The current code does not check the value range of the 'mode' parameter
and would later run into OSMO_ASSERT(), rather than rejecting such a
mode from the very beginning.
Change-Id: I10dd612487638f456d0ad59c2cca203f1e098da3
Related: OS#5751
The two functions of the SCTP socket interface we use in osmo-* are
sctp_send() and sctp_recvmsg(). We do not use sctp_sendmsg() at all,
so let's make sure the mode is named correctly.
Change-Id: Ie2d1c7ce6f211dbe025a0e843ad733443102ea15
Related: OS#5751
The function is basically a shortcut for getaddrinfo with storing the
output data into our 'struct osmo_sockaddr'.
Change-Id: I6b5c0bf8ca97e6358d992fb2ff45ffd53ba15197
Related: SYS#6657
Currently calling this function with n_ubits == 0 would result in
requesting one character from the application (via the .tx_cb()),
but not actually transmitting anything. Make it return early.
Change-Id: Icbf99a9f2f6fa64dd71a5f37922f9001577c6c97
Related: OS#4396
Whenever we encounter a parity and/or a framing error, we should
call the .rx_cb() immediately, even if this was the first
character in the receive buffer.
Change-Id: I73fab1a5c196d2dbdfe98b0c20d8dadbd22f4f64
Related: OS#4396
Let the API user pass their own default config when allocating
a soft-UART. Make the default config publicly accessible.
Change-Id: I7e78d60c747a8805064d5e4bacfd47a30bc65cba
Related: OS#4396
The problem with a single function controlling both Rx and Tx is
that enabling/disabling one of the directions requires knowing
state of the other one. In other words, disabling Tx requires
knowing the state of Rx, which may be inconvenient.
Change-Id: Ieacc7e639304eeb14fdb298c7e14d772c136ca6e
Related: OS#4396
This patch brings a Work-in-Progress implementation of the software
UART (Universal Asynchronous Receiver/Transmitter) to libosmocore.
Not only it will be useful in the context of retro-networking, but
also it's needed for the MS-side CSD implementation (see OS#4396).
It should be noted that the definition of struct osmo_soft_uart
is intentionally kept private, since the API is not stable yet.
Currently, the following limitations apply:
* Only the receiver part is implemented, the transmitter is TBD.
* Parity checking is not implemented in the receiver part.
* Software flow control is not implemented.
These missing components will be addressed in subsequent patches.
Change-Id: I2ca95963fd5852ddb89bdd35b86b31489127fe84
- Instead of using the osmo_fd API to call read() on the socket's file
descriptor each time (unused) data is received, simply open the
socket and never read
Related: OS#6213
Change-Id: I4025920d5f62d17133e9b5fe81cd34a88c4f20b5
- Adapt decl. of 'struct gsmtap_inst' for usage of Osmo IO while maintaining backwards compatibility
- Maintain legacy behavior without any message queues if osmo_io_mode is zero
Related: OS#6213
Change-Id: Iadbbef74e3add7001d84dd6b68f51eac293e44d0
This avoids an infinite recursion when sending a gsmtap log message
causes a log message.
Temporarily set target->loglevel higher than LOGL_FATAL, which
effectively disables logging for that target. Other targets like stderr
will still log this message so there is still an indication that
something went wrong.
Change-Id: I19203cadbad6019a3834793b8ac816d903fe088e
Related: OS#6213
This fixes multi-line color clobbering in logging daemons like
systemd-journald, which work with single-lines only.
Change-Id: Ia7de9d88aa5ac48ec0d5c1a931a89d21c02c5433
Closes: OS#6249
- Use forward decl. of struct gsmtap_inst in header
- Remove 'static inline' attributes from gsmtap_inst_fd() declaration,
move function definition to gsmtap_util.c and mark it as deprecated
- Add gsmtap_inst_fd2() as replacement for gsmtap_inst_fd()
Related: OS#6213
Change-Id: Ibe1a51205a6df764571b6d074e365825555609a5
These APIs are used to bind or unbind an active socket adding or
removing addresses from the existing set.
Related: OS#6077
Change-Id: Ifc6e7d643c2a0c53f479bfd0d5c36d08c0c01953
Allow the callbacks to be NULL, but then sending/receiving is disabled.
There are some cases where we only care about writing to or reading from
an fd.
Change-Id: I11ce072510b591f7881d09888524426579bd0169
Avoid uninitialized read, found with valgrind
Syscall param recvmsg(msg) points to uninitialised byte(s)
at 0x49FD865: __recvmsg_syscall (recvmsg.c:27)
by 0x49FD865: recvmsg (recvmsg.c:41)
by 0x4891FAE: iofd_poll_ofd_cb_recvmsg_sendmsg (osmo_io_poll.c:66)
by 0x48921B2: iofd_poll_ofd_cb_dispatch (osmo_io_poll.c:119)
by 0x48941F1: poll_disp_fds (select.c:419)
by 0x4894299: _osmo_select_main (select.c:457)
by 0x4894304: osmo_select_main (select.c:496)
by 0x10DC3E: test_segm_ipa_stream_srv_run (stream_test.c:628)
by 0x10E2A5: main (stream_test.c:879)
Address 0x1ffefffa68 is on thread 1's stack
in frame #1, created by iofd_poll_ofd_cb_recvmsg_sendmsg (osmo_io_poll.c:45)
Change-Id: I21114ad57784126cfdeb4a932ed44dbf23946fbe
Dequeue and free any excess messages, in case the new queue length
is shorter than the old.
Related: OS#5774
Change-Id: Ibfe51a2faf29f8ae160a9c330c9af0d09b5a9002
This allows setting several socket paramets which can only be set before
SCTP INIT phase (connect()).
Since no release with osmo_sock_init2_multiaddr2() happened yet, we are
allowed to extend struct osmo_sock_init2_multiaddr_pars without bumping
the version field.
Related: SYS#6558
Change-Id: Ibc55932d954279927d1e70ccce1e8bf32b180d44
The OSMO_SOCK_F_SCTP_ASCONF_SUPPORTED flag was added recently to enable
use of ASCONF in osmo_sock_inti2_multiaddr. Since we didn't release yet,
we are still in time to get rid of this flag.
It turns out that we'll want to add more features which must be set
before SCTP INIT (connect()) time, like setsockopt SCTP_INITMSG, which
in turn contains a struct with several parameters.
Hence, adding flags for all those makes no sense.
Instead, add a new API which allows passing an extensible struct which
allows much more fine-grained setup from the caller.
The new struct "pars" parameter is non-const on purpose, in order to be
able to extend the struct in the future if we wish to get/obtain some
data from the socket.
Related: SYS#6501
Related: SYS#6558
Change-Id: I1f6fd09a79b0a2bd794e5669d933be25bbf1eeaa
msg was made a parent of msghdr after discussion in change
I3a279b55a3adff96948120683c844e1508d0ba94
It turns out this violates some assumptions made in osmo_io,
specifically that the user read callback shall free msg, but we expect
msghdr to remain valid until after that callback returns.
In general I think it is cleaner to make iofd a parent of msghdr.
Change-Id: I41277190e3020cd8fa625bd57a743973e2a65c4b
Avoid aborting the entire socket creation, since those sockopt are fairly
new in the kernel (v5.4), and the feature can still be enabled in older
versions system-wide with sysctl. The worst it can happen is that the
ability to submit address updates (such as Peer Primary Address) doesn't
work, which is not really critical.
Related: SYS#6501
Change-Id: Iff261c8592b6b3c4237c90c84b4e8e921e3c4a65
Ensure that a msgb has the proper talloc parent:
All msgbs inside an iofd get the iofd as parent. Received msgbs are reparented
to iofd->msgb_alloc.ctx (which was set in osmo_iofd_setup()) before
being passed to the receive callback.
Before this change the code could fail for msgbs that are submitted via uring
where the (failed) write returns after the iofd has already been
osmo_iofd_free()d. free()ing the iofd is deferred until the write
completes, but the (iofd) parent context could have been free()d in the
meantime.
Change-Id: I3a279b55a3adff96948120683c844e1508d0ba94
Marked as WIP because I'm not entirely sure if negative index lookups
were allowed on purpose (although I doubt it)
Change-Id: Iba81873cbc435229599b1a8139cf4b627f8e6924
The SCTP_ASCONF_SUPPORTED sockopt, which enables ASCONF features
for the SCTP associations managed by the SCTP socket, gets its default
value from to sysctl "net.sctp.addip_enable", which, at least up to
current kernel 6.4.8 is set to "0" (disabled) by default.
As a result, ASCONF related features such as
setsockopt(SCTP_SET_PEER_PRIMARY_ADDR) fail with -EPERM
since ASNCONF extensions where not negotiated during the SCTP init
handshake.
This commit adds support for setting SCTP_ASCONF_SUPPORTED=1 during
socket creation, since that needs to be applied before the first
bind()/connect() (before first assoc is created).
Furthermore, for ASCONF features to work properly, the assoc also needs
to announce/use the AUTH extension, as per RFC5061 section 4.2.7.
Otherwise, the peer receiving an SCTP INIT with
ExtensionFeatures=ASCONF,ASCONF_ACK but without AUTH, will reject the
assoc with an ABORT since it's not complying with spec (this behavior
can be tweaked through sysctl "net.sctp.addip_noauth_enable").
The AUTH extension is enabled through sockopt SCTP_AUTH_SUPPORTED, and
its default value is that of sysctl "net.sctp.auth_enable", which is
also disabled (0) by default.
Kernel support for those is relatively new:
SCTP_FUTURE_ASSOC added in: 80df2704a375bb4b3c9c5cce9c00052361b16d61
Follows: v5.0-rc4
Precedes: v5.1-rc1
SCTP_ASCONF_SUPPORTED added in: df2c71ffdfae58961981d7cbcccea93688fc4e96
Follows: v5.3-rc5
Precedes: v5.4-rc1
SCTP_AUTH_SUPPORTED added in: 56dd525abd56f7acd7b44a52935726e3ada4916c
Follows: v5.3-rc5
Precedes: v5.4-rc1
Related: OS#6076
Change-Id: Iac07031927b66a9d32d2bb2faab817e4c922a359
Use talloc_steal() if a msg is passed in to osmo_io when sending. This
avoids the message being free()d early in case the original parent is
free()d.
Change-Id: Ie36bd68a8bd63e67d76fb41996f8fdf99f51d96c
We need to account for the fact that segmentation_cb() could have
changed the length by calling msgb_pull(). Calculate the new len
according to the new tail/data pointers.
Change-Id: I5486ddc0d3345e92b20cbc6e5bcf2cefea3958c8