| Message ID | a8ae84bd2dc4b55fe428f8e20f31438bf8bb6762.1684089931.git.christophe.jaillet@wanadoo.fr |
|---|---|
| State | Accepted |
| Commit | 89382022b370dfd34eaae9c863baa123fcd4d132 |
| Headers | show |
| Series | [v2] thermal/drivers/sun8i: Fix some error handling paths in sun8i_ths_probe() | expand |
Hi Maxime, On Mon, May 15, 2023 at 01:54:41PM +0200, Maxime Ripard wrote: > Hi, > > On Sun, May 14, 2023 at 08:46:05PM +0200, Christophe JAILLET wrote: > > Should an error occur after calling sun8i_ths_resource_init() in the probe > > function, some resources need to be released, as already done in the > > .remove() function. > > > > Switch to the devm_clk_get_enabled() helper and add a new devm_action to > > turn sun8i_ths_resource_init() into a fully managed function. > > > > Move the place where reset_control_deassert() is called so that the > > recommended order of reset release/clock enable steps is kept. > > A64 manual states that: > > > > 3.3.6.4. Gating and reset > > > > Make sure that the reset signal has been released before the release of > > module clock gating; > > > > This fixes the issue and removes some LoC at the same time. > > It should really be three different patches: > - One to fix the resource release, that can be backported to stable > releases > - One to switch to devm_clk_get_enabled > - and one to change the order of clock enable vs reset deassertion The order was correct before this patch. I don't think an incorrect order should be intorduced, even if temporarily between two patches. regards, o. > Maxime
On 15/05/2023 14:33, Ondřej Jirman wrote: > Hi Maxime, > > On Mon, May 15, 2023 at 01:54:41PM +0200, Maxime Ripard wrote: >> Hi, >> >> On Sun, May 14, 2023 at 08:46:05PM +0200, Christophe JAILLET wrote: >>> Should an error occur after calling sun8i_ths_resource_init() in the probe >>> function, some resources need to be released, as already done in the >>> .remove() function. >>> >>> Switch to the devm_clk_get_enabled() helper and add a new devm_action to >>> turn sun8i_ths_resource_init() into a fully managed function. >>> >>> Move the place where reset_control_deassert() is called so that the >>> recommended order of reset release/clock enable steps is kept. >>> A64 manual states that: >>> >>> 3.3.6.4. Gating and reset >>> >>> Make sure that the reset signal has been released before the release of >>> module clock gating; >>> >>> This fixes the issue and removes some LoC at the same time. >> >> It should really be three different patches: >> - One to fix the resource release, that can be backported to stable >> releases >> - One to switch to devm_clk_get_enabled >> - and one to change the order of clock enable vs reset deassertion > > The order was correct before this patch. I don't think an incorrect order > should be intorduced, even if temporarily between two patches. Maxime ?
On Tue, Jun 13, 2023 at 10:49:24AM +0200, Daniel Lezcano wrote: > On 15/05/2023 14:33, Ondřej Jirman wrote: > > Hi Maxime, > > > > On Mon, May 15, 2023 at 01:54:41PM +0200, Maxime Ripard wrote: > > > Hi, > > > > > > On Sun, May 14, 2023 at 08:46:05PM +0200, Christophe JAILLET wrote: > > > > Should an error occur after calling sun8i_ths_resource_init() in the probe > > > > function, some resources need to be released, as already done in the > > > > .remove() function. > > > > > > > > Switch to the devm_clk_get_enabled() helper and add a new devm_action to > > > > turn sun8i_ths_resource_init() into a fully managed function. > > > > > > > > Move the place where reset_control_deassert() is called so that the > > > > recommended order of reset release/clock enable steps is kept. > > > > A64 manual states that: > > > > > > > > 3.3.6.4. Gating and reset > > > > > > > > Make sure that the reset signal has been released before the release of > > > > module clock gating; > > > > > > > > This fixes the issue and removes some LoC at the same time. > > > > > > It should really be three different patches: > > > - One to fix the resource release, that can be backported to stable > > > releases > > > - One to switch to devm_clk_get_enabled > > > - and one to change the order of clock enable vs reset deassertion > > > > The order was correct before this patch. I don't think an incorrect order > > should be intorduced, even if temporarily between two patches. > > Maxime ? I agree with Ondřej Maxime
On 13/06/2023 11:16, Maxime Ripard wrote: > On Tue, Jun 13, 2023 at 10:49:24AM +0200, Daniel Lezcano wrote: >> On 15/05/2023 14:33, Ondřej Jirman wrote: >>> Hi Maxime, >>> >>> On Mon, May 15, 2023 at 01:54:41PM +0200, Maxime Ripard wrote: >>>> Hi, >>>> >>>> On Sun, May 14, 2023 at 08:46:05PM +0200, Christophe JAILLET wrote: >>>>> Should an error occur after calling sun8i_ths_resource_init() in the probe >>>>> function, some resources need to be released, as already done in the >>>>> .remove() function. >>>>> >>>>> Switch to the devm_clk_get_enabled() helper and add a new devm_action to >>>>> turn sun8i_ths_resource_init() into a fully managed function. >>>>> >>>>> Move the place where reset_control_deassert() is called so that the >>>>> recommended order of reset release/clock enable steps is kept. >>>>> A64 manual states that: >>>>> >>>>> 3.3.6.4. Gating and reset >>>>> >>>>> Make sure that the reset signal has been released before the release of >>>>> module clock gating; >>>>> >>>>> This fixes the issue and removes some LoC at the same time. >>>> >>>> It should really be three different patches: >>>> - One to fix the resource release, that can be backported to stable >>>> releases >>>> - One to switch to devm_clk_get_enabled >>>> - and one to change the order of clock enable vs reset deassertion >>> >>> The order was correct before this patch. I don't think an incorrect order >>> should be intorduced, even if temporarily between two patches. >> >> Maxime ? > > I agree with Ondřej Thanks, I applied the patch
diff --git a/drivers/thermal/sun8i_thermal.c b/drivers/thermal/sun8i_thermal.c index 793ddce72132..d4d241686c81 100644 --- a/drivers/thermal/sun8i_thermal.c +++ b/drivers/thermal/sun8i_thermal.c @@ -319,6 +319,11 @@ static int sun8i_ths_calibrate(struct ths_device *tmdev) return ret; } +static void sun8i_ths_reset_control_assert(void *data) +{ + reset_control_assert(data); +} + static int sun8i_ths_resource_init(struct ths_device *tmdev) { struct device *dev = tmdev->dev; @@ -339,47 +344,35 @@ static int sun8i_ths_resource_init(struct ths_device *tmdev) if (IS_ERR(tmdev->reset)) return PTR_ERR(tmdev->reset); - tmdev->bus_clk = devm_clk_get(&pdev->dev, "bus"); + ret = reset_control_deassert(tmdev->reset); + if (ret) + return ret; + + ret = devm_add_action_or_reset(dev, sun8i_ths_reset_control_assert, + tmdev->reset); + if (ret) + return ret; + + tmdev->bus_clk = devm_clk_get_enabled(&pdev->dev, "bus"); if (IS_ERR(tmdev->bus_clk)) return PTR_ERR(tmdev->bus_clk); } if (tmdev->chip->has_mod_clk) { - tmdev->mod_clk = devm_clk_get(&pdev->dev, "mod"); + tmdev->mod_clk = devm_clk_get_enabled(&pdev->dev, "mod"); if (IS_ERR(tmdev->mod_clk)) return PTR_ERR(tmdev->mod_clk); } - ret = reset_control_deassert(tmdev->reset); - if (ret) - return ret; - - ret = clk_prepare_enable(tmdev->bus_clk); - if (ret) - goto assert_reset; - ret = clk_set_rate(tmdev->mod_clk, 24000000); if (ret) - goto bus_disable; - - ret = clk_prepare_enable(tmdev->mod_clk); - if (ret) - goto bus_disable; + return ret; ret = sun8i_ths_calibrate(tmdev); if (ret) - goto mod_disable; + return ret; return 0; - -mod_disable: - clk_disable_unprepare(tmdev->mod_clk); -bus_disable: - clk_disable_unprepare(tmdev->bus_clk); -assert_reset: - reset_control_assert(tmdev->reset); - - return ret; } static int sun8i_h3_thermal_init(struct ths_device *tmdev) @@ -530,17 +523,6 @@ static int sun8i_ths_probe(struct platform_device *pdev) return 0; } -static int sun8i_ths_remove(struct platform_device *pdev) -{ - struct ths_device *tmdev = platform_get_drvdata(pdev); - - clk_disable_unprepare(tmdev->mod_clk); - clk_disable_unprepare(tmdev->bus_clk); - reset_control_assert(tmdev->reset); - - return 0; -} - static const struct ths_thermal_chip sun8i_a83t_ths = { .sensor_num = 3, .scale = 705, @@ -642,7 +624,6 @@ MODULE_DEVICE_TABLE(of, of_ths_match); static struct platform_driver ths_driver = { .probe = sun8i_ths_probe, - .remove = sun8i_ths_remove, .driver = { .name = "sun8i-thermal", .of_match_table = of_ths_match,
Should an error occur after calling sun8i_ths_resource_init() in the probe function, some resources need to be released, as already done in the .remove() function. Switch to the devm_clk_get_enabled() helper and add a new devm_action to turn sun8i_ths_resource_init() into a fully managed function. Move the place where reset_control_deassert() is called so that the recommended order of reset release/clock enable steps is kept. A64 manual states that: 3.3.6.4. Gating and reset Make sure that the reset signal has been released before the release of module clock gating; This fixes the issue and removes some LoC at the same time. Fixes: dccc5c3b6f30 ("thermal/drivers/sun8i: Add thermal driver for H6/H5/H3/A64/A83T/R40") Signed-off-by: Christophe JAILLET <christophe.jaillet@wanadoo.fr> --- Changes in v2: - move reset_control_deassert() next to devm_reset_control_get() v1: https://lore.kernel.org/all/26f9e3bb3fcd0c12ea24a44c75b7960da993b68b.1684077651.git.christophe.jaillet@wanadoo.fr/ --- drivers/thermal/sun8i_thermal.c | 55 +++++++++++---------------------- 1 file changed, 18 insertions(+), 37 deletions(-)