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, 187 insertions, 0 deletions

diff --git a/Documentation/networking/operstates.rst b/Documentation/networking/operstates.rst
new file mode 100644
index 000000000..1ee2141e8
--- /dev/null
+++ b/Documentation/networking/operstates.rst
@@ -0,0 +1,187 @@
+.. SPDX-License-Identifier: GPL-2.0
+
+==================
+Operational States
+==================
+
+
+1. Introduction
+===============
+
+Linux distinguishes between administrative and operational state of an
+interface. Administrative state is the result of "ip link set dev
+<dev> up or down" and reflects whether the administrator wants to use
+the device for traffic.
+
+However, an interface is not usable just because the admin enabled it
+- ethernet requires to be plugged into the switch and, depending on
+a site's networking policy and configuration, an 802.1X authentication
+to be performed before user data can be transferred. Operational state
+shows the ability of an interface to transmit this user data.
+
+Thanks to 802.1X, userspace must be granted the possibility to
+influence operational state. To accommodate this, operational state is
+split into two parts: Two flags that can be set by the driver only, and
+a RFC2863 compatible state that is derived from these flags, a policy,
+and changeable from userspace under certain rules.
+
+
+2. Querying from userspace
+==========================
+
+Both admin and operational state can be queried via the netlink
+operation RTM_GETLINK. It is also possible to subscribe to RTNLGRP_LINK
+to be notified of updates while the interface is admin up. This is
+important for setting from userspace.
+
+These values contain interface state:
+
+ifinfomsg::if_flags & IFF_UP:
+ Interface is admin up
+
+ifinfomsg::if_flags & IFF_RUNNING:
+ Interface is in RFC2863 operational state UP or UNKNOWN. This is for
+ backward compatibility, routing daemons, dhcp clients can use this
+ flag to determine whether they should use the interface.
+
+ifinfomsg::if_flags & IFF_LOWER_UP:
+ Driver has signaled netif_carrier_on()
+
+ifinfomsg::if_flags & IFF_DORMANT:
+ Driver has signaled netif_dormant_on()
+
+TLV IFLA_OPERSTATE
+------------------
+
+contains RFC2863 state of the interface in numeric representation:
+
+IF_OPER_UNKNOWN (0):
+ Interface is in unknown state, neither driver nor userspace has set
+ operational state. Interface must be considered for user data as
+ setting operational state has not been implemented in every driver.
+
+IF_OPER_NOTPRESENT (1):
+ Unused in current kernel (notpresent interfaces normally disappear),
+ just a numerical placeholder.
+
+IF_OPER_DOWN (2):
+ Interface is unable to transfer data on L1, f.e. ethernet is not
+ plugged or interface is ADMIN down.
+
+IF_OPER_LOWERLAYERDOWN (3):
+ Interfaces stacked on an interface that is IF_OPER_DOWN show this
+ state (f.e. VLAN).
+
+IF_OPER_TESTING (4):
+ Interface is in testing mode, for example executing driver self-tests
+ or media (cable) test. It can't be used for normal traffic until tests
+ complete.
+
+IF_OPER_DORMANT (5):
+ Interface is L1 up, but waiting for an external event, f.e. for a
+ protocol to establish. (802.1X)
+
+IF_OPER_UP (6):
+ Interface is operational up and can be used.
+
+This TLV can also be queried via sysfs.
+
+TLV IFLA_LINKMODE
+-----------------
+
+contains link policy. This is needed for userspace interaction
+described below.
+
+This TLV can also be queried via sysfs.
+
+
+3. Kernel driver API
+====================
+
+Kernel drivers have access to two flags that map to IFF_LOWER_UP and
+IFF_DORMANT. These flags can be set from everywhere, even from
+interrupts. It is guaranteed that only the driver has write access,
+however, if different layers of the driver manipulate the same flag,
+the driver has to provide the synchronisation needed.
+
+__LINK_STATE_NOCARRIER, maps to !IFF_LOWER_UP:
+
+The driver uses netif_carrier_on() to clear and netif_carrier_off() to
+set this flag. On netif_carrier_off(), the scheduler stops sending
+packets. The name 'carrier' and the inversion are historical, think of
+it as lower layer.
+
+Note that for certain kind of soft-devices, which are not managing any
+real hardware, it is possible to set this bit from userspace.  One
+should use TLV IFLA_CARRIER to do so.
+
+netif_carrier_ok() can be used to query that bit.
+
+__LINK_STATE_DORMANT, maps to IFF_DORMANT:
+
+Set by the driver to express that the device cannot yet be used
+because some driver controlled protocol establishment has to
+complete. Corresponding functions are netif_dormant_on() to set the
+flag, netif_dormant_off() to clear it and netif_dormant() to query.
+
+On device allocation, both flags __LINK_STATE_NOCARRIER and
+__LINK_STATE_DORMANT are cleared, so the effective state is equivalent
+to netif_carrier_ok() and !netif_dormant().
+
+
+Whenever the driver CHANGES one of these flags, a workqueue event is
+scheduled to translate the flag combination to IFLA_OPERSTATE as
+follows:
+
+!netif_carrier_ok():
+ IF_OPER_LOWERLAYERDOWN if the interface is stacked, IF_OPER_DOWN
+ otherwise. Kernel can recognise stacked interfaces because their
+ ifindex != iflink.
+
+netif_carrier_ok() && netif_dormant():
+ IF_OPER_DORMANT
+
+netif_carrier_ok() && !netif_dormant():
+ IF_OPER_UP if userspace interaction is disabled. Otherwise
+ IF_OPER_DORMANT with the possibility for userspace to initiate the
+ IF_OPER_UP transition afterwards.
+
+
+4. Setting from userspace
+=========================
+
+Applications have to use the netlink interface to influence the
+RFC2863 operational state of an interface. Setting IFLA_LINKMODE to 1
+via RTM_SETLINK instructs the kernel that an interface should go to
+IF_OPER_DORMANT instead of IF_OPER_UP when the combination
+netif_carrier_ok() && !netif_dormant() is set by the
+driver. Afterwards, the userspace application can set IFLA_OPERSTATE
+to IF_OPER_DORMANT or IF_OPER_UP as long as the driver does not set
+netif_carrier_off() or netif_dormant_on(). Changes made by userspace
+are multicasted on the netlink group RTNLGRP_LINK.
+
+So basically a 802.1X supplicant interacts with the kernel like this:
+
+- subscribe to RTNLGRP_LINK
+- set IFLA_LINKMODE to 1 via RTM_SETLINK
+- query RTM_GETLINK once to get initial state
+- if initial flags are not (IFF_LOWER_UP && !IFF_DORMANT), wait until
+  netlink multicast signals this state
+- do 802.1X, eventually abort if flags go down again
+- send RTM_SETLINK to set operstate to IF_OPER_UP if authentication
+  succeeds, IF_OPER_DORMANT otherwise
+- see how operstate and IFF_RUNNING is echoed via netlink multicast
+- set interface back to IF_OPER_DORMANT if 802.1X reauthentication
+  fails
+- restart if kernel changes IFF_LOWER_UP or IFF_DORMANT flag
+
+if supplicant goes down, bring back IFLA_LINKMODE to 0 and
+IFLA_OPERSTATE to a sane value.
+
+A routing daemon or dhcp client just needs to care for IFF_RUNNING or
+waiting for operstate to go IF_OPER_UP/IF_OPER_UNKNOWN before
+considering the interface / querying a DHCP address.
+
+
+For technical questions and/or comments please e-mail to Stefan Rompf
+(stefan at loplof.de).