The bt-encoding, in particular, is aimed at more efficient wallet3 rpc
interactions with a daemon.
This removes the rigidity from RPC request responses, making it easier
to return useful constructs that the current epee code doesn't support
(like lists of lists), and also adds support for bt-encoded RPC requests
(intended to eventually replace the ".bin" endpoints with Monero NIH
binary encoding).
Instead endpoints now set a nlohmann::json response object with whatever
arbitrary values they want without having to add a bunch of limited epee
serialization macro hell.
So, for example, to set the value "foo" to 20, you use:
rpc.response["height"] = 20;
Binary values (for things like hashes) are handled by going through a proxy object:
rpc.response_hex["hash"] = hash; // hash is a crypto::hash
which, for json, is equivalent to:
rpc.response["hash"] = tools::type_to_hex(hash);
but when the response is to be bt-encoded it leaves it as binary for
more efficient transfers.
There is also a `response_b64` that converts json-destined values to
base64 instead of hex (and again leaves bt-destined values as binary).
Parsing of incoming requests now moves to a new
core_rpc_server_command_parser file, which does free-form parsing from a
nlohmann::json or bt_dict_consumer, and has various templated helper
functions to make this easier.
This is preliminary: this builds the basic infrastucture for handling
requests, and converts three endpoints:
- get_info
- get_height
- ons_resolve
Currently only `make daemon` builds (there is code in the wallet that
hasn't been updated yet), and the other RPC interfaces are disabled by
this commit (until they get converted to the new style).
In short: epee's http client is garbage, standard violating, and
unreliable.
This completely removes the epee http client support and replaces it
with cpr, a curl-based C++ wrapper. rpc/http_client.h wraps cpr for RPC
requests specifically, but it is also usable directly.
This replacement has a number of advantages:
- requests are considerably more reliable. The epee http client code
assumes that a connection will be kept alive forever, and returns a
failure if a connection is ever closed. This results in some very
annoying things: for example, preparing a transaction and then waiting
a long tim before confirming it will usually result in an error
communication with the daemon. This is just terribly behaviour: the
right thing to do on a connection failure is to resubmit the request.
- epee's http client is broken in lots of other ways: for example, it
tries throwing SSL at the port to see if it is HTTPS, but this is
protocol violating and just breaks (with a several second timeout) on
anything that *isn't* epee http server (for example, when lokid is
behind a proxying server).
- even when it isn't doing the above, the client breaks in other ways:
for example, there is a comment (replaced in this PR) in the Trezor PR
code that forces a connection close after every request because epee's
http client doesn't do proper keep-alive request handling.
- it seems noticeably faster to me in practical use in this PR; both
simple requests (for example, when running `lokid status`) and
wallet<->daemon connections are faster, probably because of crappy
code in epee. (I think this is also related to the throw-ssl-at-it
junk above: the epee client always generates an ssl certificate during
static initialization because it might need one at some point).
- significantly reduces the amount of code we have to maintain.
- removes all the epee ssl option code: curl can handle all of that just
fine.
- removes the epee socks proxy code; curl can handle that just fine.
(And can do more: it also supports using HTTP/HTTPS proxies).
- When a cli wallet connection fails we know show why it failed (which
now is an error message from curl), which could have all sorts of
reasons like hostname resolution failure, bad ssl certificate, etc.
Previously you just got a useless generic error that tells you
nothing.
Other related changes in this PR:
- Drops the check-for-update and download-update code. To the best of
my knowledge these have never been supported in loki-core and so it
didn't seem worth the trouble to convert them to use cpr for the
requests.
- Cleaned up node_rpc_proxy return values: there was an inconsistent mix
of ways to return errors and how the returned strings were handled.
Instead this cleans it up to return a pair<bool, val>, which (with
C++17) can be transparently captured as:
auto [success, val] = node.whatever(req);
This drops the failure message string, but it was almost always set to
something fairly useless (if we want to resurrect it we could easily
change the first element to be a custom type with a bool operator for
success, and a `.error` attribute containing some error string, but
for the most part the current code wasn't doing much useful with the
failure string).
- changed local detection (for automatic trusted daemon determination)
to just look for localhost, and to not try to resolve anything.
Trusting non-public IPs does not work well (e.g. with lokinet where
all .loki addresses resolve to a local IP).
- ssl fingerprint option is removed; this isn't supported by curl
(because it is essentially just duplicating what a custom cainfo
bundle does)
- --daemon-ssl-allow-chained is removed; it wasn't a useful option (if
you don't want chaining, don't specify a cainfo chain).
- --daemon-address is now a URL instead of just host:port. (If you omit
the protocol, http:// is prepended).
- --daemon-host and --daemon-port are now deprecated and produce a
warning (in simplewallet) if used; the replacement is to use
--daemon-address.
- --daemon-ssl is deprecated; specify --daemon-address=https://whatever
instead.
- the above three are now hidden from --help
- reordered the wallet connection options to make more logical sense.
This replaces the NIH epee http server which does not work all that well
with an external C++ library called uWebSockets. Fundamentally this
gives the following advantages:
- Much less code to maintain
- Just one thread for handling HTTP connections versus epee's pool of
threads
- Uses existing LokiMQ job server and existing thread pool for handling
the actual tasks; they are processed/scheduled in the same "rpc" or
"admin" queues as lokimq rpc calls. One notable benefit is that "admin"
rpc commands get their own queue (and thus cannot be delayed by long rpc
commands). Currently the lokimq threads and the http rpc thread pool
and the p2p thread pool and the job queue thread pool and the dns lookup
thread pool and... are *all* different thread pools; this is a step
towards consolidating them.
- Very little mutex contention (which has been a major problem with epee
RPC in the past): there is one mutex (inside uWebSockets) for putting
responses back into the thread managing the connection; everything
internally gets handled through (lock-free) lokimq inproc sockets.
- Faster RPC performance on average, and much better worst case
performance. Epee's http interface seems to have some race condition
that ocassionally stalls a request (even a very simple one) for a dozen
or more seconds for no good reason.
- Long polling gets redone here to no longer need threads; instead we
just store the request and respond when the thread pool, or else in a
timer (that runs once/second) for timing out long polls.
---
The basic idea of how this works from a high level:
We launch a single thread to handle HTTP RPC requests and response data.
This uWebSockets thread is essentially running an event loop: it never
actually handles any logic; it only serves to shuttle data that arrives
in a request to some other thread, and then, at some later point, to
send some reply back to that waiting connection. Everything is
asynchronous and non-blocking here: the basic uWebSockets event loop
just operates as things arrive, passes it off immediately, and goes back
to waiting for the next thing to arrive.
The basic flow is like this:
0. uWS thread -- listens on localhost:22023
1. uWS thread -- incoming request on localhost:22023
2. uWS thread -- fires callback, which injects the task into the LokiMQ job queue
3. LMQ main loop -- schedules it as an RPC job
4. LMQ rpc thread -- Some LokiMQ thread runs it, gets the result
5. LMQ rpc thread -- Result gets queued up for the uWS thread
6. uWS thread -- takes the request and starts sending it
(asynchronously) back to the requestor.
In more detail:
uWebSockets has registered has registered handlers for non-jsonrpc
requests (legacy JSON or binary). If the port is restricted then admin
commands get mapped to a "Access denied" response handler, otherwise
public commands (and admin commands on an unrestricted port) go to the
rpc command handler.
POST requests to /json_rpc have their own handler; this is a little
different than the above because it has to parse the request before it
can determine whether it is allowed or not, but once this is done it
continues roughly the same as legacy/binary requests.
uWebSockets then listens on the given IP/port for new incoming requests,
and starts listening for requests in a thread (we own this thread).
When a request arrives, it fires the event handler for that request.
(This may happen multiple times, if the client is sending a bunch of
data in a POST request). Once we have the full request, we then queue
the job in LokiMQ, putting it in the "rpc" or "admin" command
categories. (The one practical different here is that "admin" is
configured to be allowed to start up its own thread if all other threads
are busy, while "rpc" commands are prioritized along with everything
else.) LokiMQ then schedules this, along with native LokiMQ "rpc." or
"admin." requests.
When a LMQ worker thread becomes available, the RPC command gets called
in it and runs. Whatever output it produces (or error message, if it
throws) then gets wrapped up in jsonrpc boilerplate (if necessary), and
delivered to the uWebSockets thread to be sent in reply to that request.
uWebSockets picks up the data and sends whatever it can without
blocking, then buffers whatever it couldn't send to be sent again in a
later event loop iteration once the requestor can accept more data.
(This part is outside lokid; we only have to give uWS the data and let
it worry about delivery).
---
PR specifics:
Things removed from this PR:
1. ssl settings; with this PR the HTTP RPC interface is plain-text. The
previous default generated a self-signed certificate for the server on
startup and then the client accepted any certificate. This is actually
*worse* than unencrypted because it is entirely MITM-readable and yet
might make people think that their RPC communication is encrypted, and
setting up actual certificates is difficult enough that I think most
people don't bother.
uWebSockets *does* support HTTPS, and we could glue the existing options
into it, but I'm not convinced it's worthwhile: it works much better to
put HTTPS in a front-end proxy holding the certificate that proxies
requests to the backend (which can then listen in restricted mode on
some localhost port). One reason this is better is that it is much
easier to reload and/or restart such a front-end server, while
certificate updates with lokid require a full restart. Another reason
is that you get an error page instead of a timeout if something is wrong
with the backend. Finally we also save having to generate a temporary
certificate on *every* lokid invocation.
2. HTTP Digest authentication. Digest authentication is obsolete (and
was already obsolete when it got added to Monero). HTTP-Digest was
originally an attempt to provide a password authentication mechanism
that does not leak the password in transit, but still required that the
server know the password. It only has marginal value against replay
attacks, and is made entirely obsolete by sending traffic over HTTPS
instead. No client out there supports Digest but *not* Basic auth, and
so given the limited usefulness it seems pointless to support more than
Basic auth for HTTP RPC login.
What's worse is that epee's HTTP Digest authentication is a terrible
implementation: it uses boost::spirit -- a recursive descent parser
meant for building complex language grammars -- just to parse a single
HTTP header for Digest auth. This is a big load of crap that should
never have been accepted upstream, and that we should get rid of (even
if we wanted to support Digest auth it takes less than 100 lines of code
to do it when *not* using a recursive descent parser).
- In simplewallet, we don't report invalid/missing commands to the
CLI because we throw and catch errors that are logged silently. Now we
always log the to the user unless it's thrown by something other than
the wallet.
- Add invalid_command exception to distinguish between an error thrown
by the console_handler (i.e. missing/empty command) versus an actual
std::out_of_range thrown by wallet code (previously would have
incorrectly reported "Unknown command ... ", when it was a known command).
- Catch exceptions thrown via commands sent by RPC over the terminal and
shut down the wallet properly.
- In console_handler don't react to cmd_handler failing. Previously, if
a command failed it'd assume invalid command and log or try the next
branch which was detecting application exit.
High-level details:
This redesigns the RPC layer to make it much easier to work with,
decouples it from an embedded HTTP server, and gets the vast majority of
the RPC serialization and dispatch code out of a very commonly included
header.
There is unfortunately rather a lot of interconnected code here that
cannot be easily separated out into separate commits. The full details
of what happens here are as follows:
Major details:
- All of the RPC code is now in a `cryptonote::rpc` namespace; this
renames quite a bit to be less verbose: e.g. CORE_RPC_STATUS_OK
becomes `rpc::STATUS_OK`, and `cryptonote::COMMAND_RPC_SOME_LONG_NAME`
becomes `rpc::SOME_LONG_NAME` (or just SOME_LONG_NAME for code already
working in the `rpc` namespace).
- `core_rpc_server` is now completely decoupled from providing any
request protocol: it is now *just* the core RPC call handler.
- The HTTP RPC interface now lives in a new rpc/http_server.h; this code
handles listening for HTTP requests and dispatching them to
core_rpc_server, then sending the results back to the caller.
- There is similarly a rpc/lmq_server.h for LMQ RPC code; more details
on this (and other LMQ specifics) below.
- RPC implementing code now returns the response object and throws when
things go wrong which simplifies much of the rpc error handling. They
can throw anything; generic exceptions get logged and a generic
"internal error" message gets returned to the caller, but there is
also an `rpc_error` class to return an error code and message used by
some json-rpc commands.
- RPC implementing functions now overload `core_rpc_server::invoke`
following the pattern:
RPC_BLAH_BLAH::response core_rpc_server::invoke(RPC_BLAH_BLAH::request&& req, rpc_context context);
This overloading makes the code vastly simpler: all instantiations are
now done with a small amount of generic instantiation code in a single
.cpp rather than needing to go to hell and back with a nest of epee
macros in a core header.
- each RPC endpoint is now defined by the RPC types themselves,
including its accessible names and permissions, in
core_rpc_server_commands_defs.h:
- every RPC structure now has a static `names()` function that returns
the names by which the end point is accessible. (The first one is
the primary, the others are for deprecated aliases).
- RPC command wrappers define their permissions and type by inheriting
from special tag classes:
- rpc::RPC_COMMAND is a basic, admin-only, JSON command, available
via JSON RPC. *All* JSON commands are now available via JSON RPC,
instead of the previous mix of some being at /foo and others at
/json_rpc. (Ones that were previously at /foo are still there for
backwards compatibility; see `rpc::LEGACY` below).
- rpc::PUBLIC specifies that the command should be available via a
restricted RPC connection.
- rpc::BINARY specifies that the command is not JSON, but rather is
accessible as /name and takes and returns values in the magic epee
binary "portable storage" (lol) data format.
- rpc::LEGACY specifies that the command should be available via the
non-json-rpc interface at `/name` for backwards compatibility (in
addition to the JSON-RPC interface).
- some epee serialization got unwrapped and de-templatized so that it
can be moved into a .cpp file with just declarations in the .h. (This
makes a *huge* difference for core_rpc_server_commands_defs.h and for
every compilation unit that includes it which previously had to
compile all the serialization code and then throw all by one copy away
at link time). This required some new macros so as to not break a ton
of places that will use the old way putting everything in the headers;
The RPC code uses this as does a few other places; there are comments
in contrib/epee/include/serialization/keyvalue_serialization.h as to
how to use it.
- Detemplatized a bunch of epee/storages code. Most of it should have
have been using templates at all (because it can only ever be called
with one type!), and now it isn't. This broke some things that didn't
properly compile because of missing headers or (in one case) a messed
up circular dependency.
- Significantly simplified a bunch of over-templatized serialization
code.
- All RPC serialization definitions is now out of
core_rpc_server_commands_defs.h and into a single .cpp file
(core_rpc_server_commands_defs.cpp).
- core RPC no longer uses the disgusting
BEGIN_URI_MAP2/MAP_URI_BLAH_BLAH macros. This was a terrible design
that forced slamming tons of code into a common header that didn't
need to be there.
- epee::struct_init is gone. It was a horrible hack that instiated
multiple templates just so the coder could be so lazy and write
`some_type var;` instead of properly value initializing with
`some_type var{};`.
- Removed a bunch of useless crap from epee. In particular, forcing
extra template instantiations all over the place in order to nest
return objects inside JSON RPC values is no longer needed, as are a
bunch of stuff related to the above de-macroization of the code.
- get_all_service_nodes, get_service_nodes, and get_n_service_nodes are
now combined into a single `get_service_nodes` (with deprecated
aliases for the others), which eliminates a fair amount of
duplication. The biggest obstacle here was getting the requested
fields reference passed through: this is now done by a new ability to
stash a context in the serialization object that can be retrieved by a
sub-serialized type.
LMQ-specifics:
- The LokiMQ instance moves into `cryptonote::core` rather than being
inside cryptonote_protocol. Currently the instance is used both for
qnet and rpc calls (and so needs to be in a common place), but I also
intend future PRs to use the batching code for job processing
(replacing the current threaded job queue).
- rpc/lmq_server.h handles the actual LMQ-request-to-core-RPC glue.
Unlike http_server it isn't technically running the whole LMQ stack
from here, but the parallel name with http_server seemed appropriate.
- All RPC endpoints are supported by LMQ under the same names as defined
generically, but prefixed with `rpc.` for public commands and `admin.`
for restricted ones.
- service node keys are now always available, even when not running in
`--service-node` mode: this is because we want the x25519 key for
being able to offer CURVE encryption for lmq RPC end-points, and
because it doesn't hurt to have them available all the time. In the
RPC layer this is now called "get_service_keys" (with
"get_service_node_key" as an alias) since they aren't strictly only
for service nodes. This also means code needs to check
m_service_node, and not m_service_node_keys, to tell if it is running
as a service node. (This is also easier to notice because
m_service_node_keys got renamed to `m_service_keys`).
- Added block and mempool monitoring LMQ RPC endpoints: `sub.block` and
`sub.mempool` subscribes the connection for new block and new mempool
TX notifications. The latter can notify on just blink txes, or all
new mempool txes (but only new ones -- txes dumped from a block don't
trigger it). The client gets pushed a [`notify.block`, `height`,
`hash`] or [`notify.tx`, `txhash`, `blob`] message when something
arrives.
Minor details:
- rpc::version_t is now a {major,minor} pair. Forcing everyone to pack
and unpack a uint32_t was gross.
- Changed some macros to constexprs (e.g. CORE_RPC_ERROR_CODE_...).
(This immediately revealed a couple of bugs in the RPC code that was
assigning CORE_RPC_ERROR_CODE_... to a string, and it worked because
the macro allows implicit conversion to a char).
- De-templatizing useless templates in epee (i.e. a bunch of templated
types that were never invoked with different types) revealed a painful
circular dependency between epee and non-epee code for tor_address and
i2p_address. This crap is now handled in a suitably named
`net/epee_network_address_hack.cpp` hack because it really isn't
trivial to extricate this mess.
- Removed `epee/include/serialization/serialize_base.h`. Amazingly the
code somehow still all works perfectly with this previously vital
header removed.
- Removed bitrotted, unused epee "crypted_storage" and
"gzipped_inmemstorage" code.
- Replaced a bunch of epee::misc_utils::auto_scope_leave_caller with
LOKI_DEFERs. The epee version involves quite a bit more instantiation
and is ugly as sin. Also made the `loki::defer` class invokable for
some edge cases that need calling before destruction in particular
conditions.
- Moved the systemd code around; it makes much more sense to do the
systemd started notification as in daemon.cpp as late as possible
rather than in core (when we can still have startup failures, e.g. if
the RPC layer can't start).
- Made the systemd short status string available in the get_info RPC
(and no longer require building with systemd).
- during startup, print (only) the x25519 when not in SN mode, and
continue to print all three when in SN mode.
- DRYed out some RPC implementation code (such as set_limit)
- Made wallet_rpc stop using a raw m_wallet pointer
This commit continues the complete replacement of the spaghetti code
mess that was inside daemon/ and daemonize/ which started in #1138, and
looked like a entry level Java programmer threw up inside the code base.
This greatly simplifies it, removing a whole pile of useless abstraction
layers that don't actually abstract anything, and results in
considerably simpler code. (Many of the changes here were also carried
out in #1138; this commit updates them with the merged result which
amends some things from that PR and goes further in some places).
In detail:
- the `--detach` (and related `--pidfile`) options are gone. (--detach
is still handled, but now just prints a fatal error). Detaching a
process is an archaic unix mechanism that has no place on a modern
system. If you *really* want to do it anyway, `nohup lokid &` will do
the job. (The Windows service control code, which is probably seldom
used, is kept because it seems potentially useful for Windows users).
- Many of the `t_whatever` classes in daemon/* are just deleted (mostly
done in #1138); each one was a bunch of junk code that wraps 3-4 lines
but forces an extra layer (not even a generic abstraction, just a
useless wrapper) for no good reason and made the daemon code painfully
hard to understand and work with.
- All of the remaining `t_whatever` classes in daemon/* are either
renamed to `whatever` (because prefixing every class with `t_` is
moronic).
- Some stupid related code (e.g. epee's command handler returning an
unsuitable "usage" string that has to be string modified into what we
want) was replaced with more generic, useful code.
- Replaced boost mutexes/cvs with std ones in epee command handler, and
deleted some commented out code.
- The `--public-node` option handling was terrible: it was being handled
in main, but main doesn't know anything about options, so then it
forced it through the spaghetti objects *beside* the pack of all
options that got passed along. Moved it to a more sane location
(core_rpc_server) and parse it out with some sanity.
- Changed a bunch of std::bind's to lambdas because, at least for small
lambdas (i.e. with only one-or-two pointers for captures) they will
generally be more efficient as the values can be stored in
std::function's without any memory allocations.
This is intended to catch traffic coming from a web browser,
so we avoid issues with a web page sending a transfer RPC to
the wallet. Requiring a particular user agent can act as a
simple password scheme, while we wait for 0MQ and proper
authentication to be merged.
It uses the async console handler differently than simplewallet,
and wasn't running the same exit code, causing it to never actually
exit after breaking out of the console entry loop.
Daemon interactive mode is now working again.
RPC mapped calls in daemon and wallet have both had connection_context
removed as an argument as that argument was not being used anywhere.
many RPC functions added by the daemonize changes
(and related changes on the upstream dev branch that were not merged)
were commented out (apart from return). Other than that, this *should*
work...at any rate, it builds, and that's something.
