[ic] [PATCH] [RFC] PGP hard failure workaround

[John Beima]

Quoting Dan Browning <db@kavod.com>:

> I have been experiencing the "PGP hard failure" error:
> 
> 	PGP failed with error level 12, status 12:
> 	PGP hard failure, command that failed: [...]
> 
> The symptom is that on a busy system, the encryption operation will 
> occasionally (1 in 100 or so) fail, resulting in no credit card at all.
> 
> Has anyone else experienced this error?  If not, could you try the steps
> to reproduce the problem (below)?  If you have, could you try the
> patch below and let me know if it helps?
> 
>  * Software: 
>    - Interchange 4.9 CVS, 05/13 and 04/04 both affected.
>    - Interchange RPC and HIGH mode both affected.
>    - Red Hat 7.3 and 9.0 both affected
>    - GPG 1.2.1
>    
>  * Hardware: 
>    - Athlon 2700 CPU
>    - Ultra160 Cheetah HDD
>    - 1GB RAM
>    
>  * Steps to reproduce the problem:
>    - Create the following test page:
>    
> 	<xmp>
> 	[calc]
> 		for (1..20) {
> 			my $results = $Tag->filter( 'encrypt', 'bob' );
> 			$out .= "Try $_: ";
> 			if ( $results !~ /^-----BEGIN PGP MESSAGE-----/ ) {
> 				$out .= "failed: results = $results\n";
> 			}
> 			else {
> 				$out .= "succeeded.\n";
> 			}
> 		}
> 		return $out;
> 	[/calc]
> 	</xmp>
> 	
>     - Access that page many times concurrently.  I was able to cause the 
>       problem to occur with apache bench (concurrency level 3):
> 	
>       ab -n 5 -c 3 http://path/to/test
>       
> I wrote a patch to workaround the problem, by retrying up to 3 times 
> (configurable).  You can get the patch to lib/Vend/Order.pm below or 
> download it here:
> 
> 	http://www.icdevgroup.org/~danb/pgp_hard_failure_workaround.patch
> 
> Feedback appreciated.
> 
> -- 
> Dan Browning, Kavod Technologies, <db@kavod.com> 360.843.4074x217
> 6700 NE 162nd Ave, Ste 611-210, Vancouver, WA.    Random Fortune:
> Deliberation, n.:
> 	The act of examining one's bread to determine which side it is
> 	buttered on.
> 		-- Ambrose Bierce, "The Devil's Dictionary"
> 
> 
> --- Order.pm	2003-04-23 09:01:02.000000000 -0700
> +++ /home/db/bin/ic49/lib/Vend/Order.pm	2003-06-02 18:35:49.000000000 -0700
> @@ -911,24 +911,38 @@
>  	my $fpre = $Vend::Cfg->{ScratchDir} . "/pgp.$Vend::Session->{id}.$$";
>  	$cmd .= " >$fpre.out";
>  	$cmd .= " 2>$fpre.err" unless $cmd =~ /2>/;
> -	open(PGP, "|$cmd")
> -			or die "Couldn't fork: $!";
> -	print PGP $body;
> -	close PGP;
> -
> -	if($?) {
> -		my $errno = $?;
> -		my $status = $errno;
> -		if($status > 255) {
> -			$status = $status >> 8;
> -			$! = $status;
> -		}
> -		logError("PGP failed with error level %s, status %s: $!", $?, $status);
> -		if($status) {
> -			logError("PGP hard failure, command that failed: %s", $cmd);
> -			return;
> +	
> +	my $tried;
> +	my $success;
> +	my $pgp_retries = ( $Vend::Cfg->{Limit}{pgp_retries} || 3 );
> +	PGP: until ( $success or $tried > $pgp_retries ) {
> +		$tried++;
> +		open(PGP, "|$cmd")
> +				or die "Couldn't fork: $!";
> +		print PGP $body;
> +		close PGP;
> +	
> +		if($?) {
> +			my $errno = $?;
> +			my $status = $errno;
> +			if($status > 255) {
> +				$status = $status >> 8;
> +				$! = $status;
> +			}
> +			logError("PGP failed with error level %s, status %s: $!", $?, $status);
> +			if($status) {
> +				logError("PGP hard failure, command that failed: %s.  " .
> +						 "Try number %s of %s.", $cmd, $tried, $pgp_retries - $tried );
> +	 			next PGP;
> +			}
>  		}
> +		$success = 1;
> +		last;
>  	}
> +	
> +	logError("PGP " . ( $success ? "succeeded " : "failed " ) . "after trying
> %s time(s).", $tried) 
> +			if $tried > 1;
> +		
>  	$body = readfile("$fpre.out");
>  	unlink "$fpre.out";
>  	unlink "$fpre.err";
> _______________________________________________
> interchange-users mailing list
> interchange-users@icdevgroup.org
> http://www.icdevgroup.org/mailman/listinfo/interchange-users
> 


Actually Dan, I may just have this one for you...

I found that the gnugpg that ships with RedHat only allows a key to be in use by
one gpg client at a time.

So if you have two orders hitting the server at the same time, the second one's
gpg will fail because the "key" has a lock on it by the first one.

Gpg does give an error back when this happens...


John Beima