| Age | Commit message (Collapse) | Author |
|---|
|
Commit 5ca5fb71 ensured that they were closed immediately at transport
removal, but in so doing broke their closing at stop service completely,
by removing the timer that caused sockets to be closed even belatedly.
Monitor on the service process to make it happen.
This could still be improved, since stop_service listening ports aren't
closed until after the service process has died. They could be closed
earlier in the case of stop_service.
|
|
The transport interface documented in diameter_transport(3) is used to
start/stop accepting/connecting transport processes: they're started
with a function call, and told to die with their parent process. In the
accepting case, both diameter_tcp and diameter_sctp start a listening
process when the first accepting transport is started. However, there's
no way for a listening process to find out that that it should stop
listening when transport configuration is removed.
Both diameter_tcp and diameter_sctp have used a timer to terminate the
listening process after all existing accepting processes have died as a
consequence of transport removal. The problem with this is that nothing
stops a new client from connecting before this, and also that no new
transport can succeed in opening the same listening port (eg.
reconfiguration) until the old listener dies.
This commit solves the problem by adding diameter_reg:subscribe/2, to
allow callers to subscribe to messages about added/removed associations.
A call to diameter:add_transport/2 results in a new child process that
registers a term that a listening process subscribes to. Transport
removal results in the death of the child, and the resulting
notification to the listener causes the latter to close its socket and
terminate.
This is still an internal interface, but the subscription mechanism
should probably be made external (eg. a diameter:subscribe/1 that can
be used to subscribe to specified messages), so that transport modules
other than diameter's own can make use of it. There is no support for
soft upgrade.
|
|
* anders/diameter/overload/OTP-13330:
Suppress dialyzer warning
Remove dead case clause
Let throttling callback send a throttle message
Acknowledge answers to notification pids when throttling
Throttle properly with TLS
Don't ask throttling callback to receive more unless needed
Let a throttling callback answer a received message
Let a throttling callback discard a received message
Let throttling callback return a notification pid
Make throttling callbacks on message reception
Add diameter_tcp option throttle_cb
|
|
This one:
diameter_tcp.erl:928: (call)
The call diameter_tcp:throttle({'timeout',_},#transport{socket::port() | {'sslsocket',_,_},parent::pid(),module::atom(),frag::binary() | {non_neg_integer(),non_neg_integer(),binary(),[binary()]},ssl::boolean() | [any()],timeout::'infinity' | non_neg_integer(),tref::'false' | reference(),flush::boolean(),throttle_cb::'false' | fun() | maybe_improper_list(fun() | maybe_improper_list(any(),[any()]) | {atom(),atom(),[any()]},[any()]) | {atom(),atom(),[any()]},throttled::'true' | binary()})
will never return since it differs in the 1st argument from the
success typing arguments:
('discard' | 'ok' | binary() | pid() | {'discard' | 'ok' | binary() | pid(),'false' | fun() | [fun() | [any()] | {atom(),atom(),[any()]}] | {atom(),atom(),[any()]}},#transport{socket::port() | {'sslsocket',_,_},parent::pid(),module::atom(),frag::binary() | {non_neg_integer(),non_neg_integer(),binary(),[binary()]},ssl::boolean() | [any()],timeout::'infinity' | non_neg_integer(),tref::'false' | reference(),flush::boolean(),throttle_cb::'false' | fun() | [fun() | [any()] | {atom(),atom(),[any()]}] | {atom(),atom(),[any()]},throttled::binary()})
It's true that the clause doesn't return, because of the throw, and
that's the intention.
|
|
Orphaned in commit 9298872b.
|
|
That is, don't assume that it's only diameter_tcp doing so: allow it to
be received when not throttling. This lets a callback module trigger a
new throttling callback itself, but it's not clear if this will be
useful in practice.
|
|
|
|
In particular, let a callback decide when to receive the initial
message.
|
|
TCP packets can contain more than one message, so only ask to receive
another message if it hasn't already been received.
|
|
As discussed in the parent commit. This is easier said than done in
practice, but there's no harm in allowing it.
|
|
This can be used as a simple form of overload protection, discarding the
message before it's passed into diameter to become one more request
process in a flood. Replying with 3004 would be more appropriate when
the request has been directed at a specific server (the RFC's
requirement) however, and possibly it should be possible for a callback
to do this as well.
|
|
In addition to returning ok or {timeout, Tmo}, let a throttling callback
for message reception return a pid(), which is then notified if the
message in question is either discarded or results in a request process.
Notification is by way of messages of the form
{diameter, discard | {request, pid()}}
where the pid is that of a request process resulting from the received
message. This allows the notification process to keep track of the
maximum number of request processes a peer connection can have given
rise to.
|
|
The callback is now applied to the atom 'false' when asking if another
message should be received on the socket, and to a received binary
message after reception. Throttling on received messages makes it
possible to distinguish between requests and answers.
There is no callback on outgoing messages since these don't have to go
through the transport process, even if they currently do.
|
|
To let a callback module decide whether or to receive another message
from the peer, so that backpressure can be applied when it's
inappropriate. This is to let a callback protect against reading more
than can be processed, which is otherwise possible since diameter_tcp
otherwise always asks for more.
A callback is made after each message, and can answer to continue
reading or to ask again after a timeout. It's each message instead of
each packet partly for simplicity, but also since this should be
sufficiently fine-grained. Per packet would require some interaction
with the fragment timer that flushes partial messages that haven't been
completely received.
|
|
Record field types have been modified due to commit 8ce35b2:
"Take out automatic insertion of 'undefined' from typed record fields".
|
|
The diffs are all about adapting to the OTP 18 time interface. The code
was previously backwards compatible, falling back on the erlang:now/0 if
erlang:monotonic_time/0 is unavailable, but this was seen to be a bad
thing in commit 9c0f2f2c. Use of erlang:now/0 is now removed.
|
|
By doing away with more wrapping that the parent commit started to
remove.
|
|
OTP-12845
* bruce/change-license:
fix errors caused by changed line numbers
Change license text to APLv2
|
|
A listener process in diameter_sctp starts accepting transport processes
as required, either as associations are established or as diameter asks
for a processes to be started. Since this can happen in any order, the
listener maintains two queues: one for processes that diameter has
requested and which are waiting to be given an association, another for
processes that have been started to become owners of an association but
are waiting for diameter to request them. Only one queue at a time is
non-empty. The first queue's length is bounded by the number of
accepting processes configured as pool_size. Entries in the second queue
are short-lived since diameter starts a replacement transport process
whenever an existing one dies or communicates that it has an
association.
The two queues were previously implemented in an ets ordered_set, whose
keys were the pid() of transport processes. Removing an element from the
queue was then done with ets:first/1. The problem with this it's not
really a queue: there's no guarantee that pid-ordering is the same as
the order in which processes are started. If it isn't then it's possible
that an established association never be given to diameter as a
transport process if there's always a newer association whose pid sorts
first. This isn't a problem in practice since it would require new
associations to be established faster than diameter starts transport
processes, but redo the implementation as a queue, with strict FIFO
semantics.
|
|
The changes in some of the previous commits assume application restart.
|
|
Don't pass an association id that's no longer used.
|
|
In particular, don't give the accepting transport process the listening
socket. It was used to match the initial sctp message received in a
peeloff message, but replace the socket in the forwarded message
instead.
|
|
The existing code was a remnant of the pre-peeloff implementation.
There's no need to close anything but the whole socket.
|
|
Listener death should have no effect on a peeled off association.
|
|
Forwarding an sctp message from the listener process at the same time
that the controlling process is changed means there's no guarantee that
the message order will be preserved. Selectively receive the peeloff
message before entering the gen_server loop to ensure the order is
preserved.
|
|
This is not the case under Solaris for one: successive
associations can receive the same association id as a result of peeloff,
the id only being unique for the controlling port, not for the listening
port as is the case under Linux for example. This made for many failures
in the diameter test suites, the traffic suite in particular.
Peeloff in diameter_sctp was introduced in 9a671bf0, before which the
assumption was fine since it was the listening process that owned all
associations. (Which obviously had other drawbacks.) Other remnants of
the pre-peeloff implementation have also been removed: that the listener
process might receive a message on a socket after peeloff for one.
Peeloff in gen_sctp became available in commit 067cfe79, after the
original implementation of diameter_sctp.
This is trace on the unpatched code showing id reuse under Solaris:
+ {trace_ts,<0.103.0>,call,
{diameter_sctp,handle_info,
[{sctp,#Port<0.1625>,
{127,0,0,1},
35904,
{[],{sctp_assoc_change,comm_up,0,32,32,1}}},
{listener,#Ref<0.0.1.948>,#Port<0.1625>,4,
57384,
{-4,61481},
#Ref<0.0.8.12>,
[]}]},
{1432,458752,612168}}
+ {trace_ts,<0.103.0>,call,
{diameter_sctp,handle_info,
[{sctp,#Port<0.1625>,
{127,0,0,1},
35905,
{[],{sctp_assoc_change,comm_up,0,32,32,1}}},
{listener,#Ref<0.0.1.948>,#Port<0.1625>,4,
57384,
{-3,61481},
#Ref<0.0.8.12>,
[]}]},
{1432,458752,613042}}
The result was this, when the second association was incorrectly
forwarded to the first association's controlling process:
** {function_clause,
[{diameter_sctp,transition,
[{peeloff,#Port<0.1635>,
{sctp,#Port<0.1625>,
{127,0,0,1},
35892,
{[],{sctp_assoc_change,comm_up,0,32,32,1}}},
[]},
{transport,<0.107.0>,accept,#Port<0.1634>,1,undefined,{32,32},0}],
[{file,"transport/diameter_sctp.erl"},{line,561}]},
{diameter_sctp,t,2,[{file,"transport/diameter_sctp.erl"},{line,549}]},
{diameter_sctp,handle_info,2,
[{file,"transport/diameter_sctp.erl"},{line,397}]},
{gen_server,try_dispatch,4,[{file,"gen_server.erl"},{line,614}]},
{gen_server,handle_msg,5,[{file,"gen_server.erl"},{line,680}]},
{proc_lib,init_p_do_apply,3,[{file,"proc_lib.erl"},{line,238}]}]}
|
|
|
|
Commit 4b691d8d made it possible for accepting transport processes to be
started concurrently, and commit 77c1b162 adapted diameter_sctp to this,
but missed that the publication of the listener process in diameter_reg
has to precede the return of its start function. As a result, concurrent
starts could result in multiple listener processes.
|
|
* anders/diameter/time/OTP-12439:
Use new time api in test suites
Use new time api in implementation
|
|
* anders/diameter/pool/OTP-12428:
Fix SCTP match blunder in suites
Be backwards compatible with diameter_sctp listener state
Add gen_tcp testcase that fails sporadically
Simplify transport suite
Remove (ancient) dead code
Don't orphan slave nodes in example suite
Refresh example code
Improve language consistency in diameter(1)
Add pool suite to test transport_opt() pool_size
Adapt tcp/sctp transport modules for pool_size > 1
Add transport_opt() pool_size
|
|
In particular, deal with the deprecation of erlang:now/0 in OTP 18. Be
backwards compatible with older releases: the new api is only used when
available.
The test suites have not been modified.
|
|
Commit 24993fc2 modified the state even in the case that the new
pool_size option the change was introduced to support was not used.
Doing so made downgrade impossible since old code would not be prepared
for the modified state. Retain a compatible state, so that simple code
replacement is enough.
|
|
Commit 9a671bf0 removed the need for diameter_sctp to send outgoing
messages through the listening process. That was prior to R5B02, so the
clause isn't need for any upgrade case.
|
|
In particular, that starts for the same transport reference can now be
concurrent. Looking up a listener process and starting a new one if not
found did handle this (more than one process could find no listener),
and diameter_sctp assumed there could only be one transport process
waiting for an association.
|
|
As suggested in supervisor(3). The leaves of the supervision tree should
determine the timeouts.
|
|
Introduced in commit ed6395a6. The {stream, Id} transport_data set upon
reception is passed back to us by default when receiving the answer to a
request message, so even capabilities exchange failed.
|
|
OTP-10229 (commit c4592b69) added these function to give access to all
addresses on a multihomed endpoint, their singular siblings not
returning anything useful in this case.
This fixes {accept, Match} config, which matches peer addresses against
configured addresses or regexps to decide whether or not a newly
established association should be retained. The functionality was added
in OTP-10893 (commit 9bbf27eb) but predated OTP-10229 by a few months.
It also fixes the addresses shown for SCTP associations in
diameter:service_info/2 output.
|
|
* anders/diameter/17.0_release/OTP-11605:
vsn -> 1.6
Remove upgrade-related code
Update appup for 17.0
Avoid type gen_sctp:open_option() until it actually exists
|
|
No longer needed to update code in runtime since the emulator is
restarted at a major release.
|
|
The type's existence is the subject of OTP-11139, which has been
gathering dust since R16B.
http://erlang.org/pipermail/erlang-bugs/2013-September/003765.html
|
|
The module uses the transport_data field of record diameter_packet to
communicate the stream on which the an incoming message is received and
on which an outgoing message should be sent, the previous interface
being that both are communicated as a tuple of the form {stream, Id}.
However, since diameter retains the value of an incoming request's
transport_data unless the corresponding answer message specifies
otherwise, the behaviour in this case is to send an answer on the
outbound stream with the same identifier as the that of the inbound
stream on which the request was received. If the inbound stream id is
greater than or equal to the number of outbound streams then this is
guaranteed to fail, causing the transport process in question to
terminate. There is no relationship between inbound and outbound stream
identifiers so diameter_sctp's imposition of one is simply wrong.
Outbound stream ids are now communicated with a different tuple:
{outstream, Id}, interpreted modulo the number of outbound streams.
Thus, retention of an inbound request's transport_data has no effect on
the selection of an outbound stream.
The change in interface is not strictly backwards compatible because of
the new atom for the outbound stream. However, as there is currently no
documented way of obtaining the available number of outbound streams for
a peer connection, there is no way for a client to have known the range
of ids from which it could reliably have chosen with the previous
interface, so any setting of the outbound stream has probably been
unintentional. Not explicitly specifying an outbound stream now results
in a round-robin selection.
|
|
Option 'accept' allows remote addresses to be configured as tuples or
regular expressions. The remote addresses for any incoming (aka
accepted) connection/association are matched against the configured
values, any non-matching address causing the connection/association to
be aborted.
|
|
The third argument to start/3 was just wrong.
|
|
|
|
Use the default address address (as selected by gen_tcp) if none is
configured, passing it in the new 'connected' message introduced by the
previous commit.
The corresponding update to diameter_sctp has to wait until problems
with inet:sockname/1 are resolved: the function currently only returns
one address, and sometimes {0,0,0,0}. See OTP-11018.
|
|
|
|
Don't start a new timer with each incoming message. Instead, start a
timer at timeout and flush after two successive timeouts with no message
reception.
|
|
|
|
Which will be the case in R16B.
|
|
|
