Fix xbcloud out of range memory read

[imrivera]

Fix a memory out of bounds read when chunk_name.size() < 21

- if (chunk_name.size() < 22 && chunk_name[chunk_name.size() - 21] != '.') {
+ if (chunk_name.size() < 22 || chunk_name[chunk_name.size() - 21] != '.') {

chunk_name.size()	Before (&&)	After (||)
< 21	- 1st condition is true - 2nd condition use negative index	- 1st condition is true - 2nd condition not evaluated
== 21	- 1st condition is true - 2nd condition using index 0, ok but useless if we want names longer than 21 chars	- 1st condition is true - 2nd condition not evaluated
>= 22	- 1st condition is false - 2nd condition is not evaluated	- 1st condition is false - 2nd condition is evaluated with an index always > 0

[it-percona-cla]

All committers have signed the CLA.

[satya-bodapati]

Thank you for the contribution @imrivera ! Fix looks good. I will verify with ASAN/vaglrind tests and merge it to the next upcoming releases.

[satya-bodapati]

Hi @imrivera, I just want to know if you reproduced this bug/behavior somehow. While I agree this is a bug from reading the code, I want to reproduce this bug. Since the default suffix name '.0000000000..' is 21 characters long, it should be impossible to get a chunk name less than this length.

Please let me know if you have a way to reproduce or how you found out this bug?

[imrivera]

Hi @imrivera, I just want to know if you reproduced this bug/behavior somehow. While I agree this is a bug from reading the code, I want to reproduce this bug. Since the default suffix name '.0000000000..' is 21 characters long, it should be impossible to get a chunk name less than this length.

Please let me know if you have a way to reproduce or how you found out this bug?

I found it just reading the code. I was trying to figure out some things about the naming scheme.

[imrivera]

@satya-bodapati any news on merging this PR?