mirror/openssl
2
0
Fork 0
mirror of https://github.com/openssl/openssl.git synced 2024-11-21 01:15:20 +08:00
Code Issues Packages Projects Releases Wiki Activity

LoongArch64 assembly pack: Fix ChaCha20 ABI breakage

Browse Source 
The [LP64D ABI][1] requires the floating-point registers f24-f31
(aka fs0-fs7) callee-saved.  The low 64 bits of a LSX/LASX vector
register aliases with the corresponding FPR, so we must save and restore
the callee-saved FPR when we writes into the corresponding vector
register.

This ABI breakage can be easily demonstrated by injecting the use of a
saved FPR into the test in bio_enc_test.c:

    static int test_bio_enc_chacha20(int idx)
    {
        register double fs7 asm("f31") = 114.514;
        asm("#optimize barrier":"+f"(fs7));
        return do_test_bio_cipher(EVP_chacha20(), idx) && fs7 == 114.514;
    }

So fix it.  To make the logic simpler, jump into the scalar
implementation earlier when LSX and LASX are not enumerated in AT_HWCAP,
or the input is too short.

[1]: https://github.com/loongson/la-abi-specs/blob/v2.20/lapcs.adoc#floating-point-registers

Reviewed-by: Neil Horman <nhorman@openssl.org>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/22817)
This commit is contained in:
Xi Ruoyao 2023-11-25 17:53:57 +08:00 committed by Tomas Mraz
parent dfd986b6f5
commit b46de72c26
1 changed files with 36 additions and 10 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

46
crypto/chacha/asm/chacha-loongarch64.pl
View File

@ -17,6 +17,14 @@ my ($a0,$a1,$a2,$a3,$a4,$a5,$a6,$a7)=map("\$r$_",(4..11));
my ($t0,$t1,$t2,$t3,$t4,$t5,$t6,$t7,$t8,$x)=map("\$r$_",(12..21));
my ($s0,$s1,$s2,$s3,$s4,$s5,$s6,$s7,$s8)=map("\$r$_",(23..31));
# The saved floating-point registers in the LP64D ABI. In LoongArch
# with vector extension, the low 64 bits of a vector register alias with
# the corresponding FPR. So we must save and restore the corresponding
# FPR if we'll write into a vector register. The ABI only requires
# saving and restoring the FPR (i.e. 64 bits of the corresponding vector
# register), not the entire vector register.
my ($fs0,$fs1,$fs2,$fs3,$fs4,$fs5,$fs6,$fs7)=map("\$f$_",(24..31));
# Here is the 128-bit vector register layout for LSX extension.
my ($vr0,$vr1,$vr2,$vr3,$vr4,$vr5,$vr6,$vr7,$vr8,$vr9,$vr10,
$vr11,$vr12,$vr13,$vr14,$vr15,$vr16,$vr17,$vr18,$vr19,
@ -66,13 +74,25 @@ ChaCha20_ctr32:
la.pcrel $t0,OPENSSL_loongarch_hwcap_P
ld.w $t0,$t0,0
bleu $len,$t3,.LChaCha20_1x # goto 1x when len <= 64
andi $t0,$t0,LOONGARCH_HWCAP_LASX | LOONGARCH_HWCAP_LSX
beqz $t0,.LChaCha20_1x
addi.d $sp,$sp,-64
fst.d $fs0,$sp,0
fst.d $fs1,$sp,8
fst.d $fs2,$sp,16
fst.d $fs3,$sp,24
fst.d $fs4,$sp,32
fst.d $fs5,$sp,40
fst.d $fs6,$sp,48
fst.d $fs7,$sp,56
andi $t1,$t0,LOONGARCH_HWCAP_LASX
bnez $t1,.LChaCha20_8x
andi $t2,$t0,LOONGARCH_HWCAP_LSX
bnez $t2,.LChaCha20_4x
b .LChaCha20_1x
b .LChaCha20_4x
EOF
@ -442,8 +462,6 @@ $code .= <<EOF;
.align 6
.LChaCha20_4x:
ori $t3,$zero,64
bleu $len,$t3,.LChaCha20_1x # goto 1x when len <= 64
addi.d $sp,$sp,-128
# Save the initial block counter in $t4
@ -783,7 +801,7 @@ $code .= <<EOF;
.Ldone_4x:
addi.d $sp,$sp,128
b .Lend
b .Lrestore_saved_fpr
EOF
}
@ -869,8 +887,6 @@ $code .= <<EOF;
.align 6
.LChaCha20_8x:
ori $t3,$zero,64
bleu $len,$t3,.LChaCha20_1x # goto 1x when len <= 64
addi.d $sp,$sp,-128
# Save the initial block counter in $t4
@ -1394,12 +1410,22 @@ $code .= <<EOF;
.Ldone_8x:
addi.d $sp,$sp,128
b .Lend
b .Lrestore_saved_fpr
EOF
}
$code .= <<EOF;
.Lrestore_saved_fpr:
fld.d $fs0,$sp,0
fld.d $fs1,$sp,8
fld.d $fs2,$sp,16
fld.d $fs3,$sp,24
fld.d $fs4,$sp,32
fld.d $fs5,$sp,40
fld.d $fs6,$sp,48
fld.d $fs7,$sp,56
addi.d $sp,$sp,64
.Lno_data:
.Lend:
jr $ra