[PATCH v6 1/6] arm64: kprobe: add checks for ARMv8.3-PAuth combined instructions

[Dave Martin]

On Mon, Sep 07, 2020 at 10:45:51PM +0100, Will Deacon wrote:
> On Fri, Sep 04, 2020 at 04:12:04PM +0530, Amit Daniel Kachhap wrote:
> > Currently the ARMv8.3-PAuth combined branch instructions (braa, retaa
> > etc.) are not simulated for out-of-line execution with a handler. Hence the
> > uprobe of such instructions leads to kernel warnings in a loop as they are
> > not explicitly checked and fall into INSN_GOOD categories. Other combined
> > instructions like LDRAA and LDRBB can be probed.
> > 
> > The issue of the combined branch instructions is fixed by adding
> > definitions of all such instructions and rejecting their probes.
> > 
> > Warning log:
> >  WARNING: CPU: 5 PID: 249 at arch/arm64/kernel/probes/uprobes.c:182 uprobe_single_step_handler+0x34/0x50
> >  Modules linked in:
> >  CPU: 5 PID: 249 Comm: func Tainted: G        W         5.8.0-rc4-00005-ge658591d66d1-dirty #160
> >  Hardware name: Foundation-v8A (DT)
> >  pstate: 204003c9 (nzCv DAIF +PAN -UAO BTYPE=--)
> >  pc : uprobe_single_step_handler+0x34/0x50
> >  lr : single_step_handler+0x70/0xf8
> >  sp : ffff800012afbe30
> >  x29: ffff800012afbe30 x28: ffff000879f00ec0
> >  x27: 0000000000000000 x26: 0000000000000000
> >  x25: 0000000000000000 x24: 0000000000000000
> >  x23: 0000000060001000 x22: 00000000cb000022
> >  x21: ffff800011fc5a68 x20: ffff800012afbec0
> >  x19: ffff800011fc86c0 x18: 0000000000000000
> >  x17: 0000000000000000 x16: 0000000000000000
> >  x15: 0000000000000000 x14: 0000000000000000
> >  x13: 0000000000000000 x12: 0000000000000000
> >  x11: 0000000000000000 x10: 0000000000000000
> >  x9 : ffff800010085d50 x8 : 0000000000000000
> >  x7 : 0000000000000000 x6 : ffff800011fba9c0
> >  x5 : ffff800011fba000 x4 : ffff800012283070
> >  x3 : ffff8000100a78e0 x2 : 00000000004005f0
> >  x1 : 0000fffffffff008 x0 : ffff800012afbec0
> >  Call trace:
> >   uprobe_single_step_handler+0x34/0x50
> >   single_step_handler+0x70/0xf8
> >   do_debug_exception+0xb8/0x130
> >   el0_sync_handler+0x7c/0x188
> >   el0_sync+0x158/0x180
> > 
> > Fixes: 74afda4016a7 ("arm64: compile the kernel with ptrauth return address signing")
> > Fixes: 04ca3204fa09 ("arm64: enable pointer authentication")
> > Signed-off-by: Amit Daniel Kachhap <amit.kachhap at arm.com>
> > Reviewed-by: Dave Martin <Dave.Martin at arm.com>
> > ---
> > Changes since v5: 
> > * Slight change in commit log.
> > * Added Reviewed-by.
> > 
> >  arch/arm64/include/asm/insn.h          | 12 ++++++++++++
> >  arch/arm64/kernel/insn.c               | 14 ++++++++++++--
> >  arch/arm64/kernel/probes/decode-insn.c |  4 +++-
> >  3 files changed, 27 insertions(+), 3 deletions(-)
> > 
> > diff --git a/arch/arm64/include/asm/insn.h b/arch/arm64/include/asm/insn.h
> > index 0bc46149e491..324234068fee 100644
> > --- a/arch/arm64/include/asm/insn.h
> > +++ b/arch/arm64/include/asm/insn.h
> > @@ -359,9 +359,21 @@ __AARCH64_INSN_FUNCS(brk,	0xFFE0001F, 0xD4200000)
> >  __AARCH64_INSN_FUNCS(exception,	0xFF000000, 0xD4000000)
> >  __AARCH64_INSN_FUNCS(hint,	0xFFFFF01F, 0xD503201F)
> >  __AARCH64_INSN_FUNCS(br,	0xFFFFFC1F, 0xD61F0000)
> > +__AARCH64_INSN_FUNCS(braaz,	0xFFFFFC1F, 0xD61F081F)
> > +__AARCH64_INSN_FUNCS(brabz,	0xFFFFFC1F, 0xD61F0C1F)
> > +__AARCH64_INSN_FUNCS(braa,	0xFFFFFC00, 0xD71F0800)
> > +__AARCH64_INSN_FUNCS(brab,	0xFFFFFC00, 0xD71F0C00)
> 
> When do we need to distinguish these variants? Can we modify the mask/value
> pair so that we catch bra* in one go? That would match how they are
> documented in the Arm ARM.
> 
> >  __AARCH64_INSN_FUNCS(blr,	0xFFFFFC1F, 0xD63F0000)
> > +__AARCH64_INSN_FUNCS(blraaz,	0xFFFFFC1F, 0xD63F081F)
> > +__AARCH64_INSN_FUNCS(blrabz,	0xFFFFFC1F, 0xD63F0C1F)
> > +__AARCH64_INSN_FUNCS(blraa,	0xFFFFFC00, 0xD73F0800)
> > +__AARCH64_INSN_FUNCS(blrab,	0xFFFFFC00, 0xD73F0C00)
> 
> Same here for blra*
> 
> >  __AARCH64_INSN_FUNCS(ret,	0xFFFFFC1F, 0xD65F0000)
> > +__AARCH64_INSN_FUNCS(retaa,	0xFFFFFFFF, 0xD65F0BFF)
> > +__AARCH64_INSN_FUNCS(retab,	0xFFFFFFFF, 0xD65F0FFF)
> >  __AARCH64_INSN_FUNCS(eret,	0xFFFFFFFF, 0xD69F03E0)
> > +__AARCH64_INSN_FUNCS(eretaa,	0xFFFFFFFF, 0xD69F0BFF)
> > +__AARCH64_INSN_FUNCS(eretab,	0xFFFFFFFF, 0xD69F0FFF)
> 
> ... and here for ereta*.

>From my side:

I thought about this myself, but I thought that this may be easier to
maintain if we avoid lumping instructions together.

Some of these cases are probably trivial enough that they can be merged
at low risk, though, and we also have some other merged instruction
classes here already.  Avoiding pointless distinctions here will also
help the efficiency of code that uses these definitions in some cases,
though I don't have a feel for how significant the difference will
be -- probably not very.

If we become concerned about performance, we would probably want a
bigger overhaul of the affected code anyway.

I guess I'm happy either way.

Cheers
---Dave