2015-04-08 06:45:16

by Florian Fainelli

[permalink] [raw]
Subject: [PATCH] of/flattree: print memory scan node results in CPU endian

Commit 51975db0b7333 ("of/flattree: merge early_init_dt_scan_memory()
common code") consolidated some code from PowerPC (typically
big-endian), and ended-up adding a pr_debug() printing reg properties in
big-endian (DT native) format, not CPU endian. Unsurprisingly, when
these messages are turned on a little-endian systems, this is confusing,
so do the conversion while printing the values.

Signed-off-by: Florian Fainelli <[email protected]>
---
drivers/of/fdt.c | 3 ++-
1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/drivers/of/fdt.c b/drivers/of/fdt.c
index 3a896c9aeb74..0d8f0e4bd107 100644
--- a/drivers/of/fdt.c
+++ b/drivers/of/fdt.c
@@ -880,7 +880,8 @@ int __init early_init_dt_scan_memory(unsigned long node, const char *uname,
endp = reg + (l / sizeof(__be32));

pr_debug("memory scan node %s, reg size %d, data: %x %x %x %x,\n",
- uname, l, reg[0], reg[1], reg[2], reg[3]);
+ uname, l, be32_to_cpu(reg[0]), be32_to_cpu(reg[1]),
+ be32_to_cpu(reg[2]), be32_to_cpu(reg[3]));

while ((endp - reg) >= (dt_root_addr_cells + dt_root_size_cells)) {
u64 base, size;
--
2.1.0


2015-04-08 21:19:19

by Frank Rowand

[permalink] [raw]
Subject: Re: [PATCH] of/flattree: print memory scan node results in CPU endian

On 4/7/2015 11:44 PM, Florian Fainelli wrote:
> Commit 51975db0b7333 ("of/flattree: merge early_init_dt_scan_memory()
> common code") consolidated some code from PowerPC (typically
> big-endian), and ended-up adding a pr_debug() printing reg properties in
> big-endian (DT native) format, not CPU endian. Unsurprisingly, when
> these messages are turned on a little-endian systems, this is confusing,
> so do the conversion while printing the values.
>
> Signed-off-by: Florian Fainelli <[email protected]>
> ---
> drivers/of/fdt.c | 3 ++-
> 1 file changed, 2 insertions(+), 1 deletion(-)
>
> diff --git a/drivers/of/fdt.c b/drivers/of/fdt.c
> index 3a896c9aeb74..0d8f0e4bd107 100644
> --- a/drivers/of/fdt.c
> +++ b/drivers/of/fdt.c
> @@ -880,7 +880,8 @@ int __init early_init_dt_scan_memory(unsigned long node, const char *uname,
> endp = reg + (l / sizeof(__be32));
>
> pr_debug("memory scan node %s, reg size %d, data: %x %x %x %x,\n",
> - uname, l, reg[0], reg[1], reg[2], reg[3]);
> + uname, l, be32_to_cpu(reg[0]), be32_to_cpu(reg[1]),
> + be32_to_cpu(reg[2]), be32_to_cpu(reg[3]));

The pr_debug() assumes that the length of reg[] is >= 4 elements, which might not be true.
Since the following while loop checks the length of reg[] and then uses pr_debug() to
print each (base, size) tuple with the correct endian, maybe it would be better to
just remove reg[0] through reg[3] from the above pr_debug() instead of fixing the
endian issue.

>
> while ((endp - reg) >= (dt_root_addr_cells + dt_root_size_cells)) {
> u64 base, size;
>

-Frank

2015-04-09 03:17:28

by Florian Fainelli

[permalink] [raw]
Subject: Re: [PATCH] of/flattree: print memory scan node results in CPU endian

Le 08/04/2015 14:19, Frank Rowand a écrit :
> On 4/7/2015 11:44 PM, Florian Fainelli wrote:
>> Commit 51975db0b7333 ("of/flattree: merge early_init_dt_scan_memory()
>> common code") consolidated some code from PowerPC (typically
>> big-endian), and ended-up adding a pr_debug() printing reg properties in
>> big-endian (DT native) format, not CPU endian. Unsurprisingly, when
>> these messages are turned on a little-endian systems, this is confusing,
>> so do the conversion while printing the values.
>>
>> Signed-off-by: Florian Fainelli <[email protected]>
>> ---
>> drivers/of/fdt.c | 3 ++-
>> 1 file changed, 2 insertions(+), 1 deletion(-)
>>
>> diff --git a/drivers/of/fdt.c b/drivers/of/fdt.c
>> index 3a896c9aeb74..0d8f0e4bd107 100644
>> --- a/drivers/of/fdt.c
>> +++ b/drivers/of/fdt.c
>> @@ -880,7 +880,8 @@ int __init early_init_dt_scan_memory(unsigned long node, const char *uname,
>> endp = reg + (l / sizeof(__be32));
>>
>> pr_debug("memory scan node %s, reg size %d, data: %x %x %x %x,\n",
>> - uname, l, reg[0], reg[1], reg[2], reg[3]);
>> + uname, l, be32_to_cpu(reg[0]), be32_to_cpu(reg[1]),
>> + be32_to_cpu(reg[2]), be32_to_cpu(reg[3]));
>
> The pr_debug() assumes that the length of reg[] is >= 4 elements, which might not be true.
> Since the following while loop checks the length of reg[] and then uses pr_debug() to
> print each (base, size) tuple with the correct endian, maybe it would be better to
> just remove reg[0] through reg[3] from the above pr_debug() instead of fixing the
> endian issue.

Right, that works too, will follow-up with a patch removing the
reg[0..3] prints.

Thanks!
--
Florian

2015-04-12 20:16:35

by Florian Fainelli

[permalink] [raw]
Subject: [PATCH] of/fdt: Remove "reg" data prints from early_init_dt_scan_memory

Commit 51975db0b7333 ("of/flattree: merge early_init_dt_scan_memory()
common code") consolidated some code from PowerPC (typically
big-endian), and ended-up adding a pr_debug() printing reg properties in
big-endian (DT native) format, not CPU endian.

This debug print suffers from two problems:
- we only print 4 "reg" values, while there could be more on typical
systems having multiple memory ranges
- we print these 4 "reg" values in FDT endianess, that is big-endian,
and these values could be confusing for little-endian configurations

Since we are already printing the base address and size of the memory
regions parsed by early_init_dt_scan_memory() later in a way that is
both endian correct, and takes into account arbitrary number of memory
banks, just remove that part of the debug print.

Suggested-by: Frank Rowand <[email protected]>
Signed-off-by: Florian Fainelli <[email protected]>
---
Patch against devicetree/next

drivers/of/fdt.c | 3 +--
1 file changed, 1 insertion(+), 2 deletions(-)

diff --git a/drivers/of/fdt.c b/drivers/of/fdt.c
index 3a896c9..34bdc4d 100644
--- a/drivers/of/fdt.c
+++ b/drivers/of/fdt.c
@@ -879,8 +879,7 @@ int __init early_init_dt_scan_memory(unsigned long node, const char *uname,

endp = reg + (l / sizeof(__be32));

- pr_debug("memory scan node %s, reg size %d, data: %x %x %x %x,\n",
- uname, l, reg[0], reg[1], reg[2], reg[3]);
+ pr_debug("memory scan node %s, reg size %d,\n", uname, l);

while ((endp - reg) >= (dt_root_addr_cells + dt_root_size_cells)) {
u64 base, size;
--
2.1.0

2015-04-15 12:52:42

by Rob Herring

[permalink] [raw]
Subject: Re: [PATCH] of/fdt: Remove "reg" data prints from early_init_dt_scan_memory

On Sun, Apr 12, 2015 at 3:16 PM, Florian Fainelli <[email protected]> wrote:
> Commit 51975db0b7333 ("of/flattree: merge early_init_dt_scan_memory()
> common code") consolidated some code from PowerPC (typically
> big-endian), and ended-up adding a pr_debug() printing reg properties in
> big-endian (DT native) format, not CPU endian.
>
> This debug print suffers from two problems:
> - we only print 4 "reg" values, while there could be more on typical
> systems having multiple memory ranges
> - we print these 4 "reg" values in FDT endianess, that is big-endian,
> and these values could be confusing for little-endian configurations
>
> Since we are already printing the base address and size of the memory
> regions parsed by early_init_dt_scan_memory() later in a way that is
> both endian correct, and takes into account arbitrary number of memory
> banks, just remove that part of the debug print.
>
> Suggested-by: Frank Rowand <[email protected]>
> Signed-off-by: Florian Fainelli <[email protected]>
> ---
> Patch against devicetree/next

Applied. Thanks.

Rob

>
> drivers/of/fdt.c | 3 +--
> 1 file changed, 1 insertion(+), 2 deletions(-)
>
> diff --git a/drivers/of/fdt.c b/drivers/of/fdt.c
> index 3a896c9..34bdc4d 100644
> --- a/drivers/of/fdt.c
> +++ b/drivers/of/fdt.c
> @@ -879,8 +879,7 @@ int __init early_init_dt_scan_memory(unsigned long node, const char *uname,
>
> endp = reg + (l / sizeof(__be32));
>
> - pr_debug("memory scan node %s, reg size %d, data: %x %x %x %x,\n",
> - uname, l, reg[0], reg[1], reg[2], reg[3]);
> + pr_debug("memory scan node %s, reg size %d,\n", uname, l);
>
> while ((endp - reg) >= (dt_root_addr_cells + dt_root_size_cells)) {
> u64 base, size;
> --
> 2.1.0
>