| Message ID | 20180223210100.86732-12-tony@atomide.com (mailing list archive) |
|---|---|
| State | Not Applicable, archived |
| Headers | show |
On 2/23/2018 10:00 PM, Tony Lindgren wrote: > We are currently probing smartreflex with omap_device while we are > already probing smartreflex related interconnect target module with > ti-sysc driver and dts data. > > Before we can flip things on for ti-sysc, we need to prepare the > smartreflex driver a bit: > > 1. The smartreflex clock is really for the whole interconnect target > module. So it may be configured at the parent device level with > ti-sysc > > 2. With ti-sysc, we have the child device manage interconnect target > module directly if pm_runtime_irq_safe() is set and there is only > one child. In that case nobody else is going to call pm_runtime_get > and put, so we need to add these calls to idle smartreflex properly > after probe if not fully configured > > 3. With ti-sysc, the parent driver may rebind. So we want to use > platform_driver_register() and don't want probe to be __init > > Note that this patch depends on the related changes to ti-sysc driver > and omap_device probing to prevent both ti-sysc and omap_device to > try to probe smartreflex. > > Cc: linux-pm@vger.kernel.org > Cc: Rafael J. Wysocki <rafael.j.wysocki@intel.com> > Signed-off-by: Tony Lindgren <tony@atomide.com> You need an ACK from Kevin Hilman on this anyway and if you receive one, please feel free to route it whatever way is convenient for you. > --- > drivers/power/avs/smartreflex.c | 41 ++++++++++++++++++++++++++++++++++------- > 1 file changed, 34 insertions(+), 7 deletions(-) > > diff --git a/drivers/power/avs/smartreflex.c b/drivers/power/avs/smartreflex.c > --- a/drivers/power/avs/smartreflex.c > +++ b/drivers/power/avs/smartreflex.c > @@ -132,12 +132,16 @@ static void sr_set_clk_length(struct omap_sr *sr) > struct clk *fck; > u32 fclk_speed; > > - fck = clk_get(&sr->pdev->dev, "fck"); > - > + /* Try interconnect target module fck first if it already exists */ > + fck = clk_get(sr->pdev->dev.parent, "fck"); > if (IS_ERR(fck)) { > - dev_err(&sr->pdev->dev, "%s: unable to get fck for device %s\n", > - __func__, dev_name(&sr->pdev->dev)); > - return; > + fck = clk_get(&sr->pdev->dev, "fck"); > + if (IS_ERR(fck)) { > + dev_err(&sr->pdev->dev, > + "%s: unable to get fck for device %s\n", > + __func__, dev_name(&sr->pdev->dev)); > + return; > + } > } > > fclk_speed = clk_get_rate(fck); > @@ -838,7 +842,7 @@ static int omap_sr_autocomp_store(void *data, u64 val) > DEFINE_SIMPLE_ATTRIBUTE(pm_sr_fops, omap_sr_autocomp_show, > omap_sr_autocomp_store, "%llu\n"); > > -static int __init omap_sr_probe(struct platform_device *pdev) > +static int omap_sr_probe(struct platform_device *pdev) > { > struct omap_sr *sr_info; > struct omap_sr_data *pdata = pdev->dev.platform_data; > @@ -898,6 +902,12 @@ static int __init omap_sr_probe(struct platform_device *pdev) > > list_add(&sr_info->node, &sr_list); > > + ret = pm_runtime_get_sync(&pdev->dev); > + if (ret < 0) { > + pm_runtime_put_noidle(&pdev->dev); > + goto err_list_del; > + } > + > /* > * Call into late init to do initializations that require > * both sr driver and sr class driver to be initiallized. > @@ -966,12 +976,17 @@ static int __init omap_sr_probe(struct platform_device *pdev) > > } > > + pm_runtime_put_sync(&pdev->dev); > + > return ret; > > err_debugfs: > debugfs_remove_recursive(sr_info->dbg_dir); > err_list_del: > list_del(&sr_info->node); > + > + pm_runtime_put_sync(&pdev->dev); > + > return ret; > } > > @@ -1025,11 +1040,23 @@ static void omap_sr_shutdown(struct platform_device *pdev) > return; > } > > +static const struct of_device_id omap_sr_match[] = { > + { .compatible = "ti,omap3-smartreflex-core", }, > + { .compatible = "ti,omap3-smartreflex-mpu-iva", }, > + { .compatible = "ti,omap4-smartreflex-core", }, > + { .compatible = "ti,omap4-smartreflex-mpu", }, > + { .compatible = "ti,omap4-smartreflex-iva", }, > + { }, > +}; > +MODULE_DEVICE_TABLE(of, omap_sr_match); > + > static struct platform_driver smartreflex_driver = { > + .probe = omap_sr_probe, > .remove = omap_sr_remove, > .shutdown = omap_sr_shutdown, > .driver = { > .name = DRIVER_NAME, > + .of_match_table = omap_sr_match, > }, > }; > > @@ -1048,7 +1075,7 @@ static int __init sr_init(void) > else > pr_warn("%s: No PMIC hook to init smartreflex\n", __func__); > > - ret = platform_driver_probe(&smartreflex_driver, omap_sr_probe); > + ret = platform_driver_register(&smartreflex_driver); > if (ret) { > pr_err("%s: platform driver register failed for SR\n", > __func__);
* Rafael J. Wysocki <rafael.j.wysocki@intel.com> [180226 10:38]: > On 2/23/2018 10:00 PM, Tony Lindgren wrote: > > We are currently probing smartreflex with omap_device while we are > > already probing smartreflex related interconnect target module with > > ti-sysc driver and dts data. ... > You need an ACK from Kevin Hilman on this anyway and if you receive one, > please feel free to route it whatever way is convenient for you. Oops sorry about that Kevin. I did the Cc list from memory and totally forgot about the MAINTAINERS entry for AVS drivers. I'll also bounce the original email to Kevin for easier reading. Regards, Tony
Tony Lindgren <tony@atomide.com> writes: > We are currently probing smartreflex with omap_device while we are > already probing smartreflex related interconnect target module with > ti-sysc driver and dts data. > > Before we can flip things on for ti-sysc, we need to prepare the > smartreflex driver a bit: > > 1. The smartreflex clock is really for the whole interconnect target > module. So it may be configured at the parent device level with > ti-sysc > > 2. With ti-sysc, we have the child device manage interconnect target > module directly if pm_runtime_irq_safe() is set and there is only > one child. In that case nobody else is going to call pm_runtime_get > and put, so we need to add these calls to idle smartreflex properly > after probe if not fully configured > > 3. With ti-sysc, the parent driver may rebind. So we want to use > platform_driver_register() and don't want probe to be __init > > Note that this patch depends on the related changes to ti-sysc driver > and omap_device probing to prevent both ti-sysc and omap_device to > try to probe smartreflex. > > Cc: linux-pm@vger.kernel.org > Cc: Rafael J. Wysocki <rafael.j.wysocki@intel.com> > Signed-off-by: Tony Lindgren <tony@atomide.com> Acked-by: Kevin Hilman <khilman@baylibre.com> I don't have anything else queued for this driver, so feel free to take it along with the rest of the series. Kevin
* Kevin Hilman <khilman@baylibre.com> [180301 03:08]: > Tony Lindgren <tony@atomide.com> writes: > > We are currently probing smartreflex with omap_device while we are > > already probing smartreflex related interconnect target module with > > ti-sysc driver and dts data. ... > Acked-by: Kevin Hilman <khilman@baylibre.com> > > I don't have anything else queued for this driver, so feel free to take > it along with the rest of the series. Thanks for looking, will do. Tony
diff --git a/drivers/power/avs/smartreflex.c b/drivers/power/avs/smartreflex.c --- a/drivers/power/avs/smartreflex.c +++ b/drivers/power/avs/smartreflex.c @@ -132,12 +132,16 @@ static void sr_set_clk_length(struct omap_sr *sr) struct clk *fck; u32 fclk_speed; - fck = clk_get(&sr->pdev->dev, "fck"); - + /* Try interconnect target module fck first if it already exists */ + fck = clk_get(sr->pdev->dev.parent, "fck"); if (IS_ERR(fck)) { - dev_err(&sr->pdev->dev, "%s: unable to get fck for device %s\n", - __func__, dev_name(&sr->pdev->dev)); - return; + fck = clk_get(&sr->pdev->dev, "fck"); + if (IS_ERR(fck)) { + dev_err(&sr->pdev->dev, + "%s: unable to get fck for device %s\n", + __func__, dev_name(&sr->pdev->dev)); + return; + } } fclk_speed = clk_get_rate(fck); @@ -838,7 +842,7 @@ static int omap_sr_autocomp_store(void *data, u64 val) DEFINE_SIMPLE_ATTRIBUTE(pm_sr_fops, omap_sr_autocomp_show, omap_sr_autocomp_store, "%llu\n"); -static int __init omap_sr_probe(struct platform_device *pdev) +static int omap_sr_probe(struct platform_device *pdev) { struct omap_sr *sr_info; struct omap_sr_data *pdata = pdev->dev.platform_data; @@ -898,6 +902,12 @@ static int __init omap_sr_probe(struct platform_device *pdev) list_add(&sr_info->node, &sr_list); + ret = pm_runtime_get_sync(&pdev->dev); + if (ret < 0) { + pm_runtime_put_noidle(&pdev->dev); + goto err_list_del; + } + /* * Call into late init to do initializations that require * both sr driver and sr class driver to be initiallized. @@ -966,12 +976,17 @@ static int __init omap_sr_probe(struct platform_device *pdev) } + pm_runtime_put_sync(&pdev->dev); + return ret; err_debugfs: debugfs_remove_recursive(sr_info->dbg_dir); err_list_del: list_del(&sr_info->node); + + pm_runtime_put_sync(&pdev->dev); + return ret; } @@ -1025,11 +1040,23 @@ static void omap_sr_shutdown(struct platform_device *pdev) return; } +static const struct of_device_id omap_sr_match[] = { + { .compatible = "ti,omap3-smartreflex-core", }, + { .compatible = "ti,omap3-smartreflex-mpu-iva", }, + { .compatible = "ti,omap4-smartreflex-core", }, + { .compatible = "ti,omap4-smartreflex-mpu", }, + { .compatible = "ti,omap4-smartreflex-iva", }, + { }, +}; +MODULE_DEVICE_TABLE(of, omap_sr_match); + static struct platform_driver smartreflex_driver = { + .probe = omap_sr_probe, .remove = omap_sr_remove, .shutdown = omap_sr_shutdown, .driver = { .name = DRIVER_NAME, + .of_match_table = omap_sr_match, }, }; @@ -1048,7 +1075,7 @@ static int __init sr_init(void) else pr_warn("%s: No PMIC hook to init smartreflex\n", __func__); - ret = platform_driver_probe(&smartreflex_driver, omap_sr_probe); + ret = platform_driver_register(&smartreflex_driver); if (ret) { pr_err("%s: platform driver register failed for SR\n", __func__);
We are currently probing smartreflex with omap_device while we are already probing smartreflex related interconnect target module with ti-sysc driver and dts data. Before we can flip things on for ti-sysc, we need to prepare the smartreflex driver a bit: 1. The smartreflex clock is really for the whole interconnect target module. So it may be configured at the parent device level with ti-sysc 2. With ti-sysc, we have the child device manage interconnect target module directly if pm_runtime_irq_safe() is set and there is only one child. In that case nobody else is going to call pm_runtime_get and put, so we need to add these calls to idle smartreflex properly after probe if not fully configured 3. With ti-sysc, the parent driver may rebind. So we want to use platform_driver_register() and don't want probe to be __init Note that this patch depends on the related changes to ti-sysc driver and omap_device probing to prevent both ti-sysc and omap_device to try to probe smartreflex. Cc: linux-pm@vger.kernel.org Cc: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Signed-off-by: Tony Lindgren <tony@atomide.com> --- drivers/power/avs/smartreflex.c | 41 ++++++++++++++++++++++++++++++++++------- 1 file changed, 34 insertions(+), 7 deletions(-)