[PATCH v2] coccinelle: device_attr_show: Adapt to the latest Documentation/filesystems/sysfs.rst

[PATCH v2] coccinelle: device_attr_show: Adapt to the latest Documentation/filesystems/sysfs.rst

[Li Zhijian]

Adapt description, warning message and MODE=patch according to the latest
Documentation/filesystems/sysfs.rst:
> show() should only use sysfs_emit() or sysfs_emit_at() when formatting
> the value to be returned to user space.

After this patch:
When MODE=report,
 $ make coccicheck COCCI=scripts/coccinelle/api/device_attr_show.cocci M=drivers/hid/hid-picolcd_core.c MODE=report
 <...snip...>
 drivers/hid/hid-picolcd_core.c:304:8-16: WARNING: please use sysfs_emit or sysfs_emit_at
 drivers/hid/hid-picolcd_core.c:259:9-17: WARNING: please use sysfs_emit or sysfs_emit_at

When MODE=patch,
 $ make coccicheck COCCI=scripts/coccinelle/api/device_attr_show.cocci M=drivers/hid/hid-picolcd_core.c MODE=patch
 <...snip...>
 diff -u -p a/drivers/hid/hid-picolcd_core.c b/drivers/hid/hid-picolcd_core.c
 --- a/drivers/hid/hid-picolcd_core.c
 +++ b/drivers/hid/hid-picolcd_core.c
 @@ -255,10 +255,12 @@ static ssize_t picolcd_operation_mode_sh
  {
         struct picolcd_data *data = dev_get_drvdata(dev);

 -       if (data->status & PICOLCD_BOOTLOADER)
 -               return snprintf(buf, PAGE_SIZE, "[bootloader] lcd\n");
 -       else
 -               return snprintf(buf, PAGE_SIZE, "bootloader [lcd]\n");
 +       if (data->status & PICOLCD_BOOTLOADER) {
 +               return sysfs_emit(buf, "[bootloader] lcd\n");
 +       }
 +       else {
 +               return sysfs_emit(buf, "bootloader [lcd]\n");
 +       }
  }

  static ssize_t picolcd_operation_mode_store(struct device *dev,
 @@ -301,7 +303,7 @@ static ssize_t picolcd_operation_mode_de
  {
         struct picolcd_data *data = dev_get_drvdata(dev);

 -       return snprintf(buf, PAGE_SIZE, "hello world\n");
 +       return sysfs_emit(buf, "hello world\n");
  }

  static ssize_t picolcd_operation_mode_delay_store(struct device *dev,

CC: Julia Lawall <Julia.Lawall@inria.fr>
CC: Nicolas Palix <nicolas.palix@imag.fr>
CC: cocci@inria.fr
Signed-off-by: Li Zhijian <lizhijian@fujitsu.com>
---
V2:
- changed title from coccinelle: device_attr_show.cocci: update description and warning message
- Fix MODE=patch
- Extract patch from the patch set[1] so that maintainer can accept it separately.
[1] https://lore.kernel.org/lkml/20240116041129.3937800-1-lizhijian@fujitsu.com/
---
 scripts/coccinelle/api/device_attr_show.cocci | 22 +++++++++++--------
 1 file changed, 13 insertions(+), 9 deletions(-)

diff --git a/scripts/coccinelle/api/device_attr_show.cocci b/scripts/coccinelle/api/device_attr_show.cocci
index a28dc061653a..634514937e63 100644
--- a/scripts/coccinelle/api/device_attr_show.cocci
+++ b/scripts/coccinelle/api/device_attr_show.cocci
@@ -1,10 +1,8 @@
 // SPDX-License-Identifier: GPL-2.0-only
 ///
 /// From Documentation/filesystems/sysfs.rst:
-///  show() must not use snprintf() when formatting the value to be
-///  returned to user space. If you can guarantee that an overflow
-///  will never happen you can use sprintf() otherwise you must use
-///  scnprintf().
+///  show() should only use sysfs_emit() or sysfs_emit_at() when formatting
+///  the value to be returned to user space.
 ///
 // Confidence: High
 // Copyright: (C) 2020 Denis Efremov ISPRAS
@@ -30,15 +28,21 @@ ssize_t show(struct device *dev, struct device_attribute *attr, char *buf)
 
 @rp depends on patch@
 identifier show, dev, attr, buf;
+expression BUF, SZ, FORMAT, STR;
 @@
 
 ssize_t show(struct device *dev, struct device_attribute *attr, char *buf)
 {
 	<...
+(
 	return
--		snprintf
-+		scnprintf
-			(...);
+-		snprintf(BUF, SZ, FORMAT, STR);
++		sysfs_emit(BUF, FORMAT, STR);
+|
+	return
+-		snprintf(BUF, SZ, STR);
++		sysfs_emit(BUF, STR);
+)
 	...>
 }
 
@@ -46,10 +50,10 @@ ssize_t show(struct device *dev, struct device_attribute *attr, char *buf)
 p << r.p;
 @@
 
-coccilib.report.print_report(p[0], "WARNING: use scnprintf or sprintf")
+coccilib.report.print_report(p[0], "WARNING: please use sysfs_emit or sysfs_emit_at")
 
 @script: python depends on org@
 p << r.p;
 @@
 
-coccilib.org.print_todo(p[0], "WARNING: use scnprintf or sprintf")
+coccilib.org.print_todo(p[0], "WARNING: please use sysfs_emit or sysfs_emit_at")
-- 
2.29.2

Re: [PATCH v2] coccinelle: device_attr_show: Adapt to the latest Documentation/filesystems/sysfs.rst

[Julia Lawall]

On Fri, 19 Jan 2024, Li Zhijian wrote:

> Adapt description, warning message and MODE=patch according to the latest
> Documentation/filesystems/sysfs.rst:
> > show() should only use sysfs_emit() or sysfs_emit_at() when formatting
> > the value to be returned to user space.
>
> After this patch:
> When MODE=report,
>  $ make coccicheck COCCI=scripts/coccinelle/api/device_attr_show.cocci M=drivers/hid/hid-picolcd_core.c MODE=report
>  <...snip...>
>  drivers/hid/hid-picolcd_core.c:304:8-16: WARNING: please use sysfs_emit or sysfs_emit_at
>  drivers/hid/hid-picolcd_core.c:259:9-17: WARNING: please use sysfs_emit or sysfs_emit_at
>
> When MODE=patch,
>  $ make coccicheck COCCI=scripts/coccinelle/api/device_attr_show.cocci M=drivers/hid/hid-picolcd_core.c MODE=patch
>  <...snip...>
>  diff -u -p a/drivers/hid/hid-picolcd_core.c b/drivers/hid/hid-picolcd_core.c
>  --- a/drivers/hid/hid-picolcd_core.c
>  +++ b/drivers/hid/hid-picolcd_core.c
>  @@ -255,10 +255,12 @@ static ssize_t picolcd_operation_mode_sh
>   {
>          struct picolcd_data *data = dev_get_drvdata(dev);
>
>  -       if (data->status & PICOLCD_BOOTLOADER)
>  -               return snprintf(buf, PAGE_SIZE, "[bootloader] lcd\n");
>  -       else
>  -               return snprintf(buf, PAGE_SIZE, "bootloader [lcd]\n");
>  +       if (data->status & PICOLCD_BOOTLOADER) {
>  +               return sysfs_emit(buf, "[bootloader] lcd\n");
>  +       }
>  +       else {
>  +               return sysfs_emit(buf, "bootloader [lcd]\n");
>  +       }
>   }
>
>   static ssize_t picolcd_operation_mode_store(struct device *dev,
>  @@ -301,7 +303,7 @@ static ssize_t picolcd_operation_mode_de
>   {
>          struct picolcd_data *data = dev_get_drvdata(dev);
>
>  -       return snprintf(buf, PAGE_SIZE, "hello world\n");
>  +       return sysfs_emit(buf, "hello world\n");
>   }
>
>   static ssize_t picolcd_operation_mode_delay_store(struct device *dev,
>
> CC: Julia Lawall <Julia.Lawall@inria.fr>
> CC: Nicolas Palix <nicolas.palix@imag.fr>
> CC: cocci@inria.fr
> Signed-off-by: Li Zhijian <lizhijian@fujitsu.com>
> ---
> V2:
> - changed title from coccinelle: device_attr_show.cocci: update description and warning message
> - Fix MODE=patch
> - Extract patch from the patch set[1] so that maintainer can accept it separately.

Applied.

I subsequently simplified the patch case to contain the following:

-              snprintf(BUF, SZ, FORMAT
+              sysfs_emit(BUF, FORMAT
                               ,...);

This also works for the case where there are only three arguments.
It has the benefit that the change is recognized as a line replacement, so
no extra {} are added when the call is in an if branch.

julia



> [1] https://lore.kernel.org/lkml/20240116041129.3937800-1-lizhijian@fujitsu.com/
> ---
>  scripts/coccinelle/api/device_attr_show.cocci | 22 +++++++++++--------
>  1 file changed, 13 insertions(+), 9 deletions(-)
>
> diff --git a/scripts/coccinelle/api/device_attr_show.cocci b/scripts/coccinelle/api/device_attr_show.cocci
> index a28dc061653a..634514937e63 100644
> --- a/scripts/coccinelle/api/device_attr_show.cocci
> +++ b/scripts/coccinelle/api/device_attr_show.cocci
> @@ -1,10 +1,8 @@
>  // SPDX-License-Identifier: GPL-2.0-only
>  ///
>  /// From Documentation/filesystems/sysfs.rst:
> -///  show() must not use snprintf() when formatting the value to be
> -///  returned to user space. If you can guarantee that an overflow
> -///  will never happen you can use sprintf() otherwise you must use
> -///  scnprintf().
> +///  show() should only use sysfs_emit() or sysfs_emit_at() when formatting
> +///  the value to be returned to user space.
>  ///
>  // Confidence: High
>  // Copyright: (C) 2020 Denis Efremov ISPRAS
> @@ -30,15 +28,21 @@ ssize_t show(struct device *dev, struct device_attribute *attr, char *buf)
>
>  @rp depends on patch@
>  identifier show, dev, attr, buf;
> +expression BUF, SZ, FORMAT, STR;
>  @@
>
>  ssize_t show(struct device *dev, struct device_attribute *attr, char *buf)
>  {
>  	<...
> +(
>  	return
> --		snprintf
> -+		scnprintf
> -			(...);
> +-		snprintf(BUF, SZ, FORMAT, STR);
> ++		sysfs_emit(BUF, FORMAT, STR);
> +|
> +	return
> +-		snprintf(BUF, SZ, STR);
> ++		sysfs_emit(BUF, STR);
> +)
>  	...>
>  }
>
> @@ -46,10 +50,10 @@ ssize_t show(struct device *dev, struct device_attribute *attr, char *buf)
>  p << r.p;
>  @@
>
> -coccilib.report.print_report(p[0], "WARNING: use scnprintf or sprintf")
> +coccilib.report.print_report(p[0], "WARNING: please use sysfs_emit or sysfs_emit_at")
>
>  @script: python depends on org@
>  p << r.p;
>  @@
>
> -coccilib.org.print_todo(p[0], "WARNING: use scnprintf or sprintf")
> +coccilib.org.print_todo(p[0], "WARNING: please use sysfs_emit or sysfs_emit_at")
> --
> 2.29.2
>
>

Re: [cocci] [v2] coccinelle: device_attr_show: Adapt to the latest Documentation/filesystems/sysfs.rst

[Markus Elfring]

> I subsequently simplified the patch case to contain the following:
>
> -              snprintf(BUF, SZ, FORMAT
> +              sysfs_emit(BUF, FORMAT
>                                ,...);
>
> This also works for the case where there are only three arguments.
> It has the benefit that the change is recognized as a line replacement, so
> no extra {} are added when the call is in an if branch.

Would you like to consider the application of the following SmPL code variant?

-snprintf
+sysfs_emit
         (BUF,
-         SZ,
          FORMAT,
          ...
         );

Regards,
Markus

Re: [cocci] [v2] coccinelle: device_attr_show: Adapt to the latest Documentation/filesystems/sysfs.rst

[Julia Lawall]

On Sun, 21 Jan 2024, Markus Elfring wrote:

> > I subsequently simplified the patch case to contain the following:
> >
> > -              snprintf(BUF, SZ, FORMAT
> > +              sysfs_emit(BUF, FORMAT
> >                                ,...);
> >
> > This also works for the case where there are only three arguments.
> > It has the benefit that the change is recognized as a line replacement, so
> > no extra {} are added when the call is in an if branch.
>
> Would you like to consider the application of the following SmPL code variant?
>
> -snprintf
> +sysfs_emit
>          (BUF,
> -         SZ,
>           FORMAT,
>           ...
>          );

It's ok too.

julia

>
> Regards,
> Markus
>

Re: [cocci] [v2] coccinelle: device_attr_show: Adapt to the latest Documentation/filesystems/sysfs.rst

[Markus Elfring]

>> Would you like to consider the application of the following SmPL code variant?
>>
>> -snprintf
>> +sysfs_emit
>>          (BUF,
>> -         SZ,
>>           FORMAT,
>>           ...
>>          );
>
> It's ok too.

Are you going to integrate such a transformation approach?

Regards,
Markus

Re: [cocci] [v2] coccinelle: device_attr_show: Adapt to the latest Documentation/filesystems/sysfs.rst

[Julia Lawall]

On Sun, 21 Jan 2024, Markus Elfring wrote:

> >> Would you like to consider the application of the following SmPL code variant?
> >>
> >> -snprintf
> >> +sysfs_emit
> >>          (BUF,
> >> -         SZ,
> >>           FORMAT,
> >>           ...
> >>          );
> >
> > It's ok too.
>
> Are you going to integrate such a transformation approach?

No.  I already pushed the commit for Linux v6.8, and I don't think the
change has any practical impact.

julia

Re: [cocci] [v2] coccinelle: device_attr_show: Adapt to the latest Documentation/filesystems/sysfs.rst

[Markus Elfring]

>>>> -snprintf
>>>> +sysfs_emit
>>>>          (BUF,
>>>> -         SZ,
>>>>           FORMAT,
>>>>           ...
>>>>          );
>>>
>>> It's ok too.
>>
>> Are you going to integrate such a transformation approach?
>
> No.

This is a pity.


> I already pushed the commit for Linux v6.8,

I hope then that the change acceptance for the “beautification”
and refinement of SmPL code can grow for subsequent versions.



> and I don't think the change has any practical impact.

I got an other development view for further fine-tuning possibilities.

Regards,
Markus

Re: [cocci] [v2] coccinelle: device_attr_show: Adapt to the latest Documentation/filesystems/sysfs.rst

[Markus Elfring]

>>> It's ok too.
>>
>> Are you going to integrate such a transformation approach?
>
> No.  I already pushed the commit for Linux v6.8,

I suggest to reconsider such a view together with the metavariable “STR”
which became obsolete in the discussed SmPL script.


> and I don't think the change has any practical impact.

Can you ever get into the mood to increase the granularity of
a source code change specification a bit more?

Regards,
Markus

Re: [PATCH v2] coccinelle: device_attr_show: Adapt to the latest Documentation/filesystems/sysfs.rst

[Zhijian Li (Fujitsu)]

On 21/01/2024 05:20, Julia Lawall wrote:
>> V2:
>> - changed title from coccinelle: device_attr_show.cocci: update description and warning message
>> - Fix MODE=patch
>> - Extract patch from the patch set[1] so that maintainer can accept it separately.
> Applied.
> 
> I subsequently simplified the patch case to contain the following:
> 
> -              snprintf(BUF, SZ, FORMAT
> +              sysfs_emit(BUF, FORMAT
>                                 ,...);
Great, that was what i want. But i didn't get the correct syntax. I wrote it wrongly:
- snprintf(BUF, SZ, FORMAT, ...);
+ sysfs_emit(BUF, SZ, FORMAT, ...);



> 
> This also works for the case where there are only three arguments.
> It has the benefit that the change is recognized as a line replacement, so
> no extra {} are added when the call is in an if branch.