crypto: qat - use strscpy_pad to simplify buffer initialization

[PATCH] crypto: qat - use strscpy_pad to simplify buffer initialization

Posted by Thorsten Blum 3 months, 2 weeks ago

Use strscpy_pad() to copy the string and zero-pad the destination buffer
in a single step instead of zero-initializing the buffer first and then
immediately overwriting it using strscpy().

Replace the magic number 16 with sizeof(buf) and remove the redundant
parentheses around kstrtoul() while we're at it.

Signed-off-by: Thorsten Blum <thorsten.blum@linux.dev>
---
 drivers/crypto/intel/qat/qat_common/qat_uclo.c | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)

diff --git a/drivers/crypto/intel/qat/qat_common/qat_uclo.c b/drivers/crypto/intel/qat/qat_common/qat_uclo.c
index 18c3e4416dc5..41a7bd434e97 100644
--- a/drivers/crypto/intel/qat/qat_common/qat_uclo.c
+++ b/drivers/crypto/intel/qat/qat_common/qat_uclo.c
@@ -200,18 +200,18 @@ qat_uclo_cleanup_batch_init_list(struct icp_qat_fw_loader_handle *handle,
 
 static int qat_uclo_parse_num(char *str, unsigned int *num)
 {
-	char buf[16] = {0};
+	char buf[16] = {};
 	unsigned long ae = 0;
 	int i;
 
-	strscpy(buf, str, sizeof(buf));
-	for (i = 0; i < 16; i++) {
+	strscpy_pad(buf, str);
+	for (i = 0; i < sizeof(buf); i++) {
 		if (!isdigit(buf[i])) {
 			buf[i] = '\0';
 			break;
 		}
 	}
-	if ((kstrtoul(buf, 10, &ae)))
+	if (kstrtoul(buf, 10, &ae))
 		return -EFAULT;
 
 	*num = (unsigned int)ae;
-- 
2.51.0

Re: [PATCH] crypto: qat - use strscpy_pad to simplify buffer initialization

Posted by Andy Shevchenko 3 months, 2 weeks ago

On Wed, Oct 22, 2025 at 02:36:19PM +0200, Thorsten Blum wrote:
> Use strscpy_pad() to copy the string and zero-pad the destination buffer
> in a single step instead of zero-initializing the buffer first and then
> immediately overwriting it using strscpy().
> 
> Replace the magic number 16 with sizeof(buf) and remove the redundant
> parentheses around kstrtoul() while we're at it.

I understand that you focused on strscpy*() conversions, but the below I think
needs a bigger refactoring, see my remarks.

...

> -	char buf[16] = {0};
> +	char buf[16] = {};
>  	unsigned long ae = 0;
>  	int i;
>  
> -	strscpy(buf, str, sizeof(buf));
> -	for (i = 0; i < 16; i++) {
> +	strscpy_pad(buf, str);

First of all, why do we need a _pad() version here? Is the data somehow being
used as a whole?

> +	for (i = 0; i < sizeof(buf); i++) {
>  		if (!isdigit(buf[i])) {
>  			buf[i] = '\0';
>  			break;
>  		}
>  	}
> -	if ((kstrtoul(buf, 10, &ae)))
> +	if (kstrtoul(buf, 10, &ae))
>  		return -EFAULT;

Looking at this, it tries to work around the kstrtoul() inability to perform
partial parses. Instead, this should do something like

	unsigned long long x;
	const char *end;

	simple_strtoull(...);
	if (x > UINT_MAX || end == buf)
		return $ERR; // wrong input / overflow

-- 
With Best Regards,
Andy Shevchenko

Re: [PATCH] crypto: qat - use strscpy_pad to simplify buffer initialization

Posted by Thorsten Blum 3 months, 2 weeks ago

On 22. Oct 2025, at 20:17, Andy Shevchenko wrote:
> On Wed, Oct 22, 2025 at 02:36:19PM +0200, Thorsten Blum wrote:
>> Use strscpy_pad() to copy the string and zero-pad the destination buffer
>> in a single step instead of zero-initializing the buffer first and then
>> immediately overwriting it using strscpy().
>> 
>> Replace the magic number 16 with sizeof(buf) and remove the redundant
>> parentheses around kstrtoul() while we're at it.
> 
> I understand that you focused on strscpy*() conversions, but the below I think
> needs a bigger refactoring, see my remarks.
> 
> ...
> 
>> -	char buf[16] = {0};
>> +	char buf[16] = {};

Sorry, this should have been just 'char buf[16];' since {} and {0} are
equivalent and both zero-initialize the array.

>> 	unsigned long ae = 0;
>> 	int i;
>> 
>> -	strscpy(buf, str, sizeof(buf));
>> -	for (i = 0; i < 16; i++) {
>> +	strscpy_pad(buf, str);
> 
> First of all, why do we need a _pad() version here? Is the data somehow being
> used as a whole?

I honestly didn't question this, but it looks like strscpy() would be
sufficient (with this approach at least).

>> +	for (i = 0; i < sizeof(buf); i++) {
>> 		if (!isdigit(buf[i])) {
>> 			buf[i] = '\0';
>> 			break;
>> 		}
>> 	}
>> -	if ((kstrtoul(buf, 10, &ae)))
>> +	if (kstrtoul(buf, 10, &ae))
>> 		return -EFAULT;
> 
> Looking at this, it tries to work around the kstrtoul() inability to perform
> partial parses. Instead, this should do something like
> 
> 	unsigned long long x;
> 	const char *end;
> 
> 	simple_strtoull(...);
> 	if (x > UINT_MAX || end == buf)
> 		return $ERR; // wrong input / overflow

How about this?

diff --git a/drivers/crypto/intel/qat/qat_common/qat_uclo.c b/drivers/crypto/intel/qat/qat_common/qat_uclo.c
index 18c3e4416dc5..04628dc01456 100644
--- a/drivers/crypto/intel/qat/qat_common/qat_uclo.c
+++ b/drivers/crypto/intel/qat/qat_common/qat_uclo.c
@@ -200,20 +200,12 @@ qat_uclo_cleanup_batch_init_list(struct icp_qat_fw_loader_handle *handle,
 
 static int qat_uclo_parse_num(char *str, unsigned int *num)
 {
-	char buf[16] = {0};
-	unsigned long ae = 0;
-	int i;
-
-	strscpy(buf, str, sizeof(buf));
-	for (i = 0; i < 16; i++) {
-		if (!isdigit(buf[i])) {
-			buf[i] = '\0';
-			break;
-		}
-	}
-	if ((kstrtoul(buf, 10, &ae)))
-		return -EFAULT;
+	unsigned long long ae;
+	char *end;
 
+	ae = simple_strtoull(str, &end, 10);
+	if (ae > UINT_MAX || str == end || (end - str) > 20)
+		return -EINVAL;
 	*num = (unsigned int)ae;
 	return 0;
 }

Re: [PATCH] crypto: qat - use strscpy_pad to simplify buffer initialization

Posted by Andy Shevchenko 3 months, 2 weeks ago

On Thu, Oct 23, 2025 at 05:35:00PM +0200, Thorsten Blum wrote:
> On 22. Oct 2025, at 20:17, Andy Shevchenko wrote:
> > On Wed, Oct 22, 2025 at 02:36:19PM +0200, Thorsten Blum wrote:

...

> How about this?

LGTM, and that's what I had in mind, but please double check the behaviour of
kstrtox() on an empty strings.

> diff --git a/drivers/crypto/intel/qat/qat_common/qat_uclo.c b/drivers/crypto/intel/qat/qat_common/qat_uclo.c
> index 18c3e4416dc5..04628dc01456 100644
> --- a/drivers/crypto/intel/qat/qat_common/qat_uclo.c
> +++ b/drivers/crypto/intel/qat/qat_common/qat_uclo.c
> @@ -200,20 +200,12 @@ qat_uclo_cleanup_batch_init_list(struct icp_qat_fw_loader_handle *handle,
>  
>  static int qat_uclo_parse_num(char *str, unsigned int *num)
>  {
> -	char buf[16] = {0};
> -	unsigned long ae = 0;
> -	int i;
> -
> -	strscpy(buf, str, sizeof(buf));
> -	for (i = 0; i < 16; i++) {
> -		if (!isdigit(buf[i])) {
> -			buf[i] = '\0';
> -			break;
> -		}
> -	}
> -	if ((kstrtoul(buf, 10, &ae)))
> -		return -EFAULT;
> +	unsigned long long ae;
> +	char *end;
>  
> +	ae = simple_strtoull(str, &end, 10);
> +	if (ae > UINT_MAX || str == end || (end - str) > 20)

I would go with >= 20, the 64-bit value is approx. 1 * 10^19.

> +		return -EINVAL;
>  	*num = (unsigned int)ae;
>  	return 0;
>  }

-- 
With Best Regards,
Andy Shevchenko

Re: [PATCH] crypto: qat - use strscpy_pad to simplify buffer initialization

Posted by Thorsten Blum 3 months, 2 weeks ago

Hi Andy,

On 23. Oct 2025, at 20:44, Andy Shevchenko wrote:
> On Thu, Oct 23, 2025 at 05:35:00PM +0200, Thorsten Blum wrote:
>> On 22. Oct 2025, at 20:17, Andy Shevchenko wrote:
>>> On Wed, Oct 22, 2025 at 02:36:19PM +0200, Thorsten Blum wrote:
> 
> ...
> 
>> How about this?
> 
> LGTM, and that's what I had in mind, ...

I was about to submit v2, but checkpatch warns me about simple_strtoull:

  WARNING: simple_strtoull is obsolete, use kstrtoull instead

Any recommendations?

Thanks,
Thorsten

Re: [PATCH] crypto: qat - use strscpy_pad to simplify buffer initialization

Posted by Andy Shevchenko 3 months, 1 week ago

On Fri, Oct 24, 2025 at 08:47:02PM +0200, Thorsten Blum wrote:
> On 23. Oct 2025, at 20:44, Andy Shevchenko wrote:
> > On Thu, Oct 23, 2025 at 05:35:00PM +0200, Thorsten Blum wrote:
> >> On 22. Oct 2025, at 20:17, Andy Shevchenko wrote:
> >>> On Wed, Oct 22, 2025 at 02:36:19PM +0200, Thorsten Blum wrote:

...

> >> How about this?
> > 
> > LGTM, and that's what I had in mind, ...
> 
> I was about to submit v2, but checkpatch warns me about simple_strtoull:
> 
>   WARNING: simple_strtoull is obsolete, use kstrtoull instead
> 
> Any recommendations?

Yes, fix checkpatch as per commit 885e68e8b7b1 ("kernel.h: update comment about
simple_strto<foo>() functions") or ignore this false positive.

-- 
With Best Regards,
Andy Shevchenko

Re: [PATCH] crypto: qat - use strscpy_pad to simplify buffer initialization

Posted by Thorsten Blum 3 months, 1 week ago

On 27. Oct 2025, at 09:19, Andy Shevchenko wrote:
> On Fri, Oct 24, 2025 at 08:47:02PM +0200, Thorsten Blum wrote:
>> [...]
>> Any recommendations?
> 
> Yes, fix checkpatch as per commit 885e68e8b7b1 ("kernel.h: update comment about
> simple_strto<foo>() functions") or ignore this false positive.

Thanks for the commit hash. I ignored it and submitted v2:

https://lore.kernel.org/lkml/20251026015710.1368-1-thorsten.blum@linux.dev/

Re: [PATCH] crypto: qat - use strscpy_pad to simplify buffer initialization

Posted by Giovanni Cabiddu 3 months, 2 weeks ago

On Thu, Oct 23, 2025 at 09:44:01PM +0300, Andy Shevchenko wrote:
> On Thu, Oct 23, 2025 at 05:35:00PM +0200, Thorsten Blum wrote:
> > On 22. Oct 2025, at 20:17, Andy Shevchenko wrote:
> > > On Wed, Oct 22, 2025 at 02:36:19PM +0200, Thorsten Blum wrote:
> 
> ...
> 
> > How about this?
> 
> LGTM, and that's what I had in mind, but please double check the behaviour of
> kstrtox() on an empty strings.

LGTM as well.

I checked the behaviour of kstrtoul() when given an empty string. It
returns -EINVAL (-22). The result variable (the third parameter) is only
modified if the conversion is succesful.

Anyway, the caller will not provide a NULL string to this function [1].

> > diff --git a/drivers/crypto/intel/qat/qat_common/qat_uclo.c b/drivers/crypto/intel/qat/qat_common/qat_uclo.c
> > index 18c3e4416dc5..04628dc01456 100644
> > --- a/drivers/crypto/intel/qat/qat_common/qat_uclo.c
> > +++ b/drivers/crypto/intel/qat/qat_common/qat_uclo.c
> > @@ -200,20 +200,12 @@ qat_uclo_cleanup_batch_init_list(struct icp_qat_fw_loader_handle *handle,
> >  
> >  static int qat_uclo_parse_num(char *str, unsigned int *num)
> >  {
> > -	char buf[16] = {0};
> > -	unsigned long ae = 0;
> > -	int i;
> > -
> > -	strscpy(buf, str, sizeof(buf));
> > -	for (i = 0; i < 16; i++) {
> > -		if (!isdigit(buf[i])) {
> > -			buf[i] = '\0';
> > -			break;
> > -		}
> > -	}
> > -	if ((kstrtoul(buf, 10, &ae)))
> > -		return -EFAULT;
> > +	unsigned long long ae;
> > +	char *end;
> >  
> > +	ae = simple_strtoull(str, &end, 10);
> > +	if (ae > UINT_MAX || str == end || (end - str) > 20)
> 
> I would go with >= 20, the 64-bit value is approx. 1 * 10^19.

Just an insight into the type of strings being parsed here. If they are
well-formed, the format looks like:

    <AE_NUMBER>!<...>

Example:

    11!l0000!lm_thread_ctrl_struct_base

This function just extract the first number. Currently, that is 2 digits [2],
and I believe it is unlikely to exceed 3 in future gens.

[1] https://elixir.bootlin.com/linux/v6.17.4/source/drivers/crypto/intel/qat/qat_common/qat_uclo.c#L237
[2] https://elixir.bootlin.com/linux/v6.17.4/source/drivers/crypto/intel/qat/qat_common/icp_qat_uclo.h#L11

Regards,

-- 
Giovanni

Re: [PATCH] crypto: qat - use strscpy_pad to simplify buffer initialization

Posted by Andy Shevchenko 3 months, 2 weeks ago

On Fri, Oct 24, 2025 at 09:50:11AM +0100, Giovanni Cabiddu wrote:
> On Thu, Oct 23, 2025 at 09:44:01PM +0300, Andy Shevchenko wrote:
> > On Thu, Oct 23, 2025 at 05:35:00PM +0200, Thorsten Blum wrote:
> > > On 22. Oct 2025, at 20:17, Andy Shevchenko wrote:
> > > > On Wed, Oct 22, 2025 at 02:36:19PM +0200, Thorsten Blum wrote:

...

> > > How about this?
> > 
> > LGTM, and that's what I had in mind, but please double check the behaviour of
> > kstrtox() on an empty strings.
> 
> LGTM as well.
> 
> I checked the behaviour of kstrtoul() when given an empty string. It
> returns -EINVAL (-22). The result variable (the third parameter) is only
> modified if the conversion is succesful.

Thanks for confirming!

> Anyway, the caller will not provide a NULL string to this function [1].
> 
> > > +	unsigned long long ae;
> > > +	char *end;
> > >  
> > > +	ae = simple_strtoull(str, &end, 10);
> > > +	if (ae > UINT_MAX || str == end || (end - str) > 20)
> > 
> > I would go with >= 20, the 64-bit value is approx. 1 * 10^19.
> 
> Just an insight into the type of strings being parsed here. If they are
> well-formed, the format looks like:
> 
>     <AE_NUMBER>!<...>
> 
> Example:
> 
>     11!l0000!lm_thread_ctrl_struct_base
> 
> This function just extract the first number. Currently, that is 2 digits [2],
> and I believe it is unlikely to exceed 3 in future gens.

Yep, but it is better to make code more robust (and esp. in case somebody
copies'n'pastes to somewhere else).

> [1] https://elixir.bootlin.com/linux/v6.17.4/source/drivers/crypto/intel/qat/qat_common/qat_uclo.c#L237
> [2] https://elixir.bootlin.com/linux/v6.17.4/source/drivers/crypto/intel/qat/qat_common/icp_qat_uclo.h#L11

-- 
With Best Regards,
Andy Shevchenko

Re: [PATCH] crypto: qat - use strscpy_pad to simplify buffer initialization

Posted by Andy Shevchenko 3 months, 2 weeks ago

On Wed, Oct 22, 2025 at 09:17:22PM +0300, Andy Shevchenko wrote:
> On Wed, Oct 22, 2025 at 02:36:19PM +0200, Thorsten Blum wrote:
> > Use strscpy_pad() to copy the string and zero-pad the destination buffer
> > in a single step instead of zero-initializing the buffer first and then
> > immediately overwriting it using strscpy().
> > 
> > Replace the magic number 16 with sizeof(buf) and remove the redundant
> > parentheses around kstrtoul() while we're at it.
> 
> I understand that you focused on strscpy*() conversions, but the below I think
> needs a bigger refactoring, see my remarks.

...

> > -	char buf[16] = {0};
> > +	char buf[16] = {};
> >  	unsigned long ae = 0;
> >  	int i;
> >  
> > -	strscpy(buf, str, sizeof(buf));
> > -	for (i = 0; i < 16; i++) {
> > +	strscpy_pad(buf, str);
> 
> First of all, why do we need a _pad() version here? Is the data somehow being
> used as a whole?
> 
> > +	for (i = 0; i < sizeof(buf); i++) {
> >  		if (!isdigit(buf[i])) {
> >  			buf[i] = '\0';
> >  			break;
> >  		}
> >  	}
> > -	if ((kstrtoul(buf, 10, &ae)))
> > +	if (kstrtoul(buf, 10, &ae))
> >  		return -EFAULT;

On top of that the function is called only from one place and returns different
error code, instead it would have returned what kstrtoul() gives...

> Looking at this, it tries to work around the kstrtoul() inability to perform
> partial parses. Instead, this should do something like
> 
> 	unsigned long long x;
> 	const char *end;
> 
> 	simple_strtoull(...);
> 	if (x > UINT_MAX || end == buf)
> 		return $ERR; // wrong input / overflow

Yeah, the overflow check here is not comprehensive, it won't catch the overflow
(wrap around) of 64-bit value. But we can add a check for the end not to be
farther than ~19 characters from the start, which would correspond the initial
copy of 16 characters.

-- 
With Best Regards,
Andy Shevchenko