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, 273 insertions, 0 deletions

diff --git a/drivers/misc/mei/dma-ring.c b/drivers/misc/mei/dma-ring.c
new file mode 100644
index 000000000..6277c4a5b
--- /dev/null
+++ b/drivers/misc/mei/dma-ring.c
@@ -0,0 +1,273 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * Copyright(c) 2016-2018 Intel Corporation. All rights reserved.
+ */
+#include <linux/dma-mapping.h>
+#include <linux/mei.h>
+
+#include "mei_dev.h"
+
+/**
+ * mei_dmam_dscr_alloc() - allocate a managed coherent buffer
+ *     for the dma descriptor
+ * @dev: mei_device
+ * @dscr: dma descriptor
+ *
+ * Return:
+ * * 0       - on success or zero allocation request
+ * * -EINVAL - if size is not power of 2
+ * * -ENOMEM - of allocation has failed
+ */
+static int mei_dmam_dscr_alloc(struct mei_device *dev,
+			       struct mei_dma_dscr *dscr)
+{
+	if (!dscr->size)
+		return 0;
+
+	if (WARN_ON(!is_power_of_2(dscr->size)))
+		return -EINVAL;
+
+	if (dscr->vaddr)
+		return 0;
+
+	dscr->vaddr = dmam_alloc_coherent(dev->parent, dscr->size, &dscr->daddr,
+					  GFP_KERNEL);
+	if (!dscr->vaddr)
+		return -ENOMEM;
+
+	return 0;
+}
+
+/**
+ * mei_dmam_dscr_free() - free a managed coherent buffer
+ *     from the dma descriptor
+ * @dev: mei_device
+ * @dscr: dma descriptor
+ */
+static void mei_dmam_dscr_free(struct mei_device *dev,
+			       struct mei_dma_dscr *dscr)
+{
+	if (!dscr->vaddr)
+		return;
+
+	dmam_free_coherent(dev->parent, dscr->size, dscr->vaddr, dscr->daddr);
+	dscr->vaddr = NULL;
+}
+
+/**
+ * mei_dmam_ring_free() - free dma ring buffers
+ * @dev: mei device
+ */
+void mei_dmam_ring_free(struct mei_device *dev)
+{
+	int i;
+
+	for (i = 0; i < DMA_DSCR_NUM; i++)
+		mei_dmam_dscr_free(dev, &dev->dr_dscr[i]);
+}
+
+/**
+ * mei_dmam_ring_alloc() - allocate dma ring buffers
+ * @dev: mei device
+ *
+ * Return: -ENOMEM on allocation failure 0 otherwise
+ */
+int mei_dmam_ring_alloc(struct mei_device *dev)
+{
+	int i;
+
+	for (i = 0; i < DMA_DSCR_NUM; i++)
+		if (mei_dmam_dscr_alloc(dev, &dev->dr_dscr[i]))
+			goto err;
+
+	return 0;
+
+err:
+	mei_dmam_ring_free(dev);
+	return -ENOMEM;
+}
+
+/**
+ * mei_dma_ring_is_allocated() - check if dma ring is allocated
+ * @dev: mei device
+ *
+ * Return: true if dma ring is allocated
+ */
+bool mei_dma_ring_is_allocated(struct mei_device *dev)
+{
+	return !!dev->dr_dscr[DMA_DSCR_HOST].vaddr;
+}
+
+static inline
+struct hbm_dma_ring_ctrl *mei_dma_ring_ctrl(struct mei_device *dev)
+{
+	return (struct hbm_dma_ring_ctrl *)dev->dr_dscr[DMA_DSCR_CTRL].vaddr;
+}
+
+/**
+ * mei_dma_ring_reset() - reset the dma control block
+ * @dev: mei device
+ */
+void mei_dma_ring_reset(struct mei_device *dev)
+{
+	struct hbm_dma_ring_ctrl *ctrl = mei_dma_ring_ctrl(dev);
+
+	if (!ctrl)
+		return;
+
+	memset(ctrl, 0, sizeof(*ctrl));
+}
+
+/**
+ * mei_dma_copy_from() - copy from dma ring into buffer
+ * @dev: mei device
+ * @buf: data buffer
+ * @offset: offset in slots.
+ * @n: number of slots to copy.
+ *
+ * Return: number of bytes copied
+ */
+static size_t mei_dma_copy_from(struct mei_device *dev, unsigned char *buf,
+				u32 offset, u32 n)
+{
+	unsigned char *dbuf = dev->dr_dscr[DMA_DSCR_DEVICE].vaddr;
+
+	size_t b_offset = offset << 2;
+	size_t b_n = n << 2;
+
+	memcpy(buf, dbuf + b_offset, b_n);
+
+	return b_n;
+}
+
+/**
+ * mei_dma_copy_to() - copy to a buffer to the dma ring
+ * @dev: mei device
+ * @buf: data buffer
+ * @offset: offset in slots.
+ * @n: number of slots to copy.
+ *
+ * Return: number of bytes copied
+ */
+static size_t mei_dma_copy_to(struct mei_device *dev, unsigned char *buf,
+			      u32 offset, u32 n)
+{
+	unsigned char *hbuf = dev->dr_dscr[DMA_DSCR_HOST].vaddr;
+
+	size_t b_offset = offset << 2;
+	size_t b_n = n << 2;
+
+	memcpy(hbuf + b_offset, buf, b_n);
+
+	return b_n;
+}
+
+/**
+ * mei_dma_ring_read() - read data from the ring
+ * @dev: mei device
+ * @buf: buffer to read into: may be NULL in case of dropping the data.
+ * @len: length to read.
+ */
+void mei_dma_ring_read(struct mei_device *dev, unsigned char *buf, u32 len)
+{
+	struct hbm_dma_ring_ctrl *ctrl = mei_dma_ring_ctrl(dev);
+	u32 dbuf_depth;
+	u32 rd_idx, rem, slots;
+
+	if (WARN_ON(!ctrl))
+		return;
+
+	dev_dbg(&dev->dev, "reading from dma %u bytes\n", len);
+
+	if (!len)
+		return;
+
+	dbuf_depth = dev->dr_dscr[DMA_DSCR_DEVICE].size >> 2;
+	rd_idx = READ_ONCE(ctrl->dbuf_rd_idx) & (dbuf_depth - 1);
+	slots = mei_data2slots(len);
+
+	/* if buf is NULL we drop the packet by advancing the pointer.*/
+	if (!buf)
+		goto out;
+
+	if (rd_idx + slots > dbuf_depth) {
+		buf += mei_dma_copy_from(dev, buf, rd_idx, dbuf_depth - rd_idx);
+		rem = slots - (dbuf_depth - rd_idx);
+		rd_idx = 0;
+	} else {
+		rem = slots;
+	}
+
+	mei_dma_copy_from(dev, buf, rd_idx, rem);
+out:
+	WRITE_ONCE(ctrl->dbuf_rd_idx, ctrl->dbuf_rd_idx + slots);
+}
+
+static inline u32 mei_dma_ring_hbuf_depth(struct mei_device *dev)
+{
+	return dev->dr_dscr[DMA_DSCR_HOST].size >> 2;
+}
+
+/**
+ * mei_dma_ring_empty_slots() - calaculate number of empty slots in dma ring
+ * @dev: mei_device
+ *
+ * Return: number of empty slots
+ */
+u32 mei_dma_ring_empty_slots(struct mei_device *dev)
+{
+	struct hbm_dma_ring_ctrl *ctrl = mei_dma_ring_ctrl(dev);
+	u32 wr_idx, rd_idx, hbuf_depth, empty;
+
+	if (!mei_dma_ring_is_allocated(dev))
+		return 0;
+
+	if (WARN_ON(!ctrl))
+		return 0;
+
+	/* easier to work in slots */
+	hbuf_depth = mei_dma_ring_hbuf_depth(dev);
+	rd_idx = READ_ONCE(ctrl->hbuf_rd_idx);
+	wr_idx = READ_ONCE(ctrl->hbuf_wr_idx);
+
+	if (rd_idx > wr_idx)
+		empty = rd_idx - wr_idx;
+	else
+		empty = hbuf_depth - (wr_idx - rd_idx);
+
+	return empty;
+}
+
+/**
+ * mei_dma_ring_write - write data to dma ring host buffer
+ *
+ * @dev: mei_device
+ * @buf: data will be written
+ * @len: data length
+ */
+void mei_dma_ring_write(struct mei_device *dev, unsigned char *buf, u32 len)
+{
+	struct hbm_dma_ring_ctrl *ctrl = mei_dma_ring_ctrl(dev);
+	u32 hbuf_depth;
+	u32 wr_idx, rem, slots;
+
+	if (WARN_ON(!ctrl))
+		return;
+
+	dev_dbg(&dev->dev, "writing to dma %u bytes\n", len);
+	hbuf_depth = mei_dma_ring_hbuf_depth(dev);
+	wr_idx = READ_ONCE(ctrl->hbuf_wr_idx) & (hbuf_depth - 1);
+	slots = mei_data2slots(len);
+
+	if (wr_idx + slots > hbuf_depth) {
+		buf += mei_dma_copy_to(dev, buf, wr_idx, hbuf_depth - wr_idx);
+		rem = slots - (hbuf_depth - wr_idx);
+		wr_idx = 0;
+	} else {
+		rem = slots;
+	}
+
+	mei_dma_copy_to(dev, buf, wr_idx, rem);
+
+	WRITE_ONCE(ctrl->hbuf_wr_idx, ctrl->hbuf_wr_idx + slots);
+}