Re: [PATCH 3/3] tracing: clean up splice code

[Eduard - Gabriel Munteanu <eduard.munteanu@linux360.ro>]

On Mon, Feb 09, 2009 at 08:26:20PM +0200, Pekka Enberg wrote:
> Eduard - Gabriel Munteanu wrote:
>>> @@ -2573,45 +2607,25 @@ static ssize_t tracing_splice_read_pipe(struct 
>>> file *filp,
>>>  		ret = iter->trace->splice_read(iter, filp,
>>>  					       ppos, pipe, len, flags);
>>>  		if (ret)
>>> -			goto out;
>>> +			goto out_err;
>>>  	}
>>>   	ret = tracing_wait_pipe(filp);
>>>  	if (ret <= 0)
>>> -		goto out;
>>> +		goto out_err;
>>>   	if (!iter->ent && !find_next_entry_inc(iter)) {
>>>  		ret = -EFAULT;
>>> -		goto out;
>>> +		goto out_err;
>>>  	}
>>>   	/* Fill as many pages as possible. */
>>>  	for (i = 0, rem = len; i < PIPE_BUFFERS && rem; i++) {
>>>  		pages[i] = alloc_page(GFP_KERNEL);
>>> +		if (!pages[i])
>>> +			break;
>> I believe you should decrement 'i' before breaking, since we fill
>> spd.nr_pages just after the loop. In case the current page couldn't be
>> allocated, spd.nr_pages will be one too many (that is, 'i').
>
> But, but, if we fail for i == 1, for example, we want ->nr_pages == 1, no?
>
> 			Pekka

Sure, my bad. I even made the same assumption myself about the outer
'for' loop when I wrote it ('i' would end up last page + 1). The patch is
okay. :-)


	Eduard