From: Andreas Dannenberg dannenberg@ti.com
Add and expose a new processor shutdown API that wraps the two TISCI messages involved in initiating a core shutdown. The API will first queue a message to have the DMSC wait for a certain processor boot status to happen followed by a message to trigger the actual shutdown- with both messages being sent without waiting or requesting for a response. Note that the processor shutdown API call will need to be followed up by user software placing the respective core into either WFE or WFI mode.
Signed-off-by: Andreas Dannenberg dannenberg@ti.com --- drivers/firmware/ti_sci.c | 191 ++++++++++++++++++++++++- drivers/firmware/ti_sci.h | 50 +++++++ include/linux/soc/ti/ti_sci_protocol.h | 4 + 3 files changed, 241 insertions(+), 4 deletions(-)
diff --git a/drivers/firmware/ti_sci.c b/drivers/firmware/ti_sci.c index eadb91e107..8c68f98788 100644 --- a/drivers/firmware/ti_sci.c +++ b/drivers/firmware/ti_sci.c @@ -101,7 +101,8 @@ struct ti_sci_info { * @msg_flags: Flag to set for the message * @buf: Buffer to be send to mailbox channel * @tx_message_size: transmit message size - * @rx_message_size: receive message size + * @rx_message_size: receive message size. may be set to zero for send-only + * transactions. * * Helper function which is used by various command functions that are * exposed to clients of this driver for allocating a message traffic event. @@ -121,7 +122,8 @@ static struct ti_sci_xfer *ti_sci_setup_one_xfer(struct ti_sci_info *info, /* Ensure we have sane transfer sizes */ if (rx_message_size > info->desc->max_msg_size || tx_message_size > info->desc->max_msg_size || - rx_message_size < sizeof(*hdr) || tx_message_size < sizeof(*hdr)) + (rx_message_size > 0 && rx_message_size < sizeof(*hdr)) || + tx_message_size < sizeof(*hdr)) return ERR_PTR(-ERANGE);
info->seq = ~info->seq; @@ -219,7 +221,9 @@ static inline int ti_sci_do_xfer(struct ti_sci_info *info,
xfer->tx_message.buf = (u32 *)secure_buf; xfer->tx_message.len += sizeof(secure_hdr); - xfer->rx_len += sizeof(secure_hdr); + + if (xfer->rx_len) + xfer->rx_len += sizeof(secure_hdr); }
/* Send the message */ @@ -230,7 +234,11 @@ static inline int ti_sci_do_xfer(struct ti_sci_info *info, return ret; }
- return ti_sci_get_response(info, xfer, &info->chan_rx); + /* Get response if requested */ + if (xfer->rx_len) + ret = ti_sci_get_response(info, xfer, &info->chan_rx); + + return ret; }
/** @@ -469,6 +477,49 @@ static int ti_sci_set_device_state(const struct ti_sci_handle *handle, return ret; }
+/** + * ti_sci_set_device_state_no_wait() - Set device state helper without + * requesting or waiting for a response. + * @handle: pointer to TI SCI handle + * @id: Device identifier + * @flags: flags to setup for the device + * @state: State to move the device to + * + * Return: 0 if all went well, else returns appropriate error value. + */ +static int ti_sci_set_device_state_no_wait(const struct ti_sci_handle *handle, + u32 id, u32 flags, u8 state) +{ + struct ti_sci_msg_req_set_device_state req; + struct ti_sci_info *info; + struct ti_sci_xfer *xfer; + int ret = 0; + + if (IS_ERR(handle)) + return PTR_ERR(handle); + if (!handle) + return -EINVAL; + + info = handle_to_ti_sci_info(handle); + + xfer = ti_sci_setup_one_xfer(info, TI_SCI_MSG_SET_DEVICE_STATE, + flags | TI_SCI_FLAG_REQ_GENERIC_NORESPONSE, + (u32 *)&req, sizeof(req), 0); + if (IS_ERR(xfer)) { + ret = PTR_ERR(xfer); + dev_err(info->dev, "Message alloc failed(%d)\n", ret); + return ret; + } + req.id = id; + req.state = state; + + ret = ti_sci_do_xfer(info, xfer); + if (ret) + dev_err(info->dev, "Mbox send fail %d\n", ret); + + return ret; +} + /** * ti_sci_get_device_state() - Get device state helper * @handle: Handle to the device @@ -2039,6 +2090,137 @@ static int ti_sci_cmd_get_proc_boot_status(const struct ti_sci_handle *handle, return ret; }
+/** + * ti_sci_proc_wait_boot_status_no_wait() - Helper function to wait for a + * processor boot status without requesting or + * waiting for a response. + * @proc_id: Processor ID this request is for + * @num_wait_iterations: Total number of iterations we will check before + * we will timeout and give up + * @num_match_iterations: How many iterations should we have continued + * status to account for status bits glitching. + * This is to make sure that match occurs for + * consecutive checks. This implies that the + * worst case should consider that the stable + * time should at the worst be num_wait_iterations + * num_match_iterations to prevent timeout. + * @delay_per_iteration_us: Specifies how long to wait (in micro seconds) + * between each status checks. This is the minimum + * duration, and overhead of register reads and + * checks are on top of this and can vary based on + * varied conditions. + * @delay_before_iterations_us: Specifies how long to wait (in micro seconds) + * before the very first check in the first + * iteration of status check loop. This is the + * minimum duration, and overhead of register + * reads and checks are. + * @status_flags_1_set_all_wait:If non-zero, Specifies that all bits of the + * status matching this field requested MUST be 1. + * @status_flags_1_set_any_wait:If non-zero, Specifies that at least one of the + * bits matching this field requested MUST be 1. + * @status_flags_1_clr_all_wait:If non-zero, Specifies that all bits of the + * status matching this field requested MUST be 0. + * @status_flags_1_clr_any_wait:If non-zero, Specifies that at least one of the + * bits matching this field requested MUST be 0. + * + * Return: 0 if all goes well, else appropriate error message + */ +static int +ti_sci_proc_wait_boot_status_no_wait(const struct ti_sci_handle *handle, + u8 proc_id, + u8 num_wait_iterations, + u8 num_match_iterations, + u8 delay_per_iteration_us, + u8 delay_before_iterations_us, + u32 status_flags_1_set_all_wait, + u32 status_flags_1_set_any_wait, + u32 status_flags_1_clr_all_wait, + u32 status_flags_1_clr_any_wait) +{ + struct ti_sci_msg_req_wait_proc_boot_status req; + struct ti_sci_info *info; + struct ti_sci_xfer *xfer; + int ret = 0; + + if (IS_ERR(handle)) + return PTR_ERR(handle); + if (!handle) + return -EINVAL; + + info = handle_to_ti_sci_info(handle); + + xfer = ti_sci_setup_one_xfer(info, TISCI_MSG_WAIT_PROC_BOOT_STATUS, + TI_SCI_FLAG_REQ_GENERIC_NORESPONSE, + (u32 *)&req, sizeof(req), 0); + if (IS_ERR(xfer)) { + ret = PTR_ERR(xfer); + dev_err(info->dev, "Message alloc failed(%d)\n", ret); + return ret; + } + req.processor_id = proc_id; + req.num_wait_iterations = num_wait_iterations; + req.num_match_iterations = num_match_iterations; + req.delay_per_iteration_us = delay_per_iteration_us; + req.delay_before_iterations_us = delay_before_iterations_us; + req.status_flags_1_set_all_wait = status_flags_1_set_all_wait; + req.status_flags_1_set_any_wait = status_flags_1_set_any_wait; + req.status_flags_1_clr_all_wait = status_flags_1_clr_all_wait; + req.status_flags_1_clr_any_wait = status_flags_1_clr_any_wait; + + ret = ti_sci_do_xfer(info, xfer); + if (ret) + dev_err(info->dev, "Mbox send fail %d\n", ret); + + return ret; +} + +/** + * ti_sci_cmd_proc_shutdown_no_wait() - Command to shutdown a core without + * requesting or waiting for a response. Note that this API call + * should be followed by placing the respective processor into + * either WFE or WFI mode. + * @handle: Pointer to TI SCI handle + * @proc_id: Processor ID this request is for + * + * Return: 0 if all went well, else returns appropriate error value. + */ +static int ti_sci_cmd_proc_shutdown_no_wait(const struct ti_sci_handle *handle, + u8 proc_id) +{ + int ret; + + /* + * Send the core boot status wait message waiting for either WFE or + * WFI without requesting or waiting for a TISCI response with the + * maximum wait time to give us the best chance to get to the WFE/WFI + * command that should follow the invocation of this API before the + * DMSC-internal processing of this command times out. Note that + * waiting for the R5 WFE/WFI flags will also work on an ARMV8 type + * core as the related flag bit positions are the same. + */ + ret = ti_sci_proc_wait_boot_status_no_wait(handle, proc_id, + U8_MAX, 100, U8_MAX, U8_MAX, + 0, PROC_BOOT_STATUS_FLAG_R5_WFE | PROC_BOOT_STATUS_FLAG_R5_WFI, + 0, 0); + if (ret) { + dev_err(info->dev, "Sending core %u wait message fail %d\n", + proc_id, ret); + return ret; + } + + /* + * Release a processor managed by TISCI without requesting or waiting + * for a response. + */ + ret = ti_sci_set_device_state_no_wait(handle, proc_id, 0, + MSG_DEVICE_SW_STATE_AUTO_OFF); + if (ret) + dev_err(info->dev, "Sending core %u shutdown message fail %d\n", + proc_id, ret); + + return ret; +} + /** * ti_sci_cmd_ring_config() - configure RA ring * @handle: pointer to TI SCI handle @@ -2687,6 +2869,7 @@ static void ti_sci_setup_ops(struct ti_sci_info *info) pops->set_proc_boot_ctrl = ti_sci_cmd_set_proc_boot_ctrl; pops->proc_auth_boot_image = ti_sci_cmd_proc_auth_boot_image; pops->get_proc_boot_status = ti_sci_cmd_get_proc_boot_status; + pops->proc_shutdown_no_wait = ti_sci_cmd_proc_shutdown_no_wait;
rops->config = ti_sci_cmd_ring_config; rops->get_config = ti_sci_cmd_ring_get_config; diff --git a/drivers/firmware/ti_sci.h b/drivers/firmware/ti_sci.h index a484b1fa40..69ff74d6a9 100644 --- a/drivers/firmware/ti_sci.h +++ b/drivers/firmware/ti_sci.h @@ -50,6 +50,7 @@ #define TISCI_MSG_SET_PROC_BOOT_CTRL 0xc101 #define TISCI_MSG_PROC_AUTH_BOOT_IMIAGE 0xc120 #define TISCI_MSG_GET_PROC_BOOT_STATUS 0xc400 +#define TISCI_MSG_WAIT_PROC_BOOT_STATUS 0xc401
/* Resource Management Requests */ #define TI_SCI_MSG_GET_RESOURCE_RANGE 0x1500 @@ -772,6 +773,55 @@ struct ti_sci_msg_resp_get_proc_boot_status { u32 status_flags; } __packed;
+/** + * struct ti_sci_msg_req_wait_proc_boot_status - Wait for a processor + * boot status + * @hdr: Generic Header + * @processor_id: ID of processor + * @num_wait_iterations: Total number of iterations we will check before + * we will timeout and give up + * @num_match_iterations: How many iterations should we have continued + * status to account for status bits glitching. + * This is to make sure that match occurs for + * consecutive checks. This implies that the + * worst case should consider that the stable + * time should at the worst be num_wait_iterations + * num_match_iterations to prevent timeout. + * @delay_per_iteration_us: Specifies how long to wait (in micro seconds) + * between each status checks. This is the minimum + * duration, and overhead of register reads and + * checks are on top of this and can vary based on + * varied conditions. + * @delay_before_iterations_us: Specifies how long to wait (in micro seconds) + * before the very first check in the first + * iteration of status check loop. This is the + * minimum duration, and overhead of register + * reads and checks are. + * @status_flags_1_set_all_wait:If non-zero, Specifies that all bits of the + * status matching this field requested MUST be 1. + * @status_flags_1_set_any_wait:If non-zero, Specifies that at least one of the + * bits matching this field requested MUST be 1. + * @status_flags_1_clr_all_wait:If non-zero, Specifies that all bits of the + * status matching this field requested MUST be 0. + * @status_flags_1_clr_any_wait:If non-zero, Specifies that at least one of the + * bits matching this field requested MUST be 0. + * + * Request type is TISCI_MSG_WAIT_PROC_BOOT_STATUS, response is appropriate + * message, or NACK in case of inability to satisfy request. + */ +struct ti_sci_msg_req_wait_proc_boot_status { + struct ti_sci_msg_hdr hdr; + u8 processor_id; + u8 num_wait_iterations; + u8 num_match_iterations; + u8 delay_per_iteration_us; + u8 delay_before_iterations_us; + u32 status_flags_1_set_all_wait; + u32 status_flags_1_set_any_wait; + u32 status_flags_1_clr_all_wait; + u32 status_flags_1_clr_any_wait; +} __packed; + /** * struct ti_sci_msg_rm_ring_cfg_req - Configure a Navigator Subsystem ring * diff --git a/include/linux/soc/ti/ti_sci_protocol.h b/include/linux/soc/ti/ti_sci_protocol.h index 842fb596f7..cd6e5853b4 100644 --- a/include/linux/soc/ti/ti_sci_protocol.h +++ b/include/linux/soc/ti/ti_sci_protocol.h @@ -266,6 +266,8 @@ struct ti_sci_core_ops { * @set_proc_boot_ctrl: Setup limited control flags in specific cases. * @proc_auth_boot_image: * @get_proc_boot_status: Get the state of physical processor + * @proc_shutdown_no_wait: Shutdown a core without requesting or waiting for a + * response. * * NOTE: for all these functions, the following parameters are generic in * nature: @@ -287,6 +289,8 @@ struct ti_sci_proc_ops { int (*get_proc_boot_status)(const struct ti_sci_handle *handle, u8 pid, u64 *bv, u32 *cfg_flags, u32 *ctrl_flags, u32 *sts_flags); + int (*proc_shutdown_no_wait)(const struct ti_sci_handle *handle, + u8 pid); };
#define TI_SCI_RING_MODE_RING (0)