public inbox for passt-dev@passt.top
 help / color / mirror / code / Atom feed
From: David Gibson <david@gibson.dropbear.id.au>
To: Stefano Brivio <sbrivio@redhat.com>
Cc: Hayato Kiwata <dev@haytok.jp>, passt-dev@passt.top
Subject: Re: [PATCH] conf, pesto: Notify client of the result of rule updates
Date: Mon, 16 Nov 2026 14:29:42 +1100	[thread overview]
Message-ID: <avp4pcdPR57uV8Wz@gractus.seuss> (raw)
In-Reply-To: <20261115012923.2502e272@elisabeth>

[-- Attachment #1: Type: text/plain, Size: 7306 bytes --]

On Sun, Nov 15, 2026 at 01:29:25AM +0100, Stefano Brivio wrote:
> Thanks, it looks good to me, and I also tested it briefly.  Just two
> minor details (which I can also fix up on merge if you prefer):
> 
> On Fri, 13 Nov 2026 17:55:05 +0900
> Hayato Kiwata <dev@haytok.jp> wrote:
> 
> > Since commit 1776dcb7317e ("fwd, conf: Roll back to old rules if
> > switching to new ones fails"), pasta tries to restore the old rules when
> > switching to new rules fails. However, pesto has no way to know that,
> > and exits successfully as if the new rules were applied.
> > 
> > In conf_handler(), write a one-byte status to the control socket after
> > trying to apply the new rules. The status has three values:
> > PESTO_STATUS_OK, PESTO_STATUS_ROLLBACK, and PESTO_STATUS_FAILED. In
> > pesto, read the status after send_conf(), and if the update failed,
> > print what happened and exit with a non-zero status.
> > 
> > Note that we bump PESTO_PROTOCOL_VERSION to 3 because this adds a
> > message from the server. If the server version is lower than that,
> > pesto doesn't read the status. Also, if the server fails to receive
> > rules from the client, it still closes the connection without sending a
> > status.
> > 
> > Link: https://archives.passt.top/passt-dev/20261025153025.600a4fe9@elisabeth/
> > Signed-off-by: Hayato Kiwata <dev@haytok.jp>
> > 
> > ---
> > Tested with:
> > 
> > $ pesto -A --tcp-ports=127.0.0.1/8085:80 /tmp/.pasta.sock
> > 
> > $ pesto -A --tcp-ports=192.0.2.1/8080:80 /tmp/.pasta.sock
> > Failed to apply new rules, old rules restored
> > 
> > $ echo $?
> > 1
> > 
> > $ pesto -s /tmp/.pasta.sock
> > passt/pasta configuration (/tmp/.pasta.sock)
> >   HOST
> >     TCP [127.0.0.1]:8085  =>  80
> >   SPLICE
> > 
> > $ curl -sI localhost:8085 | head -1
> > HTTP/1.0 200 OK
> > ---
> >  conf.c  | 10 +++++++++-
> >  pesto.c | 19 +++++++++++++++++++
> >  pesto.h | 20 +++++++++++++++++++-
> >  3 files changed, 47 insertions(+), 2 deletions(-)
> > 
> > diff --git a/conf.c b/conf.c
> > index 7be0c79..f0fc606 100644
> > --- a/conf.c
> > +++ b/conf.c
> > @@ -2273,6 +2273,7 @@ fail:
> >  void conf_handler(struct ctx *c, uint32_t events)
> >  {
> >  	if (events & EPOLLIN) {
> > +		enum pesto_status status = PESTO_STATUS_OK;
> >  		unsigned pif;
> >  
> >  		/* Clear pending tables */
> > @@ -2297,11 +2298,18 @@ void conf_handler(struct ctx *c, uint32_t events)
> >  		}
> >  
> >  		if (fwd_listen_switch(c) < 0) {
> > +			status = PESTO_STATUS_ROLLBACK;
> >  			err("Error switching to new forwarding rules");
> > +
> >  			/* Switch back to old rules (still in c->fwd_pending) */
> > -			if (fwd_listen_switch(c) < 0)
> > +			if (fwd_listen_switch(c) < 0) {
> > +				status = PESTO_STATUS_FAILED;
> >  				err("Couldn't restore old forwarding rules");
> > +			}
> >  		}
> > +
> > +		if (write_u8(c->fd_control, status) < 0)
> > +			warn_perror("Failed to write status on control socket");
> >  	}
> >  
> >  	if (events & EPOLLHUP) {
> > diff --git a/pesto.c b/pesto.c
> > index eeaf274..89ece5d 100644
> > --- a/pesto.c
> > +++ b/pesto.c
> > @@ -513,6 +513,25 @@ int main(int argc, char **argv)
> >  
> >  	send_conf(s, &conf);
> >  
> > +	if (s_version >= PESTO_VERSION_STATUS) {
> > +		uint8_t status;
> > +
> > +		if (read_u8(s, &status) < 0)
> > +			die_perror("Couldn't read configuration status");
> > +
> > +		switch (status) {
> > +		case PESTO_STATUS_OK:
> > +			break;
> > +		case PESTO_STATUS_ROLLBACK:
> > +			die("Failed to apply new rules, old rules restored");
> > +		case PESTO_STATUS_FAILED:
> > +			die(
> > +"Failed to apply new rules, couldn't restore old ones");
> > +		default:
> > +			die("Unknown configuration status %"PRIu8, status);
> 
> It doesn't impact functionality at least for the moment, but if we just
> exit on error without shutdown() this becomes inconsistent with the
> non-error path for no particular reason.

I'm not sure if that's true.  Something that came to light in
debugging a bunch of the connection-end bugs we've had is that a
close() (and thereby also an _exit()) on a TCP socket _usually_ causes
a FIN rather than an RST.  That can be changed by poking SO_LINGER,
which we use.  This is a Unix socket, rather than a TCP socket, but
I'd similarly expect a close() or _exit() to result in a clean
shutdown unless we've explicitly configured it otherwise with
SO_LINGER.

In fact, looking at the SO_LINGER entry in socket(7) it appears that
_exit(3) will always linger (and therefore close cleanly), even if
SO_LINGER *is* set.

> 
> It might become relevant the day we decide to export the configuration
> over TCP as well, or if we implement any kind of "shell" mode for pesto.

That's a reasonable point, however.

> I'd rather turn those to warn() calls (we don't export err(), it could
> be done but I don't think we really need it) and add break statements
> for the PESTO_STATUS_ROLLBACK and PESTO_STATUS_FAILED cases.
> 
> > +		}
> > +	}
> > +
> >  noupdate:
> >  	if (shutdown(s, SHUT_RDWR) < 0 || close(s) < 0)
> >  		die_perror("Error shutting down control socket");
> > diff --git a/pesto.h b/pesto.h
> > index 8db701b..5065530 100644
> > --- a/pesto.h
> > +++ b/pesto.h
> > @@ -19,12 +19,30 @@
> >   * but was little enough used that we decided not to implement backwards
> >   * compatiblity code (i.e. a v2 pesto will not work with a v1 pasta)
> >   */
> > -#define PESTO_PROTOCOL_VERSION	2
> > +/* Version 3 added a status byte, sent by the server after trying to apply the
> > + * new rules (see enum pesto_status)
> > + */
> > +#define PESTO_VERSION_STATUS	3
> 
> By the way, I think this is pretty neat.  It didn't occur to me that we
> could (also) define version numbers with specific features.
> 
> > +#define PESTO_PROTOCOL_VERSION	3
> >  
> >  /* Maximum size of a pif name, including \0 */
> >  #define	PIF_NAME_SIZE	(128)
> >  #define PIF_NONE	0
> >  
> > +/**
> > + * enum pesto_status - Result of a configuration update (protocol version 3+)
> > + *
> 
> Nit: we're trying to make all the documentation kerneldoc compatible
> (https://docs.kernel.org/doc-guide/kernel-doc.html), mostly for the sake
> of https://bugs.passt.top/show_bug.cgi?id=35, and enums should have
> their values documented in the header rather than inline, say:
> 
>  * @PESTO_STATUS_OK:		New rules applied
> 
> see also:
> 
>   $ grep -B5 "enum [a-z0-9_]* {" *.c
> 
> ...I know, we haven't been very consistent with that so far.
> 
> > + * Sent by the server as a single byte after trying to apply the new rules.
> > + */
> > +enum pesto_status {
> > +	/* New rules applied */
> > +	PESTO_STATUS_OK = 0,
> > +	/* New rules failed, old rules restored */
> > +	PESTO_STATUS_ROLLBACK = 1,
> > +	/* New rules failed, old rules couldn't be restored either */
> > +	PESTO_STATUS_FAILED = 2,
> > +};
> > +
> >  /**
> >   * struct pesto_hello - Server introduction message
> >   * @magic:		PESTO_SERVER_MAGIC
> 
> -- 
> Stefano
> 

-- 
David Gibson (he or they)	| I'll have my music baroque, and my code
david AT gibson.dropbear.id.au	| minimalist, thank you, not the other way
				| around.
http://www.ozlabs.org/~dgibson

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 833 bytes --]

      reply	other threads:[~2026-11-16  4:34 UTC|newest]

Thread overview: 3+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2026-11-13  8:55 Hayato Kiwata
2026-11-15  0:29 ` Stefano Brivio
2026-11-16  3:29   ` David Gibson [this message]

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=avp4pcdPR57uV8Wz@gractus.seuss \
    --to=david@gibson.dropbear.id.au \
    --cc=dev@haytok.jp \
    --cc=passt-dev@passt.top \
    --cc=sbrivio@redhat.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
Code repositories for project(s) associated with this public inbox

	https://passt.top/passt

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for IMAP folder(s).