| Message ID | 20211209174046.535229-6-miquel.raynal@bootlin.com (mailing list archive) |
|---|---|
| State | Superseded |
| Headers | show |
| Series | External ECC engines & Macronix support | expand |
On Thu, 9 Dec 2021 18:40:39 +0100 Miquel Raynal <miquel.raynal@bootlin.com> wrote: > The combination of checks against the number of supported operations is > going to increase exponentially each time we add a new parameter. So far > we only had a dtr parameter. Now we are introducing an ECC parameter. We > need to make this helper available for drivers with specific needs, > instead of creating another set of helpers each time we want to check > something new. In the future if we see that many different drivers use > the same parameter values, we might be tempted to create a specific > helper for that. But for now, let's just make the generic one available. > > Signed-off-by: Miquel Raynal <miquel.raynal@bootlin.com> > --- > drivers/spi/spi-mem.c | 7 ++++--- > include/linux/spi/spi-mem.h | 13 +++++++++++++ > 2 files changed, 17 insertions(+), 3 deletions(-) > > diff --git a/drivers/spi/spi-mem.c b/drivers/spi/spi-mem.c > index 9e06cd918273..48b55395178f 100644 > --- a/drivers/spi/spi-mem.c > +++ b/drivers/spi/spi-mem.c > @@ -160,9 +160,9 @@ static bool spi_mem_check_buswidth(struct spi_mem *mem, > return true; > } > > -static bool spi_mem_generic_supports_op(struct spi_mem *mem, > - const struct spi_mem_op *op, > - bool dtr, bool ecc) > +bool spi_mem_generic_supports_op(struct spi_mem *mem, > + const struct spi_mem_op *op, > + bool dtr, bool ecc) Looks like you're replacing a maintenance burden by another. Every time you'll add a new capability, you'll have to patch all the wrappers and the drivers using this generic function to pass the extra parameter. How about passing a struct spi_mem_controller_caps { bool dtr; bool ecc; }; or making the caps a bitmask, such that next time you add a new cap, you can just assume all drivers have it set to 0 by default. > { > if (!dtr) { > if (op->cmd.dtr || op->addr.dtr || > @@ -183,6 +183,7 @@ static bool spi_mem_generic_supports_op(struct spi_mem *mem, > > return spi_mem_check_buswidth(mem, op); > } > +EXPORT_SYMBOL_GPL(spi_mem_generic_supports_op); > > bool spi_mem_dtr_supports_op(struct spi_mem *mem, > const struct spi_mem_op *op) > diff --git a/include/linux/spi/spi-mem.h b/include/linux/spi/spi-mem.h > index 3be594be24c0..07f637cbe77d 100644 > --- a/include/linux/spi/spi-mem.h > +++ b/include/linux/spi/spi-mem.h > @@ -320,6 +320,10 @@ void spi_controller_dma_unmap_mem_op_data(struct spi_controller *ctlr, > const struct spi_mem_op *op, > struct sg_table *sg); > > +bool spi_mem_generic_supports_op(struct spi_mem *mem, > + const struct spi_mem_op *op, > + bool dtr, bool ecc); > + > bool spi_mem_default_supports_op(struct spi_mem *mem, > const struct spi_mem_op *op); > > @@ -327,6 +331,7 @@ bool spi_mem_dtr_supports_op(struct spi_mem *mem, > const struct spi_mem_op *op); > > #else > + > static inline int > spi_controller_dma_map_mem_op_data(struct spi_controller *ctlr, > const struct spi_mem_op *op, > @@ -342,6 +347,14 @@ spi_controller_dma_unmap_mem_op_data(struct spi_controller *ctlr, > { > } > > +static inline > +bool spi_mem_generic_supports_op(struct spi_mem *mem, > + const struct spi_mem_op *op, > + bool dtr, bool ecc) > +{ > + return false; > +} > + > static inline > bool spi_mem_default_supports_op(struct spi_mem *mem, > const struct spi_mem_op *op)
Hi Boris, boris.brezillon@collabora.com wrote on Fri, 10 Dec 2021 17:27:17 +0100: > On Thu, 9 Dec 2021 18:40:39 +0100 > Miquel Raynal <miquel.raynal@bootlin.com> wrote: > > > The combination of checks against the number of supported operations is > > going to increase exponentially each time we add a new parameter. So far > > we only had a dtr parameter. Now we are introducing an ECC parameter. We > > need to make this helper available for drivers with specific needs, > > instead of creating another set of helpers each time we want to check > > something new. In the future if we see that many different drivers use > > the same parameter values, we might be tempted to create a specific > > helper for that. But for now, let's just make the generic one available. > > > > Signed-off-by: Miquel Raynal <miquel.raynal@bootlin.com> > > --- > > drivers/spi/spi-mem.c | 7 ++++--- > > include/linux/spi/spi-mem.h | 13 +++++++++++++ > > 2 files changed, 17 insertions(+), 3 deletions(-) > > > > diff --git a/drivers/spi/spi-mem.c b/drivers/spi/spi-mem.c > > index 9e06cd918273..48b55395178f 100644 > > --- a/drivers/spi/spi-mem.c > > +++ b/drivers/spi/spi-mem.c > > @@ -160,9 +160,9 @@ static bool spi_mem_check_buswidth(struct spi_mem *mem, > > return true; > > } > > > > -static bool spi_mem_generic_supports_op(struct spi_mem *mem, > > - const struct spi_mem_op *op, > > - bool dtr, bool ecc) > > +bool spi_mem_generic_supports_op(struct spi_mem *mem, > > + const struct spi_mem_op *op, > > + bool dtr, bool ecc) > > Looks like you're replacing a maintenance burden by another. Every time > you'll add a new capability, you'll have to patch all the wrappers and > the drivers using this generic function to pass the extra parameter. > How about passing a > > struct spi_mem_controller_caps { > bool dtr; > bool ecc; > }; Oh yeah good idea, I like the structure more than the bitfield. > or making the caps a bitmask, such that next time you add a new cap, you > can just assume all drivers have it set to 0 by default. Thanks, Miquèl
diff --git a/drivers/spi/spi-mem.c b/drivers/spi/spi-mem.c index 9e06cd918273..48b55395178f 100644 --- a/drivers/spi/spi-mem.c +++ b/drivers/spi/spi-mem.c @@ -160,9 +160,9 @@ static bool spi_mem_check_buswidth(struct spi_mem *mem, return true; } -static bool spi_mem_generic_supports_op(struct spi_mem *mem, - const struct spi_mem_op *op, - bool dtr, bool ecc) +bool spi_mem_generic_supports_op(struct spi_mem *mem, + const struct spi_mem_op *op, + bool dtr, bool ecc) { if (!dtr) { if (op->cmd.dtr || op->addr.dtr || @@ -183,6 +183,7 @@ static bool spi_mem_generic_supports_op(struct spi_mem *mem, return spi_mem_check_buswidth(mem, op); } +EXPORT_SYMBOL_GPL(spi_mem_generic_supports_op); bool spi_mem_dtr_supports_op(struct spi_mem *mem, const struct spi_mem_op *op) diff --git a/include/linux/spi/spi-mem.h b/include/linux/spi/spi-mem.h index 3be594be24c0..07f637cbe77d 100644 --- a/include/linux/spi/spi-mem.h +++ b/include/linux/spi/spi-mem.h @@ -320,6 +320,10 @@ void spi_controller_dma_unmap_mem_op_data(struct spi_controller *ctlr, const struct spi_mem_op *op, struct sg_table *sg); +bool spi_mem_generic_supports_op(struct spi_mem *mem, + const struct spi_mem_op *op, + bool dtr, bool ecc); + bool spi_mem_default_supports_op(struct spi_mem *mem, const struct spi_mem_op *op); @@ -327,6 +331,7 @@ bool spi_mem_dtr_supports_op(struct spi_mem *mem, const struct spi_mem_op *op); #else + static inline int spi_controller_dma_map_mem_op_data(struct spi_controller *ctlr, const struct spi_mem_op *op, @@ -342,6 +347,14 @@ spi_controller_dma_unmap_mem_op_data(struct spi_controller *ctlr, { } +static inline +bool spi_mem_generic_supports_op(struct spi_mem *mem, + const struct spi_mem_op *op, + bool dtr, bool ecc) +{ + return false; +} + static inline bool spi_mem_default_supports_op(struct spi_mem *mem, const struct spi_mem_op *op)
The combination of checks against the number of supported operations is going to increase exponentially each time we add a new parameter. So far we only had a dtr parameter. Now we are introducing an ECC parameter. We need to make this helper available for drivers with specific needs, instead of creating another set of helpers each time we want to check something new. In the future if we see that many different drivers use the same parameter values, we might be tempted to create a specific helper for that. But for now, let's just make the generic one available. Signed-off-by: Miquel Raynal <miquel.raynal@bootlin.com> --- drivers/spi/spi-mem.c | 7 ++++--- include/linux/spi/spi-mem.h | 13 +++++++++++++ 2 files changed, 17 insertions(+), 3 deletions(-)