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, 306 insertions, 0 deletions

diff --git a/drivers/sh/intc/handle.c b/drivers/sh/intc/handle.c
new file mode 100644
index 000000000..7863a4491
--- /dev/null
+++ b/drivers/sh/intc/handle.c
@@ -0,0 +1,306 @@
+/*
+ * Shared interrupt handling code for IPR and INTC2 types of IRQs.
+ *
+ * Copyright (C) 2007, 2008 Magnus Damm
+ * Copyright (C) 2009, 2010 Paul Mundt
+ *
+ * This file is subject to the terms and conditions of the GNU General Public
+ * License.  See the file "COPYING" in the main directory of this archive
+ * for more details.
+ */
+#include <linux/init.h>
+#include <linux/irq.h>
+#include <linux/spinlock.h>
+#include "internals.h"
+
+static unsigned long ack_handle[INTC_NR_IRQS];
+
+static intc_enum __init intc_grp_id(struct intc_desc *desc,
+				    intc_enum enum_id)
+{
+	struct intc_group *g = desc->hw.groups;
+	unsigned int i, j;
+
+	for (i = 0; g && enum_id && i < desc->hw.nr_groups; i++) {
+		g = desc->hw.groups + i;
+
+		for (j = 0; g->enum_ids[j]; j++) {
+			if (g->enum_ids[j] != enum_id)
+				continue;
+
+			return g->enum_id;
+		}
+	}
+
+	return 0;
+}
+
+static unsigned int __init _intc_mask_data(struct intc_desc *desc,
+					   struct intc_desc_int *d,
+					   intc_enum enum_id,
+					   unsigned int *reg_idx,
+					   unsigned int *fld_idx)
+{
+	struct intc_mask_reg *mr = desc->hw.mask_regs;
+	unsigned int fn, mode;
+	unsigned long reg_e, reg_d;
+
+	while (mr && enum_id && *reg_idx < desc->hw.nr_mask_regs) {
+		mr = desc->hw.mask_regs + *reg_idx;
+
+		for (; *fld_idx < ARRAY_SIZE(mr->enum_ids); (*fld_idx)++) {
+			if (mr->enum_ids[*fld_idx] != enum_id)
+				continue;
+
+			if (mr->set_reg && mr->clr_reg) {
+				fn = REG_FN_WRITE_BASE;
+				mode = MODE_DUAL_REG;
+				reg_e = mr->clr_reg;
+				reg_d = mr->set_reg;
+			} else {
+				fn = REG_FN_MODIFY_BASE;
+				if (mr->set_reg) {
+					mode = MODE_ENABLE_REG;
+					reg_e = mr->set_reg;
+					reg_d = mr->set_reg;
+				} else {
+					mode = MODE_MASK_REG;
+					reg_e = mr->clr_reg;
+					reg_d = mr->clr_reg;
+				}
+			}
+
+			fn += (mr->reg_width >> 3) - 1;
+			return _INTC_MK(fn, mode,
+					intc_get_reg(d, reg_e),
+					intc_get_reg(d, reg_d),
+					1,
+					(mr->reg_width - 1) - *fld_idx);
+		}
+
+		*fld_idx = 0;
+		(*reg_idx)++;
+	}
+
+	return 0;
+}
+
+unsigned int __init
+intc_get_mask_handle(struct intc_desc *desc, struct intc_desc_int *d,
+		     intc_enum enum_id, int do_grps)
+{
+	unsigned int i = 0;
+	unsigned int j = 0;
+	unsigned int ret;
+
+	ret = _intc_mask_data(desc, d, enum_id, &i, &j);
+	if (ret)
+		return ret;
+
+	if (do_grps)
+		return intc_get_mask_handle(desc, d, intc_grp_id(desc, enum_id), 0);
+
+	return 0;
+}
+
+static unsigned int __init _intc_prio_data(struct intc_desc *desc,
+					   struct intc_desc_int *d,
+					   intc_enum enum_id,
+					   unsigned int *reg_idx,
+					   unsigned int *fld_idx)
+{
+	struct intc_prio_reg *pr = desc->hw.prio_regs;
+	unsigned int fn, n, mode, bit;
+	unsigned long reg_e, reg_d;
+
+	while (pr && enum_id && *reg_idx < desc->hw.nr_prio_regs) {
+		pr = desc->hw.prio_regs + *reg_idx;
+
+		for (; *fld_idx < ARRAY_SIZE(pr->enum_ids); (*fld_idx)++) {
+			if (pr->enum_ids[*fld_idx] != enum_id)
+				continue;
+
+			if (pr->set_reg && pr->clr_reg) {
+				fn = REG_FN_WRITE_BASE;
+				mode = MODE_PCLR_REG;
+				reg_e = pr->set_reg;
+				reg_d = pr->clr_reg;
+			} else {
+				fn = REG_FN_MODIFY_BASE;
+				mode = MODE_PRIO_REG;
+				if (!pr->set_reg)
+					BUG();
+				reg_e = pr->set_reg;
+				reg_d = pr->set_reg;
+			}
+
+			fn += (pr->reg_width >> 3) - 1;
+			n = *fld_idx + 1;
+
+			BUG_ON(n * pr->field_width > pr->reg_width);
+
+			bit = pr->reg_width - (n * pr->field_width);
+
+			return _INTC_MK(fn, mode,
+					intc_get_reg(d, reg_e),
+					intc_get_reg(d, reg_d),
+					pr->field_width, bit);
+		}
+
+		*fld_idx = 0;
+		(*reg_idx)++;
+	}
+
+	return 0;
+}
+
+unsigned int __init
+intc_get_prio_handle(struct intc_desc *desc, struct intc_desc_int *d,
+		     intc_enum enum_id, int do_grps)
+{
+	unsigned int i = 0;
+	unsigned int j = 0;
+	unsigned int ret;
+
+	ret = _intc_prio_data(desc, d, enum_id, &i, &j);
+	if (ret)
+		return ret;
+
+	if (do_grps)
+		return intc_get_prio_handle(desc, d, intc_grp_id(desc, enum_id), 0);
+
+	return 0;
+}
+
+static unsigned int intc_ack_data(struct intc_desc *desc,
+				  struct intc_desc_int *d, intc_enum enum_id)
+{
+	struct intc_mask_reg *mr = desc->hw.ack_regs;
+	unsigned int i, j, fn, mode;
+	unsigned long reg_e, reg_d;
+
+	for (i = 0; mr && enum_id && i < desc->hw.nr_ack_regs; i++) {
+		mr = desc->hw.ack_regs + i;
+
+		for (j = 0; j < ARRAY_SIZE(mr->enum_ids); j++) {
+			if (mr->enum_ids[j] != enum_id)
+				continue;
+
+			fn = REG_FN_MODIFY_BASE;
+			mode = MODE_ENABLE_REG;
+			reg_e = mr->set_reg;
+			reg_d = mr->set_reg;
+
+			fn += (mr->reg_width >> 3) - 1;
+			return _INTC_MK(fn, mode,
+					intc_get_reg(d, reg_e),
+					intc_get_reg(d, reg_d),
+					1,
+					(mr->reg_width - 1) - j);
+		}
+	}
+
+	return 0;
+}
+
+static void intc_enable_disable(struct intc_desc_int *d,
+				unsigned long handle, int do_enable)
+{
+	unsigned long addr;
+	unsigned int cpu;
+	unsigned long (*fn)(unsigned long, unsigned long,
+		   unsigned long (*)(unsigned long, unsigned long,
+				     unsigned long),
+		   unsigned int);
+
+	if (do_enable) {
+		for (cpu = 0; cpu < SMP_NR(d, _INTC_ADDR_E(handle)); cpu++) {
+			addr = INTC_REG(d, _INTC_ADDR_E(handle), cpu);
+			fn = intc_enable_noprio_fns[_INTC_MODE(handle)];
+			fn(addr, handle, intc_reg_fns[_INTC_FN(handle)], 0);
+		}
+	} else {
+		for (cpu = 0; cpu < SMP_NR(d, _INTC_ADDR_D(handle)); cpu++) {
+			addr = INTC_REG(d, _INTC_ADDR_D(handle), cpu);
+			fn = intc_disable_fns[_INTC_MODE(handle)];
+			fn(addr, handle, intc_reg_fns[_INTC_FN(handle)], 0);
+		}
+	}
+}
+
+void __init intc_enable_disable_enum(struct intc_desc *desc,
+				     struct intc_desc_int *d,
+				     intc_enum enum_id, int enable)
+{
+	unsigned int i, j, data;
+
+	/* go through and enable/disable all mask bits */
+	i = j = 0;
+	do {
+		data = _intc_mask_data(desc, d, enum_id, &i, &j);
+		if (data)
+			intc_enable_disable(d, data, enable);
+		j++;
+	} while (data);
+
+	/* go through and enable/disable all priority fields */
+	i = j = 0;
+	do {
+		data = _intc_prio_data(desc, d, enum_id, &i, &j);
+		if (data)
+			intc_enable_disable(d, data, enable);
+
+		j++;
+	} while (data);
+}
+
+unsigned int __init
+intc_get_sense_handle(struct intc_desc *desc, struct intc_desc_int *d,
+		      intc_enum enum_id)
+{
+	struct intc_sense_reg *sr = desc->hw.sense_regs;
+	unsigned int i, j, fn, bit;
+
+	for (i = 0; sr && enum_id && i < desc->hw.nr_sense_regs; i++) {
+		sr = desc->hw.sense_regs + i;
+
+		for (j = 0; j < ARRAY_SIZE(sr->enum_ids); j++) {
+			if (sr->enum_ids[j] != enum_id)
+				continue;
+
+			fn = REG_FN_MODIFY_BASE;
+			fn += (sr->reg_width >> 3) - 1;
+
+			BUG_ON((j + 1) * sr->field_width > sr->reg_width);
+
+			bit = sr->reg_width - ((j + 1) * sr->field_width);
+
+			return _INTC_MK(fn, 0, intc_get_reg(d, sr->reg),
+					0, sr->field_width, bit);
+		}
+	}
+
+	return 0;
+}
+
+
+void intc_set_ack_handle(unsigned int irq, struct intc_desc *desc,
+			 struct intc_desc_int *d, intc_enum id)
+{
+	unsigned long flags;
+
+	/*
+	 * Nothing to do for this IRQ.
+	 */
+	if (!desc->hw.ack_regs)
+		return;
+
+	raw_spin_lock_irqsave(&intc_big_lock, flags);
+	ack_handle[irq] = intc_ack_data(desc, d, id);
+	raw_spin_unlock_irqrestore(&intc_big_lock, flags);
+}
+
+unsigned long intc_get_ack_handle(unsigned int irq)
+{
+	return ack_handle[irq];
+}