Re: [PATCH 3/4] RDMA: manual changes for sysfs_emit and neatening

[Jason Gunthorpe <jgg@nvidia.com>]

On Thu, Oct 29, 2020 at 10:16:46AM -0700, Joe Perches wrote:
> On Wed, 2020-10-28 at 10:54 -0700, Joe Perches wrote:
> > On Wed, 2020-10-28 at 14:25 -0300, Jason Gunthorpe wrote:
> > > On Wed, Oct 07, 2020 at 07:36:26PM -0700, Joe Perches wrote:
> > > 
> > > > @@ -653,10 +651,7 @@ static ssize_t serial_show(struct device *device,
> > > >  		rdma_device_to_drv_device(device, struct qib_ibdev, rdi.ibdev);
> > > >  	struct qib_devdata *dd = dd_from_dev(dev);
> > > >  
> > > > 
> > > > -	buf[sizeof(dd->serial)] = '\0';
> > > > -	memcpy(buf, dd->serial, sizeof(dd->serial));
> > > > -	strcat(buf, "\n");
> > > > -	return strlen(buf);
> > > > +	return sysfs_emit(buf, "%s\n", dd->serial);
> > > >  }
> > > 
> > > This is not the same thing? dd->serial does not look null terminated,
> > > eg it is filled like this:
> > > 
> > > 		memcpy(dd->serial, ifp->if_serial, sizeof(ifp->if_serial));
> > > 
> > > From data read off the flash
> > 
> > It seems you are correct.
> > 
> > Maybe instead:
> > static ssize_t serial_show(struct device *device,
> > 			   struct device_attribute *attr, char *buf)
> > {
> > 	struct qib_ibdev *dev =
> > 		rdma_device_to_drv_device(device, struct qib_ibdev, rdi.ibdev);
> > 	struct qib_devdata *dd = dd_from_dev(dev);
> > 	const u8 *end = memchr(dd->serial, 0, ARRAY_SIZE(dd->serial));
> > 	int size = end ? end - dd->serial : ARRAY_SIZE(dd->serial);
> > 
> > 	return sysfs_emit(buf, "%*s\n", size, dd->serial);
> 
> I believe for this to actually be correct, this should be:
> 
> 	return sysfs_emit(buf, "%.*s\n", size, dd->serial);

Yes, I think so, I squished it in

Maybe Dennis can check it in the wip/jgg/for-next branch?

Thanks,
Jason