linux-kernel - Re: [PATCH] PCI: mediatek: Fix sparse warning caused to virt_to

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [day] [month] [year] [list]

Message-ID: <20231121222325.GA260057@bhelgaas>
Date:   Tue, 21 Nov 2023 16:23:25 -0600
From:   Bjorn Helgaas <helgaas@...nel.org>
To:     Stanislav Kinsburskii <skinsburskii@...ux.microsoft.com>
Cc:     ryder.lee@...iatek.com, jianjun.wang@...iatek.com,
        lpieralisi@...nel.org, kw@...ux.com, robh@...nel.org,
        bhelgaas@...gle.com, matthias.bgg@...il.com,
        angelogioacchino.delregno@...labora.com, linux-pci@...r.kernel.org,
        linux-mediatek@...ts.infradead.org, linux-kernel@...r.kernel.org,
        linux-arm-kernel@...ts.infradead.org, skinsburskii@...il.com
Subject: Re: [PATCH] PCI: mediatek: Fix sparse warning caused to
 virt_to_phys() prototype change

On Tue, Nov 21, 2023 at 02:05:56PM -0800, Stanislav Kinsburskii wrote:
> On Tue, Nov 21, 2023 at 03:37:55PM -0600, Bjorn Helgaas wrote:
> > On Mon, Nov 20, 2023 at 03:40:33PM -0800, Stanislav Kinsburskii wrote:
> > > Explicitly cast __iomem pointer to const void* with __force to fix the
> > > following warning:
> > > 
> > >   warning: incorrect type in argument 1 (different address spaces)
> > >      expected void const volatile *address
> > >      got void [noderef] __iomem *
> > 
> > I have two questions about this:
> > 
> >   1) There's no other use of __force in drivers/pci, so I don't know
> >   what's special about pcie-mediatek.c.  There should be a way to fix
> >   the types so it's not needed.
> 
> __force suppreses the following sparse warning:
> 
>     warning: cast removes address space '__iomem' of expression

I'm suggesting that the cast is a band-aid that covers up a type
mismatch, and there shouldn't be a mismatch in the first place.

> >   2) virt_to_phys() is not quite right to begin with because what we
> >   want is a *bus* address, not the CPU physical address we get from
> >   virt_to_phys().  Obviously the current platforms that use this must
> >   not apply any offset between bus and CPU physical addresses, but
> >   it's not something we should rely on.
> > 
> >   There are only three drivers (pci-aardvark.c, pcie-xilinx.c, and
> >   this one) that use virt_to_phys(), and they're all slightly wrong
> >   here.
> > 
> > The *_compose_msi_msg() methods could use a little more consistency
> > across the board.
> 
> Could you elaborate on what do you suggest?
> Should virt_to_phys() be simply removed?

The DMA API (Documentation/core-api/dma-api.rst) is the usual way to
get bus addresses, since an MSI is basically a DMA on the PCI bus.

https://lore.kernel.org/linux-pci/20230914203146.GA77870@bhelgaas/

Nobody is very motivated to fix these, I guess ;)  I sort of hate to
just throw in a cast to shut up the warning because it doesn't really
solve the problem.

> > > Signed-off-by: Stanislav Kinsburskii <skinsburskii@...ux.microsoft.com>
> > > ---
> > >  drivers/pci/controller/pcie-mediatek.c |    4 ++--
> > >  1 file changed, 2 insertions(+), 2 deletions(-)
> > > 
> > > diff --git a/drivers/pci/controller/pcie-mediatek.c b/drivers/pci/controller/pcie-mediatek.c
> > > index 66a8f73296fc..27f0f79810a1 100644
> > > --- a/drivers/pci/controller/pcie-mediatek.c
> > > +++ b/drivers/pci/controller/pcie-mediatek.c
> > > @@ -397,7 +397,7 @@ static void mtk_compose_msi_msg(struct irq_data *data, struct msi_msg *msg)
> > >  	phys_addr_t addr;
> > >  
> > >  	/* MT2712/MT7622 only support 32-bit MSI addresses */
> > > -	addr = virt_to_phys(port->base + PCIE_MSI_VECTOR);
> > > +	addr = virt_to_phys((__force const void *)port->base + PCIE_MSI_VECTOR);
> > >  	msg->address_hi = 0;
> > >  	msg->address_lo = lower_32_bits(addr);
> > >  
> > > @@ -520,7 +520,7 @@ static void mtk_pcie_enable_msi(struct mtk_pcie_port *port)
> > >  	u32 val;
> > >  	phys_addr_t msg_addr;
> > >  
> > > -	msg_addr = virt_to_phys(port->base + PCIE_MSI_VECTOR);
> > > +	msg_addr = virt_to_phys((__force const void *)port->base + PCIE_MSI_VECTOR);
> > >  	val = lower_32_bits(msg_addr);
> > >  	writel(val, port->base + PCIE_IMSI_ADDR);