diff options
author | Linus Torvalds <torvalds@linux-foundation.org> | 2015-06-29 19:44:45 +0300 |
---|---|---|
committer | Linus Torvalds <torvalds@linux-foundation.org> | 2015-06-29 19:44:45 +0300 |
commit | 1bc5e157ed2b4f5b206155fc772d860158acd201 (patch) | |
tree | 0edf76217120a05502437e775bf99cf66ae0d13f /drivers/dma/at_xdmac.c | |
parent | f199b663fc5a276af10a49c24a83d8de27a41b53 (diff) | |
parent | 657d61275dad112ff36e70ee63c71d64821db36c (diff) |
Merge tag 'dmaengine-4.2-rc1' of git://git.infradead.org/users/vkoul/slave-dma
Pull dmaengine updates from Vinod Koul:
"This time we have support for few new devices, few new features and
odd fixes spread thru the subsystem.
New devices added:
- support for CSRatlas7 dma controller
- Allwinner H3(sun8i) controller
- TI DMA crossbar driver on DRA7x
- new pxa driver
New features added:
- memset support is bought back now that we have a user in xdmac controller
- interleaved transfers support different source and destination strides
- supporting DMA routers and configuration thru DT
- support for reusing descriptors
- xdmac memset and interleaved transfer support
- hdmac support for interleaved transfers
- omap-dma support for memcpy
Others:
- Constify platform_device_id
- mv_xor fixes and improvements"
* tag 'dmaengine-4.2-rc1' of git://git.infradead.org/users/vkoul/slave-dma: (46 commits)
dmaengine: xgene: fix file permission
dmaengine: fsl-edma: clear pending interrupts on initialization
dmaengine: xdmac: Add memset support
Documentation: dmaengine: document DMA_CTRL_ACK
dmaengine: virt-dma: don't always free descriptor upon completion
dmaengine: Revert "drivers/dma: remove unused support for MEMSET operations"
dmaengine: hdmac: Implement interleaved transfers
dmaengine: Move icg helpers to global header
dmaengine: mv_xor: improve descriptors list handling and reduce locking
dmaengine: mv_xor: Enlarge descriptor pool size
dmaengine: mv_xor: add support for a38x command in descriptor mode
dmaengine: mv_xor: Rename function for consistent naming
dmaengine: mv_xor: bug fix for racing condition in descriptors cleanup
dmaengine: pl330: fix wording in mcbufsz message
dmaengine: sirf: add CSRatlas7 SoC support
dmaengine: xgene-dma: Fix "incorrect type in assignement" warnings
dmaengine: fix kernel-doc documentation
dmaengine: pxa_dma: add support for legacy transition
dmaengine: pxa_dma: add debug information
dmaengine: pxa: add pxa dmaengine driver
...
Diffstat (limited to 'drivers/dma/at_xdmac.c')
-rw-r--r-- | drivers/dma/at_xdmac.c | 394 |
1 files changed, 342 insertions, 52 deletions
diff --git a/drivers/dma/at_xdmac.c b/drivers/dma/at_xdmac.c index 7992164ea9ec..cf1213de7865 100644 --- a/drivers/dma/at_xdmac.c +++ b/drivers/dma/at_xdmac.c @@ -235,6 +235,10 @@ struct at_xdmac_lld { dma_addr_t mbr_sa; /* Source Address Member */ dma_addr_t mbr_da; /* Destination Address Member */ u32 mbr_cfg; /* Configuration Register */ + u32 mbr_bc; /* Block Control Register */ + u32 mbr_ds; /* Data Stride Register */ + u32 mbr_sus; /* Source Microblock Stride Register */ + u32 mbr_dus; /* Destination Microblock Stride Register */ }; @@ -358,6 +362,8 @@ static void at_xdmac_start_xfer(struct at_xdmac_chan *atchan, if (at_xdmac_chan_is_cyclic(atchan)) { reg = AT_XDMAC_CNDC_NDVIEW_NDV1; at_xdmac_chan_write(atchan, AT_XDMAC_CC, first->lld.mbr_cfg); + } else if (first->lld.mbr_ubc & AT_XDMAC_MBR_UBC_NDV3) { + reg = AT_XDMAC_CNDC_NDVIEW_NDV3; } else { /* * No need to write AT_XDMAC_CC reg, it will be done when the @@ -465,6 +471,33 @@ static struct at_xdmac_desc *at_xdmac_get_desc(struct at_xdmac_chan *atchan) return desc; } +static void at_xdmac_queue_desc(struct dma_chan *chan, + struct at_xdmac_desc *prev, + struct at_xdmac_desc *desc) +{ + if (!prev || !desc) + return; + + prev->lld.mbr_nda = desc->tx_dma_desc.phys; + prev->lld.mbr_ubc |= AT_XDMAC_MBR_UBC_NDE; + + dev_dbg(chan2dev(chan), "%s: chain lld: prev=0x%p, mbr_nda=%pad\n", + __func__, prev, &prev->lld.mbr_nda); +} + +static inline void at_xdmac_increment_block_count(struct dma_chan *chan, + struct at_xdmac_desc *desc) +{ + if (!desc) + return; + + desc->lld.mbr_bc++; + + dev_dbg(chan2dev(chan), + "%s: incrementing the block count of the desc 0x%p\n", + __func__, desc); +} + static struct dma_chan *at_xdmac_xlate(struct of_phandle_args *dma_spec, struct of_dma *of_dma) { @@ -656,19 +689,14 @@ at_xdmac_prep_slave_sg(struct dma_chan *chan, struct scatterlist *sgl, desc->lld.mbr_ubc = AT_XDMAC_MBR_UBC_NDV2 /* next descriptor view */ | AT_XDMAC_MBR_UBC_NDEN /* next descriptor dst parameter update */ | AT_XDMAC_MBR_UBC_NSEN /* next descriptor src parameter update */ - | (i == sg_len - 1 ? 0 : AT_XDMAC_MBR_UBC_NDE) /* descriptor fetch */ | (len >> fixed_dwidth); /* microblock length */ dev_dbg(chan2dev(chan), "%s: lld: mbr_sa=%pad, mbr_da=%pad, mbr_ubc=0x%08x\n", __func__, &desc->lld.mbr_sa, &desc->lld.mbr_da, desc->lld.mbr_ubc); /* Chain lld. */ - if (prev) { - prev->lld.mbr_nda = desc->tx_dma_desc.phys; - dev_dbg(chan2dev(chan), - "%s: chain lld: prev=0x%p, mbr_nda=%pad\n", - __func__, prev, &prev->lld.mbr_nda); - } + if (prev) + at_xdmac_queue_desc(chan, prev, desc); prev = desc; if (!first) @@ -748,7 +776,6 @@ at_xdmac_prep_dma_cyclic(struct dma_chan *chan, dma_addr_t buf_addr, desc->lld.mbr_ubc = AT_XDMAC_MBR_UBC_NDV1 | AT_XDMAC_MBR_UBC_NDEN | AT_XDMAC_MBR_UBC_NSEN - | AT_XDMAC_MBR_UBC_NDE | period_len >> at_xdmac_get_dwidth(desc->lld.mbr_cfg); dev_dbg(chan2dev(chan), @@ -756,12 +783,8 @@ at_xdmac_prep_dma_cyclic(struct dma_chan *chan, dma_addr_t buf_addr, __func__, &desc->lld.mbr_sa, &desc->lld.mbr_da, desc->lld.mbr_ubc); /* Chain lld. */ - if (prev) { - prev->lld.mbr_nda = desc->tx_dma_desc.phys; - dev_dbg(chan2dev(chan), - "%s: chain lld: prev=0x%p, mbr_nda=%pad\n", - __func__, prev, &prev->lld.mbr_nda); - } + if (prev) + at_xdmac_queue_desc(chan, prev, desc); prev = desc; if (!first) @@ -783,6 +806,215 @@ at_xdmac_prep_dma_cyclic(struct dma_chan *chan, dma_addr_t buf_addr, return &first->tx_dma_desc; } +static inline u32 at_xdmac_align_width(struct dma_chan *chan, dma_addr_t addr) +{ + u32 width; + + /* + * Check address alignment to select the greater data width we + * can use. + * + * Some XDMAC implementations don't provide dword transfer, in + * this case selecting dword has the same behavior as + * selecting word transfers. + */ + if (!(addr & 7)) { + width = AT_XDMAC_CC_DWIDTH_DWORD; + dev_dbg(chan2dev(chan), "%s: dwidth: double word\n", __func__); + } else if (!(addr & 3)) { + width = AT_XDMAC_CC_DWIDTH_WORD; + dev_dbg(chan2dev(chan), "%s: dwidth: word\n", __func__); + } else if (!(addr & 1)) { + width = AT_XDMAC_CC_DWIDTH_HALFWORD; + dev_dbg(chan2dev(chan), "%s: dwidth: half word\n", __func__); + } else { + width = AT_XDMAC_CC_DWIDTH_BYTE; + dev_dbg(chan2dev(chan), "%s: dwidth: byte\n", __func__); + } + + return width; +} + +static struct at_xdmac_desc * +at_xdmac_interleaved_queue_desc(struct dma_chan *chan, + struct at_xdmac_chan *atchan, + struct at_xdmac_desc *prev, + dma_addr_t src, dma_addr_t dst, + struct dma_interleaved_template *xt, + struct data_chunk *chunk) +{ + struct at_xdmac_desc *desc; + u32 dwidth; + unsigned long flags; + size_t ublen; + /* + * WARNING: The channel configuration is set here since there is no + * dmaengine_slave_config call in this case. Moreover we don't know the + * direction, it involves we can't dynamically set the source and dest + * interface so we have to use the same one. Only interface 0 allows EBI + * access. Hopefully we can access DDR through both ports (at least on + * SAMA5D4x), so we can use the same interface for source and dest, + * that solves the fact we don't know the direction. + */ + u32 chan_cc = AT_XDMAC_CC_DIF(0) + | AT_XDMAC_CC_SIF(0) + | AT_XDMAC_CC_MBSIZE_SIXTEEN + | AT_XDMAC_CC_TYPE_MEM_TRAN; + + dwidth = at_xdmac_align_width(chan, src | dst | chunk->size); + if (chunk->size >= (AT_XDMAC_MBR_UBC_UBLEN_MAX << dwidth)) { + dev_dbg(chan2dev(chan), + "%s: chunk too big (%d, max size %lu)...\n", + __func__, chunk->size, + AT_XDMAC_MBR_UBC_UBLEN_MAX << dwidth); + return NULL; + } + + if (prev) + dev_dbg(chan2dev(chan), + "Adding items at the end of desc 0x%p\n", prev); + + if (xt->src_inc) { + if (xt->src_sgl) + chan_cc |= AT_XDMAC_CC_SAM_UBS_DS_AM; + else + chan_cc |= AT_XDMAC_CC_SAM_INCREMENTED_AM; + } + + if (xt->dst_inc) { + if (xt->dst_sgl) + chan_cc |= AT_XDMAC_CC_DAM_UBS_DS_AM; + else + chan_cc |= AT_XDMAC_CC_DAM_INCREMENTED_AM; + } + + spin_lock_irqsave(&atchan->lock, flags); + desc = at_xdmac_get_desc(atchan); + spin_unlock_irqrestore(&atchan->lock, flags); + if (!desc) { + dev_err(chan2dev(chan), "can't get descriptor\n"); + return NULL; + } + + chan_cc |= AT_XDMAC_CC_DWIDTH(dwidth); + + ublen = chunk->size >> dwidth; + + desc->lld.mbr_sa = src; + desc->lld.mbr_da = dst; + desc->lld.mbr_sus = dmaengine_get_src_icg(xt, chunk); + desc->lld.mbr_dus = dmaengine_get_dst_icg(xt, chunk); + + desc->lld.mbr_ubc = AT_XDMAC_MBR_UBC_NDV3 + | AT_XDMAC_MBR_UBC_NDEN + | AT_XDMAC_MBR_UBC_NSEN + | ublen; + desc->lld.mbr_cfg = chan_cc; + + dev_dbg(chan2dev(chan), + "%s: lld: mbr_sa=0x%08x, mbr_da=0x%08x, mbr_ubc=0x%08x, mbr_cfg=0x%08x\n", + __func__, desc->lld.mbr_sa, desc->lld.mbr_da, + desc->lld.mbr_ubc, desc->lld.mbr_cfg); + + /* Chain lld. */ + if (prev) + at_xdmac_queue_desc(chan, prev, desc); + + return desc; +} + +static struct dma_async_tx_descriptor * +at_xdmac_prep_interleaved(struct dma_chan *chan, + struct dma_interleaved_template *xt, + unsigned long flags) +{ + struct at_xdmac_chan *atchan = to_at_xdmac_chan(chan); + struct at_xdmac_desc *prev = NULL, *first = NULL; + struct data_chunk *chunk, *prev_chunk = NULL; + dma_addr_t dst_addr, src_addr; + size_t dst_skip, src_skip, len = 0; + size_t prev_dst_icg = 0, prev_src_icg = 0; + int i; + + if (!xt || (xt->numf != 1) || (xt->dir != DMA_MEM_TO_MEM)) + return NULL; + + dev_dbg(chan2dev(chan), "%s: src=0x%08x, dest=0x%08x, numf=%d, frame_size=%d, flags=0x%lx\n", + __func__, xt->src_start, xt->dst_start, xt->numf, + xt->frame_size, flags); + + src_addr = xt->src_start; + dst_addr = xt->dst_start; + + for (i = 0; i < xt->frame_size; i++) { + struct at_xdmac_desc *desc; + size_t src_icg, dst_icg; + + chunk = xt->sgl + i; + + dst_icg = dmaengine_get_dst_icg(xt, chunk); + src_icg = dmaengine_get_src_icg(xt, chunk); + + src_skip = chunk->size + src_icg; + dst_skip = chunk->size + dst_icg; + + dev_dbg(chan2dev(chan), + "%s: chunk size=%d, src icg=%d, dst icg=%d\n", + __func__, chunk->size, src_icg, dst_icg); + + /* + * Handle the case where we just have the same + * transfer to setup, we can just increase the + * block number and reuse the same descriptor. + */ + if (prev_chunk && prev && + (prev_chunk->size == chunk->size) && + (prev_src_icg == src_icg) && + (prev_dst_icg == dst_icg)) { + dev_dbg(chan2dev(chan), + "%s: same configuration that the previous chunk, merging the descriptors...\n", + __func__); + at_xdmac_increment_block_count(chan, prev); + continue; + } + + desc = at_xdmac_interleaved_queue_desc(chan, atchan, + prev, + src_addr, dst_addr, + xt, chunk); + if (!desc) { + list_splice_init(&first->descs_list, + &atchan->free_descs_list); + return NULL; + } + + if (!first) + first = desc; + + dev_dbg(chan2dev(chan), "%s: add desc 0x%p to descs_list 0x%p\n", + __func__, desc, first); + list_add_tail(&desc->desc_node, &first->descs_list); + + if (xt->src_sgl) + src_addr += src_skip; + + if (xt->dst_sgl) + dst_addr += dst_skip; + + len += chunk->size; + prev_chunk = chunk; + prev_dst_icg = dst_icg; + prev_src_icg = src_icg; + prev = desc; + } + + first->tx_dma_desc.cookie = -EBUSY; + first->tx_dma_desc.flags = flags; + first->xfer_size = len; + + return &first->tx_dma_desc; +} + static struct dma_async_tx_descriptor * at_xdmac_prep_dma_memcpy(struct dma_chan *chan, dma_addr_t dest, dma_addr_t src, size_t len, unsigned long flags) @@ -814,24 +1046,7 @@ at_xdmac_prep_dma_memcpy(struct dma_chan *chan, dma_addr_t dest, dma_addr_t src, if (unlikely(!len)) return NULL; - /* - * Check address alignment to select the greater data width we can use. - * Some XDMAC implementations don't provide dword transfer, in this - * case selecting dword has the same behavior as selecting word transfers. - */ - if (!((src_addr | dst_addr) & 7)) { - dwidth = AT_XDMAC_CC_DWIDTH_DWORD; - dev_dbg(chan2dev(chan), "%s: dwidth: double word\n", __func__); - } else if (!((src_addr | dst_addr) & 3)) { - dwidth = AT_XDMAC_CC_DWIDTH_WORD; - dev_dbg(chan2dev(chan), "%s: dwidth: word\n", __func__); - } else if (!((src_addr | dst_addr) & 1)) { - dwidth = AT_XDMAC_CC_DWIDTH_HALFWORD; - dev_dbg(chan2dev(chan), "%s: dwidth: half word\n", __func__); - } else { - dwidth = AT_XDMAC_CC_DWIDTH_BYTE; - dev_dbg(chan2dev(chan), "%s: dwidth: byte\n", __func__); - } + dwidth = at_xdmac_align_width(chan, src_addr | dst_addr); /* Prepare descriptors. */ while (remaining_size) { @@ -861,19 +1076,8 @@ at_xdmac_prep_dma_memcpy(struct dma_chan *chan, dma_addr_t dest, dma_addr_t src, dev_dbg(chan2dev(chan), "%s: xfer_size=%zu\n", __func__, xfer_size); /* Check remaining length and change data width if needed. */ - if (!((src_addr | dst_addr | xfer_size) & 7)) { - dwidth = AT_XDMAC_CC_DWIDTH_DWORD; - dev_dbg(chan2dev(chan), "%s: dwidth: double word\n", __func__); - } else if (!((src_addr | dst_addr | xfer_size) & 3)) { - dwidth = AT_XDMAC_CC_DWIDTH_WORD; - dev_dbg(chan2dev(chan), "%s: dwidth: word\n", __func__); - } else if (!((src_addr | dst_addr | xfer_size) & 1)) { - dwidth = AT_XDMAC_CC_DWIDTH_HALFWORD; - dev_dbg(chan2dev(chan), "%s: dwidth: half word\n", __func__); - } else if ((src_addr | dst_addr | xfer_size) & 1) { - dwidth = AT_XDMAC_CC_DWIDTH_BYTE; - dev_dbg(chan2dev(chan), "%s: dwidth: byte\n", __func__); - } + dwidth = at_xdmac_align_width(chan, + src_addr | dst_addr | xfer_size); chan_cc |= AT_XDMAC_CC_DWIDTH(dwidth); ublen = xfer_size >> dwidth; @@ -884,7 +1088,6 @@ at_xdmac_prep_dma_memcpy(struct dma_chan *chan, dma_addr_t dest, dma_addr_t src, desc->lld.mbr_ubc = AT_XDMAC_MBR_UBC_NDV2 | AT_XDMAC_MBR_UBC_NDEN | AT_XDMAC_MBR_UBC_NSEN - | (remaining_size ? AT_XDMAC_MBR_UBC_NDE : 0) | ublen; desc->lld.mbr_cfg = chan_cc; @@ -893,12 +1096,8 @@ at_xdmac_prep_dma_memcpy(struct dma_chan *chan, dma_addr_t dest, dma_addr_t src, __func__, &desc->lld.mbr_sa, &desc->lld.mbr_da, desc->lld.mbr_ubc, desc->lld.mbr_cfg); /* Chain lld. */ - if (prev) { - prev->lld.mbr_nda = desc->tx_dma_desc.phys; - dev_dbg(chan2dev(chan), - "%s: chain lld: prev=0x%p, mbr_nda=0x%08x\n", - __func__, prev, prev->lld.mbr_nda); - } + if (prev) + at_xdmac_queue_desc(chan, prev, desc); prev = desc; if (!first) @@ -915,6 +1114,93 @@ at_xdmac_prep_dma_memcpy(struct dma_chan *chan, dma_addr_t dest, dma_addr_t src, return &first->tx_dma_desc; } +static struct at_xdmac_desc *at_xdmac_memset_create_desc(struct dma_chan *chan, + struct at_xdmac_chan *atchan, + dma_addr_t dst_addr, + size_t len, + int value) +{ + struct at_xdmac_desc *desc; + unsigned long flags; + size_t ublen; + u32 dwidth; + /* + * WARNING: The channel configuration is set here since there is no + * dmaengine_slave_config call in this case. Moreover we don't know the + * direction, it involves we can't dynamically set the source and dest + * interface so we have to use the same one. Only interface 0 allows EBI + * access. Hopefully we can access DDR through both ports (at least on + * SAMA5D4x), so we can use the same interface for source and dest, + * that solves the fact we don't know the direction. + */ + u32 chan_cc = AT_XDMAC_CC_DAM_INCREMENTED_AM + | AT_XDMAC_CC_SAM_INCREMENTED_AM + | AT_XDMAC_CC_DIF(0) + | AT_XDMAC_CC_SIF(0) + | AT_XDMAC_CC_MBSIZE_SIXTEEN + | AT_XDMAC_CC_MEMSET_HW_MODE + | AT_XDMAC_CC_TYPE_MEM_TRAN; + + dwidth = at_xdmac_align_width(chan, dst_addr); + + if (len >= (AT_XDMAC_MBR_UBC_UBLEN_MAX << dwidth)) { + dev_err(chan2dev(chan), + "%s: Transfer too large, aborting...\n", + __func__); + return NULL; + } + + spin_lock_irqsave(&atchan->lock, flags); + desc = at_xdmac_get_desc(atchan); + spin_unlock_irqrestore(&atchan->lock, flags); + if (!desc) { + dev_err(chan2dev(chan), "can't get descriptor\n"); + return NULL; + } + + chan_cc |= AT_XDMAC_CC_DWIDTH(dwidth); + + ublen = len >> dwidth; + + desc->lld.mbr_da = dst_addr; + desc->lld.mbr_ds = value; + desc->lld.mbr_ubc = AT_XDMAC_MBR_UBC_NDV3 + | AT_XDMAC_MBR_UBC_NDEN + | AT_XDMAC_MBR_UBC_NSEN + | ublen; + desc->lld.mbr_cfg = chan_cc; + + dev_dbg(chan2dev(chan), + "%s: lld: mbr_da=0x%08x, mbr_ds=0x%08x, mbr_ubc=0x%08x, mbr_cfg=0x%08x\n", + __func__, desc->lld.mbr_da, desc->lld.mbr_ds, desc->lld.mbr_ubc, + desc->lld.mbr_cfg); + + return desc; +} + +struct dma_async_tx_descriptor * +at_xdmac_prep_dma_memset(struct dma_chan *chan, dma_addr_t dest, int value, + size_t len, unsigned long flags) +{ + struct at_xdmac_chan *atchan = to_at_xdmac_chan(chan); + struct at_xdmac_desc *desc; + + dev_dbg(chan2dev(chan), "%s: dest=0x%08x, len=%d, pattern=0x%x, flags=0x%lx\n", + __func__, dest, len, value, flags); + + if (unlikely(!len)) + return NULL; + + desc = at_xdmac_memset_create_desc(chan, atchan, dest, len, value); + list_add_tail(&desc->desc_node, &desc->descs_list); + + desc->tx_dma_desc.cookie = -EBUSY; + desc->tx_dma_desc.flags = flags; + desc->xfer_size = len; + + return &desc->tx_dma_desc; +} + static enum dma_status at_xdmac_tx_status(struct dma_chan *chan, dma_cookie_t cookie, struct dma_tx_state *txstate) @@ -1445,7 +1731,9 @@ static int at_xdmac_probe(struct platform_device *pdev) } dma_cap_set(DMA_CYCLIC, atxdmac->dma.cap_mask); + dma_cap_set(DMA_INTERLEAVE, atxdmac->dma.cap_mask); dma_cap_set(DMA_MEMCPY, atxdmac->dma.cap_mask); + dma_cap_set(DMA_MEMSET, atxdmac->dma.cap_mask); dma_cap_set(DMA_SLAVE, atxdmac->dma.cap_mask); /* * Without DMA_PRIVATE the driver is not able to allocate more than @@ -1458,7 +1746,9 @@ static int at_xdmac_probe(struct platform_device *pdev) atxdmac->dma.device_tx_status = at_xdmac_tx_status; atxdmac->dma.device_issue_pending = at_xdmac_issue_pending; atxdmac->dma.device_prep_dma_cyclic = at_xdmac_prep_dma_cyclic; + atxdmac->dma.device_prep_interleaved_dma = at_xdmac_prep_interleaved; atxdmac->dma.device_prep_dma_memcpy = at_xdmac_prep_dma_memcpy; + atxdmac->dma.device_prep_dma_memset = at_xdmac_prep_dma_memset; atxdmac->dma.device_prep_slave_sg = at_xdmac_prep_slave_sg; atxdmac->dma.device_config = at_xdmac_device_config; atxdmac->dma.device_pause = at_xdmac_device_pause; |