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, 406 insertions, 0 deletions

diff --git a/drivers/net/usb/zaurus.c b/drivers/net/usb/zaurus.c
new file mode 100644
index 000000000..df3617c4c
--- /dev/null
+++ b/drivers/net/usb/zaurus.c
@@ -0,0 +1,406 @@
+// SPDX-License-Identifier: GPL-2.0-or-later
+/*
+ * Copyright (C) 2002 Pavel Machek <pavel@ucw.cz>
+ * Copyright (C) 2002-2005 by David Brownell
+ */
+
+// #define	DEBUG			// error path messages, extra info
+// #define	VERBOSE			// more; success messages
+
+#include <linux/module.h>
+#include <linux/netdevice.h>
+#include <linux/ethtool.h>
+#include <linux/workqueue.h>
+#include <linux/mii.h>
+#include <linux/crc32.h>
+#include <linux/usb.h>
+#include <linux/usb/cdc.h>
+#include <linux/usb/usbnet.h>
+
+
+/*
+ * All known Zaurii lie about their standards conformance.  At least
+ * the earliest SA-1100 models lie by saying they support CDC Ethernet.
+ * Some later models (especially PXA-25x and PXA-27x based ones) lie
+ * and say they support CDC MDLM (for access to cell phone modems).
+ *
+ * There are non-Zaurus products that use these same protocols too.
+ *
+ * The annoying thing is that at the same time Sharp was developing
+ * that annoying standards-breaking software, the Linux community had
+ * a simple "CDC Subset" working reliably on the same SA-1100 hardware.
+ * That is, the same functionality but not violating standards.
+ *
+ * The CDC Ethernet nonconformance points are troublesome to hosts
+ * with a true CDC Ethernet implementation:
+ *   - Framing appends a CRC, which the spec says drivers "must not" do;
+ *   - Transfers data in altsetting zero, instead of altsetting 1;
+ *   - All these peripherals use the same ethernet address.
+ *
+ * The CDC MDLM nonconformance is less immediately troublesome, since all
+ * MDLM implementations are quasi-proprietary anyway.
+ */
+
+static struct sk_buff *
+zaurus_tx_fixup(struct usbnet *dev, struct sk_buff *skb, gfp_t flags)
+{
+	int			padlen;
+	struct sk_buff		*skb2;
+
+	padlen = 2;
+	if (!skb_cloned(skb)) {
+		int	tailroom = skb_tailroom(skb);
+		if ((padlen + 4) <= tailroom)
+			goto done;
+	}
+	skb2 = skb_copy_expand(skb, 0, 4 + padlen, flags);
+	dev_kfree_skb_any(skb);
+	skb = skb2;
+	if (skb) {
+		u32		fcs;
+done:
+		fcs = crc32_le(~0, skb->data, skb->len);
+		fcs = ~fcs;
+
+		skb_put_u8(skb, fcs & 0xff);
+		skb_put_u8(skb, (fcs >> 8) & 0xff);
+		skb_put_u8(skb, (fcs >> 16) & 0xff);
+		skb_put_u8(skb, (fcs >> 24) & 0xff);
+	}
+	return skb;
+}
+
+static int zaurus_bind(struct usbnet *dev, struct usb_interface *intf)
+{
+	/* Belcarra's funky framing has other options; mostly
+	 * TRAILERS (!) with 4 bytes CRC, and maybe 2 pad bytes.
+	 */
+	dev->net->hard_header_len += 6;
+	dev->rx_urb_size = dev->net->hard_header_len + dev->net->mtu;
+	return usbnet_generic_cdc_bind(dev, intf);
+}
+
+/* PDA style devices are always connected if present */
+static int always_connected (struct usbnet *dev)
+{
+	return 0;
+}
+
+static const struct driver_info	zaurus_sl5x00_info = {
+	.description =	"Sharp Zaurus SL-5x00",
+	.flags =	FLAG_POINTTOPOINT | FLAG_FRAMING_Z,
+	.check_connect = always_connected,
+	.bind =		zaurus_bind,
+	.unbind =	usbnet_cdc_unbind,
+	.tx_fixup =	zaurus_tx_fixup,
+};
+#define	ZAURUS_STRONGARM_INFO	((unsigned long)&zaurus_sl5x00_info)
+
+static const struct driver_info	zaurus_pxa_info = {
+	.description =	"Sharp Zaurus, PXA-2xx based",
+	.flags =	FLAG_POINTTOPOINT | FLAG_FRAMING_Z,
+	.check_connect = always_connected,
+	.bind =		zaurus_bind,
+	.unbind =	usbnet_cdc_unbind,
+	.tx_fixup =	zaurus_tx_fixup,
+};
+#define	ZAURUS_PXA_INFO		((unsigned long)&zaurus_pxa_info)
+
+static const struct driver_info	olympus_mxl_info = {
+	.description =	"Olympus R1000",
+	.flags =	FLAG_POINTTOPOINT | FLAG_FRAMING_Z,
+	.check_connect = always_connected,
+	.bind =		zaurus_bind,
+	.unbind =	usbnet_cdc_unbind,
+	.tx_fixup =	zaurus_tx_fixup,
+};
+#define	OLYMPUS_MXL_INFO	((unsigned long)&olympus_mxl_info)
+
+
+/* Some more recent products using Lineo/Belcarra code will wrongly claim
+ * CDC MDLM conformance.  They aren't conformant:  data endpoints live
+ * in the control interface, there's no data interface, and it's not used
+ * to talk to a cell phone radio.  But at least we can detect these two
+ * pseudo-classes, rather than growing this product list with entries for
+ * each new nonconformant product (sigh).
+ */
+static const u8 safe_guid[16] = {
+	0x5d, 0x34, 0xcf, 0x66, 0x11, 0x18, 0x11, 0xd6,
+	0xa2, 0x1a, 0x00, 0x01, 0x02, 0xca, 0x9a, 0x7f,
+};
+static const u8 blan_guid[16] = {
+	0x74, 0xf0, 0x3d, 0xbd, 0x1e, 0xc1, 0x44, 0x70,
+	0xa3, 0x67, 0x71, 0x34, 0xc9, 0xf5, 0x54, 0x37,
+};
+
+static int blan_mdlm_bind(struct usbnet *dev, struct usb_interface *intf)
+{
+	u8				*buf = intf->cur_altsetting->extra;
+	int				len = intf->cur_altsetting->extralen;
+	struct usb_cdc_mdlm_desc	*desc = NULL;
+	struct usb_cdc_mdlm_detail_desc	*detail = NULL;
+
+	while (len > 3) {
+		if (buf [1] != USB_DT_CS_INTERFACE)
+			goto next_desc;
+
+		/* use bDescriptorSubType, and just verify that we get a
+		 * "BLAN" (or "SAFE") descriptor.
+		 */
+		switch (buf [2]) {
+		case USB_CDC_MDLM_TYPE:
+			if (desc) {
+				dev_dbg(&intf->dev, "extra MDLM\n");
+				goto bad_desc;
+			}
+			desc = (void *) buf;
+			if (desc->bLength != sizeof *desc) {
+				dev_dbg(&intf->dev, "MDLM len %u\n",
+					desc->bLength);
+				goto bad_desc;
+			}
+			/* expect bcdVersion 1.0, ignore */
+			if (memcmp(&desc->bGUID, blan_guid, 16) &&
+			    memcmp(&desc->bGUID, safe_guid, 16)) {
+				/* hey, this one might _really_ be MDLM! */
+				dev_dbg(&intf->dev, "MDLM guid\n");
+				goto bad_desc;
+			}
+			break;
+		case USB_CDC_MDLM_DETAIL_TYPE:
+			if (detail) {
+				dev_dbg(&intf->dev, "extra MDLM detail\n");
+				goto bad_desc;
+			}
+			detail = (void *) buf;
+			switch (detail->bGuidDescriptorType) {
+			case 0:			/* "SAFE" */
+				if (detail->bLength != (sizeof *detail + 2))
+					goto bad_detail;
+				break;
+			case 1:			/* "BLAN" */
+				if (detail->bLength != (sizeof *detail + 3))
+					goto bad_detail;
+				break;
+			default:
+				goto bad_detail;
+			}
+
+			/* assuming we either noticed BLAN already, or will
+			 * find it soon, there are some data bytes here:
+			 *  - bmNetworkCapabilities (unused)
+			 *  - bmDataCapabilities (bits, see below)
+			 *  - bPad (ignored, for PADAFTER -- BLAN-only)
+			 * bits are:
+			 *  - 0x01 -- Zaurus framing (add CRC)
+			 *  - 0x02 -- PADBEFORE (CRC includes some padding)
+			 *  - 0x04 -- PADAFTER (some padding after CRC)
+			 *  - 0x08 -- "fermat" packet mangling (for hw bugs)
+			 * the PADBEFORE appears not to matter; we interop
+			 * with devices that use it and those that don't.
+			 */
+			if ((detail->bDetailData[1] & ~0x02) != 0x01) {
+				/* bmDataCapabilities == 0 would be fine too,
+				 * but framing is minidriver-coupled for now.
+				 */
+bad_detail:
+				dev_dbg(&intf->dev,
+						"bad MDLM detail, %d %d %d\n",
+						detail->bLength,
+						detail->bDetailData[0],
+						detail->bDetailData[2]);
+				goto bad_desc;
+			}
+
+			/* same extra framing as for non-BLAN mode */
+			dev->net->hard_header_len += 6;
+			dev->rx_urb_size = dev->net->hard_header_len
+					+ dev->net->mtu;
+			break;
+		}
+next_desc:
+		len -= buf [0];	/* bLength */
+		buf += buf [0];
+	}
+
+	if (!desc || !detail) {
+		dev_dbg(&intf->dev, "missing cdc mdlm %s%sdescriptor\n",
+			desc ? "" : "func ",
+			detail ? "" : "detail ");
+		goto bad_desc;
+	}
+
+	/* There's probably a CDC Ethernet descriptor there, but we can't
+	 * rely on the Ethernet address it provides since not all vendors
+	 * bother to make it unique.  Likewise there's no point in tracking
+	 * of the CDC event notifications.
+	 */
+	return usbnet_get_endpoints(dev, intf);
+
+bad_desc:
+	dev_info(&dev->udev->dev, "unsupported MDLM descriptors\n");
+	return -ENODEV;
+}
+
+static const struct driver_info	bogus_mdlm_info = {
+	.description =	"pseudo-MDLM (BLAN) device",
+	.flags =	FLAG_POINTTOPOINT | FLAG_FRAMING_Z,
+	.check_connect = always_connected,
+	.tx_fixup =	zaurus_tx_fixup,
+	.bind =		blan_mdlm_bind,
+};
+
+static const struct usb_device_id	products [] = {
+#define	ZAURUS_MASTER_INTERFACE \
+	.bInterfaceClass	= USB_CLASS_COMM, \
+	.bInterfaceSubClass	= USB_CDC_SUBCLASS_ETHERNET, \
+	.bInterfaceProtocol	= USB_CDC_PROTO_NONE
+
+#define ZAURUS_FAKE_INTERFACE \
+	.bInterfaceClass	= USB_CLASS_COMM, \
+	.bInterfaceSubClass	= USB_CDC_SUBCLASS_MDLM, \
+	.bInterfaceProtocol	= USB_CDC_PROTO_NONE
+
+/* SA-1100 based Sharp Zaurus ("collie"), or compatible. */
+{
+	.match_flags	=   USB_DEVICE_ID_MATCH_INT_INFO
+			  | USB_DEVICE_ID_MATCH_DEVICE,
+	.idVendor		= 0x04DD,
+	.idProduct		= 0x8004,
+	ZAURUS_MASTER_INTERFACE,
+	.driver_info = ZAURUS_STRONGARM_INFO,
+},
+
+/* PXA-2xx based models are also lying-about-cdc.  If you add any
+ * more devices that claim to be CDC Ethernet, make sure they get
+ * added to the blacklist in cdc_ether too.
+ *
+ * NOTE:  OpenZaurus versions with 2.6 kernels won't use these entries,
+ * unlike the older ones with 2.4 "embedix" kernels.
+ */
+{
+	.match_flags	=   USB_DEVICE_ID_MATCH_INT_INFO
+			  | USB_DEVICE_ID_MATCH_DEVICE,
+	.idVendor		= 0x04DD,
+	.idProduct		= 0x8005,	/* A-300 */
+	ZAURUS_MASTER_INTERFACE,
+	.driver_info = ZAURUS_PXA_INFO,
+}, {
+	.match_flags	=   USB_DEVICE_ID_MATCH_INT_INFO
+			  | USB_DEVICE_ID_MATCH_DEVICE,
+	.idVendor		= 0x04DD,
+	.idProduct		= 0x8005,	/* A-300 */
+	ZAURUS_FAKE_INTERFACE,
+	.driver_info = (unsigned long)&bogus_mdlm_info,
+}, {
+	.match_flags    =   USB_DEVICE_ID_MATCH_INT_INFO
+			  | USB_DEVICE_ID_MATCH_DEVICE,
+	.idVendor		= 0x04DD,
+	.idProduct		= 0x8006,	/* B-500/SL-5600 */
+	ZAURUS_MASTER_INTERFACE,
+	.driver_info = ZAURUS_PXA_INFO,
+}, {
+	.match_flags    =   USB_DEVICE_ID_MATCH_INT_INFO
+			  | USB_DEVICE_ID_MATCH_DEVICE,
+	.idVendor		= 0x04DD,
+	.idProduct		= 0x8006,	/* B-500/SL-5600 */
+	ZAURUS_FAKE_INTERFACE,
+	.driver_info = (unsigned long)&bogus_mdlm_info,
+}, {
+	.match_flags    =   USB_DEVICE_ID_MATCH_INT_INFO
+	          | USB_DEVICE_ID_MATCH_DEVICE,
+	.idVendor		= 0x04DD,
+	.idProduct		= 0x8007,	/* C-700 */
+	ZAURUS_MASTER_INTERFACE,
+	.driver_info = ZAURUS_PXA_INFO,
+}, {
+	.match_flags    =   USB_DEVICE_ID_MATCH_INT_INFO
+			  | USB_DEVICE_ID_MATCH_DEVICE,
+	.idVendor		= 0x04DD,
+	.idProduct		= 0x8007,	/* C-700 */
+	ZAURUS_FAKE_INTERFACE,
+	.driver_info = (unsigned long)&bogus_mdlm_info,
+}, {
+	.match_flags    =   USB_DEVICE_ID_MATCH_INT_INFO
+		 | USB_DEVICE_ID_MATCH_DEVICE,
+	.idVendor               = 0x04DD,
+	.idProduct              = 0x9031,	/* C-750 C-760 */
+	ZAURUS_MASTER_INTERFACE,
+	.driver_info = ZAURUS_PXA_INFO,
+}, {
+	/* C-750/C-760/C-860/SL-C3000 PDA in MDLM mode */
+	USB_DEVICE_AND_INTERFACE_INFO(0x04DD, 0x9031, USB_CLASS_COMM,
+			USB_CDC_SUBCLASS_MDLM, USB_CDC_PROTO_NONE),
+	.driver_info = (unsigned long) &bogus_mdlm_info,
+}, {
+	.match_flags    =   USB_DEVICE_ID_MATCH_INT_INFO
+		 | USB_DEVICE_ID_MATCH_DEVICE,
+	.idVendor               = 0x04DD,
+	.idProduct              = 0x9032,	/* SL-6000 */
+	ZAURUS_MASTER_INTERFACE,
+	.driver_info = ZAURUS_PXA_INFO,
+}, {
+	.match_flags    =   USB_DEVICE_ID_MATCH_INT_INFO
+			    | USB_DEVICE_ID_MATCH_DEVICE,
+	.idVendor		= 0x04DD,
+	.idProduct		= 0x9032,	/* SL-6000 */
+	ZAURUS_FAKE_INTERFACE,
+	.driver_info = (unsigned long)&bogus_mdlm_info,
+}, {
+	.match_flags    =   USB_DEVICE_ID_MATCH_INT_INFO
+		 | USB_DEVICE_ID_MATCH_DEVICE,
+	.idVendor               = 0x04DD,
+	/* reported with some C860 units */
+	.idProduct              = 0x9050,	/* C-860 */
+	ZAURUS_MASTER_INTERFACE,
+	.driver_info = ZAURUS_PXA_INFO,
+},
+{
+	/* Motorola Rokr E6 */
+	USB_DEVICE_AND_INTERFACE_INFO(0x22b8, 0x6027, USB_CLASS_COMM,
+			USB_CDC_SUBCLASS_MDLM, USB_CDC_PROTO_NONE),
+	.driver_info = (unsigned long) &bogus_mdlm_info,
+}, {
+	/* Motorola MOTOMAGX phones */
+	USB_DEVICE_AND_INTERFACE_INFO(0x22b8, 0x6425, USB_CLASS_COMM,
+			USB_CDC_SUBCLASS_MDLM, USB_CDC_PROTO_NONE),
+	.driver_info = (unsigned long) &bogus_mdlm_info,
+},
+
+/* Olympus has some models with a Zaurus-compatible option.
+ * R-1000 uses a FreeScale i.MXL cpu (ARMv4T)
+ */
+{
+	.match_flags    =   USB_DEVICE_ID_MATCH_INT_INFO
+		 | USB_DEVICE_ID_MATCH_DEVICE,
+	.idVendor               = 0x07B4,
+	.idProduct              = 0x0F02,	/* R-1000 */
+	ZAURUS_MASTER_INTERFACE,
+	.driver_info = OLYMPUS_MXL_INFO,
+},
+
+/* Logitech Harmony 900 - uses the pseudo-MDLM (BLAN) driver */
+{
+	USB_DEVICE_AND_INTERFACE_INFO(0x046d, 0xc11f, USB_CLASS_COMM,
+			USB_CDC_SUBCLASS_MDLM, USB_CDC_PROTO_NONE),
+	.driver_info = (unsigned long) &bogus_mdlm_info,
+},
+	{ },		// END
+};
+MODULE_DEVICE_TABLE(usb, products);
+
+static struct usb_driver zaurus_driver = {
+	.name =		"zaurus",
+	.id_table =	products,
+	.probe =	usbnet_probe,
+	.disconnect =	usbnet_disconnect,
+	.suspend =	usbnet_suspend,
+	.resume =	usbnet_resume,
+	.disable_hub_initiated_lpm = 1,
+};
+
+module_usb_driver(zaurus_driver);
+
+MODULE_AUTHOR("Pavel Machek, David Brownell");
+MODULE_DESCRIPTION("Sharp Zaurus PDA, and compatible products");
+MODULE_LICENSE("GPL");