Re: [PATCH v4 bpf-next 00/11] Socket migration for SO_REUSEPORT.

From: Eric Dumazet
Date: Wed Apr 28 2021 - 12:33:47 EST


On Wed, Apr 28, 2021 at 5:52 PM Kuniyuki Iwashima <kuniyu@xxxxxxxxxxxx> wrote:
>
> From: Jason Baron <jbaron@xxxxxxxxxx>
> Date: Wed, 28 Apr 2021 10:44:12 -0400
> > On 4/28/21 4:13 AM, Kuniyuki Iwashima wrote:
> > > From: Jason Baron <jbaron@xxxxxxxxxx>
> > > Date: Tue, 27 Apr 2021 12:38:58 -0400
> > >> On 4/26/21 11:46 PM, Kuniyuki Iwashima wrote:
> > >>> The SO_REUSEPORT option allows sockets to listen on the same port and to
> > >>> accept connections evenly. However, there is a defect in the current
> > >>> implementation [1]. When a SYN packet is received, the connection is tied
> > >>> to a listening socket. Accordingly, when the listener is closed, in-flight
> > >>> requests during the three-way handshake and child sockets in the accept
> > >>> queue are dropped even if other listeners on the same port could accept
> > >>> such connections.
> > >>>
> > >>> This situation can happen when various server management tools restart
> > >>> server (such as nginx) processes. For instance, when we change nginx
> > >>> configurations and restart it, it spins up new workers that respect the new
> > >>> configuration and closes all listeners on the old workers, resulting in the
> > >>> in-flight ACK of 3WHS is responded by RST.
> > >>
> > >> Hi Kuniyuki,
> > >>
> > >> I had implemented a different approach to this that I wanted to get your
> > >> thoughts about. The idea is to use unix sockets and SCM_RIGHTS to pass the
> > >> listen fd (or any other fd) around. Currently, if you have an 'old' webserver
> > >> that you want to replace with a 'new' webserver, you would need a separate
> > >> process to receive the listen fd and then have that process send the fd to
> > >> the new webserver, if they are not running con-currently. So instead what
> > >> I'm proposing is a 'delayed close' for a unix socket. That is, one could do:
> > >>
> > >> 1) bind unix socket with path '/sockets'
> > >> 2) sendmsg() the listen fd via the unix socket
> > >> 2) setsockopt() some 'timeout' on the unix socket (maybe 10 seconds or so)
> > >> 3) exit/close the old webserver and the listen socket
> > >> 4) start the new webserver
> > >> 5) create new unix socket and bind to '/sockets' (if has MAY_WRITE file permissions)
> > >> 6) recvmsg() the listen fd
> > >>
> > >> So the idea is that we set a timeout on the unix socket. If the new process
> > >> does not start and bind to the unix socket, it simply closes, thus releasing
> > >> the listen socket. However, if it does bind it can now call recvmsg() and
> > >> use the listen fd as normal. It can then simply continue to use the old listen
> > >> fds and/or create new ones and drain the old ones.
> > >>
> > >> Thus, the old and new webservers do not have to run concurrently. This doesn't
> > >> involve any changes to the tcp layer and can be used to pass any type of fd.
> > >> not sure if it's actually useful for anything else though.
> > >>
> > >> I'm not sure if this solves your use-case or not but I thought I'd share it.
> > >> One can also inherit the fds like in systemd's socket activation model, but
> > >> that again requires another process to hold open the listen fd.
> > >
> > > Thank you for sharing code.
> > >
> > > It seems bit more crash-tolerant than normal fd passing, but it can still
> > > suffer if the process dies before passing fds. With this patch set, we can
> > > migrate children sockets even if the process dies.
> > >
> >
> > I don't think crashing should be much of an issue. The old server can setup the
> > unix socket patch '/sockets' when it starts up and queue the listen sockets
> > there from the start. When it dies it will close all its fds, and the new
> > server can pick anything up any fds that are in the '/sockets' queue.
> >
> >
> > > Also, as Martin said, fd passing tends to make application complicated.
> > >
> >
> > It may be but perhaps its more flexible? It gives the new server the
> > chance to re-use the existing listen fds, close, drain and/or start new
> > ones. It also addresses the non-REUSEPORT case where you can't bind right
> > away.
>
> If the flexibility is really worth the complexity, we do not care about it.
> But, SO_REUSEPORT can give enough flexibility we want.
>
> With socket migration, there is no need to reuse listener (fd passing),
> drain children (incoming connections are automatically migrated if there is
> already another listener bind()ed), and of course another listener can
> close itself and migrated children.
>
> If two different approaches resolves the same issue and one does not need
> complexity in userspace, we select the simpler one.

Kernel bloat and complexity is _not_ the simplest choice.

Touching a complex part of TCP stack is quite risky.