more information about CVE-2022-34038

[new-dream]

We are using etcd version 3.5.4 and after reviewing the related description of CVE-2022-34038 and the PR that fixes the issue, we have some questions.

1. NewPageWriter is an internal function of etcd, and pageBytes is hardcoded, value is 4096 in newEncoder. In what cases will etcd panic occur?

2. In the PR for fixing the issue, use the Verify method to intercept the scenario where pageBytes is 0. If pageBytes is 0, a panic will still occur, which seems no different from when dividing by 0 and will cause etcd to refuse service. Why fix it this way?

3. Why does this only affect versions below v3.5.5? No code fixing this issue has been found in version 3.5.5.