tjh.dev / kernel
Linux kernel mirror (for testing) git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git
kernel os linux
1
fork

Configure Feed

Select the types of activity you want to include in your feed.

crypto: x86/aegis128 - don't bother with special code for aligned data

Remove the AEGIS assembly code paths that were "optimized" to operate on
16-byte aligned data using movdqa, and instead just use the code paths
that use movdqu and can handle data with any alignment.

This does not reduce performance. movdqa is basically a historical
artifact; on aligned data, movdqu and movdqa have had the same
performance since Intel Nehalem (2008) and AMD Bulldozer (2011). And
code that requires AES-NI cannot run on CPUs older than those anyway.

Reviewed-by: Ondrej Mosnacek <omosnace@redhat.com>
Signed-off-by: Eric Biggers <ebiggers@google.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>

authored by

Eric Biggers and committed by
Herbert Xu 595bca25 b8d2e7ba

+22 -100
+22 -100
arch/x86/crypto/aegis128-aesni-asm.S
··· 245 245 movdqu 0x30(STATEP), STATE3 246 246 movdqu 0x40(STATEP), STATE4 247 247 248 - mov SRC, %r8 249 - and $0xF, %r8 250 - jnz .Lad_u_loop 251 - 252 248 .align 8 253 - .Lad_a_loop: 254 - movdqa 0x00(SRC), MSG 255 - aegis128_update 256 - pxor MSG, STATE4 257 - sub $0x10, LEN 258 - cmp $0x10, LEN 259 - jl .Lad_out_1 260 - 261 - movdqa 0x10(SRC), MSG 262 - aegis128_update 263 - pxor MSG, STATE3 264 - sub $0x10, LEN 265 - cmp $0x10, LEN 266 - jl .Lad_out_2 267 - 268 - movdqa 0x20(SRC), MSG 269 - aegis128_update 270 - pxor MSG, STATE2 271 - sub $0x10, LEN 272 - cmp $0x10, LEN 273 - jl .Lad_out_3 274 - 275 - movdqa 0x30(SRC), MSG 276 - aegis128_update 277 - pxor MSG, STATE1 278 - sub $0x10, LEN 279 - cmp $0x10, LEN 280 - jl .Lad_out_4 281 - 282 - movdqa 0x40(SRC), MSG 283 - aegis128_update 284 - pxor MSG, STATE0 285 - sub $0x10, LEN 286 - cmp $0x10, LEN 287 - jl .Lad_out_0 288 - 289 - add $0x50, SRC 290 - jmp .Lad_a_loop 291 - 292 - .align 8 293 - .Lad_u_loop: 249 + .Lad_loop: 294 250 movdqu 0x00(SRC), MSG 295 251 aegis128_update 296 252 pxor MSG, STATE4 ··· 283 327 jl .Lad_out_0 284 328 285 329 add $0x50, SRC 286 - jmp .Lad_u_loop 330 + jmp .Lad_loop 287 331 288 332 /* store the state: */ 289 333 .Lad_out_0: ··· 336 380 RET 337 381 SYM_FUNC_END(crypto_aegis128_aesni_ad) 338 382 339 - .macro encrypt_block a s0 s1 s2 s3 s4 i 340 - movdq\a (\i * 0x10)(SRC), MSG 383 + .macro encrypt_block s0 s1 s2 s3 s4 i 384 + movdqu (\i * 0x10)(SRC), MSG 341 385 movdqa MSG, T0 342 386 pxor \s1, T0 343 387 pxor \s4, T0 344 388 movdqa \s2, T1 345 389 pand \s3, T1 346 390 pxor T1, T0 347 - movdq\a T0, (\i * 0x10)(DST) 391 + movdqu T0, (\i * 0x10)(DST) 348 392 349 393 aegis128_update 350 394 pxor MSG, \s4 ··· 371 415 movdqu 0x30(STATEP), STATE3 372 416 movdqu 0x40(STATEP), STATE4 373 417 374 - mov SRC, %r8 375 - or DST, %r8 376 - and $0xF, %r8 377 - jnz .Lenc_u_loop 378 - 379 418 .align 8 380 - .Lenc_a_loop: 381 - encrypt_block a STATE0 STATE1 STATE2 STATE3 STATE4 0 382 - encrypt_block a STATE4 STATE0 STATE1 STATE2 STATE3 1 383 - encrypt_block a STATE3 STATE4 STATE0 STATE1 STATE2 2 384 - encrypt_block a STATE2 STATE3 STATE4 STATE0 STATE1 3 385 - encrypt_block a STATE1 STATE2 STATE3 STATE4 STATE0 4 419 + .Lenc_loop: 420 + encrypt_block STATE0 STATE1 STATE2 STATE3 STATE4 0 421 + encrypt_block STATE4 STATE0 STATE1 STATE2 STATE3 1 422 + encrypt_block STATE3 STATE4 STATE0 STATE1 STATE2 2 423 + encrypt_block STATE2 STATE3 STATE4 STATE0 STATE1 3 424 + encrypt_block STATE1 STATE2 STATE3 STATE4 STATE0 4 386 425 387 426 add $0x50, SRC 388 427 add $0x50, DST 389 - jmp .Lenc_a_loop 390 - 391 - .align 8 392 - .Lenc_u_loop: 393 - encrypt_block u STATE0 STATE1 STATE2 STATE3 STATE4 0 394 - encrypt_block u STATE4 STATE0 STATE1 STATE2 STATE3 1 395 - encrypt_block u STATE3 STATE4 STATE0 STATE1 STATE2 2 396 - encrypt_block u STATE2 STATE3 STATE4 STATE0 STATE1 3 397 - encrypt_block u STATE1 STATE2 STATE3 STATE4 STATE0 4 398 - 399 - add $0x50, SRC 400 - add $0x50, DST 401 - jmp .Lenc_u_loop 428 + jmp .Lenc_loop 402 429 403 430 /* store the state: */ 404 431 .Lenc_out_0: ··· 474 535 RET 475 536 SYM_FUNC_END(crypto_aegis128_aesni_enc_tail) 476 537 477 - .macro decrypt_block a s0 s1 s2 s3 s4 i 478 - movdq\a (\i * 0x10)(SRC), MSG 538 + .macro decrypt_block s0 s1 s2 s3 s4 i 539 + movdqu (\i * 0x10)(SRC), MSG 479 540 pxor \s1, MSG 480 541 pxor \s4, MSG 481 542 movdqa \s2, T1 482 543 pand \s3, T1 483 544 pxor T1, MSG 484 - movdq\a MSG, (\i * 0x10)(DST) 545 + movdqu MSG, (\i * 0x10)(DST) 485 546 486 547 aegis128_update 487 548 pxor MSG, \s4 ··· 508 569 movdqu 0x30(STATEP), STATE3 509 570 movdqu 0x40(STATEP), STATE4 510 571 511 - mov SRC, %r8 512 - or DST, %r8 513 - and $0xF, %r8 514 - jnz .Ldec_u_loop 515 - 516 572 .align 8 517 - .Ldec_a_loop: 518 - decrypt_block a STATE0 STATE1 STATE2 STATE3 STATE4 0 519 - decrypt_block a STATE4 STATE0 STATE1 STATE2 STATE3 1 520 - decrypt_block a STATE3 STATE4 STATE0 STATE1 STATE2 2 521 - decrypt_block a STATE2 STATE3 STATE4 STATE0 STATE1 3 522 - decrypt_block a STATE1 STATE2 STATE3 STATE4 STATE0 4 573 + .Ldec_loop: 574 + decrypt_block STATE0 STATE1 STATE2 STATE3 STATE4 0 575 + decrypt_block STATE4 STATE0 STATE1 STATE2 STATE3 1 576 + decrypt_block STATE3 STATE4 STATE0 STATE1 STATE2 2 577 + decrypt_block STATE2 STATE3 STATE4 STATE0 STATE1 3 578 + decrypt_block STATE1 STATE2 STATE3 STATE4 STATE0 4 523 579 524 580 add $0x50, SRC 525 581 add $0x50, DST 526 - jmp .Ldec_a_loop 527 - 528 - .align 8 529 - .Ldec_u_loop: 530 - decrypt_block u STATE0 STATE1 STATE2 STATE3 STATE4 0 531 - decrypt_block u STATE4 STATE0 STATE1 STATE2 STATE3 1 532 - decrypt_block u STATE3 STATE4 STATE0 STATE1 STATE2 2 533 - decrypt_block u STATE2 STATE3 STATE4 STATE0 STATE1 3 534 - decrypt_block u STATE1 STATE2 STATE3 STATE4 STATE0 4 535 - 536 - add $0x50, SRC 537 - add $0x50, DST 538 - jmp .Ldec_u_loop 582 + jmp .Ldec_loop 539 583 540 584 /* store the state: */ 541 585 .Ldec_out_0: