Applications can be confused slightly because we do not always return the
same error code as expected, e.g. what the TCP stack normally returns. For
example on a sock err sk->sk_err instead of returning the sock_error we
return EAGAIN. This usually means the application will 'try again'
instead of aborting immediately. Another example, when a shutdown event
is received we should immediately abort instead of waiting for data when
the user provides a timeout.
These tend to not be fatal, applications usually recover, but introduces
bogus errors to the user or introduces unexpected latency. Before
'c5d2177a72a16' we fell back to the TCP stack when no data was available
so we managed to catch many of the cases here, although with the extra
latency cost of calling tcp_msg_wait_data() first.
To fix lets duplicate the error handling in TCP stack into tcp_bpf so
that we get the same error codes.
These were found in our CI tests that run applications against sockmap
and do longer lived testing, at least compared to test_sockmap that
does short-lived ping/pong tests, and in some of our test clusters
we deploy.
Its non-trivial to do these in a shorter form CI tests that would be
appropriate for BPF selftests, but we are looking into it so we can
ensure this keeps working going forward. As a preview one idea is to
pull in the packetdrill testing which catches some of this.
Fixes: c5d2177a72 ("bpf, sockmap: Fix race in ingress receive verdict with redirect to self")
Signed-off-by: John Fastabend <john.fastabend@gmail.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Link: https://lore.kernel.org/bpf/20220104205918.286416-1-john.fastabend@gmail.com
A socket in a sockmap may have different combinations of programs attached
depending on configuration. There can be no programs in which case the socket
acts as a sink only. There can be a TX program in this case a BPF program is
attached to sending side, but no RX program is attached. There can be an RX
program only where sends have no BPF program attached, but receives are hooked
with BPF. And finally, both TX and RX programs may be attached. Giving us the
permutations:
None, Tx, Rx, and TxRx
To date most of our use cases have been TX case being used as a fast datapath
to directly copy between local application and a userspace proxy. Or Rx cases
and TxRX applications that are operating an in kernel based proxy. The traffic
in the first case where we hook applications into a userspace application looks
like this:
AppA redirect AppB
Tx <-----------> Rx
| |
+ +
TCP <--> lo <--> TCP
In this case all traffic from AppA (after 3whs) is copied into the AppB
ingress queue and no traffic is ever on the TCP recieive_queue.
In the second case the application never receives, except in some rare error
cases, traffic on the actual user space socket. Instead the send happens in
the kernel.
AppProxy socket pool
sk0 ------------->{sk1,sk2, skn}
^ |
| |
| v
ingress lb egress
TCP TCP
Here because traffic is never read off the socket with userspace recv() APIs
there is only ever one reader on the sk receive_queue. Namely the BPF programs.
However, we've started to introduce a third configuration where the BPF program
on receive should process the data, but then the normal case is to push the
data into the receive queue of AppB.
AppB
recv() (userspace)
-----------------------
tcp_bpf_recvmsg() (kernel)
| |
| |
| |
ingress_msgQ |
| |
RX_BPF |
| |
v v
sk->receive_queue
This is different from the App{A,B} redirect because traffic is first received
on the sk->receive_queue.
Now for the issue. The tcp_bpf_recvmsg() handler first checks the ingress_msg
queue for any data handled by the BPF rx program and returned with PASS code
so that it was enqueued on the ingress msg queue. Then if no data exists on
that queue it checks the socket receive queue. Unfortunately, this is the same
receive_queue the BPF program is reading data off of. So we get a race. Its
possible for the recvmsg() hook to pull data off the receive_queue before the
BPF hook has a chance to read it. It typically happens when an application is
banging on recv() and getting EAGAINs. Until they manage to race with the RX
BPF program.
To fix this we note that before this patch at attach time when the socket is
loaded into the map we check if it needs a TX program or just the base set of
proto bpf hooks. Then it uses the above general RX hook regardless of if we
have a BPF program attached at rx or not. This patch now extends this check to
handle all cases enumerated above, TX, RX, TXRX, and none. And to fix above
race when an RX program is attached we use a new hook that is nearly identical
to the old one except now we do not let the recv() call skip the RX BPF program.
Now only the BPF program pulls data from sk->receive_queue and recv() only
pulls data from the ingress msgQ post BPF program handling.
With this resolved our AppB from above has been up and running for many hours
without detecting any errors. We do this by correlating counters in RX BPF
events and the AppB to ensure data is never skipping the BPF program. Selftests,
was not able to detect this because we only run them for a short period of time
on well ordered send/recvs so we don't get any of the noise we see in real
application environments.
Fixes: 51199405f9 ("bpf: skb_verdict, support SK_PASS on RX BPF path")
Signed-off-by: John Fastabend <john.fastabend@gmail.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Tested-by: Jussi Maki <joamaki@gmail.com>
Reviewed-by: Jakub Sitnicki <jakub@cloudflare.com>
Link: https://lore.kernel.org/bpf/20211103204736.248403-4-john.fastabend@gmail.com
We do not need to handle unhash from BPF side we can simply wait for the
close to happen. The original concern was a socket could transition from
ESTABLISHED state to a new state while the BPF hook was still attached.
But, we convinced ourself this is no longer possible and we also improved
BPF sockmap to handle listen sockets so this is no longer a problem.
More importantly though there are cases where unhash is called when data is
in the receive queue. The BPF unhash logic will flush this data which is
wrong. To be correct it should keep the data in the receive queue and allow
a receiving application to continue reading the data. This may happen when
tcp_abort() is received for example. Instead of complicating the logic in
unhash simply moving all this to tcp_close() hook solves this.
Fixes: 51199405f9 ("bpf: skb_verdict, support SK_PASS on RX BPF path")
Signed-off-by: John Fastabend <john.fastabend@gmail.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Tested-by: Jussi Maki <joamaki@gmail.com>
Reviewed-by: Jakub Sitnicki <jakub@cloudflare.com>
Link: https://lore.kernel.org/bpf/20211103204736.248403-3-john.fastabend@gmail.com
The proto ops ->stream_memory_read() is currently only used
by TCP to check whether psock queue is empty or not. We need
to rename it before reusing it for non-TCP protocols, and
adjust the exsiting users accordingly.
Signed-off-by: Cong Wang <cong.wang@bytedance.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Link: https://lore.kernel.org/bpf/20211008203306.37525-2-xiyou.wangcong@gmail.com
With two Msgs, msgA and msgB and a user doing nonblocking sendmsg calls (or
multiple cores) on a single socket 'sk' we could get the following flow.
msgA, sk msgB, sk
----------- ---------------
tcp_bpf_sendmsg()
lock(sk)
psock = sk->psock
tcp_bpf_sendmsg()
lock(sk) ... blocking
tcp_bpf_send_verdict
if (psock->eval == NONE)
psock->eval = sk_psock_msg_verdict
..
< handle SK_REDIRECT case >
release_sock(sk) < lock dropped so grab here >
ret = tcp_bpf_sendmsg_redir
psock = sk->psock
tcp_bpf_send_verdict
lock_sock(sk) ... blocking on B
if (psock->eval == NONE) <- boom.
psock->eval will have msgA state
The problem here is we dropped the lock on msgA and grabbed it with msgB.
Now we have old state in psock and importantly psock->eval has not been
cleared. So msgB will run whatever action was done on A and the verdict
program may never see it.
Fixes: 604326b41a ("bpf, sockmap: convert to generic sk_msg interface")
Signed-off-by: Liu Jian <liujian56@huawei.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Acked-by: John Fastabend <john.fastabend@gmail.com>
Link: https://lore.kernel.org/bpf/20211012052019.184398-1-liujian56@huawei.com
The proc socket stats use sk_prot->inuse_idx value to record inuse sock
stats. We currently do not set this correctly from sockmap side. The
result is reading sock stats '/proc/net/sockstat' gives incorrect values.
The socket counter is incremented correctly, but because we don't set the
counter correctly when we replace sk_prot we may omit the decrement.
To get the correct inuse_idx value move the core_initcall that initializes
the TCP proto handlers to late_initcall. This way it is initialized after
TCP has the chance to assign the inuse_idx value from the register protocol
handler.
Fixes: 604326b41a ("bpf, sockmap: convert to generic sk_msg interface")
Suggested-by: Jakub Sitnicki <jakub@cloudflare.com>
Signed-off-by: John Fastabend <john.fastabend@gmail.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Reviewed-by: Cong Wang <cong.wang@bytedance.com>
Link: https://lore.kernel.org/bpf/20210712195546.423990-3-john.fastabend@gmail.com
Trivial conflict in net/netfilter/nf_tables_api.c.
Duplicate fix in tools/testing/selftests/net/devlink_port_split.py
- take the net-next version.
skmsg, and L4 bpf - keep the bpf code but remove the flags
and err params.
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
I tried to reuse sk_msg_wait_data() for different protocols,
but it turns out it can not be simply reused. For example,
UDP actually uses two queues to receive skb:
udp_sk(sk)->reader_queue and sk->sk_receive_queue. So we have
to check both of them to know whether we have received any
packet.
Also, UDP does not lock the sock during BH Rx path, it makes
no sense for its ->recvmsg() to lock the sock. It is always
possible for ->recvmsg() to be called before packets actually
arrive in the receive queue, we just use best effort to make
it accurate here.
Fixes: 1f5be6b3b0 ("udp: Implement udp_bpf_recvmsg() for sockmap")
Signed-off-by: Cong Wang <cong.wang@bytedance.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: John Fastabend <john.fastabend@gmail.com>
Acked-by: Jakub Sitnicki <jakub@cloudflare.com>
Link: https://lore.kernel.org/bpf/20210615021342.7416-2-xiyou.wangcong@gmail.com
'err' and 'flags' are not used, we can just get rid of them.
Signed-off-by: Cong Wang <cong.wang@bytedance.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Song Liu <song@kernel.org>
Acked-by: John Fastabend <john.fastabend@gmail.com>
Link: https://lore.kernel.org/bpf/20210517022348.50555-1-xiyou.wangcong@gmail.com
Alexei Starovoitov says:
====================
pull-request: bpf-next 2021-04-23
The following pull-request contains BPF updates for your *net-next* tree.
We've added 69 non-merge commits during the last 22 day(s) which contain
a total of 69 files changed, 3141 insertions(+), 866 deletions(-).
The main changes are:
1) Add BPF static linker support for extern resolution of global, from Andrii.
2) Refine retval for bpf_get_task_stack helper, from Dave.
3) Add a bpf_snprintf helper, from Florent.
4) A bunch of miscellaneous improvements from many developers.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
Conflicts:
MAINTAINERS
- keep Chandrasekar
drivers/net/ethernet/mellanox/mlx5/core/en_main.c
- simple fix + trust the code re-added to param.c in -next is fine
include/linux/bpf.h
- trivial
include/linux/ethtool.h
- trivial, fix kdoc while at it
include/linux/skmsg.h
- move to relevant place in tcp.c, comment re-wrapped
net/core/skmsg.c
- add the sk = sk // sk = NULL around calls
net/tipc/crypto.c
- trivial
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Although these two functions are only used by TCP, they are not
specific to TCP at all, both operate on skmsg and ingress_msg,
so fit in net/core/skmsg.c very well.
And we will need them for non-TCP, so rename and move them to
skmsg.c and export them to modules.
Signed-off-by: Cong Wang <cong.wang@bytedance.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Link: https://lore.kernel.org/bpf/20210331023237.41094-13-xiyou.wangcong@gmail.com
Currently sockmap calls into each protocol to update the struct
proto and replace it. This certainly won't work when the protocol
is implemented as a module, for example, AF_UNIX.
Introduce a new ops sk->sk_prot->psock_update_sk_prot(), so each
protocol can implement its own way to replace the struct proto.
This also helps get rid of symbol dependencies on CONFIG_INET.
Signed-off-by: Cong Wang <cong.wang@bytedance.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Link: https://lore.kernel.org/bpf/20210331023237.41094-11-xiyou.wangcong@gmail.com
Currently we rely on lock_sock to protect ingress_msg,
it is too big for this, we can actually just use a spinlock
to protect this list like protecting other skb queues.
__tcp_bpf_recvmsg() is still special because of peeking,
it still has to use lock_sock.
Signed-off-by: Cong Wang <cong.wang@bytedance.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Acked-by: Jakub Sitnicki <jakub@cloudflare.com>
Acked-by: John Fastabend <john.fastabend@gmail.com>
Link: https://lore.kernel.org/bpf/20210331023237.41094-3-xiyou.wangcong@gmail.com
As suggested by John, clean up sockmap related Kconfigs:
Reduce the scope of CONFIG_BPF_STREAM_PARSER down to TCP stream
parser, to reflect its name.
Make the rest sockmap code simply depend on CONFIG_BPF_SYSCALL
and CONFIG_INET, the latter is still needed at this point because
of TCP/UDP proto update. And leave CONFIG_NET_SOCK_MSG untouched,
as it is used by non-sockmap cases.
Signed-off-by: Cong Wang <cong.wang@bytedance.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Reviewed-by: Lorenz Bauer <lmb@cloudflare.com>
Acked-by: John Fastabend <john.fastabend@gmail.com>
Acked-by: Jakub Sitnicki <jakub@cloudflare.com>
Link: https://lore.kernel.org/bpf/20210223184934.6054-2-xiyou.wangcong@gmail.com
Fix sockmap sk_skb programs so that they observe sk_rcvbuf limits. This
allows users to tune SO_RCVBUF and sockmap will honor them.
We can refactor the if(charge) case out in later patches. But, keep this
fix to the point.
Fixes: 51199405f9 ("bpf: skb_verdict, support SK_PASS on RX BPF path")
Suggested-by: Jakub Sitnicki <jakub@cloudflare.com>
Signed-off-by: John Fastabend <john.fastabend@gmail.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Reviewed-by: Jakub Sitnicki <jakub@cloudflare.com>
Link: https://lore.kernel.org/bpf/160556568657.73229.8404601585878439060.stgit@john-XPS-13-9370
If copy_page_to_iter() fails or even partially completes, but with fewer
bytes copied than expected we currently reset sg.start and return EFAULT.
This proves problematic if we already copied data into the user buffer
before we return an error. Because we leave the copied data in the user
buffer and fail to unwind the scatterlist so kernel side believes data
has been copied and user side believes data has _not_ been received.
Expected behavior should be to return number of bytes copied and then
on the next read we need to return the error assuming its still there. This
can happen if we have a copy length spanning multiple scatterlist elements
and one or more complete before the error is hit.
The error is rare enough though that my normal testing with server side
programs, such as nginx, httpd, envoy, etc., I have never seen this. The
only reliable way to reproduce that I've found is to stream movies over
my browser for a day or so and wait for it to hang. Not very scientific,
but with a few extra WARN_ON()s in the code the bug was obvious.
When we review the errors from copy_page_to_iter() it seems we are hitting
a page fault from copy_page_to_iter_iovec() where the code checks
fault_in_pages_writeable(buf, copy) where buf is the user buffer. It
also seems typical server applications don't hit this case.
The other way to try and reproduce this is run the sockmap selftest tool
test_sockmap with data verification enabled, but it doesn't reproduce the
fault. Perhaps we can trigger this case artificially somehow from the
test tools. I haven't sorted out a way to do that yet though.
Fixes: 604326b41a ("bpf, sockmap: convert to generic sk_msg interface")
Signed-off-by: John Fastabend <john.fastabend@gmail.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Reviewed-by: Jakub Sitnicki <jakub@cloudflare.com>
Link: https://lore.kernel.org/bpf/160556566659.73229.15694973114605301063.stgit@john-XPS-13-9370
Initializing psock->sk_proto and other saved callbacks is only
done in sk_psock_update_proto, after sk_psock_init has returned.
The logic for this is difficult to follow, and needlessly complex.
Instead, initialize psock->sk_proto whenever we allocate a new
psock. Additionally, assert the following invariants:
* The SK has no ULP: ULP does it's own finagling of sk->sk_prot
* sk_user_data is unused: we need it to store sk_psock
Protect our access to sk_user_data with sk_callback_lock, which
is what other users like reuseport arrays, etc. do.
The result is that an sk_psock is always fully initialized, and
that psock->sk_proto is always the "original" struct proto.
The latter allows us to use psock->sk_proto when initializing
IPv6 TCP / UDP callbacks for sockmap.
Signed-off-by: Lorenz Bauer <lmb@cloudflare.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Acked-by: John Fastabend <john.fastabend@gmail.com>
Link: https://lore.kernel.org/bpf/20200821102948.21918-2-lmb@cloudflare.com
If the peer is closed, we will never get more data, so
tcp_bpf_wait_data will get stuck forever. In case we passed
MSG_DONTWAIT to recv(), we get EAGAIN but we should actually get
0.
>From man 2 recv:
RETURN VALUE
When a stream socket peer has performed an orderly shutdown, the
return value will be 0 (the traditional "end-of-file" return).
This patch makes tcp_bpf_wait_data always return 1 when the peer
socket has been shutdown. Either we have data available, and it would
have returned 1 anyway, or there isn't, in which case we'll call
tcp_recvmsg which does the right thing in this situation.
Fixes: 604326b41a ("bpf, sockmap: convert to generic sk_msg interface")
Signed-off-by: Sabrina Dubroca <sd@queasysnail.net>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Acked-by: Jakub Sitnicki <jakub@cloudflare.com>
Link: https://lore.kernel.org/bpf/26038a28c21fea5d04d4bd4744c5686d3f2e5504.1591784177.git.sd@queasysnail.net
When user application calls read() with MSG_PEEK flag to read data
of bpf sockmap socket, kernel panic happens at
__tcp_bpf_recvmsg+0x12c/0x350. sk_msg is not removed from ingress_msg
queue after read out under MSG_PEEK flag is set. Because it's not
judged whether sk_msg is the last msg of ingress_msg queue, the next
sk_msg may be the head of ingress_msg queue, whose memory address of
sg page is invalid. So it's necessary to add check codes to prevent
this problem.
[20759.125457] BUG: kernel NULL pointer dereference, address:
0000000000000008
[20759.132118] CPU: 53 PID: 51378 Comm: envoy Tainted: G E
5.4.32 #1
[20759.140890] Hardware name: Inspur SA5212M4/YZMB-00370-109, BIOS
4.1.12 06/18/2017
[20759.149734] RIP: 0010:copy_page_to_iter+0xad/0x300
[20759.270877] __tcp_bpf_recvmsg+0x12c/0x350
[20759.276099] tcp_bpf_recvmsg+0x113/0x370
[20759.281137] inet_recvmsg+0x55/0xc0
[20759.285734] __sys_recvfrom+0xc8/0x130
[20759.290566] ? __audit_syscall_entry+0x103/0x130
[20759.296227] ? syscall_trace_enter+0x1d2/0x2d0
[20759.301700] ? __audit_syscall_exit+0x1e4/0x290
[20759.307235] __x64_sys_recvfrom+0x24/0x30
[20759.312226] do_syscall_64+0x55/0x1b0
[20759.316852] entry_SYSCALL_64_after_hwframe+0x44/0xa9
Signed-off-by: dihu <anny.hu@linux.alibaba.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Acked-by: John Fastabend <john.fastabend@gmail.com>
Acked-by: Jakub Sitnicki <jakub@cloudflare.com>
Link: https://lore.kernel.org/bpf/20200605084625.9783-1-anny.hu@linux.alibaba.com
In bpf_tcp_ingress we used apply_bytes to subtract bytes from sg.size
which is used to track total bytes in a message. But this is not
correct because apply_bytes is itself modified in the main loop doing
the mem_charge.
Then at the end of this we have sg.size incorrectly set and out of
sync with actual sk values. Then we can get a splat if we try to
cork the data later and again try to redirect the msg to ingress. To
fix instead of trying to track msg.size do the easy thing and include
it as part of the sk_msg_xfer logic so that when the msg is moved the
sg.size is always correct.
To reproduce the below users will need ingress + cork and hit an
error path that will then try to 'free' the skmsg.
[ 173.699981] BUG: KASAN: null-ptr-deref in sk_msg_free_elem+0xdd/0x120
[ 173.699987] Read of size 8 at addr 0000000000000008 by task test_sockmap/5317
[ 173.700000] CPU: 2 PID: 5317 Comm: test_sockmap Tainted: G I 5.7.0-rc1+ #43
[ 173.700005] Hardware name: Dell Inc. Precision 5820 Tower/002KVM, BIOS 1.9.2 01/24/2019
[ 173.700009] Call Trace:
[ 173.700021] dump_stack+0x8e/0xcb
[ 173.700029] ? sk_msg_free_elem+0xdd/0x120
[ 173.700034] ? sk_msg_free_elem+0xdd/0x120
[ 173.700042] __kasan_report+0x102/0x15f
[ 173.700052] ? sk_msg_free_elem+0xdd/0x120
[ 173.700060] kasan_report+0x32/0x50
[ 173.700070] sk_msg_free_elem+0xdd/0x120
[ 173.700080] __sk_msg_free+0x87/0x150
[ 173.700094] tcp_bpf_send_verdict+0x179/0x4f0
[ 173.700109] tcp_bpf_sendpage+0x3ce/0x5d0
Fixes: 604326b41a ("bpf, sockmap: convert to generic sk_msg interface")
Signed-off-by: John Fastabend <john.fastabend@gmail.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Reviewed-by: Jakub Sitnicki <jakub@cloudflare.com>
Acked-by: Martin KaFai Lau <kafai@fb.com>
Link: https://lore.kernel.org/bpf/158861290407.14306.5327773422227552482.stgit@john-Precision-5820-Tower
tcp_bpf_recvmsg() invokes sk_psock_get(), which returns a reference of
the specified sk_psock object to "psock" with increased refcnt.
When tcp_bpf_recvmsg() returns, local variable "psock" becomes invalid,
so the refcount should be decreased to keep refcount balanced.
The reference counting issue happens in several exception handling paths
of tcp_bpf_recvmsg(). When those error scenarios occur such as "flags"
includes MSG_ERRQUEUE, the function forgets to decrease the refcnt
increased by sk_psock_get(), causing a refcnt leak.
Fix this issue by calling sk_psock_put() or pulling up the error queue
read handling when those error scenarios occur.
Fixes: e7a5f1f1cd ("bpf/sockmap: Read psock ingress_msg before sk_receive_queue")
Signed-off-by: Xiyu Yang <xiyuyang19@fudan.edu.cn>
Signed-off-by: Xin Tan <tanxin.ctf@gmail.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Reviewed-by: Jakub Sitnicki <jakub@cloudflare.com>
Link: https://lore.kernel.org/bpf/1587872115-42805-1-git-send-email-xiyuyang19@fudan.edu.cn
After commit f747632b60 ("bpf: sockmap: Move generic sockmap
hooks from BPF TCP"), tcp_bpf_recvmsg() is not used out of
tcp_bpf.c, so make it static and remove it from tcp.h. Also move
it to BPF_STREAM_PARSER #ifdef to fix unused function warnings.
Fixes: f747632b60 ("bpf: sockmap: Move generic sockmap hooks from BPF TCP")
Signed-off-by: YueHaibing <yuehaibing@huawei.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Yonghong Song <yhs@fb.com>
Link: https://lore.kernel.org/bpf/20200320023426.60684-3-yuehaibing@huawei.com
If BPF_STREAM_PARSER is not set, gcc warns:
net/ipv4/tcp_bpf.c:483:12: warning: 'tcp_bpf_sendpage' defined but not used [-Wunused-function]
net/ipv4/tcp_bpf.c:395:12: warning: 'tcp_bpf_sendmsg' defined but not used [-Wunused-function]
net/ipv4/tcp_bpf.c:13:13: warning: 'tcp_bpf_stream_read' defined but not used [-Wunused-function]
Moves the unused functions into the #ifdef CONFIG_BPF_STREAM_PARSER.
Fixes: f747632b60 ("bpf: sockmap: Move generic sockmap hooks from BPF TCP")
Reported-by: Hulk Robot <hulkci@huawei.com>
Signed-off-by: YueHaibing <yuehaibing@huawei.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Reviewed-by: Lorenz Bauer <lmb@cloudflare.com>
Reviewed-by: Jakub Sitnicki <jakub@cloudflare.com>
Acked-by: Yonghong Song <yhs@fb.com>
Link: https://lore.kernel.org/bpf/20200320023426.60684-2-yuehaibing@huawei.com
The init, close and unhash handlers from TCP sockmap are generic,
and can be reused by UDP sockmap. Move the helpers into the sockmap code
base and expose them. This requires tcp_bpf_get_proto and tcp_bpf_clone to
be conditional on BPF_STREAM_PARSER.
The moved functions are unmodified, except that sk_psock_unlink is
renamed to sock_map_unlink to better match its behaviour.
Signed-off-by: Lorenz Bauer <lmb@cloudflare.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Reviewed-by: Jakub Sitnicki <jakub@cloudflare.com>
Acked-by: John Fastabend <john.fastabend@gmail.com>
Link: https://lore.kernel.org/bpf/20200309111243.6982-6-lmb@cloudflare.com
We need to ensure that sk->sk_prot uses certain callbacks, so that
code that directly calls e.g. tcp_sendmsg in certain corner cases
works. To avoid spurious asserts, we must to do this only if
sk_psock_update_proto has not yet been called. The same invariants
apply for tcp_bpf_check_v6_needs_rebuild, so move the call as well.
Doing so allows us to merge tcp_bpf_init and tcp_bpf_reinit.
Signed-off-by: Lorenz Bauer <lmb@cloudflare.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Reviewed-by: Jakub Sitnicki <jakub@cloudflare.com>
Acked-by: John Fastabend <john.fastabend@gmail.com>
Link: https://lore.kernel.org/bpf/20200309111243.6982-4-lmb@cloudflare.com
Only update psock->saved_* if psock->sk_proto has not been initialized
yet. This allows us to get rid of tcp_bpf_reinit_sk_prot.
Signed-off-by: Lorenz Bauer <lmb@cloudflare.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Reviewed-by: Jakub Sitnicki <jakub@cloudflare.com>
Acked-by: John Fastabend <john.fastabend@gmail.com>
Link: https://lore.kernel.org/bpf/20200309111243.6982-3-lmb@cloudflare.com
Prepare for cloning listening sockets that have their protocol callbacks
overridden by sk_msg. Child sockets must not inherit parent callbacks that
access state stored in sk_user_data owned by the parent.
Restore the child socket protocol callbacks before it gets hashed and any
of the callbacks can get invoked.
Signed-off-by: Jakub Sitnicki <jakub@cloudflare.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: John Fastabend <john.fastabend@gmail.com>
Link: https://lore.kernel.org/bpf/20200218171023.844439-4-jakub@cloudflare.com
sk_msg and ULP frameworks override protocol callbacks pointer in
sk->sk_prot, while tcp accesses it locklessly when cloning the listening
socket, that is with neither sk_lock nor sk_callback_lock held.
Once we enable use of listening sockets with sockmap (and hence sk_msg),
there will be shared access to sk->sk_prot if socket is getting cloned
while being inserted/deleted to/from the sockmap from another CPU:
Read side:
tcp_v4_rcv
sk = __inet_lookup_skb(...)
tcp_check_req(sk)
inet_csk(sk)->icsk_af_ops->syn_recv_sock
tcp_v4_syn_recv_sock
tcp_create_openreq_child
inet_csk_clone_lock
sk_clone_lock
READ_ONCE(sk->sk_prot)
Write side:
sock_map_ops->map_update_elem
sock_map_update_elem
sock_map_update_common
sock_map_link_no_progs
tcp_bpf_init
tcp_bpf_update_sk_prot
sk_psock_update_proto
WRITE_ONCE(sk->sk_prot, ops)
sock_map_ops->map_delete_elem
sock_map_delete_elem
__sock_map_delete
sock_map_unref
sk_psock_put
sk_psock_drop
sk_psock_restore_proto
tcp_update_ulp
WRITE_ONCE(sk->sk_prot, proto)
Mark the shared access with READ_ONCE/WRITE_ONCE annotations.
Signed-off-by: Jakub Sitnicki <jakub@cloudflare.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Link: https://lore.kernel.org/bpf/20200218171023.844439-2-jakub@cloudflare.com
When user returns SK_DROP we need to reset the number of copied bytes
to indicate to the user the bytes were dropped and not sent. If we
don't reset the copied arg sendmsg will return as if those bytes were
copied giving the user a positive return value.
This works as expected today except in the case where the user also
pops bytes. In the pop case the sg.size is reduced but we don't correctly
account for this when copied bytes is reset. The popped bytes are not
accounted for and we return a small positive value potentially confusing
the user.
The reason this happens is due to a typo where we do the wrong comparison
when accounting for pop bytes. In this fix notice the if/else is not
needed and that we have a similar problem if we push data except its not
visible to the user because if delta is larger the sg.size we return a
negative value so it appears as an error regardless.
Fixes: 7246d8ed4d ("bpf: helper to pop data from messages")
Signed-off-by: John Fastabend <john.fastabend@gmail.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Jonathan Lemon <jonathan.lemon@gmail.com>
Cc: stable@vger.kernel.org
Link: https://lore.kernel.org/bpf/20200111061206.8028-9-john.fastabend@gmail.com
Right now in tcp_bpf_recvmsg, sock read data first from sk_receive_queue
if not empty than psock->ingress_msg otherwise. If a FIN packet arrives
and there's also some data in psock->ingress_msg, the data in
psock->ingress_msg will be purged. It is always happen when request to a
HTTP1.0 server like python SimpleHTTPServer since the server send FIN
packet after data is sent out.
Fixes: 604326b41a ("bpf, sockmap: convert to generic sk_msg interface")
Reported-by: Arika Chen <eaglesora@gmail.com>
Suggested-by: Arika Chen <eaglesora@gmail.com>
Signed-off-by: Lingpeng Chen <forrest0579@gmail.com>
Signed-off-by: John Fastabend <john.fastabend@gmail.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Song Liu <songliubraving@fb.com>
Link: https://lore.kernel.org/bpf/20200109014833.18951-1-forrest0579@gmail.com
TLS 1.3 started using the entry at the end of the SG array
for chaining-in the single byte content type entry. This mostly
works:
[ E E E E E E . . ]
^ ^
start end
E < content type
/
[ E E E E E E C . ]
^ ^
start end
(Where E denotes a populated SG entry; C denotes a chaining entry.)
If the array is full, however, the end will point to the start:
[ E E E E E E E E ]
^
start
end
And we end up overwriting the start:
E < content type
/
[ C E E E E E E E ]
^
start
end
The sg array is supposed to be a circular buffer with start and
end markers pointing anywhere. In case where start > end
(i.e. the circular buffer has "wrapped") there is an extra entry
reserved at the end to chain the two halves together.
[ E E E E E E . . l ]
(Where l is the reserved entry for "looping" back to front.
As suggested by John, let's reserve another entry for chaining
SG entries after the main circular buffer. Note that this entry
has to be pointed to by the end entry so its position is not fixed.
Examples of full messages:
[ E E E E E E E E . l ]
^ ^
start end
<---------------.
[ E E . E E E E E E l ]
^ ^
end start
Now the end will always point to an unused entry, so TLS 1.3
can always use it.
Fixes: 130b392c6c ("net: tls: Add tls 1.3 support")
Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Reviewed-by: Simon Horman <simon.horman@netronome.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
sk_validate_xmit_skb() and drivers depend on the sk member of
struct sk_buff to identify segments requiring encryption.
Any operation which removes or does not preserve the original TLS
socket such as skb_orphan() or skb_clone() will cause clear text
leaks.
Make the TCP socket underlying an offloaded TLS connection
mark all skbs as decrypted, if TLS TX is in offload mode.
Then in sk_validate_xmit_skb() catch skbs which have no socket
(or a socket with no validation) and decrypted flag set.
Note that CONFIG_SOCK_VALIDATE_XMIT, CONFIG_TLS_DEVICE and
sk->sk_validate_xmit_skb are slightly interchangeable right now,
they all imply TLS offload. The new checks are guarded by
CONFIG_TLS_DEVICE because that's the option guarding the
sk_buff->decrypted member.
Second, smaller issue with orphaning is that it breaks
the guarantee that packets will be delivered to device
queues in-order. All TLS offload drivers depend on that
scheduling property. This means skb_orphan_partial()'s
trick of preserving partial socket references will cause
issues in the drivers. We need a full orphan, and as a
result netem delay/throttling will cause all TLS offload
skbs to be dropped.
Reusing the sk_buff->decrypted flag also protects from
leaking clear text when incoming, decrypted skb is redirected
(e.g. by TC).
See commit 0608c69c9a ("bpf: sk_msg, sock{map|hash} redirect
through ULP") for justification why the internal flag is safe.
The only location which could leak the flag in is tcp_bpf_sendmsg(),
which is taken care of by clearing the previously unused bit.
v2:
- remove superfluous decrypted mark copy (Willem);
- remove the stale doc entry (Boris);
- rely entirely on EOR marking to prevent coalescing (Boris);
- use an internal sendpages flag instead of marking the socket
(Boris).
v3 (Willem):
- reorganize the can_skb_orphan_partial() condition;
- fix the flag leak-in through tcp_bpf_sendmsg.
Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Acked-by: Willem de Bruijn <willemb@google.com>
Reviewed-by: Boris Pismenny <borisp@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The tcp_bpf_wait_data() routine needs to check timeo != 0 before
calling sk_wait_event() otherwise we may see unexpected stalls
on receiver.
Arika did all the leg work here I just formatted, posted and ran
a few tests.
Fixes: 604326b41a ("bpf, sockmap: convert to generic sk_msg interface")
Reported-by: Arika Chen <eaglesora@gmail.com>
Suggested-by: Arika Chen <eaglesora@gmail.com>
Signed-off-by: John Fastabend <john.fastabend@gmail.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
In tcp bpf remove we free the cork list and purge the ingress msg
list. However we do this before the ref count reaches zero so it
could be possible some other access is in progress. In this case
(tcp close and/or tcp_unhash) we happen to also hold the sock
lock so no path exists but lets fix it otherwise it is extremely
fragile and breaks the reference counting rules. Also we already
check the cork list and ingress msg queue and free them once the
ref count reaches zero so its wasteful to check twice.
Fixes: 604326b41a ("bpf, sockmap: convert to generic sk_msg interface")
Signed-off-by: John Fastabend <john.fastabend@gmail.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
A sockmap program that redirects through a kTLS ULP enabled socket
will not work correctly because the ULP layer is skipped. This
fixes the behavior to call through the ULP layer on redirect to
ensure any operations required on the data stream at the ULP layer
continue to be applied.
To do this we add an internal flag MSG_SENDPAGE_NOPOLICY to avoid
calling the BPF layer on a redirected message. This is
required to avoid calling the BPF layer multiple times (possibly
recursively) which is not the current/expected behavior without
ULPs. In the future we may add a redirect flag if users _do_
want the policy applied again but this would need to work for both
ULP and non-ULP sockets and be opt-in to avoid breaking existing
programs.
Also to avoid polluting the flag space with an internal flag we
reuse the flag space overlapping MSG_SENDPAGE_NOPOLICY with
MSG_WAITFORONE. Here WAITFORONE is specific to recv path and
SENDPAGE_NOPOLICY is only used for sendpage hooks. The last thing
to verify is user space API is masked correctly to ensure the flag
can not be set by user. (Note this needs to be true regardless
because we have internal flags already in-use that user space
should not be able to set). But for completeness we have two UAPI
paths into sendpage, sendfile and splice.
In the sendfile case the function do_sendfile() zero's flags,
./fs/read_write.c:
static ssize_t do_sendfile(int out_fd, int in_fd, loff_t *ppos,
size_t count, loff_t max)
{
...
fl = 0;
#if 0
/*
* We need to debate whether we can enable this or not. The
* man page documents EAGAIN return for the output at least,
* and the application is arguably buggy if it doesn't expect
* EAGAIN on a non-blocking file descriptor.
*/
if (in.file->f_flags & O_NONBLOCK)
fl = SPLICE_F_NONBLOCK;
#endif
file_start_write(out.file);
retval = do_splice_direct(in.file, &pos, out.file, &out_pos, count, fl);
}
In the splice case the pipe_to_sendpage "actor" is used which
masks flags with SPLICE_F_MORE.
./fs/splice.c:
static int pipe_to_sendpage(struct pipe_inode_info *pipe,
struct pipe_buffer *buf, struct splice_desc *sd)
{
...
more = (sd->flags & SPLICE_F_MORE) ? MSG_MORE : 0;
...
}
Confirming what we expect that internal flags are in fact internal
to socket side.
Fixes: d3b18ad31f ("tls: add bpf support to sk_msg handling")
Signed-off-by: John Fastabend <john.fastabend@gmail.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
When a skb verdict program is in-use and either another BPF program
redirects to that socket or the new SK_PASS support is used the
data_ready callback does not wake up application. Instead because
the stream parser/verdict is using the sk data_ready callback we wake
up the stream parser/verdict block.
Fix this by adding a helper to check if the stream parser block is
enabled on the sk and if so call the saved pointer which is the
upper layers wake up function.
This fixes application stalls observed when an application is waiting
for data in a blocking read().
Fixes: d829e9c411 ("tls: convert to generic sk_msg interface")
Signed-off-by: John Fastabend <john.fastabend@gmail.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
This adds a BPF SK_MSG program helper so that we can pop data from a
msg. We use this to pop metadata from a previous push data call.
Signed-off-by: John Fastabend <john.fastabend@gmail.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
We return 0 in the case of a nonblocking socket that has no data
available. However, this is incorrect and may confuse applications.
After this patch we do the correct thing and return the error
EAGAIN.
Quoting return codes from recvmsg manpage,
EAGAIN or EWOULDBLOCK
The socket is marked nonblocking and the receive operation would
block, or a receive timeout had been set and the timeout expired
before data was received.
Fixes: 604326b41a ("bpf, sockmap: convert to generic sk_msg interface")
Signed-off-by: John Fastabend <john.fastabend@gmail.com>
Acked-by: Song Liu <songliubraving@fb.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
This adds support for the MSG_PEEK flag when doing redirect to ingress
and receiving on the sk_msg psock queue. Previously the flag was
being ignored which could confuse applications if they expected the
flag to work as normal.
Signed-off-by: John Fastabend <john.fastabend@gmail.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
When converting sockmap to new skmsg generic data structures we missed
that the recvmsg handler did not correctly use sg.size and instead was
using individual elements length. The result is if a sock is closed
with outstanding data we omit the call to sk_mem_uncharge() and can
get the warning below.
[ 66.728282] WARNING: CPU: 6 PID: 5783 at net/core/stream.c:206 sk_stream_kill_queues+0x1fa/0x210
To fix this correct the redirect handler to xfer the size along with
the scatterlist and also decrement the size from the recvmsg handler.
Now when a sock is closed the remaining 'size' will be decremented
with sk_mem_uncharge().
Signed-off-by: John Fastabend <john.fastabend@gmail.com>
Acked-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Add a generic sk_msg layer, and convert current sockmap and later
kTLS over to make use of it. While sk_buff handles network packet
representation from netdevice up to socket, sk_msg handles data
representation from application to socket layer.
This means that sk_msg framework spans across ULP users in the
kernel, and enables features such as introspection or filtering
of data with the help of BPF programs that operate on this data
structure.
Latter becomes in particular useful for kTLS where data encryption
is deferred into the kernel, and as such enabling the kernel to
perform L7 introspection and policy based on BPF for TLS connections
where the record is being encrypted after BPF has run and came to
a verdict. In order to get there, first step is to transform open
coding of scatter-gather list handling into a common core framework
that subsystems can use.
The code itself has been split and refactored into three bigger
pieces: i) the generic sk_msg API which deals with managing the
scatter gather ring, providing helpers for walking and mangling,
transferring application data from user space into it, and preparing
it for BPF pre/post-processing, ii) the plain sock map itself
where sockets can be attached to or detached from; these bits
are independent of i) which can now be used also without sock
map, and iii) the integration with plain TCP as one protocol
to be used for processing L7 application data (later this could
e.g. also be extended to other protocols like UDP). The semantics
are the same with the old sock map code and therefore no change
of user facing behavior or APIs. While pursuing this work it
also helped finding a number of bugs in the old sockmap code
that we've fixed already in earlier commits. The test_sockmap
kselftest suite passes through fine as well.
Joint work with John.
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Signed-off-by: John Fastabend <john.fastabend@gmail.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
John Fastabend