# Re: \[PATCH 05/10\] net: turn sk\_peer\_pid into an array indexed by pid type

- From: Christian Brauner \<brauner@kernel.org\>
- Date: 2026-09-04 09:03:29 
- List: netdev
- Page: https://ratatoskr.run/netdev/2026/09/17512298
- Thread: https://ratatoskr.run/netdev/2026/08/17483571/t.md

```text
On 2026-09-01 17:20 -0700, Jakub Kicinski wrote:
> On Mon, 31 Aug 2026 13:21:17 +0200 Christian Brauner wrote:
> > Currently only the struct pid of the thread-group leader is recorded
> > for a socket's peer. To make room for the struct pid of the thread that
> > called connect(), listen() or socketpair() turn sk_peer_pid into an
> > array indexed by pid type. All users, including bluetooth and the
> > coredump socket, keep using the PIDTYPE_TGID slot.
> > 
> > Nothing fills the PIDTYPE_PID slot yet. No functional changes.
> 
> This patch suffers from missing kdoc and general inability to build:

Oh, damn, the automated pipeline should have done a per-patch build.
Fixing, thanks!

> Warning: include/net/sock.h:606 struct member 'PIDTYPE_TGID' not described in 'sock'
> Warning: include/net/sock.h:606 Excess struct member 'sk_peer_pid' description in 'sock' (did you mean one of: '#define sk_portpair             __sk_common.skc_portpair', 'sk_peer_cred', 'sk_peer_lock', 'sk_peek_off')
> 
> 
> include/trace/events/landlock.h:946:13: error: incompatible pointer types assigning to 'struct pid *' from 'struct pid *const volatile *' [-Wincompatible-pointer-types]
> 
> 
> What's the merging path for these? Judging by the To: perhaps you
> expect us to take these directly?

If you have no quarrels with me taking it through the vfs tree it would
help because I have other work on top of this. Otherwise a shared branch
also works.

```