Messages in this thread | | | Date | Mon, 6 Feb 2023 09:38:29 -0800 | Subject | Re: [PATCH v9 10/27] gunyah: rsc_mgr: Add VM lifecycle RPC | From | Elliot Berman <> |
| |
On 2/6/2023 7:41 AM, Alex Elder wrote: > On 2/2/23 6:46 AM, Srinivas Kandagatla wrote: >>> + ret = gh_rm_call(rm, message_id, &req_payload, >>> sizeof(req_payload), &resp, &resp_size); >>> + if (!ret && resp_size) { >> >> Am struggling to understand these type of checks in success case, when >> a command is not expecting any response why are we checking for >> response here, This sounds like a bug in either RM or hypervisor. >> >> Or Is this something that happens due to some firmware behaviour? >> Could you elobrate on this. > > What I think you're talking about is error checking even when > it's very clear something "can't happen." It's a pattern I've > seen in Qualcomm downstream code, and I believe sometimes it > is done as "best practice" to avoid warnings from security scans. > (I might be wrong about this though.)
That's right reasoning.
> > I think your underlying point though is that we can just assume > success means "truly successful," so there's no reason to do any > additional sanity checks. We *assume* the hardware is doing the > correct thing (if it's not, we might as well assume it does > *nothing* right). > > So as a very general statement, I think all checks of this type > should go away (and I think Srini would agree). >
I'll remove the checks.
| |