diff options
author | Linus Walleij <linus.walleij@stericsson.com> | 2010-03-30 15:33:42 +0200 |
---|---|---|
committer | Dan Williams <dan.j.williams@intel.com> | 2010-04-14 14:49:20 -0700 |
commit | 8d318a50b3d72e3daf94131f91e1ab799a8d5ad4 (patch) | |
tree | ae36452931d2e836f725b3f91eebd7f4d9e27589 | |
parent | 6a3cd3ea48584d14f60dce0b3c4e9e4428beb0fe (diff) |
DMAENGINE: Support for ST-Ericssons DMA40 block v3
This is a straightforward driver for the ST-Ericsson DMA40 DMA
controller found in U8500, implemented akin to the existing
COH 901 318 driver.
Signed-off-by: Linus Walleij <linus.walleij@stericsson.com>
Acked-by: Srinidh Kasagar <srinidhi.kasagar@stericsson.com>
Cc: STEricsson_nomadik_linux@list.st.com
Cc: Alessandro Rubini <rubini@unipv.it>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
-rw-r--r-- | arch/arm/plat-nomadik/include/plat/ste_dma40.h | 239 | ||||
-rw-r--r-- | drivers/dma/Kconfig | 7 | ||||
-rw-r--r-- | drivers/dma/Makefile | 1 | ||||
-rw-r--r-- | drivers/dma/ste_dma40.c | 2596 | ||||
-rw-r--r-- | drivers/dma/ste_dma40_ll.c | 454 | ||||
-rw-r--r-- | drivers/dma/ste_dma40_ll.h | 354 |
6 files changed, 3651 insertions, 0 deletions
diff --git a/arch/arm/plat-nomadik/include/plat/ste_dma40.h b/arch/arm/plat-nomadik/include/plat/ste_dma40.h new file mode 100644 index 00000000000..4d12ea4ca36 --- /dev/null +++ b/arch/arm/plat-nomadik/include/plat/ste_dma40.h @@ -0,0 +1,239 @@ +/* + * arch/arm/plat-nomadik/include/plat/ste_dma40.h + * + * Copyright (C) ST-Ericsson 2007-2010 + * License terms: GNU General Public License (GPL) version 2 + * Author: Per Friden <per.friden@stericsson.com> + * Author: Jonas Aaberg <jonas.aberg@stericsson.com> + */ + + +#ifndef STE_DMA40_H +#define STE_DMA40_H + +#include <linux/dmaengine.h> +#include <linux/workqueue.h> +#include <linux/interrupt.h> +#include <linux/dmaengine.h> + +/* dev types for memcpy */ +#define STEDMA40_DEV_DST_MEMORY (-1) +#define STEDMA40_DEV_SRC_MEMORY (-1) + +/* + * Description of bitfields of channel_type variable is available in + * the info structure. + */ + +/* Priority */ +#define STEDMA40_INFO_PRIO_TYPE_POS 2 +#define STEDMA40_HIGH_PRIORITY_CHANNEL (0x1 << STEDMA40_INFO_PRIO_TYPE_POS) +#define STEDMA40_LOW_PRIORITY_CHANNEL (0x2 << STEDMA40_INFO_PRIO_TYPE_POS) + +/* Mode */ +#define STEDMA40_INFO_CH_MODE_TYPE_POS 6 +#define STEDMA40_CHANNEL_IN_PHY_MODE (0x1 << STEDMA40_INFO_CH_MODE_TYPE_POS) +#define STEDMA40_CHANNEL_IN_LOG_MODE (0x2 << STEDMA40_INFO_CH_MODE_TYPE_POS) +#define STEDMA40_CHANNEL_IN_OPER_MODE (0x3 << STEDMA40_INFO_CH_MODE_TYPE_POS) + +/* Mode options */ +#define STEDMA40_INFO_CH_MODE_OPT_POS 8 +#define STEDMA40_PCHAN_BASIC_MODE (0x1 << STEDMA40_INFO_CH_MODE_OPT_POS) +#define STEDMA40_PCHAN_MODULO_MODE (0x2 << STEDMA40_INFO_CH_MODE_OPT_POS) +#define STEDMA40_PCHAN_DOUBLE_DST_MODE (0x3 << STEDMA40_INFO_CH_MODE_OPT_POS) +#define STEDMA40_LCHAN_SRC_PHY_DST_LOG (0x1 << STEDMA40_INFO_CH_MODE_OPT_POS) +#define STEDMA40_LCHAN_SRC_LOG_DST_PHS (0x2 << STEDMA40_INFO_CH_MODE_OPT_POS) +#define STEDMA40_LCHAN_SRC_LOG_DST_LOG (0x3 << STEDMA40_INFO_CH_MODE_OPT_POS) + +/* Interrupt */ +#define STEDMA40_INFO_TIM_POS 10 +#define STEDMA40_NO_TIM_FOR_LINK (0x0 << STEDMA40_INFO_TIM_POS) +#define STEDMA40_TIM_FOR_LINK (0x1 << STEDMA40_INFO_TIM_POS) + +/* End of channel_type configuration */ + +#define STEDMA40_ESIZE_8_BIT 0x0 +#define STEDMA40_ESIZE_16_BIT 0x1 +#define STEDMA40_ESIZE_32_BIT 0x2 +#define STEDMA40_ESIZE_64_BIT 0x3 + +/* The value 4 indicates that PEN-reg shall be set to 0 */ +#define STEDMA40_PSIZE_PHY_1 0x4 +#define STEDMA40_PSIZE_PHY_2 0x0 +#define STEDMA40_PSIZE_PHY_4 0x1 +#define STEDMA40_PSIZE_PHY_8 0x2 +#define STEDMA40_PSIZE_PHY_16 0x3 + +/* + * The number of elements differ in logical and + * physical mode + */ +#define STEDMA40_PSIZE_LOG_1 STEDMA40_PSIZE_PHY_2 +#define STEDMA40_PSIZE_LOG_4 STEDMA40_PSIZE_PHY_4 +#define STEDMA40_PSIZE_LOG_8 STEDMA40_PSIZE_PHY_8 +#define STEDMA40_PSIZE_LOG_16 STEDMA40_PSIZE_PHY_16 + +enum stedma40_flow_ctrl { + STEDMA40_NO_FLOW_CTRL, + STEDMA40_FLOW_CTRL, +}; + +enum stedma40_endianess { + STEDMA40_LITTLE_ENDIAN, + STEDMA40_BIG_ENDIAN +}; + +enum stedma40_periph_data_width { + STEDMA40_BYTE_WIDTH = STEDMA40_ESIZE_8_BIT, + STEDMA40_HALFWORD_WIDTH = STEDMA40_ESIZE_16_BIT, + STEDMA40_WORD_WIDTH = STEDMA40_ESIZE_32_BIT, + STEDMA40_DOUBLEWORD_WIDTH = STEDMA40_ESIZE_64_BIT +}; + +struct stedma40_half_channel_info { + enum stedma40_endianess endianess; + enum stedma40_periph_data_width data_width; + int psize; + enum stedma40_flow_ctrl flow_ctrl; +}; + +enum stedma40_xfer_dir { + STEDMA40_MEM_TO_MEM, + STEDMA40_MEM_TO_PERIPH, + STEDMA40_PERIPH_TO_MEM, + STEDMA40_PERIPH_TO_PERIPH +}; + + +/** + * struct stedma40_chan_cfg - Structure to be filled by client drivers. + * + * @dir: MEM 2 MEM, PERIPH 2 MEM , MEM 2 PERIPH, PERIPH 2 PERIPH + * @channel_type: priority, mode, mode options and interrupt configuration. + * @src_dev_type: Src device type + * @dst_dev_type: Dst device type + * @src_info: Parameters for dst half channel + * @dst_info: Parameters for dst half channel + * @pre_transfer_data: Data to be passed on to the pre_transfer() function. + * @pre_transfer: Callback used if needed before preparation of transfer. + * Only called if device is set. size of bytes to transfer + * (in case of multiple element transfer size is size of the first element). + * + * + * This structure has to be filled by the client drivers. + * It is recommended to do all dma configurations for clients in the machine. + * + */ +struct stedma40_chan_cfg { + enum stedma40_xfer_dir dir; + unsigned int channel_type; + int src_dev_type; + int dst_dev_type; + struct stedma40_half_channel_info src_info; + struct stedma40_half_channel_info dst_info; + void *pre_transfer_data; + int (*pre_transfer) (struct dma_chan *chan, + void *data, + int size); +}; + +/** + * struct stedma40_platform_data - Configuration struct for the dma device. + * + * @dev_len: length of dev_tx and dev_rx + * @dev_tx: mapping between destination event line and io address + * @dev_rx: mapping between source event line and io address + * @memcpy: list of memcpy event lines + * @memcpy_len: length of memcpy + * @memcpy_conf_phy: default configuration of physical channel memcpy + * @memcpy_conf_log: default configuration of logical channel memcpy + * @llis_per_log: number of max linked list items per logical channel + * + */ +struct stedma40_platform_data { + u32 dev_len; + const dma_addr_t *dev_tx; + const dma_addr_t *dev_rx; + int *memcpy; + u32 memcpy_len; + struct stedma40_chan_cfg *memcpy_conf_phy; + struct stedma40_chan_cfg *memcpy_conf_log; + unsigned int llis_per_log; +}; + +/** + * setdma40_set_psize() - Used for changing the package size of an + * already configured dma channel. + * + * @chan: dmaengine handle + * @src_psize: new package side for src. (STEDMA40_PSIZE*) + * @src_psize: new package side for dst. (STEDMA40_PSIZE*) + * + * returns 0 on ok, otherwise negative error number. + */ +int stedma40_set_psize(struct dma_chan *chan, + int src_psize, + int dst_psize); + +/** + * stedma40_filter() - Provides stedma40_chan_cfg to the + * ste_dma40 dma driver via the dmaengine framework. + * does some checking of what's provided. + * + * Never directly called by client. It used by dmaengine. + * @chan: dmaengine handle. + * @data: Must be of type: struct stedma40_chan_cfg and is + * the configuration of the framework. + * + * + */ + +bool stedma40_filter(struct dma_chan *chan, void *data); + +/** + * stedma40_memcpy_sg() - extension of the dma framework, memcpy to/from + * scattergatter lists. + * + * @chan: dmaengine handle + * @sgl_dst: Destination scatter list + * @sgl_src: Source scatter list + * @sgl_len: The length of each scatterlist. Both lists must be of equal length + * and each element must match the corresponding element in the other scatter + * list. + * @flags: is actually enum dma_ctrl_flags. See dmaengine.h + */ + +struct dma_async_tx_descriptor *stedma40_memcpy_sg(struct dma_chan *chan, + struct scatterlist *sgl_dst, + struct scatterlist *sgl_src, + unsigned int sgl_len, + unsigned long flags); + +/** + * stedma40_slave_mem() - Transfers a raw data buffer to or from a slave + * (=device) + * + * @chan: dmaengine handle + * @addr: source or destination physicall address. + * @size: bytes to transfer + * @direction: direction of transfer + * @flags: is actually enum dma_ctrl_flags. See dmaengine.h + */ + +static inline struct +dma_async_tx_descriptor *stedma40_slave_mem(struct dma_chan *chan, + dma_addr_t addr, + unsigned int size, + enum dma_data_direction direction, + unsigned long flags) +{ + struct scatterlist sg; + sg_init_table(&sg, 1); + sg.dma_address = addr; + sg.length = size; + + return chan->device->device_prep_slave_sg(chan, &sg, 1, + direction, flags); +} + +#endif diff --git a/drivers/dma/Kconfig b/drivers/dma/Kconfig index a2fcb2ead89..1b8877922fb 100644 --- a/drivers/dma/Kconfig +++ b/drivers/dma/Kconfig @@ -141,6 +141,13 @@ config COH901318 help Enable support for ST-Ericsson COH 901 318 DMA. +config STE_DMA40 + bool "ST-Ericsson DMA40 support" + depends on ARCH_U8500 + select DMA_ENGINE + help + Support for ST-Ericsson DMA40 controller + config AMCC_PPC440SPE_ADMA tristate "AMCC PPC440SPe ADMA support" depends on 440SPe || 440SP diff --git a/drivers/dma/Makefile b/drivers/dma/Makefile index 40c627d8f73..20881426c1a 100644 --- a/drivers/dma/Makefile +++ b/drivers/dma/Makefile @@ -21,3 +21,4 @@ obj-$(CONFIG_SH_DMAE) += shdma.o obj-$(CONFIG_COH901318) += coh901318.o coh901318_lli.o obj-$(CONFIG_AMCC_PPC440SPE_ADMA) += ppc4xx/ obj-$(CONFIG_TIMB_DMA) += timb_dma.o +obj-$(CONFIG_STE_DMA40) += ste_dma40.o ste_dma40_ll.o diff --git a/drivers/dma/ste_dma40.c b/drivers/dma/ste_dma40.c new file mode 100644 index 00000000000..e4295a27672 --- /dev/null +++ b/drivers/dma/ste_dma40.c @@ -0,0 +1,2596 @@ +/* + * driver/dma/ste_dma40.c + * + * Copyright (C) ST-Ericsson 2007-2010 + * License terms: GNU General Public License (GPL) version 2 + * Author: Per Friden <per.friden@stericsson.com> + * Author: Jonas Aaberg <jonas.aberg@stericsson.com> + * + */ + +#include <linux/kernel.h> +#include <linux/slab.h> +#include <linux/dmaengine.h> +#include <linux/platform_device.h> +#include <linux/clk.h> +#include <linux/delay.h> + +#include <plat/ste_dma40.h> + +#include "ste_dma40_ll.h" + +#define D40_NAME "dma40" + +#define D40_PHY_CHAN -1 + +/* For masking out/in 2 bit channel positions */ +#define D40_CHAN_POS(chan) (2 * (chan / 2)) +#define D40_CHAN_POS_MASK(chan) (0x3 << D40_CHAN_POS(chan)) + +/* Maximum iterations taken before giving up suspending a channel */ +#define D40_SUSPEND_MAX_IT 500 + +#define D40_ALLOC_FREE (1 << 31) +#define D40_ALLOC_PHY (1 << 30) +#define D40_ALLOC_LOG_FREE 0 + +/* The number of free d40_desc to keep in memory before starting + * to kfree() them */ +#define D40_DESC_CACHE_SIZE 50 + +/* Hardware designer of the block */ +#define D40_PERIPHID2_DESIGNER 0x8 + +/** + * enum 40_command - The different commands and/or statuses. + * + * @D40_DMA_STOP: DMA channel command STOP or status STOPPED, + * @D40_DMA_RUN: The DMA channel is RUNNING of the command RUN. + * @D40_DMA_SUSPEND_REQ: Request the DMA to SUSPEND as soon as possible. + * @D40_DMA_SUSPENDED: The DMA channel is SUSPENDED. + */ +enum d40_command { + D40_DMA_STOP = 0, + D40_DMA_RUN = 1, + D40_DMA_SUSPEND_REQ = 2, + D40_DMA_SUSPENDED = 3 +}; + +/** + * struct d40_lli_pool - Structure for keeping LLIs in memory + * + * @base: Pointer to memory area when the pre_alloc_lli's are not large + * enough, IE bigger than the most common case, 1 dst and 1 src. NULL if + * pre_alloc_lli is used. + * @size: The size in bytes of the memory at base or the size of pre_alloc_lli. + * @pre_alloc_lli: Pre allocated area for the most common case of transfers, + * one buffer to one buffer. + */ +struct d40_lli_pool { + void *base; + int size; + /* Space for dst and src, plus an extra for padding */ + u8 pre_alloc_lli[3 * sizeof(struct d40_phy_lli)]; +}; + +/** + * struct d40_desc - A descriptor is one DMA job. + * + * @lli_phy: LLI settings for physical channel. Both src and dst= + * points into the lli_pool, to base if lli_len > 1 or to pre_alloc_lli if + * lli_len equals one. + * @lli_log: Same as above but for logical channels. + * @lli_pool: The pool with two entries pre-allocated. + * @lli_len: Number of LLI's in lli_pool + * @lli_tcount: Number of LLIs processed in the transfer. When equals lli_len + * then this transfer job is done. + * @txd: DMA engine struct. Used for among other things for communication + * during a transfer. + * @node: List entry. + * @dir: The transfer direction of this job. + * @is_in_client_list: true if the client owns this descriptor. + * + * This descriptor is used for both logical and physical transfers. + */ + +struct d40_desc { + /* LLI physical */ + struct d40_phy_lli_bidir lli_phy; + /* LLI logical */ + struct d40_log_lli_bidir lli_log; + + struct d40_lli_pool lli_pool; + u32 lli_len; + u32 lli_tcount; + + struct dma_async_tx_descriptor txd; + struct list_head node; + + enum dma_data_direction dir; + bool is_in_client_list; +}; + +/** + * struct d40_lcla_pool - LCLA pool settings and data. + * + * @base: The virtual address of LCLA. + * @phy: Physical base address of LCLA. + * @base_size: size of lcla. + * @lock: Lock to protect the content in this struct. + * @alloc_map: Mapping between physical channel and LCLA entries. + * @num_blocks: The number of entries of alloc_map. Equals to the + * number of physical channels. + */ +struct d40_lcla_pool { + void *base; + dma_addr_t phy; + resource_size_t base_size; + spinlock_t lock; + u32 *alloc_map; + int num_blocks; +}; + +/** + * struct d40_phy_res - struct for handling eventlines mapped to physical + * channels. + * + * @lock: A lock protection this entity. + * @num: The physical channel number of this entity. + * @allocated_src: Bit mapped to show which src event line's are mapped to + * this physical channel. Can also be free or physically allocated. + * @allocated_dst: Same as for src but is dst. + * allocated_dst and allocated_src uses the D40_ALLOC* defines as well as + * event line number. Both allocated_src and allocated_dst can not be + * allocated to a physical channel, since the interrupt handler has then + * no way of figure out which one the interrupt belongs to. + */ +struct d40_phy_res { + spinlock_t lock; + int num; + u32 allocated_src; + u32 allocated_dst; +}; + +struct d40_base; + +/** + * struct d40_chan - Struct that describes a channel. + * + * @lock: A spinlock to protect this struct. + * @log_num: The logical number, if any of this channel. + * @completed: Starts with 1, after first interrupt it is set to dma engine's + * current cookie. + * @pending_tx: The number of pending transfers. Used between interrupt handler + * and tasklet. + * @busy: Set to true when transfer is ongoing on this channel. + * @phy_chan: Pointer to physical channel which this instance runs on. + * @chan: DMA engine handle. + * @tasklet: Tasklet that gets scheduled from interrupt context to complete a + * transfer and call client callback. + * @client: Cliented owned descriptor list. + * @active: Active descriptor. + * @queue: Queued jobs. + * @free: List of free descripts, ready to be reused. + * @free_len: Number of descriptors in the free list. + * @dma_cfg: The client configuration of this dma channel. + * @base: Pointer to the device instance struct. + * @src_def_cfg: Default cfg register setting for src. + * @dst_def_cfg: Default cfg register setting for dst. + * @log_def: Default logical channel settings. + * @lcla: Space for one dst src pair for logical channel transfers. + * @lcpa: Pointer to dst and src lcpa settings. + * + * This struct can either "be" a logical or a physical channel. + */ +struct d40_chan { + spinlock_t lock; + int log_num; + /* ID of the most recent completed transfer */ + int completed; + int pending_tx; + bool busy; + struct d40_phy_res *phy_chan; + struct dma_chan chan; + struct tasklet_struct tasklet; + struct list_head client; + struct list_head active; + struct list_head queue; + struct list_head free; + int free_len; + struct stedma40_chan_cfg dma_cfg; + struct d40_base *base; + /* Default register configurations */ + u32 src_def_cfg; + u32 dst_def_cfg; + struct d40_def_lcsp log_def; + struct d40_lcla_elem lcla; + struct d40_log_lli_full *lcpa; +}; + +/** + * struct d40_base - The big global struct, one for each probe'd instance. + * + * @interrupt_lock: Lock used to make sure one interrupt is handle a time. + * @execmd_lock: Lock for execute command usage since several channels share + * the same physical register. + * @dev: The device structure. + * @virtbase: The virtual base address of the DMA's register. + * @clk: Pointer to the DMA clock structure. + * @phy_start: Physical memory start of the DMA registers. + * @phy_size: Size of the DMA register map. + * @irq: The IRQ number. + * @num_phy_chans: The number of physical channels. Read from HW. This + * is the number of available channels for this driver, not counting "Secure + * mode" allocated physical channels. + * @num_log_chans: The number of logical channels. Calculated from + * num_phy_chans. + * @dma_both: dma_device channels that can do both memcpy and slave transfers. + * @dma_slave: dma_device channels that can do only do slave transfers. + * @dma_memcpy: dma_device channels that can do only do memcpy transfers. + * @phy_chans: Room for all possible physical channels in system. + * @log_chans: Room for all possible logical channels in system. + * @lookup_log_chans: Used to map interrupt number to logical channel. Points + * to log_chans entries. + * @lookup_phy_chans: Used to map interrupt number to physical channel. Points + * to phy_chans entries. + * @plat_data: Pointer to provided platform_data which is the driver + * configuration. + * @phy_res: Vector containing all physical channels. + * @lcla_pool: lcla pool settings and data. + * @lcpa_base: The virtual mapped address of LCPA. + * @phy_lcpa: The physical address of the LCPA. + * @lcpa_size: The size of the LCPA area. + */ +struct d40_base { + spinlock_t interrupt_lock; + spinlock_t execmd_lock; + struct device *dev; + void __iomem *virtbase; + struct clk *clk; + phys_addr_t phy_start; + resource_size_t phy_size; + int irq; + int num_phy_chans; + int num_log_chans; + struct dma_device dma_both; + struct dma_device dma_slave; + struct dma_device dma_memcpy; + struct d40_chan *phy_chans; + struct d40_chan *log_chans; + struct d40_chan **lookup_log_chans; + struct d40_chan **lookup_phy_chans; + struct stedma40_platform_data *plat_data; + /* Physical half channels */ + struct d40_phy_res *phy_res; + struct d40_lcla_pool lcla_pool; + void *lcpa_base; + dma_addr_t phy_lcpa; + resource_size_t lcpa_size; +}; + +/** + * struct d40_interrupt_lookup - lookup table for interrupt handler + * + * @src: Interrupt mask register. + * @clr: Interrupt clear register. + * @is_error: true if this is an error interrupt. + * @offset: start delta in the lookup_log_chans in d40_base. If equals to + * D40_PHY_CHAN, the lookup_phy_chans shall be used instead. + */ +struct d40_interrupt_lookup { + u32 src; + u32 clr; + bool is_error; + int offset; +}; + +/** + * struct d40_reg_val - simple lookup struct + * + * @reg: The register. + * @val: The value that belongs to the register in reg. + */ +struct d40_reg_val { + unsigned int reg; + unsigned int val; +}; + +static int d40_pool_lli_alloc(struct d40_desc *d40d, + int lli_len, bool is_log) +{ + u32 align; + void *base; + + if (is_log) + align = sizeof(struct d40_log_lli); + else + align = sizeof(struct d40_phy_lli); + + if (lli_len == 1) { + base = d40d->lli_pool.pre_alloc_lli; + d40d->lli_pool.size = sizeof(d40d->lli_pool.pre_alloc_lli); + d40d->lli_pool.base = NULL; + } else { + d40d->lli_pool.size = ALIGN(lli_len * 2 * align, align); + + base = kmalloc(d40d->lli_pool.size + align, GFP_NOWAIT); + d40d->lli_pool.base = base; + + if (d40d->lli_pool.base == NULL) + return -ENOMEM; + } + + if (is_log) { + d40d->lli_log.src = PTR_ALIGN((struct d40_log_lli *) base, + align); + d40d->lli_log.dst = PTR_ALIGN(d40d->lli_log.src + lli_len, + align); + } else { + d40d->lli_phy.src = PTR_ALIGN((struct d40_phy_lli *)base, + align); + d40d->lli_phy.dst = PTR_ALIGN(d40d->lli_phy.src + lli_len, + align); + + d40d->lli_phy.src_addr = virt_to_phys(d40d->lli_phy.src); + d40d->lli_phy.dst_addr = virt_to_phys(d40d->lli_phy.dst); + } + + return 0; +} + +static void d40_pool_lli_free(struct d40_desc *d40d) +{ + kfree(d40d->lli_pool.base); + d40d->lli_pool.base = NULL; + d40d->lli_pool.size = 0; + d40d->lli_log.src = NULL; + d40d->lli_log.dst = NULL; + d40d->lli_phy.src = NULL; + d40d->lli_phy.dst = NULL; + d40d->lli_phy.src_addr = 0; + d40d->lli_phy.dst_addr = 0; +} + +static dma_cookie_t d40_assign_cookie(struct d40_chan *d40c, + struct d40_desc *desc) +{ + dma_cookie_t cookie = d40c->chan.cookie; + + if (++cookie < 0) + cookie = 1; + + d40c->chan.cookie = cookie; + desc->txd.cookie = cookie; + + return cookie; +} + +static void d40_desc_reset(struct d40_desc *d40d) +{ + d40d->lli_tcount = 0; +} + +static void d40_desc_remove(struct d40_desc *d40d) +{ + list_del(&d40d->node); +} + +static struct d40_desc *d40_desc_get(struct d40_chan *d40c) +{ + struct d40_desc *desc; + struct d40_desc *d; + struct d40_desc *_d; + + if (!list_empty(&d40c->client)) { + list_for_each_entry_safe(d, _d, &d40c->client, node) + if (async_tx_test_ack(&d->txd)) { + d40_pool_lli_free(d); + d40_desc_remove(d); + desc = d; + goto out; + } + } + + if (list_empty(&d40c->free)) { + /* Alloc new desc because we're out of used ones */ + desc = kzalloc(sizeof(struct d40_desc), GFP_NOWAIT); + if (desc == NULL) + goto out; + INIT_LIST_HEAD(&desc->node); + } else { + /* Reuse an old desc. */ + desc = list_first_entry(&d40c->free, + struct d40_desc, + node); + list_del(&desc->node); + d40c->free_len--; + } +out: + return desc; +} + +static void d40_desc_free(struct d40_chan *d40c, struct d40_desc *d40d) +{ + if (d40c->free_len < D40_DESC_CACHE_SIZE) { + list_add_tail(&d40d->node, &d40c->free); + d40c->free_len++; + } else + kfree(d40d); +} + +static void d40_desc_submit(struct d40_chan *d40c, struct d40_desc *desc) +{ + list_add_tail(&desc->node, &d40c->active); +} + +static struct d40_desc *d40_first_active_get(struct d40_chan *d40c) +{ + struct d40_desc *d; + + if (list_empty(&d40c->active)) + return NULL; + + d = list_first_entry(&d40c->active, + struct d40_desc, + node); + return d; +} + +static void d40_desc_queue(struct d40_chan *d40c, struct d40_desc *desc) +{ + list_add_tail(&desc->node, &d40c->queue); +} + +static struct d40_desc *d40_first_queued(struct d40_chan *d40c) +{ + struct d40_desc *d; + + if (list_empty(&d40c->queue)) + return NULL; + + d = list_first_entry(&d40c->queue, + struct d40_desc, + node); + return d; +} + +/* Support functions for logical channels */ + +static int d40_lcla_id_get(struct d40_chan *d40c, + struct d40_lcla_pool *pool) +{ + int src_id = 0; + int dst_id = 0; + struct d40_log_lli *lcla_lidx_base = + pool->base + d40c->phy_chan->num * 1024; + int i; + int lli_per_log = d40c->base->plat_data->llis_per_log; + + if (d40c->lcla.src_id >= 0 && d40c->lcla.dst_id >= 0) + return 0; + + if (pool->num_blocks > 32) + return -EINVAL; + + spin_lock(&pool->lock); + + for (i = 0; i < pool->num_blocks; i++) { + if (!(pool->alloc_map[d40c->phy_chan->num] & (0x1 << i))) { + pool->alloc_map[d40c->phy_chan->num] |= (0x1 << i); + break; + } + } + src_id = i; + if (src_id >= pool->num_blocks) + goto err; + + for (; i < pool->num_blocks; i++) { + if (!(pool->alloc_map[d40c->phy_chan->num] & (0x1 << i))) { + pool->alloc_map[d40c->phy_chan->num] |= (0x1 << i); + break; + } + } + + dst_id = i; + if (dst_id == src_id) + goto err; + + d40c->lcla.src_id = src_id; + d40c->lcla.dst_id = dst_id; + d40c->lcla.dst = lcla_lidx_base + dst_id * lli_per_log + 1; + d40c->lcla.src = lcla_lidx_base + src_id * lli_per_log + 1; + + + spin_unlock(&pool->lock); + return 0; +err: + spin_unlock(&pool->lock); + return -EINVAL; +} + +static void d40_lcla_id_put(struct d40_chan *d40c, + struct d40_lcla_pool *pool, + int id) +{ + if (id < 0) + return; + + d40c->lcla.src_id = -1; + d40c->lcla.dst_id = -1; + + spin_lock(&pool->lock); + pool->alloc_map[d40c->phy_chan->num] &= (~(0x1 << id)); + spin_unlock(&pool->lock); +} + +static int d40_channel_execute_command(struct d40_chan *d40c, + enum d40_command command) +{ + int status, i; + void __iomem *active_reg; + int ret = 0; + unsigned long flags; + + spin_lock_irqsave(&d40c->base->execmd_lock, flags); + + if (d40c->phy_chan->num % 2 == 0) + active_reg = d40c->base->virtbase + D40_DREG_ACTIVE; + else + active_reg = d40c->base->virtbase + D40_DREG_ACTIVO; + + if (command == D40_DMA_SUSPEND_REQ) { + status = (readl(active_reg) & + D40_CHAN_POS_MASK(d40c->phy_chan->num)) >> + D40_CHAN_POS(d40c->phy_chan->num); + + if (status == D40_DMA_SUSPENDED || status == D40_DMA_STOP) + goto done; + } + + writel(command << D40_CHAN_POS(d40c->phy_chan->num), active_reg); + + if (command == D40_DMA_SUSPEND_REQ) { + + for (i = 0 ; i < D40_SUSPEND_MAX_IT; i++) { + status = (readl(active_reg) & + D40_CHAN_POS_MASK(d40c->phy_chan->num)) >> + D40_CHAN_POS(d40c->phy_chan->num); + + cpu_relax(); + /* + * Reduce the number of bus accesses while + * waiting for the DMA to suspend. + */ + udelay(3); + + if (status == D40_DMA_STOP || + status == D40_DMA_SUSPENDED) + break; + } + + if (i == D40_SUSPEND_MAX_IT) { + dev_err(&d40c->chan.dev->device, + "[%s]: unable to suspend the chl %d (log: %d) status %x\n", + __func__, d40c->phy_chan->num, d40c->log_num, + status); + dump_stack(); + ret = -EBUSY; + } + + } +done: + spin_unlock_irqrestore(&d40c->base->execmd_lock, flags); + return ret; +} + +static void d40_term_all(struct d40_chan *d40c) +{ + struct d40_desc *d40d; + struct d40_desc *d; + struct d40_desc *_d; + + /* Release active descriptors */ + while ((d40d = d40_first_active_get(d40c))) { + d40_desc_remove(d40d); + + /* Return desc to free-list */ + d40_desc_free(d40c, d40d); + } + + /* Release queued descriptors waiting for transfer */ + while ((d40d = d40_first_queued(d40c))) { + d40_desc_remove(d40d); + + /* Return desc to free-list */ + d40_desc_free(d40c, d40d); + } + + /* Release client owned descriptors */ + if (!list_empty(&d40c->client)) + list_for_each_entry_safe(d, _d, &d40c->client, node) { + d40_pool_lli_free(d); + d40_desc_remove(d); + /* Return desc to free-list */ + d40_desc_free(d40c, d40d); + } + + d40_lcla_id_put(d40c, &d40c->base->lcla_pool, + d40c->lcla.src_id); + d40_lcla_id_put(d40c, &d40c->base->lcla_pool, + d40c->lcla.dst_id); + + d40c->pending_tx = 0; + d40c->busy = false; +} + +static void d40_config_set_event(struct d40_chan *d40c, bool do_enable) +{ + u32 val; + unsigned long flags; + + if (do_enable) + val = D40_ACTIVATE_EVENTLINE; + else + val = D40_DEACTIVATE_EVENTLINE; + + spin_lock_irqsave(&d40c->phy_chan->lock, flags); + + /* Enable event line connected to device (or memcpy) */ + if ((d40c->dma_cfg.dir == STEDMA40_PERIPH_TO_MEM) || + (d40c->dma_cfg.dir == STEDMA40_PERIPH_TO_PERIPH)) { + u32 event = D40_TYPE_TO_EVENT(d40c->dma_cfg.src_dev_type); + + writel((val << D40_EVENTLINE_POS(event)) | + ~D40_EVENTLINE_MASK(event), + d40c->base->virtbase + D40_DREG_PCBASE + + d40c->phy_chan->num * D40_DREG_PCDELTA + + D40_CHAN_REG_SSLNK); + } + if (d40c->dma_cfg.dir != STEDMA40_PERIPH_TO_MEM) { + u32 event = D40_TYPE_TO_EVENT(d40c->dma_cfg.dst_dev_type); + + writel((val << D40_EVENTLINE_POS(event)) | + ~D40_EVENTLINE_MASK(event), + d40c->base->virtbase + D40_DREG_PCBASE + + d40c->phy_chan->num * D40_DREG_PCDELTA + + D40_CHAN_REG_SDLNK); + } + + spin_unlock_irqrestore(&d40c->phy_chan->lock, flags); +} + +static bool d40_chan_has_events(struct d40_chan *d40c) +{ + u32 val = 0; + + /* If SSLNK or SDLNK is zero all events are disabled */ + if ((d40c->dma_cfg.dir == STEDMA40_PERIPH_TO_MEM) || + (d40c->dma_cfg.dir == STEDMA40_PERIPH_TO_PERIPH)) + val = readl(d40c->base->virtbase + D40_DREG_PCBASE + + d40c->phy_chan->num * D40_DREG_PCDELTA + + D40_CHAN_REG_SSLNK); + + if (d40c->dma_cfg.dir != STEDMA40_PERIPH_TO_MEM) + val = readl(d40c->base->virtbase + D40_DREG_PCBASE + + d40c->phy_chan->num * D40_DREG_PCDELTA + + D40_CHAN_REG_SDLNK); + return (bool) val; +} + +static void d40_config_enable_lidx(struct d40_chan *d40c) +{ + /* Set LIDX for lcla */ + writel((d40c->phy_chan->num << D40_SREG_ELEM_LOG_LIDX_POS) & + D40_SREG_ELEM_LOG_LIDX_MASK, + d40c->base->virtbase + D40_DREG_PCBASE + + d40c->phy_chan->num * D40_DREG_PCDELTA + D40_CHAN_REG_SDELT); + + writel((d40c->phy_chan->num << D40_SREG_ELEM_LOG_LIDX_POS) & + D40_SREG_ELEM_LOG_LIDX_MASK, + d40c->base->virtbase + D40_DREG_PCBASE + + d40c->phy_chan->num * D40_DREG_PCDELTA + D40_CHAN_REG_SSELT); +} + +static int d40_config_write(struct d40_chan *d40c) +{ + u32 addr_base; + u32 var; + int res; + + res = d40_channel_execute_command(d40c, D40_DMA_SUSPEND_REQ); + if (res) + return res; + + /* Odd addresses are even addresses + 4 */ + addr_base = (d40c->phy_chan->num % 2) * 4; + /* Setup channel mode to logical or physical */ + var = ((u32)(d40c->log_num != D40_PHY_CHAN) + 1) << + D40_CHAN_POS(d40c->phy_chan->num); + writel(var, d40c->base->virtbase + D40_DREG_PRMSE + addr_base); + + /* Setup operational mode option register */ + var = ((d40c->dma_cfg.channel_type >> STEDMA40_INFO_CH_MODE_OPT_POS) & + 0x3) << D40_CHAN_POS(d40c->phy_chan->num); + + writel(var, d40c->base->virtbase + D40_DREG_PRMOE + addr_base); + + if (d40c->log_num != D40_PHY_CHAN) { + /* Set default config for CFG reg */ + writel(d40c->src_def_cfg, + d40c->base->virtbase + D40_DREG_PCBASE + + d40c->phy_chan->num * D40_DREG_PCDELTA + + D40_CHAN_REG_SSCFG); + writel(d40c->dst_def_cfg, + d40c->base->virtbase + D40_DREG_PCBASE + + d40c->phy_chan->num * D40_DREG_PCDELTA + + D40_CHAN_REG_SDCFG); + + d40_config_enable_lidx(d40c); + } + return res; +} + +static void d40_desc_load(struct d40_chan *d40c, struct d40_desc *d40d) +{ + + if (d40d->lli_phy.dst && d40d->lli_phy.src) { + d40_phy_lli_write(d40c->base->virtbase, + d40c->phy_chan->num, + d40d->lli_phy.dst, + d40d->lli_phy.src); + d40d->lli_tcount = d40d->lli_len; + } else if (d40d->lli_log.dst && d40d->lli_log.src) { + u32 lli_len; + struct d40_log_lli *src = d40d->lli_log.src; + struct d40_log_lli *dst = d40d->lli_log.dst; + + src += d40d->lli_tcount; + dst += d40d->lli_tcount; + + if (d40d->lli_len <= d40c->base->plat_data->llis_per_log) + lli_len = d40d->lli_len; + else + lli_len = d40c->base->plat_data->llis_per_log; + d40d->lli_tcount += lli_len; + d40_log_lli_write(d40c->lcpa, d40c->lcla.src, + d40c->lcla.dst, + dst, src, + d40c->base->plat_data->llis_per_log); + } +} + +static dma_cookie_t d40_tx_submit(struct dma_async_tx_descriptor *tx) +{ + struct d40_chan *d40c = container_of(tx->chan, + struct d40_chan, + chan); + struct d40_desc *d40d = container_of(tx, struct d40_desc, txd); + unsigned long flags; + + spin_lock_irqsave(&d40c->lock, flags); + + tx->cookie = d40_assign_cookie(d40c, d40d); + + d40_desc_queue(d40c, d40d); + + spin_unlock_irqrestore(&d40c->lock, flags); + + return tx->cookie; +} + +static int d40_start(struct d40_chan *d40c) +{ + int err; + + if (d40c->log_num != D40_PHY_CHAN) { + err = d40_channel_execute_command(d40c, D40_DMA_SUSPEND_REQ); + if (err) + return err; + d40_config_set_event(d40c, true); + } + + err = d40_channel_execute_command(d40c, D40_DMA_RUN); + + return err; +} + +static struct d40_desc *d40_queue_start(struct d40_chan *d40c) +{ + struct d40_desc *d40d; + int err; + + /* Start queued jobs, if any */ + d40d = d40_first_queued(d40c); + + if (d40d != NULL) { + d40c->busy = true; + + /* Remove from queue */ + d40_desc_remove(d40d); + + /* Add to active queue */ + d40_desc_submit(d40c, d40d |