Skip to contentSkip to conversation

bpftool: Query only cgroup-related attach types

PATCH Superseded Kenta Tada 2 participants 3 messages Jun 7, 2024

Subsystems (2)

v2 of 3 Superseded Latest is v3 →

lore
Share & tools
lore

Kenta Tada

author
#0 lore raw

PATCH v2 1/1 bpftool: Query only cgroup-related attach types

When CONFIG_NETKIT=y,
bpftool-cgroup shows error even if the cgroup's path is correct:

$ bpftool cgroup tree /sys/fs/cgroup
CgroupPath
ID       AttachType      AttachFlags     Name
Error: can't query bpf programs attached to /sys/fs/cgroup: No such device or address

From strace and kernel tracing, I found netkit returned ENXIO and this command failed. I think this AttachType(BPF_NETKIT_PRIMARY) is not relevant to cgroup.

bpftool-cgroup should query just only cgroup-related attach types.

v1->v2:
- used an array of cgroup attach types

---
 tools/bpf/bpftool/cgroup.c | 38 +++++++++++++++++++++++++++++++++++---
 1 file changed, 35 insertions(+), 3 deletions(-)
tools/bpf/bpftool/cgroup.c +35 -3
index af6898c0f388..afab728468bf 100644
@@ -19,6 +19,38 @@
 
 #include "main.h"
 
+static const int cgroup_attach_types[] = {
+	BPF_CGROUP_INET_INGRESS,
+	BPF_CGROUP_INET_EGRESS,
+	BPF_CGROUP_INET_SOCK_CREATE,
+	BPF_CGROUP_INET_SOCK_RELEASE,
+	BPF_CGROUP_INET4_BIND,
+	BPF_CGROUP_INET6_BIND,
+	BPF_CGROUP_INET4_POST_BIND,
+	BPF_CGROUP_INET6_POST_BIND,
+	BPF_CGROUP_INET4_CONNECT,
+	BPF_CGROUP_INET6_CONNECT,
+	BPF_CGROUP_UNIX_CONNECT,
+	BPF_CGROUP_INET4_GETPEERNAME,
+	BPF_CGROUP_INET6_GETPEERNAME,
+	BPF_CGROUP_UNIX_GETPEERNAME,
+	BPF_CGROUP_INET4_GETSOCKNAME,
+	BPF_CGROUP_INET6_GETSOCKNAME,
+	BPF_CGROUP_UNIX_GETSOCKNAME,
+	BPF_CGROUP_UDP4_SENDMSG,
+	BPF_CGROUP_UDP6_SENDMSG,
+	BPF_CGROUP_UNIX_SENDMSG,
+	BPF_CGROUP_UDP4_RECVMSG,
+	BPF_CGROUP_UDP6_RECVMSG,
+	BPF_CGROUP_UNIX_RECVMSG,
+	BPF_CGROUP_SOCK_OPS,
+	BPF_CGROUP_DEVICE,
+	BPF_CGROUP_SYSCTL,
+	BPF_CGROUP_GETSOCKOPT,
+	BPF_CGROUP_SETSOCKOPT,
+	BPF_LSM_CGROUP
+};
+
 #define HELP_SPEC_ATTACH_FLAGS						\
 	"ATTACH_FLAGS := { multi | override }"
 
@@ -183,11 +215,11 @@ static int count_attached_bpf_progs(int cgroup_fd, enum bpf_attach_type type)
 
 static int cgroup_has_attached_progs(int cgroup_fd)
 {
-	enum bpf_attach_type type;
+	unsigned int i = 0;
 	bool no_prog = true;
 
-	for (type = 0; type < __MAX_BPF_ATTACH_TYPE; type++) {
-		int count = count_attached_bpf_progs(cgroup_fd, type);
+	for (i = 0; i < ARRAY_SIZE(cgroup_attach_types); i++) {
+		int count = count_attached_bpf_progs(cgroup_fd, cgroup_attach_types[i]);
 
 		if (count < 0 && errno != EINVAL)
 			return -1;
Signature · 1 line
2.43.0
maintainer
#1 lore raw

2024-06-07 11:21 UTC+0100 ~ Kenta Tada <[email protected]>

Quoted 65 lines

When CONFIG_NETKIT=y,
bpftool-cgroup shows error even if the cgroup's path is correct:

$ bpftool cgroup tree /sys/fs/cgroup
CgroupPath
ID       AttachType      AttachFlags     Name
Error: can't query bpf programs attached to /sys/fs/cgroup: No such device or address

From strace and kernel tracing, I found netkit returned ENXIO and this command failed. I think this AttachType(BPF_NETKIT_PRIMARY) is not relevant to cgroup.

bpftool-cgroup should query just only cgroup-related attach types.

v1->v2:
- used an array of cgroup attach types

---
 tools/bpf/bpftool/cgroup.c | 38 +++++++++++++++++++++++++++++++++++---
 1 file changed, 35 insertions(+), 3 deletions(-)
tools/bpf/bpftool/cgroup.c +32 -0
index af6898c0f388..afab728468bf 100644
@@ -19,6 +19,38 @@
 
 #include "main.h"
 
+static const int cgroup_attach_types[] = {
+	BPF_CGROUP_INET_INGRESS,
+	BPF_CGROUP_INET_EGRESS,
+	BPF_CGROUP_INET_SOCK_CREATE,
+	BPF_CGROUP_INET_SOCK_RELEASE,
+	BPF_CGROUP_INET4_BIND,
+	BPF_CGROUP_INET6_BIND,
+	BPF_CGROUP_INET4_POST_BIND,
+	BPF_CGROUP_INET6_POST_BIND,
+	BPF_CGROUP_INET4_CONNECT,
+	BPF_CGROUP_INET6_CONNECT,
+	BPF_CGROUP_UNIX_CONNECT,
+	BPF_CGROUP_INET4_GETPEERNAME,
+	BPF_CGROUP_INET6_GETPEERNAME,
+	BPF_CGROUP_UNIX_GETPEERNAME,
+	BPF_CGROUP_INET4_GETSOCKNAME,
+	BPF_CGROUP_INET6_GETSOCKNAME,
+	BPF_CGROUP_UNIX_GETSOCKNAME,
+	BPF_CGROUP_UDP4_SENDMSG,
+	BPF_CGROUP_UDP6_SENDMSG,
+	BPF_CGROUP_UNIX_SENDMSG,
+	BPF_CGROUP_UDP4_RECVMSG,
+	BPF_CGROUP_UDP6_RECVMSG,
+	BPF_CGROUP_UNIX_RECVMSG,
+	BPF_CGROUP_SOCK_OPS,
+	BPF_CGROUP_DEVICE,
+	BPF_CGROUP_SYSCTL,
+	BPF_CGROUP_GETSOCKOPT,
+	BPF_CGROUP_SETSOCKOPT,
+	BPF_LSM_CGROUP
+};
+
 #define HELP_SPEC_ATTACH_FLAGS						\
 	"ATTACH_FLAGS := { multi | override }"
 

tools/bpf/bpftool/cgroup.c static int count_attached_bpf_progs(int cgroup_fd, enum bpf_attach_type type) · line 221

Quoted 10 lines
tools/bpf/bpftool/cgroup.c +2 -3
@@ -183,11 +215,11 @@ static int count_attached_bpf_progs(int cgroup_fd, enum bpf_attach_type type)
 
 static int cgroup_has_attached_progs(int cgroup_fd)
 {
-	enum bpf_attach_type type;
+	unsigned int i = 0;
 	bool no_prog = true;
 
-	for (type = 0; type < __MAX_BPF_ATTACH_TYPE; type++) {
-		int count = count_attached_bpf_progs(cgroup_fd, type);
+	for (i = 0; i < ARRAY_SIZE(cgroup_attach_types); i++) {

Thanks, it looks better that way.

tools/bpf/bpftool/cgroup.c static int count_attached_bpf_progs(int cgroup_fd, enum bpf_attach_type type) · line 224

Quoted 3 lines
tools/bpf/bpftool/cgroup.c +1 -0
+		int count = count_attached_bpf_progs(cgroup_fd, cgroup_attach_types[i]);
 
 		if (count < 0 && errno != EINVAL)

I think the "errno != EINVAL" exception was here to allow iterating over unsupported attach types for the queries. Now that we only do supported types, we can probably remove it and return if "(count < 0)".

Quoted 1 line
tools/bpf/bpftool/cgroup.c +0 -0
 			return -1;
author
#2 lore raw
Quoted 81 lines · Quentin Monnet

2024-06-07 11:21 UTC+0100 ~ Kenta Tada <[email protected]>

Quoted 65 lines

When CONFIG_NETKIT=y,
bpftool-cgroup shows error even if the cgroup's path is correct:

$ bpftool cgroup tree /sys/fs/cgroup
CgroupPath
ID       AttachType      AttachFlags     Name
Error: can't query bpf programs attached to /sys/fs/cgroup: No such device or address

From strace and kernel tracing, I found netkit returned ENXIO and this command failed. I think this AttachType(BPF_NETKIT_PRIMARY) is not relevant to cgroup.

bpftool-cgroup should query just only cgroup-related attach types.

v1->v2:
- used an array of cgroup attach types

---
 tools/bpf/bpftool/cgroup.c | 38 +++++++++++++++++++++++++++++++++++---
 1 file changed, 35 insertions(+), 3 deletions(-)
tools/bpf/bpftool/cgroup.c +32 -0
index af6898c0f388..afab728468bf 100644
@@ -19,6 +19,38 @@
 
 #include "main.h"
 
+static const int cgroup_attach_types[] = {
+	BPF_CGROUP_INET_INGRESS,
+	BPF_CGROUP_INET_EGRESS,
+	BPF_CGROUP_INET_SOCK_CREATE,
+	BPF_CGROUP_INET_SOCK_RELEASE,
+	BPF_CGROUP_INET4_BIND,
+	BPF_CGROUP_INET6_BIND,
+	BPF_CGROUP_INET4_POST_BIND,
+	BPF_CGROUP_INET6_POST_BIND,
+	BPF_CGROUP_INET4_CONNECT,
+	BPF_CGROUP_INET6_CONNECT,
+	BPF_CGROUP_UNIX_CONNECT,
+	BPF_CGROUP_INET4_GETPEERNAME,
+	BPF_CGROUP_INET6_GETPEERNAME,
+	BPF_CGROUP_UNIX_GETPEERNAME,
+	BPF_CGROUP_INET4_GETSOCKNAME,
+	BPF_CGROUP_INET6_GETSOCKNAME,
+	BPF_CGROUP_UNIX_GETSOCKNAME,
+	BPF_CGROUP_UDP4_SENDMSG,
+	BPF_CGROUP_UDP6_SENDMSG,
+	BPF_CGROUP_UNIX_SENDMSG,
+	BPF_CGROUP_UDP4_RECVMSG,
+	BPF_CGROUP_UDP6_RECVMSG,
+	BPF_CGROUP_UNIX_RECVMSG,
+	BPF_CGROUP_SOCK_OPS,
+	BPF_CGROUP_DEVICE,
+	BPF_CGROUP_SYSCTL,
+	BPF_CGROUP_GETSOCKOPT,
+	BPF_CGROUP_SETSOCKOPT,
+	BPF_LSM_CGROUP
+};
+
 #define HELP_SPEC_ATTACH_FLAGS						\
 	"ATTACH_FLAGS := { multi | override }"
 

tools/bpf/bpftool/cgroup.c static int count_attached_bpf_progs(int cgroup_fd, enum bpf_attach_type type) · line 221

Quoted 10 lines
tools/bpf/bpftool/cgroup.c +2 -3
@@ -183,11 +215,11 @@ static int count_attached_bpf_progs(int cgroup_fd, enum bpf_attach_type type)
 
 static int cgroup_has_attached_progs(int cgroup_fd)
 {
-	enum bpf_attach_type type;
+	unsigned int i = 0;
 	bool no_prog = true;
 
-	for (type = 0; type < __MAX_BPF_ATTACH_TYPE; type++) {
-		int count = count_attached_bpf_progs(cgroup_fd, type);
+	for (i = 0; i < ARRAY_SIZE(cgroup_attach_types); i++) {

Thanks, it looks better that way.

Quoted 8 lines

tools/bpf/bpftool/cgroup.c static int count_attached_bpf_progs(int cgroup_fd, enum bpf_attach_type type) · line 224

Quoted 3 lines
tools/bpf/bpftool/cgroup.c +1 -0
+		int count = count_attached_bpf_progs(cgroup_fd, cgroup_attach_types[i]);
 
 		if (count < 0 && errno != EINVAL)

I think the "errno != EINVAL" exception was here to allow iterating over unsupported attach types for the queries. Now that we only do supported types, we can probably remove it and return if "(count < 0)".

OK.
I'll fix it quickly.
Thanks!

Quoted 4 lines
Quoted 3 lines
tools/bpf/bpftool/cgroup.c +0 -0
 			return -1;