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, 215 insertions, 0 deletions

diff --git a/net/mac80211/mesh_sync.c b/net/mac80211/mesh_sync.c
new file mode 100644
index 000000000..24a68eef7
--- /dev/null
+++ b/net/mac80211/mesh_sync.c
@@ -0,0 +1,215 @@
+// SPDX-License-Identifier: GPL-2.0-only
+/*
+ * Copyright 2011-2012, Pavel Zubarev <pavel.zubarev@gmail.com>
+ * Copyright 2011-2012, Marco Porsch <marco.porsch@s2005.tu-chemnitz.de>
+ * Copyright 2011-2012, cozybit Inc.
+ * Copyright (C) 2021,2023 Intel Corporation
+ */
+
+#include "ieee80211_i.h"
+#include "mesh.h"
+#include "driver-ops.h"
+
+/* This is not in the standard.  It represents a tolerable tsf drift below
+ * which we do no TSF adjustment.
+ */
+#define TOFFSET_MINIMUM_ADJUSTMENT 10
+
+/* This is not in the standard. It is a margin added to the
+ * Toffset setpoint to mitigate TSF overcorrection
+ * introduced by TSF adjustment latency.
+ */
+#define TOFFSET_SET_MARGIN 20
+
+/* This is not in the standard.  It represents the maximum Toffset jump above
+ * which we'll invalidate the Toffset setpoint and choose a new setpoint.  This
+ * could be, for instance, in case a neighbor is restarted and its TSF counter
+ * reset.
+ */
+#define TOFFSET_MAXIMUM_ADJUSTMENT 800		/* 0.8 ms */
+
+struct sync_method {
+	u8 method;
+	struct ieee80211_mesh_sync_ops ops;
+};
+
+/**
+ * mesh_peer_tbtt_adjusting - check if an mp is currently adjusting its TBTT
+ *
+ * @cfg: mesh config element from the mesh peer (or %NULL)
+ *
+ * Returns: If the mesh peer is currently adjusting its TBTT
+ */
+static bool mesh_peer_tbtt_adjusting(const struct ieee80211_meshconf_ie *cfg)
+{
+	return cfg &&
+	       (cfg->meshconf_cap & IEEE80211_MESHCONF_CAPAB_TBTT_ADJUSTING);
+}
+
+void mesh_sync_adjust_tsf(struct ieee80211_sub_if_data *sdata)
+{
+	struct ieee80211_local *local = sdata->local;
+	struct ieee80211_if_mesh *ifmsh = &sdata->u.mesh;
+	/* sdata->vif.bss_conf.beacon_int in 1024us units, 0.04% */
+	u64 beacon_int_fraction = sdata->vif.bss_conf.beacon_int * 1024 / 2500;
+	u64 tsf;
+	u64 tsfdelta;
+
+	spin_lock_bh(&ifmsh->sync_offset_lock);
+	if (ifmsh->sync_offset_clockdrift_max < beacon_int_fraction) {
+		msync_dbg(sdata, "TSF : max clockdrift=%lld; adjusting\n",
+			  (long long) ifmsh->sync_offset_clockdrift_max);
+		tsfdelta = -ifmsh->sync_offset_clockdrift_max;
+		ifmsh->sync_offset_clockdrift_max = 0;
+	} else {
+		msync_dbg(sdata, "TSF : max clockdrift=%lld; adjusting by %llu\n",
+			  (long long) ifmsh->sync_offset_clockdrift_max,
+			  (unsigned long long) beacon_int_fraction);
+		tsfdelta = -beacon_int_fraction;
+		ifmsh->sync_offset_clockdrift_max -= beacon_int_fraction;
+	}
+	spin_unlock_bh(&ifmsh->sync_offset_lock);
+
+	if (local->ops->offset_tsf) {
+		drv_offset_tsf(local, sdata, tsfdelta);
+	} else {
+		tsf = drv_get_tsf(local, sdata);
+		if (tsf != -1ULL)
+			drv_set_tsf(local, sdata, tsf + tsfdelta);
+	}
+}
+
+static void
+mesh_sync_offset_rx_bcn_presp(struct ieee80211_sub_if_data *sdata, u16 stype,
+			      struct ieee80211_mgmt *mgmt, unsigned int len,
+			      const struct ieee80211_meshconf_ie *mesh_cfg,
+			      struct ieee80211_rx_status *rx_status)
+{
+	struct ieee80211_if_mesh *ifmsh = &sdata->u.mesh;
+	struct ieee80211_local *local = sdata->local;
+	struct sta_info *sta;
+	u64 t_t, t_r;
+
+	WARN_ON(ifmsh->mesh_sp_id != IEEE80211_SYNC_METHOD_NEIGHBOR_OFFSET);
+
+	/* standard mentions only beacons */
+	if (stype != IEEE80211_STYPE_BEACON)
+		return;
+
+	/*
+	 * Get time when timestamp field was received.  If we don't
+	 * have rx timestamps, then use current tsf as an approximation.
+	 * drv_get_tsf() must be called before entering the rcu-read
+	 * section.
+	 */
+	if (ieee80211_have_rx_timestamp(rx_status))
+		t_r = ieee80211_calculate_rx_timestamp(&local->hw, rx_status,
+						       len + FCS_LEN, 24);
+	else
+		t_r = drv_get_tsf(local, sdata);
+
+	rcu_read_lock();
+	sta = sta_info_get(sdata, mgmt->sa);
+	if (!sta)
+		goto no_sync;
+
+	/* check offset sync conditions (13.13.2.2.1)
+	 *
+	 * TODO also sync to
+	 * dot11MeshNbrOffsetMaxNeighbor non-peer non-MBSS neighbors
+	 */
+
+	if (mesh_peer_tbtt_adjusting(mesh_cfg)) {
+		msync_dbg(sdata, "STA %pM : is adjusting TBTT\n",
+			  sta->sta.addr);
+		goto no_sync;
+	}
+
+	/* Timing offset calculation (see 13.13.2.2.2) */
+	t_t = le64_to_cpu(mgmt->u.beacon.timestamp);
+	sta->mesh->t_offset = t_t - t_r;
+
+	if (test_sta_flag(sta, WLAN_STA_TOFFSET_KNOWN)) {
+		s64 t_clockdrift = sta->mesh->t_offset_setpoint - sta->mesh->t_offset;
+		msync_dbg(sdata,
+			  "STA %pM : t_offset=%lld, t_offset_setpoint=%lld, t_clockdrift=%lld\n",
+			  sta->sta.addr, (long long) sta->mesh->t_offset,
+			  (long long) sta->mesh->t_offset_setpoint,
+			  (long long) t_clockdrift);
+
+		if (t_clockdrift > TOFFSET_MAXIMUM_ADJUSTMENT ||
+		    t_clockdrift < -TOFFSET_MAXIMUM_ADJUSTMENT) {
+			msync_dbg(sdata,
+				  "STA %pM : t_clockdrift=%lld too large, setpoint reset\n",
+				  sta->sta.addr,
+				  (long long) t_clockdrift);
+			clear_sta_flag(sta, WLAN_STA_TOFFSET_KNOWN);
+			goto no_sync;
+		}
+
+		spin_lock_bh(&ifmsh->sync_offset_lock);
+		if (t_clockdrift > ifmsh->sync_offset_clockdrift_max)
+			ifmsh->sync_offset_clockdrift_max = t_clockdrift;
+		spin_unlock_bh(&ifmsh->sync_offset_lock);
+	} else {
+		sta->mesh->t_offset_setpoint = sta->mesh->t_offset - TOFFSET_SET_MARGIN;
+		set_sta_flag(sta, WLAN_STA_TOFFSET_KNOWN);
+		msync_dbg(sdata,
+			  "STA %pM : offset was invalid, t_offset=%lld\n",
+			  sta->sta.addr,
+			  (long long) sta->mesh->t_offset);
+	}
+
+no_sync:
+	rcu_read_unlock();
+}
+
+static void mesh_sync_offset_adjust_tsf(struct ieee80211_sub_if_data *sdata,
+					 struct beacon_data *beacon)
+{
+	struct ieee80211_if_mesh *ifmsh = &sdata->u.mesh;
+
+	WARN_ON(ifmsh->mesh_sp_id != IEEE80211_SYNC_METHOD_NEIGHBOR_OFFSET);
+	WARN_ON(!rcu_read_lock_held());
+
+	spin_lock_bh(&ifmsh->sync_offset_lock);
+
+	if (ifmsh->sync_offset_clockdrift_max > TOFFSET_MINIMUM_ADJUSTMENT) {
+		/* Since adjusting the tsf here would
+		 * require a possibly blocking call
+		 * to the driver tsf setter, we punt
+		 * the tsf adjustment to the mesh tasklet
+		 */
+		msync_dbg(sdata,
+			  "TSF : kicking off TSF adjustment with clockdrift_max=%lld\n",
+			  ifmsh->sync_offset_clockdrift_max);
+		set_bit(MESH_WORK_DRIFT_ADJUST, &ifmsh->wrkq_flags);
+	} else {
+		msync_dbg(sdata,
+			  "TSF : max clockdrift=%lld; too small to adjust\n",
+			  (long long)ifmsh->sync_offset_clockdrift_max);
+		ifmsh->sync_offset_clockdrift_max = 0;
+	}
+	spin_unlock_bh(&ifmsh->sync_offset_lock);
+}
+
+static const struct sync_method sync_methods[] = {
+	{
+		.method = IEEE80211_SYNC_METHOD_NEIGHBOR_OFFSET,
+		.ops = {
+			.rx_bcn_presp = &mesh_sync_offset_rx_bcn_presp,
+			.adjust_tsf = &mesh_sync_offset_adjust_tsf,
+		}
+	},
+};
+
+const struct ieee80211_mesh_sync_ops *ieee80211_mesh_sync_ops_get(u8 method)
+{
+	int i;
+
+	for (i = 0 ; i < ARRAY_SIZE(sync_methods); ++i) {
+		if (sync_methods[i].method == method)
+			return &sync_methods[i].ops;
+	}
+	return NULL;
+}