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 --- fs/mbcache.c | 444 +++++++++++++++++++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 444 insertions(+) create mode 100644 fs/mbcache.c (limited to 'fs/mbcache.c') diff --git a/fs/mbcache.c b/fs/mbcache.c new file mode 100644 index 000000000..2a6319b40 --- /dev/null +++ b/fs/mbcache.c @@ -0,0 +1,444 @@ +// SPDX-License-Identifier: GPL-2.0-only +#include +#include +#include +#include +#include +#include +#include +#include + +/* + * Mbcache is a simple key-value store. Keys need not be unique, however + * key-value pairs are expected to be unique (we use this fact in + * mb_cache_entry_delete_or_get()). + * + * Ext2 and ext4 use this cache for deduplication of extended attribute blocks. + * Ext4 also uses it for deduplication of xattr values stored in inodes. + * They use hash of data as a key and provide a value that may represent a + * block or inode number. That's why keys need not be unique (hash of different + * data may be the same). However user provided value always uniquely + * identifies a cache entry. + * + * We provide functions for creation and removal of entries, search by key, + * and a special "delete entry with given key-value pair" operation. Fixed + * size hash table is used for fast key lookups. + */ + +struct mb_cache { + /* Hash table of entries */ + struct hlist_bl_head *c_hash; + /* log2 of hash table size */ + int c_bucket_bits; + /* Maximum entries in cache to avoid degrading hash too much */ + unsigned long c_max_entries; + /* Protects c_list, c_entry_count */ + spinlock_t c_list_lock; + struct list_head c_list; + /* Number of entries in cache */ + unsigned long c_entry_count; + struct shrinker *c_shrink; + /* Work for shrinking when the cache has too many entries */ + struct work_struct c_shrink_work; +}; + +static struct kmem_cache *mb_entry_cache; + +static unsigned long mb_cache_shrink(struct mb_cache *cache, + unsigned long nr_to_scan); + +static inline struct hlist_bl_head *mb_cache_entry_head(struct mb_cache *cache, + u32 key) +{ + return &cache->c_hash[hash_32(key, cache->c_bucket_bits)]; +} + +/* + * Number of entries to reclaim synchronously when there are too many entries + * in cache + */ +#define SYNC_SHRINK_BATCH 64 + +/* + * mb_cache_entry_create - create entry in cache + * @cache - cache where the entry should be created + * @mask - gfp mask with which the entry should be allocated + * @key - key of the entry + * @value - value of the entry + * @reusable - is the entry reusable by others? + * + * Creates entry in @cache with key @key and value @value. The function returns + * -EBUSY if entry with the same key and value already exists in cache. + * Otherwise 0 is returned. + */ +int mb_cache_entry_create(struct mb_cache *cache, gfp_t mask, u32 key, + u64 value, bool reusable) +{ + struct mb_cache_entry *entry, *dup; + struct hlist_bl_node *dup_node; + struct hlist_bl_head *head; + + /* Schedule background reclaim if there are too many entries */ + if (cache->c_entry_count >= cache->c_max_entries) + schedule_work(&cache->c_shrink_work); + /* Do some sync reclaim if background reclaim cannot keep up */ + if (cache->c_entry_count >= 2*cache->c_max_entries) + mb_cache_shrink(cache, SYNC_SHRINK_BATCH); + + entry = kmem_cache_alloc(mb_entry_cache, mask); + if (!entry) + return -ENOMEM; + + INIT_LIST_HEAD(&entry->e_list); + /* + * We create entry with two references. One reference is kept by the + * hash table, the other reference is used to protect us from + * mb_cache_entry_delete_or_get() until the entry is fully setup. This + * avoids nesting of cache->c_list_lock into hash table bit locks which + * is problematic for RT. + */ + atomic_set(&entry->e_refcnt, 2); + entry->e_key = key; + entry->e_value = value; + entry->e_flags = 0; + if (reusable) + set_bit(MBE_REUSABLE_B, &entry->e_flags); + head = mb_cache_entry_head(cache, key); + hlist_bl_lock(head); + hlist_bl_for_each_entry(dup, dup_node, head, e_hash_list) { + if (dup->e_key == key && dup->e_value == value) { + hlist_bl_unlock(head); + kmem_cache_free(mb_entry_cache, entry); + return -EBUSY; + } + } + hlist_bl_add_head(&entry->e_hash_list, head); + hlist_bl_unlock(head); + spin_lock(&cache->c_list_lock); + list_add_tail(&entry->e_list, &cache->c_list); + cache->c_entry_count++; + spin_unlock(&cache->c_list_lock); + mb_cache_entry_put(cache, entry); + + return 0; +} +EXPORT_SYMBOL(mb_cache_entry_create); + +void __mb_cache_entry_free(struct mb_cache *cache, struct mb_cache_entry *entry) +{ + struct hlist_bl_head *head; + + head = mb_cache_entry_head(cache, entry->e_key); + hlist_bl_lock(head); + hlist_bl_del(&entry->e_hash_list); + hlist_bl_unlock(head); + kmem_cache_free(mb_entry_cache, entry); +} +EXPORT_SYMBOL(__mb_cache_entry_free); + +/* + * mb_cache_entry_wait_unused - wait to be the last user of the entry + * + * @entry - entry to work on + * + * Wait to be the last user of the entry. + */ +void mb_cache_entry_wait_unused(struct mb_cache_entry *entry) +{ + wait_var_event(&entry->e_refcnt, atomic_read(&entry->e_refcnt) <= 2); +} +EXPORT_SYMBOL(mb_cache_entry_wait_unused); + +static struct mb_cache_entry *__entry_find(struct mb_cache *cache, + struct mb_cache_entry *entry, + u32 key) +{ + struct mb_cache_entry *old_entry = entry; + struct hlist_bl_node *node; + struct hlist_bl_head *head; + + head = mb_cache_entry_head(cache, key); + hlist_bl_lock(head); + if (entry && !hlist_bl_unhashed(&entry->e_hash_list)) + node = entry->e_hash_list.next; + else + node = hlist_bl_first(head); + while (node) { + entry = hlist_bl_entry(node, struct mb_cache_entry, + e_hash_list); + if (entry->e_key == key && + test_bit(MBE_REUSABLE_B, &entry->e_flags) && + atomic_inc_not_zero(&entry->e_refcnt)) + goto out; + node = node->next; + } + entry = NULL; +out: + hlist_bl_unlock(head); + if (old_entry) + mb_cache_entry_put(cache, old_entry); + + return entry; +} + +/* + * mb_cache_entry_find_first - find the first reusable entry with the given key + * @cache: cache where we should search + * @key: key to look for + * + * Search in @cache for a reusable entry with key @key. Grabs reference to the + * first reusable entry found and returns the entry. + */ +struct mb_cache_entry *mb_cache_entry_find_first(struct mb_cache *cache, + u32 key) +{ + return __entry_find(cache, NULL, key); +} +EXPORT_SYMBOL(mb_cache_entry_find_first); + +/* + * mb_cache_entry_find_next - find next reusable entry with the same key + * @cache: cache where we should search + * @entry: entry to start search from + * + * Finds next reusable entry in the hash chain which has the same key as @entry. + * If @entry is unhashed (which can happen when deletion of entry races with the + * search), finds the first reusable entry in the hash chain. The function drops + * reference to @entry and returns with a reference to the found entry. + */ +struct mb_cache_entry *mb_cache_entry_find_next(struct mb_cache *cache, + struct mb_cache_entry *entry) +{ + return __entry_find(cache, entry, entry->e_key); +} +EXPORT_SYMBOL(mb_cache_entry_find_next); + +/* + * mb_cache_entry_get - get a cache entry by value (and key) + * @cache - cache we work with + * @key - key + * @value - value + */ +struct mb_cache_entry *mb_cache_entry_get(struct mb_cache *cache, u32 key, + u64 value) +{ + struct hlist_bl_node *node; + struct hlist_bl_head *head; + struct mb_cache_entry *entry; + + head = mb_cache_entry_head(cache, key); + hlist_bl_lock(head); + hlist_bl_for_each_entry(entry, node, head, e_hash_list) { + if (entry->e_key == key && entry->e_value == value && + atomic_inc_not_zero(&entry->e_refcnt)) + goto out; + } + entry = NULL; +out: + hlist_bl_unlock(head); + return entry; +} +EXPORT_SYMBOL(mb_cache_entry_get); + +/* mb_cache_entry_delete_or_get - remove a cache entry if it has no users + * @cache - cache we work with + * @key - key + * @value - value + * + * Remove entry from cache @cache with key @key and value @value. The removal + * happens only if the entry is unused. The function returns NULL in case the + * entry was successfully removed or there's no entry in cache. Otherwise the + * function grabs reference of the entry that we failed to delete because it + * still has users and return it. + */ +struct mb_cache_entry *mb_cache_entry_delete_or_get(struct mb_cache *cache, + u32 key, u64 value) +{ + struct mb_cache_entry *entry; + + entry = mb_cache_entry_get(cache, key, value); + if (!entry) + return NULL; + + /* + * Drop the ref we got from mb_cache_entry_get() and the initial hash + * ref if we are the last user + */ + if (atomic_cmpxchg(&entry->e_refcnt, 2, 0) != 2) + return entry; + + spin_lock(&cache->c_list_lock); + if (!list_empty(&entry->e_list)) + list_del_init(&entry->e_list); + cache->c_entry_count--; + spin_unlock(&cache->c_list_lock); + __mb_cache_entry_free(cache, entry); + return NULL; +} +EXPORT_SYMBOL(mb_cache_entry_delete_or_get); + +/* mb_cache_entry_touch - cache entry got used + * @cache - cache the entry belongs to + * @entry - entry that got used + * + * Marks entry as used to give hit higher chances of surviving in cache. + */ +void mb_cache_entry_touch(struct mb_cache *cache, + struct mb_cache_entry *entry) +{ + set_bit(MBE_REFERENCED_B, &entry->e_flags); +} +EXPORT_SYMBOL(mb_cache_entry_touch); + +static unsigned long mb_cache_count(struct shrinker *shrink, + struct shrink_control *sc) +{ + struct mb_cache *cache = shrink->private_data; + + return cache->c_entry_count; +} + +/* Shrink number of entries in cache */ +static unsigned long mb_cache_shrink(struct mb_cache *cache, + unsigned long nr_to_scan) +{ + struct mb_cache_entry *entry; + unsigned long shrunk = 0; + + spin_lock(&cache->c_list_lock); + while (nr_to_scan-- && !list_empty(&cache->c_list)) { + entry = list_first_entry(&cache->c_list, + struct mb_cache_entry, e_list); + /* Drop initial hash reference if there is no user */ + if (test_bit(MBE_REFERENCED_B, &entry->e_flags) || + atomic_cmpxchg(&entry->e_refcnt, 1, 0) != 1) { + clear_bit(MBE_REFERENCED_B, &entry->e_flags); + list_move_tail(&entry->e_list, &cache->c_list); + continue; + } + list_del_init(&entry->e_list); + cache->c_entry_count--; + spin_unlock(&cache->c_list_lock); + __mb_cache_entry_free(cache, entry); + shrunk++; + cond_resched(); + spin_lock(&cache->c_list_lock); + } + spin_unlock(&cache->c_list_lock); + + return shrunk; +} + +static unsigned long mb_cache_scan(struct shrinker *shrink, + struct shrink_control *sc) +{ + struct mb_cache *cache = shrink->private_data; + return mb_cache_shrink(cache, sc->nr_to_scan); +} + +/* We shrink 1/X of the cache when we have too many entries in it */ +#define SHRINK_DIVISOR 16 + +static void mb_cache_shrink_worker(struct work_struct *work) +{ + struct mb_cache *cache = container_of(work, struct mb_cache, + c_shrink_work); + mb_cache_shrink(cache, cache->c_max_entries / SHRINK_DIVISOR); +} + +/* + * mb_cache_create - create cache + * @bucket_bits: log2 of the hash table size + * + * Create cache for keys with 2^bucket_bits hash entries. + */ +struct mb_cache *mb_cache_create(int bucket_bits) +{ + struct mb_cache *cache; + unsigned long bucket_count = 1UL << bucket_bits; + unsigned long i; + + cache = kzalloc_obj(struct mb_cache); + if (!cache) + goto err_out; + cache->c_bucket_bits = bucket_bits; + cache->c_max_entries = bucket_count << 4; + INIT_LIST_HEAD(&cache->c_list); + spin_lock_init(&cache->c_list_lock); + cache->c_hash = kmalloc_objs(struct hlist_bl_head, bucket_count); + if (!cache->c_hash) { + kfree(cache); + goto err_out; + } + for (i = 0; i < bucket_count; i++) + INIT_HLIST_BL_HEAD(&cache->c_hash[i]); + + cache->c_shrink = shrinker_alloc(0, "mbcache-shrinker"); + if (!cache->c_shrink) { + kfree(cache->c_hash); + kfree(cache); + goto err_out; + } + + cache->c_shrink->count_objects = mb_cache_count; + cache->c_shrink->scan_objects = mb_cache_scan; + cache->c_shrink->private_data = cache; + + shrinker_register(cache->c_shrink); + + INIT_WORK(&cache->c_shrink_work, mb_cache_shrink_worker); + + return cache; + +err_out: + return NULL; +} +EXPORT_SYMBOL(mb_cache_create); + +/* + * mb_cache_destroy - destroy cache + * @cache: the cache to destroy + * + * Free all entries in cache and cache itself. Caller must make sure nobody + * (except shrinker) can reach @cache when calling this. + */ +void mb_cache_destroy(struct mb_cache *cache) +{ + struct mb_cache_entry *entry, *next; + + cancel_work_sync(&cache->c_shrink_work); + shrinker_free(cache->c_shrink); + + /* + * We don't bother with any locking. Cache must not be used at this + * point. + */ + list_for_each_entry_safe(entry, next, &cache->c_list, e_list) { + list_del(&entry->e_list); + WARN_ON(atomic_read(&entry->e_refcnt) != 1); + mb_cache_entry_put(cache, entry); + } + kfree(cache->c_hash); + kfree(cache); +} +EXPORT_SYMBOL(mb_cache_destroy); + +static int __init mbcache_init(void) +{ + mb_entry_cache = KMEM_CACHE(mb_cache_entry, SLAB_RECLAIM_ACCOUNT); + if (!mb_entry_cache) + return -ENOMEM; + return 0; +} + +static void __exit mbcache_exit(void) +{ + kmem_cache_destroy(mb_entry_cache); +} + +module_init(mbcache_init) +module_exit(mbcache_exit) + +MODULE_AUTHOR("Jan Kara "); +MODULE_DESCRIPTION("Meta block cache (for extended attributes)"); +MODULE_LICENSE("GPL"); -- cgit v1.3.1