Re: [PATCH] arch:x86:coco:sev: Initialize ctxt variable

["Ragavendra B.N." <ragavendra.bn@gmail.com>]

On Mon, Nov 18, 2024 at 08:53:04AM -0600, Tom Lendacky wrote:
> On 11/18/24 08:44, Tom Lendacky wrote:
> > On 11/15/24 16:55, Ard Biesheuvel wrote:
> >> On Fri, 15 Nov 2024 at 20:53, Ragavendra B.N. <ragavendra.bn@gmail.com> wrote:
> >>>
> >>> On Fri, Nov 15, 2024 at 12:02:27PM +0100, Ard Biesheuvel wrote:
> >>>> On Fri, 15 Nov 2024 at 12:01, Ingo Molnar <mingo@kernel.org> wrote:
> >>>>>
> >>>>>
> >>>>> * Ragavendra <ragavendra.bn@gmail.com> wrote:
> >>>>>
> >>>>>> Updating the ctxt value to NULL in the svsm_perform_ghcb_protocol as
> >>>>>> it was not initialized.
> >>>>>>
> >>>>>> Fixes: 2e1b3cc9d7f7 (grafted) Merge tag 'arm-fixes-6.12-2' of git://git.kernel.org/pub/scm/linux/kernel/git/soc/soc
> >>>>>
> >>>>> This 'Fixes' tag looks bogus.
> >>>>>
> >>>>
> >>>> So does the patch itself - 'struct es_em_ctxt ctxt' is not a pointer.
> >>> Thank you very much for your response. I am relatively new to kernel development.
> >>>
> >>> I know we can use kmalloc for memory allocation. Please advice.
> >>>
> >>> struct es_em_ctxt ctxt = kmalloc(sizeof(struct es_em_ctxt), GFP_KERNEL);
> >>>
> >>> I am thinking to update like above, but like you mentioned, ctxt is not a pointer. I can update this to be a pointer if needed.
> >>>
> >>
> >> The code is fine as is. Let's end this thread here, shall we?
> > 
> > I was assuming he got some kind of warning from some compiler options or
> > a static checker. Is that the case Ragavendra?
> > 
> > When I look at the code, it is possible for ctxt->fi.error_code to be
> > left uninitialized. The simple fix is to just initialize ctxt as:
> > 
> > 	struct es_em_ctxt ctxt = {};
> 
> Although to cover all cases now and going forwared, the es_em_ctxt fi
> member should just be zeroed in verify_exception_info() when
> ES_EXCEPTION is going to be returned.
> 
> Thanks,
> Tom
> 
> > 
> > Thanks,
> > Tom

Yes Tom, that is exactly the reason I worked on it the first place. The issue was reported by the Coverity tool.

I can send the below fix if that is fine.
> > 	struct es_em_ctxt ctxt = {};

For the es_em_ctxt fi member to be zeroed, I can go ahead and assign 0 to all the three long members like below in verify_exception_info()


	if (info & SVM_EVTINJ_VALID_ERR) {
		ctxt->fi.error_code = info >> 32;
	} else {
		ctxt->fi.error_code = 0;
		ctxt->fi.vector = 0;
		ctxt->fi.cr2 = 0;

	}

return ES_EXCEPTION;

Thanks,
Ragavendra N.