Re: [PATCH] ARM: mvebu: Enable MBUS error propagation

[Chris Packham]

Hi Gregory,

On Wed, 2020-01-08 at 11:22 +0100, Gregory CLEMENT wrote:
> Hello Chris,
> 
> > U-boot disables MBUS error propagation for Armada-385. The effect of
> > this on the kernel is that any access to a mapped but inaccessible
> > address causes the system to hang.
> > 
> > By enabling MBUS error propagation the kernel can raise a Bus Error and
> > panic to restart the system.
> 
> Unless I miss it, it seems that nobody comment this patch: sorry for the
> delay.
> 

Thanks for the response.

> > 
> > Signed-off-by: Chris Packham <chris.packham@xxxxxxxxxxxxxxxxxxx>
> > ---
> > 
> > Notes:
> >     We've encountered an issue where rogue accesses to PCI-e space cause an
> >     Armada-385 system to lockup. We've found that enabling MBUS error
> >     propagation lets us get a bus error which at least gives us a panic to
> >     help identify what was accessed.
> >     
> >     U-boot clears the IO Err Prop Enable Bit[1] but so far no-one seems to
> >     know why.
> >     
> >     I wasn't sure where to put this code. There is similar code for kirwood
> >     in the equivalent dt_init function. On Armada-XP the register is part of
> >     the Core Coherency Fabric block (for A385 it's documented as part of the
> >     CCF block).
> 
> What about adding a new set of register to the mvebu mbus driver?
> 

After more testing we found that some previously "good" boards started
throwing up panics with this change. I think that this might require
handling some of the PCI-e interrupts (for correctable errors) via the
EDAC subsystem.

We're still working with Marvell to track down exactly why this is
happening on our system.

> In this case it will be called even earlier allowing to see bus error
> earlier.
> 
> In any case, you should separate the device tree change from the code
> change and at least provide 2 patches.

Agreed. If/when something solid eventuates we'll do it as a proper
series.

> 
> Gregory
> 
> >     
> >     --
> >     [1] - https://gitlab.denx.de/u-boot/u-boot/blob/master/arch/arm/mach-mvebu/cpu.c#L489
> > 
> >  arch/arm/boot/dts/armada-38x.dtsi |  5 +++++
> >  arch/arm/mach-mvebu/board-v7.c    | 27 +++++++++++++++++++++++++++
> >  2 files changed, 32 insertions(+)
> > 
> > diff --git a/arch/arm/boot/dts/armada-38x.dtsi b/arch/arm/boot/dts/armada-38x.dtsi
> > index 3f4bb44d85f0..3214c67433eb 100644
> > --- a/arch/arm/boot/dts/armada-38x.dtsi
> > +++ b/arch/arm/boot/dts/armada-38x.dtsi
> > @@ -386,6 +386,11 @@
> >  				      <0x20250 0x8>;
> >  			};
> >  
> > +			ioerrc: io-err-control@20200 {
> > +				compatible = "marvell,io-err-control";
> > +				reg = <0x20200 0x4>;
> > +			};
> > +
> >  			mpic: interrupt-controller@20a00 {
> >  				compatible = "marvell,mpic";
> >  				reg = <0x20a00 0x2d0>, <0x21070 0x58>;
> > diff --git a/arch/arm/mach-mvebu/board-v7.c b/arch/arm/mach-mvebu/board-v7.c
> > index d2df5ef9382b..fb7718386ef9 100644
> > --- a/arch/arm/mach-mvebu/board-v7.c
> > +++ b/arch/arm/mach-mvebu/board-v7.c
> > @@ -138,10 +138,36 @@ static void __init i2c_quirk(void)
> >  	}
> >  }
> >  
> > +#define MBUS_ERR_PROP_EN BIT(8)
> > +
> > +/*
> > + * U-boot disables MBUS error propagation. Re-enable it so we
> > + * can handle them as Bus Errors.
> > + */
> > +static void __init enable_mbus_error_propagation(void)
> > +{
> > +	struct device_node *np =
> > +		of_find_compatible_node(NULL, NULL, "marvell,io-err-control");
> > +
> > +	if (np) {
> > +		void __iomem *reg;
> > +
> > +		reg = of_iomap(np, 0);
> > +		if (reg) {
> > +			u32 val;
> > +
> > +			val = readl_relaxed(reg);
> > +			writel_relaxed(val | MBUS_ERR_PROP_EN, reg);
> > +		}
> > +		of_node_put(np);
> > +	}
> > +}
> > +
> >  static void __init mvebu_dt_init(void)
> >  {
> >  	if (of_machine_is_compatible("marvell,armadaxp"))
> >  		i2c_quirk();
> > +	enable_mbus_error_propagation();
> >  }
> >  
> >  static void __init armada_370_xp_dt_fixup(void)
> > @@ -191,6 +217,7 @@ DT_MACHINE_START(ARMADA_38X_DT, "Marvell Armada 380/385 (Device Tree)")
> >  	.l2c_aux_val	= 0,
> >  	.l2c_aux_mask	= ~0,
> >  	.init_irq       = mvebu_init_irq,
> > +	.init_machine	= mvebu_dt_init,
> >  	.restart	= mvebu_restart,
> >  	.dt_compat	= armada_38x_dt_compat,
> >  MACHINE_END
> > -- 
> > 2.24.0
> > 
> 
>