netinet/tcp.h and linux/tcp.h have different lenghts but overlap. LDP
uses the former while iperf the latter. Accept both lengths for now as
we do not support exposing tcp metrics via ldp.
Type: improvement
Change-Id: I13a149d68715ed9451773630a3595c09c421aa29
Signed-off-by: Florin Coras <fcoras@cisco.com>
ldp stores the libc epfds it uses in conjunction with vcl epfds in vcl
attributes. Apart from being an ldp, as opposed to vcl, specific
attribute, the epfd is retrieved on session close with vls locks that
could be acquired from multiple threads.
Avoid grabbing locks when retrieving the attribute.
Type: improvement
Signed-off-by: Florin Coras <fcoras@cisco.com>
Change-Id: I5fc52d72667a7693b1557d209c3890a968422f1e
Multi-threaded apps that do not allocate per-thread workers
(multi-thread workers vcl config) ended up sharing worker state like the
select bitmaps and time among others. Those should not be shared.
To avoid this, make ldp workers thread local variables.
Type: fix
Signed-off-by: Florin Coras <fcoras@cisco.com>
Change-Id: Iabdcc413991dbaafff33f24187f7053a9c5a1270
Make sure num threads is 1 on process create and fork. Multi-thread
locks are applied once num threads exceeds 1.
For select, follow same pattern like epoll and add check for session
migration.
Type: fix
Signed-off-by: Florin Coras <fcoras@cisco.com>
Signed-off-by: Dave Wallace <dwallacelf@gmail.com>
Change-Id: I1edcd6c4c81b6b3caf8b00781b339414e8945b0e
LDP workers is used as vector but was initialized as a pool. There was
no side effect but ASAN does not properly unpoison memory and this
triggers false used-after-poison crashes.
Type: fix
Signed-off-by: Dmitry Valter <d-valter@yandex-team.com>
Signed-off-by: Florin Coras <fcoras@cisco.com>
Change-Id: Ie769dad0e86ab970de9929800d0a4131f846e70e
GCC 12 complains about such errors while the code itself looks good.
Type: fix
Signed-off-by: Jieqiang Wang <jieqiang.wang@arm.com>
Change-Id: I021719fdbf7d9bd93a12eac76aeac8cbca13a810
Maybe some function calls the wrapper funtion like read which invokes
the ldp_init_check in the other libs as a constructor before
ldp_constructor. Then the ldp has been initialized already when
ldp_constructor is invoked. And it's normal case, we shouldn't treat it
as an error.
So ldp_init should return success if ldp is initialized already instead
of an assert.
Type: fix
Change-Id: Ifa2a7b1d5471981a3f840b14a4fa5d48fb1f1374
Signed-off-by: Gao Feng <gfree.wind@outlook.com>
Type: improvement
Support SO_ORIGINAL_DST socket option to get original dst_ip4 and dst_port if nat44 rule enabled.
Change-Id: If00e00d03e48f3b78a23a68f1b078954d79dd0f7
Signed-off-by: qinyang <qiny@yusur.tech>
When msg->msg_controllen is set in recvmsg, the caller wants to receive
additional information about the messages. However, they might not
always be available. In that case, we should clear cmsg since
the caller uses CMSG_NXTHDR which might access uninitialized fields in
cmsg.
Type: fix
Change-Id: Ifdf9634bfcb1427f7ae3812014a46bfd7f4bc473
Signed-off-by: Steven Luong <sluong@cisco.com>
This avoids printing ldp debug messages while debug is disabled and vcl
is initializing.
Type: fix
Signed-off-by: Florin Coras <fcoras@cisco.com>
Change-Id: I5dfd1d59032db937fea146b6b84b8e26307a0de0
Otherwise if vcl epoll lt events are ignored by the app, libc and vcl mq
events are never drained.
Type: fix
Signed-off-by: Florin Coras <fcoras@cisco.com>
Change-Id: I1e22f6da46d56236c52714181f6c20dcb80a33a5
Accept one spurious wakeup from vcl in epoll_pwait_eventfd to avoid
returning zero events to app without timeout.
Type: improvement
Signed-off-by: Florin Coras <fcoras@cisco.com>
Change-Id: I447c7f8176413c562be28605376a92d15e22a1f9
Nginx recreates epfds. Make sure ldp tracks the event and recreates the
libc epfd or eventfd flavor of epoll pwait will not work.
Type: fix
Signed-off-by: Florin Coras <fcoras@cisco.com>
Change-Id: I2994bead9494f0fbb85dd32767cecc1cf69ff6eb
Old distros Centos 8 / Ubuntu 18.04 header files doesn't have UDP_SEGMENT
declared, define UDP_SEGMENT to right value if not defined.
Type: fix
Fixes: eff5f7aea8 ("vcl: ldp support for ip_pktinfo")
Signed-off-by: Tianyu Li <tianyu.li@arm.com>
Change-Id: I99314b895e7d09962a36e7f5582c09d0d77563dc
For some apps(e.g. wrk2) upon vpp hoststack, ldp_epoll_pwait()
is called. In this function, epoll fd was created on one thread,
but it is now used on another thread. The vcl worker index is still
invalid, so the fetched ldp worker is also invalid and can corrupt
some already allocated memory.
Just as the ldp_epoll_pwait_eventfd(), make sure the vcl worker is valid
before getting the ldp worker in ldp_epoll_pwait().
Type: fix
Signed-off-by: Liangxing Wang <liangxing.wang@arm.com>
Change-Id: I2ec23a4b5d5b0879a06642ffd80f95e948af4274
Control use of apis that rely on _GNU_SOURCE being defined with compile
time macro.
Also fixes sendmmsg and recvmmsg which were not probably wrapped.
Type: improvement
Signed-off-by: Florin Coras <fcoras@cisco.com>
Change-Id: I207de23210d4b9dc960bb4289159502760c5614d
Some upon apps(e.g. Nginx-quic) package it's several protocol buffers into a struct msg
which is a combination of gso_buffer and gso_size.
but if HostStack regardless the gso_size to the buffer and split the buffer with default mss,
that cause peer client failed on parsing the package.
Type: improvement
Signed-off-by: Dou Chao <chao.dou@intel.com>
Change-Id: I805eb642be826038ba96d1b85dad8ec0c0f6c459
Signed-off-by: Dou Chao <chao.dou@intel.com>
The restriction of vlsh_bit_val only effect select/pselect, so move the
check to select/pselect function.
Type: fix
Signed-off-by: wanghanlin <wanghanlin@corp.netease.com>
Change-Id: I3585c83cfc2f0a2fd834450be650dccda67e3faa
vlsh may not belong to the current vcl worker.
Type: fix
Signed-off-by: liuyacan <liuyacan@corp.netease.com>
Change-Id: I75f962313e538fcd2b1cc6dffca53997648ca43e
This commit does following:
- Change the behavior of shutdown() with SHUT_RDWR flag.
- Check SHUT_RD flag when read()
- Change the errno when write() after SHUT_WR
- Remove unused code
All the above modification passed the packetdrill test.
Type: improvement
Signed-off-by: liuyacan <liuyacan@corp.netease.com>
Change-Id: I0c81f52e563562e58580d70976526b898e65e915
Make sure we don't return error if app tries to setsockopt it.
Type: improvement
Signed-off-by: Florin Coras <fcoras@cisco.com>
Change-Id: I1a9a6a8da03380b0b6a6d01efed9b672a3f0b1f9
Some app(e.g. Envoy) may call shutdown() instead of close() when
draining connection.
Type: improvement
Signed-off-by: liuyacan <liuyacan@corp.netease.com>
Change-Id: I9543b9ca3caa87b10b134fd1fc4019124e41e4d2
We call vls_epoll_wait with 0 timeout to process unhandled_evts_vector
in ldp_epoll_pwait_eventfd. But vls_epoll_wait will then call
vppcom_epoll_wait_eventfd to process events in event queue. If there are
continuous events to be handled, then libc_epoll_pwait have no chance to be called.
Type: fix
Signed-off-by: wanghanlin <wanghanlin@corp.netease.com>
Change-Id: Ia4f61d0d438a3475bfde9f6715805274671d4e75
Currently, mutlti-threads only support one dispatch thread and multiple
worker threads, eventually only dispatch thread is a vcl worker and can
interact with epoll.
This patch will register all threads as vcl worker, and then each
thread can interact with epoll now. Moreover, session migration also
supported, such as socket created in thread A and used (bind, connect
and etc.) in thread B.
Type: feature
Signed-off-by: hanlin <hanlin_wang@163.com>
Change-Id: Iab0b43a33466968c1423d7d20faf1460c8589d91
Now, libc epfd and vls epfd are independent and can only epoll_wait independently without timeout, then app calling epoll_wait will occupy high CPU. So we nest vcl_mq_epfd into libc epfd when using eventfd with VPP, and then we can only epoll_wait libc epfd with specified timeout.
Type: feature
Signed-off-by: hanlin <hanlin_wang@163.com>
Change-Id: I6b6e0f501c769e186714bfbc187cfaed2533b4c2
Signed-off-by: hanlin <hanlin_wang@163.com>
