Messages in this thread | | | Date | Fri, 10 Feb 2023 09:09:46 -0800 | Subject | Re: [PATCH v9 24/27] virt: gunyah: Add proxy-scheduled vCPUs | From | Elliot Berman <> |
| |
On 2/9/2023 10:54 PM, Srivatsa Vaddagiri wrote: > * Srivatsa Vaddagiri <quic_svaddagi@quicinc.com> [2023-02-09 16:10:06]: > >> * Elliot Berman <quic_eberman@quicinc.com> [2023-01-20 14:46:23]: >> >>> +static void gh_handle_mmio_return(struct gunyah_vcpu *vcpu, u64 *state) >>> +{ >>> + if (!vcpu->vcpu_run->mmio.is_write) >>> + memcpy(&state[0], vcpu->vcpu_run->mmio.data, vcpu->vcpu_run->mmio.len); >> >> Would be good to do a bound check on length of memcpy I think (in case >> vcpu_run_resp->state_data[1] is wrong for example).
I think this is the right place to ensure right value is passed. Right now, vcpu_run->mmio.len is user accessible and easily provides ability for buffer overrun if unchecked. I'm now preserving the length Gunyah requested in internal vcpu structure. Userspace can write whatever they want to the mmio request, but it won't change the number of bytes that we copy from user.
> > Sorry I meant to comment that for the other memcpy you had! > > + memcpy(vcpu->vcpu_run->mmio.data, &vcpu_run_resp->state_data[2], > + vcpu_run_resp->state_data[1]); > >
Per Srini and Alex, we can trust the values hypervisor sends, so I think we can skip this check.
https://lore.kernel.org/all/4a584563-1fb7-22fa-5e16-e0cf5e88b76b@linaro.org/
Thanks, Elliot
| |