All of lore.kernel.org
 help / color / mirror / Atom feed
From: Peter Zijlstra <peterz@infradead.org>
To: Dan Williams <dan.j.williams@intel.com>
Cc: linux-kernel@vger.kernel.org, linux-arch@vger.kernel.org,
	kernel-hardening@lists.openwall.com,
	Catalin Marinas <catalin.marinas@arm.com>,
	x86@kernel.org, Will Deacon <will.deacon@arm.com>,
	Alexei Starovoitov <ast@kernel.org>,
	Russell King <linux@armlinux.org.uk>,
	Ingo Molnar <mingo@redhat.com>, "H. Peter Anvin" <hpa@zytor.com>,
	tglx@linutronix.de, torvalds@linux-foundation.org,
	akpm@linux-foundation.org, alan@linux.intel.com
Subject: [kernel-hardening] Re: [PATCH v2 06/19] asm-generic/barrier: mask speculative execution flows
Date: Fri, 12 Jan 2018 10:12:46 +0100	[thread overview]
Message-ID: <20180112091246.GP32035@hirez.programming.kicks-ass.net> (raw)
In-Reply-To: <151571801681.27429.15417813964230837664.stgit@dwillia2-desk3.amr.corp.intel.com>

On Thu, Jan 11, 2018 at 04:46:56PM -0800, Dan Williams wrote:
> diff --git a/include/linux/nospec.h b/include/linux/nospec.h
> new file mode 100644
> index 000000000000..5c66fc30f919
> --- /dev/null
> +++ b/include/linux/nospec.h
> @@ -0,0 +1,71 @@
> +// SPDX-License-Identifier: GPL-2.0
> +// Copyright(c) 2018 Intel Corporation. All rights reserved.
> +
> +#ifndef __NOSPEC_H__
> +#define __NOSPEC_H__
> +
> +#include <linux/jump_label.h>
> +#include <asm/barrier.h>
> +
> +#ifndef array_ptr_mask
> +#define array_ptr_mask(idx, sz)						\
> +({									\
> +	unsigned long mask;						\
> +	unsigned long _i = (idx);					\
> +	unsigned long _s = (sz);					\
> +									\
> +	mask = ~(long)(_i | (_s - 1 - _i)) >> (BITS_PER_LONG - 1);	\
> +	mask;								\
> +})
> +#endif
> +
> +/**
> + * __array_ptr - Generate a pointer to an array element, ensuring
> + * the pointer is bounded under speculation to NULL.
> + *
> + * @base: the base of the array
> + * @idx: the index of the element, must be less than LONG_MAX
> + * @sz: the number of elements in the array, must be less than LONG_MAX
> + *
> + * If @idx falls in the interval [0, @sz), returns the pointer to
> + * @arr[@idx], otherwise returns NULL.
> + */
> +#define __array_ptr(base, idx, sz)					\
> +({									\
> +	union { typeof(*(base)) *_ptr; unsigned long _bit; } __u;	\
> +	typeof(*(base)) *_arr = (base);					\
> +	unsigned long _i = (idx);					\
> +	unsigned long _mask = array_ptr_mask(_i, (sz));			\
> +									\
> +	__u._ptr = _arr + (_i & _mask);					\
> +	__u._bit &= _mask;						\
> +	__u._ptr;							\
> +})
> +
> +#ifdef CONFIG_SPECTRE1_IFENCE
> +DECLARE_STATIC_KEY_TRUE(nospec_key);
> +#else
> +DECLARE_STATIC_KEY_FALSE(nospec_key);
> +#endif
> +
> +#ifdef ifence_array_ptr
> +/*
> + * The expectation is that no compiler or cpu will mishandle __array_ptr
> + * leading to problematic speculative execution. Bypass the ifence
> + * based implementation by default.
> + */
> +#define array_ptr(base, idx, sz)				\
> +({								\
> +	typeof(*(base)) *__ret;					\
> +								\
> +	if (static_branch_unlikely(&nospec_key))		\
> +		__ret = ifence_array_ptr(base, idx, sz);	\
> +	else							\
> +		__ret = __array_ptr(base, idx, sz);		\
> +	__ret;							\
> +})


So I think this wants:

#ifndef HAVE_JUMP_LABEL
#error Compiler lacks asm-goto, can generate unsafe code
#endif

Suppose the generic array_ptr_mask() is unsafe on some arch and they
only implement ifence_array_ptr() and they compile without asm-goto,
then the above reverts to a dynamic condition, which can be speculated.
If we then speculate into the 'bad' __array_ptr we're screwed.

> +#else
> +#define array_ptr __array_ptr
> +#endif
> +
> +#endif /* __NOSPEC_H__ */


In general I think I would write all this in a form like:

#define __array_ptr(base, idx, sz)					\
({									\
	union { typeof(*(base)) *_ptr; unsigned long _bit; } __u;	\
	typeof(*(base)) *_arr = (base);					\
	unsigned long _i = (idx);					\
	unsigned long _mask = array_ptr_mask(_i, (sz));			\
									\
	__u._ptr = _arr + (_i & _mask);					\
	__u._bit &= _mask;						\
	__u._ptr;							\
})

#if defined(array_ptr_mask) && defined(ifence_array_ptr)

#ifndef HAVE_JUMP_LABEL
#error Compiler lacks asm-goto, can generate unsafe code
#endif

#define array_ptr(base, idx, sz)				\
({								\
	typeof(*(base)) *__ret;					\
								\
	if (static_branch_unlikely(&nospec_key))		\
		__ret = ifence_array_ptr(base, idx, sz);	\
	else							\
		__ret = __array_ptr(base, idx, sz);		\
	__ret;							\
})

#elif defined(array_ptr_mask)

#define array_ptr(base, idx, sz) __array_ptr(base, idx, sz)

#elif defined(ifence_array_ptr)

#define array_ptr(base, idx, sz) ifence_array_ptr(base, idx, sz)

#else

/* XXX we want a suitable warning here ? */

#define array_ptr(base, idx, sz) (idx < sz ? base + idx : NULL)

#endif

and stick the generic array_ptr_mask into asm-generic/nospec.h or
something.

Then the static key stuff is limited to architectures that define _both_
array_ptr_mask and ifence_array_ptr.

WARNING: multiple messages have this Message-ID (diff)
From: Peter Zijlstra <peterz@infradead.org>
To: Dan Williams <dan.j.williams@intel.com>
Cc: linux-kernel@vger.kernel.org, linux-arch@vger.kernel.org,
	kernel-hardening@lists.openwall.com,
	Catalin Marinas <catalin.marinas@arm.com>,
	x86@kernel.org, Will Deacon <will.deacon@arm.com>,
	Alexei Starovoitov <ast@kernel.org>,
	Russell King <linux@armlinux.org.uk>,
	Ingo Molnar <mingo@redhat.com>, "H. Peter Anvin" <hpa@zytor.com>,
	tglx@linutronix.de, torvalds@linux-foundation.org,
	akpm@linux-foundation.org, alan@linux.intel.com
Subject: Re: [PATCH v2 06/19] asm-generic/barrier: mask speculative execution flows
Date: Fri, 12 Jan 2018 10:12:46 +0100	[thread overview]
Message-ID: <20180112091246.GP32035@hirez.programming.kicks-ass.net> (raw)
In-Reply-To: <151571801681.27429.15417813964230837664.stgit@dwillia2-desk3.amr.corp.intel.com>

On Thu, Jan 11, 2018 at 04:46:56PM -0800, Dan Williams wrote:
> diff --git a/include/linux/nospec.h b/include/linux/nospec.h
> new file mode 100644
> index 000000000000..5c66fc30f919
> --- /dev/null
> +++ b/include/linux/nospec.h
> @@ -0,0 +1,71 @@
> +// SPDX-License-Identifier: GPL-2.0
> +// Copyright(c) 2018 Intel Corporation. All rights reserved.
> +
> +#ifndef __NOSPEC_H__
> +#define __NOSPEC_H__
> +
> +#include <linux/jump_label.h>
> +#include <asm/barrier.h>
> +
> +#ifndef array_ptr_mask
> +#define array_ptr_mask(idx, sz)						\
> +({									\
> +	unsigned long mask;						\
> +	unsigned long _i = (idx);					\
> +	unsigned long _s = (sz);					\
> +									\
> +	mask = ~(long)(_i | (_s - 1 - _i)) >> (BITS_PER_LONG - 1);	\
> +	mask;								\
> +})
> +#endif
> +
> +/**
> + * __array_ptr - Generate a pointer to an array element, ensuring
> + * the pointer is bounded under speculation to NULL.
> + *
> + * @base: the base of the array
> + * @idx: the index of the element, must be less than LONG_MAX
> + * @sz: the number of elements in the array, must be less than LONG_MAX
> + *
> + * If @idx falls in the interval [0, @sz), returns the pointer to
> + * @arr[@idx], otherwise returns NULL.
> + */
> +#define __array_ptr(base, idx, sz)					\
> +({									\
> +	union { typeof(*(base)) *_ptr; unsigned long _bit; } __u;	\
> +	typeof(*(base)) *_arr = (base);					\
> +	unsigned long _i = (idx);					\
> +	unsigned long _mask = array_ptr_mask(_i, (sz));			\
> +									\
> +	__u._ptr = _arr + (_i & _mask);					\
> +	__u._bit &= _mask;						\
> +	__u._ptr;							\
> +})
> +
> +#ifdef CONFIG_SPECTRE1_IFENCE
> +DECLARE_STATIC_KEY_TRUE(nospec_key);
> +#else
> +DECLARE_STATIC_KEY_FALSE(nospec_key);
> +#endif
> +
> +#ifdef ifence_array_ptr
> +/*
> + * The expectation is that no compiler or cpu will mishandle __array_ptr
> + * leading to problematic speculative execution. Bypass the ifence
> + * based implementation by default.
> + */
> +#define array_ptr(base, idx, sz)				\
> +({								\
> +	typeof(*(base)) *__ret;					\
> +								\
> +	if (static_branch_unlikely(&nospec_key))		\
> +		__ret = ifence_array_ptr(base, idx, sz);	\
> +	else							\
> +		__ret = __array_ptr(base, idx, sz);		\
> +	__ret;							\
> +})


So I think this wants:

#ifndef HAVE_JUMP_LABEL
#error Compiler lacks asm-goto, can generate unsafe code
#endif

Suppose the generic array_ptr_mask() is unsafe on some arch and they
only implement ifence_array_ptr() and they compile without asm-goto,
then the above reverts to a dynamic condition, which can be speculated.
If we then speculate into the 'bad' __array_ptr we're screwed.

> +#else
> +#define array_ptr __array_ptr
> +#endif
> +
> +#endif /* __NOSPEC_H__ */


In general I think I would write all this in a form like:

#define __array_ptr(base, idx, sz)					\
({									\
	union { typeof(*(base)) *_ptr; unsigned long _bit; } __u;	\
	typeof(*(base)) *_arr = (base);					\
	unsigned long _i = (idx);					\
	unsigned long _mask = array_ptr_mask(_i, (sz));			\
									\
	__u._ptr = _arr + (_i & _mask);					\
	__u._bit &= _mask;						\
	__u._ptr;							\
})

#if defined(array_ptr_mask) && defined(ifence_array_ptr)

#ifndef HAVE_JUMP_LABEL
#error Compiler lacks asm-goto, can generate unsafe code
#endif

#define array_ptr(base, idx, sz)				\
({								\
	typeof(*(base)) *__ret;					\
								\
	if (static_branch_unlikely(&nospec_key))		\
		__ret = ifence_array_ptr(base, idx, sz);	\
	else							\
		__ret = __array_ptr(base, idx, sz);		\
	__ret;							\
})

#elif defined(array_ptr_mask)

#define array_ptr(base, idx, sz) __array_ptr(base, idx, sz)

#elif defined(ifence_array_ptr)

#define array_ptr(base, idx, sz) ifence_array_ptr(base, idx, sz)

#else

/* XXX we want a suitable warning here ? */

#define array_ptr(base, idx, sz) (idx < sz ? base + idx : NULL)

#endif

and stick the generic array_ptr_mask into asm-generic/nospec.h or
something.

Then the static key stuff is limited to architectures that define _both_
array_ptr_mask and ifence_array_ptr.

  parent reply	other threads:[~2018-01-12  9:12 UTC|newest]

Thread overview: 139+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2018-01-12  0:46 [kernel-hardening] [PATCH v2 00/19] prevent bounds-check bypass via speculative execution Dan Williams
2018-01-12  0:46 ` Dan Williams
2018-01-12  0:46 ` Dan Williams
2018-01-12  0:46 ` [kernel-hardening] [PATCH v2 01/19] Documentation: document array_ptr Dan Williams
2018-01-12  0:46   ` Dan Williams
2018-01-12 10:38   ` [kernel-hardening] " Geert Uytterhoeven
2018-01-12 10:38     ` Geert Uytterhoeven
2018-01-12 10:38     ` Geert Uytterhoeven
2018-01-16 21:01   ` [kernel-hardening] " Kees Cook
2018-01-16 21:01     ` Kees Cook
2018-01-12  0:46 ` [kernel-hardening] [PATCH v2 02/19] arm64: implement ifence_array_ptr() Dan Williams
2018-01-12  0:46   ` Dan Williams
2018-01-12  0:46 ` [kernel-hardening] [PATCH v2 03/19] arm: " Dan Williams
2018-01-12  0:46   ` Dan Williams
2018-01-12  0:46 ` [kernel-hardening] [PATCH v2 04/19] x86: implement ifence() Dan Williams
2018-01-12  0:46   ` Dan Williams
2018-01-12  2:27   ` [kernel-hardening] " Eric W. Biederman
2018-01-12  2:27     ` Eric W. Biederman
2018-01-12  3:39     ` [kernel-hardening] " Dan Williams
2018-01-12  3:39       ` Dan Williams
2018-01-12  0:46 ` [kernel-hardening] [PATCH v2 05/19] x86: implement ifence_array_ptr() and array_ptr_mask() Dan Williams
2018-01-12  0:46   ` Dan Williams
2018-01-12  0:46 ` [kernel-hardening] [PATCH v2 06/19] asm-generic/barrier: mask speculative execution flows Dan Williams
2018-01-12  0:46   ` Dan Williams
2018-01-12  2:42   ` [kernel-hardening] " Eric W. Biederman
2018-01-12  2:42     ` Eric W. Biederman
2018-01-12  9:12   ` Peter Zijlstra [this message]
2018-01-12  9:12     ` Peter Zijlstra
2018-01-13  0:41     ` [kernel-hardening] " Dan Williams
2018-01-13  0:41       ` Dan Williams
2018-01-15  8:46       ` [kernel-hardening] " Peter Zijlstra
2018-01-15  8:46         ` Peter Zijlstra
2018-01-12  0:47 ` [kernel-hardening] [PATCH v2 07/19] x86: introduce __uaccess_begin_nospec and ASM_IFENCE Dan Williams
2018-01-12  0:47   ` Dan Williams
2018-01-12 17:51   ` [kernel-hardening] " Josh Poimboeuf
2018-01-12 17:51     ` Josh Poimboeuf
2018-01-12 18:21     ` [kernel-hardening] " Dan Williams
2018-01-12 18:21       ` Dan Williams
2018-01-12 18:58       ` [kernel-hardening] " Josh Poimboeuf
2018-01-12 18:58         ` Josh Poimboeuf
2018-01-12 19:26         ` [kernel-hardening] " Dan Williams
2018-01-12 19:26           ` Dan Williams
2018-01-12 20:01           ` [kernel-hardening] " Linus Torvalds
2018-01-12 20:01             ` Linus Torvalds
2018-01-12 20:41             ` [kernel-hardening] " Josh Poimboeuf
2018-01-12 20:41               ` Josh Poimboeuf
2018-01-12  0:47 ` [kernel-hardening] [PATCH v2 08/19] x86: use __uaccess_begin_nospec and ASM_IFENCE in get_user paths Dan Williams
2018-01-12  0:47   ` Dan Williams
2018-01-12  1:11   ` [kernel-hardening] " Linus Torvalds
2018-01-12  1:11     ` Linus Torvalds
2018-01-12  1:14     ` [kernel-hardening] " Dan Williams
2018-01-12  1:14       ` Dan Williams
2018-01-12  0:47 ` [kernel-hardening] [PATCH v2 09/19] ipv6: prevent bounds-check bypass via speculative execution Dan Williams
2018-01-12  0:47   ` Dan Williams
2018-01-12  0:47 ` [kernel-hardening] [PATCH v2 10/19] ipv4: " Dan Williams
2018-01-12  0:47   ` Dan Williams
2018-01-12  7:59   ` [kernel-hardening] " Greg KH
2018-01-12  7:59     ` Greg KH
2018-01-12 18:47     ` [kernel-hardening] " Dan Williams
2018-01-12 18:47       ` Dan Williams
2018-01-13  8:56       ` [kernel-hardening] " Greg KH
2018-01-13  8:56         ` Greg KH
2018-01-12  0:47 ` [kernel-hardening] [PATCH v2 11/19] vfs, fdtable: " Dan Williams
2018-01-12  0:47   ` Dan Williams
2018-01-12  0:47 ` [kernel-hardening] [PATCH v2 12/19] userns: " Dan Williams
2018-01-12  0:47   ` Dan Williams
2018-01-12  0:47 ` [kernel-hardening] [PATCH v2 13/19] udf: " Dan Williams
2018-01-12  0:47   ` Dan Williams
2018-01-15 10:32   ` [kernel-hardening] " Jan Kara
2018-01-15 10:32     ` Jan Kara
2018-01-15 17:49     ` [kernel-hardening] " Dan Williams
2018-01-15 17:49       ` Dan Williams
2018-01-12  0:47 ` [kernel-hardening] [PATCH v2 14/19] [media] uvcvideo: " Dan Williams
2018-01-12  0:47   ` Dan Williams
2018-08-06 21:40   ` Laurent Pinchart
2018-01-12  0:47 ` [kernel-hardening] [PATCH v2 15/19] carl9170: " Dan Williams
2018-01-12  0:47   ` Dan Williams
2018-01-12 14:42   ` [kernel-hardening] " Christian Lamparter
2018-01-12 14:42     ` Christian Lamparter
2018-01-12 18:39     ` [kernel-hardening] " Dan Williams
2018-01-12 18:39       ` Dan Williams
2018-01-12 20:01       ` [kernel-hardening] " Christian Lamparter
2018-01-12 20:01         ` Christian Lamparter
2018-01-12 23:05         ` [kernel-hardening] " Dan Williams
2018-01-12 23:05           ` Dan Williams
2018-01-12  0:47 ` [kernel-hardening] [PATCH v2 16/19] p54: " Dan Williams
2018-01-12  0:47   ` Dan Williams
2018-01-12  0:47 ` [kernel-hardening] [PATCH v2 17/19] qla2xxx: " Dan Williams
2018-01-12  0:47   ` Dan Williams
2018-01-12  1:19   ` [kernel-hardening] " James Bottomley
2018-01-12  1:19     ` James Bottomley
2018-01-12  5:38     ` [kernel-hardening] " Dan Williams
2018-01-12  5:38       ` Dan Williams
2018-01-12  6:05       ` [kernel-hardening] " James Bottomley
2018-01-12  6:05         ` James Bottomley
2018-01-12  0:48 ` [kernel-hardening] [PATCH v2 18/19] cw1200: " Dan Williams
2018-01-12  0:48   ` Dan Williams
2018-01-12  0:48 ` [kernel-hardening] [PATCH v2 19/19] net: mpls: " Dan Williams
2018-01-12  0:48   ` Dan Williams
2018-01-12  1:19 ` [kernel-hardening] Re: [PATCH v2 00/19] " Linus Torvalds
2018-01-12  1:19   ` Linus Torvalds
2018-01-12  1:19   ` Linus Torvalds
2018-01-12  1:19   ` Linus Torvalds
2018-01-12  1:41   ` [kernel-hardening] " Dan Williams
2018-01-12  1:41     ` Dan Williams
2018-01-12  1:41     ` Dan Williams
2018-01-12  1:41     ` Dan Williams
2018-01-18 13:18     ` [kernel-hardening] " Will Deacon
2018-01-18 13:18       ` Will Deacon
2018-01-18 13:18       ` Will Deacon
2018-01-18 13:18       ` Will Deacon
2018-01-18 13:18       ` Will Deacon
2018-01-18 13:18       ` Will Deacon
2018-01-18 16:58       ` [kernel-hardening] " Dan Williams
2018-01-18 16:58         ` Dan Williams
2018-01-18 16:58         ` Dan Williams
2018-01-18 16:58         ` Dan Williams
2018-01-18 17:05         ` [kernel-hardening] " Will Deacon
2018-01-18 17:05           ` Will Deacon
2018-01-18 17:05           ` Will Deacon
2018-01-18 17:05           ` Will Deacon
2018-01-18 21:41           ` [kernel-hardening] " Laurent Pinchart
2018-01-18 21:41             ` Laurent Pinchart
2018-01-18 21:41             ` Laurent Pinchart
2018-01-18 21:41             ` Laurent Pinchart
2018-01-18 21:41             ` Laurent Pinchart
2018-01-13  0:15   ` [kernel-hardening] " Tony Luck
2018-01-13  0:15     ` Tony Luck
2018-01-13  0:15     ` Tony Luck
2018-01-13 18:51     ` [kernel-hardening] " Linus Torvalds
2018-01-13 18:51       ` Linus Torvalds
2018-01-13 18:51       ` Linus Torvalds
2018-01-16 19:21       ` [kernel-hardening] " Tony Luck
2018-01-16 19:21         ` Tony Luck
2018-01-16 19:21         ` Tony Luck
2018-01-12 10:02 ` [kernel-hardening] " Russell King - ARM Linux
2018-01-12 10:02   ` Russell King - ARM Linux
2018-01-12 10:02   ` Russell King - ARM Linux
2018-01-12 10:02   ` Russell King - ARM Linux

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20180112091246.GP32035@hirez.programming.kicks-ass.net \
    --to=peterz@infradead.org \
    --cc=akpm@linux-foundation.org \
    --cc=alan@linux.intel.com \
    --cc=ast@kernel.org \
    --cc=catalin.marinas@arm.com \
    --cc=dan.j.williams@intel.com \
    --cc=hpa@zytor.com \
    --cc=kernel-hardening@lists.openwall.com \
    --cc=linux-arch@vger.kernel.org \
    --cc=linux-kernel@vger.kernel.org \
    --cc=linux@armlinux.org.uk \
    --cc=mingo@redhat.com \
    --cc=tglx@linutronix.de \
    --cc=torvalds@linux-foundation.org \
    --cc=will.deacon@arm.com \
    --cc=x86@kernel.org \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.