On Wed, Apr 13, 2022 at 03:36:57PM +0300, David Kahurani wrote:
> On Mon, Apr 4, 2022 at 6:32 PM Dan Carpenter <dan.carpenter@xxxxxxxxxx> wrote:
>
> Hi Dan
>
> > > int ret;
> > > int (*fn)(struct usbnet *, u8, u8, u16, u16, void *, u16);
> > > @@ -201,9 +202,12 @@ static int __ax88179_read_cmd(struct usbnet *dev, u8 cmd, u16 value, u16 index,
> > > ret = fn(dev, cmd, USB_DIR_IN | USB_TYPE_VENDOR | USB_RECIP_DEVICE,
> > > value, index, data, size);
> > >
> > > - if (unlikely(ret < 0))
> > > + if (unlikely(ret < size)) {
> > > + ret = ret < 0 ? ret : -ENODATA;
> > > +
> > > netdev_warn(dev->net, "Failed to read reg index 0x%04x: %d\n",
> > > index, ret);
> > > + }
> > >
> > > return ret;
> >
> > It would be better to make __ax88179_read_cmd() return 0 on success
> > instead of returning size on success. Non-standard returns lead to bugs.
> >
>
> I don't suppose this would have much effect on the structure of the
> code and indeed plan to do this but just some minor clarification.
>
> Isn't it standard for reader functions to return the number of bytes read?
>
Not really.
There are some functions that do it, but it has historically lead to bug
after bug. For example, see commit 719b8f2850d3 ("USB: add
usb_control_msg_send() and usb_control_msg_recv()") where USB is moving
away from that to avoid bugs.
If you return zero on success then it's simple:
if (ret)
return ret;
If you return the bytes people will try call kinds of things:
if (ret)
return ret;
Bug: Now the driver is broken. (Not everyone can test the hardware).
if (ret != size)
return ret;
Bug: returns a positive.
if (ret != size)
return -EIO;
Bug: forgot to propagate the error code.
if (ret < sizeof(foo))
return -EIO;
Bug: because of type promotion negative error codes are treated as
success.
if (ret < 0)
return ret;
Bug: buffer partially filled. Information leak.
If you return the bytes then the only correct way to write error
handling is:
if (ret < 0)
return ret;
if (ret != size)
return -EIO;
regards,
dan carpenter
