aboutsummaryrefslogtreecommitdiffstats
diff options
context:
space:
mode:
authorShengjiu Wang <shengjiu.wang@nxp.com>2024-05-11 10:55:25 +0800
committerUlf Hansson <ulf.hansson@linaro.org>2024-05-21 12:21:01 +0200
commite8dc41afca161b988e6d462f4d0803d247e22250 (patch)
tree72c9ba26bd76540a4c80faf527e48cc9ea8930ae
parent8f6a15f095a63a83b096d9b29aaff4f0fbe6f6e6 (diff)
downloadlinux-pm-next.tar.gz
pmdomain: imx: gpcv2: Add delay after power up handshakenextfixes
AudioMix BLK-CTRL on i.MX8MP encountered an accessing register issue after power up. [ 2.181035] Kernel panic - not syncing: Asynchronous SError Interrupt [ 2.181038] CPU: 1 PID: 48 Comm: kworker/u16:2 Not tainted 6.9.0-rc5-next-20240424-00003-g21cec88845c6 #171 [ 2.181047] Hardware name: NXP i.MX8MPlus EVK board (DT) [ 2.181050] Workqueue: events_unbound deferred_probe_work_func [ 2.181064] Call trace: [...] [ 2.181142] arm64_serror_panic+0x6c/0x78 [ 2.181149] do_serror+0x3c/0x70 [ 2.181157] el1h_64_error_handler+0x30/0x48 [ 2.181164] el1h_64_error+0x64/0x68 [ 2.181171] clk_imx8mp_audiomix_runtime_resume+0x34/0x44 [ 2.181183] __genpd_runtime_resume+0x30/0x80 [ 2.181195] genpd_runtime_resume+0x110/0x244 [ 2.181205] __rpm_callback+0x48/0x1d8 [ 2.181213] rpm_callback+0x68/0x74 [ 2.181224] rpm_resume+0x468/0x6c0 [ 2.181234] __pm_runtime_resume+0x50/0x94 [ 2.181243] pm_runtime_get_suppliers+0x60/0x8c [ 2.181258] __driver_probe_device+0x48/0x12c [ 2.181268] driver_probe_device+0xd8/0x15c [ 2.181278] __device_attach_driver+0xb8/0x134 [ 2.181290] bus_for_each_drv+0x84/0xe0 [ 2.181302] __device_attach+0x9c/0x188 [ 2.181312] device_initial_probe+0x14/0x20 [ 2.181323] bus_probe_device+0xac/0xb0 [ 2.181334] deferred_probe_work_func+0x88/0xc0 [ 2.181344] process_one_work+0x150/0x290 [ 2.181357] worker_thread+0x2f8/0x408 [ 2.181370] kthread+0x110/0x114 [ 2.181381] ret_from_fork+0x10/0x20 [ 2.181391] SMP: stopping secondary CPUs According to comments in power up handshake: /* request the ADB400 to power up */ if (domain->bits.hskreq) { regmap_update_bits(domain->regmap, domain->regs->hsk, domain->bits.hskreq, domain->bits.hskreq); /* * ret = regmap_read_poll_timeout(domain->regmap, domain->regs->hsk, reg_val, * (reg_val & domain->bits.hskack), 0, * USEC_PER_MSEC); * Technically we need the commented code to wait handshake. But that needs * the BLK-CTL module BUS clk-en bit being set. * * There is a separate BLK-CTL module and we will have such a driver for it, * that driver will set the BUS clk-en bit and handshake will be triggered * automatically there. Just add a delay and suppose the handshake finish * after that. */ } The BLK-CTL module needs to add delay to wait for a handshake request finished. For some BLK-CTL module (eg. AudioMix on i.MX8MP) doesn't have BUS clk-en bit, it is better to add delay in this driver, as the BLK-CTL module doesn't need to care about how it is powered up. regmap_read_bypassed() is to make sure the above write IO transaction already reaches target before udelay(). Fixes: 1496dd413b2e ("clk: imx: imx8mp: Add pm_runtime support for power saving") Reported-by: Francesco Dolcini <francesco@dolcini.it> Closes: https://lore.kernel.org/all/66293535.170a0220.21fe.a2e7@mx.google.com/ Suggested-by: Frank Li <frank.li@nxp.com> Signed-off-by: Shengjiu Wang <shengjiu.wang@nxp.com> Tested-by: Adam Ford <aford173@gmail.com> Tested-by: Alexander Stein <alexander.stein@ew.tq-group.com> Link: https://lore.kernel.org/r/1715396125-3724-1-git-send-email-shengjiu.wang@nxp.com Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
-rw-r--r--drivers/pmdomain/imx/gpcv2.c11
1 files changed, 11 insertions, 0 deletions
diff --git a/drivers/pmdomain/imx/gpcv2.c b/drivers/pmdomain/imx/gpcv2.c
index 4b828d74a6064..856eaac0ec140 100644
--- a/drivers/pmdomain/imx/gpcv2.c
+++ b/drivers/pmdomain/imx/gpcv2.c
@@ -393,6 +393,17 @@ static int imx_pgc_power_up(struct generic_pm_domain *genpd)
* automatically there. Just add a delay and suppose the handshake finish
* after that.
*/
+
+ /*
+ * For some BLK-CTL module (eg. AudioMix on i.MX8MP) doesn't have BUS
+ * clk-en bit, it is better to add delay here, as the BLK-CTL module
+ * doesn't need to care about how it is powered up.
+ *
+ * regmap_read_bypassed() is to make sure the above write IO transaction
+ * already reaches target before udelay()
+ */
+ regmap_read_bypassed(domain->regmap, domain->regs->hsk, &reg_val);
+ udelay(5);
}
/* Disable reset clocks for all devices in the domain */