Disable asserts if force flag enabled

[bmx666]

I have some dtb to analyze from one of embedded device, it contains some errors, but during conversation from dtb to dts tool throws assert.

dtc -f -I dtb -O dts broken.dtb > broken.dts

dtc: livetree.c:437: propval_cell: Assertion `prop->val.len == sizeof(cell_t)' failed.

Could you please ignore assert if "force" has been enabled?

cell_t propval_cell(struct property *prop)
{
        if (!force)
                assert(prop->val.len == sizeof(cell_t));
        return fdt32_to_cpu(*((fdt32_t *)prop->val.val));
}

cell_t propval_cell_n(struct property *prop, unsigned int n)
{
        if (!force)
                assert(prop->val.len / sizeof(cell_t) > n);
        return fdt32_to_cpu(*((fdt32_t *)prop->val.val + n));
}

[dgibson]

@bmx666 that's certainly a bug, but the proposed solution isn't quite right. The problem lies in the caller of propval_cell() - it shouldn't be calling that on untrusted input without first verifying that the property has the expected form.

Can you attach the dtb causing the problem, then I can work out exactly what path is causing this and fix it.

[bmx666]

Spreadtrum_SC7731e.dtb.zip

[dgibson]

I've had a preliminary look and it sems like the problem is in the graph checks. As an interim approach, you should be able to disable those with -Wno-graph_nodes.

[bmx666]

Probably I'm not the first and not the last. The most people use dtc from package delivery system in OS. Because I had experience early it took me 5 min to find and fix this issue to generate at least some dts. My question, do you have plans to fix it or just left it as-is? If you don't want to fix it, just close the current issue.

[dgibson]

Sorry, I wasn't clear. Yes I plan to fix this, probably soon - it doesn't look too hard. But, it will be at least a few days, probably not until the new year. In the meantime you can disable the graph_nodes check as a workaround.