| Message ID | 20200905204151.25343-12-digetx@gmail.com |
|---|---|
| State | Superseded |
| Headers | show |
| Series | Improvements for Tegra I2C driver | expand |
On Sat, Sep 05, 2020 at 11:41:31PM +0300, Dmitry Osipenko wrote: > Factor out runtime PM and hardware initialization into separate function > in order have a cleaner error unwinding in the probe function. [...] > + ret = tegra_i2c_init_runtime_pm_and_hardware(i2c_dev); [...] This one doesn't improve the code for me. The problems are: 1) putting two unrelated parts in one function, 2) silently reordered initialization. Best Regards, Michał Mirosław
06.09.2020 01:10, Michał Mirosław пишет: > On Sat, Sep 05, 2020 at 11:41:31PM +0300, Dmitry Osipenko wrote: >> Factor out runtime PM and hardware initialization into separate function >> in order have a cleaner error unwinding in the probe function. > [...] >> + ret = tegra_i2c_init_runtime_pm_and_hardware(i2c_dev); > [...] > > This one doesn't improve the code for me. The problems are: 1) putting two > unrelated parts in one function, 2) silently reordered initialization. The hardware initialization depends on the resumed RPM and the rest of the probe function doesn't care about the RPM. I don't quite understand why you're saying that they are unrelated, could you please explain? The DMA/RPM initialization is intentionally reordered in order to clean up the error handling, like the commit message says. To me it's a clear improvement :)
On Sun, Sep 06, 2020 at 01:24:14AM +0300, Dmitry Osipenko wrote: > 06.09.2020 01:10, Michał Mirosław пишет: > > On Sat, Sep 05, 2020 at 11:41:31PM +0300, Dmitry Osipenko wrote: > >> Factor out runtime PM and hardware initialization into separate function > >> in order have a cleaner error unwinding in the probe function. > > [...] > >> + ret = tegra_i2c_init_runtime_pm_and_hardware(i2c_dev); > > [...] > > > > This one doesn't improve the code for me. The problems are: 1) putting two > > unrelated parts in one function, 2) silently reordered initialization. > > The hardware initialization depends on the resumed RPM and the rest of > the probe function doesn't care about the RPM. I don't quite understand > why you're saying that they are unrelated, could you please explain? > > The DMA/RPM initialization is intentionally reordered in order to clean > up the error handling, like the commit message says. To me it's a clear > improvement :) Ok, then wouldn't it be enough to just move this part in the probe()? A sign of a problem for me is how much information you had to put in the name of the new function. Best Regards, Michał Mirosław
06.09.2020 01:51, Michał Mirosław пишет: > On Sun, Sep 06, 2020 at 01:24:14AM +0300, Dmitry Osipenko wrote: >> 06.09.2020 01:10, Michał Mirosław пишет: >>> On Sat, Sep 05, 2020 at 11:41:31PM +0300, Dmitry Osipenko wrote: >>>> Factor out runtime PM and hardware initialization into separate function >>>> in order have a cleaner error unwinding in the probe function. >>> [...] >>>> + ret = tegra_i2c_init_runtime_pm_and_hardware(i2c_dev); >>> [...] >>> >>> This one doesn't improve the code for me. The problems are: 1) putting two >>> unrelated parts in one function, 2) silently reordered initialization. >> >> The hardware initialization depends on the resumed RPM and the rest of >> the probe function doesn't care about the RPM. I don't quite understand >> why you're saying that they are unrelated, could you please explain? >> >> The DMA/RPM initialization is intentionally reordered in order to clean >> up the error handling, like the commit message says. To me it's a clear >> improvement :) > > Ok, then wouldn't it be enough to just move this part in the probe()? > A sign of a problem for me is how much information you had to put in > the name of the new function. Looking at it again now, I think you're right. I also now noticed that the RPM isn't disabled now if tegra_i2c_init() fails. I'll try to take another look, but probably will lean to yours variant in the v5. Thanks!
diff --git a/drivers/i2c/busses/i2c-tegra.c b/drivers/i2c/busses/i2c-tegra.c index 8e4e72dec6ea..6e5af03d0b1d 100644 --- a/drivers/i2c/busses/i2c-tegra.c +++ b/drivers/i2c/busses/i2c-tegra.c @@ -1670,6 +1670,37 @@ static void tegra_i2c_release_clocks(struct tegra_i2c_dev *i2c_dev) clk_bulk_unprepare(i2c_dev->nclocks, i2c_dev->clocks); } +static int tegra_i2c_init_runtime_pm_and_hardware(struct tegra_i2c_dev *i2c_dev) +{ + int ret; + + /* + * VI I2C is in VE power domain which is not always on and not + * an IRQ safe. So, IRQ safe device can't be attached to a non-IRQ + * safe domain as it prevents powering off the PM domain. + * Also, VI I2C device don't need to use runtime IRQ safe as it will + * not be used for atomic transfers. + */ + if (!i2c_dev->is_vi) + pm_runtime_irq_safe(i2c_dev->dev); + + pm_runtime_enable(i2c_dev->dev); + + ret = pm_runtime_get_sync(i2c_dev->dev); + if (ret < 0) { + dev_err(i2c_dev->dev, "runtime resume failed: %d\n", ret); + pm_runtime_disable(i2c_dev->dev); + return ret; + } + + /* initialize hardware state */ + ret = tegra_i2c_init(i2c_dev); + + pm_runtime_put(i2c_dev->dev); + + return ret; +} + static int tegra_i2c_probe(struct platform_device *pdev) { struct device *dev = &pdev->dev; @@ -1728,41 +1759,23 @@ static int tegra_i2c_probe(struct platform_device *pdev) platform_set_drvdata(pdev, i2c_dev); - /* - * VI I2C is in VE power domain which is not always on and not - * an IRQ safe. So, IRQ safe device can't be attached to a non-IRQ - * safe domain as it prevents powering off the PM domain. - * Also, VI I2C device don't need to use runtime IRQ safe as it will - * not be used for atomic transfers. - */ - if (!i2c_dev->is_vi) - pm_runtime_irq_safe(&pdev->dev); - pm_runtime_enable(&pdev->dev); - ret = pm_runtime_get_sync(i2c_dev->dev); - if (ret < 0) { - dev_err(dev, "runtime resume failed\n"); - goto disable_rpm; - } - if (i2c_dev->hw->supports_bus_clear) i2c_dev->adapter.bus_recovery_info = &tegra_i2c_recovery_info; ret = tegra_i2c_init_dma(i2c_dev); if (ret < 0) - goto put_rpm; + goto release_clocks; - ret = tegra_i2c_init(i2c_dev); - if (ret) { - dev_err(&pdev->dev, "Failed to initialize i2c controller\n"); + ret = tegra_i2c_init_runtime_pm_and_hardware(i2c_dev); + if (ret) goto release_dma; - } irq_set_status_flags(i2c_dev->irq, IRQ_NOAUTOEN); ret = devm_request_irq(&pdev->dev, i2c_dev->irq, tegra_i2c_isr, IRQF_NO_SUSPEND, dev_name(&pdev->dev), i2c_dev); if (ret) - goto release_dma; + goto release_rpm; i2c_set_adapdata(&i2c_dev->adapter, i2c_dev); i2c_dev->adapter.owner = THIS_MODULE; @@ -1777,18 +1790,13 @@ static int tegra_i2c_probe(struct platform_device *pdev) if (ret) goto release_dma; - pm_runtime_put(&pdev->dev); - return 0; +release_rpm: + pm_runtime_disable(i2c_dev->dev); release_dma: tegra_i2c_release_dma(i2c_dev); - -put_rpm: - pm_runtime_put_sync(&pdev->dev); - -disable_rpm: - pm_runtime_disable(&pdev->dev); +release_clocks: tegra_i2c_release_clocks(i2c_dev); return ret;
Factor out runtime PM and hardware initialization into separate function in order have a cleaner error unwinding in the probe function. Signed-off-by: Dmitry Osipenko <digetx@gmail.com> --- drivers/i2c/busses/i2c-tegra.c | 68 +++++++++++++++++++--------------- 1 file changed, 38 insertions(+), 30 deletions(-)