diff options
Diffstat (limited to '')
-rw-r--r-- | src/spdk/intel-ipsec-mb/avx2/mb_mgr_hmac_sha_512_submit_avx2.asm | 416 |
1 files changed, 416 insertions, 0 deletions
diff --git a/src/spdk/intel-ipsec-mb/avx2/mb_mgr_hmac_sha_512_submit_avx2.asm b/src/spdk/intel-ipsec-mb/avx2/mb_mgr_hmac_sha_512_submit_avx2.asm new file mode 100644 index 000000000..a7c3e249b --- /dev/null +++ b/src/spdk/intel-ipsec-mb/avx2/mb_mgr_hmac_sha_512_submit_avx2.asm @@ -0,0 +1,416 @@ +;; +;; Copyright (c) 2012-2018, Intel Corporation +;; +;; Redistribution and use in source and binary forms, with or without +;; modification, are permitted provided that the following conditions are met: +;; +;; * Redistributions of source code must retain the above copyright notice, +;; this list of conditions and the following disclaimer. +;; * Redistributions in binary form must reproduce the above copyright +;; notice, this list of conditions and the following disclaimer in the +;; documentation and/or other materials provided with the distribution. +;; * Neither the name of Intel Corporation nor the names of its contributors +;; may be used to endorse or promote products derived from this software +;; without specific prior written permission. +;; +;; THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +;; AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +;; IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE +;; DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE +;; FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL +;; DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR +;; SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER +;; CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, +;; OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE +;; OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. +;; + +%include "include/os.asm" +%include "job_aes_hmac.asm" +%include "mb_mgr_datastruct.asm" +%include "include/reg_sizes.asm" +%include "include/memcpy.asm" +%include "include/const.inc" + +extern sha512_x4_avx2 + +section .data +default rel +align 16 +byteswap: ;ddq 0x08090a0b0c0d0e0f0001020304050607 + dq 0x0001020304050607, 0x08090a0b0c0d0e0f + +section .text + +%ifndef FUNC +%define FUNC submit_job_hmac_sha_512_avx2 +%define SHA_X_DIGEST_SIZE 512 +%endif + +%if 1 +%ifdef LINUX +%define arg1 rdi +%define arg2 rsi +%define reg3 rcx +%define reg4 rdx +%else +%define arg1 rcx +%define arg2 rdx +%define reg3 rdi +%define reg4 rsi +%endif + +%define state arg1 +%define job arg2 +%define len2 arg2 + + +; idx needs to be in rbp, r13, r14, r16 +%define last_len rbp +%define idx rbp + +%define p r11 +%define start_offset r11 + +%define unused_lanes rbx +%define tmp4 rbx + +%define job_rax rax +%define len rax + +%define size_offset reg3 +%define tmp2 reg3 + +%define lane reg4 +%define tmp3 reg4 + +%define extra_blocks r8 + +%define tmp r9 +%define p2 r9 + +%define lane_data r10 + +%endif + +; Define stack usage + +; we clobber rbx, rsi, rdi, rbp; called routine also clobbers r12 +struc STACK +_gpr_save: resq 5 +_rsp_save: resq 1 +endstruc + +; JOB* FUNC(MB_MGR_HMAC_sha_512_OOO *state, JOB_AES_HMAC *job) +; arg 1 : rcx : state +; arg 2 : rdx : job +MKGLOBAL(FUNC,function,internal) +FUNC: + + mov rax, rsp + sub rsp, STACK_size + and rsp, -32 + mov [rsp + _gpr_save + 8*0], rbx + mov [rsp + _gpr_save + 8*1], rbp + mov [rsp + _gpr_save + 8*2], r12 +%ifndef LINUX + mov [rsp + _gpr_save + 8*3], rsi + mov [rsp + _gpr_save + 8*4], rdi +%endif + mov [rsp + _rsp_save], rax ; original SP + + mov unused_lanes, [state + _unused_lanes_sha512] + movzx lane, BYTE(unused_lanes) + shr unused_lanes, 8 + imul lane_data, lane, _SHA512_LANE_DATA_size + lea lane_data, [state + _ldata_sha512 + lane_data] + mov [state + _unused_lanes_sha512], unused_lanes + mov len, [job + _msg_len_to_hash_in_bytes] + mov tmp, len + shr tmp, 7 ; divide by 128, len in terms of blocks + + mov [lane_data + _job_in_lane_sha512], job + mov dword [lane_data + _outer_done_sha512], 0 + + vmovdqa xmm0, [state + _lens_sha512] + XVPINSRW xmm0, xmm1, extra_blocks, lane, tmp, scale_x16 + vmovdqa [state + _lens_sha512], xmm0 + + + mov last_len, len + and last_len, 127 + lea extra_blocks, [last_len + 17 + 127] + shr extra_blocks, 7 + mov [lane_data + _extra_blocks_sha512], DWORD(extra_blocks) + + mov p, [job + _src] + add p, [job + _hash_start_src_offset_in_bytes] + mov [state + _args_data_ptr_sha512 + PTR_SZ*lane], p + + cmp len, 128 + jb copy_lt128 + +fast_copy: + add p, len + vmovdqu ymm0, [p - 128 + 0*32] + vmovdqu ymm1, [p - 128 + 1*32] + vmovdqu ymm2, [p - 128 + 2*32] + vmovdqu ymm3, [p - 128 + 3*32] + vmovdqu [lane_data + _extra_block_sha512 + 0*32], ymm0 + vmovdqu [lane_data + _extra_block_sha512 + 1*32], ymm1 + vmovdqu [lane_data + _extra_block_sha512 + 2*32], ymm2 + vmovdqu [lane_data + _extra_block_sha512 + 3*32], ymm3 +end_fast_copy: + + mov size_offset, extra_blocks + shl size_offset, 7 + sub size_offset, last_len + add size_offset, 128-8 + mov [lane_data + _size_offset_sha512], DWORD(size_offset) + mov start_offset, 128 + sub start_offset, last_len + mov [lane_data + _start_offset_sha512], DWORD(start_offset) + + lea tmp, [8*128 + 8*len] + bswap tmp + mov [lane_data + _extra_block_sha512 + size_offset], tmp + + mov tmp, [job + _auth_key_xor_ipad] + +%assign I 0 +%rep 4 + vmovdqu xmm0, [tmp + I * 2 * SHA512_DIGEST_WORD_SIZE] + vmovq [state + _args_digest_sha512 + SHA512_DIGEST_WORD_SIZE*lane + (2*I + 0)*SHA512_DIGEST_ROW_SIZE], xmm0 + vpextrq [state + _args_digest_sha512 + SHA512_DIGEST_WORD_SIZE*lane + (2*I + 1)*SHA512_DIGEST_ROW_SIZE], xmm0, 1 +%assign I (I+1) +%endrep + + test len, ~127 + jnz ge128_bytes + +lt128_bytes: + vmovdqa xmm0, [state + _lens_sha512] + XVPINSRW xmm0, xmm1, tmp, lane, extra_blocks, scale_x16 + vmovdqa [state + _lens_sha512], xmm0 + + lea tmp, [lane_data + _extra_block_sha512 + start_offset] + mov [state + _args_data_ptr_sha512 + PTR_SZ*lane], tmp ;; 8 to hold a UINT8 + mov dword [lane_data + _extra_blocks_sha512], 0 + +ge128_bytes: + cmp unused_lanes, 0xff + jne return_null + jmp start_loop + + align 16 +start_loop: + ; Find min length + vmovdqa xmm0, [state + _lens_sha512] + vphminposuw xmm1, xmm0 + vpextrw DWORD(len2), xmm1, 0 ; min value + vpextrw DWORD(idx), xmm1, 1 ; min index (0...1) + cmp len2, 0 + je len_is_0 + + vpshuflw xmm1, xmm1, 0x00 + vpsubw xmm0, xmm0, xmm1 + vmovdqa [state + _lens_sha512], xmm0 + + ; "state" and "args" are the same address, arg1 + ; len is arg2 + call sha512_x4_avx2 + ; state and idx are intact + +len_is_0: + ; process completed job "idx" + imul lane_data, idx, _SHA512_LANE_DATA_size + lea lane_data, [state + _ldata_sha512 + lane_data] + mov DWORD(extra_blocks), [lane_data + _extra_blocks_sha512] + cmp extra_blocks, 0 + jne proc_extra_blocks + cmp dword [lane_data + _outer_done_sha512], 0 + jne end_loop + +proc_outer: + mov dword [lane_data + _outer_done_sha512], 1 + mov DWORD(size_offset), [lane_data + _size_offset_sha512] + mov qword [lane_data + _extra_block_sha512 + size_offset], 0 + + vmovdqa xmm0, [state + _lens_sha512] + XVPINSRW xmm0, xmm1, tmp, idx, 1, scale_x16 + vmovdqa [state + _lens_sha512], xmm0 + + lea tmp, [lane_data + _outer_block_sha512] + mov job, [lane_data + _job_in_lane_sha512] + mov [state + _args_data_ptr_sha512 + PTR_SZ*idx], tmp + +%assign I 0 +%rep (SHA_X_DIGEST_SIZE / (8 * 16)) + vmovq xmm0, [state + _args_digest_sha512 + SHA512_DIGEST_WORD_SIZE*idx + (2*I + 0)*SHA512_DIGEST_ROW_SIZE] + vpinsrq xmm0, [state + _args_digest_sha512 + SHA512_DIGEST_WORD_SIZE*idx + (2*I + 1)*SHA512_DIGEST_ROW_SIZE], 1 + vpshufb xmm0, [rel byteswap] + vmovdqa [lane_data + _outer_block_sha512 + I * 2 * SHA512_DIGEST_WORD_SIZE], xmm0 +%assign I (I+1) +%endrep + + mov tmp, [job + _auth_key_xor_opad] +%assign I 0 +%rep 4 + vmovdqu xmm0, [tmp + I * 16] + vmovq [state + _args_digest_sha512 + SHA512_DIGEST_WORD_SIZE*idx + (2*I+0)*SHA512_DIGEST_ROW_SIZE], xmm0 + vpextrq [state + _args_digest_sha512 + SHA512_DIGEST_WORD_SIZE*idx + (2*I + 1)*SHA512_DIGEST_ROW_SIZE], xmm0, 1 +%assign I (I+1) +%endrep + + jmp start_loop + + align 16 +proc_extra_blocks: + mov DWORD(start_offset), [lane_data + _start_offset_sha512] + + vmovdqa xmm0, [state + _lens_sha512] + XVPINSRW xmm0, xmm1, tmp, idx, extra_blocks, scale_x16 + vmovdqa [state + _lens_sha512], xmm0 + + lea tmp, [lane_data + _extra_block_sha512 + start_offset] + mov [state + _args_data_ptr_sha512 + PTR_SZ*idx], tmp ;; idx is index of shortest length message + mov dword [lane_data + _extra_blocks_sha512], 0 + jmp start_loop + + align 16 +copy_lt128: + ;; less than one message block of data + ;; destination extra block but backwards by len from where 0x80 pre-populated + lea p2, [lane_data + _extra_block + 128] + sub p2, len + memcpy_avx2_128_1 p2, p, len, tmp4, tmp2, ymm0, ymm1, ymm2, ymm3 + mov unused_lanes, [state + _unused_lanes_sha512] + jmp end_fast_copy + +return_null: + xor job_rax, job_rax + jmp return + + align 16 +end_loop: + mov job_rax, [lane_data + _job_in_lane_sha512] + mov unused_lanes, [state + _unused_lanes_sha512] + mov qword [lane_data + _job_in_lane_sha512], 0 + or dword [job_rax + _status], STS_COMPLETED_HMAC + shl unused_lanes, 8 + or unused_lanes, idx + mov [state + _unused_lanes_sha512], unused_lanes + + mov p, [job_rax + _auth_tag_output] + + vzeroupper + +%if (SHA_X_DIGEST_SIZE != 384) + cmp qword [job_rax + _auth_tag_output_len_in_bytes], 32 + jne copy_full_digest +%else + cmp qword [job_rax + _auth_tag_output_len_in_bytes], 24 + jne copy_full_digest +%endif + ;; copy 32 bytes for SHA512 / 24 bytes for SHA384 + mov QWORD(tmp), [state + _args_digest_sha512 + SHA512_DIGEST_WORD_SIZE*idx + 0*SHA512_DIGEST_ROW_SIZE] + mov QWORD(tmp2), [state + _args_digest_sha512 + SHA512_DIGEST_WORD_SIZE*idx + 1*SHA512_DIGEST_ROW_SIZE] + mov QWORD(tmp3), [state + _args_digest_sha512 + SHA512_DIGEST_WORD_SIZE*idx + 2*SHA512_DIGEST_ROW_SIZE] +%if (SHA_X_DIGEST_SIZE != 384) + mov QWORD(tmp4), [state + _args_digest_sha512 + SHA512_DIGEST_WORD_SIZE*idx + 3*SHA512_DIGEST_ROW_SIZE] +%endif + bswap QWORD(tmp) + bswap QWORD(tmp2) + bswap QWORD(tmp3) +%if (SHA_X_DIGEST_SIZE != 384) + bswap QWORD(tmp4) +%endif + mov [p + 0*8], QWORD(tmp) + mov [p + 1*8], QWORD(tmp2) + mov [p + 2*8], QWORD(tmp3) +%if (SHA_X_DIGEST_SIZE != 384) + mov [p + 3*8], QWORD(tmp4) +%endif + jmp clear_ret + +copy_full_digest: + ;; copy 64 bytes for SHA512 / 48 bytes for SHA384 + mov QWORD(tmp), [state + _args_digest_sha512 + SHA512_DIGEST_WORD_SIZE*idx + 0*SHA512_DIGEST_ROW_SIZE] + mov QWORD(tmp2), [state + _args_digest_sha512 + SHA512_DIGEST_WORD_SIZE*idx + 1*SHA512_DIGEST_ROW_SIZE] + mov QWORD(tmp3), [state + _args_digest_sha512 + SHA512_DIGEST_WORD_SIZE*idx + 2*SHA512_DIGEST_ROW_SIZE] + mov QWORD(tmp4), [state + _args_digest_sha512 + SHA512_DIGEST_WORD_SIZE*idx + 3*SHA512_DIGEST_ROW_SIZE] + bswap QWORD(tmp) + bswap QWORD(tmp2) + bswap QWORD(tmp3) + bswap QWORD(tmp4) + mov [p + 0*8], QWORD(tmp) + mov [p + 1*8], QWORD(tmp2) + mov [p + 2*8], QWORD(tmp3) + mov [p + 3*8], QWORD(tmp4) + + mov QWORD(tmp), [state + _args_digest_sha512 + SHA512_DIGEST_WORD_SIZE*idx + 4*SHA512_DIGEST_ROW_SIZE] + mov QWORD(tmp2), [state + _args_digest_sha512 + SHA512_DIGEST_WORD_SIZE*idx + 5*SHA512_DIGEST_ROW_SIZE] +%if (SHA_X_DIGEST_SIZE != 384) + mov QWORD(tmp3), [state + _args_digest_sha512 + SHA512_DIGEST_WORD_SIZE*idx + 6*SHA512_DIGEST_ROW_SIZE] + mov QWORD(tmp4), [state + _args_digest_sha512 + SHA512_DIGEST_WORD_SIZE*idx + 7*SHA512_DIGEST_ROW_SIZE] +%endif + bswap QWORD(tmp) + bswap QWORD(tmp2) +%if (SHA_X_DIGEST_SIZE != 384) + bswap QWORD(tmp3) + bswap QWORD(tmp4) +%endif + mov [p + 4*8], QWORD(tmp) + mov [p + 5*8], QWORD(tmp2) +%if (SHA_X_DIGEST_SIZE != 384) + mov [p + 6*8], QWORD(tmp3) + mov [p + 7*8], QWORD(tmp4) +%endif + +clear_ret: + +%ifdef SAFE_DATA + ;; Clear digest (48B/64B), outer_block (48B/64B) and extra_block (128B) of returned job +%assign J 0 +%rep 6 + mov qword [state + _args_digest_sha512 + SHA512_DIGEST_WORD_SIZE*idx + J*SHA512_DIGEST_ROW_SIZE], 0 +%assign J (J+1) +%endrep +%if (SHA_X_DIGEST_SIZE != 384) + mov qword [state + _args_digest_sha512 + SHA512_DIGEST_WORD_SIZE*idx + 6*SHA256_DIGEST_ROW_SIZE], 0 + mov qword [state + _args_digest_sha512 + SHA512_DIGEST_WORD_SIZE*idx + 7*SHA256_DIGEST_ROW_SIZE], 0 +%endif + + vpxor ymm0, ymm0 + imul lane_data, idx, _SHA512_LANE_DATA_size + lea lane_data, [state + _ldata_sha512 + lane_data] + ;; Clear first 128 bytes of extra_block +%assign offset 0 +%rep 4 + vmovdqa [lane_data + _extra_block + offset], ymm0 +%assign offset (offset + 32) +%endrep + + ;; Clear first 48 bytes (SHA-384) or 64 bytes (SHA-512) of outer_block + vmovdqu [lane_data + _outer_block], ymm0 +%if (SHA_X_DIGEST_SIZE == 384) + vmovdqa [lane_data + _outer_block + 32], xmm0 +%else + vmovdqu [lane_data + _outer_block + 32], ymm0 +%endif +%endif ;; SAFE_DATA + +return: + mov rbx, [rsp + _gpr_save + 8*0] + mov rbp, [rsp + _gpr_save + 8*1] + mov r12, [rsp + _gpr_save + 8*2] +%ifndef LINUX + mov rsi, [rsp + _gpr_save + 8*3] + mov rdi, [rsp + _gpr_save + 8*4] +%endif + mov rsp, [rsp + _rsp_save] ; original SP + ret + +%ifdef LINUX +section .note.GNU-stack noalloc noexec nowrite progbits +%endif |