Re: libzstd :: non-regular file test failure

[Sergey Bugaev]

On Wed, Dec 4, 2024 at 11:26 AM Samuel Thibault <samuel.thibault@gnu.org> wrote:
> It's probably worth checking other _write methods in libstore/

That, and also glibc should make more efforts to be resilient against
servers returning bogus read/write amounts, whether by mistake like
here or maliciously. If we don't sanitize 'count', the

data = (void *) ((char *) data + count);

line in _IO_new_file_write doesn't look safe at all. That's an
arbitrary memory disclosure and a crash/DoS in plain sight.

Sergey