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, 135 insertions, 0 deletions

diff --git a/include/net/pie.h b/include/net/pie.h
new file mode 100644
index 000000000..1f3db0c35
--- /dev/null
+++ b/include/net/pie.h
@@ -0,0 +1,135 @@
+/* SPDX-License-Identifier: GPL-2.0-only */
+#ifndef __NET_SCHED_PIE_H
+#define __NET_SCHED_PIE_H
+
+#include <linux/ktime.h>
+#include <linux/skbuff.h>
+#include <linux/types.h>
+#include <net/inet_ecn.h>
+#include <net/pkt_sched.h>
+
+#define MAX_PROB	(U64_MAX >> BITS_PER_BYTE)
+#define DTIME_INVALID	U64_MAX
+#define QUEUE_THRESHOLD	16384
+#define DQCOUNT_INVALID	-1
+#define PIE_SCALE	8
+
+/**
+ * struct pie_params - contains pie parameters
+ * @target:		target delay in pschedtime
+ * @tupdate:		interval at which drop probability is calculated
+ * @limit:		total number of packets that can be in the queue
+ * @alpha:		parameter to control drop probability
+ * @beta:		parameter to control drop probability
+ * @ecn:		is ECN marking of packets enabled
+ * @bytemode:		is drop probability scaled based on pkt size
+ * @dq_rate_estimator:	is Little's law used for qdelay calculation
+ */
+struct pie_params {
+	psched_time_t target;
+	u32 tupdate;
+	u32 limit;
+	u32 alpha;
+	u32 beta;
+	u8 ecn;
+	u8 bytemode;
+	u8 dq_rate_estimator;
+};
+
+/**
+ * struct pie_vars - contains pie variables
+ * @qdelay:		current queue delay
+ * @qdelay_old:		queue delay in previous qdelay calculation
+ * @burst_time:		burst time allowance
+ * @dq_tstamp:		timestamp at which dq rate was last calculated
+ * @prob:		drop probability
+ * @accu_prob:		accumulated drop probability
+ * @dq_count:		number of bytes dequeued in a measurement cycle
+ * @avg_dq_rate:	calculated average dq rate
+ * @backlog_old:	queue backlog during previous qdelay calculation
+ */
+struct pie_vars {
+	psched_time_t qdelay;
+	psched_time_t qdelay_old;
+	psched_time_t burst_time;
+	psched_time_t dq_tstamp;
+	u64 prob;
+	u64 accu_prob;
+	u64 dq_count;
+	u32 avg_dq_rate;
+	u32 backlog_old;
+};
+
+/**
+ * struct pie_stats - contains pie stats
+ * @packets_in:	total number of packets enqueued
+ * @dropped:	packets dropped due to pie action
+ * @overlimit:	packets dropped due to lack of space in queue
+ * @ecn_mark:	packets marked with ECN
+ * @maxq:	maximum queue size
+ */
+struct pie_stats {
+	u32 packets_in;
+	u32 dropped;
+	u32 overlimit;
+	u32 ecn_mark;
+	u32 maxq;
+};
+
+/**
+ * struct pie_skb_cb - contains private skb vars
+ * @enqueue_time:	timestamp when the packet is enqueued
+ * @mem_usage:		size of the skb during enqueue
+ */
+struct pie_skb_cb {
+	psched_time_t enqueue_time;
+	u32 mem_usage;
+};
+
+static inline void pie_params_init(struct pie_params *params)
+{
+	params->target = PSCHED_NS2TICKS(15 * NSEC_PER_MSEC);	/* 15 ms */
+	params->tupdate = usecs_to_jiffies(15 * USEC_PER_MSEC);	/* 15 ms */
+	params->limit = 1000;
+	params->alpha = 2;
+	params->beta = 20;
+	params->ecn = false;
+	params->bytemode = false;
+	params->dq_rate_estimator = false;
+}
+
+static inline void pie_vars_init(struct pie_vars *vars)
+{
+	vars->burst_time = PSCHED_NS2TICKS(150 * NSEC_PER_MSEC); /* 150 ms */
+	vars->dq_tstamp = DTIME_INVALID;
+	vars->accu_prob = 0;
+	vars->dq_count = DQCOUNT_INVALID;
+	WRITE_ONCE(vars->avg_dq_rate, 0);
+}
+
+static inline struct pie_skb_cb *get_pie_cb(const struct sk_buff *skb)
+{
+	qdisc_cb_private_validate(skb, sizeof(struct pie_skb_cb));
+	return (struct pie_skb_cb *)qdisc_skb_cb(skb)->data;
+}
+
+static inline psched_time_t pie_get_enqueue_time(const struct sk_buff *skb)
+{
+	return get_pie_cb(skb)->enqueue_time;
+}
+
+static inline void pie_set_enqueue_time(struct sk_buff *skb)
+{
+	get_pie_cb(skb)->enqueue_time = psched_get_time();
+}
+
+bool pie_drop_early(struct Qdisc *sch, struct pie_params *params,
+		    struct pie_vars *vars, u32 backlog, u32 packet_size);
+
+void pie_process_dequeue(struct sk_buff *skb, struct pie_params *params,
+			 struct pie_vars *vars, u32 backlog);
+
+void pie_calculate_probability(struct pie_params *params, struct pie_vars *vars,
+			       u32 backlog);
+
+#endif