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, 190 insertions, 0 deletions

diff --git a/drivers/net/arcnet/arc-rawmode.c b/drivers/net/arcnet/arc-rawmode.c
new file mode 100644
index 000000000..57f172906
--- /dev/null
+++ b/drivers/net/arcnet/arc-rawmode.c
@@ -0,0 +1,190 @@
+/*
+ * Linux ARCnet driver - "raw mode" packet encapsulation (no soft headers)
+ *
+ * Written 1994-1999 by Avery Pennarun.
+ * Derived from skeleton.c by Donald Becker.
+ *
+ * Special thanks to Contemporary Controls, Inc. (www.ccontrols.com)
+ *  for sponsoring the further development of this driver.
+ *
+ * **********************
+ *
+ * The original copyright of skeleton.c was as follows:
+ *
+ * skeleton.c Written 1993 by Donald Becker.
+ * Copyright 1993 United States Government as represented by the
+ * Director, National Security Agency.  This software may only be used
+ * and distributed according to the terms of the GNU General Public License as
+ * modified by SRC, incorporated herein by reference.
+ *
+ * **********************
+ *
+ * For more details, see drivers/net/arcnet.c
+ *
+ * **********************
+ */
+
+#define pr_fmt(fmt) "arcnet:" KBUILD_MODNAME ": " fmt
+
+#include <linux/module.h>
+#include <linux/gfp.h>
+#include <linux/init.h>
+#include <linux/if_arp.h>
+#include <net/arp.h>
+#include <linux/netdevice.h>
+#include <linux/skbuff.h>
+#include "arcdevice.h"
+
+/* packet receiver */
+static void rx(struct net_device *dev, int bufnum,
+	       struct archdr *pkthdr, int length)
+{
+	struct arcnet_local *lp = netdev_priv(dev);
+	struct sk_buff *skb;
+	struct archdr *pkt = pkthdr;
+	int ofs;
+
+	arc_printk(D_DURING, dev, "it's a raw packet (length=%d)\n", length);
+
+	if (length > MTU)
+		ofs = 512 - length;
+	else
+		ofs = 256 - length;
+
+	skb = alloc_skb(length + ARC_HDR_SIZE, GFP_ATOMIC);
+	if (!skb) {
+		dev->stats.rx_dropped++;
+		return;
+	}
+	skb_put(skb, length + ARC_HDR_SIZE);
+	skb->dev = dev;
+
+	pkt = (struct archdr *)skb->data;
+
+	skb_reset_mac_header(skb);
+	skb_pull(skb, ARC_HDR_SIZE);
+
+	/* up to sizeof(pkt->soft) has already been copied from the card */
+	memcpy(pkt, pkthdr, sizeof(struct archdr));
+	if (length > sizeof(pkt->soft))
+		lp->hw.copy_from_card(dev, bufnum, ofs + sizeof(pkt->soft),
+				      pkt->soft.raw + sizeof(pkt->soft),
+				      length - sizeof(pkt->soft));
+
+	if (BUGLVL(D_SKB))
+		arcnet_dump_skb(dev, skb, "rx");
+
+	skb->protocol = cpu_to_be16(ETH_P_ARCNET);
+	netif_rx(skb);
+}
+
+/* Create the ARCnet hard/soft headers for raw mode.
+ * There aren't any soft headers in raw mode - not even the protocol id.
+ */
+static int build_header(struct sk_buff *skb, struct net_device *dev,
+			unsigned short type, uint8_t daddr)
+{
+	int hdr_size = ARC_HDR_SIZE;
+	struct archdr *pkt = skb_push(skb, hdr_size);
+
+	/* Set the source hardware address.
+	 *
+	 * This is pretty pointless for most purposes, but it can help in
+	 * debugging.  ARCnet does not allow us to change the source address
+	 * in the actual packet sent.
+	 */
+	pkt->hard.source = *dev->dev_addr;
+
+	/* see linux/net/ethernet/eth.c to see where I got the following */
+
+	if (dev->flags & (IFF_LOOPBACK | IFF_NOARP)) {
+		/* FIXME: fill in the last byte of the dest ipaddr here
+		 * to better comply with RFC1051 in "noarp" mode.
+		 */
+		pkt->hard.dest = 0;
+		return hdr_size;
+	}
+	/* otherwise, just fill it in and go! */
+	pkt->hard.dest = daddr;
+
+	return hdr_size;	/* success */
+}
+
+static int prepare_tx(struct net_device *dev, struct archdr *pkt, int length,
+		      int bufnum)
+{
+	struct arcnet_local *lp = netdev_priv(dev);
+	struct arc_hardware *hard = &pkt->hard;
+	int ofs;
+
+	arc_printk(D_DURING, dev, "prepare_tx: txbufs=%d/%d/%d\n",
+		   lp->next_tx, lp->cur_tx, bufnum);
+
+	/* hard header is not included in packet length */
+	length -= ARC_HDR_SIZE;
+
+	if (length > XMTU) {
+		/* should never happen! other people already check for this. */
+		arc_printk(D_NORMAL, dev, "Bug!  prepare_tx with size %d (> %d)\n",
+			   length, XMTU);
+		length = XMTU;
+	}
+	if (length >= MinTU) {
+		hard->offset[0] = 0;
+		hard->offset[1] = ofs = 512 - length;
+	} else if (length > MTU) {
+		hard->offset[0] = 0;
+		hard->offset[1] = ofs = 512 - length - 3;
+	} else {
+		hard->offset[0] = ofs = 256 - length;
+	}
+
+	arc_printk(D_DURING, dev, "prepare_tx: length=%d ofs=%d\n",
+		   length, ofs);
+
+	lp->hw.copy_to_card(dev, bufnum, 0, hard, ARC_HDR_SIZE);
+	lp->hw.copy_to_card(dev, bufnum, ofs, &pkt->soft, length);
+
+	lp->lastload_dest = hard->dest;
+
+	return 1;		/* done */
+}
+
+static struct ArcProto rawmode_proto = {
+	.suffix		= 'r',
+	.mtu		= XMTU,
+	.rx		= rx,
+	.build_header	= build_header,
+	.prepare_tx	= prepare_tx,
+	.continue_tx    = NULL,
+	.ack_tx         = NULL
+};
+
+static int __init arcnet_raw_init(void)
+{
+	int count;
+
+	pr_info("raw mode (`r') encapsulation support loaded\n");
+
+	for (count = 0; count < 256; count++)
+		if (arc_proto_map[count] == arc_proto_default)
+			arc_proto_map[count] = &rawmode_proto;
+
+	/* for raw mode, we only set the bcast proto if there's no better one */
+	if (arc_bcast_proto == arc_proto_default)
+		arc_bcast_proto = &rawmode_proto;
+
+	arc_proto_default = &rawmode_proto;
+	return 0;
+}
+
+static void __exit arcnet_raw_exit(void)
+{
+	arcnet_unregister_proto(&rawmode_proto);
+}
+
+module_init(arcnet_raw_init);
+module_exit(arcnet_raw_exit);
+
+MODULE_DESCRIPTION("ARCnet raw mode packet interface module");
+MODULE_LICENSE("GPL");