v8.9.3 proposal #17532
v8.9.3 proposal #17532
Conversation
PR-URL: #17428 Backport-PR-URL: #17467 Reviewed-By: James M Snell <[email protected]> Reviewed-By: Colin Ihrig <[email protected]> Reviewed-By: Anatoli Papirovski <[email protected]> Reviewed-By: Ali Ijaz Sheikh <[email protected]> Reviewed-By: Tobias Nießen <[email protected]> Reviewed-By: Luigi Pinca <[email protected]> Reviewed-By: Michaël Zasso <[email protected]>
Zero-fill when `Buffer.alloc()` receives invalid fill data. A solution like #17427 which switches to throwing makes sense, but is likely a breaking change. This suggestion leaves the behaviour of `buffer.fill()` untouched, since any change to it would be a breaking change, and lets `Buffer.alloc()` check whether any filling took place or not. PR-URL: #17428 Backport-PR-URL: #17467 Refs: #17427 Refs: #17423 Reviewed-By: James M Snell <[email protected]> Reviewed-By: Colin Ihrig <[email protected]> Reviewed-By: Anatoli Papirovski <[email protected]> Reviewed-By: Ali Ijaz Sheikh <[email protected]> Reviewed-By: Tobias Nießen <[email protected]> Reviewed-By: Luigi Pinca <[email protected]> Reviewed-By: Michaël Zasso <[email protected]>
This replaces all sources of openssl-1.0.2n.tar.gz into deps/openssl/openssl PR-URL: #17526 Reviewed-By: Tobias Nießen <[email protected]> Reviewed-By: Ben Noordhuis <[email protected]> Reviewed-By: Colin Ihrig <[email protected]> Reviewed-By: Myles Borins <[email protected]> Reviewed-By: Michael Dawson <[email protected]> Reviewed-By: Daniel Bevenius <[email protected]>
All symlink files in `deps/openssl/openssl/include/openssl/` are removed and replaced with real header files to avoid issues on Windows. Two files of opensslconf.h in crypto and include dir are replaced to refer config/opensslconf.h. PR-URL: #17526 Reviewed-By: Tobias Nießen <[email protected]> Reviewed-By: Ben Noordhuis <[email protected]> Reviewed-By: Colin Ihrig <[email protected]> Reviewed-By: Myles Borins <[email protected]> Reviewed-By: Michael Dawson <[email protected]> Reviewed-By: Daniel Bevenius <[email protected]>
`x86masm.pl` was mistakenly using .486 instruction set, why `cpuid` (and perhaps others) are requiring .686 . Fixes: #589 PR-URL: #1389 Reviewed-By: Fedor Indutny <[email protected]> Reviewed-By: Ben Noordhuis <[email protected]> Reviewed-By: Shigeki Ohtsu <[email protected]>
See https://mta.openssl.org/pipermail/openssl-dev/2015-February/000651.html iojs needs to stop using masm and move to nasm or yasm on Win32. Fixes: #589 PR-URL: #1389 Reviewed-By: Fedor Indutny <[email protected]> Reviewed-By: Ben Noordhuis <[email protected]>
Reapply b910613 . Fixes: #589 PR-URL: #1389 Reviewed-By: Fedor Indutny <[email protected]> Reviewed-By: Ben Noordhuis <[email protected]>
In openssl s_client on Windows, RAND_screen() is invoked to initialize random state but it takes several seconds in each connection. This added -no_rand_screen to openssl s_client on Windows to skip RAND_screen() and gets a better performance in the unit test of test-tls-server-verify. Do not enable this except to use in the unit test. Fixes: #1461 PR-URL: #1836 Reviewed-By: Ben Noordhuis <[email protected]>
Regenerate asm files with Makefile and CC=gcc and ASM=nasm where gcc version was 5.4.0 and nasm version was 2.11.08. Also asm files in asm_obsolete dir to support old compiler and assembler are regenerated without CC and ASM envs. PR-URL: #17526 Reviewed-By: Tobias Nießen <[email protected]> Reviewed-By: Ben Noordhuis <[email protected]> Reviewed-By: Colin Ihrig <[email protected]> Reviewed-By: Myles Borins <[email protected]> Reviewed-By: Michael Dawson <[email protected]> Reviewed-By: Daniel Bevenius <[email protected]>
nodejs-github-bot
added
meta
MylesBorins
force-pushed
the
v8.9.3-proposal
branch
from
e827446
to
31c2739
Compare
|
@jasnell there is a failure on linux1 related to http2. Are there commits that we should be including here that are currently missing? |
|
Shouldn't be, but it's possible that one of the other openssl changes is causing a change in the timing on that one test. I haven't seen that one fail before and this one http2 change should not be causing it. We can do a quick fix on this to remove the use of TLS in the test case (it's not strictly necessary for this particular case, but it would be a patch over). So long as the other tests are passing, another thing we can do is mark this one as flaky and I'll investigate further after the release |
|
It's failing on most of the systems. That's pretty concerning... It's also failing on 9.2.1. |
|
It's passing on OS X just fine, even under heavy load. (And seemed to pass on the CI.) This might be mostly Linux specific. |
MylesBorins
force-pushed
the
v8.9.3-proposal
branch
from
31c2739
to
77beecd
Compare
MylesBorins
force-pushed
the
v8.9.3-proposal
branch
from
77beecd
to
56fd75b
Compare
|
I've managed to backport all the commits necessary to get the http2 tests passing again. Some of these were tagged as semver-minor, but we are going to land them as semver-patch due to http2 being experimental. There are more commits here then I would generally like to include in a security release, but it is unfortunately the only way to get the http2 tests passing with the security fix that I have found so far. The specific commit that fixes things is 7ceccb1, so if someone can work out a way to get that to land without everything else and fix the timeouts I'm all for it. Plan is to get this out the door tomorrow |
|
/cc @nodejs/release @nodejs/http2 @nodejs/tsc so y'all know what is going on |
|
@jasnell @apapirovski ^^^ |
|
Hooray for whack a mole now getting failures for parallel/test-http2-client-http1-server |
MylesBorins
force-pushed
the
v8.9.3-proposal
branch
from
56fd75b
to
06b4b6c
Compare
|
@MylesBorins I would say do what you're describing but with just d607718, and push I'm looking into why the first test is failing: |
|
@refack the rest of the commits are necessary to fix more specifically it is not until 7ceccb1 that the test is fixed... the rest of the commits are necessary to avoid conflicts |
|
How about adding this PR? #14891 |
|
@XadillaX currently it appears that the error is getting swallowed internally by nghttp2, and I'm not sure the PR you are suggesting will fix that. One thing worth mentioning, 9.x is working with more or less the same change sets against the http2 implementation. |
|
@MylesBorins we're missing this PR for http2 #17183 and parts of #17328 (there seems to be two commits from it in here but not the rest) |
|
@apapirovski the challenge here is that we have things working in 9.x but not 8.x with a nearly identical changeset. #17183 has not landed on 9.x and taking only part of #17328 is intentional... none of the commits missing from that PR exist on 9.x |
|
@MylesBorins Good to know. I'll see what else could be responsible. |
MylesBorins
force-pushed
the
v8.9.3-proposal
branch
from
9214a4e
to
725d6d0
Compare
|
I've gone ahead and added the failing test to known_issues and rebased that change into CI: https://ci.nodejs.org/job/node-test-pull-request/11974/ |
Previously, we were using a shared stack allocated buffer to hold the serialized outbound data but that runs into issues if the outgoing stream does not write or copy immediately. Instead, allocate a buffer each time. Slight additional overhead here, but necessary. Later on, once we've analyzed this more, we might be able to switch to a stack allocated ring or slab buffer but that's a bit more complicated than what we strictly need right now. PR-URL: #16669 Reviewed-By: Anatoli Papirovski <[email protected]> Reviewed-By: Anna Henningsen <[email protected]> Reviewed-By: Khaidi Chu <[email protected]> Reviewed-By: Matteo Collina <[email protected]>
`WriteWrap` instances may contain extra storage space. `self_size()` returns the size of the *entire* struct, member fields as well as storage space, so it is not an accurate measure for the storage space available. Add a method `ExtraSize()` (like the existing `Extra()` for accessing the storage memory) that yields the wanted value, and use it in the HTTP2 impl to fix a crash. PR-URL: #16727 Refs: #16669 Reviewed-By: Anatoli Papirovski <[email protected]> Reviewed-By: Colin Ihrig <[email protected]> Reviewed-By: James M Snell <[email protected]>
* eliminate pooling of Nghttp2Stream instances. After testing, the pooling is not having any tangible benefit and makes things more complicated. Simplify. Simplify. * refactor inbound headers * Enforce MAX_HEADERS_LIST setting and limit the number of header pairs accepted from the peer. Use the ENHANCE_YOUR_CALM error code when receiving either too many headers or too many octets. Use a vector to store the headers instead of a queue PR-URL: #16676 Reviewed-By: Matteo Collina <[email protected]> Reviewed-By: Anna Henningsen <[email protected]>
Do not call destroy each time rstStream is called since the first call (or receipt of rst frame) will always trigger destroy. Expand existing test for this behaviour. PR-URL: #16753 Reviewed-By: Anna Henningsen <[email protected]> Reviewed-By: James M Snell <[email protected]>
The first group of tests makes one more connection and leave the server alive for longer. Otherwise the test is just catching that the server has closed the socket, depending on timing. This does not quite make the test pass yet, however. There are some quirks with how the http2 code handles errors which actually affect 1.0.2 as well. PR-URL: #16130 Reviewed-By: Ben Noordhuis <[email protected]> Reviewed-By: Rod Vagg <[email protected]>
Fix a compiler warning that was introduced in commit 4db1bc8 ("http2: allocate on every chunk send") by adding an `override` keyword. PR-URL: #16726 Reviewed-By: Anatoli Papirovski <[email protected]> Reviewed-By: Colin Ihrig <[email protected]> Reviewed-By: Daniel Bevenius <[email protected]> Reviewed-By: Franziska Hinkelmann <[email protected]> Reviewed-By: James M Snell <[email protected]> Reviewed-By: Michael Dawson <[email protected]> Reviewed-By: Michaël Zasso <[email protected]>
This update does several significant things: 1. It eliminates the base Nghttp2* classes and folds those in to node::http2::Http2Session and node::http2::Http2Stream 2. It makes node::http2::Http2Stream a StreamBase instance and sends that out to JS-land to act as the [kHandle] for the JavaScript Http2Stream class. 3. It shifts some of the callbacks from C++ off of the JavaScript Http2Session class to the Http2Stream class. 4. It refactors the data provider structure for FD and Stream based sending to help encapsulate those functions easier 5. It streamlines some of the functions at the C++ layer to eliminate now unnecessary redirections 6. It cleans up node_http2.cc for better readability and maintainability 7. It refactors some of the debug output 8. Because Http2Stream instances are now StreamBases, they are now also trackable using async-hooks 9. The Stream::OnRead algorithm has been simplified with a couple bugs fixed. 10. I've eliminated node_http2_core.h and node_http2_core-inl.h 11. Detect invalid handshake a report protocol error to session 12. Refactor out of memory error, improve other errors 13. Add Http2Session.prototype.ping PR-URL: #17105 Reviewed-By: Anna Henningsen <[email protected]> Reviewed-By: Sebastiaan Deckers <[email protected]>
* fixup js debug messages * simplify and improve rstStream * improve and simplify _read * simplify and improve priority * simplify on ready a bit * simplify and improve respond/push * reduce duplication with _unrefActive * simplify stream close handling PR-URL: #17209 Reviewed-By: Anatoli Papirovski <[email protected]>
PR-URL: #17328 Fixes: #15303 Reviewed-By: Anatoli Papirovski <[email protected]> Reviewed-By: Sebastiaan Deckers <[email protected]>
PR-URL: #17328 Fixes: #15303 Reviewed-By: Anatoli Papirovski <[email protected]> Reviewed-By: Sebastiaan Deckers <[email protected]>
This is a security release. All Node.js users should consult the security release summary at https://nodejs.org/en/blog/vulnerability/december-2017-security-releases/ for details on patched vulnerabilities. Fixes for the following CVEs are included in this release: * CVE-2017-15896 * CVE-2017-15897 * CVE-2017-3738 (from the openssl project) Notable Changes: * buffer: * buffer allocated with an invalid content will now be zero filled (Anna Henningsen) #17428 * deps: * openssl updated to 1.0.2n (Shigeki Ohtsu) #17526 PR-URL: #17532
MylesBorins
force-pushed
the
v8.9.3-proposal
branch
from
725d6d0
to
8a44289
Compare
This is a security release. All Node.js users should consult the security release summary at https://nodejs.org/en/blog/vulnerability/december-2017-security-releases/ for details on patched vulnerabilities. Fixes for the following CVEs are included in this release: * CVE-2017-15896 * CVE-2017-15897 * CVE-2017-3738 (from the openssl project) Notable Changes: * buffer: * buffer allocated with an invalid content will now be zero filled (Anna Henningsen) #17428 * deps: * openssl updated to 1.0.2n (Shigeki Ohtsu) #17526 PR-URL: #17532
2017-12-08, Version 8.9.3 'Carbon' (LTS), @MylesBorins
This is a security release. All Node.js users should consult the security release summary at https://nodejs.org/en/blog/vulnerability/december-2017-security-releases/ for details on patched vulnerabilities.
Fixes for the following CVEs are included in this release:
Notable Changes
Notable Changes
Commits
b05ef978d3] - buffer: zero-fill buffer allocated with invalid content (Anna Henningsen) #1742818652b6860] - deps: update openssl asm and asm_obsolete files (Shigeki Ohtsu) #17526e6c308e237] - deps: add -no_rand_screen to openssl s_client (Shigeki Ohtsu) nodejs/io.js#1836a85f94bd59] - deps: fix asm build error of openssl in x86_win32 (Shigeki Ohtsu) iojs/io.js#1389b5552c854c] - deps: fix openssl assembly error on ia32 win32 (Fedor Indutny) iojs/io.js#1389afad1f23a2] - deps: copy all openssl header files to include dir (Shigeki Ohtsu) #175269fdd3bddf5] - deps: upgrade openssl sources to 1.0.2n (Shigeki Ohtsu) #17526db09f245bf] - doc: warn against filling buffer with invalid data (Anna Henningsen) #174288979e0d761] - http2: use 'close' event instead of 'streamClosed' (James M Snell) #173287ceccb1f2a] - http2: general cleanups in core.js (James M Snell) #17209929f387cc8] - http2: major update to internals (James M Snell) #171052b25ad1f9a] - http2: simplify subsequent rstStream calls (Anatoli Papirovski) #16753ba127842c0] - http2: refactor multiple internals (James M Snell) #166767e663d30c3] - http2: allocate on every chunk send (James M Snell) #16669d607718448] - http2: use correct connect event for TLS Socket (James M Snell) #173287d68488524] - openssl: fix keypress requirement in apps on win32 (Shigeki Ohtsu) iojs/io.js#138973c2bb5a88] - src: fix -Winconsistent-missing-override warning (Ben Noordhuis) #167268ffb24bcf9] - src: add method to compute storage in WriteWrap (Anna Henningsen) #16727ab87282a84] - test: fix flaky test-http2-create-client-connect (David Benjamin) #16130