linux-kernel - Re: [tip: x86/urgent] lkdtm: Add a DOUBLE

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives
Hash Suite for Android: free password hash cracker in your pocket
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Date:   Wed, 27 Nov 2019 09:44:59 -0800
From:   Kees Cook <keescook@...omium.org>
To:     linux-kernel@...r.kernel.org
Cc:     linux-tip-commits@...r.kernel.org,
        Andy Lutomirski <luto@...nel.org>,
        Arnd Bergmann <arnd@...db.de>,
        Greg Kroah-Hartman <gregkh@...uxfoundation.org>,
        Borislav Petkov <bp@...en8.de>,
        Peter Zijlstra <peterz@...radead.org>,
        Thomas Gleixner <tglx@...utronix.de>,
        Linus Torvalds <torvalds@...ux-foundation.org>,
        Ingo Molnar <mingo@...nel.org>, x86 <x86@...nel.org>
Subject: Re: [tip: x86/urgent] lkdtm: Add a DOUBLE_FAULT crash type on x86

On Wed, Nov 27, 2019 at 08:19:30AM -0000, tip-bot2 for Andy Lutomirski wrote:
> The following commit has been merged into the x86/urgent branch of tip:
> 
> Commit-ID:     b09511c253e5c739a60952b97c071a93e92b2e88
> Gitweb:        https://git.kernel.org/tip/b09511c253e5c739a60952b97c071a93e92b2e88
> Author:        Andy Lutomirski <luto@...nel.org>
> AuthorDate:    Sun, 24 Nov 2019 21:18:04 -08:00
> Committer:     Ingo Molnar <mingo@...nel.org>
> CommitterDate: Tue, 26 Nov 2019 21:53:34 +01:00
> 
> lkdtm: Add a DOUBLE_FAULT crash type on x86
> 
> The DOUBLE_FAULT crash does INT $8, which is a decent approximation
> of a double fault.  This is useful for testing the double fault
> handling.  Use it like:
> 
> Signed-off-by: Andy Lutomirski <luto@...nel.org>
> Cc: Kees Cook <keescook@...omium.org>
> Cc: Arnd Bergmann <arnd@...db.de>
> Cc: Greg Kroah-Hartman <gregkh@...uxfoundation.org>
> Cc: Borislav Petkov <bp@...en8.de>
> Cc: Peter Zijlstra <peterz@...radead.org>
> Cc: Thomas Gleixner <tglx@...utronix.de>
> Cc: Linus Torvalds <torvalds@...ux-foundation.org>
> Signed-off-by: Ingo Molnar <mingo@...nel.org>
> ---
>  drivers/misc/lkdtm/bugs.c  | 39 +++++++++++++++++++++++++++++++++++++-
>  drivers/misc/lkdtm/core.c  |  3 +++-
>  drivers/misc/lkdtm/lkdtm.h |  3 +++-
>  3 files changed, 45 insertions(+)
> 
> diff --git a/drivers/misc/lkdtm/bugs.c b/drivers/misc/lkdtm/bugs.c
> index 7284a22..a4fdad0 100644
> --- a/drivers/misc/lkdtm/bugs.c
> +++ b/drivers/misc/lkdtm/bugs.c
> @@ -12,6 +12,10 @@
>  #include <linux/sched/task_stack.h>
>  #include <linux/uaccess.h>
>  
> +#ifdef CONFIG_X86_32
> +#include <asm/desc.h>
> +#endif
> +
>  struct lkdtm_list {
>  	struct list_head node;
>  };
> @@ -337,3 +341,38 @@ void lkdtm_UNSET_SMEP(void)
>  	pr_err("FAIL: this test is x86_64-only\n");
>  #endif
>  }
> +
> +#ifdef CONFIG_X86_32
> +void lkdtm_DOUBLE_FAULT(void)
> +{
> +	/*
> +	 * Trigger #DF by setting the stack limit to zero.  This clobbers
> +	 * a GDT TLS slot, which is okay because the current task will die
> +	 * anyway due to the double fault.
> +	 */
> +	struct desc_struct d = {
> +		.type = 3,	/* expand-up, writable, accessed data */
> +		.p = 1,		/* present */
> +		.d = 1,		/* 32-bit */
> +		.g = 0,		/* limit in bytes */
> +		.s = 1,		/* not system */
> +	};
> +
> +	local_irq_disable();
> +	write_gdt_entry(get_cpu_gdt_rw(smp_processor_id()),
> +			GDT_ENTRY_TLS_MIN, &d, DESCTYPE_S);
> +
> +	/*
> +	 * Put our zero-limit segment in SS and then trigger a fault.  The
> +	 * 4-byte access to (%esp) will fault with #SS, and the attempt to
> +	 * deliver the fault will recursively cause #SS and result in #DF.
> +	 * This whole process happens while NMIs and MCEs are blocked by the
> +	 * MOV SS window.  This is nice because an NMI with an invalid SS
> +	 * would also double-fault, resulting in the NMI or MCE being lost.
> +	 */
> +	asm volatile ("movw %0, %%ss; addl $0, (%%esp)" ::
> +		      "r" ((unsigned short)(GDT_ENTRY_TLS_MIN << 3)));
> +
> +	panic("tried to double fault but didn't die\n");

I'll modify this in some later patches, but I prefer the #ifdef inside
the function so that all tests are visible on all
architectures/configurations. And it should not panic on a test failure,
it should continue (see the others) with something like:

	pr_err("FAIL: did not double fault!\n");

E.g. an external system monitor would see the double-fault and the panic as
both causing a system reboot, but only the double-fault should do that.

> +}
> +#endif
> diff --git a/drivers/misc/lkdtm/core.c b/drivers/misc/lkdtm/core.c
> index cbc4c90..ee0d6e7 100644
> --- a/drivers/misc/lkdtm/core.c
> +++ b/drivers/misc/lkdtm/core.c
> @@ -171,6 +171,9 @@ static const struct crashtype crashtypes[] = {
>  	CRASHTYPE(USERCOPY_KERNEL_DS),
>  	CRASHTYPE(STACKLEAK_ERASING),
>  	CRASHTYPE(CFI_FORWARD_PROTO),
> +#ifdef CONFIG_X86_32
> +	CRASHTYPE(DOUBLE_FAULT),
> +#endif

And then ifdefs aren't needed here either.

>  };
>  
>  
> diff --git a/drivers/misc/lkdtm/lkdtm.h b/drivers/misc/lkdtm/lkdtm.h
> index ab446e0..c56d23e 100644
> --- a/drivers/misc/lkdtm/lkdtm.h
> +++ b/drivers/misc/lkdtm/lkdtm.h
> @@ -28,6 +28,9 @@ void lkdtm_CORRUPT_USER_DS(void);
>  void lkdtm_STACK_GUARD_PAGE_LEADING(void);
>  void lkdtm_STACK_GUARD_PAGE_TRAILING(void);
>  void lkdtm_UNSET_SMEP(void);
> +#ifdef CONFIG_X86_32
> +void lkdtm_DOUBLE_FAULT(void);
> +#endif

Same.

>  
>  /* lkdtm_heap.c */
>  void __init lkdtm_heap_init(void);

-- 
Kees Cook