Re: [PATCH 1/6] conf, passt: Don't try to log to stderr after we close it

[David Gibson <david@gibson.dropbear.id.au>]

[-- Attachment #1: Type: text/plain, Size: 5602 bytes --]

On Wed, Jun 19, 2024 at 10:13:46AM +0200, Stefano Brivio wrote:
> On Wed, 19 Jun 2024 12:06:01 +1000
> David Gibson <david@gibson.dropbear.id.au> wrote:
> 
> > On Tue, Jun 18, 2024 at 08:00:14AM +0200, Stefano Brivio wrote:
> > > On Tue, 18 Jun 2024 10:36:28 +1000
> > > David Gibson <david@gibson.dropbear.id.au> wrote:
> > >   
> > > > On Mon, Jun 17, 2024 at 02:03:14PM +0200, Stefano Brivio wrote:  
> > > > > If we don't run in foreground, we close standard error as we
> > > > > daemonise, so it makes no sense to check if the controlling terminal
> > > > > is an interactive terminal or if --force-stderr was given, to decide
> > > > > if we want to log to standard error.
> > > > > 
> > > > > Make --force-stderr depend on --foreground.
> > > > > 
> > > > > Signed-off-by: Stefano Brivio <sbrivio@redhat.com>    
> > > > 
> > > > Reviewed-by: David Gibson <david@gibson.dropbear.id.au>
> > > >   
> > > > > ---
> > > > >  conf.c  | 3 +++
> > > > >  passt.c | 2 +-
> > > > >  2 files changed, 4 insertions(+), 1 deletion(-)
> > > > > 
> > > > > diff --git a/conf.c b/conf.c
> > > > > index 94b3ed6..dbdbb62 100644
> > > > > --- a/conf.c
> > > > > +++ b/conf.c
> > > > > @@ -1693,6 +1693,9 @@ void conf(struct ctx *c, int argc, char **argv)
> > > > >  
> > > > >  	conf_ugid(runas, &uid, &gid);
> > > > >  
> > > > > +	if (!c->foreground && c->force_stderr)
> > > > > +		die("Can't log to standard error if not running in foreground");
> > > > > +
> > > > >  	if (logfile) {
> > > > >  		logfile_init(c->mode == MODE_PASTA ? "pasta" : "passt",
> > > > >  			     logfile, logsize);
> > > > > diff --git a/passt.c b/passt.c
> > > > > index a5e2c5a..aa9648a 100644
> > > > > --- a/passt.c
> > > > > +++ b/passt.c
> > > > > @@ -302,7 +302,7 @@ int main(int argc, char **argv)
> > > > >  	if (isolate_prefork(&c))
> > > > >  		die("Failed to sandbox process, exiting");
> > > > >  
> > > > > -	if (!c.force_stderr && !isatty(fileno(stderr)))
> > > > > +	if (!c.foreground || (!c.force_stderr && !isatty(fileno(stderr))))    
> > > > 
> > > > What's the rationale for the isatty() check in any case?  
> > > 
> > > To implement the behaviour from the man page:
> > > 
> > >        -e, --stderr
> > >               Log  to standard error too.  Default is to log to the sys‐
> > >               tem logger only, if started from an interactive  terminal,
> > >               and to both system logger and standard error otherwise.  
> > 
> > Hmm.. maybe I'm getting confused reading either the man page or the
> > code, but isn't that the opposite of what the code is doing?  The code
> > appears to be opening the log if we're *not* on an interactive terminal.
> 
> Ouch, right, this became the opposite of the original intended
> behaviour from 84a62b79a2bc ("passt: Also log to stderr, don't fork to
> background if not interactive") which again was implemented as a workaround
> for issues in the old version of the KubeVirt integration that's not used
> anymore.
> 
> The swap happened in 1e49d194d017 ("passt, pasta: Introduce command-line
> options and port re-mapping"), which makes me think that we should probably
> think of a reasonable default and meaning of --stderr regardless of the
> original workaround (which is surely not needed anymore) and implement it.

I agree.

> Probably it would make sense to log to standard error if we're running in
> foreground, unless a log file is specified. The check above would simply
> become:

I think logging to stderr when foreground even if we do have a logfile
makes even more sense.

> 	if (!c.foreground)
> 
> and we could accept -e, --stderr for compatibility only, but it wouldn't
> do anything.
> 
> > > which was needed, in turn, because of earlier versions of passt and of
> > > the KubeVirt integration where passt was running in foreground, but (of
> > > course) not attached to a terminal, and there was no option to force
> > > printing to standard error.
> > > 
> > > Given that KubeVirt doesn't have a system logger, it was otherwise
> > > impossible to get messages out of passt.
> > > 
> > > It's an abomination that just adds complexity now, and I don't think
> > > anybody uses it that way anymore. I guess we should drop that, together
> > > with qrap, in a few months from now.  
> > 
> > Right.. but what I'm after is the rationale for this depending on
> > whether it's an interactive terminal at all.  Seems to me the logical
> > thing to do is to (by default) always log to stderr before
> > daemonization (i.e. forever when in foreground mode), then to logfile
> > and/or syslog after daemonization.  Regardless of whether there's a
> > tty or not.
> 
> The original rationale was that if it was started from an interactive
> terminal, I didn't want the whole spam associated with it, but if it was
> started from a non-interactive terminal (KubeVirt integration), that was
> the only way to get messages out.
> 
> Then we had the opposite issue with KubeVirt (which is the reason why
> the log file disables stderr logging): passt started in a
> non-interactive terminal, but stderr logging would cause a lot of
> overhead in some KubeVirt component.

Again, KubeVirt can fix that with the Go equivalent of '2> /dev/null'.
I don't think we should make our semantics more complex for the sake
of that.

-- 
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 --]