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, 137 insertions, 0 deletions

diff --git a/Documentation/driver-api/phy/samsung-usb2.rst b/Documentation/driver-api/phy/samsung-usb2.rst
new file mode 100644
index 000000000..c48c8b979
--- /dev/null
+++ b/Documentation/driver-api/phy/samsung-usb2.rst
@@ -0,0 +1,137 @@
+====================================
+Samsung USB 2.0 PHY adaptation layer
+====================================
+
+1. Description
+--------------
+
+The architecture of the USB 2.0 PHY module in Samsung SoCs is similar
+among many SoCs. In spite of the similarities it proved difficult to
+create a one driver that would fit all these PHY controllers. Often
+the differences were minor and were found in particular bits of the
+registers of the PHY. In some rare cases the order of register writes or
+the PHY powering up process had to be altered. This adaptation layer is
+a compromise between having separate drivers and having a single driver
+with added support for many special cases.
+
+2. Files description
+--------------------
+
+- phy-samsung-usb2.c
+   This is the main file of the adaptation layer. This file contains
+   the probe function and provides two callbacks to the Generic PHY
+   Framework. This two callbacks are used to power on and power off the
+   phy. They carry out the common work that has to be done on all version
+   of the PHY module. Depending on which SoC was chosen they execute SoC
+   specific callbacks. The specific SoC version is selected by choosing
+   the appropriate compatible string. In addition, this file contains
+   struct of_device_id definitions for particular SoCs.
+
+- phy-samsung-usb2.h
+   This is the include file. It declares the structures used by this
+   driver. In addition it should contain extern declarations for
+   structures that describe particular SoCs.
+
+3. Supporting SoCs
+------------------
+
+To support a new SoC a new file should be added to the drivers/phy
+directory. Each SoC's configuration is stored in an instance of the
+struct samsung_usb2_phy_config::
+
+  struct samsung_usb2_phy_config {
+	const struct samsung_usb2_common_phy *phys;
+	int (*rate_to_clk)(unsigned long, u32 *);
+	unsigned int num_phys;
+	bool has_mode_switch;
+  };
+
+The num_phys is the number of phys handled by the driver. `*phys` is an
+array that contains the configuration for each phy. The has_mode_switch
+property is a boolean flag that determines whether the SoC has USB host
+and device on a single pair of pins. If so, a special register has to
+be modified to change the internal routing of these pins between a USB
+device or host module.
+
+For example the configuration for Exynos 4210 is following::
+
+  const struct samsung_usb2_phy_config exynos4210_usb2_phy_config = {
+	.has_mode_switch        = 0,
+	.num_phys		= EXYNOS4210_NUM_PHYS,
+	.phys			= exynos4210_phys,
+	.rate_to_clk		= exynos4210_rate_to_clk,
+  }
+
+- `int (*rate_to_clk)(unsigned long, u32 *)`
+
+	The rate_to_clk callback is to convert the rate of the clock
+	used as the reference clock for the PHY module to the value
+	that should be written in the hardware register.
+
+The exynos4210_phys configuration array is as follows::
+
+  static const struct samsung_usb2_common_phy exynos4210_phys[] = {
+	{
+		.label		= "device",
+		.id		= EXYNOS4210_DEVICE,
+		.power_on	= exynos4210_power_on,
+		.power_off	= exynos4210_power_off,
+	},
+	{
+		.label		= "host",
+		.id		= EXYNOS4210_HOST,
+		.power_on	= exynos4210_power_on,
+		.power_off	= exynos4210_power_off,
+	},
+	{
+		.label		= "hsic0",
+		.id		= EXYNOS4210_HSIC0,
+		.power_on	= exynos4210_power_on,
+		.power_off	= exynos4210_power_off,
+	},
+	{
+		.label		= "hsic1",
+		.id		= EXYNOS4210_HSIC1,
+		.power_on	= exynos4210_power_on,
+		.power_off	= exynos4210_power_off,
+	},
+	{},
+  };
+
+- `int (*power_on)(struct samsung_usb2_phy_instance *);`
+  `int (*power_off)(struct samsung_usb2_phy_instance *);`
+
+	These two callbacks are used to power on and power off the phy
+	by modifying appropriate registers.
+
+Final change to the driver is adding appropriate compatible value to the
+phy-samsung-usb2.c file. In case of Exynos 4210 the following lines were
+added to the struct of_device_id samsung_usb2_phy_of_match[] array::
+
+  #ifdef CONFIG_PHY_EXYNOS4210_USB2
+	{
+		.compatible = "samsung,exynos4210-usb2-phy",
+		.data = &exynos4210_usb2_phy_config,
+	},
+  #endif
+
+To add further flexibility to the driver the Kconfig file enables to
+include support for selected SoCs in the compiled driver. The Kconfig
+entry for Exynos 4210 is following::
+
+  config PHY_EXYNOS4210_USB2
+	bool "Support for Exynos 4210"
+	depends on PHY_SAMSUNG_USB2
+	depends on CPU_EXYNOS4210
+	help
+	  Enable USB PHY support for Exynos 4210. This option requires that
+	  Samsung USB 2.0 PHY driver is enabled and means that support for this
+	  particular SoC is compiled in the driver. In case of Exynos 4210 four
+	  phys are available - device, host, HSCI0 and HSCI1.
+
+The newly created file that supports the new SoC has to be also added to the
+Makefile. In case of Exynos 4210 the added line is following::
+
+  obj-$(CONFIG_PHY_EXYNOS4210_USB2)       += phy-exynos4210-usb2.o
+
+After completing these steps the support for the new SoC should be ready.