Fixes buffer underflow

[bmribler]

H5Iget_name() documents passing NULL when querying the object name length. However, passing a non-NULL buffer with size == 0 results in undefined behavior if a write is attempted.

This PR defensively treats (name != NULL, size == 0) as a length-only query by setting name = NULL at the API level.

This fixes a security issue.

---

Important

Fixes buffer underflow in H5Iget_name() by treating (name != NULL, size == 0) as a length-only query.

• Security Fix:
  • In H5Iget_name() in H5I.c, treat (name != NULL, size == 0) as a length-only query by setting name = NULL to prevent undefined behavior.
• Behavior:
  • Prevents buffer underflow by ensuring no write attempts are made when size is 0 and name is non-NULL.

^{This description was created by for 728923c. You can customize this summary. It will automatically update as commits are pushed.}

[hyoklee]

@bmribler , is GHSA-5c6x-jmgf-f5vc valid and correct?
I can't easily locate it on Internet.

[bmribler]

@hyoklee Yes, it is one of the private security issues. There are some similar ones already in the CHANGELOG.md.

[hyoklee]

@bmribler , I don't see any in release_docs/CHANGELOG.md that has GHSA.

Do you mean release_docs/HISTORY* files?
They have corresponding CVE-YYYY-XXXXX .

[bmribler]

@bmribler , I don't see any in release_docs/CHANGELOG.md that has GHSA.

Do you mean release_docs/HISTORY* files? They have corresponding CVE-YYYY-XXXXX .

@hyoklee Yes, but they were entered by mistakes. We will not use GHSA numbers anymore.

[lrknox]

Does this one have a corresponding CVE number? If so, should that be entered instead?

[bmribler]

There is no CVE number for this issue yet. It's still not public, at least.

[mattjala]

I think this should throw an error instead of making an assumption about the user's intentions. This would match the H5Iget_name documentation better, which seems like it would consider a non-NULL zero size buffer an invalid input.

[bmribler]

I think this should throw an error instead of making an assumption about the user's intentions. This would match the H5Iget_name documentation better, which seems like it would consider a non-NULL zero size buffer an invalid input.

My understanding is the documentation says "...a preliminary call to H5Iget_name() with the last two parameters set to NULL and zero respectively can be made.", and does not indicate that they must be NULL and zero to be valid.
That aside, the documentation needs correction on the return value. The function does return FAIL in some cases but the documentation doesn't say so.