Re: [PATCH] serial: 8250_pcilib: Replace deprecated PCI functions

[Florian Eckert]

On 2025-09-29 08:00, Jiri Slaby wrote:
> On 26. 09. 25, 14:31, Florian Eckert wrote:
> > > > Signed-off-by: Florian Eckert <fe@xxxxxxxxxx>
> > > ...
> > > > --- a/drivers/tty/serial/8250/8250_pci.c
> > > > +++ b/drivers/tty/serial/8250/8250_pci.c
> > > > @@ -165,7 +165,15 @@ static int
> > > >   setup_port(struct serial_private *priv, struct uart_8250_port 
> > > > *port,
> > > >          u8 bar, unsigned int offset, int regshift)
> > > >   {
> > > > -    return serial8250_pci_setup_port(priv->dev, port, bar, offset, 
> > > > regshift);
> > > > +    void __iomem *iomem = NULL;
> > > > +
> > > > +    if (pci_resource_flags(priv->dev, bar) & IORESOURCE_MEM) {
> > > > +        iomem = pcim_iomap(priv->dev, bar, 0);
> > > > +        if (IS_ERR(iomem))
> > > > +            return -ENOMEM;
> > >
> > > Why not to propagate the error?
> >
> > Most other calls in the kernel of this function return
> > -ENOMEM on error. Therefore, I thought that this is the
> > correct return value. I can fix that in v2 if you like.
> > Let me know what you prefer.
>
> Ugh, pcim_iomap() returns NULL on error, so the IS_ERR() check is all
> wrong. What other places in the kernel use IS_ERR()? They need fixing.

You're right. The function returns a pointer here and my check IS_ERR()
is wrong. I mixed it up with the 'pcim_iomap_region()' function.
Because I wanted to use this function before. As far as I can see, the 
return
value of 'pcim_iomap()' is checked correctly in the linux sources.

Sorry for the noise.

I will adjust this and send a V2 patch.

--
Florian