Revert "Reclaim one byte from HeaderString. (#6826)"

[htuch]

This is failing header_map_impl_fuzz due to memcpy heap buffer overflow and should be considered a
serious security issue. Given this was only introduced in the last day and does not affect any
release, a revert and bump is the consensus of Envoy security team.

This reverts commit 977907d.

Fixes oss-fuzz issue https://bugs.chromium.org/p/oss-fuzz/issues/detail?id=14639.

Risk level: Low
Testing: Corpus entry added, bazel test //test/common/http:header_map_impl_fuzz_test --test_output=all -c dbg --config=clang-asan

Signed-off-by: Harvey Tuch [email protected]

[htuch]

Tagging #6826 and #6580 for posterity.

[htuch]

I think the problematic line is

envoy/source/common/http/header_map_impl.cc

Line 210 in 977907d

buffer_.dynamic_[size] = 0;

. We're essentially clobbering memory beyond the buffer; it's not memcpy but the line after. This limits the possible exploit, since you can only clobber adjacent memory with a NUL rather than header defined content, but this still should be considered serious enough to bump forward given that header content is under remote control and we're writing to unrelated heap.