ofw: fix memory leak in ofwbus

[christos]

Summary

Reported in:
https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=269509

[markj]

I think this is right, but I can't see why we are calling ofw_bus_gen_setup_devinfo() at all. simplebus_add_device() -> simplebus_setup_dinfo() -> ofw_bus_gen_setup_devinfo() does the same thing. I suspect that we could remove the setup_devinfo() call and check the return value of simplebus_add_device() instead.

[christos]

In D38903#885866, @markj wrote:

I think this is right, but I can't see why we are calling ofw_bus_gen_setup_devinfo() at all. simplebus_add_device() -> simplebus_setup_dinfo() -> ofw_bus_gen_setup_devinfo() does the same thing. I suspect that we could remove the setup_devinfo() call and check the return value of simplebus_add_device() instead.

No objection to this, I'm just not sure how I can test the code so that I can implement this.

[markj]

In D38903#885879, @christos wrote:

In D38903#885866, @markj wrote:

I think this is right, but I can't see why we are calling ofw_bus_gen_setup_devinfo() at all. simplebus_add_device() -> simplebus_setup_dinfo() -> ofw_bus_gen_setup_devinfo() does the same thing. I suspect that we could remove the setup_devinfo() call and check the return value of simplebus_add_device() instead.

No objection to this, I'm just not sure how I can test the code so that I can implement this.

The easiest way might be to simply try booting a riscv kernel with this change in QEMU: https://wiki.freebsd.org/riscv/QEMU

[christos]

Seems to be working fine.

[markj]

In D38903#894894, @christos wrote:

Seems to be working fine.

Could you please describe your test? Did you use a risc-v kernel and verify that this code path gets executed?

[christos]

Yes, I booted a riscv kernel and added a few printfs to make sure the code gets executed.