Julian Margetson <runaway@xxxxxxxx> writes: > On 12/19/2015 4:41 PM, Måns Rullgård wrote: >> Andy Shevchenko <andy.shevchenko@xxxxxxxxx> writes: >> >>> On Sat, Dec 19, 2015 at 10:16 PM, Julian Margetson <runaway@xxxxxxxx> wrote: >>>> On 12/19/2015 3:07 PM, Måns Rullgård wrote: >>>>> Julian Margetson <runaway@xxxxxxxx> writes: >>>>>> Total pages: 522752 >>>>>> [ 0.000000] Kernel command line: root=/dev/sda8 console=ttyS0,115200 >>>>>> console=tty1 dw_dmac_core.dyndbg dw_dmac.dyndbg >>>>> Please add ignore_log_level. >>>>> >>>> Had to truncate the kernel command line to add it. >>> I guess Måns meant 'ignore_loglevel' >> Obviously. I can never remember where the underscores go. > > [ 18.362244] sd 3:0:0:0: [sdc] 976773168 512-byte logical blocks: (500 GB/465 GiB) > [ 18.372454] sd 3:0:0:0: Attached scsi generic sg3 type 0 > [ 18.405433] sd 3:0:0:0: [sdc] Write Protect is off > [ 18.420654] sd 3:0:0:0: [sdc] Mode Sense: 00 3a 00 00 > [ 18.461731] sd 3:0:0:0: [sdc] Write cache: enabled, read cache: enabled, doesn't support DPO or FUA > [ 18.502918] sata-dwc 4bffd1000.sata: sata_dwc_qc_prep_by_tag: port=0 dma dir=from device n_elem=1 > [ 18.511807] dma dma0chan0: dwc_prep_slave_sg > [ 18.516083] dma dma0chan0: scanned 1 descriptors on freelist > [ 18.521753] sata-dwc 4bffd1000.sata: dma_dwc_xfer_setup sg: 0xedeaa800, count: 1 addr: 0xfffffffff6a14400 > [ 18.531327] sata-dwc 4bffd1000.sata: sata_dwc_qc_issue: tag=0 ap->link.sactive = 0x00000001 sactive=0x00000001 > [ 18.541359] sata-dwc 4bffd1000.sata: sata_dwc_exec_command_by_tag cmd(0x60): READ FPDMA QUEUED tag=0 > [ 18.553703] sata-dwc 4bffd1000.sata: sata_dwc_isr intpr=0x00000082 active_tag=-84148995 > [ 18.561717] sata-dwc 4bffd1000.sata: sata_dwc_isr: NEWFP tag=0 > [ 18.567561] sata-dwc 4bffd1000.sata: sata_dwc_bmdma_start_by_tag qc=ed2340b8 tag: 0 cmd: 0x60 dma_dir: from device start_dma? 1 > [ 18.579043] sata-dwc 4bffd1000.sata: taskfile cmd: 0x60 protocol: ATA NCQ flags: 0x17 device: 40 > [ 18.587836] sata-dwc 4bffd1000.sata: feature: 0x08 nsect: 0x0 lbal: 0x0 lbam: 0x0 lbah: 0x0 > [ 18.596196] sata-dwc 4bffd1000.sata: hob_feature: 0x00 hob_nsect: 0x0 hob_lbal: 0x0 hob_lbam: 0x0 hob_lbah: 0x0 > [ 18.606292] dma dma0chan0: dwc_tx_submit: queued 2 > [ 18.611091] dma dma0chan0: dwc_dostart_first_queued: started 2 > [ 48.748614] ata3: lost interrupt (Status 0x40) Now we're getting somewhere. The dma transfer is set up and initiated, but then nothing happens. Comparing the old sata_dwc driver, from before the switch to dmaengine, with the dw_dma driver, I noticed an obvious problem: the descriptors are filled in using the wrong byte order. This patch might fix that. -- Måns Rullgård
>From 04b444b301c8b2db732dbf259dddb3dc87d622c8 Mon Sep 17 00:00:00 2001 From: Mans Rullgard <mans@xxxxxxxxx> Date: Sun, 20 Dec 2015 16:54:21 +0000 Subject: [PATCH] dmaengine: dw: fix byte order of hw descriptor fields If the DMA controller uses a different byte order than the host CPU, the hardware linked list descriptor fields need to be byte-swapped. This patch makes the driver write these fields using the same byte order it uses for mmio accesses to the DMA engine. I do not know if this is guaranteed to always be correct. Signed-off-by: Mans Rullgard <mans@xxxxxxxxx> --- drivers/dma/dw/core.c | 84 +++++++++++++++++++++++++++------------------------ drivers/dma/dw/regs.h | 26 +++++++++++----- 2 files changed, 63 insertions(+), 47 deletions(-) diff --git a/drivers/dma/dw/core.c b/drivers/dma/dw/core.c index 7067b6d..b954904 100644 --- a/drivers/dma/dw/core.c +++ b/drivers/dma/dw/core.c @@ -209,12 +209,12 @@ static inline void dwc_do_single_block(struct dw_dma_chan *dwc, * Software emulation of LLP mode relies on interrupts to continue * multi block transfer. */ - ctllo = desc->lli.ctllo | DWC_CTLL_INT_EN; + ctllo = dw_lli_read(desc->lli.ctllo) | DWC_CTLL_INT_EN; - channel_writel(dwc, SAR, desc->lli.sar); - channel_writel(dwc, DAR, desc->lli.dar); + channel_writel(dwc, SAR, dw_lli_read(desc->lli.sar)); + channel_writel(dwc, DAR, dw_lli_read(desc->lli.dar)); channel_writel(dwc, CTL_LO, ctllo); - channel_writel(dwc, CTL_HI, desc->lli.ctlhi); + channel_writel(dwc, CTL_HI, dw_lli_read(desc->lli.ctlhi)); channel_set_bit(dw, CH_EN, dwc->mask); /* Move pointer to next descriptor */ @@ -432,7 +432,7 @@ static void dwc_scan_descriptors(struct dw_dma *dw, struct dw_dma_chan *dwc) } /* Check first descriptors llp */ - if (desc->lli.llp == llp) { + if (dw_lli_read(desc->lli.llp) == llp) { /* This one is currently in progress */ dwc->residue -= dwc_get_sent(dwc); spin_unlock_irqrestore(&dwc->lock, flags); @@ -441,7 +441,7 @@ static void dwc_scan_descriptors(struct dw_dma *dw, struct dw_dma_chan *dwc) dwc->residue -= desc->len; list_for_each_entry(child, &desc->tx_list, desc_node) { - if (child->lli.llp == llp) { + if (dw_lli_read(child->lli.llp) == llp) { /* Currently in progress */ dwc->residue -= dwc_get_sent(dwc); spin_unlock_irqrestore(&dwc->lock, flags); @@ -730,16 +730,16 @@ dwc_prep_dma_memcpy(struct dma_chan *chan, dma_addr_t dest, dma_addr_t src, if (!desc) goto err_desc_get; - desc->lli.sar = src + offset; - desc->lli.dar = dest + offset; - desc->lli.ctllo = ctllo; - desc->lli.ctlhi = xfer_count; + dw_lli_write(desc->lli.sar, src + offset); + dw_lli_write(desc->lli.dar, dest + offset); + dw_lli_write(desc->lli.ctllo, ctllo); + dw_lli_write(desc->lli.ctlhi, xfer_count); desc->len = xfer_count << src_width; if (!first) { first = desc; } else { - prev->lli.llp = desc->txd.phys; + dw_lli_write(prev->lli.llp, desc->txd.phys); list_add_tail(&desc->desc_node, &first->tx_list); } @@ -748,7 +748,7 @@ dwc_prep_dma_memcpy(struct dma_chan *chan, dma_addr_t dest, dma_addr_t src, if (flags & DMA_PREP_INTERRUPT) /* Trigger interrupt after last block */ - prev->lli.ctllo |= DWC_CTLL_INT_EN; + dw_lli_or(prev->lli.ctllo, DWC_CTLL_INT_EN); prev->lli.llp = 0; first->txd.flags = flags; @@ -818,9 +818,10 @@ slave_sg_todev_fill_desc: if (!desc) goto err_desc_get; - desc->lli.sar = mem; - desc->lli.dar = reg; - desc->lli.ctllo = ctllo | DWC_CTLL_SRC_WIDTH(mem_width); + dw_lli_write(desc->lli.sar, mem); + dw_lli_write(desc->lli.dar, reg); + dw_lli_write(desc->lli.ctllo, + ctllo | DWC_CTLL_SRC_WIDTH(mem_width)); if ((len >> mem_width) > dwc->block_size) { dlen = dwc->block_size << mem_width; mem += dlen; @@ -830,13 +831,13 @@ slave_sg_todev_fill_desc: len = 0; } - desc->lli.ctlhi = dlen >> mem_width; + dw_lli_write(desc->lli.ctlhi, dlen >> mem_width); desc->len = dlen; if (!first) { first = desc; } else { - prev->lli.llp = desc->txd.phys; + dw_lli_write(prev->lli.llp, desc->txd.phys); list_add_tail(&desc->desc_node, &first->tx_list); } @@ -875,9 +876,10 @@ slave_sg_fromdev_fill_desc: if (!desc) goto err_desc_get; - desc->lli.sar = reg; - desc->lli.dar = mem; - desc->lli.ctllo = ctllo | DWC_CTLL_DST_WIDTH(mem_width); + dw_lli_write(desc->lli.sar, reg); + dw_lli_write(desc->lli.dar, mem); + dw_lli_write(desc->lli.ctllo, + ctllo | DWC_CTLL_DST_WIDTH(mem_width)); if ((len >> reg_width) > dwc->block_size) { dlen = dwc->block_size << reg_width; mem += dlen; @@ -886,13 +888,13 @@ slave_sg_fromdev_fill_desc: dlen = len; len = 0; } - desc->lli.ctlhi = dlen >> reg_width; + dw_lli_write(desc->lli.ctlhi, dlen >> reg_width); desc->len = dlen; if (!first) { first = desc; } else { - prev->lli.llp = desc->txd.phys; + dw_lli_write(prev->lli.llp, desc->txd.phys); list_add_tail(&desc->desc_node, &first->tx_list); } @@ -909,7 +911,7 @@ slave_sg_fromdev_fill_desc: if (flags & DMA_PREP_INTERRUPT) /* Trigger interrupt after last block */ - prev->lli.ctllo |= DWC_CTLL_INT_EN; + dw_lli_or(prev->lli.ctllo, DWC_CTLL_INT_EN); prev->lli.llp = 0; first->total_len = total_len; @@ -1393,50 +1395,52 @@ struct dw_cyclic_desc *dw_dma_cyclic_prep(struct dma_chan *chan, switch (direction) { case DMA_MEM_TO_DEV: - desc->lli.dar = sconfig->dst_addr; - desc->lli.sar = buf_addr + (period_len * i); - desc->lli.ctllo = (DWC_DEFAULT_CTLLO(chan) + dw_lli_write(desc->lli.dar, sconfig->dst_addr); + dw_lli_write(desc->lli.sar, + buf_addr + (period_len * i)); + dw_lli_write(desc->lli.ctllo, (DWC_DEFAULT_CTLLO(chan) | DWC_CTLL_DST_WIDTH(reg_width) | DWC_CTLL_SRC_WIDTH(reg_width) | DWC_CTLL_DST_FIX | DWC_CTLL_SRC_INC - | DWC_CTLL_INT_EN); + | DWC_CTLL_INT_EN)); - desc->lli.ctllo |= sconfig->device_fc ? - DWC_CTLL_FC(DW_DMA_FC_P_M2P) : - DWC_CTLL_FC(DW_DMA_FC_D_M2P); + dw_lli_or(desc->lli.ctllo, sconfig->device_fc ? + DWC_CTLL_FC(DW_DMA_FC_P_M2P) : + DWC_CTLL_FC(DW_DMA_FC_D_M2P)); break; case DMA_DEV_TO_MEM: - desc->lli.dar = buf_addr + (period_len * i); - desc->lli.sar = sconfig->src_addr; - desc->lli.ctllo = (DWC_DEFAULT_CTLLO(chan) + dw_lli_write(desc->lli.dar, + buf_addr + (period_len * i)); + dw_lli_write(desc->lli.sar, sconfig->src_addr); + dw_lli_write(desc->lli.ctllo, (DWC_DEFAULT_CTLLO(chan) | DWC_CTLL_SRC_WIDTH(reg_width) | DWC_CTLL_DST_WIDTH(reg_width) | DWC_CTLL_DST_INC | DWC_CTLL_SRC_FIX - | DWC_CTLL_INT_EN); + | DWC_CTLL_INT_EN)); - desc->lli.ctllo |= sconfig->device_fc ? - DWC_CTLL_FC(DW_DMA_FC_P_P2M) : - DWC_CTLL_FC(DW_DMA_FC_D_P2M); + dw_lli_or(desc->lli.ctllo, sconfig->device_fc ? + DWC_CTLL_FC(DW_DMA_FC_P_P2M) : + DWC_CTLL_FC(DW_DMA_FC_D_P2M)); break; default: break; } - desc->lli.ctlhi = (period_len >> reg_width); + dw_lli_write(desc->lli.ctlhi, (period_len >> reg_width)); cdesc->desc[i] = desc; if (last) - last->lli.llp = desc->txd.phys; + dw_lli_write(last->lli.llp, desc->txd.phys); last = desc; } /* Let's make a cyclic list */ - last->lli.llp = cdesc->desc[0]->txd.phys; + dw_lli_write(last->lli.llp, cdesc->desc[0]->txd.phys); dev_dbg(chan2dev(&dwc->chan), "cyclic prepared buf %pad len %zu period %zu periods %d\n", diff --git a/drivers/dma/dw/regs.h b/drivers/dma/dw/regs.h index 241ff2b..84f05de 100644 --- a/drivers/dma/dw/regs.h +++ b/drivers/dma/dw/regs.h @@ -308,20 +308,32 @@ static inline struct dw_dma *to_dw_dma(struct dma_device *ddev) return container_of(ddev, struct dw_dma, dma); } +#ifdef CONFIG_DW_DMAC_BIG_ENDIAN_IO +typedef __be32 dw_u32; +#define dw_lli_read(s) be32_to_cpu(s) +#define dw_lli_write(d, v) ((d) = cpu_to_be32(v)) +#else +typedef __le32 dw_u32; +#define dw_lli_read(s) le32_to_cpu(s) +#define dw_lli_write(d, v) ((d) = cpu_to_le32(v)) +#endif + +#define dw_lli_or(d, v) dw_lli_write(d, dw_lli_read(d) | (v)) + /* LLI == Linked List Item; a.k.a. DMA block descriptor */ struct dw_lli { /* values that are not changed by hardware */ - u32 sar; - u32 dar; - u32 llp; /* chain to next lli */ - u32 ctllo; + dw_u32 sar; + dw_u32 dar; + dw_u32 llp; /* chain to next lli */ + dw_u32 ctllo; /* values that may get written back: */ - u32 ctlhi; + dw_u32 ctlhi; /* sstat and dstat can snapshot peripheral register state. * silicon config may discard either or both... */ - u32 sstat; - u32 dstat; + dw_u32 sstat; + dw_u32 dstat; }; struct dw_desc { -- 2.6.3