diff options
Diffstat (limited to 'crypto/async_tx')
-rw-r--r-- | crypto/async_tx/async_memcpy.c | 7 | ||||
-rw-r--r-- | crypto/async_tx/async_memset.c | 7 | ||||
-rw-r--r-- | crypto/async_tx/async_pq.c | 5 | ||||
-rw-r--r-- | crypto/async_tx/async_raid6_recov.c | 47 | ||||
-rw-r--r-- | crypto/async_tx/async_xor.c | 11 |
5 files changed, 50 insertions, 27 deletions
diff --git a/crypto/async_tx/async_memcpy.c b/crypto/async_tx/async_memcpy.c index 98e15bd0dcb5..b38cbb3fd527 100644 --- a/crypto/async_tx/async_memcpy.c +++ b/crypto/async_tx/async_memcpy.c | |||
@@ -52,9 +52,12 @@ async_memcpy(struct page *dest, struct page *src, unsigned int dest_offset, | |||
52 | 52 | ||
53 | if (device) { | 53 | if (device) { |
54 | dma_addr_t dma_dest, dma_src; | 54 | dma_addr_t dma_dest, dma_src; |
55 | unsigned long dma_prep_flags; | 55 | unsigned long dma_prep_flags = 0; |
56 | 56 | ||
57 | dma_prep_flags = submit->cb_fn ? DMA_PREP_INTERRUPT : 0; | 57 | if (submit->cb_fn) |
58 | dma_prep_flags |= DMA_PREP_INTERRUPT; | ||
59 | if (submit->flags & ASYNC_TX_FENCE) | ||
60 | dma_prep_flags |= DMA_PREP_FENCE; | ||
58 | dma_dest = dma_map_page(device->dev, dest, dest_offset, len, | 61 | dma_dest = dma_map_page(device->dev, dest, dest_offset, len, |
59 | DMA_FROM_DEVICE); | 62 | DMA_FROM_DEVICE); |
60 | 63 | ||
diff --git a/crypto/async_tx/async_memset.c b/crypto/async_tx/async_memset.c index b896a6e5f673..a374784e3329 100644 --- a/crypto/async_tx/async_memset.c +++ b/crypto/async_tx/async_memset.c | |||
@@ -49,9 +49,12 @@ async_memset(struct page *dest, int val, unsigned int offset, size_t len, | |||
49 | 49 | ||
50 | if (device) { | 50 | if (device) { |
51 | dma_addr_t dma_dest; | 51 | dma_addr_t dma_dest; |
52 | unsigned long dma_prep_flags; | 52 | unsigned long dma_prep_flags = 0; |
53 | 53 | ||
54 | dma_prep_flags = submit->cb_fn ? DMA_PREP_INTERRUPT : 0; | 54 | if (submit->cb_fn) |
55 | dma_prep_flags |= DMA_PREP_INTERRUPT; | ||
56 | if (submit->flags & ASYNC_TX_FENCE) | ||
57 | dma_prep_flags |= DMA_PREP_FENCE; | ||
55 | dma_dest = dma_map_page(device->dev, dest, offset, len, | 58 | dma_dest = dma_map_page(device->dev, dest, offset, len, |
56 | DMA_FROM_DEVICE); | 59 | DMA_FROM_DEVICE); |
57 | 60 | ||
diff --git a/crypto/async_tx/async_pq.c b/crypto/async_tx/async_pq.c index 108b21efb499..a25e290c39fb 100644 --- a/crypto/async_tx/async_pq.c +++ b/crypto/async_tx/async_pq.c | |||
@@ -101,6 +101,7 @@ do_async_gen_syndrome(struct dma_chan *chan, struct page **blocks, | |||
101 | */ | 101 | */ |
102 | if (src_cnt > pq_src_cnt) { | 102 | if (src_cnt > pq_src_cnt) { |
103 | submit->flags &= ~ASYNC_TX_ACK; | 103 | submit->flags &= ~ASYNC_TX_ACK; |
104 | submit->flags |= ASYNC_TX_FENCE; | ||
104 | dma_flags |= DMA_COMPL_SKIP_DEST_UNMAP; | 105 | dma_flags |= DMA_COMPL_SKIP_DEST_UNMAP; |
105 | submit->cb_fn = NULL; | 106 | submit->cb_fn = NULL; |
106 | submit->cb_param = NULL; | 107 | submit->cb_param = NULL; |
@@ -111,6 +112,8 @@ do_async_gen_syndrome(struct dma_chan *chan, struct page **blocks, | |||
111 | if (cb_fn_orig) | 112 | if (cb_fn_orig) |
112 | dma_flags |= DMA_PREP_INTERRUPT; | 113 | dma_flags |= DMA_PREP_INTERRUPT; |
113 | } | 114 | } |
115 | if (submit->flags & ASYNC_TX_FENCE) | ||
116 | dma_flags |= DMA_PREP_FENCE; | ||
114 | 117 | ||
115 | /* Since we have clobbered the src_list we are committed | 118 | /* Since we have clobbered the src_list we are committed |
116 | * to doing this asynchronously. Drivers force forward | 119 | * to doing this asynchronously. Drivers force forward |
@@ -282,6 +285,8 @@ async_syndrome_val(struct page **blocks, unsigned int offset, int disks, | |||
282 | dma_flags |= DMA_PREP_PQ_DISABLE_P; | 285 | dma_flags |= DMA_PREP_PQ_DISABLE_P; |
283 | if (!Q(blocks, disks)) | 286 | if (!Q(blocks, disks)) |
284 | dma_flags |= DMA_PREP_PQ_DISABLE_Q; | 287 | dma_flags |= DMA_PREP_PQ_DISABLE_Q; |
288 | if (submit->flags & ASYNC_TX_FENCE) | ||
289 | dma_flags |= DMA_PREP_FENCE; | ||
285 | for (i = 0; i < disks; i++) | 290 | for (i = 0; i < disks; i++) |
286 | if (likely(blocks[i])) { | 291 | if (likely(blocks[i])) { |
287 | BUG_ON(is_raid6_zero_block(blocks[i])); | 292 | BUG_ON(is_raid6_zero_block(blocks[i])); |
diff --git a/crypto/async_tx/async_raid6_recov.c b/crypto/async_tx/async_raid6_recov.c index 0c14d48c9896..822a42d10061 100644 --- a/crypto/async_tx/async_raid6_recov.c +++ b/crypto/async_tx/async_raid6_recov.c | |||
@@ -44,6 +44,8 @@ async_sum_product(struct page *dest, struct page **srcs, unsigned char *coef, | |||
44 | struct dma_async_tx_descriptor *tx; | 44 | struct dma_async_tx_descriptor *tx; |
45 | enum dma_ctrl_flags dma_flags = DMA_PREP_PQ_DISABLE_P; | 45 | enum dma_ctrl_flags dma_flags = DMA_PREP_PQ_DISABLE_P; |
46 | 46 | ||
47 | if (submit->flags & ASYNC_TX_FENCE) | ||
48 | dma_flags |= DMA_PREP_FENCE; | ||
47 | dma_dest[1] = dma_map_page(dev, dest, 0, len, DMA_BIDIRECTIONAL); | 49 | dma_dest[1] = dma_map_page(dev, dest, 0, len, DMA_BIDIRECTIONAL); |
48 | dma_src[0] = dma_map_page(dev, srcs[0], 0, len, DMA_TO_DEVICE); | 50 | dma_src[0] = dma_map_page(dev, srcs[0], 0, len, DMA_TO_DEVICE); |
49 | dma_src[1] = dma_map_page(dev, srcs[1], 0, len, DMA_TO_DEVICE); | 51 | dma_src[1] = dma_map_page(dev, srcs[1], 0, len, DMA_TO_DEVICE); |
@@ -89,6 +91,8 @@ async_mult(struct page *dest, struct page *src, u8 coef, size_t len, | |||
89 | struct dma_async_tx_descriptor *tx; | 91 | struct dma_async_tx_descriptor *tx; |
90 | enum dma_ctrl_flags dma_flags = DMA_PREP_PQ_DISABLE_P; | 92 | enum dma_ctrl_flags dma_flags = DMA_PREP_PQ_DISABLE_P; |
91 | 93 | ||
94 | if (submit->flags & ASYNC_TX_FENCE) | ||
95 | dma_flags |= DMA_PREP_FENCE; | ||
92 | dma_dest[1] = dma_map_page(dev, dest, 0, len, DMA_BIDIRECTIONAL); | 96 | dma_dest[1] = dma_map_page(dev, dest, 0, len, DMA_BIDIRECTIONAL); |
93 | dma_src[0] = dma_map_page(dev, src, 0, len, DMA_TO_DEVICE); | 97 | dma_src[0] = dma_map_page(dev, src, 0, len, DMA_TO_DEVICE); |
94 | tx = dma->device_prep_dma_pq(chan, dma_dest, dma_src, 1, &coef, | 98 | tx = dma->device_prep_dma_pq(chan, dma_dest, dma_src, 1, &coef, |
@@ -138,7 +142,7 @@ __2data_recov_4(size_t bytes, int faila, int failb, struct page **blocks, | |||
138 | srcs[1] = q; | 142 | srcs[1] = q; |
139 | coef[0] = raid6_gfexi[failb-faila]; | 143 | coef[0] = raid6_gfexi[failb-faila]; |
140 | coef[1] = raid6_gfinv[raid6_gfexp[faila]^raid6_gfexp[failb]]; | 144 | coef[1] = raid6_gfinv[raid6_gfexp[faila]^raid6_gfexp[failb]]; |
141 | init_async_submit(submit, 0, tx, NULL, NULL, scribble); | 145 | init_async_submit(submit, ASYNC_TX_FENCE, tx, NULL, NULL, scribble); |
142 | tx = async_sum_product(b, srcs, coef, bytes, submit); | 146 | tx = async_sum_product(b, srcs, coef, bytes, submit); |
143 | 147 | ||
144 | /* Dy = P+Pxy+Dx */ | 148 | /* Dy = P+Pxy+Dx */ |
@@ -188,23 +192,23 @@ __2data_recov_5(size_t bytes, int faila, int failb, struct page **blocks, | |||
188 | dp = blocks[faila]; | 192 | dp = blocks[faila]; |
189 | dq = blocks[failb]; | 193 | dq = blocks[failb]; |
190 | 194 | ||
191 | init_async_submit(submit, 0, tx, NULL, NULL, scribble); | 195 | init_async_submit(submit, ASYNC_TX_FENCE, tx, NULL, NULL, scribble); |
192 | tx = async_memcpy(dp, g, 0, 0, bytes, submit); | 196 | tx = async_memcpy(dp, g, 0, 0, bytes, submit); |
193 | init_async_submit(submit, 0, tx, NULL, NULL, scribble); | 197 | init_async_submit(submit, ASYNC_TX_FENCE, tx, NULL, NULL, scribble); |
194 | tx = async_mult(dq, g, raid6_gfexp[good], bytes, submit); | 198 | tx = async_mult(dq, g, raid6_gfexp[good], bytes, submit); |
195 | 199 | ||
196 | /* compute P + Pxy */ | 200 | /* compute P + Pxy */ |
197 | srcs[0] = dp; | 201 | srcs[0] = dp; |
198 | srcs[1] = p; | 202 | srcs[1] = p; |
199 | init_async_submit(submit, ASYNC_TX_XOR_DROP_DST, tx, NULL, NULL, | 203 | init_async_submit(submit, ASYNC_TX_FENCE|ASYNC_TX_XOR_DROP_DST, tx, |
200 | scribble); | 204 | NULL, NULL, scribble); |
201 | tx = async_xor(dp, srcs, 0, 2, bytes, submit); | 205 | tx = async_xor(dp, srcs, 0, 2, bytes, submit); |
202 | 206 | ||
203 | /* compute Q + Qxy */ | 207 | /* compute Q + Qxy */ |
204 | srcs[0] = dq; | 208 | srcs[0] = dq; |
205 | srcs[1] = q; | 209 | srcs[1] = q; |
206 | init_async_submit(submit, ASYNC_TX_XOR_DROP_DST, tx, NULL, NULL, | 210 | init_async_submit(submit, ASYNC_TX_FENCE|ASYNC_TX_XOR_DROP_DST, tx, |
207 | scribble); | 211 | NULL, NULL, scribble); |
208 | tx = async_xor(dq, srcs, 0, 2, bytes, submit); | 212 | tx = async_xor(dq, srcs, 0, 2, bytes, submit); |
209 | 213 | ||
210 | /* Dx = A*(P+Pxy) + B*(Q+Qxy) */ | 214 | /* Dx = A*(P+Pxy) + B*(Q+Qxy) */ |
@@ -212,7 +216,7 @@ __2data_recov_5(size_t bytes, int faila, int failb, struct page **blocks, | |||
212 | srcs[1] = dq; | 216 | srcs[1] = dq; |
213 | coef[0] = raid6_gfexi[failb-faila]; | 217 | coef[0] = raid6_gfexi[failb-faila]; |
214 | coef[1] = raid6_gfinv[raid6_gfexp[faila]^raid6_gfexp[failb]]; | 218 | coef[1] = raid6_gfinv[raid6_gfexp[faila]^raid6_gfexp[failb]]; |
215 | init_async_submit(submit, 0, tx, NULL, NULL, scribble); | 219 | init_async_submit(submit, ASYNC_TX_FENCE, tx, NULL, NULL, scribble); |
216 | tx = async_sum_product(dq, srcs, coef, bytes, submit); | 220 | tx = async_sum_product(dq, srcs, coef, bytes, submit); |
217 | 221 | ||
218 | /* Dy = P+Pxy+Dx */ | 222 | /* Dy = P+Pxy+Dx */ |
@@ -252,7 +256,7 @@ __2data_recov_n(int disks, size_t bytes, int faila, int failb, | |||
252 | blocks[failb] = (void *)raid6_empty_zero_page; | 256 | blocks[failb] = (void *)raid6_empty_zero_page; |
253 | blocks[disks-1] = dq; | 257 | blocks[disks-1] = dq; |
254 | 258 | ||
255 | init_async_submit(submit, 0, tx, NULL, NULL, scribble); | 259 | init_async_submit(submit, ASYNC_TX_FENCE, tx, NULL, NULL, scribble); |
256 | tx = async_gen_syndrome(blocks, 0, disks, bytes, submit); | 260 | tx = async_gen_syndrome(blocks, 0, disks, bytes, submit); |
257 | 261 | ||
258 | /* Restore pointer table */ | 262 | /* Restore pointer table */ |
@@ -264,15 +268,15 @@ __2data_recov_n(int disks, size_t bytes, int faila, int failb, | |||
264 | /* compute P + Pxy */ | 268 | /* compute P + Pxy */ |
265 | srcs[0] = dp; | 269 | srcs[0] = dp; |
266 | srcs[1] = p; | 270 | srcs[1] = p; |
267 | init_async_submit(submit, ASYNC_TX_XOR_DROP_DST, tx, NULL, NULL, | 271 | init_async_submit(submit, ASYNC_TX_FENCE|ASYNC_TX_XOR_DROP_DST, tx, |
268 | scribble); | 272 | NULL, NULL, scribble); |
269 | tx = async_xor(dp, srcs, 0, 2, bytes, submit); | 273 | tx = async_xor(dp, srcs, 0, 2, bytes, submit); |
270 | 274 | ||
271 | /* compute Q + Qxy */ | 275 | /* compute Q + Qxy */ |
272 | srcs[0] = dq; | 276 | srcs[0] = dq; |
273 | srcs[1] = q; | 277 | srcs[1] = q; |
274 | init_async_submit(submit, ASYNC_TX_XOR_DROP_DST, tx, NULL, NULL, | 278 | init_async_submit(submit, ASYNC_TX_FENCE|ASYNC_TX_XOR_DROP_DST, tx, |
275 | scribble); | 279 | NULL, NULL, scribble); |
276 | tx = async_xor(dq, srcs, 0, 2, bytes, submit); | 280 | tx = async_xor(dq, srcs, 0, 2, bytes, submit); |
277 | 281 | ||
278 | /* Dx = A*(P+Pxy) + B*(Q+Qxy) */ | 282 | /* Dx = A*(P+Pxy) + B*(Q+Qxy) */ |
@@ -280,7 +284,7 @@ __2data_recov_n(int disks, size_t bytes, int faila, int failb, | |||
280 | srcs[1] = dq; | 284 | srcs[1] = dq; |
281 | coef[0] = raid6_gfexi[failb-faila]; | 285 | coef[0] = raid6_gfexi[failb-faila]; |
282 | coef[1] = raid6_gfinv[raid6_gfexp[faila]^raid6_gfexp[failb]]; | 286 | coef[1] = raid6_gfinv[raid6_gfexp[faila]^raid6_gfexp[failb]]; |
283 | init_async_submit(submit, 0, tx, NULL, NULL, scribble); | 287 | init_async_submit(submit, ASYNC_TX_FENCE, tx, NULL, NULL, scribble); |
284 | tx = async_sum_product(dq, srcs, coef, bytes, submit); | 288 | tx = async_sum_product(dq, srcs, coef, bytes, submit); |
285 | 289 | ||
286 | /* Dy = P+Pxy+Dx */ | 290 | /* Dy = P+Pxy+Dx */ |
@@ -407,13 +411,16 @@ async_raid6_datap_recov(int disks, size_t bytes, int faila, | |||
407 | int good = faila == 0 ? 1 : 0; | 411 | int good = faila == 0 ? 1 : 0; |
408 | struct page *g = blocks[good]; | 412 | struct page *g = blocks[good]; |
409 | 413 | ||
410 | init_async_submit(submit, 0, tx, NULL, NULL, scribble); | 414 | init_async_submit(submit, ASYNC_TX_FENCE, tx, NULL, NULL, |
415 | scribble); | ||
411 | tx = async_memcpy(p, g, 0, 0, bytes, submit); | 416 | tx = async_memcpy(p, g, 0, 0, bytes, submit); |
412 | 417 | ||
413 | init_async_submit(submit, 0, tx, NULL, NULL, scribble); | 418 | init_async_submit(submit, ASYNC_TX_FENCE, tx, NULL, NULL, |
419 | scribble); | ||
414 | tx = async_mult(dq, g, raid6_gfexp[good], bytes, submit); | 420 | tx = async_mult(dq, g, raid6_gfexp[good], bytes, submit); |
415 | } else { | 421 | } else { |
416 | init_async_submit(submit, 0, tx, NULL, NULL, scribble); | 422 | init_async_submit(submit, ASYNC_TX_FENCE, tx, NULL, NULL, |
423 | scribble); | ||
417 | tx = async_gen_syndrome(blocks, 0, disks, bytes, submit); | 424 | tx = async_gen_syndrome(blocks, 0, disks, bytes, submit); |
418 | } | 425 | } |
419 | 426 | ||
@@ -426,11 +433,11 @@ async_raid6_datap_recov(int disks, size_t bytes, int faila, | |||
426 | 433 | ||
427 | srcs[0] = dq; | 434 | srcs[0] = dq; |
428 | srcs[1] = q; | 435 | srcs[1] = q; |
429 | init_async_submit(submit, ASYNC_TX_XOR_DROP_DST, tx, NULL, NULL, | 436 | init_async_submit(submit, ASYNC_TX_FENCE|ASYNC_TX_XOR_DROP_DST, tx, |
430 | scribble); | 437 | NULL, NULL, scribble); |
431 | tx = async_xor(dq, srcs, 0, 2, bytes, submit); | 438 | tx = async_xor(dq, srcs, 0, 2, bytes, submit); |
432 | 439 | ||
433 | init_async_submit(submit, 0, tx, NULL, NULL, scribble); | 440 | init_async_submit(submit, ASYNC_TX_FENCE, tx, NULL, NULL, scribble); |
434 | tx = async_mult(dq, dq, coef, bytes, submit); | 441 | tx = async_mult(dq, dq, coef, bytes, submit); |
435 | 442 | ||
436 | srcs[0] = p; | 443 | srcs[0] = p; |
diff --git a/crypto/async_tx/async_xor.c b/crypto/async_tx/async_xor.c index 56b5f98da463..db279872ef3d 100644 --- a/crypto/async_tx/async_xor.c +++ b/crypto/async_tx/async_xor.c | |||
@@ -69,6 +69,7 @@ do_async_xor(struct dma_chan *chan, struct page *dest, struct page **src_list, | |||
69 | */ | 69 | */ |
70 | if (src_cnt > xor_src_cnt) { | 70 | if (src_cnt > xor_src_cnt) { |
71 | submit->flags &= ~ASYNC_TX_ACK; | 71 | submit->flags &= ~ASYNC_TX_ACK; |
72 | submit->flags |= ASYNC_TX_FENCE; | ||
72 | dma_flags = DMA_COMPL_SKIP_DEST_UNMAP; | 73 | dma_flags = DMA_COMPL_SKIP_DEST_UNMAP; |
73 | submit->cb_fn = NULL; | 74 | submit->cb_fn = NULL; |
74 | submit->cb_param = NULL; | 75 | submit->cb_param = NULL; |
@@ -78,7 +79,8 @@ do_async_xor(struct dma_chan *chan, struct page *dest, struct page **src_list, | |||
78 | } | 79 | } |
79 | if (submit->cb_fn) | 80 | if (submit->cb_fn) |
80 | dma_flags |= DMA_PREP_INTERRUPT; | 81 | dma_flags |= DMA_PREP_INTERRUPT; |
81 | 82 | if (submit->flags & ASYNC_TX_FENCE) | |
83 | dma_flags |= DMA_PREP_FENCE; | ||
82 | /* Since we have clobbered the src_list we are committed | 84 | /* Since we have clobbered the src_list we are committed |
83 | * to doing this asynchronously. Drivers force forward progress | 85 | * to doing this asynchronously. Drivers force forward progress |
84 | * in case they can not provide a descriptor | 86 | * in case they can not provide a descriptor |
@@ -264,12 +266,15 @@ async_xor_val(struct page *dest, struct page **src_list, unsigned int offset, | |||
264 | dma_src = (dma_addr_t *) src_list; | 266 | dma_src = (dma_addr_t *) src_list; |
265 | 267 | ||
266 | if (dma_src && device && src_cnt <= device->max_xor) { | 268 | if (dma_src && device && src_cnt <= device->max_xor) { |
267 | unsigned long dma_prep_flags; | 269 | unsigned long dma_prep_flags = 0; |
268 | int i; | 270 | int i; |
269 | 271 | ||
270 | pr_debug("%s: (async) len: %zu\n", __func__, len); | 272 | pr_debug("%s: (async) len: %zu\n", __func__, len); |
271 | 273 | ||
272 | dma_prep_flags = submit->cb_fn ? DMA_PREP_INTERRUPT : 0; | 274 | if (submit->cb_fn) |
275 | dma_prep_flags |= DMA_PREP_INTERRUPT; | ||
276 | if (submit->flags & ASYNC_TX_FENCE) | ||
277 | dma_prep_flags |= DMA_PREP_FENCE; | ||
273 | for (i = 0; i < src_cnt; i++) | 278 | for (i = 0; i < src_cnt; i++) |
274 | dma_src[i] = dma_map_page(device->dev, src_list[i], | 279 | dma_src[i] = dma_map_page(device->dev, src_list[i], |
275 | offset, len, DMA_TO_DEVICE); | 280 | offset, len, DMA_TO_DEVICE); |