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, 257 insertions, 0 deletions

diff --git a/drivers/hwmon/occ/sysfs.c b/drivers/hwmon/occ/sysfs.c
new file mode 100644
index 000000000..2317301fc
--- /dev/null
+++ b/drivers/hwmon/occ/sysfs.c
@@ -0,0 +1,257 @@
+// SPDX-License-Identifier: GPL-2.0+
+// Copyright IBM Corp 2019
+
+#include <linux/bitops.h>
+#include <linux/device.h>
+#include <linux/export.h>
+#include <linux/hwmon-sysfs.h>
+#include <linux/kernel.h>
+#include <linux/kstrtox.h>
+#include <linux/sysfs.h>
+
+#include "common.h"
+
+/* OCC status register */
+#define OCC_STAT_MASTER			BIT(7)
+
+/* OCC extended status register */
+#define OCC_EXT_STAT_DVFS_OT		BIT(7)
+#define OCC_EXT_STAT_DVFS_POWER		BIT(6)
+#define OCC_EXT_STAT_MEM_THROTTLE	BIT(5)
+#define OCC_EXT_STAT_QUICK_DROP		BIT(4)
+#define OCC_EXT_STAT_DVFS_VDD		BIT(3)
+#define OCC_EXT_STAT_GPU_THROTTLE	GENMASK(2, 0)
+
+static ssize_t occ_active_store(struct device *dev,
+				struct device_attribute *attr,
+				const char *buf, size_t count)
+{
+	int rc;
+	bool active;
+	struct occ *occ = dev_get_drvdata(dev);
+
+	rc = kstrtobool(buf, &active);
+	if (rc)
+		return rc;
+
+	rc = occ_active(occ, active);
+	if (rc)
+		return rc;
+
+	return count;
+}
+
+static ssize_t occ_sysfs_show(struct device *dev,
+			      struct device_attribute *attr, char *buf)
+{
+	int rc;
+	int val = 0;
+	struct occ *occ = dev_get_drvdata(dev);
+	struct occ_poll_response_header *header;
+	struct sensor_device_attribute *sattr = to_sensor_dev_attr(attr);
+
+	if (occ->active) {
+		rc = occ_update_response(occ);
+		if (rc)
+			return rc;
+
+		header = (struct occ_poll_response_header *)occ->resp.data;
+
+		switch (sattr->index) {
+		case 0:
+			val = !!(header->status & OCC_STAT_MASTER);
+			break;
+		case 1:
+			val = 1;
+			break;
+		case 2:
+			val = !!(header->ext_status & OCC_EXT_STAT_DVFS_OT);
+			break;
+		case 3:
+			val = !!(header->ext_status & OCC_EXT_STAT_DVFS_POWER);
+			break;
+		case 4:
+			val = !!(header->ext_status &
+				 OCC_EXT_STAT_MEM_THROTTLE);
+			break;
+		case 5:
+			val = !!(header->ext_status & OCC_EXT_STAT_QUICK_DROP);
+			break;
+		case 6:
+			val = header->occ_state;
+			break;
+		case 7:
+			if (header->status & OCC_STAT_MASTER)
+				val = hweight8(header->occs_present);
+			else
+				val = 1;
+			break;
+		case 8:
+			val = header->ips_status;
+			break;
+		case 9:
+			val = header->mode;
+			break;
+		case 10:
+			val = !!(header->ext_status & OCC_EXT_STAT_DVFS_VDD);
+			break;
+		case 11:
+			val = header->ext_status & OCC_EXT_STAT_GPU_THROTTLE;
+			break;
+		default:
+			return -EINVAL;
+		}
+	} else {
+		if (sattr->index == 1)
+			val = 0;
+		else if (sattr->index <= 11)
+			val = -ENODATA;
+		else
+			return -EINVAL;
+	}
+
+	return sysfs_emit(buf, "%d\n", val);
+}
+
+static ssize_t occ_error_show(struct device *dev,
+			      struct device_attribute *attr, char *buf)
+{
+	struct occ *occ = dev_get_drvdata(dev);
+
+	occ_update_response(occ);
+
+	return sysfs_emit(buf, "%d\n", occ->error);
+}
+
+static SENSOR_DEVICE_ATTR(occ_master, 0444, occ_sysfs_show, NULL, 0);
+static SENSOR_DEVICE_ATTR(occ_active, 0644, occ_sysfs_show, occ_active_store,
+			  1);
+static SENSOR_DEVICE_ATTR(occ_dvfs_overtemp, 0444, occ_sysfs_show, NULL, 2);
+static SENSOR_DEVICE_ATTR(occ_dvfs_power, 0444, occ_sysfs_show, NULL, 3);
+static SENSOR_DEVICE_ATTR(occ_mem_throttle, 0444, occ_sysfs_show, NULL, 4);
+static SENSOR_DEVICE_ATTR(occ_quick_pwr_drop, 0444, occ_sysfs_show, NULL, 5);
+static SENSOR_DEVICE_ATTR(occ_state, 0444, occ_sysfs_show, NULL, 6);
+static SENSOR_DEVICE_ATTR(occs_present, 0444, occ_sysfs_show, NULL, 7);
+static SENSOR_DEVICE_ATTR(occ_ips_status, 0444, occ_sysfs_show, NULL, 8);
+static SENSOR_DEVICE_ATTR(occ_mode, 0444, occ_sysfs_show, NULL, 9);
+static SENSOR_DEVICE_ATTR(occ_dvfs_vdd, 0444, occ_sysfs_show, NULL, 10);
+static SENSOR_DEVICE_ATTR(occ_gpu_throttle, 0444, occ_sysfs_show, NULL, 11);
+static DEVICE_ATTR_RO(occ_error);
+
+static struct attribute *occ_attributes[] = {
+	&sensor_dev_attr_occ_master.dev_attr.attr,
+	&sensor_dev_attr_occ_active.dev_attr.attr,
+	&sensor_dev_attr_occ_dvfs_overtemp.dev_attr.attr,
+	&sensor_dev_attr_occ_dvfs_power.dev_attr.attr,
+	&sensor_dev_attr_occ_mem_throttle.dev_attr.attr,
+	&sensor_dev_attr_occ_quick_pwr_drop.dev_attr.attr,
+	&sensor_dev_attr_occ_state.dev_attr.attr,
+	&sensor_dev_attr_occs_present.dev_attr.attr,
+	&sensor_dev_attr_occ_ips_status.dev_attr.attr,
+	&sensor_dev_attr_occ_mode.dev_attr.attr,
+	&sensor_dev_attr_occ_dvfs_vdd.dev_attr.attr,
+	&sensor_dev_attr_occ_gpu_throttle.dev_attr.attr,
+	&dev_attr_occ_error.attr,
+	NULL
+};
+
+static const struct attribute_group occ_sysfs = {
+	.attrs = occ_attributes,
+};
+
+void occ_sysfs_poll_done(struct occ *occ)
+{
+	const char *name;
+	struct occ_poll_response_header *header =
+		(struct occ_poll_response_header *)occ->resp.data;
+
+	/*
+	 * On the first poll response, we haven't yet created the sysfs
+	 * attributes, so don't make any notify calls.
+	 */
+	if (!occ->active)
+		goto done;
+
+	if ((header->status & OCC_STAT_MASTER) !=
+	    (occ->prev_stat & OCC_STAT_MASTER)) {
+		name = sensor_dev_attr_occ_master.dev_attr.attr.name;
+		sysfs_notify(&occ->bus_dev->kobj, NULL, name);
+	}
+
+	if ((header->ext_status & OCC_EXT_STAT_DVFS_OT) !=
+	    (occ->prev_ext_stat & OCC_EXT_STAT_DVFS_OT)) {
+		name = sensor_dev_attr_occ_dvfs_overtemp.dev_attr.attr.name;
+		sysfs_notify(&occ->bus_dev->kobj, NULL, name);
+	}
+
+	if ((header->ext_status & OCC_EXT_STAT_DVFS_POWER) !=
+	    (occ->prev_ext_stat & OCC_EXT_STAT_DVFS_POWER)) {
+		name = sensor_dev_attr_occ_dvfs_power.dev_attr.attr.name;
+		sysfs_notify(&occ->bus_dev->kobj, NULL, name);
+	}
+
+	if ((header->ext_status & OCC_EXT_STAT_MEM_THROTTLE) !=
+	    (occ->prev_ext_stat & OCC_EXT_STAT_MEM_THROTTLE)) {
+		name = sensor_dev_attr_occ_mem_throttle.dev_attr.attr.name;
+		sysfs_notify(&occ->bus_dev->kobj, NULL, name);
+	}
+
+	if ((header->ext_status & OCC_EXT_STAT_QUICK_DROP) !=
+	    (occ->prev_ext_stat & OCC_EXT_STAT_QUICK_DROP)) {
+		name = sensor_dev_attr_occ_quick_pwr_drop.dev_attr.attr.name;
+		sysfs_notify(&occ->bus_dev->kobj, NULL, name);
+	}
+
+	if ((header->ext_status & OCC_EXT_STAT_DVFS_VDD) !=
+	    (occ->prev_ext_stat & OCC_EXT_STAT_DVFS_VDD)) {
+		name = sensor_dev_attr_occ_dvfs_vdd.dev_attr.attr.name;
+		sysfs_notify(&occ->bus_dev->kobj, NULL, name);
+	}
+
+	if ((header->ext_status & OCC_EXT_STAT_GPU_THROTTLE) !=
+	    (occ->prev_ext_stat & OCC_EXT_STAT_GPU_THROTTLE)) {
+		name = sensor_dev_attr_occ_gpu_throttle.dev_attr.attr.name;
+		sysfs_notify(&occ->bus_dev->kobj, NULL, name);
+	}
+
+	if ((header->status & OCC_STAT_MASTER) &&
+	    header->occs_present != occ->prev_occs_present) {
+		name = sensor_dev_attr_occs_present.dev_attr.attr.name;
+		sysfs_notify(&occ->bus_dev->kobj, NULL, name);
+	}
+
+	if (header->ips_status != occ->prev_ips_status) {
+		name = sensor_dev_attr_occ_ips_status.dev_attr.attr.name;
+		sysfs_notify(&occ->bus_dev->kobj, NULL, name);
+	}
+
+	if (header->mode != occ->prev_mode) {
+		name = sensor_dev_attr_occ_mode.dev_attr.attr.name;
+		sysfs_notify(&occ->bus_dev->kobj, NULL, name);
+	}
+
+	if (occ->error && occ->error != occ->prev_error) {
+		name = dev_attr_occ_error.attr.name;
+		sysfs_notify(&occ->bus_dev->kobj, NULL, name);
+	}
+
+	/* no notifications for OCC state; doesn't indicate error condition */
+
+done:
+	occ->prev_error = occ->error;
+	occ->prev_stat = header->status;
+	occ->prev_ext_stat = header->ext_status;
+	occ->prev_occs_present = header->occs_present;
+	occ->prev_ips_status = header->ips_status;
+	occ->prev_mode = header->mode;
+}
+
+int occ_setup_sysfs(struct occ *occ)
+{
+	return sysfs_create_group(&occ->bus_dev->kobj, &occ_sysfs);
+}
+
+void occ_shutdown_sysfs(struct occ *occ)
+{
+	sysfs_remove_group(&occ->bus_dev->kobj, &occ_sysfs);
+}