[PATCH V4 07/11] arm64: dts: imx8mq: Enable both G1 and G2 VPU's with vpu-blk-ctrl
Martin Kepplinger
martin.kepplinger at puri.sm
Mon May 23 05:00:22 PDT 2022
Am Dienstag, dem 26.04.2022 um 12:43 +0200 schrieb Lucas Stach:
> Am Dienstag, dem 26.04.2022 um 09:38 +0200 schrieb Martin Kepplinger:
> > Am Montag, dem 25.04.2022 um 17:34 +0200 schrieb Lucas Stach:
> > > Hi Martin,
> > >
> > > Am Montag, dem 25.04.2022 um 17:22 +0200 schrieb Martin
> > > Kepplinger:
> > > > Am Dienstag, dem 25.01.2022 um 11:11 -0600 schrieb Adam Ford:
> > > > > With the Hantro G1 and G2 now setup to run independently,
> > > > > update
> > > > > the device tree to allow both to operate. This requires the
> > > > > vpu-blk-ctrl node to be configured. Since vpu-blk-ctrl needs
> > > > > certain clock enabled to handle the gating of the G1 and G2
> > > > > fuses, the clock-parents and clock-rates for the various
> > > > > VPU's
> > > > > to be moved into the pgc_vpu because they cannot get re-
> > > > > parented
> > > > > once enabled, and the pgc_vpu is the highest in the chain.
> > > > >
> > > > > Signed-off-by: Adam Ford <aford173 at gmail.com>
> > > > > Reported-by: kernel test robot <lkp at intel.com>
> > > > > Reviewed-by: Ezequiel Garcia <ezequiel at vanguardiasur.com.ar>
> > > > >
> > > > > diff --git a/arch/arm64/boot/dts/freescale/imx8mq.dtsi
> > > > > b/arch/arm64/boot/dts/freescale/imx8mq.dtsi
> > > > > index 2df2510d0118..549b2440f55d 100644
> > > > > --- a/arch/arm64/boot/dts/freescale/imx8mq.dtsi
> > > > > +++ b/arch/arm64/boot/dts/freescale/imx8mq.dtsi
> > > > > @@ -737,7 +737,21 @@ pgc_gpu: power-domain at 5 {
> > > > > pgc_vpu: power-
> > > > > domain at 6 {
> > > > > #power-
> > > > > domain-
> > > > > cells =
> > > > > <0>;
> > > > > reg =
> > > > > <IMX8M_POWER_DOMAIN_VPU>;
> > > > > - clocks =
> > > > > <&clk
> > > > > IMX8MQ_CLK_VPU_DEC_ROOT>;
> > > > > + clocks =
> > > > > <&clk
> > > > > IMX8MQ_CLK_VPU_DEC_ROOT>,
> > > > > +
> > > > > <&clk
> > > > > IMX8MQ_CLK_VPU_G1_ROOT>,
> > > > > +
> > > > > <&clk
> > > > > IMX8MQ_CLK_VPU_G2_ROOT>;
> > > > > + assigned-
> > > > > clocks =
> > > > > <&clk IMX8MQ_CLK_VPU_G1>,
> > > > > +
> > > > >
> > > > > <&clk IMX8MQ_CLK_VPU_G2>,
> > > > > +
> > > > >
> > > > > <&clk IMX8MQ_CLK_VPU_BUS>,
> > > > > +
> > > > >
> > > > > <&clk IMX8MQ_VPU_PLL_BYPASS>;
> > > > > + assigned-
> > > > > clock-
> > > > > parents = <&clk IMX8MQ_VPU_PLL_OUT>,
> > > > > +
> > > > >
> > > > >
> > > > > <&clk IMX8MQ_VPU_PLL_OUT>,
> > > > > +
> > > > >
> > > > >
> > > > > <&clk IMX8MQ_SYS1_PLL_800M>,
> > > > > +
> > > > >
> > > > >
> > > > > <&clk IMX8MQ_VPU_PLL>;
> > > > > + assigned-
> > > > > clock-
> > > > > rates
> > > > > = <600000000>,
> > > > > +
> > > > >
> > > > >
> > > > > <600000000>,
> > > > > +
> > > > >
> > > > >
> > > > > <800000000>,
> > > > > +
> > > > >
> > > > >
> > > > > <0>;
> > > > > };
> > > > >
> > > > > pgc_disp:
> > > > > power-domain at 7
> > > > > {
> > > > > @@ -1457,30 +1471,31 @@ usb3_phy1: usb-phy at 382f0040 {
> > > > > status = "disabled";
> > > > > };
> > > > >
> > > > > - vpu: video-codec at 38300000 {
> > > > > - compatible = "nxp,imx8mq-vpu";
> > > > > - reg = <0x38300000 0x10000>,
> > > > > - <0x38310000 0x10000>,
> > > > > - <0x38320000 0x10000>;
> > > > > - reg-names = "g1", "g2", "ctrl";
> > > > > - interrupts = <GIC_SPI 7
> > > > > IRQ_TYPE_LEVEL_HIGH>,
> > > > > - <GIC_SPI 8
> > > > > IRQ_TYPE_LEVEL_HIGH>;
> > > > > - interrupt-names = "g1", "g2";
> > > > > + vpu_g1: video-codec at 38300000 {
> > > > > + compatible = "nxp,imx8mq-vpu-g1";
> > > > > + reg = <0x38300000 0x10000>;
> > > > > + interrupts = <GIC_SPI 7
> > > > > IRQ_TYPE_LEVEL_HIGH>;
> > > > > + clocks = <&clk
> > > > > IMX8MQ_CLK_VPU_G1_ROOT>;
> > > > > + power-domains = <&vpu_blk_ctrl
> > > > > IMX8MQ_VPUBLK_PD_G1>;
> > > > > + };
> > > > > +
> > > > > + vpu_g2: video-codec at 38310000 {
> > > > > + compatible = "nxp,imx8mq-vpu-g2";
> > > > > + reg = <0x38310000 0x10000>;
> > > > > + interrupts = <GIC_SPI 8
> > > > > IRQ_TYPE_LEVEL_HIGH>;
> > > > > + clocks = <&clk
> > > > > IMX8MQ_CLK_VPU_G2_ROOT>;
> > > > > + power-domains = <&vpu_blk_ctrl
> > > > > IMX8MQ_VPUBLK_PD_G2>;
> > > > > + };
> > > > > +
> > > > > + vpu_blk_ctrl: blk-ctrl at 38320000 {
> > > > > + compatible = "fsl,imx8mq-vpu-blk-
> > > > > ctrl";
> > > > > + reg = <0x38320000 0x100>;
> > > > > + power-domains = <&pgc_vpu>,
> > > > > <&pgc_vpu>,
> > > > > <&pgc_vpu>;
> > > > > + power-domain-names = "bus", "g1",
> > > > > "g2";
> > > > > clocks = <&clk
> > > > > IMX8MQ_CLK_VPU_G1_ROOT>,
> > > > > - <&clk
> > > > > IMX8MQ_CLK_VPU_G2_ROOT>,
> > > > > - <&clk
> > > > > IMX8MQ_CLK_VPU_DEC_ROOT>;
> > > > > - clock-names = "g1", "g2", "bus";
> > > > > - assigned-clocks = <&clk
> > > > > IMX8MQ_CLK_VPU_G1>,
> > > > > - <&clk
> > > > > IMX8MQ_CLK_VPU_G2>,
> > > > > - <&clk
> > > > > IMX8MQ_CLK_VPU_BUS>,
> > > > > - <&clk
> > > > > IMX8MQ_VPU_PLL_BYPASS>;
> > > > > - assigned-clock-parents = <&clk
> > > > > IMX8MQ_VPU_PLL_OUT>,
> > > > > - <&clk
> > > > > IMX8MQ_VPU_PLL_OUT>,
> > > > > - <&clk
> > > > > IMX8MQ_SYS1_PLL_800M>,
> > > > > - <&clk
> > > > > IMX8MQ_VPU_PLL>;
> > > > > - assigned-clock-rates = <600000000>,
> > > > > <600000000>,
> > > > > - <800000000>,
> > > > > <0>;
> > > > > - power-domains = <&pgc_vpu>;
> > > > > + <&clk
> > > > > IMX8MQ_CLK_VPU_G2_ROOT>;
> > > > > + clock-names = "g1", "g2";
> > > > > + #power-domain-cells = <1>;
> > > > > };
> > > > >
> > > > > pcie0: pcie at 33800000 {
> > > >
> > > > With this update, when testing suspend to ram on imx8mq, I get:
> > > >
> > > > buck4: failed to disable: -ETIMEDOUT
> > > >
> > > > where buck4 is power-supply of pgc_vpu. And thus the transition
> > > > to
> > > > suspend (and resuming) fails.
> > > >
> > > > Have you tested system suspend after the imx8m-blk-ctrl update
> > > > on
> > > > imx8mq?
> > >
> > > I haven't tested system suspend, don't know if anyone else did.
> > > However
> > > I guess that this is just uncovering a preexisting issue in the
> > > system
> > > suspend sequencing, which you would also hit if the video
> > > decoders
> > > were
> > > active at system suspend time.
> > >
> > > My guess is that the regulator disable fails, due to the power
> > > domains
> > > being disabled quite late in the suspend sequence, where i2c
> > > communication with the PMIC is no longer possible due to i2c
> > > being
> > > suspended already or something like that. Maybe you can dig in a
> > > bit
> > > on
> > > the actual sequence on your system and we can see how we can
> > > rework
> > > things to suspend the power domains at a time where communication
> > > with
> > > the PMIC is still possible?
> >
> > What exactly would you like to see? Here's all gpcv2 regulators
> > disabling on suspend. (gpu (domain 5) is disabled by runtime pm
> > often):
> >
> > [ 47.138700] imx-pgc imx-pgc-domain.5: disabling regulator
> > [ 47.298071] Freezing user space processes ... (elapsed 0.008
> > seconds) done.
> > [ 47.313432] OOM killer disabled.
> > [ 47.316670] Freezing remaining freezable tasks ... (elapsed
> > 2.221
> > seconds) done.
> > [ 49.672052] imx8m-blk-ctrl 38320000.blk-ctrl:
> > imx8m_blk_ctrl_suspend
> > start
> > [ 49.704417] imx-pgc imx-pgc-domain.0: disabling regulator
> > [ 49.711114] imx-pgc imx-pgc-domain.6: disabling regulator
> > [ 49.819064] buck4: failed to disable: -ETIMEDOUT
> >
> > The stack looks pretty much the same for all of them, from
> > pm_suspend()
> > over genpd_suspend_noiry().
>
> So the GPU domain is already suspended before the system suspend,
> probably due to short runtime PM timeouts.
>
> Can you please check at which point the i2c subsystem is suspended? I
> think we are already past that point when running the PM domain
> suspend
> from a _noirq callback. I'll take a look on how we can properly
> change
> this ordering.
>
> Regards,
> Lucas
>
hi Lucas, just to update a bit: (I rmmod hantro_vpu because it adds
another bug but let's ignore for now) when I use a mainline kernel and
for testing just ignore ETIMEDOUT in gpcv2 on regulator_disable():
- return ret;
+ if (ret == -ETIMEDOUT)
+ ret = 0;
+ else
+ return ret;
I see the error message the first time I suspend. Not the consecutive
times (and the systems seems to be working after resuming).
Now what's weird is when I suspend and resume often, after some time,
resume will definitely fail. Not sure yet where, but when I add debug
output for the kernel/ directory, that's usually about where it hangs:
[ 969.620825] OOM killer enabled.
[ 969.623971] Restarting tasks ...
[ 969.624230] systemd-udevd left refrigerator
[ 969.624535] khugepaged left refrigerator
[ 969.624568] gdbus left refrigerator
[ 969.624795] alsa-sink-308b0 left refrigerator
[ 969.626864] pipewire left refrigerator
[ 969.626880] alsa-source-308 left refrigerator
[ 969.626913] alsa-source-300 left refrigerator
[ 969.626920] gmain left refrigerator
[ 969.626949] gvfs-afc-volume left refrigerator
[ 969.626949] systemd left refrigerator
[ 969.626977] gmain left refrigerator
[ 969.627009] chatty left refrigerator
[ 969.627036] gmain left refrigerator
[ 969.627060] gmain left refrigerator
[ 969.627087] gmain left refrigerator
And the above hang ("sometimes") happens for "platform" in pm_test
already, so I guess it's purely a linux bug.
just so you know, in case that tells you anything more...
thanks,
martin
More information about the Linux-rockchip
mailing list