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, 154 insertions, 0 deletions

diff --git a/drivers/misc/issei/dma.c b/drivers/misc/issei/dma.c
new file mode 100644
index 000000000..457d28f31
--- /dev/null
+++ b/drivers/misc/issei/dma.c
@@ -0,0 +1,154 @@
+// SPDX-License-Identifier: GPL-2.0
+/* Copyright (C) 2023-2026 Intel Corporation */
+#include <linux/dev_printk.h>
+#include <linux/dma-mapping.h>
+#include <linux/errno.h>
+#include <linux/slab.h>
+#include <linux/string.h>
+#include <linux/timekeeping.h>
+
+#include "issei_dev.h"
+#include "hw_msg.h"
+
+static inline size_t __issei_dma_size(const struct issei_dma *dma)
+{
+	return dma->length.h2f + dma->length.f2h + dma->length.ctl;
+}
+
+/**
+ * issei_dmam_setup - setup DMA buffer and clean it
+ * @idev: issei device object
+ *
+ * Return: 0 on success, <0 on failures
+ */
+int issei_dmam_setup(struct issei_device *idev)
+{
+	struct issei_dma *dma = &idev->dma;
+	size_t size;
+
+	size = __issei_dma_size(dma);
+	if (!size)
+		return -EINVAL;
+
+	if (!dma->vaddr)
+		dma->vaddr = dmam_alloc_coherent(idev->parent, size, &dma->daddr,
+						 GFP_KERNEL | __GFP_ZERO);
+	if (dma->vaddr)
+		memset(dma->vaddr, 0, size);
+	return dma->vaddr ? 0 : -ENOMEM;
+}
+
+static inline struct control_buffer *__dma_get_ctl_buf(struct issei_dma *dma)
+{
+	return dma->vaddr + dma->length.h2f + dma->length.f2h;
+}
+
+static bool __issei_dma_is_read_busy(struct issei_dma *dma)
+{
+	struct control_buffer *ctl = __dma_get_ctl_buf(dma);
+
+	return ctl->f2h_counter_wr != ctl->f2h_counter_rd;
+}
+
+static bool __issei_dma_is_write_busy(struct issei_dma *dma)
+{
+	struct control_buffer *ctl = __dma_get_ctl_buf(dma);
+
+	return ctl->h2f_counter_wr != ctl->h2f_counter_rd;
+}
+
+static void __issei_dma_read_finalize(struct issei_device *idev)
+{
+	struct control_buffer *ctl = __dma_get_ctl_buf(&idev->dma);
+
+	dev_dbg(&idev->dev, "ctl->f2h_counter_rd %u\n", ctl->f2h_counter_rd);
+	/* No need to check overflow - the firmware counters overflow the same way */
+	ctl->f2h_counter_rd++;
+}
+
+static void __issei_dma_write_finalize(struct issei_device *idev)
+{
+	struct control_buffer *ctl = __dma_get_ctl_buf(&idev->dma);
+
+	dev_dbg(&idev->dev, "ctl->h2f_counter_wr %u\n", ctl->h2f_counter_wr);
+	/* No need to check overflow - the firmware counters overflow the same way */
+	ctl->h2f_counter_wr++;
+}
+
+/**
+ * issei_dma_write - write data package to DMA
+ * @idev: issei device object
+ * @data: data atructure
+ *
+ * Return: 0 on success, <0 on failures
+ */
+int issei_dma_write(struct issei_device *idev, const struct issei_dma_data *data)
+{
+	u8 *write_buf = idev->dma.vaddr;
+	struct ham_message_header *hdr = (struct ham_message_header *)write_buf;
+
+	if (data->length > idev->dma.length.h2f - sizeof(*hdr)) {
+		dev_err(&idev->dev, "Message is too big\n");
+		return -EMSGSIZE;
+	}
+
+	if (__issei_dma_is_write_busy(&idev->dma)) {
+		if (ktime_ms_delta(ktime_get(), idev->last_write_ts) > ISSEI_WRITE_TIMEOUT_MSEC) {
+			dev_err(&idev->dev, "Write stuck in queue\n");
+			return -EIO;
+		}
+		dev_info(&idev->dev, "Write is busy\n");
+		return -EBUSY;
+	}
+
+	hdr->length = data->length;
+	hdr->fw_id = data->fw_id;
+	hdr->host_id = data->host_id;
+	hdr->flags = data->flags;
+	hdr->status = data->status;
+	hdr->reserved = 0;
+
+	memcpy(write_buf + sizeof(*hdr), data->buf, data->length);
+
+	__issei_dma_write_finalize(idev);
+	idev->last_write_ts = ktime_get();
+	return 0;
+}
+
+/**
+ * issei_dma_read - read data package from DMA
+ * @idev: issei device object
+ * @data: data atructure
+ *
+ * Return: %0 on success, <0 on failures
+ */
+int issei_dma_read(struct issei_device *idev, struct issei_dma_data *data)
+{
+	u8 *read_buf = idev->dma.vaddr + idev->dma.length.h2f;
+	struct ham_message_header *hdr = (struct ham_message_header *)read_buf;
+
+	if (!__issei_dma_is_read_busy(&idev->dma)) {
+		dev_dbg(&idev->dev, "Nothing to read\n");
+		return -ENODATA;
+	}
+
+	dev_dbg(&idev->dev, "Reading header\n");
+	data->length = hdr->length;
+	data->fw_id = hdr->fw_id;
+	data->host_id = hdr->host_id;
+	data->flags = hdr->flags;
+	data->status = hdr->status;
+
+	if (data->length > idev->dma.length.f2h - sizeof(*hdr)) {
+		dev_err(&idev->dev, "Message length %u is bigger than buffer %zu\n",
+			data->length, idev->dma.length.f2h - sizeof(*hdr));
+		return -EIO;
+	}
+
+	dev_dbg(&idev->dev, "Reading data (size %u)\n", data->length);
+	data->buf = kmemdup(read_buf + sizeof(*hdr), data->length, GFP_KERNEL);
+	if (!data->buf)
+		return -ENOMEM;
+	__issei_dma_read_finalize(idev);
+	return 0;
+}