bus: mhi: core: Fix power down latency
authorLoic Poulain <loic.poulain@linaro.org>
Mon, 21 Jun 2021 16:16:10 +0000 (21:46 +0530)
committerGreg Kroah-Hartman <gregkh@linuxfoundation.org>
Thu, 24 Jun 2021 13:51:49 +0000 (15:51 +0200)
On graceful power-down/disable transition, when an MHI reset is
performed, the MHI device loses its context, including interrupt
configuration. However, the current implementation is waiting for
event(irq) driven state change to confirm reset has been completed,
which never happens, and causes reset timeout, leading to unexpected
high latency of the mhi_power_down procedure (up to 45 seconds).

Fix that by moving to the recently introduced poll_reg_field method,
waiting for the reset bit to be cleared, in the same way as the
power_on procedure.

Cc: stable@vger.kernel.org
Fixes: a6e2e3522f29 ("bus: mhi: core: Add support for PM state transitions")
Signed-off-by: Loic Poulain <loic.poulain@linaro.org>
Reviewed-by: Bhaumik Bhatt <bbhatt@codeaurora.org>
Reviewed-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
Reviewed-by: Hemant Kumar <hemantk@codeaurora.org>
Link: https://lore.kernel.org/r/1620029090-8975-1-git-send-email-loic.poulain@linaro.org
Signed-off-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
Link: https://lore.kernel.org/r/20210621161616.77524-3-manivannan.sadhasivam@linaro.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
drivers/bus/mhi/core/pm.c

index e2e59a3..704a5e2 100644 (file)
@@ -465,23 +465,15 @@ static void mhi_pm_disable_transition(struct mhi_controller *mhi_cntrl)
 
        /* Trigger MHI RESET so that the device will not access host memory */
        if (!MHI_PM_IN_FATAL_STATE(mhi_cntrl->pm_state)) {
-               u32 in_reset = -1;
-               unsigned long timeout = msecs_to_jiffies(mhi_cntrl->timeout_ms);
-
                dev_dbg(dev, "Triggering MHI Reset in device\n");
                mhi_set_mhi_state(mhi_cntrl, MHI_STATE_RESET);
 
                /* Wait for the reset bit to be cleared by the device */
-               ret = wait_event_timeout(mhi_cntrl->state_event,
-                                        mhi_read_reg_field(mhi_cntrl,
-                                                           mhi_cntrl->regs,
-                                                           MHICTRL,
-                                                           MHICTRL_RESET_MASK,
-                                                           MHICTRL_RESET_SHIFT,
-                                                           &in_reset) ||
-                                       !in_reset, timeout);
-               if (!ret || in_reset)
-                       dev_err(dev, "Device failed to exit MHI Reset state\n");
+               ret = mhi_poll_reg_field(mhi_cntrl, mhi_cntrl->regs, MHICTRL,
+                                MHICTRL_RESET_MASK, MHICTRL_RESET_SHIFT, 0,
+                                25000);
+               if (ret)
+                       dev_err(dev, "Device failed to clear MHI Reset\n");
 
                /*
                 * Device will clear BHI_INTVEC as a part of RESET processing,