diff options
Diffstat (limited to 'drivers/misc/mic/cosm')
| -rw-r--r-- | drivers/misc/mic/cosm/Makefile | 10 | ||||
| -rw-r--r-- | drivers/misc/mic/cosm/cosm_debugfs.c | 156 | ||||
| -rw-r--r-- | drivers/misc/mic/cosm/cosm_main.c | 388 | ||||
| -rw-r--r-- | drivers/misc/mic/cosm/cosm_main.h | 70 | ||||
| -rw-r--r-- | drivers/misc/mic/cosm/cosm_scif_server.c | 405 | ||||
| -rw-r--r-- | drivers/misc/mic/cosm/cosm_sysfs.c | 461 |
6 files changed, 1490 insertions, 0 deletions
diff --git a/drivers/misc/mic/cosm/Makefile b/drivers/misc/mic/cosm/Makefile new file mode 100644 index 000000000..b85d4d49d --- /dev/null +++ b/drivers/misc/mic/cosm/Makefile @@ -0,0 +1,10 @@ +# +# Makefile - Intel MIC Coprocessor State Management (COSM) Driver +# Copyright(c) 2015, Intel Corporation. +# +obj-$(CONFIG_MIC_COSM) += mic_cosm.o + +mic_cosm-objs := cosm_main.o +mic_cosm-objs += cosm_debugfs.o +mic_cosm-objs += cosm_sysfs.o +mic_cosm-objs += cosm_scif_server.o diff --git a/drivers/misc/mic/cosm/cosm_debugfs.c b/drivers/misc/mic/cosm/cosm_debugfs.c new file mode 100644 index 000000000..216cb3cd2 --- /dev/null +++ b/drivers/misc/mic/cosm/cosm_debugfs.c @@ -0,0 +1,156 @@ +/* + * Intel MIC Platform Software Stack (MPSS) + * + * Copyright(c) 2015 Intel Corporation. + * + * 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. + * + * The full GNU General Public License is included in this distribution in + * the file called "COPYING". + * + * Intel MIC Coprocessor State Management (COSM) Driver + * + */ + +#include <linux/debugfs.h> +#include <linux/slab.h> +#include <linux/io.h> +#include "cosm_main.h" + +/* Debugfs parent dir */ +static struct dentry *cosm_dbg; + +/** + * cosm_log_buf_show - Display MIC kernel log buffer + * + * log_buf addr/len is read from System.map by user space + * and populated in sysfs entries. + */ +static int cosm_log_buf_show(struct seq_file *s, void *unused) +{ + void __iomem *log_buf_va; + int __iomem *log_buf_len_va; + struct cosm_device *cdev = s->private; + void *kva; + int size; + u64 aper_offset; + + if (!cdev || !cdev->log_buf_addr || !cdev->log_buf_len) + goto done; + + mutex_lock(&cdev->cosm_mutex); + switch (cdev->state) { + case MIC_BOOTING: + case MIC_ONLINE: + case MIC_SHUTTING_DOWN: + break; + default: + goto unlock; + } + + /* + * Card kernel will never be relocated and any kernel text/data mapping + * can be translated to phys address by subtracting __START_KERNEL_map. + */ + aper_offset = (u64)cdev->log_buf_len - __START_KERNEL_map; + log_buf_len_va = cdev->hw_ops->aper(cdev)->va + aper_offset; + aper_offset = (u64)cdev->log_buf_addr - __START_KERNEL_map; + log_buf_va = cdev->hw_ops->aper(cdev)->va + aper_offset; + + size = ioread32(log_buf_len_va); + kva = kmalloc(size, GFP_KERNEL); + if (!kva) + goto unlock; + + memcpy_fromio(kva, log_buf_va, size); + seq_write(s, kva, size); + kfree(kva); +unlock: + mutex_unlock(&cdev->cosm_mutex); +done: + return 0; +} + +static int cosm_log_buf_open(struct inode *inode, struct file *file) +{ + return single_open(file, cosm_log_buf_show, inode->i_private); +} + +static const struct file_operations log_buf_ops = { + .owner = THIS_MODULE, + .open = cosm_log_buf_open, + .read = seq_read, + .llseek = seq_lseek, + .release = single_release +}; + +/** + * cosm_force_reset_show - Force MIC reset + * + * Invokes the force_reset COSM bus op instead of the standard reset + * op in case a force reset of the MIC device is required + */ +static int cosm_force_reset_show(struct seq_file *s, void *pos) +{ + struct cosm_device *cdev = s->private; + + cosm_stop(cdev, true); + return 0; +} + +static int cosm_force_reset_debug_open(struct inode *inode, struct file *file) +{ + return single_open(file, cosm_force_reset_show, inode->i_private); +} + +static const struct file_operations force_reset_ops = { + .owner = THIS_MODULE, + .open = cosm_force_reset_debug_open, + .read = seq_read, + .llseek = seq_lseek, + .release = single_release +}; + +void cosm_create_debug_dir(struct cosm_device *cdev) +{ + char name[16]; + + if (!cosm_dbg) + return; + + scnprintf(name, sizeof(name), "mic%d", cdev->index); + cdev->dbg_dir = debugfs_create_dir(name, cosm_dbg); + if (!cdev->dbg_dir) + return; + + debugfs_create_file("log_buf", 0444, cdev->dbg_dir, cdev, &log_buf_ops); + debugfs_create_file("force_reset", 0444, cdev->dbg_dir, cdev, + &force_reset_ops); +} + +void cosm_delete_debug_dir(struct cosm_device *cdev) +{ + if (!cdev->dbg_dir) + return; + + debugfs_remove_recursive(cdev->dbg_dir); +} + +void cosm_init_debugfs(void) +{ + cosm_dbg = debugfs_create_dir(KBUILD_MODNAME, NULL); + if (!cosm_dbg) + pr_err("can't create debugfs dir\n"); +} + +void cosm_exit_debugfs(void) +{ + debugfs_remove(cosm_dbg); +} diff --git a/drivers/misc/mic/cosm/cosm_main.c b/drivers/misc/mic/cosm/cosm_main.c new file mode 100644 index 000000000..4b4b356c7 --- /dev/null +++ b/drivers/misc/mic/cosm/cosm_main.c @@ -0,0 +1,388 @@ +/* + * Intel MIC Platform Software Stack (MPSS) + * + * Copyright(c) 2015 Intel Corporation. + * + * 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. + * + * The full GNU General Public License is included in this distribution in + * the file called "COPYING". + * + * Intel MIC Coprocessor State Management (COSM) Driver + * + */ + +#include <linux/module.h> +#include <linux/delay.h> +#include <linux/idr.h> +#include <linux/slab.h> +#include <linux/cred.h> +#include "cosm_main.h" + +static const char cosm_driver_name[] = "mic"; + +/* COSM ID allocator */ +static struct ida g_cosm_ida; +/* Class of MIC devices for sysfs accessibility. */ +static struct class *g_cosm_class; +/* Number of MIC devices */ +static atomic_t g_num_dev; + +/** + * cosm_hw_reset - Issue a HW reset for the MIC device + * @cdev: pointer to cosm_device instance + */ +static void cosm_hw_reset(struct cosm_device *cdev, bool force) +{ + int i; + +#define MIC_RESET_TO (45) + if (force && cdev->hw_ops->force_reset) + cdev->hw_ops->force_reset(cdev); + else + cdev->hw_ops->reset(cdev); + + for (i = 0; i < MIC_RESET_TO; i++) { + if (cdev->hw_ops->ready(cdev)) { + cosm_set_state(cdev, MIC_READY); + return; + } + /* + * Resets typically take 10s of seconds to complete. + * Since an MMIO read is required to check if the + * firmware is ready or not, a 1 second delay works nicely. + */ + msleep(1000); + } + cosm_set_state(cdev, MIC_RESET_FAILED); +} + +/** + * cosm_start - Start the MIC + * @cdev: pointer to cosm_device instance + * + * This function prepares an MIC for boot and initiates boot. + * RETURNS: An appropriate -ERRNO error value on error, or 0 for success. + */ +int cosm_start(struct cosm_device *cdev) +{ + const struct cred *orig_cred; + struct cred *override_cred; + int rc; + + mutex_lock(&cdev->cosm_mutex); + if (!cdev->bootmode) { + dev_err(&cdev->dev, "%s %d bootmode not set\n", + __func__, __LINE__); + rc = -EINVAL; + goto unlock_ret; + } +retry: + if (cdev->state != MIC_READY) { + dev_err(&cdev->dev, "%s %d MIC state not READY\n", + __func__, __LINE__); + rc = -EINVAL; + goto unlock_ret; + } + if (!cdev->hw_ops->ready(cdev)) { + cosm_hw_reset(cdev, false); + /* + * The state will either be MIC_READY if the reset succeeded + * or MIC_RESET_FAILED if the firmware reset failed. + */ + goto retry; + } + + /* + * Set credentials to root to allow non-root user to download initramsfs + * with 600 permissions + */ + override_cred = prepare_creds(); + if (!override_cred) { + dev_err(&cdev->dev, "%s %d prepare_creds failed\n", + __func__, __LINE__); + rc = -ENOMEM; + goto unlock_ret; + } + override_cred->fsuid = GLOBAL_ROOT_UID; + orig_cred = override_creds(override_cred); + + rc = cdev->hw_ops->start(cdev, cdev->index); + + revert_creds(orig_cred); + put_cred(override_cred); + if (rc) + goto unlock_ret; + + /* + * If linux is being booted, card is treated 'online' only + * when the scif interface in the card is up. If anything else + * is booted, we set card to 'online' immediately. + */ + if (!strcmp(cdev->bootmode, "linux")) + cosm_set_state(cdev, MIC_BOOTING); + else + cosm_set_state(cdev, MIC_ONLINE); +unlock_ret: + mutex_unlock(&cdev->cosm_mutex); + if (rc) + dev_err(&cdev->dev, "cosm_start failed rc %d\n", rc); + return rc; +} + +/** + * cosm_stop - Prepare the MIC for reset and trigger reset + * @cdev: pointer to cosm_device instance + * @force: force a MIC to reset even if it is already reset and ready. + * + * RETURNS: None + */ +void cosm_stop(struct cosm_device *cdev, bool force) +{ + mutex_lock(&cdev->cosm_mutex); + if (cdev->state != MIC_READY || force) { + /* + * Don't call hw_ops if they have been called previously. + * stop(..) calls device_unregister and will crash the system if + * called multiple times. + */ + bool call_hw_ops = cdev->state != MIC_RESET_FAILED && + cdev->state != MIC_READY; + + if (cdev->state != MIC_RESETTING) + cosm_set_state(cdev, MIC_RESETTING); + cdev->heartbeat_watchdog_enable = false; + if (call_hw_ops) + cdev->hw_ops->stop(cdev, force); + cosm_hw_reset(cdev, force); + cosm_set_shutdown_status(cdev, MIC_NOP); + if (call_hw_ops && cdev->hw_ops->post_reset) + cdev->hw_ops->post_reset(cdev, cdev->state); + } + mutex_unlock(&cdev->cosm_mutex); + flush_work(&cdev->scif_work); +} + +/** + * cosm_reset_trigger_work - Trigger MIC reset + * @work: The work structure + * + * This work is scheduled whenever the host wants to reset the MIC. + */ +static void cosm_reset_trigger_work(struct work_struct *work) +{ + struct cosm_device *cdev = container_of(work, struct cosm_device, + reset_trigger_work); + cosm_stop(cdev, false); +} + +/** + * cosm_reset - Schedule MIC reset + * @cdev: pointer to cosm_device instance + * + * RETURNS: An -EINVAL if the card is already READY or 0 for success. + */ +int cosm_reset(struct cosm_device *cdev) +{ + int rc = 0; + + mutex_lock(&cdev->cosm_mutex); + if (cdev->state != MIC_READY) { + cosm_set_state(cdev, MIC_RESETTING); + schedule_work(&cdev->reset_trigger_work); + } else { + dev_err(&cdev->dev, "%s %d MIC is READY\n", __func__, __LINE__); + rc = -EINVAL; + } + mutex_unlock(&cdev->cosm_mutex); + return rc; +} + +/** + * cosm_shutdown - Initiate MIC shutdown. + * @cdev: pointer to cosm_device instance + * + * RETURNS: None + */ +int cosm_shutdown(struct cosm_device *cdev) +{ + struct cosm_msg msg = { .id = COSM_MSG_SHUTDOWN }; + int rc = 0; + + mutex_lock(&cdev->cosm_mutex); + if (cdev->state != MIC_ONLINE) { + rc = -EINVAL; + dev_err(&cdev->dev, "%s %d skipping shutdown in state: %s\n", + __func__, __LINE__, cosm_state_string[cdev->state]); + goto err; + } + + if (!cdev->epd) { + rc = -ENOTCONN; + dev_err(&cdev->dev, "%s %d scif endpoint not connected rc %d\n", + __func__, __LINE__, rc); + goto err; + } + + rc = scif_send(cdev->epd, &msg, sizeof(msg), SCIF_SEND_BLOCK); + if (rc < 0) { + dev_err(&cdev->dev, "%s %d scif_send failed rc %d\n", + __func__, __LINE__, rc); + goto err; + } + cdev->heartbeat_watchdog_enable = false; + cosm_set_state(cdev, MIC_SHUTTING_DOWN); + rc = 0; +err: + mutex_unlock(&cdev->cosm_mutex); + return rc; +} + +static int cosm_driver_probe(struct cosm_device *cdev) +{ + int rc; + + /* Initialize SCIF server at first probe */ + if (atomic_add_return(1, &g_num_dev) == 1) { + rc = cosm_scif_init(); + if (rc) + goto scif_exit; + } + mutex_init(&cdev->cosm_mutex); + INIT_WORK(&cdev->reset_trigger_work, cosm_reset_trigger_work); + INIT_WORK(&cdev->scif_work, cosm_scif_work); + cdev->sysfs_heartbeat_enable = true; + cosm_sysfs_init(cdev); + cdev->sdev = device_create_with_groups(g_cosm_class, cdev->dev.parent, + MKDEV(0, cdev->index), cdev, cdev->attr_group, + "mic%d", cdev->index); + if (IS_ERR(cdev->sdev)) { + rc = PTR_ERR(cdev->sdev); + dev_err(&cdev->dev, "device_create_with_groups failed rc %d\n", + rc); + goto scif_exit; + } + + cdev->state_sysfs = sysfs_get_dirent(cdev->sdev->kobj.sd, + "state"); + if (!cdev->state_sysfs) { + rc = -ENODEV; + dev_err(&cdev->dev, "sysfs_get_dirent failed rc %d\n", rc); + goto destroy_device; + } + cosm_create_debug_dir(cdev); + return 0; +destroy_device: + device_destroy(g_cosm_class, MKDEV(0, cdev->index)); +scif_exit: + if (atomic_dec_and_test(&g_num_dev)) + cosm_scif_exit(); + return rc; +} + +static void cosm_driver_remove(struct cosm_device *cdev) +{ + cosm_delete_debug_dir(cdev); + sysfs_put(cdev->state_sysfs); + device_destroy(g_cosm_class, MKDEV(0, cdev->index)); + flush_work(&cdev->reset_trigger_work); + cosm_stop(cdev, false); + if (atomic_dec_and_test(&g_num_dev)) + cosm_scif_exit(); + + /* These sysfs entries might have allocated */ + kfree(cdev->cmdline); + kfree(cdev->firmware); + kfree(cdev->ramdisk); + kfree(cdev->bootmode); +} + +static int cosm_suspend(struct device *dev) +{ + struct cosm_device *cdev = dev_to_cosm(dev); + + mutex_lock(&cdev->cosm_mutex); + switch (cdev->state) { + /** + * Suspend/freeze hooks in userspace have already shutdown the card. + * Card should be 'ready' in most cases. It is however possible that + * some userspace application initiated a boot. In those cases, we + * simply reset the card. + */ + case MIC_ONLINE: + case MIC_BOOTING: + case MIC_SHUTTING_DOWN: + mutex_unlock(&cdev->cosm_mutex); + cosm_stop(cdev, false); + break; + default: + mutex_unlock(&cdev->cosm_mutex); + break; + } + return 0; +} + +static const struct dev_pm_ops cosm_pm_ops = { + .suspend = cosm_suspend, + .freeze = cosm_suspend +}; + +static struct cosm_driver cosm_driver = { + .driver = { + .name = KBUILD_MODNAME, + .owner = THIS_MODULE, + .pm = &cosm_pm_ops, + }, + .probe = cosm_driver_probe, + .remove = cosm_driver_remove +}; + +static int __init cosm_init(void) +{ + int ret; + + cosm_init_debugfs(); + + g_cosm_class = class_create(THIS_MODULE, cosm_driver_name); + if (IS_ERR(g_cosm_class)) { + ret = PTR_ERR(g_cosm_class); + pr_err("class_create failed ret %d\n", ret); + goto cleanup_debugfs; + } + + ida_init(&g_cosm_ida); + ret = cosm_register_driver(&cosm_driver); + if (ret) { + pr_err("cosm_register_driver failed ret %d\n", ret); + goto ida_destroy; + } + return 0; +ida_destroy: + ida_destroy(&g_cosm_ida); + class_destroy(g_cosm_class); +cleanup_debugfs: + cosm_exit_debugfs(); + return ret; +} + +static void __exit cosm_exit(void) +{ + cosm_unregister_driver(&cosm_driver); + ida_destroy(&g_cosm_ida); + class_destroy(g_cosm_class); + cosm_exit_debugfs(); +} + +module_init(cosm_init); +module_exit(cosm_exit); + +MODULE_AUTHOR("Intel Corporation"); +MODULE_DESCRIPTION("Intel(R) MIC Coprocessor State Management (COSM) Driver"); +MODULE_LICENSE("GPL v2"); diff --git a/drivers/misc/mic/cosm/cosm_main.h b/drivers/misc/mic/cosm/cosm_main.h new file mode 100644 index 000000000..f01156fca --- /dev/null +++ b/drivers/misc/mic/cosm/cosm_main.h @@ -0,0 +1,70 @@ +/* + * Intel MIC Platform Software Stack (MPSS) + * + * Copyright(c) 2015 Intel Corporation. + * + * 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. + * + * The full GNU General Public License is included in this distribution in + * the file called "COPYING". + * + * Intel MIC Coprocessor State Management (COSM) Driver + * + */ +#ifndef _COSM_COSM_H_ +#define _COSM_COSM_H_ + +#include <linux/scif.h> +#include "../bus/cosm_bus.h" + +#define COSM_HEARTBEAT_SEND_SEC 30 +#define SCIF_COSM_LISTEN_PORT 201 + +/** + * enum COSM msg id's + * @COSM_MSG_SHUTDOWN: host->card trigger shutdown + * @COSM_MSG_SYNC_TIME: host->card send host time to card to sync time + * @COSM_MSG_HEARTBEAT: card->host heartbeat + * @COSM_MSG_SHUTDOWN_STATUS: card->host with shutdown status as payload + */ +enum cosm_msg_id { + COSM_MSG_SHUTDOWN, + COSM_MSG_SYNC_TIME, + COSM_MSG_HEARTBEAT, + COSM_MSG_SHUTDOWN_STATUS, +}; + +struct cosm_msg { + u64 id; + union { + u64 shutdown_status; + struct timespec64 timespec; + }; +}; + +extern const char * const cosm_state_string[]; +extern const char * const cosm_shutdown_status_string[]; + +void cosm_sysfs_init(struct cosm_device *cdev); +int cosm_start(struct cosm_device *cdev); +void cosm_stop(struct cosm_device *cdev, bool force); +int cosm_reset(struct cosm_device *cdev); +int cosm_shutdown(struct cosm_device *cdev); +void cosm_set_state(struct cosm_device *cdev, u8 state); +void cosm_set_shutdown_status(struct cosm_device *cdev, u8 status); +void cosm_init_debugfs(void); +void cosm_exit_debugfs(void); +void cosm_create_debug_dir(struct cosm_device *cdev); +void cosm_delete_debug_dir(struct cosm_device *cdev); +int cosm_scif_init(void); +void cosm_scif_exit(void); +void cosm_scif_work(struct work_struct *work); + +#endif diff --git a/drivers/misc/mic/cosm/cosm_scif_server.c b/drivers/misc/mic/cosm/cosm_scif_server.c new file mode 100644 index 000000000..5696df432 --- /dev/null +++ b/drivers/misc/mic/cosm/cosm_scif_server.c @@ -0,0 +1,405 @@ +/* + * Intel MIC Platform Software Stack (MPSS) + * + * Copyright(c) 2015 Intel Corporation. + * + * 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. + * + * The full GNU General Public License is included in this distribution in + * the file called "COPYING". + * + * Intel MIC Coprocessor State Management (COSM) Driver + * + */ +#include <linux/kthread.h> +#include "cosm_main.h" + +/* + * The COSM driver uses SCIF to communicate between the management node and the + * MIC cards. SCIF is used to (a) Send a shutdown command to the card (b) + * receive a shutdown status back from the card upon completion of shutdown and + * (c) receive periodic heartbeat messages from the card used to deduce if the + * card has crashed. + * + * A COSM server consisting of a SCIF listening endpoint waits for incoming + * connections from the card. Upon acceptance of the connection, a separate + * work-item is scheduled to handle SCIF message processing for that card. The + * life-time of this work-item is therefore the time from which the connection + * from a card is accepted to the time at which the connection is closed. A new + * work-item starts each time the card boots and is alive till the card (a) + * shuts down (b) is reset (c) crashes (d) cosm_client driver on the card is + * unloaded. + * + * From the point of view of COSM interactions with SCIF during card + * shutdown, reset and crash are as follows: + * + * Card shutdown + * ------------- + * 1. COSM client on the card invokes orderly_poweroff() in response to SHUTDOWN + * message from the host. + * 2. Card driver shutdown callback invokes scif_unregister_device(..) resulting + * in scif_remove(..) getting called on the card + * 3. scif_remove -> scif_stop -> scif_handle_remove_node -> + * scif_peer_unregister_device -> device_unregister for the host peer device + * 4. During device_unregister remove(..) method of cosm_client is invoked which + * closes the COSM SCIF endpoint on the card. This results in a SCIF_DISCNCT + * message being sent to host SCIF. SCIF_DISCNCT message processing on the + * host SCIF sets the host COSM SCIF endpoint state to DISCONNECTED and wakes + * up the host COSM thread blocked in scif_poll(..) resulting in + * scif_poll(..) returning POLLHUP. + * 5. On the card, scif_peer_release_dev is next called which results in an + * SCIF_EXIT message being sent to the host and after receiving the + * SCIF_EXIT_ACK from the host the peer device teardown on the card is + * complete. + * 6. As part of the SCIF_EXIT message processing on the host, host sends a + * SCIF_REMOVE_NODE to itself corresponding to the card being removed. This + * starts a similar SCIF peer device teardown sequence on the host + * corresponding to the card being shut down. + * + * Card reset + * ---------- + * The case of interest here is when the card has not been previously shut down + * since most of the steps below are skipped in that case: + + * 1. cosm_stop(..) invokes hw_ops->stop(..) method of the base PCIe driver + * which unregisters the SCIF HW device resulting in scif_remove(..) being + * called on the host. + * 2. scif_remove(..) calls scif_disconnect_node(..) which results in a + * SCIF_EXIT message being sent to the card. + * 3. The card executes scif_stop() as part of SCIF_EXIT message + * processing. This results in the COSM endpoint on the card being closed and + * the SCIF host peer device on the card getting unregistered similar to + * steps 3, 4 and 5 for the card shutdown case above. scif_poll(..) on the + * host returns POLLHUP as a result. + * 4. On the host, card peer device unregister and SCIF HW remove(..) also + * subsequently complete. + * + * Card crash + * ---------- + * If a reset is issued after the card has crashed, there is no SCIF_DISCNT + * message from the card which would result in scif_poll(..) returning + * POLLHUP. In this case when the host SCIF driver sends a SCIF_REMOVE_NODE + * message to itself resulting in the card SCIF peer device being unregistered, + * this results in a scif_peer_release_dev -> scif_cleanup_scifdev-> + * scif_invalidate_ep call sequence which sets the endpoint state to + * DISCONNECTED and results in scif_poll(..) returning POLLHUP. + */ + +#define COSM_SCIF_BACKLOG 16 +#define COSM_HEARTBEAT_CHECK_DELTA_SEC 10 +#define COSM_HEARTBEAT_TIMEOUT_SEC \ + (COSM_HEARTBEAT_SEND_SEC + COSM_HEARTBEAT_CHECK_DELTA_SEC) +#define COSM_HEARTBEAT_TIMEOUT_MSEC (COSM_HEARTBEAT_TIMEOUT_SEC * MSEC_PER_SEC) + +static struct task_struct *server_thread; +static scif_epd_t listen_epd; + +/* Publish MIC card's shutdown status to user space MIC daemon */ +static void cosm_update_mic_status(struct cosm_device *cdev) +{ + if (cdev->shutdown_status_int != MIC_NOP) { + cosm_set_shutdown_status(cdev, cdev->shutdown_status_int); + cdev->shutdown_status_int = MIC_NOP; + } +} + +/* Store MIC card's shutdown status internally when it is received */ +static void cosm_shutdown_status_int(struct cosm_device *cdev, + enum mic_status shutdown_status) +{ + switch (shutdown_status) { + case MIC_HALTED: + case MIC_POWER_OFF: + case MIC_RESTART: + case MIC_CRASHED: + break; + default: + dev_err(&cdev->dev, "%s %d Unexpected shutdown_status %d\n", + __func__, __LINE__, shutdown_status); + return; + }; + cdev->shutdown_status_int = shutdown_status; + cdev->heartbeat_watchdog_enable = false; + + if (cdev->state != MIC_SHUTTING_DOWN) + cosm_set_state(cdev, MIC_SHUTTING_DOWN); +} + +/* Non-blocking recv. Read and process all available messages */ +static void cosm_scif_recv(struct cosm_device *cdev) +{ + struct cosm_msg msg; + int rc; + + while (1) { + rc = scif_recv(cdev->epd, &msg, sizeof(msg), 0); + if (!rc) { + break; + } else if (rc < 0) { + dev_dbg(&cdev->dev, "%s: %d rc %d\n", + __func__, __LINE__, rc); + break; + } + dev_dbg(&cdev->dev, "%s: %d rc %d id 0x%llx\n", + __func__, __LINE__, rc, msg.id); + + switch (msg.id) { + case COSM_MSG_SHUTDOWN_STATUS: + cosm_shutdown_status_int(cdev, msg.shutdown_status); + break; + case COSM_MSG_HEARTBEAT: + /* Nothing to do, heartbeat only unblocks scif_poll */ + break; + default: + dev_err(&cdev->dev, "%s: %d unknown msg.id %lld\n", + __func__, __LINE__, msg.id); + break; + } + } +} + +/* Publish crashed status for this MIC card */ +static void cosm_set_crashed(struct cosm_device *cdev) +{ + dev_err(&cdev->dev, "node alive timeout\n"); + cosm_shutdown_status_int(cdev, MIC_CRASHED); + cosm_update_mic_status(cdev); +} + +/* Send host time to the MIC card to sync system time between host and MIC */ +static void cosm_send_time(struct cosm_device *cdev) +{ + struct cosm_msg msg = { .id = COSM_MSG_SYNC_TIME }; + int rc; + + getnstimeofday64(&msg.timespec); + rc = scif_send(cdev->epd, &msg, sizeof(msg), SCIF_SEND_BLOCK); + if (rc < 0) + dev_err(&cdev->dev, "%s %d scif_send failed rc %d\n", + __func__, __LINE__, rc); +} + +/* + * Close this cosm_device's endpoint after its peer endpoint on the card has + * been closed. In all cases except MIC card crash POLLHUP on the host is + * triggered by the client's endpoint being closed. + */ +static void cosm_scif_close(struct cosm_device *cdev) +{ + /* + * Because SHUTDOWN_STATUS message is sent by the MIC cards in the + * reboot notifier when shutdown is still not complete, we notify mpssd + * to reset the card when SCIF endpoint is closed. + */ + cosm_update_mic_status(cdev); + scif_close(cdev->epd); + cdev->epd = NULL; + dev_dbg(&cdev->dev, "%s %d\n", __func__, __LINE__); +} + +/* + * Set card state to ONLINE when a new SCIF connection from a MIC card is + * received. Normally the state is BOOTING when the connection comes in, but can + * be ONLINE if cosm_client driver on the card was unloaded and then reloaded. + */ +static int cosm_set_online(struct cosm_device *cdev) +{ + int rc = 0; + + if (MIC_BOOTING == cdev->state || MIC_ONLINE == cdev->state) { + cdev->heartbeat_watchdog_enable = cdev->sysfs_heartbeat_enable; + cdev->epd = cdev->newepd; + if (cdev->state == MIC_BOOTING) + cosm_set_state(cdev, MIC_ONLINE); + cosm_send_time(cdev); + dev_dbg(&cdev->dev, "%s %d\n", __func__, __LINE__); + } else { + dev_warn(&cdev->dev, "%s %d not going online in state: %s\n", + __func__, __LINE__, cosm_state_string[cdev->state]); + rc = -EINVAL; + } + /* Drop reference acquired by bus_find_device in the server thread */ + put_device(&cdev->dev); + return rc; +} + +/* + * Work function for handling work for a SCIF connection from a particular MIC + * card. It first sets the card state to ONLINE and then calls scif_poll to + * block on activity such as incoming messages on the SCIF endpoint. When the + * endpoint is closed, the work function exits, completing its life cycle, from + * MIC card boot to card shutdown/reset/crash. + */ +void cosm_scif_work(struct work_struct *work) +{ + struct cosm_device *cdev = container_of(work, struct cosm_device, + scif_work); + struct scif_pollepd pollepd; + int rc; + + mutex_lock(&cdev->cosm_mutex); + if (cosm_set_online(cdev)) + goto exit; + + while (1) { + pollepd.epd = cdev->epd; + pollepd.events = POLLIN; + + /* Drop the mutex before blocking in scif_poll(..) */ + mutex_unlock(&cdev->cosm_mutex); + /* poll(..) with timeout on our endpoint */ + rc = scif_poll(&pollepd, 1, COSM_HEARTBEAT_TIMEOUT_MSEC); + mutex_lock(&cdev->cosm_mutex); + if (rc < 0) { + dev_err(&cdev->dev, "%s %d scif_poll rc %d\n", + __func__, __LINE__, rc); + continue; + } + + /* There is a message from the card */ + if (pollepd.revents & POLLIN) + cosm_scif_recv(cdev); + + /* The peer endpoint is closed or this endpoint disconnected */ + if (pollepd.revents & POLLHUP) { + cosm_scif_close(cdev); + break; + } + + /* Did we timeout from poll? */ + if (!rc && cdev->heartbeat_watchdog_enable) + cosm_set_crashed(cdev); + } +exit: + dev_dbg(&cdev->dev, "%s %d exiting\n", __func__, __LINE__); + mutex_unlock(&cdev->cosm_mutex); +} + +/* + * COSM SCIF server thread function. Accepts incoming SCIF connections from MIC + * cards, finds the correct cosm_device to associate that connection with and + * schedules individual work items for each MIC card. + */ +static int cosm_scif_server(void *unused) +{ + struct cosm_device *cdev; + scif_epd_t newepd; + struct scif_port_id port_id; + int rc; + + allow_signal(SIGKILL); + + while (!kthread_should_stop()) { + rc = scif_accept(listen_epd, &port_id, &newepd, + SCIF_ACCEPT_SYNC); + if (rc < 0) { + if (-ERESTARTSYS != rc) + pr_err("%s %d rc %d\n", __func__, __LINE__, rc); + continue; + } + + /* + * Associate the incoming connection with a particular + * cosm_device, COSM device ID == SCIF node ID - 1 + */ + cdev = cosm_find_cdev_by_id(port_id.node - 1); + if (!cdev) + continue; + cdev->newepd = newepd; + schedule_work(&cdev->scif_work); + } + + pr_debug("%s %d Server thread stopped\n", __func__, __LINE__); + return 0; +} + +static int cosm_scif_listen(void) +{ + int rc; + + listen_epd = scif_open(); + if (!listen_epd) { + pr_err("%s %d scif_open failed\n", __func__, __LINE__); + return -ENOMEM; + } + + rc = scif_bind(listen_epd, SCIF_COSM_LISTEN_PORT); + if (rc < 0) { + pr_err("%s %d scif_bind failed rc %d\n", + __func__, __LINE__, rc); + goto err; + } + + rc = scif_listen(listen_epd, COSM_SCIF_BACKLOG); + if (rc < 0) { + pr_err("%s %d scif_listen rc %d\n", __func__, __LINE__, rc); + goto err; + } + pr_debug("%s %d listen_epd set up\n", __func__, __LINE__); + return 0; +err: + scif_close(listen_epd); + listen_epd = NULL; + return rc; +} + +static void cosm_scif_listen_exit(void) +{ + pr_debug("%s %d closing listen_epd\n", __func__, __LINE__); + if (listen_epd) { + scif_close(listen_epd); + listen_epd = NULL; + } +} + +/* + * Create a listening SCIF endpoint and a server kthread which accepts incoming + * SCIF connections from MIC cards + */ +int cosm_scif_init(void) +{ + int rc = cosm_scif_listen(); + + if (rc) { + pr_err("%s %d cosm_scif_listen rc %d\n", + __func__, __LINE__, rc); + goto err; + } + + server_thread = kthread_run(cosm_scif_server, NULL, "cosm_server"); + if (IS_ERR(server_thread)) { + rc = PTR_ERR(server_thread); + pr_err("%s %d kthread_run rc %d\n", __func__, __LINE__, rc); + goto listen_exit; + } + return 0; +listen_exit: + cosm_scif_listen_exit(); +err: + return rc; +} + +/* Stop the running server thread and close the listening SCIF endpoint */ +void cosm_scif_exit(void) +{ + int rc; + + if (!IS_ERR_OR_NULL(server_thread)) { + rc = send_sig(SIGKILL, server_thread, 0); + if (rc) { + pr_err("%s %d send_sig rc %d\n", + __func__, __LINE__, rc); + return; + } + kthread_stop(server_thread); + } + + cosm_scif_listen_exit(); +} diff --git a/drivers/misc/mic/cosm/cosm_sysfs.c b/drivers/misc/mic/cosm/cosm_sysfs.c new file mode 100644 index 000000000..29d6863b6 --- /dev/null +++ b/drivers/misc/mic/cosm/cosm_sysfs.c @@ -0,0 +1,461 @@ +/* + * Intel MIC Platform Software Stack (MPSS) + * + * Copyright(c) 2015 Intel Corporation. + * + * 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. + * + * The full GNU General Public License is included in this distribution in + * the file called "COPYING". + * + * Intel MIC Coprocessor State Management (COSM) Driver + * + */ +#include <linux/slab.h> +#include "cosm_main.h" + +/* + * A state-to-string lookup table, for exposing a human readable state + * via sysfs. Always keep in sync with enum cosm_states + */ +const char * const cosm_state_string[] = { + [MIC_READY] = "ready", + [MIC_BOOTING] = "booting", + [MIC_ONLINE] = "online", + [MIC_SHUTTING_DOWN] = "shutting_down", + [MIC_RESETTING] = "resetting", + [MIC_RESET_FAILED] = "reset_failed", +}; + +/* + * A shutdown-status-to-string lookup table, for exposing a human + * readable state via sysfs. Always keep in sync with enum cosm_shutdown_status + */ +const char * const cosm_shutdown_status_string[] = { + [MIC_NOP] = "nop", + [MIC_CRASHED] = "crashed", + [MIC_HALTED] = "halted", + [MIC_POWER_OFF] = "poweroff", + [MIC_RESTART] = "restart", +}; + +void cosm_set_shutdown_status(struct cosm_device *cdev, u8 shutdown_status) +{ + dev_dbg(&cdev->dev, "Shutdown Status %s -> %s\n", + cosm_shutdown_status_string[cdev->shutdown_status], + cosm_shutdown_status_string[shutdown_status]); + cdev->shutdown_status = shutdown_status; +} + +void cosm_set_state(struct cosm_device *cdev, u8 state) +{ + dev_dbg(&cdev->dev, "State %s -> %s\n", + cosm_state_string[cdev->state], + cosm_state_string[state]); + cdev->state = state; + sysfs_notify_dirent(cdev->state_sysfs); +} + +static ssize_t +family_show(struct device *dev, struct device_attribute *attr, char *buf) +{ + struct cosm_device *cdev = dev_get_drvdata(dev); + + if (!cdev) + return -EINVAL; + + return cdev->hw_ops->family(cdev, buf); +} +static DEVICE_ATTR_RO(family); + +static ssize_t +stepping_show(struct device *dev, struct device_attribute *attr, char *buf) +{ + struct cosm_device *cdev = dev_get_drvdata(dev); + + if (!cdev) + return -EINVAL; + + return cdev->hw_ops->stepping(cdev, buf); +} +static DEVICE_ATTR_RO(stepping); + +static ssize_t +state_show(struct device *dev, struct device_attribute *attr, char *buf) +{ + struct cosm_device *cdev = dev_get_drvdata(dev); + + if (!cdev || cdev->state >= MIC_LAST) + return -EINVAL; + + return scnprintf(buf, PAGE_SIZE, "%s\n", + cosm_state_string[cdev->state]); +} + +static ssize_t +state_store(struct device *dev, struct device_attribute *attr, + const char *buf, size_t count) +{ + struct cosm_device *cdev = dev_get_drvdata(dev); + int rc; + + if (!cdev) + return -EINVAL; + + if (sysfs_streq(buf, "boot")) { + rc = cosm_start(cdev); + goto done; + } + if (sysfs_streq(buf, "reset")) { + rc = cosm_reset(cdev); + goto done; + } + + if (sysfs_streq(buf, "shutdown")) { + rc = cosm_shutdown(cdev); + goto done; + } + rc = -EINVAL; +done: + if (rc) + count = rc; + return count; +} +static DEVICE_ATTR_RW(state); + +static ssize_t shutdown_status_show(struct device *dev, + struct device_attribute *attr, char *buf) +{ + struct cosm_device *cdev = dev_get_drvdata(dev); + + if (!cdev || cdev->shutdown_status >= MIC_STATUS_LAST) + return -EINVAL; + + return scnprintf(buf, PAGE_SIZE, "%s\n", + cosm_shutdown_status_string[cdev->shutdown_status]); +} +static DEVICE_ATTR_RO(shutdown_status); + +static ssize_t +heartbeat_enable_show(struct device *dev, + struct device_attribute *attr, char *buf) +{ + struct cosm_device *cdev = dev_get_drvdata(dev); + + if (!cdev) + return -EINVAL; + + return scnprintf(buf, PAGE_SIZE, "%d\n", cdev->sysfs_heartbeat_enable); +} + +static ssize_t +heartbeat_enable_store(struct device *dev, + struct device_attribute *attr, + const char *buf, size_t count) +{ + struct cosm_device *cdev = dev_get_drvdata(dev); + int enable; + int ret; + + if (!cdev) + return -EINVAL; + + mutex_lock(&cdev->cosm_mutex); + ret = kstrtoint(buf, 10, &enable); + if (ret) + goto unlock; + + cdev->sysfs_heartbeat_enable = enable; + /* if state is not online, cdev->heartbeat_watchdog_enable is 0 */ + if (cdev->state == MIC_ONLINE) + cdev->heartbeat_watchdog_enable = enable; + ret = count; +unlock: + mutex_unlock(&cdev->cosm_mutex); + return ret; +} +static DEVICE_ATTR_RW(heartbeat_enable); + +static ssize_t +cmdline_show(struct device *dev, struct device_attribute *attr, char *buf) +{ + struct cosm_device *cdev = dev_get_drvdata(dev); + char *cmdline; + + if (!cdev) + return -EINVAL; + + cmdline = cdev->cmdline; + + if (cmdline) + return scnprintf(buf, PAGE_SIZE, "%s\n", cmdline); + return 0; +} + +static ssize_t +cmdline_store(struct device *dev, struct device_attribute *attr, + const char *buf, size_t count) +{ + struct cosm_device *cdev = dev_get_drvdata(dev); + + if (!cdev) + return -EINVAL; + + mutex_lock(&cdev->cosm_mutex); + kfree(cdev->cmdline); + + cdev->cmdline = kmalloc(count + 1, GFP_KERNEL); + if (!cdev->cmdline) { + count = -ENOMEM; + goto unlock; + } + + strncpy(cdev->cmdline, buf, count); + + if (cdev->cmdline[count - 1] == '\n') + cdev->cmdline[count - 1] = '\0'; + else + cdev->cmdline[count] = '\0'; +unlock: + mutex_unlock(&cdev->cosm_mutex); + return count; +} +static DEVICE_ATTR_RW(cmdline); + +static ssize_t +firmware_show(struct device *dev, struct device_attribute *attr, char *buf) +{ + struct cosm_device *cdev = dev_get_drvdata(dev); + char *firmware; + + if (!cdev) + return -EINVAL; + + firmware = cdev->firmware; + + if (firmware) + return scnprintf(buf, PAGE_SIZE, "%s\n", firmware); + return 0; +} + +static ssize_t +firmware_store(struct device *dev, struct device_attribute *attr, + const char *buf, size_t count) +{ + struct cosm_device *cdev = dev_get_drvdata(dev); + + if (!cdev) + return -EINVAL; + + mutex_lock(&cdev->cosm_mutex); + kfree(cdev->firmware); + + cdev->firmware = kmalloc(count + 1, GFP_KERNEL); + if (!cdev->firmware) { + count = -ENOMEM; + goto unlock; + } + strncpy(cdev->firmware, buf, count); + + if (cdev->firmware[count - 1] == '\n') + cdev->firmware[count - 1] = '\0'; + else + cdev->firmware[count] = '\0'; +unlock: + mutex_unlock(&cdev->cosm_mutex); + return count; +} +static DEVICE_ATTR_RW(firmware); + +static ssize_t +ramdisk_show(struct device *dev, struct device_attribute *attr, char *buf) +{ + struct cosm_device *cdev = dev_get_drvdata(dev); + char *ramdisk; + + if (!cdev) + return -EINVAL; + + ramdisk = cdev->ramdisk; + + if (ramdisk) + return scnprintf(buf, PAGE_SIZE, "%s\n", ramdisk); + return 0; +} + +static ssize_t +ramdisk_store(struct device *dev, struct device_attribute *attr, + const char *buf, size_t count) +{ + struct cosm_device *cdev = dev_get_drvdata(dev); + + if (!cdev) + return -EINVAL; + + mutex_lock(&cdev->cosm_mutex); + kfree(cdev->ramdisk); + + cdev->ramdisk = kmalloc(count + 1, GFP_KERNEL); + if (!cdev->ramdisk) { + count = -ENOMEM; + goto unlock; + } + + strncpy(cdev->ramdisk, buf, count); + + if (cdev->ramdisk[count - 1] == '\n') + cdev->ramdisk[count - 1] = '\0'; + else + cdev->ramdisk[count] = '\0'; +unlock: + mutex_unlock(&cdev->cosm_mutex); + return count; +} +static DEVICE_ATTR_RW(ramdisk); + +static ssize_t +bootmode_show(struct device *dev, struct device_attribute *attr, char *buf) +{ + struct cosm_device *cdev = dev_get_drvdata(dev); + char *bootmode; + + if (!cdev) + return -EINVAL; + + bootmode = cdev->bootmode; + + if (bootmode) + return scnprintf(buf, PAGE_SIZE, "%s\n", bootmode); + return 0; +} + +static ssize_t +bootmode_store(struct device *dev, struct device_attribute *attr, + const char *buf, size_t count) +{ + struct cosm_device *cdev = dev_get_drvdata(dev); + + if (!cdev) + return -EINVAL; + + if (!sysfs_streq(buf, "linux") && !sysfs_streq(buf, "flash")) + return -EINVAL; + + mutex_lock(&cdev->cosm_mutex); + kfree(cdev->bootmode); + + cdev->bootmode = kmalloc(count + 1, GFP_KERNEL); + if (!cdev->bootmode) { + count = -ENOMEM; + goto unlock; + } + + strncpy(cdev->bootmode, buf, count); + + if (cdev->bootmode[count - 1] == '\n') + cdev->bootmode[count - 1] = '\0'; + else + cdev->bootmode[count] = '\0'; +unlock: + mutex_unlock(&cdev->cosm_mutex); + return count; +} +static DEVICE_ATTR_RW(bootmode); + +static ssize_t +log_buf_addr_show(struct device *dev, struct device_attribute *attr, + char *buf) +{ + struct cosm_device *cdev = dev_get_drvdata(dev); + + if (!cdev) + return -EINVAL; + + return scnprintf(buf, PAGE_SIZE, "%p\n", cdev->log_buf_addr); +} + +static ssize_t +log_buf_addr_store(struct device *dev, struct device_attribute *attr, + const char *buf, size_t count) +{ + struct cosm_device *cdev = dev_get_drvdata(dev); + int ret; + unsigned long addr; + + if (!cdev) + return -EINVAL; + + ret = kstrtoul(buf, 16, &addr); + if (ret) + goto exit; + + cdev->log_buf_addr = (void *)addr; + ret = count; +exit: + return ret; +} +static DEVICE_ATTR_RW(log_buf_addr); + +static ssize_t +log_buf_len_show(struct device *dev, struct device_attribute *attr, + char *buf) +{ + struct cosm_device *cdev = dev_get_drvdata(dev); + + if (!cdev) + return -EINVAL; + + return scnprintf(buf, PAGE_SIZE, "%p\n", cdev->log_buf_len); +} + +static ssize_t +log_buf_len_store(struct device *dev, struct device_attribute *attr, + const char *buf, size_t count) +{ + struct cosm_device *cdev = dev_get_drvdata(dev); + int ret; + unsigned long addr; + + if (!cdev) + return -EINVAL; + + ret = kstrtoul(buf, 16, &addr); + if (ret) + goto exit; + + cdev->log_buf_len = (int *)addr; + ret = count; +exit: + return ret; +} +static DEVICE_ATTR_RW(log_buf_len); + +static struct attribute *cosm_default_attrs[] = { + &dev_attr_family.attr, + &dev_attr_stepping.attr, + &dev_attr_state.attr, + &dev_attr_shutdown_status.attr, + &dev_attr_heartbeat_enable.attr, + &dev_attr_cmdline.attr, + &dev_attr_firmware.attr, + &dev_attr_ramdisk.attr, + &dev_attr_bootmode.attr, + &dev_attr_log_buf_addr.attr, + &dev_attr_log_buf_len.attr, + + NULL +}; + +ATTRIBUTE_GROUPS(cosm_default); + +void cosm_sysfs_init(struct cosm_device *cdev) +{ + cdev->attr_group = cosm_default_groups; +} |