Re: [PATCH 13/13] libnvdimm, namespace: Publish page structure init state / control

[Jeff Moyer]

Dan Williams <dan.j.williams@xxxxxxxxx> writes:

> On Thu, Jul 5, 2018 at 12:49 PM, Matthew Wilcox <willy@xxxxxxxxxxxxx> wrote:
>> On Thu, Jul 05, 2018 at 07:46:05AM -0700, Dan Williams wrote:
>>> On Thu, Jul 5, 2018 at 1:29 AM, Johannes Thumshirn <jthumshirn@xxxxxxx> wrote:
>>> > On Wed, Jul 04, 2018 at 11:50:13PM -0700, Dan Williams wrote:
>>> >> +static ssize_t memmap_state_store(struct device *dev,
>>> >> +             struct device_attribute *attr, const char *buf, size_t len)
>>> >> +{
>>> >> +     int i;
>>> >> +     struct nd_pfn *nd_pfn = to_nd_pfn_safe(dev);
>>> >> +     struct memmap_async_state *async = &nd_pfn->async;
>>> >> +
>>> >> +     if (strcmp(buf, "sync") == 0)
>>> >> +             /* pass */;
>>> >> +     else if (strcmp(buf, "sync\n") == 0)
>>> >> +             /* pass */;
>>> >> +     else
>>> >> +             return -EINVAL;
>>> >
>>> > Hmm what about:
>>> >
>>> >         if (strncmp(buf, "sync", 4))
>>> >            return -EINVAL;
>>> >
>>> > This collapses 6 lines into 4.
>>>
>>> ...but that also allows 'echo "syncAndThenSomeGarbage" >
>>> /sys/.../memmap_state' to succeed.
>>
>>         if (strncmp(buf, "sync", 4))
>>                 return -EINVAL;
>>         if (buf[4] != '\0' && buf[4] != '\n')
>>                 return -EINVAL;
>>
>
> Not sure that's a win either, I'd rather just:
>
> +       if (strcmp(buf, "sync") == 0 || strcmp(buf, "sync\n") == 0)
> +               /* pass */;
> +       else
> +               return -EINVAL;
>
> If we're trying to save those 2 lines.

WFM.  I don't like that I had to go digging around in sysfs
documentation to convince myself that strcmp was safe, but I guess
that's my problem.  ;-)

Cheers,
Jeff