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/adfs/dir_fplus.c | 287 ++++++++++++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 287 insertions(+) create mode 100644 fs/adfs/dir_fplus.c (limited to 'fs/adfs/dir_fplus.c') diff --git a/fs/adfs/dir_fplus.c b/fs/adfs/dir_fplus.c new file mode 100644 index 000000000..4a1592401 --- /dev/null +++ b/fs/adfs/dir_fplus.c @@ -0,0 +1,287 @@ +// SPDX-License-Identifier: GPL-2.0-only +/* + * linux/fs/adfs/dir_fplus.c + * + * Copyright (C) 1997-1999 Russell King + */ +#include "adfs.h" +#include "dir_fplus.h" + +/* Return the byte offset to directory entry pos */ +static unsigned int adfs_fplus_offset(const struct adfs_bigdirheader *h, + unsigned int pos) +{ + return offsetof(struct adfs_bigdirheader, bigdirname) + + ALIGN(le32_to_cpu(h->bigdirnamelen), 4) + + pos * sizeof(struct adfs_bigdirentry); +} + +static int adfs_fplus_validate_header(const struct adfs_bigdirheader *h) +{ + unsigned int size = le32_to_cpu(h->bigdirsize); + unsigned int len; + + if (h->bigdirversion[0] != 0 || h->bigdirversion[1] != 0 || + h->bigdirversion[2] != 0 || + h->bigdirstartname != cpu_to_le32(BIGDIRSTARTNAME) || + !size || size & 2047 || size > SZ_4M) + return -EIO; + + size -= sizeof(struct adfs_bigdirtail) + + offsetof(struct adfs_bigdirheader, bigdirname); + + /* Check that bigdirnamelen fits within the directory */ + len = ALIGN(le32_to_cpu(h->bigdirnamelen), 4); + if (len > size) + return -EIO; + + size -= len; + + /* Check that bigdirnamesize fits within the directory */ + len = le32_to_cpu(h->bigdirnamesize); + if (len > size) + return -EIO; + + size -= len; + + /* + * Avoid division, we know that absolute maximum number of entries + * can not be so large to cause overflow of the multiplication below. + */ + len = le32_to_cpu(h->bigdirentries); + if (len > SZ_4M / sizeof(struct adfs_bigdirentry) || + len * sizeof(struct adfs_bigdirentry) > size) + return -EIO; + + return 0; +} + +static int adfs_fplus_validate_tail(const struct adfs_bigdirheader *h, + const struct adfs_bigdirtail *t) +{ + if (t->bigdirendname != cpu_to_le32(BIGDIRENDNAME) || + t->bigdirendmasseq != h->startmasseq || + t->reserved[0] != 0 || t->reserved[1] != 0) + return -EIO; + + return 0; +} + +static u8 adfs_fplus_checkbyte(struct adfs_dir *dir) +{ + struct adfs_bigdirheader *h = dir->bighead; + struct adfs_bigdirtail *t = dir->bigtail; + unsigned int end, bs, bi, i; + __le32 *bp; + u32 dircheck; + + end = adfs_fplus_offset(h, le32_to_cpu(h->bigdirentries)) + + le32_to_cpu(h->bigdirnamesize); + + /* Accumulate the contents of the header, entries and names */ + for (dircheck = 0, bi = 0; end; bi++) { + bp = (void *)dir->bhs[bi]->b_data; + bs = dir->bhs[bi]->b_size; + if (bs > end) + bs = end; + + for (i = 0; i < bs; i += sizeof(u32)) + dircheck = ror32(dircheck, 13) ^ le32_to_cpup(bp++); + + end -= bs; + } + + /* Accumulate the contents of the tail except for the check byte */ + dircheck = ror32(dircheck, 13) ^ le32_to_cpu(t->bigdirendname); + dircheck = ror32(dircheck, 13) ^ t->bigdirendmasseq; + dircheck = ror32(dircheck, 13) ^ t->reserved[0]; + dircheck = ror32(dircheck, 13) ^ t->reserved[1]; + + return dircheck ^ dircheck >> 8 ^ dircheck >> 16 ^ dircheck >> 24; +} + +static int adfs_fplus_read(struct super_block *sb, u32 indaddr, + unsigned int size, struct adfs_dir *dir) +{ + struct adfs_bigdirheader *h; + struct adfs_bigdirtail *t; + unsigned int dirsize; + int ret; + + /* Read first buffer */ + ret = adfs_dir_read_buffers(sb, indaddr, sb->s_blocksize, dir); + if (ret) + return ret; + + dir->bighead = h = (void *)dir->bhs[0]->b_data; + ret = adfs_fplus_validate_header(h); + if (ret) { + adfs_error(sb, "dir %06x has malformed header", indaddr); + goto out; + } + + dirsize = le32_to_cpu(h->bigdirsize); + if (size && dirsize != size) { + adfs_msg(sb, KERN_WARNING, + "dir %06x header size %X does not match directory size %X", + indaddr, dirsize, size); + } + + /* Read remaining buffers */ + ret = adfs_dir_read_buffers(sb, indaddr, dirsize, dir); + if (ret) + return ret; + + dir->bigtail = t = (struct adfs_bigdirtail *) + (dir->bhs[dir->nr_buffers - 1]->b_data + (sb->s_blocksize - 8)); + + ret = adfs_fplus_validate_tail(h, t); + if (ret) { + adfs_error(sb, "dir %06x has malformed tail", indaddr); + goto out; + } + + if (adfs_fplus_checkbyte(dir) != t->bigdircheckbyte) { + adfs_error(sb, "dir %06x checkbyte mismatch\n", indaddr); + goto out; + } + + dir->parent_id = le32_to_cpu(h->bigdirparent); + return 0; + +out: + adfs_dir_relse(dir); + + return ret; +} + +static int +adfs_fplus_setpos(struct adfs_dir *dir, unsigned int fpos) +{ + int ret = -ENOENT; + + if (fpos <= le32_to_cpu(dir->bighead->bigdirentries)) { + dir->pos = fpos; + ret = 0; + } + + return ret; +} + +static int +adfs_fplus_getnext(struct adfs_dir *dir, struct object_info *obj) +{ + struct adfs_bigdirheader *h = dir->bighead; + struct adfs_bigdirentry bde; + unsigned int offset; + int ret; + + if (dir->pos >= le32_to_cpu(h->bigdirentries)) + return -ENOENT; + + offset = adfs_fplus_offset(h, dir->pos); + + ret = adfs_dir_copyfrom(&bde, dir, offset, + sizeof(struct adfs_bigdirentry)); + if (ret) + return ret; + + obj->loadaddr = le32_to_cpu(bde.bigdirload); + obj->execaddr = le32_to_cpu(bde.bigdirexec); + obj->size = le32_to_cpu(bde.bigdirlen); + obj->indaddr = le32_to_cpu(bde.bigdirindaddr); + obj->attr = le32_to_cpu(bde.bigdirattr); + obj->name_len = le32_to_cpu(bde.bigdirobnamelen); + + offset = adfs_fplus_offset(h, le32_to_cpu(h->bigdirentries)); + offset += le32_to_cpu(bde.bigdirobnameptr); + + ret = adfs_dir_copyfrom(obj->name, dir, offset, obj->name_len); + if (ret) + return ret; + + adfs_object_fixup(dir, obj); + + dir->pos += 1; + + return 0; +} + +static int adfs_fplus_iterate(struct adfs_dir *dir, struct dir_context *ctx) +{ + struct object_info obj; + + if ((ctx->pos - 2) >> 32) + return 0; + + if (adfs_fplus_setpos(dir, ctx->pos - 2)) + return 0; + + while (!adfs_fplus_getnext(dir, &obj)) { + if (!dir_emit(ctx, obj.name, obj.name_len, + obj.indaddr, DT_UNKNOWN)) + break; + ctx->pos++; + } + + return 0; +} + +static int adfs_fplus_update(struct adfs_dir *dir, struct object_info *obj) +{ + struct adfs_bigdirheader *h = dir->bighead; + struct adfs_bigdirentry bde; + int offset, end, ret; + + offset = adfs_fplus_offset(h, 0) - sizeof(bde); + end = adfs_fplus_offset(h, le32_to_cpu(h->bigdirentries)); + + do { + offset += sizeof(bde); + if (offset >= end) { + adfs_error(dir->sb, "unable to locate entry to update"); + return -ENOENT; + } + ret = adfs_dir_copyfrom(&bde, dir, offset, sizeof(bde)); + if (ret) { + adfs_error(dir->sb, "error reading directory entry"); + return -ENOENT; + } + } while (le32_to_cpu(bde.bigdirindaddr) != obj->indaddr); + + bde.bigdirload = cpu_to_le32(obj->loadaddr); + bde.bigdirexec = cpu_to_le32(obj->execaddr); + bde.bigdirlen = cpu_to_le32(obj->size); + bde.bigdirindaddr = cpu_to_le32(obj->indaddr); + bde.bigdirattr = cpu_to_le32(obj->attr); + + return adfs_dir_copyto(dir, offset, &bde, sizeof(bde)); +} + +static int adfs_fplus_commit(struct adfs_dir *dir) +{ + int ret; + + /* Increment directory sequence number */ + dir->bighead->startmasseq += 1; + dir->bigtail->bigdirendmasseq += 1; + + /* Update directory check byte */ + dir->bigtail->bigdircheckbyte = adfs_fplus_checkbyte(dir); + + /* Make sure the directory still validates correctly */ + ret = adfs_fplus_validate_header(dir->bighead); + if (ret == 0) + ret = adfs_fplus_validate_tail(dir->bighead, dir->bigtail); + + return ret; +} + +const struct adfs_dir_ops adfs_fplus_dir_ops = { + .read = adfs_fplus_read, + .iterate = adfs_fplus_iterate, + .setpos = adfs_fplus_setpos, + .getnext = adfs_fplus_getnext, + .update = adfs_fplus_update, + .commit = adfs_fplus_commit, +}; -- cgit v1.3.1