Re: [CHECKER] repetitive/contradictory comparison bugs for 2.4.7

[Neil Brown]

On Tuesday July 24, [e-mail] wrote:
> [BUG] confusing: why check twice? race condition?
> /home/eparker/tmp/linux/2.4.7/drivers/md/md.c:1633:do_md_run: ERROR:INTERNAL_NULL3:1627:1633: Repetitive check: "pers[pnum]" is NULL so the conditional "pers[pnum] == 0" will always evaluate to true! [distance=4]
> 		 */
> 		printk(BAD_CHUNKSIZE);
> 		return -EINVAL;
> 	}
> 
> Start --->
> 	if (!pers[pnum])
> 	{
> #ifdef CONFIG_KMOD
> 		char module_name[80];
> 		sprintf (module_name, "md-personality-%d", pnum);
> 		request_module (module_name);
> Error --->
> 		if (!pers[pnum])
> #endif
> 			return -EINVAL;
> 	}

This one is not a bug.  "request_module" will hopefully load a module
and run it's init routine, which should set pers[pnum] to some proper
value.
Ofcourse, it might not, so we have to check.

Note that "pers" here is a global variable, and so it should be
expected that an external function - i.e. request_module - might
change it.

Thanks anyway.

NeilBrown