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/nilfs2/direct.c | 387 +++++++++++++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 387 insertions(+) create mode 100644 fs/nilfs2/direct.c (limited to 'fs/nilfs2/direct.c') diff --git a/fs/nilfs2/direct.c b/fs/nilfs2/direct.c new file mode 100644 index 000000000..b8643d3aa --- /dev/null +++ b/fs/nilfs2/direct.c @@ -0,0 +1,387 @@ +// SPDX-License-Identifier: GPL-2.0+ +/* + * NILFS direct block pointer. + * + * Copyright (C) 2006-2008 Nippon Telegraph and Telephone Corporation. + * + * Written by Koji Sato. + */ + +#include +#include "nilfs.h" +#include "page.h" +#include "direct.h" +#include "alloc.h" +#include "dat.h" + +static inline __le64 *nilfs_direct_dptrs(const struct nilfs_bmap *direct) +{ + return (__le64 *) + ((struct nilfs_direct_node *)direct->b_u.u_data + 1); +} + +static inline __u64 +nilfs_direct_get_ptr(const struct nilfs_bmap *direct, __u64 key) +{ + return le64_to_cpu(*(nilfs_direct_dptrs(direct) + key)); +} + +static inline void nilfs_direct_set_ptr(struct nilfs_bmap *direct, + __u64 key, __u64 ptr) +{ + *(nilfs_direct_dptrs(direct) + key) = cpu_to_le64(ptr); +} + +static int nilfs_direct_lookup(const struct nilfs_bmap *direct, + __u64 key, int level, __u64 *ptrp) +{ + __u64 ptr; + + if (key > NILFS_DIRECT_KEY_MAX || level != 1) + return -ENOENT; + ptr = nilfs_direct_get_ptr(direct, key); + if (ptr == NILFS_BMAP_INVALID_PTR) + return -ENOENT; + + *ptrp = ptr; + return 0; +} + +static int nilfs_direct_lookup_contig(const struct nilfs_bmap *direct, + __u64 key, __u64 *ptrp, + unsigned int maxblocks) +{ + struct inode *dat = NULL; + __u64 ptr, ptr2; + sector_t blocknr; + int ret, cnt; + + if (key > NILFS_DIRECT_KEY_MAX) + return -ENOENT; + ptr = nilfs_direct_get_ptr(direct, key); + if (ptr == NILFS_BMAP_INVALID_PTR) + return -ENOENT; + + if (NILFS_BMAP_USE_VBN(direct)) { + dat = nilfs_bmap_get_dat(direct); + ret = nilfs_dat_translate(dat, ptr, &blocknr); + if (ret < 0) + goto dat_error; + ptr = blocknr; + } + + maxblocks = min_t(unsigned int, maxblocks, + NILFS_DIRECT_KEY_MAX - key + 1); + for (cnt = 1; cnt < maxblocks && + (ptr2 = nilfs_direct_get_ptr(direct, key + cnt)) != + NILFS_BMAP_INVALID_PTR; + cnt++) { + if (dat) { + ret = nilfs_dat_translate(dat, ptr2, &blocknr); + if (ret < 0) + goto dat_error; + ptr2 = blocknr; + } + if (ptr2 != ptr + cnt) + break; + } + *ptrp = ptr; + return cnt; + + dat_error: + if (ret == -ENOENT) + ret = -EINVAL; /* Notify bmap layer of metadata corruption */ + return ret; +} + +static __u64 +nilfs_direct_find_target_v(const struct nilfs_bmap *direct, __u64 key) +{ + __u64 ptr; + + ptr = nilfs_bmap_find_target_seq(direct, key); + if (ptr != NILFS_BMAP_INVALID_PTR) + /* sequential access */ + return ptr; + + /* block group */ + return nilfs_bmap_find_target_in_group(direct); +} + +static int nilfs_direct_insert(struct nilfs_bmap *bmap, __u64 key, __u64 ptr) +{ + union nilfs_bmap_ptr_req req; + struct inode *dat = NULL; + struct buffer_head *bh; + int ret; + + if (key > NILFS_DIRECT_KEY_MAX) + return -ENOENT; + if (nilfs_direct_get_ptr(bmap, key) != NILFS_BMAP_INVALID_PTR) + return -EEXIST; + + if (NILFS_BMAP_USE_VBN(bmap)) { + req.bpr_ptr = nilfs_direct_find_target_v(bmap, key); + dat = nilfs_bmap_get_dat(bmap); + } + ret = nilfs_bmap_prepare_alloc_ptr(bmap, &req, dat); + if (!ret) { + /* ptr must be a pointer to a buffer head. */ + bh = (struct buffer_head *)((unsigned long)ptr); + set_buffer_nilfs_volatile(bh); + + nilfs_bmap_commit_alloc_ptr(bmap, &req, dat); + nilfs_direct_set_ptr(bmap, key, req.bpr_ptr); + + if (!nilfs_bmap_dirty(bmap)) + nilfs_bmap_set_dirty(bmap); + + if (NILFS_BMAP_USE_VBN(bmap)) + nilfs_bmap_set_target_v(bmap, key, req.bpr_ptr); + + nilfs_inode_add_blocks(bmap->b_inode, 1); + } + return ret; +} + +static int nilfs_direct_delete(struct nilfs_bmap *bmap, __u64 key, bool deform) +{ + union nilfs_bmap_ptr_req req; + struct inode *dat; + int ret; + + if (key > NILFS_DIRECT_KEY_MAX || + nilfs_direct_get_ptr(bmap, key) == NILFS_BMAP_INVALID_PTR) + return -ENOENT; + + dat = NILFS_BMAP_USE_VBN(bmap) ? nilfs_bmap_get_dat(bmap) : NULL; + req.bpr_ptr = nilfs_direct_get_ptr(bmap, key); + + ret = nilfs_bmap_prepare_end_ptr(bmap, &req, dat); + if (!ret) { + nilfs_bmap_commit_end_ptr(bmap, &req, dat); + nilfs_direct_set_ptr(bmap, key, NILFS_BMAP_INVALID_PTR); + nilfs_inode_sub_blocks(bmap->b_inode, 1); + } + return ret; +} + +static int nilfs_direct_seek_key(const struct nilfs_bmap *direct, __u64 start, + __u64 *keyp) +{ + __u64 key; + + for (key = start; key <= NILFS_DIRECT_KEY_MAX; key++) { + if (nilfs_direct_get_ptr(direct, key) != + NILFS_BMAP_INVALID_PTR) { + *keyp = key; + return 0; + } + } + return -ENOENT; +} + +static int nilfs_direct_last_key(const struct nilfs_bmap *direct, __u64 *keyp) +{ + __u64 key, lastkey; + + lastkey = NILFS_DIRECT_KEY_MAX + 1; + for (key = NILFS_DIRECT_KEY_MIN; key <= NILFS_DIRECT_KEY_MAX; key++) + if (nilfs_direct_get_ptr(direct, key) != + NILFS_BMAP_INVALID_PTR) + lastkey = key; + + if (lastkey == NILFS_DIRECT_KEY_MAX + 1) + return -ENOENT; + + *keyp = lastkey; + + return 0; +} + +static int nilfs_direct_check_insert(const struct nilfs_bmap *bmap, __u64 key) +{ + return key > NILFS_DIRECT_KEY_MAX; +} + +static int nilfs_direct_gather_data(struct nilfs_bmap *direct, + __u64 *keys, __u64 *ptrs, int nitems) +{ + __u64 key; + __u64 ptr; + int n; + + if (nitems > NILFS_DIRECT_NBLOCKS) + nitems = NILFS_DIRECT_NBLOCKS; + n = 0; + for (key = 0; key < nitems; key++) { + ptr = nilfs_direct_get_ptr(direct, key); + if (ptr != NILFS_BMAP_INVALID_PTR) { + keys[n] = key; + ptrs[n] = ptr; + n++; + } + } + return n; +} + +int nilfs_direct_delete_and_convert(struct nilfs_bmap *bmap, + __u64 key, __u64 *keys, __u64 *ptrs, int n) +{ + __le64 *dptrs; + int ret, i, j; + + /* no need to allocate any resource for conversion */ + + /* delete */ + ret = bmap->b_ops->bop_delete(bmap, key, true); + if (ret < 0) + return ret; + + /* free resources */ + if (bmap->b_ops->bop_clear != NULL) + bmap->b_ops->bop_clear(bmap); + + /* convert */ + dptrs = nilfs_direct_dptrs(bmap); + for (i = 0, j = 0; i < NILFS_DIRECT_NBLOCKS; i++) { + if ((j < n) && (i == keys[j])) { + dptrs[i] = (i != key) ? + cpu_to_le64(ptrs[j]) : + NILFS_BMAP_INVALID_PTR; + j++; + } else + dptrs[i] = NILFS_BMAP_INVALID_PTR; + } + + nilfs_direct_init(bmap); + return 0; +} + +static int nilfs_direct_propagate(struct nilfs_bmap *bmap, + struct buffer_head *bh) +{ + struct nilfs_palloc_req oldreq, newreq; + struct inode *dat; + __u64 key; + __u64 ptr; + int ret; + + if (!NILFS_BMAP_USE_VBN(bmap)) + return 0; + + dat = nilfs_bmap_get_dat(bmap); + key = nilfs_bmap_data_get_key(bmap, bh); + ptr = nilfs_direct_get_ptr(bmap, key); + if (ptr == NILFS_BMAP_INVALID_PTR) + return -EINVAL; + + if (!buffer_nilfs_volatile(bh)) { + oldreq.pr_entry_nr = ptr; + newreq.pr_entry_nr = ptr; + ret = nilfs_dat_prepare_update(dat, &oldreq, &newreq); + if (ret < 0) + return ret; + nilfs_dat_commit_update(dat, &oldreq, &newreq, + bmap->b_ptr_type == NILFS_BMAP_PTR_VS); + set_buffer_nilfs_volatile(bh); + nilfs_direct_set_ptr(bmap, key, newreq.pr_entry_nr); + } else + ret = nilfs_dat_mark_dirty(dat, ptr); + + return ret; +} + +static int nilfs_direct_assign_v(struct nilfs_bmap *direct, + __u64 key, __u64 ptr, + struct buffer_head **bh, + sector_t blocknr, + union nilfs_binfo *binfo) +{ + struct inode *dat = nilfs_bmap_get_dat(direct); + union nilfs_bmap_ptr_req req; + int ret; + + req.bpr_ptr = ptr; + ret = nilfs_dat_prepare_start(dat, &req.bpr_req); + if (!ret) { + nilfs_dat_commit_start(dat, &req.bpr_req, blocknr); + binfo->bi_v.bi_vblocknr = cpu_to_le64(ptr); + binfo->bi_v.bi_blkoff = cpu_to_le64(key); + } + return ret; +} + +static int nilfs_direct_assign_p(struct nilfs_bmap *direct, + __u64 key, __u64 ptr, + struct buffer_head **bh, + sector_t blocknr, + union nilfs_binfo *binfo) +{ + nilfs_direct_set_ptr(direct, key, blocknr); + + binfo->bi_dat.bi_blkoff = cpu_to_le64(key); + binfo->bi_dat.bi_level = 0; + memset(binfo->bi_dat.bi_pad, 0, sizeof(binfo->bi_dat.bi_pad)); + + return 0; +} + +static int nilfs_direct_assign(struct nilfs_bmap *bmap, + struct buffer_head **bh, + sector_t blocknr, + union nilfs_binfo *binfo) +{ + __u64 key; + __u64 ptr; + + key = nilfs_bmap_data_get_key(bmap, *bh); + if (unlikely(key > NILFS_DIRECT_KEY_MAX)) { + nilfs_crit(bmap->b_inode->i_sb, + "%s (ino=%llu): invalid key: %llu", + __func__, + bmap->b_inode->i_ino, (unsigned long long)key); + return -EINVAL; + } + ptr = nilfs_direct_get_ptr(bmap, key); + if (unlikely(ptr == NILFS_BMAP_INVALID_PTR)) { + nilfs_crit(bmap->b_inode->i_sb, + "%s (ino=%llu): invalid pointer: %llu", + __func__, + bmap->b_inode->i_ino, (unsigned long long)ptr); + return -EINVAL; + } + + return NILFS_BMAP_USE_VBN(bmap) ? + nilfs_direct_assign_v(bmap, key, ptr, bh, blocknr, binfo) : + nilfs_direct_assign_p(bmap, key, ptr, bh, blocknr, binfo); +} + +static const struct nilfs_bmap_operations nilfs_direct_ops = { + .bop_lookup = nilfs_direct_lookup, + .bop_lookup_contig = nilfs_direct_lookup_contig, + .bop_insert = nilfs_direct_insert, + .bop_delete = nilfs_direct_delete, + .bop_clear = NULL, + + .bop_propagate = nilfs_direct_propagate, + + .bop_lookup_dirty_buffers = NULL, + + .bop_assign = nilfs_direct_assign, + .bop_mark = NULL, + + .bop_seek_key = nilfs_direct_seek_key, + .bop_last_key = nilfs_direct_last_key, + + .bop_check_insert = nilfs_direct_check_insert, + .bop_check_delete = NULL, + .bop_gather_data = nilfs_direct_gather_data, +}; + + +int nilfs_direct_init(struct nilfs_bmap *bmap) +{ + bmap->b_ops = &nilfs_direct_ops; + return 0; +} -- cgit v1.3.1