| Age | Commit message (Collapse) | Author |
|---|
|
That dialyzer hasn't noticed is broken.
|
|
Despite what the Efficiency Guide says about match being more efficient,
split_binary appears to be slightly faster. (Although this one
extraction is a drop in the bucket.) Binary optimizations aren't an
issue during decode.
|
|
Since messages are accumulated by appending binaries as of three
commits back, the accumulated binary is prone to being copied, as
discussed in the Efficiency Guide. Matching the Message Length header
as bytes are being accumulated is one cause of this, so work around it
by splitting the binary and extracting the length without a match.
This doesn't feel like something that should be necessary: that matching
a binary would cause an append to copy isn't obvious. The first attempt
at simplifying the accumulation was simply to append an incoming binary
to the current fragment, match against <<_, Len:24, _/binary>> to
extract the length, and then test if there are enough bytes for a
message. This lead to horrible performance (response times for 2 MB
messages approximately 100 times worse than previously), and it wasn't
at all obvious that the reason was the accumulated binary being copied
with each append as a result of the match. Using split_binary avoids the
match context that forces the copying.
|
|
Not with each setopts to re-activate the socket.
|
|
Only reset the fragment after all messages have been extracted.
|
|
Appending to a binary is efficient, so just append message fragments
Only match out the length once per message since doing so for every
packet from TCP causes the binary to be copied.
|
|
Create less garbage.
|
|
That dialyzer hasn't noticed is broken.
|
|
Also inline incr/8 and associated functions that were needed for the
compiler to accept the optimization, since this does make for a
measuable improvement.
|
|
With ERL_COMPILER_OPTIONS=bin_opt_info, before:
base/diameter_codec.erl:508: Warning: NOT OPTIMIZED: the binary matching instruction that follows in the same function have problems that prevent delayed sub binary optimization (probably indicated by INFO warnings)
And after:
base/diameter_codec.erl:508: Warning: OPTIMIZED: creation of sub binary delayed
This has a surprisingly large impact on the performance of
diameter_codec:collect_avps/2: about 15% faster in one testcase on a RAR
with 7 AVPs.
|
|
The index field in record diameter_avp was previously used to enumerate
AVPs so that the list could be returned in some cases, since
diameter_codec:collect_avps/2 (now /1) reversed the order. That's no
longer the case as of the grandparent commit, so use the field to
enumerate instances of the same AVP instead, and only when arities are
being checked, to save having to look them up in the map when checking
for 5009 errors, or counting AVPs at all in diameter_codec:collect_avps/1.
|
|
Extract strict_arities once and pass it through as an argument.
|
|
Decode has previously been two passes: first chunk the message into a
reversed list of toplevel diameter_avp records, then fold through the
reversed list to build the full result. Various workarounds have made it
a bit more convoluted than it should be however. Rework it completely,
turning the previous 2-pass tail-recursive implementation into a 1-pass
body recursive one.
The relay decode still exists in diameter_codec, as a stripped down
version of the full decode in diameter_gen.
|
|
|
|
To be able to disable the relatively expensive check that the number of
AVPs received in a message or grouped AVP agrees with the dictionary in
question. The may well be easier for the user in handle_request/answer
callbacks, when digesting the received message, and in some cases may
not be important.
The check at encode can also be disabled, allowing messages that don't
agree with the dictionary in question to be sent, which can be useful in
test (at least).
|
|
As noted in the parent commit, the wrong AVPs were reported, being
counted from the back of the message instead of the front. Both 5005 and
5009 errors are now detected after the message is decoded.
|
|
Stop counting when there can be no arity errors.
|
|
Count as AVPs are encoded instead.
|
|
Use the same [MsgName | Avps] representation as for the list decode, but
with Avps a map instead of a AVP name/values list. As a result, don't
set the message/AVP name on an additional key in the map, which felt a
bit odd. Messages are [MsgName :: atom() | map()], Grouped AVPs are just
map().
Fix at least one problem in the traffic suite along the way: with
decode_format false, the own decode in to_map/2 didn't know whether or
not to decode strings, resulting on some failures.
|
|
Undocumented, for transforming a map decode to record.
The record decode becomes more expensive the larger the number of AVPs
in the message definition in question, since the record is recreated
each time an AVP value is set in it. The map decode can potentially do
better.
|
|
{record_decode, map} is a bit too quirky.
|
|
That is, decode to the same format that encode already accepts. Only a
message has its name at the head of the list since AVPs are already
name/value pairs.
|
|
With {record_decode, map}. The option name is arguably a bit misleading
now, but not too objectionable given that the encode/decode in question
has historically only been of records.
One advantage of the map decode is that the map only contains values for
those AVPs existing in the message or grouped AVP in question. The name
of the message or grouped AVP is stored in with key ':name', the leading
colon ensuring that the key isn't a diameter-name.
Decoding to maps makes the hrl files generated from dictionary files
largely irrelevant. There are value defines generated into these, but
they're typically so long as to be unusable.
|
|
To control whether or not messages and grouped AVPs are decoded to
records, in #diameter_packet.msg and #diameter_avp.value respectively.
The decode became unnecessary for diameter's needs in parent commit,
which decoupled it from the checking of AVP arities.
|
|
Instead of after, during the check that AVPs have sufficient arity.
This makes the arity checks independent of the record decode, which
will allow the latter to be made optional.
|
|
Most of the contents were moved to module diameter_gen in commit
205521d3.
|
|
|
|
* anders/diameter/20.0/OTP-14398:
vsn -> 2.0
Update appup for 20.0
|
|
* anders/diameter/capx_vs_dpr/OTP-14338:
Let candidate peers be passed to diameter:call/4
Comment on RFC ambiguity regarding application identifiers
Remove trailing whitespace
|
|
* anders/diameter/performance/OTP-14343: (50 commits)
Let spawn_opt config replace erlang:spawn_opt/2 for request processes
Move (most of) diameter_gen.hrl to diameter_gen.erl
Change signature associated with dictionary @custom_type/@codecs
Avoid sending answer terms between processes unnecessarily
Refactor handling of incoming requests
Restore diameter_codec:decode/2, update diameter_codec(3)
Add diameter_codec option ordered_encode
Restore undocumented Failed-AVP setting convenience
Fix/simplify setting of one Failed-AVP
Avoid recreating records
Avoid recreating records
Avoid recreating records
Avoid recreating records
Adapt test suites to modified encode/decode
Simplify diameter_caps construction
Don't compute URI defaults unnecessarily
Don't deconstruct {TPid, Caps} unnecessarily
Remove use of process dictionary in decode
Remove minor diameter_config bloat
Fix maximum AVP arity check
...
|
|
* anders/diameter/transport/ERL-332: (35 commits)
Capitulate on SCTP vs sparc-sun-solaris2.10
Remove obsolete traffic testcase
Fix dialyzer warnings
Remove client/server string decode from traffic suite
Add diameter_sctp option packet
Add diameter_sctp send/recv callbacks
Let diameter_tcp send/recv callbacks deal in diameter_packet
Randomly select traffic testcases
Exercise diameter_tcp message callbacks in traffic suite
Exercise diameter_{tcp,sctp} sender in traffic suite
Remove upgrade from diameter_traffic
Add diameter_tcp send/recv callbacks
Make diameter_{tcp,sctp} sender configurable
Remove upgrade from diameter_sctp; tweak diameter_tcp to match
Fix incomprehensible dialyzer warning
Simplify acks to transport processes
Strip throttling callbacks from diameter_tcp
Deal with (another) SCTP association id quirk on Solaris
Use binary:copy/2 when generating largish data in test suites
Deal with SCTP association id quirk on Solaris
...
|
|
|
|
By accepting an MFA that is applied to the fun that is otherwise spawned
for each incoming request, to allow handler processes to be reused. This
is not yet documented and may change, but the motivation is to let spawn
be replaced by process pool, from which the MFA selects. A list-valued
spawn_opt is equivalent to {erlang, spawn_opt, [Opts]}.
|
|
To solve the problem of being able to send messages to a peer that
hasn't advertised support for the application in question, as discussed
in the parent commit. diameter:call/4 can be passed 'peer' options to
identify candidates, and the only requirement is that an appropriate
dictionary be configured for encode. Filters are applied as if
candidates had been selected by advertised application.
|
|
It is tempting to regard remote support for the common application as
implicit, but that leads to the problems noted, and a node could never
then expect non-intersecting application support to result in 5010. It
probably can't anyway given the different ways the RFC's intent can be
interpreted, but it's not unreasonable that a node should be able to
advertise a single Diameter application and get 5010 if the peer doesn't
support it.
The problem we have currently is that peer selection is based on the
support advertised by the peer. The application id of an outgoing
request is used to lookup peers that have advertised support, so if the
peer hasn't advertised support for Diameter common messages then the
user won't be able to send DPR and more: diameter:call/4 will just
return {error, no_connection}. This commit doesn't solve the problem.
|
|
To remove the requirement that dictionary modules be recompiled whenever
the encode/decode implementation changes. The included diameter_gen.hrl
now only contains trivial functions that call info diameter_gen.erl.
|
|
To pass the options map through the encode. This is not backwards
compatible, and dictionaries supporting @custom_types or @codecs will
need to be updated.
|
|
As in commit fb14eac9, but for outgoing answers.
|
|
To simplify the call chains and intermediate terms, that had become a
little convoluted over time.
|
|
The documentation has been out of date since the string_decode option
was added in commit 1590920c. The optionless decode/2 was removed in the
commit that removed the use of the process dictionary in decode.
|
|
To allow list-valued messaged to be encoded in the specified order,
instead of in the dictionary order by first converting the list to a
record. This is not yet exposed in configuration.
|
|
The parent commit removed the convenience of setting something like the
following in the errors field of the diameter_packet of an answer
message.
[#diameter_avp{} = A2, {5001, #diameter_avp{} = A1}]
This results in Result-Code = 5001 and Failed-AVP = [A1,A2], but is
currently undocumented. Probably useful, so restore it.
Also accept {RC, [#diameter_avp{}]} at encode, which is probably more
useful; eg. [{5001, [A || {5001, A} <- Errors]}]
Anyone who wants full control can set errors = false and formulate
Result-Code/Failed-AVP themselves. (As opposed to not setting a value
explicitly, which results in setting from the decoded errors list. A bit
quirky, but documented and historical.)
|
|
When setting the Result-Code/Failed-AVP of an outgoing answer from an
errors list either returned from or not discarded by a handle_request
callback, more than the AVP paired with the Result-Code in question
could be set in Failed-AVP.
RFC 6733:
7.5. Failed-AVP AVP
The Failed-AVP AVP (AVP Code 279) is of type Grouped and provides
debugging information in cases where a request is rejected or not
fully processed due to erroneous information in a specific AVP. The
value of the Result-Code AVP will provide information on the reason
for the Failed-AVP AVP. A Diameter answer message SHOULD contain an
instance of the Failed-AVP AVP that corresponds to the error
indicated by the Result-Code AVP. For practical purposes, this
Failed-AVP would typically refer to the first AVP processing error
that a Diameter node encounters.
|
|
In this case the diameter_packet of an answer message for encode. The
record itself could be avoided, but that requires a new interface in
diameter_codec, probably for little gain.
|
|
In the theme of the previous two commits, creating the required
diameter_header of diameter_packet record only once.
|
|
As in the parent commit, recreating the options record is relatively
costly.
|
|
This old construction is approximately two to four times slower from
best (no elements modified) to worst (all modified) case, with the new
construction having constant speed.
|
|
Replace old macro-based implementation with something more readable.
|
|
|
|
The tuple is returned from and passed to callbacks, so retain the tuple
instead of its elements.
|
