NACK: [SRU][F:linux-bluefield][PATCH v1 1/1] UBUNTU: SAUCE: Add ipmb_host.c driver

Asmaa Mnebhi asmaa at nvidia.com
Tue Jul 5 13:10:01 UTC 2022


Hi Zach,

This is not how I sent the BugLink and I already discussed this with our IT a while ago. I don't really have control over how outlook controls this.
This is how I sent the link:
"Buglink: https://bugs.launchpad.net/bugs/1980532"

This patch is targeting Elbridge 5.15 kernel, not 5.4 (which already has the ipmb_host.c driver). My understanding is that patches targeting 5.15 should be submitted the same way we did for 5.4? if the label is different for 5.15, please let me know what I should change it to.
We had a meeting couple of weeks ago with a canonical team to go over the new kernel and how to submit patches for it but I don't think this part was mentioned.

Thanks.
Asmaa


-----Original Message-----
From: Zachary Tahenakos <zachary.tahenakos at canonical.com> 
Sent: Tuesday, July 5, 2022 8:56 AM
To: Asmaa Mnebhi <asmaa at nvidia.com>; kernel-team at lists.ubuntu.com
Subject: NACK: [SRU][F:linux-bluefield][PATCH v1 1/1] UBUNTU: SAUCE: Add ipmb_host.c driver

[You don't often get email from zachary.tahenakos at canonical.com. Learn why this is important at https://aka.ms/LearnAboutSenderIdentification ]

This already appears to be in f:bluefield, perhaps the wrong series?
Also, the buglink is private.

-Zack

On 7/1/22 11:11 AM, Asmaa Mnebhi wrote:
> BugLink: 
> https://nam11.safelinks.protection.outlook.com/?url=https%3A%2F%2Fbugs
> .launchpad.net%2Fbugs%2F1980532&data=05%7C01%7Casmaa%40nvidia.com%
> 7Ca19769bd9541412819fe08da5e85c30d%7C43083d15727340c1b7db39efd9ccc17a%
> 7C0%7C0%7C637926225846070925%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwM
> DAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C3000%7C%7C%7C&
> sdata=sMtFpDuWpO9NJ6KPZqtn90D6CvVGVOE1w3OlzLrvRJ4%3D&reserved=0
>
> The ipmb_host.c driver allows sending IPMB request from the BF DPU to the BMC.
>
> Signed-off-by: Asmaa Mnebhi <asmaa at nvidia.com>
> ---
>   drivers/char/ipmi/Kconfig     |  10 +
>   drivers/char/ipmi/Makefile    |   1 +
>   drivers/char/ipmi/ipmb_host.c | 836 ++++++++++++++++++++++++++++++++++
>   3 files changed, 847 insertions(+)
>   create mode 100644 drivers/char/ipmi/ipmb_host.c
>
> diff --git a/drivers/char/ipmi/Kconfig b/drivers/char/ipmi/Kconfig 
> index 249b31197eea..2dc98a164784 100644
> --- a/drivers/char/ipmi/Kconfig
> +++ b/drivers/char/ipmi/Kconfig
> @@ -168,3 +168,13 @@ config IPMB_DEVICE_INTERFACE
>         Provides a driver for a device (Satellite MC) to
>         receive requests and send responses back to the BMC via
>         the IPMB interface. This module requires I2C support.
> +
> +config IPMB_HOST
> +     tristate 'IPMB host handler'
> +     depends on I2C
> +     depends on I2C_SLAVE
> +     help
> +          Provides a driver for an IPMB interface to a BMC. This driver
> +       is responsible of sending IPMI requests via the IPMB channel
> +       and receiving the response.
> +       This module requires I2C support.
> diff --git a/drivers/char/ipmi/Makefile b/drivers/char/ipmi/Makefile 
> index 84f47d18007f..b5fff7e9695b 100644
> --- a/drivers/char/ipmi/Makefile
> +++ b/drivers/char/ipmi/Makefile
> @@ -29,3 +29,4 @@ obj-$(CONFIG_ASPEED_BT_IPMI_BMC) += bt-bmc.o
>   obj-$(CONFIG_ASPEED_KCS_IPMI_BMC) += kcs_bmc_aspeed.o
>   obj-$(CONFIG_NPCM7XX_KCS_IPMI_BMC) += kcs_bmc_npcm7xx.o
>   obj-$(CONFIG_IPMB_DEVICE_INTERFACE) += ipmb_dev_int.o
> +obj-$(CONFIG_IPMB_HOST) += ipmb_host.o
> diff --git a/drivers/char/ipmi/ipmb_host.c 
> b/drivers/char/ipmi/ipmb_host.c new file mode 100644 index 
> 000000000000..f178a7df0389
> --- /dev/null
> +++ b/drivers/char/ipmi/ipmb_host.c
> @@ -0,0 +1,836 @@
> +// SPDX-License-Identifier: GPL-2.0+
> +
> +/*
> + * Copyright 2020, NVIDIA Corporation. All rights reserved.
> + *
> + * This was inspired by Brendan Higgins' bt-i2c driver.
> + *
> + */
> +
> +#include <linux/delay.h>
> +#include <linux/errno.h>
> +#include <linux/init.h>
> +#include <linux/ipmi_smi.h>
> +#include <linux/i2c.h>
> +#include <linux/jiffies.h>
> +#include <linux/kernel.h>
> +#include <linux/module.h>
> +#include <linux/moduleparam.h>
> +#include <linux/mutex.h>
> +#include <linux/sched/signal.h>
> +#include <linux/slab.h>
> +#include <linux/timer.h>
> +#include <linux/types.h>
> +
> +#define      IPMB_TIMEOUT                    (msecs_to_jiffies(20000))
> +
> +static bool handshake_rsp = false;
> +
> +/*
> + * The least we expect in an IPMB message is:
> + * netfn_rs_lun, checksum1, rq_sa, rq_seq_rq_lun,
> + * cmd and checksum2.
> + */
> +#define      IPMB_LEN_MIN                    6
> +
> +/*
> + * Within the response message, we need at least
> + * netfn_rq_lun, checksum1, rs_sa, rq_seq_rs_lun,
> + * cmd, completion code and checksum2.
> + */
> +#define      IPMB_RESPONSE_LEN_MIN           7
> +
> +#define      IPMB_MSG_PAYLOAD_LEN_MAX        122
> +
> +#define      IPMB_SMI_MSG_PAYLOAD_LEN_MAX    (IPMI_MAX_MSG_LENGTH - 2)
> +#define      IPMB_MAX_SMI_SIZE               125
> +#define      IPMB_SMI_MSG_HEADER_SIZE        2
> +
> +#define      IPMB_SEQ_MAX                    64
> +
> +#define      MAX_BUF_SIZE                    122
> +
> +#define      WRITE_TIMEOUT                   25
> +#define      RSP_QUEUE_MAX_LEN               256
> +
> +#define      NETFN_RSP_BIT                   0x4
> +
> +#define      GET_SEQ(lun_seq)                (lun_seq >> 2)
> +
> +struct ipmb_host_request {
> +     /*
> +      * u8 rs_sa;
> +      * rs_sa (rq_sa for rsp) is not part of the msg struct because
> +      * it is already integrated within the smbus message format.
> +      * the first data byte in the smbus message is the netfunction.
> +      */
> +     u8 netfn_rs_lun;        /* netfn_rq_lun for rsp */
> +     u8 checksum1;
> +     u8 rq_sa;               /* rs_sa for rsp */
> +     u8 rq_seq_rq_lun;       /* rq_seq_rs_lun for rsp */
> +     u8 cmd;
> +     u8 payload[IPMB_MSG_PAYLOAD_LEN_MAX];
> +     /* checksum2 is the last element of the payload */ } __packed;
> +
> +struct ipmb_host_response {
> +     /*
> +      * u8 rq_sa;
> +      * It is not part of the msg struct because
> +      * it is already integrated within the smbus message format.
> +      * the first data byte in the smbus message is the netfunction.
> +      *
> +      * 'len' is for internal use only. It is not part of the IPMB
> +      * message.
> +      */
> +     u8 len;
> +     u8 netfn_rq_lun;
> +     u8 checksum1;
> +     u8 rs_sa;
> +     u8 rq_seq_rs_lun;
> +     u8 cmd;
> +     /* completion code is the first element of the payload */
> +     u8 payload[IPMB_MSG_PAYLOAD_LEN_MAX];
> +     /* checksum2 is the last element of the payload */ } __packed;
> +
> +/*
> + * The ipmb_smi_msg struct is passed by the ipmi_smi_msg struct from
> + * in the ipmi_msghandler module. So it needs to have the same 
> +structure
> + * as ipmi_smi_msg. Refer to the linux code and libraries for
> + * more details.
> + */
> +struct ipmb_smi_msg {
> +     u8 netfn_lun;
> +     u8 cmd;
> +     u8 payload[IPMB_SMI_MSG_PAYLOAD_LEN_MAX];
> +} __packed;
> +
> +struct ipmb_seq_entry {
> +     struct ipmi_smi_msg     *msg;
> +     unsigned long           send_time;
> +};
> +
> +struct ipmb_rsp_elem {
> +     struct list_head          list;
> +     struct ipmb_host_response rsp;
> +};
> +
> +struct ipmb_master {
> +     struct ipmi_device_id           ipmi_id;
> +     /* Used to register this device as a slave device */
> +     struct i2c_client               *client;
> +     struct ipmi_smi                 *intf;
> +     spinlock_t                      lock;
> +     struct ipmb_seq_entry           seq_msg_map[IPMB_SEQ_MAX];
> +     struct work_struct              ipmb_send_work;
> +     struct ipmi_smi_msg             *msg_to_send;
> +     /* Responder's I2C slave address */
> +     u32                             rs_sa;
> +
> +     /* This is all for the response message */
> +     size_t                          msg_idx;
> +     struct ipmb_host_response       rsp;
> +     struct list_head                rsp_queue;
> +     atomic_t                        rsp_queue_len;
> +     wait_queue_head_t               wait_queue;
> +
> +     bool                            slave_registered;
> +};
> +
> +/* +1 is for the checksum integrated in payload */ #define 
> +IPMB_MSG_HDR \
> +     (sizeof(struct ipmb_host_request) - IPMB_MSG_PAYLOAD_LEN_MAX + 
> +1)
> +
> +#define IPMB_SMI_MSG_HDR \
> +     (sizeof(struct ipmb_smi_msg) - IPMB_SMI_MSG_PAYLOAD_LEN_MAX)
> +
> +/*
> + * ipmb_smi_msg contains a payload and 2 header fields: netfn_lun and cmd.
> + * Its payload does not contain checksum2.
> + *
> + * 'struct ipmb_host_request' and 'struct ipmb_host_response' contain 
> +a payload (including
> + * checksum2) and 5 header fields: netfn_r*_lun, checksum1, r*_sa,
> + * rq_seq_r*_lun, cmd. So we need to add one byte for each field 
> +which
> + * is present in the IPMB format and not in ipmb_smi_msg: checksum1,
> + * r*_sa, rq_seq_r*_lun and checksum2.
> + *
> + * Note that 'len' in 'struct ipmb_host_response' is discarded as it 
> +is for internal
> + * use only and not part of the actual IPMB message.
> + */
> +static u8 ipmi_smi_to_ipmb_len(size_t smi_msg_size) {
> +     return smi_msg_size + IPMB_MSG_HDR - IPMB_SMI_MSG_HDR; }
> +
> +/*
> + * This function is the converse of the above.
> + */
> +static u8 ipmb_to_smi_len(u8 msg_len) {
> +     return msg_len - (IPMB_MSG_HDR - IPMB_SMI_MSG_HDR); }
> +
> +/*
> + * ipmb_handle_response puts the received response message in
> + * a queue. The response will eventually be passed on to
> + * ipmitool.
> + */
> +static int ipmb_handle_response(struct ipmb_master *master)
> +{
> +     struct ipmb_rsp_elem *queue_elem;
> +     u8 seq;
> +
> +     /*
> +      * If this is a delayed response received after the ipmb_receive_rsp
> +      * timeout, discard it, don't add it to the queue.
> +      */
> +     seq = GET_SEQ(master->rsp.rq_seq_rs_lun);
> +     if (!master->seq_msg_map[seq].msg)
> +             return 0;
> +
> +     if (atomic_read(&master->rsp_queue_len) >=
> +                     RSP_QUEUE_MAX_LEN)
> +             return -EFAULT;
> +
> +     queue_elem = kmalloc(sizeof(*queue_elem), GFP_KERNEL);
> +     if (!queue_elem)
> +             return -ENOMEM;
> +     memcpy(&queue_elem->rsp, &master->rsp,
> +             sizeof(struct ipmb_host_response));
> +
> +     list_add(&queue_elem->list, &master->rsp_queue);
> +     atomic_inc(&master->rsp_queue_len);
> +     wake_up_all(&master->wait_queue);
> +     return 0;
> +}
> +
> +/*
> + * All this function does is send the request msg via I2C by calling
> + * i2c_master_send
> + */
> +static int ipmb_send_request(struct ipmb_master *master,
> +                             struct ipmb_host_request *request, u8 i2c_msg_len)
> +{
> +     struct i2c_client *client = master->client;
> +     unsigned long timeout, read_time;
> +     u8 *buf = (u8 *) request;
> +     int ret;
> +     union i2c_smbus_data data;
> +
> +     /*
> +      * skip netfn_rs_lun payload since it is passed as arg
> +      * 5 to i2c_smbus_xfer.
> +      */
> +     data.block[0] = i2c_msg_len;
> +     memcpy(&data.block[1], buf + 1, i2c_msg_len);
> +
> +     timeout = jiffies + msecs_to_jiffies(WRITE_TIMEOUT);
> +     do {
> +             read_time = jiffies;
> +
> +             ret = i2c_smbus_xfer(client->adapter, (u16)master->rs_sa,
> +                                 client->flags, I2C_SMBUS_WRITE,
> +                                 request->netfn_rs_lun,
> +                                 I2C_SMBUS_BLOCK_DATA, &data);
> +             if (ret == 0)
> +                     return ret;
> +             usleep_range(1000, 1500);
> +     } while (time_before(read_time, timeout));
> +
> +     return ret;
> +}
> +
> +static int ipmb_start_processing(void *data, struct ipmi_smi *intf)
> +{
> +     struct ipmb_master *master = data;
> +
> +     master->intf = intf;
> +
> +     return 0;
> +}
> +
> +static u8 ipmb_checksum1(u8 rs_sa, u8 netfn_rs_lun)
> +{
> +     u8 csum = rs_sa;
> +
> +     csum += netfn_rs_lun;
> +     return -csum;
> +}
> +
> +static u8 ipmb_checksum(u8 *data, int size, u8 start)
> +{
> +     u8 csum = start;
> +
> +     for (; size > 0; size--, data++)
> +             csum += *data;
> +
> +     return -csum;
> +}
> +
> +static void __ipmb_error_reply(struct ipmb_master *master,
> +                             struct ipmi_smi_msg *msg,
> +                             u8 completion_code)
> +{
> +     struct ipmb_smi_msg *response;
> +     struct ipmb_smi_msg *request;
> +
> +     response = (struct ipmb_smi_msg *) msg->rsp;
> +     request = (struct ipmb_smi_msg *) msg->data;
> +
> +     response->netfn_lun = request->netfn_lun | NETFN_RSP_BIT;
> +     response->cmd = request->cmd;
> +     response->payload[0] = completion_code;
> +     msg->rsp_size = 3;
> +     ipmi_smi_msg_received(master->intf, msg);
> +}
> +
> +static void ipmb_error_reply(struct ipmb_master *master,
> +                             struct ipmi_smi_msg *msg,
> +                             u8 completion_code)
> +{
> +     unsigned long flags;
> +
> +     spin_lock_irqsave(&master->lock, flags);
> +     __ipmb_error_reply(master, msg, completion_code);
> +     spin_unlock_irqrestore(&master->lock, flags);
> +}
> +
> +/*
> + * This function gets the length of the payload.
> + * Subtract one byte for each: netfn_rs_lun, checksum1,
> + * rq_sa, rq_seq_rq_lun, cmd and checksum2
> + */
> +static size_t ipmb_payload_len(size_t msg_len)
> +{
> +     return msg_len - 6;
> +}
> +
> +static bool ipmb_assign_seq(struct ipmb_master *master,
> +                             struct ipmi_smi_msg *msg, u8 *ret_seq)
> +{
> +     struct ipmb_seq_entry *entry;
> +     bool did_cleanup = false;
> +     unsigned long flags;
> +     u8 seq;
> +
> +     spin_lock_irqsave(&master->lock, flags);
> +retry:
> +     for (seq = 0; seq < IPMB_SEQ_MAX; seq++) {
> +             if (!master->seq_msg_map[seq].msg) {
> +                     master->seq_msg_map[seq].msg = msg;
> +                     master->seq_msg_map[seq].send_time = jiffies;
> +                     *ret_seq = seq;
> +                     spin_unlock_irqrestore(&master->lock, flags);
> +                     return true;
> +             }
> +     }
> +
> +     if (did_cleanup) {
> +             spin_unlock_irqrestore(&master->lock, flags);
> +             return false;
> +     }
> +
> +     /*
> +      * TODO: we should do cleanup at times other than only when we run out
> +      * of sequence numbers.
> +      */
> +     for (seq = 0; seq < IPMB_SEQ_MAX; seq++) {
> +             entry = &master->seq_msg_map[seq];
> +             if (entry->msg &&
> +                     time_after(entry->send_time + IPMB_TIMEOUT,
> +                     jiffies)) {
> +                     __ipmb_error_reply(master, entry->msg,
> +                                     IPMI_TIMEOUT_ERR);
> +                     entry->msg = NULL;
> +             }
> +     }
> +     did_cleanup = true;
> +     goto retry;
> +}
> +
> +static struct ipmi_smi_msg *ipmb_find_msg(
> +             struct ipmb_master *master, u8 seq)
> +{
> +     struct ipmi_smi_msg *msg;
> +     unsigned long flags;
> +
> +     spin_lock_irqsave(&master->lock, flags);
> +     msg = master->seq_msg_map[seq].msg;
> +     spin_unlock_irqrestore(&master->lock, flags);
> +     return msg;
> +}
> +
> +static void ipmb_free_seq(struct ipmb_master *master, u8 seq)
> +{
> +     unsigned long flags;
> +
> +     spin_lock_irqsave(&master->lock, flags);
> +     master->seq_msg_map[seq].msg = NULL;
> +     spin_unlock_irqrestore(&master->lock, flags);
> +}
> +
> +/*
> + * When this function is called, it waits until receiving an
> + * IPMI message in the response queue. If a response is found
> + * in the queue, it will be copied to ipmb_rsp.
> + * If no response is received after the timeout value, then
> + * the function returns with an error code. It will return:
> + * - 0 if there was no msg in the response queue after the timeout elapsed
> + * - A strictly positive number if a msg was found in the queue and
> + *   ipmb_rsp was successfully populated.
> + * - A negative value for errors.
> + */
> +static int ipmb_receive_rsp(struct ipmb_master *master,
> +                             struct ipmb_host_response *ipmb_rsp)
> +{
> +     struct ipmb_rsp_elem    *queue_elem;
> +     int                     ret = 1;
> +     unsigned long           flags;
> +
> +     spin_lock_irqsave(&master->lock, flags);
> +
> +     if (list_empty(&master->rsp_queue)) {
> +             spin_unlock_irqrestore(&master->lock, flags);
> +
> +             ret = wait_event_interruptible_timeout(master->wait_queue,
> +                     !list_empty(&master->rsp_queue), IPMB_TIMEOUT);
> +
> +             if (ret <= 0)
> +                     return ret;
> +
> +             spin_lock_irqsave(&master->lock, flags);
> +     }
> +
> +     queue_elem = list_first_entry(&master->rsp_queue,
> +                     struct ipmb_rsp_elem, list);
> +
> +     memcpy(ipmb_rsp, &queue_elem->rsp, sizeof(struct ipmb_host_response));
> +     list_del(&queue_elem->list);
> +     kfree(queue_elem);
> +     atomic_dec(&master->rsp_queue_len);
> +     spin_unlock_irqrestore(&master->lock, flags);
> +
> +     return ret;
> +}
> +
> +/*
> + * This function is called by ipmb_sender.
> + * It checks whether the message to be sent has an acceptable size,
> + * it assigns a sequence number to the msg
> + * it calls ipmb_send_request to send the msg to the receiver
> + * via I2C.
> + */
> +static void ipmb_send_workfn(struct work_struct *work)
> +{
> +     struct ipmb_master              *master;
> +
> +     struct ipmb_smi_msg             *smi_msg;
> +     struct ipmb_host_request        ipmb_req_msg;
> +     struct ipmi_smi_msg             *req_msg;
> +
> +     struct ipmb_smi_msg             *smi_rsp_msg;
> +     struct ipmb_host_response       ipmb_rsp_msg;
> +     struct ipmi_smi_msg             *rsp_msg;
> +
> +     size_t                          smi_msg_size;
> +     u8                              msg_len;
> +     unsigned long                   flags;
> +     int                             rsp_msg_len;
> +     u8                              *buf_rsp;
> +     u8                              verify_checksum;
> +     u8                              seq;
> +     u8                              i2c_msg_len;
> +
> +     u8 *buf = (u8 *) &ipmb_req_msg;
> +
> +     memset(&ipmb_req_msg, 0, sizeof(struct ipmb_host_request));
> +
> +     master = container_of(work, struct ipmb_master,
> +                          ipmb_send_work);
> +
> +     req_msg = master->msg_to_send;
> +     smi_msg_size = req_msg->data_size;
> +     smi_msg = (struct ipmb_smi_msg *) req_msg->data;
> +
> +     if (smi_msg_size > IPMB_MAX_SMI_SIZE) {
> +             ipmb_error_reply(master, req_msg, IPMI_REQ_LEN_EXCEEDED_ERR);
> +             return;
> +     }
> +
> +     if (smi_msg_size < IPMB_SMI_MSG_HEADER_SIZE) {
> +             ipmb_error_reply(master, req_msg, IPMI_REQ_LEN_INVALID_ERR);
> +             return;
> +     }
> +
> +     if (!ipmb_assign_seq(master, req_msg, &seq)) {
> +             ipmb_error_reply(master, req_msg, IPMI_NODE_BUSY_ERR);
> +             return;
> +     }
> +
> +     ipmb_req_msg.rq_seq_rq_lun = seq << 2;
> +
> +     msg_len = ipmi_smi_to_ipmb_len(smi_msg_size);
> +
> +     /* Responder  */
> +     ipmb_req_msg.netfn_rs_lun = smi_msg->netfn_lun;
> +     ipmb_req_msg.checksum1 = ipmb_checksum1((u8)(master->rs_sa << 1),
> +                                             ipmb_req_msg.netfn_rs_lun);
> +
> +     /* Requester is this device */
> +     ipmb_req_msg.rq_sa = (u8)(master->client->addr << 1);
> +     ipmb_req_msg.cmd = smi_msg->cmd;
> +
> +     memcpy(ipmb_req_msg.payload, smi_msg->payload,
> +             ipmb_payload_len((size_t)msg_len));
> +     ipmb_req_msg.payload[ipmb_payload_len((size_t)msg_len)] =
> +             ipmb_checksum(buf + 2, msg_len - 2, 0);
> +
> +     /*
> +      * subtract netfn_rs_lun payload since it is passed as arg
> +      * 5 to i2c_smbus_xfer.
> +      */
> +     i2c_msg_len = ipmi_smi_to_ipmb_len(master->msg_to_send->data_size) - 1;
> +     if (i2c_msg_len > I2C_SMBUS_BLOCK_MAX)
> +             i2c_msg_len = I2C_SMBUS_BLOCK_MAX;
> +
> +     if (ipmb_send_request(master, &ipmb_req_msg, i2c_msg_len) < 0) {
> +             ipmb_free_seq(master, (GET_SEQ(ipmb_req_msg.rq_seq_rq_lun)));
> +             ipmb_error_reply(master, req_msg, IPMI_BUS_ERR);
> +             spin_lock_irqsave(&master->lock, flags);
> +             master->msg_to_send = NULL;
> +             spin_unlock_irqrestore(&master->lock, flags);
> +             return;
> +     }
> +
> +     spin_lock_irqsave(&master->lock, flags);
> +     master->msg_to_send = NULL;
> +     spin_unlock_irqrestore(&master->lock, flags);
> +
> +     /* Done with sending request. Now handling response */
> +
> +     if (ipmb_receive_rsp(master, &ipmb_rsp_msg) <= 0) {
> +             ipmb_free_seq(master, (GET_SEQ(ipmb_req_msg.rq_seq_rq_lun)));
> +             ipmb_error_reply(master, req_msg, IPMI_TIMEOUT_ERR);
> +             return;
> +     }
> +
> +     buf_rsp = (u8 *) &ipmb_rsp_msg;
> +     /* skip len */
> +     buf_rsp += 1;
> +     rsp_msg_len = ipmb_rsp_msg.len - 1;
> +
> +     if (rsp_msg_len < IPMB_LEN_MIN) {
> +             ipmb_free_seq(master, (GET_SEQ(ipmb_req_msg.rq_seq_rq_lun)));
> +             ipmb_error_reply(master, req_msg, IPMI_ERR_MSG_TRUNCATED);
> +             return;
> +     }
> +
> +     rsp_msg = ipmb_find_msg(master, (GET_SEQ(ipmb_rsp_msg.rq_seq_rs_lun)));
> +     if (!rsp_msg) {
> +             ipmb_free_seq(master, (GET_SEQ(ipmb_req_msg.rq_seq_rq_lun)));
> +             ipmb_error_reply(master, req_msg, IPMI_ERR_UNSPECIFIED);
> +             return;
> +     }
> +
> +     ipmb_free_seq(master, (GET_SEQ(ipmb_rsp_msg.rq_seq_rs_lun)));
> +
> +     if (rsp_msg_len < IPMB_RESPONSE_LEN_MIN) {
> +             ipmb_error_reply(master, rsp_msg, IPMI_ERR_MSG_TRUNCATED);
> +             return;
> +     }
> +
> +     verify_checksum = ipmb_checksum(buf_rsp, rsp_msg_len,
> +                             (u8)(master->client->addr << 1));
> +
> +     if (verify_checksum) {
> +             ipmb_error_reply(master, req_msg, IPMI_ERR_UNSPECIFIED);
> +             return;
> +     }
> +
> +     rsp_msg->rsp_size = ipmb_to_smi_len((u8) rsp_msg_len);
> +     smi_rsp_msg = (struct ipmb_smi_msg *) rsp_msg->rsp;
> +     smi_rsp_msg->netfn_lun = ipmb_rsp_msg.netfn_rq_lun;
> +     smi_rsp_msg->cmd = ipmb_rsp_msg.cmd;
> +     memcpy(smi_rsp_msg->payload, ipmb_rsp_msg.payload,
> +             ipmb_payload_len((size_t) rsp_msg_len));
> +
> +     ipmi_smi_msg_received(master->intf, rsp_msg);
> +}
> +
> +/*
> + * Function called by smi_send in ipmi_msghandler.c
> + * It passes request message from ipmitool program
> + * to the host's kernel to the receiver via I2C.
> + */
> +static void ipmb_sender(void *data, struct ipmi_smi_msg *msg)
> +{
> +     struct ipmb_master *master = data;
> +     unsigned long flags;
> +
> +     spin_lock_irqsave(&master->lock, flags);
> +     if (master->msg_to_send) {
> +             __ipmb_error_reply(master, msg, IPMI_NODE_BUSY_ERR);
> +     } else {
> +             master->msg_to_send = msg;
> +             schedule_work(&master->ipmb_send_work);
> +     }
> +     spin_unlock_irqrestore(&master->lock, flags);
> +}
> +
> +static void ipmb_request_events(void *data)
> +{
> +}
> +
> +static void ipmb_set_run_to_completion(void *data,
> +                             bool run_to_completion)
> +{
> +}
> +
> +static void ipmb_poll(void *data)
> +{
> +}
> +
> +static struct ipmi_smi_handlers ipmb_smi_handlers = {
> +     .owner                  = THIS_MODULE,
> +     .start_processing       = ipmb_start_processing,
> +     .sender                 = ipmb_sender,
> +     .request_events         = ipmb_request_events,
> +     .set_run_to_completion  = ipmb_set_run_to_completion,
> +     .poll                   = ipmb_poll,
> +};
> +
> +static bool is_ipmb_response(u8 netfn_rq_lun, size_t msg_len)
> +{
> +     /*
> +      * First, check whether the message has the minimum IPMB response size
> +      */
> +     if (msg_len >= IPMB_RESPONSE_LEN_MIN) {
> +             /*
> +              * Then check whether this is an IPMB request or response.
> +              * Responses have an odd netfn while requests have an even
> +              * netfn.
> +              */
> +             if ((netfn_rq_lun & NETFN_RSP_BIT) == NETFN_RSP_BIT)
> +                     return true;
> +     }
> +
> +     return false;
> +}
> +
> +/*
> + * This is the callback function used to set this device as a slave
> + * and to monitor and handle only IPMB responses.
> + *
> + * This driver's purpose is to:
> + * 1) send IPMB requests,
> + * 2) then wait until it receives a response back from the responder.
> + *    This callback adds that response into a queue so that it is handled
> + *    later in ipmb_receive_rsp.
> + */
> +static int ipmb_slave_cb(struct i2c_client *client,
> +                     enum i2c_slave_event event, u8 *val)
> +{
> +     struct ipmb_master *master = i2c_get_clientdata(client);
> +     u8 *buf;
> +
> +     if (!handshake_rsp) {
> +             handshake_rsp = true;
> +             return 0;
> +     }
> +
> +     spin_lock(&master->lock);
> +
> +     switch (event) {
> +     case I2C_SLAVE_WRITE_REQUESTED:
> +             /*
> +              * The length of the msg is stored at msg_idx 0,
> +              * which maps to master->rsp.len. Skip it for now.
> +              * The len will be populated once the whole buf is
> +              * filled.
> +              */
> +             master->msg_idx = 1;
> +             memset(&master->rsp, 0,
> +                     sizeof(master->rsp));
> +             break;
> +
> +     case I2C_SLAVE_WRITE_RECEIVED:
> +             buf = (u8 *) &master->rsp;
> +
> +             if (master->msg_idx >= sizeof(struct ipmb_host_response))
> +                     break;
> +
> +             buf[master->msg_idx++] = *val;
> +             break;
> +
> +     case I2C_SLAVE_STOP:
> +             master->rsp.len = master->msg_idx;
> +             if (is_ipmb_response(master->rsp.netfn_rq_lun,
> +                             master->msg_idx))
> +                     ipmb_handle_response(master);
> +             master->msg_idx = 0;
> +             break;
> +
> +     default:
> +             break;
> +     }
> +     spin_unlock(&master->lock);
> +
> +     return 0;
> +}
> +
> +static unsigned short slave_add = 0x0;
> +module_param(slave_add, ushort, 0);
> +MODULE_PARM_DESC(slave_add, "The i2c slave address of the responding device");
> +
> +#define GET_DEVICE_ID_MSG_LEN 7
> +
> +static bool ipmb_detect(struct ipmb_master *master)
> +{
> +     struct ipmb_rsp_elem *q_elem, *tmp_q_elem;
> +     struct ipmb_host_request request;
> +     u8 *buf = (u8 *) &request;
> +     struct device dev;
> +     int retry = 2000;
> +     u8 i2c_msg_len;
> +     int ret;
> +
> +     /* Subtract rs sa and netfn */
> +     i2c_msg_len = GET_DEVICE_ID_MSG_LEN - 2;
> +
> +     dev = master->client->dev;
> +
> +     request.netfn_rs_lun = IPMI_NETFN_APP_REQUEST << 2;
> +     request.checksum1 = ipmb_checksum1((u8)(master->rs_sa << 1),
> +                                             request.netfn_rs_lun);
> +     request.rq_sa = (u8)(master->client->addr << 1);
> +     request.rq_seq_rq_lun = 0;
> +     request.cmd = IPMI_GET_DEVICE_ID_CMD;
> +     request.payload[0] = ipmb_checksum(buf + 2, 3, 0);
> +
> +     ret = ipmb_send_request(master, &request, i2c_msg_len);
> +     if (ret < 0) {
> +             dev_err(&dev, "ERROR: ipmb_send_request failed during ipmb detection\n");
> +             return false;
> +     }
> +
> +     while(!handshake_rsp && (retry > 0)) {
> +             mdelay(10);
> +             retry--;
> +     }
> +
> +     if (!retry) {
> +             dev_err(&dev, "ERROR: Response timed out during ipmb detection\n");
> +             return false;
> +     }
> +
> +     list_for_each_entry_safe(q_elem, tmp_q_elem, &master->rsp_queue, list){
> +             list_del(&q_elem->list);
> +             kfree(q_elem);
> +             atomic_dec(&master->rsp_queue_len);
> +     }
> +
> +     return true;
> +}
> +
> +static int ipmb_probe(struct i2c_client *client,
> +                     const struct i2c_device_id *id)
> +{
> +     struct ipmb_master *master;
> +     int ret;
> +
> +     master = devm_kzalloc(&client->dev, sizeof(struct ipmb_master),
> +                          GFP_KERNEL);
> +     if (!master)
> +             return -ENOMEM;
> +
> +     spin_lock_init(&master->lock);
> +     init_waitqueue_head(&master->wait_queue);
> +     atomic_set(&master->rsp_queue_len, 0);
> +     INIT_LIST_HEAD(&master->rsp_queue);
> +
> +     INIT_WORK(&master->ipmb_send_work, ipmb_send_workfn);
> +
> +     ret = device_property_read_u32(&client->dev, "slave-address",
> +                                     &master->rs_sa);
> +     if (ret) {
> +             master->rs_sa = slave_add;
> +             if (master->rs_sa == 0x0) {
> +                     dev_err(&client->dev,
> +                             "Failed to get the responder's address from user\n");
> +                     return ret;
> +             }
> +     }
> +
> +     master->client = client;
> +     i2c_set_clientdata(client, master);
> +
> +     ret = i2c_slave_register(client, ipmb_slave_cb);
> +
> +     if (ret)
> +             return ret;
> +
> +     master->slave_registered = true;
> +
> +     /*
> +      * Send a simple message "get device ID" to detect whether the BMC is responsive or not.
> +      * This is necessary before calling ipmi_register_smi which executes a handshake with the
> +      * slave device and can hold the lock for a very long if the BMC is not up. This long wait
> +      * at boot time causes the system to crash.
> +      */
> +     if (!ipmb_detect(master)) {
> +             dev_err(&client->dev, "Unable to get response from slave device at this time\n");
> +             i2c_slave_unregister(client);
> +             master->slave_registered = false;
> +             return -ENXIO;
> +     }
> +
> +     ret = ipmi_register_smi(&ipmb_smi_handlers, master,
> +                             &client->dev,
> +                             (unsigned char)master->rs_sa);
> +
> +     if (ret) {
> +             dev_err(&client->dev, "ipmi_register_smi failed with ret = %d\n", ret);
> +             i2c_slave_unregister(client);
> +             master->slave_registered = false;
> +     }
> +
> +     return ret;
> +}
> +
> +static int ipmb_remove(struct i2c_client *client)
> +{
> +     struct ipmb_master *master;
> +
> +     master = i2c_get_clientdata(client);
> +     if (!master)
> +             return 0;
> +
> +     if (master->slave_registered) {
> +             ipmi_unregister_smi(master->intf);
> +             i2c_slave_unregister(client);
> +     }
> +
> +     return 0;
> +}
> +
> +static const struct i2c_device_id ipmb_i2c_id[] = {
> +     {"ipmb-host", 0},
> +     {},
> +};
> +MODULE_DEVICE_TABLE(i2c, ipmb_i2c_id);
> +
> +static struct i2c_driver ipmb_driver = {
> +     .driver = {
> +             .owner = THIS_MODULE,
> +             .name = "ipmb-host",
> +     },
> +     .probe = ipmb_probe,
> +     .remove = ipmb_remove,
> +     .id_table = ipmb_i2c_id,
> +};
> +module_i2c_driver(ipmb_driver);
> +
> +MODULE_AUTHOR("Asmaa Mnebhi <asmaa at nvidia.com>");
> +MODULE_DESCRIPTION("Host IPMB driver");
> +MODULE_LICENSE("GPL v2");



More information about the kernel-team mailing list