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, 267 insertions, 0 deletions

diff --git a/drivers/nvmem/brcm_nvram.c b/drivers/nvmem/brcm_nvram.c
new file mode 100644
index 000000000..9f77aee37
--- /dev/null
+++ b/drivers/nvmem/brcm_nvram.c
@@ -0,0 +1,267 @@
+// SPDX-License-Identifier: GPL-2.0-only
+/*
+ * Copyright (C) 2021 Rafał Miłecki <rafal@milecki.pl>
+ */
+
+#include <linux/bcm47xx_nvram.h>
+#include <linux/etherdevice.h>
+#include <linux/hex.h>
+#include <linux/if_ether.h>
+#include <linux/io.h>
+#include <linux/module.h>
+#include <linux/nvmem-consumer.h>
+#include <linux/nvmem-provider.h>
+#include <linux/of.h>
+#include <linux/platform_device.h>
+#include <linux/slab.h>
+
+#define NVRAM_MAGIC			"FLSH"
+
+/**
+ * struct brcm_nvram - driver state internal struct
+ *
+ * @dev:		NVMEM device pointer
+ * @nvmem_size:		Size of the whole space available for NVRAM
+ * @data:		NVRAM data copy stored to avoid poking underlying flash controller
+ * @data_len:		NVRAM data size
+ * @padding_byte:	Padding value used to fill remaining space
+ * @cells:		Array of discovered NVMEM cells
+ * @ncells:		Number of elements in cells
+ */
+struct brcm_nvram {
+	struct device *dev;
+	size_t nvmem_size;
+	uint8_t *data;
+	size_t data_len;
+	uint8_t padding_byte;
+	struct nvmem_cell_info *cells;
+	int ncells;
+};
+
+struct brcm_nvram_header {
+	char magic[4];
+	__le32 len;
+	__le32 crc_ver_init;	/* 0:7 crc, 8:15 ver, 16:31 sdram_init */
+	__le32 config_refresh;	/* 0:15 sdram_config, 16:31 sdram_refresh */
+	__le32 config_ncdl;	/* ncdl values for memc */
+};
+
+static int brcm_nvram_read(void *context, unsigned int offset, void *val,
+			   size_t bytes)
+{
+	struct brcm_nvram *priv = context;
+	size_t to_copy;
+
+	if (offset + bytes > priv->data_len)
+		to_copy = max_t(ssize_t, (ssize_t)priv->data_len - offset, 0);
+	else
+		to_copy = bytes;
+
+	memcpy(val, priv->data + offset, to_copy);
+
+	memset((uint8_t *)val + to_copy, priv->padding_byte, bytes - to_copy);
+
+	return 0;
+}
+
+static int brcm_nvram_copy_data(struct brcm_nvram *priv, struct platform_device *pdev)
+{
+	struct resource *res;
+	void __iomem *base;
+
+	base = devm_platform_get_and_ioremap_resource(pdev, 0, &res);
+	if (IS_ERR(base))
+		return PTR_ERR(base);
+
+	priv->nvmem_size = resource_size(res);
+
+	priv->padding_byte = readb(base + priv->nvmem_size - 1);
+	for (priv->data_len = priv->nvmem_size;
+	     priv->data_len;
+	     priv->data_len--) {
+		if (readb(base + priv->data_len - 1) != priv->padding_byte)
+			break;
+	}
+	WARN(priv->data_len > SZ_128K, "Unexpected (big) NVRAM size: %zu B\n", priv->data_len);
+
+	if (priv->data_len < sizeof(struct brcm_nvram_header)) {
+		dev_err(priv->dev, "NVRAM data too small (%zu)\n", priv->data_len);
+		return -EINVAL;
+	}
+
+	priv->data = devm_kzalloc(priv->dev, priv->data_len, GFP_KERNEL);
+	if (!priv->data)
+		return -ENOMEM;
+
+	memcpy_fromio(priv->data, base, priv->data_len);
+
+	bcm47xx_nvram_init_from_iomem(base, priv->data_len);
+
+	return 0;
+}
+
+static int brcm_nvram_read_post_process_macaddr(void *context, const char *id, int index,
+						unsigned int offset, void *buf, size_t bytes)
+{
+	u8 mac[ETH_ALEN];
+
+	if (bytes != MAC_ADDR_STR_LEN)
+		return -EINVAL;
+
+	if (!mac_pton(buf, mac))
+		return -EINVAL;
+
+	if (index)
+		eth_addr_add(mac, index);
+
+	ether_addr_copy(buf, mac);
+
+	return 0;
+}
+
+static int brcm_nvram_add_cells(struct brcm_nvram *priv, uint8_t *data,
+				size_t len)
+{
+	struct device *dev = priv->dev;
+	char *var, *value;
+	uint8_t tmp;
+	int idx;
+	int err = 0;
+
+	tmp = priv->data[len - 1];
+	priv->data[len - 1] = '\0';
+
+	priv->ncells = 0;
+	for (var = data + sizeof(struct brcm_nvram_header);
+	     var < (char *)data + len && *var;
+	     var += strlen(var) + 1) {
+		priv->ncells++;
+	}
+
+	priv->cells = devm_kcalloc(dev, priv->ncells, sizeof(*priv->cells), GFP_KERNEL);
+	if (!priv->cells) {
+		err = -ENOMEM;
+		goto out;
+	}
+
+	for (var = data + sizeof(struct brcm_nvram_header), idx = 0;
+	     var < (char *)data + len && *var;
+	     var = value + strlen(value) + 1, idx++) {
+		char *eq, *name;
+
+		eq = strchr(var, '=');
+		if (!eq)
+			break;
+		*eq = '\0';
+		name = devm_kstrdup(dev, var, GFP_KERNEL);
+		*eq = '=';
+		if (!name) {
+			err = -ENOMEM;
+			goto out;
+		}
+		value = eq + 1;
+
+		priv->cells[idx].name = name;
+		priv->cells[idx].offset = value - (char *)data;
+		priv->cells[idx].bytes = strlen(value);
+		priv->cells[idx].np = of_get_child_by_name(dev->of_node, priv->cells[idx].name);
+		if (!strcmp(name, "et0macaddr") ||
+		    !strcmp(name, "et1macaddr") ||
+		    !strcmp(name, "et2macaddr")) {
+			priv->cells[idx].raw_len = strlen(value);
+			priv->cells[idx].bytes = ETH_ALEN;
+			priv->cells[idx].read_post_process = brcm_nvram_read_post_process_macaddr;
+		}
+	}
+
+out:
+	priv->data[len - 1] = tmp;
+	return err;
+}
+
+static int brcm_nvram_parse(struct brcm_nvram *priv)
+{
+	struct brcm_nvram_header *header = (struct brcm_nvram_header *)priv->data;
+	struct device *dev = priv->dev;
+	size_t len;
+	int err;
+
+	if (memcmp(header->magic, NVRAM_MAGIC, 4)) {
+		dev_err(dev, "Invalid NVRAM magic\n");
+		return -EINVAL;
+	}
+
+	len = le32_to_cpu(header->len);
+	if (len < sizeof(*header)) {
+		dev_err(dev, "NVRAM length (%zd) too small\n", len);
+		return -EINVAL;
+	}
+	if (len > priv->data_len) {
+		dev_err(dev, "NVRAM length (%zd) exceeds data size (%zd)\n", len,
+			priv->data_len);
+		return -EINVAL;
+	}
+
+	err = brcm_nvram_add_cells(priv, priv->data, len);
+	if (err)
+		dev_err(dev, "Failed to add cells: %d\n", err);
+
+	return 0;
+}
+
+static int brcm_nvram_probe(struct platform_device *pdev)
+{
+	struct nvmem_config config = {
+		.name = "brcm-nvram",
+		.reg_read = brcm_nvram_read,
+	};
+	struct device *dev = &pdev->dev;
+	struct brcm_nvram *priv;
+	int err;
+
+	priv = devm_kzalloc(dev, sizeof(*priv), GFP_KERNEL);
+	if (!priv)
+		return -ENOMEM;
+	priv->dev = dev;
+
+	err = brcm_nvram_copy_data(priv, pdev);
+	if (err)
+		return err;
+
+	err = brcm_nvram_parse(priv);
+	if (err)
+		return err;
+
+	config.dev = dev;
+	config.cells = priv->cells;
+	config.ncells = priv->ncells;
+	config.priv = priv;
+	config.size = priv->nvmem_size;
+
+	return PTR_ERR_OR_ZERO(devm_nvmem_register(dev, &config));
+}
+
+static const struct of_device_id brcm_nvram_of_match_table[] = {
+	{ .compatible = "brcm,nvram", },
+	{},
+};
+
+static struct platform_driver brcm_nvram_driver = {
+	.probe = brcm_nvram_probe,
+	.driver = {
+		.name = "brcm_nvram",
+		.of_match_table = brcm_nvram_of_match_table,
+	},
+};
+
+static int __init brcm_nvram_init(void)
+{
+	return platform_driver_register(&brcm_nvram_driver);
+}
+
+subsys_initcall_sync(brcm_nvram_init);
+
+MODULE_AUTHOR("Rafał Miłecki");
+MODULE_DESCRIPTION("Broadcom I/O-mapped NVRAM support driver");
+MODULE_LICENSE("GPL");
+MODULE_DEVICE_TABLE(of, brcm_nvram_of_match_table);