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, 310 insertions, 0 deletions

diff --git a/fs/afs/vl_probe.c b/fs/afs/vl_probe.c
new file mode 100644
index 000000000..3d2e0c925
--- /dev/null
+++ b/fs/afs/vl_probe.c
@@ -0,0 +1,310 @@
+// SPDX-License-Identifier: GPL-2.0-or-later
+/* AFS vlserver probing
+ *
+ * Copyright (C) 2018 Red Hat, Inc. All Rights Reserved.
+ * Written by David Howells (dhowells@redhat.com)
+ */
+
+#include <linux/sched.h>
+#include <linux/slab.h>
+#include "afs_fs.h"
+#include "internal.h"
+#include "protocol_yfs.h"
+
+
+/*
+ * Handle the completion of a set of probes.
+ */
+static void afs_finished_vl_probe(struct afs_vlserver *server)
+{
+	if (!(server->probe.flags & AFS_VLSERVER_PROBE_RESPONDED)) {
+		server->rtt = UINT_MAX;
+		clear_bit(AFS_VLSERVER_FL_RESPONDING, &server->flags);
+	}
+
+	clear_bit_unlock(AFS_VLSERVER_FL_PROBING, &server->flags);
+	wake_up_bit(&server->flags, AFS_VLSERVER_FL_PROBING);
+}
+
+/*
+ * Handle the completion of a probe RPC call.
+ */
+static void afs_done_one_vl_probe(struct afs_vlserver *server, bool wake_up)
+{
+	if (atomic_dec_and_test(&server->probe_outstanding)) {
+		afs_finished_vl_probe(server);
+		wake_up = true;
+	}
+
+	if (wake_up)
+		wake_up_all(&server->probe_wq);
+}
+
+/*
+ * Process the result of probing a vlserver.  This is called after successful
+ * or failed delivery of an VL.GetCapabilities operation.
+ */
+void afs_vlserver_probe_result(struct afs_call *call)
+{
+	struct afs_addr_list *alist = call->vl_probe;
+	struct afs_vlserver *server = call->vlserver;
+	struct afs_address *addr = &alist->addrs[call->probe_index];
+	unsigned int server_index = call->server_index;
+	unsigned int rtt_us = 0;
+	unsigned int index = call->probe_index;
+	bool have_result = false;
+	int ret = call->error;
+
+	_enter("%s,%u,%u,%d,%d", server->name, server_index, index, ret, call->abort_code);
+
+	spin_lock(&server->probe_lock);
+
+	switch (ret) {
+	case 0:
+		server->probe.error = 0;
+		goto responded;
+	case -ECONNABORTED:
+		if (!(server->probe.flags & AFS_VLSERVER_PROBE_RESPONDED)) {
+			server->probe.abort_code = call->abort_code;
+			server->probe.error = ret;
+		}
+		goto responded;
+	case -ENOMEM:
+	case -ENONET:
+	case -EKEYEXPIRED:
+	case -EKEYREVOKED:
+	case -EKEYREJECTED:
+		server->probe.flags |= AFS_VLSERVER_PROBE_LOCAL_FAILURE;
+		if (server->probe.error == 0)
+			server->probe.error = ret;
+		trace_afs_io_error(call->debug_id, ret, afs_io_error_vl_probe_fail);
+		goto out;
+	case -ECONNRESET: /* Responded, but call expired. */
+	case -ERFKILL:
+	case -EADDRNOTAVAIL:
+	case -ENETUNREACH:
+	case -EHOSTUNREACH:
+	case -EHOSTDOWN:
+	case -ECONNREFUSED:
+	case -ETIMEDOUT:
+	case -ETIME:
+	default:
+		clear_bit(index, &alist->responded);
+		set_bit(index, &alist->probe_failed);
+		if (!(server->probe.flags & AFS_VLSERVER_PROBE_RESPONDED) &&
+		    (server->probe.error == 0 ||
+		     server->probe.error == -ETIMEDOUT ||
+		     server->probe.error == -ETIME))
+			server->probe.error = ret;
+		trace_afs_io_error(call->debug_id, ret, afs_io_error_vl_probe_fail);
+		goto out;
+	}
+
+responded:
+	set_bit(index, &alist->responded);
+	clear_bit(index, &alist->probe_failed);
+
+	if (call->service_id == YFS_VL_SERVICE) {
+		server->probe.flags |= AFS_VLSERVER_PROBE_IS_YFS;
+		set_bit(AFS_VLSERVER_FL_IS_YFS, &server->flags);
+		server->service_id = call->service_id;
+	} else {
+		server->probe.flags |= AFS_VLSERVER_PROBE_NOT_YFS;
+		if (!(server->probe.flags & AFS_VLSERVER_PROBE_IS_YFS)) {
+			clear_bit(AFS_VLSERVER_FL_IS_YFS, &server->flags);
+			server->service_id = call->service_id;
+		}
+	}
+
+	rtt_us = rxrpc_kernel_get_srtt(addr->peer);
+	if (rtt_us < server->probe.rtt) {
+		server->probe.rtt = rtt_us;
+		server->rtt = rtt_us;
+		alist->preferred = index;
+	}
+
+	smp_wmb(); /* Set rtt before responded. */
+	server->probe.flags |= AFS_VLSERVER_PROBE_RESPONDED;
+	set_bit(AFS_VLSERVER_FL_PROBED, &server->flags);
+	set_bit(AFS_VLSERVER_FL_RESPONDING, &server->flags);
+	have_result = true;
+out:
+	spin_unlock(&server->probe_lock);
+
+	trace_afs_vl_probe(server, false, alist, index, call->error, call->abort_code, rtt_us);
+	_debug("probe [%u][%u] %pISpc rtt=%d ret=%d",
+	       server_index, index, rxrpc_kernel_remote_addr(addr->peer),
+	       rtt_us, ret);
+
+	afs_done_one_vl_probe(server, have_result);
+}
+
+/*
+ * Probe all of a vlserver's addresses to find out the best route and to
+ * query its capabilities.
+ */
+static bool afs_do_probe_vlserver(struct afs_net *net,
+				  struct afs_vlserver *server,
+				  struct key *key,
+				  unsigned int server_index,
+				  struct afs_error *_e)
+{
+	struct afs_addr_list *alist;
+	struct afs_call *call;
+	unsigned long unprobed;
+	unsigned int index, i;
+	bool in_progress = false;
+	int best_prio;
+
+	_enter("%s", server->name);
+
+	read_lock(&server->lock);
+	alist = rcu_dereference_protected(server->addresses,
+					  lockdep_is_held(&server->lock));
+	afs_get_addrlist(alist, afs_alist_trace_get_vlprobe);
+	read_unlock(&server->lock);
+
+	atomic_set(&server->probe_outstanding, alist->nr_addrs);
+	memset(&server->probe, 0, sizeof(server->probe));
+	server->probe.rtt = UINT_MAX;
+
+	unprobed = (1UL << alist->nr_addrs) - 1;
+	while (unprobed) {
+		best_prio = -1;
+		index = 0;
+		for (i = 0; i < alist->nr_addrs; i++) {
+			if (test_bit(i, &unprobed) &&
+			    alist->addrs[i].prio > best_prio) {
+				index = i;
+				best_prio = alist->addrs[i].prio;
+			}
+		}
+		__clear_bit(index, &unprobed);
+
+		trace_afs_vl_probe(server, true, alist, index, 0, 0, 0);
+		call = afs_vl_get_capabilities(net, alist, index, key, server,
+					       server_index);
+		if (!IS_ERR(call)) {
+			afs_prioritise_error(_e, call->error, call->abort_code);
+			afs_put_call(call);
+			in_progress = true;
+		} else {
+			afs_prioritise_error(_e, PTR_ERR(call), 0);
+			afs_done_one_vl_probe(server, false);
+		}
+	}
+
+	afs_put_addrlist(alist, afs_alist_trace_put_vlprobe);
+	return in_progress;
+}
+
+/*
+ * Send off probes to all unprobed servers.
+ */
+int afs_send_vl_probes(struct afs_net *net, struct key *key,
+		       struct afs_vlserver_list *vllist)
+{
+	struct afs_vlserver *server;
+	struct afs_error e = {};
+	bool in_progress = false;
+	int i;
+
+	for (i = 0; i < vllist->nr_servers; i++) {
+		server = vllist->servers[i].server;
+		if (test_bit(AFS_VLSERVER_FL_PROBED, &server->flags))
+			continue;
+
+		if (!test_and_set_bit_lock(AFS_VLSERVER_FL_PROBING, &server->flags) &&
+		    afs_do_probe_vlserver(net, server, key, i, &e))
+			in_progress = true;
+	}
+
+	return in_progress ? 0 : e.error;
+}
+
+/*
+ * Wait for the first as-yet untried server to respond.
+ */
+int afs_wait_for_vl_probes(struct afs_vlserver_list *vllist,
+			   unsigned long untried)
+{
+	struct wait_queue_entry *waits;
+	struct afs_vlserver *server;
+	unsigned int rtt = UINT_MAX, rtt_s;
+	bool have_responders = false;
+	int pref = -1, i;
+
+	_enter("%u,%lx", vllist->nr_servers, untried);
+
+	/* Only wait for servers that have a probe outstanding. */
+	for (i = 0; i < vllist->nr_servers; i++) {
+		if (test_bit(i, &untried)) {
+			server = vllist->servers[i].server;
+			if (!test_bit(AFS_VLSERVER_FL_PROBING, &server->flags))
+				__clear_bit(i, &untried);
+			if (server->probe.flags & AFS_VLSERVER_PROBE_RESPONDED)
+				have_responders = true;
+		}
+	}
+	if (have_responders || !untried)
+		return 0;
+
+	waits = kmalloc(array_size(vllist->nr_servers, sizeof(*waits)), GFP_KERNEL);
+	if (!waits)
+		return -ENOMEM;
+
+	for (i = 0; i < vllist->nr_servers; i++) {
+		if (test_bit(i, &untried)) {
+			server = vllist->servers[i].server;
+			init_waitqueue_entry(&waits[i], current);
+			add_wait_queue(&server->probe_wq, &waits[i]);
+		}
+	}
+
+	for (;;) {
+		bool still_probing = false;
+
+		set_current_state(TASK_INTERRUPTIBLE);
+		for (i = 0; i < vllist->nr_servers; i++) {
+			if (test_bit(i, &untried)) {
+				server = vllist->servers[i].server;
+				if (server->probe.flags & AFS_VLSERVER_PROBE_RESPONDED)
+					goto stop;
+				if (test_bit(AFS_VLSERVER_FL_PROBING, &server->flags))
+					still_probing = true;
+			}
+		}
+
+		if (!still_probing || signal_pending(current))
+			goto stop;
+		schedule();
+	}
+
+stop:
+	set_current_state(TASK_RUNNING);
+
+	for (i = 0; i < vllist->nr_servers; i++) {
+		if (test_bit(i, &untried)) {
+			server = vllist->servers[i].server;
+			rtt_s = READ_ONCE(server->rtt);
+			if (test_bit(AFS_VLSERVER_FL_RESPONDING, &server->flags) &&
+			    rtt_s < rtt) {
+				pref = i;
+				rtt = rtt_s;
+			}
+
+			remove_wait_queue(&server->probe_wq, &waits[i]);
+		}
+	}
+
+	kfree(waits);
+
+	if (pref == -1 && signal_pending(current))
+		return -ERESTARTSYS;
+
+	if (pref >= 0)
+		vllist->preferred = pref;
+
+	_leave(" = 0 [%u]", pref);
+	return 0;
+}