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/hfsplus/wrapper.c | 240 +++++++++++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 240 insertions(+) create mode 100644 fs/hfsplus/wrapper.c (limited to 'fs/hfsplus/wrapper.c') diff --git a/fs/hfsplus/wrapper.c b/fs/hfsplus/wrapper.c new file mode 100644 index 000000000..30cf4fe78 --- /dev/null +++ b/fs/hfsplus/wrapper.c @@ -0,0 +1,240 @@ +// SPDX-License-Identifier: GPL-2.0 +/* + * linux/fs/hfsplus/wrapper.c + * + * Copyright (C) 2001 + * Brad Boyer (flar@allandria.com) + * (C) 2003 Ardis Technologies + * + * Handling of HFS wrappers around HFS+ volumes + */ + +#include +#include +#include +#include + +#include "hfsplus_fs.h" +#include "hfsplus_raw.h" + +struct hfsplus_wd { + u32 ablk_size; + u16 ablk_start; + u16 embed_start; + u16 embed_count; +}; + +/** + * hfsplus_submit_bio - Perform block I/O + * @sb: super block of volume for I/O + * @sector: block to read or write, for blocks of HFSPLUS_SECTOR_SIZE bytes + * @buf: buffer for I/O + * @data: output pointer for location of requested data + * @opf: I/O operation type and flags + * + * The unit of I/O is hfsplus_min_io_size(sb), which may be bigger than + * HFSPLUS_SECTOR_SIZE, and @buf must be sized accordingly. On reads + * @data will return a pointer to the start of the requested sector, + * which may not be the same location as @buf. + * + * If @sector is not aligned to the bdev logical block size it will + * be rounded down. For writes this means that @buf should contain data + * that starts at the rounded-down address. As long as the data was + * read using hfsplus_submit_bio() and the same buffer is used things + * will work correctly. + * + * Returns: %0 on success else -errno code + */ +int hfsplus_submit_bio(struct super_block *sb, sector_t sector, + void *buf, void **data, blk_opf_t opf) +{ + u64 io_size = hfsplus_min_io_size(sb); + loff_t start = (loff_t)sector << HFSPLUS_SECTOR_SHIFT; + int offset = start & (io_size - 1); + + if ((opf & REQ_OP_MASK) != REQ_OP_WRITE && data) + *data = (u8 *)buf + offset; + + /* + * Align sector to hardware sector size and find offset. We assume that + * io_size is a power of two, which _should_ be true. + */ + sector &= ~((io_size >> HFSPLUS_SECTOR_SHIFT) - 1); + return bdev_rw_virt(sb->s_bdev, sector, buf, io_size, opf); +} + +static int hfsplus_read_mdb(void *bufptr, struct hfsplus_wd *wd) +{ + u32 extent; + u16 attrib; + __be16 sig; + + sig = *(__be16 *)(bufptr + HFSP_WRAPOFF_EMBEDSIG); + if (sig != cpu_to_be16(HFSPLUS_VOLHEAD_SIG) && + sig != cpu_to_be16(HFSPLUS_VOLHEAD_SIGX)) + return 0; + + attrib = be16_to_cpu(*(__be16 *)(bufptr + HFSP_WRAPOFF_ATTRIB)); + if (!(attrib & HFSP_WRAP_ATTRIB_SLOCK) || + !(attrib & HFSP_WRAP_ATTRIB_SPARED)) + return 0; + + wd->ablk_size = + be32_to_cpu(*(__be32 *)(bufptr + HFSP_WRAPOFF_ABLKSIZE)); + if (wd->ablk_size < HFSPLUS_SECTOR_SIZE) + return 0; + if (wd->ablk_size % HFSPLUS_SECTOR_SIZE) + return 0; + wd->ablk_start = + be16_to_cpu(*(__be16 *)(bufptr + HFSP_WRAPOFF_ABLKSTART)); + + extent = get_unaligned_be32(bufptr + HFSP_WRAPOFF_EMBEDEXT); + wd->embed_start = (extent >> 16) & 0xFFFF; + wd->embed_count = extent & 0xFFFF; + + return 1; +} + +static int hfsplus_get_last_session(struct super_block *sb, + sector_t *start, sector_t *size) +{ + struct cdrom_device_info *cdi = disk_to_cdi(sb->s_bdev->bd_disk); + + /* default values */ + *start = 0; + *size = bdev_nr_sectors(sb->s_bdev); + + if (HFSPLUS_SB(sb)->session >= 0) { + struct cdrom_tocentry te; + + if (!cdi) + return -EINVAL; + + te.cdte_track = HFSPLUS_SB(sb)->session; + te.cdte_format = CDROM_LBA; + if (cdrom_read_tocentry(cdi, &te) || + (te.cdte_ctrl & CDROM_DATA_TRACK) != 4) { + pr_err("invalid session number or type of track\n"); + return -EINVAL; + } + *start = (sector_t)te.cdte_addr.lba << 2; + } else if (cdi) { + struct cdrom_multisession ms_info; + + ms_info.addr_format = CDROM_LBA; + if (cdrom_multisession(cdi, &ms_info) == 0 && ms_info.xa_flag) + *start = (sector_t)ms_info.addr.lba << 2; + } + + return 0; +} + +/* Find the volume header and fill in some minimum bits in superblock */ +/* Takes in super block, returns true if good data read */ +int hfsplus_read_wrapper(struct super_block *sb) +{ + struct hfsplus_sb_info *sbi = HFSPLUS_SB(sb); + struct hfsplus_wd wd; + sector_t part_start, part_size; + u32 blocksize; + int error = 0; + + error = -EINVAL; + blocksize = sb_min_blocksize(sb, HFSPLUS_SECTOR_SIZE); + if (!blocksize) + goto out; + + sbi->min_io_size = blocksize; + + if (hfsplus_get_last_session(sb, &part_start, &part_size)) + goto out; + + error = -ENOMEM; + sbi->s_vhdr_buf = kmalloc(hfsplus_min_io_size(sb), GFP_KERNEL); + if (!sbi->s_vhdr_buf) + goto out; + sbi->s_backup_vhdr_buf = kmalloc(hfsplus_min_io_size(sb), GFP_KERNEL); + if (!sbi->s_backup_vhdr_buf) + goto out_free_vhdr; + +reread: + error = hfsplus_submit_bio(sb, part_start + HFSPLUS_VOLHEAD_SECTOR, + sbi->s_vhdr_buf, (void **)&sbi->s_vhdr, + REQ_OP_READ); + if (error) + goto out_free_backup_vhdr; + + error = -EINVAL; + switch (sbi->s_vhdr->signature) { + case cpu_to_be16(HFSPLUS_VOLHEAD_SIGX): + set_bit(HFSPLUS_SB_HFSX, &sbi->flags); + fallthrough; + case cpu_to_be16(HFSPLUS_VOLHEAD_SIG): + break; + case cpu_to_be16(HFSP_WRAP_MAGIC): + if (!hfsplus_read_mdb(sbi->s_vhdr, &wd)) + goto out_free_backup_vhdr; + wd.ablk_size >>= HFSPLUS_SECTOR_SHIFT; + part_start += (sector_t)wd.ablk_start + + (sector_t)wd.embed_start * wd.ablk_size; + part_size = (sector_t)wd.embed_count * wd.ablk_size; + goto reread; + default: + /* + * Check for a partition block. + * + * (should do this only for cdrom/loop though) + */ + if (hfs_part_find(sb, &part_start, &part_size)) + goto out_free_backup_vhdr; + goto reread; + } + + error = hfsplus_submit_bio(sb, part_start + part_size - 2, + sbi->s_backup_vhdr_buf, + (void **)&sbi->s_backup_vhdr, REQ_OP_READ); + if (error) + goto out_free_backup_vhdr; + + error = -EINVAL; + if (sbi->s_backup_vhdr->signature != sbi->s_vhdr->signature) { + pr_warn("invalid secondary volume header\n"); + goto out_free_backup_vhdr; + } + + blocksize = be32_to_cpu(sbi->s_vhdr->blocksize); + + /* + * Block size must be at least as large as a sector and a multiple of 2. + */ + if (blocksize < HFSPLUS_SECTOR_SIZE || ((blocksize - 1) & blocksize)) + goto out_free_backup_vhdr; + sbi->alloc_blksz = blocksize; + sbi->alloc_blksz_shift = ilog2(blocksize); + blocksize = min_t(u32, sbi->alloc_blksz, PAGE_SIZE); + + /* + * Align block size to block offset. + */ + while (part_start & ((blocksize >> HFSPLUS_SECTOR_SHIFT) - 1)) + blocksize >>= 1; + + if (sb_set_blocksize(sb, blocksize) != blocksize) { + pr_err("unable to set blocksize to %u!\n", blocksize); + goto out_free_backup_vhdr; + } + + sbi->blockoffset = + part_start >> (sb->s_blocksize_bits - HFSPLUS_SECTOR_SHIFT); + sbi->part_start = part_start; + sbi->sect_count = part_size; + sbi->fs_shift = sbi->alloc_blksz_shift - sb->s_blocksize_bits; + return 0; + +out_free_backup_vhdr: + kfree(sbi->s_backup_vhdr_buf); +out_free_vhdr: + kfree(sbi->s_vhdr_buf); +out: + return error; +} -- cgit v1.3.1