Re: [PATCH] 9pfs: fix deep path truncation in V9fsPath

[Christian Schoenebeck <qemu_oss@crudebyte.com>]

On Thursday, 30 April 2026 14:52:41 CEST Jia Jia wrote:
[...]
> If I understood your suggestion correctly, the short-term mitigation would
> be roughly:
> 
>   - make v9fs_path_sprintf() fail on both g_vasprintf() failure and overlong
> results;
>   - propagate that error through callers such as local_name_to_path();
>   - make v9fs_fix_path() return an error as well;
>   - and if a live fid path cannot be rebuilt during rename/fixup, clunk or
>     otherwise invalidate that fid immediately instead of leaving it
> reachable with invalid path state.

Yes, that's roughly what I had in mind as short-term mitigation. Additionally 
I would suggest calling error_report_once() somewhere, as it would be an 
unusual incident that should be logged.

> So this is not just a local check in v9fs_path_sprintf(), but also error
> propagation plus fid invalidation on fixup failure.

Right, unfortunately I don't see an easier way to address this. E.g. one might 
think to just add a max. path length check when creating a new file/dir, which 
would fix the reported vector, but then it could still be triggered by moving 
a dir into a subdir, and the moved dir could have a deep directory structure, 
so one simple path length check would not be sufficient.

/Christian