This series adds support for the IMG Multi-threaded DMA Controller (MDC)
which is found on IMG SoCs. Currently this driver only supports the
variant found on the MIPS-based Pistachio SoC.
The MDC supports slave and memory-to-memory transfers on up to 32 channels.
Requests from channels are handled by MDC threads. Thread assignments are
per-channel and are specified in the device-tree.
Tested on a platform based on the Pistachio SoC with additional patches.
Support for Pistachio will be submitted later. Slave DMA was tested
with the recently submitted IMG SPFI driver [1] and memory-to-memory
DMA was tested with the dmatest module.
Based on Vinod's slave-dma/topic/slave_caps_device_control_fix.
Changes from v2:
- addressed Vinod's review comments
Changes from v1:
- addressed Arnd's review comments
- rebased on Maxime's dmaengine cleanups
[1] https://lkml.org/lkml/2014/11/12/763
Andrew Bresticker (2):
dmaengine: Add binding document for IMG MDC
dmaengine: Add driver for IMG MDC
.../devicetree/bindings/dma/img-mdc-dma.txt | 57 ++
drivers/dma/Kconfig | 9 +
drivers/dma/Makefile | 1 +
drivers/dma/img-mdc-dma.c | 1011 ++++++++++++++++++++
4 files changed, 1078 insertions(+)
create mode 100644 Documentation/devicetree/bindings/dma/img-mdc-dma.txt
create mode 100644 drivers/dma/img-mdc-dma.c
--
2.2.0.rc0.207.ga3a616c
Add a binding document for the IMG Multi-threaded DMA Controller (MDC)
present on the MIPS-based Pistachio and other IMG SoCs.
Signed-off-by: Andrew Bresticker <[email protected]>
Acked-by: Arnd Bergmann <[email protected]>
---
No changes from v1/v2.
---
.../devicetree/bindings/dma/img-mdc-dma.txt | 57 ++++++++++++++++++++++
1 file changed, 57 insertions(+)
create mode 100644 Documentation/devicetree/bindings/dma/img-mdc-dma.txt
diff --git a/Documentation/devicetree/bindings/dma/img-mdc-dma.txt b/Documentation/devicetree/bindings/dma/img-mdc-dma.txt
new file mode 100644
index 0000000..28c1341
--- /dev/null
+++ b/Documentation/devicetree/bindings/dma/img-mdc-dma.txt
@@ -0,0 +1,57 @@
+* IMG Multi-threaded DMA Controller (MDC)
+
+Required properties:
+- compatible: Must be "img,pistachio-mdc-dma".
+- reg: Must contain the base address and length of the MDC registers.
+- interrupts: Must contain all the per-channel DMA interrupts.
+- clocks: Must contain an entry for each entry in clock-names.
+ See ../clock/clock-bindings.txt for details.
+- clock-names: Must include the following entries:
+ - sys: MDC system interface clock.
+- img,cr-periph: Must contain a phandle to the peripheral control syscon
+ node which contains the DMA request to channel mapping registers.
+- img,max-burst-multiplier: Must be the maximum supported burst size multiplier.
+ The maximum burst size is this value multiplied by the hardware-reported bus
+ width.
+- #dma-cells: Must be 3:
+ - The first cell is the peripheral's DMA request line.
+ - The second cell is a bitmap specifying to which channels the DMA request
+ line may be mapped (i.e. bit N set indicates channel N is usable).
+ - The third cell is the thread ID to be used by the channel.
+
+Optional properties:
+- dma-channels: Number of supported DMA channels, up to 32. If not specified
+ the number reported by the hardware is used.
+
+Example:
+
+mdc: dma-controller@18143000 {
+ compatible = "img,pistachio-mdc-dma";
+ reg = <0x18143000 0x1000>;
+ interrupts = <GIC_SHARED 27 IRQ_TYPE_LEVEL_HIGH>,
+ <GIC_SHARED 28 IRQ_TYPE_LEVEL_HIGH>,
+ <GIC_SHARED 29 IRQ_TYPE_LEVEL_HIGH>,
+ <GIC_SHARED 30 IRQ_TYPE_LEVEL_HIGH>,
+ <GIC_SHARED 31 IRQ_TYPE_LEVEL_HIGH>,
+ <GIC_SHARED 32 IRQ_TYPE_LEVEL_HIGH>,
+ <GIC_SHARED 33 IRQ_TYPE_LEVEL_HIGH>,
+ <GIC_SHARED 34 IRQ_TYPE_LEVEL_HIGH>,
+ <GIC_SHARED 35 IRQ_TYPE_LEVEL_HIGH>,
+ <GIC_SHARED 36 IRQ_TYPE_LEVEL_HIGH>,
+ <GIC_SHARED 37 IRQ_TYPE_LEVEL_HIGH>,
+ <GIC_SHARED 38 IRQ_TYPE_LEVEL_HIGH>;
+ clocks = <&system_clk>;
+ clock-names = "sys";
+
+ img,max-burst-multiplier = <16>;
+ img,cr-periph = <&cr_periph>;
+
+ #dma-cells = <3>;
+};
+
+spi@18100f00 {
+ ...
+ dmas = <&mdc 9 0xffffffff 0>, <&mdc 10 0xffffffff 0>;
+ dma-names = "tx", "rx";
+ ...
+};
--
2.2.0.rc0.207.ga3a616c
Add support for the IMG Multi-threaded DMA Controller (MDC) found on
certain IMG SoCs. Currently this driver supports the variant present
on the MIPS-based Pistachio SoC.
Signed-off-by: Andrew Bresticker <[email protected]>
Acked-by: Arnd Bergmann <[email protected]>
---
Changes from v2:
- addressed review feedback from Vinod
Changes from v1:
- split device_control
- added generic slave caps
- iterate through channels in of_xlate() instead of dma_request_channel()
- drop locking around regmap accesses
---
drivers/dma/Kconfig | 9 +
drivers/dma/Makefile | 1 +
drivers/dma/img-mdc-dma.c | 1011 +++++++++++++++++++++++++++++++++++++++++++++
3 files changed, 1021 insertions(+)
create mode 100644 drivers/dma/img-mdc-dma.c
diff --git a/drivers/dma/Kconfig b/drivers/dma/Kconfig
index f2b2c4e..8990d48 100644
--- a/drivers/dma/Kconfig
+++ b/drivers/dma/Kconfig
@@ -416,6 +416,15 @@ config NBPFAXI_DMA
help
Support for "Type-AXI" NBPF DMA IPs from Renesas
+config IMG_MDC_DMA
+ tristate "IMG MDC support"
+ depends on MIPS || COMPILE_TEST
+ depends on MFD_SYSCON
+ select DMA_ENGINE
+ select DMA_VIRTUAL_CHANNELS
+ help
+ Enable support for the IMG multi-threaded DMA controller (MDC).
+
config DMA_ENGINE
bool
diff --git a/drivers/dma/Makefile b/drivers/dma/Makefile
index 2022b54..5b8c8bc 100644
--- a/drivers/dma/Makefile
+++ b/drivers/dma/Makefile
@@ -50,3 +50,4 @@ obj-y += xilinx/
obj-$(CONFIG_INTEL_MIC_X100_DMA) += mic_x100_dma.o
obj-$(CONFIG_NBPFAXI_DMA) += nbpfaxi.o
obj-$(CONFIG_DMA_SUN6I) += sun6i-dma.o
+obj-$(CONFIG_IMG_MDC_DMA) += img-mdc-dma.o
diff --git a/drivers/dma/img-mdc-dma.c b/drivers/dma/img-mdc-dma.c
new file mode 100644
index 0000000..ed045a9
--- /dev/null
+++ b/drivers/dma/img-mdc-dma.c
@@ -0,0 +1,1011 @@
+/*
+ * IMG Multi-threaded DMA Controller (MDC)
+ *
+ * Copyright (C) 2009,2012,2013 Imagination Technologies Ltd.
+ * Copyright (C) 2014 Google, Inc.
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms and conditions of the GNU General Public License,
+ * version 2, as published by the Free Software Foundation.
+ */
+
+#include <linux/clk.h>
+#include <linux/dma-mapping.h>
+#include <linux/dmaengine.h>
+#include <linux/dmapool.h>
+#include <linux/interrupt.h>
+#include <linux/io.h>
+#include <linux/irq.h>
+#include <linux/kernel.h>
+#include <linux/mfd/syscon.h>
+#include <linux/module.h>
+#include <linux/of.h>
+#include <linux/of_device.h>
+#include <linux/of_dma.h>
+#include <linux/platform_device.h>
+#include <linux/regmap.h>
+#include <linux/slab.h>
+#include <linux/spinlock.h>
+
+#include "dmaengine.h"
+#include "virt-dma.h"
+
+#define MDC_MAX_DMA_CHANNELS 32
+
+#define MDC_GENERAL_CONFIG 0x000
+#define MDC_GENERAL_CONFIG_LIST_IEN BIT(31)
+#define MDC_GENERAL_CONFIG_IEN BIT(29)
+#define MDC_GENERAL_CONFIG_LEVEL_INT BIT(28)
+#define MDC_GENERAL_CONFIG_INC_W BIT(12)
+#define MDC_GENERAL_CONFIG_INC_R BIT(8)
+#define MDC_GENERAL_CONFIG_PHYSICAL_W BIT(7)
+#define MDC_GENERAL_CONFIG_WIDTH_W_SHIFT 4
+#define MDC_GENERAL_CONFIG_WIDTH_W_MASK 0x7
+#define MDC_GENERAL_CONFIG_PHYSICAL_R BIT(3)
+#define MDC_GENERAL_CONFIG_WIDTH_R_SHIFT 0
+#define MDC_GENERAL_CONFIG_WIDTH_R_MASK 0x7
+
+#define MDC_READ_PORT_CONFIG 0x004
+#define MDC_READ_PORT_CONFIG_STHREAD_SHIFT 28
+#define MDC_READ_PORT_CONFIG_STHREAD_MASK 0xf
+#define MDC_READ_PORT_CONFIG_RTHREAD_SHIFT 24
+#define MDC_READ_PORT_CONFIG_RTHREAD_MASK 0xf
+#define MDC_READ_PORT_CONFIG_WTHREAD_SHIFT 16
+#define MDC_READ_PORT_CONFIG_WTHREAD_MASK 0xf
+#define MDC_READ_PORT_CONFIG_BURST_SIZE_SHIFT 4
+#define MDC_READ_PORT_CONFIG_BURST_SIZE_MASK 0xff
+#define MDC_READ_PORT_CONFIG_DREQ_ENABLE BIT(1)
+
+#define MDC_READ_ADDRESS 0x008
+
+#define MDC_WRITE_ADDRESS 0x00c
+
+#define MDC_TRANSFER_SIZE 0x010
+#define MDC_TRANSFER_SIZE_MASK 0xffffff
+
+#define MDC_LIST_NODE_ADDRESS 0x014
+
+#define MDC_CMDS_PROCESSED 0x018
+#define MDC_CMDS_PROCESSED_CMDS_PROCESSED_SHIFT 16
+#define MDC_CMDS_PROCESSED_CMDS_PROCESSED_MASK 0x3f
+#define MDC_CMDS_PROCESSED_INT_ACTIVE BIT(8)
+#define MDC_CMDS_PROCESSED_CMDS_DONE_SHIFT 0
+#define MDC_CMDS_PROCESSED_CMDS_DONE_MASK 0x3f
+
+#define MDC_CONTROL_AND_STATUS 0x01c
+#define MDC_CONTROL_AND_STATUS_CANCEL BIT(20)
+#define MDC_CONTROL_AND_STATUS_LIST_EN BIT(4)
+#define MDC_CONTROL_AND_STATUS_EN BIT(0)
+
+#define MDC_ACTIVE_TRANSFER_SIZE 0x030
+
+#define MDC_GLOBAL_CONFIG_A 0x900
+#define MDC_GLOBAL_CONFIG_A_THREAD_ID_WIDTH_SHIFT 16
+#define MDC_GLOBAL_CONFIG_A_THREAD_ID_WIDTH_MASK 0xff
+#define MDC_GLOBAL_CONFIG_A_DMA_CONTEXTS_SHIFT 8
+#define MDC_GLOBAL_CONFIG_A_DMA_CONTEXTS_MASK 0xff
+#define MDC_GLOBAL_CONFIG_A_SYS_DAT_WIDTH_SHIFT 0
+#define MDC_GLOBAL_CONFIG_A_SYS_DAT_WIDTH_MASK 0xff
+
+struct mdc_hw_list_desc {
+ u32 gen_conf;
+ u32 readport_conf;
+ u32 read_addr;
+ u32 write_addr;
+ u32 xfer_size;
+ u32 node_addr;
+ u32 cmds_done;
+ u32 ctrl_status;
+ /*
+ * Not part of the list descriptor, but instead used by the CPU to
+ * traverse the list.
+ */
+ struct mdc_hw_list_desc *next_desc;
+};
+
+struct mdc_tx_desc {
+ struct mdc_chan *chan;
+ struct virt_dma_desc vd;
+ dma_addr_t list_phys;
+ struct mdc_hw_list_desc *list;
+ bool cyclic;
+ bool cmd_loaded;
+ unsigned int list_len;
+ unsigned int list_period_len;
+ size_t list_xfer_size;
+ unsigned int list_cmds_done;
+};
+
+struct mdc_chan {
+ struct mdc_dma *mdma;
+ struct virt_dma_chan vc;
+ struct dma_slave_config config;
+ struct mdc_tx_desc *desc;
+ int irq;
+ unsigned int periph;
+ unsigned int thread;
+ unsigned int chan_nr;
+};
+
+struct mdc_dma_soc_data {
+ void (*enable_chan)(struct mdc_chan *mchan);
+ void (*disable_chan)(struct mdc_chan *mchan);
+};
+
+struct mdc_dma {
+ struct dma_device dma_dev;
+ void __iomem *regs;
+ struct clk *clk;
+ struct dma_pool *desc_pool;
+ struct regmap *periph_regs;
+ spinlock_t lock;
+ unsigned int nr_threads;
+ unsigned int nr_channels;
+ unsigned int bus_width;
+ unsigned int max_burst_mult;
+ unsigned int max_xfer_size;
+ const struct mdc_dma_soc_data *soc;
+ struct mdc_chan channels[MDC_MAX_DMA_CHANNELS];
+};
+
+static inline u32 mdc_readl(struct mdc_dma *mdma, u32 reg)
+{
+ return readl(mdma->regs + reg);
+}
+
+static inline void mdc_writel(struct mdc_dma *mdma, u32 val, u32 reg)
+{
+ writel(val, mdma->regs + reg);
+}
+
+static inline u32 mdc_chan_readl(struct mdc_chan *mchan, u32 reg)
+{
+ return mdc_readl(mchan->mdma, mchan->chan_nr * 0x040 + reg);
+}
+
+static inline void mdc_chan_writel(struct mdc_chan *mchan, u32 val, u32 reg)
+{
+ mdc_writel(mchan->mdma, val, mchan->chan_nr * 0x040 + reg);
+}
+
+static inline struct mdc_chan *to_mdc_chan(struct dma_chan *c)
+{
+ return container_of(to_virt_chan(c), struct mdc_chan, vc);
+}
+
+static inline struct mdc_tx_desc *to_mdc_desc(struct dma_async_tx_descriptor *t)
+{
+ struct virt_dma_desc *vdesc = container_of(t, struct virt_dma_desc, tx);
+
+ return container_of(vdesc, struct mdc_tx_desc, vd);
+}
+
+static inline struct device *mdma2dev(struct mdc_dma *mdma)
+{
+ return mdma->dma_dev.dev;
+}
+
+static inline unsigned int to_mdc_width(unsigned int bytes)
+{
+ return ffs(bytes) - 1;
+}
+
+static inline void mdc_set_read_width(struct mdc_hw_list_desc *ldesc,
+ unsigned int bytes)
+{
+ ldesc->gen_conf |= to_mdc_width(bytes) <<
+ MDC_GENERAL_CONFIG_WIDTH_R_SHIFT;
+}
+
+static inline void mdc_set_write_width(struct mdc_hw_list_desc *ldesc,
+ unsigned int bytes)
+{
+ ldesc->gen_conf |= to_mdc_width(bytes) <<
+ MDC_GENERAL_CONFIG_WIDTH_W_SHIFT;
+}
+
+static void mdc_list_desc_config(struct mdc_chan *mchan,
+ struct mdc_hw_list_desc *ldesc,
+ enum dma_transfer_direction dir,
+ dma_addr_t src, dma_addr_t dst, size_t len)
+{
+ struct mdc_dma *mdma = mchan->mdma;
+ unsigned int max_burst, burst_size;
+
+ ldesc->gen_conf = MDC_GENERAL_CONFIG_IEN | MDC_GENERAL_CONFIG_LIST_IEN |
+ MDC_GENERAL_CONFIG_LEVEL_INT | MDC_GENERAL_CONFIG_PHYSICAL_W |
+ MDC_GENERAL_CONFIG_PHYSICAL_R;
+ ldesc->readport_conf =
+ (mchan->thread << MDC_READ_PORT_CONFIG_STHREAD_SHIFT) |
+ (mchan->thread << MDC_READ_PORT_CONFIG_RTHREAD_SHIFT) |
+ (mchan->thread << MDC_READ_PORT_CONFIG_WTHREAD_SHIFT);
+ ldesc->read_addr = src;
+ ldesc->write_addr = dst;
+ ldesc->xfer_size = len - 1;
+ ldesc->node_addr = 0;
+ ldesc->cmds_done = 0;
+ ldesc->ctrl_status = MDC_CONTROL_AND_STATUS_LIST_EN |
+ MDC_CONTROL_AND_STATUS_EN;
+ ldesc->next_desc = NULL;
+
+ if (IS_ALIGNED(dst, mdma->bus_width) &&
+ IS_ALIGNED(src, mdma->bus_width))
+ max_burst = mdma->bus_width * mdma->max_burst_mult;
+ else
+ max_burst = mdma->bus_width * (mdma->max_burst_mult - 1);
+
+ if (dir == DMA_MEM_TO_DEV) {
+ ldesc->gen_conf |= MDC_GENERAL_CONFIG_INC_R;
+ ldesc->readport_conf |= MDC_READ_PORT_CONFIG_DREQ_ENABLE;
+ mdc_set_read_width(ldesc, mdma->bus_width);
+ mdc_set_write_width(ldesc, mchan->config.dst_addr_width);
+ burst_size = min(max_burst, mchan->config.dst_maxburst *
+ mchan->config.dst_addr_width);
+ } else if (dir == DMA_DEV_TO_MEM) {
+ ldesc->gen_conf |= MDC_GENERAL_CONFIG_INC_W;
+ ldesc->readport_conf |= MDC_READ_PORT_CONFIG_DREQ_ENABLE;
+ mdc_set_read_width(ldesc, mchan->config.src_addr_width);
+ mdc_set_write_width(ldesc, mdma->bus_width);
+ burst_size = min(max_burst, mchan->config.src_maxburst *
+ mchan->config.src_addr_width);
+ } else {
+ ldesc->gen_conf |= MDC_GENERAL_CONFIG_INC_R |
+ MDC_GENERAL_CONFIG_INC_W;
+ mdc_set_read_width(ldesc, mdma->bus_width);
+ mdc_set_write_width(ldesc, mdma->bus_width);
+ burst_size = max_burst;
+ }
+ ldesc->readport_conf |= (burst_size - 1) <<
+ MDC_READ_PORT_CONFIG_BURST_SIZE_SHIFT;
+}
+
+static void mdc_list_desc_free(struct mdc_tx_desc *mdesc)
+{
+ struct mdc_dma *mdma = mdesc->chan->mdma;
+ struct mdc_hw_list_desc *curr, *next;
+ dma_addr_t curr_phys, next_phys;
+
+ curr = mdesc->list;
+ curr_phys = mdesc->list_phys;
+ while (curr) {
+ next = curr->next_desc;
+ next_phys = curr->node_addr;
+ dma_pool_free(mdma->desc_pool, curr, curr_phys);
+ curr = next;
+ curr_phys = next_phys;
+ }
+}
+
+static void mdc_desc_free(struct virt_dma_desc *vd)
+{
+ struct mdc_tx_desc *mdesc = to_mdc_desc(&vd->tx);
+
+ mdc_list_desc_free(mdesc);
+ kfree(mdesc);
+}
+
+static struct dma_async_tx_descriptor *mdc_prep_dma_memcpy(
+ struct dma_chan *chan, dma_addr_t dest, dma_addr_t src, size_t len,
+ unsigned long flags)
+{
+ struct mdc_chan *mchan = to_mdc_chan(chan);
+ struct mdc_dma *mdma = mchan->mdma;
+ struct mdc_tx_desc *mdesc;
+ struct mdc_hw_list_desc *curr, *prev = NULL;
+ dma_addr_t curr_phys, prev_phys;
+
+ if (!len)
+ return NULL;
+
+ mdesc = kzalloc(sizeof(*mdesc), GFP_NOWAIT);
+ if (!mdesc)
+ return NULL;
+ mdesc->chan = mchan;
+ mdesc->list_xfer_size = len;
+
+ while (len > 0) {
+ size_t xfer_size;
+
+ curr = dma_pool_alloc(mdma->desc_pool, GFP_NOWAIT, &curr_phys);
+ if (!curr)
+ goto free_desc;
+
+ if (prev) {
+ prev->node_addr = curr_phys;
+ prev->next_desc = curr;
+ } else {
+ mdesc->list_phys = curr_phys;
+ mdesc->list = curr;
+ }
+
+ xfer_size = min_t(size_t, mdma->max_xfer_size, len);
+
+ mdc_list_desc_config(mchan, curr, DMA_MEM_TO_MEM, src, dest,
+ xfer_size);
+
+ prev = curr;
+ prev_phys = curr_phys;
+
+ mdesc->list_len++;
+ src += xfer_size;
+ dest += xfer_size;
+ len -= xfer_size;
+ }
+
+ return vchan_tx_prep(&mchan->vc, &mdesc->vd, flags);
+
+free_desc:
+ mdc_desc_free(&mdesc->vd);
+
+ return NULL;
+}
+
+static int mdc_check_slave_width(struct mdc_chan *mchan,
+ enum dma_transfer_direction dir)
+{
+ enum dma_slave_buswidth width;
+
+ if (dir == DMA_MEM_TO_DEV)
+ width = mchan->config.dst_addr_width;
+ else
+ width = mchan->config.src_addr_width;
+
+ switch (width) {
+ case DMA_SLAVE_BUSWIDTH_1_BYTE:
+ case DMA_SLAVE_BUSWIDTH_2_BYTES:
+ case DMA_SLAVE_BUSWIDTH_4_BYTES:
+ case DMA_SLAVE_BUSWIDTH_8_BYTES:
+ break;
+ default:
+ return -EINVAL;
+ }
+
+ if (width > mchan->mdma->bus_width)
+ return -EINVAL;
+
+ return 0;
+}
+
+static struct dma_async_tx_descriptor *mdc_prep_dma_cyclic(
+ struct dma_chan *chan, dma_addr_t buf_addr, size_t buf_len,
+ size_t period_len, enum dma_transfer_direction dir,
+ unsigned long flags)
+{
+ struct mdc_chan *mchan = to_mdc_chan(chan);
+ struct mdc_dma *mdma = mchan->mdma;
+ struct mdc_tx_desc *mdesc;
+ struct mdc_hw_list_desc *curr, *prev = NULL;
+ dma_addr_t curr_phys, prev_phys;
+
+ if (!buf_len && !period_len)
+ return NULL;
+
+ if (!is_slave_direction(dir))
+ return NULL;
+
+ if (mdc_check_slave_width(mchan, dir) < 0)
+ return NULL;
+
+ mdesc = kzalloc(sizeof(*mdesc), GFP_NOWAIT);
+ if (!mdesc)
+ return NULL;
+ mdesc->chan = mchan;
+ mdesc->cyclic = true;
+ mdesc->list_xfer_size = buf_len;
+ mdesc->list_period_len = DIV_ROUND_UP(period_len,
+ mdma->max_xfer_size);
+
+ while (buf_len > 0) {
+ size_t remainder = min(period_len, buf_len);
+
+ while (remainder > 0) {
+ size_t xfer_size;
+
+ curr = dma_pool_alloc(mdma->desc_pool, GFP_NOWAIT,
+ &curr_phys);
+ if (!curr)
+ goto free_desc;
+
+ if (!prev) {
+ mdesc->list_phys = curr_phys;
+ mdesc->list = curr;
+ } else {
+ prev->node_addr = curr_phys;
+ prev->next_desc = curr;
+ }
+
+ xfer_size = min_t(size_t, mdma->max_xfer_size,
+ remainder);
+
+ if (dir == DMA_MEM_TO_DEV) {
+ mdc_list_desc_config(mchan, curr, dir,
+ buf_addr,
+ mchan->config.dst_addr,
+ xfer_size);
+ } else {
+ mdc_list_desc_config(mchan, curr, dir,
+ mchan->config.src_addr,
+ buf_addr,
+ xfer_size);
+ }
+
+ prev = curr;
+ prev_phys = curr_phys;
+
+ mdesc->list_len++;
+ buf_addr += xfer_size;
+ buf_len -= xfer_size;
+ remainder -= xfer_size;
+ }
+ }
+ prev->node_addr = mdesc->list_phys;
+
+ return vchan_tx_prep(&mchan->vc, &mdesc->vd, flags);
+
+free_desc:
+ mdc_desc_free(&mdesc->vd);
+
+ return NULL;
+}
+
+static struct dma_async_tx_descriptor *mdc_prep_slave_sg(
+ struct dma_chan *chan, struct scatterlist *sgl,
+ unsigned int sg_len, enum dma_transfer_direction dir,
+ unsigned long flags, void *context)
+{
+ struct mdc_chan *mchan = to_mdc_chan(chan);
+ struct mdc_dma *mdma = mchan->mdma;
+ struct mdc_tx_desc *mdesc;
+ struct scatterlist *sg;
+ struct mdc_hw_list_desc *curr, *prev = NULL;
+ dma_addr_t curr_phys, prev_phys;
+ unsigned int i;
+
+ if (!sgl)
+ return NULL;
+
+ if (!is_slave_direction(dir))
+ return NULL;
+
+ if (mdc_check_slave_width(mchan, dir) < 0)
+ return NULL;
+
+ mdesc = kzalloc(sizeof(*mdesc), GFP_NOWAIT);
+ if (!mdesc)
+ return NULL;
+ mdesc->chan = mchan;
+
+ for_each_sg(sgl, sg, sg_len, i) {
+ dma_addr_t buf = sg_dma_address(sg);
+ size_t buf_len = sg_dma_len(sg);
+
+ while (buf_len > 0) {
+ size_t xfer_size;
+
+ curr = dma_pool_alloc(mdma->desc_pool, GFP_NOWAIT,
+ &curr_phys);
+ if (!curr)
+ goto free_desc;
+
+ if (!prev) {
+ mdesc->list_phys = curr_phys;
+ mdesc->list = curr;
+ } else {
+ prev->node_addr = curr_phys;
+ prev->next_desc = curr;
+ }
+
+ xfer_size = min_t(size_t, mdma->max_xfer_size,
+ buf_len);
+
+ if (dir == DMA_MEM_TO_DEV) {
+ mdc_list_desc_config(mchan, curr, dir, buf,
+ mchan->config.dst_addr,
+ xfer_size);
+ } else {
+ mdc_list_desc_config(mchan, curr, dir,
+ mchan->config.src_addr,
+ buf, xfer_size);
+ }
+
+ prev = curr;
+ prev_phys = curr_phys;
+
+ mdesc->list_len++;
+ mdesc->list_xfer_size += xfer_size;
+ buf += xfer_size;
+ buf_len -= xfer_size;
+ }
+ }
+
+ return vchan_tx_prep(&mchan->vc, &mdesc->vd, flags);
+
+free_desc:
+ mdc_desc_free(&mdesc->vd);
+
+ return NULL;
+}
+
+static void mdc_issue_desc(struct mdc_chan *mchan)
+{
+ struct mdc_dma *mdma = mchan->mdma;
+ struct virt_dma_desc *vd;
+ struct mdc_tx_desc *mdesc;
+ u32 val;
+
+ vd = vchan_next_desc(&mchan->vc);
+ if (!vd)
+ return;
+
+ list_del(&vd->node);
+
+ mdesc = to_mdc_desc(&vd->tx);
+ mchan->desc = mdesc;
+
+ dev_dbg(mdma2dev(mdma), "Issuing descriptor on channel %d\n",
+ mchan->chan_nr);
+
+ mdma->soc->enable_chan(mchan);
+
+ val = mdc_chan_readl(mchan, MDC_GENERAL_CONFIG);
+ val |= MDC_GENERAL_CONFIG_LIST_IEN | MDC_GENERAL_CONFIG_IEN |
+ MDC_GENERAL_CONFIG_LEVEL_INT | MDC_GENERAL_CONFIG_PHYSICAL_W |
+ MDC_GENERAL_CONFIG_PHYSICAL_R;
+ mdc_chan_writel(mchan, val, MDC_GENERAL_CONFIG);
+ val = (mchan->thread << MDC_READ_PORT_CONFIG_STHREAD_SHIFT) |
+ (mchan->thread << MDC_READ_PORT_CONFIG_RTHREAD_SHIFT) |
+ (mchan->thread << MDC_READ_PORT_CONFIG_WTHREAD_SHIFT);
+ mdc_chan_writel(mchan, val, MDC_READ_PORT_CONFIG);
+ mdc_chan_writel(mchan, mdesc->list_phys, MDC_LIST_NODE_ADDRESS);
+ val = mdc_chan_readl(mchan, MDC_CONTROL_AND_STATUS);
+ val |= MDC_CONTROL_AND_STATUS_LIST_EN;
+ mdc_chan_writel(mchan, val, MDC_CONTROL_AND_STATUS);
+}
+
+static void mdc_issue_pending(struct dma_chan *chan)
+{
+ struct mdc_chan *mchan = to_mdc_chan(chan);
+ unsigned long flags;
+
+ spin_lock_irqsave(&mchan->vc.lock, flags);
+ if (vchan_issue_pending(&mchan->vc) && !mchan->desc)
+ mdc_issue_desc(mchan);
+ spin_unlock_irqrestore(&mchan->vc.lock, flags);
+}
+
+static enum dma_status mdc_tx_status(struct dma_chan *chan,
+ dma_cookie_t cookie, struct dma_tx_state *txstate)
+{
+ struct mdc_chan *mchan = to_mdc_chan(chan);
+ struct mdc_tx_desc *mdesc;
+ struct virt_dma_desc *vd;
+ unsigned long flags;
+ size_t bytes = 0;
+ int ret;
+
+ ret = dma_cookie_status(chan, cookie, txstate);
+ if (ret == DMA_COMPLETE)
+ return ret;
+
+ if (!txstate)
+ return ret;
+
+ spin_lock_irqsave(&mchan->vc.lock, flags);
+ vd = vchan_find_desc(&mchan->vc, cookie);
+ if (vd) {
+ mdesc = to_mdc_desc(&vd->tx);
+ bytes = mdesc->list_xfer_size;
+ } else if (mchan->desc && mchan->desc->vd.tx.cookie == cookie) {
+ struct mdc_hw_list_desc *ldesc;
+ u32 val1, val2, done, processed, residue;
+ int i, cmds;
+
+ mdesc = mchan->desc;
+
+ /*
+ * Determine the number of commands that haven't been
+ * processed (handled by the IRQ handler) yet.
+ */
+ do {
+ val1 = mdc_chan_readl(mchan, MDC_CMDS_PROCESSED) &
+ ~MDC_CMDS_PROCESSED_INT_ACTIVE;
+ residue = mdc_chan_readl(mchan,
+ MDC_ACTIVE_TRANSFER_SIZE);
+ val2 = mdc_chan_readl(mchan, MDC_CMDS_PROCESSED) &
+ ~MDC_CMDS_PROCESSED_INT_ACTIVE;
+ } while (val1 != val2);
+
+ done = (val1 >> MDC_CMDS_PROCESSED_CMDS_DONE_SHIFT) &
+ MDC_CMDS_PROCESSED_CMDS_DONE_MASK;
+ processed = (val1 >> MDC_CMDS_PROCESSED_CMDS_PROCESSED_SHIFT) &
+ MDC_CMDS_PROCESSED_CMDS_PROCESSED_MASK;
+ cmds = (done - processed) %
+ (MDC_CMDS_PROCESSED_CMDS_DONE_MASK + 1);
+
+ /*
+ * If the command loaded event hasn't been processed yet, then
+ * the difference above includes an extra command.
+ */
+ if (!mdesc->cmd_loaded)
+ cmds--;
+ else
+ cmds += mdesc->list_cmds_done;
+
+ bytes = mdesc->list_xfer_size;
+ ldesc = mdesc->list;
+ for (i = 0; i < cmds; i++) {
+ bytes -= ldesc->xfer_size + 1;
+ ldesc = ldesc->next_desc;
+ }
+ if (ldesc) {
+ if (residue != MDC_TRANSFER_SIZE_MASK)
+ bytes -= ldesc->xfer_size - residue;
+ else
+ bytes -= ldesc->xfer_size + 1;
+ }
+ }
+ spin_unlock_irqrestore(&mchan->vc.lock, flags);
+
+ dma_set_residue(txstate, bytes);
+
+ return ret;
+}
+
+static int mdc_terminate_all(struct dma_chan *chan)
+{
+ struct mdc_chan *mchan = to_mdc_chan(chan);
+ struct mdc_tx_desc *mdesc;
+ unsigned long flags;
+ LIST_HEAD(head);
+
+ spin_lock_irqsave(&mchan->vc.lock, flags);
+
+ mdc_chan_writel(mchan, MDC_CONTROL_AND_STATUS_CANCEL,
+ MDC_CONTROL_AND_STATUS);
+
+ mdesc = mchan->desc;
+ mchan->desc = NULL;
+ vchan_get_all_descriptors(&mchan->vc, &head);
+
+ spin_unlock_irqrestore(&mchan->vc.lock, flags);
+
+ if (mdesc)
+ mdc_desc_free(&mdesc->vd);
+ vchan_dma_desc_free_list(&mchan->vc, &head);
+
+ return 0;
+}
+
+static int mdc_slave_config(struct dma_chan *chan,
+ struct dma_slave_config *config)
+{
+ struct mdc_chan *mchan = to_mdc_chan(chan);
+ unsigned long flags;
+
+ spin_lock_irqsave(&mchan->vc.lock, flags);
+ mchan->config = *config;
+ spin_unlock_irqrestore(&mchan->vc.lock, flags);
+
+ return 0;
+}
+
+static int mdc_alloc_chan_resources(struct dma_chan *chan)
+{
+ return 0;
+}
+
+static void mdc_free_chan_resources(struct dma_chan *chan)
+{
+ struct mdc_chan *mchan = to_mdc_chan(chan);
+ struct mdc_dma *mdma = mchan->mdma;
+
+ mdc_terminate_all(chan);
+
+ mdma->soc->disable_chan(mchan);
+}
+
+static irqreturn_t mdc_chan_irq(int irq, void *dev_id)
+{
+ struct mdc_chan *mchan = (struct mdc_chan *)dev_id;
+ struct mdc_tx_desc *mdesc;
+ u32 val, processed, done1, done2;
+ unsigned int i;
+
+ spin_lock(&mchan->vc.lock);
+
+ val = mdc_chan_readl(mchan, MDC_CMDS_PROCESSED);
+ processed = (val >> MDC_CMDS_PROCESSED_CMDS_PROCESSED_SHIFT) &
+ MDC_CMDS_PROCESSED_CMDS_PROCESSED_MASK;
+ /*
+ * CMDS_DONE may have incremented between reading CMDS_PROCESSED
+ * and clearing INT_ACTIVE. Re-read CMDS_PROCESSED to ensure we
+ * didn't miss a command completion.
+ */
+ do {
+ val = mdc_chan_readl(mchan, MDC_CMDS_PROCESSED);
+ done1 = (val >> MDC_CMDS_PROCESSED_CMDS_DONE_SHIFT) &
+ MDC_CMDS_PROCESSED_CMDS_DONE_MASK;
+ val &= ~((MDC_CMDS_PROCESSED_CMDS_PROCESSED_MASK <<
+ MDC_CMDS_PROCESSED_CMDS_PROCESSED_SHIFT) |
+ MDC_CMDS_PROCESSED_INT_ACTIVE);
+ val |= done1 << MDC_CMDS_PROCESSED_CMDS_PROCESSED_SHIFT;
+ mdc_chan_writel(mchan, val, MDC_CMDS_PROCESSED);
+ val = mdc_chan_readl(mchan, MDC_CMDS_PROCESSED);
+ done2 = (val >> MDC_CMDS_PROCESSED_CMDS_DONE_SHIFT) &
+ MDC_CMDS_PROCESSED_CMDS_DONE_MASK;
+ } while (done1 != done2);
+
+ dev_dbg(mdma2dev(mchan->mdma), "IRQ on channel %d\n", mchan->chan_nr);
+
+ mdesc = mchan->desc;
+ if (!mdesc) {
+ dev_warn(mdma2dev(mchan->mdma),
+ "IRQ with no active descriptor on channel %d\n",
+ mchan->chan_nr);
+ goto out;
+ }
+
+ for (i = processed; i != done1;
+ i = (i + 1) % (MDC_CMDS_PROCESSED_CMDS_PROCESSED_MASK + 1)) {
+ /*
+ * The first interrupt in a transfer indicates that the
+ * command list has been loaded, not that a command has
+ * been completed.
+ */
+ if (!mdesc->cmd_loaded) {
+ mdesc->cmd_loaded = true;
+ continue;
+ }
+
+ mdesc->list_cmds_done++;
+ if (mdesc->cyclic) {
+ mdesc->list_cmds_done %= mdesc->list_len;
+ if (mdesc->list_cmds_done % mdesc->list_period_len == 0)
+ vchan_cyclic_callback(&mdesc->vd);
+ } else if (mdesc->list_cmds_done == mdesc->list_len) {
+ mchan->desc = NULL;
+ vchan_cookie_complete(&mdesc->vd);
+ mdc_issue_desc(mchan);
+ break;
+ }
+ }
+out:
+ spin_unlock(&mchan->vc.lock);
+
+ return IRQ_HANDLED;
+}
+
+static struct dma_chan *mdc_of_xlate(struct of_phandle_args *dma_spec,
+ struct of_dma *ofdma)
+{
+ struct mdc_dma *mdma = ofdma->of_dma_data;
+ struct dma_chan *chan;
+
+ if (dma_spec->args_count != 3)
+ return NULL;
+
+ list_for_each_entry(chan, &mdma->dma_dev.channels, device_node) {
+ struct mdc_chan *mchan = to_mdc_chan(chan);
+
+ if (!(dma_spec->args[1] & BIT(mchan->chan_nr)))
+ continue;
+ if (dma_get_slave_channel(chan)) {
+ mchan->periph = dma_spec->args[0];
+ mchan->thread = dma_spec->args[2];
+ return chan;
+ }
+ }
+
+ return NULL;
+}
+
+#define PISTACHIO_CR_PERIPH_DMA_ROUTE(ch) (0x120 + 0x4 * ((ch) / 4))
+#define PISTACHIO_CR_PERIPH_DMA_ROUTE_SHIFT(ch) (8 * ((ch) % 4))
+#define PISTACHIO_CR_PERIPH_DMA_ROUTE_MASK 0x3f
+
+static void pistachio_mdc_enable_chan(struct mdc_chan *mchan)
+{
+ struct mdc_dma *mdma = mchan->mdma;
+
+ regmap_update_bits(mdma->periph_regs,
+ PISTACHIO_CR_PERIPH_DMA_ROUTE(mchan->chan_nr),
+ PISTACHIO_CR_PERIPH_DMA_ROUTE_MASK <<
+ PISTACHIO_CR_PERIPH_DMA_ROUTE_SHIFT(mchan->chan_nr),
+ mchan->periph <<
+ PISTACHIO_CR_PERIPH_DMA_ROUTE_SHIFT(mchan->chan_nr));
+}
+
+static void pistachio_mdc_disable_chan(struct mdc_chan *mchan)
+{
+ struct mdc_dma *mdma = mchan->mdma;
+
+ regmap_update_bits(mdma->periph_regs,
+ PISTACHIO_CR_PERIPH_DMA_ROUTE(mchan->chan_nr),
+ PISTACHIO_CR_PERIPH_DMA_ROUTE_MASK <<
+ PISTACHIO_CR_PERIPH_DMA_ROUTE_SHIFT(mchan->chan_nr),
+ 0);
+}
+
+static const struct mdc_dma_soc_data pistachio_mdc_data = {
+ .enable_chan = pistachio_mdc_enable_chan,
+ .disable_chan = pistachio_mdc_disable_chan,
+};
+
+static const struct of_device_id mdc_dma_of_match[] = {
+ { .compatible = "img,pistachio-mdc-dma", .data = &pistachio_mdc_data, },
+ { },
+};
+MODULE_DEVICE_TABLE(of, mdc_dma_of_match);
+
+static int mdc_dma_probe(struct platform_device *pdev)
+{
+ struct mdc_dma *mdma;
+ struct resource *res;
+ const struct of_device_id *match;
+ unsigned int i;
+ u32 val;
+ int ret;
+
+ mdma = devm_kzalloc(&pdev->dev, sizeof(*mdma), GFP_KERNEL);
+ if (!mdma)
+ return -ENOMEM;
+ platform_set_drvdata(pdev, mdma);
+
+ match = of_match_device(mdc_dma_of_match, &pdev->dev);
+ mdma->soc = match->data;
+
+ res = platform_get_resource(pdev, IORESOURCE_MEM, 0);
+ mdma->regs = devm_ioremap_resource(&pdev->dev, res);
+ if (IS_ERR(mdma->regs))
+ return PTR_ERR(mdma->regs);
+
+ mdma->periph_regs = syscon_regmap_lookup_by_phandle(pdev->dev.of_node,
+ "img,cr-periph");
+ if (IS_ERR(mdma->periph_regs))
+ return PTR_ERR(mdma->periph_regs);
+
+ mdma->clk = devm_clk_get(&pdev->dev, "sys");
+ if (IS_ERR(mdma->clk))
+ return PTR_ERR(mdma->clk);
+
+ ret = clk_prepare_enable(mdma->clk);
+ if (ret)
+ return ret;
+
+ dma_cap_zero(mdma->dma_dev.cap_mask);
+ dma_cap_set(DMA_SLAVE, mdma->dma_dev.cap_mask);
+ dma_cap_set(DMA_PRIVATE, mdma->dma_dev.cap_mask);
+ dma_cap_set(DMA_CYCLIC, mdma->dma_dev.cap_mask);
+ dma_cap_set(DMA_MEMCPY, mdma->dma_dev.cap_mask);
+
+ val = mdc_readl(mdma, MDC_GLOBAL_CONFIG_A);
+ mdma->nr_channels = (val >> MDC_GLOBAL_CONFIG_A_DMA_CONTEXTS_SHIFT) &
+ MDC_GLOBAL_CONFIG_A_DMA_CONTEXTS_MASK;
+ mdma->nr_threads =
+ 1 << ((val >> MDC_GLOBAL_CONFIG_A_THREAD_ID_WIDTH_SHIFT) &
+ MDC_GLOBAL_CONFIG_A_THREAD_ID_WIDTH_MASK);
+ mdma->bus_width =
+ (1 << ((val >> MDC_GLOBAL_CONFIG_A_SYS_DAT_WIDTH_SHIFT) &
+ MDC_GLOBAL_CONFIG_A_SYS_DAT_WIDTH_MASK)) / 8;
+ /*
+ * Although transfer sizes of up to MDC_TRANSFER_SIZE_MASK + 1 bytes
+ * are supported, this makes it possible for the value reported in
+ * MDC_ACTIVE_TRANSFER_SIZE to be ambiguous - an active transfer size
+ * of MDC_TRANSFER_SIZE_MASK may indicate either that 0 bytes or
+ * MDC_TRANSFER_SIZE_MASK + 1 bytes are remaining. To eliminate this
+ * ambiguity, restrict transfer sizes to one bus-width less than the
+ * actual maximum.
+ */
+ mdma->max_xfer_size = MDC_TRANSFER_SIZE_MASK + 1 - mdma->bus_width;
+
+ of_property_read_u32(pdev->dev.of_node, "dma-channels",
+ &mdma->nr_channels);
+ ret = of_property_read_u32(pdev->dev.of_node,
+ "img,max-burst-multiplier",
+ &mdma->max_burst_mult);
+ if (ret)
+ goto disable_clk;
+
+ mdma->dma_dev.dev = &pdev->dev;
+ mdma->dma_dev.device_prep_slave_sg = mdc_prep_slave_sg;
+ mdma->dma_dev.device_prep_dma_cyclic = mdc_prep_dma_cyclic;
+ mdma->dma_dev.device_prep_dma_memcpy = mdc_prep_dma_memcpy;
+ mdma->dma_dev.device_alloc_chan_resources = mdc_alloc_chan_resources;
+ mdma->dma_dev.device_free_chan_resources = mdc_free_chan_resources;
+ mdma->dma_dev.device_tx_status = mdc_tx_status;
+ mdma->dma_dev.device_issue_pending = mdc_issue_pending;
+ mdma->dma_dev.device_terminate_all = mdc_terminate_all;
+ mdma->dma_dev.device_config = mdc_slave_config;
+
+ mdma->dma_dev.directions = BIT(DMA_DEV_TO_MEM) | BIT(DMA_MEM_TO_DEV);
+ mdma->dma_dev.residue_granularity = DMA_RESIDUE_GRANULARITY_BURST;
+ for (i = 1; i <= mdma->bus_width; i <<= 1) {
+ mdma->dma_dev.src_addr_widths |= BIT(i);
+ mdma->dma_dev.dst_addr_widths |= BIT(i);
+ }
+
+ INIT_LIST_HEAD(&mdma->dma_dev.channels);
+ for (i = 0; i < mdma->nr_channels; i++) {
+ struct mdc_chan *mchan = &mdma->channels[i];
+
+ mchan->mdma = mdma;
+ mchan->chan_nr = i;
+ mchan->irq = platform_get_irq(pdev, i);
+ if (mchan->irq < 0) {
+ ret = mchan->irq;
+ goto disable_clk;
+ }
+ ret = devm_request_irq(&pdev->dev, mchan->irq, mdc_chan_irq,
+ IRQ_TYPE_LEVEL_HIGH,
+ dev_name(&pdev->dev), mchan);
+ if (ret < 0)
+ goto disable_clk;
+
+ mchan->vc.desc_free = mdc_desc_free;
+ vchan_init(&mchan->vc, &mdma->dma_dev);
+ }
+
+ mdma->desc_pool = dmam_pool_create(dev_name(&pdev->dev), &pdev->dev,
+ sizeof(struct mdc_hw_list_desc),
+ 4, 0);
+ if (!mdma->desc_pool) {
+ ret = -ENOMEM;
+ goto disable_clk;
+ }
+
+ ret = dma_async_device_register(&mdma->dma_dev);
+ if (ret)
+ goto disable_clk;
+
+ ret = of_dma_controller_register(pdev->dev.of_node, mdc_of_xlate, mdma);
+ if (ret)
+ goto unregister;
+
+ dev_info(&pdev->dev, "MDC with %u channels and %u threads\n",
+ mdma->nr_channels, mdma->nr_threads);
+
+ return 0;
+
+unregister:
+ dma_async_device_unregister(&mdma->dma_dev);
+disable_clk:
+ clk_disable_unprepare(mdma->clk);
+ return ret;
+}
+
+static int mdc_dma_remove(struct platform_device *pdev)
+{
+ struct mdc_dma *mdma = platform_get_drvdata(pdev);
+ struct mdc_chan *mchan, *next;
+
+ of_dma_controller_free(pdev->dev.of_node);
+ dma_async_device_unregister(&mdma->dma_dev);
+
+ list_for_each_entry_safe(mchan, next, &mdma->dma_dev.channels,
+ vc.chan.device_node) {
+ list_del(&mchan->vc.chan.device_node);
+
+ synchronize_irq(mchan->irq);
+ devm_free_irq(&pdev->dev, mchan->irq, mchan);
+
+ tasklet_kill(&mchan->vc.task);
+ }
+
+ clk_disable_unprepare(mdma->clk);
+
+ return 0;
+}
+
+static struct platform_driver mdc_dma_driver = {
+ .driver = {
+ .name = "img-mdc-dma",
+ .of_match_table = of_match_ptr(mdc_dma_of_match),
+ },
+ .probe = mdc_dma_probe,
+ .remove = mdc_dma_remove,
+};
+module_platform_driver(mdc_dma_driver);
+
+MODULE_DESCRIPTION("IMG Multi-threaded DMA Controller (MDC) driver");
+MODULE_AUTHOR("Andrew Bresticker <[email protected]>");
+MODULE_LICENSE("GPL v2");
--
2.2.0.rc0.207.ga3a616c
On Thu, Dec 11, 2014 at 02:59:17PM -0800, Andrew Bresticker wrote:
> Add support for the IMG Multi-threaded DMA Controller (MDC) found on
> certain IMG SoCs. Currently this driver supports the variant present
> on the MIPS-based Pistachio SoC.a
Overall looks okay. I also need some review by DT folks on the bindings
> +static void mdc_list_desc_config(struct mdc_chan *mchan,
> + struct mdc_hw_list_desc *ldesc,
> + enum dma_transfer_direction dir,
> + dma_addr_t src, dma_addr_t dst, size_t len)
> +{
> + struct mdc_dma *mdma = mchan->mdma;
> + unsigned int max_burst, burst_size;
> +
> + ldesc->gen_conf = MDC_GENERAL_CONFIG_IEN | MDC_GENERAL_CONFIG_LIST_IEN |
> + MDC_GENERAL_CONFIG_LEVEL_INT | MDC_GENERAL_CONFIG_PHYSICAL_W |
> + MDC_GENERAL_CONFIG_PHYSICAL_R;
> + ldesc->readport_conf =
> + (mchan->thread << MDC_READ_PORT_CONFIG_STHREAD_SHIFT) |
> + (mchan->thread << MDC_READ_PORT_CONFIG_RTHREAD_SHIFT) |
> + (mchan->thread << MDC_READ_PORT_CONFIG_WTHREAD_SHIFT);
> + ldesc->read_addr = src;
> + ldesc->write_addr = dst;
> + ldesc->xfer_size = len - 1;
> + ldesc->node_addr = 0;
> + ldesc->cmds_done = 0;
> + ldesc->ctrl_status = MDC_CONTROL_AND_STATUS_LIST_EN |
> + MDC_CONTROL_AND_STATUS_EN;
> + ldesc->next_desc = NULL;
> +
> + if (IS_ALIGNED(dst, mdma->bus_width) &&
> + IS_ALIGNED(src, mdma->bus_width))
> + max_burst = mdma->bus_width * mdma->max_burst_mult;
> + else
> + max_burst = mdma->bus_width * (mdma->max_burst_mult - 1);
> +
> + if (dir == DMA_MEM_TO_DEV) {
> + ldesc->gen_conf |= MDC_GENERAL_CONFIG_INC_R;
> + ldesc->readport_conf |= MDC_READ_PORT_CONFIG_DREQ_ENABLE;
> + mdc_set_read_width(ldesc, mdma->bus_width);
> + mdc_set_write_width(ldesc, mchan->config.dst_addr_width);
> + burst_size = min(max_burst, mchan->config.dst_maxburst *
> + mchan->config.dst_addr_width);
why is this calculation done for burst size? Shouldn't we take the
config.dst_maxburst value configured by client?
> +static struct dma_async_tx_descriptor *mdc_prep_slave_sg(
> + struct dma_chan *chan, struct scatterlist *sgl,
> + unsigned int sg_len, enum dma_transfer_direction dir,
> + unsigned long flags, void *context)
> +{
> + struct mdc_chan *mchan = to_mdc_chan(chan);
> + struct mdc_dma *mdma = mchan->mdma;
> + struct mdc_tx_desc *mdesc;
> + struct scatterlist *sg;
> + struct mdc_hw_list_desc *curr, *prev = NULL;
> + dma_addr_t curr_phys, prev_phys;
> + unsigned int i;
> +
> + if (!sgl)
> + return NULL;
> +
> + if (!is_slave_direction(dir))
> + return NULL;
> +
> + if (mdc_check_slave_width(mchan, dir) < 0)
> + return NULL;
> +
> + mdesc = kzalloc(sizeof(*mdesc), GFP_NOWAIT);
> + if (!mdesc)
> + return NULL;
> + mdesc->chan = mchan;
> +
> + for_each_sg(sgl, sg, sg_len, i) {
> + dma_addr_t buf = sg_dma_address(sg);
> + size_t buf_len = sg_dma_len(sg);
> +
> + while (buf_len > 0) {
> + size_t xfer_size;
> +
> + curr = dma_pool_alloc(mdma->desc_pool, GFP_NOWAIT,
> + &curr_phys);
> + if (!curr)
> + goto free_desc;
> +
> + if (!prev) {
> + mdesc->list_phys = curr_phys;
> + mdesc->list = curr;
> + } else {
> + prev->node_addr = curr_phys;
> + prev->next_desc = curr;
> + }
> +
> + xfer_size = min_t(size_t, mdma->max_xfer_size,
> + buf_len);
> +
> + if (dir == DMA_MEM_TO_DEV) {
> + mdc_list_desc_config(mchan, curr, dir, buf,
> + mchan->config.dst_addr,
> + xfer_size);
> + } else {
> + mdc_list_desc_config(mchan, curr, dir,
> + mchan->config.src_addr,
> + buf, xfer_size);
> + }
> +
> + prev = curr;
> + prev_phys = curr_phys;
> +
> + mdesc->list_len++;
> + mdesc->list_xfer_size += xfer_size;
> + buf += xfer_size;
> + buf_len -= xfer_size;
i see this pattern is repeat in all the .prepare calls, can we make it bit
generic and use that in the these three calls..
> + dma_cap_zero(mdma->dma_dev.cap_mask);
> + dma_cap_set(DMA_SLAVE, mdma->dma_dev.cap_mask);
> + dma_cap_set(DMA_PRIVATE, mdma->dma_dev.cap_mask);
> + dma_cap_set(DMA_CYCLIC, mdma->dma_dev.cap_mask);
and you dont seen to support pause/resume, though not a blocker but is it
not supported in HW or driver doesn't?
--
~Vinod
On Tue, Dec 23, 2014 at 9:22 PM, Vinod Koul <[email protected]> wrote:
> On Thu, Dec 11, 2014 at 02:59:17PM -0800, Andrew Bresticker wrote:
>> Add support for the IMG Multi-threaded DMA Controller (MDC) found on
>> certain IMG SoCs. Currently this driver supports the variant present
>> on the MIPS-based Pistachio SoC.a
> Overall looks okay. I also need some review by DT folks on the bindings
Arnd has ack'ed the DT bindings, do you need someone else to review
them as well?
>> +static void mdc_list_desc_config(struct mdc_chan *mchan,
>> + struct mdc_hw_list_desc *ldesc,
>> + enum dma_transfer_direction dir,
>> + dma_addr_t src, dma_addr_t dst, size_t len)
>> +{
>> + struct mdc_dma *mdma = mchan->mdma;
>> + unsigned int max_burst, burst_size;
>> +
>> + ldesc->gen_conf = MDC_GENERAL_CONFIG_IEN | MDC_GENERAL_CONFIG_LIST_IEN |
>> + MDC_GENERAL_CONFIG_LEVEL_INT | MDC_GENERAL_CONFIG_PHYSICAL_W |
>> + MDC_GENERAL_CONFIG_PHYSICAL_R;
>> + ldesc->readport_conf =
>> + (mchan->thread << MDC_READ_PORT_CONFIG_STHREAD_SHIFT) |
>> + (mchan->thread << MDC_READ_PORT_CONFIG_RTHREAD_SHIFT) |
>> + (mchan->thread << MDC_READ_PORT_CONFIG_WTHREAD_SHIFT);
>> + ldesc->read_addr = src;
>> + ldesc->write_addr = dst;
>> + ldesc->xfer_size = len - 1;
>> + ldesc->node_addr = 0;
>> + ldesc->cmds_done = 0;
>> + ldesc->ctrl_status = MDC_CONTROL_AND_STATUS_LIST_EN |
>> + MDC_CONTROL_AND_STATUS_EN;
>> + ldesc->next_desc = NULL;
>> +
>> + if (IS_ALIGNED(dst, mdma->bus_width) &&
>> + IS_ALIGNED(src, mdma->bus_width))
>> + max_burst = mdma->bus_width * mdma->max_burst_mult;
>> + else
>> + max_burst = mdma->bus_width * (mdma->max_burst_mult - 1);
>> +
>> + if (dir == DMA_MEM_TO_DEV) {
>> + ldesc->gen_conf |= MDC_GENERAL_CONFIG_INC_R;
>> + ldesc->readport_conf |= MDC_READ_PORT_CONFIG_DREQ_ENABLE;
>> + mdc_set_read_width(ldesc, mdma->bus_width);
>> + mdc_set_write_width(ldesc, mchan->config.dst_addr_width);
>> + burst_size = min(max_burst, mchan->config.dst_maxburst *
>> + mchan->config.dst_addr_width);
> why is this calculation done for burst size? Shouldn't we take the
> config.dst_maxburst value configured by client?
It's possible that the client could select a burst size that is too
large (i.e. it exceeds the max_burst calculated above), so we cap it
here to max_burst.
>> +static struct dma_async_tx_descriptor *mdc_prep_slave_sg(
>> + struct dma_chan *chan, struct scatterlist *sgl,
>> + unsigned int sg_len, enum dma_transfer_direction dir,
>> + unsigned long flags, void *context)
>> +{
>> + struct mdc_chan *mchan = to_mdc_chan(chan);
>> + struct mdc_dma *mdma = mchan->mdma;
>> + struct mdc_tx_desc *mdesc;
>> + struct scatterlist *sg;
>> + struct mdc_hw_list_desc *curr, *prev = NULL;
>> + dma_addr_t curr_phys, prev_phys;
>> + unsigned int i;
>> +
>> + if (!sgl)
>> + return NULL;
>> +
>> + if (!is_slave_direction(dir))
>> + return NULL;
>> +
>> + if (mdc_check_slave_width(mchan, dir) < 0)
>> + return NULL;
>> +
>> + mdesc = kzalloc(sizeof(*mdesc), GFP_NOWAIT);
>> + if (!mdesc)
>> + return NULL;
>> + mdesc->chan = mchan;
>> +
>> + for_each_sg(sgl, sg, sg_len, i) {
>> + dma_addr_t buf = sg_dma_address(sg);
>> + size_t buf_len = sg_dma_len(sg);
>> +
>> + while (buf_len > 0) {
>> + size_t xfer_size;
>> +
>> + curr = dma_pool_alloc(mdma->desc_pool, GFP_NOWAIT,
>> + &curr_phys);
>> + if (!curr)
>> + goto free_desc;
>> +
>> + if (!prev) {
>> + mdesc->list_phys = curr_phys;
>> + mdesc->list = curr;
>> + } else {
>> + prev->node_addr = curr_phys;
>> + prev->next_desc = curr;
>> + }
>> +
>> + xfer_size = min_t(size_t, mdma->max_xfer_size,
>> + buf_len);
>> +
>> + if (dir == DMA_MEM_TO_DEV) {
>> + mdc_list_desc_config(mchan, curr, dir, buf,
>> + mchan->config.dst_addr,
>> + xfer_size);
>> + } else {
>> + mdc_list_desc_config(mchan, curr, dir,
>> + mchan->config.src_addr,
>> + buf, xfer_size);
>> + }
>> +
>> + prev = curr;
>> + prev_phys = curr_phys;
>> +
>> + mdesc->list_len++;
>> + mdesc->list_xfer_size += xfer_size;
>> + buf += xfer_size;
>> + buf_len -= xfer_size;
> i see this pattern is repeat in all the .prepare calls, can we make it bit
> generic and use that in the these three calls..
I've pulled out almost all of the common stuff into
mdc_list_desc_config(). I suppose the list manipulation could be
moved there as well, but the rest is all slightly different in each
case.
>> + dma_cap_zero(mdma->dma_dev.cap_mask);
>> + dma_cap_set(DMA_SLAVE, mdma->dma_dev.cap_mask);
>> + dma_cap_set(DMA_PRIVATE, mdma->dma_dev.cap_mask);
>> + dma_cap_set(DMA_CYCLIC, mdma->dma_dev.cap_mask);
> and you dont seen to support pause/resume, though not a blocker but is it
> not supported in HW or driver doesn't?
Someone from IMG can correct me if I'm wrong, but I don't think the HW
supports it.
> On 24 Dec 2014, at 17:21, Andrew Bresticker <[email protected]> wrote:
>
>> On Tue, Dec 23, 2014 at 9:22 PM, Vinod Koul <[email protected]> wrote:
>>> On Thu, Dec 11, 2014 at 02:59:17PM -0800, Andrew Bresticker wrote:
>>> Add support for the IMG Multi-threaded DMA Controller (MDC) found on
>>> certain IMG SoCs. Currently this driver supports the variant present
>>> on the MIPS-based Pistachio SoC.a
>
>> Overall looks okay. I also need some review by DT folks on the bindings
>
> Arnd has ack'ed the DT bindings, do you need someone else to review
> them as well?
>
>>> +static void mdc_list_desc_config(struct mdc_chan *mchan,
>>> + struct mdc_hw_list_desc *ldesc,
>>> + enum dma_transfer_direction dir,
>>> + dma_addr_t src, dma_addr_t dst, size_t len)
>>> +{
>>> + struct mdc_dma *mdma = mchan->mdma;
>>> + unsigned int max_burst, burst_size;
>>> +
>>> + ldesc->gen_conf = MDC_GENERAL_CONFIG_IEN | MDC_GENERAL_CONFIG_LIST_IEN |
>>> + MDC_GENERAL_CONFIG_LEVEL_INT | MDC_GENERAL_CONFIG_PHYSICAL_W |
>>> + MDC_GENERAL_CONFIG_PHYSICAL_R;
>>> + ldesc->readport_conf =
>>> + (mchan->thread << MDC_READ_PORT_CONFIG_STHREAD_SHIFT) |
>>> + (mchan->thread << MDC_READ_PORT_CONFIG_RTHREAD_SHIFT) |
>>> + (mchan->thread << MDC_READ_PORT_CONFIG_WTHREAD_SHIFT);
>>> + ldesc->read_addr = src;
>>> + ldesc->write_addr = dst;
>>> + ldesc->xfer_size = len - 1;
>>> + ldesc->node_addr = 0;
>>> + ldesc->cmds_done = 0;
>>> + ldesc->ctrl_status = MDC_CONTROL_AND_STATUS_LIST_EN |
>>> + MDC_CONTROL_AND_STATUS_EN;
>>> + ldesc->next_desc = NULL;
>>> +
>>> + if (IS_ALIGNED(dst, mdma->bus_width) &&
>>> + IS_ALIGNED(src, mdma->bus_width))
>>> + max_burst = mdma->bus_width * mdma->max_burst_mult;
>>> + else
>>> + max_burst = mdma->bus_width * (mdma->max_burst_mult - 1);
>>> +
>>> + if (dir == DMA_MEM_TO_DEV) {
>>> + ldesc->gen_conf |= MDC_GENERAL_CONFIG_INC_R;
>>> + ldesc->readport_conf |= MDC_READ_PORT_CONFIG_DREQ_ENABLE;
>>> + mdc_set_read_width(ldesc, mdma->bus_width);
>>> + mdc_set_write_width(ldesc, mchan->config.dst_addr_width);
>>> + burst_size = min(max_burst, mchan->config.dst_maxburst *
>>> + mchan->config.dst_addr_width);
>
>> why is this calculation done for burst size? Shouldn't we take the
>> config.dst_maxburst value configured by client?
>
> It's possible that the client could select a burst size that is too
> large (i.e. it exceeds the max_burst calculated above), so we cap it
> here to max_burst.
>
>>> +static struct dma_async_tx_descriptor *mdc_prep_slave_sg(
>>> + struct dma_chan *chan, struct scatterlist *sgl,
>>> + unsigned int sg_len, enum dma_transfer_direction dir,
>>> + unsigned long flags, void *context)
>>> +{
>>> + struct mdc_chan *mchan = to_mdc_chan(chan);
>>> + struct mdc_dma *mdma = mchan->mdma;
>>> + struct mdc_tx_desc *mdesc;
>>> + struct scatterlist *sg;
>>> + struct mdc_hw_list_desc *curr, *prev = NULL;
>>> + dma_addr_t curr_phys, prev_phys;
>>> + unsigned int i;
>>> +
>>> + if (!sgl)
>>> + return NULL;
>>> +
>>> + if (!is_slave_direction(dir))
>>> + return NULL;
>>> +
>>> + if (mdc_check_slave_width(mchan, dir) < 0)
>>> + return NULL;
>>> +
>>> + mdesc = kzalloc(sizeof(*mdesc), GFP_NOWAIT);
>>> + if (!mdesc)
>>> + return NULL;
>>> + mdesc->chan = mchan;
>>> +
>>> + for_each_sg(sgl, sg, sg_len, i) {
>>> + dma_addr_t buf = sg_dma_address(sg);
>>> + size_t buf_len = sg_dma_len(sg);
>>> +
>>> + while (buf_len > 0) {
>>> + size_t xfer_size;
>>> +
>>> + curr = dma_pool_alloc(mdma->desc_pool, GFP_NOWAIT,
>>> + &curr_phys);
>>> + if (!curr)
>>> + goto free_desc;
>>> +
>>> + if (!prev) {
>>> + mdesc->list_phys = curr_phys;
>>> + mdesc->list = curr;
>>> + } else {
>>> + prev->node_addr = curr_phys;
>>> + prev->next_desc = curr;
>>> + }
>>> +
>>> + xfer_size = min_t(size_t, mdma->max_xfer_size,
>>> + buf_len);
>>> +
>>> + if (dir == DMA_MEM_TO_DEV) {
>>> + mdc_list_desc_config(mchan, curr, dir, buf,
>>> + mchan->config.dst_addr,
>>> + xfer_size);
>>> + } else {
>>> + mdc_list_desc_config(mchan, curr, dir,
>>> + mchan->config.src_addr,
>>> + buf, xfer_size);
>>> + }
>>> +
>>> + prev = curr;
>>> + prev_phys = curr_phys;
>>> +
>>> + mdesc->list_len++;
>>> + mdesc->list_xfer_size += xfer_size;
>>> + buf += xfer_size;
>>> + buf_len -= xfer_size;
>
>> i see this pattern is repeat in all the .prepare calls, can we make it bit
>> generic and use that in the these three calls..
>
> I've pulled out almost all of the common stuff into
> mdc_list_desc_config(). I suppose the list manipulation could be
> moved there as well, but the rest is all slightly different in each
> case.
>
>>> + dma_cap_zero(mdma->dma_dev.cap_mask);
>>> + dma_cap_set(DMA_SLAVE, mdma->dma_dev.cap_mask);
>>> + dma_cap_set(DMA_PRIVATE, mdma->dma_dev.cap_mask);
>>> + dma_cap_set(DMA_CYCLIC, mdma->dma_dev.cap_mask);
>
>> and you dont seen to support pause/resume, though not a blocker but is it
>> not supported in HW or driver doesn't?
>
> Someone from IMG can correct me if I'm wrong, but I don't think the HW
> supports it.
Pause / resume is not supported by the MDC hardware.
James. -
On Thu, Dec 11, 2014 at 02:59:15PM -0800, Andrew Bresticker wrote:
> This series adds support for the IMG Multi-threaded DMA Controller (MDC)
> which is found on IMG SoCs. Currently this driver only supports the
> variant found on the MIPS-based Pistachio SoC.
>
> The MDC supports slave and memory-to-memory transfers on up to 32 channels.
> Requests from channels are handled by MDC threads. Thread assignments are
> per-channel and are specified in the device-tree.
>
> Tested on a platform based on the Pistachio SoC with additional patches.
> Support for Pistachio will be submitted later. Slave DMA was tested
> with the recently submitted IMG SPFI driver [1] and memory-to-memory
> DMA was tested with the dmatest module.
>
Applied, thanks
--
~Vinod