On Wed, 2023-12-13 at 18:00 +0100, Claudio Imbrenda wrote:
> On Wed, 13 Dec 2023 13:49:42 +0100
> Nina Schoetterl-Glausch <nsg@xxxxxxxxxxxxx> wrote:
>
> > The STFLE instruction indicates installed facilities.
> > SIE can interpretively execute STFLE.
> > Use a snippet guest executing STFLE to get the result of
> > interpretive execution and check the result.
> >
> > Signed-off-by: Nina Schoetterl-Glausch <nsg@xxxxxxxxxxxxx>
>
> [...]
>
> > static inline void setup_facilities(void)
> > diff --git a/s390x/snippets/c/stfle.c b/s390x/snippets/c/stfle.c
> > new file mode 100644
> > index 00000000..eb024a6a
> > --- /dev/null
> > +++ b/s390x/snippets/c/stfle.c
> > @@ -0,0 +1,26 @@
> > +/* SPDX-License-Identifier: GPL-2.0-only */
> > +/*
> > + * Copyright IBM Corp. 2023
> > + *
> > + * Snippet used by the STLFE interpretive execution facilities test.
> > + */
> > +#include <libcflat.h>
> > +#include <snippet-guest.h>
> > +
> > +int main(void)
> > +{
> > + const unsigned int max_fac_len = 8;
>
> why 8?
8 is a somewhat arbitrary, large number :)
I suppose I could choose an even larger one, maybe even PAGE_SIZE/8.
That would guarantee that max_fac_len >= stfle_size() (8 is enough for that today)
It's not necessary for max_fac_len >= stfle_size(), but probably good for
test coverage.
>
> > + uint64_t res[max_fac_len + 1];
> > +
> > + res[0] = max_fac_len - 1;
> > + asm volatile ( "lg 0,%[len]\n"
> > + " stfle %[fac]\n"
> > + " stg 0,%[len]\n"
> > + : [fac] "=QS"(*(uint64_t(*)[max_fac_len])&res[1]),
> > + [len] "+RT"(res[0])
> > + :
> > + : "%r0", "cc"
> > + );
> > + force_exit_value((uint64_t)&res);
> > + return 0;
> > +}
> > diff --git a/s390x/stfle-sie.c b/s390x/stfle-sie.c
> > new file mode 100644
> > index 00000000..574319ed
> > --- /dev/null
> > +++ b/s390x/stfle-sie.c
> > @@ -0,0 +1,132 @@
[...]
> > +
> > +static void test_stfle_format_0(void)
> > +{
> > + struct guest_stfle_res res;
> > +
> > + report_prefix_push("format-0");
> > + for (int j = 0; j < stfle_size(); j++)
> > + WRITE_ONCE((*fac)[j], rand64(&rand_s));
>
> do you really need random numbers? can't you use a static pattern?
> maybe something like 0x0001020304050607 etc...
Doesn't really need to be random, I need some arbitrary test pattern,
but I don't think some cumbersome constant literal improves anything.
The RNG is just initialized with the time, because why not.
>
> > + vm.sblk->fac = (uint32_t)(uint64_t)fac;
> > + res = run_guest();
> > + report(res.len == stfle_size(), "stfle len correct");
^ should be
+ report(res.len == min(stfle_size(), 8), "stfle len correct");
For the case that the guest buffer was shorter.
> > + report(!memcmp(*fac, res.mem, res.len * sizeof(uint64_t)),
> > + "Guest facility list as specified");
>
> it seems like you are comparing the full facility list (stfle_size
> doublewords long) with the result of STFLE in the guest, but the guest
> is limited to 8 double words?
Their prefixes must be the same. res.len is the guest length, so max 8 right now.
>
> > + report_prefix_pop();
> > +}
