Re: [PATCH v5 16/17] KVM: arm64: Introduce pkvm_dump_backtrace()

[Oliver Upton <oliver.upton@linux.dev>]

Hi Kalesh,

On Wed, Jul 20, 2022 at 10:57:27PM -0700, Kalesh Singh wrote:

[...]

> +/*
> + * pkvm_dump_backtrace - Dump the protected nVHE HYP backtrace.
> + *
> + * @hyp_offset: hypervisor offset, used for address translation.
> + *
> + * Dumping of the pKVM HYP backtrace is done by reading the
> + * stack addresses from the shared stacktrace buffer, since the
> + * host cannot direclty access hyperviosr memory in protected
> + * mode.
> + */
> +static void pkvm_dump_backtrace(unsigned long hyp_offset)
> +{
> +	unsigned long *stacktrace_entry
> +		= (unsigned long *)this_cpu_ptr_nvhe_sym(pkvm_stacktrace);
> +	unsigned long va_mask, pc;
> +
> +	va_mask = GENMASK_ULL(vabits_actual - 1, 0);
> +
> +	kvm_err("Protected nVHE HYP call trace:\n");

This and the footer printks should be put in respective helpers to share
between pKVM and non-pKVM backtrace implementations. I imagine users
will invariably bake some pattern matching to scrape traces, and it
should be consistent between both flavors.

> +	/* The stack trace is terminated by a null entry */
> +	for (; *stacktrace_entry; stacktrace_entry++) {

At the point we're dumping the backtrace we know that EL2 has already
soiled itself, so we shouldn't explicitly depend on it providing NULL
terminators. I believe this loop should have an explicit range && NULL
check.

--
Thanks,
Oliver