x86_64: Fix alignment bug in _nettle_ghash_update.

diff --git a/ChangeLog b/ChangeLog
index 507e0438cfd93a2940b8a13558ae3a539efc7c1d..6312ac00761fb85d6de1362726e529a08b45cb1a 100644 (file)
--- a/ChangeLog
+++ b/ChangeLog
@@ -1,3 +1,10 @@
+2023-05-16  Niels Möller  <nisse@lysator.liu.se>
+
+       * x86_64/ghash-update.asm: Use separate unaligned load
+       instructions (movups) to load the tabulated values, since they are
+       only 8-byte aligned and pand memory operands require 16-byte
+       alignment.
+
 2023-05-15  Niels Möller  <nisse@lysator.liu.se>
 
        * eccdata.c (output_bignum_redc): Add missing mpz_clear, reported

diff --git a/x86_64/ghash-update.asm b/x86_64/ghash-update.asm
index b3417e45349da4abd7dcd0c206b497b8fb24ca40..0a72f071d980841150500c5f640730ee20675638 100644 (file)
--- a/x86_64/ghash-update.asm
+++ b/x86_64/ghash-update.asm
@@ -44,6 +44,8 @@ define(`M0', `%xmm2')
 define(`M1', `%xmm3')
 define(`M2', `%xmm4')
 define(`M3', `%xmm5')
+define(`T0', `%xmm6')
+define(`T1', `%xmm7')
 
        .file "ghash-update.asm"
 
@@ -54,7 +56,7 @@ define(`M3', `%xmm5')
        .text
        ALIGN(16)
 PROLOGUE(_nettle_ghash_update)
-       W64_ENTRY(4, 6)
+       W64_ENTRY(4, 8)
        sub     $1, BLOCKS
        movups  (XP), X
        jc      .Ldone
@@ -77,10 +79,18 @@ ALIGN(16)
        pshufd  $0xaa, M3, M2
        pshufd  $0xff, M3, M3
        pslld   $1, X
-       pand    (KEY, CNT), M0
-       pand    (KEY32, CNT), M1
-       pand    16(KEY, CNT), M2
-       pand    16(KEY32, CNT), M3
+       C Tabulated values are only 8-byte aligned, and therefore
+       C can't be used as memory operands to pand. It would be nice
+       C if we could force 16-byte alignment on nettle_block16, using
+       C C11 alignas.
+       movups  (KEY, CNT), T0
+       movups  (KEY32, CNT), T1
+       pand    T0, M0
+       pand    T1, M1
+       movups  16(KEY, CNT), T0
+       movups  16(KEY32, CNT), T1
+       pand    T0, M2
+       pand    T1, M3
        pxor    M0, M1
        pxor    M2, M3
        pxor    M1, R
@@ -98,6 +108,6 @@ ALIGN(16)
 .Ldone:
        movups  X, (XP)
        mov     SRC, %rax
-       W64_EXIT(4, 6)
+       W64_EXIT(4, 8)
        ret
 EPILOGUE(_nettle_ghash_update)