Question : PLIC address check at fdt node parse

[vmorales-oc]

Hi,

I would like to ask about a check that is performed during FDT parsing, specifically for the PLIC node. The check in question is:

if (rc < 0 || !reg_addr || !reg_size)
                return SBI_ENODEV;

This code is located in lib/utils/fdt/fdt_helper.c:

int fdt_parse_plic_node(const void *fdt, int nodeoffset, struct plic_data *plic)
{
        int len, rc;
        const fdt32_t *val;
        uint64_t reg_addr, reg_size;

        if (nodeoffset < 0 || !plic || !fdt)
                return SBI_ENODEV;

        rc = fdt_get_node_addr_size(fdt, nodeoffset, 0,
                                    &reg_addr, &reg_size);
        if (rc < 0 || !reg_addr || !reg_size)
                return SBI_ENODEV;
        plic->addr = reg_addr;
        plic->size = reg_size;

        val = fdt_getprop(fdt, nodeoffset, "riscv,ndev", &len);
        if (len > 0)
                plic->num_src = fdt32_to_cpu(*val);

        return 0;
}

My question concerns the need to check reg_addr != 0. This condition causes an error when the PLIC base address is 0x0.

I am aware that placing the PLIC at address 0x0 is generally not considered good practice, but has this possibility been explicitly considered? As far as I know, there is no restriction in the PLIC specification that forbids a base address of zero.

Thank you in advance.

[avpatel]

I agree, the reg_addr != 0 is paranoid check. Feel free to send patch to OpenSBI mailing list.