fix: bound DWARF .eh_frame_hdr parsing to the section

[edznux-dd]

What does this PR do?:

Fix bound checking on untrusted section size, avoid infinite loops, and fix small fuzz-harness only memory leak.

Motivation:

Make the dwarf parser more reliable.

Additional Notes:

How to test the change?:

For Datadog employees:

• If this PR touches code that signs or publishes builds or packages, or handles
  credentials of any kind, I've requested a review from @DataDog/security-design-and-guidance.
• This PR doesn't touch any of that.
• JIRA: [JIRA-XXXX]

Unsure? Have a question? Request a review!

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: c6e3fc8393

ℹ️ About Codex in GitHub

Codex has been enabled to automatically review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

When you sign up for Codex through ChatGPT, Codex can also answer questions or update the PR, like "@codex address that feedback".

[jbachorik]

Sphinx review — issues in unchanged parseFde() body

Two findings that couldn't be anchored as inline comments because they fall in the unchanged part of parseFde():

dwarf.cpp:301 — unguarded cie_offset underflow (HIGH)

cie_offset is a u32 read directly from the untrusted, memory-mapped ELF image. When cie_offset > (fde_start - _section_start), the subtraction fde_start - cie_offset underflows past the start of the mapped allocation — undefined behaviour. A crafted small value also redirects CIE parsing to an attacker-controlled position inside the section, defeating the bounds model this PR establishes.

if (cie_offset > (size_t)(fde_start - _section_start)) {
    return;
}
_ptr = fde_start - cie_offset;

dwarf.cpp:308 — unclamped _ptr advance after untrusted LEB (MEDIUM)

_ptr += getLeb() advances by a decoded value from untrusted input without clamping to _section_end. A large value pushes _ptr past the section boundary, silently truncating the rest of the FDE. Same issue in parseInstructions() at the DW_CFA_def_cfa_expression and DW_CFA_expression branches (lines ~404, ~409).

After each unclamped advance: if (_ptr > _section_end) _ptr = _section_end;

Raised by Sphinx review.

[jbachorik]

LGTM