PCI: mvebu: Use for_each_of_range() iterator for parsing "ranges"

[PATCH] PCI: mvebu: Use for_each_of_range() iterator for parsing "ranges"

Posted by Rob Herring (Arm) 2 weeks, 2 days ago

The mvebu "ranges" is a bit unusual with its own encoding of addresses,
but it's still just normal "ranges" as far as parsing is concerned.
Convert mvebu_get_tgt_attr() to use the for_each_of_range() iterator
instead of open coding the parsing.

Signed-off-by: Rob Herring (Arm) <robh@kernel.org>
---
Compile tested only.
---
 drivers/pci/controller/pci-mvebu.c | 26 +++++++++-----------------
 1 file changed, 9 insertions(+), 17 deletions(-)

diff --git a/drivers/pci/controller/pci-mvebu.c b/drivers/pci/controller/pci-mvebu.c
index 29fe09c99e7d..d4e3f1e76f84 100644
--- a/drivers/pci/controller/pci-mvebu.c
+++ b/drivers/pci/controller/pci-mvebu.c
@@ -1179,37 +1179,29 @@ static int mvebu_get_tgt_attr(struct device_node *np, int devfn,
 			      unsigned int *tgt,
 			      unsigned int *attr)
 {
-	const int na = 3, ns = 2;
-	const __be32 *range;
-	int rlen, nranges, rangesz, pna, i;
+	struct of_range range;
+	struct of_range_parser parser;
 
 	*tgt = -1;
 	*attr = -1;
 
-	range = of_get_property(np, "ranges", &rlen);
-	if (!range)
+	if (of_pci_range_parser_init(&parser, np))
 		return -EINVAL;
 
-	pna = of_n_addr_cells(np);
-	rangesz = pna + na + ns;
-	nranges = rlen / sizeof(__be32) / rangesz;
-
-	for (i = 0; i < nranges; i++, range += rangesz) {
-		u32 flags = of_read_number(range, 1);
-		u32 slot = of_read_number(range + 1, 1);
-		u64 cpuaddr = of_read_number(range + na, pna);
+	for_each_of_range(&parser, &range) {
 		unsigned long rtype;
+		u32 slot = upper_32_bits(range.bus_addr);
 
-		if (DT_FLAGS_TO_TYPE(flags) == DT_TYPE_IO)
+		if (DT_FLAGS_TO_TYPE(range.flags) == DT_TYPE_IO)
 			rtype = IORESOURCE_IO;
-		else if (DT_FLAGS_TO_TYPE(flags) == DT_TYPE_MEM32)
+		else if (DT_FLAGS_TO_TYPE(range.flags) == DT_TYPE_MEM32)
 			rtype = IORESOURCE_MEM;
 		else
 			continue;
 
 		if (slot == PCI_SLOT(devfn) && type == rtype) {
-			*tgt = DT_CPUADDR_TO_TARGET(cpuaddr);
-			*attr = DT_CPUADDR_TO_ATTR(cpuaddr);
+			*tgt = DT_CPUADDR_TO_TARGET(range.cpu_addr);
+			*attr = DT_CPUADDR_TO_ATTR(range.cpu_addr);
 			return 0;
 		}
 	}
-- 
2.45.2

Re: [PATCH] PCI: mvebu: Use for_each_of_range() iterator for parsing "ranges"

Posted by Manivannan Sadhasivam 1 week, 1 day ago

On Thu, Nov 07, 2024 at 09:32:55AM -0600, Rob Herring (Arm) wrote:
> The mvebu "ranges" is a bit unusual with its own encoding of addresses,
> but it's still just normal "ranges" as far as parsing is concerned.
> Convert mvebu_get_tgt_attr() to use the for_each_of_range() iterator
> instead of open coding the parsing.
> 
> Signed-off-by: Rob Herring (Arm) <robh@kernel.org>

LGTM!

Reviewed-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>

Could someone please verify it on mvebu machine?

- Mani

> ---
> Compile tested only.
> ---
>  drivers/pci/controller/pci-mvebu.c | 26 +++++++++-----------------
>  1 file changed, 9 insertions(+), 17 deletions(-)
> 
> diff --git a/drivers/pci/controller/pci-mvebu.c b/drivers/pci/controller/pci-mvebu.c
> index 29fe09c99e7d..d4e3f1e76f84 100644
> --- a/drivers/pci/controller/pci-mvebu.c
> +++ b/drivers/pci/controller/pci-mvebu.c
> @@ -1179,37 +1179,29 @@ static int mvebu_get_tgt_attr(struct device_node *np, int devfn,
>  			      unsigned int *tgt,
>  			      unsigned int *attr)
>  {
> -	const int na = 3, ns = 2;
> -	const __be32 *range;
> -	int rlen, nranges, rangesz, pna, i;
> +	struct of_range range;
> +	struct of_range_parser parser;
>  
>  	*tgt = -1;
>  	*attr = -1;
>  
> -	range = of_get_property(np, "ranges", &rlen);
> -	if (!range)
> +	if (of_pci_range_parser_init(&parser, np))
>  		return -EINVAL;
>  
> -	pna = of_n_addr_cells(np);
> -	rangesz = pna + na + ns;
> -	nranges = rlen / sizeof(__be32) / rangesz;
> -
> -	for (i = 0; i < nranges; i++, range += rangesz) {
> -		u32 flags = of_read_number(range, 1);
> -		u32 slot = of_read_number(range + 1, 1);
> -		u64 cpuaddr = of_read_number(range + na, pna);
> +	for_each_of_range(&parser, &range) {
>  		unsigned long rtype;
> +		u32 slot = upper_32_bits(range.bus_addr);
>  
> -		if (DT_FLAGS_TO_TYPE(flags) == DT_TYPE_IO)
> +		if (DT_FLAGS_TO_TYPE(range.flags) == DT_TYPE_IO)
>  			rtype = IORESOURCE_IO;
> -		else if (DT_FLAGS_TO_TYPE(flags) == DT_TYPE_MEM32)
> +		else if (DT_FLAGS_TO_TYPE(range.flags) == DT_TYPE_MEM32)
>  			rtype = IORESOURCE_MEM;
>  		else
>  			continue;
>  
>  		if (slot == PCI_SLOT(devfn) && type == rtype) {
> -			*tgt = DT_CPUADDR_TO_TARGET(cpuaddr);
> -			*attr = DT_CPUADDR_TO_ATTR(cpuaddr);
> +			*tgt = DT_CPUADDR_TO_TARGET(range.cpu_addr);
> +			*attr = DT_CPUADDR_TO_ATTR(range.cpu_addr);
>  			return 0;
>  		}
>  	}
> -- 
> 2.45.2
> 

-- 
மணிவண்ணன் சதாசிவம்

Re: [PATCH] PCI: mvebu: Use for_each_of_range() iterator for parsing "ranges"

Posted by Pali Rohár 2 weeks, 2 days ago

On Thursday 07 November 2024 09:32:55 Rob Herring (Arm) wrote:
> The mvebu "ranges" is a bit unusual with its own encoding of addresses,
> but it's still just normal "ranges" as far as parsing is concerned.
> Convert mvebu_get_tgt_attr() to use the for_each_of_range() iterator
> instead of open coding the parsing.
> 
> Signed-off-by: Rob Herring (Arm) <robh@kernel.org>
> ---
> Compile tested only.

I see no reason for such change, which was even non tested at all and
does not fix any issue. There are more important issues in the driver,
it was decided that bug fixes are not going to be included (yet).

> ---
>  drivers/pci/controller/pci-mvebu.c | 26 +++++++++-----------------
>  1 file changed, 9 insertions(+), 17 deletions(-)
> 
> diff --git a/drivers/pci/controller/pci-mvebu.c b/drivers/pci/controller/pci-mvebu.c
> index 29fe09c99e7d..d4e3f1e76f84 100644
> --- a/drivers/pci/controller/pci-mvebu.c
> +++ b/drivers/pci/controller/pci-mvebu.c
> @@ -1179,37 +1179,29 @@ static int mvebu_get_tgt_attr(struct device_node *np, int devfn,
>  			      unsigned int *tgt,
>  			      unsigned int *attr)
>  {
> -	const int na = 3, ns = 2;
> -	const __be32 *range;
> -	int rlen, nranges, rangesz, pna, i;
> +	struct of_range range;
> +	struct of_range_parser parser;
>  
>  	*tgt = -1;
>  	*attr = -1;
>  
> -	range = of_get_property(np, "ranges", &rlen);
> -	if (!range)
> +	if (of_pci_range_parser_init(&parser, np))
>  		return -EINVAL;
>  
> -	pna = of_n_addr_cells(np);
> -	rangesz = pna + na + ns;
> -	nranges = rlen / sizeof(__be32) / rangesz;
> -
> -	for (i = 0; i < nranges; i++, range += rangesz) {
> -		u32 flags = of_read_number(range, 1);
> -		u32 slot = of_read_number(range + 1, 1);
> -		u64 cpuaddr = of_read_number(range + na, pna);
> +	for_each_of_range(&parser, &range) {
>  		unsigned long rtype;
> +		u32 slot = upper_32_bits(range.bus_addr);
>  
> -		if (DT_FLAGS_TO_TYPE(flags) == DT_TYPE_IO)
> +		if (DT_FLAGS_TO_TYPE(range.flags) == DT_TYPE_IO)
>  			rtype = IORESOURCE_IO;
> -		else if (DT_FLAGS_TO_TYPE(flags) == DT_TYPE_MEM32)
> +		else if (DT_FLAGS_TO_TYPE(range.flags) == DT_TYPE_MEM32)
>  			rtype = IORESOURCE_MEM;
>  		else
>  			continue;
>  
>  		if (slot == PCI_SLOT(devfn) && type == rtype) {
> -			*tgt = DT_CPUADDR_TO_TARGET(cpuaddr);
> -			*attr = DT_CPUADDR_TO_ATTR(cpuaddr);
> +			*tgt = DT_CPUADDR_TO_TARGET(range.cpu_addr);
> +			*attr = DT_CPUADDR_TO_ATTR(range.cpu_addr);
>  			return 0;
>  		}
>  	}
> -- 
> 2.45.2
>

Re: [PATCH] PCI: mvebu: Use for_each_of_range() iterator for parsing "ranges"

Posted by Rob Herring 2 weeks, 2 days ago

On Thu, Nov 7, 2024 at 11:00 AM Pali Rohár <pali@kernel.org> wrote:
>
> On Thursday 07 November 2024 09:32:55 Rob Herring (Arm) wrote:
> > The mvebu "ranges" is a bit unusual with its own encoding of addresses,
> > but it's still just normal "ranges" as far as parsing is concerned.
> > Convert mvebu_get_tgt_attr() to use the for_each_of_range() iterator
> > instead of open coding the parsing.
> >
> > Signed-off-by: Rob Herring (Arm) <robh@kernel.org>
> > ---
> > Compile tested only.
>
> I see no reason for such change, which was even non tested at all and
> does not fix any issue.

Maintenance of the kernel is an issue. Maybe not for you, but for some of us.

>  There are more important issues in the driver,
> it was decided that bug fixes are not going to be included (yet).

The larger reason is to get rid of custom parsing of address
properties throughout the kernel. As well as remove and consolidate
uses of_n_addr_cells() and remove its support of long since deprecated
behavior. Also, I intend to make of_get_property/of_find_property()
warn about leaking data since it just returns a pointer into the raw
DT data and we have no control over the lifetime. Then we can't free
those properties. Why do we care? Overlays and Rust.

Rob