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, 415 insertions, 0 deletions

diff --git a/net/sched/act_ctinfo.c b/net/sched/act_ctinfo.c
new file mode 100644
index 000000000..fced4b109
--- /dev/null
+++ b/net/sched/act_ctinfo.c
@@ -0,0 +1,415 @@
+// SPDX-License-Identifier: GPL-2.0+
+/* net/sched/act_ctinfo.c  netfilter ctinfo connmark actions
+ *
+ * Copyright (c) 2019 Kevin Darbyshire-Bryant <ldir@darbyshire-bryant.me.uk>
+ */
+
+#include <linux/module.h>
+#include <linux/init.h>
+#include <linux/kernel.h>
+#include <linux/skbuff.h>
+#include <linux/rtnetlink.h>
+#include <linux/pkt_cls.h>
+#include <linux/ip.h>
+#include <linux/ipv6.h>
+#include <net/netlink.h>
+#include <net/pkt_sched.h>
+#include <net/act_api.h>
+#include <net/pkt_cls.h>
+#include <net/inet_ecn.h>
+#include <uapi/linux/tc_act/tc_ctinfo.h>
+#include <net/tc_act/tc_ctinfo.h>
+#include <net/tc_wrapper.h>
+
+#include <net/netfilter/nf_conntrack.h>
+#include <net/netfilter/nf_conntrack_core.h>
+#include <net/netfilter/nf_conntrack_ecache.h>
+#include <net/netfilter/nf_conntrack_zones.h>
+
+static struct tc_action_ops act_ctinfo_ops;
+
+static void tcf_ctinfo_dscp_set(struct nf_conn *ct, struct tcf_ctinfo *ca,
+				struct tcf_ctinfo_params *cp,
+				struct sk_buff *skb, int wlen, int proto)
+{
+	u8 dscp, newdscp;
+
+	newdscp = (((READ_ONCE(ct->mark) & cp->dscpmask) >> cp->dscpmaskshift) << 2) &
+		     ~INET_ECN_MASK;
+
+	switch (proto) {
+	case NFPROTO_IPV4:
+		dscp = ipv4_get_dsfield(ip_hdr(skb)) & ~INET_ECN_MASK;
+		if (dscp != newdscp) {
+			if (likely(!skb_try_make_writable(skb, wlen))) {
+				ipv4_change_dsfield(ip_hdr(skb),
+						    INET_ECN_MASK,
+						    newdscp);
+				atomic64_inc(&ca->stats_dscp_set);
+			} else {
+				atomic64_inc(&ca->stats_dscp_error);
+			}
+		}
+		break;
+	case NFPROTO_IPV6:
+		dscp = ipv6_get_dsfield(ipv6_hdr(skb)) & ~INET_ECN_MASK;
+		if (dscp != newdscp) {
+			if (likely(!skb_try_make_writable(skb, wlen))) {
+				ipv6_change_dsfield(ipv6_hdr(skb),
+						    INET_ECN_MASK,
+						    newdscp);
+				atomic64_inc(&ca->stats_dscp_set);
+			} else {
+				atomic64_inc(&ca->stats_dscp_error);
+			}
+		}
+		break;
+	default:
+		break;
+	}
+}
+
+static void tcf_ctinfo_cpmark_set(struct nf_conn *ct, struct tcf_ctinfo *ca,
+				  struct tcf_ctinfo_params *cp,
+				  struct sk_buff *skb)
+{
+	atomic64_inc(&ca->stats_cpmark_set);
+	skb->mark = READ_ONCE(ct->mark) & cp->cpmarkmask;
+}
+
+TC_INDIRECT_SCOPE int tcf_ctinfo_act(struct sk_buff *skb,
+				     const struct tc_action *a,
+				     struct tcf_result *res)
+{
+	const struct nf_conntrack_tuple_hash *thash = NULL;
+	struct tcf_ctinfo *ca = to_ctinfo(a);
+	struct nf_conntrack_tuple tuple;
+	struct nf_conntrack_zone zone;
+	enum ip_conntrack_info ctinfo;
+	struct tcf_ctinfo_params *cp;
+	struct nf_conn *ct;
+	int proto, wlen;
+
+	cp = rcu_dereference_bh(ca->params);
+
+	tcf_lastuse_update(&ca->tcf_tm);
+	tcf_action_update_bstats(&ca->common, skb);
+
+	wlen = skb_network_offset(skb);
+	switch (skb_protocol(skb, true)) {
+	case htons(ETH_P_IP):
+		wlen += sizeof(struct iphdr);
+		if (!pskb_may_pull(skb, wlen))
+			goto out;
+
+		proto = NFPROTO_IPV4;
+		break;
+	case htons(ETH_P_IPV6):
+		wlen += sizeof(struct ipv6hdr);
+		if (!pskb_may_pull(skb, wlen))
+			goto out;
+
+		proto = NFPROTO_IPV6;
+		break;
+	default:
+		goto out;
+	}
+
+	ct = nf_ct_get(skb, &ctinfo);
+	if (!ct) { /* look harder, usually ingress */
+		if (!nf_ct_get_tuplepr(skb, skb_network_offset(skb),
+				       proto, cp->net, &tuple))
+			goto out;
+		zone.id = cp->zone;
+		zone.dir = NF_CT_DEFAULT_ZONE_DIR;
+
+		thash = nf_conntrack_find_get(cp->net, &zone, &tuple);
+		if (!thash)
+			goto out;
+
+		ct = nf_ct_tuplehash_to_ctrack(thash);
+	}
+
+	if (cp->mode & CTINFO_MODE_DSCP)
+		if (!cp->dscpstatemask || (READ_ONCE(ct->mark) & cp->dscpstatemask))
+			tcf_ctinfo_dscp_set(ct, ca, cp, skb, wlen, proto);
+
+	if (cp->mode & CTINFO_MODE_CPMARK)
+		tcf_ctinfo_cpmark_set(ct, ca, cp, skb);
+
+	if (thash)
+		nf_ct_put(ct);
+out:
+	return cp->action;
+}
+
+static const struct nla_policy ctinfo_policy[TCA_CTINFO_MAX + 1] = {
+	[TCA_CTINFO_ACT]		  =
+		NLA_POLICY_EXACT_LEN(sizeof(struct tc_ctinfo)),
+	[TCA_CTINFO_ZONE]		  = { .type = NLA_U16 },
+	[TCA_CTINFO_PARMS_DSCP_MASK]	  = { .type = NLA_U32 },
+	[TCA_CTINFO_PARMS_DSCP_STATEMASK] = { .type = NLA_U32 },
+	[TCA_CTINFO_PARMS_CPMARK_MASK]	  = { .type = NLA_U32 },
+};
+
+static int tcf_ctinfo_init(struct net *net, struct nlattr *nla,
+			   struct nlattr *est, struct tc_action **a,
+			   struct tcf_proto *tp, u32 flags,
+			   struct netlink_ext_ack *extack)
+{
+	struct tc_action_net *tn = net_generic(net, act_ctinfo_ops.net_id);
+	bool bind = flags & TCA_ACT_FLAGS_BIND;
+	u32 dscpmask = 0, dscpstatemask, index;
+	struct nlattr *tb[TCA_CTINFO_MAX + 1];
+	struct tcf_ctinfo_params *cp_new;
+	struct tcf_chain *goto_ch = NULL;
+	struct tc_ctinfo *actparm;
+	struct tcf_ctinfo *ci;
+	u8 dscpmaskshift;
+	int ret = 0, err;
+
+	if (!nla) {
+		NL_SET_ERR_MSG_MOD(extack, "ctinfo requires attributes to be passed");
+		return -EINVAL;
+	}
+
+	err = nla_parse_nested(tb, TCA_CTINFO_MAX, nla, ctinfo_policy, extack);
+	if (err < 0)
+		return err;
+
+	if (!tb[TCA_CTINFO_ACT]) {
+		NL_SET_ERR_MSG_MOD(extack,
+				   "Missing required TCA_CTINFO_ACT attribute");
+		return -EINVAL;
+	}
+	actparm = nla_data(tb[TCA_CTINFO_ACT]);
+
+	/* do some basic validation here before dynamically allocating things */
+	/* that we would otherwise have to clean up.			      */
+	if (tb[TCA_CTINFO_PARMS_DSCP_MASK]) {
+		dscpmask = nla_get_u32(tb[TCA_CTINFO_PARMS_DSCP_MASK]);
+		/* need contiguous 6 bit mask */
+		dscpmaskshift = dscpmask ? __ffs(dscpmask) : 0;
+		if ((~0 & (dscpmask >> dscpmaskshift)) != 0x3f) {
+			NL_SET_ERR_MSG_ATTR(extack,
+					    tb[TCA_CTINFO_PARMS_DSCP_MASK],
+					    "dscp mask must be 6 contiguous bits");
+			return -EINVAL;
+		}
+		dscpstatemask =
+			nla_get_u32_default(tb[TCA_CTINFO_PARMS_DSCP_STATEMASK],
+					    0);
+		/* mask & statemask must not overlap */
+		if (dscpmask & dscpstatemask) {
+			NL_SET_ERR_MSG_ATTR(extack,
+					    tb[TCA_CTINFO_PARMS_DSCP_STATEMASK],
+					    "dscp statemask must not overlap dscp mask");
+			return -EINVAL;
+		}
+	}
+
+	/* done the validation:now to the actual action allocation */
+	index = actparm->index;
+	err = tcf_idr_check_alloc(tn, &index, a, bind);
+	if (!err) {
+		ret = tcf_idr_create_from_flags(tn, index, est, a,
+						&act_ctinfo_ops, bind, flags);
+		if (ret) {
+			tcf_idr_cleanup(tn, index);
+			return ret;
+		}
+		ret = ACT_P_CREATED;
+	} else if (err > 0) {
+		if (bind) /* don't override defaults */
+			return ACT_P_BOUND;
+		if (!(flags & TCA_ACT_FLAGS_REPLACE)) {
+			tcf_idr_release(*a, bind);
+			return -EEXIST;
+		}
+	} else {
+		return err;
+	}
+
+	err = tcf_action_check_ctrlact(actparm->action, tp, &goto_ch, extack);
+	if (err < 0)
+		goto release_idr;
+
+	ci = to_ctinfo(*a);
+
+	cp_new = kzalloc_obj(*cp_new);
+	if (unlikely(!cp_new)) {
+		err = -ENOMEM;
+		goto put_chain;
+	}
+
+	cp_new->net = net;
+	cp_new->zone = nla_get_u16_default(tb[TCA_CTINFO_ZONE], 0);
+	if (dscpmask) {
+		cp_new->dscpmask = dscpmask;
+		cp_new->dscpmaskshift = dscpmaskshift;
+		cp_new->dscpstatemask = dscpstatemask;
+		cp_new->mode |= CTINFO_MODE_DSCP;
+	}
+
+	if (tb[TCA_CTINFO_PARMS_CPMARK_MASK]) {
+		cp_new->cpmarkmask =
+				nla_get_u32(tb[TCA_CTINFO_PARMS_CPMARK_MASK]);
+		cp_new->mode |= CTINFO_MODE_CPMARK;
+	}
+
+	cp_new->action = actparm->action;
+
+	spin_lock_bh(&ci->tcf_lock);
+	goto_ch = tcf_action_set_ctrlact(*a, actparm->action, goto_ch);
+	cp_new = rcu_replace_pointer(ci->params, cp_new,
+				     lockdep_is_held(&ci->tcf_lock));
+	spin_unlock_bh(&ci->tcf_lock);
+
+	if (goto_ch)
+		tcf_chain_put_by_act(goto_ch);
+	if (cp_new)
+		kfree_rcu(cp_new, rcu);
+
+	return ret;
+
+put_chain:
+	if (goto_ch)
+		tcf_chain_put_by_act(goto_ch);
+release_idr:
+	tcf_idr_release(*a, bind);
+	return err;
+}
+
+static int tcf_ctinfo_dump(struct sk_buff *skb, struct tc_action *a,
+			   int bind, int ref)
+{
+	const struct tcf_ctinfo *ci = to_ctinfo(a);
+	unsigned char *b = skb_tail_pointer(skb);
+	const struct tcf_ctinfo_params *cp;
+	struct tc_ctinfo opt = {
+		.index   = ci->tcf_index,
+		.refcnt  = refcount_read(&ci->tcf_refcnt) - ref,
+		.bindcnt = atomic_read(&ci->tcf_bindcnt) - bind,
+	};
+	struct tcf_t t;
+
+	rcu_read_lock();
+	cp = rcu_dereference(ci->params);
+
+	tcf_tm_dump(&t, &ci->tcf_tm);
+	if (nla_put_64bit(skb, TCA_CTINFO_TM, sizeof(t), &t, TCA_CTINFO_PAD))
+		goto nla_put_failure;
+
+	opt.action = cp->action;
+	if (nla_put(skb, TCA_CTINFO_ACT, sizeof(opt), &opt))
+		goto nla_put_failure;
+
+	if (nla_put_u16(skb, TCA_CTINFO_ZONE, cp->zone))
+		goto nla_put_failure;
+
+	if (cp->mode & CTINFO_MODE_DSCP) {
+		if (nla_put_u32(skb, TCA_CTINFO_PARMS_DSCP_MASK,
+				cp->dscpmask))
+			goto nla_put_failure;
+		if (nla_put_u32(skb, TCA_CTINFO_PARMS_DSCP_STATEMASK,
+				cp->dscpstatemask))
+			goto nla_put_failure;
+	}
+
+	if (cp->mode & CTINFO_MODE_CPMARK) {
+		if (nla_put_u32(skb, TCA_CTINFO_PARMS_CPMARK_MASK,
+				cp->cpmarkmask))
+			goto nla_put_failure;
+	}
+
+	if (nla_put_u64_64bit(skb, TCA_CTINFO_STATS_DSCP_SET,
+			      atomic64_read(&ci->stats_dscp_set),
+			      TCA_CTINFO_PAD))
+		goto nla_put_failure;
+
+	if (nla_put_u64_64bit(skb, TCA_CTINFO_STATS_DSCP_ERROR,
+			      atomic64_read(&ci->stats_dscp_error),
+			      TCA_CTINFO_PAD))
+		goto nla_put_failure;
+
+	if (nla_put_u64_64bit(skb, TCA_CTINFO_STATS_CPMARK_SET,
+			      atomic64_read(&ci->stats_cpmark_set),
+			      TCA_CTINFO_PAD))
+		goto nla_put_failure;
+
+	rcu_read_unlock();
+	return skb->len;
+
+nla_put_failure:
+	rcu_read_unlock();
+	nlmsg_trim(skb, b);
+	return -1;
+}
+
+static void tcf_ctinfo_cleanup(struct tc_action *a)
+{
+	struct tcf_ctinfo *ci = to_ctinfo(a);
+	struct tcf_ctinfo_params *cp;
+
+	cp = rcu_dereference_protected(ci->params, 1);
+	if (cp)
+		kfree_rcu(cp, rcu);
+}
+
+static size_t tcf_ctinfo_get_fill_size(const struct tc_action *act)
+{
+	return nla_total_size(sizeof(struct tc_ctinfo)) /* TCA_CTINFO_ACT */
+		+ nla_total_size(sizeof(u16)) /* TCA_CTINFO_ZONE */
+		/* TCA_CTINFO_PARMS_{DSCP_MASK,DSCP_STATEMASK,CPMARK_MASK} */
+		+ 3 * nla_total_size(sizeof(u32))
+		/* TCA_CTINFO_STATS_{DSCP_SET,DSCP_ERROR,CPMARK_SET} */
+		+ 3 * nla_total_size_64bit(sizeof(u64));
+}
+
+static struct tc_action_ops act_ctinfo_ops = {
+	.kind	= "ctinfo",
+	.id	= TCA_ID_CTINFO,
+	.owner	= THIS_MODULE,
+	.act	= tcf_ctinfo_act,
+	.dump	= tcf_ctinfo_dump,
+	.init	= tcf_ctinfo_init,
+	.cleanup= tcf_ctinfo_cleanup,
+	.get_fill_size = tcf_ctinfo_get_fill_size,
+	.size	= sizeof(struct tcf_ctinfo),
+};
+MODULE_ALIAS_NET_ACT("ctinfo");
+
+static __net_init int ctinfo_init_net(struct net *net)
+{
+	struct tc_action_net *tn = net_generic(net, act_ctinfo_ops.net_id);
+
+	return tc_action_net_init(net, tn, &act_ctinfo_ops);
+}
+
+static void __net_exit ctinfo_exit_net(struct list_head *net_list)
+{
+	tc_action_net_exit(net_list, act_ctinfo_ops.net_id);
+}
+
+static struct pernet_operations ctinfo_net_ops = {
+	.init		= ctinfo_init_net,
+	.exit_batch	= ctinfo_exit_net,
+	.id		= &act_ctinfo_ops.net_id,
+	.size		= sizeof(struct tc_action_net),
+};
+
+static int __init ctinfo_init_module(void)
+{
+	return tcf_register_action(&act_ctinfo_ops, &ctinfo_net_ops);
+}
+
+static void __exit ctinfo_cleanup_module(void)
+{
+	tcf_unregister_action(&act_ctinfo_ops, &ctinfo_net_ops);
+}
+
+module_init(ctinfo_init_module);
+module_exit(ctinfo_cleanup_module);
+MODULE_AUTHOR("Kevin Darbyshire-Bryant <ldir@darbyshire-bryant.me.uk>");
+MODULE_DESCRIPTION("Connection tracking mark actions");
+MODULE_LICENSE("GPL");