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/exfat/cache.c | 357 +++++++++++++++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 357 insertions(+) create mode 100644 fs/exfat/cache.c (limited to 'fs/exfat/cache.c') diff --git a/fs/exfat/cache.c b/fs/exfat/cache.c new file mode 100644 index 000000000..7c8b4182f --- /dev/null +++ b/fs/exfat/cache.c @@ -0,0 +1,357 @@ +// SPDX-License-Identifier: GPL-2.0-or-later +/* + * linux/fs/fat/cache.c + * + * Written 1992,1993 by Werner Almesberger + * + * Mar 1999. AV. Changed cache, so that it uses the starting cluster instead + * of inode number. + * May 1999. AV. Fixed the bogosity with FAT32 (read "FAT28"). Fscking lusers. + * Copyright (C) 2012-2013 Samsung Electronics Co., Ltd. + */ + +#include +#include +#include + +#include "exfat_raw.h" +#include "exfat_fs.h" + +#define EXFAT_MAX_CACHE 16 + +struct exfat_cache { + struct list_head cache_list; + unsigned int nr_contig; /* number of contiguous clusters */ + unsigned int fcluster; /* cluster number in the file. */ + unsigned int dcluster; /* cluster number on disk. */ +}; + +struct exfat_cache_id { + unsigned int id; + unsigned int nr_contig; + unsigned int fcluster; + unsigned int dcluster; +}; + +static struct kmem_cache *exfat_cachep; + +static void exfat_cache_init_once(void *c) +{ + struct exfat_cache *cache = (struct exfat_cache *)c; + + INIT_LIST_HEAD(&cache->cache_list); +} + +int exfat_cache_init(void) +{ + exfat_cachep = kmem_cache_create("exfat_cache", + sizeof(struct exfat_cache), + 0, SLAB_RECLAIM_ACCOUNT, + exfat_cache_init_once); + if (!exfat_cachep) + return -ENOMEM; + return 0; +} + +void exfat_cache_shutdown(void) +{ + if (!exfat_cachep) + return; + kmem_cache_destroy(exfat_cachep); +} + +static inline struct exfat_cache *exfat_cache_alloc(void) +{ + return kmem_cache_alloc(exfat_cachep, GFP_NOFS); +} + +static inline void exfat_cache_free(struct exfat_cache *cache) +{ + WARN_ON(!list_empty(&cache->cache_list)); + kmem_cache_free(exfat_cachep, cache); +} + +static inline void exfat_cache_update_lru(struct inode *inode, + struct exfat_cache *cache) +{ + struct exfat_inode_info *ei = EXFAT_I(inode); + + if (ei->cache_lru.next != &cache->cache_list) + list_move(&cache->cache_list, &ei->cache_lru); +} + +/* + * Find the cache that covers or precedes 'fclus' and return the last + * cluster before the next cache range. + */ +static inline unsigned int +exfat_cache_lookup(struct inode *inode, struct exfat_cache_id *cid, + unsigned int fclus, unsigned int end, + unsigned int *cached_fclus, unsigned int *cached_dclus) +{ + struct exfat_inode_info *ei = EXFAT_I(inode); + static struct exfat_cache nohit = { .fcluster = 0, }; + struct exfat_cache *hit = &nohit, *p; + unsigned int tail = 0; /* End boundary of hit cache */ + + /* + * Search range [fclus, end]. Stop early if: + * 1. Cache covers entire range, or + * 2. Next cache starts at current cache tail + */ + spin_lock(&ei->cache_lru_lock); + list_for_each_entry(p, &ei->cache_lru, cache_list) { + /* Find the cache of "fclus" or nearest cache. */ + if (p->fcluster <= fclus) { + if (p->fcluster < hit->fcluster) + continue; + + hit = p; + tail = hit->fcluster + hit->nr_contig; + + /* Current cache covers [fclus, end] completely */ + if (tail >= end) + break; + } else if (p->fcluster <= end) { + end = p->fcluster - 1; + + /* + * If we have a hit and next cache starts within/at + * its tail, caches are contiguous, stop searching. + */ + if (tail && tail >= end) + break; + } + } + if (hit != &nohit) { + unsigned int offset; + + exfat_cache_update_lru(inode, hit); + cid->id = ei->cache_valid_id; + cid->nr_contig = hit->nr_contig; + cid->fcluster = hit->fcluster; + cid->dcluster = hit->dcluster; + + offset = min(cid->nr_contig, fclus - cid->fcluster); + *cached_fclus = cid->fcluster + offset; + *cached_dclus = cid->dcluster + offset; + } + spin_unlock(&ei->cache_lru_lock); + + /* Return next cache start or 'end' if no more caches */ + return end; +} + +static struct exfat_cache *exfat_cache_merge(struct inode *inode, + struct exfat_cache_id *new) +{ + struct exfat_inode_info *ei = EXFAT_I(inode); + struct exfat_cache *p; + + list_for_each_entry(p, &ei->cache_lru, cache_list) { + /* Find the same part as "new" in cluster-chain. */ + if (p->fcluster == new->fcluster) { + if (new->nr_contig > p->nr_contig) + p->nr_contig = new->nr_contig; + return p; + } + } + return NULL; +} + +static void exfat_cache_add(struct inode *inode, + struct exfat_cache_id *new) +{ + struct exfat_inode_info *ei = EXFAT_I(inode); + struct exfat_cache *cache, *tmp; + + if (new->fcluster == EXFAT_EOF_CLUSTER) /* dummy cache */ + return; + + spin_lock(&ei->cache_lru_lock); + if (new->id != EXFAT_CACHE_VALID && + new->id != ei->cache_valid_id) + goto unlock; /* this cache was invalidated */ + + cache = exfat_cache_merge(inode, new); + if (cache == NULL) { + if (ei->nr_caches < EXFAT_MAX_CACHE) { + ei->nr_caches++; + spin_unlock(&ei->cache_lru_lock); + + tmp = exfat_cache_alloc(); + if (!tmp) { + spin_lock(&ei->cache_lru_lock); + ei->nr_caches--; + spin_unlock(&ei->cache_lru_lock); + return; + } + + spin_lock(&ei->cache_lru_lock); + cache = exfat_cache_merge(inode, new); + if (cache != NULL) { + ei->nr_caches--; + exfat_cache_free(tmp); + goto out_update_lru; + } + cache = tmp; + } else { + struct list_head *p = ei->cache_lru.prev; + + cache = list_entry(p, + struct exfat_cache, cache_list); + } + cache->fcluster = new->fcluster; + cache->dcluster = new->dcluster; + cache->nr_contig = new->nr_contig; + } +out_update_lru: + exfat_cache_update_lru(inode, cache); +unlock: + spin_unlock(&ei->cache_lru_lock); +} + +/* + * Cache invalidation occurs rarely, thus the LRU chain is not updated. It + * fixes itself after a while. + */ +static void __exfat_cache_inval_inode(struct inode *inode) +{ + struct exfat_inode_info *ei = EXFAT_I(inode); + struct exfat_cache *cache; + + while (!list_empty(&ei->cache_lru)) { + cache = list_entry(ei->cache_lru.next, + struct exfat_cache, cache_list); + list_del_init(&cache->cache_list); + ei->nr_caches--; + exfat_cache_free(cache); + } + /* Update. The copy of caches before this id is discarded. */ + ei->cache_valid_id++; + if (ei->cache_valid_id == EXFAT_CACHE_VALID) + ei->cache_valid_id++; +} + +void exfat_cache_inval_inode(struct inode *inode) +{ + struct exfat_inode_info *ei = EXFAT_I(inode); + + spin_lock(&ei->cache_lru_lock); + __exfat_cache_inval_inode(inode); + spin_unlock(&ei->cache_lru_lock); +} + +static inline int cache_contiguous(struct exfat_cache_id *cid, + unsigned int dclus) +{ + cid->nr_contig++; + return cid->dcluster + cid->nr_contig == dclus; +} + +static inline void cache_init(struct exfat_cache_id *cid, + unsigned int fclus, unsigned int dclus) +{ + cid->id = EXFAT_CACHE_VALID; + cid->fcluster = fclus; + cid->dcluster = dclus; + cid->nr_contig = 0; +} + +int exfat_get_cluster(struct inode *inode, unsigned int cluster, + unsigned int *dclus, unsigned int *count, + unsigned int *last_dclus) +{ + struct super_block *sb = inode->i_sb; + struct exfat_inode_info *ei = EXFAT_I(inode); + struct buffer_head *bh = NULL; + struct exfat_cache_id cid; + unsigned int content, fclus; + unsigned int end = cluster + *count - 1; + + if (ei->start_clu == EXFAT_FREE_CLUSTER) { + exfat_fs_error(sb, + "invalid access to exfat cache (entry 0x%08x)", + ei->start_clu); + return -EIO; + } + + fclus = 0; + *dclus = ei->start_clu; + *last_dclus = *dclus; + + /* + * This case should not exist, as exfat_map_cluster function doesn't + * call this routine when start_clu == EXFAT_EOF_CLUSTER. + * This case is retained here for routine completeness. + */ + if (*dclus == EXFAT_EOF_CLUSTER) { + *count = 0; + return 0; + } + + /* If only the first cluster is needed, return now. */ + if (fclus == cluster && *count == 1) + return 0; + + cache_init(&cid, fclus, *dclus); + /* + * Update the 'end' to exclude the next cache range, as clusters in + * different cache are typically not contiguous. + */ + end = exfat_cache_lookup(inode, &cid, cluster, end, &fclus, dclus); + + /* Return if the cache covers the entire range. */ + if (cid.fcluster + cid.nr_contig >= end) { + *count = end - cluster + 1; + return 0; + } + + /* Find the first cluster we need. */ + while (fclus < cluster) { + if (exfat_ent_get(sb, *dclus, &content, &bh)) + return -EIO; + + *last_dclus = *dclus; + *dclus = content; + fclus++; + + if (content == EXFAT_EOF_CLUSTER) + break; + + if (!cache_contiguous(&cid, *dclus)) + cache_init(&cid, fclus, *dclus); + } + + /* + * Now the cid cache contains the first cluster requested, collect + * the remaining clusters of this contiguous extent. + */ + if (*dclus != EXFAT_EOF_CLUSTER) { + unsigned int clu = *dclus; + + while (fclus < end) { + if (exfat_ent_get(sb, clu, &content, &bh)) + return -EIO; + if (++clu != content) + break; + fclus++; + } + cid.nr_contig = fclus - cid.fcluster; + *count = fclus - cluster + 1; + + /* + * Cache this discontiguous cluster, we'll definitely need + * it later + */ + if (fclus < end && content != EXFAT_EOF_CLUSTER) { + exfat_cache_add(inode, &cid); + cache_init(&cid, fclus + 1, content); + } + } else { + *count = 0; + } + brelse(bh); + exfat_cache_add(inode, &cid); + return 0; +} -- cgit v1.3.1