[PATCH v2 0/3] arm64: kprobes: Fix bugs in kprobes for arm64

[PATCH v2 0/3] arm64: kprobes: Fix bugs in kprobes for arm64

[Masami Hiramatsu (Google)]

Hi,

Here is the 2nd version of the kprobe bugfixes on arm64. The previous
series is here;

https://lore.kernel.org/all/166990553243.253128.13594802750635478633.stgit@devnote3/T/#u

The first one is a critical issue, which will cause a kernel crach
easily with lockdep[1/3]. Others are minor issues and rare cases.
The [2/3] let do_page_fault() fixup the page fault in kprobes user
handler, and [3/3] is more like code cleanup and returns
DBG_HOOK_ERROR if it can not handle kprobe's BRK (but that
should not happen.)

In this version I update to use noinstr and __always_inline [1/3]
and update comments[3/3]. Also add Mark's Ack (Thanks!)

Thank you,

---

Masami Hiramatsu (Google) (3):
      arm64: Prohibit instrumentation on arch_stack_walk()
      arm64: kprobes: Let arch do_page_fault() fix up page fault in user handler
      arm64: kprobes: Return DBG_HOOK_ERROR if kprobes can not handle a BRK


 arch/arm64/kernel/probes/kprobes.c |   86 +++++++++++++++---------------------
 arch/arm64/kernel/stacktrace.c     |   10 ++--
 2 files changed, 41 insertions(+), 55 deletions(-)

--
Masami Hiramatsu (Google) <mhiramat@kernel.org>

_______________________________________________
linux-arm-kernel mailing list
linux-arm-kernel@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel

[PATCH v2 1/3] arm64: Prohibit instrumentation on arch_stack_walk()

[Masami Hiramatsu (Google)]

From: Masami Hiramatsu (Google) <mhiramat@kernel.org>

Mark arch_stack_walk() as noinstr instead of notrace and inline functions
called from arch_stack_walk() as __always_inline so that user does not
put any instrumentations on it, because this function can be used from
return_address() which is used by lockdep.

Without this, if the kernel built with CONFIG_LOCKDEP=y, just probing
arch_stack_walk() via <tracefs>/kprobe_events will crash the kernel on
arm64.

 # echo p arch_stack_walk >> ${TRACEFS}/kprobe_events
 # echo 1 > ${TRACEFS}/events/kprobes/enable
  kprobes: Failed to recover from reentered kprobes.
  kprobes: Dump kprobe:
  .symbol_name = arch_stack_walk, .offset = 0, .addr = arch_stack_walk+0x0/0x1c0
  ------------[ cut here ]------------
  kernel BUG at arch/arm64/kernel/probes/kprobes.c:241!
  kprobes: Failed to recover from reentered kprobes.
  kprobes: Dump kprobe:
  .symbol_name = arch_stack_walk, .offset = 0, .addr = arch_stack_walk+0x0/0x1c0
  ------------[ cut here ]------------
  kernel BUG at arch/arm64/kernel/probes/kprobes.c:241!
  PREEMPT SMP
  Modules linked in:
  CPU: 0 PID: 17 Comm: migration/0 Tainted: G                 N 6.1.0-rc5+ #6
  Hardware name: linux,dummy-virt (DT)
  Stopper: 0x0 <- 0x0
  pstate: 600003c5 (nZCv DAIF -PAN -UAO -TCO -DIT -SSBS BTYPE=--)
  pc : kprobe_breakpoint_handler+0x178/0x17c
  lr : kprobe_breakpoint_handler+0x178/0x17c
  sp : ffff8000080d3090
  x29: ffff8000080d3090 x28: ffff0df5845798c0 x27: ffffc4f59057a774
  x26: ffff0df5ffbba770 x25: ffff0df58f420f18 x24: ffff49006f641000
  x23: ffffc4f590579768 x22: ffff0df58f420f18 x21: ffff8000080d31c0
  x20: ffffc4f590579768 x19: ffffc4f590579770 x18: 0000000000000006
  x17: 5f6b636174735f68 x16: 637261203d207264 x15: 64612e202c30203d
  x14: 2074657366666f2e x13: 30633178302f3078 x12: 302b6b6c61775f6b
  x11: 636174735f686372 x10: ffffc4f590dc5bd8 x9 : ffffc4f58eb31958
  x8 : 00000000ffffefff x7 : ffffc4f590dc5bd8 x6 : 80000000fffff000
  x5 : 000000000000bff4 x4 : 0000000000000000 x3 : 0000000000000000
  x2 : 0000000000000000 x1 : ffff0df5845798c0 x0 : 0000000000000064
  Call trace:
  kprobes: Failed to recover from reentered kprobes.
  kprobes: Dump kprobe:
  .symbol_name = arch_stack_walk, .offset = 0, .addr = arch_stack_walk+0x0/0x1c0
  ------------[ cut here ]------------
  kernel BUG at arch/arm64/kernel/probes/kprobes.c:241!

Fixes: 39ef362d2d45 ("arm64: Make return_address() use arch_stack_walk()")
Cc: stable@vger.kernel.org
Signed-off-by: Masami Hiramatsu (Google) <mhiramat@kernel.org>
---
 Changes in v2:
  - Use noinstr instead of NOKPROBE_SYMBOL()
  - Use __always_inline because nokprobe_inline will be changed if
    CONFIG_KPROBES=n.
  - Fix indentation.
---
 arch/arm64/kernel/stacktrace.c |   10 +++++-----
 1 file changed, 5 insertions(+), 5 deletions(-)

diff --git a/arch/arm64/kernel/stacktrace.c b/arch/arm64/kernel/stacktrace.c
index 634279b3b03d..117e2c180f3c 100644
--- a/arch/arm64/kernel/stacktrace.c
+++ b/arch/arm64/kernel/stacktrace.c
@@ -23,8 +23,8 @@
  *
  * The regs must be on a stack currently owned by the calling task.
  */
-static inline void unwind_init_from_regs(struct unwind_state *state,
-					 struct pt_regs *regs)
+static __always_inline void unwind_init_from_regs(struct unwind_state *state,
+						  struct pt_regs *regs)
 {
 	unwind_init_common(state, current);
 
@@ -58,8 +58,8 @@ static __always_inline void unwind_init_from_caller(struct unwind_state *state)
  * duration of the unwind, or the unwind will be bogus. It is never valid to
  * call this for the current task.
  */
-static inline void unwind_init_from_task(struct unwind_state *state,
-					 struct task_struct *task)
+static __always_inline void unwind_init_from_task(struct unwind_state *state,
+						  struct task_struct *task)
 {
 	unwind_init_common(state, task);
 
@@ -186,7 +186,7 @@ void show_stack(struct task_struct *tsk, unsigned long *sp, const char *loglvl)
 			: stackinfo_get_unknown();		\
 	})
 
-noinline notrace void arch_stack_walk(stack_trace_consume_fn consume_entry,
+noinline noinstr void arch_stack_walk(stack_trace_consume_fn consume_entry,
 			      void *cookie, struct task_struct *task,
 			      struct pt_regs *regs)
 {


_______________________________________________
linux-arm-kernel mailing list
linux-arm-kernel@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel

Re: [PATCH v2 1/3] arm64: Prohibit instrumentation on arch_stack_walk()

[Mark Rutland]

On Fri, Dec 02, 2022 at 11:18:33AM +0900, Masami Hiramatsu (Google) wrote:
> From: Masami Hiramatsu (Google) <mhiramat@kernel.org>
> 
> Mark arch_stack_walk() as noinstr instead of notrace and inline functions
> called from arch_stack_walk() as __always_inline so that user does not
> put any instrumentations on it, because this function can be used from
> return_address() which is used by lockdep.

Hmmm... since arch_stack_walk is marked as notrace, that will be prohibited by
default unless the kernel was built with CONFIG_KPROBE_EVENTS_ON_NOTRACE=y,
and the commit message for that says:

│ This option disables such protection and allows you to put kprobe                                                                                                                                   │
│ events on ftrace functions for debugging ftrace by itself.                                                                                                                                          │
│ Note that this might let you shoot yourself in the foot.

... and IIUC we generally don't expect people to set that, and that might be
worth calling out in the commit message.

> Without this, if the kernel built with CONFIG_LOCKDEP=y, just probing
> arch_stack_walk() via <tracefs>/kprobe_events will crash the kernel on
> arm64.
> 
>  # echo p arch_stack_walk >> ${TRACEFS}/kprobe_events
>  # echo 1 > ${TRACEFS}/events/kprobes/enable

I had a go at testing this patch, and it fixes the crash with the reproducer
above, but there are plenty of other instances in stacktrace.c that lead to the
same sort of crash, e.g.

# echo p stackinfo_get_task >> ${TRACEFS}/kprobe_events
# echo 1 > ${TRACEFS}/events/kprobes/enable

... so I think there's more that we need to do to fix this generally.

Note: I found interesting functions to try tracing by looking at the output of:

  aarch64-linux-objdump -t arch/arm64/kernel/stacktrace.o | grep -w '.text'

That all said, I think this patch is nice-to-have, and that we can address the
other cases as a follow-up, so for this patch as-is (with or without some
wording regarding CONFIG_KPROBE_EVENTS_ON_NOTRACE):

Acked-by: Mark Rutland <mark.rutland@arm.com>

Mark.

>   kprobes: Failed to recover from reentered kprobes.
>   kprobes: Dump kprobe:
>   .symbol_name = arch_stack_walk, .offset = 0, .addr = arch_stack_walk+0x0/0x1c0
>   ------------[ cut here ]------------
>   kernel BUG at arch/arm64/kernel/probes/kprobes.c:241!
>   kprobes: Failed to recover from reentered kprobes.
>   kprobes: Dump kprobe:
>   .symbol_name = arch_stack_walk, .offset = 0, .addr = arch_stack_walk+0x0/0x1c0
>   ------------[ cut here ]------------
>   kernel BUG at arch/arm64/kernel/probes/kprobes.c:241!
>   PREEMPT SMP
>   Modules linked in:
>   CPU: 0 PID: 17 Comm: migration/0 Tainted: G                 N 6.1.0-rc5+ #6
>   Hardware name: linux,dummy-virt (DT)
>   Stopper: 0x0 <- 0x0
>   pstate: 600003c5 (nZCv DAIF -PAN -UAO -TCO -DIT -SSBS BTYPE=--)
>   pc : kprobe_breakpoint_handler+0x178/0x17c
>   lr : kprobe_breakpoint_handler+0x178/0x17c
>   sp : ffff8000080d3090
>   x29: ffff8000080d3090 x28: ffff0df5845798c0 x27: ffffc4f59057a774
>   x26: ffff0df5ffbba770 x25: ffff0df58f420f18 x24: ffff49006f641000
>   x23: ffffc4f590579768 x22: ffff0df58f420f18 x21: ffff8000080d31c0
>   x20: ffffc4f590579768 x19: ffffc4f590579770 x18: 0000000000000006
>   x17: 5f6b636174735f68 x16: 637261203d207264 x15: 64612e202c30203d
>   x14: 2074657366666f2e x13: 30633178302f3078 x12: 302b6b6c61775f6b
>   x11: 636174735f686372 x10: ffffc4f590dc5bd8 x9 : ffffc4f58eb31958
>   x8 : 00000000ffffefff x7 : ffffc4f590dc5bd8 x6 : 80000000fffff000
>   x5 : 000000000000bff4 x4 : 0000000000000000 x3 : 0000000000000000
>   x2 : 0000000000000000 x1 : ffff0df5845798c0 x0 : 0000000000000064
>   Call trace:
>   kprobes: Failed to recover from reentered kprobes.
>   kprobes: Dump kprobe:
>   .symbol_name = arch_stack_walk, .offset = 0, .addr = arch_stack_walk+0x0/0x1c0
>   ------------[ cut here ]------------
>   kernel BUG at arch/arm64/kernel/probes/kprobes.c:241!
> 
> Fixes: 39ef362d2d45 ("arm64: Make return_address() use arch_stack_walk()")
> Cc: stable@vger.kernel.org
> Signed-off-by: Masami Hiramatsu (Google) <mhiramat@kernel.org>
> ---
>  Changes in v2:
>   - Use noinstr instead of NOKPROBE_SYMBOL()
>   - Use __always_inline because nokprobe_inline will be changed if
>     CONFIG_KPROBES=n.
>   - Fix indentation.
> ---
>  arch/arm64/kernel/stacktrace.c |   10 +++++-----
>  1 file changed, 5 insertions(+), 5 deletions(-)
> 
> diff --git a/arch/arm64/kernel/stacktrace.c b/arch/arm64/kernel/stacktrace.c
> index 634279b3b03d..117e2c180f3c 100644
> --- a/arch/arm64/kernel/stacktrace.c
> +++ b/arch/arm64/kernel/stacktrace.c
> @@ -23,8 +23,8 @@
>   *
>   * The regs must be on a stack currently owned by the calling task.
>   */
> -static inline void unwind_init_from_regs(struct unwind_state *state,
> -					 struct pt_regs *regs)
> +static __always_inline void unwind_init_from_regs(struct unwind_state *state,
> +						  struct pt_regs *regs)
>  {
>  	unwind_init_common(state, current);
>  
> @@ -58,8 +58,8 @@ static __always_inline void unwind_init_from_caller(struct unwind_state *state)
>   * duration of the unwind, or the unwind will be bogus. It is never valid to
>   * call this for the current task.
>   */
> -static inline void unwind_init_from_task(struct unwind_state *state,
> -					 struct task_struct *task)
> +static __always_inline void unwind_init_from_task(struct unwind_state *state,
> +						  struct task_struct *task)
>  {
>  	unwind_init_common(state, task);
>  
> @@ -186,7 +186,7 @@ void show_stack(struct task_struct *tsk, unsigned long *sp, const char *loglvl)
>  			: stackinfo_get_unknown();		\
>  	})
>  
> -noinline notrace void arch_stack_walk(stack_trace_consume_fn consume_entry,
> +noinline noinstr void arch_stack_walk(stack_trace_consume_fn consume_entry,
>  			      void *cookie, struct task_struct *task,
>  			      struct pt_regs *regs)
>  {
> 

_______________________________________________
linux-arm-kernel mailing list
linux-arm-kernel@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel

Re: [PATCH v2 1/3] arm64: Prohibit instrumentation on arch_stack_walk()

[Masami Hiramatsu]

On Fri, 2 Dec 2022 13:17:30 +0000
Mark Rutland <mark.rutland@arm.com> wrote:

> On Fri, Dec 02, 2022 at 11:18:33AM +0900, Masami Hiramatsu (Google) wrote:
> > From: Masami Hiramatsu (Google) <mhiramat@kernel.org>
> > 
> > Mark arch_stack_walk() as noinstr instead of notrace and inline functions
> > called from arch_stack_walk() as __always_inline so that user does not
> > put any instrumentations on it, because this function can be used from
> > return_address() which is used by lockdep.
> 
> Hmmm... since arch_stack_walk is marked as notrace, that will be prohibited by
> default unless the kernel was built with CONFIG_KPROBE_EVENTS_ON_NOTRACE=y,
> and the commit message for that says:
> 
> │ This option disables such protection and allows you to put kprobe                                                                                                                                   │
> │ events on ftrace functions for debugging ftrace by itself.                                                                                                                                          │
> │ Note that this might let you shoot yourself in the foot.
> 
> ... and IIUC we generally don't expect people to set that, and that might be
> worth calling out in the commit message.

Yeah, but that is the feature for kprobe events, and not applied to
the kprobes itself. Maybe I should use a test kprobe module to
put a probe on it.

> 
> > Without this, if the kernel built with CONFIG_LOCKDEP=y, just probing
> > arch_stack_walk() via <tracefs>/kprobe_events will crash the kernel on
> > arm64.
> > 
> >  # echo p arch_stack_walk >> ${TRACEFS}/kprobe_events
> >  # echo 1 > ${TRACEFS}/events/kprobes/enable
> 
> I had a go at testing this patch, and it fixes the crash with the reproducer
> above, but there are plenty of other instances in stacktrace.c that lead to the
> same sort of crash, e.g.
> 
> # echo p stackinfo_get_task >> ${TRACEFS}/kprobe_events
> # echo 1 > ${TRACEFS}/events/kprobes/enable

Oops, thanks for pointing! Hmm, I thought stackinfo_get_task() is an
inlined function usually. Maybe we should make it nokprobe_inline.
But this is just one case. I need to scan all symbols to trace...

> 
> ... so I think there's more that we need to do to fix this generally.
> 
> Note: I found interesting functions to try tracing by looking at the output of:
> 
>   aarch64-linux-objdump -t arch/arm64/kernel/stacktrace.o | grep -w '.text'
> 
> That all said, I think this patch is nice-to-have, and that we can address the
> other cases as a follow-up, so for this patch as-is (with or without some
> wording regarding CONFIG_KPROBE_EVENTS_ON_NOTRACE):
> 
> Acked-by: Mark Rutland <mark.rutland@arm.com>

Thanks!

> 
> Mark.
> 
> >   kprobes: Failed to recover from reentered kprobes.
> >   kprobes: Dump kprobe:
> >   .symbol_name = arch_stack_walk, .offset = 0, .addr = arch_stack_walk+0x0/0x1c0
> >   ------------[ cut here ]------------
> >   kernel BUG at arch/arm64/kernel/probes/kprobes.c:241!
> >   kprobes: Failed to recover from reentered kprobes.
> >   kprobes: Dump kprobe:
> >   .symbol_name = arch_stack_walk, .offset = 0, .addr = arch_stack_walk+0x0/0x1c0
> >   ------------[ cut here ]------------
> >   kernel BUG at arch/arm64/kernel/probes/kprobes.c:241!
> >   PREEMPT SMP
> >   Modules linked in:
> >   CPU: 0 PID: 17 Comm: migration/0 Tainted: G                 N 6.1.0-rc5+ #6
> >   Hardware name: linux,dummy-virt (DT)
> >   Stopper: 0x0 <- 0x0
> >   pstate: 600003c5 (nZCv DAIF -PAN -UAO -TCO -DIT -SSBS BTYPE=--)
> >   pc : kprobe_breakpoint_handler+0x178/0x17c
> >   lr : kprobe_breakpoint_handler+0x178/0x17c
> >   sp : ffff8000080d3090
> >   x29: ffff8000080d3090 x28: ffff0df5845798c0 x27: ffffc4f59057a774
> >   x26: ffff0df5ffbba770 x25: ffff0df58f420f18 x24: ffff49006f641000
> >   x23: ffffc4f590579768 x22: ffff0df58f420f18 x21: ffff8000080d31c0
> >   x20: ffffc4f590579768 x19: ffffc4f590579770 x18: 0000000000000006
> >   x17: 5f6b636174735f68 x16: 637261203d207264 x15: 64612e202c30203d
> >   x14: 2074657366666f2e x13: 30633178302f3078 x12: 302b6b6c61775f6b
> >   x11: 636174735f686372 x10: ffffc4f590dc5bd8 x9 : ffffc4f58eb31958
> >   x8 : 00000000ffffefff x7 : ffffc4f590dc5bd8 x6 : 80000000fffff000
> >   x5 : 000000000000bff4 x4 : 0000000000000000 x3 : 0000000000000000
> >   x2 : 0000000000000000 x1 : ffff0df5845798c0 x0 : 0000000000000064
> >   Call trace:
> >   kprobes: Failed to recover from reentered kprobes.
> >   kprobes: Dump kprobe:
> >   .symbol_name = arch_stack_walk, .offset = 0, .addr = arch_stack_walk+0x0/0x1c0
> >   ------------[ cut here ]------------
> >   kernel BUG at arch/arm64/kernel/probes/kprobes.c:241!
> > 
> > Fixes: 39ef362d2d45 ("arm64: Make return_address() use arch_stack_walk()")
> > Cc: stable@vger.kernel.org
> > Signed-off-by: Masami Hiramatsu (Google) <mhiramat@kernel.org>
> > ---
> >  Changes in v2:
> >   - Use noinstr instead of NOKPROBE_SYMBOL()
> >   - Use __always_inline because nokprobe_inline will be changed if
> >     CONFIG_KPROBES=n.
> >   - Fix indentation.
> > ---
> >  arch/arm64/kernel/stacktrace.c |   10 +++++-----
> >  1 file changed, 5 insertions(+), 5 deletions(-)
> > 
> > diff --git a/arch/arm64/kernel/stacktrace.c b/arch/arm64/kernel/stacktrace.c
> > index 634279b3b03d..117e2c180f3c 100644
> > --- a/arch/arm64/kernel/stacktrace.c
> > +++ b/arch/arm64/kernel/stacktrace.c
> > @@ -23,8 +23,8 @@
> >   *
> >   * The regs must be on a stack currently owned by the calling task.
> >   */
> > -static inline void unwind_init_from_regs(struct unwind_state *state,
> > -					 struct pt_regs *regs)
> > +static __always_inline void unwind_init_from_regs(struct unwind_state *state,
> > +						  struct pt_regs *regs)
> >  {
> >  	unwind_init_common(state, current);
> >  
> > @@ -58,8 +58,8 @@ static __always_inline void unwind_init_from_caller(struct unwind_state *state)
> >   * duration of the unwind, or the unwind will be bogus. It is never valid to
> >   * call this for the current task.
> >   */
> > -static inline void unwind_init_from_task(struct unwind_state *state,
> > -					 struct task_struct *task)
> > +static __always_inline void unwind_init_from_task(struct unwind_state *state,
> > +						  struct task_struct *task)
> >  {
> >  	unwind_init_common(state, task);
> >  
> > @@ -186,7 +186,7 @@ void show_stack(struct task_struct *tsk, unsigned long *sp, const char *loglvl)
> >  			: stackinfo_get_unknown();		\
> >  	})
> >  
> > -noinline notrace void arch_stack_walk(stack_trace_consume_fn consume_entry,
> > +noinline noinstr void arch_stack_walk(stack_trace_consume_fn consume_entry,
> >  			      void *cookie, struct task_struct *task,
> >  			      struct pt_regs *regs)
> >  {
> > 


-- 
Masami Hiramatsu (Google) <mhiramat@kernel.org>

_______________________________________________
linux-arm-kernel mailing list
linux-arm-kernel@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel

Re: [PATCH v2 1/3] arm64: Prohibit instrumentation on arch_stack_walk()

[Mark Rutland]

On Tue, Dec 06, 2022 at 09:41:37AM +0900, Masami Hiramatsu wrote:
> On Fri, 2 Dec 2022 13:17:30 +0000
> Mark Rutland <mark.rutland@arm.com> wrote:
> > I had a go at testing this patch, and it fixes the crash with the reproducer
> > above, but there are plenty of other instances in stacktrace.c that lead to the
> > same sort of crash, e.g.
> > 
> > # echo p stackinfo_get_task >> ${TRACEFS}/kprobe_events
> > # echo 1 > ${TRACEFS}/events/kprobes/enable
> 
> Oops, thanks for pointing! Hmm, I thought stackinfo_get_task() is an
> inlined function usually. Maybe we should make it nokprobe_inline.
> But this is just one case. I need to scan all symbols to trace...

FWIW, due to other instrumentation issues I've started doing a larger noinstr
cleanup on arm64 which should address this, as fixing this properly involves
also modifying a bunch of underlying helpers (e.g. preempt_count()).

I can Cc you when sending that out, if you'd like? That'll probably be in the
new year.

Thanks,
Mark.

_______________________________________________
linux-arm-kernel mailing list
linux-arm-kernel@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel

Re: [PATCH v2 1/3] arm64: Prohibit instrumentation on arch_stack_walk()

[Masami Hiramatsu]

On Fri, 9 Dec 2022 11:16:34 +0000
Mark Rutland <mark.rutland@arm.com> wrote:

> On Tue, Dec 06, 2022 at 09:41:37AM +0900, Masami Hiramatsu wrote:
> > On Fri, 2 Dec 2022 13:17:30 +0000
> > Mark Rutland <mark.rutland@arm.com> wrote:
> > > I had a go at testing this patch, and it fixes the crash with the reproducer
> > > above, but there are plenty of other instances in stacktrace.c that lead to the
> > > same sort of crash, e.g.
> > > 
> > > # echo p stackinfo_get_task >> ${TRACEFS}/kprobe_events
> > > # echo 1 > ${TRACEFS}/events/kprobes/enable
> > 
> > Oops, thanks for pointing! Hmm, I thought stackinfo_get_task() is an
> > inlined function usually. Maybe we should make it nokprobe_inline.
> > But this is just one case. I need to scan all symbols to trace...
> 
> FWIW, due to other instrumentation issues I've started doing a larger noinstr
> cleanup on arm64 which should address this, as fixing this properly involves
> also modifying a bunch of underlying helpers (e.g. preempt_count()).
> 
> I can Cc you when sending that out, if you'd like? That'll probably be in the
> new year.

Thanks Mark, yes, I'm interested in that work.
Until that, I'll finish making a script for testing kprobes widely on
kernel symbols.

Thank you,

> 
> Thanks,
> Mark.


-- 
Masami Hiramatsu (Google) <mhiramat@kernel.org>

_______________________________________________
linux-arm-kernel mailing list
linux-arm-kernel@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel

[PATCH v2 2/3] arm64: kprobes: Let arch do_page_fault() fix up page fault in user handler

[Masami Hiramatsu (Google)]

From: Masami Hiramatsu (Google) <mhiramat@kernel.org>

Since arm64's do_page_fault() can handle the page fault correctly
than kprobe_fault_handler() according to the context, let it handle
the page fault instead of simply call fixup_exception() in the
kprobe_fault_handler().

Signed-off-by: Masami Hiramatsu (Google) <mhiramat@kernel.org>
Acked-by: Mark Rutland <mark.rutland@arm.com>
---
 arch/arm64/kernel/probes/kprobes.c |    8 --------
 1 file changed, 8 deletions(-)

diff --git a/arch/arm64/kernel/probes/kprobes.c b/arch/arm64/kernel/probes/kprobes.c
index c9e4d0720285..d2ae37f89774 100644
--- a/arch/arm64/kernel/probes/kprobes.c
+++ b/arch/arm64/kernel/probes/kprobes.c
@@ -294,14 +294,6 @@ int __kprobes kprobe_fault_handler(struct pt_regs *regs, unsigned int fsr)
 		}
 
 		break;
-	case KPROBE_HIT_ACTIVE:
-	case KPROBE_HIT_SSDONE:
-		/*
-		 * In case the user-specified fault handler returned
-		 * zero, try to fix up.
-		 */
-		if (fixup_exception(regs))
-			return 1;
 	}
 	return 0;
 }


_______________________________________________
linux-arm-kernel mailing list
linux-arm-kernel@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel

[PATCH v2 3/3] arm64: kprobes: Return DBG_HOOK_ERROR if kprobes can not handle a BRK

[Masami Hiramatsu (Google)]

From: Masami Hiramatsu (Google) <mhiramat@kernel.org>

Return DBG_HOOK_ERROR if kprobes can not handle a BRK because it
fails to find a kprobe corresponding to the address.

Since arm64 kprobes uses stop_machine based text patching for removing
BRK, it ensures all running kprobe_break_handler() is done at that point.
And after removing the BRK, it removes the kprobe from its hash list.
Thus, if the kprobe_break_handler() fails to find kprobe from hash list,
there is a bug.

Signed-off-by: Masami Hiramatsu (Google) <mhiramat@kernel.org>
Acked-by: Mark Rutland <mark.rutland@arm.com>
---
 Changes in v2:
  - Update comments in the code.
---
 arch/arm64/kernel/probes/kprobes.c |   78 +++++++++++++++++-------------------
 1 file changed, 36 insertions(+), 42 deletions(-)

diff --git a/arch/arm64/kernel/probes/kprobes.c b/arch/arm64/kernel/probes/kprobes.c
index d2ae37f89774..f35d059a9a36 100644
--- a/arch/arm64/kernel/probes/kprobes.c
+++ b/arch/arm64/kernel/probes/kprobes.c
@@ -298,7 +298,8 @@ int __kprobes kprobe_fault_handler(struct pt_regs *regs, unsigned int fsr)
 	return 0;
 }
 
-static void __kprobes kprobe_handler(struct pt_regs *regs)
+static int __kprobes
+kprobe_breakpoint_handler(struct pt_regs *regs, unsigned long esr)
 {
 	struct kprobe *p, *cur_kprobe;
 	struct kprobe_ctlblk *kcb;
@@ -308,39 +309,44 @@ static void __kprobes kprobe_handler(struct pt_regs *regs)
 	cur_kprobe = kprobe_running();
 
 	p = get_kprobe((kprobe_opcode_t *) addr);
+	if (WARN_ON_ONCE(!p)) {
+		/*
+		 * Something went wrong. This BRK used an immediate reserved
+		 * for kprobes, but we couldn't find any corresponding probe.
+		 */
+		return DBG_HOOK_ERROR;
+	}
 
-	if (p) {
-		if (cur_kprobe) {
-			if (reenter_kprobe(p, regs, kcb))
-				return;
-		} else {
-			/* Probe hit */
-			set_current_kprobe(p);
-			kcb->kprobe_status = KPROBE_HIT_ACTIVE;
-
-			/*
-			 * If we have no pre-handler or it returned 0, we
-			 * continue with normal processing.  If we have a
-			 * pre-handler and it returned non-zero, it will
-			 * modify the execution path and no need to single
-			 * stepping. Let's just reset current kprobe and exit.
-			 */
-			if (!p->pre_handler || !p->pre_handler(p, regs)) {
-				setup_singlestep(p, regs, kcb, 0);
-			} else
-				reset_current_kprobe();
-		}
+	if (cur_kprobe) {
+		/* Hit a kprobe inside another kprobe */
+		if (!reenter_kprobe(p, regs, kcb))
+			return DBG_HOOK_ERROR;
+	} else {
+		/* Probe hit */
+		set_current_kprobe(p);
+		kcb->kprobe_status = KPROBE_HIT_ACTIVE;
+
+		/*
+		 * If we have no pre-handler or it returned 0, we
+		 * continue with normal processing.  If we have a
+		 * pre-handler and it returned non-zero, it will
+		 * modify the execution path and not need to single-step
+		 * Let's just reset current kprobe and exit.
+		 */
+		if (!p->pre_handler || !p->pre_handler(p, regs))
+			setup_singlestep(p, regs, kcb, 0);
+		else
+			reset_current_kprobe();
 	}
-	/*
-	 * The breakpoint instruction was removed right
-	 * after we hit it.  Another cpu has removed
-	 * either a probepoint or a debugger breakpoint
-	 * at this address.  In either case, no further
-	 * handling of this interrupt is appropriate.
-	 * Return back to original instruction, and continue.
-	 */
+
+	return DBG_HOOK_HANDLED;
 }
 
+static struct break_hook kprobes_break_hook = {
+	.imm = KPROBES_BRK_IMM,
+	.fn = kprobe_breakpoint_handler,
+};
+
 static int __kprobes
 kprobe_breakpoint_ss_handler(struct pt_regs *regs, unsigned long esr)
 {
@@ -365,18 +371,6 @@ static struct break_hook kprobes_break_ss_hook = {
 	.fn = kprobe_breakpoint_ss_handler,
 };
 
-static int __kprobes
-kprobe_breakpoint_handler(struct pt_regs *regs, unsigned long esr)
-{
-	kprobe_handler(regs);
-	return DBG_HOOK_HANDLED;
-}
-
-static struct break_hook kprobes_break_hook = {
-	.imm = KPROBES_BRK_IMM,
-	.fn = kprobe_breakpoint_handler,
-};
-
 /*
  * Provide a blacklist of symbols identifying ranges which cannot be kprobed.
  * This blacklist is exposed to userspace via debugfs (kprobes/blacklist).


_______________________________________________
linux-arm-kernel mailing list
linux-arm-kernel@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel

Re: [PATCH v2 0/3] arm64: kprobes: Fix bugs in kprobes for arm64

[Will Deacon]

On Fri, 2 Dec 2022 11:18:23 +0900, Masami Hiramatsu (Google) wrote:
> Here is the 2nd version of the kprobe bugfixes on arm64. The previous
> series is here;
> 
> https://lore.kernel.org/all/166990553243.253128.13594802750635478633.stgit@devnote3/T/#u
> 
> The first one is a critical issue, which will cause a kernel crach
> easily with lockdep[1/3]. Others are minor issues and rare cases.
> The [2/3] let do_page_fault() fixup the page fault in kprobes user
> handler, and [3/3] is more like code cleanup and returns
> DBG_HOOK_ERROR if it can not handle kprobe's BRK (but that
> should not happen.)
> 
> [...]

Applied to arm64 (for-next/kprobes), thanks!

[1/3] arm64: Prohibit instrumentation on arch_stack_walk()
      https://git.kernel.org/arm64/c/0fbcd8abf337
[2/3] arm64: kprobes: Let arch do_page_fault() fix up page fault in user handler
      https://git.kernel.org/arm64/c/30a421552373
[3/3] arm64: kprobes: Return DBG_HOOK_ERROR if kprobes can not handle a BRK
      https://git.kernel.org/arm64/c/3b84efc066f8

Cheers,
-- 
Will

https://fixes.arm64.dev
https://next.arm64.dev
https://will.arm64.dev

_______________________________________________
linux-arm-kernel mailing list
linux-arm-kernel@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel