Added fuzz test case and initialized the boolean that previously was uninitialized

[KBaichoo]

Commit Message: Fixes a filter related fuzz bug.
Additional Description: Initialized the boolean member to false, based on observations of usage in unit tests (https://github.com/envoyproxy/envoy/blob/8614e837cf0e4ba35eea8a45085953274b88fa03/test/extensions/filters/http/admission_control/admission_control_filter_test.cc)
Risk Level: Low
Testing: Corpus entries added as regression.
Docs Changes: N/A
Release Notes: N/A
Fixes: https://bugs.chromium.org/p/oss-fuzz/issues/detail?id=23843

Signed-off-by: Kevin Baichoo kbaichoo@google.com

[dio]

#12018 is merged. Please sync your branch with head to fix the CI (Linux-x64 asan) issue. Thanks!

[KBaichoo]

Done, thanks for the heads up.

[asraa]

Thanks Kevin!! This looks great, glad to see more fuzz bugs fixed :)

Just passing the review to @tonya11en to address this question because my read on it may be that it needs to be flipped.
if

envoy/source/extensions/filters/http/admission_control/admission_control.cc

Line 82 in af1c70e

if (Grpc::Common::isGrpcResponseHeaders(headers, end_stream)) {

isn't executed (and so the bool isn't initialized there) because of an early return/false condition, then we probably expect the grpc status to be in trailers?

[tonya11en]

Thanks Kevin!! This looks great, glad to see more fuzz bugs fixed :)

Just passing the review to @tonya11en to address this question because my read on it may be that it needs to be flipped.
if

envoy/source/extensions/filters/http/admission_control/admission_control.cc

Line 82 in af1c70e

if (Grpc::Common::isGrpcResponseHeaders(headers, end_stream)) {

isn't executed (and so the bool isn't initialized there) because of an early return/false condition, then we probably expect the grpc status to be in trailers?

You can always expect a gRPC content-type header when it's a gRPC reply, so in any scenario where that's not parsed, the boolean would not be set and we shouldn't expect to parse a gRPC status in the trailers. Setting the boolean to false here would maintain this behavior, where setting it true would cause the filter to attempt to parse a gRPC status from the trailers in all of the early return cases.

This LGTM as-is. Thanks for fixing!

[asraa]

Thanks @tonya11en for the feedback!