Re: panic inducing use of gwlist_extract_first in dlr_pgsql.c

Hi Alan,

unfortunately your patch is partially wrong, BUT thank you for spotting this BUG.
I commited corrected patch to SVN but here are comments to your patch:

     if (result == NULL || gwlist_len(result) < 1) {
         debug("dlr.pgsql", 0, "no rows found");
-        while((row = gwlist_extract_first(result)))
+        while(result && (row = gwlist_extract_first(result)) != NULL)

while loop doesn't make any sense here because result is ether NULL or just empty therefore no need to loop at all...

             gwlist_destroy(row, octstr_destroy_item);
         gwlist_destroy(result, NULL);
         return NULL;
@@ -282,7 +282,8 @@
         ret = atol(octstr_get_cstr(gwlist_get(gwlist_get(res, 0), 0)));
     }
 
-    gwlist_destroy(gwlist_extract_first(res), octstr_destroy_item);
+    if ( res != NULL ) 
+      gwlist_destroy(gwlist_extract_first(res), octstr_destroy_item);

I put this inside else because there we know the list is not null...

     gwlist_destroy(res, NULL);


Thanks,
Alex


Am 10.08.2011 um 06:51 schrieb Alan McNatty:

> Hi All,
> 
> I'm finding what I think is incorrect use of gwlist_extract_first in the
> postgres dlr implementations (it may also exist in others - I've not
> checked yet). The DLR methods issue 'error's when they fail to return
> results, etc but subsequent calls to gwlist_extract_first on NULL lists
> cause 'panic's.
> 
> What I'm testing is the situation when the DLR DB is available on
> start-up (we panic if it is not). If during during normal operation the
> database is shutdown or temporarily unavailable (network issue, etc).
> The select fail is an error but results in a panic.
> 
> 2011-08-10 16:37:43 [18552] [3] ERROR: PGSQL: SELECT count(*) FROM
> "dlr";
> 2011-08-10 16:37:43 [18552] [3] ERROR: PGSQL: FATAL:  terminating
> connection due to administrator command
> server closed the connection unexpectedly
> 	This probably means the server terminated abnormally
> 	before or while processing the request.
> 
> 2011-08-10 16:37:43 [18552] [3] ERROR: PGSQL: Select failed!
> 2011-08-10 16:37:43 [18552] [3] ERROR: PGSQL: Could not get count of DLR
> table
> 2011-08-10 16:37:43 [18552] [3] PANIC: gwlib/list.c:309:
> gwlist_extract_first: Assertion `list != NULL' failed.
> 2011-08-10 16:37:43 [18552] [3] PANIC: /usr/sbin/bearerbox(gw_panic
> +0x14b) [0x48b55b]
> 2011-08-10 16:37:43 [18552] [3]
> PANIC: /usr/sbin/bearerbox(gwlist_extract_first+0x94) [0x489874]
> 2011-08-10 16:37:43 [18552] [3] PANIC: /usr/sbin/bearerbox [0x41e3d3]
> 2011-08-10 16:37:43 [18552] [3]
> PANIC: /usr/sbin/bearerbox(bb_print_status+0x11d) [0x40edfd]
> 2011-08-10 16:37:43 [18552] [3] PANIC: /usr/sbin/bearerbox [0x415075]
> 2011-08-10 16:37:43 [18552] [3] PANIC: /usr/sbin/bearerbox [0x4823cf]
> 2011-08-10 16:37:43 [18552] [3] PANIC: /lib/libpthread.so.0
> [0x2b0e670a9fc7]
> 2011-08-10 16:37:43 [18552] [3] PANIC: /lib/libc.so.6(clone+0x6d)
> [0x2b0e67a8664d]
> 
> The attached patch addresses this (for postgres implementation only - I
> can check the others if required). Once applied The result on the status
> page is ..
> 
> DLR: -1 queued, using pgsql storage
> 
> And when a DLR is received ...
> 
> 2011-08-10 16:44:53 [18889] [11] ERROR: PGSQL: FATAL:  terminating
> connection due to administrator command
> server closed the connection unexpectedly
> 	This probably means the server terminated abnormally
> 	before or while processing the request.
> 
> 2011-08-10 16:44:53 [18889] [11] ERROR: PGSQL: Select failed!
> 2011-08-10 16:44:53 [18889] [11] DEBUG: no rows found
> 2011-08-10 16:44:53 [18889] [11] WARNING: DLR[pgsql]: DLR from SMSC<FOO>
> for DST<02xxxxxxxxx> not found.
> 2011-08-10 16:44:53 [18889] [11] ERROR: SMPP[FOO]: got DLR but could not
> find message or was not interested in it id<534001841355>
> dst<02xxxxxxxxx>, type<1>
> 
> Cheers,
> Alan
> <dlr_pgsql.c.patch>