Message-ID: <CAMuHMdW-E4ykVdCugCyJt7_uKZZHyc=jStiL7DOiq2RZr6GTvQ@mail.gmail.com>
Date:   Mon, 17 Oct 2022 17:49:01 +0200
From:   Geert Uytterhoeven <geert@...ux-m68k.org>
To:     Sean Anderson <sean.anderson@...o.com>
Cc:     "David S . Miller" <davem@...emloft.net>,
        Eric Dumazet <edumazet@...gle.com>,
        Jakub Kicinski <kuba@...nel.org>,
        Paolo Abeni <pabeni@...hat.com>, netdev@...r.kernel.org,
        Camelia Groza <camelia.groza@....com>,
        Madalin Bucur <madalin.bucur@....com>,
        open list <linux-kernel@...r.kernel.org>,
        linux-arm-kernel@...ts.infradead.org, linuxppc-dev@...ts.ozlabs.org
Subject: Re: [PATCH net-next v5 05/14] net: fman: Map the base address once

Hi Sean,

On Mon, Oct 17, 2022 at 5:34 PM Sean Anderson <sean.anderson@...o.com> wrote:
> On 10/17/22 11:15 AM, Geert Uytterhoeven wrote:
> > On Sat, Sep 3, 2022 at 12:00 AM Sean Anderson <sean.anderson@...o.com> wrote:
> >> We don't need to remap the base address from the resource twice (once in
> >> mac_probe() and again in set_fman_mac_params()). We still need the
> >> resource to get the end address, but we can use a single function call
> >> to get both at once.
> >>
> >> While we're at it, use platform_get_mem_or_io and devm_request_resource
> >> to map the resource. I think this is the more "correct" way to do things
> >> here, since we use the pdev resource, instead of creating a new one.
> >> It's still a bit tricky, since we need to ensure that the resource is a
> >> child of the fman region when it gets requested.
> >>
> >> Signed-off-by: Sean Anderson <sean.anderson@...o.com>
> >> Acked-by: Camelia Groza <camelia.groza@....com>
> >
> > Thanks for your patch, which is now commit 262f2b782e255b79
> > ("net: fman: Map the base address once") in v6.1-rc1.
> >
> >> --- a/drivers/net/ethernet/freescale/dpaa/dpaa_eth_sysfs.c
> >> +++ b/drivers/net/ethernet/freescale/dpaa/dpaa_eth_sysfs.c
> >> @@ -18,7 +18,7 @@ static ssize_t dpaa_eth_show_addr(struct device *dev,
> >>
> >>         if (mac_dev)
> >>                 return sprintf(buf, "%llx",
> >> -                               (unsigned long long)mac_dev->res->start);
> >> +                               (unsigned long long)mac_dev->vaddr);
> >
> > On 32-bit:
> >
> >     warning: cast from pointer to integer of different size
> > [-Wpointer-to-int-cast]
> >
> > Obviously you should cast to "uintptr_t" or "unsigned long" instead,
> > and change the "%llx" to "%p" or "%lx"...
>
> Isn't there a %px for this purpose?

Yes there is.  But if it makes sense to use that depends on the
still to be answered questions at the bottom...

> > However, taking a closer look:
> >   1. The old code exposed a physical address to user space, the new
> >      code exposes the mapped virtual address.
> >      Is that change intentional?
>
> No, this is not intentional. So to make this backwards-compatible, I
> suppose I need a virt_to_phys?

I think virt_to_phys() will work only on real memory, not on MMIO,
so you may need to reintroduce the resource again.

> >   2. Virtual addresses are useless in user space.
> >      Moreover, addresses printed by "%p" are obfuscated, as this is
> >      considered a security issue. Likewise for working around this by
> >      casting to an integer.
>
> Yes, you're right that this probably shouldn't be exposed to userspace.
>
> > What's the real purpose of dpaa_eth_show_addr()?
>
> I have no idea. This is a question for Madalin.
>
> > Perhaps it should be removed?
>
> That would be reasonable IMO.

Gr{oetje,eeting}s,

                        Geert

--
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@...ux-m68k.org

In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
                                -- Linus Torvalds

Powered by blists - more mailing lists