bpf: Force to MPTCP
Subsystems (9)
v11 cover bpf: Force to MPTCP
As is described in the "How to use MPTCP?" section in MPTCP wiki [1]:
"Your app should create sockets with IPPROTO_MPTCP as the proto:
( socket(AF_INET, SOCK_STREAM, IPPROTO_MPTCP); ). Legacy apps can be forced to create and use MPTCP sockets instead of TCP ones via the mptcpize command bundled with the mptcpd daemon."
But the mptcpize (LD_PRELOAD technique) command has some limitations [2]:
- it doesn't work if the application is not using libc (e.g. GoLang
apps)
- in some envs, it might not be easy to set env vars / change the way
apps are launched, e.g. on Android
- mptcpize needs to be launched with all apps that want MPTCP: we could
have more control from BPF to enable MPTCP only for some apps or all the ones of a netns or a cgroup, etc.
- it is not in BPF, we cannot talk about it at netdev conf.
So this patchset attempts to use BPF to implement functions similer to mptcpize.
The main idea is to add a hook in sys_socket() to change the protocol id from IPPROTO_TCP (or 0) to IPPROTO_MPTCP.
[1]
https://github.com/multipath-tcp/mptcp_net-next/wiki
[2]
https://github.com/multipath-tcp/mptcp_net-next/issues/79
v11:
- add comments about outputs of 'ss' and 'nstat'.
- use "err = verify_mptcpify()" instead of using =+.
v10:
- drop "#ifdef CONFIG_BPF_JIT".
- include vmlinux.h and bpf_tracing_net.h to avoid defining some
macros.
- drop unneeded checks for mptcp.
v9:
- update comment for 'update_socket_protocol'.
v8:
- drop the additional checks on the 'protocol' value after the
'update_socket_protocol()' call.
v7:
- add __weak and __diag_* for update_socket_protocol.
v6:
- add update_socket_protocol.
v5:
- add bpf_mptcpify helper.
v4:
- use lsm_cgroup/socket_create
v3:
- patch 8: char cmd[128]; -> char cmd[256];
v2:
- Fix build selftests errors reported by CI
Geliang Tang (5):
bpf: Add update_socket_protocol hook
selftests/bpf: Use random netns name for mptcp
selftests/bpf: Add two mptcp netns helpers
selftests/bpf: Drop unneeded checks for mptcp
selftests/bpf: Add mptcpify test
net/mptcp/bpf.c | 15 ++ net/socket.c | 24 +++ .../testing/selftests/bpf/prog_tests/mptcp.c | 139 +++++++++++++++--- tools/testing/selftests/bpf/progs/mptcpify.c | 20 +++ 4 files changed, 179 insertions(+), 19 deletions(-) create mode 100644 tools/testing/selftests/bpf/progs/mptcpify.c
Signature · 1 line
2.35.3
PATCH v11 1/5 bpf: Add update_socket_protocol hook
Add a hook named update_socket_protocol in __sys_socket(), for bpf progs to attach to and update socket protocol. One user case is to force legacy TCP apps to create and use MPTCP sockets instead of TCP ones.
Define a mod_ret set named bpf_mptcp_fmodret_ids, add the hook update_socket_protocol into this set, and register it in bpf_mptcp_kfunc_init().
--- net/mptcp/bpf.c | 15 +++++++++++++++ net/socket.c | 24 ++++++++++++++++++++++++ 2 files changed, 39 insertions(+)
net/mptcp/bpf.c +15 -0
@@ -19,3 +19,18 @@ struct mptcp_sock *bpf_mptcp_sock_from_subflow(struct sock *sk)return NULL ; } + +BTF_SET8_START (bpf_mptcp_fmodret_ids) +BTF_ID_FLAGS (func ,update_socket_protocol ) +BTF_SET8_END (bpf_mptcp_fmodret_ids) + +static const struct btf_kfunc_id_set bpf_mptcp_fmodret_set = { + .owner =THIS_MODULE , + .set = &bpf_mptcp_fmodret_ids, +}; + +static int __init bpf_mptcp_kfunc_init (void ) +{ +return register_btf_fmodret_id_set (&bpf_mptcp_fmodret_set); +} +late_initcall (bpf_mptcp_kfunc_init);
net/socket.c +24 -0
@@ -1644,11 +1644,35 @@ struct file *__sys_socket_file(int family, int type, int protocol)return sock_alloc_file (sock, flags,NULL ); } +/* A hook for bpf progs to attach to and update socket protocol. +* +* A static noinline declaration here could cause the compiler to +* optimize away the function. A global noinline declaration will +* keep the definition, but may optimize away the callsite. +* Therefore, __weak is needed to ensure that the call is still +* emitted, by telling the compiler that we don't know what the +* function might eventually be. +* +* __diag_* below are needed to dismiss the missing prototype warning. +*/ + +__diag_push (); +__diag_ignore_all ("-Wmissing-prototypes" , +"kfuncs which will be used in BPF programs" ); + +__weak noinline int update_socket_protocol(int family, int type, int protocol) +{ +return protocol; +} + +__diag_pop (); +int __sys_socket (int family,int type,int protocol) {struct socket *sock;int flags; + protocol =update_socket_protocol (family, type, protocol); sock =__sys_socket_create (family, type, protocol);if (IS_ERR (sock))return PTR_ERR (sock);
Signature · 1 line
2.35.3
Martin KaFai Lau
reviewerRe: [PATCH bpf-next v11 1/5] bpf: Add update_socket_protocol hook
Quoted 54 lines · Geliang Tang
Add a hook named
update_socket_protocolin__sys_socket(), for bpf progs to attach to and update socket protocol. One user case is to force legacy TCP apps to create and use MPTCP sockets instead of TCP ones.Define a
mod_retset namedbpf_mptcp_fmodret_ids, add the hookupdate_socket_protocolinto this set, and register it inbpf_mptcp_kfunc_init().Acked-by Matthieu Baerts <[email protected]>Acked-by Yonghong Song <[email protected]>Signed-off-by Geliang Tang <[email protected]>--- net/mptcp/bpf.c | 15 +++++++++++++++ net/socket.c | 24 ++++++++++++++++++++++++ 2 files changed, 39 insertions(+)net/mptcp/bpf.c +15
-0@@ -19,3 +19,18 @@ struct mptcp_sock *bpf_mptcp_sock_from_subflow(struct sock *sk)return NULL ; } + +BTF_SET8_START (bpf_mptcp_fmodret_ids) +BTF_ID_FLAGS (func ,update_socket_protocol ) +BTF_SET8_END (bpf_mptcp_fmodret_ids) + +static const struct btf_kfunc_id_set bpf_mptcp_fmodret_set = { + .owner =THIS_MODULE , + .set = &bpf_mptcp_fmodret_ids, +}; + +static int __init bpf_mptcp_kfunc_init (void ) +{ +return register_btf_fmodret_id_set (&bpf_mptcp_fmodret_set); +} +late_initcall (bpf_mptcp_kfunc_init);net/socket.c +5
-0@@ -1644,11 +1644,35 @@ struct file *__sys_socket_file(int family, int type, int protocol)return sock_alloc_file (sock, flags,NULL ); } +/* A hook for bpf progs to attach to and update socket protocol. +* +* A static noinline declaration here could cause the compiler to +* optimize away the function. A global noinline declaration will +* keep the definition, but may optimize away the callsite.
net/socket.c struct file *__sys_socket_file(int family, int type, int protocol) · line 1661
Quoted 10 lines
net/socket.c +10
-0+* Therefore, __weak is needed to ensure that the call is still +* emitted, by telling the compiler that we don't know what the +* function might eventually be. +* +* __diag_* below are needed to dismiss the missing prototype warning. +*/ + +__diag_push (); +__diag_ignore_all ("-Wmissing-prototypes" , +"kfuncs which will be used in BPF programs" );
This "kfuns which will be used in BPF programs" piece is not accurate. It is a fmod_ret entry point for bpf prog.
Quoted 4 lines
net/socket.c +4
-0+ +__weak noinline int update_socket_protocol(int family, int type, int protocol) +{ +return protocol;
net/socket.c struct file *__sys_socket_file(int family, int type, int protocol) · line 1675
Quoted 10 lines
net/socket.c +5
-0+} + +__diag_pop (); +int __sys_socket (int family,int type,int protocol) {struct socket *sock;int flags; + protocol =update_socket_protocol (family, type, protocol);
Paolo, could you help to take another look and ack this patch if it has addressed your earlier comment ?
Quoted 3 lines
net/socket.c +0
-0sock =__sys_socket_create (family, type, protocol);if (IS_ERR (sock))return PTR_ERR (sock);
PATCH v11 2/5 selftests/bpf: Use random netns name for mptcp
In reply to Geliang TangUse rand() to generate a random netns name instead of using the fixed name "mptcp_ns" for every test.
By doing that, we can re-launch the test even if there was an issue removing the previous netns or if by accident, a netns with this generic name already existed on the system.
Note that using a different name each will also help adding more subtests in future commits.
--- tools/testing/selftests/bpf/prog_tests/mptcp.c | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-)
tools/testing/selftests/bpf/prog_tests/mptcp.c +3 -1
@@ -7,7 +7,7 @@#include "network_helpers.h" #include "mptcp_sock.skel.h" -#define NS_TEST "mptcp_ns" +char NS_TEST [32 ];#ifndef TCP_CA_NAME_MAX #define TCP_CA_NAME_MAX 16
@@ -147,6 +147,8 @@ static void test_base(void)if (!ASSERT_GE (cgroup_fd,0 ,"test__join_cgroup" ))return ; +srand (time (NULL )); +snprintf (NS_TEST ,sizeof (NS_TEST ),"mptcp_ns_%d" ,rand ());SYS (fail,"ip netns add %s" ,NS_TEST );SYS (fail,"ip -net %s link set dev lo up" ,NS_TEST );
Signature · 1 line
2.35.3
Martin KaFai Lau
reviewerRe: [PATCH bpf-next v11 2/5] selftests/bpf: Use random netns name for mptcp
Quoted 9 lines · Geliang Tang
Use
rand()to generate a random netns name instead of using the fixed name "mptcp_ns" for every test.By doing that, we can re-launch the test even if there was an issue removing the previous netns or if by accident, a netns with this generic name already existed on the system.
Note that using a different name each will also help adding more subtests in future commits.
I run test_progs repeatedly without rebooting qemu to save time. If there is a test did not clean up its netns, I would rather uncover it earlier and fix it instead. Randomizing the name is hiding the issue and does not help to uncover the broken test sooner. Although this change is to mptcp test alone, this could be referred in other future tests.
afaik, I don't remember bpf CI ever run into a test failure because the picked name had already been used by the system. It seems you ran into this issue a lot with the mptcp test in your setup. Could you explain a little more?
Re: [PATCH bpf-next v11 2/5] selftests/bpf: Use random netns name for mptcp
Quoted 10 lines · Martin KaFai Lau
Quoted 9 lines · Geliang Tang
Use
rand()to generate a random netns name instead of using the fixed name "mptcp_ns" for every test.By doing that, we can re-launch the test even if there was an issue removing the previous netns or if by accident, a netns with this generic name already existed on the system.
Note that using a different name each will also help adding more subtests in future commits.
Hi Martin,
I tried to run mptcp tests simultaneously, and got "Cannot create namespace file "/var/run/netns/mptcp_ns": File exists" errors sometimes. So I add this patch to fix it.
It's easy to reproduce, just run this commands in multiple terminals:
Quoted 8 lines
for i in
seq 1 100; do sudo ./test_progs -t mptcp; doneI run
test_progsrepeatedly without rebooting qemu to save time. If there is a test did not clean up its netns, I would rather uncover it earlier and fix it instead. Randomizing the name is hiding the issue and does not help to uncover the broken test sooner. Although this change is to mptcp test alone, this could be referred in other future tests.
I added "ip netns show" after "ip netns del" in v12 to check if there is a test did not clean up its netns.
Thanks,
-Geliang
Quoted 5 lines
afaik, I don't remember bpf CI ever run into a test failure because the picked name had already been used by the system. It seems you ran into this issue a lot with the mptcp test in your setup. Could you explain a little more?
Martin KaFai Lau
reviewerRe: [PATCH bpf-next v11 2/5] selftests/bpf: Use random netns name for mptcp
Quoted 12 lines · Geliang Tang
Quoted 10 lines · Martin KaFai Lau
Quoted 9 lines · Geliang Tang
Use
rand()to generate a random netns name instead of using the fixed name "mptcp_ns" for every test.By doing that, we can re-launch the test even if there was an issue removing the previous netns or if by accident, a netns with this generic name already existed on the system.
Note that using a different name each will also help adding more subtests in future commits.
Quoted 8 lines
Hi Martin,
I tried to run mptcp tests simultaneously, and got "Cannot create namespace file "
/var/run/netns/mptcp_ns": File exists" errors sometimes. So I add this patch to fix it.It's easy to reproduce, just run this commands in multiple terminals:
Quoted 1 line
for i in
seq 1 100; do sudo ./test_progs -t mptcp; done
Not only the "-t mptcp" test. Other tests in test_progs also don't support running parallel in multiple terminals. Does it really help to test the bpf part of the prog_tests/mptcp.c test by running like this? If it wants to exercise the other mptcp networking specific code like this, a separate mptcp test is needed outside of test_progs and it won't be run in the bpf CI.
If you agree, can you please avoid introducing unnecessary randomness to the test_progs where bpf CI and most users don't run in this way?
Also, please don't resend the patches too fast until the discussion is concluded. Please give reasonable time for others to reply.
I have a high level question. In LPC 2022 (https://lpc.events/event/16/contributions/1354/), I recall there was idea in using bpf to make other mptcp decision/policy. Any thought and progress on this? This set which only uses bpf to change the protocol feels like an incomplete solution.
Re: [PATCH bpf-next v11 2/5] selftests/bpf: Use random netns name for mptcp
Quoted 29 lines · Martin KaFai Lau
Quoted 12 lines · Geliang Tang
Quoted 10 lines · Martin KaFai Lau
Quoted 9 lines · Geliang Tang
Use
rand()to generate a random netns name instead of using the fixed name "mptcp_ns" for every test.By doing that, we can re-launch the test even if there was an issue removing the previous netns or if by accident, a netns with this generic name already existed on the system.
Note that using a different name each will also help adding more subtests in future commits.
Quoted 8 lines
Hi Martin,
I tried to run mptcp tests simultaneously, and got "Cannot create namespace file "
/var/run/netns/mptcp_ns": File exists" errors sometimes. So I add this patch to fix it.It's easy to reproduce, just run this commands in multiple terminals:
Quoted 1 line
for i in
seq 1 100; do sudo ./test_progs -t mptcp; doneNot only the "-t mptcp" test. Other tests in
test_progsalso don't support running parallel in multiple terminals. Does it really help to test the bpf part of theprog_tests/mptcp.ctest by running like this? If it wants to exercise the other mptcp networking specific code like this, a separate mptcp test is needed outside oftest_progsand it won't be run in the bpf CI.
Quoted 2 lines
If you agree, can you please avoid introducing unnecessary randomness to the
test_progswhere bpf CI and most users don't run in this way?
Thanks Martin. Sure, I agree. Let's drop this patch.
Quoted 3 lines
Also, please don't resend the patches too fast until the discussion is concluded. Please give reasonable time for others to reply.
Sure. Please give me a clear reminder next time that it's time to resend a new version of the patches.
Quoted 6 lines
I have a high level question. In LPC 2022
(https://lpc.events/event/16/contributions/1354/), I recall there was idea in using bpf to make other mptcp decision/policy. Any thought and progress on this? This set which only uses bpf to change the protocol feels like an incomplete solution.
We are implementing MPTCP packet scheduler using BPF. Patches aren't sent to BPF mail list yet, only temporarily on our mptcp repo[1].
Here are the patches:
selftests/bpf: Add bpf_burst test
selftests/bpf: Add bpf_burst scheduler
bpf: Export more bpf_burst related functions
selftests/bpf: Add bpf_red test
selftests/bpf: Add bpf_red scheduler
selftests/bpf: Add bpf_rr test
selftests/bpf: Add bpf_rr scheduler
selftests/bpf: Add bpf_bkup test
selftests/bpf: Add bpf_bkup scheduler
selftests/bpf: Add bpf_first test
selftests/bpf: Add bpf_first scheduler
selftests/bpf: Add bpf scheduler test
selftests/bpf: add two mptcp netns helpers
selftests/bpf: use random netns name for mptcp
selftests/bpf: Add mptcp sched structs
bpf: Add bpf_mptcp_sched_kfunc_set
bpf: Add bpf_mptcp_sched_ops
If you could take a look at these patches in advance, I would greatly appreciate it. Any feedback is welcome.
[1]
https://github.com/multipath-tcp/mptcp_net-next.git
-Geliang
Quoted 1 line
Martin KaFai Lau
reviewerRe: [PATCH bpf-next v11 2/5] selftests/bpf: Use random netns name for mptcp
Quoted 30 lines · Geliang Tang
Quoted 29 lines · Martin KaFai Lau
Quoted 12 lines · Geliang Tang
Quoted 10 lines · Martin KaFai Lau
Quoted 9 lines · Geliang Tang
Use
rand()to generate a random netns name instead of using the fixed name "mptcp_ns" for every test.By doing that, we can re-launch the test even if there was an issue removing the previous netns or if by accident, a netns with this generic name already existed on the system.
Note that using a different name each will also help adding more subtests in future commits.
Quoted 8 lines
Hi Martin,
I tried to run mptcp tests simultaneously, and got "Cannot create namespace file "
/var/run/netns/mptcp_ns": File exists" errors sometimes. So I add this patch to fix it.It's easy to reproduce, just run this commands in multiple terminals:
Quoted 1 line
for i in
seq 1 100; do sudo ./test_progs -t mptcp; doneNot only the "-t mptcp" test. Other tests in
test_progsalso don't support running parallel in multiple terminals. Does it really help to test the bpf part of theprog_tests/mptcp.ctest by running like this? If it wants to exercise the other mptcp networking specific code like this, a separate mptcp test is needed outside oftest_progsand it won't be run in the bpf CI.
Quoted 4 lines
Quoted 2 lines
If you agree, can you please avoid introducing unnecessary randomness to the
test_progswhere bpf CI and most users don't run in this way?Thanks Martin. Sure, I agree. Let's drop this patch.
Thanks you.
Quoted 29 lines
Quoted 5 lines
I have a high level question. In LPC 2022
(https://lpc.events/event/16/contributions/1354/), I recall there was idea in using bpf to make other mptcp decision/policy. Any thought and progress on this? This set which only uses bpf to change the protocol feels like an incomplete solution.We are implementing MPTCP packet scheduler using BPF. Patches aren't sent to BPF mail list yet, only temporarily on our mptcp repo[1].
Here are the patches:
selftests/bpf: Add
bpf_bursttest
selftests/bpf: Addbpf_burstscheduler
bpf: Export morebpf_burstrelated functions
selftests/bpf: Addbpf_redtest
selftests/bpf: Addbpf_redscheduler
selftests/bpf: Addbpf_rrtest
selftests/bpf: Addbpf_rrscheduler
selftests/bpf: Addbpf_bkuptest
selftests/bpf: Addbpf_bkupscheduler
selftests/bpf: Addbpf_firsttest
selftests/bpf: Addbpf_firstscheduler
selftests/bpf: Add bpf scheduler test
selftests/bpf: add two mptcp netns helpers
selftests/bpf: use random netns name for mptcp
selftests/bpf: Add mptcp sched structs
bpf: Addbpf_mptcp_sched_kfunc_set
bpf: Addbpf_mptcp_sched_ops
Quoted 5 lines
If you could take a look at these patches in advance, I would greatly appreciate it. Any feedback is welcome.
Thanks for sharing. I did not go into the details. iiuc, the scheduler is specific to a namespace. Do you see if it is useful to have more finer control like depending on what IP address it is connected to? BPF policy is usually found more useful to have finer policy control than global or per-netns.
The same question goes for the fmod_ret here in this patch. The progs/mptcpify.c selftest is as good as upgrading all TCP connections. Is it your only use case and no need for finer selection?
Re: [PATCH bpf-next v11 2/5] selftests/bpf: Use random netns name for mptcp
Quoted 79 lines · Martin KaFai Lau
Quoted 30 lines · Geliang Tang
Quoted 29 lines · Martin KaFai Lau
Quoted 12 lines · Geliang Tang
Quoted 10 lines · Martin KaFai Lau
Quoted 9 lines · Geliang Tang
Use
rand()to generate a random netns name instead of using the fixed name "mptcp_ns" for every test.By doing that, we can re-launch the test even if there was an issue removing the previous netns or if by accident, a netns with this generic name already existed on the system.
Note that using a different name each will also help adding more subtests in future commits.
Quoted 8 lines
Hi Martin,
I tried to run mptcp tests simultaneously, and got "Cannot create namespace file "
/var/run/netns/mptcp_ns": File exists" errors sometimes. So I add this patch to fix it.It's easy to reproduce, just run this commands in multiple terminals:
Quoted 1 line
for i in
seq 1 100; do sudo ./test_progs -t mptcp; doneNot only the "-t mptcp" test. Other tests in
test_progsalso don't support running parallel in multiple terminals. Does it really help to test the bpf part of theprog_tests/mptcp.ctest by running like this? If it wants to exercise the other mptcp networking specific code like this, a separate mptcp test is needed outside oftest_progsand it won't be run in the bpf CI.Quoted 4 lines
Quoted 2 lines
If you agree, can you please avoid introducing unnecessary randomness to the
test_progswhere bpf CI and most users don't run in this way?Thanks Martin. Sure, I agree. Let's drop this patch.
Thanks you.
Quoted 29 lines
Quoted 5 lines
I have a high level question. In LPC 2022
(https://lpc.events/event/16/contributions/1354/), I recall there was idea in using bpf to make other mptcp decision/policy. Any thought and progress on this? This set which only uses bpf to change the protocol feels like an incomplete solution.We are implementing MPTCP packet scheduler using BPF. Patches aren't sent to BPF mail list yet, only temporarily on our mptcp repo[1].
Here are the patches:
selftests/bpf: Add
bpf_bursttest
selftests/bpf: Addbpf_burstscheduler
bpf: Export morebpf_burstrelated functions
selftests/bpf: Addbpf_redtest
selftests/bpf: Addbpf_redscheduler
selftests/bpf: Addbpf_rrtest
selftests/bpf: Addbpf_rrscheduler
selftests/bpf: Addbpf_bkuptest
selftests/bpf: Addbpf_bkupscheduler
selftests/bpf: Addbpf_firsttest
selftests/bpf: Addbpf_firstscheduler
selftests/bpf: Add bpf scheduler test
selftests/bpf: add two mptcp netns helpers
selftests/bpf: use random netns name for mptcp
selftests/bpf: Add mptcp sched structs
bpf: Addbpf_mptcp_sched_kfunc_set
bpf: Addbpf_mptcp_sched_opsQuoted 5 lines
If you could take a look at these patches in advance, I would greatly appreciate it. Any feedback is welcome.
Thanks for sharing. I did not go into the details. iiuc, the scheduler is specific to a namespace. Do you see if it is useful to have more finer control like depending on what IP address it is connected to? BPF policy is usually found more useful to have finer policy control than global or per-netns.
Quoted 3 lines
The same question goes for the
fmod_rethere in this patch. Theprogs/mptcpify.cselftest is as good as upgrading all TCP connections. Is it your only use case and no need for finer selection?
This per-netns control is just the first step. We do need finer selection. The most ideal mode is to select one app to upgrade it's TCP connections only. So per-cgroup control is much better than per-netns. But we haven't found a good per-cgroup solution yet.
Thanks,
-Geliang
Quoted 1 line
Martin KaFai Lau
reviewerRe: [PATCH bpf-next v11 2/5] selftests/bpf: Use random netns name for mptcp
Quoted 80 lines · Geliang Tang
Quoted 79 lines · Martin KaFai Lau
Quoted 30 lines · Geliang Tang
Quoted 29 lines · Martin KaFai Lau
Quoted 12 lines · Geliang Tang
Quoted 10 lines · Martin KaFai Lau
Quoted 9 lines · Geliang Tang
Use
rand()to generate a random netns name instead of using the fixed name "mptcp_ns" for every test.By doing that, we can re-launch the test even if there was an issue removing the previous netns or if by accident, a netns with this generic name already existed on the system.
Note that using a different name each will also help adding more subtests in future commits.
Quoted 8 lines
Hi Martin,
I tried to run mptcp tests simultaneously, and got "Cannot create namespace file "
/var/run/netns/mptcp_ns": File exists" errors sometimes. So I add this patch to fix it.It's easy to reproduce, just run this commands in multiple terminals:
Quoted 1 line
for i in
seq 1 100; do sudo ./test_progs -t mptcp; doneNot only the "-t mptcp" test. Other tests in
test_progsalso don't support running parallel in multiple terminals. Does it really help to test the bpf part of theprog_tests/mptcp.ctest by running like this? If it wants to exercise the other mptcp networking specific code like this, a separate mptcp test is needed outside oftest_progsand it won't be run in the bpf CI.Quoted 4 lines
Quoted 2 lines
If you agree, can you please avoid introducing unnecessary randomness to the
test_progswhere bpf CI and most users don't run in this way?Thanks Martin. Sure, I agree. Let's drop this patch.
Thanks you.
Quoted 29 lines
Quoted 5 lines
I have a high level question. In LPC 2022
(https://lpc.events/event/16/contributions/1354/), I recall there was idea in using bpf to make other mptcp decision/policy. Any thought and progress on this? This set which only uses bpf to change the protocol feels like an incomplete solution.We are implementing MPTCP packet scheduler using BPF. Patches aren't sent to BPF mail list yet, only temporarily on our mptcp repo[1].
Here are the patches:
selftests/bpf: Add
bpf_bursttest
selftests/bpf: Addbpf_burstscheduler
bpf: Export morebpf_burstrelated functions
selftests/bpf: Addbpf_redtest
selftests/bpf: Addbpf_redscheduler
selftests/bpf: Addbpf_rrtest
selftests/bpf: Addbpf_rrscheduler
selftests/bpf: Addbpf_bkuptest
selftests/bpf: Addbpf_bkupscheduler
selftests/bpf: Addbpf_firsttest
selftests/bpf: Addbpf_firstscheduler
selftests/bpf: Add bpf scheduler test
selftests/bpf: add two mptcp netns helpers
selftests/bpf: use random netns name for mptcp
selftests/bpf: Add mptcp sched structs
bpf: Addbpf_mptcp_sched_kfunc_set
bpf: Addbpf_mptcp_sched_opsQuoted 5 lines
If you could take a look at these patches in advance, I would greatly appreciate it. Any feedback is welcome.
Thanks for sharing. I did not go into the details. iiuc, the scheduler is specific to a namespace. Do you see if it is useful to have more finer control like depending on what IP address it is connected to? BPF policy is usually found more useful to have finer policy control than global or per-netns.
Quoted 8 lines
Quoted 3 lines
The same question goes for the
fmod_rethere in this patch. Theprogs/mptcpify.cselftest is as good as upgrading all TCP connections. Is it your only use case and no need for finer selection?This per-netns control is just the first step. We do need finer selection. The most ideal mode is to select one app to upgrade it's TCP connections only. So per-cgroup control is much better than per-netns. But we haven't found a good per-cgroup solution yet.
Selecting an app or cgroup can sort of be done by getting the current task or current cgroup (there is helper to do that). I am imagining eventually it will want to decide the protocol upgrade and/or the mptcp-scheduler when the destination IP is decided. This fmod_ret upgrade for all acts like a global knob (sysctl) and feels like a hack or at least incomplete. However, I also don't see a clean way to do that for now in the current shape.
Please respin another revision to address the earlier selftest comment on the netns name. Thanks.
Re: [PATCH bpf-next v11 2/5] selftests/bpf: Use random netns name for mptcp
In reply to Geliang TangHi Geliang,
(only with MPTCP list)
Quoted 30 lines · Geliang Tang
Quoted 29 lines · Martin KaFai Lau
Quoted 12 lines · Geliang Tang
Quoted 10 lines · Martin KaFai Lau
Quoted 9 lines · Geliang Tang
Use
rand()to generate a random netns name instead of using the fixed name "mptcp_ns" for every test.By doing that, we can re-launch the test even if there was an issue removing the previous netns or if by accident, a netns with this generic name already existed on the system.
Note that using a different name each will also help adding more subtests in future commits.
Quoted 8 lines
Hi Martin,
I tried to run mptcp tests simultaneously, and got "Cannot create namespace file "
/var/run/netns/mptcp_ns": File exists" errors sometimes. So I add this patch to fix it.It's easy to reproduce, just run this commands in multiple terminals:
Quoted 1 line
for i in
seq 1 100; do sudo ./test_progs -t mptcp; doneNot only the "-t mptcp" test. Other tests in
test_progsalso don't support running parallel in multiple terminals. Does it really help to test the bpf part of theprog_tests/mptcp.ctest by running like this? If it wants to exercise the other mptcp networking specific code like this, a separate mptcp test is needed outside oftest_progsand it won't be run in the bpf CI.
Quoted 4 lines
Quoted 2 lines
If you agree, can you please avoid introducing unnecessary randomness to the
test_progswhere bpf CI and most users don't run in this way?Thanks Martin. Sure, I agree. Let's drop this patch.
FYI, I'm also going to drop this patch from our tree as it doesn't seem needed and it causes conflicts with bpf-next (now in net-next).
If needed, we can still re-add it later.
Cheers,
Matt
Signature · 2 lines
Tessares | Belgium | Hybrid Access Solutions www.tessares.net
PATCH v11 3/5 selftests/bpf: Add two mptcp netns helpers
In reply to Geliang TangAdd two netns helpers for mptcp tests: create_netns() and cleanup_netns(). Use them in test_base().
These new helpers will be re-used in the following commits introducing new tests.
--- .../testing/selftests/bpf/prog_tests/mptcp.c | 35 ++++++++++++------- 1 file changed, 23 insertions(+), 12 deletions(-)
tools/testing/selftests/bpf/prog_tests/mptcp.c +23 -12
@@ -22,6 +22,26 @@ struct mptcp_storage {char ca_name[TCP_CA_NAME_MAX ]; }; +static struct nstoken *create_netns (void ) +{ +srand (time (NULL )); +snprintf (NS_TEST ,sizeof (NS_TEST ),"mptcp_ns_%d" ,rand ()); +SYS (fail,"ip netns add %s" ,NS_TEST ); +SYS (fail,"ip -net %s link set dev lo up" ,NS_TEST ); + +return open_netns (NS_TEST ); +fail: +return NULL ; +} + +static void cleanup_netns (struct nstoken *nstoken) +{ +if (nstoken) +close_netns (nstoken); + +SYS_NOFAIL ("ip netns del %s &> /dev/null" ,NS_TEST ); +} +static int verify_tsk (int map_fd,int client_fd) {int err, cfd = client_fd;
@@ -147,13 +167,8 @@ static void test_base(void)if (!ASSERT_GE (cgroup_fd,0 ,"test__join_cgroup" ))return ; -srand (time (NULL )); -snprintf (NS_TEST ,sizeof (NS_TEST ),"mptcp_ns_%d" ,rand ()); -SYS (fail,"ip netns add %s" ,NS_TEST ); -SYS (fail,"ip -net %s link set dev lo up" ,NS_TEST ); - - nstoken =open_netns (NS_TEST ); -if (!ASSERT_OK_PTR (nstoken,"open_netns" )) + nstoken =create_netns (); +if (!ASSERT_OK_PTR (nstoken,"create_netns" )) goto fail;/* without MPTCP */
@@ -176,11 +191,7 @@ static void test_base(void)close (server_fd); fail: -if (nstoken) -close_netns (nstoken); - -SYS_NOFAIL ("ip netns del " NS_TEST " &> /dev/null" ); - +cleanup_netns (nstoken);close (cgroup_fd); }
Signature · 1 line
2.35.3
PATCH v11 4/5 selftests/bpf: Drop unneeded checks for mptcp
In reply to Geliang TangLoad success means prog_fd and map_fd are always valid. So drop these unneeded ASSERT_GE checks for them in mptcp run_test().
--- tools/testing/selftests/bpf/prog_tests/mptcp.c | 10 ---------- 1 file changed, 10 deletions(-)
tools/testing/selftests/bpf/prog_tests/mptcp.c +0 -10
@@ -127,17 +127,7 @@ static int run_test(int cgroup_fd, int server_fd, bool is_mptcp) goto out; prog_fd =bpf_program__fd (sock_skel->progs._sockops); -if (!ASSERT_GE (prog_fd,0 ,"bpf_program__fd" )) { - err = -EIO ; - goto out; - } - map_fd =bpf_map__fd (sock_skel->maps.socket_storage_map); -if (!ASSERT_GE (map_fd,0 ,"bpf_map__fd" )) { - err = -EIO ; - goto out; - } - err =bpf_prog_attach (prog_fd, cgroup_fd,BPF_CGROUP_SOCK_OPS ,0 );if (!ASSERT_OK (err,"bpf_prog_attach" )) goto out;
Signature · 1 line
2.35.3
PATCH v11 5/5 selftests/bpf: Add mptcpify test
In reply to Geliang TangImplement a new test program mptcpify: if the family is AF_INET or AF_INET6, the type is SOCK_STREAM, and the protocol ID is 0 or IPPROTO_TCP, set it to IPPROTO_MPTCP. It will be hooked in update_socket_protocol().
Extend the MPTCP test base, add a selftest test_mptcpify() for the mptcpify case. Open and load the mptcpify test prog to mptcpify the TCP sockets dynamically, then use start_server() and connect_to_fd() to create a TCP socket, but actually what's created is an MPTCP socket, which can be verified through the outputs of 'ss' and 'nstat' commands.
--- .../testing/selftests/bpf/prog_tests/mptcp.c | 98 +++++++++++++++++++ tools/testing/selftests/bpf/progs/mptcpify.c | 20 ++++ 2 files changed, 118 insertions(+) create mode 100644 tools/testing/selftests/bpf/progs/mptcpify.c
tools/testing/selftests/bpf/prog_tests/mptcp.c +98 -0
@@ -6,6 +6,7 @@#include "cgroup_helpers.h" #include "network_helpers.h" #include "mptcp_sock.skel.h" +#include "mptcpify.skel.h" char NS_TEST [32 ];
@@ -185,8 +186,105 @@ static void test_base(void)close (cgroup_fd); } +static void send_byte (int fd) +{ +char b =0x55 ; + +ASSERT_EQ (write (fd, &b,sizeof (b)),1 ,"send single byte" ); +} + +static int verify_mptcpify (void ) +{ +char cmd[256 ]; +int err =0 ; + +/* Output of ss: +* +* ESTAB 0 0 127.0.0.1:44180 127.0.0.1:42225 cubic +* ... tcp-ulp-mptcp flags:Mmec ... +*/ +snprintf (cmd,sizeof (cmd), +"ip netns exec %s ss -tOni | grep -q '%s'" , +NS_TEST ,"tcp-ulp-mptcp" ); +if (!ASSERT_OK (system (cmd),"No tcp-ulp-mptcp found!" )) + err++; + +/* Output of nstat: +* +* #kernel +* MPTcpExtMPCapableSYNACKRX 1 0.0 +*/ +snprintf (cmd,sizeof (cmd), +"ip netns exec %s nstat -asz %s | awk '%s' | grep -q '%s'" , +NS_TEST ,"MPTcpExtMPCapableSYNACKRX" , +"NR==1 {next} {print $2}" ,"1" ); +if (!ASSERT_OK (system (cmd),"No MPTcpExtMPCapableSYNACKRX found!" )) + err++; + +return err; +} + +static int run_mptcpify (int cgroup_fd) +{ +int server_fd, client_fd, err =0 ; +struct mptcpify *mptcpify_skel; + + mptcpify_skel =mptcpify__open_and_load (); +if (!ASSERT_OK_PTR (mptcpify_skel,"skel_open_load" )) +return -EIO ; + + err =mptcpify__attach (mptcpify_skel); +if (!ASSERT_OK (err,"skel_attach" )) + goto out; + +/* without MPTCP */ + server_fd =start_server (AF_INET ,SOCK_STREAM ,NULL ,0 ,0 ); +if (!ASSERT_GE (server_fd,0 ,"start_server" )) { + err = -EIO ; + goto out; + } + + client_fd =connect_to_fd (server_fd,0 ); +if (!ASSERT_GE (client_fd,0 ,"connect to fd" )) { + err = -EIO ; + goto close_server; + } + +send_byte (client_fd); + err =verify_mptcpify (); + +close (client_fd); +close_server: +close (server_fd); +out: +mptcpify__destroy (mptcpify_skel); +return err; +} + +static void test_mptcpify (void ) +{ +struct nstoken *nstoken =NULL ; +int cgroup_fd; + + cgroup_fd =test__join_cgroup ("/mptcpify" ); +if (!ASSERT_GE (cgroup_fd,0 ,"test__join_cgroup" )) +return ; + + nstoken =create_netns (); +if (!ASSERT_OK_PTR (nstoken,"create_netns" )) + goto fail; + +ASSERT_OK (run_mptcpify (cgroup_fd),"run_mptcpify" ); + +fail: +cleanup_netns (nstoken); +close (cgroup_fd); +} +void test_mptcp (void ) {if (test__start_subtest ("base" ))test_base (); +if (test__start_subtest ("mptcpify" )) +test_mptcpify (); }
tools/testing/selftests/bpf/progs/mptcpify.c new file +20 -0
@@ -0,0 +1,20 @@ +// SPDX-License-Identifier: GPL-2.0 +/* Copyright (c) 2023, SUSE. */ + +#include "vmlinux.h" +#include <bpf/bpf_tracing.h> +#include "bpf_tracing_net.h" + +char _license[] SEC("license") = "GPL"; + +SEC("fmod_ret/update_socket_protocol") +int BPF_PROG (mptcpify ,int family,int type,int protocol) +{ +if ((family ==AF_INET || family ==AF_INET6 ) && + type ==SOCK_STREAM && + (!protocol || protocol ==IPPROTO_TCP )) { +return IPPROTO_MPTCP ; + } + +return protocol; +}
Signature · 1 line
2.35.3
Re: selftests/bpf: Add mptcpify test: Tests Results
Hi Geliang,
Thank you for your modifications, that's great!
Our CI did some validations and here is its report:
- KVM Validation: normal (except selftest_mptcp_join):
- Success! ✅:
- Task: https://cirrus-ci.com/task/6058655376736256
- Summary: https://api.cirrus-ci.com/v1/artifact/task/6058655376736256/summary/summary.txt
- KVM Validation: debug (except selftest_mptcp_join):
- Success! ✅:
- Task: https://cirrus-ci.com/task/6621605330157568
- Summary: https://api.cirrus-ci.com/v1/artifact/task/6621605330157568/summary/summary.txt
- KVM Validation: normal (only selftest_mptcp_join):
- Success! ✅:
- Task: https://cirrus-ci.com/task/5495705423314944
- Summary: https://api.cirrus-ci.com/v1/artifact/task/5495705423314944/summary/summary.txt
- KVM Validation: debug (only selftest_mptcp_join):
- Unstable: 1 failed test(s): selftest_mptcp_join 🔴:
- Task: https://cirrus-ci.com/task/5233749210693632
- Summary: https://api.cirrus-ci.com/v1/artifact/task/5233749210693632/summary/summary.txt
Initiator: Matthieu Baerts
Commits: https://github.com/multipath-tcp/mptcp_net-next/commits/352004bd53be
If there are some issues, you can reproduce them using the same environment as the one used by the CI thanks to a docker image, e.g.:
$ cd [kernel source code]
$ docker run -v "${PWD}:${PWD}:rw" -w "${PWD}" --privileged --rm -it \
--pull always mptcp/mptcp-upstream-virtme-docker:latest \
auto-debugFor more details:
https://github.com/multipath-tcp/mptcp-upstream-virtme-docker
Please note that despite all the efforts that have been already done to have a stable tests suite when executed on a public CI like here, it is possible some reported issues are not due to your modifications. Still, do not hesitate to help us improve that ;-)
Cheers,
MPTCP GH Action bot
Bot operated by Matthieu Baerts (Tessares)
Martin KaFai Lau
reviewerRe: [PATCH bpf-next v11 5/5] selftests/bpf: Add mptcpify test
In reply to Geliang TangQuoted 40 lines · Geliang Tang
tools/testing/selftests/bpf/prog_tests/mptcp.c +25
-0@@ -6,6 +6,7 @@#include "cgroup_helpers.h" #include "network_helpers.h" #include "mptcp_sock.skel.h" +#include "mptcpify.skel.h" char NS_TEST [32 ];@@ -185,8 +186,105 @@ static void test_base(void)close (cgroup_fd); } +static void send_byte (int fd) +{ +char b =0x55 ; + +ASSERT_EQ (write (fd, &b,sizeof (b)),1 ,"send single byte" ); +} + +static int verify_mptcpify (void ) +{ +char cmd[256 ]; +int err =0 ; + +/* Output of ss: +* +* ESTAB 0 0 127.0.0.1:44180 127.0.0.1:42225 cubic +* ... tcp-ulp-mptcp flags:Mmec ... +*/ +snprintf (cmd,sizeof (cmd), +"ip netns exec %s ss -tOni | grep -q '%s'" , +NS_TEST ,"tcp-ulp-mptcp" ); +if (!ASSERT_OK (system (cmd),"No tcp-ulp-mptcp found!" )) + err++; + +/* Output of nstat:
tools/testing/selftests/bpf/prog_tests/mptcp.c static void test_base(void) · line 222
Quoted 10 lines
tools/testing/selftests/bpf/prog_tests/mptcp.c +10
-0+* +* #kernel +* MPTcpExtMPCapableSYNACKRX 1 0.0 +*/ +snprintf (cmd,sizeof (cmd), +"ip netns exec %s nstat -asz %s | awk '%s' | grep -q '%s'" , +NS_TEST ,"MPTcpExtMPCapableSYNACKRX" , +"NR==1 {next} {print $2}" ,"1" ); +if (!ASSERT_OK (system (cmd),"No MPTcpExtMPCapableSYNACKRX found!" )) + err++;
The idea is to confirm the protocol has been changed. Is it more direct to use getsockopt(SO_PROTOCOL) on the created fd(s)?
tools/testing/selftests/bpf/prog_tests/mptcp.c static void test_base(void) · line 234
Quoted 12 lines
tools/testing/selftests/bpf/prog_tests/mptcp.c +12
-0+ +return err; +} + +static int run_mptcpify (int cgroup_fd) +{ +int server_fd, client_fd, err =0 ; +struct mptcpify *mptcpify_skel; + + mptcpify_skel =mptcpify__open_and_load (); +if (!ASSERT_OK_PTR (mptcpify_skel,"skel_open_load" )) +return -EIO ;
Although the return value does not matter much, -EIO looks weird for the error from mptcpify__open_and_load(). May be 'return libbpf_get_error(mptcpify_skel);'
--
pw-bot: cr
Quoted 29 lines
tools/testing/selftests/bpf/prog_tests/mptcp.c +29
-0+ + err =mptcpify__attach (mptcpify_skel); +if (!ASSERT_OK (err,"skel_attach" )) + goto out; + +/* without MPTCP */ + server_fd =start_server (AF_INET ,SOCK_STREAM ,NULL ,0 ,0 ); +if (!ASSERT_GE (server_fd,0 ,"start_server" )) { + err = -EIO ; + goto out; + } + + client_fd =connect_to_fd (server_fd,0 ); +if (!ASSERT_GE (client_fd,0 ,"connect to fd" )) { + err = -EIO ; + goto close_server; + } + +send_byte (client_fd); + err =verify_mptcpify (); + +close (client_fd); +close_server: +close (server_fd); +out: +mptcpify__destroy (mptcpify_skel); +return err; +} +