On Fri, Jul 18, 2014 at 03:51:31PM -0700, Omar Sandoval wrote:
> Don't break into kgdb when userspace executes the kernel break instructions
> (KGDB_BREAKINST and KGDB_COMPILED_BREAK). The kernel will oops in
> kgdb_handle_exception.
> 
> Signed-off-by: Omar Sandoval <osan...@osandov.com>
> ---
> The following program will immediately cause a kernel oops:
> .globl _start
> _start:
>       udf     #65006  @ KGDB_BREAKINST
> 
>  arch/arm/kernel/kgdb.c | 4 ++++
>  1 file changed, 4 insertions(+)
> 
> diff --git a/arch/arm/kernel/kgdb.c b/arch/arm/kernel/kgdb.c
> index 778c2f7..a74b53c 100644
> --- a/arch/arm/kernel/kgdb.c
> +++ b/arch/arm/kernel/kgdb.c
> @@ -160,12 +160,16 @@ static int kgdb_compiled_brk_fn(struct pt_regs *regs, 
> unsigned int instr)
>  static struct undef_hook kgdb_brkpt_hook = {
>       .instr_mask             = 0xffffffff,
>       .instr_val              = KGDB_BREAKINST,
> +     .cpsr_mask              = MODE_MASK,
> +     .cpsr_val               = SVC_MODE,
>       .fn                     = kgdb_brk_fn
>  };
>  
>  static struct undef_hook kgdb_compiled_brkpt_hook = {
>       .instr_mask             = 0xffffffff,
>       .instr_val              = KGDB_COMPILED_BREAK,
> +     .cpsr_mask              = MODE_MASK,
> +     .cpsr_val               = SVC_MODE,
>       .fn                     = kgdb_compiled_brk_fn
>  };
>  
> -- 
> 2.0.1

-- 

Following up/clarifying this. This only happens when the kernel is compiled
with CONFIG_KGDB. When a userspace program executes KGDB_BREAKINST or
KGDB_COMPILED_BREAK, the undef_hook for kgdb catches it. The reason in kdb_stub
defaults to KDB_REASON_OOPS, so the bug manifests itself as an oops caused by
userspace (a better description for the patch would be "Don't enter KGDB when
userspace executes kgdb break instructions"). This means that a buggy/malicious
program can take down the system just by executing an instruction.

ARM64 might have the same issue, but I don't have a board to test that on.

I verified that breaking normally (e.g., with kgdbwait or through
/proc/sysrq-trigger) still works.
—
Omar
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majord...@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/

Reply via email to