Re: [PATCH mlx5-next 2/2] net/mlx5: Factor out HCA capabilities functions

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



On Mon, 2019-02-11 at 20:32 +0000, Jason Gunthorpe wrote:
> On Mon, Feb 11, 2019 at 10:02:07PM +0200, Leon Romanovsky wrote:
> > On Mon, Feb 11, 2019 at 07:50:55PM +0000, Jason Gunthorpe wrote:
> > > On Mon, Feb 11, 2019 at 01:56:08PM +0200, Leon Romanovsky wrote:
> > > > From: Leon Romanovsky <leonro@xxxxxxxxxxxx>
> > > > 
> > > > Combine all HCA capabilities setters under one function
> > > > and compile out the ODP related function in case kernel
> > > > was compiled without ODP support.
> > > > 
> > > > Signed-off-by: Leon Romanovsky <leonro@xxxxxxxxxxxx>
> > > >  .../net/ethernet/mellanox/mlx5/core/main.c    | 47
> > > > +++++++++++++------
> > > >  1 file changed, 33 insertions(+), 14 deletions(-)
> > > > 
> > > > diff --git a/drivers/net/ethernet/mellanox/mlx5/core/main.c
> > > > b/drivers/net/ethernet/mellanox/mlx5/core/main.c
> > > > index 6d45518edbdc..d7145ab6105d 100644
> > > > +++ b/drivers/net/ethernet/mellanox/mlx5/core/main.c
> > > > @@ -459,6 +459,7 @@ static int handle_hca_cap_atomic(struct
> > > > mlx5_core_dev *dev)
> > > >  	return err;
> > > >  }
> > > > 
> > > > +#ifdef CONFIG_INFINIBAND_ON_DEMAND_PAGING
> > > >  static int handle_hca_cap_odp(struct mlx5_core_dev *dev)
> > > >  {
> > > >  	void *set_hca_cap;
> > > > @@ -502,6 +503,7 @@ static int handle_hca_cap_odp(struct
> > > > mlx5_core_dev *dev)
> > > >  	kfree(set_ctx);
> > > >  	return err;
> > > >  }
> > > > +#endif
> > > > 
> > > >  static int handle_hca_cap(struct mlx5_core_dev *dev)
> > > >  {
> > > > @@ -576,6 +578,35 @@ static int handle_hca_cap(struct
> > > > mlx5_core_dev *dev)
> > > >  	return err;
> > > >  }
> > > > 
> > > > +static int set_hca_cap(struct mlx5_core_dev *dev)
> > > > +{
> > > > +	struct pci_dev *pdev = dev->pdev;
> > > > +	int err;
> > > > +
> > > > +	err = handle_hca_cap(dev);
> > > > +	if (err) {
> > > > +		dev_err(&pdev->dev, "handle_hca_cap failed\n");
> > > > +		goto out;
> > > > +	}
> > > > +
> > > > +	err = handle_hca_cap_atomic(dev);
> > > > +	if (err) {
> > > > +		dev_err(&pdev->dev, "handle_hca_cap_atomic
> > > > failed\n");
> > > > +		goto out;
> > > > +	}
> > > > +
> > > > +#ifdef CONFIG_INFINIBAND_ON_DEMAND_PAGING
> > > > +	err = handle_hca_cap_odp(dev);
> > > > +	if (err) {
> > > > +		dev_err(&pdev->dev, "handle_hca_cap_odp
> > > > failed\n");
> > > > +		goto out;
> > > > +	}
> > > > +#endif
> > > 
> > > Adding
> > >   if (IS_ENABLED..)
> > >     return 0;
> > > 
> > > To the top of handle_hca_cap_odp is alot better.
> > 
> > Saeed gave comment that he prefers code to be compiled-out in case
> > config is not set. In you suggestion, the code will exist and only
> > with some optimizations enabled it will be thrown.
> 
> the kernel always compiles with optimizations that will throw away
> this code, it is a widely used pattern.
> 
> #ifdef creates a compilation test matrix that is very undesired.
> 

It is more about code separation and readability rather than compiler
optimizations, optimally i want to have a file(s) that will include all
IB stuff and compile out  directly from Makefile, of course we still
need stub functions in a header file which will yield the same
compilation test matrix, but code readability and separation will
become more clear.

-Saeed.

> Jason




[Index of Archives]     [Linux USB Devel]     [Video for Linux]     [Linux Audio Users]     [Photo]     [Yosemite News]     [Yosemite Photos]     [Linux Kernel]     [Linux SCSI]     [XFree86]

  Powered by Linux