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, 236 insertions, 0 deletions

diff --git a/arch/powerpc/boot/rs6000.h b/arch/powerpc/boot/rs6000.h
new file mode 100644
index 000000000..970754a1f
--- /dev/null
+++ b/arch/powerpc/boot/rs6000.h
@@ -0,0 +1,236 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+/* IBM RS/6000 "XCOFF" file definitions for BFD.
+   Copyright (C) 1990, 1991 Free Software Foundation, Inc.
+   Written by Mimi Phuong-Thao Vo of IBM
+   and John Gilmore of Cygnus Support.  */
+
+/********************** FILE HEADER **********************/
+
+struct external_filehdr {
+	char f_magic[2];	/* magic number			*/
+	char f_nscns[2];	/* number of sections		*/
+	char f_timdat[4];	/* time & date stamp		*/
+	char f_symptr[4];	/* file pointer to symtab	*/
+	char f_nsyms[4];	/* number of symtab entries	*/
+	char f_opthdr[2];	/* sizeof(optional hdr)		*/
+	char f_flags[2];	/* flags			*/
+};
+
+        /* IBM RS/6000 */
+#define U802WRMAGIC     0730    /* writeable text segments **chh**      */
+#define U802ROMAGIC     0735    /* readonly sharable text segments      */
+#define U802TOCMAGIC    0737    /* readonly text segments and TOC       */
+
+#define BADMAG(x)	\
+	((x).f_magic != U802ROMAGIC && (x).f_magic != U802WRMAGIC && \
+	 (x).f_magic != U802TOCMAGIC)
+
+#define	FILHDR	struct external_filehdr
+#define	FILHSZ	20
+
+
+/********************** AOUT "OPTIONAL HEADER" **********************/
+
+
+struct aouthdr {
+	unsigned char	magic[2];	/* type of file			*/
+	unsigned char	vstamp[2];	/* version stamp		*/
+	unsigned char	tsize[4];	/* text size in bytes, padded to FW bdry */
+	unsigned char	dsize[4];	/* initialized data "  "	*/
+	unsigned char	bsize[4];	/* uninitialized data "   "	*/
+	unsigned char	entry[4];	/* entry pt.			*/
+	unsigned char	text_start[4];	/* base of text used for this file */
+	unsigned char	data_start[4];	/* base of data used for this file */
+	unsigned char	o_toc[4];	/* address of TOC */
+	unsigned char	o_snentry[2];	/* section number of entry point */
+	unsigned char	o_sntext[2];	/* section number of .text section */
+	unsigned char	o_sndata[2];	/* section number of .data section */
+	unsigned char	o_sntoc[2];	/* section number of TOC */
+	unsigned char	o_snloader[2];	/* section number of .loader section */
+	unsigned char	o_snbss[2];	/* section number of .bss section */
+	unsigned char	o_algntext[2];	/* .text alignment */
+	unsigned char	o_algndata[2];	/* .data alignment */
+	unsigned char	o_modtype[2];	/* module type (??) */
+	unsigned char	o_cputype[2];	/* cpu type */
+	unsigned char	o_maxstack[4];	/* max stack size (??) */
+	unsigned char	o_maxdata[4];	/* max data size (??) */
+	unsigned char	o_resv2[12];	/* reserved */
+};
+
+#define AOUTSZ 72
+#define SMALL_AOUTSZ (28)
+#define AOUTHDRSZ 72
+
+#define	RS6K_AOUTHDR_OMAGIC	0x0107	/* old: text & data writeable */
+#define	RS6K_AOUTHDR_NMAGIC	0x0108	/* new: text r/o, data r/w */
+#define	RS6K_AOUTHDR_ZMAGIC	0x010B	/* paged: text r/o, both page-aligned */
+
+
+/********************** SECTION HEADER **********************/
+
+
+struct external_scnhdr {
+	char		s_name[8];	/* section name			*/
+	char		s_paddr[4];	/* physical address, aliased s_nlib */
+	char		s_vaddr[4];	/* virtual address		*/
+	char		s_size[4];	/* section size			*/
+	char		s_scnptr[4];	/* file ptr to raw data for section */
+	char		s_relptr[4];	/* file ptr to relocation	*/
+	char		s_lnnoptr[4];	/* file ptr to line numbers	*/
+	char		s_nreloc[2];	/* number of relocation entries	*/
+	char		s_nlnno[2];	/* number of line number entries*/
+	char		s_flags[4];	/* flags			*/
+};
+
+/*
+ * names of "special" sections
+ */
+#define _TEXT	".text"
+#define _DATA	".data"
+#define _BSS	".bss"
+#define _PAD	".pad"
+#define _LOADER	".loader"
+
+#define	SCNHDR	struct external_scnhdr
+#define	SCNHSZ	40
+
+/* XCOFF uses a special .loader section with type STYP_LOADER.  */
+#define STYP_LOADER 0x1000
+
+/* XCOFF uses a special .debug section with type STYP_DEBUG.  */
+#define STYP_DEBUG 0x2000
+
+/* XCOFF handles line number or relocation overflow by creating
+   another section header with STYP_OVRFLO set.  */
+#define STYP_OVRFLO 0x8000
+
+/********************** LINE NUMBERS **********************/
+
+/* 1 line number entry for every "breakpointable" source line in a section.
+ * Line numbers are grouped on a per function basis; first entry in a function
+ * grouping will have l_lnno = 0 and in place of physical address will be the
+ * symbol table index of the function name.
+ */
+struct external_lineno {
+	union {
+		char l_symndx[4];	/* function name symbol index, iff l_lnno == 0 */
+		char l_paddr[4];	/* (physical) address of line number	*/
+	} l_addr;
+	char l_lnno[2];	/* line number */
+};
+
+
+#define	LINENO	struct external_lineno
+#define	LINESZ	6
+
+
+/********************** SYMBOLS **********************/
+
+#define E_SYMNMLEN	8	/* # characters in a symbol name	*/
+#define E_FILNMLEN	14	/* # characters in a file name		*/
+#define E_DIMNUM	4	/* # array dimensions in auxiliary entry */
+
+struct external_syment {
+	union {
+		char e_name[E_SYMNMLEN];
+		struct {
+			char e_zeroes[4];
+			char e_offset[4];
+		} e;
+	} e;
+	char e_value[4];
+	char e_scnum[2];
+	char e_type[2];
+	char e_sclass[1];
+	char e_numaux[1];
+};
+
+
+
+#define N_BTMASK	(017)
+#define N_TMASK		(060)
+#define N_BTSHFT	(4)
+#define N_TSHIFT	(2)
+
+
+union external_auxent {
+	struct {
+		char x_tagndx[4];	/* str, un, or enum tag indx */
+		union {
+			struct {
+			    char  x_lnno[2]; /* declaration line number */
+			    char  x_size[2]; /* str/union/array size */
+			} x_lnsz;
+			char x_fsize[4];	/* size of function */
+		} x_misc;
+		union {
+			struct {		/* if ISFCN, tag, or .bb */
+			    char x_lnnoptr[4];	/* ptr to fcn line # */
+			    char x_endndx[4];	/* entry ndx past block end */
+			} x_fcn;
+			struct {		/* if ISARY, up to 4 dimen. */
+			    char x_dimen[E_DIMNUM][2];
+			} x_ary;
+		} x_fcnary;
+		char x_tvndx[2];		/* tv index */
+	} x_sym;
+
+	union {
+		char x_fname[E_FILNMLEN];
+		struct {
+			char x_zeroes[4];
+			char x_offset[4];
+		} x_n;
+	} x_file;
+
+	struct {
+		char x_scnlen[4];	/* section length */
+		char x_nreloc[2];	/* # relocation entries */
+		char x_nlinno[2];	/* # line numbers */
+	} x_scn;
+
+        struct {
+		char x_tvfill[4];	/* tv fill value */
+		char x_tvlen[2];	/* length of .tv */
+		char x_tvran[2][2];	/* tv range */
+	} x_tv;		/* info about .tv section (in auxent of symbol .tv)) */
+
+	struct {
+		unsigned char x_scnlen[4];
+		unsigned char x_parmhash[4];
+		unsigned char x_snhash[2];
+		unsigned char x_smtyp[1];
+		unsigned char x_smclas[1];
+		unsigned char x_stab[4];
+		unsigned char x_snstab[2];
+	} x_csect;
+};
+
+#define	SYMENT	struct external_syment
+#define	SYMESZ	18
+#define	AUXENT	union external_auxent
+#define	AUXESZ	18
+#define DBXMASK 0x80		/* for dbx storage mask */
+#define SYMNAME_IN_DEBUG(symptr) ((symptr)->n_sclass & DBXMASK)
+
+
+
+/********************** RELOCATION DIRECTIVES **********************/
+
+
+struct external_reloc {
+  char r_vaddr[4];
+  char r_symndx[4];
+  char r_size[1];
+  char r_type[1];
+};
+
+
+#define RELOC struct external_reloc
+#define RELSZ 10
+
+#define DEFAULT_DATA_SECTION_ALIGNMENT 4
+#define DEFAULT_BSS_SECTION_ALIGNMENT 4
+#define DEFAULT_TEXT_SECTION_ALIGNMENT 4
+/* For new sections we haven't heard of before */
+#define DEFAULT_SECTION_ALIGNMENT 4