2019-11-14 14:18:00

by Asmaa Mnebhi

[permalink] [raw]
Subject: RE: [PATCH v4] drivers: ipmi: Support raw i2c packet in IPMB

Hi Vijay,

You haven't addressed some of my comments.

-----Original Message-----
From: Vijay Khemka <[email protected]>
Sent: Wednesday, November 13, 2019 6:42 PM
To: Corey Minyard <[email protected]>; Arnd Bergmann <[email protected]>; Greg Kroah-Hartman <[email protected]>; [email protected]; [email protected]
Cc: [email protected]; [email protected]; Asmaa Mnebhi <[email protected]>; [email protected]; [email protected]; [email protected]
Subject: [PATCH v4] drivers: ipmi: Support raw i2c packet in IPMB

Many IPMB devices doesn't support smbus protocol and current driver support only smbus devices. Added support for raw i2c packets.

User can define i2c-protocol in device tree to use i2c raw transfer.

A>> Please fix the description as suggested in previous comments

Signed-off-by: Vijay Khemka <[email protected]>
---
drivers/char/ipmi/ipmb_dev_int.c | 32 ++++++++++++++++++++++++++++++++
1 file changed, 32 insertions(+)

diff --git a/drivers/char/ipmi/ipmb_dev_int.c b/drivers/char/ipmi/ipmb_dev_int.c
index ae3bfba27526..10904bec1de0 100644
--- a/drivers/char/ipmi/ipmb_dev_int.c
+++ b/drivers/char/ipmi/ipmb_dev_int.c
@@ -63,6 +63,7 @@ struct ipmb_dev {
spinlock_t lock;
wait_queue_head_t wait_queue;
struct mutex file_mutex;
+ bool is_i2c_protocol;
};

static inline struct ipmb_dev *to_ipmb_dev(struct file *file) @@ -112,6 +113,25 @@ static ssize_t ipmb_read(struct file *file, char __user *buf, size_t count,
return ret < 0 ? ret : count;
}

+static int ipmb_i2c_write(struct i2c_client *client, u8 *msg) {
+ struct i2c_msg i2c_msg;
+
+ /*
+ * subtract 1 byte (rq_sa) from the length of the msg passed to
+ * raw i2c_transfer
+ */
+ i2c_msg.len = msg[IPMB_MSG_LEN_IDX] - 1;
+
+ /* Assign message to buffer except first 2 bytes (length and address) */
+ i2c_msg.buf = msg + 2;
+
+ i2c_msg.addr = GET_7BIT_ADDR(msg[RQ_SA_8BIT_IDX]);

You can have:
i2c_msg.flags = addr;
addr being an argument of the ipmb_i2c_write function and passed in ipmb_write. We already define it.

+ i2c_msg.flags = client->flags & I2C_CLIENT_PEC;
+
+ return i2c_transfer(client->adapter, &i2c_msg, 1); }
+
static ssize_t ipmb_write(struct file *file, const char __user *buf,
size_t count, loff_t *ppos)
{
@@ -133,6 +153,12 @@ static ssize_t ipmb_write(struct file *file, const char __user *buf,
rq_sa = GET_7BIT_ADDR(msg[RQ_SA_8BIT_IDX]);
netf_rq_lun = msg[NETFN_LUN_IDX];

+ /* Check i2c block transfer vs smbus */
+ if (ipmb_dev->is_i2c_protocol) {
+ ret = ipmb_i2c_write(ipmb_dev->client, msg);
+ return (ret == 1) ? count : ret;
+ }
+
/*
* subtract rq_sa and netf_rq_lun from the length of the msg passed to
* i2c_smbus_xfer
@@ -277,6 +303,7 @@ static int ipmb_probe(struct i2c_client *client,
const struct i2c_device_id *id)
{
struct ipmb_dev *ipmb_dev;
+ struct device_node *np;
int ret;

ipmb_dev = devm_kzalloc(&client->dev, sizeof(*ipmb_dev), @@ -302,6 +329,11 @@ static int ipmb_probe(struct i2c_client *client,
if (ret)
return ret;

+ /* Check if i2c block xmit needs to use instead of smbus */
+ np = client->dev.of_node;
+ if (np && of_get_property(np, "i2c-protocol", NULL))
+ ipmb_dev->is_i2c_protocol = true;

I know Corey said that ACPI is not a priority but many companies (including mine) use ACPI and I would prefer if we implement it. All you need to do is use
device_property_read_u32 function instead of of_get_property:
ret = device_property_read_u32(&client->dev, "i2c-protocol", &ipmb_dev->is_i2c_protocol);
This function does the job for both dt and acpi.

+
ipmb_dev->client = client;
i2c_set_clientdata(client, ipmb_dev);
ret = i2c_slave_register(client, ipmb_slave_cb);
--
2.17.1


2019-11-14 18:07:57

by Vijay Khemka

[permalink] [raw]
Subject: Re: [PATCH v4] drivers: ipmi: Support raw i2c packet in IPMB

Ok I will send another patch with acpi support.

On 11/14/19, 6:16 AM, "Asmaa Mnebhi" <[email protected]> wrote:

Hi Vijay,

You haven't addressed some of my comments.

-----Original Message-----
From: Vijay Khemka <[email protected]>
Sent: Wednesday, November 13, 2019 6:42 PM
To: Corey Minyard <[email protected]>; Arnd Bergmann <[email protected]>; Greg Kroah-Hartman <[email protected]>; [email protected]; [email protected]
Cc: [email protected]; [email protected]; Asmaa Mnebhi <[email protected]>; [email protected]; [email protected]; [email protected]
Subject: [PATCH v4] drivers: ipmi: Support raw i2c packet in IPMB

Many IPMB devices doesn't support smbus protocol and current driver support only smbus devices. Added support for raw i2c packets.

User can define i2c-protocol in device tree to use i2c raw transfer.

A>> Please fix the description as suggested in previous comments

Signed-off-by: Vijay Khemka <[email protected]>
---
drivers/char/ipmi/ipmb_dev_int.c | 32 ++++++++++++++++++++++++++++++++
1 file changed, 32 insertions(+)

diff --git a/drivers/char/ipmi/ipmb_dev_int.c b/drivers/char/ipmi/ipmb_dev_int.c
index ae3bfba27526..10904bec1de0 100644
--- a/drivers/char/ipmi/ipmb_dev_int.c
+++ b/drivers/char/ipmi/ipmb_dev_int.c
@@ -63,6 +63,7 @@ struct ipmb_dev {
spinlock_t lock;
wait_queue_head_t wait_queue;
struct mutex file_mutex;
+ bool is_i2c_protocol;
};

static inline struct ipmb_dev *to_ipmb_dev(struct file *file) @@ -112,6 +113,25 @@ static ssize_t ipmb_read(struct file *file, char __user *buf, size_t count,
return ret < 0 ? ret : count;
}

+static int ipmb_i2c_write(struct i2c_client *client, u8 *msg) {
+ struct i2c_msg i2c_msg;
+
+ /*
+ * subtract 1 byte (rq_sa) from the length of the msg passed to
+ * raw i2c_transfer
+ */
+ i2c_msg.len = msg[IPMB_MSG_LEN_IDX] - 1;
+
+ /* Assign message to buffer except first 2 bytes (length and address) */
+ i2c_msg.buf = msg + 2;
+
+ i2c_msg.addr = GET_7BIT_ADDR(msg[RQ_SA_8BIT_IDX]);

You can have:
i2c_msg.flags = addr;
addr being an argument of the ipmb_i2c_write function and passed in ipmb_write. We already define it.

+ i2c_msg.flags = client->flags & I2C_CLIENT_PEC;
+
+ return i2c_transfer(client->adapter, &i2c_msg, 1); }
+
static ssize_t ipmb_write(struct file *file, const char __user *buf,
size_t count, loff_t *ppos)
{
@@ -133,6 +153,12 @@ static ssize_t ipmb_write(struct file *file, const char __user *buf,
rq_sa = GET_7BIT_ADDR(msg[RQ_SA_8BIT_IDX]);
netf_rq_lun = msg[NETFN_LUN_IDX];

+ /* Check i2c block transfer vs smbus */
+ if (ipmb_dev->is_i2c_protocol) {
+ ret = ipmb_i2c_write(ipmb_dev->client, msg);
+ return (ret == 1) ? count : ret;
+ }
+
/*
* subtract rq_sa and netf_rq_lun from the length of the msg passed to
* i2c_smbus_xfer
@@ -277,6 +303,7 @@ static int ipmb_probe(struct i2c_client *client,
const struct i2c_device_id *id)
{
struct ipmb_dev *ipmb_dev;
+ struct device_node *np;
int ret;

ipmb_dev = devm_kzalloc(&client->dev, sizeof(*ipmb_dev), @@ -302,6 +329,11 @@ static int ipmb_probe(struct i2c_client *client,
if (ret)
return ret;

+ /* Check if i2c block xmit needs to use instead of smbus */
+ np = client->dev.of_node;
+ if (np && of_get_property(np, "i2c-protocol", NULL))
+ ipmb_dev->is_i2c_protocol = true;

I know Corey said that ACPI is not a priority but many companies (including mine) use ACPI and I would prefer if we implement it. All you need to do is use
device_property_read_u32 function instead of of_get_property:
ret = device_property_read_u32(&client->dev, "i2c-protocol", &ipmb_dev->is_i2c_protocol);
This function does the job for both dt and acpi.

+
ipmb_dev->client = client;
i2c_set_clientdata(client, ipmb_dev);
ret = i2c_slave_register(client, ipmb_slave_cb);
--
2.17.1