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

3 files changed, 316 insertions, 0 deletions

diff --git a/drivers/pmdomain/xilinx/Kconfig b/drivers/pmdomain/xilinx/Kconfig
new file mode 100644
index 000000000..5242753d8
--- /dev/null
+++ b/drivers/pmdomain/xilinx/Kconfig
@@ -0,0 +1,10 @@
+# SPDX-License-Identifier: GPL-2.0
+
+config ZYNQMP_PM_DOMAINS
+	bool "Enable Zynq MPSoC generic PM domains"
+	default y
+	depends on PM && ZYNQMP_FIRMWARE
+	select PM_GENERIC_DOMAINS
+	help
+	  Say yes to enable device power management through PM domains
+	  If in doubt, say N.
diff --git a/drivers/pmdomain/xilinx/Makefile b/drivers/pmdomain/xilinx/Makefile
new file mode 100644
index 000000000..a706ab699
--- /dev/null
+++ b/drivers/pmdomain/xilinx/Makefile
@@ -0,0 +1,2 @@
+# SPDX-License-Identifier: GPL-2.0
+obj-$(CONFIG_ZYNQMP_PM_DOMAINS)		+= zynqmp-pm-domains.o
diff --git a/drivers/pmdomain/xilinx/zynqmp-pm-domains.c b/drivers/pmdomain/xilinx/zynqmp-pm-domains.c
new file mode 100644
index 000000000..b5aedd6e3
--- /dev/null
+++ b/drivers/pmdomain/xilinx/zynqmp-pm-domains.c
@@ -0,0 +1,304 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * ZynqMP Generic PM domain support
+ *
+ *  Copyright (C) 2015-2019 Xilinx, Inc.
+ *
+ *  Davorin Mista <davorin.mista@aggios.com>
+ *  Jolly Shah <jollys@xilinx.com>
+ *  Rajan Vaja <rajan.vaja@xilinx.com>
+ */
+
+#include <linux/err.h>
+#include <linux/list.h>
+#include <linux/module.h>
+#include <linux/of_platform.h>
+#include <linux/platform_device.h>
+#include <linux/pm_domain.h>
+#include <linux/slab.h>
+
+#include <linux/firmware/xlnx-zynqmp.h>
+
+#define ZYNQMP_NUM_DOMAINS		(100)
+
+static int min_capability;
+
+/**
+ * struct zynqmp_pm_domain - Wrapper around struct generic_pm_domain
+ * @gpd:		Generic power domain
+ * @node_id:		PM node ID corresponding to device inside PM domain
+ * @requested:		The PM node mapped to the PM domain has been requested
+ */
+struct zynqmp_pm_domain {
+	struct generic_pm_domain gpd;
+	u32 node_id;
+	bool requested;
+};
+
+#define to_zynqmp_pm_domain(pm_domain) \
+	container_of(pm_domain, struct zynqmp_pm_domain, gpd)
+
+/**
+ * zynqmp_gpd_is_active_wakeup_path() - Check if device is in wakeup source
+ *					path
+ * @dev:	Device to check for wakeup source path
+ * @not_used:	Data member (not required)
+ *
+ * This function is checks device's child hierarchy and checks if any device is
+ * set as wakeup source.
+ *
+ * Return: 1 if device is in wakeup source path else 0
+ */
+static int zynqmp_gpd_is_active_wakeup_path(struct device *dev, void *not_used)
+{
+	int may_wakeup;
+
+	may_wakeup = device_may_wakeup(dev);
+	if (may_wakeup)
+		return may_wakeup;
+
+	return device_for_each_child(dev, NULL,
+			zynqmp_gpd_is_active_wakeup_path);
+}
+
+/**
+ * zynqmp_gpd_power_on() - Power on PM domain
+ * @domain:	Generic PM domain
+ *
+ * This function is called before devices inside a PM domain are resumed, to
+ * power on PM domain.
+ *
+ * Return: 0 on success, error code otherwise
+ */
+static int zynqmp_gpd_power_on(struct generic_pm_domain *domain)
+{
+	struct zynqmp_pm_domain *pd = to_zynqmp_pm_domain(domain);
+	int ret;
+
+	ret = zynqmp_pm_set_requirement(pd->node_id,
+					ZYNQMP_PM_CAPABILITY_ACCESS,
+					ZYNQMP_PM_MAX_QOS,
+					ZYNQMP_PM_REQUEST_ACK_BLOCKING);
+	if (ret) {
+		dev_err(&domain->dev,
+			"failed to set requirement to 0x%x for PM node id %d: %d\n",
+			ZYNQMP_PM_CAPABILITY_ACCESS, pd->node_id, ret);
+		return ret;
+	}
+
+	dev_dbg(&domain->dev, "set requirement to 0x%x for PM node id %d\n",
+		ZYNQMP_PM_CAPABILITY_ACCESS, pd->node_id);
+
+	return 0;
+}
+
+/**
+ * zynqmp_gpd_power_off() - Power off PM domain
+ * @domain:	Generic PM domain
+ *
+ * This function is called after devices inside a PM domain are suspended, to
+ * power off PM domain.
+ *
+ * Return: 0 on success, error code otherwise
+ */
+static int zynqmp_gpd_power_off(struct generic_pm_domain *domain)
+{
+	struct zynqmp_pm_domain *pd = to_zynqmp_pm_domain(domain);
+	int ret;
+	struct pm_domain_data *pdd, *tmp;
+	u32 capabilities = min_capability;
+	bool may_wakeup;
+
+	/* If domain is already released there is nothing to be done */
+	if (!pd->requested) {
+		dev_dbg(&domain->dev, "PM node id %d is already released\n",
+			pd->node_id);
+		return 0;
+	}
+
+	list_for_each_entry_safe(pdd, tmp, &domain->dev_list, list_node) {
+		/* If device is in wakeup path, set capability to WAKEUP */
+		may_wakeup = zynqmp_gpd_is_active_wakeup_path(pdd->dev, NULL);
+		if (may_wakeup) {
+			dev_dbg(pdd->dev, "device is in wakeup path in %s\n",
+				domain->name);
+			capabilities = ZYNQMP_PM_CAPABILITY_WAKEUP;
+			break;
+		}
+	}
+
+	ret = zynqmp_pm_set_requirement(pd->node_id, capabilities, 0,
+					ZYNQMP_PM_REQUEST_ACK_NO);
+	if (ret) {
+		dev_err(&domain->dev,
+			"failed to set requirement to 0x%x for PM node id %d: %d\n",
+			capabilities, pd->node_id, ret);
+		return ret;
+	}
+
+	dev_dbg(&domain->dev, "set requirement to 0x%x for PM node id %d\n",
+		capabilities, pd->node_id);
+
+	return 0;
+}
+
+/**
+ * zynqmp_gpd_attach_dev() - Attach device to the PM domain
+ * @domain:	Generic PM domain
+ * @dev:	Device to attach
+ *
+ * Return: 0 on success, error code otherwise
+ */
+static int zynqmp_gpd_attach_dev(struct generic_pm_domain *domain,
+				 struct device *dev)
+{
+	struct zynqmp_pm_domain *pd = to_zynqmp_pm_domain(domain);
+	int ret;
+
+	/* If this is not the first device to attach there is nothing to do */
+	if (domain->device_count)
+		return 0;
+
+	ret = zynqmp_pm_request_node(pd->node_id, 0, 0,
+				     ZYNQMP_PM_REQUEST_ACK_BLOCKING);
+	if (ret) {
+		dev_err(&domain->dev, "%s request failed for node %d: %d\n",
+			domain->name, pd->node_id, ret);
+		return ret;
+	}
+
+	pd->requested = true;
+
+	dev_dbg(&domain->dev, "%s requested PM node id %d\n",
+		dev_name(dev), pd->node_id);
+
+	return 0;
+}
+
+/**
+ * zynqmp_gpd_detach_dev() - Detach device from the PM domain
+ * @domain:	Generic PM domain
+ * @dev:	Device to detach
+ */
+static void zynqmp_gpd_detach_dev(struct generic_pm_domain *domain,
+				  struct device *dev)
+{
+	struct zynqmp_pm_domain *pd = to_zynqmp_pm_domain(domain);
+	int ret;
+
+	/* If this is not the last device to detach there is nothing to do */
+	if (domain->device_count)
+		return;
+
+	ret = zynqmp_pm_release_node(pd->node_id);
+	if (ret) {
+		dev_err(&domain->dev, "failed to release PM node id %d: %d\n",
+			pd->node_id, ret);
+		return;
+	}
+
+	pd->requested = false;
+
+	dev_dbg(&domain->dev, "%s released PM node id %d\n",
+		dev_name(dev), pd->node_id);
+}
+
+static struct generic_pm_domain *zynqmp_gpd_xlate
+				(const struct of_phandle_args *genpdspec, void *data)
+{
+	struct genpd_onecell_data *genpd_data = data;
+	unsigned int i, idx = genpdspec->args[0];
+	struct zynqmp_pm_domain *pd;
+
+	pd = to_zynqmp_pm_domain(genpd_data->domains[0]);
+
+	if (genpdspec->args_count != 1)
+		return ERR_PTR(-EINVAL);
+
+	/* Check for existing pm domains */
+	for (i = 0; i < ZYNQMP_NUM_DOMAINS; i++) {
+		if (pd[i].node_id == idx)
+			goto done;
+	}
+
+	/*
+	 * Add index in empty node_id of power domain list as no existing
+	 * power domain found for current index.
+	 */
+	for (i = 0; i < ZYNQMP_NUM_DOMAINS; i++) {
+		if (pd[i].node_id == 0) {
+			pd[i].node_id = idx;
+			break;
+		}
+	}
+
+done:
+	if (!genpd_data->domains[i] || i == ZYNQMP_NUM_DOMAINS)
+		return ERR_PTR(-ENOENT);
+
+	return genpd_data->domains[i];
+}
+
+static int zynqmp_gpd_probe(struct platform_device *pdev)
+{
+	int i;
+	struct genpd_onecell_data *zynqmp_pd_data;
+	struct generic_pm_domain **domains;
+	struct zynqmp_pm_domain *pd;
+	struct device *dev = &pdev->dev;
+
+	pd = devm_kcalloc(dev, ZYNQMP_NUM_DOMAINS, sizeof(*pd), GFP_KERNEL);
+	if (!pd)
+		return -ENOMEM;
+
+	zynqmp_pd_data = devm_kzalloc(dev, sizeof(*zynqmp_pd_data), GFP_KERNEL);
+	if (!zynqmp_pd_data)
+		return -ENOMEM;
+
+	zynqmp_pd_data->xlate = zynqmp_gpd_xlate;
+
+	domains = devm_kcalloc(dev, ZYNQMP_NUM_DOMAINS, sizeof(*domains),
+			       GFP_KERNEL);
+	if (!domains)
+		return -ENOMEM;
+
+	if (!of_device_is_compatible(dev->parent->of_node,
+				     "xlnx,zynqmp-firmware"))
+		min_capability = ZYNQMP_PM_CAPABILITY_UNUSABLE;
+
+	for (i = 0; i < ZYNQMP_NUM_DOMAINS; i++, pd++) {
+		pd->node_id = 0;
+		pd->gpd.name = kasprintf(GFP_KERNEL, "domain%d", i);
+		pd->gpd.power_off = zynqmp_gpd_power_off;
+		pd->gpd.power_on = zynqmp_gpd_power_on;
+		pd->gpd.attach_dev = zynqmp_gpd_attach_dev;
+		pd->gpd.detach_dev = zynqmp_gpd_detach_dev;
+
+		domains[i] = &pd->gpd;
+
+		/* Mark all PM domains as initially powered off */
+		pm_genpd_init(&pd->gpd, NULL, true);
+	}
+
+	zynqmp_pd_data->domains = domains;
+	zynqmp_pd_data->num_domains = ZYNQMP_NUM_DOMAINS;
+	of_genpd_add_provider_onecell(dev->parent->of_node, zynqmp_pd_data);
+
+	return 0;
+}
+
+static void zynqmp_gpd_remove(struct platform_device *pdev)
+{
+	of_genpd_del_provider(pdev->dev.parent->of_node);
+}
+
+static struct platform_driver zynqmp_power_domain_driver = {
+	.driver	= {
+		.name = "zynqmp_power_controller",
+	},
+	.probe = zynqmp_gpd_probe,
+	.remove = zynqmp_gpd_remove,
+};
+module_platform_driver(zynqmp_power_domain_driver);
+
+MODULE_ALIAS("platform:zynqmp_power_controller");