The "by8" CTR AVX implementation fails to propperly handle counter
overflows. That was the reason it got disabled in commit 7da4b29d496b
("crypto: aesni - disable "by8" AVX CTR optimization").
Fix the overflow handling by incrementing the counter block as a double
quad word, i.e. a 128 bit, and testing for overflows afterwards. We need
to use VPTEST to do so as VPADD* does not set the flags itself and
silently drops the carry bit.
As this change adds branches to the hot path, minor performance
regressions might be a side effect. But, OTOH, we now have a conforming
implementation -- the preferable goal.
A tcrypt test on a SandyBridge system (i7-2620M) showed almost identical
numbers for the old and this version with differences within the noise
range. A dm-crypt test with the fixed version gave even slightly better
results for this version. So the performance impact might not be as big
as expected.
Tested-by: Romain Francoise <rom...@orebokech.com>
Signed-off-by: Mathias Krause <mini...@googlemail.com>
Cc: Chandramouli Narayanan <mo...@linux.intel.com>
---
arch/x86/crypto/aes_ctrby8_avx-x86_64.S | 17 +++++++++++++++--
1 file changed, 15 insertions(+), 2 deletions(-)
diff --git a/arch/x86/crypto/aes_ctrby8_avx-x86_64.S
b/arch/x86/crypto/aes_ctrby8_avx-x86_64.S
index f091f122ed24..a029bc744244 100644
--- a/arch/x86/crypto/aes_ctrby8_avx-x86_64.S
+++ b/arch/x86/crypto/aes_ctrby8_avx-x86_64.S
@@ -108,6 +108,10 @@
byteswap_const:
.octa 0x000102030405060708090A0B0C0D0E0F
+ddq_low_msk:
+ .octa 0x0000000000000000FFFFFFFFFFFFFFFF
+ddq_high_add_1:
+ .octa 0x00000000000000010000000000000000
ddq_add_1:
.octa 0x00000000000000000000000000000001
ddq_add_2:
@@ -169,7 +173,12 @@ ddq_add_8:
.rept (by - 1)
club DDQ_DATA, i
club XDATA, i
- vpaddd var_ddq_add(%rip), xcounter, var_xdata
+ vpaddq var_ddq_add(%rip), xcounter, var_xdata
+ vptest ddq_low_msk(%rip), var_xdata
+ jnz 1f
+ vpaddq ddq_high_add_1(%rip), var_xdata, var_xdata
+ vpaddq ddq_high_add_1(%rip), xcounter, xcounter
+ 1:
vpshufb xbyteswap, var_xdata, var_xdata
.set i, (i +1)
.endr
@@ -178,7 +187,11 @@ ddq_add_8:
vpxor xkey0, xdata0, xdata0
club DDQ_DATA, by
- vpaddd var_ddq_add(%rip), xcounter, xcounter
+ vpaddq var_ddq_add(%rip), xcounter, xcounter
+ vptest ddq_low_msk(%rip), xcounter
+ jnz 1f
+ vpaddq ddq_high_add_1(%rip), xcounter, xcounter
+ 1:
.set i, 1
.rept (by - 1)
--
1.7.10.4
--
To unsubscribe from this list: send the line "unsubscribe linux-crypto" in
the body of a message to majord...@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
