[Nut-upsuser] upsd crashes with a "broken pipe" error

[Zach La Celle]

On 02/15/2011 03:27 PM, Arjen de Korte wrote:
> Citeren Zach La Celle <lacelle at roboticresearch.com>:
>
>> You can see where the problem happens in parseconf.c, on line 125 
>> with the code:
>> /* resize the lists */
>> ctx->arglist = realloc(ctx->arglist,
>>                                    sizeof(char *) * ctx->numargs);
>
> With the given arguments, this boils down to
>
>     ctx->arglist = realloc(NULL, sizeof(char *));
>
> This is all normal. Upon the first invocation of add_arg_word, 
> ctx->arglist will be a NULL pointer (since there is nothing in the 
> list yet). This should then allocate a one element array of a pointer 
> to char (to store the
>
> "If ptr is a null pointer, realloc() shall be equivalent to malloc() 
> for the specified size."
>
> After that, all hell breaks loose, but that's out of our control.
>
> There is a slight problem in lines 131-132
>
>     ctx->argsize = realloc(ctx->argsize, sizeof(int *) * ctx->numargs);
>
> which should really read
>
>     ctx->argsize = realloc(ctx->argsize, sizeof(size_t) * ctx->numargs);
>
> but I doubt that sizeof(size_t) will be smaller that sizeof(int *), so 
> this just wastes a few bytes of memory.
>
>> This also might help:
>> (gdb) p *ctx
>> $4 = {f = 0x0, state = 5, ch = 9, arglist = 0x0, argsize = 0x0, 
>> numargs = 1, maxargs = 1, wordbuf = 0x61f2e0 "Z", wordptr = 0x61f2fd 
>> "", wordbufsize = 16, linenum = 0, error = 0, errmsg = '\000' 
>> <repeats 255 times>, errhandler = 0, magic = 7497264, arg_limit = 32, 
>> wordlen_limit = 512}
>
> None of these values is suspect.
>
>> If I go "up" in GDB to the pconf_char function, here is the character 
>> which is killing it:
>> (gdb) p ch
>> $6 = 9 '\t'
>
> This is expected. Any whitespace character ends the collection of 
> characters for the current argument and will start a new one. Nothing 
> out of the ordinary. If it was, 100% of the NUT installations would 
> suffer the same problems as you're seeing 100% of the time they start 
> the upsd server. This is not the case and even in your case, the 
> problem seems to occur intermittently, which is more an indication 
> you're either running out of memory or the system is suffering from 
> bad memory. Did you run a memory check lately?
>
> Best regards, Arjen
I will run memtest86 tonight when I can garner some time to bring the 
server down.  It has 4GB of memory and is currently utilizing only 
500MB, so I doubt it runs out.  Nothing else is crashing.  It only runs 
email and apache.

Thank you for your help,

-Zach