Invoke threadsafe_function during the same tick and avoid marshalling
costs between threads and/or churning event loop if either:
1. There's a queued call already
2. `Push()` is called while the main thread was running
threadsafe_function
PR-URL: https://github.com/nodejs/node/pull/38506
Reviewed-By: Anna Henningsen <anna@addaleax.net>
Reviewed-By: Rich Trott <rtrott@gmail.com>
Reviewed-By: James M Snell <jasnell@gmail.com>
Prior to this patch `session` event was emitted after `secure` event on
TLSSocket, but before `secureConnect` event. This is problematic for
`https.Agent` because it must cache session only after verifying the
remote peer's certificate.
Connecting to a server that presents an invalid certificate resulted
in the session being cached after the handshake with the server and
evicted right after a certifiate validation error and socket's
destruction. A request initiated during this narrow window would pick
the faulty session, send it to the malicious server and skip the
verification of the server's certificate.
Fixes: https://hackerone.com/reports/811502
CVE-ID: CVE-2020-8172
PR-URL: https://github.com/nodejs-private/node-private/pull/200
Reviewed-By: Sam Roberts <vieuxtech@gmail.com>
Reviewed-By: Matteo Collina <matteo.collina@gmail.com>
Reviewed-By: Ben Noordhuis <info@bnoordhuis.nl>
Reviewed-By: James M Snell <jasnell@gmail.com>
Reviewed-By: Colin Ihrig <cjihrig@gmail.com>
Reviewed-By: Michael Dawson <michael_dawson@ca.ibm.com>
There is no way to disable SNI extension when sending a request to HTTPS
server. Setting `options.servername` to a falsy value would make Node.js
core override it with either hostname or ip address.
This change introduces a way to disable SNI completely if this is
required for user's application. Setting `options.servername` to ``
in `https.request` would disable overrides and thus disable the
extension.
PR-URL: https://github.com/nodejs/node/pull/27316
Reviewed-By: Steven R Loomis <srloomis@us.ibm.com>
Reviewed-By: Luigi Pinca <luigipinca@gmail.com>
Reviewed-By: Ruben Bridgewater <ruben@bridgewater.de>
Reviewed-By: Anatoli Papirovski <apapirovski@mac.com>
Reviewed-By: Ben Noordhuis <info@bnoordhuis.nl>
Reviewed-By: Rich Trott <rtrott@gmail.com>
Reviewed-By: Sam Roberts <vieuxtech@gmail.com>
Reviewed-By: Trivikram Kamat <trivikr.dev@gmail.com>
`http_parser_execute(..., nullptr, 0)` returns either `0` or `1`. The
expectation is that no error must be returned if it is `0`, and if
it is `1` - a `Error` object must be returned back to user.
The introduction of `llhttp` and the refactor that happened during it
accidentally removed the error-returning code. This commit reverts it
back to its original state.
Fix: #24585
PR-URL: https://github.com/nodejs/node/pull/24738
Reviewed-By: Luigi Pinca <luigipinca@gmail.com>
Reviewed-By: James M Snell <jasnell@gmail.com>
Introduce two overridable `Agent` methods:
* `keepSocketAlive(socket)`
* `reuseSocket(socket, req)`
These methods can be overridden by particular `Agent` class child to
make keep-alive behavior customizable.
Motivation: destroy persisted sockets after some configurable timeout.
It is very non-trivial to do it with available primitives. Such program
will most likely need to poke with undocumented events and methods of
`Agent`. With introduced API such behavior is easy to implement.
PR-URL: https://github.com/nodejs/node/pull/13005
Reviewed-By: Matteo Collina <matteo.collina@gmail.com>
Reviewed-By: Refael Ackermann <refack@gmail.com>
Reviewed-By: Colin Ihrig <cjihrig@gmail.com>
Reviewed-By: Brian White <mscdex@mscdex.net>
Reviewed-By: Michael Dawson <michael_dawson@ca.ibm.com>
Writing data to TLSWrap instance during handshake will result in it
being queued in `write_item_queue_`. This queue won't get cleared up
until the end of the handshake.
Technically, it gets cleared on `~TLSWrap` invocation, however this
won't ever happen because every `WriteWrap` holds a reference to the
`TLSWrap` through JS object, meaning that they are doomed to be alive
for eternity.
To breach this dreadful contract a knight shall embark from the
`close` function to kill the dragon of memory leak with his magic
spear of `destroySSL`.
`destroySSL` cleans up `write_item_queue_` and frees `SSL` structure,
both are good for memory usage.
PR-URL: https://github.com/nodejs/node/pull/9586
Reviewed-By: Ben Noordhuis <info@bnoordhuis.nl>
Previously `TLSSocket#_emitTLSError` was used as an `error` event
handler. However that function can emit `error` event itself, so it is
not suitable for such use. Luckily the event can be emitted only when
the control is released, so this looping-error can't happen.
Replace the error handler for clarity and correctness.
PR-URL: https://github.com/nodejs/node/pull/8889
Reviewed-By: Franziska Hinkelmann <franziska.hinkelmann@gmail.com>
Reviewed-By: Colin Ihrig <cjihrig@gmail.com>
`/bin/sh -c` trick wasn't working for several reasons:
* `/bin/sh -c "..."` expects the first argument after `"..."` to be a
`$0`, not a `$1`. Previously `-n` wasn't passed to `nm` because of
this, and many symbols were ordered improperly
* `c++filt` was applied not only to the names of the functions but to
their `nm` prefixes like `t` and `a` (`t xxx` turns into
`unsigned char xxx`).
Instead of applying `c++filt` wide and using `sh -c`, execute `nm` as
requested by `deps/v8/tools/tickprocessor.js` and apply `c++filt` to all
matching entries manually.
Included test demonstrates where previous approach failed: all builtins
were merged into `v8::internal::Builtins::~Builtins`, because they were
prefixed by `t` in `nm` output.
PR-URL: https://github.com/nodejs/node/pull/8480
Reviewed-By: Matthew Loring <mattloring@google.com>
`git-secure-tag` recursively constructs an SHA-512 digest out of the
git tree, and puts the hash from the tree's root into the tag
annotation. This hash provides better integrity guarantees than the
default SHA-1 merkle tree that git uses.
Fix: #7579
PR-URL: https://github.com/nodejs/node/pull/7603
Reviewed-By: Rod Vagg <rod@vagg.org>
Reviewed-By: James M Snell <jasnell@gmail.com>
Reviewed-By: Evan Lucas <evanlucas@me.com>
`offset` is user supplied variable and may be bigger than
`ts_obj_length`. There is no need to subtract them and pass along, so
just throw when the subtraction result would overflow.
PR-URL: https://github.com/nodejs/node/pull/7494
Reviewed-By: Ben Noordhuis <info@bnoordhuis.nl>
Use `msvs_settings.MASM.UseSafeExceptionHandlers` when building OpenSSL
assembly code on Windows. This option appends `/safeseh` to the list of
assembler flags when building `.asm` files on Windows.
Having this option in place, separate rules in `masm_compile.gypi` are
no longer needed.
Fix: #7426
PR-URL: https://github.com/nodejs/node/pull/7427
Reviewed-By: Ben Noordhuis <info@bnoordhuis.nl>
Reviewed-By: James M Snell <jasnell@gmail.com>
Reviewed-By: Bert Belder <bertbelder@gmail.com>
When `maybeReadMore` kicks in on a first bytes of incoming data, the
`req.read(0)` will be invoked and the `req._consuming` will be set to
`true`. This seemingly harmless property leads to a dire consequences:
the server won't call `req._dump()` and the whole HTTP/1.1 pipeline will
hang (single connection).
PR-URL: https://github.com/nodejs/node/pull/7211
Reviewed-By: Ben Noordhuis <info@bnoordhuis.nl>
Reviewed-By: Matteo Collina <matteo.collina@gmail.com>
When `free`ing the socket to be reused in keep-alive Agent wait for
both `prefinish` and `end` events. Otherwise the next request may be
written before the previous one has finished sending the body, leading
to a parser errors.
PR-URL: https://github.com/nodejs/node/pull/7149
Reviewed-By: Matteo Collina <matteo.collina@gmail.com>
Reviewed-By: James M Snell <jasnell@gmail.com>
There is no official way to figure out if the socket that you have on
hand is still connecting to the remote host. Introduce
`Socket#connecting`, which is essentially an unprefixed `_connecting`
property that we already had.
PR-URL: https://github.com/nodejs/node/pull/6404
Reviewed-By: Evan Lucas <evanlucas@me.com>
Reviewed-By: Sakthipriyan Vairamani <thechargingvolcano@gmail.com>
When underlying `net.Socket` instance is consumed in http server - no
`data` events are emitted, and thus `socket.setTimeout` fires the
callback even if the data is constantly flowing into the socket.
Fix this by calling `socket._unrefTimer()` on every `onParserExecute`
call.
Fix: #5899
PR-URL: https://github.com/nodejs/node/pull/6286
Reviewed-By: James M Snell <jasnell@gmail.com>
Original commit message:
tools: fix tickprocessor Cpp symbols on mac
Despite man page documentation:
-f Display the symbol table of a dynamic library flat (as one
file not separate modules).
`nm` on mac treats `-f` as a shorthand for `-format`. The `-f` argument
does not seem to be required, so just remove it completely.
(For `-format` documentation - see `nm --help` on mac).
BUG=
Review URL: https://codereview.chromium.org/1840633002
Cr-Commit-Position: refs/heads/master@{#35445}
Fix: #5903
PR-URL: https://github.com/nodejs/node/pull/6179
Reviewed-By: James M Snell <jasnell@gmail.com>
Reviewed-By: Ben Noordhuis <info@bnoordhuis.nl>
Reviewed-By: Michaël Zasso <mic.besace@gmail.com>
Introduce `--no-browser-globals` configure flag. With this flag set, following
globals won't be exported:
- `setTimeout`, `clearTimeout`, `setInterval`, `clearInterval`,
`setImmediate`, `clearImmediate`
- `console`
These are provided by the DOM implementation in browser, so the
`--no-browser-globals` flag may be helpful when embedding node.js within
chromium/webkit.
Inspired-By: https://github.com/atom/node/commit/82e10ce94f3c90234dac187f04a47d4d357ffd31
PR-URL: https://github.com/nodejs/node/pull/5853
Reviewed-By: James M Snell <jasnell@gmail.com>
Reviewed-By: Rod Vagg <rod@vagg.org>
It looks like github appends commit log right after the template
contents. Seeing this it does not look logical to ask for details first.
This commit puts `Description of change` part to the end of the
template.
PR-URL: https://github.com/nodejs/node/pull/5378
Reviewed-By: Jeremiah Senkpiel <fishrock123@rocketmail.com>
The role of `this.server` is now split between `this._server` and
`this.server`. Where the first one is used for counting active
connections of `net.Server`, and the latter one is just a public API for
users' consumption.
The reasoning for this is simple, `TLSSocket` instances wrap
`net.Socket` instances, thus both refer to the `net.Server` through the
`this.server` property. However, only one of them should be used for
`net.Server` connection count book-keeping, otherwise double-decrement
will happen on socket destruction.
Fix: #5083
PR-URL: https://github.com/nodejs/node/pull/5262
Reviewed-By: James M Snell <jasnell@gmail.com>
This is an intermediate fix for an issue of accessing `TLSWrap` fields
after the parent handle was destroyed. While `close` listener cleans up
this field automatically, it can be done even earlier at the
`TLSWrap.close` call.
Proper fix is going to be submitted and landed after this one.
Fix: #5108
PR-URL: https://github.com/nodejs/node/pull/5168
Reviewed-By: Shigeki Ohtsu <ohtsu@iij.ad.jp>
According to documentation all certificates specified in `pfx` option
should be treated as a CA certificates too. While it doesn't seem to be
logically correct to me, we can't afford to break API stability at this
point.
Fix: #5100
PR-URL: https://github.com/nodejs/node/pull/5109
Reviewed-By: James M Snell <jasnell@gmail.com>
Reviewed-By: Shigeki Ohtsu <ohtsu@iij.ad.jp>
Do not blindly take data from underlying `ArrayBuffer`, use
`ByteOffset`/`ByteLength` of `Uint8Array` itself.
Additionally, fix tests that weren't actually properly running because
of V8's internal code cache. The code should be different, otherwise the
cached data won't be used at all.
Fix: #4939
PR-URL: https://github.com/nodejs/node/pull/4947
Reviewed-By: Ben Noordhuis <info@bnoordhuis.nl>
Introduce `cachedData`/`produceCachedData` options for `v8.Script`.
Could be used to consume/produce V8's code cache for speeding up
compilation of known code.
PR-URL: https://github.com/nodejs/node/pull/4777
Reviewed-By: Ben Noordhuis <info@bnoordhuis.nl>
Call `obj.postSend` in error case of `process.send()`. The
`net.Socket`'s handle should not be leaked.
Note that there are two callbacks invoked on handles
when they are sent to the child process over IPC pipes.
These callbacks are specified by `handleConversion` object, and
during send two of them are invoked:
* `send`
* `postSend`
Now for `net.Socket` in particular, `postSend` performs clean up by
closing the actual uv handle. However this clean up will not happen in
one of the branches. This pull request aims to fix this.
PR-URL: https://github.com/nodejs/node/pull/4752
Reviewed-By: Ben Noordhuis <info@bnoordhuis.nl>
Reviewed-By: James M Snell <jasnell@gmail.com>
Make default `clientError` behavior (close socket immediately)
overridable. With this APIs it is possible to write a custom error
handler, and to send, for example, a 400 HTTP response.
http.createServer(...).on('clientError', function(err, socket) {
socket.end('HTTP/1.1 400 Bad Request\r\n\r\n');
socket.destroy();
});
Fix: #4543
PR-URL: https://github.com/nodejs/node/pull/4557
Reviewed-By: Brian White <mscdex@mscdex.net>
https requests with different SNI values should not be sent over the
same connection, even if the `host` is the same. Server may want to
present different certificate or route the incoming TLS connection
differently, depending on the received servername extension.
Fix: https://github.com/nodejs/node/issues/3940
PR-URL: https://github.com/nodejs/node/pull/4389
Reviewed-By: Ben Noordhuis <info@bnoordhuis.nl>
Do not emit `upgrade` if the server is just advertising its protocols
support as per RFC 7230 Section 6.7.
A server MAY send an Upgrade header field in any other response
to advertise that it implements support for upgrading to the
listed protocols, in order of descending preference, when
appropriate for a future request.
Fix: https://github.com/nodejs/node/issues/4334
PR-URL: https://github.com/nodejs/node/pull/4337
Reviewed-By: Ben Noordhuis <info@bnoordhuis.nl>
Original commit message:
[heap] fix crash during the scavenge of ArrayBuffer
Scavenger should not attempt to visit ArrayBuffer's storage, it is a
user-supplied pointer that may have any alignment. Visiting it, may
result in a crash.
BUG=
R=jochen
Review URL: https://codereview.chromium.org/1406133003
Cr-Commit-Position: refs/heads/master@{#31611}
PR-URL: https://github.com/nodejs/node/pull/4259
Reviewed-By: Ali Ijaz Sheikh <ofrobots@google.com>
Reviewed-By: James M Snell <jasnell@gmail.com>
`WrapperInfo` casts pointer in JS object's internal field to
`AsyncWrap`. This approach fails miserably for `TLSWrap` because it was
inhereted from the `StreamBase` first, creating different kind of
`vtable` for the whole class.
Reorder parent classes to put `AsyncWrap` first.
Fix: https://github.com/nodejs/node/issues/4250
PR-URL: https://github.com/nodejs/node/pull/4268
Reviewed-By: James M Snell <jasnell@gmail.com>
Copy client CA certs and cert store when asynchronously selecting
`SecureContext` during `SNICallback`. We already copy private key,
certificate, and certificate chain, but the client CA certs were
missing.
Fix: #2772
PR-URL: https://github.com/nodejs/node/pull/3537
Reviewed-By: Ben Noordhuis <info@bnoordhuis.nl>
Neuter external `nullptr` buffers, otherwise their contents will be
materialized on access, and the buffer instance will be internalized.
This leads to a crash like this:
v8::ArrayBuffer::Neuter Only externalized ArrayBuffers can be
neutered
Fix: #3619
PR-URL: https://github.com/nodejs/node/pull/3624
Reviewed-By: Ben Noordhuis <info@bnoordhuis.nl>
Reviewed-By: Trevor Norris <trev.norris@gmail.com>
Original commit message:
[heap] fix crash during the scavenge of ArrayBuffer
Scavenger should not attempt to visit ArrayBuffer's storage, it is a
user-supplied pointer that may have any alignment. Visiting it, may
result in a crash.
BUG=
R=jochen
Review URL: https://codereview.chromium.org/1406133003
Cr-Commit-Position: refs/heads/master@{#31611}
PR-URL: https://github.com/nodejs/node/pull/3549
Reviewed-By: Trevor Norris <trev.norris@gmail.com>
This a backport of the following commits from the v8's upstream:
* 1a8c38c50513f9af07ada479629a653e1cf36ff3
* 206f12abee3f1e7eda8fc6521d48f3c319460ee1
* 9e3676da9ab1aaf7de3e8582cb3fdefcc3dbaf33
Original commit message:
heap: make array buffer maps disjoint
Remove intersection from the `std::map`s representing current live
ArrayBuffers. While being simpler to understand, it poses
significant performance issue for the active ArrayBuffer users (like
node.js).
Store buffers separately, and process them together during
mark-sweep phase.
The results of benchmarks are:
$ ./node-slow bench && ./node-fast bench
4997.4 ns/op
4685.7 ns/op
NOTE: `fast` - was a patched node.js, `slow` - unpatched node.js
with vanilla v8.
Ref: https://github.com/nodejs/node/pull/2732
PR-URL: https://github.com/nodejs/node/pull/3351
Reviewed-By: indutny - Fedor Indutny <fedor.indutny@gmail.com>
Reviewed-By: bnoordhuis - Ben Noordhuis <info@bnoordhuis.nl>
Fedor Indutny