroup *cgrp) {}
 
 static inline int cgroup_bpf_prog_attach(const union bpf_attr *attr,
 					 enum bpf_prog_type ptype,
include/linux/cgroup.h +18 -0
index c0077adeea83..49e8facf7c4a 100644
@@ -924,4 +924,22 @@ static inline bool cgroup_task_frozen(struct task_struct *task)
 
 #endif /* !CONFIG_CGROUPS */
 
+#ifdef CONFIG_CGROUP_BPF
+static inline void cgroup_bpf_get(struct cgroup *cgrp)
+{
+	percpu_ref_get(&cgrp->bpf.refcnt);
+}
+
+static inline void cgroup_bpf_put(struct cgroup *cgrp)
+{
+	percpu_ref_put(&cgrp->bpf.refcnt);
+}
+
+#else /* CONFIG_CGROUP_BPF */
+
+static inline void cgroup_bpf_get(struct cgroup *cgrp) {}
+static inline void cgroup_bpf_put(struct cgroup *cgrp) {}
+
+#endif /* CONFIG_CGROUP_BPF */
+
 #endif /* _LINUX_CGROUP_H */
kernel/bpf/cgroup.c +37 -4

68 lines not on this page · raw message

kernel/cgroup/cgroup.c +8 -3
index 217cec4e22c6..ef9cfbfc82a9 100644
@@ -4955,8 +4955,6 @@ static void css_release_work_fn(struct work_struct *work)
 		if (cgrp->kn)
 			RCU_INIT_POINTER(*(void __rcu __force **)&cgrp->kn->priv,
 					 NULL);
-
-		cgroup_bpf_put(cgrp);
 	}
 
 	mutex_unlock(&cgroup_mutex);
@@ -5482,6 +5480,8 @@ static int cgroup_destroy_locked(struct cgroup *cgrp)
 
 	cgroup1_check_for_release(parent);
 
+	cgroup_bpf_offline(cgrp);
+
 	/* put the base reference */
 	percpu_ref_kill(&cgrp->self.refcnt);
 
@@ -6221,6 +6221,7 @@ void cgroup_sk_alloc(struct sock_cgroup_data *skcd)
 		 * Don't use cgroup_get_live().
 		 */
 		cgroup_get(sock_cgroup_ptr(skcd));
+		cgroup_bpf_get(sock_cgroup_ptr(skcd));
 		return;
 	}
 
@@ -6232,6 +6233,7 @@ void cgroup_sk_alloc(struct sock_cgroup_data *skcd)
 		cset = task_css_set(current);
 		if (likely(cgroup_tryget(cset->dfl_cgrp))) {
 			skcd->val = (unsigned long)cset->dfl_cgrp;
+			cgroup_bpf_get(cset->dfl_cgrp);
 			break;
 		}
 		cpu_relax();
@@ -6242,7 +6244,10 @@ void cgroup_sk_alloc(struct sock_cgroup_data *skcd)
 
 void cgroup_sk_free(struct sock_cgroup_data *skcd)
 {
-	cgroup_put(sock_cgroup_ptr(skcd));
+	struct cgroup *cgrp = sock_cgroup_ptr(skcd);
+
+	cgroup_bpf_put(cgrp);
+	cgroup_put(cgrp);
 }
 
 #endif	/* CONFIG_SOCK_CGROUP_DATA */
Signature · 1 line
2.21.0
#2 lore raw

PATCH v5 2/4 selftests/bpf: convert test_cgrp2_attach2 example into kselftest

In reply to Roman Gushchin

Convert test_cgrp2_attach2 example into a proper test_cgroup_attach kselftest. It's better because we do run kselftest on a constant basis, so there are better chances to spot a potential regression.

Also make it slightly less verbose to conform kselftests output style.

Output example:

  $ ./test_cgroup_attach
  #override:PASS
  #multi:PASS
  test_cgroup_attach:PASS
---
 samples/bpf/Makefile                          |  2 -
 tools/testing/selftests/bpf/.gitignore        |  1 +
 tools/testing/selftests/bpf/Makefile          |  3 +-
 .../selftests/bpf/test_cgroup_attach.c        | 50 ++++++++++++-------
 4 files changed, 36 insertions(+), 20 deletions(-)
 rename samples/bpf/test_cgrp2_attach2.c => tools/testing/selftests/bpf/test_cgroup_attach.c (91%)
samples/bpf/Makefile +0 -2
index 4f0a1cdbfe7c..253e5a2856be 100644
@@ -26,7 +26,6 @@ hostprogs-y += map_perf_test
 hostprogs-y += test_overhead
 hostprogs-y += test_cgrp2_array_pin
 hostprogs-y += test_cgrp2_attach
-hostprogs-y += test_cgrp2_attach2
 hostprogs-y += test_cgrp2_sock
 hostprogs-y += test_cgrp2_sock2
 hostprogs-y += xdp1
@@ -81,7 +80,6 @@ map_perf_test-objs := bpf_load.o map_perf_test_user.o
 test_overhead-objs := bpf_load.o test_overhead_user.o
 test_cgrp2_array_pin-objs := test_cgrp2_array_pin.o
 test_cgrp2_attach-objs := test_cgrp2_attach.o
-test_cgrp2_attach2-objs := test_cgrp2_attach2.o $(CGROUP_HELPERS)
 test_cgrp2_sock-objs := test_cgrp2_sock.o
 test_cgrp2_sock2-objs := bpf_load.o test_cgrp2_sock2.o
 xdp1-objs := xdp1_user.o
tools/testing/selftests/bpf/.gitignore +1 -0
index b3da2ffdc158..b2a9902f11c5 100644
@@ -22,6 +22,7 @@ test_lirc_mode2_user
 get_cgroup_id_user
 test_skb_cgroup_id_user
 test_socket_cookie
+test_cgroup_attach
 test_cgroup_storage
 test_select_reuseport
 test_flow_dissector
tools/testing/selftests/bpf/Makefile +2 -1
index fa002da36d0d..9b21391c4966 100644
@@ -26,7 +26,7 @@ TEST_GEN_PROGS = test_verifier test_tag test_maps test_lru_map test_lpm_map test
 	test_sock test_btf test_sockmap test_lirc_mode2_user get_cgroup_id_user \
 	test_socket_cookie test_cgroup_storage test_select_reuseport test_section_names \
 	test_netcnt test_tcpnotify_user test_sock_fields test_sysctl test_hashmap \
-	test_btf_dump
+	test_btf_dump test_cgroup_attach
 
 BPF_OBJ_FILES = $(patsubst %.c,%.o, $(notdir $(wildcard progs/*.c)))
 TEST_GEN_FILES = $(BPF_OBJ_FILES)
@@ -99,6 +99,7 @@ $(OUTPUT)/test_cgroup_storage: cgroup_helpers.c
 $(OUTPUT)/test_netcnt: cgroup_helpers.c
 $(OUTPUT)/test_sock_fields: cgroup_helpers.c
 $(OUTPUT)/test_sysctl: cgroup_helpers.c
+$(OUTPUT)/test_cgroup_attach: cgroup_helpers.c
 
 .PHONY: force
 
tools/testing/selftests/bpf/test_cgroup_attach.c renamed from samples/bpf/test_cgrp2_attach2.c +33 -17

119 lines not on this page · raw message

Signature · 1 line
2.21.0
#3 lore raw

PATCH v5 3/4 selftests/bpf: enable all available cgroup v2 controllers

In reply to Roman Gushchin

Enable all available cgroup v2 controllers when setting up the environment for the bpf kselftests. It's required to properly test the bpf prog auto-detach feature. Also it will generally increase the code coverage.

---
 tools/testing/selftests/bpf/cgroup_helpers.c | 57 ++++++++++++++++++++
 1 file changed, 57 insertions(+)
tools/testing/selftests/bpf/cgroup_helpers.c +57 -0
index 6692a40a6979..0d89f0396be4 100644
@@ -33,6 +33,60 @@
 	snprintf(buf, sizeof(buf), "%s%s%s", CGROUP_MOUNT_PATH, \
 		 CGROUP_WORK_DIR, path)
 
+/**
+ * enable_all_controllers() - Enable all available cgroup v2 controllers
+ *
+ * Enable all available cgroup v2 controllers in order to increase
+ * the code coverage.
+ *
+ * If successful, 0 is returned.
+ */
+int enable_all_controllers(char *cgroup_path)
+{
+	char path[PATH_MAX + 1];
+	char buf[PATH_MAX];
+	char *c, *c2;
+	int fd, cfd;
+	size_t len;
+
+	snprintf(path, sizeof(path), "%s/cgroup.controllers", cgroup_path);
+	fd = open(path, O_RDONLY);
+	if (fd < 0) {
+		log_err("Opening cgroup.controllers: %s", path);
+		return 1;
+	}
+
+	len = read(fd, buf, sizeof(buf) - 1);
+	if (len < 0) {
+		close(fd);
+		log_err("Reading cgroup.controllers: %s", path);
+		return 1;
+	}
+	buf[len] = 0;
+	close(fd);
+
+	/* No controllers available? We're probably on cgroup v1. */
+	if (len == 0)
+		return 0;
+
+	snprintf(path, sizeof(path), "%s/cgroup.subtree_control", cgroup_path);
+	cfd = open(path, O_RDWR);
+	if (cfd < 0) {
+		log_err("Opening cgroup.subtree_control: %s", path);
+		return 1;
+	}
+
+	for (c = strtok_r(buf, " ", &c2); c; c = strtok_r(NULL, " ", &c2)) {
+		if (dprintf(cfd, "+%s\n", c) <= 0) {
+			log_err("Enabling controller %s: %s", c, path);
+			close(cfd);
+			return 1;
+		}
+	}
+	close(cfd);
+	return 0;
+}
+
 /**
  * setup_cgroup_environment() - Setup the cgroup environment
  *
@@ -71,6 +125,9 @@ int setup_cgroup_environment(void)
 		return 1;
 	}
 
+	if (enable_all_controllers(cgroup_workdir))
+		return 1;
+
 	return 0;
 }
 
Signature · 1 line
2.21.0
#4 lore raw

PATCH v5 4/4 selftests/bpf: add auto-detach test

In reply to Roman Gushchin

Add a kselftest to cover bpf auto-detachment functionality. The test creates a cgroup, associates some resources with it, attaches a couple of bpf programs and deletes the cgroup.

Then it checks that bpf programs are going away in 5 seconds.

Expected output:

  $ ./test_cgroup_attach
  #override:PASS
  #multi:PASS
  #autodetach:PASS
  test_cgroup_attach:PASS

On a kernel without auto-detaching:

  $ ./test_cgroup_attach
  #override:PASS
  #multi:PASS
  #autodetach:FAIL
  test_cgroup_attach:FAIL
---
 .../selftests/bpf/test_cgroup_attach.c        | 98 ++++++++++++++++++-
 1 file changed, 97 insertions(+), 1 deletion(-)
tools/testing/selftests/bpf/test_cgroup_attach.c +97 -1
index 2d6d57f50e10..7671909ee1cb 100644
@@ -456,9 +456,105 @@ static int test_multiprog(void)
 	return rc;
 }
 
+static int test_autodetach(void)
+{
+	__u32 prog_cnt = 4, attach_flags;
+	int allow_prog[2] = {0};
+	__u32 prog_ids[2] = {0};
+	int cg = 0, i, rc = -1;
+	void *ptr = NULL;
+	int attempts;
+
+	for (i = 0; i < ARRAY_SIZE(allow_prog); i++) {
+		allow_prog[i] = prog_load_cnt(1, 1 << i);
+		if (!allow_prog[i])
+			goto err;
+	}
+
+	if (setup_cgroup_environment())
+		goto err;
+
+	/* create a cgroup, attach two programs and remember their ids */
+	cg = create_and_get_cgroup("/cg_autodetach");
+	if (cg < 0)
+		goto err;
+
+	if (join_cgroup("/cg_autodetach"))
+		goto err;
+
+	for (i = 0; i < ARRAY_SIZE(allow_prog); i++) {
+		if (bpf_prog_attach(allow_prog[i], cg, BPF_CGROUP_INET_EGRESS,
+				    BPF_F_ALLOW_MULTI)) {
+			log_err("Attaching prog[%d] to cg:egress", i);
+			goto err;
+		}
+	}
+
+	/* make sure that programs are attached and run some traffic */
+	assert(bpf_prog_query(cg, BPF_CGROUP_INET_EGRESS, 0, &attach_flags,
+			      prog_ids, &prog_cnt) == 0);
+	assert(system(PING_CMD) == 0);
+
+	/* allocate some memory (4Mb) to pin the original cgroup */
+	ptr = malloc(4 * (1 << 20));
+	if (!ptr)
+		goto err;
+
+	/* close programs and cgroup fd */
+	for (i = 0; i < ARRAY_SIZE(allow_prog); i++) {
+		close(allow_prog[i]);
+		allow_prog[i] = 0;
+	}
+
+	close(cg);
+	cg = 0;
+
+	/* leave the cgroup and remove it. don't detach programs */
+	cleanup_cgroup_environment();
+
+	/* wait for the asynchronous auto-detachment.
+	 * wait for no more than 5 sec and give up.
+	 */
+	for (i = 0; i < ARRAY_SIZE(prog_ids); i++) {
+		for (attempts = 5; attempts >= 0; attempts--) {
+			int fd = bpf_prog_get_fd_by_id(prog_ids[i]);
+
+			if (fd < 0)
+				break;
+
+			/* don't leave the fd open */
+			close(fd);
+
+			if (!attempts)
+				goto err;
+
+			sleep(1);
+		}
+	}
+
+	rc = 0;
+err:
+	for (i = 0; i < ARRAY_SIZE(allow_prog); i++)
+		if (allow_prog[i] > 0)
+			close(allow_prog[i]);
+	if (cg)
+		close(cg);
+	free(ptr);
+	cleanup_cgroup_environment();
+	if (!rc)
+		printf("#autodetach:PASS\n");
+	else
+		printf("#autodetach:FAIL\n");
+	return rc;
+}
+
 int main(void)
 {
-	int (*tests[])(void) = {test_foo_bar, test_multiprog};
+	int (*tests[])(void) = {
+		test_foo_bar,
+		test_multiprog,
+		test_autodetach,
+	};
 	int errors = 0;
 	int i;
 
Signature · 1 line
2.21.0
#5 lore raw
In reply to Roman Gushchin
Quoted 15 lines · Roman Gushchin

This patchset implements a cgroup bpf auto-detachment functionality:
bpf programs are detached as soon as possible after removal of the cgroup, without waiting for the release of all associated resources.

Patches 2 and 3 are required to implement a corresponding kselftest in patch 4.

v5:
1) rebase

v4:
1) release cgroup bpf data using a workqueue
2) add test_cgroup_attach to .gitignore

Quoted 6 lines

v3:
1) some minor changes and typo fixes

v2:
1) removed a bogus check in patch 4
2) moved buf[len] = 0 in patch 2

Applied. Thanks!