From 099f53cb50e45ef617a9f1d63ceec799e489418b Mon Sep 17 00:00:00 2001 From: Dan Williams Date: Wed, 8 Apr 2009 14:28:37 -0700 Subject: async_tx: rename zero_sum to val 'zero_sum' does not properly describe the operation of generating parity and checking that it validates against an existing buffer. Change the name of the operation to 'val' (for 'validate'). This is in anticipation of the p+q case where it is a requirement to identify the target parity buffers separately from the source buffers, because the target parity buffers will not have corresponding pq coefficients. Reviewed-by: Andre Noll Acked-by: Maciej Sosnowski Signed-off-by: Dan Williams --- include/linux/dmaengine.h | 8 ++++---- 1 file changed, 4 insertions(+), 4 deletions(-) (limited to 'include/linux/dmaengine.h') diff --git a/include/linux/dmaengine.h b/include/linux/dmaengine.h index 2e2aa3df170c..6768727d00d7 100644 --- a/include/linux/dmaengine.h +++ b/include/linux/dmaengine.h @@ -55,8 +55,8 @@ enum dma_transaction_type { DMA_PQ_XOR, DMA_DUAL_XOR, DMA_PQ_UPDATE, - DMA_ZERO_SUM, - DMA_PQ_ZERO_SUM, + DMA_XOR_VAL, + DMA_PQ_VAL, DMA_MEMSET, DMA_MEMCPY_CRC32C, DMA_INTERRUPT, @@ -214,7 +214,7 @@ struct dma_async_tx_descriptor { * @device_free_chan_resources: release DMA channel's resources * @device_prep_dma_memcpy: prepares a memcpy operation * @device_prep_dma_xor: prepares a xor operation - * @device_prep_dma_zero_sum: prepares a zero_sum operation + * @device_prep_dma_xor_val: prepares a xor validation operation * @device_prep_dma_memset: prepares a memset operation * @device_prep_dma_interrupt: prepares an end of chain interrupt operation * @device_prep_slave_sg: prepares a slave dma operation @@ -243,7 +243,7 @@ struct dma_device { struct dma_async_tx_descriptor *(*device_prep_dma_xor)( struct dma_chan *chan, dma_addr_t dest, dma_addr_t *src, unsigned int src_cnt, size_t len, unsigned long flags); - struct dma_async_tx_descriptor *(*device_prep_dma_zero_sum)( + struct dma_async_tx_descriptor *(*device_prep_dma_xor_val)( struct dma_chan *chan, dma_addr_t *src, unsigned int src_cnt, size_t len, u32 *result, unsigned long flags); struct dma_async_tx_descriptor *(*device_prep_dma_memset)( -- cgit v1.2.3 From ad283ea4a3ce82cda2efe33163748a397b31b1eb Mon Sep 17 00:00:00 2001 From: Dan Williams Date: Sat, 29 Aug 2009 19:09:26 -0700 Subject: async_tx: add sum check flags Replace the flat zero_sum_result with a collection of flags to contain the P (xor) zero-sum result, and the soon to be utilized Q (raid6 reed solomon syndrome) zero-sum result. Use the SUM_CHECK_ namespace instead of DMA_ since these flags will be used on non-dma-zero-sum enabled platforms. Reviewed-by: Andre Noll Acked-by: Maciej Sosnowski Signed-off-by: Dan Williams --- include/linux/dmaengine.h | 21 ++++++++++++++++++++- 1 file changed, 20 insertions(+), 1 deletion(-) (limited to 'include/linux/dmaengine.h') diff --git a/include/linux/dmaengine.h b/include/linux/dmaengine.h index 6768727d00d7..02447afcebad 100644 --- a/include/linux/dmaengine.h +++ b/include/linux/dmaengine.h @@ -86,6 +86,25 @@ enum dma_ctrl_flags { DMA_COMPL_SKIP_DEST_UNMAP = (1 << 3), }; +/** + * enum sum_check_bits - bit position of pq_check_flags + */ +enum sum_check_bits { + SUM_CHECK_P = 0, + SUM_CHECK_Q = 1, +}; + +/** + * enum pq_check_flags - result of async_{xor,pq}_zero_sum operations + * @SUM_CHECK_P_RESULT - 1 if xor zero sum error, 0 otherwise + * @SUM_CHECK_Q_RESULT - 1 if reed-solomon zero sum error, 0 otherwise + */ +enum sum_check_flags { + SUM_CHECK_P_RESULT = (1 << SUM_CHECK_P), + SUM_CHECK_Q_RESULT = (1 << SUM_CHECK_Q), +}; + + /** * dma_cap_mask_t - capabilities bitmap modeled after cpumask_t. * See linux/cpumask.h @@ -245,7 +264,7 @@ struct dma_device { unsigned int src_cnt, size_t len, unsigned long flags); struct dma_async_tx_descriptor *(*device_prep_dma_xor_val)( struct dma_chan *chan, dma_addr_t *src, unsigned int src_cnt, - size_t len, u32 *result, unsigned long flags); + size_t len, enum sum_check_flags *result, unsigned long flags); struct dma_async_tx_descriptor *(*device_prep_dma_memset)( struct dma_chan *chan, dma_addr_t dest, int value, size_t len, unsigned long flags); -- cgit v1.2.3 From b2f46fd8ef3dff2ab30f31126833f78b7480283a Mon Sep 17 00:00:00 2001 From: Dan Williams Date: Tue, 14 Jul 2009 12:20:36 -0700 Subject: async_tx: add support for asynchronous GF multiplication [ Based on an original patch by Yuri Tikhonov ] This adds support for doing asynchronous GF multiplication by adding two additional functions to the async_tx API: async_gen_syndrome() does simultaneous XOR and Galois field multiplication of sources. async_syndrome_val() validates the given source buffers against known P and Q values. When a request is made to run async_pq against more than the hardware maximum number of supported sources we need to reuse the previous generated P and Q values as sources into the next operation. Care must be taken to remove Q from P' and P from Q'. For example to perform a 5 source pq op with hardware that only supports 4 sources at a time the following approach is taken: p, q = PQ(src0, src1, src2, src3, COEF({01}, {02}, {04}, {08})) p', q' = PQ(p, q, q, src4, COEF({00}, {01}, {00}, {10})) p' = p + q + q + src4 = p + src4 q' = {00}*p + {01}*q + {00}*q + {10}*src4 = q + {10}*src4 Note: 4 is the minimum acceptable maxpq otherwise we punt to synchronous-software path. The DMA_PREP_CONTINUE flag indicates to the driver to reuse p and q as sources (in the above manner) and fill the remaining slots up to maxpq with the new sources/coefficients. Note1: Some devices have native support for P+Q continuation and can skip this extra work. Devices with this capability can advertise it with dma_set_maxpq. It is up to each driver how to handle the DMA_PREP_CONTINUE flag. Note2: The api supports disabling the generation of P when generating Q, this is ignored by the synchronous path but is implemented by some dma devices to save unnecessary writes. In this case the continuation algorithm is simplified to only reuse Q as a source. Cc: H. Peter Anvin Cc: David Woodhouse Signed-off-by: Yuri Tikhonov Signed-off-by: Ilya Yanok Reviewed-by: Andre Noll Acked-by: Maciej Sosnowski Signed-off-by: Dan Williams --- include/linux/dmaengine.h | 87 +++++++++++++++++++++++++++++++++++++++++++---- 1 file changed, 81 insertions(+), 6 deletions(-) (limited to 'include/linux/dmaengine.h') diff --git a/include/linux/dmaengine.h b/include/linux/dmaengine.h index 02447afcebad..ce010cd991d2 100644 --- a/include/linux/dmaengine.h +++ b/include/linux/dmaengine.h @@ -52,7 +52,7 @@ enum dma_status { enum dma_transaction_type { DMA_MEMCPY, DMA_XOR, - DMA_PQ_XOR, + DMA_PQ, DMA_DUAL_XOR, DMA_PQ_UPDATE, DMA_XOR_VAL, @@ -70,20 +70,28 @@ enum dma_transaction_type { /** * enum dma_ctrl_flags - DMA flags to augment operation preparation, - * control completion, and communicate status. + * control completion, and communicate status. * @DMA_PREP_INTERRUPT - trigger an interrupt (callback) upon completion of - * this transaction + * this transaction * @DMA_CTRL_ACK - the descriptor cannot be reused until the client - * acknowledges receipt, i.e. has has a chance to establish any - * dependency chains + * acknowledges receipt, i.e. has has a chance to establish any dependency + * chains * @DMA_COMPL_SKIP_SRC_UNMAP - set to disable dma-unmapping the source buffer(s) * @DMA_COMPL_SKIP_DEST_UNMAP - set to disable dma-unmapping the destination(s) + * @DMA_PREP_PQ_DISABLE_P - prevent generation of P while generating Q + * @DMA_PREP_PQ_DISABLE_Q - prevent generation of Q while generating P + * @DMA_PREP_CONTINUE - indicate to a driver that it is reusing buffers as + * sources that were the result of a previous operation, in the case of a PQ + * operation it continues the calculation with new sources */ enum dma_ctrl_flags { DMA_PREP_INTERRUPT = (1 << 0), DMA_CTRL_ACK = (1 << 1), DMA_COMPL_SKIP_SRC_UNMAP = (1 << 2), DMA_COMPL_SKIP_DEST_UNMAP = (1 << 3), + DMA_PREP_PQ_DISABLE_P = (1 << 4), + DMA_PREP_PQ_DISABLE_Q = (1 << 5), + DMA_PREP_CONTINUE = (1 << 6), }; /** @@ -226,6 +234,7 @@ struct dma_async_tx_descriptor { * @global_node: list_head for global dma_device_list * @cap_mask: one or more dma_capability flags * @max_xor: maximum number of xor sources, 0 if no capability + * @max_pq: maximum number of PQ sources and PQ-continue capability * @dev_id: unique device ID * @dev: struct device reference for dma mapping api * @device_alloc_chan_resources: allocate resources and return the @@ -234,6 +243,8 @@ struct dma_async_tx_descriptor { * @device_prep_dma_memcpy: prepares a memcpy operation * @device_prep_dma_xor: prepares a xor operation * @device_prep_dma_xor_val: prepares a xor validation operation + * @device_prep_dma_pq: prepares a pq operation + * @device_prep_dma_pq_val: prepares a pqzero_sum operation * @device_prep_dma_memset: prepares a memset operation * @device_prep_dma_interrupt: prepares an end of chain interrupt operation * @device_prep_slave_sg: prepares a slave dma operation @@ -248,7 +259,9 @@ struct dma_device { struct list_head channels; struct list_head global_node; dma_cap_mask_t cap_mask; - int max_xor; + unsigned short max_xor; + unsigned short max_pq; + #define DMA_HAS_PQ_CONTINUE (1 << 15) int dev_id; struct device *dev; @@ -265,6 +278,14 @@ struct dma_device { struct dma_async_tx_descriptor *(*device_prep_dma_xor_val)( struct dma_chan *chan, dma_addr_t *src, unsigned int src_cnt, size_t len, enum sum_check_flags *result, unsigned long flags); + struct dma_async_tx_descriptor *(*device_prep_dma_pq)( + struct dma_chan *chan, dma_addr_t *dst, dma_addr_t *src, + unsigned int src_cnt, const unsigned char *scf, + size_t len, unsigned long flags); + struct dma_async_tx_descriptor *(*device_prep_dma_pq_val)( + struct dma_chan *chan, dma_addr_t *pq, dma_addr_t *src, + unsigned int src_cnt, const unsigned char *scf, size_t len, + enum sum_check_flags *pqres, unsigned long flags); struct dma_async_tx_descriptor *(*device_prep_dma_memset)( struct dma_chan *chan, dma_addr_t dest, int value, size_t len, unsigned long flags); @@ -283,6 +304,60 @@ struct dma_device { void (*device_issue_pending)(struct dma_chan *chan); }; +static inline void +dma_set_maxpq(struct dma_device *dma, int maxpq, int has_pq_continue) +{ + dma->max_pq = maxpq; + if (has_pq_continue) + dma->max_pq |= DMA_HAS_PQ_CONTINUE; +} + +static inline bool dmaf_continue(enum dma_ctrl_flags flags) +{ + return (flags & DMA_PREP_CONTINUE) == DMA_PREP_CONTINUE; +} + +static inline bool dmaf_p_disabled_continue(enum dma_ctrl_flags flags) +{ + enum dma_ctrl_flags mask = DMA_PREP_CONTINUE | DMA_PREP_PQ_DISABLE_P; + + return (flags & mask) == mask; +} + +static inline bool dma_dev_has_pq_continue(struct dma_device *dma) +{ + return (dma->max_pq & DMA_HAS_PQ_CONTINUE) == DMA_HAS_PQ_CONTINUE; +} + +static unsigned short dma_dev_to_maxpq(struct dma_device *dma) +{ + return dma->max_pq & ~DMA_HAS_PQ_CONTINUE; +} + +/* dma_maxpq - reduce maxpq in the face of continued operations + * @dma - dma device with PQ capability + * @flags - to check if DMA_PREP_CONTINUE and DMA_PREP_PQ_DISABLE_P are set + * + * When an engine does not support native continuation we need 3 extra + * source slots to reuse P and Q with the following coefficients: + * 1/ {00} * P : remove P from Q', but use it as a source for P' + * 2/ {01} * Q : use Q to continue Q' calculation + * 3/ {00} * Q : subtract Q from P' to cancel (2) + * + * In the case where P is disabled we only need 1 extra source: + * 1/ {01} * Q : use Q to continue Q' calculation + */ +static inline int dma_maxpq(struct dma_device *dma, enum dma_ctrl_flags flags) +{ + if (dma_dev_has_pq_continue(dma) || !dmaf_continue(flags)) + return dma_dev_to_maxpq(dma); + else if (dmaf_p_disabled_continue(flags)) + return dma_dev_to_maxpq(dma) - 1; + else if (dmaf_continue(flags)) + return dma_dev_to_maxpq(dma) - 3; + BUG(); +} + /* --- public DMA engine API --- */ #ifdef CONFIG_DMA_ENGINE -- cgit v1.2.3 From 0403e3827788d878163f9ef0541b748b0f88ca5d Mon Sep 17 00:00:00 2001 From: Dan Williams Date: Tue, 8 Sep 2009 17:42:50 -0700 Subject: dmaengine: add fence support Some engines optimize operation by reading ahead in the descriptor chain such that descriptor2 may start execution before descriptor1 completes. If descriptor2 depends on the result from descriptor1 then a fence is required (on descriptor2) to disable this optimization. The async_tx api could implicitly identify dependencies via the 'depend_tx' parameter, but that would constrain cases where the dependency chain only specifies a completion order rather than a data dependency. So, provide an ASYNC_TX_FENCE to explicitly identify data dependencies. Signed-off-by: Dan Williams --- include/linux/dmaengine.h | 3 +++ 1 file changed, 3 insertions(+) (limited to 'include/linux/dmaengine.h') diff --git a/include/linux/dmaengine.h b/include/linux/dmaengine.h index 1012f1abcb54..4d6c1c925fd4 100644 --- a/include/linux/dmaengine.h +++ b/include/linux/dmaengine.h @@ -87,6 +87,8 @@ enum dma_transaction_type { * @DMA_PREP_CONTINUE - indicate to a driver that it is reusing buffers as * sources that were the result of a previous operation, in the case of a PQ * operation it continues the calculation with new sources + * @DMA_PREP_FENCE - tell the driver that subsequent operations depend + * on the result of this operation */ enum dma_ctrl_flags { DMA_PREP_INTERRUPT = (1 << 0), @@ -98,6 +100,7 @@ enum dma_ctrl_flags { DMA_PREP_PQ_DISABLE_P = (1 << 6), DMA_PREP_PQ_DISABLE_Q = (1 << 7), DMA_PREP_CONTINUE = (1 << 8), + DMA_PREP_FENCE = (1 << 9), }; /** -- cgit v1.2.3 From 138f4c359d23d2ec38d18bd70dd9613ae515fe93 Mon Sep 17 00:00:00 2001 From: Dan Williams Date: Tue, 8 Sep 2009 17:42:51 -0700 Subject: dmaengine, async_tx: add a "no channel switch" allocator Channel switching is problematic for some dmaengine drivers as the architecture precludes separating the ->prep from ->submit. In these cases the driver can select ASYNC_TX_DISABLE_CHANNEL_SWITCH to modify the async_tx allocator to only return channels that support all of the required asynchronous operations. For example MD_RAID456=y selects support for asynchronous xor, xor validate, pq, pq validate, and memcpy. When ASYNC_TX_DISABLE_CHANNEL_SWITCH=y any channel with all these capabilities is marked DMA_ASYNC_TX allowing async_tx_find_channel() to quickly locate compatible channels with the guarantee that dependency chains will remain on one channel. When ASYNC_TX_DISABLE_CHANNEL_SWITCH=n async_tx_find_channel() may select channels that lead to operation chains that need to cross channel boundaries using the async_tx channel switch capability. Signed-off-by: Dan Williams --- include/linux/dmaengine.h | 10 +++++++++- 1 file changed, 9 insertions(+), 1 deletion(-) (limited to 'include/linux/dmaengine.h') diff --git a/include/linux/dmaengine.h b/include/linux/dmaengine.h index 4d6c1c925fd4..86853ed7970b 100644 --- a/include/linux/dmaengine.h +++ b/include/linux/dmaengine.h @@ -48,6 +48,9 @@ enum dma_status { /** * enum dma_transaction_type - DMA transaction types/indexes + * + * Note: The DMA_ASYNC_TX capability is not to be set by drivers. It is + * automatically set as dma devices are registered. */ enum dma_transaction_type { DMA_MEMCPY, @@ -61,6 +64,7 @@ enum dma_transaction_type { DMA_MEMCPY_CRC32C, DMA_INTERRUPT, DMA_PRIVATE, + DMA_ASYNC_TX, DMA_SLAVE, }; @@ -396,7 +400,11 @@ static inline void net_dmaengine_put(void) #ifdef CONFIG_ASYNC_TX_DMA #define async_dmaengine_get() dmaengine_get() #define async_dmaengine_put() dmaengine_put() +#ifdef CONFIG_ASYNC_TX_DISABLE_CHANNEL_SWITCH +#define async_dma_find_channel(type) dma_find_channel(DMA_ASYNC_TX) +#else #define async_dma_find_channel(type) dma_find_channel(type) +#endif /* CONFIG_ASYNC_TX_DISABLE_CHANNEL_SWITCH */ #else static inline void async_dmaengine_get(void) { @@ -409,7 +417,7 @@ async_dma_find_channel(enum dma_transaction_type type) { return NULL; } -#endif +#endif /* CONFIG_ASYNC_TX_DMA */ dma_cookie_t dma_async_memcpy_buf_to_buf(struct dma_chan *chan, void *dest, void *src, size_t len); -- cgit v1.2.3 From 9308add6ea4fedeba37b0d7c4630a542bd34f214 Mon Sep 17 00:00:00 2001 From: Dan Williams Date: Tue, 8 Sep 2009 17:42:52 -0700 Subject: dmaengine: cleanup unused transaction types No drivers currently implement these operation types, so they can be deleted. Signed-off-by: Dan Williams --- include/linux/dmaengine.h | 3 --- 1 file changed, 3 deletions(-) (limited to 'include/linux/dmaengine.h') diff --git a/include/linux/dmaengine.h b/include/linux/dmaengine.h index 86853ed7970b..db23fd583f98 100644 --- a/include/linux/dmaengine.h +++ b/include/linux/dmaengine.h @@ -56,12 +56,9 @@ enum dma_transaction_type { DMA_MEMCPY, DMA_XOR, DMA_PQ, - DMA_DUAL_XOR, - DMA_PQ_UPDATE, DMA_XOR_VAL, DMA_PQ_VAL, DMA_MEMSET, - DMA_MEMCPY_CRC32C, DMA_INTERRUPT, DMA_PRIVATE, DMA_ASYNC_TX, -- cgit v1.2.3 From 83544ae9f3991bfc7d5e0fe9a3008cd05a8d57b7 Mon Sep 17 00:00:00 2001 From: Dan Williams Date: Tue, 8 Sep 2009 17:42:53 -0700 Subject: dmaengine, async_tx: support alignment checks Some engines have transfer size and address alignment restrictions. Add a per-operation alignment property to struct dma_device that the async routines and dmatest can use to check alignment capabilities. Signed-off-by: Dan Williams --- include/linux/dmaengine.h | 44 ++++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 44 insertions(+) (limited to 'include/linux/dmaengine.h') diff --git a/include/linux/dmaengine.h b/include/linux/dmaengine.h index db23fd583f98..835b9c7bf1c2 100644 --- a/include/linux/dmaengine.h +++ b/include/linux/dmaengine.h @@ -245,6 +245,10 @@ struct dma_async_tx_descriptor { * @cap_mask: one or more dma_capability flags * @max_xor: maximum number of xor sources, 0 if no capability * @max_pq: maximum number of PQ sources and PQ-continue capability + * @copy_align: alignment shift for memcpy operations + * @xor_align: alignment shift for xor operations + * @pq_align: alignment shift for pq operations + * @fill_align: alignment shift for memset operations * @dev_id: unique device ID * @dev: struct device reference for dma mapping api * @device_alloc_chan_resources: allocate resources and return the @@ -271,6 +275,10 @@ struct dma_device { dma_cap_mask_t cap_mask; unsigned short max_xor; unsigned short max_pq; + u8 copy_align; + u8 xor_align; + u8 pq_align; + u8 fill_align; #define DMA_HAS_PQ_CONTINUE (1 << 15) int dev_id; @@ -314,6 +322,42 @@ struct dma_device { void (*device_issue_pending)(struct dma_chan *chan); }; +static inline bool dmaengine_check_align(u8 align, size_t off1, size_t off2, size_t len) +{ + size_t mask; + + if (!align) + return true; + mask = (1 << align) - 1; + if (mask & (off1 | off2 | len)) + return false; + return true; +} + +static inline bool is_dma_copy_aligned(struct dma_device *dev, size_t off1, + size_t off2, size_t len) +{ + return dmaengine_check_align(dev->copy_align, off1, off2, len); +} + +static inline bool is_dma_xor_aligned(struct dma_device *dev, size_t off1, + size_t off2, size_t len) +{ + return dmaengine_check_align(dev->xor_align, off1, off2, len); +} + +static inline bool is_dma_pq_aligned(struct dma_device *dev, size_t off1, + size_t off2, size_t len) +{ + return dmaengine_check_align(dev->pq_align, off1, off2, len); +} + +static inline bool is_dma_fill_aligned(struct dma_device *dev, size_t off1, + size_t off2, size_t len) +{ + return dmaengine_check_align(dev->fill_align, off1, off2, len); +} + static inline void dma_set_maxpq(struct dma_device *dma, int maxpq, int has_pq_continue) { -- cgit v1.2.3 From 0803172778901e24a75ab074798d98c2b7411559 Mon Sep 17 00:00:00 2001 From: Dan Williams Date: Tue, 8 Sep 2009 17:53:04 -0700 Subject: dmaengine: kill tx_list The tx_list attribute of struct dma_async_tx_descriptor is common to most, but not all dma driver implementations. None of the upper level code (dmaengine/async_tx) uses it, so allow drivers to implement it locally if they need it. This saves sizeof(struct list_head) bytes for drivers that do not manage descriptors with a linked list (e.g.: ioatdma v2,3). Signed-off-by: Dan Williams --- include/linux/dmaengine.h | 3 --- 1 file changed, 3 deletions(-) (limited to 'include/linux/dmaengine.h') diff --git a/include/linux/dmaengine.h b/include/linux/dmaengine.h index ffefba81c818..f114bc7790bc 100644 --- a/include/linux/dmaengine.h +++ b/include/linux/dmaengine.h @@ -180,8 +180,6 @@ typedef void (*dma_async_tx_callback)(void *dma_async_param); * @flags: flags to augment operation preparation, control completion, and * communicate status * @phys: physical address of the descriptor - * @tx_list: driver common field for operations that require multiple - * descriptors * @chan: target channel for this operation * @tx_submit: set the prepared descriptor(s) to be executed by the engine * @callback: routine to call after this operation is complete @@ -195,7 +193,6 @@ struct dma_async_tx_descriptor { dma_cookie_t cookie; enum dma_ctrl_flags flags; /* not a 'long' to pack with cookie */ dma_addr_t phys; - struct list_head tx_list; struct dma_chan *chan; dma_cookie_t (*tx_submit)(struct dma_async_tx_descriptor *tx); dma_async_tx_callback callback; -- cgit v1.2.3