Re: [PATCH 3/6] log, passt: Always print to stderr before initialisation is complete

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

[-- Attachment #1: Type: text/plain, Size: 5731 bytes --]

On Wed, Jun 19, 2024 at 10:17:21AM +0200, Stefano Brivio wrote:
> On Wed, 19 Jun 2024 12:10:38 +1000
> David Gibson <david@gibson.dropbear.id.au> wrote:
> 
> > On Tue, Jun 18, 2024 at 08:01:31AM +0200, Stefano Brivio wrote:
> > > On Tue, 18 Jun 2024 10:44:20 +1000
> > > David Gibson <david@gibson.dropbear.id.au> wrote:
> > >   
> > > > On Mon, Jun 17, 2024 at 02:03:16PM +0200, Stefano Brivio wrote:  
> > > > > After commit 15001b39ef1d ("conf: set the log level much earlier"), we
> > > > > had a phase during initialisation when messages wouldn't be printed to
> > > > > standard error anymore.
> > > > > 
> > > > > Commit f67238aa864d ("passt, log: Call __openlog() earlier, log to
> > > > > stderr until we detach") fixed that, but only for the case where no
> > > > > log files are given.
> > > > > 
> > > > > If a log file is configured, vlogmsg() will not call passt_vsyslog(),
> > > > > but during initialisation, LOG_PERROR is set, so to avoid duplicated
> > > > > prints (which would result from passt_vsyslog() printing to stderr),
> > > > > we don't call fprintf() from vlogmsg() either.
> > > > > 
> > > > > This is getting a bit too complicated. Instead of abusing LOG_PERROR,
> > > > > define an internal logging flag that clearly represents that we're not
> > > > > done with the initialisation phase yet.
> > > > > 
> > > > > If this flag is not set, make sure we always print to stderr, if the
> > > > > log mask matches. Then, set LOG_PERROR only as we set this internal
> > > > > flag, to make sure we don't duplicate messages.
> > > > > 
> > > > > Reported-by: Yalan Zhang <yalzhang@redhat.com>
> > > > > Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
> > > > > ---
> > > > >  log.c   |  4 +++-
> > > > >  log.h   |  1 +
> > > > >  passt.1 |  3 ++-
> > > > >  passt.c | 17 ++++++++++-------
> > > > >  4 files changed, 16 insertions(+), 9 deletions(-)
> > > > > 
> > > > > diff --git a/log.c b/log.c
> > > > > index 3b5a1c6..939bb93 100644
> > > > > --- a/log.c
> > > > > +++ b/log.c
> > > > > @@ -49,6 +49,7 @@ int		log_trace;		/* --trace mode enabled */
> > > > >  void vlogmsg(int pri, const char *format, va_list ap)
> > > > >  {
> > > > >  	bool debug_print = (log_mask & LOG_MASK(LOG_DEBUG)) && log_file == -1;
> > > > > +	bool before_daemon = !(log_flags & LOG_FLAG_DAEMON_READY);    
> > > > 
> > > > As in 2/6 would just a global bool be simpler than flags.
> > > >   
> > > > >  	bool early_print = !(log_flags & LOG_FLAG_CONF_PARSED);
> > > > >  	struct timespec tp;
> > > > >  
> > > > > @@ -71,7 +72,8 @@ void vlogmsg(int pri, const char *format, va_list ap)
> > > > >  		va_end(ap2);
> > > > >  	}
> > > > >  
> > > > > -	if (debug_print || (early_print && !(log_opt & LOG_PERROR))) {
> > > > > +	if (debug_print || early_print ||
> > > > > +	    (before_daemon && (log_mask & LOG_MASK(LOG_PRI(pri))))) {
> > > > >  		(void)vfprintf(stderr, format, ap);
> > > > >  		if (format[strlen(format)] != '\n')
> > > > >  			fprintf(stderr, "\n");
> > > > > diff --git a/log.h b/log.h
> > > > > index 6a3224a..680baab 100644
> > > > > --- a/log.h
> > > > > +++ b/log.h
> > > > > @@ -10,6 +10,7 @@
> > > > >  #include <syslog.h>
> > > > >  
> > > > >  #define LOG_FLAG_CONF_PARSED	BIT(0)	/* We already parsed logging options */
> > > > > +#define LOG_FLAG_DAEMON_READY	BIT(1)	/* Daemonised, or ready in foreground */
> > > > >  
> > > > >  #define LOGFILE_SIZE_DEFAULT		(1024 * 1024UL)
> > > > >  #define LOGFILE_CUT_RATIO		30	/* When full, cut ~30% size */
> > > > > diff --git a/passt.1 b/passt.1
> > > > > index 3a23a43..31e528e 100644
> > > > > --- a/passt.1
> > > > > +++ b/passt.1
> > > > > @@ -99,7 +99,8 @@ terminal, and to both system logger and standard error otherwise.
> > > > >  
> > > > >  .TP
> > > > >  .BR \-l ", " \-\-log-file " " \fIPATH\fR
> > > > > -Log to file \fIPATH\fR, not to standard error, and not to the system logger.
> > > > > +Log to file \fIPATH\fR, not to standard error (once initialisation is complete),
> > > > > +and not to the system logger.    
> > > > 
> > > > IIUC when -l is set we'll log to the logfile _as well as_ stderr
> > > > before we daemonize.  The description above doesn't exactly contradict
> > > > that, but seems to imply something different.  
> > > 
> > > Is that because "(once initialisation is complete)" doesn't clearly
> > > refer to "not to standard error"?  
> > 
> > Yes, or rather that it's not entirely clear it refers *only* to that
> > clause and not to the "Log to file" part at the beginning.
> > 
> > > I could go with something slightly more verbose:
> > > 
> > >   Log to file \fIPATH\fR, not to standard error, and not to the system
> > >   logger.
> > > 
> > >   During initialisation phase, that is, before forking to background,
> > >   or before being ready for communication when running in foreground,
> > >   messages are always printed to standard error as well.  
> > 
> > Hrm... so I want to say:
> > 
> >     Log to file PATH instead of to system logger.
> > 
> > Which may not be totally accurate for the current behaviour... but
> > seems like it might be a sensible behaviour.  That is, we typically
> > log to syslog, but -l replaces it with a logfile.  Regardless of
> > which, under some circumstances we'll also log to stderr.
> 
> No, not really: it's not regardless of that. If the log file is given,
> we don't want to log to standard error (once we're up and running)

But... why?

-- 
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 --]