x86/kprobes: Fix crash when probing CS CALL instructionsgrafted

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) <mhiramat@kernel.org>
Signed-off-by: Jinke Han <jinkehan@didiglobal.com>
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Reviewed-by: Masami Hiramatsu (Google) <mhiramat@kernel.org>
Acked-by: Yafang Shao <laoar.shao@gmail.com>
Acked-by: Borislav Petkov <bp@alien8.de>
Cc: Peter Zijlstra <peterz@infradead.org>
Link: https://patch.msgid.link/20260908073742.GA10517@didi-ThinkCentre-M920t-N000

1 files changed, 335 insertions, 0 deletions

diff --git a/fs/hpfs/map.c b/fs/hpfs/map.c
new file mode 100644
index 000000000..be7323350
--- /dev/null
+++ b/fs/hpfs/map.c
@@ -0,0 +1,335 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ *  linux/fs/hpfs/map.c
+ *
+ *  Mikulas Patocka (mikulas@artax.karlin.mff.cuni.cz), 1998-1999
+ *
+ *  mapping structures to memory with some minimal checks
+ */
+
+#include "hpfs_fn.h"
+
+__le32 *hpfs_map_dnode_bitmap(struct super_block *s, struct quad_buffer_head *qbh)
+{
+	return hpfs_map_4sectors(s, hpfs_sb(s)->sb_dmap, qbh, 0);
+}
+
+__le32 *hpfs_map_bitmap(struct super_block *s, unsigned bmp_block,
+			 struct quad_buffer_head *qbh, char *id)
+{
+	secno sec;
+	__le32 *ret;
+	unsigned n_bands = (hpfs_sb(s)->sb_fs_size + 0x3fff) >> 14;
+	if (hpfs_sb(s)->sb_chk) if (bmp_block >= n_bands) {
+		hpfs_error(s, "hpfs_map_bitmap called with bad parameter: %08x at %s", bmp_block, id);
+		return NULL;
+	}
+	sec = le32_to_cpu(hpfs_sb(s)->sb_bmp_dir[bmp_block]);
+	if (!sec || sec > hpfs_sb(s)->sb_fs_size-4) {
+		hpfs_error(s, "invalid bitmap block pointer %08x -> %08x at %s", bmp_block, sec, id);
+		return NULL;
+	}
+	ret = hpfs_map_4sectors(s, sec, qbh, 4);
+	if (ret) hpfs_prefetch_bitmap(s, bmp_block + 1);
+	return ret;
+}
+
+void hpfs_prefetch_bitmap(struct super_block *s, unsigned bmp_block)
+{
+	unsigned to_prefetch, next_prefetch;
+	unsigned n_bands = (hpfs_sb(s)->sb_fs_size + 0x3fff) >> 14;
+	if (unlikely(bmp_block >= n_bands))
+		return;
+	to_prefetch = le32_to_cpu(hpfs_sb(s)->sb_bmp_dir[bmp_block]);
+	if (unlikely(bmp_block + 1 >= n_bands))
+		next_prefetch = 0;
+	else
+		next_prefetch = le32_to_cpu(hpfs_sb(s)->sb_bmp_dir[bmp_block + 1]);
+	hpfs_prefetch_sectors(s, to_prefetch, 4 + 4 * (to_prefetch + 4 == next_prefetch));
+}
+
+/*
+ * Load first code page into kernel memory, return pointer to 256-byte array,
+ * first 128 bytes are uppercasing table for chars 128-255, next 128 bytes are
+ * lowercasing table
+ */
+
+unsigned char *hpfs_load_code_page(struct super_block *s, secno cps)
+{
+	struct buffer_head *bh;
+	secno cpds;
+	unsigned cpi;
+	unsigned char *ptr;
+	unsigned char *cp_table;
+	int i;
+	struct code_page_data *cpd;
+	struct code_page_directory *cp = hpfs_map_sector(s, cps, &bh, 0);
+	if (!cp) return NULL;
+	if (le32_to_cpu(cp->magic) != CP_DIR_MAGIC) {
+		pr_err("Code page directory magic doesn't match (magic = %08x)\n",
+			le32_to_cpu(cp->magic));
+		brelse(bh);
+		return NULL;
+	}
+	if (!le32_to_cpu(cp->n_code_pages)) {
+		pr_err("n_code_pages == 0\n");
+		brelse(bh);
+		return NULL;
+	}
+	cpds = le32_to_cpu(cp->array[0].code_page_data);
+	cpi = le16_to_cpu(cp->array[0].index);
+	brelse(bh);
+
+	if (cpi >= 3) {
+		pr_err("Code page index out of array\n");
+		return NULL;
+	}
+	
+	if (!(cpd = hpfs_map_sector(s, cpds, &bh, 0))) return NULL;
+	if (le16_to_cpu(cpd->offs[cpi]) > 0x178) {
+		pr_err("Code page index out of sector\n");
+		brelse(bh);
+		return NULL;
+	}
+	ptr = (unsigned char *)cpd + le16_to_cpu(cpd->offs[cpi]) + 6;
+	if (!(cp_table = kmalloc(256, GFP_KERNEL))) {
+		pr_err("out of memory for code page table\n");
+		brelse(bh);
+		return NULL;
+	}
+	memcpy(cp_table, ptr, 128);
+	brelse(bh);
+
+	/* Try to build lowercasing table from uppercasing one */
+
+	for (i=128; i<256; i++) cp_table[i]=i;
+	for (i=128; i<256; i++) if (cp_table[i-128]!=i && cp_table[i-128]>=128)
+		cp_table[cp_table[i-128]] = i;
+	
+	return cp_table;
+}
+
+__le32 *hpfs_load_bitmap_directory(struct super_block *s, secno bmp)
+{
+	struct buffer_head *bh;
+	int n = (hpfs_sb(s)->sb_fs_size + 0x200000 - 1) >> 21;
+	int i;
+	__le32 *b;
+	if (!(b = kmalloc_array(n, 512, GFP_KERNEL))) {
+		pr_err("can't allocate memory for bitmap directory\n");
+		return NULL;
+	}	
+	for (i=0;i<n;i++) {
+		__le32 *d = hpfs_map_sector(s, bmp+i, &bh, n - i - 1);
+		if (!d) {
+			kfree(b);
+			return NULL;
+		}
+		memcpy((char *)b + 512 * i, d, 512);
+		brelse(bh);
+	}
+	return b;
+}
+
+void hpfs_load_hotfix_map(struct super_block *s, struct hpfs_spare_block *spareblock)
+{
+	struct quad_buffer_head qbh;
+	__le32 *directory;
+	u32 n_hotfixes, n_used_hotfixes;
+	unsigned i;
+
+	n_hotfixes = le32_to_cpu(spareblock->n_spares);
+	n_used_hotfixes = le32_to_cpu(spareblock->n_spares_used);
+
+	if (n_hotfixes > 256 || n_used_hotfixes > n_hotfixes) {
+		hpfs_error(s, "invalid number of hotfixes: %u, used: %u", n_hotfixes, n_used_hotfixes);
+		return;
+	}
+	if (!(directory = hpfs_map_4sectors(s, le32_to_cpu(spareblock->hotfix_map), &qbh, 0))) {
+		hpfs_error(s, "can't load hotfix map");
+		return;
+	}
+	for (i = 0; i < n_used_hotfixes; i++) {
+		hpfs_sb(s)->hotfix_from[i] = le32_to_cpu(directory[i]);
+		hpfs_sb(s)->hotfix_to[i] = le32_to_cpu(directory[n_hotfixes + i]);
+	}
+	hpfs_sb(s)->n_hotfixes = n_used_hotfixes;
+	hpfs_brelse4(&qbh);
+}
+
+/*
+ * Load fnode to memory
+ */
+
+struct fnode *hpfs_map_fnode(struct super_block *s, ino_t ino, struct buffer_head **bhp)
+{
+	struct fnode *fnode;
+	if (hpfs_sb(s)->sb_chk) if (hpfs_chk_sectors(s, ino, 1, "fnode")) {
+		return NULL;
+	}
+	if ((fnode = hpfs_map_sector(s, ino, bhp, FNODE_RD_AHEAD))) {
+		if (hpfs_sb(s)->sb_chk) {
+			struct extended_attribute *ea;
+			struct extended_attribute *ea_end;
+			if (le32_to_cpu(fnode->magic) != FNODE_MAGIC) {
+				hpfs_error(s, "bad magic on fnode %08lx",
+					(unsigned long)ino);
+				goto bail;
+			}
+			if (!fnode_is_dir(fnode)) {
+				if ((unsigned)fnode->btree.n_used_nodes + (unsigned)fnode->btree.n_free_nodes !=
+				    (bp_internal(GET_BTREE_PTR(&fnode->btree)) ? 12 : 8)) {
+					hpfs_error(s,
+					   "bad number of nodes in fnode %08lx",
+					    (unsigned long)ino);
+					goto bail;
+				}
+				if (le16_to_cpu(fnode->btree.first_free) !=
+				    8 + fnode->btree.n_used_nodes * (bp_internal(GET_BTREE_PTR(&fnode->btree)) ? 8 : 12)) {
+					hpfs_error(s,
+					    "bad first_free pointer in fnode %08lx",
+					    (unsigned long)ino);
+					goto bail;
+				}
+			}
+			if (le16_to_cpu(fnode->ea_size_s) && (le16_to_cpu(fnode->ea_offs) < 0xc4 ||
+			   le16_to_cpu(fnode->ea_offs) + le16_to_cpu(fnode->acl_size_s) + le16_to_cpu(fnode->ea_size_s) > 0x200)) {
+				hpfs_error(s,
+					"bad EA info in fnode %08lx: ea_offs == %04x ea_size_s == %04x",
+					(unsigned long)ino,
+					le16_to_cpu(fnode->ea_offs), le16_to_cpu(fnode->ea_size_s));
+				goto bail;
+			}
+			ea = fnode_ea(fnode);
+			ea_end = fnode_end_ea(fnode);
+			while (ea != ea_end) {
+				if (ea > ea_end) {
+					hpfs_error(s, "bad EA in fnode %08lx",
+						(unsigned long)ino);
+					goto bail;
+				}
+				ea = next_ea(ea);
+			}
+		}
+	}
+	return fnode;
+	bail:
+	brelse(*bhp);
+	return NULL;
+}
+
+struct anode *hpfs_map_anode(struct super_block *s, anode_secno ano, struct buffer_head **bhp)
+{
+	struct anode *anode;
+	if (hpfs_sb(s)->sb_chk) if (hpfs_chk_sectors(s, ano, 1, "anode")) return NULL;
+	if ((anode = hpfs_map_sector(s, ano, bhp, ANODE_RD_AHEAD)))
+		if (hpfs_sb(s)->sb_chk) {
+			if (le32_to_cpu(anode->magic) != ANODE_MAGIC) {
+				hpfs_error(s, "bad magic on anode %08x", ano);
+				goto bail;
+			}
+			if (le32_to_cpu(anode->self) != ano) {
+				hpfs_error(s, "self pointer invalid on anode %08x", ano);
+				goto bail;
+			}
+			if ((unsigned)anode->btree.n_used_nodes + (unsigned)anode->btree.n_free_nodes !=
+			    (bp_internal(GET_BTREE_PTR(&anode->btree)) ? 60 : 40)) {
+				hpfs_error(s, "bad number of nodes in anode %08x", ano);
+				goto bail;
+			}
+			if (le16_to_cpu(anode->btree.first_free) !=
+			    8 + anode->btree.n_used_nodes * (bp_internal(GET_BTREE_PTR(&anode->btree)) ? 8 : 12)) {
+				hpfs_error(s, "bad first_free pointer in anode %08x", ano);
+				goto bail;
+			}
+		}
+	return anode;
+	bail:
+	brelse(*bhp);
+	return NULL;
+}
+
+/*
+ * Load dnode to memory and do some checks
+ */
+
+struct dnode *hpfs_map_dnode(struct super_block *s, unsigned secno,
+			     struct quad_buffer_head *qbh)
+{
+	struct dnode *dnode;
+	if (hpfs_sb(s)->sb_chk) {
+		if (hpfs_chk_sectors(s, secno, 4, "dnode")) return NULL;
+		if (secno & 3) {
+			hpfs_error(s, "dnode %08x not byte-aligned", secno);
+			return NULL;
+		}	
+	}
+	if ((dnode = hpfs_map_4sectors(s, secno, qbh, DNODE_RD_AHEAD)))
+		if (hpfs_sb(s)->sb_chk) {
+			unsigned p, pp = 0;
+			unsigned char *d = (unsigned char *)dnode;
+			int b = 0;
+			if (le32_to_cpu(dnode->magic) != DNODE_MAGIC) {
+				hpfs_error(s, "bad magic on dnode %08x", secno);
+				goto bail;
+			}
+			if (le32_to_cpu(dnode->self) != secno)
+				hpfs_error(s, "bad self pointer on dnode %08x self = %08x", secno, le32_to_cpu(dnode->self));
+			/* Check dirents - bad dirents would cause infinite
+			   loops or shooting to memory */
+			if (le32_to_cpu(dnode->first_free) > 2048) {
+				hpfs_error(s, "dnode %08x has first_free == %08x", secno, le32_to_cpu(dnode->first_free));
+				goto bail;
+			}
+			for (p = 20; p < le32_to_cpu(dnode->first_free); p += d[p] + (d[p+1] << 8)) {
+				struct hpfs_dirent *de = (struct hpfs_dirent *)((char *)dnode + p);
+				if (le16_to_cpu(de->length) > 292 || (le16_to_cpu(de->length) < 32) || (le16_to_cpu(de->length) & 3) || p + le16_to_cpu(de->length) > 2048) {
+					hpfs_error(s, "bad dirent size in dnode %08x, dirent %03x, last %03x", secno, p, pp);
+					goto bail;
+				}
+				if (((31 + de->namelen + de->down*4 + 3) & ~3) != le16_to_cpu(de->length)) {
+					if (((31 + de->namelen + de->down*4 + 3) & ~3) < le16_to_cpu(de->length) && s->s_flags & SB_RDONLY) goto ok;
+					hpfs_error(s, "namelen does not match dirent size in dnode %08x, dirent %03x, last %03x", secno, p, pp);
+					goto bail;
+				}
+				ok:
+				if (hpfs_sb(s)->sb_chk >= 2) b |= 1 << de->down;
+				if (de->down) if (de_down_pointer(de) < 0x10) {
+					hpfs_error(s, "bad down pointer in dnode %08x, dirent %03x, last %03x", secno, p, pp);
+					goto bail;
+				}
+				pp = p;
+				
+			}
+			if (p != le32_to_cpu(dnode->first_free)) {
+				hpfs_error(s, "size on last dirent does not match first_free; dnode %08x", secno);
+				goto bail;
+			}
+			if (d[pp + 30] != 1 || d[pp + 31] != 255) {
+				hpfs_error(s, "dnode %08x does not end with \\377 entry", secno);
+				goto bail;
+			}
+			if (b == 3)
+				pr_err("unbalanced dnode tree, dnode %08x; see hpfs.txt 4 more info\n",
+					secno);
+		}
+	return dnode;
+	bail:
+	hpfs_brelse4(qbh);
+	return NULL;
+}
+
+dnode_secno hpfs_fnode_dno(struct super_block *s, ino_t ino)
+{
+	struct buffer_head *bh;
+	struct fnode *fnode;
+	dnode_secno dno;
+
+	fnode = hpfs_map_fnode(s, ino, &bh);
+	if (!fnode)
+		return 0;
+
+	dno = le32_to_cpu(fnode->u.external[0].disk_secno);
+	brelse(bh);
+	return dno;
+}