A missing or empty dma-ranges in a DT node implies a 1:1 mapping for dma
translations. In this specific case, the current behaviour is to zero out
the entire specifier so that the translation could be carried on as an
offset from zero.  This includes address specifier that has flags (e.g.
PCI ranges).
Once the flags portion has been zeroed, the translation chain is broken
since the mapping functions will check the upcoming address specifier
against mismatching flags, always failing the 1:1 mapping and its entire
purpose of always succeeding.
Set to zero only the address portion while passing the flags through.

Fixes: dbbdee94734b ("of/address: Merge all of the bus translation code")
Cc: stable@vger.kernel.org
Signed-off-by: Andrea della Porta <andrea.porta@suse.com>
Tested-by: Herve Codina <herve.codina@bootlin.com>
---
 drivers/of/address.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/drivers/of/address.c b/drivers/of/address.c
index 286f0c161e33..72b6accff21c 100644
--- a/drivers/of/address.c
+++ b/drivers/of/address.c
@@ -455,7 +455,8 @@ static int of_translate_one(struct device_node *parent, struct of_bus *bus,
 	}
 	if (ranges == NULL || rlen == 0) {
 		offset = of_read_number(addr, na);
-		memset(addr, 0, pna * 4);
+		/* copy the address while preserving the flags */
+		memset(addr + pbus->flag_cells, 0, (pna - pbus->flag_cells) * 4);
 		pr_debug("empty ranges; 1:1 translation\n");
 		goto finish;
 	}
-- 
2.35.3

On Fri, Nov 08, 2024 at 11:39:21AM +0100, Andrea della Porta wrote:
> A missing or empty dma-ranges in a DT node implies a 1:1 mapping for dma
> translations. In this specific case, the current behaviour is to zero out
> the entire specifier so that the translation could be carried on as an
> offset from zero.  This includes address specifier that has flags (e.g.
> PCI ranges).
> Once the flags portion has been zeroed, the translation chain is broken
> since the mapping functions will check the upcoming address specifier
> against mismatching flags, always failing the 1:1 mapping and its entire
> purpose of always succeeding.
> Set to zero only the address portion while passing the flags through.

Add blank lines between paragraphs.

> Fixes: dbbdee94734b ("of/address: Merge all of the bus translation code")
> Cc: stable@vger.kernel.org
> Signed-off-by: Andrea della Porta <andrea.porta@suse.com>
> Tested-by: Herve Codina <herve.codina@bootlin.com>
> ---
>  drivers/of/address.c | 3 ++-
>  1 file changed, 2 insertions(+), 1 deletion(-)
> 
> diff --git a/drivers/of/address.c b/drivers/of/address.c
> index 286f0c161e33..72b6accff21c 100644
> --- a/drivers/of/address.c
> +++ b/drivers/of/address.c
> @@ -455,7 +455,8 @@ static int of_translate_one(struct device_node *parent, struct of_bus *bus,
>  	}
>  	if (ranges == NULL || rlen == 0) {
>  		offset = of_read_number(addr, na);
> -		memset(addr, 0, pna * 4);
> +		/* copy the address while preserving the flags */

Not knowing the surrounding code, it seems strange to say "copy the
address" when the memset() fills with zero and does no copying.

The commit log says "set address to zero, pass flags through," and I
could believe *that* matches the memset().

> +		memset(addr + pbus->flag_cells, 0, (pna - pbus->flag_cells) * 4);
>  		pr_debug("empty ranges; 1:1 translation\n");
>  		goto finish;
>  	}
> -- 
> 2.35.3
>

Hi Bjorn,

On 10:56 Fri 08 Nov     , Bjorn Helgaas wrote:
> On Fri, Nov 08, 2024 at 11:39:21AM +0100, Andrea della Porta wrote:
> > A missing or empty dma-ranges in a DT node implies a 1:1 mapping for dma
> > translations. In this specific case, the current behaviour is to zero out
> > the entire specifier so that the translation could be carried on as an
> > offset from zero.  This includes address specifier that has flags (e.g.
> > PCI ranges).
> > Once the flags portion has been zeroed, the translation chain is broken
> > since the mapping functions will check the upcoming address specifier
> > against mismatching flags, always failing the 1:1 mapping and its entire
> > purpose of always succeeding.
> > Set to zero only the address portion while passing the flags through.
> 
> Add blank lines between paragraphs.

Ack.

> 
> > Fixes: dbbdee94734b ("of/address: Merge all of the bus translation code")
> > Cc: stable@vger.kernel.org
> > Signed-off-by: Andrea della Porta <andrea.porta@suse.com>
> > Tested-by: Herve Codina <herve.codina@bootlin.com>
> > ---
> >  drivers/of/address.c | 3 ++-
> >  1 file changed, 2 insertions(+), 1 deletion(-)
> > 
> > diff --git a/drivers/of/address.c b/drivers/of/address.c
> > index 286f0c161e33..72b6accff21c 100644
> > --- a/drivers/of/address.c
> > +++ b/drivers/of/address.c
> > @@ -455,7 +455,8 @@ static int of_translate_one(struct device_node *parent, struct of_bus *bus,
> >  	}
> >  	if (ranges == NULL || rlen == 0) {
> >  		offset = of_read_number(addr, na);
> > -		memset(addr, 0, pna * 4);
> > +		/* copy the address while preserving the flags */
> 
> Not knowing the surrounding code, it seems strange to say "copy the
> address" when the memset() fills with zero and does no copying.
> 
> The commit log says "set address to zero, pass flags through," and I
> could believe *that* matches the memset().

Ack.

Many thanks,
Andrea

> 
> > +		memset(addr + pbus->flag_cells, 0, (pna - pbus->flag_cells) * 4);
> >  		pr_debug("empty ranges; 1:1 translation\n");
> >  		goto finish;
> >  	}
> > -- 
> > 2.35.3
> >