[v7,2/2] crypto: s5p-sss: Add HASH support for Exynos

Message ID	20171017112824.10231-3-k.konieczny@partner.samsung.com (mailing list archive)
State	Not Applicable
Headers	show Return-Path: <linux-samsung-soc-owner@kernel.org> From: Kamil Konieczny <k.konieczny@partner.samsung.com> To: linux-crypto@vger.kernel.org Cc: Kamil Konieczny <k.konieczny@partner.samsung.com>, Herbert Xu <herbert@gondor.apana.org.au>, Krzysztof Kozlowski <krzk@kernel.org>, Vladimir Zapolskiy <vladimir_zapolskiy@mentor.com>, Vladimir Zapolskiy <vz@mleia.com>, "David S. Miller" <davem@davemloft.net>, Bartlomiej Zolnierkiewicz <b.zolnierkie@samsung.com>, linux-samsung-soc@vger.kernel.org, linux-kernel@vger.kernel.org Subject: [PATCH v7 2/2] crypto: s5p-sss: Add HASH support for Exynos Date: Tue, 17 Oct 2017 13:28:24 +0200 Message-id: <20171017112824.10231-3-k.konieczny@partner.samsung.com> In-reply-to: <20171017112824.10231-1-k.konieczny@partner.samsung.com> CMS-TYPE: 201P References: <20171017112824.10231-1-k.konieczny@partner.samsung.com> <CGME20171017112846eucas1p17185d93eb5f837643c880357b3a8053b@eucas1p1.samsung.com> Sender: linux-samsung-soc-owner@vger.kernel.org Precedence: bulk

Hi Vladimir, Thank you for review, I will apply almost all of your remarks, see answers below. On 22.10.2017 12:18, Vladimir Zapolskiy wrote: > Hi Kamil, > > thank you for updates, I have just a few more comments. > > On 10/17/2017 02:28 PM, Kamil Konieczny wrote: >> [...] >> - Select sw algorithms MD5, SHA1 and SHA256 in EXYNOS_HASH >> as they are nedded for fallback. > > Typo, s/nedded/needed/ ok, Thank you for spotting this. > [snip] > >> >> #include <linux/clk.h> >> #include <linux/crypto.h> >> +#include <linux/delay.h> > > I can not find which interfaces from linux/delay.h are needed. > I suppose it should not be added. > > [snip] Yes, you are right, I will delete this 'include delay.h' >> -static struct s5p_aes_dev *s5p_dev; >> +/** >> + * struct s5p_hash_reqctx - HASH request context >> + * @dev: Associated device > > dev or dd? dd >> + * @op_update: Current request operation (OP_UPDATE or OP_FINAL) >> + * @digcnt: Number of bytes processed by HW (without buffer[] ones) >> + * @digest: Digest message or IV for partial result >> + * @nregs: Number of HW registers for digest or IV read/write >> + * @engine: Bits for selecting type of HASH in SSS block >> + * @sg: sg for DMA transfer >> + * @sg_len: Length of sg for DMA transfer >> + * @sgl[]: sg for joining buffer and req->src scatterlist >> + * @skip: Skip offset in req->src for current op >> + * @total: Total number of bytes for current request >> + * @finup: Keep state for finup or final. >> + * @error: Keep track of error. >> + * @bufcnt: Number of bytes holded in buffer[] >> + * @buffer[]: For byte(s) from end of req->src in UPDATE op >> + */ >> +struct s5p_hash_reqctx { >> + struct s5p_aes_dev *dd; >> + bool op_update; >> + > > [snip] > >> + >> +/** >> + * s5p_hash_shash_digest() - calculate shash digest >> + * @tfm: crypto transformation >> + * @flags: tfm flags >> + * @data: input data >> + * @len: length of data >> + * @out: output buffer >> + */ >> +static int s5p_hash_shash_digest(struct crypto_shash *tfm, u32 flags, >> + const u8 *data, unsigned int len, u8 *out) >> +{ >> + SHASH_DESC_ON_STACK(shash, tfm); > > Just for information, this line produces a spatch warning: > > drivers/crypto/s5p-sss.c:1534:9: warning: Variable length array is used. > > I think it can be ignored. I also think it can be ignored, it uses 104 bytes on stack in case of SHA256 , sizeof struct sha256_state for SW generic implementation, found in include/crypto/sha.h >> + >> + shash->tfm = tfm; >> + shash->flags = flags & ~CRYPTO_TFM_REQ_MAY_SLEEP; >> + >> + return crypto_shash_digest(shash, data, len, out); >> +} >> + > > [snip] > >> +/** >> + * s5p_hash_final() - close up hash and calculate digest >> + * @req: AHASH request >> + * >> + * Note: in final req->src do not have any data, and req->nbytes can be >> + * non-zero. >> + * >> + * If there were no input data processed yet and the buffered hash data is >> + * less than BUFLEN (64) then calculate the final hash immediately by using >> + * SW algorithm fallback. >> + * >> + * Otherwise enqueues the current AHASH request with OP_FINAL operation op >> + * and finalize hash message in HW. Note that if digcnt!=0 then there were >> + * previous update op, so there are always some buffered bytes in ctx->buffer, >> + * which means that ctx->bufcnt!=0 >> + * >> + * Returns: >> + * 0 if the request has been processed immediately, >> + * -EINPROGRESS if the operation has been queued for later execution or is set >> + * to processing by HW, >> + * -EBUSY if queue is full and request should be resubmitted later, other >> + * negative values on error. > > Do you want to add -EINVAL into the list also? Here I made bad formatting, it should read: * -EBUSY if queue is full and request should be resubmitted later, * other negative values on error. I do not want to specify explicitly all error negative codes here, as it also uses s5p_hash_enqueue and these return codes are referenced from other places. I intended to listing success values, 0, -EINPROGRESS, then explaining -EBUSY, then adding all other negative as error. The other values can be -EINVAL, -ENOMEM or maybe other, when this module gets extended to HMAC implementation. >> + */ >> +static int s5p_hash_final(struct ahash_request *req) >> +{ >> + struct s5p_hash_reqctx *ctx = ahash_request_ctx(req); >> + >> + ctx->finup = true; >> + if (ctx->error) >> + return -EINVAL; /* uncompleted hash is not needed */ >> + >> + if (!ctx->digcnt && ctx->bufcnt < BUFLEN) >> + return s5p_hash_final_shash(req); >> + >> + return s5p_hash_enqueue(req, false); /* HASH_OP_FINAL */ >> +} >> + > > [snip] > >> +/** >> + * s5p_hash_finup() - process last req->src and calculate digest >> + * @req: AHASH request containing the last update data >> + * >> + * Return values: see s5p_hash_final above. >> + */ >> +static int s5p_hash_finup(struct ahash_request *req) >> +{ >> + struct s5p_hash_reqctx *ctx = ahash_request_ctx(req); >> + int err1, err2; >> + >> + ctx->finup = true; >> + >> + err1 = s5p_hash_update(req); >> + if (err1 == -EINPROGRESS || err1 == -EBUSY) >> + return err1; >> + >> + /* >> + * final() has to be always called to cleanup resources even if >> + * update() failed, except EINPROGRESS or calculate digest for small >> + * size >> + */ >> + err2 = s5p_hash_final(req); >> + >> + return err1 ?: err2; > > See a note below. > > [snip] > >> +/** >> + * s5p_hash_digest - calculate digest from req->src >> + * @req: AHASH request >> + * >> + * Return values: see s5p_hash_final above. >> + */ >> +static int s5p_hash_digest(struct ahash_request *req) >> +{ >> + return s5p_hash_init(req) ?: s5p_hash_finup(req); > > Now I got it, this ?: notation is invalid according to C99 and thus it > confused me, but GCC has an extension to support it, so, it's up to you > if you leave it as is or change it to comply with C99. If it ever causes > any troubles, it'll be easy to fix, and I'm fine if you leave it as is, > because it is understandable to GCC. > > [snip] I agree and I prefer to keep it in current shape, this patch is already big. >> +/** >> + * s5p_hash_import - import hash state >> + * @req: AHASH request >> + * @in: buffer with state to be imported from >> + */ >> +static int s5p_hash_import(struct ahash_request *req, const void *in) >> +{ >> + struct s5p_hash_reqctx *ctx = ahash_request_ctx(req); >> + struct crypto_ahash *tfm = crypto_ahash_reqtfm(req); >> + struct s5p_hash_ctx *tctx = crypto_ahash_ctx(tfm); >> + const struct s5p_hash_reqctx *ctx_in = in; >> + >> + memcpy(ctx, in, sizeof(*ctx) + BUFLEN); >> + if (ctx_in->bufcnt < 0 || ctx_in->bufcnt > BUFLEN) { > > Now "ctx_in->bufcnt < 0" condition can be removed, moreover it > will eliminate a warning reproted by the compiler: > > drivers/crypto/s5p-sss.c:1748:21: warning: comparison of unsigned expression < 0 is always false [-Wtype-limits] > if (ctx_in->bufcnt < 0 || ctx_in->bufcnt > BUFLEN) { > ^ You are right, I will drop first condition. BTW what compiler options are you using ? This one was not reported by my compilation process.

diff --git a/drivers/crypto/Kconfig b/drivers/crypto/Kconfig index 4b75084fabad..dea4d33d9c7f 100644 --- a/drivers/crypto/Kconfig +++ b/drivers/crypto/Kconfig @@ -439,6 +439,20 @@ config CRYPTO_DEV_S5P Select this to offload Samsung S5PV210 or S5PC110, Exynos from AES algorithms execution. +config CRYPTO_DEV_EXYNOS_HASH + bool "Support for Samsung Exynos HASH accelerator" + depends on CRYPTO_DEV_S5P + depends on !CRYPTO_DEV_EXYNOS_RNG && CRYPTO_DEV_EXYNOS_RNG!=m + select CRYPTO_SHA1 + select CRYPTO_MD5 + select CRYPTO_SHA256 + help + Select this to offload Exynos from HASH MD5/SHA1/SHA256. + This will select software SHA1, MD5 and SHA256 as they are + needed for small and zero-size messages. + HASH algorithms will be disabled if EXYNOS_RNG + is enabled due to hw conflict. + config CRYPTO_DEV_NX bool "Support for IBM PowerPC Nest (NX) cryptographic acceleration" depends on PPC64 diff --git a/drivers/crypto/s5p-sss.c b/drivers/crypto/s5p-sss.c index dfae1865c384..c9feccf21efe 100644 --- a/drivers/crypto/s5p-sss.c +++ b/drivers/crypto/s5p-sss.c @@ -1,18 +1,21 @@ /* * Cryptographic API. * - * Support for Samsung S5PV210 HW acceleration. + * Support for Samsung S5PV210 and Exynos HW acceleration. * * Copyright (C) 2011 NetUP Inc. All rights reserved. + * Copyright (c) 2017 Samsung Electronics Co., Ltd. All rights reserved. * * This program is free software; you can redistribute it and/or modify * it under the terms of the GNU General Public License version 2 as published * by the Free Software Foundation. * + * Hash part based on omap-sham.c driver. */ #include <linux/clk.h> #include <linux/crypto.h> +#include <linux/delay.h> #include <linux/dma-mapping.h> #include <linux/err.h> #include <linux/errno.h> @@ -30,28 +33,41 @@ #include <crypto/algapi.h> #include <crypto/scatterwalk.h> +#include <crypto/hash.h> +#include <crypto/md5.h> +#include <crypto/sha.h> +#include <crypto/internal/hash.h> + #define _SBF(s, v) ((v) << (s)) /* Feed control registers */ #define SSS_REG_FCINTSTAT 0x0000 +#define SSS_FCINTSTAT_HPARTINT BIT(7) +#define SSS_FCINTSTAT_HDONEINT BIT(5) #define SSS_FCINTSTAT_BRDMAINT BIT(3) #define SSS_FCINTSTAT_BTDMAINT BIT(2) #define SSS_FCINTSTAT_HRDMAINT BIT(1) #define SSS_FCINTSTAT_PKDMAINT BIT(0) #define SSS_REG_FCINTENSET 0x0004 +#define SSS_FCINTENSET_HPARTINTENSET BIT(7) +#define SSS_FCINTENSET_HDONEINTENSET BIT(5) #define SSS_FCINTENSET_BRDMAINTENSET BIT(3) #define SSS_FCINTENSET_BTDMAINTENSET BIT(2) #define SSS_FCINTENSET_HRDMAINTENSET BIT(1) #define SSS_FCINTENSET_PKDMAINTENSET BIT(0) #define SSS_REG_FCINTENCLR 0x0008 +#define SSS_FCINTENCLR_HPARTINTENCLR BIT(7) +#define SSS_FCINTENCLR_HDONEINTENCLR BIT(5) #define SSS_FCINTENCLR_BRDMAINTENCLR BIT(3) #define SSS_FCINTENCLR_BTDMAINTENCLR BIT(2) #define SSS_FCINTENCLR_HRDMAINTENCLR BIT(1) #define SSS_FCINTENCLR_PKDMAINTENCLR BIT(0) #define SSS_REG_FCINTPEND 0x000C +#define SSS_FCINTPEND_HPARTINTP BIT(7) +#define SSS_FCINTPEND_HDONEINTP BIT(5) #define SSS_FCINTPEND_BRDMAINTP BIT(3) #define SSS_FCINTPEND_BTDMAINTP BIT(2) #define SSS_FCINTPEND_HRDMAINTP BIT(1) @@ -72,6 +88,7 @@ #define SSS_HASHIN_INDEPENDENT _SBF(0, 0x00) #define SSS_HASHIN_CIPHER_INPUT _SBF(0, 0x01) #define SSS_HASHIN_CIPHER_OUTPUT _SBF(0, 0x02) +#define SSS_HASHIN_MASK _SBF(0, 0x03) #define SSS_REG_FCBRDMAS 0x0020 #define SSS_REG_FCBRDMAL 0x0024 @@ -146,9 +163,80 @@ #define AES_KEY_LEN 16 #define CRYPTO_QUEUE_LEN 1 +/* HASH registers */ +#define SSS_REG_HASH_CTRL 0x00 + +#define SSS_HASH_USER_IV_EN BIT(5) +#define SSS_HASH_INIT_BIT BIT(4) +#define SSS_HASH_ENGINE_SHA1 _SBF(1, 0x00) +#define SSS_HASH_ENGINE_MD5 _SBF(1, 0x01) +#define SSS_HASH_ENGINE_SHA256 _SBF(1, 0x02) + +#define SSS_HASH_ENGINE_MASK _SBF(1, 0x03) + +#define SSS_REG_HASH_CTRL_PAUSE 0x04 + +#define SSS_HASH_PAUSE BIT(0) + +#define SSS_REG_HASH_CTRL_FIFO 0x08 + +#define SSS_HASH_FIFO_MODE_DMA BIT(0) +#define SSS_HASH_FIFO_MODE_CPU 0 + +#define SSS_REG_HASH_CTRL_SWAP 0x0C + +#define SSS_HASH_BYTESWAP_DI BIT(3) +#define SSS_HASH_BYTESWAP_DO BIT(2) +#define SSS_HASH_BYTESWAP_IV BIT(1) +#define SSS_HASH_BYTESWAP_KEY BIT(0) + +#define SSS_REG_HASH_STATUS 0x10 + +#define SSS_HASH_STATUS_MSG_DONE BIT(6) +#define SSS_HASH_STATUS_PARTIAL_DONE BIT(4) +#define SSS_HASH_STATUS_BUFFER_READY BIT(0) + +#define SSS_REG_HASH_MSG_SIZE_LOW 0x20 +#define SSS_REG_HASH_MSG_SIZE_HIGH 0x24 + +#define SSS_REG_HASH_PRE_MSG_SIZE_LOW 0x28 +#define SSS_REG_HASH_PRE_MSG_SIZE_HIGH 0x2C + +#define SSS_REG_HASH_IV(s) (0xB0 + ((s) << 2)) +#define SSS_REG_HASH_OUT(s) (0x100 + ((s) << 2)) + +#define HASH_BLOCK_SIZE 64 +#define HASH_REG_SIZEOF 4 +#define HASH_MD5_MAX_REG (MD5_DIGEST_SIZE / HASH_REG_SIZEOF) +#define HASH_SHA1_MAX_REG (SHA1_DIGEST_SIZE / HASH_REG_SIZEOF) +#define HASH_SHA256_MAX_REG (SHA256_DIGEST_SIZE / HASH_REG_SIZEOF) + +/* + * HASH bit numbers, used by device, setting in dev->hash_flags with + * functions set_bit(), clear_bit() or tested with test_bit() or BIT(), + * to keep HASH state BUSY or FREE, or to signal state from irq_handler + * to hash_tasklet. SGS keep track of allocated memory for scatterlist + */ +#define HASH_FLAGS_BUSY 0 +#define HASH_FLAGS_FINAL 1 +#define HASH_FLAGS_DMA_ACTIVE 2 +#define HASH_FLAGS_OUTPUT_READY 3 +#define HASH_FLAGS_DMA_READY 4 +#define HASH_FLAGS_SGS_COPIED 5 +#define HASH_FLAGS_SGS_ALLOCED 6 + +/* HASH HW constants */ +#define BUFLEN HASH_BLOCK_SIZE + +#define SSS_HASH_DMA_LEN_ALIGN 8 +#define SSS_HASH_DMA_ALIGN_MASK (SSS_HASH_DMA_LEN_ALIGN - 1) + +#define SSS_HASH_QUEUE_LENGTH 10 + /** * struct samsung_aes_variant - platform specific SSS driver data * @aes_offset: AES register offset from SSS module's base. + * @hash_offset: HASH register offset from SSS module's base. * * Specifies platform specific configuration of SSS module. * Note: A structure for driver specific platform data is used for future @@ -156,6 +244,7 @@ */ struct samsung_aes_variant { unsigned int aes_offset; + unsigned int hash_offset; }; struct s5p_aes_reqctx { @@ -195,6 +284,19 @@ struct s5p_aes_ctx { * protects against concurrent access to these fields. * @lock: Lock for protecting both access to device hardware registers * and fields related to current request (including the busy field). + * @res: Resources for hash. + * @io_hash_base: Per-variant offset for HASH block IO memory. + * @hash_lock: Lock for protecting hash_req, hash_queue and hash_flags + * variable. + * @hash_flags: Flags for current HASH op. + * @hash_queue: Async hash queue. + * @hash_tasklet: New HASH request scheduling job. + * @xmit_buf: Buffer for current HASH request transfer into SSS block. + * @hash_req: Current request sending to SSS HASH block. + * @hash_sg_iter: Scatterlist transferred through DMA into SSS HASH block. + * @hash_sg_cnt: Counter for hash_sg_iter. + * + * @use_hash: true if HASH algs enabled */ struct s5p_aes_dev { struct device *dev; @@ -215,16 +317,83 @@ struct s5p_aes_dev { struct crypto_queue queue; bool busy; spinlock_t lock; + + struct resource *res; + void __iomem *io_hash_base; + + spinlock_t hash_lock; /* protect hash_ vars */ + unsigned long hash_flags; + struct crypto_queue hash_queue; + struct tasklet_struct hash_tasklet; + + u8 xmit_buf[BUFLEN]; + struct ahash_request *hash_req; + struct scatterlist *hash_sg_iter; + unsigned int hash_sg_cnt; + + bool use_hash; }; -static struct s5p_aes_dev *s5p_dev; +/** + * struct s5p_hash_reqctx - HASH request context + * @dev: Associated device + * @op_update: Current request operation (OP_UPDATE or OP_FINAL) + * @digcnt: Number of bytes processed by HW (without buffer[] ones) + * @digest: Digest message or IV for partial result + * @nregs: Number of HW registers for digest or IV read/write + * @engine: Bits for selecting type of HASH in SSS block + * @sg: sg for DMA transfer + * @sg_len: Length of sg for DMA transfer + * @sgl[]: sg for joining buffer and req->src scatterlist + * @skip: Skip offset in req->src for current op + * @total: Total number of bytes for current request + * @finup: Keep state for finup or final. + * @error: Keep track of error. + * @bufcnt: Number of bytes holded in buffer[] + * @buffer[]: For byte(s) from end of req->src in UPDATE op + */ +struct s5p_hash_reqctx { + struct s5p_aes_dev *dd; + bool op_update; + + u64 digcnt; + u8 digest[SHA256_DIGEST_SIZE]; + + unsigned int nregs; /* digest_size / sizeof(reg) */ + u32 engine; + + struct scatterlist *sg; + unsigned int sg_len; + struct scatterlist sgl[2]; + unsigned int skip; + unsigned int total; + bool finup; + bool error; + + u32 bufcnt; + u8 buffer[0]; +}; + +/** + * struct s5p_hash_ctx - HASH transformation context + * @dd: Associated device + * @flags: Bits for algorithm HASH. + * @fallback: Software transformation for zero message or size < BUFLEN. + */ +struct s5p_hash_ctx { + struct s5p_aes_dev *dd; + unsigned long flags; + struct crypto_shash *fallback; +}; static const struct samsung_aes_variant s5p_aes_data = { .aes_offset = 0x4000, + .hash_offset = 0x6000, }; static const struct samsung_aes_variant exynos_aes_data = { .aes_offset = 0x200, + .hash_offset = 0x400, }; static const struct of_device_id s5p_sss_dt_match[] = { @@ -254,6 +423,8 @@ static inline struct samsung_aes_variant *find_s5p_sss_version platform_get_device_id(pdev)->driver_data; } +static struct s5p_aes_dev *s5p_dev; + static void s5p_set_dma_indata(struct s5p_aes_dev *dev, struct scatterlist *sg) { SSS_WRITE(dev, FCBRDMAS, sg_dma_address(sg)); @@ -436,15 +607,65 @@ static int s5p_aes_rx(struct s5p_aes_dev *dev/*, bool *set_dma*/) return ret; } +static inline u32 s5p_hash_read(struct s5p_aes_dev *dd, u32 offset) +{ + return __raw_readl(dd->io_hash_base + offset); +} + +static inline void s5p_hash_write(struct s5p_aes_dev *dd, + u32 offset, u32 value) +{ + __raw_writel(value, dd->io_hash_base + offset); +} + +/** + * s5p_set_dma_hashdata() - start DMA with sg + * @dev: device + * @sg: scatterlist ready to DMA transmit + */ +static void s5p_set_dma_hashdata(struct s5p_aes_dev *dev, + struct scatterlist *sg) +{ + dev->hash_sg_cnt--; + SSS_WRITE(dev, FCHRDMAS, sg_dma_address(sg)); + SSS_WRITE(dev, FCHRDMAL, sg_dma_len(sg)); /* DMA starts */ +} + +/** + * s5p_hash_rx() - get next hash_sg_iter + * @dev: device + * + * Return: + * 2 if there is no more data and it is UPDATE op + * 1 if new receiving (input) data is ready and can be written to device + * 0 if there is no more data and it is FINAL op + */ +static int s5p_hash_rx(struct s5p_aes_dev *dev) +{ + if (dev->hash_sg_cnt > 0) { + dev->hash_sg_iter = sg_next(dev->hash_sg_iter); + return 1; + } + + set_bit(HASH_FLAGS_DMA_READY, &dev->hash_flags); + if (test_bit(HASH_FLAGS_FINAL, &dev->hash_flags)) + return 0; + + return 2; +} + static irqreturn_t s5p_aes_interrupt(int irq, void *dev_id) { struct platform_device *pdev = dev_id; struct s5p_aes_dev *dev = platform_get_drvdata(pdev); int err_dma_tx = 0; int err_dma_rx = 0; + int err_dma_hx = 0; bool tx_end = false; + bool hx_end = false; unsigned long flags; uint32_t status; + u32 st_bits; int err; spin_lock_irqsave(&dev->lock, flags); @@ -456,6 +677,8 @@ static irqreturn_t s5p_aes_interrupt(int irq, void *dev_id) * * If there is no more data in tx scatter list, call s5p_aes_complete() * and schedule new tasklet. + * + * Handle hx interrupt. If there is still data map next entry. */ status = SSS_READ(dev, FCINTSTAT); if (status & SSS_FCINTSTAT_BRDMAINT) @@ -467,7 +690,29 @@ static irqreturn_t s5p_aes_interrupt(int irq, void *dev_id) err_dma_tx = s5p_aes_tx(dev); } - SSS_WRITE(dev, FCINTPEND, status); + if (status & SSS_FCINTSTAT_HRDMAINT) + err_dma_hx = s5p_hash_rx(dev); + + st_bits = status & (SSS_FCINTSTAT_BRDMAINT | SSS_FCINTSTAT_BTDMAINT | + SSS_FCINTSTAT_HRDMAINT); + /* clear DMA bits */ + SSS_WRITE(dev, FCINTPEND, st_bits); + + /* clear HASH irq bits */ + if (status & (SSS_FCINTSTAT_HDONEINT | SSS_FCINTSTAT_HPARTINT)) { + /* cannot have both HPART and HDONE */ + if (status & SSS_FCINTSTAT_HPARTINT) + st_bits = SSS_HASH_STATUS_PARTIAL_DONE; + + if (status & SSS_FCINTSTAT_HDONEINT) + st_bits = SSS_HASH_STATUS_MSG_DONE; + + set_bit(HASH_FLAGS_OUTPUT_READY, &dev->hash_flags); + s5p_hash_write(dev, SSS_REG_HASH_STATUS, st_bits); + hx_end = true; + /* when DONE or PART, do not handle HASH DMA */ + err_dma_hx = 0; + } if (err_dma_rx < 0) { err = err_dma_rx; @@ -480,6 +725,8 @@ static irqreturn_t s5p_aes_interrupt(int irq, void *dev_id) if (tx_end) { s5p_sg_done(dev); + if (err_dma_hx == 1) + s5p_set_dma_hashdata(dev, dev->hash_sg_iter); spin_unlock_irqrestore(&dev->lock, flags); @@ -497,21 +744,1100 @@ static irqreturn_t s5p_aes_interrupt(int irq, void *dev_id) s5p_set_dma_outdata(dev, dev->sg_dst); if (err_dma_rx == 1) s5p_set_dma_indata(dev, dev->sg_src); + if (err_dma_hx == 1) + s5p_set_dma_hashdata(dev, dev->hash_sg_iter); spin_unlock_irqrestore(&dev->lock, flags); } - return IRQ_HANDLED; + goto hash_irq_end; error: s5p_sg_done(dev); dev->busy = false; + if (err_dma_hx == 1) + s5p_set_dma_hashdata(dev, dev->hash_sg_iter); + spin_unlock_irqrestore(&dev->lock, flags); s5p_aes_complete(dev, err); +hash_irq_end: + /* + * Note about else if: + * when hash_sg_iter reaches end and its UPDATE op, + * issue SSS_HASH_PAUSE and wait for HPART irq + */ + if (hx_end) + tasklet_schedule(&dev->hash_tasklet); + else if (err_dma_hx == 2) + s5p_hash_write(dev, SSS_REG_HASH_CTRL_PAUSE, + SSS_HASH_PAUSE); + return IRQ_HANDLED; } +/** + * s5p_hash_read_msg() - read message or IV from HW + * @req: AHASH request + */ +static void s5p_hash_read_msg(struct ahash_request *req) +{ + struct s5p_hash_reqctx *ctx = ahash_request_ctx(req); + struct s5p_aes_dev *dd = ctx->dd; + u32 *hash = (u32 *)ctx->digest; + unsigned int i; + + for (i = 0; i < ctx->nregs; i++) + hash[i] = s5p_hash_read(dd, SSS_REG_HASH_OUT(i)); +} + +/** + * s5p_hash_write_ctx_iv() - write IV for next partial/finup op. + * @dd: device + * @ctx: request context + */ +static void s5p_hash_write_ctx_iv(struct s5p_aes_dev *dd, + struct s5p_hash_reqctx *ctx) +{ + u32 *hash = (u32 *)ctx->digest; + unsigned int i; + + for (i = 0; i < ctx->nregs; i++) + s5p_hash_write(dd, SSS_REG_HASH_IV(i), hash[i]); +} + +/** + * s5p_hash_write_iv() - write IV for next partial/finup op. + * @req: AHASH request + */ +static void s5p_hash_write_iv(struct ahash_request *req) +{ + struct s5p_hash_reqctx *ctx = ahash_request_ctx(req); + + s5p_hash_write_ctx_iv(ctx->dd, ctx); +} + +/** + * s5p_hash_copy_result() - copy digest into req->result + * @req: AHASH request + */ +static void s5p_hash_copy_result(struct ahash_request *req) +{ + struct s5p_hash_reqctx *ctx = ahash_request_ctx(req); + + if (!req->result) + return; + + memcpy(req->result, ctx->digest, ctx->nregs * HASH_REG_SIZEOF); +} + +/** + * s5p_hash_dma_flush() - flush HASH DMA + * @dev: secss device + */ +static void s5p_hash_dma_flush(struct s5p_aes_dev *dev) +{ + SSS_WRITE(dev, FCHRDMAC, SSS_FCHRDMAC_FLUSH); +} + +/** + * s5p_hash_dma_enable() - enable DMA mode for HASH + * @dev: secss device + * + * enable DMA mode for HASH + */ +static void s5p_hash_dma_enable(struct s5p_aes_dev *dev) +{ + s5p_hash_write(dev, SSS_REG_HASH_CTRL_FIFO, SSS_HASH_FIFO_MODE_DMA); +} + +/** + * s5p_hash_irq_disable() - disable irq HASH signals + * @dev: secss device + * @flags: bitfield with irq's to be disabled + */ +static void s5p_hash_irq_disable(struct s5p_aes_dev *dev, u32 flags) +{ + SSS_WRITE(dev, FCINTENCLR, flags); +} + +/** + * s5p_hash_irq_enable() - enable irq signals + * @dev: secss device + * @flags: bitfield with irq's to be enabled + */ +static void s5p_hash_irq_enable(struct s5p_aes_dev *dev, int flags) +{ + SSS_WRITE(dev, FCINTENSET, flags); +} + +/** + * s5p_hash_set_flow() - set flow inside SecSS AES/DES with/without HASH + * @dev: secss device + * @hashflow: HASH stream flow with/without crypto AES/DES + */ +static void s5p_hash_set_flow(struct s5p_aes_dev *dev, u32 hashflow) +{ + unsigned long flags; + u32 flow; + + spin_lock_irqsave(&dev->lock, flags); + + flow = SSS_READ(dev, FCFIFOCTRL); + flow &= ~SSS_HASHIN_MASK; + flow |= hashflow; + SSS_WRITE(dev, FCFIFOCTRL, flow); + + spin_unlock_irqrestore(&dev->lock, flags); +} + +/** + * s5p_ahash_dma_init() - enable DMA and set HASH flow inside SecSS + * @dev: secss device + * @hashflow: HASH stream flow with/without AES/DES + * + * flush HASH DMA and enable DMA, set HASH stream flow inside SecSS HW, + * enable HASH irq's HRDMA, HDONE, HPART + */ +static void s5p_ahash_dma_init(struct s5p_aes_dev *dev, u32 hashflow) +{ + s5p_hash_irq_disable(dev, SSS_FCINTENCLR_HRDMAINTENCLR | + SSS_FCINTENCLR_HDONEINTENCLR | + SSS_FCINTENCLR_HPARTINTENCLR); + s5p_hash_dma_flush(dev); + + s5p_hash_dma_enable(dev); + s5p_hash_set_flow(dev, hashflow & SSS_HASHIN_MASK); + s5p_hash_irq_enable(dev, SSS_FCINTENSET_HRDMAINTENSET | + SSS_FCINTENSET_HDONEINTENSET | + SSS_FCINTENSET_HPARTINTENSET); +} + +/** + * s5p_hash_write_ctrl() - prepare HASH block in SecSS for processing + * @dd: secss device + * @length: length for request + * @final: true if final op + * + * Prepare SSS HASH block for processing bytes in DMA mode. If it is called + * after previous updates, fill up IV words. For final, calculate and set + * lengths for HASH so SecSS can finalize hash. For partial, set SSS HASH + * length as 2^63 so it will be never reached and set to zero prelow and + * prehigh. + * + * This function does not start DMA transfer. + */ +static void s5p_hash_write_ctrl(struct s5p_aes_dev *dd, size_t length, + bool final) +{ + struct s5p_hash_reqctx *ctx = ahash_request_ctx(dd->hash_req); + u32 prelow, prehigh, low, high; + u32 configflags, swapflags; + u64 tmplen; + + configflags = ctx->engine | SSS_HASH_INIT_BIT; + + if (likely(ctx->digcnt)) { + s5p_hash_write_ctx_iv(dd, ctx); + configflags |= SSS_HASH_USER_IV_EN; + } + + if (final) { + /* number of bytes for last part */ + low = length; + high = 0; + /* total number of bits prev hashed */ + tmplen = ctx->digcnt * 8; + prelow = (u32)tmplen; + prehigh = (u32)(tmplen >> 32); + } else { + prelow = 0; + prehigh = 0; + low = 0; + high = BIT(31); + } + + swapflags = SSS_HASH_BYTESWAP_DI | SSS_HASH_BYTESWAP_DO | + SSS_HASH_BYTESWAP_IV | SSS_HASH_BYTESWAP_KEY; + + s5p_hash_write(dd, SSS_REG_HASH_MSG_SIZE_LOW, low); + s5p_hash_write(dd, SSS_REG_HASH_MSG_SIZE_HIGH, high); + s5p_hash_write(dd, SSS_REG_HASH_PRE_MSG_SIZE_LOW, prelow); + s5p_hash_write(dd, SSS_REG_HASH_PRE_MSG_SIZE_HIGH, prehigh); + + s5p_hash_write(dd, SSS_REG_HASH_CTRL_SWAP, swapflags); + s5p_hash_write(dd, SSS_REG_HASH_CTRL, configflags); +} + +/** + * s5p_hash_xmit_dma() - start DMA hash processing + * @dd: secss device + * @length: length for request + * @final: true if final op + * + * Update digcnt here, as it is needed for finup/final op. + */ +static int s5p_hash_xmit_dma(struct s5p_aes_dev *dd, size_t length, + bool final) +{ + struct s5p_hash_reqctx *ctx = ahash_request_ctx(dd->hash_req); + unsigned int cnt; + + cnt = dma_map_sg(dd->dev, ctx->sg, ctx->sg_len, DMA_TO_DEVICE); + if (!cnt) { + dev_err(dd->dev, "dma_map_sg error\n"); + ctx->error = true; + return -EINVAL; + } + + set_bit(HASH_FLAGS_DMA_ACTIVE, &dd->hash_flags); + dd->hash_sg_iter = ctx->sg; + dd->hash_sg_cnt = cnt; + s5p_hash_write_ctrl(dd, length, final); + ctx->digcnt += length; + ctx->total -= length; + + /* catch last interrupt */ + if (final) + set_bit(HASH_FLAGS_FINAL, &dd->hash_flags); + + s5p_set_dma_hashdata(dd, dd->hash_sg_iter); /* DMA starts */ + + return -EINPROGRESS; +} + +/** + * s5p_hash_copy_sgs() - copy request's bytes into new buffer + * @ctx: request context + * @sg: source scatterlist request + * @new_len: number of bytes to process from sg + * + * Allocate new buffer, copy data for HASH into it. If there was xmit_buf + * filled, copy it first, then copy data from sg into it. Prepare one sgl[0] + * with allocated buffer. + * + * Set bit in dd->hash_flag so we can free it after irq ends processing. + */ +static int s5p_hash_copy_sgs(struct s5p_hash_reqctx *ctx, + struct scatterlist *sg, unsigned int new_len) +{ + unsigned int pages, len; + void *buf; + + len = new_len + ctx->bufcnt; + pages = get_order(len); + + buf = (void *)__get_free_pages(GFP_ATOMIC, pages); + if (!buf) { + dev_err(ctx->dd->dev, "alloc pages for unaligned case.\n"); + ctx->error = true; + return -ENOMEM; + } + + if (ctx->bufcnt) + memcpy(buf, ctx->dd->xmit_buf, ctx->bufcnt); + + scatterwalk_map_and_copy(buf + ctx->bufcnt, sg, ctx->skip, + new_len, 0); + sg_init_table(ctx->sgl, 1); + sg_set_buf(ctx->sgl, buf, len); + ctx->sg = ctx->sgl; + ctx->sg_len = 1; + ctx->bufcnt = 0; + ctx->skip = 0; + set_bit(HASH_FLAGS_SGS_COPIED, &ctx->dd->hash_flags); + + return 0; +} + +/** + * s5p_hash_copy_sg_lists() - copy sg list and make fixes in copy + * @ctx: request context + * @sg: source scatterlist request + * @new_len: number of bytes to process from sg + * + * Allocate new scatterlist table, copy data for HASH into it. If there was + * xmit_buf filled, prepare it first, then copy page, length and offset from + * source sg into it, adjusting begin and/or end for skip offset and + * hash_later value. + * + * Resulting sg table will be assigned to ctx->sg. Set flag so we can free + * it after irq ends processing. + */ +static int s5p_hash_copy_sg_lists(struct s5p_hash_reqctx *ctx, + struct scatterlist *sg, unsigned int new_len) +{ + unsigned int skip = ctx->skip, n = sg_nents(sg); + struct scatterlist *tmp; + unsigned int len; + + if (ctx->bufcnt) + n++; + + ctx->sg = kmalloc_array(n, sizeof(*sg), GFP_KERNEL); + if (!ctx->sg) { + ctx->error = true; + return -ENOMEM; + } + + sg_init_table(ctx->sg, n); + + tmp = ctx->sg; + + ctx->sg_len = 0; + + if (ctx->bufcnt) { + sg_set_buf(tmp, ctx->dd->xmit_buf, ctx->bufcnt); + tmp = sg_next(tmp); + ctx->sg_len++; + } + + while (sg && skip >= sg->length) { + skip -= sg->length; + sg = sg_next(sg); + } + + while (sg && new_len) { + len = sg->length - skip; + if (new_len < len) + len = new_len; + + new_len -= len; + sg_set_page(tmp, sg_page(sg), len, sg->offset + skip); + skip = 0; + if (new_len <= 0) + sg_mark_end(tmp); + + tmp = sg_next(tmp); + ctx->sg_len++; + sg = sg_next(sg); + } + + set_bit(HASH_FLAGS_SGS_ALLOCED, &ctx->dd->hash_flags); + + return 0; +} + +/** + * s5p_hash_prepare_sgs() - prepare sg for processing + * @ctx: request context + * @sg: source scatterlist request + * @nbytes: number of bytes to process from sg + * @final: final flag + * + * Check two conditions: (1) if buffers in sg have len aligned data, and (2) + * sg table have good aligned elements (list_ok). If one of this checks fails, + * then either (1) allocates new buffer for data with s5p_hash_copy_sgs, copy + * data into this buffer and prepare request in sgl, or (2) allocates new sg + * table and prepare sg elements. + * + * For digest or finup all conditions can be good, and we may not need any + * fixes. + */ +static int s5p_hash_prepare_sgs(struct s5p_hash_reqctx *ctx, + struct scatterlist *sg, + unsigned int new_len, bool final) +{ + unsigned int skip = ctx->skip, nbytes = new_len, n = 0; + bool aligned = true, list_ok = true; + struct scatterlist *sg_tmp = sg; + + if (!sg || !sg->length || !new_len) + return 0; + + if (skip || !final) + list_ok = false; + + while (nbytes > 0 && sg_tmp) { + n++; + if (skip >= sg_tmp->length) { + skip -= sg_tmp->length; + if (!sg_tmp->length) { + aligned = false; + break; + } + } else { + if (!IS_ALIGNED(sg_tmp->length - skip, BUFLEN)) { + aligned = false; + break; + } + + if (nbytes < sg_tmp->length - skip) { + list_ok = false; + break; + } + + nbytes -= sg_tmp->length - skip; + skip = 0; + } + + sg_tmp = sg_next(sg_tmp); + } + + if (!aligned) + return s5p_hash_copy_sgs(ctx, sg, new_len); + else if (!list_ok) + return s5p_hash_copy_sg_lists(ctx, sg, new_len); + + /* + * Have aligned data from previous operation and/or current + * Note: will enter here only if (digest or finup) and aligned + */ + if (ctx->bufcnt) { + ctx->sg_len = n; + sg_init_table(ctx->sgl, 2); + sg_set_buf(ctx->sgl, ctx->dd->xmit_buf, ctx->bufcnt); + sg_chain(ctx->sgl, 2, sg); + ctx->sg = ctx->sgl; + ctx->sg_len++; + } else { + ctx->sg = sg; + ctx->sg_len = n; + } + + return 0; +} + +/** + * s5p_hash_prepare_request() - prepare request for processing + * @req: AHASH request + * @update: true if UPDATE op + * + * Note 1: we can have update flag _and_ final flag at the same time. + * Note 2: we enter here when digcnt > BUFLEN (=HASH_BLOCK_SIZE) or + * either req->nbytes or ctx->bufcnt + req->nbytes is > BUFLEN or + * we have final op + */ +static int s5p_hash_prepare_request(struct ahash_request *req, bool update) +{ + struct s5p_hash_reqctx *ctx = ahash_request_ctx(req); + bool final = ctx->finup; + int xmit_len, hash_later, nbytes; + int ret; + + if (!req) + return 0; + + if (update) + nbytes = req->nbytes; + else + nbytes = 0; + + ctx->total = nbytes + ctx->bufcnt; + if (!ctx->total) + return 0; + + if (nbytes && (!IS_ALIGNED(ctx->bufcnt, BUFLEN))) { + /* bytes left from previous request, so fill up to BUFLEN */ + int len = BUFLEN - ctx->bufcnt % BUFLEN; + + if (len > nbytes) + len = nbytes; + + scatterwalk_map_and_copy(ctx->buffer + ctx->bufcnt, req->src, + 0, len, 0); + ctx->bufcnt += len; + nbytes -= len; + ctx->skip = len; + } else { + ctx->skip = 0; + } + + if (ctx->bufcnt) + memcpy(ctx->dd->xmit_buf, ctx->buffer, ctx->bufcnt); + + xmit_len = ctx->total; + if (final) { + hash_later = 0; + } else { + if (IS_ALIGNED(xmit_len, BUFLEN)) + xmit_len -= BUFLEN; + else + xmit_len -= xmit_len & (BUFLEN - 1); + + hash_later = ctx->total - xmit_len; + /* copy hash_later bytes from end of req->src */ + /* previous bytes are in xmit_buf, so no overwrite */ + scatterwalk_map_and_copy(ctx->buffer, req->src, + req->nbytes - hash_later, + hash_later, 0); + } + + if (xmit_len > BUFLEN) { + ret = s5p_hash_prepare_sgs(ctx, req->src, nbytes - hash_later, + final); + if (ret) + return ret; + } else { + /* have buffered data only */ + if (unlikely(!ctx->bufcnt)) { + /* first update didn't fill up buffer */ + scatterwalk_map_and_copy(ctx->dd->xmit_buf, req->src, + 0, xmit_len, 0); + } + + sg_init_table(ctx->sgl, 1); + sg_set_buf(ctx->sgl, ctx->dd->xmit_buf, xmit_len); + + ctx->sg = ctx->sgl; + ctx->sg_len = 1; + } + + ctx->bufcnt = hash_later; + if (!final) + ctx->total = xmit_len; + + return 0; +} + +/** + * s5p_hash_update_dma_stop() - unmap DMA + * @dd: secss device + * + * Unmap scatterlist ctx->sg. + */ +static void s5p_hash_update_dma_stop(struct s5p_aes_dev *dd) +{ + struct s5p_hash_reqctx *ctx = ahash_request_ctx(dd->hash_req); + + dma_unmap_sg(dd->dev, ctx->sg, ctx->sg_len, DMA_TO_DEVICE); + clear_bit(HASH_FLAGS_DMA_ACTIVE, &dd->hash_flags); +} + +/** + * s5p_hash_finish() - copy calculated digest to crypto layer + * @req: AHASH request + */ +static void s5p_hash_finish(struct ahash_request *req) +{ + struct s5p_hash_reqctx *ctx = ahash_request_ctx(req); + struct s5p_aes_dev *dd = ctx->dd; + + if (ctx->digcnt) + s5p_hash_copy_result(req); + + dev_dbg(dd->dev, "hash_finish digcnt: %lld\n", ctx->digcnt); +} + +/** + * s5p_hash_finish_req() - finish request + * @req: AHASH request + * @err: error + */ +static void s5p_hash_finish_req(struct ahash_request *req, int err) +{ + struct s5p_hash_reqctx *ctx = ahash_request_ctx(req); + struct s5p_aes_dev *dd = ctx->dd; + unsigned long flags; + + if (test_bit(HASH_FLAGS_SGS_COPIED, &dd->hash_flags)) + free_pages((unsigned long)sg_virt(ctx->sg), + get_order(ctx->sg->length)); + + if (test_bit(HASH_FLAGS_SGS_ALLOCED, &dd->hash_flags)) + kfree(ctx->sg); + + ctx->sg = NULL; + dd->hash_flags &= ~(BIT(HASH_FLAGS_SGS_ALLOCED) | + BIT(HASH_FLAGS_SGS_COPIED)); + + if (!err && !ctx->error) { + s5p_hash_read_msg(req); + if (test_bit(HASH_FLAGS_FINAL, &dd->hash_flags)) + s5p_hash_finish(req); + } else { + ctx->error = true; + } + + spin_lock_irqsave(&dd->hash_lock, flags); + dd->hash_flags &= ~(BIT(HASH_FLAGS_BUSY) | BIT(HASH_FLAGS_FINAL) | + BIT(HASH_FLAGS_DMA_READY) | + BIT(HASH_FLAGS_OUTPUT_READY)); + spin_unlock_irqrestore(&dd->hash_lock, flags); + + if (req->base.complete) + req->base.complete(&req->base, err); +} + +/** + * s5p_hash_handle_queue() - handle hash queue + * @dd: device s5p_aes_dev + * @req: AHASH request + * + * If req!=NULL enqueue it on dd->queue, if FLAGS_BUSY is not set on the + * device then processes the first request from the dd->queue + * + * Returns: see s5p_hash_final below. + */ +static int s5p_hash_handle_queue(struct s5p_aes_dev *dd, + struct ahash_request *req) +{ + struct crypto_async_request *async_req, *backlog; + struct s5p_hash_reqctx *ctx; + unsigned long flags; + int err = 0, ret = 0; + +retry: + spin_lock_irqsave(&dd->hash_lock, flags); + if (req) + ret = ahash_enqueue_request(&dd->hash_queue, req); + + if (test_bit(HASH_FLAGS_BUSY, &dd->hash_flags)) { + spin_unlock_irqrestore(&dd->hash_lock, flags); + return ret; + } + + backlog = crypto_get_backlog(&dd->hash_queue); + async_req = crypto_dequeue_request(&dd->hash_queue); + if (async_req) + set_bit(HASH_FLAGS_BUSY, &dd->hash_flags); + + spin_unlock_irqrestore(&dd->hash_lock, flags); + + if (!async_req) + return ret; + + if (backlog) + backlog->complete(backlog, -EINPROGRESS); + + req = ahash_request_cast(async_req); + dd->hash_req = req; + ctx = ahash_request_ctx(req); + + err = s5p_hash_prepare_request(req, ctx->op_update); + if (err || !ctx->total) + goto out; + + dev_dbg(dd->dev, "handling new req, op_update: %u, nbytes: %d\n", + ctx->op_update, req->nbytes); + + s5p_ahash_dma_init(dd, SSS_HASHIN_INDEPENDENT); + if (ctx->digcnt) + s5p_hash_write_iv(req); /* restore hash IV */ + + if (ctx->op_update) { /* HASH_OP_UPDATE */ + err = s5p_hash_xmit_dma(dd, ctx->total, ctx->finup); + if (err != -EINPROGRESS && ctx->finup && !ctx->error) + /* no final() after finup() */ + err = s5p_hash_xmit_dma(dd, ctx->total, true); + } else { /* HASH_OP_FINAL */ + err = s5p_hash_xmit_dma(dd, ctx->total, true); + } +out: + if (err != -EINPROGRESS) { + /* hash_tasklet_cb will not finish it, so do it here */ + s5p_hash_finish_req(req, err); + req = NULL; + + /* + * Execute next request immediately if there is anything + * in queue. + */ + goto retry; + } + + return ret; +} + +/** + * s5p_hash_tasklet_cb() - hash tasklet + * @data: ptr to s5p_aes_dev + */ +static void s5p_hash_tasklet_cb(unsigned long data) +{ + struct s5p_aes_dev *dd = (struct s5p_aes_dev *)data; + + if (!test_bit(HASH_FLAGS_BUSY, &dd->hash_flags)) { + s5p_hash_handle_queue(dd, NULL); + return; + } + + if (test_bit(HASH_FLAGS_DMA_READY, &dd->hash_flags)) { + if (test_and_clear_bit(HASH_FLAGS_DMA_ACTIVE, + &dd->hash_flags)) { + s5p_hash_update_dma_stop(dd); + } + + if (test_and_clear_bit(HASH_FLAGS_OUTPUT_READY, + &dd->hash_flags)) { + /* hash or semi-hash ready */ + clear_bit(HASH_FLAGS_DMA_READY, &dd->hash_flags); + goto finish; + } + } + + return; + +finish: + /* finish curent request */ + s5p_hash_finish_req(dd->hash_req, 0); + + /* If we are not busy, process next req */ + if (!test_bit(HASH_FLAGS_BUSY, &dd->hash_flags)) + s5p_hash_handle_queue(dd, NULL); +} + +/** + * s5p_hash_enqueue() - enqueue request + * @req: AHASH request + * @op: operation UPDATE (true) or FINAL (false) + * + * Returns: see s5p_hash_final below. + */ +static int s5p_hash_enqueue(struct ahash_request *req, bool op) +{ + struct s5p_hash_reqctx *ctx = ahash_request_ctx(req); + struct s5p_hash_ctx *tctx = crypto_tfm_ctx(req->base.tfm); + + ctx->op_update = op; + + return s5p_hash_handle_queue(tctx->dd, req); +} + +/** + * s5p_hash_update() - process the hash input data + * @req: AHASH request + * + * If request will fit in buffer, copy it and return immediately + * else enqueue it with OP_UPDATE. + * + * Returns: see s5p_hash_final below. + */ +static int s5p_hash_update(struct ahash_request *req) +{ + struct s5p_hash_reqctx *ctx = ahash_request_ctx(req); + + if (!req->nbytes) + return 0; + + if (ctx->bufcnt + req->nbytes <= BUFLEN) { + scatterwalk_map_and_copy(ctx->buffer + ctx->bufcnt, req->src, + 0, req->nbytes, 0); + ctx->bufcnt += req->nbytes; + return 0; + } + + return s5p_hash_enqueue(req, true); /* HASH_OP_UPDATE */ +} + +/** + * s5p_hash_shash_digest() - calculate shash digest + * @tfm: crypto transformation + * @flags: tfm flags + * @data: input data + * @len: length of data + * @out: output buffer + */ +static int s5p_hash_shash_digest(struct crypto_shash *tfm, u32 flags, + const u8 *data, unsigned int len, u8 *out) +{ + SHASH_DESC_ON_STACK(shash, tfm); + + shash->tfm = tfm; + shash->flags = flags & ~CRYPTO_TFM_REQ_MAY_SLEEP; + + return crypto_shash_digest(shash, data, len, out); +} + +/** + * s5p_hash_final_shash() - calculate shash digest + * @req: AHASH request + */ +static int s5p_hash_final_shash(struct ahash_request *req) +{ + struct s5p_hash_ctx *tctx = crypto_tfm_ctx(req->base.tfm); + struct s5p_hash_reqctx *ctx = ahash_request_ctx(req); + + return s5p_hash_shash_digest(tctx->fallback, req->base.flags, + ctx->buffer, ctx->bufcnt, req->result); +} + +/** + * s5p_hash_final() - close up hash and calculate digest + * @req: AHASH request + * + * Note: in final req->src do not have any data, and req->nbytes can be + * non-zero. + * + * If there were no input data processed yet and the buffered hash data is + * less than BUFLEN (64) then calculate the final hash immediately by using + * SW algorithm fallback. + * + * Otherwise enqueues the current AHASH request with OP_FINAL operation op + * and finalize hash message in HW. Note that if digcnt!=0 then there were + * previous update op, so there are always some buffered bytes in ctx->buffer, + * which means that ctx->bufcnt!=0 + * + * Returns: + * 0 if the request has been processed immediately, + * -EINPROGRESS if the operation has been queued for later execution or is set + * to processing by HW, + * -EBUSY if queue is full and request should be resubmitted later, other + * negative values on error. + */ +static int s5p_hash_final(struct ahash_request *req) +{ + struct s5p_hash_reqctx *ctx = ahash_request_ctx(req); + + ctx->finup = true; + if (ctx->error) + return -EINVAL; /* uncompleted hash is not needed */ + + if (!ctx->digcnt && ctx->bufcnt < BUFLEN) + return s5p_hash_final_shash(req); + + return s5p_hash_enqueue(req, false); /* HASH_OP_FINAL */ +} + +/** + * s5p_hash_finup() - process last req->src and calculate digest + * @req: AHASH request containing the last update data + * + * Return values: see s5p_hash_final above. + */ +static int s5p_hash_finup(struct ahash_request *req) +{ + struct s5p_hash_reqctx *ctx = ahash_request_ctx(req); + int err1, err2; + + ctx->finup = true; + + err1 = s5p_hash_update(req); + if (err1 == -EINPROGRESS || err1 == -EBUSY) + return err1; + + /* + * final() has to be always called to cleanup resources even if + * update() failed, except EINPROGRESS or calculate digest for small + * size + */ + err2 = s5p_hash_final(req); + + return err1 ?: err2; +} + +/** + * s5p_hash_init() - initialize AHASH request contex + * @req: AHASH request + * + * Init async hash request context. + */ +static int s5p_hash_init(struct ahash_request *req) +{ + struct s5p_hash_reqctx *ctx = ahash_request_ctx(req); + struct crypto_ahash *tfm = crypto_ahash_reqtfm(req); + struct s5p_hash_ctx *tctx = crypto_ahash_ctx(tfm); + + ctx->dd = tctx->dd; + ctx->error = false; + ctx->finup = false; + ctx->bufcnt = 0; + ctx->digcnt = 0; + ctx->total = 0; + ctx->skip = 0; + + dev_dbg(tctx->dd->dev, "init: digest size: %d\n", + crypto_ahash_digestsize(tfm)); + + switch (crypto_ahash_digestsize(tfm)) { + case MD5_DIGEST_SIZE: + ctx->engine = SSS_HASH_ENGINE_MD5; + ctx->nregs = HASH_MD5_MAX_REG; + break; + case SHA1_DIGEST_SIZE: + ctx->engine = SSS_HASH_ENGINE_SHA1; + ctx->nregs = HASH_SHA1_MAX_REG; + break; + case SHA256_DIGEST_SIZE: + ctx->engine = SSS_HASH_ENGINE_SHA256; + ctx->nregs = HASH_SHA256_MAX_REG; + break; + default: + ctx->error = true; + return -EINVAL; + } + + return 0; +} + +/** + * s5p_hash_digest - calculate digest from req->src + * @req: AHASH request + * + * Return values: see s5p_hash_final above. + */ +static int s5p_hash_digest(struct ahash_request *req) +{ + return s5p_hash_init(req) ?: s5p_hash_finup(req); +} + +/** + * s5p_hash_cra_init_alg - init crypto alg transformation + * @tfm: crypto transformation + */ +static int s5p_hash_cra_init_alg(struct crypto_tfm *tfm) +{ + struct s5p_hash_ctx *tctx = crypto_tfm_ctx(tfm); + const char *alg_name = crypto_tfm_alg_name(tfm); + + tctx->dd = s5p_dev; + /* Allocate a fallback and abort if it failed. */ + tctx->fallback = crypto_alloc_shash(alg_name, 0, + CRYPTO_ALG_NEED_FALLBACK); + if (IS_ERR(tctx->fallback)) { + pr_err("fallback alloc fails for '%s'\n", alg_name); + return PTR_ERR(tctx->fallback); + } + + crypto_ahash_set_reqsize(__crypto_ahash_cast(tfm), + sizeof(struct s5p_hash_reqctx) + BUFLEN); + + return 0; +} + +/** + * s5p_hash_cra_init - init crypto tfm + * @tfm: crypto transformation + */ +static int s5p_hash_cra_init(struct crypto_tfm *tfm) +{ + return s5p_hash_cra_init_alg(tfm); +} + +/** + * s5p_hash_cra_exit - exit crypto tfm + * @tfm: crypto transformation + * + * free allocated fallback + */ +static void s5p_hash_cra_exit(struct crypto_tfm *tfm) +{ + struct s5p_hash_ctx *tctx = crypto_tfm_ctx(tfm); + + crypto_free_shash(tctx->fallback); + tctx->fallback = NULL; +} + +/** + * s5p_hash_export - export hash state + * @req: AHASH request + * @out: buffer for exported state + */ +static int s5p_hash_export(struct ahash_request *req, void *out) +{ + struct s5p_hash_reqctx *ctx = ahash_request_ctx(req); + + memcpy(out, ctx, sizeof(*ctx) + ctx->bufcnt); + + return 0; +} + +/** + * s5p_hash_import - import hash state + * @req: AHASH request + * @in: buffer with state to be imported from + */ +static int s5p_hash_import(struct ahash_request *req, const void *in) +{ + struct s5p_hash_reqctx *ctx = ahash_request_ctx(req); + struct crypto_ahash *tfm = crypto_ahash_reqtfm(req); + struct s5p_hash_ctx *tctx = crypto_ahash_ctx(tfm); + const struct s5p_hash_reqctx *ctx_in = in; + + memcpy(ctx, in, sizeof(*ctx) + BUFLEN); + if (ctx_in->bufcnt < 0 || ctx_in->bufcnt > BUFLEN) { + ctx->error = true; + return -EINVAL; + } + + ctx->dd = tctx->dd; + ctx->error = false; + + return 0; +} + +static struct ahash_alg algs_sha1_md5_sha256[] = { +{ + .init = s5p_hash_init, + .update = s5p_hash_update, + .final = s5p_hash_final, + .finup = s5p_hash_finup, + .digest = s5p_hash_digest, + .export = s5p_hash_export, + .import = s5p_hash_import, + .halg.statesize = sizeof(struct s5p_hash_reqctx) + BUFLEN, + .halg.digestsize = SHA1_DIGEST_SIZE, + .halg.base = { + .cra_name = "sha1", + .cra_driver_name = "exynos-sha1", + .cra_priority = 100, + .cra_flags = CRYPTO_ALG_TYPE_AHASH | + CRYPTO_ALG_KERN_DRIVER_ONLY | + CRYPTO_ALG_ASYNC | + CRYPTO_ALG_NEED_FALLBACK, + .cra_blocksize = HASH_BLOCK_SIZE, + .cra_ctxsize = sizeof(struct s5p_hash_ctx), + .cra_alignmask = SSS_HASH_DMA_ALIGN_MASK, + .cra_module = THIS_MODULE, + .cra_init = s5p_hash_cra_init, + .cra_exit = s5p_hash_cra_exit, + } +}, +{ + .init = s5p_hash_init, + .update = s5p_hash_update, + .final = s5p_hash_final, + .finup = s5p_hash_finup, + .digest = s5p_hash_digest, + .export = s5p_hash_export, + .import = s5p_hash_import, + .halg.statesize = sizeof(struct s5p_hash_reqctx) + BUFLEN, + .halg.digestsize = MD5_DIGEST_SIZE, + .halg.base = { + .cra_name = "md5", + .cra_driver_name = "exynos-md5", + .cra_priority = 100, + .cra_flags = CRYPTO_ALG_TYPE_AHASH | + CRYPTO_ALG_KERN_DRIVER_ONLY | + CRYPTO_ALG_ASYNC | + CRYPTO_ALG_NEED_FALLBACK, + .cra_blocksize = HASH_BLOCK_SIZE, + .cra_ctxsize = sizeof(struct s5p_hash_ctx), + .cra_alignmask = SSS_HASH_DMA_ALIGN_MASK, + .cra_module = THIS_MODULE, + .cra_init = s5p_hash_cra_init, + .cra_exit = s5p_hash_cra_exit, + } +}, +{ + .init = s5p_hash_init, + .update = s5p_hash_update, + .final = s5p_hash_final, + .finup = s5p_hash_finup, + .digest = s5p_hash_digest, + .export = s5p_hash_export, + .import = s5p_hash_import, + .halg.statesize = sizeof(struct s5p_hash_reqctx) + BUFLEN, + .halg.digestsize = SHA256_DIGEST_SIZE, + .halg.base = { + .cra_name = "sha256", + .cra_driver_name = "exynos-sha256", + .cra_priority = 100, + .cra_flags = CRYPTO_ALG_TYPE_AHASH | + CRYPTO_ALG_KERN_DRIVER_ONLY | + CRYPTO_ALG_ASYNC | + CRYPTO_ALG_NEED_FALLBACK, + .cra_blocksize = HASH_BLOCK_SIZE, + .cra_ctxsize = sizeof(struct s5p_hash_ctx), + .cra_alignmask = SSS_HASH_DMA_ALIGN_MASK, + .cra_module = THIS_MODULE, + .cra_init = s5p_hash_cra_init, + .cra_exit = s5p_hash_cra_exit, + } +} + +}; + static void s5p_set_aes(struct s5p_aes_dev *dev, uint8_t *key, uint8_t *iv, unsigned int keylen) { @@ -829,6 +2155,7 @@ static int s5p_aes_probe(struct platform_device *pdev) struct samsung_aes_variant *variant; struct s5p_aes_dev *pdata; struct resource *res; + unsigned int hash_i; if (s5p_dev) return -EEXIST; @@ -837,12 +2164,34 @@ static int s5p_aes_probe(struct platform_device *pdev) if (!pdata) return -ENOMEM; + variant = find_s5p_sss_version(pdev); res = platform_get_resource(pdev, IORESOURCE_MEM, 0); - pdata->ioaddr = devm_ioremap_resource(&pdev->dev, res); - if (IS_ERR(pdata->ioaddr)) - return PTR_ERR(pdata->ioaddr); - variant = find_s5p_sss_version(pdev); + /* + * Note: HASH and PRNG uses the same registers in secss, avoid + * overwrite each other. This will drop HASH when CONFIG_EXYNOS_RNG + * is enabled in config. We need larger size for HASH registers in + * secss, current describe only AES/DES + */ + if (IS_ENABLED(CONFIG_CRYPTO_DEV_EXYNOS_HASH)) { + if (variant == &exynos_aes_data) { + res->end += 0x300; + pdata->use_hash = true; + } + } + + pdata->res = res; + pdata->ioaddr = devm_ioremap_resource(&pdev->dev, res); + if (IS_ERR(pdata->ioaddr)) { + if (!pdata->use_hash) + return PTR_ERR(pdata->ioaddr); + /* try AES without HASH */ + res->end -= 0x300; + pdata->use_hash = false; + pdata->ioaddr = devm_ioremap_resource(&pdev->dev, res); + if (IS_ERR(pdata->ioaddr)) + return PTR_ERR(pdata->ioaddr); + } pdata->clk = devm_clk_get(dev, "secss"); if (IS_ERR(pdata->clk)) { @@ -857,8 +2206,10 @@ static int s5p_aes_probe(struct platform_device *pdev) } spin_lock_init(&pdata->lock); + spin_lock_init(&pdata->hash_lock); pdata->aes_ioaddr = pdata->ioaddr + variant->aes_offset; + pdata->io_hash_base = pdata->ioaddr + variant->hash_offset; pdata->irq_fc = platform_get_irq(pdev, 0); if (pdata->irq_fc < 0) { @@ -888,12 +2239,40 @@ static int s5p_aes_probe(struct platform_device *pdev) goto err_algs; } + if (pdata->use_hash) { + tasklet_init(&pdata->hash_tasklet, s5p_hash_tasklet_cb, + (unsigned long)pdata); + crypto_init_queue(&pdata->hash_queue, SSS_HASH_QUEUE_LENGTH); + + for (hash_i = 0; hash_i < ARRAY_SIZE(algs_sha1_md5_sha256); + hash_i++) { + struct ahash_alg *alg; + + alg = &algs_sha1_md5_sha256[hash_i]; + err = crypto_register_ahash(alg); + if (err) { + dev_err(dev, "can't register '%s': %d\n", + alg->halg.base.cra_driver_name, err); + goto err_hash; + } + } + } + dev_info(dev, "s5p-sss driver registered\n"); return 0; +err_hash: + for (j = hash_i - 1; j >= 0; j--) + crypto_unregister_ahash(&algs_sha1_md5_sha256[j]); + + tasklet_kill(&pdata->hash_tasklet); + res->end -= 0x300; + err_algs: - dev_err(dev, "can't register '%s': %d\n", algs[i].cra_name, err); + if (i < ARRAY_SIZE(algs)) + dev_err(dev, "can't register '%s': %d\n", algs[i].cra_name, + err); for (j = 0; j < i; j++) crypto_unregister_alg(&algs[j]); @@ -920,9 +2299,16 @@ static int s5p_aes_remove(struct platform_device *pdev) crypto_unregister_alg(&algs[i]); tasklet_kill(&pdata->tasklet); + if (pdata->use_hash) { + for (i = ARRAY_SIZE(algs_sha1_md5_sha256) - 1; i >= 0; i--) + crypto_unregister_ahash(&algs_sha1_md5_sha256[i]); - clk_disable_unprepare(pdata->clk); + pdata->res->end -= 0x300; + tasklet_kill(&pdata->hash_tasklet); + pdata->use_hash = false; + } + clk_disable_unprepare(pdata->clk); s5p_dev = NULL; return 0; @@ -942,3 +2328,4 @@ module_platform_driver(s5p_aes_crypto); MODULE_DESCRIPTION("S5PV210 AES hw acceleration support."); MODULE_LICENSE("GPL v2"); MODULE_AUTHOR("Vladimir Zapolskiy <vzapolskiy@gmail.com>"); +MODULE_AUTHOR("Kamil Konieczny <k.konieczny@partner.samsung.com>");

[v7,2/2] crypto: s5p-sss: Add HASH support for Exynos

Commit Message

Comments

Patch