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, 481 insertions, 0 deletions

diff --git a/drivers/ptp/ptp_sysfs.c b/drivers/ptp/ptp_sysfs.c
new file mode 100644
index 000000000..dc398c6b7
--- /dev/null
+++ b/drivers/ptp/ptp_sysfs.c
@@ -0,0 +1,481 @@
+// SPDX-License-Identifier: GPL-2.0-or-later
+/*
+ * PTP 1588 clock support - sysfs interface.
+ *
+ * Copyright (C) 2010 OMICRON electronics GmbH
+ * Copyright 2021 NXP
+ */
+#include <linux/capability.h>
+#include <linux/slab.h>
+
+#include "ptp_private.h"
+
+static ssize_t clock_name_show(struct device *dev,
+			       struct device_attribute *attr, char *page)
+{
+	struct ptp_clock *ptp = dev_get_drvdata(dev);
+	return sysfs_emit(page, "%s\n", ptp->info->name);
+}
+static DEVICE_ATTR_RO(clock_name);
+
+static ssize_t max_phase_adjustment_show(struct device *dev,
+					 struct device_attribute *attr,
+					 char *page)
+{
+	struct ptp_clock *ptp = dev_get_drvdata(dev);
+
+	return sysfs_emit(page, "%d\n", ptp->info->getmaxphase(ptp->info));
+}
+static DEVICE_ATTR_RO(max_phase_adjustment);
+
+#define PTP_SHOW_INT(name, var)						\
+static ssize_t var##_show(struct device *dev,				\
+			   struct device_attribute *attr, char *page)	\
+{									\
+	struct ptp_clock *ptp = dev_get_drvdata(dev);			\
+	return sysfs_emit(page, "%d\n", ptp->info->var);	\
+}									\
+static DEVICE_ATTR(name, 0444, var##_show, NULL);
+
+PTP_SHOW_INT(max_adjustment, max_adj);
+PTP_SHOW_INT(n_alarms, n_alarm);
+PTP_SHOW_INT(n_external_timestamps, n_ext_ts);
+PTP_SHOW_INT(n_periodic_outputs, n_per_out);
+PTP_SHOW_INT(n_programmable_pins, n_pins);
+PTP_SHOW_INT(pps_available, pps);
+
+static ssize_t extts_enable_store(struct device *dev,
+				  struct device_attribute *attr,
+				  const char *buf, size_t count)
+{
+	struct ptp_clock *ptp = dev_get_drvdata(dev);
+	struct ptp_clock_info *ops = ptp->info;
+	struct ptp_clock_request req = { .type = PTP_CLK_REQ_EXTTS };
+	int cnt, enable;
+	int err = -EINVAL;
+
+	cnt = sscanf(buf, "%u %d", &req.extts.index, &enable);
+	if (cnt != 2)
+		goto out;
+	if (req.extts.index >= ops->n_ext_ts)
+		goto out;
+
+	err = ops->enable(ops, &req, enable ? 1 : 0);
+	if (err)
+		goto out;
+
+	return count;
+out:
+	return err;
+}
+static DEVICE_ATTR(extts_enable, 0220, NULL, extts_enable_store);
+
+static ssize_t extts_fifo_show(struct device *dev,
+			       struct device_attribute *attr, char *page)
+{
+	struct ptp_clock *ptp = dev_get_drvdata(dev);
+	struct timestamp_event_queue *queue;
+	struct ptp_extts_event event;
+	unsigned long flags;
+	size_t qcnt;
+	int cnt = 0;
+
+	cnt = list_count_nodes(&ptp->tsevqs);
+	if (cnt <= 0)
+		goto out;
+
+	/* The sysfs fifo will always draw from the fist queue */
+	queue = list_first_entry(&ptp->tsevqs, struct timestamp_event_queue,
+				 qlist);
+
+	memset(&event, 0, sizeof(event));
+	spin_lock_irqsave(&queue->lock, flags);
+	qcnt = queue_cnt(queue);
+	if (qcnt) {
+		event = queue->buf[queue->head];
+		/* Paired with READ_ONCE() in queue_cnt() */
+		WRITE_ONCE(queue->head, (queue->head + 1) % PTP_MAX_TIMESTAMPS);
+	}
+	spin_unlock_irqrestore(&queue->lock, flags);
+
+	if (!qcnt)
+		goto out;
+
+	cnt = sysfs_emit(page, "%u %lld %u\n",
+			 event.index, event.t.sec, event.t.nsec);
+out:
+	return cnt;
+}
+static DEVICE_ATTR(fifo, 0444, extts_fifo_show, NULL);
+
+static ssize_t period_store(struct device *dev,
+			    struct device_attribute *attr,
+			    const char *buf, size_t count)
+{
+	struct ptp_clock *ptp = dev_get_drvdata(dev);
+	struct ptp_clock_info *ops = ptp->info;
+	struct ptp_clock_request req = { .type = PTP_CLK_REQ_PEROUT };
+	int cnt, enable, err = -EINVAL;
+
+	cnt = sscanf(buf, "%u %lld %u %lld %u", &req.perout.index,
+		     &req.perout.start.sec, &req.perout.start.nsec,
+		     &req.perout.period.sec, &req.perout.period.nsec);
+	if (cnt != 5)
+		goto out;
+	if (req.perout.index >= ops->n_per_out)
+		goto out;
+
+	enable = req.perout.period.sec || req.perout.period.nsec;
+	err = ops->enable(ops, &req, enable);
+	if (err)
+		goto out;
+
+	return count;
+out:
+	return err;
+}
+static DEVICE_ATTR(period, 0220, NULL, period_store);
+
+static ssize_t pps_enable_store(struct device *dev,
+				struct device_attribute *attr,
+				const char *buf, size_t count)
+{
+	struct ptp_clock *ptp = dev_get_drvdata(dev);
+	struct ptp_clock_info *ops = ptp->info;
+	struct ptp_clock_request req = { .type = PTP_CLK_REQ_PPS };
+	int cnt, enable;
+	int err = -EINVAL;
+
+	if (!capable(CAP_SYS_TIME))
+		return -EPERM;
+
+	cnt = sscanf(buf, "%d", &enable);
+	if (cnt != 1)
+		goto out;
+
+	err = ops->enable(ops, &req, enable ? 1 : 0);
+	if (err)
+		goto out;
+
+	return count;
+out:
+	return err;
+}
+static DEVICE_ATTR(pps_enable, 0220, NULL, pps_enable_store);
+
+static int unregister_vclock(struct device *dev, void *data)
+{
+	struct ptp_clock *ptp = dev_get_drvdata(dev);
+	struct ptp_clock_info *info = ptp->info;
+	struct ptp_vclock *vclock;
+	u32 *num = data;
+
+	vclock = info_to_vclock(info);
+	dev_info(dev->parent, "delete virtual clock ptp%d\n",
+		 vclock->clock->index);
+
+	ptp_vclock_unregister(vclock);
+	(*num)--;
+
+	/* For break. Not error. */
+	if (*num == 0)
+		return -EINVAL;
+
+	return 0;
+}
+
+static ssize_t n_vclocks_show(struct device *dev,
+			      struct device_attribute *attr, char *page)
+{
+	struct ptp_clock *ptp = dev_get_drvdata(dev);
+	ssize_t size;
+
+	if (mutex_lock_interruptible(&ptp->n_vclocks_mux))
+		return -ERESTARTSYS;
+
+	size = sysfs_emit(page, "%u\n", ptp->n_vclocks);
+
+	mutex_unlock(&ptp->n_vclocks_mux);
+
+	return size;
+}
+
+static ssize_t n_vclocks_store(struct device *dev,
+			       struct device_attribute *attr,
+			       const char *buf, size_t count)
+{
+	struct ptp_clock *ptp = dev_get_drvdata(dev);
+	struct ptp_vclock *vclock;
+	int err = -EINVAL;
+	u32 num, i;
+
+	if (kstrtou32(buf, 0, &num))
+		return err;
+
+	if (mutex_lock_interruptible(&ptp->n_vclocks_mux))
+		return -ERESTARTSYS;
+
+	if (num > ptp->max_vclocks) {
+		dev_err(dev, "max value is %d\n", ptp->max_vclocks);
+		goto out;
+	}
+
+	/* Need to create more vclocks */
+	if (num > ptp->n_vclocks) {
+		for (i = 0; i < num - ptp->n_vclocks; i++) {
+			vclock = ptp_vclock_register(ptp);
+			if (!vclock)
+				goto out;
+
+			*(ptp->vclock_index + ptp->n_vclocks + i) =
+				vclock->clock->index;
+
+			dev_info(dev, "new virtual clock ptp%d\n",
+				 vclock->clock->index);
+		}
+	}
+
+	/* Need to delete vclocks */
+	if (num < ptp->n_vclocks) {
+		i = ptp->n_vclocks - num;
+		device_for_each_child_reverse(dev, &i,
+					      unregister_vclock);
+
+		for (i = 1; i <= ptp->n_vclocks - num; i++)
+			*(ptp->vclock_index + ptp->n_vclocks - i) = -1;
+	}
+
+	/* Need to inform about changed physical clock behavior */
+	if (!ptp->has_cycles) {
+		if (num == 0)
+			dev_info(dev, "only physical clock in use now\n");
+		else
+			dev_info(dev, "guarantee physical clock free running\n");
+	}
+
+	ptp->n_vclocks = num;
+	mutex_unlock(&ptp->n_vclocks_mux);
+
+	return count;
+out:
+	mutex_unlock(&ptp->n_vclocks_mux);
+	return err;
+}
+static DEVICE_ATTR_RW(n_vclocks);
+
+static ssize_t max_vclocks_show(struct device *dev,
+				struct device_attribute *attr, char *page)
+{
+	struct ptp_clock *ptp = dev_get_drvdata(dev);
+	ssize_t size;
+
+	size = sysfs_emit(page, "%u\n", ptp->max_vclocks);
+
+	return size;
+}
+
+static ssize_t max_vclocks_store(struct device *dev,
+				 struct device_attribute *attr,
+				 const char *buf, size_t count)
+{
+	struct ptp_clock *ptp = dev_get_drvdata(dev);
+	unsigned int *vclock_index;
+	int err = -EINVAL;
+	size_t size;
+	u32 max;
+
+	if (kstrtou32(buf, 0, &max) || max == 0 || max > PTP_MAX_VCLOCKS_LIMIT)
+		return -EINVAL;
+
+	if (max == ptp->max_vclocks)
+		return count;
+
+	if (mutex_lock_interruptible(&ptp->n_vclocks_mux))
+		return -ERESTARTSYS;
+
+	if (max < ptp->n_vclocks)
+		goto out;
+
+	vclock_index = kcalloc(max, sizeof(int), GFP_KERNEL);
+	if (!vclock_index) {
+		err = -ENOMEM;
+		goto out;
+	}
+
+	size = sizeof(int) * ptp->n_vclocks;
+	memcpy(vclock_index, ptp->vclock_index, size);
+
+	kfree(ptp->vclock_index);
+	ptp->vclock_index = vclock_index;
+	ptp->max_vclocks = max;
+
+	mutex_unlock(&ptp->n_vclocks_mux);
+
+	return count;
+out:
+	mutex_unlock(&ptp->n_vclocks_mux);
+	return err;
+}
+static DEVICE_ATTR_RW(max_vclocks);
+
+static struct attribute *ptp_attrs[] = {
+	&dev_attr_clock_name.attr,
+
+	&dev_attr_max_adjustment.attr,
+	&dev_attr_max_phase_adjustment.attr,
+	&dev_attr_n_alarms.attr,
+	&dev_attr_n_external_timestamps.attr,
+	&dev_attr_n_periodic_outputs.attr,
+	&dev_attr_n_programmable_pins.attr,
+	&dev_attr_pps_available.attr,
+
+	&dev_attr_extts_enable.attr,
+	&dev_attr_fifo.attr,
+	&dev_attr_period.attr,
+	&dev_attr_pps_enable.attr,
+	&dev_attr_n_vclocks.attr,
+	&dev_attr_max_vclocks.attr,
+	NULL
+};
+
+static umode_t ptp_is_attribute_visible(struct kobject *kobj,
+					struct attribute *attr, int n)
+{
+	struct device *dev = kobj_to_dev(kobj);
+	struct ptp_clock *ptp = dev_get_drvdata(dev);
+	struct ptp_clock_info *info = ptp->info;
+	umode_t mode = attr->mode;
+
+	if (attr == &dev_attr_extts_enable.attr ||
+	    attr == &dev_attr_fifo.attr) {
+		if (!info->n_ext_ts)
+			mode = 0;
+	} else if (attr == &dev_attr_period.attr) {
+		if (!info->n_per_out)
+			mode = 0;
+	} else if (attr == &dev_attr_pps_enable.attr) {
+		if (!info->pps)
+			mode = 0;
+	} else if (attr == &dev_attr_n_vclocks.attr ||
+		   attr == &dev_attr_max_vclocks.attr) {
+		if (ptp->is_virtual_clock)
+			mode = 0;
+	} else if (attr == &dev_attr_max_phase_adjustment.attr) {
+		if (!info->adjphase || !info->getmaxphase)
+			mode = 0;
+	}
+
+	return mode;
+}
+
+static const struct attribute_group ptp_group = {
+	.is_visible	= ptp_is_attribute_visible,
+	.attrs		= ptp_attrs,
+};
+
+const struct attribute_group *ptp_groups[] = {
+	&ptp_group,
+	NULL
+};
+
+static int ptp_pin_name2index(struct ptp_clock *ptp, const char *name)
+{
+	int i;
+	for (i = 0; i < ptp->info->n_pins; i++) {
+		if (!strcmp(ptp->info->pin_config[i].name, name))
+			return i;
+	}
+	return -1;
+}
+
+static ssize_t ptp_pin_show(struct device *dev, struct device_attribute *attr,
+			    char *page)
+{
+	struct ptp_clock *ptp = dev_get_drvdata(dev);
+	unsigned int func, chan;
+	int index;
+
+	index = ptp_pin_name2index(ptp, attr->attr.name);
+	if (index < 0)
+		return -EINVAL;
+
+	if (mutex_lock_interruptible(&ptp->pincfg_mux))
+		return -ERESTARTSYS;
+
+	func = ptp->info->pin_config[index].func;
+	chan = ptp->info->pin_config[index].chan;
+
+	mutex_unlock(&ptp->pincfg_mux);
+
+	return sysfs_emit(page, "%u %u\n", func, chan);
+}
+
+static ssize_t ptp_pin_store(struct device *dev, struct device_attribute *attr,
+			     const char *buf, size_t count)
+{
+	struct ptp_clock *ptp = dev_get_drvdata(dev);
+	unsigned int func, chan;
+	int cnt, err, index;
+
+	cnt = sscanf(buf, "%u %u", &func, &chan);
+	if (cnt != 2)
+		return -EINVAL;
+
+	index = ptp_pin_name2index(ptp, attr->attr.name);
+	if (index < 0)
+		return -EINVAL;
+
+	if (mutex_lock_interruptible(&ptp->pincfg_mux))
+		return -ERESTARTSYS;
+	err = ptp_set_pinfunc(ptp, index, func, chan);
+	mutex_unlock(&ptp->pincfg_mux);
+	if (err)
+		return err;
+
+	return count;
+}
+
+int ptp_populate_pin_groups(struct ptp_clock *ptp)
+{
+	struct ptp_clock_info *info = ptp->info;
+	int err = -ENOMEM, i, n_pins = info->n_pins;
+
+	if (!n_pins)
+		return 0;
+
+	ptp->pin_dev_attr = kzalloc_objs(*ptp->pin_dev_attr, n_pins);
+	if (!ptp->pin_dev_attr)
+		goto no_dev_attr;
+
+	ptp->pin_attr = kzalloc_objs(*ptp->pin_attr, 1 + n_pins);
+	if (!ptp->pin_attr)
+		goto no_pin_attr;
+
+	for (i = 0; i < n_pins; i++) {
+		struct device_attribute *da = &ptp->pin_dev_attr[i];
+		sysfs_attr_init(&da->attr);
+		da->attr.name = info->pin_config[i].name;
+		da->attr.mode = 0644;
+		da->show = ptp_pin_show;
+		da->store = ptp_pin_store;
+		ptp->pin_attr[i] = &da->attr;
+	}
+
+	ptp->pin_attr_group.name = "pins";
+	ptp->pin_attr_group.attrs = ptp->pin_attr;
+
+	ptp->pin_attr_groups[0] = &ptp->pin_attr_group;
+
+	return 0;
+
+no_pin_attr:
+	kfree(ptp->pin_dev_attr);
+no_dev_attr:
+	return err;
+}
+
+void ptp_cleanup_pin_groups(struct ptp_clock *ptp)
+{
+	kfree(ptp->pin_attr);
+	kfree(ptp->pin_dev_attr);
+}