dm: spi: Return 0 if driver does not implement ops->cs_info
Subsystems (2)
PATCH v2 1/4 dm: spi: Return 0 if driver does not implement ops->cs_info
If an SPI controller driver does not implement ops->cs_info, that probably means any chip select number could be valid, hence let's return 0 for spi_cs_info().
---
Changes in v2:
- update spi-howto.rst to reflect the code changes
doc/driver-model/spi-howto.rst | 4 ++-- drivers/spi/spi-uclass.c | 7 +++---- 2 files changed, 5 insertions(+), 6 deletions(-)
doc/driver-model/spi-howto.rst +2 -2
@@ -634,8 +634,8 @@ method for cs_info() to deal with this. If you don't provide it, then the device tree will be used to determine what chip selects are valid. Return -ENODEV if the supplied chip select is invalid, or 0 if it is valid. -If you don't provide the cs_info() method, -ENODEV is assumed for all -chip selects that do not appear in the device tree. +If you don't provide the cs_info() method, 0 is assumed for all chip selects +that do not appear in the device tree. Test it
drivers/spi/spi-uclass.c +3 -4
@@ -237,11 +237,10 @@ int spi_cs_info(struct udevice *bus, uint cs, struct spi_cs_info *info)return ops->cs_info (bus, cs, info);/* -* We could assume there is at least one valid chip select, but best -* to be sure and return an error in this case. The driver didn't -* care enough to tell us. +* We could assume there is at least one valid chip select. +* The driver didn't care enough to tell us. */ -return -ENODEV ; +return 0 ; } int spi_find_bus_and_cs(int busnum, int cs, struct udevice **busp,
Signature · 1 line
2.7.4
PATCH v2 1/4 dm: spi: Return 0 if driver does not implement ops->cs_info
Hi Jagan,
Quoted 8 lines · Bin Meng
If an SPI controller driver does not implement
ops->cs_info, that probably means any chip select number could be valid, hence let's return 0 forspi_cs_info().Signed-off-by Bin Meng <[email protected]>Reviewed-by Jagan Teki <[email protected]>
Quoted 9 lines
---
Changes in v2:
- updatespi-howto.rstto reflect the code changesdoc/driver-model/spi-howto.rst | 4 ++-- drivers/spi/spi-uclass.c | 7 +++---- 2 files changed, 5 insertions(+), 6 deletions(-)
Ping for this series?
Regards,
Bin
PATCH v2 1/4 dm: spi: Return 0 if driver does not implement ops->cs_info
Quoted 17 lines · Bin Meng
Hi Jagan,
Quoted 13 lines · Bin Meng
If an SPI controller driver does not implement
ops->cs_info, that probably means any chip select number could be valid, hence let's return 0 forspi_cs_info().Signed-off-by Bin Meng <[email protected]>Reviewed-by Jagan Teki <[email protected]>---
Changes in v2:
- updatespi-howto.rstto reflect the code changes
Quoted 7 lines
Quoted 4 lines
doc/driver-model/spi-howto.rst | 4 ++-- drivers/spi/spi-uclass.c | 7 +++---- 2 files changed, 5 insertions(+), 6 deletions(-)Ping for this series?
Another ping?
PATCH v2 1/4 dm: spi: Return 0 if driver does not implement ops->cs_info
Quoted 24 lines · Bin Meng
Quoted 22 lines · Bin Meng
Hi Jagan,
Quoted 17 lines · Bin Meng
If an SPI controller driver does not implement
ops->cs_info, that probably means any chip select number could be valid, hence let's return 0 forspi_cs_info().Signed-off-by Bin Meng <[email protected]>Reviewed-by Jagan Teki <[email protected]>---
Changes in v2:
- updatespi-howto.rstto reflect the code changesdoc/driver-model/spi-howto.rst | 4 ++-- drivers/spi/spi-uclass.c | 7 +++---- 2 files changed, 5 insertions(+), 6 deletions(-)
Quoted 4 lines
Quoted 2 lines
Ping for this series?
Another ping?
Ping!
PATCH v2 1/4 dm: spi: Return 0 if driver does not implement ops->cs_info
In reply to Bin MengQuoted 17 lines · Bin Meng
Hi Jagan,
Quoted 13 lines · Bin Meng
If an SPI controller driver does not implement
ops->cs_info, that probably means any chip select number could be valid, hence let's return 0 forspi_cs_info().Signed-off-by Bin Meng <[email protected]>Reviewed-by Jagan Teki <[email protected]>---
Changes in v2:
- updatespi-howto.rstto reflect the code changes
Quoted 6 lines
Quoted 4 lines
doc/driver-model/spi-howto.rst | 4 ++-- drivers/spi/spi-uclass.c | 7 +++---- 2 files changed, 5 insertions(+), 6 deletions(-)Ping for this series?
Sorry, been in vacations. will try to give an update on this.
PATCH v2 2/4 dm: spi: Change cs_info op to return -EINVAL for invalid cs num
In reply to Bin MengWe need distinguish the following two situations in various SPI APIs:
- given chip select num is invalid
- given chip select num is valid, but no device is attached
Currently -ENODEV is returned for both cases.
For the first case, it's more reasonable to return -EINVAL instead of -ENODEV for invalid chip select numbers.
---
Changes in v2:
- new patch to change cs_info op to return -EINVAL for invalid cs num
doc/driver-model/spi-howto.rst | 4 ++-- drivers/spi/ath79_spi.c | 2 +- drivers/spi/bcm63xx_hsspi.c | 2 +- drivers/spi/bcm63xx_spi.c | 2 +- drivers/spi/sandbox_spi.c | 2 +- drivers/spi/tegra20_sflash.c | 2 +- include/spi.h | 2 +- 7 files changed, 8 insertions(+), 8 deletions(-)
doc/driver-model/spi-howto.rst +2 -2
@@ -116,7 +116,7 @@ Put this code at the bottom of your existing driver file: static int exynos_cs_info(struct udevice *bus, uint cs, struct spi_cs_info *info) { - return -ENODEV; + return -EINVAL; } static const struct dm_spi_ops exynos_spi_ops = {
@@ -633,7 +633,7 @@ is not obvious from outside the driver. In this case you can provide a method for cs_info() to deal with this. If you don't provide it, then the device tree will be used to determine what chip selects are valid. -Return -ENODEV if the supplied chip select is invalid, or 0 if it is valid. +Return -EINVAL if the supplied chip select is invalid, or 0 if it is valid. If you don't provide the cs_info() method, 0 is assumed for all chip selects that do not appear in the device tree.
drivers/spi/ath79_spi.c +1 -1
@@ -198,7 +198,7 @@ static int ath79_cs_info(struct udevice *bus, uint cs, {/* Always allow activity on CS 0/1/2 */ if (cs >=3 ) -return -ENODEV ; +return -EINVAL ;return 0 ; }
drivers/spi/bcm63xx_hsspi.c +1 -1
@@ -108,7 +108,7 @@ static int bcm63xx_hsspi_cs_info(struct udevice *bus, uint cs,if (cs >= priv->num_cs) {printf ("no cs %u\n" , cs); -return -ENODEV ; +return -EINVAL ; }return 0 ;
drivers/spi/bcm63xx_spi.c +1 -1
@@ -130,7 +130,7 @@ static int bcm63xx_spi_cs_info(struct udevice *bus, uint cs,if (cs >= priv->num_cs) {printf ("no cs %u\n" , cs); -return -ENODEV ; +return -EINVAL ; }return 0 ;
drivers/spi/sandbox_spi.c +1 -1
@@ -117,7 +117,7 @@ static int sandbox_cs_info(struct udevice *bus, uint cs, {/* Always allow activity on CS 0 */ if (cs >=1 ) -return -ENODEV ; +return -EINVAL ;return 0 ; }
drivers/spi/tegra20_sflash.c +1 -1
@@ -78,7 +78,7 @@ int tegra20_sflash_cs_info(struct udevice *bus, unsigned int cs, {/* Tegra20 SPI-Flash - only 1 device ('bus/cs') */ if (cs !=0 ) -return -ENODEV ; +return -EINVAL ;else return 0 ; }
include/spi.h +1 -1
@@ -438,7 +438,7 @@ struct dm_spi_ops {* @cs: The chip select (0..n-1) * @info: Returns information about the chip select, if valid. * On entry info->dev is NULL -* @return 0 if OK (and @info is set up), -ENODEV if the chip select +* @return 0 if OK (and @info is set up), -EINVAL if the chip select * is invalid, other -ve value on error */ int (*cs_info)(struct udevice *bus,uint cs,struct spi_cs_info *info);
Signature · 1 line
2.7.4
PATCH v2 3/4 dm: spi: Check cs number before accessing slaves
In reply to Bin MengAdd chip select number check in spi_find_chip_select().
---
Changes in v2:
- move the chip select number check to spi_find_chip_select()
drivers/spi/spi-uclass.c | 45 ++++++++++++++++++++++++++------------------- include/spi.h | 3 ++- 2 files changed, 28 insertions(+), 20 deletions(-)
drivers/spi/spi-uclass.c +26 -19
@@ -179,7 +179,32 @@ int spi_chip_select(struct udevice *dev)int spi_find_chip_select (struct udevice *bus,int cs,struct udevice **devp) { +struct dm_spi_ops *ops; +struct spi_cs_info info;struct udevice *dev; +int ret; + +/* +* Ask the driver. For the moment we don't have CS info. +* When we do we could provide the driver with a helper function +* to figure out what chip selects are valid, or just handle the +* request. +*/ + ops =spi_get_ops (bus); +if (ops->cs_info) { + ret = ops->cs_info (bus, cs, &info); + }else { +/* +* We could assume there is at least one valid chip select. +* The driver didn't care enough to tell us. +*/ + ret =0 ; + } + +if (ret) { +printf ("Invalid cs %d (err=%d)\n" , cs, ret); +return ret; + }for (device_find_first_child (bus, &dev); dev;device_find_next_child (&dev)) {
@@ -214,7 +239,6 @@ int spi_cs_is_valid(unsigned int busnum, unsigned int cs)int spi_cs_info (struct udevice *bus,uint cs,struct spi_cs_info *info) {struct spi_cs_info local_info; -struct dm_spi_ops *ops;int ret;if (!info)
@@ -223,24 +247,7 @@ int spi_cs_info(struct udevice *bus, uint cs, struct spi_cs_info *info)/* If there is a device attached, return it */ info->dev =NULL ; ret =spi_find_chip_select (bus, cs, &info->dev); -if (!ret) -return 0 ; - -/* -* Otherwise ask the driver. For the moment we don't have CS info. -* When we do we could provide the driver with a helper function -* to figure out what chip selects are valid, or just handle the -* request. -*/ - ops =spi_get_ops (bus); -if (ops->cs_info) -return ops->cs_info (bus, cs, info); - -/* -* We could assume there is at least one valid chip select. -* The driver didn't care enough to tell us. -*/ -return 0 ; +return ret == -ENODEV ?0 : ret; } int spi_find_bus_and_cs(int busnum, int cs, struct udevice **busp,
include/spi.h +2 -1
@@ -528,7 +528,8 @@ int spi_chip_select(struct udevice *slave);* @bus: SPI bus to search * @cs: Chip select to look for * @devp: Returns the slave device if found -* @return 0 if found, -ENODEV on error +* @return 0 if found, -EINVAL if cs is invalid, -ENODEV if no device attached, +* other -ve value on error */ int spi_find_chip_select (struct udevice *bus,int cs,struct udevice **devp);
Signature · 1 line
2.7.4
PATCH v2 3/4 dm: spi: Check cs number before accessing slaves
Quoted 2 lines · Bin Meng
Add chip select number check in
spi_find_chip_select().
Since we discussed the cs check to common call in spi_find_chip_select. Would you please add some more commit description.
Quoted 4 lines
Signed-off-by Bin Meng <[email protected]>---
PATCH v2 3/4 dm: spi: Check cs number before accessing slaves
In reply to Bin MengHi Bin,
Quoted 39 lines · Bin Meng
Add chip select number check in
spi_find_chip_select().Signed-off-by Bin Meng <[email protected]>---
Changes in v2:
- move the chip select number check tospi_find_chip_select()drivers/spi/spi-uclass.c | 45 ++++++++++++++++++++++++++------------------- include/spi.h | 3 ++- 2 files changed, 28 insertions(+), 20 deletions(-)drivers/spi/spi-uclass.c +16
-0@@ -179,7 +179,32 @@ int spi_chip_select(struct udevice *dev)int spi_find_chip_select (struct udevice *bus,int cs,struct udevice **devp) { +struct dm_spi_ops *ops; +struct spi_cs_info info;struct udevice *dev; +int ret; + +/* +* Ask the driver. For the moment we don't have CS info. +* When we do we could provide the driver with a helper function +* to figure out what chip selects are valid, or just handle the +* request. +*/ + ops =spi_get_ops (bus); +if (ops->cs_info) { + ret = ops->cs_info (bus, cs, &info); + }else { +/* +* We could assume there is at least one valid chip select.
drivers/spi/spi-uclass.c int spi_chip_select(struct udevice *dev) · line 208
Quoted 10 lines
drivers/spi/spi-uclass.c +9
-0+* The driver didn't care enough to tell us. +*/ + ret =0 ; + } + +if (ret) { +printf ("Invalid cs %d (err=%d)\n" , cs, ret); +return ret; + }
This is breaking 'sf probe' with associated bus and cs on SiFive.
PATCH v2 3/4 dm: spi: Check cs number before accessing slaves
Hi Jagan,
Quoted 47 lines · Jagan Teki
Hi Bin,
Quoted 43 lines · Bin Meng
Add chip select number check in
spi_find_chip_select().Signed-off-by Bin Meng <[email protected]>---
Changes in v2:
- move the chip select number check tospi_find_chip_select()drivers/spi/spi-uclass.c | 45 ++++++++++++++++++++++++++------------------- include/spi.h | 3 ++- 2 files changed, 28 insertions(+), 20 deletions(-)drivers/spi/spi-uclass.c +20
-0@@ -179,7 +179,32 @@ int spi_chip_select(struct udevice *dev)int spi_find_chip_select (struct udevice *bus,int cs,struct udevice **devp) { +struct dm_spi_ops *ops; +struct spi_cs_info info;struct udevice *dev; +int ret; + +/* +* Ask the driver. For the moment we don't have CS info. +* When we do we could provide the driver with a helper function +* to figure out what chip selects are valid, or just handle the +* request. +*/ + ops =spi_get_ops (bus); +if (ops->cs_info) { + ret = ops->cs_info (bus, cs, &info); + }else { +/* +* We could assume there is at least one valid chip select. +* The driver didn't care enough to tell us. +*/ + ret =0 ; + }
Quoted 8 lines
drivers/spi/spi-uclass.c int spi_chip_select(struct udevice *dev) · line 208
Quoted 6 lines
drivers/spi/spi-uclass.c +5
-0+ +if (ret) { +printf ("Invalid cs %d (err=%d)\n" , cs, ret); +return ret; + }This is breaking 'sf probe' with associated bus and cs on SiFive.
No this does not break anything on SiFive. It enforces the 'sf probe' to use the correct CS number on SiFive. And something is wrong when CS is 0 on SiFive.
Regards,
Bin
PATCH v2 3/4 dm: spi: Check cs number before accessing slaves
Hi Jagan,
Quoted 60 lines · Bin Meng
Hi Jagan,
Quoted 47 lines · Jagan Teki
Hi Bin,
Quoted 43 lines · Bin Meng
Add chip select number check in
spi_find_chip_select().Signed-off-by Bin Meng <[email protected]>---
Changes in v2:
- move the chip select number check tospi_find_chip_select()drivers/spi/spi-uclass.c | 45 ++++++++++++++++++++++++++------------------- include/spi.h | 3 ++- 2 files changed, 28 insertions(+), 20 deletions(-)drivers/spi/spi-uclass.c +20
-0@@ -179,7 +179,32 @@ int spi_chip_select(struct udevice *dev)int spi_find_chip_select (struct udevice *bus,int cs,struct udevice **devp) { +struct dm_spi_ops *ops; +struct spi_cs_info info;struct udevice *dev; +int ret; + +/* +* Ask the driver. For the moment we don't have CS info. +* When we do we could provide the driver with a helper function +* to figure out what chip selects are valid, or just handle the +* request. +*/ + ops =spi_get_ops (bus); +if (ops->cs_info) { + ret = ops->cs_info (bus, cs, &info); + }else { +/* +* We could assume there is at least one valid chip select. +* The driver didn't care enough to tell us. +*/ + ret =0 ; + }Quoted 8 lines
drivers/spi/spi-uclass.c int spi_chip_select(struct udevice *dev) · line 208
Quoted 6 lines
drivers/spi/spi-uclass.c +5
-0+ +if (ret) { +printf ("Invalid cs %d (err=%d)\n" , cs, ret); +return ret; + }This is breaking 'sf probe' with associated bus and cs on SiFive.
Quoted 3 lines
No this does not break anything on SiFive. It enforces the 'sf probe' to use the correct CS number on SiFive. And something is wrong when CS is 0 on SiFive.
As discussed in another thread, this patch does not break SiFive. Could you please merge the remaining 2 patches in this series? thanks!
Regards,
Bin
PATCH v2 3/4 dm: spi: Check cs number before accessing slaves
Hi Jagan,
Quoted 64 lines · Bin Meng
Hi Jagan,
Quoted 60 lines · Bin Meng
Hi Jagan,
Quoted 55 lines · Jagan Teki
Hi Bin,
Quoted 39 lines · Bin Meng
Add chip select number check in
spi_find_chip_select().Signed-off-by Bin Meng <[email protected]>---
Changes in v2:
- move the chip select number check tospi_find_chip_select()drivers/spi/spi-uclass.c | 45 ++++++++++++++++++++++++++------------------- include/spi.h | 3 ++- 2 files changed, 28 insertions(+), 20 deletions(-)drivers/spi/spi-uclass.c +16
-0@@ -179,7 +179,32 @@ int spi_chip_select(struct udevice *dev)int spi_find_chip_select (struct udevice *bus,int cs,struct udevice **devp) { +struct dm_spi_ops *ops; +struct spi_cs_info info;struct udevice *dev; +int ret; + +/* +* Ask the driver. For the moment we don't have CS info. +* When we do we could provide the driver with a helper function +* to figure out what chip selects are valid, or just handle the +* request. +*/ + ops =spi_get_ops (bus); +if (ops->cs_info) { + ret = ops->cs_info (bus, cs, &info); + }else { +/* +* We could assume there is at least one valid chip select. drivers/spi/spi-uclass.c int spi_chip_select(struct udevice *dev) · line 208
Quoted 10 lines
drivers/spi/spi-uclass.c +9
-0+* The driver didn't care enough to tell us. +*/ + ret =0 ; + } + +if (ret) { +printf ("Invalid cs %d (err=%d)\n" , cs, ret); +return ret; + }This is breaking 'sf probe' with associated bus and cs on SiFive.
Quoted 6 lines
Quoted 3 lines
No this does not break anything on SiFive. It enforces the 'sf probe' to use the correct CS number on SiFive. And something is wrong when CS is 0 on SiFive.
As discussed in another thread, this patch does not break SiFive. Could you please merge the remaining 2 patches in this series? thanks!
Ping?
Regards,
Bin
PATCH v2 3/4 dm: spi: Check cs number before accessing slaves
Quoted 68 lines · Bin Meng
Hi Jagan,
Quoted 64 lines · Bin Meng
Hi Jagan,
Quoted 60 lines · Bin Meng
Hi Jagan,
Quoted 55 lines · Jagan Teki
Hi Bin,
Quoted 39 lines · Bin Meng
Add chip select number check in
spi_find_chip_select().Signed-off-by Bin Meng <[email protected]>---
Changes in v2:
- move the chip select number check tospi_find_chip_select()drivers/spi/spi-uclass.c | 45 ++++++++++++++++++++++++++------------------- include/spi.h | 3 ++- 2 files changed, 28 insertions(+), 20 deletions(-)drivers/spi/spi-uclass.c +16
-0@@ -179,7 +179,32 @@ int spi_chip_select(struct udevice *dev)int spi_find_chip_select (struct udevice *bus,int cs,struct udevice **devp) { +struct dm_spi_ops *ops; +struct spi_cs_info info;struct udevice *dev; +int ret; + +/* +* Ask the driver. For the moment we don't have CS info. +* When we do we could provide the driver with a helper function +* to figure out what chip selects are valid, or just handle the +* request. +*/ + ops =spi_get_ops (bus); +if (ops->cs_info) { + ret = ops->cs_info (bus, cs, &info); + }else { +/* +* We could assume there is at least one valid chip select. drivers/spi/spi-uclass.c int spi_chip_select(struct udevice *dev) · line 208
Quoted 10 lines
drivers/spi/spi-uclass.c +9
-0+* The driver didn't care enough to tell us. +*/ + ret =0 ; + } + +if (ret) { +printf ("Invalid cs %d (err=%d)\n" , cs, ret); +return ret; + }This is breaking 'sf probe' with associated bus and cs on SiFive.
Quoted 8 lines
Quoted 6 lines
Quoted 3 lines
No this does not break anything on SiFive. It enforces the 'sf probe' to use the correct CS number on SiFive. And something is wrong when CS is 0 on SiFive.
As discussed in another thread, this patch does not break SiFive. Could you please merge the remaining 2 patches in this series? thanks!
Ping?
I took it already will send it PR soon.
Jagan.
PATCH v2 4/4 test: dm: spi: Fix sandbox dm_test_spi_find()
In reply to Bin MengPer sandbox_cs_info(), sandbox spi controller only supports chip select 0. Current test case tries to locate devices on chip select 1, and any call to spi_get_bus_and_cs() or spi_cs_info() with cs number 1 should not return 0.
This updates the test case to handle it correctly.
---
Changes in v2:
- new patch to fix sandbox dm_test_spi_find()
test/dm/spi.c | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-)
test/dm/spi.c +3 -3
@@ -77,10 +77,10 @@ static int dm_test_spi_find(struct unit_test_state *uts)/* We should be able to add something to another chip select */ ut_assertok (sandbox_sf_bind_emul (state, busnum, cs_b, bus, node,"name" )); -ut_assertok (spi_get_bus_and_cs (busnum, cs_b, speed, mode, +ut_asserteq (-EINVAL ,spi_get_bus_and_cs (busnum, cs_b, speed, mode,"spi_flash_std" ,"name" , &bus, &slave)); -ut_assertok (spi_cs_info (bus, cs_b, &info)); -ut_asserteq_ptr (info.dev, slave->dev); +ut_asserteq (-EINVAL ,spi_cs_info (bus, cs_b, &info)); +ut_asserteq_ptr (NULL , info.dev);/* * Since we are about to destroy all devices, we must tell sandbox
Signature · 1 line
2.7.4
PATCH v2 4/4 test: dm: spi: Fix sandbox dm_test_spi_find()
Quoted 23 lines · Bin Meng
Per
sandbox_cs_info(), sandbox spi controller only supports chip select 0. Current test case tries to locate devices on chip select 1, and any call tospi_get_bus_and_cs()orspi_cs_info()with cs number 1 should not return 0.This updates the test case to handle it correctly.
Signed-off-by Bin Meng <[email protected]>---
Changes in v2:
- new patch to fix sandboxdm_test_spi_find()test/dm/spi.c | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-)test/dm/spi.c +0
-0
test/dm/spi.c static int dm_test_spi_find(struct unit_test_state *uts) · line 83
Quoted 10 lines
test/dm/spi.c +3
-3@@ -77,10 +77,10 @@ static int dm_test_spi_find(struct unit_test_state *uts)/* We should be able to add something to another chip select */ ut_assertok (sandbox_sf_bind_emul (state, busnum, cs_b, bus, node,"name" )); -ut_assertok (spi_get_bus_and_cs (busnum, cs_b, speed, mode, +ut_asserteq (-EINVAL ,spi_get_bus_and_cs (busnum, cs_b, speed, mode,"spi_flash_std" ,"name" , &bus, &slave)); -ut_assertok (spi_cs_info (bus, cs_b, &info)); -ut_asserteq_ptr (info.dev, slave->dev); +ut_asserteq (-EINVAL ,spi_cs_info (bus, cs_b, &info)); +ut_asserteq_ptr (NULL , info.dev);
I assume you ran the sandbox test for this, if yes
PATCH v2 1/4 dm: spi: Return 0 if driver does not implement ops->cs_info
In reply to Bin MengQuoted 9 lines · Bin Meng
If an SPI controller driver does not implement
ops->cs_info, that probably means any chip select number could be valid, hence let's return 0 forspi_cs_info().Signed-off-by Bin Meng <[email protected]>Reviewed-by Jagan Teki <[email protected]>---
Applied to u-boot-spi/master