* Re: [PATCH v24 5/5] remoteproc: Add initial zynqmp R5 remoteproc driver
2020-11-30 15:57 ` [PATCH v24 5/5] remoteproc: Add initial zynqmp R5 remoteproc driver Ben Levinsky
@ 2020-12-07 21:08 ` Mathieu Poirier
2020-12-08 16:48 ` Mathieu Poirier
` (2 subsequent siblings)
3 siblings, 0 replies; 13+ messages in thread
From: Mathieu Poirier @ 2020-12-07 21:08 UTC (permalink / raw)
To: Ben Levinsky; +Cc: devicetree, linux-remoteproc, linux-kernel, linux-arm-kernel
On Mon, Nov 30, 2020 at 07:57:17AM -0800, Ben Levinsky wrote:
> R5 is included in Xilinx Zynq UltraScale MPSoC so by adding this
> remotproc driver, we can boot the R5 sub-system in two different
> configurations -
> * Split
> * Lockstep
>
> The Xilinx R5 Remoteproc Driver boots the R5's via calls to the Xilinx
> Platform Management Unit that handles the R5 configuration, memory access
> and R5 lifecycle management. The interface to this manager is done in this
> driver via zynqmp_pm_* function calls.
>
> Signed-off-by: Wendy Liang <wendy.liang@xilinx.com>
> Signed-off-by: Michal Simek <michal.simek@xilinx.com>
> Signed-off-by: Ed Mooring <ed.mooring@xilinx.com>
> Signed-off-by: Jason Wu <j.wu@xilinx.com>
> Signed-off-by: Ben Levinsky <ben.levinsky@xilinx.com>
> ---
> drivers/remoteproc/Kconfig | 8 +
> drivers/remoteproc/Makefile | 1 +
> drivers/remoteproc/zynqmp_r5_remoteproc.c | 872 ++++++++++++++++++++++
> 3 files changed, 881 insertions(+)
> create mode 100644 drivers/remoteproc/zynqmp_r5_remoteproc.c
>
> diff --git a/drivers/remoteproc/Kconfig b/drivers/remoteproc/Kconfig
> index c6659dfea7c7..c2fe54b1d94f 100644
> --- a/drivers/remoteproc/Kconfig
> +++ b/drivers/remoteproc/Kconfig
> @@ -275,6 +275,14 @@ config TI_K3_DSP_REMOTEPROC
> It's safe to say N here if you're not interested in utilizing
> the DSP slave processors.
>
> +config ZYNQMP_R5_REMOTEPROC
> + tristate "ZynqMP R5 remoteproc support"
> + depends on PM && ARCH_ZYNQMP
> + select RPMSG_VIRTIO
> + select ZYNQMP_IPI_MBOX
> + help
> + Say y or m here to support ZynqMP R5 remote processors via the remote
> + processor framework.
> endif # REMOTEPROC
>
> endmenu
> diff --git a/drivers/remoteproc/Makefile b/drivers/remoteproc/Makefile
> index 3dfa28e6c701..ef1abff654c2 100644
> --- a/drivers/remoteproc/Makefile
> +++ b/drivers/remoteproc/Makefile
> @@ -33,3 +33,4 @@ obj-$(CONFIG_ST_REMOTEPROC) += st_remoteproc.o
> obj-$(CONFIG_ST_SLIM_REMOTEPROC) += st_slim_rproc.o
> obj-$(CONFIG_STM32_RPROC) += stm32_rproc.o
> obj-$(CONFIG_TI_K3_DSP_REMOTEPROC) += ti_k3_dsp_remoteproc.o
> +obj-$(CONFIG_ZYNQMP_R5_REMOTEPROC) += zynqmp_r5_remoteproc.o
> diff --git a/drivers/remoteproc/zynqmp_r5_remoteproc.c b/drivers/remoteproc/zynqmp_r5_remoteproc.c
> new file mode 100644
> index 000000000000..2593de618409
> --- /dev/null
> +++ b/drivers/remoteproc/zynqmp_r5_remoteproc.c
> @@ -0,0 +1,872 @@
> +// SPDX-License-Identifier: GPL-2.0
> +/*
> + * Zynq R5 Remote Processor driver
> + *
> + * Based on origin OMAP and Zynq Remote Processor driver
> + *
> + */
> +
> +#include <linux/firmware/xlnx-zynqmp.h>
> +#include <linux/interrupt.h>
> +#include <linux/kernel.h>
> +#include <linux/list.h>
> +#include <linux/mailbox_client.h>
> +#include <linux/mailbox/zynqmp-ipi-message.h>
> +#include <linux/module.h>
> +#include <linux/of_address.h>
> +#include <linux/of_platform.h>
> +#include <linux/of_reserved_mem.h>
> +#include <linux/platform_device.h>
> +#include <linux/remoteproc.h>
> +#include <linux/skbuff.h>
> +#include <linux/sysfs.h>
> +
> +#include "remoteproc_internal.h"
> +
> +#define MAX_RPROCS 2 /* Support up to 2 RPU */
> +#define MAX_MEM_PNODES 4 /* Max power nodes for one RPU memory instance */
> +
> +#define BANK_LIST_PROP "sram"
> +#define DDR_LIST_PROP "memory-region"
> +
> +/* IPI buffer MAX length */
> +#define IPI_BUF_LEN_MAX 32U
> +/* RX mailbox client buffer max length */
> +#define RX_MBOX_CLIENT_BUF_MAX (IPI_BUF_LEN_MAX + \
> + sizeof(struct zynqmp_ipi_message))
> +
> +/*
> + * Map each Xilinx on-chip SRAM Bank address to their own respective
> + * pm_node_id.
> + */
> +struct sram_addr_data {
> + phys_addr_t addr;
> + enum pm_node_id id;
> +};
> +
> +#define NUM_SRAMS 4U
> +static const struct sram_addr_data zynqmp_banks[NUM_SRAMS] = {
> + {0xffe00000UL, NODE_TCM_0_A},
> + {0xffe20000UL, NODE_TCM_0_B},
> + {0xffe90000UL, NODE_TCM_1_A},
> + {0xffeb0000UL, NODE_TCM_1_B},
> +};
> +
> +/**
> + * struct zynqmp_r5_rproc - ZynqMP R5 core structure
> + *
> + * @rx_mc_buf: rx mailbox client buffer to save the rx message
> + * @tx_mc: tx mailbox client
> + * @rx_mc: rx mailbox client
> + * @mbox_work: mbox_work for the RPU remoteproc
> + * @tx_mc_skbs: socket buffers for tx mailbox client
> + * @dev: device of RPU instance
> + * @rproc: rproc handle
> + * @tx_chan: tx mailbox channel
> + * @rx_chan: rx mailbox channel
> + * @pnode_id: RPU CPU power domain id
> + * @elem: linked list item
> + */
> +struct zynqmp_r5_rproc {
> + unsigned char rx_mc_buf[RX_MBOX_CLIENT_BUF_MAX];
> + struct mbox_client tx_mc;
> + struct mbox_client rx_mc;
> + struct work_struct mbox_work;
> + struct sk_buff_head tx_mc_skbs;
> + struct device *dev;
> + struct rproc *rproc;
> + struct mbox_chan *tx_chan;
> + struct mbox_chan *rx_chan;
> + u32 pnode_id;
> + struct list_head elem;
> +};
> +
> +/*
> + * r5_set_mode - set RPU operation mode
> + * @z_rproc: Remote processor private data
> + * @rpu_mode: mode specified by device tree to configure the RPU to
> + *
> + * set RPU operation mode
> + *
> + * Return: 0 for success, negative value for failure
> + */
> +static int r5_set_mode(struct zynqmp_r5_rproc *z_rproc,
> + enum rpu_oper_mode rpu_mode)
> +{
> + enum rpu_tcm_comb tcm_mode;
> + enum rpu_oper_mode cur_rpu_mode;
> + int ret;
> +
> + ret = zynqmp_pm_get_rpu_mode(z_rproc->pnode_id, &cur_rpu_mode);
> + if (ret < 0)
> + return ret;
> +
> + if (rpu_mode != cur_rpu_mode) {
> + ret = zynqmp_pm_set_rpu_mode(z_rproc->pnode_id,
> + rpu_mode);
> + if (ret < 0)
> + return ret;
> + }
> +
> + tcm_mode = (rpu_mode == PM_RPU_MODE_LOCKSTEP) ?
> + PM_RPU_TCM_COMB : PM_RPU_TCM_SPLIT;
> + return zynqmp_pm_set_tcm_config(z_rproc->pnode_id, tcm_mode);
> +}
> +
> +/*
> + * zynqmp_r5_rproc_mem_release
> + * @rproc: single R5 core's corresponding rproc instance
> + * @mem: mem entry to unmap
> + *
> + * Unmap TCM banks when powering down R5 core.
> + *
> + * return 0 on success, otherwise non-zero value on failure
> + */
> +static int tcm_mem_release(struct rproc *rproc, struct rproc_mem_entry *mem)
> +{
> + u32 pnode_id = (u64)mem->priv;
> +
> + iounmap(mem->va);
> + return zynqmp_pm_release_node(pnode_id);
> +}
> +
> +/*
> + * zynqmp_r5_rproc_start
> + * @rproc: single R5 core's corresponding rproc instance
> + *
> + * Start R5 Core from designated boot address.
> + *
> + * return 0 on success, otherwise non-zero value on failure
> + */
> +static int zynqmp_r5_rproc_start(struct rproc *rproc)
> +{
> + struct zynqmp_r5_rproc *z_rproc = rproc->priv;
> + enum rpu_boot_mem bootmem;
> +
> + bootmem = (rproc->bootaddr & 0xF0000000) == 0xF0000000 ?
> + PM_RPU_BOOTMEM_HIVEC : PM_RPU_BOOTMEM_LOVEC;
> +
> + dev_dbg(rproc->dev.parent, "RPU boot from %s.",
> + bootmem == PM_RPU_BOOTMEM_HIVEC ? "OCM" : "TCM");
> +
> + return zynqmp_pm_request_wake(z_rproc->pnode_id, 1,
> + bootmem, ZYNQMP_PM_REQUEST_ACK_NO);
> +}
> +
> +/*
> + * zynqmp_r5_rproc_stop
> + * @rproc: single R5 core's corresponding rproc instance
> + *
> + * Power down R5 Core.
> + *
> + * return 0 on success, otherwise non-zero value on failure
> + */
> +static int zynqmp_r5_rproc_stop(struct rproc *rproc)
> +{
> + struct zynqmp_r5_rproc *z_rproc = rproc->priv;
> +
> + return zynqmp_pm_force_pwrdwn(z_rproc->pnode_id,
> + ZYNQMP_PM_REQUEST_ACK_BLOCKING);
> +}
> +
> +/*
> + * zynqmp_r5_rproc_mem_alloc
> + * @rproc: single R5 core's corresponding rproc instance
> + * @mem: mem entry to map
> + *
> + * Callback to map va for memory-region's carveout.
> + *
> + * return 0 on success, otherwise non-zero value on failure
> + */
> +static int zynqmp_r5_rproc_mem_alloc(struct rproc *rproc,
> + struct rproc_mem_entry *mem)
> +{
> + void *va;
> +
> + va = ioremap_wc(mem->dma, mem->len);
> + if (IS_ERR_OR_NULL(va))
> + return -ENOMEM;
> +
> + mem->va = va;
> +
> + return 0;
> +}
> +
> +/*
> + * zynqmp_r5_rproc_mem_release
> + * @rproc: single R5 core's corresponding rproc instance
> + * @mem: mem entry to unmap
> + *
> + * Unmap memory-region carveout
> + *
> + * return 0 on success, otherwise non-zero value on failure
> + */
> +static int zynqmp_r5_rproc_mem_release(struct rproc *rproc,
> + struct rproc_mem_entry *mem)
> +{
> + iounmap(mem->va);
> + return 0;
> +}
> +
> +/*
> + * parse_mem_regions
> + * @rproc: single R5 core's corresponding rproc instance
> + *
> + * Construct rproc mem carveouts from carveout provided in
> + * memory-region property
> + *
> + * return 0 on success, otherwise non-zero value on failure
> + */
> +static int parse_mem_regions(struct rproc *rproc)
> +{
> + int num_mems, i;
> + struct zynqmp_r5_rproc *z_rproc = rproc->priv;
> + struct device *dev = &rproc->dev;
> + struct device_node *np = z_rproc->dev->of_node;
> + struct rproc_mem_entry *mem;
> +
> + num_mems = of_count_phandle_with_args(np, DDR_LIST_PROP, NULL);
> + if (num_mems <= 0)
> + return 0;
> +
> + for (i = 0; i < num_mems; i++) {
> + struct device_node *node;
> + struct reserved_mem *rmem;
> +
> + node = of_parse_phandle(np, DDR_LIST_PROP, i);
> + if (!node)
> + return -EINVAL;
> +
> + rmem = of_reserved_mem_lookup(node);
> + if (!rmem)
> + return -EINVAL;
> +
> + if (strstr(node->name, "vdev0vring")) {
> + int vring_id;
> + char name[16];
> +
> + /*
> + * expecting form of "rpuXvdev0vringX as documented
> + * in xilinx remoteproc device tree binding
> + */
> + if (strlen(node->name) < 15) {
> + dev_err(dev, "%pOF is less than 14 chars",
> + node);
> + return -EINVAL;
> + }
> +
> + /*
> + * can be 1 of multiple vring IDs per IPC channel
> + * e.g. 'vdev0vring0' and 'vdev0vring1'
> + */
> + vring_id = node->name[14] - '0';
> + snprintf(name, sizeof(name), "vdev0vring%d", vring_id);
> + /* Register vring */
> + mem = rproc_mem_entry_init(dev, NULL,
> + (dma_addr_t)rmem->base,
> + rmem->size, rmem->base,
> + zynqmp_r5_rproc_mem_alloc,
> + zynqmp_r5_rproc_mem_release,
> + name);
> + } else {
> + /* Register DMA region */
> + int (*alloc)(struct rproc *r,
> + struct rproc_mem_entry *rme);
> + int (*release)(struct rproc *r,
> + struct rproc_mem_entry *rme);
> + char name[20];
> +
> + if (strstr(node->name, "vdev0buffer")) {
> + alloc = NULL;
> + release = NULL;
> + strcpy(name, "vdev0buffer");
> + } else {
> + alloc = zynqmp_r5_rproc_mem_alloc;
> + release = zynqmp_r5_rproc_mem_release;
> + strcpy(name, node->name);
> + }
> +
> + mem = rproc_mem_entry_init(dev, NULL,
> + (dma_addr_t)rmem->base,
> + rmem->size, rmem->base,
> + alloc, release, name);
> + }
> + if (!mem)
> + return -ENOMEM;
> +
> + rproc_add_carveout(rproc, mem);
> + }
> +
> + return 0;
> +}
> +
> +/*
> + * zynqmp_r5_pm_request_tcm
> + * @addr: base address of mem provided in R5 core's sram property.
> + *
> + * Given sram base address, determine its corresponding Xilinx
> + * Platform Management ID and then request access to this node
> + * so that it can be power up.
> + *
> + * return 0 on success, otherwise non-zero value on failure
> + */
> +static int zynqmp_r5_pm_request_sram(phys_addr_t addr)
> +{
> + unsigned int i;
> +
> + for (i = 0; i < NUM_SRAMS; i++) {
> + if (zynqmp_banks[i].addr == addr)
> + return zynqmp_pm_request_node(zynqmp_banks[i].id,
> + ZYNQMP_PM_CAPABILITY_ACCESS,
> + 0,
> + ZYNQMP_PM_REQUEST_ACK_BLOCKING);
> + }
> +
> + return -EINVAL;
> +}
> +
> +/*
> + * tcm_mem_alloc
> + * @rproc: single R5 core's corresponding rproc instance
> + * @mem: mem entry to initialize the va and da fields of
> + *
> + * Given TCM bank entry,
> + * this callback will set device address for R5 running on TCM
> + * and also setup virtual address for TCM bank remoteproc carveout
> + *
> + * return 0 on success, otherwise non-zero value on failure
> + */
> +static int tcm_mem_alloc(struct rproc *rproc,
> + struct rproc_mem_entry *mem)
> +{
> + void *va;
> + struct device *dev = rproc->dev.parent;
> +
> + va = ioremap_wc(mem->dma, mem->len);
> + if (IS_ERR_OR_NULL(va))
> + return -ENOMEM;
> +
> + /* Update memory entry va */
> + mem->va = va;
> +
> + va = devm_ioremap_wc(dev, mem->da, mem->len);
> + if (!va)
> + return -ENOMEM;
> + /* As R5 is 32 bit, wipe out extra high bits */
> + mem->da &= 0x000fffff;
> + /*
> + * The R5s expect their TCM banks to be at address 0x0 and 0x2000,
> + * while on the Linux side they are at 0xffexxxxx. Zero out the high
> + * 12 bits of the address.
> + */
> +
> + /*
> + * TCM Banks 1A and 1B (0xffe90000 and 0xffeb0000) still
> + * need to be translated to 0x0 and 0x20000
> + */
> + if (mem->da == 0x90000 || mem->da == 0xB0000)
> + mem->da -= 0x90000;
> +
> + /* if translated TCM bank address is not valid report error */
> + if (mem->da != 0x0 && mem->da != 0x20000) {
> + dev_err(dev, "invalid TCM bank address: %x\n", mem->da);
> + return -EINVAL;
> + }
> +
> + return 0;
> +}
> +
> +/*
> + * parse_tcm_banks()
> + * @rproc: single R5 core's corresponding rproc instance
> + *
> + * Given R5 node in remoteproc instance
> + * allocate remoteproc carveout for TCM memory
> + * needed for firmware to be loaded
> + *
> + * return 0 on success, otherwise non-zero value on failure
> + */
> +static int parse_tcm_banks(struct rproc *rproc)
> +{
> + int i, num_banks;
> + struct zynqmp_r5_rproc *z_rproc = rproc->priv;
> + struct device *dev = &rproc->dev;
> + struct device_node *r5_node = z_rproc->dev->of_node;
> +
> + /* go through TCM banks for r5 node */
> + num_banks = of_count_phandle_with_args(r5_node, BANK_LIST_PROP, NULL);
> + if (num_banks <= 0) {
> + dev_err(dev, "need to specify TCM banks\n");
> + return -EINVAL;
> + }
> + for (i = 0; i < num_banks; i++) {
> + struct resource rsc;
> + resource_size_t size;
> + struct device_node *dt_node;
> + struct rproc_mem_entry *mem;
> + int ret;
> + u32 pnode_id; /* zynqmp_pm* fn's expect u32 */
> +
> + dt_node = of_parse_phandle(r5_node, BANK_LIST_PROP, i);
> + if (!dt_node)
> + return -EINVAL;
> +
> + if (of_device_is_available(dt_node)) {
> + ret = of_address_to_resource(dt_node, 0, &rsc);
> + if (ret < 0)
> + return ret;
> + ret = zynqmp_r5_pm_request_sram(rsc.start);
> + if (ret < 0)
> + return ret;
> +
> + /* add carveout */
> + size = resource_size(&rsc);
> + mem = rproc_mem_entry_init(dev, NULL, rsc.start,
> + (int)size, rsc.start,
> + tcm_mem_alloc,
> + tcm_mem_release,
> + rsc.name);
> + if (!mem)
> + return -ENOMEM;
> +
> + mem->priv = (void *)(u64)pnode_id;
> + rproc_add_carveout(rproc, mem);
> + }
> + }
> +
> + return 0;
> +}
> +
> +/*
> + * zynqmp_r5_parse_fw()
> + * @rproc: single R5 core's corresponding rproc instance
> + * @fw: ptr to firmware to be loaded onto r5 core
> + *
> + * When loading firmware, ensure the necessary carveouts are in remoteproc
> + *
> + * return 0 on success, otherwise non-zero value on failure
> + */
> +static int zynqmp_r5_parse_fw(struct rproc *rproc, const struct firmware *fw)
> +{
> + int ret;
> +
> + ret = parse_tcm_banks(rproc);
> + if (ret)
> + return ret;
> +
> + ret = parse_mem_regions(rproc);
> + if (ret)
> + return ret;
> +
> + ret = rproc_elf_load_rsc_table(rproc, fw);
> + if (ret == -EINVAL) {
> + /*
> + * resource table only required for IPC.
> + * if not present, this is not necessarily an error;
> + * for example, loading r5 hello world application
> + * so simply inform user and keep going.
> + */
> + dev_info(&rproc->dev, "no resource table found.\n");
> + ret = 0;
> + }
> + return ret;
> +}
> +
> +/*
> + * zynqmp_r5_rproc_kick() - kick a firmware if mbox is provided
> + * @rproc: r5 core's corresponding rproc structure
> + * @vqid: virtqueue ID
> + */
> +static void zynqmp_r5_rproc_kick(struct rproc *rproc, int vqid)
> +{
> + struct sk_buff *skb;
> + unsigned int skb_len;
> + struct zynqmp_ipi_message *mb_msg;
> + int ret;
> +
> + struct device *dev = rproc->dev.parent;
> + struct zynqmp_r5_rproc *z_rproc = rproc->priv;
> +
> + if (of_property_read_bool(dev->of_node, "mboxes")) {
> + skb_len = (unsigned int)(sizeof(vqid) + sizeof(mb_msg));
> + skb = alloc_skb(skb_len, GFP_ATOMIC);
> + if (!skb)
> + return;
> +
> + mb_msg = (struct zynqmp_ipi_message *)skb_put(skb, skb_len);
> + mb_msg->len = sizeof(vqid);
> + memcpy(mb_msg->data, &vqid, sizeof(vqid));
> +
> + skb_queue_tail(&z_rproc->tx_mc_skbs, skb);
> + ret = mbox_send_message(z_rproc->tx_chan, mb_msg);
> + if (ret < 0) {
> + dev_warn(dev, "Failed to kick remote.\n");
> + skb_dequeue_tail(&z_rproc->tx_mc_skbs);
> + kfree_skb(skb);
> + }
> + } else {
> + (void)skb;
> + (void)skb_len;
> + (void)mb_msg;
> + (void)ret;
> + (void)vqid;
> + }
> +}
> +
> +static struct rproc_ops zynqmp_r5_rproc_ops = {
> + .start = zynqmp_r5_rproc_start,
> + .stop = zynqmp_r5_rproc_stop,
> + .load = rproc_elf_load_segments,
> + .parse_fw = zynqmp_r5_parse_fw,
> + .find_loaded_rsc_table = rproc_elf_find_loaded_rsc_table,
> + .sanity_check = rproc_elf_sanity_check,
> + .get_boot_addr = rproc_elf_get_boot_addr,
> + .kick = zynqmp_r5_rproc_kick,
> +};
> +
> +/**
> + * event_notified_idr_cb() - event notified idr callback
> + * @id: idr id
> + * @ptr: pointer to idr private data
> + * @data: data passed to idr_for_each callback
> + *
> + * Pass notification to remoteproc virtio
> + *
> + * Return: 0. having return is to satisfy the idr_for_each() function
> + * pointer input argument requirement.
> + **/
> +static int event_notified_idr_cb(int id, void *ptr, void *data)
> +{
> + struct rproc *rproc = data;
> +
> + (void)rproc_vq_interrupt(rproc, id);
> + return 0;
> +}
> +
> +/**
> + * handle_event_notified() - remoteproc notification work function
> + * @work: pointer to the work structure
> + *
> + * It checks each registered remoteproc notify IDs.
> + */
> +static void handle_event_notified(struct work_struct *work)
> +{
> + struct rproc *rproc;
> + struct zynqmp_r5_rproc *z_rproc;
> +
> + z_rproc = container_of(work, struct zynqmp_r5_rproc, mbox_work);
> +
> + (void)mbox_send_message(z_rproc->rx_chan, NULL);
> + rproc = z_rproc->rproc;
> + /*
> + * We only use IPI for interrupt. The firmware side may or may
> + * not write the notifyid when it trigger IPI.
> + * And thus, we scan through all the registered notifyids.
> + */
> + idr_for_each(&rproc->notifyids, event_notified_idr_cb, rproc);
> +}
> +
> +/**
> + * zynqmp_r5_mb_rx_cb() - Receive channel mailbox callback
> + * @cl: mailbox client
> + * @msg: message pointer
> + *
> + * It will schedule the R5 notification work.
> + */
> +static void zynqmp_r5_mb_rx_cb(struct mbox_client *cl, void *msg)
> +{
> + struct zynqmp_r5_rproc *z_rproc;
> +
> + z_rproc = container_of(cl, struct zynqmp_r5_rproc, rx_mc);
> + if (msg) {
> + struct zynqmp_ipi_message *ipi_msg, *buf_msg;
> + size_t len;
> +
> + ipi_msg = (struct zynqmp_ipi_message *)msg;
> + buf_msg = (struct zynqmp_ipi_message *)z_rproc->rx_mc_buf;
> + len = (ipi_msg->len >= IPI_BUF_LEN_MAX) ?
> + IPI_BUF_LEN_MAX : ipi_msg->len;
> + buf_msg->len = len;
> + memcpy(buf_msg->data, ipi_msg->data, len);
> + }
> + schedule_work(&z_rproc->mbox_work);
> +}
> +
> +/**
> + * zynqmp_r5_mb_tx_done() - Request has been sent to the remote
> + * @cl: mailbox client
> + * @msg: pointer to the message which has been sent
> + * @r: status of last TX - OK or error
> + *
> + * It will be called by the mailbox framework when the last TX has done.
> + */
> +static void zynqmp_r5_mb_tx_done(struct mbox_client *cl, void *msg, int r)
> +{
> + struct zynqmp_r5_rproc *z_rproc;
> + struct sk_buff *skb;
> +
> + if (!msg)
> + return;
> + z_rproc = container_of(cl, struct zynqmp_r5_rproc, tx_mc);
> + skb = skb_dequeue(&z_rproc->tx_mc_skbs);
> + kfree_skb(skb);
> +}
> +
> +/**
> + * zynqmp_r5_setup_mbox() - Setup mailboxes
> + * this is used for each individual R5 core
> + *
> + * @z_rproc: pointer to the ZynqMP R5 processor platform data
> + * @node: pointer of the device node
> + *
> + * Function to setup mailboxes to talk to RPU.
> + *
> + * Return: 0 for success, negative value for failure.
> + */
> +static int zynqmp_r5_setup_mbox(struct zynqmp_r5_rproc *z_rproc,
> + struct device_node *node)
> +{
> + struct mbox_client *mclient;
> +
> + /* Setup TX mailbox channel client */
> + mclient = &z_rproc->tx_mc;
> + mclient->rx_callback = NULL;
> + mclient->tx_block = false;
> + mclient->knows_txdone = false;
> + mclient->tx_done = zynqmp_r5_mb_tx_done;
> + mclient->dev = z_rproc->dev;
> +
> + /* Setup TX mailbox channel client */
> + mclient = &z_rproc->rx_mc;
> + mclient->dev = z_rproc->dev;
> + mclient->rx_callback = zynqmp_r5_mb_rx_cb;
> + mclient->tx_block = false;
> + mclient->knows_txdone = false;
It would be much easier to read (and review) if the initialistion of tx_mc and
rx_mc was done in the same order.
> +
> + INIT_WORK(&z_rproc->mbox_work, handle_event_notified);
> +
> + /* Request TX and RX channels */
> + z_rproc->tx_chan = mbox_request_channel_byname(&z_rproc->tx_mc, "tx");
> + if (IS_ERR(z_rproc->tx_chan)) {
> + dev_err(z_rproc->dev, "failed to request mbox tx channel.\n");
> + z_rproc->tx_chan = NULL;
> + return -EINVAL;
> + }
> +
> + z_rproc->rx_chan = mbox_request_channel_byname(&z_rproc->rx_mc, "rx");
> + if (IS_ERR(z_rproc->rx_chan)) {
> + dev_err(z_rproc->dev, "failed to request mbox rx channel.\n");
> + z_rproc->rx_chan = NULL;
Functions should cleanup after themselves when things go wrong. That way we
know exactly the state of the system and how to cleanup successful allocations.
> + return -EINVAL;
> + }
> + skb_queue_head_init(&z_rproc->tx_mc_skbs);
> +
> + return 0;
> +}
> +
> +/**
> + * zynqmp_r5_probe() - Probes ZynqMP R5 processor device node
> + * this is called for each individual R5 core to
> + * set up mailbox, Xilinx platform manager unique ID,
> + * add to rproc core
> + *
> + * @pdev: domain platform device for current R5 core
> + * @node: pointer of the device node for current R5 core
> + * @rpu_mode: mode to configure RPU, split or lockstep
> + * @z_rproc: Xilinx specific remoteproc structure used later to link
> + * in to cluster of cores
> + *
> + * Return: 0 for success, negative value for failure.
> + */
> +static int zynqmp_r5_probe(struct platform_device *pdev,
> + struct device_node *node,
> + enum rpu_oper_mode rpu_mode,
> + struct zynqmp_r5_rproc **z_rproc)
> +{
This double pointer thing is adding needless complexity to the driver. I
suggest removing the z_proc parameter completely and return a struct
*zynqmp_r5_rproc instead of an int. Errors can be encoded in the ERR_PTR()
macro.
> + int ret;
> + struct device *dev = &pdev->dev;
> + struct rproc *rproc_ptr;
> +
> + /* Allocate remoteproc instance */
> + rproc_ptr = devm_rproc_alloc(dev, dev_name(dev), &zynqmp_r5_rproc_ops,
> + NULL, sizeof(struct zynqmp_r5_rproc));
> + if (!rproc_ptr) {
> + ret = -ENOMEM;
> + goto error;
> + }
> +
> + rproc_ptr->auto_boot = false;
> + *z_rproc = rproc_ptr->priv;
> + (*z_rproc)->rproc = rproc_ptr;
> + (*z_rproc)->dev = dev;
> + /* Set up DMA mask */
> + ret = dma_set_coherent_mask(dev, DMA_BIT_MASK(32));
> + if (ret)
> + goto error;
> +
> + /* Get R5 power domain node */
> + ret = of_property_read_u32(node, "power-domain", &(*z_rproc)->pnode_id);
> + if (ret)
> + goto error;
> +
> + ret = r5_set_mode(*z_rproc, rpu_mode);
> + if (ret)
> + goto error;
> +
> + if (of_property_read_bool(node, "mboxes")) {
> + ret = zynqmp_r5_setup_mbox(*z_rproc, node);
> + if (ret)
> + goto error;
> + }
> +
> + /* Add R5 remoteproc */
> + ret = devm_rproc_add(dev, rproc_ptr);
> + if (ret) {
> + goto error;
if (ret) {
zynqmp_r5_cleanup_mbox(*z_rproc);
goto error;
}
> +
> + return 0;
> +error:
> + *z_rproc = NULL;
> + return ret;
> +}
> +
> +/*
> + * zynqmp_r5_remoteproc_probe()
> + *
> + * @pdev: domain platform device for R5 cluster
> + *
> + * called when driver is probed, for each R5 core specified in DT,
> + * setup as needed to do remoteproc-related operations
> + *
> + * Return: 0 for success, negative value for failure.
> + */
> +static int zynqmp_r5_remoteproc_probe(struct platform_device *pdev)
> +{
> + int ret, core_count;
> + struct device *dev = &pdev->dev;
> + struct device_node *nc;
> + enum rpu_oper_mode rpu_mode = PM_RPU_MODE_LOCKSTEP;
> + struct list_head *cluster; /* list to track each core's rproc */
> + struct zynqmp_r5_rproc *z_rproc;
> + struct platform_device *child_pdev;
> + struct list_head *pos;
> +
> + ret = of_property_read_u32(dev->of_node, "xlnx,cluster-mode", &rpu_mode);
> + if (ret < 0 || (rpu_mode != PM_RPU_MODE_LOCKSTEP &&
> + rpu_mode != PM_RPU_MODE_SPLIT)) {
> + dev_err(dev, "invalid format cluster mode: ret %d mode %x\n",
> + ret, rpu_mode);
> + return ret;
> + }
> +
> + dev_dbg(dev, "RPU configuration: %s\n",
> + rpu_mode == PM_RPU_MODE_LOCKSTEP ? "lockstep" : "split");
> +
> + /*
> + * if 2 RPUs provided but one is lockstep, then we have an
> + * invalid configuration.
> + */
> +
> + core_count = of_get_available_child_count(dev->of_node);
> + if ((rpu_mode == PM_RPU_MODE_LOCKSTEP && core_count != 1) ||
> + core_count > MAX_RPROCS)
> + return -EINVAL;
> +
> + cluster = devm_kzalloc(dev, sizeof(*cluster), GFP_KERNEL);
> + if (!cluster)
> + return -ENOMEM;
> + INIT_LIST_HEAD(cluster);
> +
> + ret = devm_of_platform_populate(dev);
> + if (ret) {
> + dev_err(dev, "devm_of_platform_populate failed, ret = %d\n",
> + ret);
> + return ret;
> + }
> +
> + /* probe each individual r5 core's remoteproc-related info */
> + for_each_available_child_of_node(dev->of_node, nc) {
> + child_pdev = of_find_device_by_node(nc);
> + if (!child_pdev) {
> + dev_err(dev, "could not get R5 core platform device\n");
> + ret = -ENODEV;
> + goto out;
> + }
> +
> + ret = zynqmp_r5_probe(child_pdev, nc, rpu_mode, &z_rproc);
> + dev_dbg(dev, "%s to probe rpu %pOF\n",
> + ret ? "Failed" : "Able",
> + nc);
> + if (!z_rproc)
> + ret = -EINVAL;
> + if (ret)
> + goto out;
> + list_add_tail(&z_rproc->elem, cluster);
> + }
> + /* wire in so each core can be cleaned up at driver remove */
> + platform_set_drvdata(pdev, cluster);
> + return 0;
> +out:
> + /*
> + * undo core0 upon any failures on core1 in split-mode
> + *
> + * in zynqmp_r5_probe z_rproc is set to null
> + * and ret to non-zero value if error
> + */
> + if (ret && !z_rproc && rpu_mode == PM_RPU_MODE_SPLIT &&
> + !list_empty(cluster)) {
This is needlessly complex and if I'm correct, mailbox channels won't be cleaned
up properly if devm_rproc_add() fails on core0. As far as I can tell the entire
if() above is not needed.
> + list_for_each(pos, cluster) {
> + z_rproc = list_entry(pos, struct zynqmp_r5_rproc, elem);
> + if (of_property_read_bool(z_rproc->dev->of_node, "mboxes")) {
This if() isn't needed either since calling mbox_free_channel(NULL) is
supported.
> + mbox_free_channel(z_rproc->tx_chan);
> + mbox_free_channel(z_rproc->rx_chan);
> + }
> + }
With the above comments the error patch should be:
out:
list_for_each(pos, cluster) {
z_rproc = list_entry(pos, struct zynqmp_r5_rproc, elem);
zynqmp_r5_cleanup_mbox(z_rproc);
}
I'm out of time for today, I will continue tomorrow.
Thanks,
Mathieu
> + }
> + return ret;
> +}
> +
> +/*
> + * zynqmp_r5_remoteproc_remove()
> + *
> + * @pdev: domain platform device for R5 cluster
> + *
> + * When the driver is unloaded, clean up the mailboxes for each
> + * remoteproc that was initially probed.
> + */
> +static int zynqmp_r5_remoteproc_remove(struct platform_device *pdev)
> +{
> + struct list_head *pos, *temp, *cluster = (struct list_head *)
> + platform_get_drvdata(pdev);
> + struct zynqmp_r5_rproc *z_rproc = NULL;
> +
> + list_for_each_safe(pos, temp, cluster) {
> + z_rproc = list_entry(pos, struct zynqmp_r5_rproc, elem);
> + if (of_property_read_bool(z_rproc->dev->of_node, "mboxes")) {
> + mbox_free_channel(z_rproc->tx_chan);
> + mbox_free_channel(z_rproc->rx_chan);
> + }
> + list_del(pos);
> + }
> + return 0;
> +}
> +
> +/* Match table for OF platform binding */
> +static const struct of_device_id zynqmp_r5_remoteproc_match[] = {
> + { .compatible = "xlnx,zynqmp-r5-remoteproc", },
> + { /* end of list */ },
> +};
> +MODULE_DEVICE_TABLE(of, zynqmp_r5_remoteproc_match);
> +
> +static struct platform_driver zynqmp_r5_remoteproc_driver = {
> + .probe = zynqmp_r5_remoteproc_probe,
> + .remove = zynqmp_r5_remoteproc_remove,
> + .driver = {
> + .name = "zynqmp_r5_remoteproc",
> + .of_match_table = zynqmp_r5_remoteproc_match,
> + },
> +};
> +module_platform_driver(zynqmp_r5_remoteproc_driver);
> +
> +MODULE_AUTHOR("Ben Levinsky <ben.levinsky@xilinx.com>");
> +MODULE_LICENSE("GPL v2");
> --
> 2.17.1
>
^ permalink raw reply [flat|nested] 13+ messages in thread* Re: [PATCH v24 5/5] remoteproc: Add initial zynqmp R5 remoteproc driver
2020-11-30 15:57 ` [PATCH v24 5/5] remoteproc: Add initial zynqmp R5 remoteproc driver Ben Levinsky
2020-12-07 21:08 ` Mathieu Poirier
@ 2020-12-08 16:48 ` Mathieu Poirier
2020-12-09 19:00 ` Mathieu Poirier
2020-12-10 20:56 ` Mathieu Poirier
3 siblings, 0 replies; 13+ messages in thread
From: Mathieu Poirier @ 2020-12-08 16:48 UTC (permalink / raw)
To: Ben Levinsky; +Cc: devicetree, linux-remoteproc, linux-kernel, linux-arm-kernel
On Mon, Nov 30, 2020 at 07:57:17AM -0800, Ben Levinsky wrote:
> R5 is included in Xilinx Zynq UltraScale MPSoC so by adding this
> remotproc driver, we can boot the R5 sub-system in two different
> configurations -
> * Split
> * Lockstep
>
> The Xilinx R5 Remoteproc Driver boots the R5's via calls to the Xilinx
> Platform Management Unit that handles the R5 configuration, memory access
> and R5 lifecycle management. The interface to this manager is done in this
> driver via zynqmp_pm_* function calls.
>
> Signed-off-by: Wendy Liang <wendy.liang@xilinx.com>
> Signed-off-by: Michal Simek <michal.simek@xilinx.com>
> Signed-off-by: Ed Mooring <ed.mooring@xilinx.com>
> Signed-off-by: Jason Wu <j.wu@xilinx.com>
> Signed-off-by: Ben Levinsky <ben.levinsky@xilinx.com>
> ---
> drivers/remoteproc/Kconfig | 8 +
> drivers/remoteproc/Makefile | 1 +
> drivers/remoteproc/zynqmp_r5_remoteproc.c | 872 ++++++++++++++++++++++
> 3 files changed, 881 insertions(+)
> create mode 100644 drivers/remoteproc/zynqmp_r5_remoteproc.c
>
> diff --git a/drivers/remoteproc/Kconfig b/drivers/remoteproc/Kconfig
> index c6659dfea7c7..c2fe54b1d94f 100644
> --- a/drivers/remoteproc/Kconfig
> +++ b/drivers/remoteproc/Kconfig
> @@ -275,6 +275,14 @@ config TI_K3_DSP_REMOTEPROC
> It's safe to say N here if you're not interested in utilizing
> the DSP slave processors.
>
> +config ZYNQMP_R5_REMOTEPROC
> + tristate "ZynqMP R5 remoteproc support"
> + depends on PM && ARCH_ZYNQMP
> + select RPMSG_VIRTIO
> + select ZYNQMP_IPI_MBOX
> + help
> + Say y or m here to support ZynqMP R5 remote processors via the remote
> + processor framework.
> endif # REMOTEPROC
>
> endmenu
> diff --git a/drivers/remoteproc/Makefile b/drivers/remoteproc/Makefile
> index 3dfa28e6c701..ef1abff654c2 100644
> --- a/drivers/remoteproc/Makefile
> +++ b/drivers/remoteproc/Makefile
> @@ -33,3 +33,4 @@ obj-$(CONFIG_ST_REMOTEPROC) += st_remoteproc.o
> obj-$(CONFIG_ST_SLIM_REMOTEPROC) += st_slim_rproc.o
> obj-$(CONFIG_STM32_RPROC) += stm32_rproc.o
> obj-$(CONFIG_TI_K3_DSP_REMOTEPROC) += ti_k3_dsp_remoteproc.o
> +obj-$(CONFIG_ZYNQMP_R5_REMOTEPROC) += zynqmp_r5_remoteproc.o
> diff --git a/drivers/remoteproc/zynqmp_r5_remoteproc.c b/drivers/remoteproc/zynqmp_r5_remoteproc.c
> new file mode 100644
> index 000000000000..2593de618409
> --- /dev/null
> +++ b/drivers/remoteproc/zynqmp_r5_remoteproc.c
> @@ -0,0 +1,872 @@
> +// SPDX-License-Identifier: GPL-2.0
> +/*
> + * Zynq R5 Remote Processor driver
> + *
> + * Based on origin OMAP and Zynq Remote Processor driver
> + *
> + */
> +
> +#include <linux/firmware/xlnx-zynqmp.h>
> +#include <linux/interrupt.h>
> +#include <linux/kernel.h>
> +#include <linux/list.h>
> +#include <linux/mailbox_client.h>
> +#include <linux/mailbox/zynqmp-ipi-message.h>
> +#include <linux/module.h>
> +#include <linux/of_address.h>
> +#include <linux/of_platform.h>
> +#include <linux/of_reserved_mem.h>
> +#include <linux/platform_device.h>
> +#include <linux/remoteproc.h>
> +#include <linux/skbuff.h>
> +#include <linux/sysfs.h>
> +
> +#include "remoteproc_internal.h"
> +
> +#define MAX_RPROCS 2 /* Support up to 2 RPU */
> +#define MAX_MEM_PNODES 4 /* Max power nodes for one RPU memory instance */
> +
> +#define BANK_LIST_PROP "sram"
> +#define DDR_LIST_PROP "memory-region"
> +
> +/* IPI buffer MAX length */
> +#define IPI_BUF_LEN_MAX 32U
> +/* RX mailbox client buffer max length */
> +#define RX_MBOX_CLIENT_BUF_MAX (IPI_BUF_LEN_MAX + \
> + sizeof(struct zynqmp_ipi_message))
> +
> +/*
> + * Map each Xilinx on-chip SRAM Bank address to their own respective
> + * pm_node_id.
> + */
> +struct sram_addr_data {
> + phys_addr_t addr;
> + enum pm_node_id id;
> +};
> +
> +#define NUM_SRAMS 4U
> +static const struct sram_addr_data zynqmp_banks[NUM_SRAMS] = {
> + {0xffe00000UL, NODE_TCM_0_A},
> + {0xffe20000UL, NODE_TCM_0_B},
> + {0xffe90000UL, NODE_TCM_1_A},
> + {0xffeb0000UL, NODE_TCM_1_B},
> +};
> +
> +/**
> + * struct zynqmp_r5_rproc - ZynqMP R5 core structure
> + *
> + * @rx_mc_buf: rx mailbox client buffer to save the rx message
> + * @tx_mc: tx mailbox client
> + * @rx_mc: rx mailbox client
> + * @mbox_work: mbox_work for the RPU remoteproc
> + * @tx_mc_skbs: socket buffers for tx mailbox client
> + * @dev: device of RPU instance
> + * @rproc: rproc handle
> + * @tx_chan: tx mailbox channel
> + * @rx_chan: rx mailbox channel
> + * @pnode_id: RPU CPU power domain id
> + * @elem: linked list item
> + */
> +struct zynqmp_r5_rproc {
> + unsigned char rx_mc_buf[RX_MBOX_CLIENT_BUF_MAX];
> + struct mbox_client tx_mc;
> + struct mbox_client rx_mc;
> + struct work_struct mbox_work;
> + struct sk_buff_head tx_mc_skbs;
> + struct device *dev;
> + struct rproc *rproc;
> + struct mbox_chan *tx_chan;
> + struct mbox_chan *rx_chan;
> + u32 pnode_id;
> + struct list_head elem;
> +};
> +
> +/*
> + * r5_set_mode - set RPU operation mode
> + * @z_rproc: Remote processor private data
> + * @rpu_mode: mode specified by device tree to configure the RPU to
> + *
> + * set RPU operation mode
> + *
> + * Return: 0 for success, negative value for failure
> + */
> +static int r5_set_mode(struct zynqmp_r5_rproc *z_rproc,
> + enum rpu_oper_mode rpu_mode)
> +{
> + enum rpu_tcm_comb tcm_mode;
> + enum rpu_oper_mode cur_rpu_mode;
> + int ret;
> +
> + ret = zynqmp_pm_get_rpu_mode(z_rproc->pnode_id, &cur_rpu_mode);
> + if (ret < 0)
> + return ret;
> +
> + if (rpu_mode != cur_rpu_mode) {
> + ret = zynqmp_pm_set_rpu_mode(z_rproc->pnode_id,
> + rpu_mode);
No need for stacking here, @rpu_node can easily fit on the above line.
> + if (ret < 0)
> + return ret;
> + }
> +
> + tcm_mode = (rpu_mode == PM_RPU_MODE_LOCKSTEP) ?
> + PM_RPU_TCM_COMB : PM_RPU_TCM_SPLIT;
> + return zynqmp_pm_set_tcm_config(z_rproc->pnode_id, tcm_mode);
> +}
> +
> +/*
> + * zynqmp_r5_rproc_mem_release
> + * @rproc: single R5 core's corresponding rproc instance
> + * @mem: mem entry to unmap
> + *
> + * Unmap TCM banks when powering down R5 core.
> + *
> + * return 0 on success, otherwise non-zero value on failure
> + */
> +static int tcm_mem_release(struct rproc *rproc, struct rproc_mem_entry *mem)
> +{
> + u32 pnode_id = (u64)mem->priv;
> +
> + iounmap(mem->va);
> + return zynqmp_pm_release_node(pnode_id);
> +}
> +
> +/*
> + * zynqmp_r5_rproc_start
> + * @rproc: single R5 core's corresponding rproc instance
> + *
> + * Start R5 Core from designated boot address.
> + *
> + * return 0 on success, otherwise non-zero value on failure
> + */
> +static int zynqmp_r5_rproc_start(struct rproc *rproc)
> +{
> + struct zynqmp_r5_rproc *z_rproc = rproc->priv;
> + enum rpu_boot_mem bootmem;
> +
> + bootmem = (rproc->bootaddr & 0xF0000000) == 0xF0000000 ?
> + PM_RPU_BOOTMEM_HIVEC : PM_RPU_BOOTMEM_LOVEC;
Indentation
> +
> + dev_dbg(rproc->dev.parent, "RPU boot from %s.",
> + bootmem == PM_RPU_BOOTMEM_HIVEC ? "OCM" : "TCM");
> +
> + return zynqmp_pm_request_wake(z_rproc->pnode_id, 1,
> + bootmem, ZYNQMP_PM_REQUEST_ACK_NO);
Indentation
> +}
> +
> +/*
> + * zynqmp_r5_rproc_stop
> + * @rproc: single R5 core's corresponding rproc instance
> + *
> + * Power down R5 Core.
> + *
> + * return 0 on success, otherwise non-zero value on failure
> + */
> +static int zynqmp_r5_rproc_stop(struct rproc *rproc)
> +{
> + struct zynqmp_r5_rproc *z_rproc = rproc->priv;
> +
> + return zynqmp_pm_force_pwrdwn(z_rproc->pnode_id,
> + ZYNQMP_PM_REQUEST_ACK_BLOCKING);
Indentation
> +}
> +
> +/*
> + * zynqmp_r5_rproc_mem_alloc
> + * @rproc: single R5 core's corresponding rproc instance
> + * @mem: mem entry to map
> + *
> + * Callback to map va for memory-region's carveout.
> + *
> + * return 0 on success, otherwise non-zero value on failure
> + */
> +static int zynqmp_r5_rproc_mem_alloc(struct rproc *rproc,
> + struct rproc_mem_entry *mem)
Indentation
> +{
> + void *va;
> +
> + va = ioremap_wc(mem->dma, mem->len);
> + if (IS_ERR_OR_NULL(va))
> + return -ENOMEM;
> +
> + mem->va = va;
> +
> + return 0;
> +}
> +
> +/*
> + * zynqmp_r5_rproc_mem_release
> + * @rproc: single R5 core's corresponding rproc instance
> + * @mem: mem entry to unmap
> + *
> + * Unmap memory-region carveout
> + *
> + * return 0 on success, otherwise non-zero value on failure
> + */
> +static int zynqmp_r5_rproc_mem_release(struct rproc *rproc,
> + struct rproc_mem_entry *mem)
> +{
> + iounmap(mem->va);
> + return 0;
> +}
> +
> +/*
> + * parse_mem_regions
> + * @rproc: single R5 core's corresponding rproc instance
> + *
> + * Construct rproc mem carveouts from carveout provided in
> + * memory-region property
> + *
> + * return 0 on success, otherwise non-zero value on failure
> + */
> +static int parse_mem_regions(struct rproc *rproc)
> +{
> + int num_mems, i;
> + struct zynqmp_r5_rproc *z_rproc = rproc->priv;
> + struct device *dev = &rproc->dev;
> + struct device_node *np = z_rproc->dev->of_node;
> + struct rproc_mem_entry *mem;
> +
> + num_mems = of_count_phandle_with_args(np, DDR_LIST_PROP, NULL);
> + if (num_mems <= 0)
> + return 0;
> +
> + for (i = 0; i < num_mems; i++) {
> + struct device_node *node;
> + struct reserved_mem *rmem;
> +
> + node = of_parse_phandle(np, DDR_LIST_PROP, i);
> + if (!node)
> + return -EINVAL;
> +
> + rmem = of_reserved_mem_lookup(node);
> + if (!rmem)
> + return -EINVAL;
> +
> + if (strstr(node->name, "vdev0vring")) {
> + int vring_id;
> + char name[16];
> +
> + /*
> + * expecting form of "rpuXvdev0vringX as documented
> + * in xilinx remoteproc device tree binding
> + */
> + if (strlen(node->name) < 15) {
> + dev_err(dev, "%pOF is less than 14 chars",
> + node);
> + return -EINVAL;
> + }
> +
> + /*
> + * can be 1 of multiple vring IDs per IPC channel
> + * e.g. 'vdev0vring0' and 'vdev0vring1'
> + */
> + vring_id = node->name[14] - '0';
> + snprintf(name, sizeof(name), "vdev0vring%d", vring_id);
> + /* Register vring */
> + mem = rproc_mem_entry_init(dev, NULL,
> + (dma_addr_t)rmem->base,
> + rmem->size, rmem->base,
> + zynqmp_r5_rproc_mem_alloc,
> + zynqmp_r5_rproc_mem_release,
> + name);
> + } else {
> + /* Register DMA region */
> + int (*alloc)(struct rproc *r,
> + struct rproc_mem_entry *rme);
> + int (*release)(struct rproc *r,
> + struct rproc_mem_entry *rme);
> + char name[20];
> +
> + if (strstr(node->name, "vdev0buffer")) {
> + alloc = NULL;
> + release = NULL;
> + strcpy(name, "vdev0buffer");
> + } else {
> + alloc = zynqmp_r5_rproc_mem_alloc;
> + release = zynqmp_r5_rproc_mem_release;
> + strcpy(name, node->name);
> + }
> +
> + mem = rproc_mem_entry_init(dev, NULL,
> + (dma_addr_t)rmem->base,
> + rmem->size, rmem->base,
> + alloc, release, name);
> + }
> + if (!mem)
> + return -ENOMEM;
> +
> + rproc_add_carveout(rproc, mem);
> + }
> +
> + return 0;
> +}
> +
> +/*
> + * zynqmp_r5_pm_request_tcm
> + * @addr: base address of mem provided in R5 core's sram property.
> + *
> + * Given sram base address, determine its corresponding Xilinx
> + * Platform Management ID and then request access to this node
> + * so that it can be power up.
> + *
> + * return 0 on success, otherwise non-zero value on failure
> + */
> +static int zynqmp_r5_pm_request_sram(phys_addr_t addr)
> +{
> + unsigned int i;
> +
> + for (i = 0; i < NUM_SRAMS; i++) {
> + if (zynqmp_banks[i].addr == addr)
> + return zynqmp_pm_request_node(zynqmp_banks[i].id,
> + ZYNQMP_PM_CAPABILITY_ACCESS,
> + 0,
> + ZYNQMP_PM_REQUEST_ACK_BLOCKING);
> + }
> +
> + return -EINVAL;
> +}
> +
> +/*
> + * tcm_mem_alloc
> + * @rproc: single R5 core's corresponding rproc instance
> + * @mem: mem entry to initialize the va and da fields of
> + *
> + * Given TCM bank entry,
> + * this callback will set device address for R5 running on TCM
> + * and also setup virtual address for TCM bank remoteproc carveout
> + *
> + * return 0 on success, otherwise non-zero value on failure
> + */
> +static int tcm_mem_alloc(struct rproc *rproc,
> + struct rproc_mem_entry *mem)
> +{
> + void *va;
> + struct device *dev = rproc->dev.parent;
> +
> + va = ioremap_wc(mem->dma, mem->len);
> + if (IS_ERR_OR_NULL(va))
> + return -ENOMEM;
> +
> + /* Update memory entry va */
> + mem->va = va;
> +
> + va = devm_ioremap_wc(dev, mem->da, mem->len);
> + if (!va)
> + return -ENOMEM;
> + /* As R5 is 32 bit, wipe out extra high bits */
> + mem->da &= 0x000fffff;
> + /*
> + * The R5s expect their TCM banks to be at address 0x0 and 0x2000,
> + * while on the Linux side they are at 0xffexxxxx. Zero out the high
> + * 12 bits of the address.
> + */
> +
> + /*
> + * TCM Banks 1A and 1B (0xffe90000 and 0xffeb0000) still
> + * need to be translated to 0x0 and 0x20000
> + */
> + if (mem->da == 0x90000 || mem->da == 0xB0000)
> + mem->da -= 0x90000;
> +
> + /* if translated TCM bank address is not valid report error */
> + if (mem->da != 0x0 && mem->da != 0x20000) {
> + dev_err(dev, "invalid TCM bank address: %x\n", mem->da);
> + return -EINVAL;
> + }
> +
> + return 0;
> +}
> +
> +/*
> + * parse_tcm_banks()
> + * @rproc: single R5 core's corresponding rproc instance
> + *
> + * Given R5 node in remoteproc instance
> + * allocate remoteproc carveout for TCM memory
> + * needed for firmware to be loaded
> + *
> + * return 0 on success, otherwise non-zero value on failure
> + */
> +static int parse_tcm_banks(struct rproc *rproc)
> +{
> + int i, num_banks;
> + struct zynqmp_r5_rproc *z_rproc = rproc->priv;
> + struct device *dev = &rproc->dev;
> + struct device_node *r5_node = z_rproc->dev->of_node;
> +
> + /* go through TCM banks for r5 node */
> + num_banks = of_count_phandle_with_args(r5_node, BANK_LIST_PROP, NULL);
> + if (num_banks <= 0) {
> + dev_err(dev, "need to specify TCM banks\n");
> + return -EINVAL;
> + }
> + for (i = 0; i < num_banks; i++) {
> + struct resource rsc;
> + resource_size_t size;
> + struct device_node *dt_node;
> + struct rproc_mem_entry *mem;
> + int ret;
> + u32 pnode_id; /* zynqmp_pm* fn's expect u32 */
> +
> + dt_node = of_parse_phandle(r5_node, BANK_LIST_PROP, i);
> + if (!dt_node)
> + return -EINVAL;
> +
> + if (of_device_is_available(dt_node)) {
> + ret = of_address_to_resource(dt_node, 0, &rsc);
> + if (ret < 0)
> + return ret;
> + ret = zynqmp_r5_pm_request_sram(rsc.start);
> + if (ret < 0)
> + return ret;
> +
> + /* add carveout */
> + size = resource_size(&rsc);
> + mem = rproc_mem_entry_init(dev, NULL, rsc.start,
> + (int)size, rsc.start,
> + tcm_mem_alloc,
> + tcm_mem_release,
> + rsc.name);
> + if (!mem)
> + return -ENOMEM;
> +
> + mem->priv = (void *)(u64)pnode_id;
> + rproc_add_carveout(rproc, mem);
> + }
> + }
> +
> + return 0;
> +}
> +
> +/*
> + * zynqmp_r5_parse_fw()
> + * @rproc: single R5 core's corresponding rproc instance
> + * @fw: ptr to firmware to be loaded onto r5 core
> + *
> + * When loading firmware, ensure the necessary carveouts are in remoteproc
> + *
> + * return 0 on success, otherwise non-zero value on failure
> + */
> +static int zynqmp_r5_parse_fw(struct rproc *rproc, const struct firmware *fw)
> +{
> + int ret;
> +
> + ret = parse_tcm_banks(rproc);
> + if (ret)
> + return ret;
> +
> + ret = parse_mem_regions(rproc);
> + if (ret)
> + return ret;
> +
> + ret = rproc_elf_load_rsc_table(rproc, fw);
> + if (ret == -EINVAL) {
> + /*
> + * resource table only required for IPC.
> + * if not present, this is not necessarily an error;
> + * for example, loading r5 hello world application
> + * so simply inform user and keep going.
> + */
> + dev_info(&rproc->dev, "no resource table found.\n");
> + ret = 0;
> + }
> + return ret;
> +}
> +
> +/*
> + * zynqmp_r5_rproc_kick() - kick a firmware if mbox is provided
> + * @rproc: r5 core's corresponding rproc structure
> + * @vqid: virtqueue ID
> + */
> +static void zynqmp_r5_rproc_kick(struct rproc *rproc, int vqid)
> +{
> + struct sk_buff *skb;
> + unsigned int skb_len;
> + struct zynqmp_ipi_message *mb_msg;
> + int ret;
> +
> + struct device *dev = rproc->dev.parent;
> + struct zynqmp_r5_rproc *z_rproc = rproc->priv;
> +
struct zynqmp_r5_rproc *z_rproc = rproc->priv;
struct device *dev = rproc->dev.parent;
struct zynqmp_ipi_message *mb_msg;
unsigned int skb_len;
struct sk_buff *skb;
int ret;
> + if (of_property_read_bool(dev->of_node, "mboxes")) {
Checking for a mailbox property here is weird - I will come back to this later.
> + skb_len = (unsigned int)(sizeof(vqid) + sizeof(mb_msg));
> + skb = alloc_skb(skb_len, GFP_ATOMIC);
GFP_ATOMIC?
> + if (!skb)
> + return;
> +
> + mb_msg = (struct zynqmp_ipi_message *)skb_put(skb, skb_len);
> + mb_msg->len = sizeof(vqid);
> + memcpy(mb_msg->data, &vqid, sizeof(vqid));
> +
> + skb_queue_tail(&z_rproc->tx_mc_skbs, skb);
> + ret = mbox_send_message(z_rproc->tx_chan, mb_msg);
> + if (ret < 0) {
> + dev_warn(dev, "Failed to kick remote.\n");
> + skb_dequeue_tail(&z_rproc->tx_mc_skbs);
> + kfree_skb(skb);
> + }
> + } else {
> + (void)skb;
> + (void)skb_len;
> + (void)mb_msg;
> + (void)ret;
> + (void)vqid;
What is that?
> + }
> +}
> +
> +static struct rproc_ops zynqmp_r5_rproc_ops = {
> + .start = zynqmp_r5_rproc_start,
> + .stop = zynqmp_r5_rproc_stop,
> + .load = rproc_elf_load_segments,
> + .parse_fw = zynqmp_r5_parse_fw,
> + .find_loaded_rsc_table = rproc_elf_find_loaded_rsc_table,
> + .sanity_check = rproc_elf_sanity_check,
> + .get_boot_addr = rproc_elf_get_boot_addr,
> + .kick = zynqmp_r5_rproc_kick,
> +};
> +
> +/**
> + * event_notified_idr_cb() - event notified idr callback
> + * @id: idr id
> + * @ptr: pointer to idr private data
> + * @data: data passed to idr_for_each callback
> + *
> + * Pass notification to remoteproc virtio
> + *
> + * Return: 0. having return is to satisfy the idr_for_each() function
> + * pointer input argument requirement.
> + **/
> +static int event_notified_idr_cb(int id, void *ptr, void *data)
> +{
> + struct rproc *rproc = data;
> +
> + (void)rproc_vq_interrupt(rproc, id);
You really don't want to know if this fails? Not event a dev_dbg()?
> + return 0;
> +}
> +
> +/**
> + * handle_event_notified() - remoteproc notification work function
> + * @work: pointer to the work structure
> + *
> + * It checks each registered remoteproc notify IDs.
> + */
> +static void handle_event_notified(struct work_struct *work)
> +{
> + struct rproc *rproc;
> + struct zynqmp_r5_rproc *z_rproc;
> +
> + z_rproc = container_of(work, struct zynqmp_r5_rproc, mbox_work);
> +
> + (void)mbox_send_message(z_rproc->rx_chan, NULL);
What is going on here? Is this some kind of ACK needed by the IPI protocol? It
would be interesting to have a comment that explains the logic.
> + rproc = z_rproc->rproc;
> + /*
> + * We only use IPI for interrupt. The firmware side may or may
> + * not write the notifyid when it trigger IPI.
> + * And thus, we scan through all the registered notifyids.
> + */
> + idr_for_each(&rproc->notifyids, event_notified_idr_cb, rproc);
> +}
> +
> +/**
> + * zynqmp_r5_mb_rx_cb() - Receive channel mailbox callback
> + * @cl: mailbox client
> + * @msg: message pointer
> + *
> + * It will schedule the R5 notification work.
> + */
> +static void zynqmp_r5_mb_rx_cb(struct mbox_client *cl, void *msg)
> +{
> + struct zynqmp_r5_rproc *z_rproc;
> +
> + z_rproc = container_of(cl, struct zynqmp_r5_rproc, rx_mc);
> + if (msg) {
It is possible to received a notification without a message?
> + struct zynqmp_ipi_message *ipi_msg, *buf_msg;
> + size_t len;
> +
> + ipi_msg = (struct zynqmp_ipi_message *)msg;
> + buf_msg = (struct zynqmp_ipi_message *)z_rproc->rx_mc_buf;
> + len = (ipi_msg->len >= IPI_BUF_LEN_MAX) ?
> + IPI_BUF_LEN_MAX : ipi_msg->len;
> + buf_msg->len = len;
> + memcpy(buf_msg->data, ipi_msg->data, len);
> + }
> + schedule_work(&z_rproc->mbox_work);
And when that happens the work function still needs to execute?
I will continue tomorrow...
> +}
> +
> +/**
> + * zynqmp_r5_mb_tx_done() - Request has been sent to the remote
> + * @cl: mailbox client
> + * @msg: pointer to the message which has been sent
> + * @r: status of last TX - OK or error
> + *
> + * It will be called by the mailbox framework when the last TX has done.
> + */
> +static void zynqmp_r5_mb_tx_done(struct mbox_client *cl, void *msg, int r)
> +{
> + struct zynqmp_r5_rproc *z_rproc;
> + struct sk_buff *skb;
> +
> + if (!msg)
> + return;
> + z_rproc = container_of(cl, struct zynqmp_r5_rproc, tx_mc);
> + skb = skb_dequeue(&z_rproc->tx_mc_skbs);
> + kfree_skb(skb);
> +}
> +
> +/**
> + * zynqmp_r5_setup_mbox() - Setup mailboxes
> + * this is used for each individual R5 core
> + *
> + * @z_rproc: pointer to the ZynqMP R5 processor platform data
> + * @node: pointer of the device node
> + *
> + * Function to setup mailboxes to talk to RPU.
> + *
> + * Return: 0 for success, negative value for failure.
> + */
> +static int zynqmp_r5_setup_mbox(struct zynqmp_r5_rproc *z_rproc,
> + struct device_node *node)
> +{
> + struct mbox_client *mclient;
> +
> + /* Setup TX mailbox channel client */
> + mclient = &z_rproc->tx_mc;
> + mclient->rx_callback = NULL;
> + mclient->tx_block = false;
> + mclient->knows_txdone = false;
> + mclient->tx_done = zynqmp_r5_mb_tx_done;
> + mclient->dev = z_rproc->dev;
> +
> + /* Setup TX mailbox channel client */
> + mclient = &z_rproc->rx_mc;
> + mclient->dev = z_rproc->dev;
> + mclient->rx_callback = zynqmp_r5_mb_rx_cb;
> + mclient->tx_block = false;
> + mclient->knows_txdone = false;
> +
> + INIT_WORK(&z_rproc->mbox_work, handle_event_notified);
> +
> + /* Request TX and RX channels */
> + z_rproc->tx_chan = mbox_request_channel_byname(&z_rproc->tx_mc, "tx");
> + if (IS_ERR(z_rproc->tx_chan)) {
> + dev_err(z_rproc->dev, "failed to request mbox tx channel.\n");
> + z_rproc->tx_chan = NULL;
> + return -EINVAL;
> + }
> +
> + z_rproc->rx_chan = mbox_request_channel_byname(&z_rproc->rx_mc, "rx");
> + if (IS_ERR(z_rproc->rx_chan)) {
> + dev_err(z_rproc->dev, "failed to request mbox rx channel.\n");
> + z_rproc->rx_chan = NULL;
> + return -EINVAL;
> + }
> + skb_queue_head_init(&z_rproc->tx_mc_skbs);
> +
> + return 0;
> +}
> +
> +/**
> + * zynqmp_r5_probe() - Probes ZynqMP R5 processor device node
> + * this is called for each individual R5 core to
> + * set up mailbox, Xilinx platform manager unique ID,
> + * add to rproc core
> + *
> + * @pdev: domain platform device for current R5 core
> + * @node: pointer of the device node for current R5 core
> + * @rpu_mode: mode to configure RPU, split or lockstep
> + * @z_rproc: Xilinx specific remoteproc structure used later to link
> + * in to cluster of cores
> + *
> + * Return: 0 for success, negative value for failure.
> + */
> +static int zynqmp_r5_probe(struct platform_device *pdev,
> + struct device_node *node,
> + enum rpu_oper_mode rpu_mode,
> + struct zynqmp_r5_rproc **z_rproc)
> +{
> + int ret;
> + struct device *dev = &pdev->dev;
> + struct rproc *rproc_ptr;
> +
> + /* Allocate remoteproc instance */
> + rproc_ptr = devm_rproc_alloc(dev, dev_name(dev), &zynqmp_r5_rproc_ops,
> + NULL, sizeof(struct zynqmp_r5_rproc));
> + if (!rproc_ptr) {
> + ret = -ENOMEM;
> + goto error;
> + }
> +
> + rproc_ptr->auto_boot = false;
> + *z_rproc = rproc_ptr->priv;
> + (*z_rproc)->rproc = rproc_ptr;
> + (*z_rproc)->dev = dev;
> + /* Set up DMA mask */
> + ret = dma_set_coherent_mask(dev, DMA_BIT_MASK(32));
> + if (ret)
> + goto error;
> +
> + /* Get R5 power domain node */
> + ret = of_property_read_u32(node, "power-domain", &(*z_rproc)->pnode_id);
> + if (ret)
> + goto error;
> +
> + ret = r5_set_mode(*z_rproc, rpu_mode);
> + if (ret)
> + goto error;
> +
> + if (of_property_read_bool(node, "mboxes")) {
> + ret = zynqmp_r5_setup_mbox(*z_rproc, node);
> + if (ret)
> + goto error;
> + }
> +
> + /* Add R5 remoteproc */
> + ret = devm_rproc_add(dev, rproc_ptr);
> + if (ret)
> + goto error;
> +
> + return 0;
> +error:
> + *z_rproc = NULL;
> + return ret;
> +}
> +
> +/*
> + * zynqmp_r5_remoteproc_probe()
> + *
> + * @pdev: domain platform device for R5 cluster
> + *
> + * called when driver is probed, for each R5 core specified in DT,
> + * setup as needed to do remoteproc-related operations
> + *
> + * Return: 0 for success, negative value for failure.
> + */
> +static int zynqmp_r5_remoteproc_probe(struct platform_device *pdev)
> +{
> + int ret, core_count;
> + struct device *dev = &pdev->dev;
> + struct device_node *nc;
> + enum rpu_oper_mode rpu_mode = PM_RPU_MODE_LOCKSTEP;
> + struct list_head *cluster; /* list to track each core's rproc */
> + struct zynqmp_r5_rproc *z_rproc;
> + struct platform_device *child_pdev;
> + struct list_head *pos;
> +
> + ret = of_property_read_u32(dev->of_node, "xlnx,cluster-mode", &rpu_mode);
> + if (ret < 0 || (rpu_mode != PM_RPU_MODE_LOCKSTEP &&
> + rpu_mode != PM_RPU_MODE_SPLIT)) {
> + dev_err(dev, "invalid format cluster mode: ret %d mode %x\n",
> + ret, rpu_mode);
> + return ret;
> + }
> +
> + dev_dbg(dev, "RPU configuration: %s\n",
> + rpu_mode == PM_RPU_MODE_LOCKSTEP ? "lockstep" : "split");
> +
> + /*
> + * if 2 RPUs provided but one is lockstep, then we have an
> + * invalid configuration.
> + */
> +
> + core_count = of_get_available_child_count(dev->of_node);
> + if ((rpu_mode == PM_RPU_MODE_LOCKSTEP && core_count != 1) ||
> + core_count > MAX_RPROCS)
> + return -EINVAL;
> +
> + cluster = devm_kzalloc(dev, sizeof(*cluster), GFP_KERNEL);
> + if (!cluster)
> + return -ENOMEM;
> + INIT_LIST_HEAD(cluster);
> +
> + ret = devm_of_platform_populate(dev);
> + if (ret) {
> + dev_err(dev, "devm_of_platform_populate failed, ret = %d\n",
> + ret);
> + return ret;
> + }
> +
> + /* probe each individual r5 core's remoteproc-related info */
> + for_each_available_child_of_node(dev->of_node, nc) {
> + child_pdev = of_find_device_by_node(nc);
> + if (!child_pdev) {
> + dev_err(dev, "could not get R5 core platform device\n");
> + ret = -ENODEV;
> + goto out;
> + }
> +
> + ret = zynqmp_r5_probe(child_pdev, nc, rpu_mode, &z_rproc);
> + dev_dbg(dev, "%s to probe rpu %pOF\n",
> + ret ? "Failed" : "Able",
> + nc);
> + if (!z_rproc)
> + ret = -EINVAL;
> + if (ret)
> + goto out;
> + list_add_tail(&z_rproc->elem, cluster);
> + }
> + /* wire in so each core can be cleaned up at driver remove */
> + platform_set_drvdata(pdev, cluster);
> + return 0;
> +out:
> + /*
> + * undo core0 upon any failures on core1 in split-mode
> + *
> + * in zynqmp_r5_probe z_rproc is set to null
> + * and ret to non-zero value if error
> + */
> + if (ret && !z_rproc && rpu_mode == PM_RPU_MODE_SPLIT &&
> + !list_empty(cluster)) {
> + list_for_each(pos, cluster) {
> + z_rproc = list_entry(pos, struct zynqmp_r5_rproc, elem);
> + if (of_property_read_bool(z_rproc->dev->of_node, "mboxes")) {
> + mbox_free_channel(z_rproc->tx_chan);
> + mbox_free_channel(z_rproc->rx_chan);
> + }
> + }
> + }
> + return ret;
> +}
> +
> +/*
> + * zynqmp_r5_remoteproc_remove()
> + *
> + * @pdev: domain platform device for R5 cluster
> + *
> + * When the driver is unloaded, clean up the mailboxes for each
> + * remoteproc that was initially probed.
> + */
> +static int zynqmp_r5_remoteproc_remove(struct platform_device *pdev)
> +{
> + struct list_head *pos, *temp, *cluster = (struct list_head *)
> + platform_get_drvdata(pdev);
> + struct zynqmp_r5_rproc *z_rproc = NULL;
> +
> + list_for_each_safe(pos, temp, cluster) {
> + z_rproc = list_entry(pos, struct zynqmp_r5_rproc, elem);
> + if (of_property_read_bool(z_rproc->dev->of_node, "mboxes")) {
> + mbox_free_channel(z_rproc->tx_chan);
> + mbox_free_channel(z_rproc->rx_chan);
> + }
> + list_del(pos);
> + }
> + return 0;
> +}
> +
> +/* Match table for OF platform binding */
> +static const struct of_device_id zynqmp_r5_remoteproc_match[] = {
> + { .compatible = "xlnx,zynqmp-r5-remoteproc", },
> + { /* end of list */ },
> +};
> +MODULE_DEVICE_TABLE(of, zynqmp_r5_remoteproc_match);
> +
> +static struct platform_driver zynqmp_r5_remoteproc_driver = {
> + .probe = zynqmp_r5_remoteproc_probe,
> + .remove = zynqmp_r5_remoteproc_remove,
> + .driver = {
> + .name = "zynqmp_r5_remoteproc",
> + .of_match_table = zynqmp_r5_remoteproc_match,
> + },
> +};
> +module_platform_driver(zynqmp_r5_remoteproc_driver);
> +
> +MODULE_AUTHOR("Ben Levinsky <ben.levinsky@xilinx.com>");
> +MODULE_LICENSE("GPL v2");
> --
> 2.17.1
>
^ permalink raw reply [flat|nested] 13+ messages in thread* Re: [PATCH v24 5/5] remoteproc: Add initial zynqmp R5 remoteproc driver
2020-11-30 15:57 ` [PATCH v24 5/5] remoteproc: Add initial zynqmp R5 remoteproc driver Ben Levinsky
2020-12-07 21:08 ` Mathieu Poirier
2020-12-08 16:48 ` Mathieu Poirier
@ 2020-12-09 19:00 ` Mathieu Poirier
2020-12-10 20:56 ` Mathieu Poirier
3 siblings, 0 replies; 13+ messages in thread
From: Mathieu Poirier @ 2020-12-09 19:00 UTC (permalink / raw)
To: Ben Levinsky; +Cc: devicetree, linux-remoteproc, linux-kernel, linux-arm-kernel
On Mon, Nov 30, 2020 at 07:57:17AM -0800, Ben Levinsky wrote:
> R5 is included in Xilinx Zynq UltraScale MPSoC so by adding this
> remotproc driver, we can boot the R5 sub-system in two different
> configurations -
> * Split
> * Lockstep
>
> The Xilinx R5 Remoteproc Driver boots the R5's via calls to the Xilinx
> Platform Management Unit that handles the R5 configuration, memory access
> and R5 lifecycle management. The interface to this manager is done in this
> driver via zynqmp_pm_* function calls.
>
> Signed-off-by: Wendy Liang <wendy.liang@xilinx.com>
> Signed-off-by: Michal Simek <michal.simek@xilinx.com>
> Signed-off-by: Ed Mooring <ed.mooring@xilinx.com>
> Signed-off-by: Jason Wu <j.wu@xilinx.com>
> Signed-off-by: Ben Levinsky <ben.levinsky@xilinx.com>
> ---
> drivers/remoteproc/Kconfig | 8 +
> drivers/remoteproc/Makefile | 1 +
> drivers/remoteproc/zynqmp_r5_remoteproc.c | 872 ++++++++++++++++++++++
> 3 files changed, 881 insertions(+)
> create mode 100644 drivers/remoteproc/zynqmp_r5_remoteproc.c
>
> diff --git a/drivers/remoteproc/Kconfig b/drivers/remoteproc/Kconfig
> index c6659dfea7c7..c2fe54b1d94f 100644
> --- a/drivers/remoteproc/Kconfig
> +++ b/drivers/remoteproc/Kconfig
> @@ -275,6 +275,14 @@ config TI_K3_DSP_REMOTEPROC
> It's safe to say N here if you're not interested in utilizing
> the DSP slave processors.
>
> +config ZYNQMP_R5_REMOTEPROC
> + tristate "ZynqMP R5 remoteproc support"
> + depends on PM && ARCH_ZYNQMP
> + select RPMSG_VIRTIO
> + select ZYNQMP_IPI_MBOX
> + help
> + Say y or m here to support ZynqMP R5 remote processors via the remote
> + processor framework.
> endif # REMOTEPROC
>
> endmenu
> diff --git a/drivers/remoteproc/Makefile b/drivers/remoteproc/Makefile
> index 3dfa28e6c701..ef1abff654c2 100644
> --- a/drivers/remoteproc/Makefile
> +++ b/drivers/remoteproc/Makefile
> @@ -33,3 +33,4 @@ obj-$(CONFIG_ST_REMOTEPROC) += st_remoteproc.o
> obj-$(CONFIG_ST_SLIM_REMOTEPROC) += st_slim_rproc.o
> obj-$(CONFIG_STM32_RPROC) += stm32_rproc.o
> obj-$(CONFIG_TI_K3_DSP_REMOTEPROC) += ti_k3_dsp_remoteproc.o
> +obj-$(CONFIG_ZYNQMP_R5_REMOTEPROC) += zynqmp_r5_remoteproc.o
> diff --git a/drivers/remoteproc/zynqmp_r5_remoteproc.c b/drivers/remoteproc/zynqmp_r5_remoteproc.c
> new file mode 100644
> index 000000000000..2593de618409
> --- /dev/null
> +++ b/drivers/remoteproc/zynqmp_r5_remoteproc.c
> @@ -0,0 +1,872 @@
> +// SPDX-License-Identifier: GPL-2.0
> +/*
> + * Zynq R5 Remote Processor driver
> + *
> + * Based on origin OMAP and Zynq Remote Processor driver
> + *
> + */
> +
> +#include <linux/firmware/xlnx-zynqmp.h>
> +#include <linux/interrupt.h>
> +#include <linux/kernel.h>
> +#include <linux/list.h>
> +#include <linux/mailbox_client.h>
> +#include <linux/mailbox/zynqmp-ipi-message.h>
> +#include <linux/module.h>
> +#include <linux/of_address.h>
> +#include <linux/of_platform.h>
> +#include <linux/of_reserved_mem.h>
> +#include <linux/platform_device.h>
> +#include <linux/remoteproc.h>
> +#include <linux/skbuff.h>
> +#include <linux/sysfs.h>
> +
> +#include "remoteproc_internal.h"
> +
> +#define MAX_RPROCS 2 /* Support up to 2 RPU */
> +#define MAX_MEM_PNODES 4 /* Max power nodes for one RPU memory instance */
> +
> +#define BANK_LIST_PROP "sram"
> +#define DDR_LIST_PROP "memory-region"
> +
> +/* IPI buffer MAX length */
> +#define IPI_BUF_LEN_MAX 32U
> +/* RX mailbox client buffer max length */
> +#define RX_MBOX_CLIENT_BUF_MAX (IPI_BUF_LEN_MAX + \
> + sizeof(struct zynqmp_ipi_message))
> +
> +/*
> + * Map each Xilinx on-chip SRAM Bank address to their own respective
> + * pm_node_id.
> + */
> +struct sram_addr_data {
> + phys_addr_t addr;
> + enum pm_node_id id;
> +};
> +
> +#define NUM_SRAMS 4U
> +static const struct sram_addr_data zynqmp_banks[NUM_SRAMS] = {
> + {0xffe00000UL, NODE_TCM_0_A},
> + {0xffe20000UL, NODE_TCM_0_B},
> + {0xffe90000UL, NODE_TCM_1_A},
> + {0xffeb0000UL, NODE_TCM_1_B},
> +};
> +
> +/**
> + * struct zynqmp_r5_rproc - ZynqMP R5 core structure
> + *
> + * @rx_mc_buf: rx mailbox client buffer to save the rx message
> + * @tx_mc: tx mailbox client
> + * @rx_mc: rx mailbox client
> + * @mbox_work: mbox_work for the RPU remoteproc
> + * @tx_mc_skbs: socket buffers for tx mailbox client
> + * @dev: device of RPU instance
> + * @rproc: rproc handle
> + * @tx_chan: tx mailbox channel
> + * @rx_chan: rx mailbox channel
> + * @pnode_id: RPU CPU power domain id
> + * @elem: linked list item
> + */
> +struct zynqmp_r5_rproc {
> + unsigned char rx_mc_buf[RX_MBOX_CLIENT_BUF_MAX];
> + struct mbox_client tx_mc;
> + struct mbox_client rx_mc;
> + struct work_struct mbox_work;
> + struct sk_buff_head tx_mc_skbs;
> + struct device *dev;
> + struct rproc *rproc;
> + struct mbox_chan *tx_chan;
> + struct mbox_chan *rx_chan;
> + u32 pnode_id;
> + struct list_head elem;
> +};
> +
> +/*
> + * r5_set_mode - set RPU operation mode
> + * @z_rproc: Remote processor private data
> + * @rpu_mode: mode specified by device tree to configure the RPU to
> + *
> + * set RPU operation mode
> + *
> + * Return: 0 for success, negative value for failure
> + */
> +static int r5_set_mode(struct zynqmp_r5_rproc *z_rproc,
> + enum rpu_oper_mode rpu_mode)
> +{
> + enum rpu_tcm_comb tcm_mode;
> + enum rpu_oper_mode cur_rpu_mode;
> + int ret;
> +
> + ret = zynqmp_pm_get_rpu_mode(z_rproc->pnode_id, &cur_rpu_mode);
> + if (ret < 0)
> + return ret;
> +
> + if (rpu_mode != cur_rpu_mode) {
> + ret = zynqmp_pm_set_rpu_mode(z_rproc->pnode_id,
> + rpu_mode);
> + if (ret < 0)
> + return ret;
> + }
> +
> + tcm_mode = (rpu_mode == PM_RPU_MODE_LOCKSTEP) ?
> + PM_RPU_TCM_COMB : PM_RPU_TCM_SPLIT;
> + return zynqmp_pm_set_tcm_config(z_rproc->pnode_id, tcm_mode);
> +}
> +
> +/*
> + * zynqmp_r5_rproc_mem_release
> + * @rproc: single R5 core's corresponding rproc instance
> + * @mem: mem entry to unmap
> + *
> + * Unmap TCM banks when powering down R5 core.
> + *
> + * return 0 on success, otherwise non-zero value on failure
> + */
> +static int tcm_mem_release(struct rproc *rproc, struct rproc_mem_entry *mem)
> +{
> + u32 pnode_id = (u64)mem->priv;
> +
> + iounmap(mem->va);
> + return zynqmp_pm_release_node(pnode_id);
> +}
> +
> +/*
> + * zynqmp_r5_rproc_start
> + * @rproc: single R5 core's corresponding rproc instance
> + *
> + * Start R5 Core from designated boot address.
> + *
> + * return 0 on success, otherwise non-zero value on failure
> + */
> +static int zynqmp_r5_rproc_start(struct rproc *rproc)
> +{
> + struct zynqmp_r5_rproc *z_rproc = rproc->priv;
> + enum rpu_boot_mem bootmem;
> +
> + bootmem = (rproc->bootaddr & 0xF0000000) == 0xF0000000 ?
> + PM_RPU_BOOTMEM_HIVEC : PM_RPU_BOOTMEM_LOVEC;
> +
> + dev_dbg(rproc->dev.parent, "RPU boot from %s.",
> + bootmem == PM_RPU_BOOTMEM_HIVEC ? "OCM" : "TCM");
> +
> + return zynqmp_pm_request_wake(z_rproc->pnode_id, 1,
> + bootmem, ZYNQMP_PM_REQUEST_ACK_NO);
> +}
> +
> +/*
> + * zynqmp_r5_rproc_stop
> + * @rproc: single R5 core's corresponding rproc instance
> + *
> + * Power down R5 Core.
> + *
> + * return 0 on success, otherwise non-zero value on failure
> + */
> +static int zynqmp_r5_rproc_stop(struct rproc *rproc)
> +{
> + struct zynqmp_r5_rproc *z_rproc = rproc->priv;
> +
> + return zynqmp_pm_force_pwrdwn(z_rproc->pnode_id,
> + ZYNQMP_PM_REQUEST_ACK_BLOCKING);
> +}
> +
> +/*
> + * zynqmp_r5_rproc_mem_alloc
> + * @rproc: single R5 core's corresponding rproc instance
> + * @mem: mem entry to map
> + *
> + * Callback to map va for memory-region's carveout.
> + *
> + * return 0 on success, otherwise non-zero value on failure
> + */
> +static int zynqmp_r5_rproc_mem_alloc(struct rproc *rproc,
> + struct rproc_mem_entry *mem)
> +{
> + void *va;
> +
> + va = ioremap_wc(mem->dma, mem->len);
> + if (IS_ERR_OR_NULL(va))
> + return -ENOMEM;
> +
> + mem->va = va;
> +
> + return 0;
> +}
> +
> +/*
> + * zynqmp_r5_rproc_mem_release
> + * @rproc: single R5 core's corresponding rproc instance
> + * @mem: mem entry to unmap
> + *
> + * Unmap memory-region carveout
> + *
> + * return 0 on success, otherwise non-zero value on failure
> + */
> +static int zynqmp_r5_rproc_mem_release(struct rproc *rproc,
> + struct rproc_mem_entry *mem)
> +{
> + iounmap(mem->va);
> + return 0;
> +}
> +
> +/*
> + * parse_mem_regions
> + * @rproc: single R5 core's corresponding rproc instance
> + *
> + * Construct rproc mem carveouts from carveout provided in
> + * memory-region property
> + *
> + * return 0 on success, otherwise non-zero value on failure
> + */
> +static int parse_mem_regions(struct rproc *rproc)
> +{
> + int num_mems, i;
> + struct zynqmp_r5_rproc *z_rproc = rproc->priv;
> + struct device *dev = &rproc->dev;
> + struct device_node *np = z_rproc->dev->of_node;
> + struct rproc_mem_entry *mem;
> +
> + num_mems = of_count_phandle_with_args(np, DDR_LIST_PROP, NULL);
> + if (num_mems <= 0)
> + return 0;
> +
> + for (i = 0; i < num_mems; i++) {
> + struct device_node *node;
> + struct reserved_mem *rmem;
> +
> + node = of_parse_phandle(np, DDR_LIST_PROP, i);
> + if (!node)
> + return -EINVAL;
> +
> + rmem = of_reserved_mem_lookup(node);
> + if (!rmem)
> + return -EINVAL;
> +
> + if (strstr(node->name, "vdev0vring")) {
> + int vring_id;
> + char name[16];
> +
> + /*
> + * expecting form of "rpuXvdev0vringX as documented
> + * in xilinx remoteproc device tree binding
> + */
> + if (strlen(node->name) < 15) {
> + dev_err(dev, "%pOF is less than 14 chars",
> + node);
> + return -EINVAL;
> + }
> +
> + /*
> + * can be 1 of multiple vring IDs per IPC channel
> + * e.g. 'vdev0vring0' and 'vdev0vring1'
> + */
> + vring_id = node->name[14] - '0';
> + snprintf(name, sizeof(name), "vdev0vring%d", vring_id);
> + /* Register vring */
> + mem = rproc_mem_entry_init(dev, NULL,
> + (dma_addr_t)rmem->base,
> + rmem->size, rmem->base,
> + zynqmp_r5_rproc_mem_alloc,
> + zynqmp_r5_rproc_mem_release,
> + name);
> + } else {
> + /* Register DMA region */
> + int (*alloc)(struct rproc *r,
> + struct rproc_mem_entry *rme);
> + int (*release)(struct rproc *r,
> + struct rproc_mem_entry *rme);
> + char name[20];
> +
> + if (strstr(node->name, "vdev0buffer")) {
> + alloc = NULL;
> + release = NULL;
> + strcpy(name, "vdev0buffer");
> + } else {
> + alloc = zynqmp_r5_rproc_mem_alloc;
> + release = zynqmp_r5_rproc_mem_release;
> + strcpy(name, node->name);
> + }
> +
> + mem = rproc_mem_entry_init(dev, NULL,
> + (dma_addr_t)rmem->base,
> + rmem->size, rmem->base,
> + alloc, release, name);
> + }
> + if (!mem)
> + return -ENOMEM;
> +
> + rproc_add_carveout(rproc, mem);
> + }
> +
> + return 0;
> +}
> +
> +/*
> + * zynqmp_r5_pm_request_tcm
> + * @addr: base address of mem provided in R5 core's sram property.
> + *
> + * Given sram base address, determine its corresponding Xilinx
> + * Platform Management ID and then request access to this node
> + * so that it can be power up.
> + *
> + * return 0 on success, otherwise non-zero value on failure
> + */
> +static int zynqmp_r5_pm_request_sram(phys_addr_t addr)
> +{
> + unsigned int i;
> +
unsigned int?
> + for (i = 0; i < NUM_SRAMS; i++) {
> + if (zynqmp_banks[i].addr == addr)
> + return zynqmp_pm_request_node(zynqmp_banks[i].id,
> + ZYNQMP_PM_CAPABILITY_ACCESS,
> + 0,
> + ZYNQMP_PM_REQUEST_ACK_BLOCKING);
> + }
> +
for (i = 0; i < NUM_SRAMS; i++) {
if (zynqmp_banks[i].addr != addr)
continue;
return zynqmp_pm_request_node(zynqmp_banks[i].id,
ZYNQMP_PM_CAPABILITY_ACCESS, 0,
ZYNQMP_PM_REQUEST_ACK_BLOCKING);
}
> + return -EINVAL;
> +}
> +
> +/*
> + * tcm_mem_alloc
> + * @rproc: single R5 core's corresponding rproc instance
> + * @mem: mem entry to initialize the va and da fields of
> + *
> + * Given TCM bank entry,
Line break in the middle of a sentence.
> + * this callback will set device address for R5 running on TCM
> + * and also setup virtual address for TCM bank remoteproc carveout
> + *
> + * return 0 on success, otherwise non-zero value on failure
> + */
> +static int tcm_mem_alloc(struct rproc *rproc,
> + struct rproc_mem_entry *mem)
> +{
> + void *va;
> + struct device *dev = rproc->dev.parent;
> +
> + va = ioremap_wc(mem->dma, mem->len);
> + if (IS_ERR_OR_NULL(va))
> + return -ENOMEM;
> +
> + /* Update memory entry va */
> + mem->va = va;
> +
> + va = devm_ioremap_wc(dev, mem->da, mem->len);
What does that do?
> + if (!va)
> + return -ENOMEM;
> + /* As R5 is 32 bit, wipe out extra high bits */
> + mem->da &= 0x000fffff;
And yet only 20 bits are kept...
> + /*
> + * The R5s expect their TCM banks to be at address 0x0 and 0x2000,
> + * while on the Linux side they are at 0xffexxxxx. Zero out the high
> + * 12 bits of the address.
> + */
> +
> + /*
> + * TCM Banks 1A and 1B (0xffe90000 and 0xffeb0000) still
> + * need to be translated to 0x0 and 0x20000
> + */
> + if (mem->da == 0x90000 || mem->da == 0xB0000)
> + mem->da -= 0x90000;
> +
> + /* if translated TCM bank address is not valid report error */
> + if (mem->da != 0x0 && mem->da != 0x20000) {
> + dev_err(dev, "invalid TCM bank address: %x\n", mem->da);
> + return -EINVAL;
> + }
> +
> + return 0;
> +}
> +
> +/*
> + * parse_tcm_banks()
> + * @rproc: single R5 core's corresponding rproc instance
> + *
> + * Given R5 node in remoteproc instance
> + * allocate remoteproc carveout for TCM memory
> + * needed for firmware to be loaded
> + *
> + * return 0 on success, otherwise non-zero value on failure
> + */
> +static int parse_tcm_banks(struct rproc *rproc)
> +{
> + int i, num_banks;
> + struct zynqmp_r5_rproc *z_rproc = rproc->priv;
> + struct device *dev = &rproc->dev;
> + struct device_node *r5_node = z_rproc->dev->of_node;
> +
> + /* go through TCM banks for r5 node */
> + num_banks = of_count_phandle_with_args(r5_node, BANK_LIST_PROP, NULL);
> + if (num_banks <= 0) {
> + dev_err(dev, "need to specify TCM banks\n");
> + return -EINVAL;
> + }
> + for (i = 0; i < num_banks; i++) {
> + struct resource rsc;
> + resource_size_t size;
> + struct device_node *dt_node;
> + struct rproc_mem_entry *mem;
> + int ret;
> + u32 pnode_id; /* zynqmp_pm* fn's expect u32 */
> +
> + dt_node = of_parse_phandle(r5_node, BANK_LIST_PROP, i);
Please have a look at the documentation for of_parse_phandle(), especially the
part about the need to use of_node_put().
> + if (!dt_node)
> + return -EINVAL;
> +
> + if (of_device_is_available(dt_node)) {
> + ret = of_address_to_resource(dt_node, 0, &rsc);
Is there a scenario where SRAMs would be declared but explicitly marked as
unusable?
> + if (ret < 0)
> + return ret;
> + ret = zynqmp_r5_pm_request_sram(rsc.start);
> + if (ret < 0)
> + return ret;
> +
> + /* add carveout */
> + size = resource_size(&rsc);
> + mem = rproc_mem_entry_init(dev, NULL, rsc.start,
> + (int)size, rsc.start,
(int)size? I'm pretty sure that will make sparse unhappy. Speaking of which, I
advise to run the sparse checker on you set if you haven't already done so.
More comments to follow tomorrow.
> + tcm_mem_alloc,
> + tcm_mem_release,
> + rsc.name);
> + if (!mem)
> + return -ENOMEM;
> +
> + mem->priv = (void *)(u64)pnode_id;
> + rproc_add_carveout(rproc, mem);
> + }
> + }
> +
> + return 0;
> +}
> +
> +/*
> + * zynqmp_r5_parse_fw()
> + * @rproc: single R5 core's corresponding rproc instance
> + * @fw: ptr to firmware to be loaded onto r5 core
> + *
> + * When loading firmware, ensure the necessary carveouts are in remoteproc
> + *
> + * return 0 on success, otherwise non-zero value on failure
> + */
> +static int zynqmp_r5_parse_fw(struct rproc *rproc, const struct firmware *fw)
> +{
> + int ret;
> +
> + ret = parse_tcm_banks(rproc);
> + if (ret)
> + return ret;
> +
> + ret = parse_mem_regions(rproc);
> + if (ret)
> + return ret;
> +
> + ret = rproc_elf_load_rsc_table(rproc, fw);
> + if (ret == -EINVAL) {
> + /*
> + * resource table only required for IPC.
> + * if not present, this is not necessarily an error;
> + * for example, loading r5 hello world application
> + * so simply inform user and keep going.
> + */
> + dev_info(&rproc->dev, "no resource table found.\n");
> + ret = 0;
> + }
> + return ret;
> +}
> +
> +/*
> + * zynqmp_r5_rproc_kick() - kick a firmware if mbox is provided
> + * @rproc: r5 core's corresponding rproc structure
> + * @vqid: virtqueue ID
> + */
> +static void zynqmp_r5_rproc_kick(struct rproc *rproc, int vqid)
> +{
> + struct sk_buff *skb;
> + unsigned int skb_len;
> + struct zynqmp_ipi_message *mb_msg;
> + int ret;
> +
> + struct device *dev = rproc->dev.parent;
> + struct zynqmp_r5_rproc *z_rproc = rproc->priv;
> +
> + if (of_property_read_bool(dev->of_node, "mboxes")) {
> + skb_len = (unsigned int)(sizeof(vqid) + sizeof(mb_msg));
> + skb = alloc_skb(skb_len, GFP_ATOMIC);
> + if (!skb)
> + return;
> +
> + mb_msg = (struct zynqmp_ipi_message *)skb_put(skb, skb_len);
> + mb_msg->len = sizeof(vqid);
> + memcpy(mb_msg->data, &vqid, sizeof(vqid));
> +
> + skb_queue_tail(&z_rproc->tx_mc_skbs, skb);
> + ret = mbox_send_message(z_rproc->tx_chan, mb_msg);
> + if (ret < 0) {
> + dev_warn(dev, "Failed to kick remote.\n");
> + skb_dequeue_tail(&z_rproc->tx_mc_skbs);
> + kfree_skb(skb);
> + }
> + } else {
> + (void)skb;
> + (void)skb_len;
> + (void)mb_msg;
> + (void)ret;
> + (void)vqid;
> + }
> +}
> +
> +static struct rproc_ops zynqmp_r5_rproc_ops = {
> + .start = zynqmp_r5_rproc_start,
> + .stop = zynqmp_r5_rproc_stop,
> + .load = rproc_elf_load_segments,
> + .parse_fw = zynqmp_r5_parse_fw,
> + .find_loaded_rsc_table = rproc_elf_find_loaded_rsc_table,
> + .sanity_check = rproc_elf_sanity_check,
> + .get_boot_addr = rproc_elf_get_boot_addr,
> + .kick = zynqmp_r5_rproc_kick,
> +};
> +
> +/**
> + * event_notified_idr_cb() - event notified idr callback
> + * @id: idr id
> + * @ptr: pointer to idr private data
> + * @data: data passed to idr_for_each callback
> + *
> + * Pass notification to remoteproc virtio
> + *
> + * Return: 0. having return is to satisfy the idr_for_each() function
> + * pointer input argument requirement.
> + **/
> +static int event_notified_idr_cb(int id, void *ptr, void *data)
> +{
> + struct rproc *rproc = data;
> +
> + (void)rproc_vq_interrupt(rproc, id);
> + return 0;
> +}
> +
> +/**
> + * handle_event_notified() - remoteproc notification work function
> + * @work: pointer to the work structure
> + *
> + * It checks each registered remoteproc notify IDs.
> + */
> +static void handle_event_notified(struct work_struct *work)
> +{
> + struct rproc *rproc;
> + struct zynqmp_r5_rproc *z_rproc;
> +
> + z_rproc = container_of(work, struct zynqmp_r5_rproc, mbox_work);
> +
> + (void)mbox_send_message(z_rproc->rx_chan, NULL);
> + rproc = z_rproc->rproc;
> + /*
> + * We only use IPI for interrupt. The firmware side may or may
> + * not write the notifyid when it trigger IPI.
> + * And thus, we scan through all the registered notifyids.
> + */
> + idr_for_each(&rproc->notifyids, event_notified_idr_cb, rproc);
> +}
> +
> +/**
> + * zynqmp_r5_mb_rx_cb() - Receive channel mailbox callback
> + * @cl: mailbox client
> + * @msg: message pointer
> + *
> + * It will schedule the R5 notification work.
> + */
> +static void zynqmp_r5_mb_rx_cb(struct mbox_client *cl, void *msg)
> +{
> + struct zynqmp_r5_rproc *z_rproc;
> +
> + z_rproc = container_of(cl, struct zynqmp_r5_rproc, rx_mc);
> + if (msg) {
> + struct zynqmp_ipi_message *ipi_msg, *buf_msg;
> + size_t len;
> +
> + ipi_msg = (struct zynqmp_ipi_message *)msg;
> + buf_msg = (struct zynqmp_ipi_message *)z_rproc->rx_mc_buf;
> + len = (ipi_msg->len >= IPI_BUF_LEN_MAX) ?
> + IPI_BUF_LEN_MAX : ipi_msg->len;
> + buf_msg->len = len;
> + memcpy(buf_msg->data, ipi_msg->data, len);
> + }
> + schedule_work(&z_rproc->mbox_work);
> +}
> +
> +/**
> + * zynqmp_r5_mb_tx_done() - Request has been sent to the remote
> + * @cl: mailbox client
> + * @msg: pointer to the message which has been sent
> + * @r: status of last TX - OK or error
> + *
> + * It will be called by the mailbox framework when the last TX has done.
> + */
> +static void zynqmp_r5_mb_tx_done(struct mbox_client *cl, void *msg, int r)
> +{
> + struct zynqmp_r5_rproc *z_rproc;
> + struct sk_buff *skb;
> +
> + if (!msg)
> + return;
> + z_rproc = container_of(cl, struct zynqmp_r5_rproc, tx_mc);
> + skb = skb_dequeue(&z_rproc->tx_mc_skbs);
> + kfree_skb(skb);
> +}
> +
> +/**
> + * zynqmp_r5_setup_mbox() - Setup mailboxes
> + * this is used for each individual R5 core
> + *
> + * @z_rproc: pointer to the ZynqMP R5 processor platform data
> + * @node: pointer of the device node
> + *
> + * Function to setup mailboxes to talk to RPU.
> + *
> + * Return: 0 for success, negative value for failure.
> + */
> +static int zynqmp_r5_setup_mbox(struct zynqmp_r5_rproc *z_rproc,
> + struct device_node *node)
> +{
> + struct mbox_client *mclient;
> +
> + /* Setup TX mailbox channel client */
> + mclient = &z_rproc->tx_mc;
> + mclient->rx_callback = NULL;
> + mclient->tx_block = false;
> + mclient->knows_txdone = false;
> + mclient->tx_done = zynqmp_r5_mb_tx_done;
> + mclient->dev = z_rproc->dev;
> +
> + /* Setup TX mailbox channel client */
> + mclient = &z_rproc->rx_mc;
> + mclient->dev = z_rproc->dev;
> + mclient->rx_callback = zynqmp_r5_mb_rx_cb;
> + mclient->tx_block = false;
> + mclient->knows_txdone = false;
> +
> + INIT_WORK(&z_rproc->mbox_work, handle_event_notified);
> +
> + /* Request TX and RX channels */
> + z_rproc->tx_chan = mbox_request_channel_byname(&z_rproc->tx_mc, "tx");
> + if (IS_ERR(z_rproc->tx_chan)) {
> + dev_err(z_rproc->dev, "failed to request mbox tx channel.\n");
> + z_rproc->tx_chan = NULL;
> + return -EINVAL;
> + }
> +
> + z_rproc->rx_chan = mbox_request_channel_byname(&z_rproc->rx_mc, "rx");
> + if (IS_ERR(z_rproc->rx_chan)) {
> + dev_err(z_rproc->dev, "failed to request mbox rx channel.\n");
> + z_rproc->rx_chan = NULL;
> + return -EINVAL;
> + }
> + skb_queue_head_init(&z_rproc->tx_mc_skbs);
> +
> + return 0;
> +}
> +
> +/**
> + * zynqmp_r5_probe() - Probes ZynqMP R5 processor device node
> + * this is called for each individual R5 core to
> + * set up mailbox, Xilinx platform manager unique ID,
> + * add to rproc core
> + *
> + * @pdev: domain platform device for current R5 core
> + * @node: pointer of the device node for current R5 core
> + * @rpu_mode: mode to configure RPU, split or lockstep
> + * @z_rproc: Xilinx specific remoteproc structure used later to link
> + * in to cluster of cores
> + *
> + * Return: 0 for success, negative value for failure.
> + */
> +static int zynqmp_r5_probe(struct platform_device *pdev,
> + struct device_node *node,
> + enum rpu_oper_mode rpu_mode,
> + struct zynqmp_r5_rproc **z_rproc)
> +{
> + int ret;
> + struct device *dev = &pdev->dev;
> + struct rproc *rproc_ptr;
> +
> + /* Allocate remoteproc instance */
> + rproc_ptr = devm_rproc_alloc(dev, dev_name(dev), &zynqmp_r5_rproc_ops,
> + NULL, sizeof(struct zynqmp_r5_rproc));
> + if (!rproc_ptr) {
> + ret = -ENOMEM;
> + goto error;
> + }
> +
> + rproc_ptr->auto_boot = false;
> + *z_rproc = rproc_ptr->priv;
> + (*z_rproc)->rproc = rproc_ptr;
> + (*z_rproc)->dev = dev;
> + /* Set up DMA mask */
> + ret = dma_set_coherent_mask(dev, DMA_BIT_MASK(32));
> + if (ret)
> + goto error;
> +
> + /* Get R5 power domain node */
> + ret = of_property_read_u32(node, "power-domain", &(*z_rproc)->pnode_id);
> + if (ret)
> + goto error;
> +
> + ret = r5_set_mode(*z_rproc, rpu_mode);
> + if (ret)
> + goto error;
> +
> + if (of_property_read_bool(node, "mboxes")) {
> + ret = zynqmp_r5_setup_mbox(*z_rproc, node);
> + if (ret)
> + goto error;
> + }
> +
> + /* Add R5 remoteproc */
> + ret = devm_rproc_add(dev, rproc_ptr);
> + if (ret)
> + goto error;
> +
> + return 0;
> +error:
> + *z_rproc = NULL;
> + return ret;
> +}
> +
> +/*
> + * zynqmp_r5_remoteproc_probe()
> + *
> + * @pdev: domain platform device for R5 cluster
> + *
> + * called when driver is probed, for each R5 core specified in DT,
> + * setup as needed to do remoteproc-related operations
> + *
> + * Return: 0 for success, negative value for failure.
> + */
> +static int zynqmp_r5_remoteproc_probe(struct platform_device *pdev)
> +{
> + int ret, core_count;
> + struct device *dev = &pdev->dev;
> + struct device_node *nc;
> + enum rpu_oper_mode rpu_mode = PM_RPU_MODE_LOCKSTEP;
> + struct list_head *cluster; /* list to track each core's rproc */
> + struct zynqmp_r5_rproc *z_rproc;
> + struct platform_device *child_pdev;
> + struct list_head *pos;
> +
> + ret = of_property_read_u32(dev->of_node, "xlnx,cluster-mode", &rpu_mode);
> + if (ret < 0 || (rpu_mode != PM_RPU_MODE_LOCKSTEP &&
> + rpu_mode != PM_RPU_MODE_SPLIT)) {
> + dev_err(dev, "invalid format cluster mode: ret %d mode %x\n",
> + ret, rpu_mode);
> + return ret;
> + }
> +
> + dev_dbg(dev, "RPU configuration: %s\n",
> + rpu_mode == PM_RPU_MODE_LOCKSTEP ? "lockstep" : "split");
> +
> + /*
> + * if 2 RPUs provided but one is lockstep, then we have an
> + * invalid configuration.
> + */
> +
> + core_count = of_get_available_child_count(dev->of_node);
> + if ((rpu_mode == PM_RPU_MODE_LOCKSTEP && core_count != 1) ||
> + core_count > MAX_RPROCS)
> + return -EINVAL;
> +
> + cluster = devm_kzalloc(dev, sizeof(*cluster), GFP_KERNEL);
> + if (!cluster)
> + return -ENOMEM;
> + INIT_LIST_HEAD(cluster);
> +
> + ret = devm_of_platform_populate(dev);
> + if (ret) {
> + dev_err(dev, "devm_of_platform_populate failed, ret = %d\n",
> + ret);
> + return ret;
> + }
> +
> + /* probe each individual r5 core's remoteproc-related info */
> + for_each_available_child_of_node(dev->of_node, nc) {
> + child_pdev = of_find_device_by_node(nc);
> + if (!child_pdev) {
> + dev_err(dev, "could not get R5 core platform device\n");
> + ret = -ENODEV;
> + goto out;
> + }
> +
> + ret = zynqmp_r5_probe(child_pdev, nc, rpu_mode, &z_rproc);
> + dev_dbg(dev, "%s to probe rpu %pOF\n",
> + ret ? "Failed" : "Able",
> + nc);
> + if (!z_rproc)
> + ret = -EINVAL;
> + if (ret)
> + goto out;
> + list_add_tail(&z_rproc->elem, cluster);
> + }
> + /* wire in so each core can be cleaned up at driver remove */
> + platform_set_drvdata(pdev, cluster);
> + return 0;
> +out:
> + /*
> + * undo core0 upon any failures on core1 in split-mode
> + *
> + * in zynqmp_r5_probe z_rproc is set to null
> + * and ret to non-zero value if error
> + */
> + if (ret && !z_rproc && rpu_mode == PM_RPU_MODE_SPLIT &&
> + !list_empty(cluster)) {
> + list_for_each(pos, cluster) {
> + z_rproc = list_entry(pos, struct zynqmp_r5_rproc, elem);
> + if (of_property_read_bool(z_rproc->dev->of_node, "mboxes")) {
> + mbox_free_channel(z_rproc->tx_chan);
> + mbox_free_channel(z_rproc->rx_chan);
> + }
> + }
> + }
> + return ret;
> +}
> +
> +/*
> + * zynqmp_r5_remoteproc_remove()
> + *
> + * @pdev: domain platform device for R5 cluster
> + *
> + * When the driver is unloaded, clean up the mailboxes for each
> + * remoteproc that was initially probed.
> + */
> +static int zynqmp_r5_remoteproc_remove(struct platform_device *pdev)
> +{
> + struct list_head *pos, *temp, *cluster = (struct list_head *)
> + platform_get_drvdata(pdev);
> + struct zynqmp_r5_rproc *z_rproc = NULL;
> +
> + list_for_each_safe(pos, temp, cluster) {
> + z_rproc = list_entry(pos, struct zynqmp_r5_rproc, elem);
> + if (of_property_read_bool(z_rproc->dev->of_node, "mboxes")) {
> + mbox_free_channel(z_rproc->tx_chan);
> + mbox_free_channel(z_rproc->rx_chan);
> + }
> + list_del(pos);
> + }
> + return 0;
> +}
> +
> +/* Match table for OF platform binding */
> +static const struct of_device_id zynqmp_r5_remoteproc_match[] = {
> + { .compatible = "xlnx,zynqmp-r5-remoteproc", },
> + { /* end of list */ },
> +};
> +MODULE_DEVICE_TABLE(of, zynqmp_r5_remoteproc_match);
> +
> +static struct platform_driver zynqmp_r5_remoteproc_driver = {
> + .probe = zynqmp_r5_remoteproc_probe,
> + .remove = zynqmp_r5_remoteproc_remove,
> + .driver = {
> + .name = "zynqmp_r5_remoteproc",
> + .of_match_table = zynqmp_r5_remoteproc_match,
> + },
> +};
> +module_platform_driver(zynqmp_r5_remoteproc_driver);
> +
> +MODULE_AUTHOR("Ben Levinsky <ben.levinsky@xilinx.com>");
> +MODULE_LICENSE("GPL v2");
> --
> 2.17.1
>
^ permalink raw reply [flat|nested] 13+ messages in thread* Re: [PATCH v24 5/5] remoteproc: Add initial zynqmp R5 remoteproc driver
2020-11-30 15:57 ` [PATCH v24 5/5] remoteproc: Add initial zynqmp R5 remoteproc driver Ben Levinsky
` (2 preceding siblings ...)
2020-12-09 19:00 ` Mathieu Poirier
@ 2020-12-10 20:56 ` Mathieu Poirier
3 siblings, 0 replies; 13+ messages in thread
From: Mathieu Poirier @ 2020-12-10 20:56 UTC (permalink / raw)
To: Ben Levinsky; +Cc: devicetree, linux-remoteproc, linux-kernel, linux-arm-kernel
On Mon, Nov 30, 2020 at 07:57:17AM -0800, Ben Levinsky wrote:
> R5 is included in Xilinx Zynq UltraScale MPSoC so by adding this
> remotproc driver, we can boot the R5 sub-system in two different
> configurations -
> * Split
> * Lockstep
>
> The Xilinx R5 Remoteproc Driver boots the R5's via calls to the Xilinx
> Platform Management Unit that handles the R5 configuration, memory access
> and R5 lifecycle management. The interface to this manager is done in this
> driver via zynqmp_pm_* function calls.
>
> Signed-off-by: Wendy Liang <wendy.liang@xilinx.com>
> Signed-off-by: Michal Simek <michal.simek@xilinx.com>
> Signed-off-by: Ed Mooring <ed.mooring@xilinx.com>
> Signed-off-by: Jason Wu <j.wu@xilinx.com>
> Signed-off-by: Ben Levinsky <ben.levinsky@xilinx.com>
> ---
> drivers/remoteproc/Kconfig | 8 +
> drivers/remoteproc/Makefile | 1 +
> drivers/remoteproc/zynqmp_r5_remoteproc.c | 872 ++++++++++++++++++++++
> 3 files changed, 881 insertions(+)
> create mode 100644 drivers/remoteproc/zynqmp_r5_remoteproc.c
>
> diff --git a/drivers/remoteproc/Kconfig b/drivers/remoteproc/Kconfig
> index c6659dfea7c7..c2fe54b1d94f 100644
> --- a/drivers/remoteproc/Kconfig
> +++ b/drivers/remoteproc/Kconfig
> @@ -275,6 +275,14 @@ config TI_K3_DSP_REMOTEPROC
> It's safe to say N here if you're not interested in utilizing
> the DSP slave processors.
>
> +config ZYNQMP_R5_REMOTEPROC
> + tristate "ZynqMP R5 remoteproc support"
> + depends on PM && ARCH_ZYNQMP
> + select RPMSG_VIRTIO
> + select ZYNQMP_IPI_MBOX
> + help
> + Say y or m here to support ZynqMP R5 remote processors via the remote
> + processor framework.
> endif # REMOTEPROC
>
> endmenu
> diff --git a/drivers/remoteproc/Makefile b/drivers/remoteproc/Makefile
> index 3dfa28e6c701..ef1abff654c2 100644
> --- a/drivers/remoteproc/Makefile
> +++ b/drivers/remoteproc/Makefile
> @@ -33,3 +33,4 @@ obj-$(CONFIG_ST_REMOTEPROC) += st_remoteproc.o
> obj-$(CONFIG_ST_SLIM_REMOTEPROC) += st_slim_rproc.o
> obj-$(CONFIG_STM32_RPROC) += stm32_rproc.o
> obj-$(CONFIG_TI_K3_DSP_REMOTEPROC) += ti_k3_dsp_remoteproc.o
> +obj-$(CONFIG_ZYNQMP_R5_REMOTEPROC) += zynqmp_r5_remoteproc.o
> diff --git a/drivers/remoteproc/zynqmp_r5_remoteproc.c b/drivers/remoteproc/zynqmp_r5_remoteproc.c
> new file mode 100644
> index 000000000000..2593de618409
> --- /dev/null
> +++ b/drivers/remoteproc/zynqmp_r5_remoteproc.c
> @@ -0,0 +1,872 @@
> +// SPDX-License-Identifier: GPL-2.0
> +/*
> + * Zynq R5 Remote Processor driver
> + *
> + * Based on origin OMAP and Zynq Remote Processor driver
> + *
> + */
> +
> +#include <linux/firmware/xlnx-zynqmp.h>
> +#include <linux/interrupt.h>
> +#include <linux/kernel.h>
> +#include <linux/list.h>
> +#include <linux/mailbox_client.h>
> +#include <linux/mailbox/zynqmp-ipi-message.h>
> +#include <linux/module.h>
> +#include <linux/of_address.h>
> +#include <linux/of_platform.h>
> +#include <linux/of_reserved_mem.h>
> +#include <linux/platform_device.h>
> +#include <linux/remoteproc.h>
> +#include <linux/skbuff.h>
> +#include <linux/sysfs.h>
> +
> +#include "remoteproc_internal.h"
> +
> +#define MAX_RPROCS 2 /* Support up to 2 RPU */
> +#define MAX_MEM_PNODES 4 /* Max power nodes for one RPU memory instance */
> +
> +#define BANK_LIST_PROP "sram"
> +#define DDR_LIST_PROP "memory-region"
> +
> +/* IPI buffer MAX length */
> +#define IPI_BUF_LEN_MAX 32U
> +/* RX mailbox client buffer max length */
> +#define RX_MBOX_CLIENT_BUF_MAX (IPI_BUF_LEN_MAX + \
> + sizeof(struct zynqmp_ipi_message))
> +
> +/*
> + * Map each Xilinx on-chip SRAM Bank address to their own respective
> + * pm_node_id.
> + */
> +struct sram_addr_data {
> + phys_addr_t addr;
> + enum pm_node_id id;
> +};
> +
> +#define NUM_SRAMS 4U
> +static const struct sram_addr_data zynqmp_banks[NUM_SRAMS] = {
> + {0xffe00000UL, NODE_TCM_0_A},
> + {0xffe20000UL, NODE_TCM_0_B},
> + {0xffe90000UL, NODE_TCM_1_A},
> + {0xffeb0000UL, NODE_TCM_1_B},
> +};
> +
> +/**
> + * struct zynqmp_r5_rproc - ZynqMP R5 core structure
> + *
> + * @rx_mc_buf: rx mailbox client buffer to save the rx message
> + * @tx_mc: tx mailbox client
> + * @rx_mc: rx mailbox client
> + * @mbox_work: mbox_work for the RPU remoteproc
> + * @tx_mc_skbs: socket buffers for tx mailbox client
> + * @dev: device of RPU instance
> + * @rproc: rproc handle
> + * @tx_chan: tx mailbox channel
> + * @rx_chan: rx mailbox channel
> + * @pnode_id: RPU CPU power domain id
> + * @elem: linked list item
> + */
> +struct zynqmp_r5_rproc {
> + unsigned char rx_mc_buf[RX_MBOX_CLIENT_BUF_MAX];
> + struct mbox_client tx_mc;
> + struct mbox_client rx_mc;
> + struct work_struct mbox_work;
> + struct sk_buff_head tx_mc_skbs;
> + struct device *dev;
> + struct rproc *rproc;
> + struct mbox_chan *tx_chan;
> + struct mbox_chan *rx_chan;
> + u32 pnode_id;
> + struct list_head elem;
> +};
> +
> +/*
> + * r5_set_mode - set RPU operation mode
> + * @z_rproc: Remote processor private data
> + * @rpu_mode: mode specified by device tree to configure the RPU to
> + *
> + * set RPU operation mode
> + *
> + * Return: 0 for success, negative value for failure
> + */
> +static int r5_set_mode(struct zynqmp_r5_rproc *z_rproc,
> + enum rpu_oper_mode rpu_mode)
> +{
> + enum rpu_tcm_comb tcm_mode;
> + enum rpu_oper_mode cur_rpu_mode;
> + int ret;
> +
> + ret = zynqmp_pm_get_rpu_mode(z_rproc->pnode_id, &cur_rpu_mode);
> + if (ret < 0)
> + return ret;
> +
> + if (rpu_mode != cur_rpu_mode) {
> + ret = zynqmp_pm_set_rpu_mode(z_rproc->pnode_id,
> + rpu_mode);
> + if (ret < 0)
> + return ret;
> + }
> +
> + tcm_mode = (rpu_mode == PM_RPU_MODE_LOCKSTEP) ?
> + PM_RPU_TCM_COMB : PM_RPU_TCM_SPLIT;
> + return zynqmp_pm_set_tcm_config(z_rproc->pnode_id, tcm_mode);
> +}
> +
> +/*
> + * zynqmp_r5_rproc_mem_release
> + * @rproc: single R5 core's corresponding rproc instance
> + * @mem: mem entry to unmap
> + *
> + * Unmap TCM banks when powering down R5 core.
> + *
> + * return 0 on success, otherwise non-zero value on failure
> + */
> +static int tcm_mem_release(struct rproc *rproc, struct rproc_mem_entry *mem)
> +{
> + u32 pnode_id = (u64)mem->priv;
> +
> + iounmap(mem->va);
> + return zynqmp_pm_release_node(pnode_id);
> +}
> +
> +/*
> + * zynqmp_r5_rproc_start
> + * @rproc: single R5 core's corresponding rproc instance
> + *
> + * Start R5 Core from designated boot address.
> + *
> + * return 0 on success, otherwise non-zero value on failure
> + */
> +static int zynqmp_r5_rproc_start(struct rproc *rproc)
> +{
> + struct zynqmp_r5_rproc *z_rproc = rproc->priv;
> + enum rpu_boot_mem bootmem;
> +
> + bootmem = (rproc->bootaddr & 0xF0000000) == 0xF0000000 ?
> + PM_RPU_BOOTMEM_HIVEC : PM_RPU_BOOTMEM_LOVEC;
> +
> + dev_dbg(rproc->dev.parent, "RPU boot from %s.",
> + bootmem == PM_RPU_BOOTMEM_HIVEC ? "OCM" : "TCM");
> +
> + return zynqmp_pm_request_wake(z_rproc->pnode_id, 1,
> + bootmem, ZYNQMP_PM_REQUEST_ACK_NO);
> +}
> +
> +/*
> + * zynqmp_r5_rproc_stop
> + * @rproc: single R5 core's corresponding rproc instance
> + *
> + * Power down R5 Core.
> + *
> + * return 0 on success, otherwise non-zero value on failure
> + */
> +static int zynqmp_r5_rproc_stop(struct rproc *rproc)
> +{
> + struct zynqmp_r5_rproc *z_rproc = rproc->priv;
> +
> + return zynqmp_pm_force_pwrdwn(z_rproc->pnode_id,
> + ZYNQMP_PM_REQUEST_ACK_BLOCKING);
> +}
> +
> +/*
> + * zynqmp_r5_rproc_mem_alloc
> + * @rproc: single R5 core's corresponding rproc instance
> + * @mem: mem entry to map
> + *
> + * Callback to map va for memory-region's carveout.
> + *
> + * return 0 on success, otherwise non-zero value on failure
> + */
> +static int zynqmp_r5_rproc_mem_alloc(struct rproc *rproc,
> + struct rproc_mem_entry *mem)
> +{
> + void *va;
> +
> + va = ioremap_wc(mem->dma, mem->len);
> + if (IS_ERR_OR_NULL(va))
> + return -ENOMEM;
> +
> + mem->va = va;
> +
> + return 0;
> +}
> +
> +/*
> + * zynqmp_r5_rproc_mem_release
> + * @rproc: single R5 core's corresponding rproc instance
> + * @mem: mem entry to unmap
> + *
> + * Unmap memory-region carveout
> + *
> + * return 0 on success, otherwise non-zero value on failure
> + */
> +static int zynqmp_r5_rproc_mem_release(struct rproc *rproc,
> + struct rproc_mem_entry *mem)
> +{
> + iounmap(mem->va);
> + return 0;
> +}
> +
> +/*
> + * parse_mem_regions
> + * @rproc: single R5 core's corresponding rproc instance
> + *
> + * Construct rproc mem carveouts from carveout provided in
> + * memory-region property
> + *
> + * return 0 on success, otherwise non-zero value on failure
> + */
> +static int parse_mem_regions(struct rproc *rproc)
> +{
> + int num_mems, i;
> + struct zynqmp_r5_rproc *z_rproc = rproc->priv;
> + struct device *dev = &rproc->dev;
> + struct device_node *np = z_rproc->dev->of_node;
> + struct rproc_mem_entry *mem;
> +
> + num_mems = of_count_phandle_with_args(np, DDR_LIST_PROP, NULL);
> + if (num_mems <= 0)
> + return 0;
> +
> + for (i = 0; i < num_mems; i++) {
> + struct device_node *node;
> + struct reserved_mem *rmem;
> +
> + node = of_parse_phandle(np, DDR_LIST_PROP, i);
> + if (!node)
> + return -EINVAL;
> +
> + rmem = of_reserved_mem_lookup(node);
> + if (!rmem)
> + return -EINVAL;
> +
> + if (strstr(node->name, "vdev0vring")) {
> + int vring_id;
> + char name[16];
> +
> + /*
> + * expecting form of "rpuXvdev0vringX as documented
> + * in xilinx remoteproc device tree binding
> + */
> + if (strlen(node->name) < 15) {
> + dev_err(dev, "%pOF is less than 14 chars",
14, 15 or 16 characters?
> + node);
> + return -EINVAL;
> + }
> +
> + /*
> + * can be 1 of multiple vring IDs per IPC channel
> + * e.g. 'vdev0vring0' and 'vdev0vring1'
> + */
> + vring_id = node->name[14] - '0';
> + snprintf(name, sizeof(name), "vdev0vring%d", vring_id);
> + /* Register vring */
> + mem = rproc_mem_entry_init(dev, NULL,
> + (dma_addr_t)rmem->base,
> + rmem->size, rmem->base,
> + zynqmp_r5_rproc_mem_alloc,
> + zynqmp_r5_rproc_mem_release,
> + name);
> + } else {
> + /* Register DMA region */
> + int (*alloc)(struct rproc *r,
> + struct rproc_mem_entry *rme);
> + int (*release)(struct rproc *r,
> + struct rproc_mem_entry *rme);
> + char name[20];
> +
> + if (strstr(node->name, "vdev0buffer")) {
> + alloc = NULL;
> + release = NULL;
> + strcpy(name, "vdev0buffer");
Any reason for not using rproc_of_resm_mem_entry_init() for vdev buffers? All
the hard work is already done.
> + } else {
> + alloc = zynqmp_r5_rproc_mem_alloc;
> + release = zynqmp_r5_rproc_mem_release;
> + strcpy(name, node->name);
> + }
> +
> + mem = rproc_mem_entry_init(dev, NULL,
> + (dma_addr_t)rmem->base,
> + rmem->size, rmem->base,
> + alloc, release, name);
> + }
> + if (!mem)
> + return -ENOMEM;
> +
> + rproc_add_carveout(rproc, mem);
> + }
> +
> + return 0;
> +}
> +
> +/*
> + * zynqmp_r5_pm_request_tcm
> + * @addr: base address of mem provided in R5 core's sram property.
> + *
> + * Given sram base address, determine its corresponding Xilinx
> + * Platform Management ID and then request access to this node
> + * so that it can be power up.
> + *
> + * return 0 on success, otherwise non-zero value on failure
> + */
> +static int zynqmp_r5_pm_request_sram(phys_addr_t addr)
> +{
> + unsigned int i;
> +
> + for (i = 0; i < NUM_SRAMS; i++) {
> + if (zynqmp_banks[i].addr == addr)
> + return zynqmp_pm_request_node(zynqmp_banks[i].id,
> + ZYNQMP_PM_CAPABILITY_ACCESS,
> + 0,
> + ZYNQMP_PM_REQUEST_ACK_BLOCKING);
> + }
> +
> + return -EINVAL;
> +}
> +
> +/*
> + * tcm_mem_alloc
> + * @rproc: single R5 core's corresponding rproc instance
> + * @mem: mem entry to initialize the va and da fields of
> + *
> + * Given TCM bank entry,
> + * this callback will set device address for R5 running on TCM
> + * and also setup virtual address for TCM bank remoteproc carveout
> + *
> + * return 0 on success, otherwise non-zero value on failure
> + */
> +static int tcm_mem_alloc(struct rproc *rproc,
> + struct rproc_mem_entry *mem)
> +{
> + void *va;
> + struct device *dev = rproc->dev.parent;
> +
> + va = ioremap_wc(mem->dma, mem->len);
> + if (IS_ERR_OR_NULL(va))
> + return -ENOMEM;
> +
> + /* Update memory entry va */
> + mem->va = va;
> +
> + va = devm_ioremap_wc(dev, mem->da, mem->len);
> + if (!va)
> + return -ENOMEM;
> + /* As R5 is 32 bit, wipe out extra high bits */
> + mem->da &= 0x000fffff;
> + /*
> + * The R5s expect their TCM banks to be at address 0x0 and 0x2000,
> + * while on the Linux side they are at 0xffexxxxx. Zero out the high
> + * 12 bits of the address.
> + */
> +
> + /*
> + * TCM Banks 1A and 1B (0xffe90000 and 0xffeb0000) still
> + * need to be translated to 0x0 and 0x20000
> + */
> + if (mem->da == 0x90000 || mem->da == 0xB0000)
> + mem->da -= 0x90000;
> +
> + /* if translated TCM bank address is not valid report error */
> + if (mem->da != 0x0 && mem->da != 0x20000) {
> + dev_err(dev, "invalid TCM bank address: %x\n", mem->da);
> + return -EINVAL;
> + }
> +
> + return 0;
> +}
> +
> +/*
> + * parse_tcm_banks()
> + * @rproc: single R5 core's corresponding rproc instance
> + *
> + * Given R5 node in remoteproc instance
> + * allocate remoteproc carveout for TCM memory
> + * needed for firmware to be loaded
> + *
> + * return 0 on success, otherwise non-zero value on failure
> + */
> +static int parse_tcm_banks(struct rproc *rproc)
> +{
> + int i, num_banks;
> + struct zynqmp_r5_rproc *z_rproc = rproc->priv;
> + struct device *dev = &rproc->dev;
> + struct device_node *r5_node = z_rproc->dev->of_node;
> +
> + /* go through TCM banks for r5 node */
> + num_banks = of_count_phandle_with_args(r5_node, BANK_LIST_PROP, NULL);
> + if (num_banks <= 0) {
> + dev_err(dev, "need to specify TCM banks\n");
> + return -EINVAL;
> + }
> + for (i = 0; i < num_banks; i++) {
> + struct resource rsc;
> + resource_size_t size;
> + struct device_node *dt_node;
> + struct rproc_mem_entry *mem;
> + int ret;
> + u32 pnode_id; /* zynqmp_pm* fn's expect u32 */
> +
> + dt_node = of_parse_phandle(r5_node, BANK_LIST_PROP, i);
> + if (!dt_node)
> + return -EINVAL;
> +
> + if (of_device_is_available(dt_node)) {
> + ret = of_address_to_resource(dt_node, 0, &rsc);
> + if (ret < 0)
> + return ret;
> + ret = zynqmp_r5_pm_request_sram(rsc.start);
> + if (ret < 0)
> + return ret;
> +
> + /* add carveout */
> + size = resource_size(&rsc);
> + mem = rproc_mem_entry_init(dev, NULL, rsc.start,
> + (int)size, rsc.start,
> + tcm_mem_alloc,
> + tcm_mem_release,
> + rsc.name);
> + if (!mem)
> + return -ENOMEM;
> +
> + mem->priv = (void *)(u64)pnode_id;
> + rproc_add_carveout(rproc, mem);
> + }
> + }
> +
> + return 0;
> +}
> +
> +/*
> + * zynqmp_r5_parse_fw()
> + * @rproc: single R5 core's corresponding rproc instance
> + * @fw: ptr to firmware to be loaded onto r5 core
> + *
> + * When loading firmware, ensure the necessary carveouts are in remoteproc
> + *
> + * return 0 on success, otherwise non-zero value on failure
> + */
> +static int zynqmp_r5_parse_fw(struct rproc *rproc, const struct firmware *fw)
> +{
> + int ret;
> +
> + ret = parse_tcm_banks(rproc);
> + if (ret)
> + return ret;
> +
> + ret = parse_mem_regions(rproc);
> + if (ret)
> + return ret;
> +
> + ret = rproc_elf_load_rsc_table(rproc, fw);
> + if (ret == -EINVAL) {
> + /*
> + * resource table only required for IPC.
> + * if not present, this is not necessarily an error;
> + * for example, loading r5 hello world application
> + * so simply inform user and keep going.
> + */
> + dev_info(&rproc->dev, "no resource table found.\n");
> + ret = 0;
> + }
> + return ret;
> +}
> +
> +/*
> + * zynqmp_r5_rproc_kick() - kick a firmware if mbox is provided
> + * @rproc: r5 core's corresponding rproc structure
> + * @vqid: virtqueue ID
> + */
> +static void zynqmp_r5_rproc_kick(struct rproc *rproc, int vqid)
> +{
> + struct sk_buff *skb;
> + unsigned int skb_len;
> + struct zynqmp_ipi_message *mb_msg;
> + int ret;
> +
> + struct device *dev = rproc->dev.parent;
> + struct zynqmp_r5_rproc *z_rproc = rproc->priv;
> +
> + if (of_property_read_bool(dev->of_node, "mboxes")) {
I'm still very puzzled here. If we get to this point we should have a mailbox
infrastructure. Otherwise there is no point in having reserved memory and
setting up the virtqueues.
I am done reviewing this set. There is no point in rushing a new revision as I
won't have the bandwidth to look at it before the holidays.
Thanks,
Mathieu
> + skb_len = (unsigned int)(sizeof(vqid) + sizeof(mb_msg));
> + skb = alloc_skb(skb_len, GFP_ATOMIC);
> + if (!skb)
> + return;
> +
> + mb_msg = (struct zynqmp_ipi_message *)skb_put(skb, skb_len);
> + mb_msg->len = sizeof(vqid);
> + memcpy(mb_msg->data, &vqid, sizeof(vqid));
> +
> + skb_queue_tail(&z_rproc->tx_mc_skbs, skb);
> + ret = mbox_send_message(z_rproc->tx_chan, mb_msg);
> + if (ret < 0) {
> + dev_warn(dev, "Failed to kick remote.\n");
> + skb_dequeue_tail(&z_rproc->tx_mc_skbs);
> + kfree_skb(skb);
> + }
> + } else {
> + (void)skb;
> + (void)skb_len;
> + (void)mb_msg;
> + (void)ret;
> + (void)vqid;
> + }
> +}
> +
> +static struct rproc_ops zynqmp_r5_rproc_ops = {
> + .start = zynqmp_r5_rproc_start,
> + .stop = zynqmp_r5_rproc_stop,
> + .load = rproc_elf_load_segments,
> + .parse_fw = zynqmp_r5_parse_fw,
> + .find_loaded_rsc_table = rproc_elf_find_loaded_rsc_table,
> + .sanity_check = rproc_elf_sanity_check,
> + .get_boot_addr = rproc_elf_get_boot_addr,
> + .kick = zynqmp_r5_rproc_kick,
> +};
> +
> +/**
> + * event_notified_idr_cb() - event notified idr callback
> + * @id: idr id
> + * @ptr: pointer to idr private data
> + * @data: data passed to idr_for_each callback
> + *
> + * Pass notification to remoteproc virtio
> + *
> + * Return: 0. having return is to satisfy the idr_for_each() function
> + * pointer input argument requirement.
> + **/
> +static int event_notified_idr_cb(int id, void *ptr, void *data)
> +{
> + struct rproc *rproc = data;
> +
> + (void)rproc_vq_interrupt(rproc, id);
> + return 0;
> +}
> +
> +/**
> + * handle_event_notified() - remoteproc notification work function
> + * @work: pointer to the work structure
> + *
> + * It checks each registered remoteproc notify IDs.
> + */
> +static void handle_event_notified(struct work_struct *work)
> +{
> + struct rproc *rproc;
> + struct zynqmp_r5_rproc *z_rproc;
> +
> + z_rproc = container_of(work, struct zynqmp_r5_rproc, mbox_work);
> +
> + (void)mbox_send_message(z_rproc->rx_chan, NULL);
> + rproc = z_rproc->rproc;
> + /*
> + * We only use IPI for interrupt. The firmware side may or may
> + * not write the notifyid when it trigger IPI.
> + * And thus, we scan through all the registered notifyids.
> + */
> + idr_for_each(&rproc->notifyids, event_notified_idr_cb, rproc);
> +}
> +
> +/**
> + * zynqmp_r5_mb_rx_cb() - Receive channel mailbox callback
> + * @cl: mailbox client
> + * @msg: message pointer
> + *
> + * It will schedule the R5 notification work.
> + */
> +static void zynqmp_r5_mb_rx_cb(struct mbox_client *cl, void *msg)
> +{
> + struct zynqmp_r5_rproc *z_rproc;
> +
> + z_rproc = container_of(cl, struct zynqmp_r5_rproc, rx_mc);
> + if (msg) {
> + struct zynqmp_ipi_message *ipi_msg, *buf_msg;
> + size_t len;
> +
> + ipi_msg = (struct zynqmp_ipi_message *)msg;
> + buf_msg = (struct zynqmp_ipi_message *)z_rproc->rx_mc_buf;
> + len = (ipi_msg->len >= IPI_BUF_LEN_MAX) ?
> + IPI_BUF_LEN_MAX : ipi_msg->len;
> + buf_msg->len = len;
> + memcpy(buf_msg->data, ipi_msg->data, len);
> + }
> + schedule_work(&z_rproc->mbox_work);
> +}
> +
> +/**
> + * zynqmp_r5_mb_tx_done() - Request has been sent to the remote
> + * @cl: mailbox client
> + * @msg: pointer to the message which has been sent
> + * @r: status of last TX - OK or error
> + *
> + * It will be called by the mailbox framework when the last TX has done.
> + */
> +static void zynqmp_r5_mb_tx_done(struct mbox_client *cl, void *msg, int r)
> +{
> + struct zynqmp_r5_rproc *z_rproc;
> + struct sk_buff *skb;
> +
> + if (!msg)
> + return;
> + z_rproc = container_of(cl, struct zynqmp_r5_rproc, tx_mc);
> + skb = skb_dequeue(&z_rproc->tx_mc_skbs);
> + kfree_skb(skb);
> +}
> +
> +/**
> + * zynqmp_r5_setup_mbox() - Setup mailboxes
> + * this is used for each individual R5 core
> + *
> + * @z_rproc: pointer to the ZynqMP R5 processor platform data
> + * @node: pointer of the device node
> + *
> + * Function to setup mailboxes to talk to RPU.
> + *
> + * Return: 0 for success, negative value for failure.
> + */
> +static int zynqmp_r5_setup_mbox(struct zynqmp_r5_rproc *z_rproc,
> + struct device_node *node)
> +{
> + struct mbox_client *mclient;
> +
> + /* Setup TX mailbox channel client */
> + mclient = &z_rproc->tx_mc;
> + mclient->rx_callback = NULL;
> + mclient->tx_block = false;
> + mclient->knows_txdone = false;
> + mclient->tx_done = zynqmp_r5_mb_tx_done;
> + mclient->dev = z_rproc->dev;
> +
> + /* Setup TX mailbox channel client */
> + mclient = &z_rproc->rx_mc;
> + mclient->dev = z_rproc->dev;
> + mclient->rx_callback = zynqmp_r5_mb_rx_cb;
> + mclient->tx_block = false;
> + mclient->knows_txdone = false;
> +
> + INIT_WORK(&z_rproc->mbox_work, handle_event_notified);
> +
> + /* Request TX and RX channels */
> + z_rproc->tx_chan = mbox_request_channel_byname(&z_rproc->tx_mc, "tx");
> + if (IS_ERR(z_rproc->tx_chan)) {
> + dev_err(z_rproc->dev, "failed to request mbox tx channel.\n");
> + z_rproc->tx_chan = NULL;
> + return -EINVAL;
> + }
> +
> + z_rproc->rx_chan = mbox_request_channel_byname(&z_rproc->rx_mc, "rx");
> + if (IS_ERR(z_rproc->rx_chan)) {
> + dev_err(z_rproc->dev, "failed to request mbox rx channel.\n");
> + z_rproc->rx_chan = NULL;
> + return -EINVAL;
> + }
> + skb_queue_head_init(&z_rproc->tx_mc_skbs);
> +
> + return 0;
> +}
> +
> +/**
> + * zynqmp_r5_probe() - Probes ZynqMP R5 processor device node
> + * this is called for each individual R5 core to
> + * set up mailbox, Xilinx platform manager unique ID,
> + * add to rproc core
> + *
> + * @pdev: domain platform device for current R5 core
> + * @node: pointer of the device node for current R5 core
> + * @rpu_mode: mode to configure RPU, split or lockstep
> + * @z_rproc: Xilinx specific remoteproc structure used later to link
> + * in to cluster of cores
> + *
> + * Return: 0 for success, negative value for failure.
> + */
> +static int zynqmp_r5_probe(struct platform_device *pdev,
> + struct device_node *node,
> + enum rpu_oper_mode rpu_mode,
> + struct zynqmp_r5_rproc **z_rproc)
> +{
> + int ret;
> + struct device *dev = &pdev->dev;
> + struct rproc *rproc_ptr;
> +
> + /* Allocate remoteproc instance */
> + rproc_ptr = devm_rproc_alloc(dev, dev_name(dev), &zynqmp_r5_rproc_ops,
> + NULL, sizeof(struct zynqmp_r5_rproc));
> + if (!rproc_ptr) {
> + ret = -ENOMEM;
> + goto error;
> + }
> +
> + rproc_ptr->auto_boot = false;
> + *z_rproc = rproc_ptr->priv;
> + (*z_rproc)->rproc = rproc_ptr;
> + (*z_rproc)->dev = dev;
> + /* Set up DMA mask */
> + ret = dma_set_coherent_mask(dev, DMA_BIT_MASK(32));
> + if (ret)
> + goto error;
> +
> + /* Get R5 power domain node */
> + ret = of_property_read_u32(node, "power-domain", &(*z_rproc)->pnode_id);
> + if (ret)
> + goto error;
> +
> + ret = r5_set_mode(*z_rproc, rpu_mode);
> + if (ret)
> + goto error;
> +
> + if (of_property_read_bool(node, "mboxes")) {
> + ret = zynqmp_r5_setup_mbox(*z_rproc, node);
> + if (ret)
> + goto error;
> + }
> +
> + /* Add R5 remoteproc */
> + ret = devm_rproc_add(dev, rproc_ptr);
> + if (ret)
> + goto error;
> +
> + return 0;
> +error:
> + *z_rproc = NULL;
> + return ret;
> +}
> +
> +/*
> + * zynqmp_r5_remoteproc_probe()
> + *
> + * @pdev: domain platform device for R5 cluster
> + *
> + * called when driver is probed, for each R5 core specified in DT,
> + * setup as needed to do remoteproc-related operations
> + *
> + * Return: 0 for success, negative value for failure.
> + */
> +static int zynqmp_r5_remoteproc_probe(struct platform_device *pdev)
> +{
> + int ret, core_count;
> + struct device *dev = &pdev->dev;
> + struct device_node *nc;
> + enum rpu_oper_mode rpu_mode = PM_RPU_MODE_LOCKSTEP;
> + struct list_head *cluster; /* list to track each core's rproc */
> + struct zynqmp_r5_rproc *z_rproc;
> + struct platform_device *child_pdev;
> + struct list_head *pos;
> +
> + ret = of_property_read_u32(dev->of_node, "xlnx,cluster-mode", &rpu_mode);
> + if (ret < 0 || (rpu_mode != PM_RPU_MODE_LOCKSTEP &&
> + rpu_mode != PM_RPU_MODE_SPLIT)) {
> + dev_err(dev, "invalid format cluster mode: ret %d mode %x\n",
> + ret, rpu_mode);
> + return ret;
> + }
> +
> + dev_dbg(dev, "RPU configuration: %s\n",
> + rpu_mode == PM_RPU_MODE_LOCKSTEP ? "lockstep" : "split");
> +
> + /*
> + * if 2 RPUs provided but one is lockstep, then we have an
> + * invalid configuration.
> + */
> +
> + core_count = of_get_available_child_count(dev->of_node);
> + if ((rpu_mode == PM_RPU_MODE_LOCKSTEP && core_count != 1) ||
> + core_count > MAX_RPROCS)
> + return -EINVAL;
> +
> + cluster = devm_kzalloc(dev, sizeof(*cluster), GFP_KERNEL);
> + if (!cluster)
> + return -ENOMEM;
> + INIT_LIST_HEAD(cluster);
> +
> + ret = devm_of_platform_populate(dev);
> + if (ret) {
> + dev_err(dev, "devm_of_platform_populate failed, ret = %d\n",
> + ret);
> + return ret;
> + }
> +
> + /* probe each individual r5 core's remoteproc-related info */
> + for_each_available_child_of_node(dev->of_node, nc) {
> + child_pdev = of_find_device_by_node(nc);
> + if (!child_pdev) {
> + dev_err(dev, "could not get R5 core platform device\n");
> + ret = -ENODEV;
> + goto out;
> + }
> +
> + ret = zynqmp_r5_probe(child_pdev, nc, rpu_mode, &z_rproc);
> + dev_dbg(dev, "%s to probe rpu %pOF\n",
> + ret ? "Failed" : "Able",
> + nc);
> + if (!z_rproc)
> + ret = -EINVAL;
> + if (ret)
> + goto out;
> + list_add_tail(&z_rproc->elem, cluster);
> + }
> + /* wire in so each core can be cleaned up at driver remove */
> + platform_set_drvdata(pdev, cluster);
> + return 0;
> +out:
> + /*
> + * undo core0 upon any failures on core1 in split-mode
> + *
> + * in zynqmp_r5_probe z_rproc is set to null
> + * and ret to non-zero value if error
> + */
> + if (ret && !z_rproc && rpu_mode == PM_RPU_MODE_SPLIT &&
> + !list_empty(cluster)) {
> + list_for_each(pos, cluster) {
> + z_rproc = list_entry(pos, struct zynqmp_r5_rproc, elem);
> + if (of_property_read_bool(z_rproc->dev->of_node, "mboxes")) {
> + mbox_free_channel(z_rproc->tx_chan);
> + mbox_free_channel(z_rproc->rx_chan);
> + }
> + }
> + }
> + return ret;
> +}
> +
> +/*
> + * zynqmp_r5_remoteproc_remove()
> + *
> + * @pdev: domain platform device for R5 cluster
> + *
> + * When the driver is unloaded, clean up the mailboxes for each
> + * remoteproc that was initially probed.
> + */
> +static int zynqmp_r5_remoteproc_remove(struct platform_device *pdev)
> +{
> + struct list_head *pos, *temp, *cluster = (struct list_head *)
> + platform_get_drvdata(pdev);
> + struct zynqmp_r5_rproc *z_rproc = NULL;
> +
> + list_for_each_safe(pos, temp, cluster) {
> + z_rproc = list_entry(pos, struct zynqmp_r5_rproc, elem);
> + if (of_property_read_bool(z_rproc->dev->of_node, "mboxes")) {
> + mbox_free_channel(z_rproc->tx_chan);
> + mbox_free_channel(z_rproc->rx_chan);
> + }
> + list_del(pos);
> + }
> + return 0;
> +}
> +
> +/* Match table for OF platform binding */
> +static const struct of_device_id zynqmp_r5_remoteproc_match[] = {
> + { .compatible = "xlnx,zynqmp-r5-remoteproc", },
> + { /* end of list */ },
> +};
> +MODULE_DEVICE_TABLE(of, zynqmp_r5_remoteproc_match);
> +
> +static struct platform_driver zynqmp_r5_remoteproc_driver = {
> + .probe = zynqmp_r5_remoteproc_probe,
> + .remove = zynqmp_r5_remoteproc_remove,
> + .driver = {
> + .name = "zynqmp_r5_remoteproc",
> + .of_match_table = zynqmp_r5_remoteproc_match,
> + },
> +};
> +module_platform_driver(zynqmp_r5_remoteproc_driver);
> +
> +MODULE_AUTHOR("Ben Levinsky <ben.levinsky@xilinx.com>");
> +MODULE_LICENSE("GPL v2");
> --
> 2.17.1
>
^ permalink raw reply [flat|nested] 13+ messages in thread