re: bin/57476: dhcpd dumps core

[matthew green <mrg%eterna.com.au@localhost>]

>  From your debug info it looks as if in this loop:
>  
>  	for (bptr = hash [hashix]; bptr; bptr = bptr -> cdr) {
>  		if (((struct option_cache *)(bptr -> car)) -> option -> code ==
>  		    code)
>  			return (struct option_cache *)(bptr -> car);
>  	}
>  
>  bptr->car == 0   (NULL).   I can't tell on which iteration,

when i first saw this PR, it made me ponder, but having you
debug this far i'm convinced i have seen this problem on big
endian arm64 the last year and a half (ever since switching
my host over from a sparc64.)

IIRC, this system went live in late feb last year, and i've
seen this crash twice so far, so it does not happen often,
and i did not see any obvious errors in the code here, but
i suspect you looked further going on the next message.  my
last crash was april, and i haven't seen the log trigger.
 
i still have the latest core file lying around, but the
binary is newer since i'm running with this patch now:


Index: common/options.c
===================================================================
RCS file: /cvsroot/src/external/mpl/dhcp/dist/common/options.c,v
retrieving revision 1.5
diff -p -r1.5 options.c
*** common/options.c	13 Jan 2021 17:01:31 -0000	1.5
--- common/options.c	21 Jun 2023 20:12:29 -0000
*************** struct option_cache *lookup_hashed_optio
*** 2497,2502 ****
--- 2497,2507 ----
  
  	hashix = compute_option_hash (code);
  	for (bptr = hash [hashix]; bptr; bptr = bptr -> cdr) {
+ 		if (bptr -> car == NULL) {
+ 			log_error ("Null pointer in bptr -> cdr: %s(%d)",
+ 			   __func__, __LINE__);
+ 			continue;
+ 		}
  		if (((struct option_cache *)(bptr -> car)) -> option -> code ==
  		    code)
  			return (struct option_cache *)(bptr -> car);