[PATCH v4 15/19] fpga: dfl: convert is_feature_dev_detected() to use FIU type

Peter Colberg posted 19 patches 1 month ago
There is a newer version of this series
[PATCH v4 15/19] fpga: dfl: convert is_feature_dev_detected() to use FIU type
Posted by Peter Colberg 1 month ago
Use binfo->type instead of binfo->feature_dev to decide whether a
feature device was detected during feature parsing. A subsequent
commit will delay the allocation of the feature platform device
to feature_dev_register() and remove binfo->feature_dev.

This commit does not introduce any functional changes.

Signed-off-by: Peter Colberg <peter.colberg@intel.com>
Reviewed-by: Matthew Gerlach <matthew.gerlach@linux.intel.com>
Reviewed-by: Basheer Ahmed Muddebihal <basheer.ahmed.muddebihal@linux.intel.com>
---
Changes since v3:
- New patch extracted from last patch of v3 series.
---
 drivers/fpga/dfl.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/fpga/dfl.c b/drivers/fpga/dfl.c
index 758673b0290a..a9ec37278b2d 100644
--- a/drivers/fpga/dfl.c
+++ b/drivers/fpga/dfl.c
@@ -1248,7 +1248,7 @@ static int parse_feature_port_afu(struct build_feature_devs_info *binfo,
 	return create_feature_instance(binfo, ofst, size, FEATURE_ID_AFU);
 }
 
-#define is_feature_dev_detected(binfo) (!!(binfo)->feature_dev)
+#define is_feature_dev_detected(binfo) ((binfo)->type != DFL_ID_MAX)
 
 static int parse_feature_afu(struct build_feature_devs_info *binfo,
 			     resource_size_t ofst)
-- 
2.47.0
Re: [PATCH v4 15/19] fpga: dfl: convert is_feature_dev_detected() to use FIU type
Posted by Xu Yilun 6 days, 22 hours ago
On Fri, Oct 25, 2024 at 06:37:10PM -0400, Peter Colberg wrote:
> Use binfo->type instead of binfo->feature_dev to decide whether a
> feature device was detected during feature parsing. A subsequent
> commit will delay the allocation of the feature platform device
> to feature_dev_register() and remove binfo->feature_dev.
> 
> This commit does not introduce any functional changes.
> 
> Signed-off-by: Peter Colberg <peter.colberg@intel.com>
> Reviewed-by: Matthew Gerlach <matthew.gerlach@linux.intel.com>
> Reviewed-by: Basheer Ahmed Muddebihal <basheer.ahmed.muddebihal@linux.intel.com>
> ---
> Changes since v3:
> - New patch extracted from last patch of v3 series.
> ---
>  drivers/fpga/dfl.c | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/drivers/fpga/dfl.c b/drivers/fpga/dfl.c
> index 758673b0290a..a9ec37278b2d 100644
> --- a/drivers/fpga/dfl.c
> +++ b/drivers/fpga/dfl.c
> @@ -1248,7 +1248,7 @@ static int parse_feature_port_afu(struct build_feature_devs_info *binfo,
>  	return create_feature_instance(binfo, ofst, size, FEATURE_ID_AFU);
>  }
>  
> -#define is_feature_dev_detected(binfo) (!!(binfo)->feature_dev)
> +#define is_feature_dev_detected(binfo) ((binfo)->type != DFL_ID_MAX)

I still doesn't get why put the change here. How it resolves my concern
compared to v3?

Thanks,
Yilun

>  
>  static int parse_feature_afu(struct build_feature_devs_info *binfo,
>  			     resource_size_t ofst)
> -- 
> 2.47.0
> 
>
Re: [PATCH v4 15/19] fpga: dfl: convert is_feature_dev_detected() to use FIU type
Posted by Colberg, Peter 6 days, 14 hours ago
On Mon, 2024-11-18 at 22:32 +0800, Xu Yilun wrote:
> On Fri, Oct 25, 2024 at 06:37:10PM -0400, Peter Colberg wrote:
> > Use binfo->type instead of binfo->feature_dev to decide whether a
> > feature device was detected during feature parsing. A subsequent
> > commit will delay the allocation of the feature platform device
> > to feature_dev_register() and remove binfo->feature_dev.
> > 
> > This commit does not introduce any functional changes.
> > 
> > Signed-off-by: Peter Colberg <peter.colberg@intel.com>
> > Reviewed-by: Matthew Gerlach <matthew.gerlach@linux.intel.com>
> > Reviewed-by: Basheer Ahmed Muddebihal <basheer.ahmed.muddebihal@linux.intel.com>
> > ---
> > Changes since v3:
> > - New patch extracted from last patch of v3 series.
> > ---
> >  drivers/fpga/dfl.c | 2 +-
> >  1 file changed, 1 insertion(+), 1 deletion(-)
> > 
> > diff --git a/drivers/fpga/dfl.c b/drivers/fpga/dfl.c
> > index 758673b0290a..a9ec37278b2d 100644
> > --- a/drivers/fpga/dfl.c
> > +++ b/drivers/fpga/dfl.c
> > @@ -1248,7 +1248,7 @@ static int parse_feature_port_afu(struct build_feature_devs_info *binfo,
> >  return create_feature_instance(binfo, ofst, size, FEATURE_ID_AFU);
> >  }
> >  
> > -#define is_feature_dev_detected(binfo) (!!(binfo)->feature_dev)
> > +#define is_feature_dev_detected(binfo) ((binfo)->type != DFL_ID_MAX)
> 
> I still doesn't get why put the change here. How it resolves my concern
> compared to v3?

Could you elaborate on your concern? I moved this change into a
separate commit so that it is not lost in other changes, but I don't
see how the two definitions would not be functionally equivalent. Would
it help to extend the commit description along the following lines?

   1. Before this series, binfo->feature_dev was initialized to NULL by
      devm_kzalloc() in dfl_fpga_feature_devs_enumerate(). After this
      series, binfo->type is initialized to DFL_ID_MAX in
      dfl_fpga_feature_devs_enumerate().
   2. Before this series, binfo->feature_dev was set to a non-NULL
      pointer in build_info_create_dev(), which in turn was called from
      parse_feature_fiu(). After this series, binfo->type is set to a
      non-DFL_ID_MAX value, as returned by dfh_id_to_type(), in
      parse_feature_fiu().
   3. Before this series, binfo->feature_dev was reset to NULL at the
      end of build_info_commit_dev(). After this series, binfo->type is
      reset to DFL_ID_MAX at the end of build_info_commit_dev().

Thanks,
Peter

> 
> Thanks,
> Yilun
> 
> >  
> >  static int parse_feature_afu(struct build_feature_devs_info *binfo,
> >  resource_size_t ofst)
> > -- 
> > 2.47.0
> > 
> > 

Re: [PATCH v4 15/19] fpga: dfl: convert is_feature_dev_detected() to use FIU type
Posted by Xu Yilun 6 days, 11 hours ago
On Mon, Nov 18, 2024 at 11:00:28PM +0000, Colberg, Peter wrote:
> On Mon, 2024-11-18 at 22:32 +0800, Xu Yilun wrote:
> > On Fri, Oct 25, 2024 at 06:37:10PM -0400, Peter Colberg wrote:
> > > Use binfo->type instead of binfo->feature_dev to decide whether a
> > > feature device was detected during feature parsing. A subsequent
> > > commit will delay the allocation of the feature platform device
> > > to feature_dev_register() and remove binfo->feature_dev.
> > > 
> > > This commit does not introduce any functional changes.
> > > 
> > > Signed-off-by: Peter Colberg <peter.colberg@intel.com>
> > > Reviewed-by: Matthew Gerlach <matthew.gerlach@linux.intel.com>
> > > Reviewed-by: Basheer Ahmed Muddebihal <basheer.ahmed.muddebihal@linux.intel.com>
> > > ---
> > > Changes since v3:
> > > - New patch extracted from last patch of v3 series.
> > > ---
> > >  drivers/fpga/dfl.c | 2 +-
> > >  1 file changed, 1 insertion(+), 1 deletion(-)
> > > 
> > > diff --git a/drivers/fpga/dfl.c b/drivers/fpga/dfl.c
> > > index 758673b0290a..a9ec37278b2d 100644
> > > --- a/drivers/fpga/dfl.c
> > > +++ b/drivers/fpga/dfl.c
> > > @@ -1248,7 +1248,7 @@ static int parse_feature_port_afu(struct build_feature_devs_info *binfo,
> > >  return create_feature_instance(binfo, ofst, size, FEATURE_ID_AFU);
> > >  }
> > >  
> > > -#define is_feature_dev_detected(binfo) (!!(binfo)->feature_dev)
> > > +#define is_feature_dev_detected(binfo) ((binfo)->type != DFL_ID_MAX)
> > 
> > I still doesn't get why put the change here. How it resolves my concern
> > compared to v3?
> 
> Could you elaborate on your concern? I moved this change into a
> separate commit so that it is not lost in other changes, but I don't

I did't mean this change had to be separated, I doubt it impacts the
functionality when it was applied. After the series were all applied the
issue may be fixed but people review patches one by one.

> see how the two definitions would not be functionally equivalent. Would
> it help to extend the commit description along the following lines?
> 
>    1. Before this series, binfo->feature_dev was initialized to NULL by
>       devm_kzalloc() in dfl_fpga_feature_devs_enumerate(). After this
>       series, binfo->type is initialized to DFL_ID_MAX in
>       dfl_fpga_feature_devs_enumerate().
>    2. Before this series, binfo->feature_dev was set to a non-NULL
>       pointer in build_info_create_dev(), which in turn was called from

So at the point of *this patch* is applied, binfo->feature_dev & binfo->type
don't initialize at the same time, there is some gap the caller of
is_feature_dev_detected() would get a different result, e.g.
when build_info_create_dev() fails, binfo->type holds valid DFL_ID but
binfo->feature_dev is NULL.

>       parse_feature_fiu(). After this series, binfo->type is set to a
>       non-DFL_ID_MAX value, as returned by dfh_id_to_type(), in
>       parse_feature_fiu().
>    3. Before this series, binfo->feature_dev was reset to NULL at the
>       end of build_info_commit_dev(). After this series, binfo->type is
>       reset to DFL_ID_MAX at the end of build_info_commit_dev().
> 
> Thanks,
> Peter
> 
> > 
> > Thanks,
> > Yilun
> > 
> > >  
> > >  static int parse_feature_afu(struct build_feature_devs_info *binfo,
> > >  resource_size_t ofst)
> > > -- 
> > > 2.47.0
> > > 
> > > 
>
Re: [PATCH v4 15/19] fpga: dfl: convert is_feature_dev_detected() to use FIU type
Posted by Colberg, Peter 5 days, 13 hours ago
On Tue, 2024-11-19 at 09:19 +0800, Xu Yilun wrote:
> On Mon, Nov 18, 2024 at 11:00:28PM +0000, Colberg, Peter wrote:
> > On Mon, 2024-11-18 at 22:32 +0800, Xu Yilun wrote:
> > > On Fri, Oct 25, 2024 at 06:37:10PM -0400, Peter Colberg wrote:
> > > > Use binfo->type instead of binfo->feature_dev to decide whether a
> > > > feature device was detected during feature parsing. A subsequent
> > > > commit will delay the allocation of the feature platform device
> > > > to feature_dev_register() and remove binfo->feature_dev.
> > > > 
> > > > This commit does not introduce any functional changes.
> > > > 
> > > > Signed-off-by: Peter Colberg <peter.colberg@intel.com>
> > > > Reviewed-by: Matthew Gerlach <matthew.gerlach@linux.intel.com>
> > > > Reviewed-by: Basheer Ahmed Muddebihal <basheer.ahmed.muddebihal@linux.intel.com>
> > > > ---
> > > > Changes since v3:
> > > > - New patch extracted from last patch of v3 series.
> > > > ---
> > > >  drivers/fpga/dfl.c | 2 +-
> > > >  1 file changed, 1 insertion(+), 1 deletion(-)
> > > > 
> > > > diff --git a/drivers/fpga/dfl.c b/drivers/fpga/dfl.c
> > > > index 758673b0290a..a9ec37278b2d 100644
> > > > --- a/drivers/fpga/dfl.c
> > > > +++ b/drivers/fpga/dfl.c
> > > > @@ -1248,7 +1248,7 @@ static int parse_feature_port_afu(struct build_feature_devs_info *binfo,
> > > >  return create_feature_instance(binfo, ofst, size, FEATURE_ID_AFU);
> > > >  }
> > > >  
> > > > -#define is_feature_dev_detected(binfo) (!!(binfo)->feature_dev)
> > > > +#define is_feature_dev_detected(binfo) ((binfo)->type != DFL_ID_MAX)
> > > 
> > > I still doesn't get why put the change here. How it resolves my concern
> > > compared to v3?
> > 
> > Could you elaborate on your concern? I moved this change into a
> > separate commit so that it is not lost in other changes, but I don't
> 
> I did't mean this change had to be separated, I doubt it impacts the
> functionality when it was applied. After the series were all applied the
> issue may be fixed but people review patches one by one.
> 
> > see how the two definitions would not be functionally equivalent. Would
> > it help to extend the commit description along the following lines?
> > 
> >    1. Before this series, binfo->feature_dev was initialized to NULL by
> >       devm_kzalloc() in dfl_fpga_feature_devs_enumerate(). After this
> >       series, binfo->type is initialized to DFL_ID_MAX in
> >       dfl_fpga_feature_devs_enumerate().
> >    2. Before this series, binfo->feature_dev was set to a non-NULL
> >       pointer in build_info_create_dev(), which in turn was called from
> 
> So at the point of *this patch* is applied, binfo->feature_dev & binfo->type
> don't initialize at the same time, there is some gap the caller of
> is_feature_dev_detected() would get a different result, e.g.
> when build_info_create_dev() fails, binfo->type holds valid DFL_ID but
> binfo->feature_dev is NULL.

Thank you for the clarification. This and the subsequent patch have
been merged in [PATCH v5 15/18], to avoid a window between binfo->type
already holding a valid DFL id but binfo->feature_dev still NULL.

Peter

> 
> >       parse_feature_fiu(). After this series, binfo->type is set to a
> >       non-DFL_ID_MAX value, as returned by dfh_id_to_type(), in
> >       parse_feature_fiu().
> >    3. Before this series, binfo->feature_dev was reset to NULL at the
> >       end of build_info_commit_dev(). After this series, binfo->type is
> >       reset to DFL_ID_MAX at the end of build_info_commit_dev().
> > 
> > Thanks,
> > Peter
> > 
> > > 
> > > Thanks,
> > > Yilun
> > > 
> > > >  
> > > >  static int parse_feature_afu(struct build_feature_devs_info *binfo,
> > > >  resource_size_t ofst)
> > > > -- 
> > > > 2.47.0
> > > > 
> > > > 
> >