On 7/9/2018 5:49 PM, Evan Green wrote: > pm_runtime_suspended can return 0 even if the last runtime power > management function called in the device was a suspend call. This > trips up the i2c-qcom-geni's suspend_noirq accounting, which was > trying to use the function to determine if it should disable clocks > or if that had already been done. > > The fix is to track whether or not clocks have been enabled explicitly > in the driver with a new member. While in there, also honor the return > value of geni_se_resources_off, which can technically fail. > > An example of the warning generated during suspend: > [ 68.314733] WARNING: CPU: 0 PID: 1990 at drivers/clk/clk.c:595 clk_core_disable+0x28/0x200 > [ 68.464576] Call trace: > [ 68.554410] [<ffffff80083f0ed0>] clk_core_disable+0x28/0x200 > [ 68.560244] [<ffffff80083f349c>] clk_disable+0x38/0x4c > [ 68.565541] [<ffffff8008408910>] geni_se_resources_off+0x40/0x74 > [ 68.571731] [<ffffff800862aadc>] geni_i2c_runtime_suspend+0x2c/0x3c > [ 68.578185] [<ffffff800862ae6c>] geni_i2c_suspend_noirq+0x38/0x68 > [ 68.584475] [<ffffff800852bd40>] dpm_run_callback+0x104/0x210 > [ 68.590404] [<ffffff800852c3d4>] __device_suspend_noirq+0x158/0x1a8 > [ 68.596859] [<ffffff800852dff0>] dpm_noirq_suspend_devices+0x180/0x38c > [ 68.603594] [<ffffff80080fdebc>] suspend_devices_and_enter+0x49c/0x964 > [ 68.610321] [<ffffff80080fea14>] pm_suspend+0x690/0x6e0 > [ 68.615712] [<ffffff80080fc930>] state_store+0xd4/0xf8 > [ 68.621014] [<ffffff80088a57f0>] kobj_attr_store+0x18/0x28 > [ 68.626672] [<ffffff8008297a54>] sysfs_kf_write+0x5c/0x68 > [ 68.632240] [<ffffff8008296a00>] kernfs_fop_write+0x174/0x1b8 > [ 68.638177] [<ffffff800821a2c4>] __vfs_write+0x58/0x160 > [ 68.643567] [<ffffff800821a5d8>] vfs_write+0xcc/0x184 > [ 68.648780] [<ffffff800821a850>] SyS_write+0x64/0xb4 > > Signed-off-by: Evan Green <evgreen@xxxxxxxxxxxx> > --- > This change applies atop Karthik's patch [1], which as far as I can tell > technically hasn't landed yet, but also hasn't had feedback in awhile > either. I opted to submit this as a separate patch, but if the maintainers > prefer it's fine with me if this get folded in to another spin of the > base series. > > [1] https://patchwork.kernel.org/patch/10460819/ > > drivers/i2c/busses/i2c-qcom-geni.c | 25 ++++++++++++++++++++++--- > 1 file changed, 22 insertions(+), 3 deletions(-) > > diff --git a/drivers/i2c/busses/i2c-qcom-geni.c b/drivers/i2c/busses/i2c-qcom-geni.c > index 54bcd1a79cc5..90c4cbfa22d8 100644 > --- a/drivers/i2c/busses/i2c-qcom-geni.c > +++ b/drivers/i2c/busses/i2c-qcom-geni.c > @@ -85,6 +85,7 @@ struct geni_i2c_dev { > spinlock_t lock; > u32 clk_freq_out; > const struct geni_i2c_clk_fld *clk_fld; > + int suspended; > }; > > struct geni_i2c_err_log { > @@ -567,9 +568,15 @@ static int geni_i2c_probe(struct platform_device *pdev) > geni_se_init(&gi2c->se, gi2c->tx_wm, tx_depth); > geni_se_config_packing(&gi2c->se, BITS_PER_BYTE, PACKING_BYTES_PW, > true, true, true); > - geni_se_resources_off(&gi2c->se); > + ret = geni_se_resources_off(&gi2c->se); > + if (ret) { > + dev_err(&pdev->dev, "Error turning off resources %d\n", ret); > + return ret; > + } > + > dev_dbg(&pdev->dev, "i2c fifo/se-dma mode. fifo depth:%d\n", tx_depth); > > + gi2c->suspended = 1; > pm_runtime_set_suspended(gi2c->se.dev); > pm_runtime_set_autosuspend_delay(gi2c->se.dev, I2C_AUTO_SUSPEND_DELAY); > pm_runtime_use_autosuspend(gi2c->se.dev); > @@ -590,10 +597,19 @@ static int geni_i2c_remove(struct platform_device *pdev) > > static int __maybe_unused geni_i2c_runtime_suspend(struct device *dev) > { > + int ret; > struct geni_i2c_dev *gi2c = dev_get_drvdata(dev); > > disable_irq(gi2c->irq); > - geni_se_resources_off(&gi2c->se); > + ret = geni_se_resources_off(&gi2c->se); > + if (ret) { > + enable_irq(gi2c->irq); > + return ret; > + > + } else { > + gi2c->suspended = 1; > + } > + > return 0; > } > > @@ -607,12 +623,15 @@ static int __maybe_unused geni_i2c_runtime_resume(struct device *dev) > return ret; > > enable_irq(gi2c->irq); > + gi2c->suspended = 0; > return 0; > } > > static int __maybe_unused geni_i2c_suspend_noirq(struct device *dev) > { > - if (!pm_runtime_suspended(dev)) { > + struct geni_i2c_dev *gi2c = dev_get_drvdata(dev); > + > + if (!gi2c->suspended) { > geni_i2c_runtime_suspend(dev); > pm_runtime_disable(dev); > pm_runtime_set_suspended(dev); > Reviewed-by: Karthikeyan Ramasubramanian <kramasub@xxxxxxxxxxxxxx> -- Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum, a Linux Foundation Collaborative Project
