-
Notifications
You must be signed in to change notification settings - Fork 29.8k
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
crypto: don't reach into OpenSSL internals for ThrowCryptoError #16701
Conversation
Consistent error on windows
other failures are known flakes (fixed in master) |
I don't think that's this PR, it's failing on others too |
src/node_crypto.cc
Outdated
es->top = (((es->top - 1) % ERR_NUM_ERRORS) + ERR_NUM_ERRORS) % | ||
ERR_NUM_ERRORS; | ||
std::vector<Local<String>> errors; | ||
while (true) { |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Most minor of nits but could you use for (;;) {
here? That's what we most commonly use for indefinite loops.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Done.
There is a perfectly serviceable ERR_get_error function which avoids having to sniff through the OpenSSL ring buffer like that. It does return the errors in the opposite order, but that's easily fixed with std::reverse. Note this behavior is slightly different in that an ERR_get_error loop will ultimately clear the error queue, but this is desireable. Leaving the error queue uncleared means errors in subsequent operations may get mixed up and cause issues.
8aba98b
to
7844116
Compare
A new Coverity defect report came out, listing this:
I was going to PR a fix, but it looks like this PR will take care of it by removing that code. |
ping @nodejs/crypto & @bnoordhuis — can/should this land? |
@apapirovski Yes, it can land, unless more people want/should sign off on it. The freebsd failure seems to be an unrelated flake. |
Landed in 7db5370 |
There is a perfectly serviceable ERR_get_error function which avoids having to sniff through the OpenSSL ring buffer like that. It does return the errors in the opposite order, but that's easily fixed with std::reverse. Note this behavior is slightly different in that an ERR_get_error loop will ultimately clear the error queue, but this is desirable. Leaving the error queue uncleared means errors in subsequent operations may get mixed up and cause issues. PR-URL: #16701 Reviewed-By: Ben Noordhuis <info@bnoordhuis.nl>
There is a perfectly serviceable ERR_get_error function which avoids having to sniff through the OpenSSL ring buffer like that. It does return the errors in the opposite order, but that's easily fixed with std::reverse. Note this behavior is slightly different in that an ERR_get_error loop will ultimately clear the error queue, but this is desirable. Leaving the error queue uncleared means errors in subsequent operations may get mixed up and cause issues. PR-URL: #16701 Reviewed-By: Ben Noordhuis <info@bnoordhuis.nl>
There is a perfectly serviceable ERR_get_error function which avoids having to sniff through the OpenSSL ring buffer like that. It does return the errors in the opposite order, but that's easily fixed with std::reverse. Note this behavior is slightly different in that an ERR_get_error loop will ultimately clear the error queue, but this is desirable. Leaving the error queue uncleared means errors in subsequent operations may get mixed up and cause issues. PR-URL: #16701 Reviewed-By: Ben Noordhuis <info@bnoordhuis.nl>
There is a perfectly serviceable ERR_get_error function which avoids having to sniff through the OpenSSL ring buffer like that. It does return the errors in the opposite order, but that's easily fixed with std::reverse. Note this behavior is slightly different in that an ERR_get_error loop will ultimately clear the error queue, but this is desirable. Leaving the error queue uncleared means errors in subsequent operations may get mixed up and cause issues. PR-URL: #16701 Reviewed-By: Ben Noordhuis <info@bnoordhuis.nl>
There is a perfectly serviceable ERR_get_error function which avoids having to sniff through the OpenSSL ring buffer like that. It does return the errors in the opposite order, but that's easily fixed with std::reverse. Note this behavior is slightly different in that an ERR_get_error loop will ultimately clear the error queue, but this is desirable. Leaving the error queue uncleared means errors in subsequent operations may get mixed up and cause issues. PR-URL: #16701 Reviewed-By: Ben Noordhuis <info@bnoordhuis.nl>
I backported this to v8.x-staging, and I'm getting this error on Ubuntu 16.04: ../src/node_crypto.cc: In function ‘void node::crypto::PBKDF2(const v8::FunctionCallbackInfo<v8::Value>&)’:
../src/node_crypto.cc:5416:43: error: ‘isnan’ was not declared in this scope
if (raw_keylen < 0.0 || isnan(raw_keylen) || isinf(raw_keylen) ||
^
../src/node_crypto.cc:5416:43: note: suggested alternative:
In file included from /usr/include/c++/5/random:38:0,
from /usr/include/c++/5/bits/stl_algo.h:66,
from /usr/include/c++/5/algorithm:62,
from ../src/node_crypto.cc:46:
/usr/include/c++/5/cmath:641:5: note: ‘std::isnan’
isnan(_Tp __x)
^
../src/node_crypto.cc:5416:64: error: ‘isinf’ was not declared in this scope
if (raw_keylen < 0.0 || isnan(raw_keylen) || isinf(raw_keylen) ||
^
../src/node_crypto.cc:5416:64: note: suggested alternative:
In file included from /usr/include/c++/5/random:38:0,
from /usr/include/c++/5/bits/stl_algo.h:66,
from /usr/include/c++/5/algorithm:62,
from ../src/node_crypto.cc:46:
/usr/include/c++/5/cmath:621:5: note: ‘std::isinf’
isinf(_Tp __x)
^
g++ '-DNODE_ARCH="x64"' '-DNODE_PLATFORM="linux"' '-DNODE_WANT_INTERNALS=1' '-DV8_DEPRECATION_WARNINGS=1' '-DHAVE_INSPECTOR=1' '-D__POSIX__' '-DHAVE_OPENSSL=1' '-DNODE_USE_V8_PLATFORM=1' '-DNODE_HAVE_I18N_SUPPORT=1' '-DNODE_HAVE_SMALL_ICU=1' '-DUCONFIG_NO_SERVICE=1' '-DUCONFIG_NO_REGULAR_EXPRESSIONS=1' '-DU_ENABLE_DYLOAD=0' '-DU_STATIC_IMPLEMENTATION=1' '-DU_HAVE_STD_STRING=0' '-DUCONFIG_NO_BREAK_ITERATION=0' '-DHTTP_PARSER_STRICT=0' '-D_LARGEFILE_SOURCE' '-D_FILE_OFFSET_BITS=64' '-D_POSIX_C_SOURCE=200112' '-DNGHTTP2_STATICLIB' -I../src -I../tools/msvs/genfiles -I/home/gib/node/out/Release/obj/gen -I/home/gib/node/out/Release/obj/gen/include -I../deps/openssl/openssl/include -I../deps/v8/include -I../deps/icu-small/source/i18n -I../deps/icu-small/source/common -I../deps/zlib -I../deps/http_parser -I../deps/cares/include -I../deps/uv/include -I../deps/nghttp2/lib/includes -pthread -Wall -Wextra -Wno-unused-parameter -m64 -O3 -fno-omit-frame-pointer -fno-rtti -fno-exceptions -std=gnu++0x -MMD -MF /home/gib/node/out/Release/.deps//home/gib/node/out/Release/obj.target/node/src/node_crypto_clienthello.o.d.raw -c -o /home/gib/node/out/Release/obj.target/node/src/node_crypto_clienthello.o ../src/node_crypto_clienthello.cc
g++ '-DNODE_ARCH="x64"' '-DNODE_PLATFORM="linux"' '-DNODE_WANT_INTERNALS=1' '-DV8_DEPRECATION_WARNINGS=1' '-DHAVE_INSPECTOR=1' '-D__POSIX__' '-DHAVE_OPENSSL=1' '-DNODE_USE_V8_PLATFORM=1' '-DNODE_HAVE_I18N_SUPPORT=1' '-DNODE_HAVE_SMALL_ICU=1' '-DUCONFIG_NO_SERVICE=1' '-DUCONFIG_NO_REGULAR_EXPRESSIONS=1' '-DU_ENABLE_DYLOAD=0' '-DU_STATIC_IMPLEMENTATION=1' '-DU_HAVE_STD_STRING=0' '-DUCONFIG_NO_BREAK_ITERATION=0' '-DHTTP_PARSER_STRICT=0' '-D_LARGEFILE_SOURCE' '-D_FILE_OFFSET_BITS=64' '-D_POSIX_C_SOURCE=200112' '-DNGHTTP2_STATICLIB' -I../src -I../tools/msvs/genfiles -I/home/gib/node/out/Release/obj/gen -I/home/gib/node/out/Release/obj/gen/include -I../deps/openssl/openssl/include -I../deps/v8/include -I../deps/icu-small/source/i18n -I../deps/icu-small/source/common -I../deps/zlib -I../deps/http_parser -I../deps/cares/include -I../deps/uv/include -I../deps/nghttp2/lib/includes -pthread -Wall -Wextra -Wno-unused-parameter -m64 -O3 -fno-omit-frame-pointer -fno-rtti -fno-exceptions -std=gnu++0x -MMD -MF /home/gib/node/out/Release/.deps//home/gib/node/out/Release/obj.target/node/src/tls_wrap.o.d.raw -c -o /home/gib/node/out/Release/obj.target/node/src/tls_wrap.o ../src/tls_wrap.cc
node.target.mk:199: recipe for target '/home/gib/node/out/Release/obj.target/node/src/node_crypto.o' failed
make[1]: *** [/home/gib/node/out/Release/obj.target/node/src/node_crypto.o] Error 1 See the CI run for more info https://ci.nodejs.org/job/node-test-commit/14987/ @yhwang took a look, and said:
I've removed it from staging for now. Assuming this should be backported, could someone raise a PR to backport to |
There is a perfectly serviceable ERR_get_error function which avoids having to sniff through the OpenSSL ring buffer like that. It does return the errors in the opposite order, but that's easily fixed with std::reverse. Note this behavior is slightly different in that an ERR_get_error loop will ultimately clear the error queue, but this is desirable. Leaving the error queue uncleared means errors in subsequent operations may get mixed up and cause issues. PR-URL: nodejs#16701 Reviewed-By: Ben Noordhuis <info@bnoordhuis.nl>
There is a perfectly serviceable ERR_get_error function which avoids having to sniff through the OpenSSL ring buffer like that. It does return the errors in the opposite order, but that's easily fixed with std::reverse. Note this behavior is slightly different in that an ERR_get_error loop will ultimately clear the error queue, but this is desirable. Leaving the error queue uncleared means errors in subsequent operations may get mixed up and cause issues. PR-URL: #16701 Backport-PR-URL: #18327 Reviewed-By: Ben Noordhuis <info@bnoordhuis.nl>
There is a perfectly serviceable
ERR_get_error
function which avoidshaving to sniff through the OpenSSL ring buffer like that. It does
return the errors in the opposite order, but that's easily fixed with
std::reverse
.Note this behavior is slightly different in that an
ERR_get_error
loopwill ultimately clear the error queue, but this is desireable. Leaving
the error queue uncleared means errors in subsequent operations may get
mixed up and cause issues.
PS: In future, I'm happy to look over changes to crypto stuff for you all, especially when they reach into internals like that. That's rarely actually necessary. :-)
Checklist
make -j4 test
(UNIX), orvcbuild test
(Windows) passesAffected core subsystem(s)
crypto