The original function was already naturally split into two chunks:
First, we parse and validate the users' RPC configuration for IPs and
ports. Next we bind libevent's http server to the appropriate
endpoints.
This commit splits these chunks into two separate functions, leaving
the argument parsing in the common space of the module and moving the
libevent-specific binding into the http_libevent namespace.
A future commit will implement http_bitcoin::HTTPBindAddresses to
bind the validate list of endpoints by the new HTTP server.
The original function is passed to libevent as a callback when HTTP
requests are received and processed. It wrapped the libevent request
object in a http_libevent::HTTPRequest and then handed that off to
bitcoin for basic checks and finally dispatch to worker threads.
In this commit we split the function after the
http_libevent::HTTPRequest is created, and pass that object to a new
function that maintains the logic of checking and dispatching.
This will be the merge point for http_libevent and http_bitcoin,
where HTTPRequest objects from either namespace have the same
downstream lifecycle.
This is a refactor to prepare for matching the API of HTTPRequest
definitions in both namespaces http_bitcoin and http_libevent. In
particular, to provide a consistent return type for GetRequestMethod()
in both classes.
See https://www.rfc-editor.org/rfc/rfc7230#section-6.3.2
> A server MAY process a sequence of pipelined requests in
parallel if they all have safe methods (Section 4.2.1 of [RFC7231]),
but it MUST send the corresponding responses in the same order that
the requests were received.
We choose NOT to process requests in parallel. They are executed in
the order recevied as well as responded to in the order received.
This prevents race conditions where old state may get sent in response
to requests that are very quick to process but were requested later on
in the queue.
This commit is a no-op to isolate HTTP methods and objects that
depend on libevent. Following commits will add replacement objects
and methods in a new namespace for testing and review before
switching over the server.
Treat specifying -norpcbind and -norpcallowip the same as not specifying
-rpcbind or -rpcallowip, instead of failing to bind to localhost and failing to
show warnings.
Also add code comment to clarify what intent of existing code is.
This makes code more consistent and makes it easier to add compile-time checking to
enforce that format strings contain the right specifiers, because it stops
using Untranslated() to create the format string, so the Untranslated()
function will not need to get involved in formatting.
-BEGIN VERIFY SCRIPT-
quote='"[^"]+"'
quotes="(?:$quote|\\s)*"
nonparens="[^()]*"
single_level_paren="\($nonparens\)"
double_level_paren="\($nonparens\($nonparens\)$nonparens\)"
exprs="(?:$double_level_paren|$single_level_paren|$nonparens)*"
git grep -l 'Untranslated' | xargs perl -0777 -i -pe "s/strprintf\((\\W*)Untranslated\(($quotes)\)($exprs)(\))/Untranslated(\1strprintf(\2\3))/gs"
-END VERIFY SCRIPT-
Adds invalid rpcbind port checking to
`HTTPBindAddresses()`. While movement of
`CheckHostPortOptions()` in the previous
commit handles rcpbind port errors, updating
`HTTPBindAddresses()` port checking adds
a defensive measure for potential future
changes.
03d49d0f25 http: set TCP_NODELAY when creating HTTP server (Roman Zeyde)
Pull request description:
Otherwise, the default HTTP server config may result in high latency, due to Nagle's algorithm (on the server) and delayed ACK (on the client):
[1] https://www.extrahop.com/blog/tcp-nodelay-nagle-quickack-best-practices
[2] https://eklitzke.org/the-caveats-of-tcp-nodelay
Without the fix, fetching a small block takes ~40ms (when connection keep-alive is enabled):
```
$ ab -k -c 1 -n 100 http://localhost:8332/rest/block/00000000000002b5898f7cdc80d9c84e9747bc6b9388cc989971d443f05713ee.bin
Server Software:
Server Hostname: localhost
Server Port: 8332
Document Path: /rest/block/00000000000002b5898f7cdc80d9c84e9747bc6b9388cc989971d443f05713ee.bin
Document Length: 25086 bytes
Concurrency Level: 1
Time taken for tests: 4.075 seconds
Complete requests: 100
Failed requests: 0
Keep-Alive requests: 100
Total transferred: 2519200 bytes
HTML transferred: 2508600 bytes
Requests per second: 24.54 [#/sec] (mean)
Time per request: 40.747 [ms] (mean)
Time per request: 40.747 [ms] (mean, across all concurrent requests)
Transfer rate: 603.76 [Kbytes/sec] received
Connection Times (ms)
min mean[+/-sd] median max
Connect: 0 0 0.0 0 0
Processing: 0 41 4.1 41 42
Waiting: 0 0 0.1 0 1
Total: 0 41 4.1 41 42
Percentage of the requests served within a certain time (ms)
50% 41
66% 41
75% 41
80% 41
90% 42
95% 42
98% 42
99% 42
100% 42 (longest request)
```
With the fix, it takes ~0.2ms:
```
$ ab -k -c 1 -n 1000 http://localhost:8332/rest/block/00000000000002b5898f7cdc80d9c84e9747bc6b9388cc989971d443f05713ee.bin
Benchmarking localhost (be patient)
Completed 100 requests
Completed 200 requests
Completed 300 requests
Completed 400 requests
Completed 500 requests
Completed 600 requests
Completed 700 requests
Completed 800 requests
Completed 900 requests
Completed 1000 requests
Finished 1000 requests
Server Software:
Server Hostname: localhost
Server Port: 8332
Document Path: /rest/block/00000000000002b5898f7cdc80d9c84e9747bc6b9388cc989971d443f05713ee.bin
Document Length: 25086 bytes
Concurrency Level: 1
Time taken for tests: 0.194 seconds
Complete requests: 1000
Failed requests: 0
Keep-Alive requests: 1000
Total transferred: 25192000 bytes
HTML transferred: 25086000 bytes
Requests per second: 5147.05 [#/sec] (mean)
Time per request: 0.194 [ms] (mean)
Time per request: 0.194 [ms] (mean, across all concurrent requests)
Transfer rate: 126625.50 [Kbytes/sec] received
Connection Times (ms)
min mean[+/-sd] median max
Connect: 0 0 0.0 0 0
Processing: 0 0 0.0 0 0
Waiting: 0 0 0.0 0 0
Total: 0 0 0.0 0 0
Percentage of the requests served within a certain time (ms)
50% 0
66% 0
75% 0
80% 0
90% 0
95% 0
98% 0
99% 0
100% 0 (longest request)
```
ACKs for top commit:
achow101:
ACK 03d49d0f25
theStack:
re-ACK 03d49d0f25
tdb3:
ACK 03d49d0f25
Tree-SHA512: bbf3d78b8521f569430850ec4315a75711303547df1a3de213a4ad34c9700105e374e0a649352fd05f8e4badb5b59debd3720e1c5d392c5113d7816648f7fcaa
e60fc7d5d3 logging: Replace uses of LogPrintfCategory (Anthony Towns)
f7ce5ac08c logging: add LogError, LogWarning, LogInfo, LogDebug, LogTrace (Anthony Towns)
fbd7642c8e logging: add -loglevelalways=1 option (Anthony Towns)
782bb6a056 logging: treat BCLog::ALL like BCLog::NONE (Anthony Towns)
667ce3e329 logging: Drop BCLog::Level::None (Anthony Towns)
ab34dc6012 logging: Log Info messages unconditionally (Anthony Towns)
dfe98b6874 logging: make [cat:debug] and [info] implicit (Anthony Towns)
c5c76dc615 logging: refactor: pull prefix code out (Anthony Towns)
Pull request description:
Replace `LogPrint*` functions with severity based logging functions:
* `LogInfo(...)`, `LogWarning(...)`, `LogError(...)` for unconditional (uncategorised) logging (replaces `LogPrintf`)
* `LogDebug(CATEGORY, ...)` and `LogTrace(CATEGORY, ...)` for conditional logging (replaces `LogPrint`)
* `LogPrintLevel(CATEGORY, LEVEL, ...)` for when the level isn't known in advance, or a category needs to be added for an info/warning/error log message (mostly unchanged, but rarely needed)
Logs look roughly as they do now with `LogInfo` not having an `[info]` prefix, and `LogDebug` having a `[cat]` prefix, rather than a `[cat:debug]` prefix. This removes `BCLog::Level::None` entirely -- for `LogFlags::NONE` just use `Level::Info`, for any actual category, use `Level::Debug`.
Adds docs to developer-notes about when to use which level.
Adds `-loglevelalways=1` option so that you get `[net:debug]`, `[all:info]`, `[all:warning]` etc, which might be helpful for automated parsing, or just if you like everything to be consistent. Defaults to off to reduce noise in the default config, and to avoid unnecessary changes on upgrades.
Changes the behaviour of `LogPrintLevel(CATEGORY, BCLog::Level::Info, ...)` to be logged unconditionally, rather than only being an additional optional logging level in addition to trace and debug. Does not change the behaviour of `LogPrintLevel(NONE, Debug, ...)` and `LogPrintLevel(NONE, Trace, ...)` being no-ops.
ACKs for top commit:
maflcko:
re-ACK e60fc7d5d3🌚
achow101:
ACK e60fc7d5d3
stickies-v:
ACK e60fc7d5d3
jamesob:
ACK e60fc7d5d3 ([`jamesob/ackr/28318.1.ajtowns.logging_simplify_api_for`](https://github.com/jamesob/bitcoin/tree/ackr/28318.1.ajtowns.logging_simplify_api_for))
Tree-SHA512: e7a4588779b148242495b7b6f64198a00c314cd57100affab11c43e9d39c9bbf85118ee2002792087fdcffdea08c84576e20844b3079f27083e26ddd7ca15d7f
Pass HTTP server an interrupt object instead of having it depend on shutdown.h
and global shutdown state.
There is no change in behavior in this commit.
fb3e812277 p2p: return `CSubNet` in `LookupSubNet` (brunoerg)
Pull request description:
Analyzing the usage of `LookupSubNet`, noticed that most cases uses check if the subnet is valid by calling `subnet.IsValid()`, and the boolean returned by `LookupSubNet` hasn't been used so much, see:
29d540b7ad/src/httpserver.cpp (L172-L174)29d540b7ad/src/net_permissions.cpp (L114-L116)
It makes sense to return `CSubNet` instead of `bool`.
ACKs for top commit:
achow101:
ACK fb3e812277
vasild:
ACK fb3e812277
theStack:
Code-review ACK fb3e812277
stickies-v:
Concept ACK, but Approach ~0 (for now). Reviewed the code (fb3e812277) and it all looks good to me.
Tree-SHA512: ba50d6bd5d58dfdbe1ce1faebd80dd8cf8c92ac53ef33519860b83399afffab482d5658cb6921b849d7a3df6d5cea911412850e08f3f4e27f7af510fbde4b254
0e6f6ebc06 net: remove unused CConnman::FindNode(const CSubNet&) (Vasil Dimov)
9482cb780f netbase: possibly change the result of LookupSubNet() to CJDNS (Vasil Dimov)
53afa68026 net: move MaybeFlipIPv6toCJDNS() from net to netbase (Vasil Dimov)
6e308651c4 net: move IsReachable() code to netbase and encapsulate it (Vasil Dimov)
c42ded3d9b fuzz: ConsumeNetAddr(): avoid IPv6 addresses that look like CJDNS (Vasil Dimov)
64d6f77907 net: put CJDNS prefix byte in a constant (Vasil Dimov)
Pull request description:
`LookupSubNet()` would treat addresses that start with `fc` as IPv6 even if `-cjdnsreachable` is set. This creates the following problems where it is called:
* `NetWhitelistPermissions::TryParse()`: otherwise `-whitelist=` fails to white list CJDNS addresses: when a CJDNS peer connects to us, it will be matched against IPv6 `fc...` subnet and the match will never succeed.
* `BanMapFromJson()`: CJDNS bans are stored as just IPv6 addresses in `banlist.json`. Upon reading from disk they have to be converted back to CJDNS, otherwise, after restart, a ban entry like (`fc00::1`, IPv6) would not match a peer (`fc00::1`, CJDNS).
* `RPCConsole::unbanSelectedNode()`: in the GUI the ban entries go through `CSubNet::ToString()` and back via `LookupSubNet()`. Then it must match whatever is stored in `BanMan`, otherwise it is impossible to unban via the GUI.
These were uncovered by https://github.com/bitcoin/bitcoin/pull/26859.
Thus, flip the result of `LookupSubNet()` to CJDNS if the network base address starts with `fc` and `-cjdnsreachable` is set. Since subnetting/masking does not make sense for CJDNS (the address is "random" bytes, like Tor and I2P, there is no hierarchy) treat `fc.../mask` as an invalid `CSubNet`.
To achieve that, `MaybeFlipIPv6toCJDNS()` has to be moved from `net` to `netbase` and thus also `IsReachable()`. In the process of moving `IsReachable()`, `SetReachable()` and `vfLimited[]` encapsulate those in a class.
ACKs for top commit:
jonatack:
Code review ACK 0e6f6ebc06
achow101:
ACK 0e6f6ebc06
mzumsande:
re-ACK 0e6f6ebc06
Tree-SHA512: 4767a60dc882916de4c8b110ce8de208ff3f58daaa0b560e6547d72e604d07c4157e72cf98b237228310fc05c0a3922f446674492e2ba02e990a272d288bd566
All callers of `LookupSubNet()` need the result to be of CJDNS type if
`-cjdnsreachable` is set and the address begins with `fc`:
* `NetWhitelistPermissions::TryParse()`: otherwise `-whitelist=` fails
to white list CJDNS addresses: when a CJDNS peer connects to us, it
will be matched against IPv6 `fc...` subnet and the match will never
succeed.
* `BanMapFromJson()`: CJDNS bans are stored as just IPv6 addresses in
`banlist.json`. Upon reading from disk they have to be converted back
to CJDNS, otherwise, after restart, a ban entry like (`fc00::1`, IPv6)
would not match a peer (`fc00::1`, CJDNS).
* `setban()` (in `rpc/net.cpp`): otherwise `setban fc.../mask add` would
add an IPv6 entry to BanMan. Subnetting does not make sense for CJDNS
addresses, thus treat `fc.../mask` as invalid `CSubNet`. The result of
`LookupHost()` has to be converted for the case of banning a single
host.
* `InitHTTPAllowList()`: not necessary since before this change
`-rpcallowip=fc...` would match IPv6 subnets against IPv6 peers even
if they started with `fc`. But because it is necessary for the above,
`HTTPRequest::GetPeer()` also has to be adjusted to return CJDNS peer,
so that now CJDNS peers are matched against CJDNS subnets.
It is possible that the client disconnects before the request is
handled. In those cases, evhttp_request_set_on_complete_cb is never
called, which means that on shutdown the server we'll keep waiting
endlessly.
By adding evhttp_connection_set_closecb, libevent automatically
cleans up those dead connections at latest when we shutdown, and
depending on the libevent version already at the moment of remote
client disconnect. In both cases, the bug is fixed.
Introduces and uses a HTTPRequestTracker class to keep track of
how many HTTP requests are currently active, so we don't stop the
server before they're all handled.
This has two purposes:
1. In a next commit, allows us to untrack all requests associated
with a connection without running into lifetime issues of the
connection living longer than the request
(see https://github.com/bitcoin/bitcoin/pull/27909#discussion_r1265614783)
2. Improve encapsulation by making the mutex and cv internal members,
and exposing just the WaitUntilEmpty() method that can be safely
used.
79d343a642 http: update libevent workaround to correct version (stickies-v)
Pull request description:
The libevent bug described in 5ff8eb2637 was already patched in [release-2.1.9-beta](https://github.com/libevent/libevent/releases/tag/release-2.1.9-beta), with cherry-picked commits [5b40744d1581447f5b4496ee8d4807383e468e7a](5b40744d15) and [b25813800f97179b2355a7b4b3557e6a7f568df2](b25813800f).
There should be no side-effects by re-applying the workaround on an already patched version of libevent (as is currently done in master for people running libevent between 2.1.9 and 2.1.12), but it is best to just set the correct version number to avoid confusion.
This will prevent situations like e.g. in https://github.com/bitcoin/bitcoin/pull/27909#discussion_r1238858604, where a reverse workaround was incorrectly applied to the wrong version range.
ACKs for top commit:
fanquake:
ACK 79d343a642
Tree-SHA512: 56d2576411cf38e56d0976523fec951e032a48e35af293ed1ef3af820af940b26f779b9197baaed6d8b79bd1c7f7334646b9d73f80610d63cffbc955958ca8a0
The libevent bug described in 5ff8eb2637
was already patched in release-2.1.9-beta, with cherry-picked
commits 5b40744d1581447f5b4496ee8d4807383e468e7a and
b25813800f97179b2355a7b4b3557e6a7f568df2.
There should be no side-effects by re-applying the workaround on
an already patched version of libevent, but it is best to set the
correct version number to avoid confusion.
After initially being merged in #20487, it's no-longer clear that an
internal syscall sandboxing mechanism is something that Bitcoin Core
should have/maintain, especially when compared to better
maintained/supported alterantives, i.e firejail.
Note that given where it's used, the sandbox also gets dragged into the
kernel.
There is some related discussion in #24771.
This should not require any sort of deprecation, as this was only ever
an opt-in, experimental feature.
Closes#24771.
be55f545d5 move-only: Extract common/args and common/config.cpp from util/system (TheCharlatan)
Pull request description:
This pull request is part of the `libbitcoinkernel` project https://github.com/bitcoin/bitcoin/issues/24303https://github.com/bitcoin/bitcoin/projects/18 and more specifically its "Step 2: Decouple most non-consensus code from libbitcoinkernel". It is part of a series of patches splitting up the `util/system` files. Its preceding pull request is https://github.com/bitcoin/bitcoin/pull/27254.
The pull request contains an extraction of ArgsManager related functions from util/system into their own common/ file.
The background of this commit is an ongoing effort to decouple the libbitcoinkernel library from the ArgsManager. The ArgsManager belongs into the common library, since the kernel library should not depend on it. See [doc/design/libraries.md](https://github.com/bitcoin/bitcoin/blob/master/doc/design/libraries.md) for more information on this rationale.
ACKs for top commit:
MarcoFalke:
re-ACK be55f545d5🚲
ryanofsky:
Code review ACK be55f545d5. Just small cleanups since the last review.
hebasto:
ACK be55f545d5, I have reviewed the code and it looks OK, I agree it can be merged.
Tree-SHA512: 90eb03334af0155b823030b4f2ecf286d35058d700ee2ddbbaa445be19e31eb0fe982656f35bd14ecee3ad2c3d0db3746855cb8f3777eff7253713e42873e111
This is an extraction of ArgsManager related functions from util/system
into their own common file.
Config file related functions are moved to common/config.cpp.
The background of this commit is an ongoing effort to decouple the
libbitcoinkernel library from the ArgsManager. The ArgsManager belongs
into the common library, since the kernel library should not depend on
it. See doc/design/libraries.md for more information on this rationale.
`evhttp_uri_parse` can return a nullptr, for example when the URI
contains invalid characters (e.g. "%").
`GetQueryParameterFromUri` passes the output of `evhttp_uri_parse`
straight into `evhttp_uri_get_query`, which means that anyone calling
a REST endpoint in which query parameters are used (e.g. `rest_headers`)
can cause a segfault.
This bugfix is designed to be minimal and without additional behaviour change.
Follow-up work should be done to resolve this in a more general and robust way,
so not every endpoint has to handle it individually.
60978c8080 test: Reduce extended timeout on abortnode test (Fabian Jahr)
660bdbf785 http: Release server before waiting for event base loop exit (João Barbosa)
8c6d007c80 http: Track active requests and wait for last to finish (João Barbosa)
Pull request description:
This revives #19420. Since promag is not so active at the moment, I can support this to finally get it merged.
The PR is rebased and comments by jonatack have been addressed.
Once this is merged, I will also reopen#19434.
ACKs for top commit:
achow101:
ACK 60978c8080
stickies-v:
re-ACK [60978c8](60978c8080)
hebasto:
ACK 60978c8080
Tree-SHA512: eef0fe1081e9331b95cfafc71d82f2398abd1d3439dac5b2fa5c6d9c0a3f63ef19adde1c38c88d3b4e7fb41ce7c097943f1815c10e33d165918ccbdec512fe1c
c9d548c91f net: remove CService::ToStringPort() (Vasil Dimov)
fd4f0f41e9 gui: simplify OptionsDialog::updateDefaultProxyNets() (Vasil Dimov)
96c791dd20 net: remove CService::ToString() use ToStringAddrPort() instead (Vasil Dimov)
944a9de08a net: remove CNetAddr::ToString() and use ToStringAddr() instead (Vasil Dimov)
043b9de59a scripted-diff: rename ToStringIP[Port]() to ToStringAddr[Port]() (Vasil Dimov)
Pull request description:
Before this PR we had the somewhat confusing combination of methods:
`CNetAddr::ToStringIP()`
`CNetAddr::ToString()` (duplicate of the above)
`CService::ToStringIPPort()`
`CService::ToString()` (duplicate of the above, overrides a non-virtual method from `CNetAddr`)
`CService::ToStringPort()`
Avoid [overriding non-virtual methods](https://github.com/bitcoin/bitcoin/pull/25349/#issuecomment-1185226396).
"IP" stands for "Internet Protocol" and while sometimes "IP addresses" are called just "IPs", it is incorrect to call Tor or I2P addresses "IPs". Thus use "Addr" instead of "IP".
Change the above to:
`CNetAddr::ToStringAddr()`
`CService::ToStringAddrPort()`
The changes touch a lot of files, but are mostly mechanical.
ACKs for top commit:
sipa:
utACK c9d548c91f
achow101:
ACK c9d548c91f
jonatack:
re-ACK c9d548c91f only change since my previous reviews is rebase, but as a sanity check rebased to current master and at each commit quickly re-reviewed and re-verified clean build and green unit tests
LarryRuane:
ACK c9d548c91f
Tree-SHA512: 633fb044bdecf9f551b5e3314c385bf10e2b78e8027dc51ec324b66b018da35e5b01f3fbe6295bbc455ea1bcd1a3629de1918d28de510693afaf6a52693f2157
