| Message ID | 6e471922-2850-97b8-c555-34e1708cbc38@free.fr (mailing list archive) |
|---|---|
| State | Superseded |
| Headers | show |
| Series | [v1] regulator: core: Log forbidden DRMS operation | expand |
On Tue, Feb 19, 2019 at 05:02:46PM +0100, Marc Gonzalez wrote: > When REGULATOR_CHANGE_DRMS is not set, drms_uA_update is a no-op. > It used to print a debug message, which was dropped in commit > 8a34e979f684 ("regulator: refactor valid_ops_mask checking code") > Let's bring the debug message back as KERN_INFO, because it is very > useful to diagnose missing regulator-allow-set-load properties. No, it's perfectly normal for machine constraints to stop drivers from doing things so we shouldn't warn on this - it would get incredibly noisy if we started printing every time constraints didn't let us do something at info level. Debug level might be viable, or definitely vdbg or trace points.
On 19/02/2019 17:39, Mark Brown wrote: > On Tue, Feb 19, 2019 at 05:02:46PM +0100, Marc Gonzalez wrote: > >> When REGULATOR_CHANGE_DRMS is not set, drms_uA_update is a no-op. >> It used to print a debug message, which was dropped in commit >> 8a34e979f684 ("regulator: refactor valid_ops_mask checking code") >> >> Let's bring the debug message back as KERN_INFO, because it is very >> useful to diagnose missing regulator-allow-set-load properties. > > No, it's perfectly normal for machine constraints to stop drivers from > doing things so we shouldn't warn on this - it would get incredibly > noisy if we started printing every time constraints didn't let us do > something at info level. Debug level might be viable, or definitely > vdbg or trace points. I had a look at the various REGULATOR_CHANGE_* flags. #define REGULATOR_CHANGE_VOLTAGE 0x1 #define REGULATOR_CHANGE_CURRENT 0x2 #define REGULATOR_CHANGE_MODE 0x4 #define REGULATOR_CHANGE_STATUS 0x8 #define REGULATOR_CHANGE_DRMS 0x10 #define REGULATOR_CHANGE_BYPASS 0x20 Several functions return an error (and log a KERN_ERR message) if their corresponding flag is not set: regulator_check_voltage() REGULATOR_CHANGE_VOLTAGE regulator_check_current_limit() REGULATOR_CHANGE_CURRENT regulator_mode_constrain() REGULATOR_CHANGE_MODE Others succeed silently even if their corresponding flag is not set: drms_uA_update() REGULATOR_CHANGE_DRMS regulator_allow_bypass() REGULATOR_CHANGE_BYPASS Why is setting the voltage handled differently than setting the load? Regards.
[ SMTP server blocked by spamcop -- trying a different one ] On 19/02/2019 17:39, Mark Brown wrote: > On Tue, Feb 19, 2019 at 05:02:46PM +0100, Marc Gonzalez wrote: > >> When REGULATOR_CHANGE_DRMS is not set, drms_uA_update is a no-op. >> It used to print a debug message, which was dropped in commit >> 8a34e979f684 ("regulator: refactor valid_ops_mask checking code") >> >> Let's bring the debug message back as KERN_INFO, because it is very >> useful to diagnose missing regulator-allow-set-load properties. > > No, it's perfectly normal for machine constraints to stop drivers from > doing things so we shouldn't warn on this - it would get incredibly > noisy if we started printing every time constraints didn't let us do > something at info level. Debug level might be viable, or definitely > vdbg or trace points. I had a look at the various REGULATOR_CHANGE_* flags. #define REGULATOR_CHANGE_VOLTAGE 0x1 #define REGULATOR_CHANGE_CURRENT 0x2 #define REGULATOR_CHANGE_MODE 0x4 #define REGULATOR_CHANGE_STATUS 0x8 #define REGULATOR_CHANGE_DRMS 0x10 #define REGULATOR_CHANGE_BYPASS 0x20 Several functions return an error (and log a KERN_ERR message) if their corresponding flag is not set: regulator_check_voltage() REGULATOR_CHANGE_VOLTAGE regulator_check_current_limit() REGULATOR_CHANGE_CURRENT regulator_mode_constrain() REGULATOR_CHANGE_MODE Others succeed silently even if their corresponding flag is not set: drms_uA_update() REGULATOR_CHANGE_DRMS regulator_allow_bypass() REGULATOR_CHANGE_BYPASS Why is setting the voltage handled differently than setting the load? Regards.
On Thu, Feb 21, 2019 at 10:42:01AM +0100, Marc Gonzalez wrote: > On 19/02/2019 17:39, Mark Brown wrote: > > On Tue, Feb 19, 2019 at 05:02:46PM +0100, Marc Gonzalez wrote: > > No, it's perfectly normal for machine constraints to stop drivers from > > doing things so we shouldn't warn on this - it would get incredibly > > noisy if we started printing every time constraints didn't let us do > > something at info level. Debug level might be viable, or definitely > > vdbg or trace points. > Several functions return an error (and log a KERN_ERR message) if their > corresponding flag is not set: > > regulator_check_voltage() REGULATOR_CHANGE_VOLTAGE That's a bug due to an incomplete implementation, what it's supposed to be doing there is checking if the voltage request is satisfied by the current voltage and returning an error only if the requested voltage is out of range. > regulator_check_current_limit() REGULATOR_CHANGE_CURRENT This one is less clear than anything else - we might want to error out here if the device is somehow relying on being able to lower the current limit for safety. We'd need an audit of users. > regulator_mode_constrain() REGULATOR_CHANGE_MODE This one is also a bug, modes barely mean anything anyway.
diff --git a/drivers/regulator/core.c b/drivers/regulator/core.c index 59782a706ac4..2ff9b55cfcd8 100644 --- a/drivers/regulator/core.c +++ b/drivers/regulator/core.c @@ -930,8 +930,10 @@ static int drms_uA_update(struct regulator_dev *rdev) * first check to see if we can set modes at all, otherwise just * tell the consumer everything is OK. */ - if (!regulator_ops_is_valid(rdev, REGULATOR_CHANGE_DRMS)) + if (!regulator_ops_is_valid(rdev, REGULATOR_CHANGE_DRMS)) { + rdev_info(rdev, "DRMS operation not allowed\n"); return 0; + } if (!rdev->desc->ops->get_optimum_mode && !rdev->desc->ops->set_load)
When REGULATOR_CHANGE_DRMS is not set, drms_uA_update is a no-op. It used to print a debug message, which was dropped in commit 8a34e979f684 ("regulator: refactor valid_ops_mask checking code") Let's bring the debug message back as KERN_INFO, because it is very useful to diagnose missing regulator-allow-set-load properties. Signed-off-by: Marc Gonzalez <marc.w.gonzalez@free.fr> --- drivers/regulator/core.c | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-)