2010-12-14 10:09:07

by Carlos Chinea

[permalink] [raw]
Subject: [RFC PATCHv4 0/7] HSI framework and drivers

Hi !

Here you have the fourth round of the HSI framework patches.

The patch series introduces the HSI framework, an SSI driver
for OMAP and a generic character device for HSI/SSI devices.

SSI, which is a legacy version of HSI, is used to connect the application
engine with the cellular modem on the Nokia N900.

In this round we have added/fixed:

- Address some comments from Tony Lindgren and Peter Henn.
- Add module support to the HSI framework.
- Fix sparse warnings in OMAP SSI driver.
- Fix module license.
- Remove spurious udev events generated by the HSI framework.
- Avoid removal of a controller claimed by a client.
- Fix hsi_char client device removal.

TODO:

- Move OMAP SSI to use omap hwmod

I would very glad to continue getting feedback about this proposal.

This patch series is based on 2.6.37-rc1.

Version 3 of the patch set: https://lkml.org/lkml/2010/10/11/60

Note: checkpatch reports a false positive on patch 1/7.

Andras Domokos (3):
HSI: hsi_char: Add HSI char device driver
HSI: hsi_char: Add HSI char device kernel configuration
HSI: hsi_char: Update ioctl-number.txt

Carlos Chinea (4):
HSI: hsi: Introducing HSI framework
HSI: omap_ssi: Introducing OMAP SSI driver
HSI: omap_ssi: Add OMAP SSI to the kernel configuration
HSI: Add HSI API documentation

Documentation/DocBook/device-drivers.tmpl | 17 +
Documentation/ioctl/ioctl-number.txt | 1 +
arch/arm/mach-omap2/Makefile | 2 +
arch/arm/mach-omap2/ssi.c | 134 +++
arch/arm/plat-omap/include/plat/ssi.h | 204 ++++
drivers/Kconfig | 2 +
drivers/Makefile | 1 +
drivers/hsi/Kconfig | 20 +
drivers/hsi/Makefile | 6 +
drivers/hsi/clients/Kconfig | 13 +
drivers/hsi/clients/Makefile | 5 +
drivers/hsi/clients/hsi_char.c | 1094 +++++++++++++++++
drivers/hsi/controllers/Kconfig | 23 +
drivers/hsi/controllers/Makefile | 5 +
drivers/hsi/controllers/omap_ssi.c | 1853 +++++++++++++++++++++++++++++
drivers/hsi/hsi.c | 496 ++++++++
drivers/hsi/hsi_boardinfo.c | 64 +
drivers/hsi/hsi_core.h | 37 +
include/linux/Kbuild | 1 +
include/linux/hsi/Kbuild | 1 +
include/linux/hsi/hsi.h | 389 ++++++
include/linux/hsi/hsi_char.h | 65 +
22 files changed, 4433 insertions(+), 0 deletions(-)
create mode 100644 arch/arm/mach-omap2/ssi.c
create mode 100644 arch/arm/plat-omap/include/plat/ssi.h
create mode 100644 drivers/hsi/Kconfig
create mode 100644 drivers/hsi/Makefile
create mode 100644 drivers/hsi/clients/Kconfig
create mode 100644 drivers/hsi/clients/Makefile
create mode 100644 drivers/hsi/clients/hsi_char.c
create mode 100644 drivers/hsi/controllers/Kconfig
create mode 100644 drivers/hsi/controllers/Makefile
create mode 100644 drivers/hsi/controllers/omap_ssi.c
create mode 100644 drivers/hsi/hsi.c
create mode 100644 drivers/hsi/hsi_boardinfo.c
create mode 100644 drivers/hsi/hsi_core.h
create mode 100644 include/linux/hsi/Kbuild
create mode 100644 include/linux/hsi/hsi.h
create mode 100644 include/linux/hsi/hsi_char.h


2010-12-14 10:09:15

by Carlos Chinea

[permalink] [raw]
Subject: [RFC PATCHv4 6/7] HSI: Add HSI API documentation

Add an entry for HSI in the device-drivers section of
the kernel documentation.

Signed-off-by: Carlos Chinea <[email protected]>
---
Documentation/DocBook/device-drivers.tmpl | 17 +++++++++++++++++
1 files changed, 17 insertions(+), 0 deletions(-)

diff --git a/Documentation/DocBook/device-drivers.tmpl b/Documentation/DocBook/device-drivers.tmpl
index 22edcbb..877028a 100644
--- a/Documentation/DocBook/device-drivers.tmpl
+++ b/Documentation/DocBook/device-drivers.tmpl
@@ -433,4 +433,21 @@ X!Idrivers/video/console/fonts.c
!Edrivers/i2c/i2c-core.c
</chapter>

+ <chapter id="hsi">
+ <title>High Speed Synchronous Serial Interface (HSI)</title>
+
+ <para>
+ High Speed Synchronous Serial Interface (HSI) is a
+ serial interface mainly used for connecting application
+ engines (APE) with cellular modem engines (CMT) in cellular
+ handsets.
+
+ HSI provides multiplexing for up to 16 logical channels,
+ low-latency and full duplex communication.
+ </para>
+
+!Iinclude/linux/hsi/hsi.h
+!Edrivers/hsi/hsi.c
+ </chapter>
+
</book>
--
1.7.1

2010-12-14 10:09:24

by Carlos Chinea

[permalink] [raw]
Subject: [RFC PATCHv4 4/7] HSI: hsi_char: Add HSI char device driver

From: Andras Domokos <[email protected]>

Add HSI char device driver to the kernel.

Signed-off-by: Andras Domokos <[email protected]>
Signed-off-by: Carlos Chinea <[email protected]>
---
drivers/hsi/clients/hsi_char.c | 1094 ++++++++++++++++++++++++++++++++++++++++
include/linux/hsi/hsi_char.h | 65 +++
2 files changed, 1159 insertions(+), 0 deletions(-)
create mode 100644 drivers/hsi/clients/hsi_char.c
create mode 100644 include/linux/hsi/hsi_char.h

diff --git a/drivers/hsi/clients/hsi_char.c b/drivers/hsi/clients/hsi_char.c
new file mode 100644
index 0000000..19aeb24
--- /dev/null
+++ b/drivers/hsi/clients/hsi_char.c
@@ -0,0 +1,1094 @@
+/*
+ * hsi-char.c
+ *
+ * HSI character device driver, implements the character device
+ * interface.
+ *
+ * Copyright (C) 2010 Nokia Corporation. All rights reserved.
+ *
+ * Contact: Andras Domokos <[email protected]>
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public License
+ * 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.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA
+ * 02110-1301 USA
+ */
+
+#include <linux/errno.h>
+#include <linux/types.h>
+#include <asm/atomic.h>
+#include <linux/kernel.h>
+#include <linux/init.h>
+#include <linux/module.h>
+#include <linux/list.h>
+#include <linux/slab.h>
+#include <linux/poll.h>
+#include <linux/ioctl.h>
+#include <linux/wait.h>
+#include <linux/fs.h>
+#include <linux/sched.h>
+#include <linux/device.h>
+#include <linux/cdev.h>
+#include <linux/uaccess.h>
+#include <linux/scatterlist.h>
+#include <linux/hsi/hsi.h>
+#include <linux/hsi/hsi_char.h>
+
+#define HSI_CHAR_CHANNELS 8
+#define HSI_CHAR_DEVS 8
+#define HSI_CHAR_MSGS 4
+
+#define HSI_CHST_UNAVAIL 0 /* SBZ! */
+#define HSI_CHST_AVAIL 1
+
+#define HSI_CHST_CLOSED (0 << 4)
+#define HSI_CHST_CLOSING (1 << 4)
+#define HSI_CHST_OPENING (2 << 4)
+#define HSI_CHST_OPENED (3 << 4)
+
+#define HSI_CHST_READOFF (0 << 8)
+#define HSI_CHST_READON (1 << 8)
+#define HSI_CHST_READING (2 << 8)
+
+#define HSI_CHST_WRITEOFF (0 << 12)
+#define HSI_CHST_WRITEON (1 << 12)
+#define HSI_CHST_WRITING (2 << 12)
+
+#define HSI_CHST_OC_MASK 0xf0
+#define HSI_CHST_RD_MASK 0xf00
+#define HSI_CHST_WR_MASK 0xf000
+
+#define HSI_CHST_OC(c) ((c)->state & HSI_CHST_OC_MASK)
+#define HSI_CHST_RD(c) ((c)->state & HSI_CHST_RD_MASK)
+#define HSI_CHST_WR(c) ((c)->state & HSI_CHST_WR_MASK)
+
+#define HSI_CHST_OC_SET(c, v) \
+ do { \
+ (c)->state &= ~HSI_CHST_OC_MASK; \
+ (c)->state |= v; \
+ } while (0);
+
+#define HSI_CHST_RD_SET(c, v) \
+ do { \
+ (c)->state &= ~HSI_CHST_RD_MASK; \
+ (c)->state |= v; \
+ } while (0);
+
+#define HSI_CHST_WR_SET(c, v) \
+ do { \
+ (c)->state &= ~HSI_CHST_WR_MASK; \
+ (c)->state |= v; \
+ } while (0);
+
+#define HSI_CHAR_POLL_RST (-1)
+#define HSI_CHAR_POLL_OFF 0
+#define HSI_CHAR_POLL_ON 1
+
+#define HSI_CHAR_RX 0
+#define HSI_CHAR_TX 1
+
+struct hsi_char_channel {
+ unsigned int ch;
+ unsigned int state;
+ int wlrefcnt;
+ int rxpoll;
+ struct hsi_client *cl;
+ struct list_head free_msgs_list;
+ struct list_head rx_msgs_queue;
+ struct list_head tx_msgs_queue;
+ int poll_event;
+ spinlock_t lock;
+ struct fasync_struct *async_queue;
+ wait_queue_head_t rx_wait;
+ wait_queue_head_t tx_wait;
+};
+
+struct hsi_char_client_data {
+ atomic_t refcnt;
+ int attached;
+ atomic_t breq;
+ struct hsi_char_channel channels[HSI_CHAR_DEVS];
+};
+
+static unsigned int max_data_size = 0x1000;
+module_param(max_data_size, uint, 1);
+MODULE_PARM_DESC(max_data_size, "max read/write data size [4,8..65536] (^2)");
+
+static int channels_map[HSI_CHAR_DEVS] = {0, -1, -1 , -1, -1, -1, -1, -1};
+module_param_array(channels_map, int, NULL, 0);
+MODULE_PARM_DESC(channels_map, "Array of HSI channels ([0...7]) to be probed");
+
+static dev_t hsi_char_dev;
+static struct hsi_char_client_data hsi_char_cl_data;
+
+static inline void hsi_char_msg_free(struct hsi_msg *msg)
+{
+ msg->complete = NULL;
+ msg->destructor = NULL;
+ kfree(sg_virt(msg->sgt.sgl));
+ hsi_free_msg(msg);
+}
+
+static inline void hsi_char_msgs_free(struct hsi_char_channel *channel)
+{
+ struct hsi_msg *msg, *tmp;
+
+ list_for_each_entry_safe(msg, tmp, &channel->free_msgs_list, link) {
+ list_del(&msg->link);
+ hsi_char_msg_free(msg);
+ }
+ list_for_each_entry_safe(msg, tmp, &channel->rx_msgs_queue, link) {
+ list_del(&msg->link);
+ hsi_char_msg_free(msg);
+ }
+ list_for_each_entry_safe(msg, tmp, &channel->tx_msgs_queue, link) {
+ list_del(&msg->link);
+ hsi_char_msg_free(msg);
+ }
+}
+
+static inline struct hsi_msg *hsi_char_msg_alloc(unsigned int alloc_size)
+{
+ struct hsi_msg *msg;
+ void *buf;
+
+ msg = hsi_alloc_msg(1, GFP_KERNEL);
+ if (!msg)
+ goto out;
+ buf = kmalloc(alloc_size, GFP_KERNEL);
+ if (!buf) {
+ hsi_free_msg(msg);
+ goto out;
+ }
+ sg_init_one(msg->sgt.sgl, buf, alloc_size);
+ msg->context = buf;
+ return msg;
+out:
+ return NULL;
+}
+
+static inline int hsi_char_msgs_alloc(struct hsi_char_channel *channel)
+{
+ struct hsi_msg *msg;
+ int i;
+
+ for (i = 0; i < HSI_CHAR_MSGS; i++) {
+ msg = hsi_char_msg_alloc(max_data_size);
+ if (!msg)
+ goto out;
+ msg->channel = channel->ch;
+ list_add_tail(&msg->link, &channel->free_msgs_list);
+ }
+ return 0;
+out:
+ hsi_char_msgs_free(channel);
+
+ return -ENOMEM;
+}
+
+static int _hsi_char_release(struct hsi_char_channel *channel, int remove)
+{
+ struct hsi_char_client_data *cl_data = hsi_client_drvdata(channel->cl);
+ int ret = 0, refcnt;
+
+ spin_lock_bh(&channel->lock);
+ if (HSI_CHST_OC(channel) != HSI_CHST_OPENED)
+ goto out;
+ HSI_CHST_OC_SET(channel, HSI_CHST_CLOSING);
+ spin_unlock_bh(&channel->lock);
+
+ while (channel->wlrefcnt > 0) {
+ hsi_stop_tx(channel->cl);
+ channel->wlrefcnt--;
+ }
+
+ if (channel->rxpoll == HSI_CHAR_POLL_ON)
+ channel->poll_event |= POLLERR;
+
+ wake_up_interruptible(&channel->rx_wait);
+ wake_up_interruptible(&channel->tx_wait);
+
+ refcnt = atomic_dec_return(&cl_data->refcnt);
+ if (!refcnt) {
+ hsi_flush(channel->cl);
+ hsi_release_port(channel->cl);
+ cl_data->attached = 0;
+ }
+ hsi_char_msgs_free(channel);
+
+ spin_lock_bh(&channel->lock);
+ HSI_CHST_OC_SET(channel, HSI_CHST_CLOSED);
+ HSI_CHST_RD_SET(channel, HSI_CHST_READOFF);
+ HSI_CHST_WR_SET(channel, HSI_CHST_WRITEOFF);
+out:
+ if (remove)
+ channel->cl = NULL;
+ spin_unlock_bh(&channel->lock);
+
+ return ret;
+}
+
+static int __devinit hsi_char_probe(struct device *dev)
+{
+ struct hsi_char_client_data *cl_data = &hsi_char_cl_data;
+ struct hsi_char_channel *channel = cl_data->channels;
+ struct hsi_client *cl = to_hsi_client(dev);
+ int i;
+
+ for (i = 0; i < HSI_CHAR_DEVS; i++, channel++) {
+ if (channel->state == HSI_CHST_AVAIL)
+ channel->cl = cl;
+ }
+ cl->hsi_start_rx = NULL;
+ cl->hsi_stop_rx = NULL;
+ atomic_set(&cl_data->refcnt, 0);
+ atomic_set(&cl_data->breq, 1);
+ cl_data->attached = 0;
+ hsi_client_set_drvdata(cl, cl_data);
+
+ return 0;
+}
+
+static int __devexit hsi_char_remove(struct device *dev)
+{
+ struct hsi_client *cl = to_hsi_client(dev);
+ struct hsi_char_client_data *cl_data = hsi_client_drvdata(cl);
+ struct hsi_char_channel *channel = cl_data->channels;
+ int i;
+
+ for (i = 0; i < HSI_CHAR_DEVS; i++, channel++) {
+ if (!(channel->state & HSI_CHST_AVAIL))
+ continue;
+ _hsi_char_release(channel, 1);
+ }
+
+ return 0;
+}
+
+static inline unsigned int hsi_char_msg_len_get(struct hsi_msg *msg)
+{
+ return msg->sgt.sgl->length;
+}
+
+static inline void hsi_char_msg_len_set(struct hsi_msg *msg, unsigned int len)
+{
+ msg->sgt.sgl->length = len;
+}
+
+static void hsi_char_data_available(struct hsi_msg *msg)
+{
+ struct hsi_char_client_data *cl_data = hsi_client_drvdata(msg->cl);
+ struct hsi_char_channel *channel = cl_data->channels + msg->channel;
+ int ret;
+
+ if (msg->status == HSI_STATUS_ERROR) {
+ ret = hsi_async_read(channel->cl, msg);
+ if (ret < 0) {
+ list_add_tail(&msg->link, &channel->free_msgs_list);
+ spin_lock_bh(&channel->lock);
+ list_add_tail(&msg->link, &channel->free_msgs_list);
+ channel->rxpoll = HSI_CHAR_POLL_OFF;
+ spin_unlock_bh(&channel->lock);
+ }
+ } else {
+ spin_lock_bh(&channel->lock);
+ channel->rxpoll = HSI_CHAR_POLL_OFF;
+ channel->poll_event |= (POLLIN | POLLRDNORM);
+ spin_unlock_bh(&channel->lock);
+ spin_lock_bh(&channel->lock);
+ list_add_tail(&msg->link, &channel->free_msgs_list);
+ spin_unlock_bh(&channel->lock);
+ wake_up_interruptible(&channel->rx_wait);
+ }
+}
+
+static void hsi_char_rx_completed(struct hsi_msg *msg)
+{
+ struct hsi_char_client_data *cl_data = hsi_client_drvdata(msg->cl);
+ struct hsi_char_channel *channel = cl_data->channels + msg->channel;
+
+ spin_lock_bh(&channel->lock);
+ list_add_tail(&msg->link, &channel->rx_msgs_queue);
+ spin_unlock_bh(&channel->lock);
+ wake_up_interruptible(&channel->rx_wait);
+}
+
+static void hsi_char_rx_msg_destructor(struct hsi_msg *msg)
+{
+ struct hsi_char_client_data *cl_data = hsi_client_drvdata(msg->cl);
+ struct hsi_char_channel *channel = cl_data->channels + msg->channel;
+
+ spin_lock_bh(&channel->lock);
+ list_add_tail(&msg->link, &channel->free_msgs_list);
+ HSI_CHST_RD_SET(channel, HSI_CHST_READOFF);
+ spin_unlock_bh(&channel->lock);
+}
+
+static void hsi_char_rx_poll_destructor(struct hsi_msg *msg)
+{
+ struct hsi_char_client_data *cl_data = hsi_client_drvdata(msg->cl);
+ struct hsi_char_channel *channel = cl_data->channels + msg->channel;
+
+ spin_lock_bh(&channel->lock);
+ list_add_tail(&msg->link, &channel->free_msgs_list);
+ channel->rxpoll = HSI_CHAR_POLL_RST;
+ spin_unlock_bh(&channel->lock);
+}
+
+static int hsi_char_rx_poll(struct hsi_char_channel *channel)
+{
+ struct hsi_msg *msg;
+ int ret = 0;
+
+ spin_lock_bh(&channel->lock);
+ if (list_empty(&channel->free_msgs_list)) {
+ ret = -ENOMEM;
+ goto out;
+ }
+ if (channel->rxpoll == HSI_CHAR_POLL_ON)
+ goto out;
+ msg = list_first_entry(&channel->free_msgs_list, struct hsi_msg, link);
+ list_del(&msg->link);
+ channel->rxpoll = HSI_CHAR_POLL_ON;
+ spin_unlock_bh(&channel->lock);
+ hsi_char_msg_len_set(msg, 0);
+ msg->complete = hsi_char_data_available;
+ msg->destructor = hsi_char_rx_poll_destructor;
+ /* don't touch msg->context! */
+ ret = hsi_async_read(channel->cl, msg);
+ spin_lock_bh(&channel->lock);
+ if (ret < 0) {
+ list_add_tail(&msg->link, &channel->free_msgs_list);
+ channel->rxpoll = HSI_CHAR_POLL_OFF;
+ goto out;
+ }
+out:
+ spin_unlock_bh(&channel->lock);
+
+ return ret;
+}
+
+static void hsi_char_tx_completed(struct hsi_msg *msg)
+{
+ struct hsi_char_client_data *cl_data = hsi_client_drvdata(msg->cl);
+ struct hsi_char_channel *channel = cl_data->channels + msg->channel;
+
+ spin_lock_bh(&channel->lock);
+ list_add_tail(&msg->link, &channel->tx_msgs_queue);
+ channel->poll_event |= (POLLOUT | POLLWRNORM);
+ spin_unlock_bh(&channel->lock);
+ wake_up_interruptible(&channel->tx_wait);
+}
+
+static void hsi_char_tx_msg_destructor(struct hsi_msg *msg)
+{
+ struct hsi_char_client_data *cl_data = hsi_client_drvdata(msg->cl);
+ struct hsi_char_channel *channel = cl_data->channels + msg->channel;
+
+ spin_lock_bh(&channel->lock);
+ list_add_tail(&msg->link, &channel->free_msgs_list);
+ HSI_CHST_WR_SET(channel, HSI_CHST_WRITEOFF);
+ spin_unlock_bh(&channel->lock);
+}
+
+static void hsi_char_rx_poll_rst(struct hsi_client *cl)
+{
+ struct hsi_char_client_data *cl_data = hsi_client_drvdata(cl);
+ struct hsi_char_channel *channel = cl_data->channels;
+ int i;
+
+ for (i = 0; i < HSI_CHAR_DEVS; i++, channel++) {
+ if ((HSI_CHST_OC(channel) == HSI_CHST_OPENED) &&
+ (channel->rxpoll == HSI_CHAR_POLL_RST))
+ hsi_char_rx_poll(channel);
+ }
+}
+
+static void hsi_char_reset(struct hsi_client *cl)
+{
+ hsi_flush(cl);
+ hsi_char_rx_poll_rst(cl);
+}
+
+static void hsi_char_rx_cancel(struct hsi_char_channel *channel)
+{
+ hsi_flush(channel->cl);
+ hsi_char_rx_poll_rst(channel->cl);
+}
+
+static void hsi_char_tx_cancel(struct hsi_char_channel *channel)
+{
+ hsi_flush(channel->cl);
+ hsi_char_rx_poll_rst(channel->cl);
+}
+
+static void hsi_char_bcast_break(struct hsi_client *cl)
+{
+ struct hsi_char_client_data *cl_data = hsi_client_drvdata(cl);
+ struct hsi_char_channel *channel = cl_data->channels;
+ int i;
+
+ for (i = 0; i < HSI_CHAR_DEVS; i++, channel++) {
+ if (HSI_CHST_OC(channel) != HSI_CHST_OPENED)
+ continue;
+ channel->poll_event |= POLLPRI;
+ wake_up_interruptible(&channel->rx_wait);
+ wake_up_interruptible(&channel->tx_wait);
+ }
+}
+
+static void hsi_char_break_received(struct hsi_msg *msg)
+{
+ struct hsi_char_client_data *cl_data = hsi_client_drvdata(msg->cl);
+ int ret;
+
+ hsi_char_bcast_break(msg->cl);
+ ret = hsi_async_read(msg->cl, msg);
+ if (ret < 0) {
+ hsi_free_msg(msg);
+ atomic_inc(&cl_data->breq);
+ }
+}
+
+static void hsi_char_break_req_destructor(struct hsi_msg *msg)
+{
+ struct hsi_char_client_data *cl_data = hsi_client_drvdata(msg->cl);
+
+ hsi_free_msg(msg);
+ atomic_inc(&cl_data->breq);
+}
+
+static int hsi_char_break_request(struct hsi_client *cl)
+{
+ struct hsi_char_client_data *cl_data = hsi_client_drvdata(cl);
+ struct hsi_msg *msg;
+ int ret = 0;
+
+ if (!atomic_dec_and_test(&cl_data->breq)) {
+ atomic_inc(&cl_data->breq);
+ return -EBUSY;
+ }
+ msg = hsi_alloc_msg(0, GFP_KERNEL);
+ if (!msg)
+ return -ENOMEM;
+ msg->break_frame = 1;
+ msg->complete = hsi_char_break_received;
+ msg->destructor = hsi_char_break_req_destructor;
+ ret = hsi_async_read(cl, msg);
+ if (ret < 0)
+ hsi_free_msg(msg);
+
+ return ret;
+}
+
+static int hsi_char_break_send(struct hsi_client *cl)
+{
+ struct hsi_msg *msg;
+ int ret = 0;
+
+ msg = hsi_alloc_msg(0, GFP_ATOMIC);
+ if (!msg)
+ return -ENOMEM;
+ msg->break_frame = 1;
+ msg->complete = hsi_free_msg;
+ msg->destructor = hsi_free_msg;
+ ret = hsi_async_write(cl, msg);
+ if (ret < 0)
+ hsi_free_msg(msg);
+
+ return ret;
+}
+
+static inline int ssi_check_common_cfg(struct hsi_config *cfg)
+{
+ if ((cfg->mode != HSI_MODE_STREAM) && (cfg->mode != HSI_MODE_FRAME))
+ return -EINVAL;
+ if ((cfg->channels == 0) || (cfg->channels > HSI_CHAR_CHANNELS))
+ return -EINVAL;
+ if (cfg->channels & (cfg->channels - 1))
+ return -EINVAL;
+
+ return 0;
+}
+
+static inline int ssi_check_rx_cfg(struct hsi_config *cfg)
+{
+ int ret;
+
+ ret = ssi_check_common_cfg(cfg);
+ if (ret < 0)
+ return ret;
+ if ((cfg->flow != HSI_FLOW_SYNC) && (cfg->flow != HSI_FLOW_PIPE))
+ return -EINVAL;
+
+ return 0;
+}
+
+static inline int ssi_check_tx_cfg(struct hsi_config *cfg)
+{
+ int ret;
+
+ ret = ssi_check_common_cfg(cfg);
+ if (ret < 0)
+ return ret;
+ if ((cfg->arb_mode != HSI_ARB_RR) && (cfg->arb_mode != HSI_ARB_PRIO))
+ return -EINVAL;
+
+ return 0;
+}
+
+static inline int hsi_char_cfg_set(struct hsi_client *cl,
+ struct hsi_config *cfg, int dir)
+{
+ struct hsi_config *rxtx_cfg;
+ int ret = 0;
+
+ if (dir == HSI_CHAR_RX) {
+ rxtx_cfg = &cl->rx_cfg;
+ ret = ssi_check_rx_cfg(cfg);
+ } else {
+ rxtx_cfg = &cl->tx_cfg;
+ ret = ssi_check_tx_cfg(cfg);
+ }
+ if (ret < 0)
+ return ret;
+
+ *rxtx_cfg = *cfg;
+ ret = hsi_setup(cl);
+ if (ret < 0)
+ return ret;
+
+ if ((dir == HSI_CHAR_RX) && (cfg->mode == HSI_MODE_FRAME))
+ hsi_char_break_request(cl);
+
+ return ret;
+}
+
+static inline void hsi_char_cfg_get(struct hsi_client *cl,
+ struct hsi_config *cfg, int dir)
+{
+ struct hsi_config *rxtx_cfg;
+
+ if (dir == HSI_CHAR_RX)
+ rxtx_cfg = &cl->rx_cfg;
+ else
+ rxtx_cfg = &cl->tx_cfg;
+ *cfg = *rxtx_cfg;
+}
+
+static inline void hsi_char_rx2icfg(struct hsi_config *cfg,
+ struct hsc_rx_config *rx_cfg)
+{
+ cfg->mode = rx_cfg->mode;
+ cfg->flow = rx_cfg->flow;
+ cfg->channels = rx_cfg->channels;
+ cfg->speed = 0;
+}
+
+static inline void hsi_char_tx2icfg(struct hsi_config *cfg,
+ struct hsc_tx_config *tx_cfg)
+{
+ cfg->mode = tx_cfg->mode;
+ cfg->channels = tx_cfg->channels;
+ cfg->speed = tx_cfg->speed;
+ cfg->arb_mode = tx_cfg->arb_mode;
+}
+
+static inline void hsi_char_rx2ecfg(struct hsc_rx_config *rx_cfg,
+ struct hsi_config *cfg)
+{
+ rx_cfg->mode = cfg->mode;
+ rx_cfg->flow = cfg->flow;
+ rx_cfg->channels = cfg->channels;
+}
+
+static inline void hsi_char_tx2ecfg(struct hsc_tx_config *tx_cfg,
+ struct hsi_config *cfg)
+{
+ tx_cfg->mode = cfg->mode;
+ tx_cfg->channels = cfg->channels;
+ tx_cfg->speed = cfg->speed;
+ tx_cfg->arb_mode = cfg->arb_mode;
+}
+
+static ssize_t hsi_char_read(struct file *file, char __user *buf,
+ size_t len, loff_t *ppos)
+{
+ struct hsi_char_channel *channel = file->private_data;
+ struct hsi_msg *msg = NULL;
+ ssize_t ret;
+
+ if (len == 0) {
+ channel->poll_event &= ~POLLPRI;
+ return 0;
+ }
+ channel->poll_event &= ~POLLPRI;
+
+ if (!IS_ALIGNED(len, sizeof(u32)))
+ return -EINVAL;
+
+ if (len > max_data_size)
+ len = max_data_size;
+
+ spin_lock_bh(&channel->lock);
+ if (HSI_CHST_OC(channel) != HSI_CHST_OPENED) {
+ ret = -ENODEV;
+ goto out;
+ }
+ if (HSI_CHST_RD(channel) != HSI_CHST_READOFF) {
+ ret = -EBUSY;
+ goto out;
+ }
+ if (channel->ch >= channel->cl->rx_cfg.channels) {
+ ret = -ENODEV;
+ goto out;
+ }
+ if (list_empty(&channel->free_msgs_list)) {
+ ret = -ENOMEM;
+ goto out;
+ }
+ msg = list_first_entry(&channel->free_msgs_list, struct hsi_msg, link);
+ list_del(&msg->link);
+ spin_unlock_bh(&channel->lock);
+ hsi_char_msg_len_set(msg, len);
+ msg->complete = hsi_char_rx_completed;
+ msg->destructor = hsi_char_rx_msg_destructor;
+ ret = hsi_async_read(channel->cl, msg);
+ spin_lock_bh(&channel->lock);
+ if (ret < 0)
+ goto out;
+ HSI_CHST_RD_SET(channel, HSI_CHST_READING);
+ msg = NULL;
+
+ for ( ; ; ) {
+ DEFINE_WAIT(wait);
+
+ if (!list_empty(&channel->rx_msgs_queue)) {
+ msg = list_first_entry(&channel->rx_msgs_queue,
+ struct hsi_msg, link);
+ HSI_CHST_RD_SET(channel, HSI_CHST_READOFF);
+ channel->poll_event &= ~(POLLIN | POLLRDNORM);
+ list_del(&msg->link);
+ spin_unlock_bh(&channel->lock);
+ if (msg->status == HSI_STATUS_ERROR) {
+ ret = -EIO;
+ } else {
+ ret = copy_to_user((void __user *)buf,
+ msg->context,
+ hsi_char_msg_len_get(msg));
+ if (ret)
+ ret = -EFAULT;
+ else
+ ret = hsi_char_msg_len_get(msg);
+ }
+ spin_lock_bh(&channel->lock);
+ break;
+ } else if (signal_pending(current)) {
+ spin_unlock_bh(&channel->lock);
+ hsi_char_rx_cancel(channel);
+ spin_lock_bh(&channel->lock);
+ HSI_CHST_RD_SET(channel, HSI_CHST_READOFF);
+ ret = -EINTR;
+ break;
+ } else if ((HSI_CHST_OC(channel) == HSI_CHST_CLOSING) ||
+ (HSI_CHST_OC(channel) == HSI_CHST_CLOSING)) {
+ ret = -EIO;
+ break;
+ }
+ prepare_to_wait(&channel->rx_wait, &wait, TASK_INTERRUPTIBLE);
+ spin_unlock_bh(&channel->lock);
+
+ schedule();
+
+ spin_lock_bh(&channel->lock);
+ finish_wait(&channel->rx_wait, &wait);
+ }
+out:
+ if (msg)
+ list_add_tail(&msg->link, &channel->free_msgs_list);
+ spin_unlock_bh(&channel->lock);
+
+ return ret;
+}
+
+static ssize_t hsi_char_write(struct file *file, const char __user *buf,
+ size_t len, loff_t *ppos)
+{
+ struct hsi_char_channel *channel = file->private_data;
+ struct hsi_msg *msg = NULL;
+ ssize_t ret;
+
+ if ((len == 0) || !IS_ALIGNED(len, sizeof(u32)))
+ return -EINVAL;
+
+ if (len > max_data_size)
+ len = max_data_size;
+
+ spin_lock_bh(&channel->lock);
+ if (HSI_CHST_OC(channel) != HSI_CHST_OPENED) {
+ ret = -ENODEV;
+ goto out;
+ }
+ if (HSI_CHST_WR(channel) != HSI_CHST_WRITEOFF) {
+ ret = -EBUSY;
+ goto out;
+ }
+ if (channel->ch >= channel->cl->tx_cfg.channels) {
+ ret = -ENODEV;
+ goto out;
+ }
+ if (list_empty(&channel->free_msgs_list)) {
+ ret = -ENOMEM;
+ goto out;
+ }
+ msg = list_first_entry(&channel->free_msgs_list, struct hsi_msg, link);
+ list_del(&msg->link);
+ HSI_CHST_WR_SET(channel, HSI_CHST_WRITEON);
+ spin_unlock_bh(&channel->lock);
+
+ if (copy_from_user(msg->context, (void __user *)buf, len)) {
+ spin_lock_bh(&channel->lock);
+ HSI_CHST_WR_SET(channel, HSI_CHST_WRITEOFF);
+ ret = -EFAULT;
+ goto out;
+ }
+
+ hsi_char_msg_len_set(msg, len);
+ msg->complete = hsi_char_tx_completed;
+ msg->destructor = hsi_char_tx_msg_destructor;
+ channel->poll_event &= ~(POLLOUT | POLLWRNORM);
+ ret = hsi_async_write(channel->cl, msg);
+ spin_lock_bh(&channel->lock);
+ if (ret < 0) {
+ channel->poll_event |= (POLLOUT | POLLWRNORM);
+ HSI_CHST_WR_SET(channel, HSI_CHST_WRITEOFF);
+ goto out;
+ }
+ HSI_CHST_WR_SET(channel, HSI_CHST_WRITING);
+ msg = NULL;
+
+ for ( ; ; ) {
+ DEFINE_WAIT(wait);
+
+ if (!list_empty(&channel->tx_msgs_queue)) {
+ msg = list_first_entry(&channel->tx_msgs_queue,
+ struct hsi_msg, link);
+ list_del(&msg->link);
+ HSI_CHST_WR_SET(channel, HSI_CHST_WRITEOFF);
+ if (msg->status == HSI_STATUS_ERROR)
+ ret = -EIO;
+ else
+ ret = hsi_char_msg_len_get(msg);
+ break;
+ } else if (signal_pending(current)) {
+ spin_unlock_bh(&channel->lock);
+ hsi_char_tx_cancel(channel);
+ spin_lock_bh(&channel->lock);
+ HSI_CHST_WR_SET(channel, HSI_CHST_WRITEOFF);
+ ret = -EINTR;
+ break;
+ } else if ((HSI_CHST_OC(channel) == HSI_CHST_CLOSING) ||
+ (HSI_CHST_OC(channel) == HSI_CHST_CLOSING)) {
+ ret = -EIO;
+ break;
+ }
+ prepare_to_wait(&channel->tx_wait, &wait, TASK_INTERRUPTIBLE);
+ spin_unlock_bh(&channel->lock);
+
+ schedule();
+
+ spin_lock_bh(&channel->lock);
+ finish_wait(&channel->tx_wait, &wait);
+ }
+out:
+ if (msg)
+ list_add_tail(&msg->link, &channel->free_msgs_list);
+
+ spin_unlock_bh(&channel->lock);
+
+ return ret;
+}
+
+static unsigned int hsi_char_poll(struct file *file, poll_table *wait)
+{
+ struct hsi_char_channel *channel = file->private_data;
+ unsigned int ret;
+
+ spin_lock_bh(&channel->lock);
+ if ((HSI_CHST_OC(channel) != HSI_CHST_OPENED) ||
+ (channel->ch >= channel->cl->rx_cfg.channels)) {
+ spin_unlock_bh(&channel->lock);
+ return -ENODEV;
+ }
+ poll_wait(file, &channel->rx_wait, wait);
+ poll_wait(file, &channel->tx_wait, wait);
+ ret = channel->poll_event;
+ spin_unlock_bh(&channel->lock);
+ hsi_char_rx_poll(channel);
+
+ return ret;
+}
+
+static long hsi_char_ioctl(struct file *file, unsigned int cmd,
+ unsigned long arg)
+{
+ struct hsi_char_channel *channel = file->private_data;
+ unsigned int state;
+ struct hsi_config cfg;
+ struct hsc_rx_config rx_cfg;
+ struct hsc_tx_config tx_cfg;
+ long ret = 0;
+
+ if (HSI_CHST_OC(channel) != HSI_CHST_OPENED)
+ return -ENODEV;
+
+ switch (cmd) {
+ case HSC_RESET:
+ hsi_char_reset(channel->cl);
+ break;
+ case HSC_SET_PM:
+ if (copy_from_user(&state, (void __user *)arg, sizeof(state)))
+ return -EFAULT;
+ if (state == HSC_PM_DISABLE) {
+ ret = hsi_start_tx(channel->cl);
+ if (!ret)
+ channel->wlrefcnt++;
+ } else if ((state == HSC_PM_ENABLE)
+ && (channel->wlrefcnt > 0)) {
+ ret = hsi_stop_tx(channel->cl);
+ if (!ret)
+ channel->wlrefcnt--;
+ } else {
+ ret = -EINVAL;
+ }
+ break;
+ case HSC_SEND_BREAK:
+ return hsi_char_break_send(channel->cl);
+ case HSC_SET_RX:
+ if (copy_from_user(&rx_cfg, (void __user *)arg, sizeof(rx_cfg)))
+ return -EFAULT;
+ hsi_char_rx2icfg(&cfg, &rx_cfg);
+ return hsi_char_cfg_set(channel->cl, &cfg, HSI_CHAR_RX);
+ case HSC_GET_RX:
+ hsi_char_cfg_get(channel->cl, &cfg, HSI_CHAR_RX);
+ hsi_char_rx2ecfg(&rx_cfg, &cfg);
+ if (copy_to_user((void __user *)arg, &rx_cfg, sizeof(rx_cfg)))
+ return -EFAULT;
+ break;
+ case HSC_SET_TX:
+ if (copy_from_user(&tx_cfg, (void __user *)arg, sizeof(tx_cfg)))
+ return -EFAULT;
+ hsi_char_tx2icfg(&cfg, &tx_cfg);
+ return hsi_char_cfg_set(channel->cl, &cfg, HSI_CHAR_TX);
+ case HSC_GET_TX:
+ hsi_char_cfg_get(channel->cl, &cfg, HSI_CHAR_TX);
+ hsi_char_tx2ecfg(&tx_cfg, &cfg);
+ if (copy_to_user((void __user *)arg, &tx_cfg, sizeof(tx_cfg)))
+ return -EFAULT;
+ break;
+ default:
+ return -ENOIOCTLCMD;
+ }
+
+ return ret;
+}
+
+static int hsi_char_open(struct inode *inode, struct file *file)
+{
+ struct hsi_char_client_data *cl_data = &hsi_char_cl_data;
+ struct hsi_char_channel *channel = cl_data->channels + iminor(inode);
+ int ret = 0, refcnt;
+
+ spin_lock_bh(&channel->lock);
+ if ((channel->state == HSI_CHST_UNAVAIL) || (!channel->cl)) {
+ ret = -ENODEV;
+ goto out;
+ }
+ if (HSI_CHST_OC(channel) != HSI_CHST_CLOSED) {
+ ret = -EBUSY;
+ goto out;
+ }
+ HSI_CHST_OC_SET(channel, HSI_CHST_OPENING);
+ spin_unlock_bh(&channel->lock);
+
+ refcnt = atomic_inc_return(&cl_data->refcnt);
+ if (refcnt == 1) {
+ if (cl_data->attached) {
+ atomic_dec(&cl_data->refcnt);
+ spin_lock_bh(&channel->lock);
+ HSI_CHST_OC_SET(channel, HSI_CHST_CLOSED);
+ ret = -EBUSY;
+ goto out;
+ }
+ ret = hsi_claim_port(channel->cl, 0);
+ if (ret < 0) {
+ atomic_dec(&cl_data->refcnt);
+ spin_lock_bh(&channel->lock);
+ HSI_CHST_OC_SET(channel, HSI_CHST_CLOSED);
+ goto out;
+ }
+ hsi_setup(channel->cl);
+ } else if (!cl_data->attached) {
+ atomic_dec(&cl_data->refcnt);
+ spin_lock_bh(&channel->lock);
+ HSI_CHST_OC_SET(channel, HSI_CHST_CLOSED);
+ ret = -ENODEV;
+ goto out;
+ }
+ ret = hsi_char_msgs_alloc(channel);
+
+ if (ret < 0) {
+ refcnt = atomic_dec_return(&cl_data->refcnt);
+ if (!refcnt)
+ hsi_release_port(channel->cl);
+ spin_lock_bh(&channel->lock);
+ HSI_CHST_OC_SET(channel, HSI_CHST_CLOSED);
+ goto out;
+ }
+ if (refcnt == 1)
+ cl_data->attached = 1;
+ channel->wlrefcnt = 0;
+ channel->rxpoll = HSI_CHAR_POLL_OFF;
+ channel->poll_event = (POLLOUT | POLLWRNORM);
+ file->private_data = channel;
+ spin_lock_bh(&channel->lock);
+ HSI_CHST_OC_SET(channel, HSI_CHST_OPENED);
+out:
+ spin_unlock_bh(&channel->lock);
+
+ return ret;
+}
+
+static int hsi_char_release(struct inode *inode, struct file *file)
+{
+ struct hsi_char_channel *channel = file->private_data;
+ return _hsi_char_release(channel, 0);
+}
+
+static int hsi_char_fasync(int fd, struct file *file, int on)
+{
+ struct hsi_char_channel *channel = file->private_data;
+
+ if (fasync_helper(fd, file, on, &channel->async_queue) < 0)
+ return -EIO;
+
+ return 0;
+}
+
+static const struct file_operations hsi_char_fops = {
+ .owner = THIS_MODULE,
+ .read = hsi_char_read,
+ .write = hsi_char_write,
+ .poll = hsi_char_poll,
+ .unlocked_ioctl = hsi_char_ioctl,
+ .open = hsi_char_open,
+ .release = hsi_char_release,
+ .fasync = hsi_char_fasync,
+};
+
+static struct hsi_client_driver hsi_char_driver = {
+ .driver = {
+ .name = "hsi_char",
+ .owner = THIS_MODULE,
+ .probe = hsi_char_probe,
+ .remove = hsi_char_remove,
+ },
+};
+
+static inline void hsi_char_channel_init(struct hsi_char_channel *channel)
+{
+ channel->state = HSI_CHST_AVAIL;
+ INIT_LIST_HEAD(&channel->free_msgs_list);
+ init_waitqueue_head(&channel->rx_wait);
+ init_waitqueue_head(&channel->tx_wait);
+ spin_lock_init(&channel->lock);
+ INIT_LIST_HEAD(&channel->rx_msgs_queue);
+ INIT_LIST_HEAD(&channel->tx_msgs_queue);
+}
+
+static struct cdev hsi_char_cdev;
+
+static int __init hsi_char_init(void)
+{
+ char devname[] = "hsi_char";
+ struct hsi_char_client_data *cl_data = &hsi_char_cl_data;
+ struct hsi_char_channel *channel = cl_data->channels;
+ unsigned long ch_mask = 0;
+ unsigned int i;
+ int ret;
+
+ if ((max_data_size < 4) || (max_data_size > 0x10000) ||
+ (max_data_size & (max_data_size - 1))) {
+ pr_err("Invalid max read/write data size");
+ return -EINVAL;
+ }
+
+ for (i = 0; i < HSI_CHAR_DEVS && channels_map[i] >= 0; i++) {
+ if (channels_map[i] >= HSI_CHAR_DEVS) {
+ pr_err("Invalid HSI/SSI channel specified");
+ return -EINVAL;
+ }
+ set_bit(channels_map[i], &ch_mask);
+ }
+
+ if (i == 0) {
+ pr_err("No HSI channels available");
+ return -EINVAL;
+ }
+
+ memset(cl_data->channels, 0, sizeof(cl_data->channels));
+ for (i = 0; i < HSI_CHAR_DEVS; i++, channel++) {
+ channel->ch = i;
+ channel->state = HSI_CHST_UNAVAIL;
+ if (test_bit(i, &ch_mask))
+ hsi_char_channel_init(channel);
+ }
+
+ ret = hsi_register_client_driver(&hsi_char_driver);
+ if (ret) {
+ pr_err("Error while registering HSI/SSI driver %d", ret);
+ return ret;
+ }
+
+ ret = alloc_chrdev_region(&hsi_char_dev, 0, HSI_CHAR_DEVS, devname);
+ if (ret < 0) {
+ hsi_unregister_client_driver(&hsi_char_driver);
+ return ret;
+ }
+
+ cdev_init(&hsi_char_cdev, &hsi_char_fops);
+ ret = cdev_add(&hsi_char_cdev, hsi_char_dev, HSI_CHAR_DEVS);
+ if (ret) {
+ unregister_chrdev_region(hsi_char_dev, HSI_CHAR_DEVS);
+ hsi_unregister_client_driver(&hsi_char_driver);
+ return ret;
+ }
+
+ pr_info("HSI/SSI char device loaded\n");
+
+ return 0;
+}
+module_init(hsi_char_init);
+
+static void __exit hsi_char_exit(void)
+{
+ cdev_del(&hsi_char_cdev);
+ unregister_chrdev_region(hsi_char_dev, HSI_CHAR_DEVS);
+ hsi_unregister_client_driver(&hsi_char_driver);
+ pr_info("HSI char device removed\n");
+}
+module_exit(hsi_char_exit);
+
+MODULE_AUTHOR("Andras Domokos <[email protected]>");
+MODULE_ALIAS("hsi:hsi_char");
+MODULE_DESCRIPTION("HSI character device");
+MODULE_LICENSE("GPL v2");
diff --git a/include/linux/hsi/hsi_char.h b/include/linux/hsi/hsi_char.h
new file mode 100644
index 0000000..fc49897
--- /dev/null
+++ b/include/linux/hsi/hsi_char.h
@@ -0,0 +1,65 @@
+/*
+ * hsi_char.h
+ *
+ * Part of the HSI character device driver.
+ *
+ * Copyright (C) 2010 Nokia Corporation. All rights reserved.
+ *
+ * Contact: Andras Domokos <andras.domokos at nokia.com>
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public License
+ * 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.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA
+ * 02110-1301 USA
+ */
+
+
+#ifndef __HSI_CHAR_H
+#define __HSI_CHAR_H
+
+#define HSI_CHAR_MAGIC 'k'
+#define HSC_IOW(num, dtype) _IOW(HSI_CHAR_MAGIC, num, dtype)
+#define HSC_IOR(num, dtype) _IOR(HSI_CHAR_MAGIC, num, dtype)
+#define HSC_IOWR(num, dtype) _IOWR(HSI_CHAR_MAGIC, num, dtype)
+#define HSC_IO(num) _IO(HSI_CHAR_MAGIC, num)
+
+#define HSC_RESET HSC_IO(16)
+#define HSC_SET_PM HSC_IO(17)
+#define HSC_SEND_BREAK HSC_IO(18)
+#define HSC_SET_RX HSC_IOW(19, struct hsc_rx_config)
+#define HSC_GET_RX HSC_IOW(20, struct hsc_rx_config)
+#define HSC_SET_TX HSC_IOW(21, struct hsc_tx_config)
+#define HSC_GET_TX HSC_IOW(22, struct hsc_tx_config)
+
+#define HSC_PM_DISABLE 0
+#define HSC_PM_ENABLE 1
+
+#define HSC_MODE_STREAM 1
+#define HSC_MODE_FRAME 2
+#define HSC_FLOW_SYNC 0
+#define HSC_ARB_RR 0
+#define HSC_ARB_PRIO 1
+
+struct hsc_rx_config {
+ uint32_t mode;
+ uint32_t flow;
+ uint32_t channels;
+};
+
+struct hsc_tx_config {
+ uint32_t mode;
+ uint32_t channels;
+ uint32_t speed;
+ uint32_t arb_mode;
+};
+
+#endif /* __HSI_CHAR_H */
--
1.7.1

2010-12-14 10:09:26

by Carlos Chinea

[permalink] [raw]
Subject: [RFC PATCHv4 5/7] HSI: hsi_char: Add HSI char device kernel configuration

From: Andras Domokos <[email protected]>

Add HSI character device kernel configuration

Signed-off-by: Andras Domokos <[email protected]>
Signed-off-by: Carlos Chinea <[email protected]>
---
drivers/hsi/Kconfig | 1 +
drivers/hsi/Makefile | 2 +-
drivers/hsi/clients/Kconfig | 13 +++++++++++++
drivers/hsi/clients/Makefile | 5 +++++
include/linux/Kbuild | 1 +
include/linux/hsi/Kbuild | 1 +
6 files changed, 22 insertions(+), 1 deletions(-)
create mode 100644 drivers/hsi/clients/Kconfig
create mode 100644 drivers/hsi/clients/Makefile
create mode 100644 include/linux/hsi/Kbuild

diff --git a/drivers/hsi/Kconfig b/drivers/hsi/Kconfig
index bba73ea..2c76de4 100644
--- a/drivers/hsi/Kconfig
+++ b/drivers/hsi/Kconfig
@@ -15,5 +15,6 @@ config HSI_BOARDINFO
default y

source "drivers/hsi/controllers/Kconfig"
+source "drivers/hsi/clients/Kconfig"

endif # HSI
diff --git a/drivers/hsi/Makefile b/drivers/hsi/Makefile
index 0de87bd..d47ca5d 100644
--- a/drivers/hsi/Makefile
+++ b/drivers/hsi/Makefile
@@ -3,4 +3,4 @@
#
obj-$(CONFIG_HSI_BOARDINFO) += hsi_boardinfo.o
obj-$(CONFIG_HSI) += hsi.o
-obj-y += controllers/
+obj-y += controllers/ clients/
diff --git a/drivers/hsi/clients/Kconfig b/drivers/hsi/clients/Kconfig
new file mode 100644
index 0000000..3bacd27
--- /dev/null
+++ b/drivers/hsi/clients/Kconfig
@@ -0,0 +1,13 @@
+#
+# HSI clients configuration
+#
+
+comment "HSI clients"
+
+config HSI_CHAR
+ tristate "HSI/SSI character driver"
+ depends on HSI
+ ---help---
+ If you say Y here, you will enable the HSI/SSI character driver.
+ This driver provides a simple character device interface for
+ serial communication with the cellular modem over HSI/SSI bus.
diff --git a/drivers/hsi/clients/Makefile b/drivers/hsi/clients/Makefile
new file mode 100644
index 0000000..327c0e2
--- /dev/null
+++ b/drivers/hsi/clients/Makefile
@@ -0,0 +1,5 @@
+#
+# Makefile for HSI clients
+#
+
+obj-$(CONFIG_HSI_CHAR) += hsi_char.o
diff --git a/include/linux/Kbuild b/include/linux/Kbuild
index 97319a8..b1eafe6 100644
--- a/include/linux/Kbuild
+++ b/include/linux/Kbuild
@@ -2,6 +2,7 @@ header-y += byteorder/
header-y += can/
header-y += dvb/
header-y += hdlc/
+header-y += hsi/
header-y += isdn/
header-y += nfsd/
header-y += raid/
diff --git a/include/linux/hsi/Kbuild b/include/linux/hsi/Kbuild
new file mode 100644
index 0000000..271a770
--- /dev/null
+++ b/include/linux/hsi/Kbuild
@@ -0,0 +1 @@
+header-y += hsi_char.h
--
1.7.1

2010-12-14 10:09:22

by Carlos Chinea

[permalink] [raw]
Subject: [RFC PATCHv4 7/7] HSI: hsi_char: Update ioctl-number.txt

From: Andras Domokos <[email protected]>

Added ioctl range for HSI char devices to the documentation

Signed-off-by: Andras Domokos <[email protected]>
Signed-off-by: Carlos Chinea <[email protected]>
---
Documentation/ioctl/ioctl-number.txt | 1 +
1 files changed, 1 insertions(+), 0 deletions(-)

diff --git a/Documentation/ioctl/ioctl-number.txt b/Documentation/ioctl/ioctl-number.txt
index 63ffd78..955bf5e 100644
--- a/Documentation/ioctl/ioctl-number.txt
+++ b/Documentation/ioctl/ioctl-number.txt
@@ -225,6 +225,7 @@ Code Seq#(hex) Include File Comments
'j' 00-3F linux/joystick.h
'k' 00-0F linux/spi/spidev.h conflict!
'k' 00-05 video/kyro.h conflict!
+'k' 10-17 linux/hsi/hsi_char.h HSI character device
'l' 00-3F linux/tcfs_fs.h transparent cryptographic file system
<http://web.archive.org/web/*/http://mikonos.dia.unisa.it/tcfs>
'l' 40-7F linux/udf_fs_i.h in development:
--
1.7.1

2010-12-14 10:10:38

by Carlos Chinea

[permalink] [raw]
Subject: [RFC PATCHv4 3/7] HSI: omap_ssi: Add OMAP SSI to the kernel configuration

Add OMAP SSI device and driver to the kernel configuration

Signed-off-by: Carlos Chinea <[email protected]>
---
arch/arm/mach-omap2/Makefile | 2 ++
drivers/hsi/Kconfig | 2 ++
drivers/hsi/Makefile | 1 +
drivers/hsi/controllers/Kconfig | 23 +++++++++++++++++++++++
drivers/hsi/controllers/Makefile | 5 +++++
5 files changed, 33 insertions(+), 0 deletions(-)
create mode 100644 drivers/hsi/controllers/Kconfig
create mode 100644 drivers/hsi/controllers/Makefile

diff --git a/arch/arm/mach-omap2/Makefile b/arch/arm/mach-omap2/Makefile
index b0810b9..1a1c0d9 100644
--- a/arch/arm/mach-omap2/Makefile
+++ b/arch/arm/mach-omap2/Makefile
@@ -111,6 +111,8 @@ obj-y += $(i2c-omap-m) $(i2c-omap-y)
ifneq ($(CONFIG_TIDSPBRIDGE),)
obj-y += dsp.o
endif
+omap-ssi-$(CONFIG_OMAP_SSI) := ssi.o
+obj-y += $(omap-ssi-m) $(omap-ssi-y)

# Specific board support
obj-$(CONFIG_MACH_OMAP_GENERIC) += board-generic.o
diff --git a/drivers/hsi/Kconfig b/drivers/hsi/Kconfig
index 937062e..bba73ea 100644
--- a/drivers/hsi/Kconfig
+++ b/drivers/hsi/Kconfig
@@ -14,4 +14,6 @@ config HSI_BOARDINFO
bool
default y

+source "drivers/hsi/controllers/Kconfig"
+
endif # HSI
diff --git a/drivers/hsi/Makefile b/drivers/hsi/Makefile
index ed94a3a..0de87bd 100644
--- a/drivers/hsi/Makefile
+++ b/drivers/hsi/Makefile
@@ -3,3 +3,4 @@
#
obj-$(CONFIG_HSI_BOARDINFO) += hsi_boardinfo.o
obj-$(CONFIG_HSI) += hsi.o
+obj-y += controllers/
diff --git a/drivers/hsi/controllers/Kconfig b/drivers/hsi/controllers/Kconfig
new file mode 100644
index 0000000..3efe0f0
--- /dev/null
+++ b/drivers/hsi/controllers/Kconfig
@@ -0,0 +1,23 @@
+#
+# HSI controllers configuration
+#
+comment "HSI controllers"
+
+config OMAP_SSI
+ tristate "OMAP SSI hardware driver"
+ depends on ARCH_OMAP && HSI
+ default n
+ ---help---
+ SSI is a legacy version of HSI. It is usually used to connect
+ an application engine with a cellular modem.
+ If you say Y here, you will enable the OMAP SSI hardware driver.
+
+ If unsure, say N.
+
+if OMAP_SSI
+
+config OMAP_SSI_CONFIG
+ boolean
+ default y
+
+endif # OMAP_SSI
diff --git a/drivers/hsi/controllers/Makefile b/drivers/hsi/controllers/Makefile
new file mode 100644
index 0000000..c4ba2c2
--- /dev/null
+++ b/drivers/hsi/controllers/Makefile
@@ -0,0 +1,5 @@
+#
+# Makefile for HSI controllers drivers
+#
+
+obj-$(CONFIG_OMAP_SSI) += omap_ssi.o
--
1.7.1

2010-12-14 10:10:37

by Carlos Chinea

[permalink] [raw]
Subject: [RFC PATCHv4 2/7] HSI: omap_ssi: Introducing OMAP SSI driver

Introduces the OMAP SSI driver in the kernel.

The Synchronous Serial Interface (SSI) is a legacy version
of HSI. As in the case of HSI, it is mainly used to connect
Application engines (APE) with cellular modem engines (CMT)
in cellular handsets.

It provides a multichannel, full-duplex, multi-core communication
with no reference clock. The OMAP SSI block is capable of reaching
speeds of 110 Mbit/s.

Signed-off-by: Carlos Chinea <[email protected]>
---
arch/arm/mach-omap2/ssi.c | 134 +++
arch/arm/plat-omap/include/plat/ssi.h | 204 ++++
drivers/hsi/controllers/omap_ssi.c | 1853 +++++++++++++++++++++++++++++++++
3 files changed, 2191 insertions(+), 0 deletions(-)
create mode 100644 arch/arm/mach-omap2/ssi.c
create mode 100644 arch/arm/plat-omap/include/plat/ssi.h
create mode 100644 drivers/hsi/controllers/omap_ssi.c

diff --git a/arch/arm/mach-omap2/ssi.c b/arch/arm/mach-omap2/ssi.c
new file mode 100644
index 0000000..e822a77
--- /dev/null
+++ b/arch/arm/mach-omap2/ssi.c
@@ -0,0 +1,134 @@
+/*
+ * linux/arch/arm/mach-omap2/ssi.c
+ *
+ * Copyright (C) 2010 Nokia Corporation. All rights reserved.
+ *
+ * Contact: Carlos Chinea <[email protected]>
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public License
+ * 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.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA
+ * 02110-1301 USA
+ */
+
+#include <linux/kernel.h>
+#include <linux/init.h>
+#include <linux/err.h>
+#include <linux/gpio.h>
+#include <linux/platform_device.h>
+#include <plat/omap-pm.h>
+#include <plat/ssi.h>
+
+static struct omap_ssi_platform_data ssi_pdata = {
+ .num_ports = SSI_NUM_PORTS,
+ .get_dev_context_loss_count = omap_pm_get_dev_context_loss_count,
+};
+
+static struct resource ssi_resources[] = {
+ /* SSI controller */
+ [0] = {
+ .start = 0x48058000,
+ .end = 0x48058fff,
+ .name = "omap_ssi_sys",
+ .flags = IORESOURCE_MEM,
+ },
+ /* GDD */
+ [1] = {
+ .start = 0x48059000,
+ .end = 0x48059fff,
+ .name = "omap_ssi_gdd",
+ .flags = IORESOURCE_MEM,
+ },
+ [2] = {
+ .start = 71,
+ .end = 71,
+ .name = "ssi_gdd",
+ .flags = IORESOURCE_IRQ,
+ },
+ /* SSI port 1 */
+ [3] = {
+ .start = 0x4805a000,
+ .end = 0x4805a7ff,
+ .name = "omap_ssi_sst1",
+ .flags = IORESOURCE_MEM,
+ },
+ [4] = {
+ .start = 0x4805a800,
+ .end = 0x4805afff,
+ .name = "omap_ssi_ssr1",
+ .flags = IORESOURCE_MEM,
+ },
+ [5] = {
+ .start = 67,
+ .end = 67,
+ .name = "ssi_p1_mpu_irq0",
+ .flags = IORESOURCE_IRQ,
+ },
+ [6] = {
+ .start = 68,
+ .end = 68,
+ .name = "ssi_p1_mpu_irq1",
+ .flags = IORESOURCE_IRQ,
+ },
+ [7] = {
+ .start = 0,
+ .end = 0,
+ .name = "ssi_p1_cawake",
+ .flags = IORESOURCE_IRQ | IORESOURCE_UNSET,
+ },
+};
+
+static struct platform_device ssi_pdev = {
+ .name = "omap_ssi",
+ .id = 0,
+ .num_resources = ARRAY_SIZE(ssi_resources),
+ .resource = ssi_resources,
+ .dev = {
+ .platform_data = &ssi_pdata,
+ },
+};
+
+int __init omap_ssi_config(struct omap_ssi_board_config *ssi_config)
+{
+ unsigned int port, offset, cawake_gpio;
+ int err;
+
+ ssi_pdata.num_ports = ssi_config->num_ports;
+ for (port = 0, offset = 7; port < ssi_config->num_ports;
+ port++, offset += 5) {
+ cawake_gpio = ssi_config->cawake_gpio[port];
+ if (!cawake_gpio)
+ continue; /* Nothing to do */
+ err = gpio_request(cawake_gpio, "cawake");
+ if (err < 0)
+ goto rback;
+ gpio_direction_input(cawake_gpio);
+ ssi_resources[offset].start = gpio_to_irq(cawake_gpio);
+ ssi_resources[offset].flags &= ~IORESOURCE_UNSET;
+ ssi_resources[offset].flags |= IORESOURCE_IRQ_HIGHEDGE |
+ IORESOURCE_IRQ_LOWEDGE;
+ }
+
+ return 0;
+rback:
+ dev_err(&ssi_pdev.dev, "Request cawake (gpio%d) failed\n", cawake_gpio);
+ while (port > 0)
+ gpio_free(ssi_config->cawake_gpio[--port]);
+
+ return err;
+}
+
+static int __init ssi_init(void)
+{
+ return platform_device_register(&ssi_pdev);
+}
+subsys_initcall(ssi_init);
diff --git a/arch/arm/plat-omap/include/plat/ssi.h b/arch/arm/plat-omap/include/plat/ssi.h
new file mode 100644
index 0000000..eb84c3a
--- /dev/null
+++ b/arch/arm/plat-omap/include/plat/ssi.h
@@ -0,0 +1,204 @@
+/*
+ * plat/ssi.h
+ *
+ * Hardware definitions for SSI.
+ *
+ * Copyright (C) 2010 Nokia Corporation. All rights reserved.
+ *
+ * Contact: Carlos Chinea <[email protected]>
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public License
+ * 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.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA
+ * 02110-1301 USA
+ */
+
+#ifndef __OMAP_SSI_REGS_H__
+#define __OMAP_SSI_REGS_H__
+
+#define SSI_NUM_PORTS 1
+/*
+ * SSI SYS registers
+ */
+#define SSI_REVISION_REG 0
+# define SSI_REV_MAJOR 0xf0
+# define SSI_REV_MINOR 0xf
+#define SSI_SYSCONFIG_REG 0x10
+# define SSI_AUTOIDLE (1 << 0)
+# define SSI_SOFTRESET (1 << 1)
+# define SSI_SIDLEMODE_FORCE 0
+# define SSI_SIDLEMODE_NO (1 << 3)
+# define SSI_SIDLEMODE_SMART (1 << 4)
+# define SSI_SIDLEMODE_MASK 0x18
+# define SSI_MIDLEMODE_FORCE 0
+# define SSI_MIDLEMODE_NO (1 << 12)
+# define SSI_MIDLEMODE_SMART (1 << 13)
+# define SSI_MIDLEMODE_MASK 0x3000
+#define SSI_SYSSTATUS_REG 0x14
+# define SSI_RESETDONE 1
+#define SSI_MPU_STATUS_REG(port, irq) (0x808 + ((port) * 0x10) + ((irq) * 2))
+#define SSI_MPU_ENABLE_REG(port, irq) (0x80c + ((port) * 0x10) + ((irq) * 8))
+# define SSI_DATAACCEPT(channel) (1 << (channel))
+# define SSI_DATAAVAILABLE(channel) (1 << ((channel) + 8))
+# define SSI_DATAOVERRUN(channel) (1 << ((channel) + 16))
+# define SSI_ERROROCCURED (1 << 24)
+# define SSI_BREAKDETECTED (1 << 25)
+#define SSI_GDD_MPU_IRQ_STATUS_REG 0x0800
+#define SSI_GDD_MPU_IRQ_ENABLE_REG 0x0804
+# define SSI_GDD_LCH(channel) (1 << (channel))
+#define SSI_WAKE_REG(port) (0xc00 + ((port) * 0x10))
+#define SSI_CLEAR_WAKE_REG(port) (0xc04 + ((port) * 0x10))
+#define SSI_SET_WAKE_REG(port) (0xc08 + ((port) * 0x10))
+# define SSI_WAKE(channel) (1 << (channel))
+# define SSI_WAKE_MASK 0xff
+
+/*
+ * SSI SST registers
+ */
+#define SSI_SST_ID_REG 0
+#define SSI_SST_MODE_REG 4
+# define SSI_MODE_VAL_MASK 3
+# define SSI_MODE_SLEEP 0
+# define SSI_MODE_STREAM 1
+# define SSI_MODE_FRAME 2
+# define SSI_MODE_MULTIPOINTS 3
+#define SSI_SST_FRAMESIZE_REG 8
+# define SSI_FRAMESIZE_DEFAULT 31
+#define SSI_SST_TXSTATE_REG 0xc
+# define SSI_TXSTATE_IDLE 0
+#define SSI_SST_BUFSTATE_REG 0x10
+# define SSI_FULL(channel) (1 << (channel))
+#define SSI_SST_DIVISOR_REG 0x18
+# define SSI_MAX_DIVISOR 127
+#define SSI_SST_BREAK_REG 0x20
+#define SSI_SST_CHANNELS_REG 0x24
+# define SSI_CHANNELS_DEFAULT 4
+#define SSI_SST_ARBMODE_REG 0x28
+# define SSI_ARBMODE_ROUNDROBIN 0
+# define SSI_ARBMODE_PRIORITY 1
+#define SSI_SST_BUFFER_CH_REG(channel) (0x80 + ((channel) * 4))
+#define SSI_SST_SWAPBUF_CH_REG(channel) (0xc0 + ((channel) * 4))
+
+/*
+ * SSI SSR registers
+ */
+#define SSI_SSR_ID_REG 0
+#define SSI_SSR_MODE_REG 4
+#define SSI_SSR_FRAMESIZE_REG 8
+#define SSI_SSR_RXSTATE_REG 0xc
+#define SSI_SSR_BUFSTATE_REG 0x10
+# define SSI_NOTEMPTY(channel) (1 << (channel))
+#define SSI_SSR_BREAK_REG 0x1c
+#define SSI_SSR_ERROR_REG 0x20
+#define SSI_SSR_ERRORACK_REG 0x24
+#define SSI_SSR_OVERRUN_REG 0x2c
+#define SSI_SSR_OVERRUNACK_REG 0x30
+#define SSI_SSR_TIMEOUT_REG 0x34
+# define SSI_TIMEOUT_DEFAULT 0
+#define SSI_SSR_CHANNELS_REG 0x28
+#define SSI_SSR_BUFFER_CH_REG(channel) (0x80 + ((channel) * 4))
+#define SSI_SSR_SWAPBUF_CH_REG(channel) (0xc0 + ((channel) * 4))
+
+/*
+ * SSI GDD registers
+ */
+#define SSI_GDD_HW_ID_REG 0
+#define SSI_GDD_PPORT_ID_REG 0x10
+#define SSI_GDD_MPORT_ID_REG 0x14
+#define SSI_GDD_PPORT_SR_REG 0x20
+#define SSI_GDD_MPORT_SR_REG 0x24
+# define SSI_ACTIVE_LCH_NUM_MASK 0xff
+#define SSI_GDD_TEST_REG 0x40
+# define SSI_TEST 1
+#define SSI_GDD_GCR_REG 0x100
+# define SSI_CLK_AUTOGATING_ON (1 << 3)
+# define SSI_FREE (1 << 2)
+# define SSI_SWITCH_OFF (1 << 0)
+#define SSI_GDD_GRST_REG 0x200
+# define SSI_SWRESET 1
+#define SSI_GDD_CSDP_REG(channel) (0x800 + ((channel) * 0x40))
+# define SSI_DST_BURST_EN_MASK 0xc000
+# define SSI_DST_SINGLE_ACCESS0 0
+# define SSI_DST_SINGLE_ACCESS (1 << 14)
+# define SSI_DST_BURST_4x32_BIT (2 << 14)
+# define SSI_DST_BURST_8x32_BIT (3 << 14)
+# define SSI_DST_MASK 0x1e00
+# define SSI_DST_MEMORY_PORT (8 << 9)
+# define SSI_DST_PERIPHERAL_PORT (9 << 9)
+# define SSI_SRC_BURST_EN_MASK 0x180
+# define SSI_SRC_SINGLE_ACCESS0 0
+# define SSI_SRC_SINGLE_ACCESS (1 << 7)
+# define SSI_SRC_BURST_4x32_BIT (2 << 7)
+# define SSI_SRC_BURST_8x32_BIT (3 << 7)
+# define SSI_SRC_MASK 0x3c
+# define SSI_SRC_MEMORY_PORT (8 << 2)
+# define SSI_SRC_PERIPHERAL_PORT (9 << 2)
+# define SSI_DATA_TYPE_MASK 3
+# define SSI_DATA_TYPE_S32 2
+#define SSI_GDD_CCR_REG(channel) (0x802 + ((channel) * 0x40))
+# define SSI_DST_AMODE_MASK (3 << 14)
+# define SSI_DST_AMODE_CONST 0
+# define SSI_DST_AMODE_POSTINC (1 << 12)
+# define SSI_SRC_AMODE_MASK (3 << 12)
+# define SSI_SRC_AMODE_CONST 0
+# define SSI_SRC_AMODE_POSTINC (1 << 12)
+# define SSI_CCR_ENABLE (1 << 7)
+# define SSI_CCR_SYNC_MASK 0x1f
+#define SSI_GDD_CICR_REG(channel) (0x804 + ((channel) * 0x40))
+# define SSI_BLOCK_IE (1 << 5)
+# define SSI_HALF_IE (1 << 2)
+# define SSI_TOUT_IE (1 << 0)
+#define SSI_GDD_CSR_REG(channel) (0x806 + ((channel) * 0x40))
+# define SSI_CSR_SYNC (1 << 6)
+# define SSI_CSR_BLOCK (1 << 5)
+# define SSI_CSR_HALF (1 << 2)
+# define SSI_CSR_TOUR (1 << 0)
+#define SSI_GDD_CSSA_REG(channel) (0x808 + ((channel) * 0x40))
+#define SSI_GDD_CDSA_REG(channel) (0x80c + ((channel) * 0x40))
+#define SSI_GDD_CEN_REG(channel) (0x810 + ((channel) * 0x40))
+#define SSI_GDD_CSAC_REG(channel) (0x818 + ((channel) * 0x40))
+#define SSI_GDD_CDAC_REG(channel) (0x81a + ((channel) * 0x40))
+#define SSI_GDD_CLNK_CTRL_REG(channel) (0x828 + ((channel) * 0x40))
+# define SSI_ENABLE_LNK (1 << 15)
+# define SSI_STOP_LNK (1 << 14)
+# define SSI_NEXT_CH_ID_MASK 0xf
+
+/**
+ * struct omap_ssi_platform_data - OMAP SSI platform data
+ * @num_ports: Number of ports on the controller
+ * @ctxt_loss_count: Pointer to omap_pm_get_dev_context_loss_count
+ */
+struct omap_ssi_platform_data {
+ unsigned int num_ports;
+ int (*get_dev_context_loss_count)(struct device *dev);
+};
+
+/**
+ * struct omap_ssi_config - SSI board configuration
+ * @num_ports: Number of ports in use
+ * @cawake_line: Array of cawake gpio lines
+ */
+struct omap_ssi_board_config {
+ unsigned int num_ports;
+ int cawake_gpio[SSI_NUM_PORTS];
+};
+
+#ifdef CONFIG_OMAP_SSI_CONFIG
+extern int omap_ssi_config(struct omap_ssi_board_config *ssi_config);
+#else
+static inline int omap_ssi_config(struct omap_ssi_board_config *ssi_config)
+{
+ return 0;
+}
+#endif /* CONFIG_OMAP_SSI_CONFIG */
+
+#endif /* __OMAP_SSI_REGS_H__ */
diff --git a/drivers/hsi/controllers/omap_ssi.c b/drivers/hsi/controllers/omap_ssi.c
new file mode 100644
index 0000000..a82ea0e
--- /dev/null
+++ b/drivers/hsi/controllers/omap_ssi.c
@@ -0,0 +1,1853 @@
+/*
+ * omap_ssi.c
+ *
+ * Implements the OMAP SSI driver.
+ *
+ * Copyright (C) 2010 Nokia Corporation. All rights reserved.
+ *
+ * Contact: Carlos Chinea <[email protected]>
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public License
+ * 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.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA
+ * 02110-1301 USA
+ */
+#include <linux/compiler.h>
+#include <linux/err.h>
+#include <linux/ioport.h>
+#include <linux/io.h>
+#include <linux/gpio.h>
+#include <linux/clk.h>
+#include <linux/device.h>
+#include <linux/platform_device.h>
+#include <linux/dma-mapping.h>
+#include <linux/delay.h>
+#include <linux/seq_file.h>
+#include <linux/scatterlist.h>
+#include <linux/interrupt.h>
+#include <linux/spinlock.h>
+#include <linux/hsi/hsi.h>
+#include <linux/debugfs.h>
+#include <plat/omap-pm.h>
+#include <plat/clock.h>
+#include <plat/ssi.h>
+
+#define SSI_MAX_CHANNELS 8
+#define SSI_MAX_GDD_LCH 8
+#define SSI_BYTES_TO_FRAMES(x) ((((x) - 1) >> 2) + 1)
+
+/**
+ * struct ssi_clk_res - Device resource data for the SSI clocks
+ * @clk: Pointer to the clock
+ * @nb: Pointer to the clock notifier for clk, if any
+ */
+struct ssi_clk_res {
+ struct clk *clk;
+ struct notifier_block *nb;
+};
+
+/**
+ * struct gdd_trn - GDD transaction data
+ * @msg: Pointer to the HSI message being served
+ * @sg: Pointer to the current sg entry being served
+ */
+struct gdd_trn {
+ struct hsi_msg *msg;
+ struct scatterlist *sg;
+};
+
+/**
+ * struct omap_ssm_ctx - OMAP synchronous serial module (TX/RX) context
+ * @mode: Bit transmission mode
+ * @channels: Number of channels
+ * @framesize: Frame size in bits
+ * @timeout: RX frame timeout
+ * @divisor: TX divider
+ * @arb_mode: Arbitration mode for TX frame (Round robin, priority)
+ */
+struct omap_ssm_ctx {
+ u32 mode;
+ u32 channels;
+ u32 frame_size;
+ union {
+ u32 timeout; /* Rx Only */
+ struct {
+ u32 arb_mode;
+ u32 divisor;
+ }; /* Tx only */
+ };
+};
+
+/**
+ * struct omap_ssi_port - OMAP SSI port data
+ * @dev: device associated to the port (HSI port)
+ * @sst_dma: SSI transmitter physical base address
+ * @ssr_dma: SSI receiver physical base address
+ * @sst_base: SSI transmitter base address
+ * @ssr_base: SSI receiver base address
+ * @wk_lock: spin lock to serialize access to the wake lines
+ * @lock: Spin lock to serialize access to the SSI port
+ * @channels: Current number of channels configured (1,2,4 or 8)
+ * @txqueue: TX message queues
+ * @rxqueue: RX message queues
+ * @brkqueue: Queue of incoming HWBREAK requests (FRAME mode)
+ * @irq: IRQ number
+ * @wake_irq: IRQ number for incoming wake line (-1 if none)
+ * @pio_tasklet: Bottom half for PIO transfers and events
+ * @wake_tasklet: Bottom half for incoming wake events
+ * @wkin_cken: Keep track of clock references due to the incoming wake line
+ * @wk_refcount: Reference count for output wake line
+ * @sys_mpu_enable: Context for the interrupt enable register for irq 0
+ * @sst: Context for the synchronous serial transmitter
+ * @ssr: Context for the synchronous serial receiver
+ */
+struct omap_ssi_port {
+ struct device *dev;
+ dma_addr_t sst_dma;
+ dma_addr_t ssr_dma;
+ void __iomem *sst_base;
+ void __iomem *ssr_base;
+ spinlock_t wk_lock;
+ spinlock_t lock;
+ unsigned int channels;
+ struct list_head txqueue[SSI_MAX_CHANNELS];
+ struct list_head rxqueue[SSI_MAX_CHANNELS];
+ struct list_head brkqueue;
+ unsigned int irq;
+ int wake_irq;
+ struct tasklet_struct pio_tasklet;
+ struct tasklet_struct wake_tasklet;
+ unsigned int wkin_cken:1; /* Workaround */
+ int wk_refcount;
+ /* OMAP SSI port context */
+ u32 sys_mpu_enable; /* We use only one irq */
+ struct omap_ssm_ctx sst;
+ struct omap_ssm_ctx ssr;
+};
+
+/**
+ * struct omap_ssi_controller - OMAP SSI controller data
+ * @dev: device associated to the controller (HSI controller)
+ * @sys: SSI I/O base address
+ * @gdd: GDD I/O base address
+ * @ick: SSI interconnect clock
+ * @fck: SSI functional clock
+ * @ck_refcount: References count for clocks
+ * @gdd_irq: IRQ line for GDD
+ * @gdd_tasklet: bottom half for DMA transfers
+ * @gdd_trn: Array of GDD transaction data for ongoing GDD transfers
+ * @lock: lock to serialize access to GDD
+ * @ck_lock: lock to serialize access to the clocks
+ * @loss_count: To follow if we need to restore context or not
+ * @max_speed: Maximum TX speed (Kb/s) set by the clients.
+ * @sysconfig: SSI controller saved context
+ * @gdd_gcr: SSI GDD saved context
+ * @get_loss: Pointer to omap_pm_get_dev_context_loss_count, if any
+ * @port: Array of pointers of the ports of the controller
+ * @dir: Debugfs SSI root directory
+ */
+struct omap_ssi_controller {
+ struct device *dev;
+ void __iomem *sys;
+ void __iomem *gdd;
+ struct clk *ick;
+ struct clk *fck;
+ int ck_refcount;
+ unsigned int gdd_irq;
+ struct tasklet_struct gdd_tasklet;
+ struct gdd_trn gdd_trn[SSI_MAX_GDD_LCH];
+ spinlock_t lock;
+ spinlock_t ck_lock;
+ unsigned long fck_rate;
+ int loss_count;
+ u32 max_speed;
+ /* OMAP SSI Controller context */
+ u32 sysconfig;
+ u32 gdd_gcr;
+ int (*get_loss)(struct device *dev);
+ struct omap_ssi_port **port;
+#ifdef CONFIG_DEBUG_FS
+ struct dentry *dir;
+#endif
+};
+
+static inline unsigned int ssi_wakein(struct hsi_port *port)
+{
+ struct omap_ssi_port *omap_port = hsi_port_drvdata(port);
+
+ return gpio_get_value(irq_to_gpio(omap_port->wake_irq));
+}
+
+static int ssi_for_each_port(struct hsi_controller *ssi, void *data,
+ int (*fn)(struct omap_ssi_port *p, void *data))
+{
+ struct omap_ssi_controller *omap_ssi = hsi_controller_drvdata(ssi);
+ unsigned int i = 0;
+ int err = 0;
+
+ for (i = 0; ((i < ssi->num_ports) && !err); i++)
+ err = (*fn)(omap_ssi->port[i], data);
+
+ return err;
+}
+
+static int ssi_set_port_mode(struct omap_ssi_port *omap_port, void *data)
+{
+ u32 *mode = data;
+
+ __raw_writel(*mode, omap_port->sst_base + SSI_SST_MODE_REG);
+ __raw_writel(*mode, omap_port->ssr_base + SSI_SSR_MODE_REG);
+ /* OCP barrier */
+ *mode = __raw_readl(omap_port->ssr_base + SSI_SSR_MODE_REG);
+
+ return 0;
+}
+
+static inline void ssi_set_mode(struct hsi_controller *ssi, u32 mode)
+{
+ ssi_for_each_port(ssi, &mode, ssi_set_port_mode);
+}
+
+static int ssi_restore_port_mode(struct omap_ssi_port *omap_port,
+ void *data __maybe_unused)
+{
+ u32 mode;
+
+ __raw_writel(omap_port->sst.mode,
+ omap_port->sst_base + SSI_SST_MODE_REG);
+ __raw_writel(omap_port->ssr.mode,
+ omap_port->ssr_base + SSI_SSR_MODE_REG);
+ /* OCP barrier */
+ mode = __raw_readl(omap_port->ssr_base + SSI_SSR_MODE_REG);
+
+ return 0;
+}
+
+static int ssi_restore_divisor(struct omap_ssi_port *omap_port,
+ void *data __maybe_unused)
+{
+ __raw_writel(omap_port->sst.divisor,
+ omap_port->sst_base + SSI_SST_DIVISOR_REG);
+
+ return 0;
+}
+
+static int ssi_restore_port_ctx(struct omap_ssi_port *omap_port,
+ void *data __maybe_unused)
+{
+ struct hsi_port *port = to_hsi_port(omap_port->dev);
+ struct hsi_controller *ssi = to_hsi_controller(port->device.parent);
+ struct omap_ssi_controller *omap_ssi = hsi_controller_drvdata(ssi);
+ void __iomem *base = omap_port->sst_base;
+
+ __raw_writel(omap_port->sys_mpu_enable,
+ omap_ssi->sys + SSI_MPU_ENABLE_REG(port->num, 0));
+ /* SST context */
+ __raw_writel(omap_port->sst.frame_size, base + SSI_SST_FRAMESIZE_REG);
+ __raw_writel(omap_port->sst.channels, base + SSI_SST_CHANNELS_REG);
+ __raw_writel(omap_port->sst.arb_mode, base + SSI_SST_ARBMODE_REG);
+ /* SSR context */
+ base = omap_port->ssr_base;
+ __raw_writel(omap_port->ssr.frame_size, base + SSI_SSR_FRAMESIZE_REG);
+ __raw_writel(omap_port->ssr.channels, base + SSI_SSR_CHANNELS_REG);
+ __raw_writel(omap_port->ssr.timeout, base + SSI_SSR_TIMEOUT_REG);
+
+ return 0;
+}
+
+static int ssi_save_port_ctx(struct omap_ssi_port *omap_port,
+ void *data __maybe_unused)
+{
+ struct hsi_port *port = to_hsi_port(omap_port->dev);
+ struct hsi_controller *ssi = to_hsi_controller(port->device.parent);
+ struct omap_ssi_controller *omap_ssi = hsi_controller_drvdata(ssi);
+
+ omap_port->sys_mpu_enable = __raw_readl(omap_ssi->sys +
+ SSI_MPU_ENABLE_REG(port->num, 0));
+
+ return 0;
+}
+
+static int ssi_clk_enable(struct hsi_controller *ssi)
+{
+ struct omap_ssi_controller *omap_ssi = hsi_controller_drvdata(ssi);
+ int err = 0;
+
+ spin_lock_bh(&omap_ssi->ck_lock);
+ if (omap_ssi->ck_refcount++)
+ goto out;
+ err = clk_enable(omap_ssi->fck);
+ if (unlikely(err < 0))
+ goto out;
+ err = clk_enable(omap_ssi->ick);
+ if (unlikely(err < 0)) {
+ clk_disable(omap_ssi->fck);
+ goto out;
+ }
+ if ((omap_ssi->get_loss) && (omap_ssi->loss_count ==
+ (*omap_ssi->get_loss)(ssi->device.parent)))
+ goto mode; /* We always need to restore the mode & TX divisor */
+
+ __raw_writel(omap_ssi->sysconfig, omap_ssi->sys + SSI_SYSCONFIG_REG);
+ __raw_writel(omap_ssi->gdd_gcr, omap_ssi->gdd + SSI_GDD_GCR_REG);
+
+ ssi_for_each_port(ssi, NULL, ssi_restore_port_ctx);
+mode:
+ ssi_for_each_port(ssi, NULL, ssi_restore_divisor);
+ ssi_for_each_port(ssi, NULL, ssi_restore_port_mode);
+out:
+ spin_unlock_bh(&omap_ssi->ck_lock);
+
+ return err;
+}
+
+static void ssi_clk_disable(struct hsi_controller *ssi)
+{
+ struct omap_ssi_controller *omap_ssi = hsi_controller_drvdata(ssi);
+
+ spin_lock_bh(&omap_ssi->ck_lock);
+ WARN_ON(omap_ssi->ck_refcount <= 0);
+ if (--omap_ssi->ck_refcount)
+ goto out;
+
+ ssi_set_mode(ssi, SSI_MODE_SLEEP);
+
+ if (omap_ssi->get_loss)
+ omap_ssi->loss_count =
+ (*omap_ssi->get_loss)(ssi->device.parent);
+
+ ssi_for_each_port(ssi, NULL, ssi_save_port_ctx);
+ clk_disable(omap_ssi->ick);
+ clk_disable(omap_ssi->fck);
+out:
+ spin_unlock_bh(&omap_ssi->ck_lock);
+}
+
+#ifdef CONFIG_DEBUG_FS
+static int ssi_debug_show(struct seq_file *m, void *p __maybe_unused)
+{
+ struct hsi_controller *ssi = m->private;
+ struct omap_ssi_controller *omap_ssi = hsi_controller_drvdata(ssi);
+ void __iomem *sys = omap_ssi->sys;
+
+ ssi_clk_enable(ssi);
+ seq_printf(m, "REVISION\t: 0x%08x\n",
+ __raw_readl(sys + SSI_REVISION_REG));
+ seq_printf(m, "SYSCONFIG\t: 0x%08x\n",
+ __raw_readl(sys + SSI_SYSCONFIG_REG));
+ seq_printf(m, "SYSSTATUS\t: 0x%08x\n",
+ __raw_readl(sys + SSI_SYSSTATUS_REG));
+ ssi_clk_disable(ssi);
+
+ return 0;
+}
+
+static int ssi_debug_port_show(struct seq_file *m, void *p __maybe_unused)
+{
+ struct hsi_port *port = m->private;
+ struct omap_ssi_port *omap_port = hsi_port_drvdata(port);
+ struct hsi_controller *ssi = to_hsi_controller(port->device.parent);
+ struct omap_ssi_controller *omap_ssi = hsi_controller_drvdata(ssi);
+ void __iomem *base = omap_ssi->sys;
+ unsigned int ch;
+
+ ssi_clk_enable(ssi);
+ if (omap_port->wake_irq > 0)
+ seq_printf(m, "CAWAKE\t\t: %d\n", ssi_wakein(port));
+ seq_printf(m, "WAKE\t\t: 0x%08x\n",
+ __raw_readl(base + SSI_WAKE_REG(port->num)));
+ seq_printf(m, "MPU_ENABLE_IRQ%d\t: 0x%08x\n", 0,
+ __raw_readl(base + SSI_MPU_ENABLE_REG(port->num, 0)));
+ seq_printf(m, "MPU_STATUS_IRQ%d\t: 0x%08x\n", 0,
+ __raw_readl(base + SSI_MPU_STATUS_REG(port->num, 0)));
+ /* SST */
+ base = omap_port->sst_base;
+ seq_printf(m, "\nSST\n===\n");
+ seq_printf(m, "ID SST\t\t: 0x%08x\n",
+ __raw_readl(base + SSI_SST_ID_REG));
+ seq_printf(m, "MODE\t\t: 0x%08x\n",
+ __raw_readl(base + SSI_SST_MODE_REG));
+ seq_printf(m, "FRAMESIZE\t: 0x%08x\n",
+ __raw_readl(base + SSI_SST_FRAMESIZE_REG));
+ seq_printf(m, "DIVISOR\t\t: 0x%08x\n",
+ __raw_readl(base + SSI_SST_DIVISOR_REG));
+ seq_printf(m, "CHANNELS\t: 0x%08x\n",
+ __raw_readl(base + SSI_SST_CHANNELS_REG));
+ seq_printf(m, "ARBMODE\t\t: 0x%08x\n",
+ __raw_readl(base + SSI_SST_ARBMODE_REG));
+ seq_printf(m, "TXSTATE\t\t: 0x%08x\n",
+ __raw_readl(base + SSI_SST_TXSTATE_REG));
+ seq_printf(m, "BUFSTATE\t: 0x%08x\n",
+ __raw_readl(base + SSI_SST_BUFSTATE_REG));
+ seq_printf(m, "BREAK\t\t: 0x%08x\n",
+ __raw_readl(base + SSI_SST_BREAK_REG));
+ for (ch = 0; ch < omap_port->channels; ch++) {
+ seq_printf(m, "BUFFER_CH%d\t: 0x%08x\n", ch,
+ __raw_readl(base + SSI_SST_BUFFER_CH_REG(ch)));
+ }
+ /* SSR */
+ base = omap_port->ssr_base;
+ seq_printf(m, "\nSSR\n===\n");
+ seq_printf(m, "ID SSR\t\t: 0x%08x\n",
+ __raw_readl(base + SSI_SSR_ID_REG));
+ seq_printf(m, "MODE\t\t: 0x%08x\n",
+ __raw_readl(base + SSI_SSR_MODE_REG));
+ seq_printf(m, "FRAMESIZE\t: 0x%08x\n",
+ __raw_readl(base + SSI_SSR_FRAMESIZE_REG));
+ seq_printf(m, "CHANNELS\t: 0x%08x\n",
+ __raw_readl(base + SSI_SSR_CHANNELS_REG));
+ seq_printf(m, "TIMEOUT\t\t: 0x%08x\n",
+ __raw_readl(base + SSI_SSR_TIMEOUT_REG));
+ seq_printf(m, "RXSTATE\t\t: 0x%08x\n",
+ __raw_readl(base + SSI_SSR_RXSTATE_REG));
+ seq_printf(m, "BUFSTATE\t: 0x%08x\n",
+ __raw_readl(base + SSI_SSR_BUFSTATE_REG));
+ seq_printf(m, "BREAK\t\t: 0x%08x\n",
+ __raw_readl(base + SSI_SSR_BREAK_REG));
+ seq_printf(m, "ERROR\t\t: 0x%08x\n",
+ __raw_readl(base + SSI_SSR_ERROR_REG));
+ seq_printf(m, "ERRORACK\t: 0x%08x\n",
+ __raw_readl(base + SSI_SSR_ERRORACK_REG));
+ for (ch = 0; ch < omap_port->channels; ch++) {
+ seq_printf(m, "BUFFER_CH%d\t: 0x%08x\n", ch,
+ __raw_readl(base + SSI_SSR_BUFFER_CH_REG(ch)));
+ }
+ ssi_clk_disable(ssi);
+
+ return 0;
+}
+
+static int ssi_debug_gdd_show(struct seq_file *m, void *p __maybe_unused)
+{
+ struct hsi_controller *ssi = m->private;
+ struct omap_ssi_controller *omap_ssi = hsi_controller_drvdata(ssi);
+ void __iomem *gdd = omap_ssi->gdd;
+ int lch;
+
+ ssi_clk_enable(ssi);
+ seq_printf(m, "GDD_MPU_STATUS\t: 0x%08x\n",
+ __raw_readl(omap_ssi->sys + SSI_GDD_MPU_IRQ_STATUS_REG));
+ seq_printf(m, "GDD_MPU_ENABLE\t: 0x%08x\n\n",
+ __raw_readl(omap_ssi->sys + SSI_GDD_MPU_IRQ_ENABLE_REG));
+ seq_printf(m, "HW_ID\t\t: 0x%08x\n",
+ __raw_readl(gdd + SSI_GDD_HW_ID_REG));
+ seq_printf(m, "PPORT_ID\t: 0x%08x\n",
+ __raw_readl(gdd + SSI_GDD_PPORT_ID_REG));
+ seq_printf(m, "MPORT_ID\t: 0x%08x\n",
+ __raw_readl(gdd + SSI_GDD_MPORT_ID_REG));
+ seq_printf(m, "TEST\t\t: 0x%08x\n",
+ __raw_readl(gdd + SSI_GDD_TEST_REG));
+ seq_printf(m, "GCR\t\t: 0x%08x\n",
+ __raw_readl(gdd + SSI_GDD_GCR_REG));
+
+ for (lch = 0; lch < SSI_MAX_GDD_LCH; lch++) {
+ seq_printf(m, "\nGDD LCH %d\n=========\n", lch);
+ seq_printf(m, "CSDP\t\t: 0x%04x\n",
+ __raw_readw(gdd + SSI_GDD_CSDP_REG(lch)));
+ seq_printf(m, "CCR\t\t: 0x%04x\n",
+ __raw_readw(gdd + SSI_GDD_CCR_REG(lch)));
+ seq_printf(m, "CICR\t\t: 0x%04x\n",
+ __raw_readw(gdd + SSI_GDD_CICR_REG(lch)));
+ seq_printf(m, "CSR\t\t: 0x%04x\n",
+ __raw_readw(gdd + SSI_GDD_CSR_REG(lch)));
+ seq_printf(m, "CSSA\t\t: 0x%08x\n",
+ __raw_readl(gdd + SSI_GDD_CSSA_REG(lch)));
+ seq_printf(m, "CDSA\t\t: 0x%08x\n",
+ __raw_readl(gdd + SSI_GDD_CDSA_REG(lch)));
+ seq_printf(m, "CEN\t\t: 0x%04x\n",
+ __raw_readw(gdd + SSI_GDD_CEN_REG(lch)));
+ seq_printf(m, "CSAC\t\t: 0x%04x\n",
+ __raw_readw(gdd + SSI_GDD_CSAC_REG(lch)));
+ seq_printf(m, "CDAC\t\t: 0x%04x\n",
+ __raw_readw(gdd + SSI_GDD_CDAC_REG(lch)));
+ seq_printf(m, "CLNK_CTRL\t: 0x%04x\n",
+ __raw_readw(gdd + SSI_GDD_CLNK_CTRL_REG(lch)));
+ }
+ ssi_clk_disable(ssi);
+
+ return 0;
+}
+
+static int ssi_div_get(void *data, u64 *val)
+{
+ struct hsi_port *port = data;
+ struct omap_ssi_port *omap_port = hsi_port_drvdata(port);
+ struct hsi_controller *ssi = to_hsi_controller(port->device.parent);
+
+ ssi_clk_enable(ssi);
+ *val = __raw_readl(omap_port->sst_base + SSI_SST_DIVISOR_REG);
+ ssi_clk_disable(ssi);
+
+ return 0;
+}
+
+static int ssi_div_set(void *data, u64 val)
+{
+ struct hsi_port *port = data;
+ struct omap_ssi_port *omap_port = hsi_port_drvdata(port);
+ struct hsi_controller *ssi = to_hsi_controller(port->device.parent);
+
+ if (val > 127)
+ return -EINVAL;
+
+ ssi_clk_enable(ssi);
+ __raw_writel(val, omap_port->sst_base + SSI_SST_DIVISOR_REG);
+ omap_port->sst.divisor = val;
+ ssi_clk_disable(ssi);
+
+ return 0;
+}
+
+static int ssi_regs_open(struct inode *inode, struct file *file)
+{
+ return single_open(file, ssi_debug_show, inode->i_private);
+}
+
+static int ssi_port_regs_open(struct inode *inode, struct file *file)
+{
+ return single_open(file, ssi_debug_port_show, inode->i_private);
+}
+
+static int ssi_gdd_regs_open(struct inode *inode, struct file *file)
+{
+ return single_open(file, ssi_debug_gdd_show, inode->i_private);
+}
+
+static const struct file_operations ssi_regs_fops = {
+ .open = ssi_regs_open,
+ .read = seq_read,
+ .llseek = seq_lseek,
+ .release = single_release,
+};
+
+static const struct file_operations ssi_port_regs_fops = {
+ .open = ssi_port_regs_open,
+ .read = seq_read,
+ .llseek = seq_lseek,
+ .release = single_release,
+};
+
+static const struct file_operations ssi_gdd_regs_fops = {
+ .open = ssi_gdd_regs_open,
+ .read = seq_read,
+ .llseek = seq_lseek,
+ .release = single_release,
+};
+
+DEFINE_SIMPLE_ATTRIBUTE(ssi_sst_div_fops, ssi_div_get, ssi_div_set, "%llu\n");
+
+static int __init ssi_debug_add_port(struct omap_ssi_port *omap_port,
+ void *data)
+{
+ struct hsi_port *port = to_hsi_port(omap_port->dev);
+ struct dentry *dir = data;
+
+ dir = debugfs_create_dir(dev_name(omap_port->dev), dir);
+ if (IS_ERR(dir))
+ return PTR_ERR(dir);
+ debugfs_create_file("regs", S_IRUGO, dir, port, &ssi_port_regs_fops);
+ dir = debugfs_create_dir("sst", dir);
+ if (IS_ERR(dir))
+ return PTR_ERR(dir);
+ debugfs_create_file("divisor", S_IRUGO | S_IWUSR, dir, port,
+ &ssi_sst_div_fops);
+
+ return 0;
+}
+
+static int __init ssi_debug_add_ctrl(struct hsi_controller *ssi)
+{
+ struct omap_ssi_controller *omap_ssi = hsi_controller_drvdata(ssi);
+ struct dentry *dir;
+ int err;
+
+ /* SSI controller */
+ omap_ssi->dir = debugfs_create_dir(dev_name(&ssi->device), NULL);
+ if (IS_ERR(omap_ssi->dir))
+ return PTR_ERR(omap_ssi->dir);
+
+ debugfs_create_file("regs", S_IRUGO, omap_ssi->dir, ssi,
+ &ssi_regs_fops);
+ /* SSI GDD (DMA) */
+ dir = debugfs_create_dir("gdd", omap_ssi->dir);
+ if (IS_ERR(dir))
+ goto rback;
+ debugfs_create_file("regs", S_IRUGO, dir, ssi, &ssi_gdd_regs_fops);
+ /* SSI ports */
+ err = ssi_for_each_port(ssi, omap_ssi->dir, ssi_debug_add_port);
+ if (err < 0)
+ goto rback;
+
+ return 0;
+rback:
+ debugfs_remove_recursive(omap_ssi->dir);
+
+ return PTR_ERR(dir);
+}
+
+static void ssi_debug_remove_ctrl(struct hsi_controller *ssi)
+{
+ struct omap_ssi_controller *omap_ssi = hsi_controller_drvdata(ssi);
+
+ debugfs_remove_recursive(omap_ssi->dir);
+}
+#endif /* CONFIG_DEBUG_FS */
+
+static int ssi_claim_lch(struct hsi_msg *msg)
+{
+
+ struct hsi_port *port = hsi_get_port(msg->cl);
+ struct hsi_controller *ssi = to_hsi_controller(port->device.parent);
+ struct omap_ssi_controller *omap_ssi = hsi_controller_drvdata(ssi);
+ int lch;
+
+ for (lch = 0; lch < SSI_MAX_GDD_LCH; lch++)
+ if (!omap_ssi->gdd_trn[lch].msg) {
+ omap_ssi->gdd_trn[lch].msg = msg;
+ omap_ssi->gdd_trn[lch].sg = msg->sgt.sgl;
+ return lch;
+ }
+
+ return -EBUSY;
+}
+
+static int ssi_start_pio(struct hsi_msg *msg)
+{
+ struct hsi_port *port = hsi_get_port(msg->cl);
+ struct hsi_controller *ssi = to_hsi_controller(port->device.parent);
+ struct omap_ssi_controller *omap_ssi = hsi_controller_drvdata(ssi);
+ u32 val;
+
+ ssi_clk_enable(ssi);
+ if (msg->ttype == HSI_MSG_WRITE) {
+ val = SSI_DATAACCEPT(msg->channel);
+ ssi_clk_enable(ssi); /* Hold clocks for pio writes */
+ } else {
+ val = SSI_DATAAVAILABLE(msg->channel) | SSI_ERROROCCURED;
+ }
+ dev_dbg(&port->device, "Single %s transfer\n",
+ msg->ttype ? "write" : "read");
+ val |= __raw_readl(omap_ssi->sys + SSI_MPU_ENABLE_REG(port->num, 0));
+ __raw_writel(val, omap_ssi->sys + SSI_MPU_ENABLE_REG(port->num, 0));
+ ssi_clk_disable(ssi);
+ msg->actual_len = 0;
+ msg->status = HSI_STATUS_PROCEEDING;
+
+ return 0;
+}
+
+static int ssi_start_dma(struct hsi_msg *msg, int lch)
+{
+ struct hsi_port *port = hsi_get_port(msg->cl);
+ struct omap_ssi_port *omap_port = hsi_port_drvdata(port);
+ struct hsi_controller *ssi = to_hsi_controller(port->device.parent);
+ struct omap_ssi_controller *omap_ssi = hsi_controller_drvdata(ssi);
+ void __iomem *gdd = omap_ssi->gdd;
+ int err;
+ u16 csdp;
+ u16 ccr;
+ u32 s_addr;
+ u32 d_addr;
+ u32 tmp;
+
+ if (msg->ttype == HSI_MSG_READ) {
+ err = dma_map_sg(&ssi->device, msg->sgt.sgl, msg->sgt.nents,
+ DMA_FROM_DEVICE);
+ if (err < 0) {
+ dev_dbg(&ssi->device, "DMA map SG failed !\n");
+ return err;
+ }
+ csdp = SSI_DST_BURST_4x32_BIT | SSI_DST_MEMORY_PORT |
+ SSI_SRC_SINGLE_ACCESS0 | SSI_SRC_PERIPHERAL_PORT |
+ SSI_DATA_TYPE_S32;
+ ccr = msg->channel + 0x10 + (port->num * 8); /* Sync */
+ ccr |= SSI_DST_AMODE_POSTINC | SSI_SRC_AMODE_CONST |
+ SSI_CCR_ENABLE;
+ s_addr = omap_port->ssr_dma +
+ SSI_SSR_BUFFER_CH_REG(msg->channel);
+ d_addr = sg_dma_address(msg->sgt.sgl);
+ } else {
+ err = dma_map_sg(&ssi->device, msg->sgt.sgl, msg->sgt.nents,
+ DMA_TO_DEVICE);
+ if (err < 0) {
+ dev_dbg(&ssi->device, "DMA map SG failed !\n");
+ return err;
+ }
+ csdp = SSI_SRC_BURST_4x32_BIT | SSI_SRC_MEMORY_PORT |
+ SSI_DST_SINGLE_ACCESS0 | SSI_DST_PERIPHERAL_PORT |
+ SSI_DATA_TYPE_S32;
+ ccr = (msg->channel + 1 + (port->num * 8)) & 0xf; /* Sync */
+ ccr |= SSI_SRC_AMODE_POSTINC | SSI_DST_AMODE_CONST |
+ SSI_CCR_ENABLE;
+ s_addr = sg_dma_address(msg->sgt.sgl);
+ d_addr = omap_port->sst_dma +
+ SSI_SST_BUFFER_CH_REG(msg->channel);
+ }
+ dev_dbg(&ssi->device, "lch %d cdsp %08x ccr %04x s_addr %08x"
+ " d_addr %08x\n", lch, csdp, ccr, s_addr, d_addr);
+ ssi_clk_enable(ssi); /* Hold clocks during the transfer */
+ __raw_writew(csdp, gdd + SSI_GDD_CSDP_REG(lch));
+ __raw_writew(SSI_BLOCK_IE | SSI_TOUT_IE, gdd + SSI_GDD_CICR_REG(lch));
+ __raw_writel(d_addr, gdd + SSI_GDD_CDSA_REG(lch));
+ __raw_writel(s_addr, gdd + SSI_GDD_CSSA_REG(lch));
+ __raw_writew(SSI_BYTES_TO_FRAMES(msg->sgt.sgl->length),
+ gdd + SSI_GDD_CEN_REG(lch));
+
+ spin_lock_bh(&omap_ssi->lock);
+ tmp = __raw_readl(omap_ssi->sys + SSI_GDD_MPU_IRQ_ENABLE_REG);
+ tmp |= SSI_GDD_LCH(lch);
+ __raw_writel(tmp, omap_ssi->sys + SSI_GDD_MPU_IRQ_ENABLE_REG);
+ spin_unlock_bh(&omap_ssi->lock);
+ __raw_writew(ccr, gdd + SSI_GDD_CCR_REG(lch));
+ msg->status = HSI_STATUS_PROCEEDING;
+
+ return 0;
+}
+
+static int ssi_start_transfer(struct list_head *queue)
+{
+ struct hsi_msg *msg;
+ int lch = -1;
+
+ if (list_empty(queue))
+ return 0;
+ msg = list_first_entry(queue, struct hsi_msg, link);
+ if (msg->status != HSI_STATUS_QUEUED)
+ return 0;
+ if ((msg->sgt.nents) && (msg->sgt.sgl->length > sizeof(u32)))
+ lch = ssi_claim_lch(msg);
+ if (lch >= 0)
+ return ssi_start_dma(msg, lch);
+ else
+ return ssi_start_pio(msg);
+}
+
+static void ssi_transfer(struct omap_ssi_port *omap_port,
+ struct list_head *queue)
+{
+ struct hsi_msg *msg;
+ int err = -1;
+
+ spin_lock_bh(&omap_port->lock);
+ while (err < 0) {
+ err = ssi_start_transfer(queue);
+ if (err < 0) {
+ msg = list_first_entry(queue, struct hsi_msg, link);
+ msg->status = HSI_STATUS_ERROR;
+ msg->actual_len = 0;
+ list_del(&msg->link);
+ spin_unlock_bh(&omap_port->lock);
+ msg->complete(msg);
+ spin_lock_bh(&omap_port->lock);
+ }
+ }
+ spin_unlock_bh(&omap_port->lock);
+}
+
+static u32 ssi_calculate_div(struct hsi_controller *ssi)
+{
+ struct omap_ssi_controller *omap_ssi = hsi_controller_drvdata(ssi);
+ u32 tx_fckrate = (u32) omap_ssi->fck_rate;
+
+ /* / 2 : SSI TX clock is always half of the SSI functional clock */
+ tx_fckrate >>= 1;
+ /* Round down when tx_fckrate % omap_ssi->max_speed == 0 */
+ tx_fckrate--;
+ dev_dbg(&ssi->device, "TX div %d for fck_rate %lu Khz speed %d Kb/s\n",
+ tx_fckrate / omap_ssi->max_speed, omap_ssi->fck_rate,
+ omap_ssi->max_speed);
+
+ return tx_fckrate / omap_ssi->max_speed;
+}
+
+static void ssi_error(struct hsi_port *port)
+{
+ struct omap_ssi_port *omap_port = hsi_port_drvdata(port);
+ struct hsi_controller *ssi = to_hsi_controller(port->device.parent);
+ struct omap_ssi_controller *omap_ssi = hsi_controller_drvdata(ssi);
+ struct hsi_msg *msg;
+ unsigned int i;
+ u32 err;
+ u32 val;
+ u32 tmp;
+
+ /* ACK error */
+ err = __raw_readl(omap_port->ssr_base + SSI_SSR_ERROR_REG);
+ dev_err(&port->device, "SSI error: 0x%02x\n", err);
+ if (!err) {
+ dev_dbg(&port->device, "spurious SSI error ignored!\n");
+ return;
+ }
+ spin_lock(&omap_ssi->lock);
+ /* Cancel all GDD read transfers */
+ for (i = 0, val = 0; i < SSI_MAX_GDD_LCH; i++) {
+ msg = omap_ssi->gdd_trn[i].msg;
+ if ((msg) && (msg->ttype == HSI_MSG_READ)) {
+ __raw_writew(0, omap_ssi->gdd + SSI_GDD_CCR_REG(i));
+ val |= (1 << i);
+ omap_ssi->gdd_trn[i].msg = NULL;
+ }
+ }
+ tmp = __raw_readl(omap_ssi->sys + SSI_GDD_MPU_IRQ_ENABLE_REG);
+ tmp &= ~val;
+ __raw_writel(tmp, omap_ssi->sys + SSI_GDD_MPU_IRQ_ENABLE_REG);
+ spin_unlock(&omap_ssi->lock);
+ /* Cancel all PIO read transfers */
+ spin_lock(&omap_port->lock);
+ tmp = __raw_readl(omap_ssi->sys + SSI_MPU_ENABLE_REG(port->num, 0));
+ tmp &= 0xfeff00ff; /* Disable error & all dataavailable interrupts */
+ __raw_writel(tmp, omap_ssi->sys + SSI_MPU_ENABLE_REG(port->num, 0));
+ /* ACK error */
+ __raw_writel(err, omap_port->ssr_base + SSI_SSR_ERRORACK_REG);
+ __raw_writel(SSI_ERROROCCURED,
+ omap_ssi->sys + SSI_MPU_STATUS_REG(port->num, 0));
+ /* Signal the error all current pending read requests */
+ for (i = 0; i < omap_port->channels; i++) {
+ if (list_empty(&omap_port->rxqueue[i]))
+ continue;
+ msg = list_first_entry(&omap_port->rxqueue[i], struct hsi_msg,
+ link);
+ list_del(&msg->link);
+ msg->status = HSI_STATUS_ERROR;
+ spin_unlock(&omap_port->lock);
+ msg->complete(msg);
+ /* Now restart queued reads if any */
+ ssi_transfer(omap_port, &omap_port->rxqueue[i]);
+ spin_lock(&omap_port->lock);
+ }
+ spin_unlock(&omap_port->lock);
+}
+
+static void ssi_break_complete(struct hsi_port *port)
+{
+ struct omap_ssi_port *omap_port = hsi_port_drvdata(port);
+ struct hsi_controller *ssi = to_hsi_controller(port->device.parent);
+ struct omap_ssi_controller *omap_ssi = hsi_controller_drvdata(ssi);
+ struct hsi_msg *msg;
+ struct hsi_msg *tmp;
+ u32 val;
+
+ dev_dbg(&port->device, "HWBREAK received\n");
+
+ spin_lock(&omap_port->lock);
+ val = __raw_readl(omap_ssi->sys + SSI_MPU_ENABLE_REG(port->num, 0));
+ val &= ~SSI_BREAKDETECTED;
+ __raw_writel(val, omap_ssi->sys + SSI_MPU_ENABLE_REG(port->num, 0));
+ __raw_writel(0, omap_port->ssr_base + SSI_SSR_BREAK_REG);
+ __raw_writel(SSI_BREAKDETECTED,
+ omap_ssi->sys + SSI_MPU_STATUS_REG(port->num, 0));
+ spin_unlock(&omap_port->lock);
+
+ list_for_each_entry_safe(msg, tmp, &omap_port->brkqueue, link) {
+ msg->status = HSI_STATUS_COMPLETED;
+ spin_lock(&omap_port->lock);
+ list_del(&msg->link);
+ spin_unlock(&omap_port->lock);
+ msg->complete(msg);
+ }
+
+}
+
+static int ssi_async_break(struct hsi_msg *msg)
+{
+ struct hsi_port *port = hsi_get_port(msg->cl);
+ struct omap_ssi_port *omap_port = hsi_port_drvdata(port);
+ struct hsi_controller *ssi = to_hsi_controller(port->device.parent);
+ struct omap_ssi_controller *omap_ssi = hsi_controller_drvdata(ssi);
+ int err = 0;
+ u32 tmp;
+
+ ssi_clk_enable(ssi);
+ if (msg->ttype == HSI_MSG_WRITE) {
+ if (omap_port->sst.mode != SSI_MODE_FRAME) {
+ err = -EINVAL;
+ goto out;
+ }
+ __raw_writel(1, omap_port->sst_base + SSI_SST_BREAK_REG);
+ msg->status = HSI_STATUS_COMPLETED;
+ msg->complete(msg);
+ } else {
+ if (omap_port->ssr.mode != SSI_MODE_FRAME) {
+ err = -EINVAL;
+ goto out;
+ }
+ spin_lock_bh(&omap_port->lock);
+ tmp = __raw_readl(omap_ssi->sys +
+ SSI_MPU_ENABLE_REG(port->num, 0));
+ __raw_writel(tmp | SSI_BREAKDETECTED,
+ omap_ssi->sys + SSI_MPU_ENABLE_REG(port->num, 0));
+ msg->status = HSI_STATUS_PROCEEDING;
+ list_add_tail(&msg->link, &omap_port->brkqueue);
+ spin_unlock_bh(&omap_port->lock);
+ }
+out:
+ ssi_clk_disable(ssi);
+
+ return err;
+}
+
+static int ssi_async(struct hsi_msg *msg)
+{
+ struct hsi_port *port = hsi_get_port(msg->cl);
+ struct omap_ssi_port *omap_port = hsi_port_drvdata(port);
+ struct list_head *queue;
+ int err = 0;
+
+ BUG_ON(!msg);
+
+ if (msg->sgt.nents > 1)
+ return -ENOSYS; /* TODO: Add sg support */
+
+ if (msg->break_frame)
+ return ssi_async_break(msg);
+
+ if (msg->ttype) {
+ BUG_ON(msg->channel >= omap_port->sst.channels);
+ queue = &omap_port->txqueue[msg->channel];
+ } else {
+ BUG_ON(msg->channel >= omap_port->ssr.channels);
+ queue = &omap_port->rxqueue[msg->channel];
+ }
+ msg->status = HSI_STATUS_QUEUED;
+ spin_lock_bh(&omap_port->lock);
+ list_add_tail(&msg->link, queue);
+ err = ssi_start_transfer(queue);
+ if (err < 0) {
+ list_del(&msg->link);
+ msg->status = HSI_STATUS_ERROR;
+ }
+ spin_unlock_bh(&omap_port->lock);
+ dev_dbg(&port->device, "msg status %d ttype %d ch %d\n",
+ msg->status, msg->ttype, msg->channel);
+
+ return err;
+}
+
+static void ssi_flush_queue(struct list_head *queue, struct hsi_client *cl)
+{
+ struct list_head *node, *tmp;
+ struct hsi_msg *msg;
+
+ list_for_each_safe(node, tmp, queue) {
+ msg = list_entry(node, struct hsi_msg, link);
+ if ((cl) && (cl != msg->cl))
+ continue;
+ list_del(node);
+ pr_debug("flush queue: ch %d, msg %p len %d type %d ctxt %p\n",
+ msg->channel, msg, msg->sgt.sgl->length,
+ msg->ttype, msg->context);
+ if (msg->destructor)
+ msg->destructor(msg);
+ else
+ hsi_free_msg(msg);
+ }
+}
+
+static int ssi_setup(struct hsi_client *cl)
+{
+ struct hsi_port *port = to_hsi_port(cl->device.parent);
+ struct omap_ssi_port *omap_port = hsi_port_drvdata(port);
+ struct hsi_controller *ssi = to_hsi_controller(port->device.parent);
+ struct omap_ssi_controller *omap_ssi = hsi_controller_drvdata(ssi);
+ void __iomem *sst = omap_port->sst_base;
+ void __iomem *ssr = omap_port->ssr_base;
+ u32 div;
+ u32 val;
+ int err = 0;
+
+ ssi_clk_enable(ssi);
+ spin_lock_bh(&omap_port->lock);
+ if (cl->tx_cfg.speed)
+ omap_ssi->max_speed = cl->tx_cfg.speed;
+ div = ssi_calculate_div(ssi);
+ if (div > SSI_MAX_DIVISOR) {
+ dev_err(&cl->device, "Invalid TX speed %d Mb/s (div %d)\n",
+ cl->tx_cfg.speed, div);
+ err = -EINVAL;
+ goto out;
+ }
+ /* Set TX/RX module to sleep to stop TX/RX during cfg update */
+ __raw_writel(SSI_MODE_SLEEP, sst + SSI_SST_MODE_REG);
+ __raw_writel(SSI_MODE_SLEEP, ssr + SSI_SSR_MODE_REG);
+ /* Flush posted write */
+ val = __raw_readl(ssr + SSI_SSR_MODE_REG);
+ /* TX */
+ __raw_writel(31, sst + SSI_SST_FRAMESIZE_REG);
+ __raw_writel(div, sst + SSI_SST_DIVISOR_REG);
+ __raw_writel(cl->tx_cfg.channels, sst + SSI_SST_CHANNELS_REG);
+ __raw_writel(cl->tx_cfg.arb_mode, sst + SSI_SST_ARBMODE_REG);
+ __raw_writel(cl->tx_cfg.mode, sst + SSI_SST_MODE_REG);
+ /* RX */
+ __raw_writel(31, ssr + SSI_SSR_FRAMESIZE_REG);
+ __raw_writel(cl->rx_cfg.channels, ssr + SSI_SSR_CHANNELS_REG);
+ __raw_writel(0, ssr + SSI_SSR_TIMEOUT_REG);
+ /* Cleanup the break queue if we leave FRAME mode */
+ if ((omap_port->ssr.mode == SSI_MODE_FRAME) &&
+ (cl->rx_cfg.mode != SSI_MODE_FRAME))
+ ssi_flush_queue(&omap_port->brkqueue, cl);
+ __raw_writel(cl->rx_cfg.mode, ssr + SSI_SSR_MODE_REG);
+ omap_port->channels = max(cl->rx_cfg.channels, cl->tx_cfg.channels);
+ /* Shadow registering for OFF mode */
+ /* SST */
+ omap_port->sst.divisor = div;
+ omap_port->sst.frame_size = 31;
+ omap_port->sst.channels = cl->tx_cfg.channels;
+ omap_port->sst.arb_mode = cl->tx_cfg.arb_mode;
+ omap_port->sst.mode = cl->tx_cfg.mode;
+ /* SSR */
+ omap_port->ssr.frame_size = 31;
+ omap_port->ssr.timeout = 0;
+ omap_port->ssr.channels = cl->rx_cfg.channels;
+ omap_port->ssr.mode = cl->rx_cfg.mode;
+out:
+ spin_unlock_bh(&omap_port->lock);
+ ssi_clk_disable(ssi);
+
+ return err;
+}
+
+static void ssi_cleanup_queues(struct hsi_client *cl)
+{
+ struct hsi_port *port = hsi_get_port(cl);
+ struct omap_ssi_port *omap_port = hsi_port_drvdata(port);
+ struct hsi_controller *ssi = to_hsi_controller(port->device.parent);
+ struct omap_ssi_controller *omap_ssi = hsi_controller_drvdata(ssi);
+ struct hsi_msg *msg;
+ unsigned int i;
+ u32 rxbufstate = 0;
+ u32 txbufstate = 0;
+ u32 status = SSI_ERROROCCURED;
+ u32 tmp;
+
+ ssi_flush_queue(&omap_port->brkqueue, cl);
+ if (list_empty(&omap_port->brkqueue))
+ status |= SSI_BREAKDETECTED;
+
+ for (i = 0; i < omap_port->channels; i++) {
+ if (list_empty(&omap_port->txqueue[i]))
+ continue;
+ msg = list_first_entry(&omap_port->txqueue[i], struct hsi_msg,
+ link);
+ if ((msg->cl == cl) && (msg->status == HSI_STATUS_PROCEEDING)) {
+ txbufstate |= (1 << i);
+ status |= SSI_DATAACCEPT(i);
+ /* Release the clocks writes, also GDD ones */
+ ssi_clk_disable(ssi);
+ }
+ ssi_flush_queue(&omap_port->txqueue[i], cl);
+ }
+ for (i = 0; i < omap_port->channels; i++) {
+ if (list_empty(&omap_port->rxqueue[i]))
+ continue;
+ msg = list_first_entry(&omap_port->rxqueue[i], struct hsi_msg,
+ link);
+ if ((msg->cl == cl) && (msg->status == HSI_STATUS_PROCEEDING)) {
+ rxbufstate |= (1 << i);
+ status |= SSI_DATAAVAILABLE(i);
+ }
+ ssi_flush_queue(&omap_port->rxqueue[i], cl);
+ /* Check if we keep the error detection interrupt armed */
+ if (!list_empty(&omap_port->rxqueue[i]))
+ status &= ~SSI_ERROROCCURED;
+ }
+ /* Cleanup write buffers */
+ tmp = __raw_readl(omap_port->sst_base + SSI_SST_BUFSTATE_REG);
+ tmp &= ~txbufstate;
+ __raw_writel(tmp, omap_port->sst_base + SSI_SST_BUFSTATE_REG);
+ /* Cleanup read buffers */
+ tmp = __raw_readl(omap_port->ssr_base + SSI_SSR_BUFSTATE_REG);
+ tmp &= ~rxbufstate;
+ __raw_writel(tmp, omap_port->ssr_base + SSI_SSR_BUFSTATE_REG);
+ /* Disarm and ack pending interrupts */
+ tmp = __raw_readl(omap_ssi->sys + SSI_MPU_ENABLE_REG(port->num, 0));
+ tmp &= ~status;
+ __raw_writel(tmp, omap_ssi->sys + SSI_MPU_ENABLE_REG(port->num, 0));
+ __raw_writel(status, omap_ssi->sys + SSI_MPU_STATUS_REG(port->num, 0));
+}
+
+static void ssi_cleanup_gdd(struct hsi_controller *ssi, struct hsi_client *cl)
+{
+ struct omap_ssi_controller *omap_ssi = hsi_controller_drvdata(ssi);
+ struct hsi_msg *msg;
+ unsigned int i;
+ u32 val = 0;
+ u32 tmp;
+
+ for (i = 0; i < SSI_MAX_GDD_LCH; i++) {
+ msg = omap_ssi->gdd_trn[i].msg;
+ if ((!msg) || (msg->cl != cl))
+ continue;
+ __raw_writew(0, omap_ssi->gdd + SSI_GDD_CCR_REG(i));
+ val |= (1 << i);
+ /*
+ * Clock references for write will be handled in
+ * ssi_cleanup_queues
+ */
+ if (msg->ttype == HSI_MSG_READ)
+ ssi_clk_disable(ssi);
+ omap_ssi->gdd_trn[i].msg = NULL;
+ }
+ tmp = __raw_readl(omap_ssi->sys + SSI_GDD_MPU_IRQ_ENABLE_REG);
+ tmp &= ~val;
+ __raw_writel(tmp, omap_ssi->sys + SSI_GDD_MPU_IRQ_ENABLE_REG);
+ __raw_writel(val, omap_ssi->sys + SSI_GDD_MPU_IRQ_STATUS_REG);
+}
+
+static int ssi_release(struct hsi_client *cl)
+{
+ struct hsi_port *port = hsi_get_port(cl);
+ struct omap_ssi_port *omap_port = hsi_port_drvdata(port);
+ struct hsi_controller *ssi = to_hsi_controller(port->device.parent);
+ struct omap_ssi_controller *omap_ssi = hsi_controller_drvdata(ssi);
+
+ spin_lock_bh(&omap_port->lock);
+ ssi_clk_enable(ssi);
+ /* Stop all the pending DMA requests for that client */
+ ssi_cleanup_gdd(ssi, cl);
+ /* Now cleanup all the queues */
+ ssi_cleanup_queues(cl);
+ ssi_clk_disable(ssi);
+ /* If it is the last client of the port, do extra checks and cleanup */
+ if (port->claimed <= 1) {
+ /*
+ * Drop the clock reference for the incoming wake line
+ * if it is still kept high by the other side.
+ */
+ if (omap_port->wkin_cken) {
+ ssi_clk_disable(ssi);
+ omap_port->wkin_cken = 0;
+ }
+ ssi_clk_enable(ssi);
+ /* Stop any SSI TX/RX without a client */
+ ssi_set_mode(ssi, SSI_MODE_SLEEP);
+ omap_port->sst.mode = SSI_MODE_SLEEP;
+ omap_port->ssr.mode = SSI_MODE_SLEEP;
+ ssi_clk_disable(ssi);
+ WARN_ON(omap_port->wk_refcount != 0);
+ WARN_ON(omap_ssi->ck_refcount != 0);
+ }
+ spin_unlock_bh(&omap_port->lock);
+
+ return 0;
+}
+
+static int ssi_flush(struct hsi_client *cl)
+{
+ struct hsi_port *port = hsi_get_port(cl);
+ struct omap_ssi_port *omap_port = hsi_port_drvdata(port);
+ struct hsi_controller *ssi = to_hsi_controller(port->device.parent);
+ struct omap_ssi_controller *omap_ssi = hsi_controller_drvdata(ssi);
+ struct hsi_msg *msg;
+ void __iomem *sst = omap_port->sst_base;
+ void __iomem *ssr = omap_port->ssr_base;
+ unsigned int i;
+ u32 err;
+
+ ssi_clk_enable(ssi);
+ spin_lock_bh(&omap_port->lock);
+ /* Stop all DMA transfers */
+ for (i = 0; i < SSI_MAX_GDD_LCH; i++) {
+ msg = omap_ssi->gdd_trn[i].msg;
+ if (!msg || (port != hsi_get_port(msg->cl)))
+ continue;
+ __raw_writew(0, omap_ssi->gdd + SSI_GDD_CCR_REG(i));
+ if (msg->ttype == HSI_MSG_READ)
+ ssi_clk_disable(ssi);
+ omap_ssi->gdd_trn[i].msg = NULL;
+ }
+ /* Flush all SST buffers */
+ __raw_writel(0, sst + SSI_SST_BUFSTATE_REG);
+ __raw_writel(0, sst + SSI_SST_TXSTATE_REG);
+ /* Flush all SSR buffers */
+ __raw_writel(0, ssr + SSI_SSR_RXSTATE_REG);
+ __raw_writel(0, ssr + SSI_SSR_BUFSTATE_REG);
+ /* Flush all errors */
+ err = __raw_readl(ssr + SSI_SSR_ERROR_REG);
+ __raw_writel(err, ssr + SSI_SSR_ERRORACK_REG);
+ /* Flush break */
+ __raw_writel(0, ssr + SSI_SSR_BREAK_REG);
+ /* Clear interrupts */
+ __raw_writel(0, omap_ssi->sys + SSI_MPU_ENABLE_REG(port->num, 0));
+ __raw_writel(0xffffff00,
+ omap_ssi->sys + SSI_MPU_STATUS_REG(port->num, 0));
+ __raw_writel(0, omap_ssi->sys + SSI_GDD_MPU_IRQ_ENABLE_REG);
+ __raw_writel(0xff, omap_ssi->sys + SSI_GDD_MPU_IRQ_STATUS_REG);
+ /* Dequeue all pending requests */
+ for (i = 0; i < omap_port->channels; i++) {
+ /* Release write clocks */
+ if (!list_empty(&omap_port->txqueue[i]))
+ ssi_clk_disable(ssi);
+ ssi_flush_queue(&omap_port->txqueue[i], NULL);
+ ssi_flush_queue(&omap_port->rxqueue[i], NULL);
+ }
+ ssi_flush_queue(&omap_port->brkqueue, NULL);
+ spin_unlock_bh(&omap_port->lock);
+ ssi_clk_disable(ssi);
+
+ return 0;
+}
+
+static int ssi_start_tx(struct hsi_client *cl)
+{
+ struct hsi_port *port = hsi_get_port(cl);
+ struct omap_ssi_port *omap_port = hsi_port_drvdata(port);
+ struct hsi_controller *ssi = to_hsi_controller(port->device.parent);
+ struct omap_ssi_controller *omap_ssi = hsi_controller_drvdata(ssi);
+
+ dev_dbg(&port->device, "Wake out high %d\n", omap_port->wk_refcount);
+
+ spin_lock_bh(&omap_port->wk_lock);
+ if (omap_port->wk_refcount++) {
+ spin_unlock_bh(&omap_port->wk_lock);
+ return 0;
+ }
+ ssi_clk_enable(ssi); /* Grab clocks */
+ __raw_writel(SSI_WAKE(0), omap_ssi->sys + SSI_SET_WAKE_REG(port->num));
+ spin_unlock_bh(&omap_port->wk_lock);
+
+ return 0;
+}
+
+static int ssi_stop_tx(struct hsi_client *cl)
+{
+ struct hsi_port *port = hsi_get_port(cl);
+ struct omap_ssi_port *omap_port = hsi_port_drvdata(port);
+ struct hsi_controller *ssi = to_hsi_controller(port->device.parent);
+ struct omap_ssi_controller *omap_ssi = hsi_controller_drvdata(ssi);
+
+ dev_dbg(&port->device, "Wake out low %d\n", omap_port->wk_refcount);
+
+ spin_lock_bh(&omap_port->wk_lock);
+ BUG_ON(!omap_port->wk_refcount);
+ if (--omap_port->wk_refcount) {
+ spin_unlock_bh(&omap_port->wk_lock);
+ return 0;
+ }
+ __raw_writel(SSI_WAKE(0),
+ omap_ssi->sys + SSI_CLEAR_WAKE_REG(port->num));
+ ssi_clk_disable(ssi); /* Release clocks */
+ spin_unlock_bh(&omap_port->wk_lock);
+
+ return 0;
+}
+
+static void ssi_pio_complete(struct hsi_port *port, struct list_head *queue)
+{
+ struct hsi_controller *ssi = to_hsi_controller(port->device.parent);
+ struct omap_ssi_controller *omap_ssi = hsi_controller_drvdata(ssi);
+ struct omap_ssi_port *omap_port = hsi_port_drvdata(port);
+ struct hsi_msg *msg;
+ u32 *buf;
+ u32 reg;
+ u32 val;
+
+ spin_lock(&omap_port->lock);
+ msg = list_first_entry(queue, struct hsi_msg, link);
+ if ((!msg->sgt.nents) || (!msg->sgt.sgl->length)) {
+ msg->actual_len = 0;
+ msg->status = HSI_STATUS_PENDING;
+ }
+ if (msg->ttype == HSI_MSG_WRITE)
+ val = SSI_DATAACCEPT(msg->channel);
+ else
+ val = SSI_DATAAVAILABLE(msg->channel);
+ if (msg->status == HSI_STATUS_PROCEEDING) {
+ buf = sg_virt(msg->sgt.sgl) + msg->actual_len;
+ if (msg->ttype == HSI_MSG_WRITE)
+ __raw_writel(*buf, omap_port->sst_base +
+ SSI_SST_BUFFER_CH_REG(msg->channel));
+ else
+ *buf = __raw_readl(omap_port->ssr_base +
+ SSI_SSR_BUFFER_CH_REG(msg->channel));
+ dev_dbg(&port->device, "ch %d ttype %d 0x%08x\n", msg->channel,
+ msg->ttype, *buf);
+ msg->actual_len += sizeof(*buf);
+ if (msg->actual_len >= msg->sgt.sgl->length)
+ msg->status = HSI_STATUS_COMPLETED;
+ /*
+ * Wait for the last written frame to be really sent before
+ * we call the complete callback
+ */
+ if ((msg->status == HSI_STATUS_PROCEEDING) ||
+ ((msg->status == HSI_STATUS_COMPLETED) &&
+ (msg->ttype == HSI_MSG_WRITE))) {
+ __raw_writel(val, omap_ssi->sys +
+ SSI_MPU_STATUS_REG(port->num, 0));
+ spin_unlock(&omap_port->lock);
+
+ return;
+ }
+
+ }
+ /* Transfer completed at this point */
+ reg = __raw_readl(omap_ssi->sys + SSI_MPU_ENABLE_REG(port->num, 0));
+ if (msg->ttype == HSI_MSG_WRITE)
+ ssi_clk_disable(ssi); /* Release clocks for write transfer */
+ reg &= ~val;
+ __raw_writel(reg, omap_ssi->sys + SSI_MPU_ENABLE_REG(port->num, 0));
+ __raw_writel(val, omap_ssi->sys + SSI_MPU_STATUS_REG(port->num, 0));
+ list_del(&msg->link);
+ spin_unlock(&omap_port->lock);
+ msg->complete(msg);
+ ssi_transfer(omap_port, queue);
+}
+
+static void ssi_gdd_complete(struct hsi_controller *ssi, unsigned int lch)
+{
+ struct omap_ssi_controller *omap_ssi = hsi_controller_drvdata(ssi);
+ struct hsi_msg *msg = omap_ssi->gdd_trn[lch].msg;
+ struct hsi_port *port = to_hsi_port(msg->cl->device.parent);
+ struct omap_ssi_port *omap_port = hsi_port_drvdata(port);
+ unsigned int dir;
+ u32 csr;
+ u32 val;
+
+ spin_lock(&omap_ssi->lock);
+
+ val = __raw_readl(omap_ssi->sys + SSI_GDD_MPU_IRQ_ENABLE_REG);
+ val &= ~SSI_GDD_LCH(lch);
+ __raw_writel(val, omap_ssi->sys + SSI_GDD_MPU_IRQ_ENABLE_REG);
+
+ if (msg->ttype == HSI_MSG_READ) {
+ dir = DMA_FROM_DEVICE;
+ val = SSI_DATAAVAILABLE(msg->channel);
+ ssi_clk_disable(ssi);
+ } else {
+ dir = DMA_TO_DEVICE;
+ val = SSI_DATAACCEPT(msg->channel);
+ /* Keep clocks reference for write pio event */
+ }
+ dma_unmap_sg(&ssi->device, msg->sgt.sgl, msg->sgt.nents, dir);
+ csr = __raw_readw(omap_ssi->gdd + SSI_GDD_CSR_REG(lch));
+ omap_ssi->gdd_trn[lch].msg = NULL; /* release GDD lch */
+ dev_dbg(&port->device, "DMA completed ch %d ttype %d\n",
+ msg->channel, msg->ttype);
+ spin_unlock(&omap_ssi->lock);
+ if (csr & SSI_CSR_TOUR) { /* Timeout error */
+ msg->status = HSI_STATUS_ERROR;
+ msg->actual_len = 0;
+ spin_lock(&omap_port->lock);
+ list_del(&msg->link); /* Dequeue msg */
+ spin_unlock(&omap_port->lock);
+ msg->complete(msg);
+ return;
+ }
+ spin_lock(&omap_port->lock);
+ val |= __raw_readl(omap_ssi->sys + SSI_MPU_ENABLE_REG(port->num, 0));
+ __raw_writel(val, omap_ssi->sys + SSI_MPU_ENABLE_REG(port->num, 0));
+ spin_unlock(&omap_port->lock);
+
+ msg->status = HSI_STATUS_COMPLETED;
+ msg->actual_len = sg_dma_len(msg->sgt.sgl);
+}
+
+static void ssi_gdd_tasklet(unsigned long dev)
+{
+ struct hsi_controller *ssi = (struct hsi_controller *)dev;
+ struct omap_ssi_controller *omap_ssi = hsi_controller_drvdata(ssi);
+ void __iomem *sys = omap_ssi->sys;
+ unsigned int lch;
+ u32 status_reg;
+
+ ssi_clk_enable(ssi);
+
+ status_reg = __raw_readl(sys + SSI_GDD_MPU_IRQ_STATUS_REG);
+ for (lch = 0; lch < SSI_MAX_GDD_LCH; lch++) {
+ if (status_reg & SSI_GDD_LCH(lch))
+ ssi_gdd_complete(ssi, lch);
+ }
+ __raw_writel(status_reg, sys + SSI_GDD_MPU_IRQ_STATUS_REG);
+ status_reg = __raw_readl(sys + SSI_GDD_MPU_IRQ_STATUS_REG);
+ ssi_clk_disable(ssi);
+ if (status_reg)
+ tasklet_hi_schedule(&omap_ssi->gdd_tasklet);
+ else
+ enable_irq(omap_ssi->gdd_irq);
+
+}
+
+static irqreturn_t ssi_gdd_isr(int irq, void *ssi)
+{
+ struct omap_ssi_controller *omap_ssi = hsi_controller_drvdata(ssi);
+
+ tasklet_hi_schedule(&omap_ssi->gdd_tasklet);
+ disable_irq_nosync(irq);
+
+ return IRQ_HANDLED;
+}
+
+static void ssi_pio_tasklet(unsigned long ssi_port)
+{
+ struct hsi_port *port = (struct hsi_port *)ssi_port;
+ struct hsi_controller *ssi = to_hsi_controller(port->device.parent);
+ struct omap_ssi_port *omap_port = hsi_port_drvdata(port);
+ struct omap_ssi_controller *omap_ssi = hsi_controller_drvdata(ssi);
+ void __iomem *sys = omap_ssi->sys;
+ unsigned int ch;
+ u32 status_reg;
+
+ ssi_clk_enable(ssi);
+ status_reg = __raw_readl(sys + SSI_MPU_STATUS_REG(port->num, 0));
+ status_reg &= __raw_readl(sys + SSI_MPU_ENABLE_REG(port->num, 0));
+
+ for (ch = 0; ch < omap_port->channels; ch++) {
+ if (status_reg & SSI_DATAACCEPT(ch))
+ ssi_pio_complete(port, &omap_port->txqueue[ch]);
+ if (status_reg & SSI_DATAAVAILABLE(ch))
+ ssi_pio_complete(port, &omap_port->rxqueue[ch]);
+ }
+ if (status_reg & SSI_BREAKDETECTED)
+ ssi_break_complete(port);
+ if (status_reg & SSI_ERROROCCURED)
+ ssi_error(port);
+
+ status_reg = __raw_readl(sys + SSI_MPU_STATUS_REG(port->num, 0));
+ status_reg &= __raw_readl(sys + SSI_MPU_ENABLE_REG(port->num, 0));
+ ssi_clk_disable(ssi);
+
+ if (status_reg)
+ tasklet_hi_schedule(&omap_port->pio_tasklet);
+ else
+ enable_irq(omap_port->irq);
+}
+
+static irqreturn_t ssi_pio_isr(int irq, void *port)
+{
+ struct omap_ssi_port *omap_port = hsi_port_drvdata(port);
+
+ tasklet_hi_schedule(&omap_port->pio_tasklet);
+ disable_irq_nosync(irq);
+
+ return IRQ_HANDLED;
+}
+
+static void ssi_wake_tasklet(unsigned long ssi_port)
+{
+ struct hsi_port *port = (struct hsi_port *)ssi_port;
+ struct hsi_controller *ssi = to_hsi_controller(port->device.parent);
+ struct omap_ssi_port *omap_port = hsi_port_drvdata(port);
+
+ if (ssi_wakein(port)) {
+ /**
+ * We can have a quick High-Low-High transition in the line.
+ * In such a case if we have long interrupt latencies,
+ * we can miss the low event or get twice a high event.
+ * This workaround will avoid breaking the clock reference
+ * count when such a situation ocurrs.
+ */
+ spin_lock(&omap_port->lock);
+ if (!omap_port->wkin_cken) {
+ omap_port->wkin_cken = 1;
+ ssi_clk_enable(ssi);
+ }
+ spin_unlock(&omap_port->lock);
+ dev_dbg(&ssi->device, "Wake in high\n");
+ hsi_event(port, HSI_EVENT_START_RX);
+ } else {
+ dev_dbg(&ssi->device, "Wake in low\n");
+ hsi_event(port, HSI_EVENT_STOP_RX);
+ spin_lock(&omap_port->lock);
+ if (omap_port->wkin_cken) {
+ ssi_clk_disable(ssi);
+ omap_port->wkin_cken = 0;
+ }
+ spin_unlock(&omap_port->lock);
+ }
+}
+
+static irqreturn_t ssi_wake_isr(int irq __maybe_unused, void *ssi_port)
+{
+ struct omap_ssi_port *omap_port = hsi_port_drvdata(ssi_port);
+
+ tasklet_hi_schedule(&omap_port->wake_tasklet);
+
+ return IRQ_HANDLED;
+}
+
+static int __init ssi_port_irq(struct hsi_port *port,
+ struct platform_device *pd)
+{
+ struct omap_ssi_port *omap_port = hsi_port_drvdata(port);
+ struct resource *irq;
+ int err;
+
+ irq = platform_get_resource(pd, IORESOURCE_IRQ, (port->num * 3) + 1);
+ if (!irq) {
+ dev_err(&port->device, "Port IRQ resource missing\n");
+ return -ENXIO;
+ }
+ omap_port->irq = irq->start;
+ tasklet_init(&omap_port->pio_tasklet, ssi_pio_tasklet,
+ (unsigned long)port);
+ err = devm_request_irq(&pd->dev, omap_port->irq, ssi_pio_isr,
+ IRQF_DISABLED, irq->name, port);
+ if (err < 0)
+ dev_err(&port->device, "Request IRQ %d failed (%d)\n",
+ omap_port->irq, err);
+ return err;
+}
+
+static int __init ssi_wake_irq(struct hsi_port *port,
+ struct platform_device *pd)
+{
+ struct omap_ssi_port *omap_port = hsi_port_drvdata(port);
+ struct resource *irq;
+ int err;
+
+ irq = platform_get_resource(pd, IORESOURCE_IRQ, (port->num * 3) + 3);
+ if (!irq) {
+ dev_err(&port->device, "Wake in IRQ resource missing");
+ return -ENXIO;
+ }
+ if (irq->flags & IORESOURCE_UNSET) {
+ dev_info(&port->device, "No Wake in support\n");
+ omap_port->wake_irq = -1;
+ return 0;
+ }
+ omap_port->wake_irq = irq->start;
+ tasklet_init(&omap_port->wake_tasklet, ssi_wake_tasklet,
+ (unsigned long)port);
+ err = devm_request_irq(&pd->dev, omap_port->wake_irq, ssi_wake_isr,
+ IRQF_DISABLED | IRQF_TRIGGER_RISING | IRQF_TRIGGER_FALLING,
+ irq->name, port);
+ if (err < 0)
+ dev_err(&port->device, "Request Wake in IRQ %d failed %d\n",
+ omap_port->wake_irq, err);
+ err = enable_irq_wake(omap_port->wake_irq);
+ if (err < 0)
+ dev_err(&port->device, "Enable wake on the wakeline in irq %d"
+ " failed %d\n", omap_port->wake_irq, err);
+
+ return err;
+}
+
+static void __init ssi_queues_init(struct omap_ssi_port *omap_port)
+{
+ unsigned int ch;
+
+ for (ch = 0; ch < SSI_MAX_CHANNELS; ch++) {
+ INIT_LIST_HEAD(&omap_port->txqueue[ch]);
+ INIT_LIST_HEAD(&omap_port->rxqueue[ch]);
+ }
+ INIT_LIST_HEAD(&omap_port->brkqueue);
+}
+
+static int __init ssi_get_iomem(struct platform_device *pd,
+ unsigned int num, void __iomem **pbase, dma_addr_t *phy)
+{
+ struct resource *mem;
+ struct resource *ioarea;
+ void __iomem *base;
+
+ mem = platform_get_resource(pd, IORESOURCE_MEM, num);
+ if (!mem) {
+ dev_err(&pd->dev, "IO memory region missing (%d)\n", num);
+ return -ENXIO;
+ }
+ ioarea = devm_request_mem_region(&pd->dev, mem->start,
+ resource_size(mem), dev_name(&pd->dev));
+ if (!ioarea) {
+ dev_err(&pd->dev, "%s IO memory region request failed\n",
+ mem->name);
+ return -ENXIO;
+ }
+ base = devm_ioremap(&pd->dev, mem->start, resource_size(mem));
+ if (!base) {
+ dev_err(&pd->dev, "%s IO remap failed\n", mem->name);
+ return -ENXIO;
+ }
+ *pbase = base;
+
+ if (phy)
+ *phy = mem->start;
+
+ return 0;
+}
+
+static int __init ssi_ports_init(struct hsi_controller *ssi,
+ struct platform_device *pd)
+{
+ struct hsi_port *port;
+ struct omap_ssi_controller *omap_ssi = hsi_controller_drvdata(ssi);
+ struct omap_ssi_port *omap_port;
+ unsigned int i;
+ int err;
+
+ omap_ssi->port = devm_kzalloc(&pd->dev,
+ sizeof(omap_port) * ssi->num_ports, GFP_KERNEL);
+ if (!omap_ssi->port)
+ return -ENOMEM;
+
+ for (i = 0; i < ssi->num_ports; i++) {
+ port = &ssi->port[i];
+ omap_port = devm_kzalloc(&pd->dev, sizeof(*omap_port),
+ GFP_KERNEL);
+ if (!omap_port)
+ return -ENOMEM;
+ port->async = ssi_async;
+ port->setup = ssi_setup;
+ port->flush = ssi_flush;
+ port->start_tx = ssi_start_tx;
+ port->stop_tx = ssi_stop_tx;
+ port->release = ssi_release;
+ hsi_port_set_drvdata(port, omap_port);
+ /* Get SST base addresses*/
+ err = ssi_get_iomem(pd, ((i * 2) + 2), &omap_port->sst_base,
+ &omap_port->sst_dma);
+ if (err < 0)
+ return err;
+ /* Get SSR base addresses */
+ err = ssi_get_iomem(pd, ((i * 2) + 3), &omap_port->ssr_base,
+ &omap_port->ssr_dma);
+ if (err < 0)
+ return err;
+ err = ssi_port_irq(port, pd);
+ if (err < 0)
+ return err;
+ err = ssi_wake_irq(port, pd);
+ if (err < 0)
+ return err;
+ ssi_queues_init(omap_port);
+ spin_lock_init(&omap_port->lock);
+ spin_lock_init(&omap_port->wk_lock);
+ omap_port->dev = &port->device;
+ omap_ssi->port[i] = omap_port;
+ }
+
+ return 0;
+}
+
+static void ssi_ports_exit(struct hsi_controller *ssi)
+{
+ struct omap_ssi_port *omap_port;
+ unsigned int i;
+
+ for (i = 0; i < ssi->num_ports; i++) {
+ omap_port = hsi_port_drvdata(&ssi->port[i]);
+ tasklet_kill(&omap_port->wake_tasklet);
+ tasklet_kill(&omap_port->pio_tasklet);
+ }
+}
+
+static void ssi_clk_release(struct device *dev __maybe_unused, void *res)
+{
+ struct ssi_clk_res *r = res;
+
+ clk_put(r->clk);
+}
+
+static struct clk *__init ssi_devm_clk_get(struct device *dev, const char *id)
+{
+ struct ssi_clk_res *pclk;
+ struct clk *clk;
+
+ pclk = devres_alloc(ssi_clk_release, sizeof(*pclk), GFP_KERNEL);
+ if (!pclk) {
+ dev_err(dev, "Could not allocate the device resource entry\n");
+ return ERR_PTR(-ENOMEM);
+ }
+ clk = clk_get(dev, id);
+ if (IS_ERR(clk)) {
+ dev_err(dev, "clock get %s failed %li\n", id, PTR_ERR(clk));
+ devres_free(pclk);
+ } else {
+ pclk->clk = clk;
+ devres_add(dev, pclk);
+ }
+
+ return clk;
+}
+
+static int __init ssi_add_controller(struct hsi_controller *ssi,
+ struct platform_device *pd)
+{
+ struct omap_ssi_platform_data *omap_ssi_pdata = pd->dev.platform_data;
+ struct omap_ssi_controller *omap_ssi;
+ struct resource *irq;
+ int err;
+
+ omap_ssi = devm_kzalloc(&pd->dev, sizeof(*omap_ssi), GFP_KERNEL);
+ if (!omap_ssi) {
+ dev_err(&pd->dev, "not enough memory for omap ssi\n");
+ return -ENOMEM;
+ }
+ ssi->id = pd->id;
+ ssi->owner = THIS_MODULE;
+ ssi->device.parent = &pd->dev;
+ dev_set_name(&ssi->device, "ssi%d", ssi->id);
+ hsi_controller_set_drvdata(ssi, omap_ssi);
+ omap_ssi->dev = &ssi->device;
+ err = ssi_get_iomem(pd, 0, &omap_ssi->sys, NULL);
+ if (err < 0)
+ return err;
+ err = ssi_get_iomem(pd, 1, &omap_ssi->gdd, NULL);
+ if (err < 0)
+ return err;
+ irq = platform_get_resource(pd, IORESOURCE_IRQ, 0);
+ if (!irq) {
+ dev_err(&pd->dev, "GDD IRQ resource missing\n");
+ return -ENXIO;
+ }
+ omap_ssi->gdd_irq = irq->start;
+ tasklet_init(&omap_ssi->gdd_tasklet, ssi_gdd_tasklet,
+ (unsigned long)ssi);
+ err = devm_request_irq(&pd->dev, omap_ssi->gdd_irq, ssi_gdd_isr,
+ IRQF_DISABLED, irq->name, ssi);
+ if (err < 0) {
+ dev_err(&ssi->device, "Request GDD IRQ %d failed (%d)",
+ omap_ssi->gdd_irq, err);
+ return err;
+ }
+ err = ssi_ports_init(ssi, pd);
+ if (err < 0)
+ return err;
+ omap_ssi->get_loss = omap_ssi_pdata->get_dev_context_loss_count;
+ omap_ssi->max_speed = UINT_MAX;
+ spin_lock_init(&omap_ssi->lock);
+ spin_lock_init(&omap_ssi->ck_lock);
+ omap_ssi->ick = ssi_devm_clk_get(&pd->dev, "ssi_ick");
+ if (IS_ERR(omap_ssi->ick))
+ return PTR_ERR(omap_ssi->ick);
+ omap_ssi->fck = ssi_devm_clk_get(&pd->dev, "ssi_ssr_fck");
+ if (IS_ERR(omap_ssi->fck))
+ return PTR_ERR(omap_ssi->fck);
+ err = hsi_register_controller(ssi);
+
+ return err;
+}
+
+static int __init ssi_hw_init(struct hsi_controller *ssi)
+{
+ struct omap_ssi_controller *omap_ssi = hsi_controller_drvdata(ssi);
+ unsigned int i;
+ u32 val;
+ int err;
+
+ err = ssi_clk_enable(ssi);
+ if (err < 0) {
+ dev_err(&ssi->device, "Failed to enable the clocks %d\n", err);
+ return err;
+ }
+ /* Reseting SSI controller */
+ __raw_writel(SSI_SOFTRESET, omap_ssi->sys + SSI_SYSCONFIG_REG);
+ val = __raw_readl(omap_ssi->sys + SSI_SYSSTATUS_REG);
+ for (i = 0; ((i < 20) && !(val & SSI_RESETDONE)); i++) {
+ msleep(20);
+ val = __raw_readl(omap_ssi->sys + SSI_SYSSTATUS_REG);
+ }
+ if (!(val & SSI_RESETDONE)) {
+ dev_err(&ssi->device, "SSI HW reset failed\n");
+ ssi_clk_disable(ssi);
+ return -EIO;
+ }
+ /* Reseting GDD */
+ __raw_writel(SSI_SWRESET, omap_ssi->gdd + SSI_GDD_GRST_REG);
+ /* Get FCK rate */
+ omap_ssi->fck_rate = clk_get_rate(omap_ssi->fck) / 1000; /* KHz */
+ dev_dbg(&ssi->device, "SSI fck rate %lu KHz\n", omap_ssi->fck_rate);
+ /* Set default PM settings */
+ val = SSI_AUTOIDLE | SSI_SIDLEMODE_SMART | SSI_MIDLEMODE_SMART;
+ __raw_writel(val, omap_ssi->sys + SSI_SYSCONFIG_REG);
+ omap_ssi->sysconfig = val;
+ __raw_writel(SSI_CLK_AUTOGATING_ON, omap_ssi->sys + SSI_GDD_GCR_REG);
+ omap_ssi->gdd_gcr = SSI_CLK_AUTOGATING_ON;
+ ssi_clk_disable(ssi);
+
+ return 0;
+}
+
+static void ssi_remove_controller(struct hsi_controller *ssi)
+{
+ struct omap_ssi_controller *omap_ssi = hsi_controller_drvdata(ssi);
+
+ ssi_ports_exit(ssi);
+ tasklet_kill(&omap_ssi->gdd_tasklet);
+ hsi_unregister_controller(ssi);
+}
+
+static int __init ssi_probe(struct platform_device *pd)
+{
+ struct omap_ssi_platform_data *omap_ssi_pdata = pd->dev.platform_data;
+ struct hsi_controller *ssi;
+ int err;
+
+ if (!omap_ssi_pdata) {
+ dev_err(&pd->dev, "No OMAP SSI platform data\n");
+ return -EINVAL;
+ }
+ ssi = hsi_alloc_controller(omap_ssi_pdata->num_ports, GFP_KERNEL);
+ if (!ssi) {
+ dev_err(&pd->dev, "No memory for controller\n");
+ return -ENOMEM;
+ }
+ platform_set_drvdata(pd, ssi);
+ err = ssi_add_controller(ssi, pd);
+ if (err < 0)
+ goto out1;
+ err = ssi_hw_init(ssi);
+ if (err < 0)
+ goto out2;
+#ifdef CONFIG_DEBUG_FS
+ err = ssi_debug_add_ctrl(ssi);
+ if (err < 0)
+ goto out2;
+#endif
+ return err;
+out2:
+ ssi_remove_controller(ssi);
+out1:
+ platform_set_drvdata(pd, NULL);
+ hsi_free_controller(ssi);
+
+ return err;
+}
+
+static int __exit ssi_remove(struct platform_device *pd)
+{
+ struct hsi_controller *ssi = platform_get_drvdata(pd);
+
+#ifdef CONFIG_DEBUG_FS
+ ssi_debug_remove_ctrl(ssi);
+#endif
+ ssi_remove_controller(ssi);
+ platform_set_drvdata(pd, NULL);
+ hsi_free_controller(ssi);
+
+ return 0;
+}
+
+static struct platform_driver ssi_pdriver = {
+ .remove = __exit_p(ssi_remove),
+ .driver = {
+ .name = "omap_ssi",
+ .owner = THIS_MODULE,
+ },
+};
+
+static int __init omap_ssi_init(void)
+{
+ pr_info("OMAP SSI hw driver loaded\n");
+ return platform_driver_probe(&ssi_pdriver, ssi_probe);
+}
+module_init(omap_ssi_init);
+
+static void __exit omap_ssi_exit(void)
+{
+ platform_driver_unregister(&ssi_pdriver);
+ pr_info("OMAP SSI driver removed\n");
+}
+module_exit(omap_ssi_exit);
+
+MODULE_ALIAS("platform:omap_ssi");
+MODULE_AUTHOR("Carlos Chinea <[email protected]>");
+MODULE_DESCRIPTION("Synchronous Serial Interface Driver");
+MODULE_LICENSE("GPL v2");
--
1.7.1

2010-12-14 10:11:21

by Carlos Chinea

[permalink] [raw]
Subject: [RFC PATCHv4 1/7] HSI: hsi: Introducing HSI framework

Adds HSI framework in to the linux kernel.

High Speed Synchronous Serial Interface (HSI) is a
serial interface mainly used for connecting application
engines (APE) with cellular modem engines (CMT) in cellular
handsets.

HSI provides multiplexing for up to 16 logical channels,
low-latency and full duplex communication.

Signed-off-by: Carlos Chinea <[email protected]>
---
drivers/Kconfig | 2 +
drivers/Makefile | 1 +
drivers/hsi/Kconfig | 17 ++
drivers/hsi/Makefile | 5 +
drivers/hsi/hsi.c | 496 +++++++++++++++++++++++++++++++++++++++++++
drivers/hsi/hsi_boardinfo.c | 64 ++++++
drivers/hsi/hsi_core.h | 37 ++++
include/linux/hsi/hsi.h | 389 +++++++++++++++++++++++++++++++++
8 files changed, 1011 insertions(+), 0 deletions(-)
create mode 100644 drivers/hsi/Kconfig
create mode 100644 drivers/hsi/Makefile
create mode 100644 drivers/hsi/hsi.c
create mode 100644 drivers/hsi/hsi_boardinfo.c
create mode 100644 drivers/hsi/hsi_core.h
create mode 100644 include/linux/hsi/hsi.h

diff --git a/drivers/Kconfig b/drivers/Kconfig
index a2b902f..4fe39f9 100644
--- a/drivers/Kconfig
+++ b/drivers/Kconfig
@@ -50,6 +50,8 @@ source "drivers/i2c/Kconfig"

source "drivers/spi/Kconfig"

+source "drivers/hsi/Kconfig"
+
source "drivers/pps/Kconfig"

source "drivers/gpio/Kconfig"
diff --git a/drivers/Makefile b/drivers/Makefile
index dabaabc..5f9d0da 100644
--- a/drivers/Makefile
+++ b/drivers/Makefile
@@ -47,6 +47,7 @@ obj-$(CONFIG_SCSI) += scsi/
obj-$(CONFIG_ATA) += ata/
obj-$(CONFIG_MTD) += mtd/
obj-$(CONFIG_SPI) += spi/
+obj-y += hsi/
obj-y += net/
obj-$(CONFIG_ATM) += atm/
obj-$(CONFIG_FUSION) += message/
diff --git a/drivers/hsi/Kconfig b/drivers/hsi/Kconfig
new file mode 100644
index 0000000..937062e
--- /dev/null
+++ b/drivers/hsi/Kconfig
@@ -0,0 +1,17 @@
+#
+# HSI driver configuration
+#
+menuconfig HSI
+ tristate "HSI support"
+ ---help---
+ The "High speed synchronous Serial Interface" is
+ synchronous serial interface used mainly to connect
+ application engines and cellular modems.
+
+if HSI
+
+config HSI_BOARDINFO
+ bool
+ default y
+
+endif # HSI
diff --git a/drivers/hsi/Makefile b/drivers/hsi/Makefile
new file mode 100644
index 0000000..ed94a3a
--- /dev/null
+++ b/drivers/hsi/Makefile
@@ -0,0 +1,5 @@
+#
+# Makefile for HSI
+#
+obj-$(CONFIG_HSI_BOARDINFO) += hsi_boardinfo.o
+obj-$(CONFIG_HSI) += hsi.o
diff --git a/drivers/hsi/hsi.c b/drivers/hsi/hsi.c
new file mode 100644
index 0000000..06b5743
--- /dev/null
+++ b/drivers/hsi/hsi.c
@@ -0,0 +1,496 @@
+/*
+ * hsi.c
+ *
+ * HSI core.
+ *
+ * Copyright (C) 2010 Nokia Corporation. All rights reserved.
+ *
+ * Contact: Carlos Chinea <[email protected]>
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public License
+ * 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.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA
+ * 02110-1301 USA
+ */
+#include <linux/hsi/hsi.h>
+#include <linux/compiler.h>
+#include <linux/rwsem.h>
+#include <linux/list.h>
+#include <linux/spinlock.h>
+#include <linux/kobject.h>
+#include <linux/slab.h>
+#include <linux/string.h>
+#include "hsi_core.h"
+
+static struct device_type hsi_ctrl = {
+ .name = "hsi_controller",
+};
+
+static struct device_type hsi_cl = {
+ .name = "hsi_client",
+};
+
+static struct device_type hsi_port = {
+ .name = "hsi_port",
+};
+
+static ssize_t modalias_show(struct device *dev,
+ struct device_attribute *a __maybe_unused, char *buf)
+{
+ return sprintf(buf, "hsi:%s\n", dev_name(dev));
+}
+
+static struct device_attribute hsi_bus_dev_attrs[] = {
+ __ATTR_RO(modalias),
+ __ATTR_NULL,
+};
+
+static int hsi_bus_uevent(struct device *dev, struct kobj_uevent_env *env)
+{
+ if (dev->type == &hsi_cl)
+ add_uevent_var(env, "MODALIAS=hsi:%s", dev_name(dev));
+
+ return 0;
+}
+
+static int hsi_bus_match(struct device *dev, struct device_driver *driver)
+{
+ return strcmp(dev_name(dev), driver->name) == 0;
+}
+
+static struct bus_type hsi_bus_type = {
+ .name = "hsi",
+ .dev_attrs = hsi_bus_dev_attrs,
+ .match = hsi_bus_match,
+ .uevent = hsi_bus_uevent,
+};
+
+static void hsi_client_release(struct device *dev)
+{
+ kfree(to_hsi_client(dev));
+}
+
+static void hsi_new_client(struct hsi_port *port, struct hsi_board_info *info)
+{
+ struct hsi_client *cl;
+ unsigned long flags;
+
+ cl = kzalloc(sizeof(*cl), GFP_KERNEL);
+ if (!cl)
+ return;
+ cl->device.type = &hsi_cl;
+ cl->tx_cfg = info->tx_cfg;
+ cl->rx_cfg = info->rx_cfg;
+ cl->device.bus = &hsi_bus_type;
+ cl->device.parent = &port->device;
+ cl->device.release = hsi_client_release;
+ dev_set_name(&cl->device, info->name);
+ cl->device.platform_data = info->platform_data;
+ spin_lock_irqsave(&port->clock, flags);
+ list_add_tail(&cl->link, &port->clients);
+ spin_unlock_irqrestore(&port->clock, flags);
+ if (info->archdata)
+ cl->device.archdata = *info->archdata;
+ if (device_register(&cl->device) < 0) {
+ pr_err("hsi: failed to register client: %s\n", info->name);
+ kfree(cl);
+ }
+}
+
+static void hsi_scan_board_info(struct hsi_controller *hsi)
+{
+ struct hsi_cl_info *cl_info;
+ struct hsi_port *p;
+
+ list_for_each_entry(cl_info, &hsi_board_list, list)
+ if (cl_info->info.hsi_id == hsi->id) {
+ p = hsi_find_port_num(hsi, cl_info->info.port);
+ if (!p)
+ continue;
+ hsi_new_client(p, &cl_info->info);
+ }
+}
+
+static int hsi_remove_client(struct device *dev, void *data __maybe_unused)
+{
+ struct hsi_client *cl = to_hsi_client(dev);
+ struct hsi_port *port = to_hsi_port(dev->parent);
+ unsigned long flags;
+
+ spin_lock_irqsave(&port->clock, flags);
+ list_del(&cl->link);
+ spin_unlock_irqrestore(&port->clock, flags);
+ device_unregister(dev);
+
+ return 0;
+}
+
+static int hsi_remove_port(struct device *dev, void *data __maybe_unused)
+{
+ device_for_each_child(dev, NULL, hsi_remove_client);
+ device_unregister(dev);
+
+ return 0;
+}
+
+static void hsi_controller_release(struct device *dev __maybe_unused)
+{
+}
+
+static void hsi_port_release(struct device *dev __maybe_unused)
+{
+}
+
+/**
+ * hsi_unregister_controller - Unregister an HSI controller
+ * @hsi: The HSI controller to register
+ */
+void hsi_unregister_controller(struct hsi_controller *hsi)
+{
+ device_for_each_child(&hsi->device, NULL, hsi_remove_port);
+ device_unregister(&hsi->device);
+}
+EXPORT_SYMBOL_GPL(hsi_unregister_controller);
+
+/**
+ * hsi_register_controller - Register an HSI controller and its ports
+ * @hsi: The HSI controller to register
+ *
+ * Returns -errno on failure, 0 on success.
+ */
+int hsi_register_controller(struct hsi_controller *hsi)
+{
+ unsigned int i;
+ int err;
+
+ hsi->device.type = &hsi_ctrl;
+ hsi->device.bus = &hsi_bus_type;
+ hsi->device.release = hsi_controller_release;
+ err = device_register(&hsi->device);
+ if (err < 0)
+ return err;
+ for (i = 0; i < hsi->num_ports; i++) {
+ hsi->port[i].device.parent = &hsi->device;
+ hsi->port[i].device.bus = &hsi_bus_type;
+ hsi->port[i].device.release = hsi_port_release;
+ hsi->port[i].device.type = &hsi_port;
+ INIT_LIST_HEAD(&hsi->port[i].clients);
+ spin_lock_init(&hsi->port[i].clock);
+ err = device_register(&hsi->port[i].device);
+ if (err < 0)
+ goto out;
+ }
+ /* Populate HSI bus with HSI clients */
+ hsi_scan_board_info(hsi);
+
+ return 0;
+out:
+ hsi_unregister_controller(hsi);
+
+ return err;
+}
+EXPORT_SYMBOL_GPL(hsi_register_controller);
+
+/**
+ * hsi_register_client_driver - Register an HSI client to the HSI bus
+ * @drv: HSI client driver to register
+ *
+ * Returns -errno on failure, 0 on success.
+ */
+int hsi_register_client_driver(struct hsi_client_driver *drv)
+{
+ drv->driver.bus = &hsi_bus_type;
+
+ return driver_register(&drv->driver);
+}
+EXPORT_SYMBOL_GPL(hsi_register_client_driver);
+
+static inline int hsi_dummy_msg(struct hsi_msg *msg __maybe_unused)
+{
+ return 0;
+}
+
+static inline int hsi_dummy_cl(struct hsi_client *cl __maybe_unused)
+{
+ return 0;
+}
+
+/**
+ * hsi_alloc_controller - Allocate an HSI controller and its ports
+ * @n_ports: Number of ports on the HSI controller
+ * @flags: Kernel allocation flags
+ *
+ * Return NULL on failure or a pointer to an hsi_controller on success.
+ */
+struct hsi_controller *hsi_alloc_controller(unsigned int n_ports, gfp_t flags)
+{
+ struct hsi_controller *hsi;
+ struct hsi_port *port;
+ unsigned int i;
+
+ if (!n_ports)
+ return NULL;
+
+ port = kzalloc(sizeof(*port)*n_ports, flags);
+ if (!port)
+ return NULL;
+ hsi = kzalloc(sizeof(*hsi), flags);
+ if (!hsi)
+ goto out;
+ for (i = 0; i < n_ports; i++) {
+ dev_set_name(&port[i].device, "port%d", i);
+ port[i].num = i;
+ port[i].async = hsi_dummy_msg;
+ port[i].setup = hsi_dummy_cl;
+ port[i].flush = hsi_dummy_cl;
+ port[i].start_tx = hsi_dummy_cl;
+ port[i].stop_tx = hsi_dummy_cl;
+ port[i].release = hsi_dummy_cl;
+ mutex_init(&port[i].lock);
+ }
+ hsi->num_ports = n_ports;
+ hsi->port = port;
+
+ return hsi;
+out:
+ kfree(port);
+
+ return NULL;
+}
+EXPORT_SYMBOL_GPL(hsi_alloc_controller);
+
+/**
+ * hsi_free_controller - Free an HSI controller
+ * @hsi: Pointer to HSI controller
+ */
+void hsi_free_controller(struct hsi_controller *hsi)
+{
+ if (!hsi)
+ return;
+
+ kfree(hsi->port);
+ kfree(hsi);
+}
+EXPORT_SYMBOL_GPL(hsi_free_controller);
+
+/**
+ * hsi_free_msg - Free an HSI message
+ * @msg: Pointer to the HSI message
+ *
+ * Client is responsible to free the buffers pointed by the scatterlists.
+ */
+void hsi_free_msg(struct hsi_msg *msg)
+{
+ if (!msg)
+ return;
+ sg_free_table(&msg->sgt);
+ kfree(msg);
+}
+EXPORT_SYMBOL_GPL(hsi_free_msg);
+
+/**
+ * hsi_alloc_msg - Allocate an HSI message
+ * @nents: Number of memory entries
+ * @flags: Kernel allocation flags
+ *
+ * nents can be 0. This mainly makes sense for read transfer.
+ * In that case, HSI drivers will call the complete callback when
+ * there is data to be read without consuming it.
+ *
+ * Return NULL on failure or a pointer to an hsi_msg on success.
+ */
+struct hsi_msg *hsi_alloc_msg(unsigned int nents, gfp_t flags)
+{
+ struct hsi_msg *msg;
+ int err;
+
+ msg = kzalloc(sizeof(*msg), flags);
+ if (!msg)
+ return NULL;
+
+ if (!nents)
+ return msg;
+
+ err = sg_alloc_table(&msg->sgt, nents, flags);
+ if (unlikely(err)) {
+ kfree(msg);
+ msg = NULL;
+ }
+
+ return msg;
+}
+EXPORT_SYMBOL_GPL(hsi_alloc_msg);
+
+/**
+ * hsi_async - Submit an HSI transfer to the controller
+ * @cl: HSI client sending the transfer
+ * @msg: The HSI transfer passed to controller
+ *
+ * The HSI message must have the channel, ttype, complete and destructor
+ * fields set beforehand. If nents > 0 then the client has to initialize
+ * also the scatterlists to point to the buffers to write to or read from.
+ *
+ * HSI controllers relay on pre-allocated buffers from their clients and they
+ * do not allocate buffers on their own.
+ *
+ * Once the HSI message transfer finishes, the HSI controller calls the
+ * complete callback with the status and actual_len fields of the HSI message
+ * updated. The complete callback can be called before returning from
+ * hsi_async.
+ *
+ * Returns -errno on failure or 0 on success
+ */
+int hsi_async(struct hsi_client *cl, struct hsi_msg *msg)
+{
+ struct hsi_port *port = hsi_get_port(cl);
+
+ if (!hsi_port_claimed(cl))
+ return -EACCES;
+
+ WARN_ON_ONCE(!msg->destructor || !msg->complete);
+ msg->cl = cl;
+
+ return port->async(msg);
+}
+EXPORT_SYMBOL_GPL(hsi_async);
+
+/**
+ * hsi_claim_port - Claim the HSI client's port
+ * @cl: HSI client that wants to claim its port
+ * @share: Flag to indicate if the client wants to share the port or not.
+ *
+ * Returns -errno on failure, 0 on success.
+ */
+int hsi_claim_port(struct hsi_client *cl, unsigned int share)
+{
+ struct hsi_port *port = hsi_get_port(cl);
+ int err = 0;
+
+ mutex_lock(&port->lock);
+ if ((port->claimed) && (!port->shared || !share)) {
+ err = -EBUSY;
+ goto out;
+ }
+ if (!try_module_get(to_hsi_controller(port->device.parent)->owner)) {
+ err = -ENODEV;
+ goto out;
+ }
+ port->claimed++;
+ port->shared = !!share;
+ cl->pclaimed = 1;
+out:
+ mutex_unlock(&port->lock);
+
+ return err;
+}
+EXPORT_SYMBOL_GPL(hsi_claim_port);
+
+/**
+ * hsi_release_port - Release the HSI client's port
+ * @cl: HSI client which previously claimed its port
+ */
+void hsi_release_port(struct hsi_client *cl)
+{
+ struct hsi_port *port = hsi_get_port(cl);
+
+ mutex_lock(&port->lock);
+ /* Allow HW driver to do some cleanup */
+ port->release(cl);
+ if (cl->pclaimed)
+ port->claimed--;
+ BUG_ON(port->claimed < 0);
+ cl->pclaimed = 0;
+ if (!port->claimed)
+ port->shared = 0;
+ module_put(to_hsi_controller(port->device.parent)->owner);
+ mutex_unlock(&port->lock);
+}
+EXPORT_SYMBOL_GPL(hsi_release_port);
+
+static int hsi_start_rx(struct hsi_client *cl, void *data __maybe_unused)
+{
+ if (cl->hsi_start_rx)
+ (*cl->hsi_start_rx)(cl);
+
+ return 0;
+}
+
+static int hsi_stop_rx(struct hsi_client *cl, void *data __maybe_unused)
+{
+ if (cl->hsi_stop_rx)
+ (*cl->hsi_stop_rx)(cl);
+
+ return 0;
+}
+
+static int hsi_port_for_each_client(struct hsi_port *port, void *data,
+ int (*fn)(struct hsi_client *cl, void *data))
+{
+ struct hsi_client *cl;
+
+ spin_lock(&port->clock);
+ list_for_each_entry(cl, &port->clients, link) {
+ spin_unlock(&port->clock);
+ (*fn)(cl, data);
+ spin_lock(&port->clock);
+ }
+ spin_unlock(&port->clock);
+
+ return 0;
+}
+
+/**
+ * hsi_event -Notifies clients about port events
+ * @port: Port where the event occurred
+ * @event: The event type
+ *
+ * Clients should not be concerned about wake line behavior. However, due
+ * to a race condition in HSI HW protocol, clients need to be notified
+ * about wake line changes, so they can implement a workaround for it.
+ *
+ * Events:
+ * HSI_EVENT_START_RX - Incoming wake line high
+ * HSI_EVENT_STOP_RX - Incoming wake line down
+ */
+void hsi_event(struct hsi_port *port, unsigned int event)
+{
+ int (*fn)(struct hsi_client *cl, void *data);
+
+ switch (event) {
+ case HSI_EVENT_START_RX:
+ fn = hsi_start_rx;
+ break;
+ case HSI_EVENT_STOP_RX:
+ fn = hsi_stop_rx;
+ break;
+ default:
+ return;
+ }
+ hsi_port_for_each_client(port, NULL, fn);
+}
+EXPORT_SYMBOL_GPL(hsi_event);
+
+static int __init hsi_init(void)
+{
+ return bus_register(&hsi_bus_type);
+}
+postcore_initcall(hsi_init);
+
+static void __exit hsi_exit(void)
+{
+ bus_unregister(&hsi_bus_type);
+}
+module_exit(hsi_exit);
+
+MODULE_AUTHOR("Carlos Chinea <[email protected]>");
+MODULE_DESCRIPTION("High-speed Synchronous Serial Interface (HSI) framework");
+MODULE_LICENSE("GPL v2");
diff --git a/drivers/hsi/hsi_boardinfo.c b/drivers/hsi/hsi_boardinfo.c
new file mode 100644
index 0000000..3a9e4e8
--- /dev/null
+++ b/drivers/hsi/hsi_boardinfo.c
@@ -0,0 +1,64 @@
+/*
+ * hsi_boardinfo.c
+ *
+ * HSI clients registration interface
+ *
+ * Copyright (C) 2010 Nokia Corporation. All rights reserved.
+ *
+ * Contact: Carlos Chinea <[email protected]>
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public License
+ * 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.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA
+ * 02110-1301 USA
+ */
+#include <linux/hsi/hsi.h>
+#include <linux/list.h>
+#include <linux/slab.h>
+#include "hsi_core.h"
+
+/*
+ * hsi_board_list is only used internally by the HSI framework.
+ * No one else is allowed to make use of it.
+ */
+LIST_HEAD(hsi_board_list);
+EXPORT_SYMBOL_GPL(hsi_board_list);
+
+/**
+ * hsi_register_board_info - Register HSI clients information
+ * @info: Array of HSI clients on the board
+ * @len: Length of the array
+ *
+ * HSI clients are statically declared and registered on board files.
+ *
+ * HSI clients will be automatically registered to the HSI bus once the
+ * controller and the port where the clients wishes to attach are registered
+ * to it.
+ *
+ * Return -errno on failure, 0 on success.
+ */
+int __init hsi_register_board_info(struct hsi_board_info const *info,
+ unsigned int len)
+{
+ struct hsi_cl_info *cl_info;
+
+ cl_info = kzalloc(sizeof(*cl_info) * len, GFP_KERNEL);
+ if (!cl_info)
+ return -ENOMEM;
+
+ for (; len; len--, info++, cl_info++) {
+ cl_info->info = *info;
+ list_add_tail(&cl_info->list, &hsi_board_list);
+ }
+
+ return 0;
+}
diff --git a/drivers/hsi/hsi_core.h b/drivers/hsi/hsi_core.h
new file mode 100644
index 0000000..8005509
--- /dev/null
+++ b/drivers/hsi/hsi_core.h
@@ -0,0 +1,37 @@
+/*
+ * hsi_core.h
+ *
+ * HSI framework internal interfaces,
+ *
+ * Copyright (C) 2010 Nokia Corporation. All rights reserved.
+ *
+ * Contact: Carlos Chinea <[email protected]>
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public License
+ * 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.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA
+ * 02110-1301 USA
+ */
+
+#ifndef __LINUX_HSI_CORE_H__
+#define __LINUX_HSI_CORE_H__
+
+#include <linux/hsi/hsi.h>
+
+struct hsi_cl_info {
+ struct list_head list;
+ struct hsi_board_info info;
+};
+
+extern struct list_head hsi_board_list;
+
+#endif /* __LINUX_HSI_CORE_H__ */
diff --git a/include/linux/hsi/hsi.h b/include/linux/hsi/hsi.h
new file mode 100644
index 0000000..1a52623
--- /dev/null
+++ b/include/linux/hsi/hsi.h
@@ -0,0 +1,389 @@
+/*
+ * hsi.h
+ *
+ * HSI core header file.
+ *
+ * Copyright (C) 2010 Nokia Corporation. All rights reserved.
+ *
+ * Contact: Carlos Chinea <[email protected]>
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public License
+ * 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.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA
+ * 02110-1301 USA
+ */
+
+#ifndef __LINUX_HSI_H__
+#define __LINUX_HSI_H__
+
+#include <linux/device.h>
+#include <linux/mutex.h>
+#include <linux/scatterlist.h>
+#include <linux/spinlock.h>
+#include <linux/list.h>
+#include <linux/module.h>
+
+/* HSI message ttype */
+#define HSI_MSG_READ 0
+#define HSI_MSG_WRITE 1
+
+/* HSI configuration values */
+enum {
+ HSI_MODE_STREAM = 1,
+ HSI_MODE_FRAME,
+};
+
+enum {
+ HSI_FLOW_SYNC, /* Synchronized flow */
+ HSI_FLOW_PIPE, /* Pipelined flow */
+};
+
+enum {
+ HSI_ARB_RR, /* Round-robin arbitration */
+ HSI_ARB_PRIO, /* Channel priority arbitration */
+};
+
+#define HSI_MAX_CHANNELS 16
+
+/* HSI message status codes */
+enum {
+ HSI_STATUS_COMPLETED, /* Message transfer is completed */
+ HSI_STATUS_PENDING, /* Message pending to be read/write (POLL) */
+ HSI_STATUS_PROCEEDING, /* Message transfer is ongoing */
+ HSI_STATUS_QUEUED, /* Message waiting to be served */
+ HSI_STATUS_ERROR, /* Error when message transfer was ongoing */
+};
+
+/* HSI port event codes */
+enum {
+ HSI_EVENT_START_RX,
+ HSI_EVENT_STOP_RX,
+};
+
+/**
+ * struct hsi_config - Configuration for RX/TX HSI modules
+ * @mode: Bit transmission mode (STREAM or FRAME)
+ * @channels: Number of channels to use [1..16]
+ * @speed: Max bit transmission speed (Kbit/s)
+ * @flow: RX flow type (SYNCHRONIZED or PIPELINE)
+ * @arb_mode: Arbitration mode for TX frame (Round robin, priority)
+ */
+struct hsi_config {
+ unsigned int mode;
+ unsigned int channels;
+ unsigned int speed;
+ union {
+ unsigned int flow; /* RX only */
+ unsigned int arb_mode; /* TX only */
+ };
+};
+
+/**
+ * struct hsi_board_info - HSI client board info
+ * @name: Name for the HSI device
+ * @hsi_id: HSI controller id where the client sits
+ * @port: Port number in the controller where the client sits
+ * @tx_cfg: HSI TX configuration
+ * @rx_cfg: HSI RX configuration
+ * @platform_data: Platform related data
+ * @archdata: Architecture-dependent device data
+ */
+struct hsi_board_info {
+ const char *name;
+ int hsi_id;
+ unsigned int port;
+ struct hsi_config tx_cfg;
+ struct hsi_config rx_cfg;
+ void *platform_data;
+ struct dev_archdata *archdata;
+};
+
+#ifdef CONFIG_HSI_BOARDINFO
+extern int hsi_register_board_info(struct hsi_board_info const *info,
+ unsigned int len);
+#else
+static inline int hsi_register_board_info(struct hsi_board_info const *info,
+ unsigned int len)
+{
+ return 0;
+}
+#endif /* CONFIG_HSI_BOARDINFO */
+
+/**
+ * struct hsi_client - HSI client attached to an HSI port
+ * @device: Driver model representation of the device
+ * @tx_cfg: HSI TX configuration
+ * @rx_cfg: HSI RX configuration
+ * @hsi_start_rx: Called after incoming wake line goes high
+ * @hsi_stop_rx: Called after incoming wake line goes low
+ */
+struct hsi_client {
+ struct device device;
+ struct hsi_config tx_cfg;
+ struct hsi_config rx_cfg;
+ void (*hsi_start_rx)(struct hsi_client *cl);
+ void (*hsi_stop_rx)(struct hsi_client *cl);
+ /* private: */
+ unsigned int pclaimed:1;
+ struct list_head link;
+};
+
+#define to_hsi_client(dev) container_of(dev, struct hsi_client, device)
+
+static inline void hsi_client_set_drvdata(struct hsi_client *cl, void *data)
+{
+ dev_set_drvdata(&cl->device, data);
+}
+
+static inline void *hsi_client_drvdata(struct hsi_client *cl)
+{
+ return dev_get_drvdata(&cl->device);
+}
+
+/**
+ * struct hsi_client_driver - Driver associated to an HSI client
+ * @driver: Driver model representation of the driver
+ */
+struct hsi_client_driver {
+ struct device_driver driver;
+};
+
+#define to_hsi_client_driver(drv) container_of(drv, struct hsi_client_driver,\
+ driver)
+
+int hsi_register_client_driver(struct hsi_client_driver *drv);
+
+static inline void hsi_unregister_client_driver(struct hsi_client_driver *drv)
+{
+ driver_unregister(&drv->driver);
+}
+
+/**
+ * struct hsi_msg - HSI message descriptor
+ * @link: Free to use by the current descriptor owner
+ * @cl: HSI device client that issues the transfer
+ * @sgt: Head of the scatterlist array
+ * @context: Client context data associated to the transfer
+ * @complete: Transfer completion callback
+ * @destructor: Destructor to free resources when flushing
+ * @status: Status of the transfer when completed
+ * @actual_len: Actual length of data transfered on completion
+ * @channel: Channel were to TX/RX the message
+ * @ttype: Transfer type (TX if set, RX otherwise)
+ * @break_frame: if true HSI will send/receive a break frame (FRAME MODE)
+ */
+struct hsi_msg {
+ struct list_head link;
+ struct hsi_client *cl;
+ struct sg_table sgt;
+ void *context;
+
+ void (*complete)(struct hsi_msg *msg);
+ void (*destructor)(struct hsi_msg *msg);
+
+ int status;
+ unsigned int actual_len;
+ unsigned int channel;
+ unsigned int ttype:1;
+ unsigned int break_frame:1;
+};
+
+struct hsi_msg *hsi_alloc_msg(unsigned int n_frag, gfp_t flags);
+void hsi_free_msg(struct hsi_msg *msg);
+
+/**
+ * struct hsi_port - HSI port device
+ * @device: Driver model representation of the device
+ * @tx_cfg: Current TX path configuration
+ * @rx_cfg: Current RX path configuration
+ * @num: Port number
+ * @shared: Set when port can be shared by different clients
+ * @claimed: Reference count of clients which claimed the port
+ * @lock: Serialize port claim
+ * @async: Asynchronous transfer callback
+ * @setup: Callback to set the HSI client configuration
+ * @flush: Callback to clean the HW state and destroy all pending transfers
+ * @start_tx: Callback to inform that a client wants to TX data
+ * @stop_tx: Callback to inform that a client no longer wishes to TX data
+ * @release: Callback to inform that a client no longer uses the port
+ * @clients: List of hsi_clients using the port.
+ * @clock: Lock to serialize access to the clients list.
+ */
+struct hsi_port {
+ struct device device;
+ struct hsi_config tx_cfg;
+ struct hsi_config rx_cfg;
+ unsigned int num;
+ unsigned int shared:1;
+ int claimed;
+ struct mutex lock;
+ int (*async)(struct hsi_msg *msg);
+ int (*setup)(struct hsi_client *cl);
+ int (*flush)(struct hsi_client *cl);
+ int (*start_tx)(struct hsi_client *cl);
+ int (*stop_tx)(struct hsi_client *cl);
+ int (*release)(struct hsi_client *cl);
+ struct list_head clients;
+ spinlock_t clock;
+};
+
+#define to_hsi_port(dev) container_of(dev, struct hsi_port, device)
+#define hsi_get_port(cl) to_hsi_port((cl)->device.parent)
+
+void hsi_event(struct hsi_port *port, unsigned int event);
+int hsi_claim_port(struct hsi_client *cl, unsigned int share);
+void hsi_release_port(struct hsi_client *cl);
+
+static inline int hsi_port_claimed(struct hsi_client *cl)
+{
+ return cl->pclaimed;
+}
+
+static inline void hsi_port_set_drvdata(struct hsi_port *port, void *data)
+{
+ dev_set_drvdata(&port->device, data);
+}
+
+static inline void *hsi_port_drvdata(struct hsi_port *port)
+{
+ return dev_get_drvdata(&port->device);
+}
+
+/**
+ * struct hsi_controller - HSI controller device
+ * @device: Driver model representation of the device
+ * @owner: Pointer to the module owning the controller
+ * @id: HSI controller ID
+ * @num_ports: Number of ports in the HSI controller
+ * @port: Array of HSI ports
+ */
+struct hsi_controller {
+ struct device device;
+ struct module *owner;
+ int id;
+ unsigned int num_ports;
+ struct hsi_port *port;
+};
+
+#define to_hsi_controller(dev) container_of(dev, struct hsi_controller, device)
+
+struct hsi_controller *hsi_alloc_controller(unsigned int n_ports, gfp_t flags);
+void hsi_free_controller(struct hsi_controller *hsi);
+int hsi_register_controller(struct hsi_controller *hsi);
+void hsi_unregister_controller(struct hsi_controller *hsi);
+
+static inline void hsi_controller_set_drvdata(struct hsi_controller *hsi,
+ void *data)
+{
+ dev_set_drvdata(&hsi->device, data);
+}
+
+static inline void *hsi_controller_drvdata(struct hsi_controller *hsi)
+{
+ return dev_get_drvdata(&hsi->device);
+}
+
+static inline struct hsi_port *hsi_find_port_num(struct hsi_controller *hsi,
+ unsigned int num)
+{
+ return (num < hsi->num_ports) ? &hsi->port[num] : NULL;
+}
+
+/*
+ * API for HSI clients
+ */
+int hsi_async(struct hsi_client *cl, struct hsi_msg *msg);
+
+/**
+ * hsi_setup - Configure the client's port
+ * @cl: Pointer to the HSI client
+ *
+ * When sharing ports, clients should either relay on a single
+ * client setup or have the same setup for all of them.
+ *
+ * Return -errno on failure, 0 on success
+ */
+static inline int hsi_setup(struct hsi_client *cl)
+{
+ if (!hsi_port_claimed(cl))
+ return -EACCES;
+ return hsi_get_port(cl)->setup(cl);
+}
+
+/**
+ * hsi_flush - Flush all pending transactions on the client's port
+ * @cl: Pointer to the HSI client
+ *
+ * This function will destroy all pending hsi_msg in the port and reset
+ * the HW port so it is ready to receive and transmit from a clean state.
+ *
+ * Return -errno on failure, 0 on success
+ */
+static inline int hsi_flush(struct hsi_client *cl)
+{
+ if (!hsi_port_claimed(cl))
+ return -EACCES;
+ return hsi_get_port(cl)->flush(cl);
+}
+
+/**
+ * hsi_async_read - Submit a read transfer
+ * @cl: Pointer to the HSI client
+ * @msg: HSI message descriptor of the transfer
+ *
+ * Return -errno on failure, 0 on success
+ */
+static inline int hsi_async_read(struct hsi_client *cl, struct hsi_msg *msg)
+{
+ msg->ttype = HSI_MSG_READ;
+ return hsi_async(cl, msg);
+}
+
+/**
+ * hsi_async_write - Submit a write transfer
+ * @cl: Pointer to the HSI client
+ * @msg: HSI message descriptor of the transfer
+ *
+ * Return -errno on failure, 0 on success
+ */
+static inline int hsi_async_write(struct hsi_client *cl, struct hsi_msg *msg)
+{
+ msg->ttype = HSI_MSG_WRITE;
+ return hsi_async(cl, msg);
+}
+
+/**
+ * hsi_start_tx - Signal the port that the client wants to start a TX
+ * @cl: Pointer to the HSI client
+ *
+ * Return -errno on failure, 0 on success
+ */
+static inline int hsi_start_tx(struct hsi_client *cl)
+{
+ if (!hsi_port_claimed(cl))
+ return -EACCES;
+ return hsi_get_port(cl)->start_tx(cl);
+}
+
+/**
+ * hsi_stop_tx - Signal the port that the client no longer wants to transmit
+ * @cl: Pointer to the HSI client
+ *
+ * Return -errno on failure, 0 on success
+ */
+static inline int hsi_stop_tx(struct hsi_client *cl)
+{
+ if (!hsi_port_claimed(cl))
+ return -EACCES;
+ return hsi_get_port(cl)->stop_tx(cl);
+}
+#endif /* __LINUX_HSI_H__ */
--
1.7.1

2010-12-14 11:57:19

by Alan

[permalink] [raw]
Subject: Re: [RFC PATCHv4 4/7] HSI: hsi_char: Add HSI char device driver

> +#define HSI_CHST_RD(c) ((c)->state & HSI_CHST_RD_MASK)
> +#define HSI_CHST_WR(c) ((c)->state & HSI_CHST_WR_MASK)
> +
> +#define HSI_CHST_OC_SET(c, v) \
> + do { \
> + (c)->state &= ~HSI_CHST_OC_MASK; \
> + (c)->state |= v; \
> + } while (0);
> +
> +#define HSI_CHST_RD_SET(c, v) \
> + do { \
> + (c)->state &= ~HSI_CHST_RD_MASK; \
> + (c)->state |= v; \
> + } while (0);
> +
> +#define HSI_CHST_WR_SET(c, v) \
> + do { \
> + (c)->state &= ~HSI_CHST_WR_MASK; \
> + (c)->state |= v; \
> + } while (0);


These sort of macros just hide detail - eg the lack of internal locking,
which can lead to problems later, plus they reference their arguments
multiple times so may have weird side effects.

They should probably be inline functions so they at least type check and
behave sanely, and their locking rules defintiely need documenting



> +static long hsi_char_ioctl(struct file *file, unsigned int cmd,
> + unsigned long arg)
> +{
> + struct hsi_char_channel *channel = file->private_data;
> + unsigned int state;
> + struct hsi_config cfg;
> + struct hsc_rx_config rx_cfg;
> + struct hsc_tx_config tx_cfg;
> + long ret = 0;
> +
> + if (HSI_CHST_OC(channel) != HSI_CHST_OPENED)
> + return -ENODEV;

-EIO is the traditional response in this case if the channel has been
closed by the other end - ENODEV indicates there is no physical device
present - not sure which is right here from the code.


> + } else if ((state == HSC_PM_ENABLE)
> + && (channel->wlrefcnt > 0)) {
> + ret = hsi_stop_tx(channel->cl);
> + if (!ret)
> + channel->wlrefcnt--;

What locks this lot against races (ditto some of the other ioctl code)

> + refcnt = atomic_inc_return(&cl_data->refcnt);
> + if (refcnt == 1) {

You seem to construct a lot of clever stuff using atomic_inc_return and
friends where it looks like test_and_set_bit - or even a mutex over
open/close would be far easier to understand ?

> + ret = hsi_char_msgs_alloc(channel);
> +
> + if (ret < 0) {
> + refcnt = atomic_dec_return(&cl_data->refcnt);
> + if (!refcnt)
> + hsi_release_port(channel->cl);
> + spin_lock_bh(&channel->lock);
> + HSI_CHST_OC_SET(channel, HSI_CHST_CLOSED);
> + goto out;
> + }
> + if (refcnt == 1)
> + cl_data->attached = 1;

What happens here if a second open passes the first, the attached will
stay zero and other stuff will be in strange states but the open flag
will be set ?


> + for (i = 0; i < HSI_CHAR_DEVS && channels_map[i] >= 0; i++) {
> + if (channels_map[i] >= HSI_CHAR_DEVS) {
> + pr_err("Invalid HSI/SSI channel specified");
> + return -EINVAL;
> + }
> + set_bit(channels_map[i], &ch_mask);

How will this integrate with hot discovery of SSI supporting devices ?

2010-12-14 17:24:07

by Carlos Chinea

[permalink] [raw]
Subject: Re: [RFC PATCHv4 4/7] HSI: hsi_char: Add HSI char device driver

On Tue, 2010-12-14 at 11:56 +0000, ext Alan Cox wrote:
> > +#define HSI_CHST_RD(c) ((c)->state & HSI_CHST_RD_MASK)
> > +#define HSI_CHST_WR(c) ((c)->state & HSI_CHST_WR_MASK)
> > +
> > +#define HSI_CHST_OC_SET(c, v) \
> > + do { \
> > + (c)->state &= ~HSI_CHST_OC_MASK; \
> > + (c)->state |= v; \
> > + } while (0);
> > +
> > +#define HSI_CHST_RD_SET(c, v) \
> > + do { \
> > + (c)->state &= ~HSI_CHST_RD_MASK; \
> > + (c)->state |= v; \
> > + } while (0);
> > +
> > +#define HSI_CHST_WR_SET(c, v) \
> > + do { \
> > + (c)->state &= ~HSI_CHST_WR_MASK; \
> > + (c)->state |= v; \
> > + } while (0);
>
>
> These sort of macros just hide detail - eg the lack of internal locking,
> which can lead to problems later, plus they reference their arguments
> multiple times so may have weird side effects.
>
> They should probably be inline functions so they at least type check and
> behave sanely, and their locking rules defintiely need documenting

Agree. We would change that.

>
>
>
> > +static long hsi_char_ioctl(struct file *file, unsigned int cmd,
> > + unsigned long arg)
> > +{
> > + struct hsi_char_channel *channel = file->private_data;
> > + unsigned int state;
> > + struct hsi_config cfg;
> > + struct hsc_rx_config rx_cfg;
> > + struct hsc_tx_config tx_cfg;
> > + long ret = 0;
> > +
> > + if (HSI_CHST_OC(channel) != HSI_CHST_OPENED)
> > + return -ENODEV;
>
> -EIO is the traditional response in this case if the channel has been
> closed by the other end - ENODEV indicates there is no physical device
> present - not sure which is right here from the code.
>

The initial idea of this code was to guard against underneath removal of
the device. That was the reason for the -ENODEV.

Now after, rechecking the code I think that we need also to lock the
device to avoid that the removal happening in the middle of the
function.

>
> > + } else if ((state == HSC_PM_ENABLE)
> > + && (channel->wlrefcnt > 0)) {
> > + ret = hsi_stop_tx(channel->cl);
> > + if (!ret)
> > + channel->wlrefcnt--;
>
> What locks this lot against races (ditto some of the other ioctl code)

Right. We will fix this also.

>
> > + refcnt = atomic_inc_return(&cl_data->refcnt);
> > + if (refcnt == 1) {
>
> You seem to construct a lot of clever stuff using atomic_inc_return and
> friends where it looks like test_and_set_bit - or even a mutex over
> open/close would be far easier to understand ?

Yep a mutex will be a better option here. It would also take care of the
next comment.

>
> > + ret = hsi_char_msgs_alloc(channel);
> > +
> > + if (ret < 0) {
> > + refcnt = atomic_dec_return(&cl_data->refcnt);
> > + if (!refcnt)
> > + hsi_release_port(channel->cl);
> > + spin_lock_bh(&channel->lock);
> > + HSI_CHST_OC_SET(channel, HSI_CHST_CLOSED);
> > + goto out;
> > + }
> > + if (refcnt == 1)
> > + cl_data->attached = 1;
>
> What happens here if a second open passes the first, the attached will
> stay zero and other stuff will be in strange states but the open flag
> will be set ?
>
>
> > + for (i = 0; i < HSI_CHAR_DEVS && channels_map[i] >= 0; i++) {
> > + if (channels_map[i] >= HSI_CHAR_DEVS) {
> > + pr_err("Invalid HSI/SSI channel specified");
> > + return -EINVAL;
> > + }
> > + set_bit(channels_map[i], &ch_mask);
>
> How will this integrate with hot discovery of SSI supporting devices ?
>

For every SSI port there is an HSI char client device, which is the
device received through probe. The driver will export for every HSI char
client device, HSI_CHAR_DEVS char devices, one char device per channel.

The intention of this piece of code is to allow the user to configure
which channels the user wants to be available to userspace.

Thanks for the comments,
Carlos

2010-12-18 00:30:29

by Tony Lindgren

[permalink] [raw]
Subject: Re: [RFC PATCHv4 2/7] HSI: omap_ssi: Introducing OMAP SSI driver

* Carlos Chinea <[email protected]> [101214 02:13]:
> Introduces the OMAP SSI driver in the kernel.
>
> The Synchronous Serial Interface (SSI) is a legacy version
> of HSI. As in the case of HSI, it is mainly used to connect
> Application engines (APE) with cellular modem engines (CMT)
> in cellular handsets.
>
> It provides a multichannel, full-duplex, multi-core communication
> with no reference clock. The OMAP SSI block is capable of reaching
> speeds of 110 Mbit/s.
...

> +static int __init ssi_init(void)
> +{
> + return platform_device_register(&ssi_pdev);
> +}
> +subsys_initcall(ssi_init);

Is this safe to do for all omaps2+ machines?

Regards,

Tony

2010-12-20 16:38:10

by Carlos Chinea

[permalink] [raw]
Subject: Re: [RFC PATCHv4 2/7] HSI: omap_ssi: Introducing OMAP SSI driver

Hi Tony,

On Fri, 2010-12-17 at 16:30 -0800, ext Tony Lindgren wrote:
> * Carlos Chinea <[email protected]> [101214 02:13]:
> > Introduces the OMAP SSI driver in the kernel.
> >
> > The Synchronous Serial Interface (SSI) is a legacy version
> > of HSI. As in the case of HSI, it is mainly used to connect
> > Application engines (APE) with cellular modem engines (CMT)
> > in cellular handsets.
> >
> > It provides a multichannel, full-duplex, multi-core communication
> > with no reference clock. The OMAP SSI block is capable of reaching
> > speeds of 110 Mbit/s.
> ...
>
> > +static int __init ssi_init(void)
> > +{
> > + return platform_device_register(&ssi_pdev);
> > +}
> > +subsys_initcall(ssi_init);
>
> Is this safe to do for all omaps2+ machines?
>

Well, I think it is safe. SSI will be only built in, if it is explicitly
configured for that specific machine. And AFAIK, at susbsys initcall
time all other initializations that the OMAP SSI driver may need, will
be already in place.

But I may missed something here ?

Br,
--
Carlos Chinea <[email protected]>

2011-03-21 08:16:45

by Linus Walleij

[permalink] [raw]
Subject: Re: [RFC PATCHv4 0/7] HSI framework and drivers

2010/12/14 Carlos Chinea <[email protected]>:

> Here you have the fourth round of the HSI framework patches.

I'd like to notify that this out-of-tree framework is also in use for
the ST-Ericsson
Ux500s, and we've had great success with it. You can add my:

Acked-by: Linus Walleij <[email protected]>

Possibly some ST-Ericsson people can add Tested-by tags as well.

We will integrate Ux500 support whenever it's suitable, either after the merge
or as add-on patches for Carlos patch series. (If Carlos would like to keep
them in his series.)

Yours,
Linus Walleij

2011-03-22 09:51:46

by Carlos Chinea

[permalink] [raw]
Subject: Re: [RFC PATCHv4 0/7] HSI framework and drivers

Hi,

On Mon, 2011-03-21 at 09:16 +0100, ext Linus Walleij wrote:
> 2010/12/14 Carlos Chinea <[email protected]>:
>
> > Here you have the fourth round of the HSI framework patches.
>
> I'd like to notify that this out-of-tree framework is also in use for
> the ST-Ericsson
> Ux500s, and we've had great success with it. You can add my:
>
> Acked-by: Linus Walleij <[email protected]>
>

Thanks I will add the Ack to the next round of patches.

We are currently working on the hsi_char module. Apart from addressing
the comments from Alan, we are planning to drop also the non-blocking
support from it. The current version is not fully honoring the
non-blocking behavior.

> Possibly some ST-Ericsson people can add Tested-by tags as well.
>
> We will integrate Ux500 support whenever it's suitable, either after the merge
> or as add-on patches for Carlos patch series. (If Carlos would like to keep
> them in his series.)
>

Either way is fine, for me. Could you send the patches for review ?

Br,
Carlos
--
Carlos Chinea <[email protected]>

2011-05-12 12:17:53

by Govindraj

[permalink] [raw]
Subject: Re: [RFC PATCHv4 0/7] HSI framework and drivers

On Tue, Mar 22, 2011 at 3:20 PM, Carlos Chinea <[email protected]> wrote:
> Hi,
>
> On Mon, 2011-03-21 at 09:16 +0100, ext Linus Walleij wrote:
>> 2010/12/14 Carlos Chinea <[email protected]>:
>>
>> > Here you have the fourth round of the HSI framework patches.
>>
>> I'd like to notify that this out-of-tree framework is also in use for
>> the ST-Ericsson
>> Ux500s, and we've had great success with it. You can add my:
>>
>> Acked-by: Linus Walleij <[email protected]>
>>
>

Hi Carlos,

Will you posting a next version of HSI framework.

I am planning to adapt omap-hsi driver to this framework.

--
Thanks,
Govindraj.R


> Thanks I will add the Ack to the next round of patches.
>
> We are currently working on the hsi_char module. Apart from addressing
> the comments from Alan, ?we are planning to drop also the non-blocking
> support from it. The current version is not fully honoring the
> non-blocking behavior.
>
>> Possibly some ST-Ericsson people can add Tested-by tags as well.
>>
>> We will integrate Ux500 support whenever it's suitable, either after the merge
>> or as add-on patches for Carlos patch series. (If Carlos would like to keep
>> them in his series.)
>>
>
> Either way is fine, for me. Could you send the patches for review ?
>
> Br,
> Carlos
> --
> Carlos Chinea <[email protected]>
>
> --
> To unsubscribe from this list: send the line "unsubscribe linux-omap" in
> the body of a message to [email protected]
> More majordomo info at ?http://vger.kernel.org/majordomo-info.html
>

2011-05-16 12:04:12

by Linus Walleij

[permalink] [raw]
Subject: Re: [RFC PATCHv4 0/7] HSI framework and drivers

On Thu, May 12, 2011 at 2:17 PM, Govindraj <[email protected]> wrote:
> On Tue, Mar 22, 2011 at 3:20 PM, Carlos Chinea <[email protected]> wrote:
>> Hi,
>>
>> On Mon, 2011-03-21 at 09:16 +0100, ext Linus Walleij wrote:
>>> 2010/12/14 Carlos Chinea <[email protected]>:
>>>
>>> > Here you have the fourth round of the HSI framework patches.
>>>
>>> I'd like to notify that this out-of-tree framework is also in use for
>>> the ST-Ericsson
>>> Ux500s, and we've had great success with it. You can add my:
>>>
>>> Acked-by: Linus Walleij <[email protected]>
>>>
>>
>
> Hi Carlos,
>
> Will you posting a next version of HSI framework.
>
> I am planning to adapt omap-hsi driver to this framework.

I was sort of asking the same thing.

I expected this to be queued up for 2.6.40 so we don't proliferate
out-of-tree drivers :-(

If Nokia will not pursue this patch set then tell me and Govindraj
and we can surely take it over.

Linus Walleij

2011-05-17 08:42:26

by Carlos Chinea

[permalink] [raw]
Subject: Re: [RFC PATCHv4 0/7] HSI framework and drivers

On Mon, 2011-05-16 at 14:04 +0200, ext Linus Walleij wrote:
> On Thu, May 12, 2011 at 2:17 PM, Govindraj <[email protected]> wrote:
> > On Tue, Mar 22, 2011 at 3:20 PM, Carlos Chinea <[email protected]> wrote:
> >> Hi,
> >>
> >> On Mon, 2011-03-21 at 09:16 +0100, ext Linus Walleij wrote:
> >>> 2010/12/14 Carlos Chinea <[email protected]>:
> >>>
> >>> > Here you have the fourth round of the HSI framework patches.
> >>>
> >>> I'd like to notify that this out-of-tree framework is also in use for
> >>> the ST-Ericsson
> >>> Ux500s, and we've had great success with it. You can add my:
> >>>
> >>> Acked-by: Linus Walleij <[email protected]>
> >>>
> >>
> >
> > Hi Carlos,
> >
> > Will you posting a next version of HSI framework.
> >
> > I am planning to adapt omap-hsi driver to this framework.
>
> I was sort of asking the same thing.

> I expected this to be queued up for 2.6.40 so we don't proliferate
> out-of-tree drivers :-(
>
> If Nokia will not pursue this patch set then tell me and Govindraj
> and we can surely take it over.
>

We are still committed to these patches. We are aiming to release a new
version in 3 weeks max, sooner if we can.

I am very sorry for the delay and I agree that starting to have several
out-of-tree drivers is not acceptable. I'll try my best to fix this
asap.

Br,
--
Carlos Chinea <[email protected]>

2011-06-07 15:26:54

by Sebastian Reichel

[permalink] [raw]
Subject: Re: [RFC PATCHv4 0/7] HSI framework and drivers

On Tue, May 17, 2011 at 11:39:13AM +0300, Carlos Chinea wrote:
> We are still committed to these patches. We are aiming to release a new
> version in 3 weeks max, sooner if we can.
>
> I am very sorry for the delay and I agree that starting to have several
> out-of-tree drivers is not acceptable. I'll try my best to fix this
> asap.

Do you also plan to add the platform code for rx51 to the mainline
kernel?

P.S.: 3 weeks are over today :)

-- Sebastian


Attachments:
(No filename) (459.00 B)
signature.asc (836.00 B)
Digital signature
Download all attachments

2011-06-08 07:16:45

by Carlos Chinea

[permalink] [raw]
Subject: Re: [RFC PATCHv4 0/7] HSI framework and drivers

Hi,

On Tue, 2011-06-07 at 17:09 +0200, ext Sebastian Reichel wrote:
> On Tue, May 17, 2011 at 11:39:13AM +0300, Carlos Chinea wrote:
> > We are still committed to these patches. We are aiming to release a new
> > version in 3 weeks max, sooner if we can.
> >
> > I am very sorry for the delay and I agree that starting to have several
> > out-of-tree drivers is not acceptable. I'll try my best to fix this
> > asap.
>
> Do you also plan to add the platform code for rx51 to the mainline
> kernel?

Not yet in this patch set. If the HSI framework makes it into the
kernel, then I will try to push a couple of new HSI clients needed for
rx51 and the platform related code.

>
> P.S.: 3 weeks are over today :)
>

Yep, we are almost there with the patches :P A small problem in a last
minute test is keeping us from releasing the patches already.

Br,
--
Carlos Chinea <[email protected]>