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, 280 insertions, 0 deletions

diff --git a/drivers/net/netdevsim/ethtool.c b/drivers/net/netdevsim/ethtool.c
new file mode 100644
index 000000000..025ea7987
--- /dev/null
+++ b/drivers/net/netdevsim/ethtool.c
@@ -0,0 +1,280 @@
+// SPDX-License-Identifier: GPL-2.0
+// Copyright (c) 2020 Facebook
+
+#include <linux/debugfs.h>
+#include <linux/random.h>
+#include <net/netdev_queues.h>
+
+#include "netdevsim.h"
+
+static void
+nsim_get_pause_stats(struct net_device *dev,
+		     struct ethtool_pause_stats *pause_stats)
+{
+	struct netdevsim *ns = netdev_priv(dev);
+
+	if (ns->ethtool.pauseparam.report_stats_rx)
+		pause_stats->rx_pause_frames = 1;
+	if (ns->ethtool.pauseparam.report_stats_tx)
+		pause_stats->tx_pause_frames = 2;
+}
+
+static void
+nsim_get_pauseparam(struct net_device *dev, struct ethtool_pauseparam *pause)
+{
+	struct netdevsim *ns = netdev_priv(dev);
+
+	pause->autoneg = 0; /* We don't support ksettings, so can't pretend */
+	pause->rx_pause = ns->ethtool.pauseparam.rx;
+	pause->tx_pause = ns->ethtool.pauseparam.tx;
+}
+
+static int
+nsim_set_pauseparam(struct net_device *dev, struct ethtool_pauseparam *pause)
+{
+	struct netdevsim *ns = netdev_priv(dev);
+
+	if (pause->autoneg)
+		return -EINVAL;
+
+	ns->ethtool.pauseparam.rx = pause->rx_pause;
+	ns->ethtool.pauseparam.tx = pause->tx_pause;
+	return 0;
+}
+
+static int nsim_get_coalesce(struct net_device *dev,
+			     struct ethtool_coalesce *coal,
+			     struct kernel_ethtool_coalesce *kernel_coal,
+			     struct netlink_ext_ack *extack)
+{
+	struct netdevsim *ns = netdev_priv(dev);
+
+	memcpy(coal, &ns->ethtool.coalesce, sizeof(ns->ethtool.coalesce));
+	return 0;
+}
+
+static int nsim_set_coalesce(struct net_device *dev,
+			     struct ethtool_coalesce *coal,
+			     struct kernel_ethtool_coalesce *kernel_coal,
+			     struct netlink_ext_ack *extack)
+{
+	struct netdevsim *ns = netdev_priv(dev);
+
+	memcpy(&ns->ethtool.coalesce, coal, sizeof(ns->ethtool.coalesce));
+	return 0;
+}
+
+static void nsim_get_ringparam(struct net_device *dev,
+			       struct ethtool_ringparam *ring,
+			       struct kernel_ethtool_ringparam *kernel_ring,
+			       struct netlink_ext_ack *extack)
+{
+	struct netdevsim *ns = netdev_priv(dev);
+
+	memcpy(ring, &ns->ethtool.ring, sizeof(ns->ethtool.ring));
+	kernel_ring->hds_thresh_max = NSIM_HDS_THRESHOLD_MAX;
+
+	if (dev->cfg->hds_config == ETHTOOL_TCP_DATA_SPLIT_UNKNOWN)
+		kernel_ring->tcp_data_split = ETHTOOL_TCP_DATA_SPLIT_ENABLED;
+}
+
+static int nsim_set_ringparam(struct net_device *dev,
+			      struct ethtool_ringparam *ring,
+			      struct kernel_ethtool_ringparam *kernel_ring,
+			      struct netlink_ext_ack *extack)
+{
+	struct netdevsim *ns = netdev_priv(dev);
+
+	ns->ethtool.ring.rx_pending = ring->rx_pending;
+	ns->ethtool.ring.rx_jumbo_pending = ring->rx_jumbo_pending;
+	ns->ethtool.ring.rx_mini_pending = ring->rx_mini_pending;
+	ns->ethtool.ring.tx_pending = ring->tx_pending;
+	return 0;
+}
+
+static void
+nsim_get_channels(struct net_device *dev, struct ethtool_channels *ch)
+{
+	struct netdevsim *ns = netdev_priv(dev);
+
+	ch->max_combined = ns->nsim_bus_dev->num_queues;
+	ch->combined_count = ns->ethtool.channels;
+}
+
+static void
+nsim_wake_queues(struct net_device *dev)
+{
+	struct netdevsim *ns = netdev_priv(dev);
+	struct netdevsim *peer;
+
+	synchronize_net();
+	netif_tx_wake_all_queues(dev);
+
+	rcu_read_lock();
+	peer = rcu_dereference(ns->peer);
+	if (peer)
+		netif_tx_wake_all_queues(peer->netdev);
+	rcu_read_unlock();
+}
+
+static int
+nsim_set_channels(struct net_device *dev, struct ethtool_channels *ch)
+{
+	struct netdevsim *ns = netdev_priv(dev);
+	int err;
+
+	err = netif_set_real_num_queues(dev, ch->combined_count,
+					ch->combined_count);
+	if (err)
+		return err;
+
+	ns->ethtool.channels = ch->combined_count;
+
+	/* Only wake up queues if devices are linked */
+	if (rcu_access_pointer(ns->peer))
+		nsim_wake_queues(dev);
+
+	return 0;
+}
+
+static int
+nsim_get_fecparam(struct net_device *dev, struct ethtool_fecparam *fecparam)
+{
+	struct netdevsim *ns = netdev_priv(dev);
+
+	if (ns->ethtool.get_err)
+		return -ns->ethtool.get_err;
+	memcpy(fecparam, &ns->ethtool.fec, sizeof(ns->ethtool.fec));
+	return 0;
+}
+
+static int
+nsim_set_fecparam(struct net_device *dev, struct ethtool_fecparam *fecparam)
+{
+	struct netdevsim *ns = netdev_priv(dev);
+	u32 fec;
+
+	if (ns->ethtool.set_err)
+		return -ns->ethtool.set_err;
+	memcpy(&ns->ethtool.fec, fecparam, sizeof(ns->ethtool.fec));
+	fec = fecparam->fec;
+	if (fec == ETHTOOL_FEC_AUTO)
+		fec |= ETHTOOL_FEC_OFF;
+	fec |= ETHTOOL_FEC_NONE;
+	ns->ethtool.fec.active_fec = 1 << (fls(fec) - 1);
+	return 0;
+}
+
+static const struct ethtool_fec_hist_range netdevsim_fec_ranges[] = {
+	{ 0, 0},
+	{ 1, 3},
+	{ 4, 7},
+	{ 0, 0}
+};
+
+static void
+nsim_get_fec_stats(struct net_device *dev, struct ethtool_fec_stats *fec_stats,
+		   struct ethtool_fec_hist *hist)
+{
+	struct ethtool_fec_hist_value *values = hist->values;
+
+	hist->ranges = netdevsim_fec_ranges;
+
+	fec_stats->corrected_blocks.total = 123;
+	fec_stats->uncorrectable_blocks.total = 4;
+
+	values[0].per_lane[0] = 125;
+	values[0].per_lane[1] = 120;
+	values[0].per_lane[2] = 100;
+	values[0].per_lane[3] = 100;
+	values[1].sum = 12;
+	values[2].sum = 2;
+	values[2].per_lane[0] = 2;
+	values[2].per_lane[1] = 0;
+	values[2].per_lane[2] = 0;
+	values[2].per_lane[3] = 0;
+}
+
+static int nsim_get_ts_info(struct net_device *dev,
+			    struct kernel_ethtool_ts_info *info)
+{
+	struct netdevsim *ns = netdev_priv(dev);
+
+	info->phc_index = mock_phc_index(ns->phc);
+
+	return 0;
+}
+
+static const struct ethtool_ops nsim_ethtool_ops = {
+	.supported_coalesce_params	= ETHTOOL_COALESCE_ALL_PARAMS,
+	.supported_ring_params		= ETHTOOL_RING_USE_TCP_DATA_SPLIT |
+					  ETHTOOL_RING_USE_HDS_THRS,
+	.op_needs_rtnl			= ETHTOOL_OP_NEEDS_RTNL_SCHANNELS,
+	.get_pause_stats	        = nsim_get_pause_stats,
+	.get_pauseparam		        = nsim_get_pauseparam,
+	.set_pauseparam		        = nsim_set_pauseparam,
+	.set_coalesce			= nsim_set_coalesce,
+	.get_coalesce			= nsim_get_coalesce,
+	.get_ringparam			= nsim_get_ringparam,
+	.set_ringparam			= nsim_set_ringparam,
+	.get_channels			= nsim_get_channels,
+	.set_channels			= nsim_set_channels,
+	.get_fecparam			= nsim_get_fecparam,
+	.set_fecparam			= nsim_set_fecparam,
+	.get_fec_stats			= nsim_get_fec_stats,
+	.get_ts_info			= nsim_get_ts_info,
+};
+
+static void nsim_ethtool_ring_init(struct netdevsim *ns)
+{
+	ns->ethtool.ring.rx_pending = 512;
+	ns->ethtool.ring.rx_max_pending = 4096;
+	ns->ethtool.ring.rx_jumbo_max_pending = 4096;
+	ns->ethtool.ring.rx_mini_max_pending = 4096;
+	ns->ethtool.ring.tx_pending = 512;
+	ns->ethtool.ring.tx_max_pending = 4096;
+}
+
+void nsim_ethtool_init(struct netdevsim *ns)
+{
+	struct dentry *ethtool, *dir;
+
+	ns->netdev->ethtool_ops = &nsim_ethtool_ops;
+
+	nsim_ethtool_ring_init(ns);
+
+	ns->ethtool.pauseparam.report_stats_rx = true;
+	ns->ethtool.pauseparam.report_stats_tx = true;
+
+	ns->ethtool.fec.fec = ETHTOOL_FEC_NONE;
+	ns->ethtool.fec.active_fec = ETHTOOL_FEC_NONE;
+
+	ns->ethtool.channels = ns->nsim_bus_dev->num_queues;
+
+	ethtool = debugfs_create_dir("ethtool", ns->nsim_dev_port->ddir);
+	ns->ethtool_ddir = ethtool;
+
+	debugfs_create_u32("get_err", 0600, ethtool, &ns->ethtool.get_err);
+	debugfs_create_u32("set_err", 0600, ethtool, &ns->ethtool.set_err);
+
+	dir = debugfs_create_dir("pause", ethtool);
+	debugfs_create_bool("report_stats_rx", 0600, dir,
+			    &ns->ethtool.pauseparam.report_stats_rx);
+	debugfs_create_bool("report_stats_tx", 0600, dir,
+			    &ns->ethtool.pauseparam.report_stats_tx);
+
+	dir = debugfs_create_dir("ring", ethtool);
+	debugfs_create_u32("rx_max_pending", 0600, dir,
+			   &ns->ethtool.ring.rx_max_pending);
+	debugfs_create_u32("rx_jumbo_max_pending", 0600, dir,
+			   &ns->ethtool.ring.rx_jumbo_max_pending);
+	debugfs_create_u32("rx_mini_max_pending", 0600, dir,
+			   &ns->ethtool.ring.rx_mini_max_pending);
+	debugfs_create_u32("tx_max_pending", 0600, dir,
+			   &ns->ethtool.ring.tx_max_pending);
+}
+
+void nsim_ethtool_fini(struct netdevsim *ns)
+{
+	debugfs_remove(ns->ethtool_ddir);
+}