On Wed, Aug 28, 2024 at 11:08:31PM +0200, Philipp Hortmann wrote:
> > diff --git a/drivers/staging/rtl8712/rtl871x_io.c b/drivers/staging/rtl8712/rtl871x_io.c
> > index 6789a4c98564..6311ac15c581 100644
> > --- a/drivers/staging/rtl8712/rtl871x_io.c
> > +++ b/drivers/staging/rtl8712/rtl871x_io.c
> > @@ -48,10 +48,10 @@ static uint _init_intf_hdl(struct _adapter *padapter,
> > set_intf_funs = &(r8712_usb_set_intf_funs);
> > set_intf_ops = &r8712_usb_set_intf_ops;
> > init_intf_priv = &r8712_usb_init_intf_priv;
> > - pintf_priv = pintf_hdl->pintfpriv = kmalloc(sizeof(struct intf_priv),
> > - GFP_ATOMIC);
> > + pintf_priv = kmalloc(sizeof(struct intf_priv), GFP_ATOMIC);
> > if (!pintf_priv)
> > goto _init_intf_hdl_fail;
>
> By pushing the below statement after the "if (!pintf_priv)" you change the
> logic. Is this really wanted? Why do you think it is better? I would avoid
> this and it would be a separate patch anyhow.
>
> > + pintf_hdl->pintfpriv = pintf_priv;
I liked moving it. And I feel like it should be done in this patch, not as a
separate patch. But it should have some justification as you say. The commit
message could say something like:
Checkpatch complains that we should avoid multiple assignments on the same
line for readability purposes. Generally, we would allocate, check and then
assign. It doesn't matter what is assigned to "pintf_hdl->pintfpriv" on the
error path. For example, on subsequent error paths "pintf_hdl->pintfpriv"
is a freed pointer. So this code is okay as-is and it's also okay to move
the pintf_hdl->pintfpriv = pintf_priv assignment after the NULL check.
(Notice how I sold this as one thing, moving the "pintf_hdl->pintfpriv"
assignment, not silencing checkpatch and then moving it. Notice how I avoided
using the word "also".)
regards,
dan carpenter
