From a5f7a5bb3b7f28ba7e4fa246775b29a0e5537255 Mon Sep 17 00:00:00 2001 From: Jinke Han Date: Tue, 8 Sep 2026 15:37:42 +0800 Subject: x86/kprobes: Fix crash when probing CS CALL instructions When using eBPF to probe CS CALL instructions within a function, a crash can be triggered. The eBPF tool probes offset 257 of the __hrtimer_run_queues() function: <__hrtimer_run_queues+249>: nopl 0x0(%rax,%rax,1) <__hrtimer_run_queues+254>: mov %r14,%rdi <__hrtimer_run_queues+257>: cs call <__x86_indirect_thunk_r12> <__hrtimer_run_queues+263>: mov %eax,%r12d <__hrtimer_run_queues+266>: xchg %ax,%ax <__hrtimer_run_queues+268>: mov %r13,%rdi Which triggers this crash: BUG: unable to handle page fault for address: 00000000000f41c9 #PF: supervisor write access in kernel mode #PF: error_code(0x0002) - not-present page PGD 0 P4D 0 Oops: 0002 [#1] SMP NOPTI CPU: 1 PID: 0 Comm: swapper/1 Kdump: loaded Tainted: P RIP: 0010:__hrtimer_run_queues+0x106/0x230 Note that __hrtimer_run_queues+0x106 is __hrtimer_run_queues+262, which is at the 6th byte of the above CS CALL instruction. Since the CS CALL instruction occupies 6 bytes, the exception occurred in the middle of that call instruction. The root cause is that when using eBPF tools to probe in the middle of a function, a kprobe with INT3 is used as the underlying implementation. During single-step emulation of the original CALL instruction, int3_emulate_call() assumes that the probed CALL instruction is 5 bytes long. However, the actual CS-prefixed CALL instruction occupies 6 bytes, so it constructs an incorrect exception return address. When the CPU returns from the kprobe handler, the next instruction to be executed is at the address of the last byte of that CS CALL instruction. Coincidentally, starting from that address, the CPU fetches and decodes a completely different instruction, which ultimately triggers a kernel crash. Fix the issue by using the actual instruction length obtained from the instruction decoder when constructing the exception return address, rather than relying on the hardcoded CALL_INSN_SIZE macro. [ mingo: Refined the changelog ] Fixes: 6256e668b7af ("x86/kprobes: Use int3 instead of debug trap for single-step") Suggested-by: Masami Hiramatsu (Google) Signed-off-by: Jinke Han Signed-off-by: Ingo Molnar Reviewed-by: Masami Hiramatsu (Google) Acked-by: Yafang Shao Acked-by: Borislav Petkov Cc: Peter Zijlstra Link: https://patch.msgid.link/20260908073742.GA10517@didi-ThinkCentre-M920t-N000 --- drivers/base/regmap/regmap-debugfs.c | 704 +++++++++++++++++++++++++++++++++++ 1 file changed, 704 insertions(+) create mode 100644 drivers/base/regmap/regmap-debugfs.c (limited to 'drivers/base/regmap/regmap-debugfs.c') diff --git a/drivers/base/regmap/regmap-debugfs.c b/drivers/base/regmap/regmap-debugfs.c new file mode 100644 index 000000000..18f1c6074 --- /dev/null +++ b/drivers/base/regmap/regmap-debugfs.c @@ -0,0 +1,704 @@ +// SPDX-License-Identifier: GPL-2.0 +// +// Register map access API - debugfs +// +// Copyright 2011 Wolfson Microelectronics plc +// +// Author: Mark Brown + +#include +#include +#include +#include +#include +#include +#include + +#include "internal.h" + +struct regmap_debugfs_node { + struct regmap *map; + struct list_head link; +}; + +static DEFINE_IDA(dummy_ida); +static struct dentry *regmap_debugfs_root; +static LIST_HEAD(regmap_debugfs_early_list); +static DEFINE_MUTEX(regmap_debugfs_early_lock); + +/* Calculate the length of a fixed format */ +static size_t regmap_calc_reg_len(int max_val) +{ + return snprintf(NULL, 0, "%x", max_val); +} + +static ssize_t regmap_name_read_file(struct file *file, + char __user *user_buf, size_t count, + loff_t *ppos) +{ + struct regmap *map = file->private_data; + const char *name = "nodev"; + int ret; + char *buf; + + buf = kmalloc(PAGE_SIZE, GFP_KERNEL); + if (!buf) + return -ENOMEM; + + if (map->dev && map->dev->driver) + name = map->dev->driver->name; + + ret = snprintf(buf, PAGE_SIZE, "%s\n", name); + if (ret >= PAGE_SIZE) { + kfree(buf); + return ret; + } + + ret = simple_read_from_buffer(user_buf, count, ppos, buf, ret); + kfree(buf); + return ret; +} + +static const struct file_operations regmap_name_fops = { + .open = simple_open, + .read = regmap_name_read_file, + .llseek = default_llseek, +}; + +static void regmap_debugfs_free_dump_cache(struct regmap *map) +{ + struct regmap_debugfs_off_cache *c; + + while (!list_empty(&map->debugfs_off_cache)) { + c = list_first_entry(&map->debugfs_off_cache, + struct regmap_debugfs_off_cache, + list); + list_del(&c->list); + kfree(c); + } +} + +static bool regmap_printable(struct regmap *map, unsigned int reg) +{ + if (regmap_precious(map, reg)) + return false; + + if (!regmap_readable(map, reg) && !regmap_cached(map, reg)) + return false; + + return true; +} + +/* + * Work out where the start offset maps into register numbers, bearing + * in mind that we suppress hidden registers. + */ +static unsigned int regmap_debugfs_get_dump_start(struct regmap *map, + unsigned int base, + loff_t from, + loff_t *pos) +{ + struct regmap_debugfs_off_cache *c = NULL; + loff_t p = 0; + unsigned int i, ret; + unsigned int fpos_offset; + unsigned int reg_offset; + + /* Suppress the cache if we're using a subrange */ + if (base) + return base; + + /* + * If we don't have a cache build one so we don't have to do a + * linear scan each time. + */ + mutex_lock(&map->cache_lock); + i = base; + if (list_empty(&map->debugfs_off_cache)) { + for (; i <= map->max_register; i += map->reg_stride) { + /* Skip unprinted registers, closing off cache entry */ + if (!regmap_printable(map, i)) { + if (c) { + c->max = p - 1; + c->max_reg = i - map->reg_stride; + list_add_tail(&c->list, + &map->debugfs_off_cache); + c = NULL; + } + + continue; + } + + /* No cache entry? Start a new one */ + if (!c) { + c = kzalloc_obj(*c); + if (!c) { + regmap_debugfs_free_dump_cache(map); + mutex_unlock(&map->cache_lock); + return base; + } + c->min = p; + c->base_reg = i; + } + + p += map->debugfs_tot_len; + } + } + + /* Close the last entry off if we didn't scan beyond it */ + if (c) { + c->max = p - 1; + c->max_reg = i - map->reg_stride; + list_add_tail(&c->list, + &map->debugfs_off_cache); + } + + /* + * This should never happen; we return above if we fail to + * allocate and we should never be in this code if there are + * no registers at all. + */ + WARN_ON(list_empty(&map->debugfs_off_cache)); + ret = base; + + /* Find the relevant block:offset */ + list_for_each_entry(c, &map->debugfs_off_cache, list) { + if (from >= c->min && from <= c->max) { + fpos_offset = from - c->min; + reg_offset = fpos_offset / map->debugfs_tot_len; + *pos = c->min + (reg_offset * map->debugfs_tot_len); + mutex_unlock(&map->cache_lock); + return c->base_reg + (reg_offset * map->reg_stride); + } + + *pos = c->max; + ret = c->max_reg; + } + mutex_unlock(&map->cache_lock); + + return ret; +} + +static inline void regmap_calc_tot_len(struct regmap *map, + void *buf, size_t count) +{ + /* Calculate the length of a fixed format */ + if (!map->debugfs_tot_len) { + map->debugfs_reg_len = regmap_calc_reg_len(map->max_register); + map->debugfs_val_len = 2 * map->format.val_bytes; + map->debugfs_tot_len = map->debugfs_reg_len + + map->debugfs_val_len + 3; /* : \n */ + } +} + +static int regmap_next_readable_reg(struct regmap *map, int reg) +{ + struct regmap_debugfs_off_cache *c; + int ret = -EINVAL; + + if (regmap_printable(map, reg + map->reg_stride)) { + ret = reg + map->reg_stride; + } else { + mutex_lock(&map->cache_lock); + list_for_each_entry(c, &map->debugfs_off_cache, list) { + if (reg > c->max_reg) + continue; + if (reg < c->base_reg) { + ret = c->base_reg; + break; + } + } + mutex_unlock(&map->cache_lock); + } + return ret; +} + +static ssize_t regmap_read_debugfs(struct regmap *map, unsigned int from, + unsigned int to, char __user *user_buf, + size_t count, loff_t *ppos) +{ + size_t buf_pos = 0; + loff_t p = *ppos; + ssize_t ret; + int i; + char *buf; + unsigned int val, start_reg; + + if (*ppos < 0 || !count) + return -EINVAL; + + if (count > (PAGE_SIZE << MAX_PAGE_ORDER)) + count = PAGE_SIZE << MAX_PAGE_ORDER; + + buf = kmalloc(count, GFP_KERNEL); + if (!buf) + return -ENOMEM; + + regmap_calc_tot_len(map, buf, count); + + /* Work out which register we're starting at */ + start_reg = regmap_debugfs_get_dump_start(map, from, *ppos, &p); + + for (i = start_reg; i >= 0 && i <= to; + i = regmap_next_readable_reg(map, i)) { + + /* If we're in the region the user is trying to read */ + if (p >= *ppos) { + /* ...but not beyond it */ + if (buf_pos + map->debugfs_tot_len > count) + break; + + /* Format the register */ + snprintf(buf + buf_pos, count - buf_pos, "%.*x: ", + map->debugfs_reg_len, i - from); + buf_pos += map->debugfs_reg_len + 2; + + /* Format the value, write all X if we can't read */ + ret = regmap_read(map, i, &val); + if (ret == 0) + snprintf(buf + buf_pos, count - buf_pos, + "%.*x", map->debugfs_val_len, val); + else + memset(buf + buf_pos, 'X', + map->debugfs_val_len); + buf_pos += 2 * map->format.val_bytes; + + buf[buf_pos++] = '\n'; + } + p += map->debugfs_tot_len; + } + + ret = buf_pos; + + if (copy_to_user(user_buf, buf, buf_pos)) { + ret = -EFAULT; + goto out; + } + + *ppos += buf_pos; + +out: + kfree(buf); + return ret; +} + +static ssize_t regmap_map_read_file(struct file *file, char __user *user_buf, + size_t count, loff_t *ppos) +{ + struct regmap *map = file->private_data; + + return regmap_read_debugfs(map, 0, map->max_register, user_buf, + count, ppos); +} + +#undef REGMAP_ALLOW_WRITE_DEBUGFS +#ifdef REGMAP_ALLOW_WRITE_DEBUGFS +/* + * This can be dangerous especially when we have clients such as + * PMICs, therefore don't provide any real compile time configuration option + * for this feature, people who want to use this will need to modify + * the source code directly. + */ +static ssize_t regmap_map_write_file(struct file *file, + const char __user *user_buf, + size_t count, loff_t *ppos) +{ + char buf[32]; + size_t buf_size; + char *start = buf; + unsigned long reg, value; + struct regmap *map = file->private_data; + int ret; + + buf_size = min(count, (sizeof(buf)-1)); + if (copy_from_user(buf, user_buf, buf_size)) + return -EFAULT; + buf[buf_size] = 0; + + while (*start == ' ') + start++; + reg = simple_strtoul(start, &start, 16); + while (*start == ' ') + start++; + if (kstrtoul(start, 16, &value)) + return -EINVAL; + + /* Userspace has been fiddling around behind the kernel's back */ + add_taint(TAINT_USER, LOCKDEP_STILL_OK); + + ret = regmap_write(map, reg, value); + if (ret < 0) + return ret; + return buf_size; +} +#else +#define regmap_map_write_file NULL +#endif + +static const struct file_operations regmap_map_fops = { + .open = simple_open, + .read = regmap_map_read_file, + .write = regmap_map_write_file, + .llseek = default_llseek, +}; + +static ssize_t regmap_range_read_file(struct file *file, char __user *user_buf, + size_t count, loff_t *ppos) +{ + struct regmap_range_node *range = file->private_data; + struct regmap *map = range->map; + + return regmap_read_debugfs(map, range->range_min, range->range_max, + user_buf, count, ppos); +} + +static const struct file_operations regmap_range_fops = { + .open = simple_open, + .read = regmap_range_read_file, + .llseek = default_llseek, +}; + +static ssize_t regmap_reg_ranges_read_file(struct file *file, + char __user *user_buf, size_t count, + loff_t *ppos) +{ + struct regmap *map = file->private_data; + struct regmap_debugfs_off_cache *c; + loff_t p = 0; + size_t buf_pos = 0; + char *buf; + char *entry; + int ret; + unsigned int entry_len; + + if (*ppos < 0 || !count) + return -EINVAL; + + if (count > (PAGE_SIZE << MAX_PAGE_ORDER)) + count = PAGE_SIZE << MAX_PAGE_ORDER; + + buf = kmalloc(count, GFP_KERNEL); + if (!buf) + return -ENOMEM; + + entry = kmalloc(PAGE_SIZE, GFP_KERNEL); + if (!entry) { + kfree(buf); + return -ENOMEM; + } + + /* While we are at it, build the register dump cache + * now so the read() operation on the `registers' file + * can benefit from using the cache. We do not care + * about the file position information that is contained + * in the cache, just about the actual register blocks */ + regmap_calc_tot_len(map, buf, count); + regmap_debugfs_get_dump_start(map, 0, *ppos, &p); + + /* Reset file pointer as the fixed-format of the `registers' + * file is not compatible with the `range' file */ + p = 0; + mutex_lock(&map->cache_lock); + list_for_each_entry(c, &map->debugfs_off_cache, list) { + entry_len = snprintf(entry, PAGE_SIZE, "%x-%x\n", + c->base_reg, c->max_reg); + if (p >= *ppos) { + if (buf_pos + entry_len > count) + break; + memcpy(buf + buf_pos, entry, entry_len); + buf_pos += entry_len; + } + p += entry_len; + } + mutex_unlock(&map->cache_lock); + + kfree(entry); + ret = buf_pos; + + if (copy_to_user(user_buf, buf, buf_pos)) { + ret = -EFAULT; + goto out_buf; + } + + *ppos += buf_pos; +out_buf: + kfree(buf); + return ret; +} + +static const struct file_operations regmap_reg_ranges_fops = { + .open = simple_open, + .read = regmap_reg_ranges_read_file, + .llseek = default_llseek, +}; + +static int regmap_access_show(struct seq_file *s, void *ignored) +{ + struct regmap *map = s->private; + int i, reg_len; + + reg_len = regmap_calc_reg_len(map->max_register); + + for (i = 0; i <= map->max_register; i += map->reg_stride) { + /* Ignore registers which are neither readable nor writable */ + if (!regmap_readable(map, i) && !regmap_writeable(map, i)) + continue; + + /* Format the register */ + seq_printf(s, "%.*x: %c %c %c %c\n", reg_len, i, + regmap_readable(map, i) ? 'y' : 'n', + regmap_writeable(map, i) ? 'y' : 'n', + regmap_volatile(map, i) ? 'y' : 'n', + regmap_precious(map, i) ? 'y' : 'n'); + } + + return 0; +} + +DEFINE_SHOW_ATTRIBUTE(regmap_access); + +static ssize_t regmap_cache_only_write_file(struct file *file, + const char __user *user_buf, + size_t count, loff_t *ppos) +{ + struct regmap *map = container_of(file->private_data, + struct regmap, cache_only); + bool new_val, require_sync = false; + int err; + + err = kstrtobool_from_user(user_buf, count, &new_val); + /* Ignore malforned data like debugfs_write_file_bool() */ + if (err) + return count; + + map->lock(map->lock_arg); + + if (new_val && !map->cache_only) { + dev_warn(map->dev, "debugfs cache_only=Y forced\n"); + add_taint(TAINT_USER, LOCKDEP_STILL_OK); + } else if (!new_val && map->cache_only) { + dev_warn(map->dev, "debugfs cache_only=N forced: syncing cache\n"); + require_sync = true; + } + map->cache_only = new_val; + + map->unlock(map->lock_arg); + + if (require_sync) { + err = regcache_sync(map); + if (err) + dev_err(map->dev, "Failed to sync cache %d\n", err); + } + + return count; +} + +static const struct file_operations regmap_cache_only_fops = { + .open = simple_open, + .read = debugfs_read_file_bool, + .write = regmap_cache_only_write_file, +}; + +static ssize_t regmap_cache_bypass_write_file(struct file *file, + const char __user *user_buf, + size_t count, loff_t *ppos) +{ + struct regmap *map = container_of(file->private_data, + struct regmap, cache_bypass); + bool new_val; + int err; + + err = kstrtobool_from_user(user_buf, count, &new_val); + /* Ignore malforned data like debugfs_write_file_bool() */ + if (err) + return count; + + map->lock(map->lock_arg); + + if (new_val && !map->cache_bypass) { + dev_warn(map->dev, "debugfs cache_bypass=Y forced\n"); + add_taint(TAINT_USER, LOCKDEP_STILL_OK); + } else if (!new_val && map->cache_bypass) { + dev_warn(map->dev, "debugfs cache_bypass=N forced\n"); + } + map->cache_bypass = new_val; + + map->unlock(map->lock_arg); + + return count; +} + +static const struct file_operations regmap_cache_bypass_fops = { + .open = simple_open, + .read = debugfs_read_file_bool, + .write = regmap_cache_bypass_write_file, +}; + +void regmap_debugfs_init(struct regmap *map) +{ + struct rb_node *next; + struct regmap_range_node *range_node; + const char *devname = "dummy"; + const char *name = map->name; + int id; + + /* + * Userspace can initiate reads from the hardware over debugfs. + * Normally internal regmap structures and buffers are protected with + * a mutex or a spinlock, but if the regmap owner decided to disable + * all locking mechanisms, this is no longer the case. For safety: + * don't create the debugfs entries if locking is disabled. + */ + if (map->debugfs_disable) { + dev_dbg(map->dev, "regmap locking disabled - not creating debugfs entries\n"); + return; + } + + /* If we don't have the debugfs root yet, postpone init */ + if (!regmap_debugfs_root) { + struct regmap_debugfs_node *node; + node = kzalloc_obj(*node); + if (!node) + return; + node->map = map; + mutex_lock(®map_debugfs_early_lock); + list_add(&node->link, ®map_debugfs_early_list); + mutex_unlock(®map_debugfs_early_lock); + return; + } + + INIT_LIST_HEAD(&map->debugfs_off_cache); + mutex_init(&map->cache_lock); + map->debugfs_dummy_id = -1; + + if (map->dev) + devname = dev_name(map->dev); + + if (name) { + if (!map->debugfs_name) { + map->debugfs_name = kasprintf(GFP_KERNEL, "%s-%s", + devname, name); + if (!map->debugfs_name) + return; + } + name = map->debugfs_name; + } else { + name = devname; + } + + if (!strcmp(name, "dummy")) { + kfree(map->debugfs_name); + id = ida_alloc(&dummy_ida, GFP_KERNEL); + if (id < 0) + return; + map->debugfs_name = kasprintf(GFP_KERNEL, "dummy%d", id); + if (!map->debugfs_name) { + ida_free(&dummy_ida, id); + return; + } + map->debugfs_dummy_id = id; + name = map->debugfs_name; + } + + map->debugfs = debugfs_create_dir(name, regmap_debugfs_root); + + debugfs_create_file("name", 0400, map->debugfs, + map, ®map_name_fops); + + debugfs_create_file("range", 0400, map->debugfs, + map, ®map_reg_ranges_fops); + + if (map->max_register || regmap_readable(map, 0)) { + umode_t registers_mode; + +#if defined(REGMAP_ALLOW_WRITE_DEBUGFS) + registers_mode = 0600; +#else + registers_mode = 0400; +#endif + + debugfs_create_file("registers", registers_mode, map->debugfs, + map, ®map_map_fops); + debugfs_create_file("access", 0400, map->debugfs, + map, ®map_access_fops); + } + + if (map->cache_type) { + debugfs_create_file("cache_only", 0600, map->debugfs, + &map->cache_only, ®map_cache_only_fops); + debugfs_create_bool("cache_dirty", 0400, map->debugfs, + &map->cache_dirty); + debugfs_create_file("cache_bypass", 0600, map->debugfs, + &map->cache_bypass, + ®map_cache_bypass_fops); + } + + /* + * This could interfere with driver operation. Therefore, don't provide + * any real compile time configuration option for this feature. One will + * have to modify the source code directly in order to use it. + */ +#undef REGMAP_ALLOW_FORCE_WRITE_FIELD_DEBUGFS +#ifdef REGMAP_ALLOW_FORCE_WRITE_FIELD_DEBUGFS + debugfs_create_bool("force_write_field", 0600, map->debugfs, + &map->force_write_field); +#endif + + next = rb_first(&map->range_tree); + while (next) { + range_node = rb_entry(next, struct regmap_range_node, node); + + if (range_node->name) + debugfs_create_file(range_node->name, 0400, + map->debugfs, range_node, + ®map_range_fops); + + next = rb_next(&range_node->node); + } + + if (map->cache_ops && map->cache_ops->debugfs_init) + map->cache_ops->debugfs_init(map); +} + +void regmap_debugfs_exit(struct regmap *map) +{ + if (map->debugfs) { + debugfs_remove_recursive(map->debugfs); + mutex_lock(&map->cache_lock); + regmap_debugfs_free_dump_cache(map); + mutex_unlock(&map->cache_lock); + if (map->debugfs_dummy_id >= 0) { + ida_free(&dummy_ida, map->debugfs_dummy_id); + map->debugfs_dummy_id = -1; + } + kfree(map->debugfs_name); + map->debugfs_name = NULL; + } else { + struct regmap_debugfs_node *node, *tmp; + + mutex_lock(®map_debugfs_early_lock); + list_for_each_entry_safe(node, tmp, ®map_debugfs_early_list, + link) { + if (node->map == map) { + list_del(&node->link); + kfree(node); + } + } + mutex_unlock(®map_debugfs_early_lock); + } +} + +void regmap_debugfs_initcall(void) +{ + struct regmap_debugfs_node *node, *tmp; + + regmap_debugfs_root = debugfs_create_dir("regmap", NULL); + + mutex_lock(®map_debugfs_early_lock); + list_for_each_entry_safe(node, tmp, ®map_debugfs_early_list, link) { + regmap_debugfs_init(node->map); + list_del(&node->link); + kfree(node); + } + mutex_unlock(®map_debugfs_early_lock); +} -- cgit v1.3.1