Re: [Qemu-devel] [PULL 23/44] target/ppc: Use vector variable shifts for VSL, VSR, VSRA

[David Gibson <david@gibson.dropbear.id.au>]

[-- Attachment #1: Type: text/plain, Size: 3449 bytes --]

On Fri, Jun 07, 2019 at 09:28:49AM -0500, Richard Henderson wrote:
> On 6/7/19 9:09 AM, Laurent Vivier wrote:
> > On 07/06/2019 11:29, Laurent Vivier wrote:
> >> On 29/05/2019 08:49, David Gibson wrote:
> >>> From: Richard Henderson <richard.henderson@linaro.org>
> >>>
> >>> The gvec expanders take care of masking the shift amount
> >>> against the element width.
> >>>
> >>> Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
> >>> Message-Id: <20190518191430.21686-2-richard.henderson@linaro.org>
> >>> Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
> >>> ---
> >>>  target/ppc/helper.h                 | 12 ----------
> >>>  target/ppc/int_helper.c             | 37 -----------------------------
> >>>  target/ppc/translate/vmx-impl.inc.c | 24 +++++++++----------
> >>>  3 files changed, 12 insertions(+), 61 deletions(-)
> >>
> >> This patch introduces a regressions
> >>  with Fedora 29 guest:
> >>
> >> - during kernel boot:
> >>
> >> [   40.397876] crypto_register_alg 'aes' = 0
> >> [   40.577517] crypto_register_alg 'cbc(aes)' = 0
> >> [   40.743576] crypto_register_alg 'ctr(aes)' = 0
> >> [   41.061379] alg: skcipher: Test 1 failed (invalid result) on encryption for p8_aes_xts
> >> [   41.062054] 00000000: 91 7c f6 9e bd 68 b2 ec 9b 9f e9 a3 ea dd a6 92
> >> [   41.062163] 00000010: 98 10 35 57 5e dc 36 1e 9a f7 bc ba 39 f2 5c eb
> >> [   41.062834] crypto_register_alg 'xts(aes)' = 0
> >> [   41.077358] alg: hash: Test 2 failed for p8_ghash
> >> [   41.077553] 00000000: 5f 89 ab f7 20 57 20 57 20 57 20 57 20 57 20 57
> >>
> >> - with libssl:
> >>
> >> # curl -o /dev/null https://www.google.com
> >>   % Total    % Received % Xferd  Average Speed   Time    Time     Time  Current
> >>                                  Dload  Upload   Total   Spent    Left  Speed
> >>   0     0    0     0    0     0      0      0 --:--:-- --:--:-- --:--:--     0
> >> curl: (35) error:1408F119:SSL routines:ssl3_get_record:decryption failed or bad record mac
> >>
> >> [before, this one fails with:
> >> curl: (35) error:04091068:rsa routines:int_rsa_verify:bad signature ]
> >>
> >> If I revert this patch on top of 0d74f3b427 + "target/ppc: Fix lxvw4x, lxvh8x and lxvb16x", all works fine.
> >>
> >> Thanks,
> >> Laurent
> >>
> > 
> > This seems to fix the problem:
> > 
> > diff --git a/accel/tcg/tcg-runtime-gvec.c b/accel/tcg/tcg-runtime-gvec.c
> > index 3b6052fe97..6f0709b307 100644
> > --- a/accel/tcg/tcg-runtime-gvec.c
> > +++ b/accel/tcg/tcg-runtime-gvec.c
> > @@ -874,7 +874,7 @@ void HELPER(gvec_sar8v)(void *d, void *a, void *b,
> > uint32_t desc)
> >      intptr_t oprsz = simd_oprsz(desc);
> >      intptr_t i;
> > 
> > -    for (i = 0; i < oprsz; i += sizeof(vec8)) {
> > +    for (i = 0; i < oprsz; i += sizeof(uint8_t)) {
> >          uint8_t sh = *(uint8_t *)(b + i) & 7;
> >          *(int8_t *)(d + i) = *(int8_t *)(a + i) >> sh;
> >      }
> 
> Grr.  I really really need to come up with a solution for testing that allows
> me to test paths that the host cpu would not ordinarily take.  This bug is
> hidden on a host with AVX2.
> 
> Thanks for the digging.

Can one of you send this fix formally with a S-o-b and so forth?

-- 
David Gibson			| I'll have my music baroque, and my code
david AT gibson.dropbear.id.au	| minimalist, thank you.  NOT _the_ _other_
				| _way_ _around_!
http://www.ozlabs.org/~dgibson

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 833 bytes --]