diff options
Diffstat (limited to 'drivers/base/devcoredump.c')
| -rw-r--r-- | drivers/base/devcoredump.c | 244 |
1 files changed, 176 insertions, 68 deletions
diff --git a/drivers/base/devcoredump.c b/drivers/base/devcoredump.c index 7be310f7db73..55bdc7f5e59d 100644 --- a/drivers/base/devcoredump.c +++ b/drivers/base/devcoredump.c @@ -1,27 +1,8 @@ +// SPDX-License-Identifier: GPL-2.0 /* - * This file is provided under the GPLv2 license. - * - * GPL LICENSE SUMMARY - * * Copyright(c) 2014 Intel Mobile Communications GmbH * Copyright(c) 2015 Intel Deutschland GmbH * - * This program is free software; you can redistribute it and/or modify - * it under the terms of version 2 of the GNU General Public License 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. - * - * Contact Information: - * Intel Linux Wireless <ilw@linux.intel.com> - * Intel Corporation, 5200 N.E. Elam Young Parkway, Hillsboro, OR 97124-6497 - * * Author: Johannes Berg <johannes@sipsolutions.net> */ #include <linux/module.h> @@ -37,17 +18,51 @@ static struct class devcd_class; /* global disable flag, for security purposes */ static bool devcd_disabled; -/* if data isn't read by userspace after 5 minutes then delete it */ -#define DEVCD_TIMEOUT (HZ * 60 * 5) - struct devcd_entry { struct device devcd_dev; void *data; size_t datalen; + /* + * There are 2 races for which mutex is required. + * + * The first race is between device creation and userspace writing to + * schedule immediately destruction. + * + * This race is handled by arming the timer before device creation, but + * when device creation fails the timer still exists. + * + * To solve this, hold the mutex during device_add(), and set + * init_completed on success before releasing the mutex. + * + * That way the timer will never fire until device_add() is called, + * it will do nothing if init_completed is not set. The timer is also + * cancelled in that case. + * + * The second race involves multiple parallel invocations of devcd_free(), + * add a deleted flag so only 1 can call the destructor. + */ + struct mutex mutex; + bool init_completed, deleted; struct module *owner; ssize_t (*read)(char *buffer, loff_t offset, size_t count, void *data, size_t datalen); void (*free)(void *data); + /* + * If nothing interferes and device_add() was returns success, + * del_wk will destroy the device after the timer fires. + * + * Multiple userspace processes can interfere in the working of the timer: + * - Writing to the coredump will reschedule the timer to run immediately, + * if still armed. + * + * This is handled by using "if (cancel_delayed_work()) { + * schedule_delayed_work() }", to prevent re-arming after having + * been previously fired. + * - Writing to /sys/class/devcoredump/disabled will destroy the + * coredump synchronously. + * This is handled by using disable_delayed_work_sync(), and then + * checking if deleted flag is set with &devcd->mutex held. + */ struct delayed_work del_wk; struct device *failing_dev; }; @@ -76,18 +91,31 @@ static void devcd_dev_release(struct device *dev) kfree(devcd); } +static void __devcd_del(struct devcd_entry *devcd) +{ + devcd->deleted = true; + device_del(&devcd->devcd_dev); + put_device(&devcd->devcd_dev); +} + static void devcd_del(struct work_struct *wk) { struct devcd_entry *devcd; + bool init_completed; devcd = container_of(wk, struct devcd_entry, del_wk.work); - device_del(&devcd->devcd_dev); - put_device(&devcd->devcd_dev); + /* devcd->mutex serializes against dev_coredumpm_timeout */ + mutex_lock(&devcd->mutex); + init_completed = devcd->init_completed; + mutex_unlock(&devcd->mutex); + + if (init_completed) + __devcd_del(devcd); } static ssize_t devcd_data_read(struct file *filp, struct kobject *kobj, - struct bin_attribute *bin_attr, + const struct bin_attribute *bin_attr, char *buffer, loff_t offset, size_t count) { struct device *dev = kobj_to_dev(kobj); @@ -97,25 +125,26 @@ static ssize_t devcd_data_read(struct file *filp, struct kobject *kobj, } static ssize_t devcd_data_write(struct file *filp, struct kobject *kobj, - struct bin_attribute *bin_attr, + const struct bin_attribute *bin_attr, char *buffer, loff_t offset, size_t count) { struct device *dev = kobj_to_dev(kobj); struct devcd_entry *devcd = dev_to_devcd(dev); - mod_delayed_work(system_wq, &devcd->del_wk, 0); + /* + * Although it's tempting to use mod_delayed work here, + * that will cause a reschedule if the timer already fired. + */ + if (cancel_delayed_work(&devcd->del_wk)) + schedule_delayed_work(&devcd->del_wk, 0); return count; } -static struct bin_attribute devcd_attr_data = { - .attr = { .name = "data", .mode = S_IRUSR | S_IWUSR, }, - .size = 0, - .read = devcd_data_read, - .write = devcd_data_write, -}; +static const struct bin_attribute devcd_attr_data = + __BIN_ATTR(data, 0600, devcd_data_read, devcd_data_write, 0); -static struct bin_attribute *devcd_dev_bin_attrs[] = { +static const struct bin_attribute *const devcd_dev_bin_attrs[] = { &devcd_attr_data, NULL, }; @@ -131,17 +160,54 @@ static int devcd_free(struct device *dev, void *data) { struct devcd_entry *devcd = dev_to_devcd(dev); - flush_delayed_work(&devcd->del_wk); + /* + * To prevent a race with devcd_data_write(), disable work and + * complete manually instead. + * + * We cannot rely on the return value of + * disable_delayed_work_sync() here, because it might be in the + * middle of a cancel_delayed_work + schedule_delayed_work pair. + * + * devcd->mutex here guards against multiple parallel invocations + * of devcd_free(). + */ + disable_delayed_work_sync(&devcd->del_wk); + mutex_lock(&devcd->mutex); + if (!devcd->deleted) + __devcd_del(devcd); + mutex_unlock(&devcd->mutex); return 0; } -static ssize_t disabled_show(struct class *class, struct class_attribute *attr, +static ssize_t disabled_show(const struct class *class, const struct class_attribute *attr, char *buf) { - return sprintf(buf, "%d\n", devcd_disabled); + return sysfs_emit(buf, "%d\n", devcd_disabled); } -static ssize_t disabled_store(struct class *class, struct class_attribute *attr, +/* + * + * disabled_store() worker() + * class_for_each_device(&devcd_class, + * NULL, NULL, devcd_free) + * ... + * ... + * while ((dev = class_dev_iter_next(&iter)) + * devcd_del() + * device_del() + * put_device() <- last reference + * error = fn(dev, data) devcd_dev_release() + * devcd_free(dev, data) kfree(devcd) + * + * + * In the above diagram, it looks like disabled_store() would be racing with parallelly + * running devcd_del() and result in memory abort after dropping its last reference with + * put_device(). However, this will not happens as fn(dev, data) runs + * with its own reference to device via klist_node so it is not its last reference. + * so, above situation would not occur. + */ + +static ssize_t disabled_store(const struct class *class, const struct class_attribute *attr, const char *buf, size_t count) { long tmp = simple_strtol(buf, NULL, 10); @@ -170,7 +236,6 @@ ATTRIBUTE_GROUPS(devcd_class); static struct class devcd_class = { .name = "devcoredump", - .owner = THIS_MODULE, .dev_release = devcd_dev_release, .dev_groups = devcd_dev_groups, .class_groups = devcd_class_groups, @@ -179,16 +244,7 @@ static struct class devcd_class = { static ssize_t devcd_readv(char *buffer, loff_t offset, size_t count, void *data, size_t datalen) { - if (offset > datalen) - return -EINVAL; - - if (offset + count > datalen) - count = datalen - offset; - - if (count) - memcpy(buffer, ((u8 *)data) + offset, count); - - return count; + return memory_read_from_buffer(buffer, count, &offset, data, datalen); } static void devcd_freev(void *data) @@ -226,7 +282,7 @@ static int devcd_match_failing(struct device *dev, const void *failing) * NOTE: if two tables allocated with devcd_alloc_sgtable and then chained * using the sg_chain function then that function should be called only once * on the chained table - * @table: pointer to sg_table to free + * @data: pointer to sg_table to free */ static void devcd_free_sgtable(void *data) { @@ -234,7 +290,7 @@ static void devcd_free_sgtable(void *data) } /** - * devcd_read_from_table - copy data from sg_table to a given buffer + * devcd_read_from_sgtable - copy data from sg_table to a given buffer * and return the number of bytes read * @buffer: the buffer to copy the data to it * @buf_len: the length of the buffer @@ -242,6 +298,8 @@ static void devcd_free_sgtable(void *data) * @offset: start copy from @offset@ bytes from the head of the data * in the given scatterlist * @data_len: the length of the data in the sg_table + * + * Returns: the number of bytes copied */ static ssize_t devcd_read_from_sgtable(char *buffer, loff_t offset, size_t buf_len, void *data, @@ -259,7 +317,31 @@ static ssize_t devcd_read_from_sgtable(char *buffer, loff_t offset, } /** - * dev_coredumpm - create device coredump with read/free methods + * dev_coredump_put - remove device coredump + * @dev: the struct device for the crashed device + * + * dev_coredump_put() removes coredump, if exists, for a given device from + * the file system and free its associated data otherwise, does nothing. + * + * It is useful for modules that do not want to keep coredump + * available after its unload. + */ +void dev_coredump_put(struct device *dev) +{ + struct device *existing; + + existing = class_find_device(&devcd_class, NULL, dev, + devcd_match_failing); + if (existing) { + devcd_free(existing, NULL); + put_device(existing); + } +} +EXPORT_SYMBOL_GPL(dev_coredump_put); + +/** + * dev_coredumpm_timeout - create device coredump with read/free methods with a + * custom timeout. * @dev: the struct device for the crashed device * @owner: the module that contains the read/free functions, use %THIS_MODULE * @data: data cookie for the @read/@free functions @@ -267,17 +349,20 @@ static ssize_t devcd_read_from_sgtable(char *buffer, loff_t offset, * @gfp: allocation flags * @read: function to read from the given buffer * @free: function to free the given buffer + * @timeout: time in jiffies to remove coredump * * Creates a new device coredump for the given device. If a previous one hasn't * been read yet, the new coredump is discarded. The data lifetime is determined * by the device coredump framework and when it is no longer needed the @free * function will be called to free the data. */ -void dev_coredumpm(struct device *dev, struct module *owner, - void *data, size_t datalen, gfp_t gfp, - ssize_t (*read)(char *buffer, loff_t offset, size_t count, - void *data, size_t datalen), - void (*free)(void *data)) +void dev_coredumpm_timeout(struct device *dev, struct module *owner, + void *data, size_t datalen, gfp_t gfp, + ssize_t (*read)(char *buffer, loff_t offset, + size_t count, void *data, + size_t datalen), + void (*free)(void *data), + unsigned long timeout) { static atomic_t devcd_count = ATOMIC_INIT(0); struct devcd_entry *devcd; @@ -306,39 +391,62 @@ void dev_coredumpm(struct device *dev, struct module *owner, devcd->read = read; devcd->free = free; devcd->failing_dev = get_device(dev); + devcd->deleted = false; + mutex_init(&devcd->mutex); device_initialize(&devcd->devcd_dev); dev_set_name(&devcd->devcd_dev, "devcd%d", atomic_inc_return(&devcd_count)); devcd->devcd_dev.class = &devcd_class; + dev_set_uevent_suppress(&devcd->devcd_dev, true); + + /* devcd->mutex prevents devcd_del() completing until init finishes */ + mutex_lock(&devcd->mutex); + devcd->init_completed = false; + INIT_DELAYED_WORK(&devcd->del_wk, devcd_del); + schedule_delayed_work(&devcd->del_wk, timeout); + if (device_add(&devcd->devcd_dev)) goto put_device; + /* + * These should normally not fail, but there is no problem + * continuing without the links, so just warn instead of + * failing. + */ if (sysfs_create_link(&devcd->devcd_dev.kobj, &dev->kobj, - "failing_device")) - /* nothing - symlink will be missing */; + "failing_device") || + sysfs_create_link(&dev->kobj, &devcd->devcd_dev.kobj, + "devcoredump")) + dev_warn(dev, "devcoredump create_link failed\n"); - if (sysfs_create_link(&dev->kobj, &devcd->devcd_dev.kobj, - "devcoredump")) - /* nothing - symlink will be missing */; - - INIT_DELAYED_WORK(&devcd->del_wk, devcd_del); - schedule_delayed_work(&devcd->del_wk, DEVCD_TIMEOUT); + dev_set_uevent_suppress(&devcd->devcd_dev, false); + kobject_uevent(&devcd->devcd_dev.kobj, KOBJ_ADD); + /* + * Safe to run devcd_del() now that we are done with devcd_dev. + * Alternatively we could have taken a ref on devcd_dev before + * dropping the lock. + */ + devcd->init_completed = true; + mutex_unlock(&devcd->mutex); return; put_device: + mutex_unlock(&devcd->mutex); + cancel_delayed_work_sync(&devcd->del_wk); put_device(&devcd->devcd_dev); + put_module: module_put(owner); free: free(data); } -EXPORT_SYMBOL_GPL(dev_coredumpm); +EXPORT_SYMBOL_GPL(dev_coredumpm_timeout); /** - * dev_coredumpmsg - create device coredump that uses scatterlist as data + * dev_coredumpsg - create device coredump that uses scatterlist as data * parameter * @dev: the struct device for the crashed device * @table: the dump data |