Skip to contentSkip to conversation

drivers:staging: sources for ST core

PATCH Pavan Savoy 3 participants 4 messages Apr 1, 2010 – Apr 5, 2010 Staging

lore
Share & tools
lore
#0 lore raw
Quoted 50 lines · Alan Cox

From: Alan Cox <[email protected]>
Subject: Re: [PATCH] drivers:staging: sources for ST core
To: [email protected]
Cc: "Greg KH" <[email protected]>, "Marcel Holtmann" <[email protected]>, [email protected]
Date: Thursday, 1 April, 2010, 2:50 PM

Quoted 2 lines

+/*
+ * function to return whether the firmware response

was proper

Quoted 1 line

+ * in case of error don't complete so that waiting

for proper

Review

Quoted 5 lines

+ * response times out
+ */

(quoted diff) +3 -0
+void validate_firmware_response(struct sk_buff *skb)
+{
+    if (unlikely(skb->data[5] !=

0)) {

Quoted 1 line

+        pr_err("no

proper response during fw download");

Quoted 1 line

+        pr_err("data6

%x", skb->data[5]);

In this driver you do know the device so you need to be using dev_ and
passing around dev (or something that gives you dev).

Review

Quoted 5 lines
(quoted diff) +5 -0
+static int kim_probe(struct platform_device *pdev)
+{
+    long status;
+    long proto;
+    long *gpios =

pdev->dev.platform_data;

Quoted 2 lines

+
+    status =

st_core_init(&kim_gdata->core_data);

I would expect any truely global data to be configured in the module init
and then device specific data you want to do something like this

    kim_data = kzalloc(sizeof(something),
GFP_KERNEL);

    ..

    kim_data_init(&pdev->dev,
kim_data);
    dev_set_drvdata(&pdev->dev,
kim_data);

Quoted 6 lines

Elsewhere you can now do

    kim_data =
dev_get_drvdata(&pdev->dev);

to get it back

There are 2 sets of data structures here (after removing the un-necessary 3rd one),
1. st_gdata - which I would want to tie to tty
2. kim_gdata - which I "would" like to tie to the pdev.

Now the problem being, I have reference of st_gdata in kim_gdata, and there are about 4 functions in the st_core where I would need the st_gdata, as in,

EXPORTED symbol st_register - because I need to add in entries as to who registered, +long st_register(struct st_proto_s *new_proto)
+{
+    struct st_data_s    *st_gdata;
+    long err = ST_SUCCESS;
+    unsigned long flags = 0;
+
+    st_kim_ref(&st_gdata);
+    pr_info("%s(%d) ", __func__, new_proto->type);
+    if (st_gdata == NULL || new_proto == NULL || new_proto->recv == NULL +        || new_proto->reg_complete_cb == NULL) {
+        pr_err("gdata/new_proto/recv or reg_complete_cb not ready"); +        return ST_ERR_FAILURE;
+    }
Also in st_unregister and st_write for similar purposes,
But I also need it in tty_open, to link it to the disc_data,

+static int st_tty_open(struct tty_struct *tty)
+{
+    int err = ST_SUCCESS;
+    struct st_data_s *st_gdata;
+    pr_info("%s ", __func__);
+
+    st_kim_ref(&st_gdata);
+    st_gdata->tty = tty;
+    tty->disc_data = st_gdata;

So, shouldn't some function like st_kim_ref be enough ?

+void st_kim_ref(struct st_data_s **core_data)
+{
+    *core_data = kim_gdata->core_data;
+}

So Now st_gdata is tied to tty, and kim_gdata being purely global, as in only 1 platform device of such kind can exist.

Suppose 2 platform devices want to exist - then who's EXPORT of st_register is considered ?
And If bluetooth or FM wants to use this transport then, how would it tell onto which platform device it wants to attach to ?

Why should I tie kim_gdata to a pdev ?

Quoted 7 lines

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to [email protected]
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/

The INTERNET now has a personality. YOURS! See your Yahoo! Homepage. http://in.yahoo.com/

#1 lore raw
Quoted 126 lines · Pavan Savoy

From: Pavan Savoy <[email protected]>
Subject: Re: [PATCH] drivers:staging: sources for ST core
To: "Alan Cox" <[email protected]>
Cc: "Greg KH" <[email protected]>, "Marcel Holtmann" <[email protected]>, [email protected]
Date: Thursday, 1 April, 2010, 10:50 PM
--- On Thu, 1/4/10, Alan Cox <[email protected]>

Quoted 2 lines · wrote:

From: Alan Cox <[email protected]>
Subject: Re: [PATCH] drivers:staging: sources for ST

core

Quoted 2 lines

To: [email protected]
Cc: "Greg KH" <[email protected]>,

"Marcel Holtmann" <[email protected]>,
[email protected]

Quoted 3 lines

Date: Thursday, 1 April, 2010, 2:50 PM

Quoted 2 lines

+/*
+ * function to return whether the firmware

response

Quoted 2 lines

was proper

Quoted 1 line

+ * in case of error don't complete so that

waiting

Quoted 4 lines

for proper

Quoted 3 lines

+ * response times out
+ */
+void validate_firmware_response(struct sk_buff

*skb)

Quoted 9 lines
Quoted 2 lines

+{
+    if (unlikely(skb->data[5] !=

0)) {

Quoted 1 line

+        pr_err("no

proper response during fw download");

Quoted 1 line

+        pr_err("data6

%x", skb->data[5]);

In this driver you do know the device so you need to

be

Quoted 4 lines

using dev_ and
passing around dev (or something that gives you dev).

Quoted 1 line

+static int kim_probe(struct platform_device

*pdev)

Quoted 10 lines
Quoted 4 lines

+{
+    long status;
+    long proto;
+    long *gpios =

pdev->dev.platform_data;

Quoted 2 lines

+
+    status =

st_core_init(&kim_gdata->core_data);

I would expect any truely global data to be configured

in

Quoted 2 lines

the module init
and then device specific data you want to do something

like

Quoted 12 lines

this

    kim_data = kzalloc(sizeof(something),
GFP_KERNEL);

    ..

    kim_data_init(&pdev->dev,
kim_data);
    dev_set_drvdata(&pdev->dev,
kim_data);

Quoted 6 lines

Elsewhere you can now do

    kim_data =
dev_get_drvdata(&pdev->dev);

to get it back

There are 2 sets of data structures here (after removing the un-necessary 3rd one),
1. st_gdata - which I would want to tie to tty
2. kim_gdata - which I "would" like to tie to the pdev.

Now the problem being, I have reference of st_gdata in kim_gdata, and there are about 4 functions in the st_core where I would need the st_gdata, as in,

EXPORTED symbol st_register - because I need to add in entries as to who registered,

(quoted diff) +9 -0
+long st_register(struct st_proto_s *new_proto)
+{
+    struct st_data_s    *st_gdata;
+    long err = ST_SUCCESS;
+    unsigned long flags = 0;
+
+    st_kim_ref(&st_gdata);
+    pr_info("%s(%d) ", __func__, new_proto->type);
+    if (st_gdata == NULL || new_proto == NULL ||

new_proto->recv == NULL
+        || new_proto->reg_complete_cb == NULL) { +        pr_err("gdata/new_proto/recv or reg_complete_cb not ready");
+        return ST_ERR_FAILURE;
+    }
Also in st_unregister and st_write for similar purposes, But I also need it in tty_open, to link it to the disc_data,

(quoted diff) +9 -0
+static int st_tty_open(struct tty_struct *tty)
+{
+    int err = ST_SUCCESS;
+    struct st_data_s *st_gdata;
+    pr_info("%s ", __func__);
+
+    st_kim_ref(&st_gdata);
+    st_gdata->tty = tty;
+    tty->disc_data = st_gdata;

So, shouldn't some function like st_kim_ref be enough ?

(quoted diff) +4 -0
+void st_kim_ref(struct st_data_s **core_data)
+{
+    *core_data = kim_gdata->core_data;
+}

So Now st_gdata is tied to tty, and kim_gdata being purely global, as in only 1 platform device of such kind can exist.

Quoted 7 lines

Suppose 2 platform devices want to exist - then who's EXPORT of st_register is considered ?
And If bluetooth or FM wants to use this transport then, how would it tell onto which platform device it wants to attach to ?

Why should I tie kim_gdata to a pdev ?

Well, because of the comments on the architecture of this driver, I tried out the bus_ driver method, where I register a new bus type as ST (which also registers the N_TI_WL line discipline) and the different protocols on it, are registered to as devices with the N_TI_WL as bus type.

However, I ended up in some similar mess there on too, where during an st_register when I need to register devices to the bus, status-es like how many devices currently registered, and allowing other devices to register during firmware download all would then on require a data which isn't bound to any device ...

So any ideas ? What should be done ? Find below the test patch...

---
 drivers/staging/ti-st/bus_drv/bt_test_bus.c |   55 +++++++++++
 drivers/staging/ti-st/bus_drv/fm_test_bus.c |   55 +++++++++++
 drivers/staging/ti-st/bus_drv/test_bus.c    |  134 +++++++++++++++++++++++++++
 3 files changed, 244 insertions(+), 0 deletions(-)
 create mode 100644 drivers/staging/ti-st/bus_drv/bt_test_bus.c
 create mode 100644 drivers/staging/ti-st/bus_drv/fm_test_bus.c
 create mode 100644 drivers/staging/ti-st/bus_drv/test_bus.c
drivers/staging/ti-st/bus_drv/bt_test_bus.c new file +55 -0
new file mode 100644
index 0000000..02ae812
@@ -0,0 +1,55 @@
+#include <linux/module.h>
+#include <linux/kernel.h>
+#include <linux/init.h>
+#include <linux/device.h>
+
+#include "st_bus.h"
+
+static int __init st_user_probe(struct device *dev, int cur_bus, int cur_slot)
+{
+	pr_info("%s\n", __func__);
+	return 0;
+}
+static int __exit st_user_remove(struct device *dev, int cur_bus, int cur_slot)
+{
+	pr_info("%s\n", __func__);
+	return 0;
+}
+
+void bt_release(struct device *dev)
+{
+	pr_info("%s\n", __func__);
+}
+
+static struct st_driver st_user_driver = {
+	.dev = {
+		.release = bt_release,
+	},
+	.id = ST_BT,
+	.name = "bluetooth",
+};
+
+static int __init st_user_init(void)
+{
+	int retval;
+	retval = st_register(&st_user_driver);
+	if (retval != 0)
+		pr_err("%s: error registering driver\n", __func__);
+
+	return retval;
+
+}
+
+static void __exit st_user_exit(void)
+{
+	pr_info("%s\n", __func__);
+	st_unregister(&st_user_driver);
+}
+
+
+
+MODULE_LICENSE("GPL");
+
+module_init(st_user_init);
+module_exit(st_user_exit);
+
drivers/staging/ti-st/bus_drv/fm_test_bus.c new file +55 -0

55 lines not on this page · raw message

drivers/staging/ti-st/bus_drv/test_bus.c new file +37 -0
new file mode 100644
index 0000000..87aaa94
@@ -0,0 +1,134 @@
+#include <linux/module.h>
+#include <linux/kernel.h>
+#include <linux/init.h>
+#include <linux/device.h>
+#include <linux/platform_device.h>
+
+#include "st_bus.h"
+#include "st_core.h"
+
+struct bus_type st_bus_type;
+
+static int check_if_registered(struct device *dev, void *data)
+{
+	struct st_driver *drv;
+
+	drv = dev_get_drvdata(dev);
+	pr_info("device is %s, type %d\n", drv->name, drv->id);
+	return 1;
+}
+
+int st_register(struct st_driver *drv)
+{
+	int retval;
+	struct device *dev = &drv->dev;
+
+	if (bus_find_device_by_name(&st_bus_type, NULL,
+				drv->name) != NULL) {
+		pr_err("already registered\n");
+	}
+
+	retval = bus_for_each_dev(&st_bus_type, NULL, NULL,
+		check_if_registered);
+	if (retval != 1) { /* as returned by check_if_registered */
+		/* download_firmware here */
+		pr_info("downloading firmware..\n");
+		/* HOWTO carry on with device_register ? */
+	}

This is 1 place, I would be totally lost because, I should return the flow back to the driver which called the st_register, and allow other driver to do an st_register and return them as pending - and when firmware download completes (takes around 10secs - worst case), I then have to notify all the drivers(devices) registered about the completion of firmware download.

+ dev->bus = &(st_bus_type);
+ dev_set_name(dev, drv->name);
+ dev_set_drvdata(dev, drv);
+
+ /* add device here */
+ retval = device_register(dev);
+ if (retval) {
+ pr_err("device register failed\n");
+ return -1;
+ }
+
+ pr_info("%s: %s\n", __func__, drv->name);
+ return 0;
+}
+EXPORT_SYMBOL(st_register);
+
+void st_unregister(struct st_driver *drv)
+{
+ pr_info("%s\n", __func__);
+ device_unregister(&drv->dev);
+}
+EXPORT_SYMBOL(st_unregister);
+
+
+static int st_bus_probe(struct platform_device *pdev)
+{
+ struct st_data_s *core_data;
+ pr_info("%s\n", __func__);
+ bus_register(&st_bus_type);
+
+ st_core_init(&core_data);
+ dev_set_drvdata(&pdev->dev, core_data);
+ return 0;
+}
+
+static int st_bus_remove(struct platform_device *pdev)
+{
+ struct st_data_s *core_data;
+ pr_info("%s\n", __func__);
+
+ core_data = dev_get_drvdata(&pdev->dev);
+ st_core_exit(core_data);
+ bus_unregister(&st_bus_type);
+ return 0;
+}
+
+#if 0
+static int st_bus_match(struct device *dev, struct device_driver *drv) +{
+ return 1;
+}
+
+int st_notify(struct notifier_block *notify, unsigned long a, void *b) +{
+ pr_info("%s \n", __func__);
+ return 0;
+}
+#endif
+
+struct bus_type st_bus_type = {
+ .name = "ti-st",
+};
+EXPORT_SYMBOL(st_bus_type);
+
+struct platform_driver st_bus_driver = {
+ .driver = {
+ .name = "kim",
+ .owner = THIS_MODULE,
+ },
+ .probe = st_bus_probe,
+ .remove = st_bus_remove,
+};
+
+#if 0
+struct notifier_block st_bus_notify = {
+ .notifier_call = st_notify,
+};
+#endif
+
+static int __init test_bus_init(void)
+{
+ platform_driver_register(&st_bus_driver);
+ return 0;
+}
+
+static void __exit test_bus_exit(void)
+{
+ pr_info("%s\n", __func__);
+ platform_driver_unregister(&st_bus_driver);
+}
+
+module_init(test_bus_init);
+module_exit(test_bus_exit);
+MODULE_AUTHOR("Pavan Savoy <[email protected]>"); +MODULE_DESCRIPTION("Shared Transport Driver for TI BT/FM/GPS combo chips "); +MODULE_LICENSE("GPL");

Signature · 1 line
1.5.4.3
Quoted 11 lines
Quoted 2 lines

--
To unsubscribe from this list: send the line

"unsubscribe

Quoted 5 lines

linux-kernel" in
the body of a message to [email protected]
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/

Quoted 9 lines

      The INTERNET now has a personality.
YOURS! See your Yahoo! Homepage. http://in.yahoo.com/
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to [email protected]
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/

Your Mail works best with the New Yahoo Optimized IE8. Get it NOW! http://downloads.yahoo.com/in/internetexplorer/

#2 lore raw

Sorry but I can't make head or tail of this and the code flow you are trying to achieve.

The usual way you do stuff is to put per device stuff in a per device struct, driver wide stuff in a driver struct (or static variables) and then run everything from the device end.

I'd expect a low level driver to do something like

probe()
create device entry
initialise device (eg download firmware) register itself with anything higher level

#3 lore raw
Quoted 18 lines · Alan Cox

From: Alan Cox <[email protected]>
Subject: Re: [PATCH] drivers:staging: sources for ST core
To: "Pavan Savoy" <[email protected]>
Cc: "Greg KH" <[email protected]>, "Marcel Holtmann" <[email protected]>, [email protected]
Date: Friday, 2 April, 2010, 4:57 AM
Sorry but I can't make head or tail
of this and the code flow you are
trying to achieve.

The usual way you do stuff is to put per device stuff in a per device
struct, driver wide stuff in a driver struct (or static variables) and
then run everything from the device end.

I'd expect a low level driver to do something like

Quoted 7 lines

    probe()
        create device entry
        initialise device (eg
download firmware)
        register itself with
anything higher level

What I am trying to achieve is something like this,

HCI-core V4L2-radio Char-device=/dev/tigps for fops
^ ^ ^

 |		|		|
 |		|		|

BT FM GPS [these register themselves to ST]
\ | /
\ | /
\ | /
Shared Transport Ldisc driver
|
TTY Layer <-- UART driver has already registered.

So, when a BT device try and registers itself to ST (shared transport) driver, I need to toggle chip enable line and 'download_firmware', and in case another _register from FM or GPS happens at the same time, I need to signal it pending, and call a callback upon completion of firmware download.

Now which are to be identified as per-device or bus or driver ?
Because when an st_register is called, I need to do some tty_* operations from the ldisc driver itself - i.e I cannot embed tty into any of BT, FM or GPS per-device structures.

I also expose a st_write function, where in any of BT, FM and GPS driver upon being ready (fw download completed) sends across an SKB, which I queue up and write.

All of what I wanted to do, could not be done when I tried ST ldisc as a bus and each of BT, FM and GPS as per-device structs registering with type ST as bus. So again, Now which are to be identified as per-device or bus or driver ?

The INTERNET now has a personality. YOURS! See your Yahoo! Homepage. http://in.yahoo.com/