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, 236 insertions, 0 deletions

diff --git a/scripts/kconfig/Makefile b/scripts/kconfig/Makefile
new file mode 100644
index 000000000..5baf1c44f
--- /dev/null
+++ b/scripts/kconfig/Makefile
@@ -0,0 +1,236 @@
+# SPDX-License-Identifier: GPL-2.0
+# ===========================================================================
+# Kernel configuration targets
+# These targets are used from top-level makefile
+
+ifdef KBUILD_KCONFIG
+Kconfig := $(KBUILD_KCONFIG)
+else
+Kconfig := Kconfig
+endif
+
+ifndef KBUILD_DEFCONFIG
+KBUILD_DEFCONFIG := defconfig
+endif
+
+ifeq ($(quiet),silent_)
+silent := -s
+endif
+
+export KCONFIG_DEFCONFIG_LIST :=
+ifndef cross_compiling
+kernel-release := $(shell uname -r)
+KCONFIG_DEFCONFIG_LIST += \
+	/lib/modules/$(kernel-release)/.config \
+	/etc/kernel-config \
+	/boot/config-$(kernel-release)
+endif
+KCONFIG_DEFCONFIG_LIST += arch/$(SRCARCH)/configs/$(KBUILD_DEFCONFIG)
+
+ifneq ($(findstring c, $(KBUILD_EXTRA_WARN)),)
+export KCONFIG_WARN_UNKNOWN_SYMBOLS=1
+endif
+
+ifneq ($(findstring e, $(KBUILD_EXTRA_WARN)),)
+export KCONFIG_WERROR=1
+endif
+
+# We need this, in case the user has it in its environment
+unexport CONFIG_
+
+config-prog	:= conf
+menuconfig-prog	:= mconf
+nconfig-prog	:= nconf
+gconfig-prog	:= gconf
+xconfig-prog	:= qconf
+
+define config_rule
+PHONY += $(1)
+$(1): $(obj)/$($(1)-prog)
+	$(Q)$$< $(silent) $(Kconfig)
+
+PHONY += build_$(1)
+build_$(1): $(obj)/$($(1)-prog)
+endef
+
+$(foreach c, config menuconfig nconfig gconfig xconfig, $(eval $(call config_rule,$(c))))
+
+PHONY += localmodconfig localyesconfig
+localyesconfig localmodconfig: $(obj)/conf
+	$(Q)$(PERL) $(src)/streamline_config.pl --$@ $(srctree) $(Kconfig) > .tmp.config
+	$(Q)if [ -f .config ]; then 				\
+		cmp -s .tmp.config .config ||			\
+		(mv -f .config .config.old.1;			\
+		 mv -f .tmp.config .config;			\
+		 $< $(silent) --oldconfig $(Kconfig);		\
+		 mv -f .config.old.1 .config.old)		\
+	else							\
+		mv -f .tmp.config .config;			\
+		$< $(silent) --oldconfig $(Kconfig);		\
+	fi
+	$(Q)rm -f .tmp.config
+
+# These targets map 1:1 to the commandline options of 'conf'
+#
+# Note:
+#  syncconfig has become an internal implementation detail and is now
+#  deprecated for external use
+simple-targets := oldconfig allnoconfig allyesconfig allmodconfig \
+	alldefconfig randconfig listnewconfig olddefconfig syncconfig \
+	helpnewconfig yes2modconfig mod2yesconfig mod2noconfig
+
+PHONY += $(simple-targets)
+
+$(simple-targets): $(obj)/conf
+	$(Q)$< $(silent) --$@ $(Kconfig)
+
+PHONY += savedefconfig defconfig
+
+savedefconfig: $(obj)/conf
+	$(Q)$< $(silent) --$@=defconfig $(Kconfig)
+
+defconfig: $(obj)/conf
+ifneq ($(wildcard $(srctree)/arch/$(SRCARCH)/configs/$(KBUILD_DEFCONFIG)),)
+	@$(kecho) "*** Default configuration is based on '$(KBUILD_DEFCONFIG)'"
+	$(Q)$< $(silent) --defconfig=arch/$(SRCARCH)/configs/$(KBUILD_DEFCONFIG) $(Kconfig)
+else
+	@$(kecho) "*** Default configuration is based on target '$(KBUILD_DEFCONFIG)'"
+	$(Q)$(MAKE) -f $(srctree)/Makefile $(KBUILD_DEFCONFIG)
+endif
+
+%_defconfig: $(obj)/conf
+	$(Q)$< $(silent) --defconfig=arch/$(SRCARCH)/configs/$@ $(Kconfig)
+
+configfiles = $(wildcard $(srctree)/kernel/configs/$(1) $(srctree)/arch/$(SRCARCH)/configs/$(1))
+all-config-fragments = $(call configfiles,*.config)
+config-fragments = $(call configfiles,$@)
+
+cmd_merge_fragments = $(srctree)/scripts/kconfig/merge_config.sh -m $(KCONFIG_CONFIG) $(config-fragments)
+
+%.config: $(obj)/conf
+	$(if $(config-fragments),, $(error $@ fragment does not exists on this architecture))
+	$(call cmd,merge_fragments)
+	$(Q)$(MAKE) -f $(srctree)/Makefile olddefconfig
+
+PHONY += tinyconfig
+tinyconfig:
+	$(Q)KCONFIG_ALLCONFIG=kernel/configs/tiny-base.config $(MAKE) -f $(srctree)/Makefile allnoconfig
+	$(Q)$(MAKE) -f $(srctree)/Makefile tiny.config
+
+# CHECK: -o cache_dir=<path> working?
+PHONY += testconfig
+testconfig: $(obj)/conf
+	$(Q)$(PYTHON3) -B -m pytest $(src)/tests \
+	-o cache_dir=$(abspath $(obj)/tests/.cache) \
+	$(if $(findstring 1,$(KBUILD_VERBOSE)),--capture=no)
+clean-files += tests/.cache
+
+# Help text used by make help
+help:
+	@echo  'Configuration targets:'
+	@echo  '  config	  - Update current config utilising a line-oriented program'
+	@echo  '  nconfig         - Update current config utilising a ncurses menu based program'
+	@echo  '  menuconfig	  - Update current config utilising a menu based program'
+	@echo  '  xconfig	  - Update current config utilising a Qt based front-end'
+	@echo  '  gconfig	  - Update current config utilising a GTK+ based front-end'
+	@echo  '  oldconfig	  - Update current config utilising a provided .config as base'
+	@echo  '  localmodconfig  - Update current config disabling modules not loaded'
+	@echo  '                    except those preserved by LMC_KEEP environment variable'
+	@echo  '  localyesconfig  - Update current config converting local mods to core'
+	@echo  '                    except those preserved by LMC_KEEP environment variable'
+	@echo  '  defconfig	  - New config with default from ARCH supplied defconfig'
+	@echo  '  savedefconfig   - Save current config as ./defconfig (minimal config)'
+	@echo  '  allnoconfig	  - New config where all options are answered with no'
+	@echo  '  allyesconfig	  - New config where all options are accepted with yes'
+	@echo  '  allmodconfig	  - New config selecting modules when possible'
+	@echo  '  alldefconfig    - New config with all symbols set to default'
+	@echo  '  randconfig	  - New config with random answer to all options'
+	@echo  '  yes2modconfig	  - Change answers from yes to mod if possible'
+	@echo  '  mod2yesconfig	  - Change answers from mod to yes if possible'
+	@echo  '  mod2noconfig	  - Change answers from mod to no if possible'
+	@echo  '  listnewconfig   - List new options'
+	@echo  '  helpnewconfig   - List new options and help text'
+	@echo  '  olddefconfig	  - Same as oldconfig but sets new symbols to their'
+	@echo  '                    default value without prompting'
+	@echo  '  tinyconfig	  - Configure the tiniest possible kernel'
+	@echo  '  testconfig	  - Run Kconfig unit tests (requires python3 and pytest)'
+	@echo  ''
+	@echo  'Configuration topic targets:'
+	@$(foreach f, $(all-config-fragments), \
+		if help=$$(grep -m1 '^# Help: ' $(f)); then \
+			printf '  %-25s - %s\n' '$(notdir $(f))' "$${help#*: }"; \
+		fi;)
+
+# ===========================================================================
+# object files used by all kconfig flavours
+common-objs	:= confdata.o expr.o lexer.lex.o menu.o parser.tab.o \
+		   preprocess.o symbol.o util.o
+
+$(obj)/lexer.lex.o: $(obj)/parser.tab.h
+HOSTCFLAGS_lexer.lex.o	:= -I $(src)
+HOSTCFLAGS_parser.tab.o	:= -I $(src)
+
+# conf: Used for defconfig, oldconfig and related targets
+hostprogs	+= conf
+conf-objs	:= conf.o $(common-objs)
+
+# nconf: Used for the nconfig target based on ncurses
+hostprogs	+= nconf
+nconf-objs	:= nconf.o nconf.gui.o mnconf-common.o $(common-objs)
+
+HOSTLDLIBS_nconf       = $(call read-file, $(obj)/nconf-libs)
+HOSTCFLAGS_nconf.o     = $(call read-file, $(obj)/nconf-cflags)
+HOSTCFLAGS_nconf.gui.o = $(call read-file, $(obj)/nconf-cflags)
+
+$(obj)/nconf: | $(obj)/nconf-libs
+$(obj)/nconf.o $(obj)/nconf.gui.o: | $(obj)/nconf-cflags
+
+# mconf: Used for the menuconfig target based on lxdialog
+hostprogs	+= mconf
+lxdialog	:= $(addprefix lxdialog/, \
+		     checklist.o inputbox.o menubox.o textbox.o util.o yesno.o)
+mconf-objs	:= mconf.o $(lxdialog) mnconf-common.o $(common-objs)
+
+HOSTLDLIBS_mconf = $(call read-file, $(obj)/mconf-libs)
+$(foreach f, mconf.o $(lxdialog), \
+  $(eval HOSTCFLAGS_$f = $$(call read-file, $(obj)/mconf-cflags)))
+
+$(obj)/mconf: | $(obj)/mconf-libs
+$(addprefix $(obj)/, mconf.o $(lxdialog)): | $(obj)/mconf-cflags
+
+# qconf: Used for the xconfig target based on Qt
+hostprogs	+= qconf
+qconf-cxxobjs	:= qconf.o qconf-moc.o
+qconf-objs	:= $(common-objs)
+
+HOSTLDLIBS_qconf         = $(call read-file, $(obj)/qconf-libs)
+HOSTCXXFLAGS_qconf.o     = -std=c++11 -fPIC $(call read-file, $(obj)/qconf-cflags)
+HOSTCXXFLAGS_qconf-moc.o = -std=c++11 -fPIC $(call read-file, $(obj)/qconf-cflags)
+$(obj)/qconf: | $(obj)/qconf-libs
+$(obj)/qconf.o $(obj)/qconf-moc.o: | $(obj)/qconf-cflags
+
+quiet_cmd_moc = MOC     $@
+      cmd_moc = $(call read-file, $(obj)/qconf-bin)/moc $< -o $@
+
+$(obj)/qconf-moc.cc: $(src)/qconf.h FORCE | $(obj)/qconf-bin
+	$(call if_changed,moc)
+
+targets += qconf-moc.cc
+
+# gconf: Used for the gconfig target based on GTK+
+hostprogs	+= gconf
+gconf-objs	:= gconf.o $(common-objs)
+
+HOSTLDLIBS_gconf   = $(call read-file, $(obj)/gconf-libs)
+HOSTCFLAGS_gconf.o = $(call read-file, $(obj)/gconf-cflags)
+
+$(obj)/gconf: | $(obj)/gconf-libs
+$(obj)/gconf.o: | $(obj)/gconf-cflags
+
+# check if necessary packages are available, and configure build flags
+cmd_conf_cfg = $< $(addprefix $(obj)/$*conf-, cflags libs bin); touch $(obj)/$*conf-bin
+
+$(obj)/%conf-cflags $(obj)/%conf-libs $(obj)/%conf-bin: $(src)/%conf-cfg.sh
+	$(call cmd,conf_cfg)
+
+clean-files += *conf-cflags *conf-libs *conf-bin