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, 324 insertions, 0 deletions

diff --git a/net/sched/sch_codel.c b/net/sched/sch_codel.c
new file mode 100644
index 000000000..6aa5829d6
--- /dev/null
+++ b/net/sched/sch_codel.c
@@ -0,0 +1,324 @@
+// SPDX-License-Identifier: GPL-2.0 OR BSD-3-Clause
+/*
+ * Codel - The Controlled-Delay Active Queue Management algorithm
+ *
+ *  Copyright (C) 2011-2012 Kathleen Nichols <nichols@pollere.com>
+ *  Copyright (C) 2011-2012 Van Jacobson <van@pollere.net>
+ *
+ *  Implemented on linux by :
+ *  Copyright (C) 2012 Michael D. Taht <dave.taht@bufferbloat.net>
+ *  Copyright (C) 2012,2015 Eric Dumazet <edumazet@google.com>
+ */
+
+#include <linux/module.h>
+#include <linux/slab.h>
+#include <linux/types.h>
+#include <linux/kernel.h>
+#include <linux/errno.h>
+#include <linux/skbuff.h>
+#include <linux/prefetch.h>
+#include <net/pkt_sched.h>
+#include <net/codel.h>
+#include <net/codel_impl.h>
+#include <net/codel_qdisc.h>
+
+
+#define DEFAULT_CODEL_LIMIT 1000
+
+struct codel_sched_data {
+	struct codel_params	params;
+	struct codel_vars	vars;
+	struct codel_stats	stats;
+	u32			drop_overlimit;
+};
+
+/* This is the specific function called from codel_dequeue()
+ * to dequeue a packet from queue. Note: backlog is handled in
+ * codel, we dont need to reduce it here.
+ */
+static struct sk_buff *dequeue_func(struct codel_vars *vars, void *ctx)
+{
+	struct Qdisc *sch = ctx;
+	struct sk_buff *skb = __qdisc_dequeue_head(&sch->q);
+
+	if (skb) {
+		qstats_backlog_sub(sch, qdisc_pkt_len(skb));
+		prefetch(&skb->end); /* we'll need skb_shinfo() */
+	}
+	return skb;
+}
+
+static void drop_func(struct sk_buff *skb, void *ctx)
+{
+	struct Qdisc *sch = ctx;
+
+	qdisc_dequeue_drop(sch, skb, QDISC_DROP_CONGESTED);
+	qdisc_qstats_drop(sch);
+}
+
+static struct sk_buff *__codel_qdisc_dequeue(struct Qdisc *sch)
+{
+	struct codel_sched_data *q = qdisc_priv(sch);
+	struct sk_buff *skb;
+
+	skb = codel_dequeue(sch, &sch->qstats.backlog, &q->params, &q->vars,
+			    &q->stats, qdisc_pkt_len, codel_get_enqueue_time,
+			    drop_func, dequeue_func);
+
+	if (skb)
+		qdisc_bstats_update(sch, skb);
+	return skb;
+}
+
+static void codel_dequeue_drop(struct Qdisc *sch)
+{
+	struct codel_sched_data *q = qdisc_priv(sch);
+
+	if (q->stats.drop_count) {
+		qdisc_tree_reduce_backlog(sch, q->stats.drop_count,
+					  q->stats.drop_len);
+		q->stats.drop_count = 0;
+		q->stats.drop_len = 0;
+	}
+}
+
+static struct sk_buff *codel_qdisc_dequeue(struct Qdisc *sch)
+{
+	struct sk_buff *skb;
+
+	skb = __codel_qdisc_dequeue(sch);
+
+	codel_dequeue_drop(sch);
+
+	return skb;
+}
+
+static struct sk_buff *codel_peek(struct Qdisc *sch)
+{
+	struct sk_buff *skb = skb_peek(&sch->gso_skb);
+
+	if (!skb) {
+		skb = __codel_qdisc_dequeue(sch);
+
+		if (skb) {
+			__skb_queue_head(&sch->gso_skb, skb);
+			/* it's still part of the queue */
+			qdisc_qstats_backlog_inc(sch, skb);
+			sch->q.qlen++;
+		}
+
+		codel_dequeue_drop(sch);
+	}
+
+	return skb;
+}
+
+static int codel_qdisc_enqueue(struct sk_buff *skb, struct Qdisc *sch,
+			       struct sk_buff **to_free)
+{
+	struct codel_sched_data *q;
+
+	if (likely(qdisc_qlen(sch) < sch->limit)) {
+		codel_set_enqueue_time(skb);
+		return qdisc_enqueue_tail(skb, sch);
+	}
+	q = qdisc_priv(sch);
+	WRITE_ONCE(q->drop_overlimit, q->drop_overlimit + 1);
+	return qdisc_drop_reason(skb, sch, to_free, QDISC_DROP_OVERLIMIT);
+}
+
+static const struct nla_policy codel_policy[TCA_CODEL_MAX + 1] = {
+	[TCA_CODEL_TARGET]	= { .type = NLA_U32 },
+	[TCA_CODEL_LIMIT]	= { .type = NLA_U32 },
+	[TCA_CODEL_INTERVAL]	= { .type = NLA_U32 },
+	[TCA_CODEL_ECN]		= { .type = NLA_U32 },
+	[TCA_CODEL_CE_THRESHOLD]= { .type = NLA_U32 },
+};
+
+static int codel_change(struct Qdisc *sch, struct nlattr *opt,
+			struct netlink_ext_ack *extack)
+{
+	unsigned int dropped_pkts = 0, dropped_bytes = 0;
+	struct codel_sched_data *q = qdisc_priv(sch);
+	struct nlattr *tb[TCA_CODEL_MAX + 1];
+	int err;
+
+	err = nla_parse_nested_deprecated(tb, TCA_CODEL_MAX, opt,
+					  codel_policy, NULL);
+	if (err < 0)
+		return err;
+
+	sch_tree_lock(sch);
+
+	if (tb[TCA_CODEL_TARGET]) {
+		u32 target = nla_get_u32(tb[TCA_CODEL_TARGET]);
+
+		WRITE_ONCE(q->params.target,
+			   ((u64)target * NSEC_PER_USEC) >> CODEL_SHIFT);
+	}
+
+	if (tb[TCA_CODEL_CE_THRESHOLD]) {
+		u64 val = nla_get_u32(tb[TCA_CODEL_CE_THRESHOLD]);
+
+		WRITE_ONCE(q->params.ce_threshold,
+			   (val * NSEC_PER_USEC) >> CODEL_SHIFT);
+	}
+
+	if (tb[TCA_CODEL_INTERVAL]) {
+		u32 interval = nla_get_u32(tb[TCA_CODEL_INTERVAL]);
+
+		WRITE_ONCE(q->params.interval,
+			   ((u64)interval * NSEC_PER_USEC) >> CODEL_SHIFT);
+	}
+
+	if (tb[TCA_CODEL_LIMIT])
+		WRITE_ONCE(sch->limit,
+			   nla_get_u32(tb[TCA_CODEL_LIMIT]));
+
+	if (tb[TCA_CODEL_ECN])
+		WRITE_ONCE(q->params.ecn,
+			   !!nla_get_u32(tb[TCA_CODEL_ECN]));
+
+	while (sch->q.qlen > sch->limit) {
+		struct sk_buff *skb = qdisc_dequeue_internal(sch, true);
+
+		if (!skb)
+			break;
+
+		dropped_pkts++;
+		dropped_bytes += qdisc_pkt_len(skb);
+		rtnl_qdisc_drop(skb, sch);
+	}
+	qdisc_tree_reduce_backlog(sch, dropped_pkts, dropped_bytes);
+
+	sch_tree_unlock(sch);
+	return 0;
+}
+
+static int codel_init(struct Qdisc *sch, struct nlattr *opt,
+		      struct netlink_ext_ack *extack)
+{
+	struct codel_sched_data *q = qdisc_priv(sch);
+
+	sch->limit = DEFAULT_CODEL_LIMIT;
+
+	codel_params_init(&q->params);
+	codel_vars_init(&q->vars);
+	codel_stats_init(&q->stats);
+	q->params.mtu = clamp_t(u32, psched_mtu(qdisc_dev(sch)), 256, 1 << 20);
+
+	if (opt) {
+		int err = codel_change(sch, opt, extack);
+
+		if (err)
+			return err;
+	}
+
+	if (sch->limit >= 1)
+		sch->flags |= TCQ_F_CAN_BYPASS;
+	else
+		sch->flags &= ~TCQ_F_CAN_BYPASS;
+
+	sch->flags |= TCQ_F_DEQUEUE_DROPS;
+
+	return 0;
+}
+
+static int codel_dump(struct Qdisc *sch, struct sk_buff *skb)
+{
+	struct codel_sched_data *q = qdisc_priv(sch);
+	codel_time_t ce_threshold;
+	struct nlattr *opts;
+
+	opts = nla_nest_start_noflag(skb, TCA_OPTIONS);
+	if (opts == NULL)
+		goto nla_put_failure;
+
+	if (nla_put_u32(skb, TCA_CODEL_TARGET,
+			codel_time_to_us(READ_ONCE(q->params.target))) ||
+	    nla_put_u32(skb, TCA_CODEL_LIMIT,
+			READ_ONCE(sch->limit)) ||
+	    nla_put_u32(skb, TCA_CODEL_INTERVAL,
+			codel_time_to_us(READ_ONCE(q->params.interval))) ||
+	    nla_put_u32(skb, TCA_CODEL_ECN,
+			READ_ONCE(q->params.ecn)))
+		goto nla_put_failure;
+	ce_threshold = READ_ONCE(q->params.ce_threshold);
+	if (ce_threshold != CODEL_DISABLED_THRESHOLD &&
+	    nla_put_u32(skb, TCA_CODEL_CE_THRESHOLD,
+			codel_time_to_us(ce_threshold)))
+		goto nla_put_failure;
+	return nla_nest_end(skb, opts);
+
+nla_put_failure:
+	nla_nest_cancel(skb, opts);
+	return -1;
+}
+
+static int codel_dump_stats(struct Qdisc *sch, struct gnet_dump *d)
+{
+	const struct codel_sched_data *q = qdisc_priv(sch);
+	struct tc_codel_xstats st = {
+		.maxpacket	= READ_ONCE(q->stats.maxpacket),
+		.count		= READ_ONCE(q->vars.count),
+		.lastcount	= READ_ONCE(q->vars.lastcount),
+		.drop_overlimit = READ_ONCE(q->drop_overlimit),
+		.ldelay		= codel_time_to_us(READ_ONCE(q->vars.ldelay)),
+		.dropping	= READ_ONCE(q->vars.dropping),
+		.ecn_mark	= READ_ONCE(q->stats.ecn_mark),
+		.ce_mark	= READ_ONCE(q->stats.ce_mark),
+	};
+
+	if (st.dropping) {
+		codel_tdiff_t delta = READ_ONCE(q->vars.drop_next) - codel_get_time();
+
+		if (delta >= 0)
+			st.drop_next = codel_time_to_us(delta);
+		else
+			st.drop_next = -codel_time_to_us(-delta);
+	}
+
+	return gnet_stats_copy_app(d, &st, sizeof(st));
+}
+
+static void codel_reset(struct Qdisc *sch)
+{
+	struct codel_sched_data *q = qdisc_priv(sch);
+
+	qdisc_reset_queue(sch);
+	codel_vars_init(&q->vars);
+}
+
+static struct Qdisc_ops codel_qdisc_ops __read_mostly = {
+	.id		=	"codel",
+	.priv_size	=	sizeof(struct codel_sched_data),
+
+	.enqueue	=	codel_qdisc_enqueue,
+	.dequeue	=	codel_qdisc_dequeue,
+	.peek		=	codel_peek,
+	.init		=	codel_init,
+	.reset		=	codel_reset,
+	.change 	=	codel_change,
+	.dump		=	codel_dump,
+	.dump_stats	=	codel_dump_stats,
+	.owner		=	THIS_MODULE,
+};
+MODULE_ALIAS_NET_SCH("codel");
+
+static int __init codel_module_init(void)
+{
+	return register_qdisc(&codel_qdisc_ops);
+}
+
+static void __exit codel_module_exit(void)
+{
+	unregister_qdisc(&codel_qdisc_ops);
+}
+
+module_init(codel_module_init)
+module_exit(codel_module_exit)
+
+MODULE_DESCRIPTION("Controlled Delay queue discipline");
+MODULE_AUTHOR("Dave Taht");
+MODULE_AUTHOR("Eric Dumazet");
+MODULE_LICENSE("Dual BSD/GPL");