| Message ID | 20240423233622.1494708-3-florian.fainelli@broadcom.com |
|---|---|
| State | New |
| Headers | show |
| Series | [1/4] i2c: designware: Create shared header hosting driver name | expand |
On Tue, 23 Apr 2024, Florian Fainelli wrote: > > > On 4/23/2024 5:00 PM, Andy Shevchenko wrote: > > Tue, Apr 23, 2024 at 04:36:20PM -0700, Florian Fainelli kirjoitti: > > > Rather than open code the i2c_designware string, utilize the newly > > > defined constant in i2c-designware.h. > > > > ... > > > > > static const struct mfd_cell intel_lpss_i2c_cell = { > > > - .name = "i2c_designware", > > > + .name = I2C_DESIGNWARE_NAME, > > > .num_resources = ARRAY_SIZE(intel_lpss_dev_resources), > > > .resources = intel_lpss_dev_resources, > > > }; > > > > We have tons of drivers that are using explicit naming, why is this case > > special? > > > > It is not special, just one of the 3 cases outside of drivers/i2c/busses > that reference a driver living under drivers/i2c/busses, as I replied in the > cover letter, this is a contract between the various device drivers and > their users, so we should have a central place where it is defined, not > repeated. I have always held the opinion that replacing user-facing strings with defines harms debugability, since grepping becomes a multi-stage process, often with ambiguous results (in the case of multiple definitions with the same name. Please keep the string in-place.
On 5/2/24 00:17, Lee Jones wrote: > On Tue, 23 Apr 2024, Florian Fainelli wrote: > >> >> >> On 4/23/2024 5:00 PM, Andy Shevchenko wrote: >>> Tue, Apr 23, 2024 at 04:36:20PM -0700, Florian Fainelli kirjoitti: >>>> Rather than open code the i2c_designware string, utilize the newly >>>> defined constant in i2c-designware.h. >>> >>> ... >>> >>>> static const struct mfd_cell intel_lpss_i2c_cell = { >>>> - .name = "i2c_designware", >>>> + .name = I2C_DESIGNWARE_NAME, >>>> .num_resources = ARRAY_SIZE(intel_lpss_dev_resources), >>>> .resources = intel_lpss_dev_resources, >>>> }; >>> >>> We have tons of drivers that are using explicit naming, why is this case >>> special? >>> >> >> It is not special, just one of the 3 cases outside of drivers/i2c/busses >> that reference a driver living under drivers/i2c/busses, as I replied in the >> cover letter, this is a contract between the various device drivers and >> their users, so we should have a central place where it is defined, not >> repeated. > > I have always held the opinion that replacing user-facing strings with > defines harms debugability, since grepping becomes a multi-stage > process, often with ambiguous results (in the case of multiple > definitions with the same name. Please keep the string in-place. I am not buying into that argument and the fact that Duangiang was able to trip over the lack of an explicit contract between drivers seems like a bigger obstacle than doing a multi-stage grep. Anyway, I have no skin in this game, I just don't like seeing repetition and not stating contracts between drivers more explicitly.
On Thu, 02 May 2024, Florian Fainelli wrote: > On 5/2/24 00:17, Lee Jones wrote: > > On Tue, 23 Apr 2024, Florian Fainelli wrote: > > > > > > > > > > > On 4/23/2024 5:00 PM, Andy Shevchenko wrote: > > > > Tue, Apr 23, 2024 at 04:36:20PM -0700, Florian Fainelli kirjoitti: > > > > > Rather than open code the i2c_designware string, utilize the newly > > > > > defined constant in i2c-designware.h. > > > > > > > > ... > > > > > > > > > static const struct mfd_cell intel_lpss_i2c_cell = { > > > > > - .name = "i2c_designware", > > > > > + .name = I2C_DESIGNWARE_NAME, > > > > > .num_resources = ARRAY_SIZE(intel_lpss_dev_resources), > > > > > .resources = intel_lpss_dev_resources, > > > > > }; > > > > > > > > We have tons of drivers that are using explicit naming, why is this case > > > > special? > > > > > > > > > > It is not special, just one of the 3 cases outside of drivers/i2c/busses > > > that reference a driver living under drivers/i2c/busses, as I replied in the > > > cover letter, this is a contract between the various device drivers and > > > their users, so we should have a central place where it is defined, not > > > repeated. > > > > I have always held the opinion that replacing user-facing strings with > > defines harms debugability, since grepping becomes a multi-stage > > process, often with ambiguous results (in the case of multiple > > definitions with the same name. Please keep the string in-place. > > I am not buying into that argument and the fact that Duangiang was able to > trip over the lack of an explicit contract between drivers seems like a > bigger obstacle than doing a multi-stage grep. Anyway, I have no skin in > this game, I just don't like seeing repetition and not stating contracts > between drivers more explicitly. Good thing no one is asking you to buy into it then. :) I'm not sure how or if the code that failed to match was tested or what went wrong exactly and I'm pleased that the bug was caught and fixed. However, swapping out matching strings with a define is a regression from a development perspective. One which I've felt the pain of personally in the past. I've pushed back on it before and I'm pushing back again. We're not swapping out matching strings for defines.
diff --git a/drivers/mfd/intel-lpss.c b/drivers/mfd/intel-lpss.c index 2a9018112dfc..4d2c5250046f 100644 --- a/drivers/mfd/intel-lpss.c +++ b/drivers/mfd/intel-lpss.c @@ -19,6 +19,7 @@ #include <linux/device.h> #include <linux/err.h> #include <linux/gfp_types.h> +#include <linux/i2c-designware.h> #include <linux/idr.h> #include <linux/io.h> #include <linux/ioport.h> @@ -116,7 +117,7 @@ static const struct mfd_cell intel_lpss_idma64_cell = { }; static const struct mfd_cell intel_lpss_i2c_cell = { - .name = "i2c_designware", + .name = I2C_DESIGNWARE_NAME, .num_resources = ARRAY_SIZE(intel_lpss_dev_resources), .resources = intel_lpss_dev_resources, };
Rather than open code the i2c_designware string, utilize the newly defined constant in i2c-designware.h. Signed-off-by: Florian Fainelli <florian.fainelli@broadcom.com> --- drivers/mfd/intel-lpss.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-)