Re: [Xen-devel] [PATCH RFC 11/15] xen/arm: generate a simple device tree for domUs

[Stefano Stabellini <sstabellini@xxxxxxxxxx>]

On Thu, 14 Jun 2018, Julien Grall wrote:
> Hi Stefano,
> 
> On 13/06/18 23:15, Stefano Stabellini wrote:
> > Introduce functions to generate a basic domU device tree, similar to the
> > existing functions in tools/libxl/libxl_arm.c.
> > 
> > Rename existing prepare_dtb to prepare_dtb_dom0 to avoid confusion.
> 
> Ah this is were the rename is. It might have make sense to do that in patch
> #9. Also, you want to name it "hwdom" as this is the preferred name nowadays.

OK


> > 
> > Signed-off-by: Stefano Stabellini <stefanos@xxxxxxxxxx>
> > ---
> >   xen/arch/arm/domain_build.c | 195
> > +++++++++++++++++++++++++++++++++++++++++++-
> >   1 file changed, 193 insertions(+), 2 deletions(-)
> > 
> > diff --git a/xen/arch/arm/domain_build.c b/xen/arch/arm/domain_build.c
> > index 02a7f94..b4f560f 100644
> > --- a/xen/arch/arm/domain_build.c
> > +++ b/xen/arch/arm/domain_build.c
> > @@ -1360,7 +1360,194 @@ static int handle_node(struct domain *d, struct
> > kernel_info *kinfo,
> >       return res;
> >   }
> >   -static int prepare_dtb(struct domain *d, struct kernel_info *kinfo)
> > +static int make_gic_domU_node(const struct domain *d, void *fdt, int
> > addrcells, int sizecells)
> > +{
> > +    int res = 0;
> > +    int reg_size = addrcells + sizecells;
> > +    int nr_cells = reg_size * 2;
> > +    __be32 reg[nr_cells]; > +    __be32 *cells;
> > +
> > +    res = fdt_begin_node(fdt, "interrupt-controller");
> 
> Per the DT spec, node name should contain @unit-address when a "reg" property
> present. "unit-address" been the first address of the "reg" property.
> 
> Per the DT spec, node should have @base when a "regs" property is present.

OK, I'll add it


> > +    if ( res )
> > +        return res;
> > +
> > +    res = fdt_property_cell(fdt, "#address-cells", 0);
> > +    if ( res )
> > +        return res;
> > +
> > +    res = fdt_property_cell(fdt, "#interrupt-cells", 3);
> > +    if ( res )
> > +        return res;
> > +
> > +    res = fdt_property(fdt, "interrupt-controller", NULL, 0);
> > +    if ( res )
> > +        return res;
> > +
> > +    if (gic_hw_version() == GIC_V3)
> 
> Could we use a switch please? This would allow to catch easily new hardware
> version.

Sure


> > +    {
> > +        const uint64_t gicd_base = GUEST_GICV3_GICD_BASE;
> > +        const uint64_t gicd_size = GUEST_GICV3_GICD_SIZE;
> > +        const uint64_t gicr0_base = GUEST_GICV3_GICR0_BASE;
> > +        const uint64_t gicr0_size = GUEST_GICV3_GICR0_SIZE;
> > +
> > +        res = fdt_property_string(fdt, "compatible", "arm,gic-v3");
> > +        if ( res )
> > +            return res;
> > +
> > +        cells = &reg[0];
> > +        dt_child_set_range(&cells, addrcells, sizecells, gicd_base,
> > gicd_size);
> > +        dt_child_set_range(&cells, addrcells, sizecells, gicr0_base,
> > gicr0_size);
> > +        res = fdt_property(fdt, "reg", reg, sizeof(reg));
> > +    }
> > +    else if (gic_hw_version() == GIC_V3)
> 
> I think this should be GIC_V2 here.

Yep


> > +    {
> > +        const uint64_t gicd_base = GUEST_GICD_BASE;
> > +        const uint64_t gicd_size = GUEST_GICD_SIZE;
> > +        const uint64_t gicc_base = GUEST_GICC_BASE;
> > +        const uint64_t gicc_size = GUEST_GICC_SIZE;
> > +
> > +        res = fdt_property_string(fdt, "compatible", "arm,cortex-a15-gic");
> 
> I know that we use that property in libxl. But this is a bit weird to use it
> for Arm64 guest :). It would be better to use arm,gic-400 here.

OK

> > +        if ( res )
> > +            return res;
> > +
> > +        cells = &reg[0];
> > +        dt_child_set_range(&cells, addrcells, sizecells, gicd_base,
> > gicd_size);
> > +        dt_child_set_range(&cells, addrcells, sizecells, gicc_base,
> > gicc_size);
> > +    }
> > +
> > +    res = fdt_property(fdt, "reg", reg, sizeof(reg));
> > +    if (res)
> > +        return res;
> > +
> > +    res = fdt_property_cell(fdt, "linux,phandle", PHANDLE_GIC);
> > +    if (res)
> > +        return res;
> > +
> > +    res = fdt_property_cell(fdt, "phandle", PHANDLE_GIC);
> > +    if (res)
> > +        return res;
> > +
> > +    res = fdt_end_node(fdt);
> > +
> > +    return res;
> > +}
> > +
> > +static int make_timer_domU_node(const struct domain *d, void *fdt)
> > +{
> > +    int res;
> > +    gic_interrupt_t intrs[3];
> > +
> > +    res = fdt_begin_node(fdt, "timer");
> > +    if ( res )
> > +        return res;
> > +
> > +    if (!is_64bit_domain(d))
> > +    {
> > +        res = fdt_property_string(fdt, "compatible", "arm,armv7-timer");
> > +        if ( res )
> > +            return res;
> > +    } else {
> > +        res = fdt_property_string(fdt, "compatible", "arm,armv8-timer");
> > +        if ( res )
> > +            return res;
> > +    }
> > +
> > +    set_interrupt_ppi(intrs[0], GUEST_TIMER_PHYS_S_PPI, 0xf,
> > DT_IRQ_TYPE_LEVEL_LOW);
> > +    set_interrupt_ppi(intrs[1], GUEST_TIMER_PHYS_NS_PPI, 0xf,
> > DT_IRQ_TYPE_LEVEL_LOW);
> > +    set_interrupt_ppi(intrs[2], GUEST_TIMER_VIRT_PPI, 0xf,
> > DT_IRQ_TYPE_LEVEL_LOW);
> > +
> > +    res = fdt_property(fdt, "interrupts", intrs, sizeof (intrs[0]) * 3);
> > +    if ( res )
> > +        return res;
> > +
> > +    res = fdt_property_cell(fdt, "interrupt-parent",
> > +                            PHANDLE_GIC);
> > +    if (res)
> > +        return res;
> > +
> > +    res = fdt_end_node(fdt);
> > +    return res;
> > +}
> > +
> > +#define DOMU_DTB_SIZE 4096
> 
> Is this going to be enough? Per the documentation, the maximum size of a DT is
> 2MB.

That is what we use today for the small generated DTB for ACPI dom0.
Certainly it is enough memory from my empirical tests so far.  I suggest
to keep as is in this series, then potentially change it when we
introduce larger DomU DTBs (device assignment might require it, for
example).


> > +static int prepare_dtb_domU(struct domain *d, struct kernel_info *kinfo)
> > +{
> > +    int addrcells, sizecells;
> > +    int ret;
> > +
> > +    addrcells = dt_child_n_addr_cells(dt_host);
> > +    sizecells = dt_child_n_size_cells(dt_host);
> > +
> > +    kinfo->fdt = xmalloc_bytes(DOMU_DTB_SIZE);
> > +    if ( kinfo->fdt == NULL )
> > +        return -ENOMEM;
> > +
> > +    ret = fdt_create(kinfo->fdt, DOMU_DTB_SIZE);
> > +    if ( ret < 0 )
> > +        goto err;
> > +
> > +    ret = fdt_finish_reservemap(kinfo->fdt);
> > +    if ( ret < 0 )
> > +        goto err;
> > +
> > +    ret = fdt_begin_node(kinfo->fdt, "/");
> > +    if ( ret < 0 )
> > +        goto err;
> > +
> > +    ret = fdt_property_cell(kinfo->fdt, "#address-cells", addrcells);
> > +    if ( ret )
> > +        goto err;
> > +
> > +    ret = fdt_property_cell(kinfo->fdt, "#size-cells", sizecells);
> > +    if ( ret )
> > +        goto err;
> > +
> > +    ret = make_chosen_node(kinfo);
> > +    if ( ret )
> > +        goto err;
> > +
> > +    ret = make_hypervisor_node(d, kinfo, addrcells, sizecells);
> 
> make_hypervisor_node() will allocate an PPIfor the event channel based on
> what's available. However, the PPI for the guests is static and the timer is
> allocated afterwards.
> 
> So you probably want to rework make_hypervisor_node to pass the interrupt in
> parameter.

I'll fix


> > +    if ( ret )
> > +        goto err;
> > +
> > +    ret = make_psci_node(kinfo->fdt, NULL);
> > +    if ( ret )
> > +        goto err;
> > +
> > +    ret = make_cpus_node(d, kinfo->fdt, NULL);
> > +    if ( ret )
> > +        goto err;
> > +
> > +    ret = make_memory_node(d, kinfo->fdt, addrcells, sizecells, kinfo);
> > +    if ( ret )
> > +        goto err;
> > +
> > +    ret = make_gic_domU_node(d, kinfo->fdt, addrcells, sizecells);
> > +    if ( ret )
> > +        goto err;
> > +
> > +    ret = make_timer_domU_node(d, kinfo->fdt);
> > +    if ( ret )
> > +        goto err;
> > +
> > +    ret = fdt_end_node(kinfo->fdt);
> > +    if ( ret < 0 )
> > +        goto err;
> > +
> > +    ret = fdt_finish(kinfo->fdt);
> > +    if ( ret < 0 )
> > +        goto err;
> > +
> > +    return 0;
> > +
> > +  err:
> > +    printk("Device tree generation failed (%d).\n", ret);
> > +    xfree(kinfo->fdt);
> > +    return -EINVAL;
> > +}
> > +
> > +static int prepare_dtb_dom0(struct domain *d, struct kernel_info *kinfo)
> >   {
> >       const p2m_type_t default_p2mt = p2m_mmio_direct_c;
> >       const void *fdt;
> > @@ -2221,6 +2408,10 @@ int __init construct_domU(struct domain *d, struct
> > dt_device_node *node)
> >       d->arch.type = kinfo.type;
> >       allocate_memory(d, &kinfo);
> >   +    rc = prepare_dtb_domU(d, &kinfo);
> > +    if ( rc < 0 )
> > +        return rc;
> > +
> >       return __construct_domain(d, &kinfo);
> >   }
> >   @@ -2270,7 +2461,7 @@ int __init construct_dom0(struct domain *d)
> >       d->arch.type = kinfo.type;
> >         if ( acpi_disabled )
> > -        rc = prepare_dtb(d, &kinfo);
> > +        rc = prepare_dtb_dom0(d, &kinfo);
> >       else
> >           rc = prepare_acpi(d, &kinfo);
> >   
> 
> Cheers,
> 
> -- 
> Julien Grall
> 

_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxxxxxxxxx
https://lists.xenproject.org/mailman/listinfo/xen-devel