From patchwork Tue Sep 11 07:40:08 2018 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Ondrej Mosnacek X-Patchwork-Id: 10595205 X-Patchwork-Delegate: herbert@gondor.apana.org.au Return-Path: Received: from mail.wl.linuxfoundation.org (pdx-wl-mail.web.codeaurora.org [172.30.200.125]) by pdx-korg-patchwork-2.web.codeaurora.org (Postfix) with ESMTP id 7765514E5 for ; Tue, 11 Sep 2018 07:40:21 +0000 (UTC) Received: from mail.wl.linuxfoundation.org (localhost [127.0.0.1]) by mail.wl.linuxfoundation.org (Postfix) with ESMTP id 626532921F for ; Tue, 11 Sep 2018 07:40:21 +0000 (UTC) Received: by mail.wl.linuxfoundation.org (Postfix, from userid 486) id 544BA29238; Tue, 11 Sep 2018 07:40:21 +0000 (UTC) X-Spam-Checker-Version: SpamAssassin 3.3.1 (2010-03-16) on pdx-wl-mail.web.codeaurora.org X-Spam-Level: X-Spam-Status: No, score=-7.9 required=2.0 tests=BAYES_00,MAILING_LIST_MULTI, RCVD_IN_DNSWL_HI autolearn=ham version=3.3.1 Received: from vger.kernel.org (vger.kernel.org [209.132.180.67]) by mail.wl.linuxfoundation.org (Postfix) with ESMTP id 503B52921F for ; Tue, 11 Sep 2018 07:40:20 +0000 (UTC) Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1726387AbeIKMiU (ORCPT ); Tue, 11 Sep 2018 08:38:20 -0400 Received: from mail-wm0-f65.google.com ([74.125.82.65]:35368 "EHLO mail-wm0-f65.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1726325AbeIKMiT (ORCPT ); Tue, 11 Sep 2018 08:38:19 -0400 Received: by mail-wm0-f65.google.com with SMTP id o18-v6so273719wmc.0 for ; Tue, 11 Sep 2018 00:40:17 -0700 (PDT) X-Google-DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=1e100.net; s=20161025; h=x-gm-message-state:from:to:cc:subject:date:message-id; bh=84nRa7WlloI2zHSoDwq/d21DFSZ4+FYNsdqSL9l1qWg=; b=Ke69Yi6IHXMeFoQ2a0HbfrDsqvTnKwE8sKBn9NQcG3pyRbtsOLkCJNp9zQbFg8CMcJ HrkBnx5C2/GvnLHTR33g5pCp42LbwIcuNk73UjIKzqPAQ31WTBB0YOTc0Y8m6BOjeR9+ PXRDw+6zI8ez/aTZCR467mIGzUJStyr708D4wrQzJCi4A0akoHCXdHKPMfx58kYnnFsg BhIe0lO9yZklCKpH3YmlMoxl0p4gIrRud1zKB7H5CHedD4VFfBduQTLHUvvoA0WQiFAT XEN6UJCTFRjFmtVwOmNclGK4vcn1uh/bYl51hVvbWyCbe5iJK/mB795LWiMEigbaA8gc Qnqw== X-Gm-Message-State: APzg51CJnIf86LpdZSyl0ouR/Aj0RGWbx8c8/RWKsh1lGzLdQ/ppci+L WYcDENZ+tmNNaGRSBxXJZoZkMw== X-Google-Smtp-Source: ANB0VdaDb5gst4rJy1KDvsOi9Q5t3DRWJmZoOkujVYPuzTE8k/ZXfY/dJHvuWzZq1xg7NlkYoArYQQ== X-Received: by 2002:a1c:aa8f:: with SMTP id t137-v6mr412892wme.54.1536651616453; Tue, 11 Sep 2018 00:40:16 -0700 (PDT) Received: from localhost.localdomain.com (nat-pool-brq-t.redhat.com. [213.175.37.10]) by smtp.gmail.com with ESMTPSA id q16-v6sm12289744wrn.41.2018.09.11.00.40.15 (version=TLS1_2 cipher=ECDHE-RSA-CHACHA20-POLY1305 bits=256/256); Tue, 11 Sep 2018 00:40:15 -0700 (PDT) From: Ondrej Mosnacek To: Herbert Xu Cc: linux-crypto@vger.kernel.org, dm-devel@redhat.com, Mikulas Patocka , Eric Biggers , Ondrej Mosnacek Subject: [PATCH v5] crypto: xts - Drop use of auxiliary buffer Date: Tue, 11 Sep 2018 09:40:08 +0200 Message-Id: <20180911074008.2249-1-omosnace@redhat.com> X-Mailer: git-send-email 2.17.1 Sender: linux-crypto-owner@vger.kernel.org Precedence: bulk List-ID: X-Mailing-List: linux-crypto@vger.kernel.org X-Virus-Scanned: ClamAV using ClamSMTP Since commit acb9b159c784 ("crypto: gf128mul - define gf128mul_x_* in gf128mul.h"), the gf128mul_x_*() functions are very fast and therefore caching the computed XTS tweaks has only negligible advantage over computing them twice. In fact, since the current caching implementation limits the size of the calls to the child ecb(...) algorithm to PAGE_SIZE (usually 4096 B), it is often actually slower than the simple recomputing implementation. This patch simplifies the XTS template to recompute the XTS tweaks from scratch in the second pass and thus also removes the need to allocate a dynamic buffer using kmalloc(). As discussed at [1], the use of kmalloc causes deadlocks with dm-crypt. PERFORMANCE RESULTS I measured time to encrypt/decrypt a memory buffer of varying sizes with xts(ecb-aes-aesni) using a tool I wrote ([2]) and the results suggest that after this patch the performance is either better or comparable for both small and large buffers. Note that there is a lot of noise in the measurements, but the overall difference is easy to see. Old code: ALGORITHM KEY (b) DATA (B) TIME ENC (ns) TIME DEC (ns) xts(aes) 256 64 331 328 xts(aes) 384 64 332 333 xts(aes) 512 64 338 348 xts(aes) 256 512 889 920 xts(aes) 384 512 1019 993 xts(aes) 512 512 1032 990 xts(aes) 256 4096 2152 2292 xts(aes) 384 4096 2453 2597 xts(aes) 512 4096 3041 2641 xts(aes) 256 16384 9443 8027 xts(aes) 384 16384 8536 8925 xts(aes) 512 16384 9232 9417 xts(aes) 256 32768 16383 14897 xts(aes) 384 32768 17527 16102 xts(aes) 512 32768 18483 17322 New code: ALGORITHM KEY (b) DATA (B) TIME ENC (ns) TIME DEC (ns) xts(aes) 256 64 328 324 xts(aes) 384 64 324 319 xts(aes) 512 64 320 322 xts(aes) 256 512 476 473 xts(aes) 384 512 509 492 xts(aes) 512 512 531 514 xts(aes) 256 4096 2132 1829 xts(aes) 384 4096 2357 2055 xts(aes) 512 4096 2178 2027 xts(aes) 256 16384 6920 6983 xts(aes) 384 16384 8597 7505 xts(aes) 512 16384 7841 8164 xts(aes) 256 32768 13468 12307 xts(aes) 384 32768 14808 13402 xts(aes) 512 32768 15753 14636 [1] https://lkml.org/lkml/2018/8/23/1315 [2] https://gitlab.com/omos/linux-crypto-bench Signed-off-by: Ondrej Mosnacek --- crypto/xts.c | 269 +++++++++------------------------------------------ 1 file changed, 46 insertions(+), 223 deletions(-) Changes in v5: - fix dumb mistakes v4: https://www.spinics.net/lists/linux-crypto/msg34889.html Changes in v4: - small cleanup suggested by Eric Biggers v3: https://www.spinics.net/lists/linux-crypto/msg34823.html Changes in v3: - add comment explaining the new approach as suggested by Eric - ensure correct alignment in second xor_tweak() pass - align performance results table header to the right v2: https://www.spinics.net/lists/linux-crypto/msg34799.html Changes in v2: - rebase to latest cryptodev tree v1: https://www.spinics.net/lists/linux-crypto/msg34776.html diff --git a/crypto/xts.c b/crypto/xts.c index ccf55fbb8bc2..847f54f76789 100644 --- a/crypto/xts.c +++ b/crypto/xts.c @@ -26,8 +26,6 @@ #include #include -#define XTS_BUFFER_SIZE 128u - struct priv { struct crypto_skcipher *child; struct crypto_cipher *tweak; @@ -39,19 +37,7 @@ struct xts_instance_ctx { }; struct rctx { - le128 buf[XTS_BUFFER_SIZE / sizeof(le128)]; - le128 t; - - le128 *ext; - - struct scatterlist srcbuf[2]; - struct scatterlist dstbuf[2]; - struct scatterlist *src; - struct scatterlist *dst; - - unsigned int left; - struct skcipher_request subreq; }; @@ -96,81 +82,27 @@ static int setkey(struct crypto_skcipher *parent, const u8 *key, return err; } -static int post_crypt(struct skcipher_request *req) +/* + * We compute the tweak masks twice (both before and after the ECB encryption or + * decryption) to avoid having to allocate a temporary buffer and/or make + * mutliple calls to the 'ecb(..)' instance, which usually would be slower than + * just doing the gf128mul_x_ble() calls again. + */ +static int xor_tweak(struct skcipher_request *req, bool second_pass) { struct rctx *rctx = skcipher_request_ctx(req); - le128 *buf = rctx->ext ?: rctx->buf; - struct skcipher_request *subreq; + struct crypto_skcipher *tfm = crypto_skcipher_reqtfm(req); const int bs = XTS_BLOCK_SIZE; struct skcipher_walk w; - struct scatterlist *sg; - unsigned offset; + le128 t = rctx->t; int err; - subreq = &rctx->subreq; - err = skcipher_walk_virt(&w, subreq, false); - - while (w.nbytes) { - unsigned int avail = w.nbytes; - le128 *wdst; - - wdst = w.dst.virt.addr; - - do { - le128_xor(wdst, buf++, wdst); - wdst++; - } while ((avail -= bs) >= bs); - - err = skcipher_walk_done(&w, avail); + if (second_pass) { + req = &rctx->subreq; + /* set to our TFM to enforce correct alignment: */ + skcipher_request_set_tfm(req, tfm); } - - rctx->left -= subreq->cryptlen; - - if (err || !rctx->left) - goto out; - - rctx->dst = rctx->dstbuf; - - scatterwalk_done(&w.out, 0, 1); - sg = w.out.sg; - offset = w.out.offset; - - if (rctx->dst != sg) { - rctx->dst[0] = *sg; - sg_unmark_end(rctx->dst); - scatterwalk_crypto_chain(rctx->dst, sg_next(sg), 2); - } - rctx->dst[0].length -= offset - sg->offset; - rctx->dst[0].offset = offset; - -out: - return err; -} - -static int pre_crypt(struct skcipher_request *req) -{ - struct rctx *rctx = skcipher_request_ctx(req); - le128 *buf = rctx->ext ?: rctx->buf; - struct skcipher_request *subreq; - const int bs = XTS_BLOCK_SIZE; - struct skcipher_walk w; - struct scatterlist *sg; - unsigned cryptlen; - unsigned offset; - bool more; - int err; - - subreq = &rctx->subreq; - cryptlen = subreq->cryptlen; - - more = rctx->left > cryptlen; - if (!more) - cryptlen = rctx->left; - - skcipher_request_set_crypt(subreq, rctx->src, rctx->dst, - cryptlen, NULL); - - err = skcipher_walk_virt(&w, subreq, false); + err = skcipher_walk_virt(&w, req, false); while (w.nbytes) { unsigned int avail = w.nbytes; @@ -181,180 +113,71 @@ static int pre_crypt(struct skcipher_request *req) wdst = w.dst.virt.addr; do { - *buf++ = rctx->t; - le128_xor(wdst++, &rctx->t, wsrc++); - gf128mul_x_ble(&rctx->t, &rctx->t); + le128_xor(wdst++, &t, wsrc++); + gf128mul_x_ble(&t, &t); } while ((avail -= bs) >= bs); err = skcipher_walk_done(&w, avail); } - skcipher_request_set_crypt(subreq, rctx->dst, rctx->dst, - cryptlen, NULL); - - if (err || !more) - goto out; - - rctx->src = rctx->srcbuf; - - scatterwalk_done(&w.in, 0, 1); - sg = w.in.sg; - offset = w.in.offset; - - if (rctx->src != sg) { - rctx->src[0] = *sg; - sg_unmark_end(rctx->src); - scatterwalk_crypto_chain(rctx->src, sg_next(sg), 2); - } - rctx->src[0].length -= offset - sg->offset; - rctx->src[0].offset = offset; - -out: return err; } -static int init_crypt(struct skcipher_request *req, crypto_completion_t done) +static int xor_tweak_pre(struct skcipher_request *req) { - struct priv *ctx = crypto_skcipher_ctx(crypto_skcipher_reqtfm(req)); - struct rctx *rctx = skcipher_request_ctx(req); - struct skcipher_request *subreq; - gfp_t gfp; - - subreq = &rctx->subreq; - skcipher_request_set_tfm(subreq, ctx->child); - skcipher_request_set_callback(subreq, req->base.flags, done, req); - - gfp = req->base.flags & CRYPTO_TFM_REQ_MAY_SLEEP ? GFP_KERNEL : - GFP_ATOMIC; - rctx->ext = NULL; - - subreq->cryptlen = XTS_BUFFER_SIZE; - if (req->cryptlen > XTS_BUFFER_SIZE) { - unsigned int n = min(req->cryptlen, (unsigned int)PAGE_SIZE); - - rctx->ext = kmalloc(n, gfp); - if (rctx->ext) - subreq->cryptlen = n; - } - - rctx->src = req->src; - rctx->dst = req->dst; - rctx->left = req->cryptlen; - - /* calculate first value of T */ - crypto_cipher_encrypt_one(ctx->tweak, (u8 *)&rctx->t, req->iv); - - return 0; + return xor_tweak(req, false); } -static void exit_crypt(struct skcipher_request *req) +static int xor_tweak_post(struct skcipher_request *req) { - struct rctx *rctx = skcipher_request_ctx(req); - - rctx->left = 0; - - if (rctx->ext) - kzfree(rctx->ext); + return xor_tweak(req, true); } -static int do_encrypt(struct skcipher_request *req, int err) -{ - struct rctx *rctx = skcipher_request_ctx(req); - struct skcipher_request *subreq; - - subreq = &rctx->subreq; - - while (!err && rctx->left) { - err = pre_crypt(req) ?: - crypto_skcipher_encrypt(subreq) ?: - post_crypt(req); - - if (err == -EINPROGRESS || err == -EBUSY) - return err; - } - - exit_crypt(req); - return err; -} - -static void encrypt_done(struct crypto_async_request *areq, int err) +static void crypt_done(struct crypto_async_request *areq, int err) { struct skcipher_request *req = areq->data; - struct skcipher_request *subreq; - struct rctx *rctx; - - rctx = skcipher_request_ctx(req); - - if (err == -EINPROGRESS) { - if (rctx->left != req->cryptlen) - return; - goto out; - } - - subreq = &rctx->subreq; - subreq->base.flags &= CRYPTO_TFM_REQ_MAY_BACKLOG; - err = do_encrypt(req, err ?: post_crypt(req)); - if (rctx->left) - return; + if (!err) + err = xor_tweak_post(req); -out: skcipher_request_complete(req, err); } -static int encrypt(struct skcipher_request *req) -{ - return do_encrypt(req, init_crypt(req, encrypt_done)); -} - -static int do_decrypt(struct skcipher_request *req, int err) +static void init_crypt(struct skcipher_request *req) { + struct priv *ctx = crypto_skcipher_ctx(crypto_skcipher_reqtfm(req)); struct rctx *rctx = skcipher_request_ctx(req); - struct skcipher_request *subreq; - - subreq = &rctx->subreq; + struct skcipher_request *subreq = &rctx->subreq; - while (!err && rctx->left) { - err = pre_crypt(req) ?: - crypto_skcipher_decrypt(subreq) ?: - post_crypt(req); - - if (err == -EINPROGRESS || err == -EBUSY) - return err; - } + skcipher_request_set_tfm(subreq, ctx->child); + skcipher_request_set_callback(subreq, req->base.flags, crypt_done, req); + skcipher_request_set_crypt(subreq, req->dst, req->dst, + req->cryptlen, NULL); - exit_crypt(req); - return err; + /* calculate first value of T */ + crypto_cipher_encrypt_one(ctx->tweak, (u8 *)&rctx->t, req->iv); } -static void decrypt_done(struct crypto_async_request *areq, int err) +static int encrypt(struct skcipher_request *req) { - struct skcipher_request *req = areq->data; - struct skcipher_request *subreq; - struct rctx *rctx; - - rctx = skcipher_request_ctx(req); - - if (err == -EINPROGRESS) { - if (rctx->left != req->cryptlen) - return; - goto out; - } - - subreq = &rctx->subreq; - subreq->base.flags &= CRYPTO_TFM_REQ_MAY_BACKLOG; - - err = do_decrypt(req, err ?: post_crypt(req)); - if (rctx->left) - return; + struct rctx *rctx = skcipher_request_ctx(req); + struct skcipher_request *subreq = &rctx->subreq; -out: - skcipher_request_complete(req, err); + init_crypt(req); + return xor_tweak_pre(req) ?: + crypto_skcipher_encrypt(subreq) ?: + xor_tweak_post(req); } static int decrypt(struct skcipher_request *req) { - return do_decrypt(req, init_crypt(req, decrypt_done)); + struct rctx *rctx = skcipher_request_ctx(req); + struct skcipher_request *subreq = &rctx->subreq; + + init_crypt(req); + return xor_tweak_pre(req) ?: + crypto_skcipher_decrypt(subreq) ?: + xor_tweak_post(req); } static int init_tfm(struct crypto_skcipher *tfm)