Message ID | 20230620171409.166001-1-kuifeng@meta.com |
---|---|
Headers | show |
Series | Fix missing synack in BPF cgroup_skb filters | expand |
On 6/20/23 10:14 AM, Kui-Feng Lee wrote: > Always call BPF filters if CGROUP BPF is enabled for EGRESS without > checking skb->sk against sk. > > The filters were called only if skb is owned by the sock that the > skb is sent out through. In another words, skb->sk should point to > the sock that it is sending through its egress. However, the filters would > miss SYNACK skbs that they are owned by a request_sock but sent through > the listening sock, that is the socket listening incoming connections. > This is an unnecessary restrict. The original patch which introduced 'sk == skb->sk' is 3007098494be cgroup: add support for eBPF programs There are no mentioning in commit message why 'sk == skb->sk' is needed. So it is possible that this is just restricted for use cases at that moment. Now there are use cases where 'sk != skb->sk' so removing this check can enable the new use case. Maybe you can add this into your commit message so people can understand the history of 'sk == skb->sk'. > > Signed-off-by: Kui-Feng Lee <kuifeng@meta.com> > --- > include/linux/bpf-cgroup.h | 2 +- > 1 file changed, 1 insertion(+), 1 deletion(-) > > diff --git a/include/linux/bpf-cgroup.h b/include/linux/bpf-cgroup.h > index 57e9e109257e..e656da531f9f 100644 > --- a/include/linux/bpf-cgroup.h > +++ b/include/linux/bpf-cgroup.h > @@ -199,7 +199,7 @@ static inline bool cgroup_bpf_sock_enabled(struct sock *sk, > #define BPF_CGROUP_RUN_PROG_INET_EGRESS(sk, skb) \ > ({ \ > int __ret = 0; \ > - if (cgroup_bpf_enabled(CGROUP_INET_EGRESS) && sk && sk == skb->sk) { \ > + if (cgroup_bpf_enabled(CGROUP_INET_EGRESS) && sk) { \ > typeof(sk) __sk = sk_to_full_sk(sk); \ > if (sk_fullsock(__sk) && \ > cgroup_bpf_sock_enabled(__sk, CGROUP_INET_EGRESS)) \
On 6/21/23 20:37, Yonghong Song wrote: > > > On 6/20/23 10:14 AM, Kui-Feng Lee wrote: >> Always call BPF filters if CGROUP BPF is enabled for EGRESS without >> checking skb->sk against sk. >> >> The filters were called only if skb is owned by the sock that the >> skb is sent out through. In another words, skb->sk should point to >> the sock that it is sending through its egress. However, the filters >> would >> miss SYNACK skbs that they are owned by a request_sock but sent through >> the listening sock, that is the socket listening incoming connections. >> This is an unnecessary restrict. > > The original patch which introduced 'sk == skb->sk' is > 3007098494be cgroup: add support for eBPF programs > There are no mentioning in commit message why 'sk == skb->sk' > is needed. So it is possible that this is just restricted > for use cases at that moment. Now there are use cases > where 'sk != skb->sk' so removing this check can enable > the new use case. Maybe you can add this into your commit > message so people can understand the history of 'sk == skb->sk'. I will put it down on the next version. > >> >> Signed-off-by: Kui-Feng Lee <kuifeng@meta.com> >> --- >> include/linux/bpf-cgroup.h | 2 +- >> 1 file changed, 1 insertion(+), 1 deletion(-) >> >> diff --git a/include/linux/bpf-cgroup.h b/include/linux/bpf-cgroup.h >> index 57e9e109257e..e656da531f9f 100644 >> --- a/include/linux/bpf-cgroup.h >> +++ b/include/linux/bpf-cgroup.h >> @@ -199,7 +199,7 @@ static inline bool cgroup_bpf_sock_enabled(struct >> sock *sk, >> #define BPF_CGROUP_RUN_PROG_INET_EGRESS(sk, skb) \ >> ({ \ >> int __ret = 0; \ >> - if (cgroup_bpf_enabled(CGROUP_INET_EGRESS) && sk && sk == >> skb->sk) { \ >> + if (cgroup_bpf_enabled(CGROUP_INET_EGRESS) && sk) { \ >> typeof(sk) __sk = sk_to_full_sk(sk); \ >> if (sk_fullsock(__sk) && \ >> cgroup_bpf_sock_enabled(__sk, >> CGROUP_INET_EGRESS)) \ >
On 6/21/23 20:37, Yonghong Song wrote: > > > On 6/20/23 10:14 AM, Kui-Feng Lee wrote: >> Always call BPF filters if CGROUP BPF is enabled for EGRESS without >> checking skb->sk against sk. >> >> The filters were called only if skb is owned by the sock that the >> skb is sent out through. In another words, skb->sk should point to >> the sock that it is sending through its egress. However, the filters >> would >> miss SYNACK skbs that they are owned by a request_sock but sent through >> the listening sock, that is the socket listening incoming connections. >> This is an unnecessary restrict. > > The original patch which introduced 'sk == skb->sk' is > 3007098494be cgroup: add support for eBPF programs > There are no mentioning in commit message why 'sk == skb->sk' > is needed. So it is possible that this is just restricted > for use cases at that moment. Now there are use cases > where 'sk != skb->sk' so removing this check can enable > the new use case. Maybe you can add this into your commit > message so people can understand the history of 'sk == skb->sk'. After checking the code and the Alexei's comment[1] again, this check may be different from what I thought. In another post[2], Daniel Borkmann mentioned Wouldn't that mean however, when you go through stacked devices that you'd run the same eBPF cgroup program for skb->sk multiple times? I read this paragraph several times. This check ensures the filters are only called for the device on the top of a stack. So, I probably should change the check to sk == skb_to_full_sk(skb) instead of removing it. If we remove the check, egress filters could be called multiple times for a skb, just like what Daniel said. Does that make sense? [1] https://lore.kernel.org/all/CAADnVQKi0c=Mf3b=z43=b6n2xBVhwPw4QoV_au5+pFE29iLkaQ@mail.gmail.com/ [2] https://lore.kernel.org/all/58193E9D.7040201@iogearbox.net/ > >> >> Signed-off-by: Kui-Feng Lee <kuifeng@meta.com> >> --- >> include/linux/bpf-cgroup.h | 2 +- >> 1 file changed, 1 insertion(+), 1 deletion(-) >> >> diff --git a/include/linux/bpf-cgroup.h b/include/linux/bpf-cgroup.h >> index 57e9e109257e..e656da531f9f 100644 >> --- a/include/linux/bpf-cgroup.h >> +++ b/include/linux/bpf-cgroup.h >> @@ -199,7 +199,7 @@ static inline bool cgroup_bpf_sock_enabled(struct >> sock *sk, >> #define BPF_CGROUP_RUN_PROG_INET_EGRESS(sk, skb) \ >> ({ \ >> int __ret = 0; \ >> - if (cgroup_bpf_enabled(CGROUP_INET_EGRESS) && sk && sk == >> skb->sk) { \ >> + if (cgroup_bpf_enabled(CGROUP_INET_EGRESS) && sk) { \ >> typeof(sk) __sk = sk_to_full_sk(sk); \ >> if (sk_fullsock(__sk) && \ >> cgroup_bpf_sock_enabled(__sk, >> CGROUP_INET_EGRESS)) \ >
On 6/22/23 10:15 AM, Kui-Feng Lee wrote: > > > On 6/21/23 20:37, Yonghong Song wrote: >> >> >> On 6/20/23 10:14 AM, Kui-Feng Lee wrote: >>> Always call BPF filters if CGROUP BPF is enabled for EGRESS without >>> checking skb->sk against sk. >>> >>> The filters were called only if skb is owned by the sock that the >>> skb is sent out through. In another words, skb->sk should point to >>> the sock that it is sending through its egress. However, the filters >>> would >>> miss SYNACK skbs that they are owned by a request_sock but sent through >>> the listening sock, that is the socket listening incoming connections. >>> This is an unnecessary restrict. >> >> The original patch which introduced 'sk == skb->sk' is >> 3007098494be cgroup: add support for eBPF programs >> There are no mentioning in commit message why 'sk == skb->sk' >> is needed. So it is possible that this is just restricted >> for use cases at that moment. Now there are use cases >> where 'sk != skb->sk' so removing this check can enable >> the new use case. Maybe you can add this into your commit >> message so people can understand the history of 'sk == skb->sk'. > > After checking the code and the Alexei's comment[1] again, this check > may be different from what I thought. In another post[2], > Daniel Borkmann mentioned > > Wouldn't that mean however, when you go through stacked devices that > you'd run the same eBPF cgroup program for skb->sk multiple times? > > I read this paragraph several times. > This check ensures the filters are only called for the device on > the top of a stack. So, I probably should change the check to > > sk == skb_to_full_sk(skb) I think this should work. It exactly covers your use case: they are owned by a request_sock but sent through the listening sock, that is the socket listening incoming connections and sk == skb->sk for non request_sock/listening_sock case. I originally though whether you could do sk == skb->sk || skb->sk->sk_state == TCP_NEW_SYN_RECV but obviously your approach is better. > > instead of removing it. If we remove the check, egress filters > could be called multiple times for a skb, just like what Daniel said. > > Does that make sense? > > [1] > https://lore.kernel.org/all/CAADnVQKi0c=Mf3b=z43=b6n2xBVhwPw4QoV_au5+pFE29iLkaQ@mail.gmail.com/ > [2] https://lore.kernel.org/all/58193E9D.7040201@iogearbox.net/ > >> >>> >>> Signed-off-by: Kui-Feng Lee <kuifeng@meta.com> >>> --- >>> include/linux/bpf-cgroup.h | 2 +- >>> 1 file changed, 1 insertion(+), 1 deletion(-) >>> >>> diff --git a/include/linux/bpf-cgroup.h b/include/linux/bpf-cgroup.h >>> index 57e9e109257e..e656da531f9f 100644 >>> --- a/include/linux/bpf-cgroup.h >>> +++ b/include/linux/bpf-cgroup.h >>> @@ -199,7 +199,7 @@ static inline bool cgroup_bpf_sock_enabled(struct >>> sock *sk, >>> #define BPF_CGROUP_RUN_PROG_INET_EGRESS(sk, skb) \ >>> ({ \ >>> int __ret = 0; \ >>> - if (cgroup_bpf_enabled(CGROUP_INET_EGRESS) && sk && sk == >>> skb->sk) { \ >>> + if (cgroup_bpf_enabled(CGROUP_INET_EGRESS) && sk) { \ >>> typeof(sk) __sk = sk_to_full_sk(sk); \ >>> if (sk_fullsock(__sk) && \ >>> cgroup_bpf_sock_enabled(__sk, >>> CGROUP_INET_EGRESS)) \ >>
On 6/22/23 8:28 PM, Yonghong Song wrote: > On 6/22/23 10:15 AM, Kui-Feng Lee wrote: >> On 6/21/23 20:37, Yonghong Song wrote: >>> On 6/20/23 10:14 AM, Kui-Feng Lee wrote: >>>> Always call BPF filters if CGROUP BPF is enabled for EGRESS without >>>> checking skb->sk against sk. >>>> >>>> The filters were called only if skb is owned by the sock that the >>>> skb is sent out through. In another words, skb->sk should point to >>>> the sock that it is sending through its egress. However, the filters would >>>> miss SYNACK skbs that they are owned by a request_sock but sent through >>>> the listening sock, that is the socket listening incoming connections. >>>> This is an unnecessary restrict. >>> >>> The original patch which introduced 'sk == skb->sk' is >>> 3007098494be cgroup: add support for eBPF programs >>> There are no mentioning in commit message why 'sk == skb->sk' >>> is needed. So it is possible that this is just restricted >>> for use cases at that moment. Now there are use cases >>> where 'sk != skb->sk' so removing this check can enable >>> the new use case. Maybe you can add this into your commit >>> message so people can understand the history of 'sk == skb->sk'. >> >> After checking the code and the Alexei's comment[1] again, this check >> may be different from what I thought. In another post[2], >> Daniel Borkmann mentioned >> >> Wouldn't that mean however, when you go through stacked devices that >> you'd run the same eBPF cgroup program for skb->sk multiple times? >> >> I read this paragraph several times. >> This check ensures the filters are only called for the device on >> the top of a stack. So, I probably should change the check to >> >> sk == skb_to_full_sk(skb) > > I think this should work. It exactly covers your use case: > they are owned by a request_sock but sent through > the listening sock, that is the socket listening incoming connections > and sk == skb->sk for non request_sock/listening_sock case. Just a thought, should the test look like the below? int __ret = 0; \ if (cgroup_bpf_enabled(CGROUP_INET_EGRESS) && sk) { \ typeof(sk) __sk = sk_to_full_sk(sk); \ if (sk_fullsock(__sk) && __sk == skb_to_full_sk(skb) && \ cgroup_bpf_sock_enabled(__sk, CGROUP_INET_EGRESS)) \ __ret = __cgroup_bpf_run_filter_skb(__sk, skb, \ CGROUP_INET_EGRESS); \ } \ Iow, we do already convert __sk to full sk, so we should then also use that for the test with skb_to_full_sk(skb). Thanks, Daniel
On 6/22/23 13:06, Daniel Borkmann wrote: > On 6/22/23 8:28 PM, Yonghong Song wrote: >> On 6/22/23 10:15 AM, Kui-Feng Lee wrote: >>> On 6/21/23 20:37, Yonghong Song wrote: >>>> On 6/20/23 10:14 AM, Kui-Feng Lee wrote: >>>>> Always call BPF filters if CGROUP BPF is enabled for EGRESS without >>>>> checking skb->sk against sk. >>>>> >>>>> The filters were called only if skb is owned by the sock that the >>>>> skb is sent out through. In another words, skb->sk should point to >>>>> the sock that it is sending through its egress. However, the >>>>> filters would >>>>> miss SYNACK skbs that they are owned by a request_sock but sent >>>>> through >>>>> the listening sock, that is the socket listening incoming connections. >>>>> This is an unnecessary restrict. >>>> >>>> The original patch which introduced 'sk == skb->sk' is >>>> 3007098494be cgroup: add support for eBPF programs >>>> There are no mentioning in commit message why 'sk == skb->sk' >>>> is needed. So it is possible that this is just restricted >>>> for use cases at that moment. Now there are use cases >>>> where 'sk != skb->sk' so removing this check can enable >>>> the new use case. Maybe you can add this into your commit >>>> message so people can understand the history of 'sk == skb->sk'. >>> >>> After checking the code and the Alexei's comment[1] again, this check >>> may be different from what I thought. In another post[2], >>> Daniel Borkmann mentioned >>> >>> Wouldn't that mean however, when you go through stacked devices >>> that >>> you'd run the same eBPF cgroup program for skb->sk multiple times? >>> >>> I read this paragraph several times. >>> This check ensures the filters are only called for the device on >>> the top of a stack. So, I probably should change the check to >>> >>> sk == skb_to_full_sk(skb) >> >> I think this should work. It exactly covers your use case: >> they are owned by a request_sock but sent through >> the listening sock, that is the socket listening incoming connections >> and sk == skb->sk for non request_sock/listening_sock case. > > Just a thought, should the test look like the below? > > int __ret = > 0; \ > if (cgroup_bpf_enabled(CGROUP_INET_EGRESS) && sk) > { \ > typeof(sk) __sk = > sk_to_full_sk(sk); \ > if (sk_fullsock(__sk) && __sk == skb_to_full_sk(skb) > && \ > cgroup_bpf_sock_enabled(__sk, > CGROUP_INET_EGRESS)) \ > __ret = __cgroup_bpf_run_filter_skb(__sk, > skb, \ > > CGROUP_INET_EGRESS); \ > > } \ > > Iow, we do already convert __sk to full sk, so we should then also use that > for the test with skb_to_full_sk(skb). Agree! > > Thanks, > Daniel
On 6/23/23 1:55 AM, Kui-Feng Lee wrote: > On 6/22/23 13:06, Daniel Borkmann wrote: >> On 6/22/23 8:28 PM, Yonghong Song wrote: >>> On 6/22/23 10:15 AM, Kui-Feng Lee wrote: >>>> On 6/21/23 20:37, Yonghong Song wrote: >>>>> On 6/20/23 10:14 AM, Kui-Feng Lee wrote: >>>>>> Always call BPF filters if CGROUP BPF is enabled for EGRESS without >>>>>> checking skb->sk against sk. >>>>>> >>>>>> The filters were called only if skb is owned by the sock that the >>>>>> skb is sent out through. In another words, skb->sk should point to >>>>>> the sock that it is sending through its egress. However, the filters would >>>>>> miss SYNACK skbs that they are owned by a request_sock but sent through >>>>>> the listening sock, that is the socket listening incoming connections. >>>>>> This is an unnecessary restrict. >>>>> >>>>> The original patch which introduced 'sk == skb->sk' is >>>>> 3007098494be cgroup: add support for eBPF programs >>>>> There are no mentioning in commit message why 'sk == skb->sk' >>>>> is needed. So it is possible that this is just restricted >>>>> for use cases at that moment. Now there are use cases >>>>> where 'sk != skb->sk' so removing this check can enable >>>>> the new use case. Maybe you can add this into your commit >>>>> message so people can understand the history of 'sk == skb->sk'. >>>> >>>> After checking the code and the Alexei's comment[1] again, this check >>>> may be different from what I thought. In another post[2], >>>> Daniel Borkmann mentioned >>>> >>>> Wouldn't that mean however, when you go through stacked devices that >>>> you'd run the same eBPF cgroup program for skb->sk multiple times? >>>> >>>> I read this paragraph several times. >>>> This check ensures the filters are only called for the device on >>>> the top of a stack. So, I probably should change the check to >>>> >>>> sk == skb_to_full_sk(skb) >>> >>> I think this should work. It exactly covers your use case: >>> they are owned by a request_sock but sent through >>> the listening sock, that is the socket listening incoming connections >>> and sk == skb->sk for non request_sock/listening_sock case. >> >> Just a thought, should the test look like the below? >> >> int __ret = 0; \ >> if (cgroup_bpf_enabled(CGROUP_INET_EGRESS) && sk) { \ >> typeof(sk) __sk = sk_to_full_sk(sk); \ >> if (sk_fullsock(__sk) && __sk == skb_to_full_sk(skb) && \ >> cgroup_bpf_sock_enabled(__sk, CGROUP_INET_EGRESS)) \ >> __ret = __cgroup_bpf_run_filter_skb(__sk, skb, \ >> CGROUP_INET_EGRESS); \ >> } \ >> >> Iow, we do already convert __sk to full sk, so we should then also use that >> for the test with skb_to_full_sk(skb). > > Agree! It would also be useful to do an in-depth analysis for the commit msg in which cases the sk == skb->sk matches and sk was not a full sock (but __sk is) given the __sk = sk_to_full_sk(sk) exists in the code to document which situation this is covering in the existing code (... perhaps it used to work back then for synack just that later changes altered it without anyone noticing until now). Thanks, Daniel
On 6/23/23 01:50, Daniel Borkmann wrote: > On 6/23/23 1:55 AM, Kui-Feng Lee wrote: >> On 6/22/23 13:06, Daniel Borkmann wrote: >>> On 6/22/23 8:28 PM, Yonghong Song wrote: >>>> On 6/22/23 10:15 AM, Kui-Feng Lee wrote: >>>>> On 6/21/23 20:37, Yonghong Song wrote: >>>>>> On 6/20/23 10:14 AM, Kui-Feng Lee wrote: >>>>>>> Always call BPF filters if CGROUP BPF is enabled for EGRESS without >>>>>>> checking skb->sk against sk. >>>>>>> >>>>>>> The filters were called only if skb is owned by the sock that the >>>>>>> skb is sent out through. In another words, skb->sk should point to >>>>>>> the sock that it is sending through its egress. However, the >>>>>>> filters would >>>>>>> miss SYNACK skbs that they are owned by a request_sock but sent >>>>>>> through >>>>>>> the listening sock, that is the socket listening incoming >>>>>>> connections. >>>>>>> This is an unnecessary restrict. >>>>>> >>>>>> The original patch which introduced 'sk == skb->sk' is >>>>>> 3007098494be cgroup: add support for eBPF programs >>>>>> There are no mentioning in commit message why 'sk == skb->sk' >>>>>> is needed. So it is possible that this is just restricted >>>>>> for use cases at that moment. Now there are use cases >>>>>> where 'sk != skb->sk' so removing this check can enable >>>>>> the new use case. Maybe you can add this into your commit >>>>>> message so people can understand the history of 'sk == skb->sk'. >>>>> >>>>> After checking the code and the Alexei's comment[1] again, this check >>>>> may be different from what I thought. In another post[2], >>>>> Daniel Borkmann mentioned >>>>> >>>>> Wouldn't that mean however, when you go through stacked >>>>> devices that >>>>> you'd run the same eBPF cgroup program for skb->sk multiple >>>>> times? >>>>> >>>>> I read this paragraph several times. >>>>> This check ensures the filters are only called for the device on >>>>> the top of a stack. So, I probably should change the check to >>>>> >>>>> sk == skb_to_full_sk(skb) >>>> >>>> I think this should work. It exactly covers your use case: >>>> they are owned by a request_sock but sent through >>>> the listening sock, that is the socket listening incoming >>>> connections >>>> and sk == skb->sk for non request_sock/listening_sock case. >>> >>> Just a thought, should the test look like the below? >>> >>> int __ret = >>> 0; \ >>> if (cgroup_bpf_enabled(CGROUP_INET_EGRESS) && sk) >>> { \ >>> typeof(sk) __sk = >>> sk_to_full_sk(sk); \ >>> if (sk_fullsock(__sk) && __sk == skb_to_full_sk(skb) >>> && \ >>> cgroup_bpf_sock_enabled(__sk, >>> CGROUP_INET_EGRESS)) \ >>> __ret = __cgroup_bpf_run_filter_skb(__sk, >>> skb, \ >>> CGROUP_INET_EGRESS); \ >>> } \ >>> >>> Iow, we do already convert __sk to full sk, so we should then also >>> use that >>> for the test with skb_to_full_sk(skb). >> >> Agree! > > It would also be useful to do an in-depth analysis for the commit msg in > which > cases the sk == skb->sk matches and sk was not a full sock (but __sk is) > given > the __sk = sk_to_full_sk(sk) exists in the code to document which > situation this > is covering in the existing code (... perhaps it used to work back then for > synack just that later changes altered it without anyone noticing until > now). I did a test that trace how a packet going through L2TP devices. I am going to include the analysis of the test and other related links of discussions in the commit log.