blob: f34467b7b9e98eb26976ff050b00a06b5be7c04d [file] [log] [blame]
/* Copyright (c) 2011-2017, The Linux Foundation. All rights reserved.
*
* This program is free software; you can redistribute it and/or modify
* it under the terms of the GNU General Public License version 2 and
* only version 2 as published by the Free Software Foundation.
*
* This program is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
* GNU General Public License for more details.
*/
#include <linux/kernel.h>
#include <linux/errno.h>
#include <linux/slab.h>
#include <linux/init.h>
#include <linux/completion.h>
#include <linux/idr.h>
#include <linux/pm_runtime.h>
#include <linux/slimbus/slimbus.h>
#define SLIM_PORT_HDL(la, f, p) ((la)<<24 | (f) << 16 | (p))
#define SLIM_HDL_TO_LA(hdl) ((u32)((hdl) & 0xFF000000) >> 24)
#define SLIM_HDL_TO_FLOW(hdl) (((u32)(hdl) & 0xFF0000) >> 16)
#define SLIM_HDL_TO_PORT(hdl) ((u32)(hdl) & 0xFF)
#define SLIM_HDL_TO_CHIDX(hdl) ((u16)(hdl) & 0xFF)
#define SLIM_GRP_TO_NCHAN(hdl) ((u16)(hdl >> 8) & 0xFF)
#define SLIM_SLAVE_PORT(p, la) (((la)<<16) | (p))
#define SLIM_MGR_PORT(p) ((0xFF << 16) | (p))
#define SLIM_LA_MANAGER 0xFF
#define SLIM_START_GRP (1 << 8)
#define SLIM_END_GRP (1 << 9)
#define SLIM_MAX_INTR_COEFF_3 (SLIM_SL_PER_SUPERFRAME/3)
#define SLIM_MAX_INTR_COEFF_1 SLIM_SL_PER_SUPERFRAME
static DEFINE_MUTEX(slim_lock);
static DEFINE_IDR(ctrl_idr);
static struct device_type slim_dev_type;
static struct device_type slim_ctrl_type;
#define DEFINE_SLIM_LDEST_TXN(name, mc, len, rl, rbuf, wbuf, la) \
struct slim_msg_txn name = { rl, 0, mc, SLIM_MSG_DEST_LOGICALADDR, 0,\
len, 0, la, false, rbuf, wbuf, NULL, }
#define DEFINE_SLIM_BCAST_TXN(name, mc, len, rl, rbuf, wbuf, la) \
struct slim_msg_txn name = { rl, 0, mc, SLIM_MSG_DEST_BROADCAST, 0,\
len, 0, la, false, rbuf, wbuf, NULL, }
static const struct slim_device_id *slim_match(const struct slim_device_id *id,
const struct slim_device *slim_dev)
{
while (id->name[0]) {
if (strcmp(slim_dev->name, id->name) == 0)
return id;
id++;
}
return NULL;
}
const struct slim_device_id *slim_get_device_id(const struct slim_device *sdev)
{
const struct slim_driver *sdrv = to_slim_driver(sdev->dev.driver);
return slim_match(sdrv->id_table, sdev);
}
EXPORT_SYMBOL(slim_get_device_id);
static int slim_device_match(struct device *dev, struct device_driver *driver)
{
struct slim_device *slim_dev;
struct slim_driver *drv = to_slim_driver(driver);
if (dev->type == &slim_dev_type)
slim_dev = to_slim_device(dev);
else
return 0;
if (drv->id_table)
return slim_match(drv->id_table, slim_dev) != NULL;
if (driver->name)
return strcmp(slim_dev->name, driver->name) == 0;
return 0;
}
#ifdef CONFIG_PM_SLEEP
static int slim_legacy_suspend(struct device *dev, pm_message_t mesg)
{
struct slim_device *slim_dev = NULL;
struct slim_driver *driver;
if (dev->type == &slim_dev_type)
slim_dev = to_slim_device(dev);
if (!slim_dev || !dev->driver)
return 0;
driver = to_slim_driver(dev->driver);
if (!driver->suspend)
return 0;
return driver->suspend(slim_dev, mesg);
}
static int slim_legacy_resume(struct device *dev)
{
struct slim_device *slim_dev = NULL;
struct slim_driver *driver;
if (dev->type == &slim_dev_type)
slim_dev = to_slim_device(dev);
if (!slim_dev || !dev->driver)
return 0;
driver = to_slim_driver(dev->driver);
if (!driver->resume)
return 0;
return driver->resume(slim_dev);
}
static int slim_pm_suspend(struct device *dev)
{
const struct dev_pm_ops *pm = dev->driver ? dev->driver->pm : NULL;
if (pm)
return pm_generic_suspend(dev);
else
return slim_legacy_suspend(dev, PMSG_SUSPEND);
}
static int slim_pm_resume(struct device *dev)
{
const struct dev_pm_ops *pm = dev->driver ? dev->driver->pm : NULL;
if (pm)
return pm_generic_resume(dev);
else
return slim_legacy_resume(dev);
}
#else
#define slim_pm_suspend NULL
#define slim_pm_resume NULL
#endif
static const struct dev_pm_ops slimbus_pm = {
.suspend = slim_pm_suspend,
.resume = slim_pm_resume,
SET_RUNTIME_PM_OPS(
pm_generic_suspend,
pm_generic_resume,
NULL
)
};
struct bus_type slimbus_type = {
.name = "slimbus",
.match = slim_device_match,
.pm = &slimbus_pm,
};
EXPORT_SYMBOL(slimbus_type);
struct device slimbus_dev = {
.init_name = "slimbus",
};
static void __exit slimbus_exit(void)
{
device_unregister(&slimbus_dev);
bus_unregister(&slimbus_type);
}
static int __init slimbus_init(void)
{
int retval;
retval = bus_register(&slimbus_type);
if (!retval)
retval = device_register(&slimbus_dev);
if (retval)
bus_unregister(&slimbus_type);
return retval;
}
postcore_initcall(slimbus_init);
module_exit(slimbus_exit);
static int slim_drv_probe(struct device *dev)
{
const struct slim_driver *sdrv = to_slim_driver(dev->driver);
struct slim_device *sbdev = to_slim_device(dev);
struct slim_controller *ctrl = sbdev->ctrl;
if (sdrv->probe) {
int ret;
ret = sdrv->probe(sbdev);
if (ret)
return ret;
if (sdrv->device_up)
queue_work(ctrl->wq, &sbdev->wd);
return 0;
}
return -ENODEV;
}
static int slim_drv_remove(struct device *dev)
{
const struct slim_driver *sdrv = to_slim_driver(dev->driver);
struct slim_device *sbdev = to_slim_device(dev);
sbdev->notified = false;
if (sdrv->remove)
return sdrv->remove(to_slim_device(dev));
return -ENODEV;
}
static void slim_drv_shutdown(struct device *dev)
{
const struct slim_driver *sdrv = to_slim_driver(dev->driver);
if (sdrv->shutdown)
sdrv->shutdown(to_slim_device(dev));
}
/*
* slim_driver_register: Client driver registration with slimbus
* @drv:Client driver to be associated with client-device.
* This API will register the client driver with the slimbus
* It is called from the driver's module-init function.
*/
int slim_driver_register(struct slim_driver *drv)
{
drv->driver.bus = &slimbus_type;
if (drv->probe)
drv->driver.probe = slim_drv_probe;
if (drv->remove)
drv->driver.remove = slim_drv_remove;
if (drv->shutdown)
drv->driver.shutdown = slim_drv_shutdown;
return driver_register(&drv->driver);
}
EXPORT_SYMBOL(slim_driver_register);
/*
* slim_driver_unregister: Undo effects of slim_driver_register
* @drv: Client driver to be unregistered
*/
void slim_driver_unregister(struct slim_driver *drv)
{
if (drv)
driver_unregister(&drv->driver);
}
EXPORT_SYMBOL(slim_driver_unregister);
#define slim_ctrl_attr_gr NULL
static void slim_ctrl_release(struct device *dev)
{
struct slim_controller *ctrl = to_slim_controller(dev);
complete(&ctrl->dev_released);
}
static struct device_type slim_ctrl_type = {
.groups = slim_ctrl_attr_gr,
.release = slim_ctrl_release,
};
static struct slim_controller *slim_ctrl_get(struct slim_controller *ctrl)
{
if (!ctrl || !get_device(&ctrl->dev))
return NULL;
return ctrl;
}
static void slim_ctrl_put(struct slim_controller *ctrl)
{
if (ctrl)
put_device(&ctrl->dev);
}
#define slim_device_attr_gr NULL
#define slim_device_uevent NULL
static void slim_dev_release(struct device *dev)
{
struct slim_device *sbdev = to_slim_device(dev);
slim_ctrl_put(sbdev->ctrl);
}
static struct device_type slim_dev_type = {
.groups = slim_device_attr_gr,
.uevent = slim_device_uevent,
.release = slim_dev_release,
};
static void slim_report(struct work_struct *work)
{
struct slim_driver *sbdrv;
struct slim_device *sbdev =
container_of(work, struct slim_device, wd);
if (!sbdev->dev.driver)
return;
/* check if device-up or down needs to be called */
if ((!sbdev->reported && !sbdev->notified) ||
(sbdev->reported && sbdev->notified))
return;
sbdrv = to_slim_driver(sbdev->dev.driver);
/*
* address no longer valid, means device reported absent, whereas
* address valid, means device reported present
*/
if (sbdev->notified && !sbdev->reported) {
sbdev->notified = false;
if (sbdrv->device_down)
sbdrv->device_down(sbdev);
} else if (!sbdev->notified && sbdev->reported) {
sbdev->notified = true;
if (sbdrv->device_up)
sbdrv->device_up(sbdev);
}
}
static void slim_device_reset(struct work_struct *work)
{
struct slim_driver *sbdrv;
struct slim_device *sbdev =
container_of(work, struct slim_device, device_reset);
if (!sbdev->dev.driver)
return;
sbdrv = to_slim_driver(sbdev->dev.driver);
if (sbdrv && sbdrv->reset_device)
sbdrv->reset_device(sbdev);
}
/*
* slim_add_device: Add a new device without register board info.
* @ctrl: Controller to which this device is to be added to.
* Called when device doesn't have an explicit client-driver to be probed, or
* the client-driver is a module installed dynamically.
*/
int slim_add_device(struct slim_controller *ctrl, struct slim_device *sbdev)
{
sbdev->dev.bus = &slimbus_type;
sbdev->dev.parent = ctrl->dev.parent;
sbdev->dev.type = &slim_dev_type;
sbdev->dev.driver = NULL;
sbdev->ctrl = ctrl;
slim_ctrl_get(ctrl);
dev_set_name(&sbdev->dev, "%s", sbdev->name);
mutex_init(&sbdev->sldev_reconf);
INIT_LIST_HEAD(&sbdev->mark_define);
INIT_LIST_HEAD(&sbdev->mark_suspend);
INIT_LIST_HEAD(&sbdev->mark_removal);
INIT_WORK(&sbdev->wd, slim_report);
INIT_WORK(&sbdev->device_reset, slim_device_reset);
mutex_lock(&ctrl->m_ctrl);
list_add_tail(&sbdev->dev_list, &ctrl->devs);
mutex_unlock(&ctrl->m_ctrl);
/* probe slave on this controller */
return device_register(&sbdev->dev);
}
EXPORT_SYMBOL(slim_add_device);
struct sbi_boardinfo {
struct list_head list;
struct slim_boardinfo board_info;
};
static LIST_HEAD(board_list);
static LIST_HEAD(slim_ctrl_list);
static DEFINE_MUTEX(board_lock);
/* If controller is not present, only add to boards list */
static void slim_match_ctrl_to_boardinfo(struct slim_controller *ctrl,
struct slim_boardinfo *bi)
{
int ret;
if (ctrl->nr != bi->bus_num)
return;
ret = slim_add_device(ctrl, bi->slim_slave);
if (ret != 0)
dev_err(ctrl->dev.parent, "can't create new device for %s\n",
bi->slim_slave->name);
}
/*
* slim_register_board_info: Board-initialization routine.
* @info: List of all devices on all controllers present on the board.
* @n: number of entries.
* API enumerates respective devices on corresponding controller.
* Called from board-init function.
*/
int slim_register_board_info(struct slim_boardinfo const *info, unsigned int n)
{
struct sbi_boardinfo *bi;
int i;
bi = kcalloc(n, sizeof(*bi), GFP_KERNEL);
if (!bi)
return -ENOMEM;
for (i = 0; i < n; i++, bi++, info++) {
struct slim_controller *ctrl;
memcpy(&bi->board_info, info, sizeof(*info));
mutex_lock(&board_lock);
list_add_tail(&bi->list, &board_list);
list_for_each_entry(ctrl, &slim_ctrl_list, list)
slim_match_ctrl_to_boardinfo(ctrl, &bi->board_info);
mutex_unlock(&board_lock);
}
return 0;
}
EXPORT_SYMBOL(slim_register_board_info);
/*
* slim_ctrl_add_boarddevs: Add devices registered by board-info
* @ctrl: Controller to which these devices are to be added to.
* This API is called by controller when it is up and running.
* If devices on a controller were registered before controller,
* this will make sure that they get probed when controller is up.
*/
void slim_ctrl_add_boarddevs(struct slim_controller *ctrl)
{
struct sbi_boardinfo *bi;
mutex_lock(&board_lock);
list_add_tail(&ctrl->list, &slim_ctrl_list);
list_for_each_entry(bi, &board_list, list)
slim_match_ctrl_to_boardinfo(ctrl, &bi->board_info);
mutex_unlock(&board_lock);
}
EXPORT_SYMBOL(slim_ctrl_add_boarddevs);
/*
* slim_busnum_to_ctrl: Map bus number to controller
* @busnum: Bus number
* Returns controller representing this bus number
*/
struct slim_controller *slim_busnum_to_ctrl(u32 bus_num)
{
struct slim_controller *ctrl;
mutex_lock(&board_lock);
list_for_each_entry(ctrl, &slim_ctrl_list, list)
if (bus_num == ctrl->nr) {
mutex_unlock(&board_lock);
return ctrl;
}
mutex_unlock(&board_lock);
return NULL;
}
EXPORT_SYMBOL(slim_busnum_to_ctrl);
static int slim_register_controller(struct slim_controller *ctrl)
{
int ret = 0;
/* Can't register until after driver model init */
if (WARN_ON(!slimbus_type.p)) {
ret = -EPROBE_DEFER;
goto out_list;
}
dev_set_name(&ctrl->dev, "sb-%d", ctrl->nr);
ctrl->dev.bus = &slimbus_type;
ctrl->dev.type = &slim_ctrl_type;
ctrl->num_dev = 0;
if (!ctrl->min_cg)
ctrl->min_cg = SLIM_MIN_CLK_GEAR;
if (!ctrl->max_cg)
ctrl->max_cg = SLIM_MAX_CLK_GEAR;
spin_lock_init(&ctrl->txn_lock);
mutex_init(&ctrl->m_ctrl);
mutex_init(&ctrl->sched.m_reconf);
ret = device_register(&ctrl->dev);
if (ret)
goto out_list;
dev_dbg(&ctrl->dev, "Bus [%s] registered:dev:%p\n", ctrl->name,
&ctrl->dev);
if (ctrl->nports) {
ctrl->ports = kcalloc(ctrl->nports, sizeof(struct slim_port),
GFP_KERNEL);
if (!ctrl->ports) {
ret = -ENOMEM;
goto err_port_failed;
}
}
if (ctrl->nchans) {
ctrl->chans = kcalloc(ctrl->nchans, sizeof(struct slim_ich),
GFP_KERNEL);
if (!ctrl->chans) {
ret = -ENOMEM;
goto err_chan_failed;
}
ctrl->sched.chc1 = kcalloc(ctrl->nchans,
sizeof(struct slim_ich *), GFP_KERNEL);
if (!ctrl->sched.chc1) {
kfree(ctrl->chans);
ret = -ENOMEM;
goto err_chan_failed;
}
ctrl->sched.chc3 = kcalloc(ctrl->nchans,
sizeof(struct slim_ich *), GFP_KERNEL);
if (!ctrl->sched.chc3) {
kfree(ctrl->sched.chc1);
kfree(ctrl->chans);
ret = -ENOMEM;
goto err_chan_failed;
}
}
#ifdef DEBUG
ctrl->sched.slots = kzalloc(SLIM_SL_PER_SUPERFRAME, GFP_KERNEL);
#endif
init_completion(&ctrl->pause_comp);
INIT_LIST_HEAD(&ctrl->devs);
ctrl->wq = create_singlethread_workqueue(dev_name(&ctrl->dev));
if (!ctrl->wq)
goto err_workq_failed;
return 0;
err_workq_failed:
kfree(ctrl->sched.chc3);
kfree(ctrl->sched.chc1);
kfree(ctrl->chans);
err_chan_failed:
kfree(ctrl->ports);
err_port_failed:
device_unregister(&ctrl->dev);
out_list:
mutex_lock(&slim_lock);
idr_remove(&ctrl_idr, ctrl->nr);
mutex_unlock(&slim_lock);
return ret;
}
/* slim_remove_device: Remove the effect of slim_add_device() */
void slim_remove_device(struct slim_device *sbdev)
{
struct slim_controller *ctrl = sbdev->ctrl;
mutex_lock(&ctrl->m_ctrl);
list_del_init(&sbdev->dev_list);
mutex_unlock(&ctrl->m_ctrl);
device_unregister(&sbdev->dev);
}
EXPORT_SYMBOL(slim_remove_device);
static void slim_ctrl_remove_device(struct slim_controller *ctrl,
struct slim_boardinfo *bi)
{
if (ctrl->nr == bi->bus_num)
slim_remove_device(bi->slim_slave);
}
/*
* slim_del_controller: Controller tear-down.
* Controller added with the above API is teared down using this API.
*/
int slim_del_controller(struct slim_controller *ctrl)
{
struct slim_controller *found;
struct sbi_boardinfo *bi;
/* First make sure that this bus was added */
mutex_lock(&slim_lock);
found = idr_find(&ctrl_idr, ctrl->nr);
mutex_unlock(&slim_lock);
if (found != ctrl)
return -EINVAL;
/* Remove all clients */
mutex_lock(&board_lock);
list_for_each_entry(bi, &board_list, list)
slim_ctrl_remove_device(ctrl, &bi->board_info);
mutex_unlock(&board_lock);
init_completion(&ctrl->dev_released);
device_unregister(&ctrl->dev);
wait_for_completion(&ctrl->dev_released);
list_del(&ctrl->list);
destroy_workqueue(ctrl->wq);
/* free bus id */
mutex_lock(&slim_lock);
idr_remove(&ctrl_idr, ctrl->nr);
mutex_unlock(&slim_lock);
kfree(ctrl->sched.chc1);
kfree(ctrl->sched.chc3);
#ifdef DEBUG
kfree(ctrl->sched.slots);
#endif
kfree(ctrl->chans);
kfree(ctrl->ports);
return 0;
}
EXPORT_SYMBOL(slim_del_controller);
/*
* slim_add_numbered_controller: Controller bring-up.
* @ctrl: Controller to be registered.
* A controller is registered with the framework using this API. ctrl->nr is the
* desired number with which slimbus framework registers the controller.
* Function will return -EBUSY if the number is in use.
*/
int slim_add_numbered_controller(struct slim_controller *ctrl)
{
int id;
mutex_lock(&slim_lock);
id = idr_alloc(&ctrl_idr, ctrl, ctrl->nr, ctrl->nr + 1, GFP_KERNEL);
mutex_unlock(&slim_lock);
if (id < 0)
return id;
ctrl->nr = id;
return slim_register_controller(ctrl);
}
EXPORT_SYMBOL(slim_add_numbered_controller);
/*
* slim_report_absent: Controller calls this function when a device
* reports absent, OR when the device cannot be communicated with
* @sbdev: Device that cannot be reached, or sent report absent
*/
void slim_report_absent(struct slim_device *sbdev)
{
struct slim_controller *ctrl;
int i;
if (!sbdev)
return;
ctrl = sbdev->ctrl;
if (!ctrl)
return;
/* invalidate logical addresses */
mutex_lock(&ctrl->m_ctrl);
for (i = 0; i < ctrl->num_dev; i++) {
if (sbdev->laddr == ctrl->addrt[i].laddr)
ctrl->addrt[i].valid = false;
}
mutex_unlock(&ctrl->m_ctrl);
sbdev->reported = false;
queue_work(ctrl->wq, &sbdev->wd);
}
EXPORT_SYMBOL(slim_report_absent);
static int slim_remove_ch(struct slim_controller *ctrl, struct slim_ich *slc);
/*
* slim_framer_booted: This function is called by controller after the active
* framer has booted (using Bus Reset sequence, or after it has shutdown and has
* come back up). Components, devices on the bus may be in undefined state,
* and this function triggers their drivers to do the needful
* to bring them back in Reset state so that they can acquire sync, report
* present and be operational again.
*/
void slim_framer_booted(struct slim_controller *ctrl)
{
struct slim_device *sbdev;
struct list_head *pos, *next;
int i;
if (!ctrl)
return;
/* Since framer has rebooted, reset all data channels */
mutex_lock(&ctrl->sched.m_reconf);
for (i = 0; i < ctrl->nchans; i++) {
struct slim_ich *slc = &ctrl->chans[i];
if (slc->state > SLIM_CH_DEFINED)
slim_remove_ch(ctrl, slc);
}
mutex_unlock(&ctrl->sched.m_reconf);
mutex_lock(&ctrl->m_ctrl);
list_for_each_safe(pos, next, &ctrl->devs) {
sbdev = list_entry(pos, struct slim_device, dev_list);
if (sbdev)
queue_work(ctrl->wq, &sbdev->device_reset);
}
mutex_unlock(&ctrl->m_ctrl);
}
EXPORT_SYMBOL(slim_framer_booted);
/*
* slim_msg_response: Deliver Message response received from a device to the
* framework.
* @ctrl: Controller handle
* @reply: Reply received from the device
* @len: Length of the reply
* @tid: Transaction ID received with which framework can associate reply.
* Called by controller to inform framework about the response received.
* This helps in making the API asynchronous, and controller-driver doesn't need
* to manage 1 more table other than the one managed by framework mapping TID
* with buffers
*/
void slim_msg_response(struct slim_controller *ctrl, u8 *reply, u8 tid, u8 len)
{
int i;
unsigned long flags;
bool async;
struct slim_msg_txn *txn;
spin_lock_irqsave(&ctrl->txn_lock, flags);
txn = ctrl->txnt[tid];
if (txn == NULL || txn->rbuf == NULL) {
spin_unlock_irqrestore(&ctrl->txn_lock, flags);
if (txn == NULL)
dev_err(&ctrl->dev, "Got response to invalid TID:%d, len:%d",
tid, len);
else
dev_err(&ctrl->dev, "Invalid client buffer passed\n");
return;
}
async = txn->async;
for (i = 0; i < len; i++)
txn->rbuf[i] = reply[i];
if (txn->comp)
complete(txn->comp);
ctrl->txnt[tid] = NULL;
spin_unlock_irqrestore(&ctrl->txn_lock, flags);
if (async)
kfree(txn);
}
EXPORT_SYMBOL(slim_msg_response);
static int slim_processtxn(struct slim_controller *ctrl,
struct slim_msg_txn *txn, bool need_tid)
{
u8 i = 0;
int ret = 0;
unsigned long flags;
if (need_tid) {
spin_lock_irqsave(&ctrl->txn_lock, flags);
for (i = 0; i < ctrl->last_tid; i++) {
if (ctrl->txnt[i] == NULL)
break;
}
if (i >= ctrl->last_tid) {
if (ctrl->last_tid == 255) {
spin_unlock_irqrestore(&ctrl->txn_lock, flags);
return -ENOMEM;
}
ctrl->last_tid++;
}
ctrl->txnt[i] = txn;
txn->tid = i;
spin_unlock_irqrestore(&ctrl->txn_lock, flags);
}
ret = ctrl->xfer_msg(ctrl, txn);
return ret;
}
static int ctrl_getlogical_addr(struct slim_controller *ctrl, const u8 *eaddr,
u8 e_len, u8 *entry)
{
u8 i;
for (i = 0; i < ctrl->num_dev; i++) {
if (ctrl->addrt[i].valid &&
memcmp(ctrl->addrt[i].eaddr, eaddr, e_len) == 0) {
*entry = i;
return 0;
}
}
return -ENXIO;
}
/*
* slim_assign_laddr: Assign logical address to a device enumerated.
* @ctrl: Controller with which device is enumerated.
* @e_addr: 6-byte elemental address of the device.
* @e_len: buffer length for e_addr
* @laddr: Return logical address (if valid flag is false)
* @valid: true if laddr holds a valid address that controller wants to
* set for this enumeration address. Otherwise framework sets index into
* address table as logical address.
* Called by controller in response to REPORT_PRESENT. Framework will assign
* a logical address to this enumeration address.
* Function returns -EXFULL to indicate that all logical addresses are already
* taken.
*/
int slim_assign_laddr(struct slim_controller *ctrl, const u8 *e_addr,
u8 e_len, u8 *laddr, bool valid)
{
int ret;
u8 i = 0;
bool exists = false;
struct slim_device *sbdev;
struct list_head *pos, *next;
void *new_addrt = NULL;
mutex_lock(&ctrl->m_ctrl);
/* already assigned */
if (ctrl_getlogical_addr(ctrl, e_addr, e_len, &i) == 0) {
*laddr = ctrl->addrt[i].laddr;
exists = true;
} else {
if (ctrl->num_dev >= 254) {
ret = -EXFULL;
goto ret_assigned_laddr;
}
for (i = 0; i < ctrl->num_dev; i++) {
if (ctrl->addrt[i].valid == false)
break;
}
if (i == ctrl->num_dev) {
new_addrt = krealloc(ctrl->addrt,
(ctrl->num_dev + 1) *
sizeof(struct slim_addrt),
GFP_KERNEL);
if (!new_addrt) {
ret = -ENOMEM;
goto ret_assigned_laddr;
}
ctrl->addrt = new_addrt;
ctrl->num_dev++;
}
memcpy(ctrl->addrt[i].eaddr, e_addr, e_len);
ctrl->addrt[i].valid = true;
/* Preferred address is index into table */
if (!valid)
*laddr = i;
}
ret = ctrl->set_laddr(ctrl, (const u8 *)&ctrl->addrt[i].eaddr, 6,
*laddr);
if (ret) {
ctrl->addrt[i].valid = false;
goto ret_assigned_laddr;
}
ctrl->addrt[i].laddr = *laddr;
dev_dbg(&ctrl->dev, "setting slimbus l-addr:%x\n", *laddr);
ret_assigned_laddr:
mutex_unlock(&ctrl->m_ctrl);
if (exists || ret)
return ret;
pr_info("slimbus:%d laddr:0x%x, EAPC:0x%x:0x%x", ctrl->nr, *laddr,
e_addr[1], e_addr[2]);
mutex_lock(&ctrl->m_ctrl);
list_for_each_safe(pos, next, &ctrl->devs) {
sbdev = list_entry(pos, struct slim_device, dev_list);
if (memcmp(sbdev->e_addr, e_addr, 6) == 0) {
struct slim_driver *sbdrv;
sbdev->laddr = *laddr;
sbdev->reported = true;
if (sbdev->dev.driver) {
sbdrv = to_slim_driver(sbdev->dev.driver);
if (sbdrv->device_up)
queue_work(ctrl->wq, &sbdev->wd);
}
break;
}
}
mutex_unlock(&ctrl->m_ctrl);
return 0;
}
EXPORT_SYMBOL(slim_assign_laddr);
/*
* slim_get_logical_addr: Return the logical address of a slimbus device.
* @sb: client handle requesting the adddress.
* @e_addr: Elemental address of the device.
* @e_len: Length of e_addr
* @laddr: output buffer to store the address
* context: can sleep
* -EINVAL is returned in case of invalid parameters, and -ENXIO is returned if
* the device with this elemental address is not found.
*/
int slim_get_logical_addr(struct slim_device *sb, const u8 *e_addr,
u8 e_len, u8 *laddr)
{
int ret = 0;
u8 entry;
struct slim_controller *ctrl = sb->ctrl;
if (!ctrl || !laddr || !e_addr || e_len != 6)
return -EINVAL;
mutex_lock(&ctrl->m_ctrl);
ret = ctrl_getlogical_addr(ctrl, e_addr, e_len, &entry);
if (!ret)
*laddr = ctrl->addrt[entry].laddr;
mutex_unlock(&ctrl->m_ctrl);
if (ret == -ENXIO && ctrl->get_laddr) {
ret = ctrl->get_laddr(ctrl, e_addr, e_len, laddr);
if (!ret)
ret = slim_assign_laddr(ctrl, e_addr, e_len, laddr,
true);
}
return ret;
}
EXPORT_SYMBOL(slim_get_logical_addr);
static int slim_ele_access_sanity(struct slim_ele_access *msg, int oper,
u8 *rbuf, const u8 *wbuf, u8 len)
{
if (!msg || msg->num_bytes > 16 || msg->start_offset + len > 0xC00)
return -EINVAL;
switch (oper) {
case SLIM_MSG_MC_REQUEST_VALUE:
case SLIM_MSG_MC_REQUEST_INFORMATION:
if (rbuf == NULL)
return -EINVAL;
return 0;
case SLIM_MSG_MC_CHANGE_VALUE:
case SLIM_MSG_MC_CLEAR_INFORMATION:
if (wbuf == NULL)
return -EINVAL;
return 0;
case SLIM_MSG_MC_REQUEST_CHANGE_VALUE:
case SLIM_MSG_MC_REQUEST_CLEAR_INFORMATION:
if (rbuf == NULL || wbuf == NULL)
return -EINVAL;
return 0;
default:
return -EINVAL;
}
}
static u16 slim_slicecodefromsize(u32 req)
{
u8 codetosize[8] = {1, 2, 3, 4, 6, 8, 12, 16};
if (req >= 8)
return 0;
else
return codetosize[req];
}
static u16 slim_slicesize(u32 code)
{
u8 sizetocode[16] = {0, 1, 2, 3, 3, 4, 4, 5, 5, 5, 5, 6, 6, 6, 6, 7};
if (code == 0)
code = 1;
if (code > 16)
code = 16;
return sizetocode[code - 1];
}
/* Message APIs Unicast message APIs used by slimbus slave drivers */
/*
* Message API access routines.
* @sb: client handle requesting elemental message reads, writes.
* @msg: Input structure for start-offset, number of bytes to read.
* @rbuf: data buffer to be filled with values read.
* @len: data buffer size
* @wbuf: data buffer containing value/information to be written
* context: can sleep
* Returns:
* -EINVAL: Invalid parameters
* -ETIMEDOUT: If controller could not complete the request. This may happen if
* the bus lines are not clocked, controller is not powered-on, slave with
* given address is not enumerated/responding.
*/
int slim_request_val_element(struct slim_device *sb,
struct slim_ele_access *msg, u8 *buf, u8 len)
{
struct slim_controller *ctrl = sb->ctrl;
if (!ctrl)
return -EINVAL;
return slim_xfer_msg(ctrl, sb, msg, SLIM_MSG_MC_REQUEST_VALUE, buf,
NULL, len);
}
EXPORT_SYMBOL(slim_request_val_element);
int slim_request_inf_element(struct slim_device *sb,
struct slim_ele_access *msg, u8 *buf, u8 len)
{
struct slim_controller *ctrl = sb->ctrl;
if (!ctrl)
return -EINVAL;
return slim_xfer_msg(ctrl, sb, msg, SLIM_MSG_MC_REQUEST_INFORMATION,
buf, NULL, len);
}
EXPORT_SYMBOL(slim_request_inf_element);
int slim_change_val_element(struct slim_device *sb, struct slim_ele_access *msg,
const u8 *buf, u8 len)
{
struct slim_controller *ctrl = sb->ctrl;
if (!ctrl)
return -EINVAL;
return slim_xfer_msg(ctrl, sb, msg, SLIM_MSG_MC_CHANGE_VALUE, NULL, buf,
len);
}
EXPORT_SYMBOL(slim_change_val_element);
int slim_clear_inf_element(struct slim_device *sb, struct slim_ele_access *msg,
u8 *buf, u8 len)
{
struct slim_controller *ctrl = sb->ctrl;
if (!ctrl)
return -EINVAL;
return slim_xfer_msg(ctrl, sb, msg, SLIM_MSG_MC_CLEAR_INFORMATION, NULL,
buf, len);
}
EXPORT_SYMBOL(slim_clear_inf_element);
int slim_request_change_val_element(struct slim_device *sb,
struct slim_ele_access *msg, u8 *rbuf,
const u8 *wbuf, u8 len)
{
struct slim_controller *ctrl = sb->ctrl;
if (!ctrl)
return -EINVAL;
return slim_xfer_msg(ctrl, sb, msg, SLIM_MSG_MC_REQUEST_CHANGE_VALUE,
rbuf, wbuf, len);
}
EXPORT_SYMBOL(slim_request_change_val_element);
int slim_request_clear_inf_element(struct slim_device *sb,
struct slim_ele_access *msg, u8 *rbuf,
const u8 *wbuf, u8 len)
{
struct slim_controller *ctrl = sb->ctrl;
if (!ctrl)
return -EINVAL;
return slim_xfer_msg(ctrl, sb, msg,
SLIM_MSG_MC_REQUEST_CLEAR_INFORMATION,
rbuf, wbuf, len);
}
EXPORT_SYMBOL(slim_request_clear_inf_element);
/*
* Broadcast message API:
* call this API directly with sbdev = NULL.
* For broadcast reads, make sure that buffers are big-enough to incorporate
* replies from all logical addresses.
* All controllers may not support broadcast
*/
int slim_xfer_msg(struct slim_controller *ctrl, struct slim_device *sbdev,
struct slim_ele_access *msg, u16 mc, u8 *rbuf,
const u8 *wbuf, u8 len)
{
DECLARE_COMPLETION_ONSTACK(complete);
DEFINE_SLIM_LDEST_TXN(txn_stack, mc, len, 6, rbuf, wbuf, sbdev->laddr);
struct slim_msg_txn *txn;
int ret;
u16 sl, cur;
if (msg->comp && rbuf) {
txn = kmalloc(sizeof(struct slim_msg_txn),
GFP_KERNEL);
if (IS_ERR_OR_NULL(txn))
return PTR_ERR(txn);
*txn = txn_stack;
txn->async = true;
txn->comp = msg->comp;
} else {
txn = &txn_stack;
if (rbuf)
txn->comp = &complete;
}
ret = slim_ele_access_sanity(msg, mc, rbuf, wbuf, len);
if (ret)
goto xfer_err;
sl = slim_slicesize(len);
dev_dbg(&ctrl->dev, "SB xfer msg:os:%x, len:%d, MC:%x, sl:%x\n",
msg->start_offset, len, mc, sl);
cur = slim_slicecodefromsize(sl);
txn->ec = ((sl | (1 << 3)) | ((msg->start_offset & 0xFFF) << 4));
if (wbuf)
txn->rl += len;
if (rbuf) {
unsigned long flags;
txn->rl++;
ret = slim_processtxn(ctrl, txn, true);
/* sync read */
if (!ret && !msg->comp) {
ret = wait_for_completion_timeout(&complete, HZ);
if (!ret) {
dev_err(&ctrl->dev, "slimbus Read timed out");
spin_lock_irqsave(&ctrl->txn_lock, flags);
/* Invalidate the transaction */
ctrl->txnt[txn->tid] = NULL;
spin_unlock_irqrestore(&ctrl->txn_lock, flags);
ret = -ETIMEDOUT;
} else
ret = 0;
} else if (ret < 0 && !msg->comp) {
dev_err(&ctrl->dev, "slimbus Read error");
spin_lock_irqsave(&ctrl->txn_lock, flags);
/* Invalidate the transaction */
ctrl->txnt[txn->tid] = NULL;
spin_unlock_irqrestore(&ctrl->txn_lock, flags);
}
} else
ret = slim_processtxn(ctrl, txn, false);
xfer_err:
return ret;
}
EXPORT_SYMBOL(slim_xfer_msg);
/*
* User message:
* slim_user_msg: Send user message that is interpreted by destination device
* @sb: Client handle sending the message
* @la: Destination device for this user message
* @mt: Message Type (Soruce-referred, or Destination-referred)
* @mc: Message Code
* @msg: Message structure (start offset, number of bytes) to be sent
* @buf: data buffer to be sent
* @len: data buffer size in bytes
*/
int slim_user_msg(struct slim_device *sb, u8 la, u8 mt, u8 mc,
struct slim_ele_access *msg, u8 *buf, u8 len)
{
if (!sb || !sb->ctrl || !msg || mt == SLIM_MSG_MT_CORE)
return -EINVAL;
if (!sb->ctrl->xfer_user_msg)
return -EPROTONOSUPPORT;
return sb->ctrl->xfer_user_msg(sb->ctrl, la, mt, mc, msg, buf, len);
}
EXPORT_SYMBOL(slim_user_msg);
/*
* Queue bulk of message writes:
* slim_bulk_msg_write: Write bulk of messages (e.g. downloading FW)
* @sb: Client handle sending these messages
* @la: Destination device for these messages
* @mt: Message Type
* @mc: Message Code
* @msgs: List of messages to be written in bulk
* @n: Number of messages in the list
* @cb: Callback if client needs this to be non-blocking
* @ctx: Context for this callback
* If supported by controller, this message list will be sent in bulk to the HW
* If the client specifies this to be non-blocking, the callback will be
* called from atomic context.
*/
int slim_bulk_msg_write(struct slim_device *sb, u8 mt, u8 mc,
struct slim_val_inf msgs[], int n,
int (*comp_cb)(void *ctx, int err), void *ctx)
{
int i, ret = 0;
if (!sb || !sb->ctrl || !msgs || n <= 0)
return -EINVAL;
if (!sb->ctrl->xfer_bulk_wr) {
pr_warn("controller does not support bulk WR, serializing");
for (i = 0; i < n; i++) {
struct slim_ele_access ele;
ele.comp = NULL;
ele.start_offset = msgs[i].start_offset;
ele.num_bytes = msgs[i].num_bytes;
ret = slim_xfer_msg(sb->ctrl, sb, &ele, mc,
msgs[i].rbuf, msgs[i].wbuf,
ele.num_bytes);
if (ret)
return ret;
}
return ret;
}
return sb->ctrl->xfer_bulk_wr(sb->ctrl, sb->laddr, mt, mc, msgs, n,
comp_cb, ctx);
}
EXPORT_SYMBOL(slim_bulk_msg_write);
/*
* slim_alloc_mgrports: Allocate port on manager side.
* @sb: device/client handle.
* @req: Port request type.
* @nports: Number of ports requested
* @rh: output buffer to store the port handles
* @hsz: size of buffer storing handles
* context: can sleep
* This port will be typically used by SW. e.g. client driver wants to receive
* some data from audio codec HW using a data channel.
* Port allocated using this API will be used to receive the data.
* If half-duplex ports are requested, two adjacent ports are allocated for
* 1 half-duplex port. So the handle-buffer size should be twice the number
* of half-duplex ports to be allocated.
* -EDQUOT is returned if all ports are in use.
*/
int slim_alloc_mgrports(struct slim_device *sb, enum slim_port_req req,
int nports, u32 *rh, int hsz)
{
int i, j;
int ret = -EINVAL;
int nphysp = nports;
struct slim_controller *ctrl = sb->ctrl;
if (!rh || !ctrl)
return -EINVAL;
if (req == SLIM_REQ_HALF_DUP)
nphysp *= 2;
if (hsz/sizeof(u32) < nphysp)
return -EINVAL;
mutex_lock(&ctrl->m_ctrl);
for (i = 0; i < ctrl->nports; i++) {
bool multiok = true;
if (ctrl->ports[i].state != SLIM_P_FREE)
continue;
/* Start half duplex channel at even port */
if (req == SLIM_REQ_HALF_DUP && (i % 2))
continue;
/* Allocate ports contiguously for multi-ch */
if (ctrl->nports < (i + nphysp)) {
i = ctrl->nports;
break;
}
if (req == SLIM_REQ_MULTI_CH) {
multiok = true;
for (j = i; j < i + nphysp; j++) {
if (ctrl->ports[j].state != SLIM_P_FREE) {
multiok = false;
break;
}
}
if (!multiok)
continue;
}
break;
}
if (i >= ctrl->nports) {
ret = -EDQUOT;
goto alloc_err;
}
ret = 0;
for (j = i; j < i + nphysp; j++) {
ctrl->ports[j].state = SLIM_P_UNCFG;
ctrl->ports[j].req = req;
if (req == SLIM_REQ_HALF_DUP && (j % 2))
ctrl->ports[j].flow = SLIM_SINK;
else
ctrl->ports[j].flow = SLIM_SRC;
if (ctrl->alloc_port)
ret = ctrl->alloc_port(ctrl, j);
if (ret) {
for (; j >= i; j--)
ctrl->ports[j].state = SLIM_P_FREE;
goto alloc_err;
}
*rh++ = SLIM_PORT_HDL(SLIM_LA_MANAGER, 0, j);
}
alloc_err:
mutex_unlock(&ctrl->m_ctrl);
return ret;
}
EXPORT_SYMBOL(slim_alloc_mgrports);
/* Deallocate the port(s) allocated using the API above */
int slim_dealloc_mgrports(struct slim_device *sb, u32 *hdl, int nports)
{
int i;
struct slim_controller *ctrl = sb->ctrl;
if (!ctrl || !hdl)
return -EINVAL;
mutex_lock(&ctrl->m_ctrl);
for (i = 0; i < nports; i++) {
u8 pn;
pn = SLIM_HDL_TO_PORT(hdl[i]);
if (pn >= ctrl->nports || ctrl->ports[pn].state == SLIM_P_CFG) {
int j, ret;
if (pn >= ctrl->nports) {
dev_err(&ctrl->dev, "invalid port number");
ret = -EINVAL;
} else {
dev_err(&ctrl->dev,
"Can't dealloc connected port:%d", i);
ret = -EISCONN;
}
for (j = i - 1; j >= 0; j--) {
pn = SLIM_HDL_TO_PORT(hdl[j]);
ctrl->ports[pn].state = SLIM_P_UNCFG;
}
mutex_unlock(&ctrl->m_ctrl);
return ret;
}
if (ctrl->dealloc_port)
ctrl->dealloc_port(ctrl, pn);
ctrl->ports[pn].state = SLIM_P_FREE;
}
mutex_unlock(&ctrl->m_ctrl);
return 0;
}
EXPORT_SYMBOL(slim_dealloc_mgrports);
/*
* slim_config_mgrports: Configure manager side ports
* @sb: device/client handle.
* @ph: array of port handles for which this configuration is valid
* @nports: Number of ports in ph
* @cfg: configuration requested for port(s)
* Configure port settings if they are different than the default ones.
* Returns success if the config could be applied. Returns -EISCONN if the
* port is in use
*/
int slim_config_mgrports(struct slim_device *sb, u32 *ph, int nports,
struct slim_port_cfg *cfg)
{
int i;
struct slim_controller *ctrl;
if (!sb || !ph || !nports || !sb->ctrl || !cfg)
return -EINVAL;
ctrl = sb->ctrl;
mutex_lock(&ctrl->sched.m_reconf);
for (i = 0; i < nports; i++) {
u8 pn = SLIM_HDL_TO_PORT(ph[i]);
if (ctrl->ports[pn].state == SLIM_P_CFG)
return -EISCONN;
ctrl->ports[pn].cfg = *cfg;
}
mutex_unlock(&ctrl->sched.m_reconf);
return 0;
}
EXPORT_SYMBOL(slim_config_mgrports);
/*
* slim_get_slaveport: Get slave port handle
* @la: slave device logical address.
* @idx: port index at slave
* @rh: return handle
* @flw: Flow type (source or destination)
* This API only returns a slave port's representation as expected by slimbus
* driver. This port is not managed by the slimbus driver. Caller is expected
* to have visibility of this port since it's a device-port.
*/
int slim_get_slaveport(u8 la, int idx, u32 *rh, enum slim_port_flow flw)
{
if (rh == NULL)
return -EINVAL;
*rh = SLIM_PORT_HDL(la, flw, idx);
return 0;
}
EXPORT_SYMBOL(slim_get_slaveport);
static int connect_port_ch(struct slim_controller *ctrl, u8 ch, u32 ph,
enum slim_port_flow flow)
{
int ret;
u8 buf[2];
u32 la = SLIM_HDL_TO_LA(ph);
u8 pn = (u8)SLIM_HDL_TO_PORT(ph);
DEFINE_SLIM_LDEST_TXN(txn, 0, 2, 6, NULL, buf, la);
if (flow == SLIM_SRC)
txn.mc = SLIM_MSG_MC_CONNECT_SOURCE;
else
txn.mc = SLIM_MSG_MC_CONNECT_SINK;
buf[0] = pn;
buf[1] = ctrl->chans[ch].chan;
if (la == SLIM_LA_MANAGER) {
ctrl->ports[pn].flow = flow;
ctrl->ports[pn].ch = &ctrl->chans[ch].prop;
}
ret = slim_processtxn(ctrl, &txn, false);
if (!ret && la == SLIM_LA_MANAGER)
ctrl->ports[pn].state = SLIM_P_CFG;
return ret;
}
static int disconnect_port_ch(struct slim_controller *ctrl, u32 ph)
{
int ret;
u32 la = SLIM_HDL_TO_LA(ph);
u8 pn = (u8)SLIM_HDL_TO_PORT(ph);
DEFINE_SLIM_LDEST_TXN(txn, 0, 1, 5, NULL, &pn, la);
txn.mc = SLIM_MSG_MC_DISCONNECT_PORT;
ret = slim_processtxn(ctrl, &txn, false);
if (ret)
return ret;
if (la == SLIM_LA_MANAGER) {
ctrl->ports[pn].state = SLIM_P_UNCFG;
ctrl->ports[pn].cfg.watermark = 0;
ctrl->ports[pn].cfg.port_opts = 0;
ctrl->ports[pn].ch = NULL;
}
return 0;
}
/*
* slim_connect_src: Connect source port to channel.
* @sb: client handle
* @srch: source handle to be connected to this channel
* @chanh: Channel with which the ports need to be associated with.
* Per slimbus specification, a channel may have 1 source port.
* Channel specified in chanh needs to be allocated first.
* Returns -EALREADY if source is already configured for this channel.
* Returns -ENOTCONN if channel is not allocated
* Returns -EINVAL if invalid direction is specified for non-manager port,
* or if the manager side port number is out of bounds, or in incorrect state
*/
int slim_connect_src(struct slim_device *sb, u32 srch, u16 chanh)
{
struct slim_controller *ctrl = sb->ctrl;
int ret;
u8 chan = SLIM_HDL_TO_CHIDX(chanh);
struct slim_ich *slc = &ctrl->chans[chan];
enum slim_port_flow flow = SLIM_HDL_TO_FLOW(srch);
u8 la = SLIM_HDL_TO_LA(srch);
u8 pn = SLIM_HDL_TO_PORT(srch);
/* manager ports don't have direction when they are allocated */
if (la != SLIM_LA_MANAGER && flow != SLIM_SRC)
return -EINVAL;
mutex_lock(&ctrl->sched.m_reconf);
if (la == SLIM_LA_MANAGER) {
if (pn >= ctrl->nports ||
ctrl->ports[pn].state != SLIM_P_UNCFG) {
ret = -EINVAL;
goto connect_src_err;
}
}
if (slc->state == SLIM_CH_FREE) {
ret = -ENOTCONN;
goto connect_src_err;
}
/*
* Once channel is removed, its ports can be considered disconnected
* So its ports can be reassigned. Source port is zeroed
* when channel is deallocated.
*/
if (slc->srch) {
ret = -EALREADY;
goto connect_src_err;
}
ret = connect_port_ch(ctrl, chan, srch, SLIM_SRC);
if (!ret)
slc->srch = srch;
connect_src_err:
mutex_unlock(&ctrl->sched.m_reconf);
return ret;
}
EXPORT_SYMBOL(slim_connect_src);
/*
* slim_connect_sink: Connect sink port(s) to channel.
* @sb: client handle
* @sinkh: sink handle(s) to be connected to this channel
* @nsink: number of sinks
* @chanh: Channel with which the ports need to be associated with.
* Per slimbus specification, a channel may have multiple sink-ports.
* Channel specified in chanh needs to be allocated first.
* Returns -EALREADY if sink is already configured for this channel.
* Returns -ENOTCONN if channel is not allocated
* Returns -EINVAL if invalid parameters are passed, or invalid direction is
* specified for non-manager port, or if the manager side port number is out of
* bounds, or in incorrect state
*/
int slim_connect_sink(struct slim_device *sb, u32 *sinkh, int nsink, u16 chanh)
{
struct slim_controller *ctrl = sb->ctrl;
int j;
int ret = 0;
u8 chan = SLIM_HDL_TO_CHIDX(chanh);
struct slim_ich *slc = &ctrl->chans[chan];
void *new_sinkh = NULL;
if (!sinkh || !nsink)
return -EINVAL;
mutex_lock(&ctrl->sched.m_reconf);
/*
* Once channel is removed, its ports can be considered disconnected
* So its ports can be reassigned. Sink ports are freed when channel
* is deallocated.
*/
if (slc->state == SLIM_CH_FREE) {
ret = -ENOTCONN;
goto connect_sink_err;
}
for (j = 0; j < nsink; j++) {
enum slim_port_flow flow = SLIM_HDL_TO_FLOW(sinkh[j]);
u8 la = SLIM_HDL_TO_LA(sinkh[j]);
u8 pn = SLIM_HDL_TO_PORT(sinkh[j]);
if (la != SLIM_LA_MANAGER && flow != SLIM_SINK)
ret = -EINVAL;
else if (la == SLIM_LA_MANAGER &&
(pn >= ctrl->nports ||
ctrl->ports[pn].state != SLIM_P_UNCFG))
ret = -EINVAL;
else
ret = connect_port_ch(ctrl, chan, sinkh[j], SLIM_SINK);
if (ret) {
for (j = j - 1; j >= 0; j--)
disconnect_port_ch(ctrl, sinkh[j]);
goto connect_sink_err;
}
}
new_sinkh = krealloc(slc->sinkh, (sizeof(u32) * (slc->nsink + nsink)),
GFP_KERNEL);
if (!new_sinkh) {
ret = -ENOMEM;
for (j = 0; j < nsink; j++)
disconnect_port_ch(ctrl, sinkh[j]);
goto connect_sink_err;
}
slc->sinkh = new_sinkh;
memcpy(slc->sinkh + slc->nsink, sinkh, (sizeof(u32) * nsink));
slc->nsink += nsink;
connect_sink_err:
mutex_unlock(&ctrl->sched.m_reconf);
return ret;
}
EXPORT_SYMBOL(slim_connect_sink);
/*
* slim_disconnect_ports: Disconnect port(s) from channel
* @sb: client handle
* @ph: ports to be disconnected
* @nph: number of ports.
* Disconnects ports from a channel.
*/
int slim_disconnect_ports(struct slim_device *sb, u32 *ph, int nph)
{
struct slim_controller *ctrl = sb->ctrl;
int i;
mutex_lock(&ctrl->sched.m_reconf);
for (i = 0; i < nph; i++)
disconnect_port_ch(ctrl, ph[i]);
mutex_unlock(&ctrl->sched.m_reconf);
return 0;
}
EXPORT_SYMBOL(slim_disconnect_ports);
/*
* slim_port_xfer: Schedule buffer to be transferred/received using port-handle.
* @sb: client handle
* @ph: port-handle
* @iobuf: buffer to be transferred or populated
* @len: buffer size.
* @comp: completion signal to indicate transfer done or error.
* context: can sleep
* Returns number of bytes transferred/received if used synchronously.
* Will return 0 if used asynchronously.
* Client will call slim_port_get_xfer_status to get error and/or number of
* bytes transferred if used asynchronously.
*/
int slim_port_xfer(struct slim_device *sb, u32 ph, phys_addr_t iobuf, u32 len,
struct completion *comp)
{
struct slim_controller *ctrl = sb->ctrl;
u8 pn = SLIM_HDL_TO_PORT(ph);
dev_dbg(&ctrl->dev, "port xfer: num:%d", pn);
return ctrl->port_xfer(ctrl, pn, iobuf, len, comp);
}
EXPORT_SYMBOL(slim_port_xfer);
/*
* slim_port_get_xfer_status: Poll for port transfers, or get transfer status
* after completion is done.
* @sb: client handle
* @ph: port-handle
* @done_buf: return pointer (iobuf from slim_port_xfer) which is processed.
* @done_len: Number of bytes transferred.
* This can be called when port_xfer complition is signalled.
* The API will return port transfer error (underflow/overflow/disconnect)
* and/or done_len will reflect number of bytes transferred. Note that
* done_len may be valid even if port error (overflow/underflow) has happened.
* e.g. If the transfer was scheduled with a few bytes to be transferred and
* client has not supplied more data to be transferred, done_len will indicate
* number of bytes transferred with underflow error. To avoid frequent underflow
* errors, multiple transfers can be queued (e.g. ping-pong buffers) so that
* channel has data to be transferred even if client is not ready to transfer
* data all the time. done_buf will indicate address of the last buffer
* processed from the multiple transfers.
*/
enum slim_port_err slim_port_get_xfer_status(struct slim_device *sb, u32 ph,
phys_addr_t *done_buf, u32 *done_len)
{
struct slim_controller *ctrl = sb->ctrl;
u8 pn = SLIM_HDL_TO_PORT(ph);
u32 la = SLIM_HDL_TO_LA(ph);
enum slim_port_err err;
dev_dbg(&ctrl->dev, "get status port num:%d", pn);
/*
* Framework only has insight into ports managed by ported device
* used by the manager and not slave
*/
if (la != SLIM_LA_MANAGER) {
if (done_buf)
*done_buf = 0;
if (done_len)
*done_len = 0;
return SLIM_P_NOT_OWNED;
}
err = ctrl->port_xfer_status(ctrl, pn, done_buf, done_len);
if (err == SLIM_P_INPROGRESS)
err = ctrl->ports[pn].err;
return err;
}
EXPORT_SYMBOL(slim_port_get_xfer_status);
static void slim_add_ch(struct slim_controller *ctrl, struct slim_ich *slc)
{
struct slim_ich **arr;
int i, j;
int *len;
int sl = slc->seglen << slc->rootexp;
/* Channel is already active and other end is transmitting data */
if (slc->state >= SLIM_CH_ACTIVE)
return;
if (slc->coeff == SLIM_COEFF_1) {
arr = ctrl->sched.chc1;
len = &ctrl->sched.num_cc1;
} else {
arr = ctrl->sched.chc3;
len = &ctrl->sched.num_cc3;
sl *= 3;
}
*len += 1;
/* Insert the channel based on rootexp and seglen */
for (i = 0; i < *len - 1; i++) {
/*
* Primary key: exp low to high.
* Secondary key: seglen: high to low
*/
if ((slc->rootexp > arr[i]->rootexp) ||
((slc->rootexp == arr[i]->rootexp) &&
(slc->seglen < arr[i]->seglen)))
continue;
else
break;
}
for (j = *len - 1; j > i; j--)
arr[j] = arr[j - 1];
arr[i] = slc;
if (!ctrl->allocbw)
ctrl->sched.usedslots += sl;
}
static int slim_remove_ch(struct slim_controller *ctrl, struct slim_ich *slc)
{
struct slim_ich **arr;
int i;
u32 la, ph;
int *len;
if (slc->coeff == SLIM_COEFF_1) {
arr = ctrl->sched.chc1;
len = &ctrl->sched.num_cc1;
} else {
arr = ctrl->sched.chc3;
len = &ctrl->sched.num_cc3;
}
for (i = 0; i < *len; i++) {
if (arr[i] == slc)
break;
}
if (i >= *len)
return -EXFULL;
for (; i < *len - 1; i++)
arr[i] = arr[i + 1];
*len -= 1;
arr[*len] = NULL;
slc->state = SLIM_CH_ALLOCATED;
slc->def = 0;
slc->newintr = 0;
slc->newoff = 0;
for (i = 0; i < slc->nsink; i++) {
ph = slc->sinkh[i];
la = SLIM_HDL_TO_LA(ph);
/*
* For ports managed by manager's ported device, no need to send
* disconnect. It is client's responsibility to call disconnect
* on ports owned by the slave device
*/
if (la == SLIM_LA_MANAGER) {
ctrl->ports[SLIM_HDL_TO_PORT(ph)].state = SLIM_P_UNCFG;
ctrl->ports[SLIM_HDL_TO_PORT(ph)].ch = NULL;
}
}
ph = slc->srch;
la = SLIM_HDL_TO_LA(ph);
if (la == SLIM_LA_MANAGER) {
u8 pn = SLIM_HDL_TO_PORT(ph);
ctrl->ports[pn].state = SLIM_P_UNCFG;
ctrl->ports[pn].cfg.watermark = 0;
ctrl->ports[pn].cfg.port_opts = 0;
}
kfree(slc->sinkh);
slc->sinkh = NULL;
slc->srch = 0;
slc->nsink = 0;
return 0;
}
static u32 slim_calc_prrate(struct slim_controller *ctrl, struct slim_ch *prop)
{
u32 rate = 0, rate4k = 0, rate11k = 0;
u32 exp = 0;
u32 pr = 0;
bool exact = true;
bool done = false;
enum slim_ch_rate ratefam;
if (prop->prot >= SLIM_ASYNC_SMPLX)
return 0;
if (prop->baser == SLIM_RATE_1HZ) {
rate = prop->ratem / 4000;
rate4k = rate;
if (rate * 4000 == prop->ratem)
ratefam = SLIM_RATE_4000HZ;
else {
rate = prop->ratem / 11025;
rate11k = rate;
if (rate * 11025 == prop->ratem)
ratefam = SLIM_RATE_11025HZ;
else
ratefam = SLIM_RATE_1HZ;
}
} else {
ratefam = prop->baser;
rate = prop->ratem;
}
if (ratefam == SLIM_RATE_1HZ) {
exact = false;
if ((rate4k + 1) * 4000 < (rate11k + 1) * 11025) {
rate = rate4k + 1;
ratefam = SLIM_RATE_4000HZ;
} else {
rate = rate11k + 1;
ratefam = SLIM_RATE_11025HZ;
}
}
/* covert rate to coeff-exp */
while (!done) {
while ((rate & 0x1) != 0x1) {
rate >>= 1;
exp++;
}
if (rate > 3) {
/* roundup if not exact */
rate++;
exact = false;
} else
done = true;
}
if (ratefam == SLIM_RATE_4000HZ) {
if (rate == 1)
pr = 0x10;
else {
pr = 0;
exp++;
}
} else {
pr = 8;
exp++;
}
if (exp <= 7) {
pr |= exp;
if (exact)
pr |= 0x80;
} else
pr = 0;
return pr;
}
static int slim_nextdefine_ch(struct slim_device *sb, u8 chan)
{
struct slim_controller *ctrl = sb->ctrl;
u32 chrate = 0;
u32 exp = 0;
u32 coeff = 0;
bool exact = true;
bool done = false;
int ret = 0;
struct slim_ich *slc = &ctrl->chans[chan];
struct slim_ch *prop = &slc->prop;
slc->prrate = slim_calc_prrate(ctrl, prop);
dev_dbg(&ctrl->dev, "ch:%d, chan PR rate:%x\n", chan, slc->prrate);
if (prop->baser == SLIM_RATE_4000HZ)
chrate = 4000 * prop->ratem;
else if (prop->baser == SLIM_RATE_11025HZ)
chrate = 11025 * prop->ratem;
else
chrate = prop->ratem;
/* max allowed sample freq = 768 seg/frame */
if (chrate > 3600000)
return -EDQUOT;
if (prop->baser == SLIM_RATE_4000HZ &&
ctrl->a_framer->superfreq == 4000)
coeff = prop->ratem;
else if (prop->baser == SLIM_RATE_11025HZ &&
ctrl->a_framer->superfreq == 3675)
coeff = 3 * prop->ratem;
else {
u32 tempr = 0;
tempr = chrate * SLIM_CL_PER_SUPERFRAME_DIV8;
coeff = tempr / ctrl->a_framer->rootfreq;
if (coeff * ctrl->a_framer->rootfreq != tempr) {
coeff++;
exact = false;
}
}
/* convert coeff to coeff-exponent */
exp = 0;
while (!done) {
while ((coeff & 0x1) != 0x1) {
coeff >>= 1;
exp++;
}
if (coeff > 3) {
coeff++;
exact = false;
} else
done = true;
}
if (prop->prot == SLIM_HARD_ISO && !exact)
return -EPROTONOSUPPORT;
else if (prop->prot == SLIM_AUTO_ISO) {
if (exact)
prop->prot = SLIM_HARD_ISO;
else
prop->prot = SLIM_PUSH;
}
slc->rootexp = exp;
slc->seglen = prop->sampleszbits/SLIM_CL_PER_SL;
if (prop->prot != SLIM_HARD_ISO)
slc->seglen++;
if (prop->prot >= SLIM_EXT_SMPLX)
slc->seglen++;
/* convert coeff to enum */
if (coeff == 1) {
if (exp > 9)
ret = -EIO;
coeff = SLIM_COEFF_1;
} else {
if (exp > 8)
ret = -EIO;
coeff = SLIM_COEFF_3;
}
slc->coeff = coeff;
return ret;
}
/*
* slim_alloc_ch: Allocate a slimbus channel and return its handle.
* @sb: client handle.
* @chanh: return channel handle
* Slimbus channels are limited to 256 per specification.
* -EXFULL is returned if all channels are in use.
* Although slimbus specification supports 256 channels, a controller may not
* support that many channels.
*/
int slim_alloc_ch(struct slim_device *sb, u16 *chanh)
{
struct slim_controller *ctrl = sb->ctrl;
u16 i;
if (!ctrl)
return -EINVAL;
mutex_lock(&ctrl->sched.m_reconf);
for (i = 0; i < ctrl->nchans; i++) {
if (ctrl->chans[i].state == SLIM_CH_FREE)
break;
}
if (i >= ctrl->nchans) {
mutex_unlock(&ctrl->sched.m_reconf);
return -EXFULL;
}
*chanh = i;
ctrl->chans[i].nextgrp = 0;
ctrl->chans[i].state = SLIM_CH_ALLOCATED;
ctrl->chans[i].chan = (u8)(ctrl->reserved + i);
mutex_unlock(&ctrl->sched.m_reconf);
return 0;
}
EXPORT_SYMBOL(slim_alloc_ch);
/*
* slim_query_ch: Get reference-counted handle for a channel number. Every
* channel is reference counted by upto one as producer and the others as
* consumer)
* @sb: client handle
* @chan: slimbus channel number
* @chanh: return channel handle
* If request channel number is not in use, it is allocated, and reference
* count is set to one. If the channel was was already allocated, this API
* will return handle to that channel and reference count is incremented.
* -EXFULL is returned if all channels are in use
*/
int slim_query_ch(struct slim_device *sb, u8 ch, u16 *chanh)
{
struct slim_controller *ctrl = sb->ctrl;
u16 i, j;
int ret = 0;
if (!ctrl || !chanh)
return -EINVAL;
mutex_lock(&ctrl->sched.m_reconf);
/* start with modulo number */
i = ch % ctrl->nchans;
for (j = 0; j < ctrl->nchans; j++) {
if (ctrl->chans[i].chan == ch) {
*chanh = i;
ctrl->chans[i].ref++;
if (ctrl->chans[i].state == SLIM_CH_FREE)
ctrl->chans[i].state = SLIM_CH_ALLOCATED;
goto query_out;
}
i = (i + 1) % ctrl->nchans;
}
/* Channel not in table yet */
ret = -EXFULL;
for (j = 0; j < ctrl->nchans; j++) {
if (ctrl->chans[i].state == SLIM_CH_FREE) {
ctrl->chans[i].state =
SLIM_CH_ALLOCATED;
*chanh = i;
ctrl->chans[i].ref++;
ctrl->chans[i].chan = ch;
ctrl->chans[i].nextgrp = 0;
ret = 0;
break;
}
i = (i + 1) % ctrl->nchans;
}
query_out:
mutex_unlock(&ctrl->sched.m_reconf);
dev_dbg(&ctrl->dev, "query ch:%d,hdl:%d,ref:%d,ret:%d",
ch, i, ctrl->chans[i].ref, ret);
return ret;
}
EXPORT_SYMBOL(slim_query_ch);
/*
* slim_dealloc_ch: Deallocate channel allocated using the API above
* -EISCONN is returned if the channel is tried to be deallocated without
* being removed first.
* -ENOTCONN is returned if deallocation is tried on a channel that's not
* allocated.
*/
int slim_dealloc_ch(struct slim_device *sb, u16 chanh)
{
struct slim_controller *ctrl = sb->ctrl;
u8 chan = SLIM_HDL_TO_CHIDX(chanh);
struct slim_ich *slc = &ctrl->chans[chan];
if (!ctrl)
return -EINVAL;
mutex_lock(&ctrl->sched.m_reconf);
if (slc->state == SLIM_CH_FREE) {
mutex_unlock(&ctrl->sched.m_reconf);
return -ENOTCONN;
}
if (slc->ref > 1) {
slc->ref--;
mutex_unlock(&ctrl->sched.m_reconf);
dev_dbg(&ctrl->dev, "remove chan:%d,hdl:%d,ref:%d",
slc->chan, chanh, slc->ref);
return 0;
}
if (slc->state >= SLIM_CH_PENDING_ACTIVE) {
dev_err(&ctrl->dev, "Channel:%d should be removed first", chan);
mutex_unlock(&ctrl->sched.m_reconf);
return -EISCONN;
}
slc->ref--;
slc->state = SLIM_CH_FREE;
mutex_unlock(&ctrl->sched.m_reconf);
dev_dbg(&ctrl->dev, "remove chan:%d,hdl:%d,ref:%d",
slc->chan, chanh, slc->ref);
return 0;
}
EXPORT_SYMBOL(slim_dealloc_ch);
/*
* slim_get_ch_state: Channel state.
* This API returns the channel's state (active, suspended, inactive etc)
*/
enum slim_ch_state slim_get_ch_state(struct slim_device *sb, u16 chanh)
{
u8 chan = SLIM_HDL_TO_CHIDX(chanh);
struct slim_ich *slc = &sb->ctrl->chans[chan];
return slc->state;
}
EXPORT_SYMBOL(slim_get_ch_state);
/*
* slim_define_ch: Define a channel.This API defines channel parameters for a
* given channel.
* @sb: client handle.
* @prop: slim_ch structure with channel parameters desired to be used.
* @chanh: list of channels to be defined.
* @nchan: number of channels in a group (1 if grp is false)
* @grp: Are the channels grouped
* @grph: return group handle if grouping of channels is desired.
* Channels can be grouped if multiple channels use same parameters
* (e.g. 5.1 audio has 6 channels with same parameters. They will all be grouped
* and given 1 handle for simplicity and avoid repeatedly calling the API)
* -EISCONN is returned if channel is already used with different parameters.
* -ENXIO is returned if the channel is not yet allocated.
*/
int slim_define_ch(struct slim_device *sb, struct slim_ch *prop, u16 *chanh,
u8 nchan, bool grp, u16 *grph)
{
struct slim_controller *ctrl = sb->ctrl;
int i, ret = 0;
if (!ctrl || !chanh || !prop || !nchan)
return -EINVAL;
mutex_lock(&ctrl->sched.m_reconf);
for (i = 0; i < nchan; i++) {
u8 chan = SLIM_HDL_TO_CHIDX(chanh[i]);
struct slim_ich *slc = &ctrl->chans[chan];
dev_dbg(&ctrl->dev, "define_ch: ch:%d, state:%d", chan,
(int)ctrl->chans[chan].state);
if (slc->state < SLIM_CH_ALLOCATED) {
ret = -ENXIO;
goto err_define_ch;
}
if (slc->state >= SLIM_CH_DEFINED && slc->ref >= 2) {
if (prop->ratem != slc->prop.ratem ||
prop->sampleszbits != slc->prop.sampleszbits ||
prop->baser != slc->prop.baser) {
ret = -EISCONN;
goto err_define_ch;
}
} else if (slc->state > SLIM_CH_DEFINED) {
ret = -EISCONN;
goto err_define_ch;
} else {
ctrl->chans[chan].prop = *prop;
ret = slim_nextdefine_ch(sb, chan);
if (ret)
goto err_define_ch;
}
if (i < (nchan - 1))
ctrl->chans[chan].nextgrp = chanh[i + 1];
if (i == 0)
ctrl->chans[chan].nextgrp |= SLIM_START_GRP;
if (i == (nchan - 1))
ctrl->chans[chan].nextgrp |= SLIM_END_GRP;
}
if (grp)
*grph = ((nchan << 8) | SLIM_HDL_TO_CHIDX(chanh[0]));
for (i = 0; i < nchan; i++) {
u8 chan = SLIM_HDL_TO_CHIDX(chanh[i]);
struct slim_ich *slc = &ctrl->chans[chan];
if (slc->state == SLIM_CH_ALLOCATED)
slc->state = SLIM_CH_DEFINED;
}
err_define_ch:
dev_dbg(&ctrl->dev, "define_ch: ch:%d, ret:%d", *chanh, ret);
mutex_unlock(&ctrl->sched.m_reconf);
return ret;
}
EXPORT_SYMBOL(slim_define_ch);
static u32 getsubfrmcoding(u32 *ctrlw, u32 *subfrml, u32 *msgsl)
{
u32 code = 0;
if (*ctrlw == *subfrml) {
*ctrlw = 8;
*subfrml = 8;
*msgsl = SLIM_SL_PER_SUPERFRAME - SLIM_FRM_SLOTS_PER_SUPERFRAME
- SLIM_GDE_SLOTS_PER_SUPERFRAME;
return 0;
}
if (*subfrml == 6) {
code = 0;
*msgsl = 256;
} else if (*subfrml == 8) {
code = 1;
*msgsl = 192;
} else if (*subfrml == 24) {
code = 2;
*msgsl = 64;
} else { /* 32 */
code = 3;
*msgsl = 48;
}
if (*ctrlw < 8) {
if (*ctrlw >= 6) {
*ctrlw = 6;
code |= 0x14;
} else {
if (*ctrlw == 5)
*ctrlw = 4;
code |= (*ctrlw << 2);
}
} else {
code -= 2;
if (*ctrlw >= 24) {
*ctrlw = 24;
code |= 0x1e;
} else if (*ctrlw >= 16) {
*ctrlw = 16;
code |= 0x1c;
} else if (*ctrlw >= 12) {
*ctrlw = 12;
code |= 0x1a;
} else {
*ctrlw = 8;
code |= 0x18;
}
}
*msgsl = (*msgsl * *ctrlw) - SLIM_FRM_SLOTS_PER_SUPERFRAME -
SLIM_GDE_SLOTS_PER_SUPERFRAME;
return code;
}
static void shiftsegoffsets(struct slim_controller *ctrl, struct slim_ich **ach,
int sz, u32 shft)
{
int i;
u32 oldoff;
for (i = 0; i < sz; i++) {
struct slim_ich *slc;
if (ach[i] == NULL)
continue;
slc = ach[i];
if (slc->state == SLIM_CH_PENDING_REMOVAL)
continue;
oldoff = slc->newoff;
slc->newoff += shft;
/* seg. offset must be <= interval */
if (slc->newoff >= slc->newintr)
slc->newoff -= slc->newintr;
}
}
static inline int slim_sched_4k_coeff1_chans(struct slim_controller *ctrl,
struct slim_ich **slc, int *coeff, int *opensl1,
u32 expshft, u32 curintr, u32 curmaxsl,
int curexp, int finalexp)
{
int coeff1;
struct slim_ich *slc1;
if (unlikely(!coeff || !slc || !ctrl || !opensl1))
return -EINVAL;
coeff1 = *coeff;
slc1 = *slc;
while ((coeff1 < ctrl->sched.num_cc1) &&
(curexp == (int)(slc1->rootexp + expshft))) {
if (slc1->state == SLIM_CH_PENDING_REMOVAL) {
coeff1++;
slc1 = ctrl->sched.chc1[coeff1];
continue;
}
if (opensl1[1] >= opensl1[0] ||
(finalexp == (int)slc1->rootexp &&
curintr <= 24 && opensl1[0] == curmaxsl)) {
opensl1[1] -= slc1->seglen;
slc1->newoff = curmaxsl + opensl1[1];
if (opensl1[1] < 0 && opensl1[0] == curmaxsl) {
opensl1[0] += opensl1[1];
opensl1[1] = 0;
if (opensl1[0] < 0) {
dev_dbg(&ctrl->dev,
"reconfig failed:%d\n",
__LINE__);
return -EXFULL;
}
}
} else {
if (slc1->seglen > opensl1[0]) {
dev_dbg(&ctrl->dev,
"reconfig failed:%d\n", __LINE__);
return -EXFULL;
}
slc1->newoff = opensl1[0] - slc1->seglen;
opensl1[0] = slc1->newoff;
}
slc1->newintr = curintr;
coeff1++;
slc1 = ctrl->sched.chc1[coeff1];
}
*coeff = coeff1;
*slc = slc1;
return 0;
}
static int slim_sched_chans(struct slim_device *sb, u32 clkgear,
u32 *ctrlw, u32 *subfrml)
{
int coeff1, coeff3;
enum slim_ch_coeff bias;
struct slim_controller *ctrl = sb->ctrl;
int last1 = ctrl->sched.num_cc1 - 1;
int last3 = ctrl->sched.num_cc3 - 1;
/*
* Find first channels with coeff 1 & 3 as starting points for
* scheduling
*/
for (coeff3 = 0; coeff3 < ctrl->sched.num_cc3; coeff3++) {
struct slim_ich *slc = ctrl->sched.chc3[coeff3];
if (slc->state == SLIM_CH_PENDING_REMOVAL)
continue;
else
break;
}
for (coeff1 = 0; coeff1 < ctrl->sched.num_cc1; coeff1++) {
struct slim_ich *slc = ctrl->sched.chc1[coeff1];
if (slc->state == SLIM_CH_PENDING_REMOVAL)
continue;
else
break;
}
if (coeff3 == ctrl->sched.num_cc3 && coeff1 == ctrl->sched.num_cc1) {
*ctrlw = 8;
*subfrml = 8;
return 0;
} else if (coeff3 == ctrl->sched.num_cc3)
bias = SLIM_COEFF_1;
else
bias = SLIM_COEFF_3;
/*
* Find last chan in coeff1, 3 list, we will use to know when we
* have done scheduling all coeff1 channels
*/
while (last1 >= 0) {
if (ctrl->sched.chc1[last1] != NULL &&
(ctrl->sched.chc1[last1])->state !=
SLIM_CH_PENDING_REMOVAL)
break;
last1--;
}
while (last3 >= 0) {
if (ctrl->sched.chc3[last3] != NULL &&
(ctrl->sched.chc3[last3])->state !=
SLIM_CH_PENDING_REMOVAL)
break;
last3--;
}
if (bias == SLIM_COEFF_1) {
struct slim_ich *slc1 = ctrl->sched.chc1[coeff1];
u32 expshft = SLIM_MAX_CLK_GEAR - clkgear;
int curexp, finalexp;
u32 curintr, curmaxsl;
int opensl1[2];
int maxctrlw1;
int ret;
finalexp = (ctrl->sched.chc1[last1])->rootexp;
curexp = (int)expshft - 1;
curintr = (SLIM_MAX_INTR_COEFF_1 * 2) >> (curexp + 1);
curmaxsl = curintr >> 1;
opensl1[0] = opensl1[1] = curmaxsl;
while ((coeff1 < ctrl->sched.num_cc1) || (curintr > 24)) {
curintr >>= 1;
curmaxsl >>= 1;
/* update 4K family open slot records */
if (opensl1[1] < opensl1[0])
opensl1[1] -= curmaxsl;
else
opensl1[1] = opensl1[0] - curmaxsl;
opensl1[0] = curmaxsl;
if (opensl1[1] < 0) {
opensl1[0] += opensl1[1];
opensl1[1] = 0;
}
if (opensl1[0] <= 0) {
dev_dbg(&ctrl->dev, "reconfig failed:%d\n",
__LINE__);
return -EXFULL;
}
curexp++;
/* schedule 4k family channels */
ret = slim_sched_4k_coeff1_chans(ctrl, &slc1, &coeff1,
opensl1, expshft, curintr, curmaxsl,
curexp, finalexp);
if (ret)
return ret;
}
/* Leave some slots for messaging space */
if (opensl1[1] <= 0 && opensl1[0] <= 0)
return -EXFULL;
if (opensl1[1] > opensl1[0]) {
int temp = opensl1[0];
opensl1[0] = opensl1[1];
opensl1[1] = temp;
shiftsegoffsets(ctrl, ctrl->sched.chc1,
ctrl->sched.num_cc1, curmaxsl);
}
/* choose subframe mode to maximize bw */
maxctrlw1 = opensl1[0];
if (opensl1[0] == curmaxsl)
maxctrlw1 += opensl1[1];
if (curintr >= 24) {
*subfrml = 24;
*ctrlw = maxctrlw1;
} else if (curintr == 12) {
if (maxctrlw1 > opensl1[1] * 4) {
*subfrml = 24;
*ctrlw = maxctrlw1;
} else {
*subfrml = 6;
*ctrlw = opensl1[1];
}
} else {
*subfrml = 6;
*ctrlw = maxctrlw1;
}
} else {
struct slim_ich *slc1 = NULL;
struct slim_ich *slc3 = ctrl->sched.chc3[coeff3];
u32 expshft = SLIM_MAX_CLK_GEAR - clkgear;
int curexp, finalexp, exp1;
u32 curintr, curmaxsl;
int opensl3[2];
int opensl1[6];
bool opensl1valid = false;
int maxctrlw1, maxctrlw3, i;
/* intitalize array to zero */
memset(opensl1, 0x0, sizeof(opensl1));
finalexp = (ctrl->sched.chc3[last3])->rootexp;
if (last1 >= 0) {
slc1 = ctrl->sched.chc1[coeff1];
exp1 = (ctrl->sched.chc1[last1])->rootexp;
if (exp1 > finalexp)
finalexp = exp1;
}
curexp = (int)expshft - 1;
curintr = (SLIM_MAX_INTR_COEFF_3 * 2) >> (curexp + 1);
curmaxsl = curintr >> 1;
opensl3[0] = opensl3[1] = curmaxsl;
while (coeff1 < ctrl->sched.num_cc1 ||
coeff3 < ctrl->sched.num_cc3 ||
curintr > 32) {
curintr >>= 1;
curmaxsl >>= 1;
/* update 12k family open slot records */
if (opensl3[1] < opensl3[0])
opensl3[1] -= curmaxsl;
else
opensl3[1] = opensl3[0] - curmaxsl;
opensl3[0] = curmaxsl;
if (opensl3[1] < 0) {
opensl3[0] += opensl3[1];
opensl3[1] = 0;
}
if (opensl3[0] <= 0) {
dev_dbg(&ctrl->dev, "reconfig failed:%d\n",
__LINE__);
return -EXFULL;
}
curexp++;
/* schedule 12k family channels */
while (coeff3 < ctrl->sched.num_cc3 &&
curexp == (int)slc3->rootexp + expshft) {
if (slc3->state == SLIM_CH_PENDING_REMOVAL) {
coeff3++;
slc3 = ctrl->sched.chc3[coeff3];
continue;
}
opensl1valid = false;
if (opensl3[1] >= opensl3[0] ||
(finalexp == (int)slc3->rootexp &&
curintr <= 32 &&
opensl3[0] == curmaxsl &&
last1 < 0)) {
opensl3[1] -= slc3->seglen;
slc3->newoff = curmaxsl + opensl3[1];
if (opensl3[1] < 0 &&
opensl3[0] == curmaxsl) {
opensl3[0] += opensl3[1];
opensl3[1] = 0;
}
if (opensl3[0] < 0) {
dev_dbg(&ctrl->dev,
"reconfig failed:%d\n",
__LINE__);
return -EXFULL;
}
} else {
if (slc3->seglen > opensl3[0]) {
dev_dbg(&ctrl->dev,
"reconfig failed:%d\n",
__LINE__);
return -EXFULL;
}
slc3->newoff = opensl3[0] -
slc3->seglen;
opensl3[0] = slc3->newoff;
}
slc3->newintr = curintr;
coeff3++;
slc3 = ctrl->sched.chc3[coeff3];
}
/* update 4k openslot records */
if (opensl1valid == false) {
for (i = 0; i < 3; i++) {
opensl1[i * 2] = opensl3[0];
opensl1[(i * 2) + 1] = opensl3[1];
}
} else {
int opensl1p[6];
memcpy(opensl1p, opensl1, sizeof(opensl1));
for (i = 0; i < 3; i++) {
if (opensl1p[i] < opensl1p[i + 3])
opensl1[(i * 2) + 1] =
opensl1p[i];
else
opensl1[(i * 2) + 1] =
opensl1p[i + 3];
}
for (i = 0; i < 3; i++) {
opensl1[(i * 2) + 1] -= curmaxsl;
opensl1[i * 2] = curmaxsl;
if (opensl1[(i * 2) + 1] < 0) {
opensl1[i * 2] +=
opensl1[(i * 2) + 1];
opensl1[(i * 2) + 1] = 0;
}
if (opensl1[i * 2] < 0) {
dev_dbg(&ctrl->dev,
"reconfig failed:%d\n",
__LINE__);
return -EXFULL;
}
}
}
/* schedule 4k family channels */
while (coeff1 < ctrl->sched.num_cc1 && slc1 &&
curexp == (int)slc1->rootexp + expshft) {
/* searchorder effective when opensl valid */
static const int srcho[] = { 5, 2, 4, 1, 3, 0 };
int maxopensl = 0;
int maxi = 0;
if (slc1->state == SLIM_CH_PENDING_REMOVAL) {
coeff1++;
slc1 = ctrl->sched.chc1[coeff1];
continue;
}
opensl1valid = true;
for (i = 0; i < 6; i++) {
if (opensl1[srcho[i]] > maxopensl) {
maxopensl = opensl1[srcho[i]];
maxi = srcho[i];
}
}
opensl1[maxi] -= slc1->seglen;
slc1->newoff = (curmaxsl * maxi) +
opensl1[maxi];
if (opensl1[maxi] < 0 && (maxi & 1) == 1 &&
opensl1[maxi - 1] == curmaxsl) {
opensl1[maxi - 1] += opensl1[maxi];
if (opensl3[0] > opensl1[maxi - 1])
opensl3[0] = opensl1[maxi - 1];
opensl3[1] = 0;
opensl1[maxi] = 0;
if (opensl1[maxi - 1] < 0) {
dev_dbg(&ctrl->dev,
"reconfig failed:%d\n",
__LINE__);
return -EXFULL;
}
} else if (opensl1[maxi] < 0) {
dev_dbg(&ctrl->dev,
"reconfig failed:%d\n",
__LINE__);
return -EXFULL;
} else if (opensl3[maxi & 1] > opensl1[maxi]) {
opensl3[maxi & 1] = opensl1[maxi];
}
slc1->newintr = curintr * 3;
coeff1++;
slc1 = ctrl->sched.chc1[coeff1];
}
}
/* Leave some slots for messaging space */
if (opensl3[1] <= 0 && opensl3[0] <= 0)
return -EXFULL;
/* swap 1st and 2nd bucket if 2nd bucket has more open slots */
if (opensl3[1] > opensl3[0]) {
int temp = opensl3[0];
opensl3[0] = opensl3[1];
opensl3[1] = temp;
temp = opensl1[5];
opensl1[5] = opensl1[4];
opensl1[4] = opensl1[3];
opensl1[3] = opensl1[2];
opensl1[2] = opensl1[1];
opensl1[1] = opensl1[0];
opensl1[0] = temp;
shiftsegoffsets(ctrl, ctrl->sched.chc1,
ctrl->sched.num_cc1, curmaxsl);
shiftsegoffsets(ctrl, ctrl->sched.chc3,
ctrl->sched.num_cc3, curmaxsl);
}
/* subframe mode to maximize BW */
maxctrlw3 = opensl3[0];
maxctrlw1 = opensl1[0];
if (opensl3[0] == curmaxsl)
maxctrlw3 += opensl3[1];
for (i = 0; i < 5 && opensl1[i] == curmaxsl; i++)
maxctrlw1 += opensl1[i + 1];
if (curintr >= 32) {
*subfrml = 32;
*ctrlw = maxctrlw3;
} else if (curintr == 16) {
if (maxctrlw3 > (opensl3[1] * 4)) {
*subfrml = 32;
*ctrlw = maxctrlw3;
} else {
*subfrml = 8;
*ctrlw = opensl3[1];
}
} else {
if ((maxctrlw1 * 8) >= (maxctrlw3 * 24)) {
*subfrml = 24;
*ctrlw = maxctrlw1;
} else {
*subfrml = 8;
*ctrlw = maxctrlw3;
}
}
}
return 0;
}
#ifdef DEBUG
static int slim_verifychansched(struct slim_controller *ctrl, u32 ctrlw,
u32 subfrml, u32 clkgear)
{
int sl, i;
int cc1 = 0;
int cc3 = 0;
struct slim_ich *slc = NULL;
if (!ctrl->sched.slots)
return 0;
memset(ctrl->sched.slots, 0, SLIM_SL_PER_SUPERFRAME);
dev_dbg(&ctrl->dev, "Clock gear is:%d\n", clkgear);
for (sl = 0; sl < SLIM_SL_PER_SUPERFRAME; sl += subfrml) {
for (i = 0; i < ctrlw; i++)
ctrl->sched.slots[sl + i] = 33;
}
while (cc1 < ctrl->sched.num_cc1) {
slc = ctrl->sched.chc1[cc1];
if (slc == NULL) {
dev_err(&ctrl->dev, "SLC1 null in verify: chan%d\n",
cc1);
return -EIO;
}
dev_dbg(&ctrl->dev, "chan:%d, offset:%d, intr:%d, seglen:%d\n",
(slc - ctrl->chans), slc->newoff,
slc->newintr, slc->seglen);
if (slc->state != SLIM_CH_PENDING_REMOVAL) {
for (sl = slc->newoff;
sl < SLIM_SL_PER_SUPERFRAME;
sl += slc->newintr) {
for (i = 0; i < slc->seglen; i++) {
if (ctrl->sched.slots[sl + i])
return -EXFULL;
ctrl->sched.slots[sl + i] = cc1 + 1;
}
}
}
cc1++;
}
while (cc3 < ctrl->sched.num_cc3) {
slc = ctrl->sched.chc3[cc3];
if (slc == NULL) {
dev_err(&ctrl->dev, "SLC3 null in verify: chan%d\n",
cc3);
return -EIO;
}
dev_dbg(&ctrl->dev, "chan:%d, offset:%d, intr:%d, seglen:%d\n",
(slc - ctrl->chans), slc->newoff,
slc->newintr, slc->seglen);
if (slc->state != SLIM_CH_PENDING_REMOVAL) {
for (sl = slc->newoff;
sl < SLIM_SL_PER_SUPERFRAME;
sl += slc->newintr) {
for (i = 0; i < slc->seglen; i++) {
if (ctrl->sched.slots[sl + i])
return -EXFULL;
ctrl->sched.slots[sl + i] = cc3 + 1;
}
}
}
cc3++;
}
return 0;
}
#else
static int slim_verifychansched(struct slim_controller *ctrl, u32 ctrlw,
u32 subfrml, u32 clkgear)
{
return 0;
}
#endif
static void slim_sort_chan_grp(struct slim_controller *ctrl,
struct slim_ich *slc)
{
u8 last = (u8)-1;
u8 second = 0;
for (; last > 0; last--) {
struct slim_ich *slc1 = slc;
struct slim_ich *slc2;
u8 next = SLIM_HDL_TO_CHIDX(slc1->nextgrp);
slc2 = &ctrl->chans[next];
for (second = 1; second <= last && slc2 &&
(slc2->state == SLIM_CH_ACTIVE ||
slc2->state == SLIM_CH_PENDING_ACTIVE); second++) {
if (slc1->newoff > slc2->newoff) {
u32 temp = slc2->newoff;
slc2->newoff = slc1->newoff;
slc1->newoff = temp;
}
if (slc2->nextgrp & SLIM_END_GRP) {
last = second;
break;
}
slc1 = slc2;
next = SLIM_HDL_TO_CHIDX(slc1->nextgrp);
slc2 = &ctrl->chans[next];
}
if (slc2 == NULL)
last = second - 1;
}
}
static int slim_allocbw(struct slim_device *sb, int *subfrmc, int *clkgear)
{
u32 msgsl = 0;
u32 ctrlw = 0;
u32 subfrml = 0;
int ret = -EIO;
struct slim_controller *ctrl = sb->ctrl;
u32 usedsl = ctrl->sched.usedslots + ctrl->sched.pending_msgsl;
u32 availsl = SLIM_SL_PER_SUPERFRAME - SLIM_FRM_SLOTS_PER_SUPERFRAME -
SLIM_GDE_SLOTS_PER_SUPERFRAME;
*clkgear = SLIM_MAX_CLK_GEAR;
dev_dbg(&ctrl->dev, "used sl:%u, availlable sl:%u\n", usedsl, availsl);
dev_dbg(&ctrl->dev, "pending:chan sl:%u, :msg sl:%u, clkgear:%u\n",
ctrl->sched.usedslots,
ctrl->sched.pending_msgsl, *clkgear);
/*
* If number of slots are 0, that means channels are inactive.
* It is very likely that the manager will call clock pause very soon.
* By making sure that bus is in MAX_GEAR, clk pause sequence will take
* minimum amount of time.
*/
if (ctrl->sched.usedslots != 0) {
while ((usedsl * 2 <= availsl) && (*clkgear > ctrl->min_cg)) {
*clkgear -= 1;
usedsl *= 2;
}
}
/*
* Try scheduling data channels at current clock gear, if all channels
* can be scheduled, or reserved BW can't be satisfied, increase clock
* gear and try again
*/
for (; *clkgear <= ctrl->max_cg; (*clkgear)++) {
ret = slim_sched_chans(sb, *clkgear, &ctrlw, &subfrml);
if (ret == 0) {
*subfrmc = getsubfrmcoding(&ctrlw, &subfrml, &msgsl);
if ((msgsl >> (ctrl->max_cg - *clkgear) <
ctrl->sched.pending_msgsl) &&
(*clkgear < ctrl->max_cg))
continue;
else
break;
}
}
if (ret == 0) {
int i;
/* Sort channel-groups */
for (i = 0; i < ctrl->sched.num_cc1; i++) {
struct slim_ich *slc = ctrl->sched.chc1[i];
if (slc->state == SLIM_CH_PENDING_REMOVAL)
continue;
if ((slc->nextgrp & SLIM_START_GRP) &&
!(slc->nextgrp & SLIM_END_GRP)) {
slim_sort_chan_grp(ctrl, slc);
}
}
for (i = 0; i < ctrl->sched.num_cc3; i++) {
struct slim_ich *slc = ctrl->sched.chc3[i];
if (slc->state == SLIM_CH_PENDING_REMOVAL)
continue;
if ((slc->nextgrp & SLIM_START_GRP) &&
!(slc->nextgrp & SLIM_END_GRP)) {
slim_sort_chan_grp(ctrl, slc);
}
}
ret = slim_verifychansched(ctrl, ctrlw, subfrml, *clkgear);
}
return ret;
}
static void slim_change_existing_chans(struct slim_controller *ctrl, int coeff)
{
struct slim_ich **arr;
int len, i;
if (coeff == SLIM_COEFF_1) {
arr = ctrl->sched.chc1;
len = ctrl->sched.num_cc1;
} else {
arr = ctrl->sched.chc3;
len = ctrl->sched.num_cc3;
}
for (i = 0; i < len; i++) {
struct slim_ich *slc = arr[i];
if (slc->state == SLIM_CH_ACTIVE ||
slc->state == SLIM_CH_SUSPENDED) {
slc->offset = slc->newoff;
slc->interval = slc->newintr;
}
}
}
static void slim_chan_changes(struct slim_device *sb, bool revert)
{
struct slim_controller *ctrl = sb->ctrl;
while (!list_empty(&sb->mark_define)) {
struct slim_ich *slc;
struct slim_pending_ch *pch =
list_entry(sb->mark_define.next,
struct slim_pending_ch, pending);
slc = &ctrl->chans[pch->chan];
if (revert) {
if (slc->state == SLIM_CH_PENDING_ACTIVE) {
u32 sl = slc->seglen << slc->rootexp;
if (slc->coeff == SLIM_COEFF_3)
sl *= 3;
if (!ctrl->allocbw)
ctrl->sched.usedslots -= sl;
slim_remove_ch(ctrl, slc);
slc->state = SLIM_CH_DEFINED;
}
} else {
slc->state = SLIM_CH_ACTIVE;
slc->def++;
}
list_del_init(&pch->pending);
kfree(pch);
}
while (!list_empty(&sb->mark_removal)) {
struct slim_pending_ch *pch =
list_entry(sb->mark_removal.next,
struct slim_pending_ch, pending);
struct slim_ich *slc = &ctrl->chans[pch->chan];
u32 sl = slc->seglen << slc->rootexp;
if (revert || slc->def > 0) {
if (slc->coeff == SLIM_COEFF_3)
sl *= 3;
if (!ctrl->allocbw)
ctrl->sched.usedslots += sl;
if (revert)
slc->def++;
slc->state = SLIM_CH_ACTIVE;
} else
slim_remove_ch(ctrl, slc);
list_del_init(&pch->pending);
kfree(pch);
}
while (!list_empty(&sb->mark_suspend)) {
struct slim_pending_ch *pch =
list_entry(sb->mark_suspend.next,
struct slim_pending_ch, pending);
struct slim_ich *slc = &ctrl->chans[pch->chan];
if (revert)
slc->state = SLIM_CH_ACTIVE;
list_del_init(&pch->pending);
kfree(pch);
}
/* Change already active channel if reconfig succeeded */
if (!revert) {
slim_change_existing_chans(ctrl, SLIM_COEFF_1);
slim_change_existing_chans(ctrl, SLIM_COEFF_3);
}
}
/*
* slim_reconfigure_now: Request reconfiguration now.
* @sb: client handle
* This API does what commit flag in other scheduling APIs do.
* -EXFULL is returned if there is no space in TDM to reserve the
* bandwidth. -EBUSY is returned if reconfiguration request is already in
* progress.
*/
int slim_reconfigure_now(struct slim_device *sb)
{
u8 i;
u8 wbuf[4];
u32 clkgear, subframe;
u32 curexp;
int ret;
struct slim_controller *ctrl = sb->ctrl;
u32 expshft;
u32 segdist;
struct slim_pending_ch *pch;
DEFINE_SLIM_BCAST_TXN(txn, SLIM_MSG_MC_BEGIN_RECONFIGURATION, 0, 3,
NULL, NULL, sb->laddr);
mutex_lock(&ctrl->sched.m_reconf);
/*
* If there are no pending changes from this client, avoid sending
* the reconfiguration sequence
*/
if (sb->pending_msgsl == sb->cur_msgsl &&
list_empty(&sb->mark_define) &&
list_empty(&sb->mark_suspend)) {
struct list_head *pos, *next;
list_for_each_safe(pos, next, &sb->mark_removal) {
struct slim_ich *slc;
pch = list_entry(pos, struct slim_pending_ch, pending);
slc = &ctrl->chans[pch->chan];
if (slc->def > 0)
slc->def--;
/* Disconnect source port to free it up */
if (SLIM_HDL_TO_LA(slc->srch) == sb->laddr)
slc->srch = 0;
/*
* If controller overrides BW allocation,
* delete this in remove channel itself
*/
if (slc->def != 0 && !ctrl->allocbw) {
list_del(&pch->pending);
kfree(pch);
}
}
if (list_empty(&sb->mark_removal)) {
mutex_unlock(&ctrl->sched.m_reconf);
pr_info("SLIM_CL: skip reconfig sequence");
return 0;
}
}
ctrl->sched.pending_msgsl += sb->pending_msgsl - sb->cur_msgsl;
list_for_each_entry(pch, &sb->mark_define, pending) {
struct slim_ich *slc = &ctrl->chans[pch->chan];
slim_add_ch(ctrl, slc);
if (slc->state < SLIM_CH_ACTIVE)
slc->state = SLIM_CH_PENDING_ACTIVE;
}
list_for_each_entry(pch, &sb->mark_removal, pending) {
struct slim_ich *slc = &ctrl->chans[pch->chan];
u32 sl = slc->seglen << slc->rootexp;
if (slc->coeff == SLIM_COEFF_3)
sl *= 3;
if (!ctrl->allocbw)
ctrl->sched.usedslots -= sl;
slc->state = SLIM_CH_PENDING_REMOVAL;
}
list_for_each_entry(pch, &sb->mark_suspend, pending) {
struct slim_ich *slc = &ctrl->chans[pch->chan];
slc->state = SLIM_CH_SUSPENDED;
}
/*
* Controller can override default channel scheduling algorithm.
* (e.g. if controller needs to use fixed channel scheduling based
* on number of channels)
*/
if (ctrl->allocbw)
ret = ctrl->allocbw(sb, &subframe, &clkgear);
else
ret = slim_allocbw(sb, &subframe, &clkgear);
if (!ret) {
ret = slim_processtxn(ctrl, &txn, false);
dev_dbg(&ctrl->dev, "sending begin_reconfig:ret:%d\n", ret);
}
if (!ret && subframe != ctrl->sched.subfrmcode) {
wbuf[0] = (u8)(subframe & 0xFF);
txn.mc = SLIM_MSG_MC_NEXT_SUBFRAME_MODE;
txn.len = 1;
txn.rl = 4;
txn.wbuf = wbuf;
ret = slim_processtxn(ctrl, &txn, false);
dev_dbg(&ctrl->dev, "sending subframe:%d,ret:%d\n",
(int)wbuf[0], ret);
}
if (!ret && clkgear != ctrl->clkgear) {
wbuf[0] = (u8)(clkgear & 0xFF);
txn.mc = SLIM_MSG_MC_NEXT_CLOCK_GEAR;
txn.len = 1;
txn.rl = 4;
txn.wbuf = wbuf;
ret = slim_processtxn(ctrl, &txn, false);
dev_dbg(&ctrl->dev, "sending clkgear:%d,ret:%d\n",
(int)wbuf[0], ret);
}
if (ret)
goto revert_reconfig;
expshft = SLIM_MAX_CLK_GEAR - clkgear;
/* activate/remove channel */
list_for_each_entry(pch, &sb->mark_define, pending) {
struct slim_ich *slc = &ctrl->chans[pch->chan];
/* Define content */
wbuf[0] = slc->chan;
wbuf[1] = slc->prrate;
wbuf[2] = slc->prop.dataf | (slc->prop.auxf << 4);
wbuf[3] = slc->prop.sampleszbits / SLIM_CL_PER_SL;
txn.mc = SLIM_MSG_MC_NEXT_DEFINE_CONTENT;
txn.len = 4;
txn.rl = 7;
txn.wbuf = wbuf;
dev_dbg(&ctrl->dev, "define content, activate:%x, %x, %x, %x\n",
wbuf[0], wbuf[1], wbuf[2], wbuf[3]);
/* Right now, channel link bit is not supported */
ret = slim_processtxn(ctrl, &txn, false);
if (ret)
goto revert_reconfig;
txn.mc = SLIM_MSG_MC_NEXT_ACTIVATE_CHANNEL;
txn.len = 1;
txn.rl = 4;
ret = slim_processtxn(ctrl, &txn, false);
if (ret)
goto revert_reconfig;
}
list_for_each_entry(pch, &sb->mark_removal, pending) {
struct slim_ich *slc = &ctrl->chans[pch->chan];
dev_dbg(&ctrl->dev, "remove chan:%x\n", pch->chan);
wbuf[0] = slc->chan;
txn.mc = SLIM_MSG_MC_NEXT_REMOVE_CHANNEL;
txn.len = 1;
txn.rl = 4;
txn.wbuf = wbuf;
ret = slim_processtxn(ctrl, &txn, false);
if (ret)
goto revert_reconfig;
}
list_for_each_entry(pch, &sb->mark_suspend, pending) {
struct slim_ich *slc = &ctrl->chans[pch->chan];
dev_dbg(&ctrl->dev, "suspend chan:%x\n", pch->chan);
wbuf[0] = slc->chan;
txn.mc = SLIM_MSG_MC_NEXT_DEACTIVATE_CHANNEL;
txn.len = 1;
txn.rl = 4;
txn.wbuf = wbuf;
ret = slim_processtxn(ctrl, &txn, false);
if (ret)
goto revert_reconfig;
}
/* Define CC1 channel */
for (i = 0; i < ctrl->sched.num_cc1; i++) {
struct slim_ich *slc = ctrl->sched.chc1[i];
if (slc->state == SLIM_CH_PENDING_REMOVAL)
continue;
curexp = slc->rootexp + expshft;
segdist = (slc->newoff << curexp) & 0x1FF;
expshft = SLIM_MAX_CLK_GEAR - clkgear;
dev_dbg(&ctrl->dev, "new-intr:%d, old-intr:%d, dist:%d\n",
slc->newintr, slc->interval, segdist);
dev_dbg(&ctrl->dev, "new-off:%d, old-off:%d\n",
slc->newoff, slc->offset);
if (slc->state < SLIM_CH_ACTIVE || slc->def < slc->ref ||
slc->newintr != slc->interval ||
slc->newoff != slc->offset) {
segdist |= 0x200;
segdist >>= curexp;
segdist |= (slc->newoff << (curexp + 1)) & 0xC00;
wbuf[0] = slc->chan;
wbuf[1] = (u8)(segdist & 0xFF);
wbuf[2] = (u8)((segdist & 0xF00) >> 8) |
(slc->prop.prot << 4);
wbuf[3] = slc->seglen;
txn.mc = SLIM_MSG_MC_NEXT_DEFINE_CHANNEL;
txn.len = 4;
txn.rl = 7;
txn.wbuf = wbuf;
ret = slim_processtxn(ctrl, &txn, false);
if (ret)
goto revert_reconfig;
}
}
/* Define CC3 channels */
for (i = 0; i < ctrl->sched.num_cc3; i++) {
struct slim_ich *slc = ctrl->sched.chc3[i];
if (slc->state == SLIM_CH_PENDING_REMOVAL)
continue;
curexp = slc->rootexp + expshft;
segdist = (slc->newoff << curexp) & 0x1FF;
expshft = SLIM_MAX_CLK_GEAR - clkgear;
dev_dbg(&ctrl->dev, "new-intr:%d, old-intr:%d, dist:%d\n",
slc->newintr, slc->interval, segdist);
dev_dbg(&ctrl->dev, "new-off:%d, old-off:%d\n",
slc->newoff, slc->offset);
if (slc->state < SLIM_CH_ACTIVE || slc->def < slc->ref ||
slc->newintr != slc->interval ||
slc->newoff != slc->offset) {
segdist |= 0x200;
segdist >>= curexp;
segdist |= 0xC00;
wbuf[0] = slc->chan;
wbuf[1] = (u8)(segdist & 0xFF);
wbuf[2] = (u8)((segdist & 0xF00) >> 8) |
(slc->prop.prot << 4);
wbuf[3] = (u8)(slc->seglen);
txn.mc = SLIM_MSG_MC_NEXT_DEFINE_CHANNEL;
txn.len = 4;
txn.rl = 7;
txn.wbuf = wbuf;
ret = slim_processtxn(ctrl, &txn, false);
if (ret)
goto revert_reconfig;
}
}
txn.mc = SLIM_MSG_MC_RECONFIGURE_NOW;
txn.len = 0;
txn.rl = 3;
txn.wbuf = NULL;
ret = slim_processtxn(ctrl, &txn, false);
dev_dbg(&ctrl->dev, "reconfig now:ret:%d\n", ret);
if (!ret) {
ctrl->sched.subfrmcode = subframe;
ctrl->clkgear = clkgear;
ctrl->sched.msgsl = ctrl->sched.pending_msgsl;
sb->cur_msgsl = sb->pending_msgsl;
slim_chan_changes(sb, false);
mutex_unlock(&ctrl->sched.m_reconf);
return 0;
}
revert_reconfig:
/* Revert channel changes */
slim_chan_changes(sb, true);
mutex_unlock(&ctrl->sched.m_reconf);
return ret;
}
EXPORT_SYMBOL(slim_reconfigure_now);
static int add_pending_ch(struct list_head *listh, u8 chan)
{
struct slim_pending_ch *pch;
pch = kmalloc(sizeof(struct slim_pending_ch), GFP_KERNEL);
if (!pch)
return -ENOMEM;
pch->chan = chan;
list_add_tail(&pch->pending, listh);
return 0;
}
/*
* slim_control_ch: Channel control API.
* @sb: client handle
* @chanh: group or channel handle to be controlled
* @chctrl: Control command (activate/suspend/remove)
* @commit: flag to indicate whether the control should take effect right-away.
* This API activates, removes or suspends a channel (or group of channels)
* chanh indicates the channel or group handle (returned by the define_ch API).
* Reconfiguration may be time-consuming since it can change all other active
* channel allocations on the bus, change in clock gear used by the slimbus,
* and change in the control space width used for messaging.
* commit makes sure that multiple channels can be activated/deactivated before
* reconfiguration is started.
* -EXFULL is returned if there is no space in TDM to reserve the bandwidth.
* -EISCONN/-ENOTCONN is returned if the channel is already connected or not
* yet defined.
* -EINVAL is returned if individual control of a grouped-channel is attempted.
*/
int slim_control_ch(struct slim_device *sb, u16 chanh,
enum slim_ch_control chctrl, bool commit)
{
struct slim_controller *ctrl = sb->ctrl;
int ret = 0;
/* Get rid of the group flag in MSB if any */
u8 chan = SLIM_HDL_TO_CHIDX(chanh);
u8 nchan = 0;
struct slim_ich *slc = &ctrl->chans[chan];
if (!(slc->nextgrp & SLIM_START_GRP))
return -EINVAL;
mutex_lock(&sb->sldev_reconf);
mutex_lock(&ctrl->sched.m_reconf);
do {
struct slim_pending_ch *pch;
u8 add_mark_removal = true;
slc = &ctrl->chans[chan];
dev_dbg(&ctrl->dev, "chan:%d,ctrl:%d,def:%d", chan, chctrl,
slc->def);
if (slc->state < SLIM_CH_DEFINED) {
ret = -ENOTCONN;
break;
}
if (chctrl == SLIM_CH_SUSPEND) {
ret = add_pending_ch(&sb->mark_suspend, chan);
if (ret)
break;
} else if (chctrl == SLIM_CH_ACTIVATE) {
if (slc->state > SLIM_CH_ACTIVE) {
ret = -EISCONN;
break;
}
ret = add_pending_ch(&sb->mark_define, chan);
if (ret)
break;
} else {
if (slc->state < SLIM_CH_ACTIVE) {
ret = -ENOTCONN;
break;
}
/* If channel removal request comes when pending
* in the mark_define, remove it from the define
* list instead of adding it to removal list
*/
if (!list_empty(&sb->mark_define)) {
struct list_head *pos, *next;
list_for_each_safe(pos, next,
&sb->mark_define) {
pch = list_entry(pos,
struct slim_pending_ch,
pending);
if (pch->chan == chan) {
list_del(&pch->pending);
kfree(pch);
add_mark_removal = false;
break;
}
}
}
if (add_mark_removal == true) {
ret = add_pending_ch(&sb->mark_removal, chan);
if (ret)
break;
}
}
nchan++;
if (nchan < SLIM_GRP_TO_NCHAN(chanh))
chan = SLIM_HDL_TO_CHIDX(slc->nextgrp);
} while (nchan < SLIM_GRP_TO_NCHAN(chanh));
mutex_unlock(&ctrl->sched.m_reconf);
if (!ret && commit == true)
ret = slim_reconfigure_now(sb);
mutex_unlock(&sb->sldev_reconf);
return ret;
}
EXPORT_SYMBOL(slim_control_ch);
/*
* slim_reservemsg_bw: Request to reserve bandwidth for messages.
* @sb: client handle
* @bw_bps: message bandwidth in bits per second to be requested
* @commit: indicates whether the reconfiguration needs to be acted upon.
* This API call can be grouped with slim_control_ch API call with only one of
* the APIs specifying the commit flag to avoid reconfiguration being called too
* frequently. -EXFULL is returned if there is no space in TDM to reserve the
* bandwidth. -EBUSY is returned if reconfiguration is requested, but a request
* is already in progress.
*/
int slim_reservemsg_bw(struct slim_device *sb, u32 bw_bps, bool commit)
{
struct slim_controller *ctrl = sb->ctrl;
int ret = 0;
int sl;
mutex_lock(&sb->sldev_reconf);
if ((bw_bps >> 3) >= ctrl->a_framer->rootfreq)
sl = SLIM_SL_PER_SUPERFRAME;
else {
sl = (bw_bps * (SLIM_CL_PER_SUPERFRAME_DIV8/SLIM_CL_PER_SL/2) +
(ctrl->a_framer->rootfreq/2 - 1)) /
(ctrl->a_framer->rootfreq/2);
}
dev_dbg(&ctrl->dev, "request:bw:%d, slots:%d, current:%d\n", bw_bps, sl,
sb->cur_msgsl);
sb->pending_msgsl = sl;
if (commit == true)
ret = slim_reconfigure_now(sb);
mutex_unlock(&sb->sldev_reconf);
return ret;
}
EXPORT_SYMBOL(slim_reservemsg_bw);
/*
* slim_ctrl_clk_pause: Called by slimbus controller to request clock to be
* paused or woken up out of clock pause
* or woken up from clock pause
* @ctrl: controller requesting bus to be paused or woken up
* @wakeup: Wakeup this controller from clock pause.
* @restart: Restart time value per spec used for clock pause. This value
* isn't used when controller is to be woken up.
* This API executes clock pause reconfiguration sequence if wakeup is false.
* If wakeup is true, controller's wakeup is called
* Slimbus clock is idle and can be disabled by the controller later.
*/
int slim_ctrl_clk_pause(struct slim_controller *ctrl, bool wakeup, u8 restart)
{
int ret = 0;
int i;
DEFINE_SLIM_BCAST_TXN(txn, SLIM_MSG_CLK_PAUSE_SEQ_FLG |
SLIM_MSG_MC_BEGIN_RECONFIGURATION, 0, 3,
NULL, NULL, 0);
if (wakeup == false && restart > SLIM_CLK_UNSPECIFIED)
return -EINVAL;
mutex_lock(&ctrl->m_ctrl);
if (wakeup) {
if (ctrl->clk_state == SLIM_CLK_ACTIVE) {
mutex_unlock(&ctrl->m_ctrl);
return 0;
}
wait_for_completion(&ctrl->pause_comp);
/*
* Slimbus framework will call controller wakeup
* Controller should make sure that it sets active framer
* out of clock pause by doing appropriate setting
*/
if (ctrl->clk_state == SLIM_CLK_PAUSED && ctrl->wakeup)
ret = ctrl->wakeup(ctrl);
/*
* If wakeup fails, make sure that next attempt can succeed.
* Since we already consumed pause_comp, complete it so
* that next wakeup isn't blocked forever
*/
if (!ret)
ctrl->clk_state = SLIM_CLK_ACTIVE;
else
complete(&ctrl->pause_comp);
mutex_unlock(&ctrl->m_ctrl);
return ret;
}
switch (ctrl->clk_state) {
case SLIM_CLK_ENTERING_PAUSE:
case SLIM_CLK_PAUSE_FAILED:
/*
* If controller is already trying to enter clock pause,
* let it finish.
* In case of error, retry
* In both cases, previous clock pause has signalled
* completion.
*/
wait_for_completion(&ctrl->pause_comp);
/* retry upon failure */
if (ctrl->clk_state == SLIM_CLK_PAUSE_FAILED) {
ctrl->clk_state = SLIM_CLK_ACTIVE;
} else {
mutex_unlock(&ctrl->m_ctrl);
/*
* Signal completion so that wakeup can wait on
* it.
*/
complete(&ctrl->pause_comp);
return 0;
}
break;
case SLIM_CLK_PAUSED:
/* already paused */
mutex_unlock(&ctrl->m_ctrl);
return 0;
case SLIM_CLK_ACTIVE:
default:
break;
}
/* Pending response for a message */
for (i = 0; i < ctrl->last_tid; i++) {
if (ctrl->txnt[i]) {
ret = -EBUSY;
pr_info("slim_clk_pause: txn-rsp for %d pending", i);
mutex_unlock(&ctrl->m_ctrl);
return -EBUSY;
}
}
ctrl->clk_state = SLIM_CLK_ENTERING_PAUSE;
mutex_unlock(&ctrl->m_ctrl);
mutex_lock(&ctrl->sched.m_reconf);
/* Data channels active */
if (ctrl->sched.usedslots) {
pr_info("slim_clk_pause: data channel active");
ret = -EBUSY;
goto clk_pause_ret;
}
ret = slim_processtxn(ctrl, &txn, false);
if (ret)
goto clk_pause_ret;
txn.mc = SLIM_MSG_CLK_PAUSE_SEQ_FLG | SLIM_MSG_MC_NEXT_PAUSE_CLOCK;
txn.len = 1;
txn.rl = 4;
txn.wbuf = &restart;
ret = slim_processtxn(ctrl, &txn, false);
if (ret)
goto clk_pause_ret;
txn.mc = SLIM_MSG_CLK_PAUSE_SEQ_FLG | SLIM_MSG_MC_RECONFIGURE_NOW;
txn.len = 0;
txn.rl = 3;
txn.wbuf = NULL;
ret = slim_processtxn(ctrl, &txn, false);
if (ret)
goto clk_pause_ret;
clk_pause_ret:
if (ret)
ctrl->clk_state = SLIM_CLK_PAUSE_FAILED;
else
ctrl->clk_state = SLIM_CLK_PAUSED;
complete(&ctrl->pause_comp);
mutex_unlock(&ctrl->sched.m_reconf);
return ret;
}
EXPORT_SYMBOL(slim_ctrl_clk_pause);
MODULE_LICENSE("GPL v2");
MODULE_DESCRIPTION("Slimbus module");
MODULE_ALIAS("platform:slimbus");