Re: [PATCH] ath6kl: Use vmalloc to allocate ar->fw for api1 method

[Souptick Joarder]

Brent,

On Tue, Dec 22, 2015 at 2:42 PM, Kalle Valo  wrote:
> Souptick Joarder  writes:
>
>> Hi Brent,
>>
>> On Tue, Dec 22, 2015 at 3:23 AM, Brent Taylor  wrote:
>>> On Mon, Dec 21, 2015 at 1:23 PM, Souptick Joarder  
>>> wrote:
 Hi Brent,

 On Tue, Dec 1, 2015 at 11:11 AM, Brent Taylor  wrote:

> --- a/drivers/net/wireless/ath/ath6kl/init.c
> +++ b/drivers/net/wireless/ath/ath6kl/init.c
> @@ -673,10 +673,15 @@ static int ath6kl_get_fw(struct ath6kl *ar, const 
> char *filename,
> return ret;
>
> *fw_len = fw_entry->size;
> -   *fw = kmemdup(fw_entry->data, fw_entry->size, GFP_KERNEL);
> +   if (>fw == fw)
> +   *fw = vmalloc(fw_entry->size);
> +   else
> +   *fw = kmalloc(fw_entry->size, GFP_KERNEL)

   Why vmalloc and kmalloc both are required? can't use either
 vmalloc or kmalloc?
>>>
>>> My original problem was that kmemdup (which uses kmalloc) could not
>>> allocate enough memory
>>
>> If kmemdump ( which uses kmalloc) could not allocate memory then
>> using kmalloc again can lead to same problem.
>> I guess it will be correct to use
>> *fw = vmalloc(fw_entry->size);
>> Correct me if i am wrong.
>
> That sounds best. But remember take into account DMA requirements, IIRC
> you cannot DMA from vmalloc memory on all platforms.

Is it possible to modify the patch as per feedback from Kalle.

>
> --
> Kalle Valo

-Souptick
--
To unsubscribe from this list: send the line "unsubscribe linux-wireless" in
the body of a message to majord...@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: [PATCH] ath6kl: Use vmalloc to allocate ar->fw for api1 method

[Souptick Joarder]

Hi Brent,

On Tue, Dec 22, 2015 at 3:23 AM, Brent Taylor  wrote:
> On Mon, Dec 21, 2015 at 1:23 PM, Souptick Joarder  
> wrote:
>> Hi Brent,
>>
>> On Tue, Dec 1, 2015 at 11:11 AM, Brent Taylor  wrote:
>>> Since commit 8437754c83351d6213c1a47ff029c1126d6042a7, ar->fw is expected 
>>> to be pointing to memory allocated by vmalloc.  If the api1 method (via 
>>> ath6kl_fetch_fw_api1) is used to allocate memory for ar->fw, then kmemdup 
>>> is used.  This patch checks if the firmware being loaded is the 'fw' image, 
>>> then use vmalloc, otherwise use kmalloc.
>>>
>>> Signed-off-by: Brent Taylor 
>>> ---
>>>  drivers/net/wireless/ath/ath6kl/init.c | 7 ++-
>>>  1 file changed, 6 insertions(+), 1 deletion(-)
>>>
>>> diff --git a/drivers/net/wireless/ath/ath6kl/init.c 
>>> b/drivers/net/wireless/ath/ath6kl/init.c
>>> index 6ae0734..4f2b124d 100644
>>> --- a/drivers/net/wireless/ath/ath6kl/init.c
>>> +++ b/drivers/net/wireless/ath/ath6kl/init.c
>>> @@ -673,10 +673,15 @@ static int ath6kl_get_fw(struct ath6kl *ar, const 
>>> char *filename,
>>> return ret;
>>>
>>> *fw_len = fw_entry->size;
>>> -   *fw = kmemdup(fw_entry->data, fw_entry->size, GFP_KERNEL);
>>> +   if (>fw == fw)
>>> +   *fw = vmalloc(fw_entry->size);
>>> +   else
>>> +   *fw = kmalloc(fw_entry->size, GFP_KERNEL)
>>
>>   Why vmalloc and kmalloc both are required? can't use either
>> vmalloc or kmalloc?
>
> My original problem was that kmemdup (which uses kmalloc) could not
> allocate enough memory

If kmemdump ( which uses kmalloc) could not allocate memory then
using kmalloc again can lead to same problem.
I guess it will be correct to use
*fw = vmalloc(fw_entry->size);
Correct me if i am wrong.

> to hold the firmware that is placed into "ar->fw".  In the function
> ath6kl_core_cleanup (in core.c),
> the "ar->fw" pointer is the only one that uses vfree which was changed in 
> commit
> 8437754c83351d6213c1a47ff029c1126d6042a7.
>
> I was trying to change as little as possible and I wasn't sure if
> there was a reason that any of the
> other firmware items needed to be allocated with kmalloc or if they
> could be changed to use vmalloc.

-Souptick
--
To unsubscribe from this list: send the line "unsubscribe linux-wireless" in
the body of a message to majord...@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: [PATCH] ath6kl: Use vmalloc to allocate ar->fw for api1 method

[Kalle Valo]

Souptick Joarder  writes:

> Hi Brent,
>
> On Tue, Dec 22, 2015 at 3:23 AM, Brent Taylor  wrote:
>> On Mon, Dec 21, 2015 at 1:23 PM, Souptick Joarder  
>> wrote:
>>> Hi Brent,
>>>
>>> On Tue, Dec 1, 2015 at 11:11 AM, Brent Taylor  wrote:
>>>
 --- a/drivers/net/wireless/ath/ath6kl/init.c
 +++ b/drivers/net/wireless/ath/ath6kl/init.c
 @@ -673,10 +673,15 @@ static int ath6kl_get_fw(struct ath6kl *ar, const 
 char *filename,
 return ret;

 *fw_len = fw_entry->size;
 -   *fw = kmemdup(fw_entry->data, fw_entry->size, GFP_KERNEL);
 +   if (>fw == fw)
 +   *fw = vmalloc(fw_entry->size);
 +   else
 +   *fw = kmalloc(fw_entry->size, GFP_KERNEL)
>>>
>>>   Why vmalloc and kmalloc both are required? can't use either
>>> vmalloc or kmalloc?
>>
>> My original problem was that kmemdup (which uses kmalloc) could not
>> allocate enough memory
>
> If kmemdump ( which uses kmalloc) could not allocate memory then
> using kmalloc again can lead to same problem.
> I guess it will be correct to use
> *fw = vmalloc(fw_entry->size);
> Correct me if i am wrong.

That sounds best. But remember take into account DMA requirements, IIRC
you cannot DMA from vmalloc memory on all platforms.

-- 
Kalle Valo
--
To unsubscribe from this list: send the line "unsubscribe linux-wireless" in
the body of a message to majord...@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: [PATCH] ath6kl: Use vmalloc to allocate ar->fw for api1 method

[Souptick Joarder]

Hi Brent,

On Tue, Dec 1, 2015 at 11:11 AM, Brent Taylor  wrote:
> Since commit 8437754c83351d6213c1a47ff029c1126d6042a7, ar->fw is expected to 
> be pointing to memory allocated by vmalloc.  If the api1 method (via 
> ath6kl_fetch_fw_api1) is used to allocate memory for ar->fw, then kmemdup is 
> used.  This patch checks if the firmware being loaded is the 'fw' image, then 
> use vmalloc, otherwise use kmalloc.
>
> Signed-off-by: Brent Taylor 
> ---
>  drivers/net/wireless/ath/ath6kl/init.c | 7 ++-
>  1 file changed, 6 insertions(+), 1 deletion(-)
>
> diff --git a/drivers/net/wireless/ath/ath6kl/init.c 
> b/drivers/net/wireless/ath/ath6kl/init.c
> index 6ae0734..4f2b124d 100644
> --- a/drivers/net/wireless/ath/ath6kl/init.c
> +++ b/drivers/net/wireless/ath/ath6kl/init.c
> @@ -673,10 +673,15 @@ static int ath6kl_get_fw(struct ath6kl *ar, const char 
> *filename,
> return ret;
>
> *fw_len = fw_entry->size;
> -   *fw = kmemdup(fw_entry->data, fw_entry->size, GFP_KERNEL);
> +   if (>fw == fw)
> +   *fw = vmalloc(fw_entry->size);
> +   else
> +   *fw = kmalloc(fw_entry->size, GFP_KERNEL);

  Why vmalloc and kmalloc both are required? can't use either
vmalloc or kmalloc?
>
> if (*fw == NULL)
> ret = -ENOMEM;
> +   else
> +   memcpy(*fw, fw_entry->data, fw_entry->size);
>
> release_firmware(fw_entry);
>
> --
> 2.6.3
>
> --
> To unsubscribe from this list: send the line "unsubscribe linux-wireless" in
> the body of a message to majord...@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html

-Souptick
--
To unsubscribe from this list: send the line "unsubscribe linux-wireless" in
the body of a message to majord...@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: [PATCH] ath6kl: Use vmalloc to allocate ar->fw for api1 method

[Brent Taylor]

On Mon, Dec 21, 2015 at 1:23 PM, Souptick Joarder  wrote:
> Hi Brent,
>
> On Tue, Dec 1, 2015 at 11:11 AM, Brent Taylor  wrote:
>> Since commit 8437754c83351d6213c1a47ff029c1126d6042a7, ar->fw is expected to 
>> be pointing to memory allocated by vmalloc.  If the api1 method (via 
>> ath6kl_fetch_fw_api1) is used to allocate memory for ar->fw, then kmemdup is 
>> used.  This patch checks if the firmware being loaded is the 'fw' image, 
>> then use vmalloc, otherwise use kmalloc.
>>
>> Signed-off-by: Brent Taylor 
>> ---
>>  drivers/net/wireless/ath/ath6kl/init.c | 7 ++-
>>  1 file changed, 6 insertions(+), 1 deletion(-)
>>
>> diff --git a/drivers/net/wireless/ath/ath6kl/init.c 
>> b/drivers/net/wireless/ath/ath6kl/init.c
>> index 6ae0734..4f2b124d 100644
>> --- a/drivers/net/wireless/ath/ath6kl/init.c
>> +++ b/drivers/net/wireless/ath/ath6kl/init.c
>> @@ -673,10 +673,15 @@ static int ath6kl_get_fw(struct ath6kl *ar, const char 
>> *filename,
>> return ret;
>>
>> *fw_len = fw_entry->size;
>> -   *fw = kmemdup(fw_entry->data, fw_entry->size, GFP_KERNEL);
>> +   if (>fw == fw)
>> +   *fw = vmalloc(fw_entry->size);
>> +   else
>> +   *fw = kmalloc(fw_entry->size, GFP_KERNEL);
>
>   Why vmalloc and kmalloc both are required? can't use either
> vmalloc or kmalloc?

My original problem was that kmemdup (which uses kmalloc) could not
allocate enough memory
to hold the firmware that is placed into "ar->fw".  In the function
ath6kl_core_cleanup (in core.c),
the "ar->fw" pointer is the only one that uses vfree which was changed in commit
8437754c83351d6213c1a47ff029c1126d6042a7.

I was trying to change as little as possible and I wasn't sure if
there was a reason that any of the
other firmware items needed to be allocated with kmalloc or if they
could be changed to use vmalloc.
--
To unsubscribe from this list: send the line "unsubscribe linux-wireless" in
the body of a message to majord...@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: [PATCH] ath6kl: Use vmalloc to allocate ar->fw for api1 method

[Dan Kephart]

Hi Brent and Souptick,




On 12/21/15, 2:23 PM, "ath6kl on behalf of Souptick Joarder" 
 wrote:

>Hi Brent,
>
>On Tue, Dec 1, 2015 at 11:11 AM, Brent Taylor  wrote:
>> Since commit 8437754c83351d6213c1a47ff029c1126d6042a7, ar->fw is expected to 
>> be pointing to memory allocated by vmalloc.  If the api1 method (via 
>> ath6kl_fetch_fw_api1) is used to allocate memory for ar->fw, then kmemdup is 
>> used.  This patch checks if the firmware being loaded is the 'fw' image, 
>> then use vmalloc, otherwise use kmalloc.
>>
>> Signed-off-by: Brent Taylor 
>> ---
>>  drivers/net/wireless/ath/ath6kl/init.c | 7 ++-
>>  1 file changed, 6 insertions(+), 1 deletion(-)
>>
>> diff --git a/drivers/net/wireless/ath/ath6kl/init.c 
>> b/drivers/net/wireless/ath/ath6kl/init.c
>> index 6ae0734..4f2b124d 100644
>> --- a/drivers/net/wireless/ath/ath6kl/init.c
>> +++ b/drivers/net/wireless/ath/ath6kl/init.c
>> @@ -673,10 +673,15 @@ static int ath6kl_get_fw(struct ath6kl *ar, const char 
>> *filename,
>> return ret;
>>
>> *fw_len = fw_entry->size;
>> -   *fw = kmemdup(fw_entry->data, fw_entry->size, GFP_KERNEL);
>> +   if (>fw == fw)
>> +   *fw = vmalloc(fw_entry->size);
>> +   else
>> +   *fw = kmalloc(fw_entry->size, GFP_KERNEL);
>
>  Why vmalloc and kmalloc both are required? can't use either
>vmalloc or kmalloc?

My guess is the reason to use both vmalloc and kmalloc is that the firmware 
blob can be near 128KB.  I know our ar6003 firmwares approach that.  So vmalloc 
must have been chosen to avoid any issues if it was greater than 128KB.  So 
kmalloc is used for all the small firmware pieces (board data, otp.bin, etc) 
but vmalloc for the firmware itself.  

I personally fixed this issue for loading the firmware (but not board data, 
opt.bin, etc) in the api1 and testmode functions by have it call a new helper 
function:

static int ath6kl_get_fw_vm(struct ath6kl *ar, const char *filename,
 u8 **fw, size_t *fw_len)
{
const struct firmware *fw_entry;
int ret;

ret = request_firmware(_entry, filename, ar->dev);
if (ret)
return ret;

*fw_len = fw_entry->size;
*fw = vmalloc(*fw_len);

if (*fw == NULL)
ret = -ENOMEM;

memcpy(*fw, fw_entry->data, *fw_len);

release_firmware(fw_entry);

return ret;
}



>>
>> if (*fw == NULL)
>> ret = -ENOMEM;
>> +   else
>> +   memcpy(*fw, fw_entry->data, fw_entry->size);
>>
>> release_firmware(fw_entry);
>>
>> --
>> 2.6.3
>>
>> --
>> To unsubscribe from this list: send the line "unsubscribe linux-wireless" in
>> the body of a message to majord...@vger.kernel.org
>> More majordomo info at  http://vger.kernel.org/majordomo-info.html
>
>-Souptick
>
>___
>ath6kl mailing list
>ath...@lists.infradead.org
>http://lists.infradead.org/mailman/listinfo/ath6kl

- Dan Kephart
N�r��yb�X��ǧv�^�)޺{.n�+{��*ޕ�,�{ay�ʇڙ�,j��f���h���z��w���
���j:+v���w�j�mzZ+�ݢj"��!�i

Re: [PATCH] ath6kl: Use vmalloc to allocate ar->fw for api1 method

[Andy Shevchenko]

On Tue, Dec 1, 2015 at 7:41 AM, Brent Taylor  wrote:
> Since commit 8437754c83351d6213c1a47ff029c1126d6042a7, ar->fw is expected to 
> be pointing to memory allocated by vmalloc.  If the api1 method (via 
> ath6kl_fetch_fw_api1) is used to allocate memory for ar->fw, then kmemdup is 
> used.  This patch checks if the firmware being loaded is the 'fw' image, then 
> use vmalloc, otherwise use kmalloc.
>

I think I already told someone that this kind of fixes miss the
symmetric kvfree() calls. Please, fix.


> Signed-off-by: Brent Taylor 
> ---
>  drivers/net/wireless/ath/ath6kl/init.c | 7 ++-
>  1 file changed, 6 insertions(+), 1 deletion(-)
>
> diff --git a/drivers/net/wireless/ath/ath6kl/init.c 
> b/drivers/net/wireless/ath/ath6kl/init.c
> index 6ae0734..4f2b124d 100644
> --- a/drivers/net/wireless/ath/ath6kl/init.c
> +++ b/drivers/net/wireless/ath/ath6kl/init.c
> @@ -673,10 +673,15 @@ static int ath6kl_get_fw(struct ath6kl *ar, const char 
> *filename,
> return ret;
>
> *fw_len = fw_entry->size;
> -   *fw = kmemdup(fw_entry->data, fw_entry->size, GFP_KERNEL);
> +   if (>fw == fw)
> +   *fw = vmalloc(fw_entry->size);
> +   else
> +   *fw = kmalloc(fw_entry->size, GFP_KERNEL);
>
> if (*fw == NULL)
> ret = -ENOMEM;
> +   else
> +   memcpy(*fw, fw_entry->data, fw_entry->size);
>
> release_firmware(fw_entry);
>
> --
> 2.6.3
>
> --
> To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
> the body of a message to majord...@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
> Please read the FAQ at  http://www.tux.org/lkml/



-- 
With Best Regards,
Andy Shevchenko
--
To unsubscribe from this list: send the line "unsubscribe linux-wireless" in
the body of a message to majord...@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: [PATCH] ath6kl: Use vmalloc to allocate ar->fw for api1 method

[Sergei Shtylyov]

Hello.

On 12/1/2015 8:41 AM, Brent Taylor wrote:


Since commit 8437754c83351d6213c1a47ff029c1126d6042a7,


   The scripts/checkpatch.pl now enforces certain format of the commit citing.


ar->fw is expected to be pointing to memory allocated by vmalloc.  If the api1 
method (via ath6kl_fetch_fw_api1) is used to allocate memory for ar->fw, then 
kmemdup is used.  This patch checks if the firmware being loaded is the 'fw' image, 
then use vmalloc, otherwise use kmalloc.


   This script also checks for unwrapped changelogs. Please run your patches 
thru it before posting.



Signed-off-by: Brent Taylor 

[...]

MBR, Sergei

--
To unsubscribe from this list: send the line "unsubscribe linux-wireless" in
the body of a message to majord...@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html