Re: [PATCH 4/5] man-pages: cap_rights_limit.2: limit FD rights for Capsicum
From: David Drysdale <hidden>
Date: 2014-06-30 16:32:38
Also in:
lkml
On Mon, Jun 30, 2014 at 09:06:41AM -0700, Andy Lutomirski wrote:
On Mon, Jun 30, 2014 at 8:35 AM, David Drysdale [off-list ref] wrote:quoted
On Mon, Jun 30, 2014 at 07:53:57AM -0700, Andy Lutomirski wrote:quoted
On Mon, Jun 30, 2014 at 3:28 AM, David Drysdale [off-list ref] wrote:quoted
Signed-off-by: David Drysdale <redacted> --- man2/cap_rights_limit.2 | 171 ++++++++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 171 insertions(+) create mode 100644 man2/cap_rights_limit.2diff --git a/man2/cap_rights_limit.2 b/man2/cap_rights_limit.2 new file mode 100644 index 000000000000..3484ee1076aa --- /dev/null +++ b/man2/cap_rights_limit.2@@ -0,0 +1,171 @@ +.\" +.\" Copyright (c) 2008-2010 Robert N. M. Watson +.\" Copyright (c) 2012-2013 The FreeBSD Foundation +.\" Copyright (c) 2013-2014 Google, Inc. +.\" All rights reserved. +.\" +.\" %%%LICENSE_START(BSD_2_CLAUSE) +.\" Redistribution and use in source and binary forms, with or without +.\" modification, are permitted provided that the following conditions +.\" are met: +.\" 1. Redistributions of source code must retain the above copyright +.\" notice, this list of conditions and the following disclaimer. +.\" 2. Redistributions in binary form must reproduce the above copyright +.\" notice, this list of conditions and the following disclaimer in the +.\" documentation and/or other materials provided with the distribution. +.\" +.\" THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND +.\" ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +.\" IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +.\" ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE +.\" FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL +.\" DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS +.\" OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) +.\" HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT +.\" LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY +.\" OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF +.\" SUCH DAMAGE. +.\" %%%LICENSE_END +.\" +.TH CAP_RIGHTS_LIMIT 2 2014-05-07 "Linux" "Linux Programmer's Manual" +.SH NAME +cap_rights_limit \- limit Capsicum capability rights +.SH SYNOPSIS +.nf +.B #include <sys/capsicum.h> +.sp +.BI "int cap_rights_limit(int " fd ", const struct cap_rights *" rights , +.BI " unsigned int " fcntls , +.BI " int " nioctls ", unsigned int *" ioctls );Am I missing the docs for struct cap_rights somewhere?There's a little bit of discussion in rights.7 (mail 3/5 of the man-pages set), but there isn't a structure description. I was trying to keep the structure opaque to userspace, which would be expected to manipulate the rights with various utility functions rather than directly. But I now realize this leaves a gap -- the description of this syscall doesn't include a full description of its ABI. So I'll add in a description of the structure to this page -- basically: struct cap_rights { __u64 cr_rights[2]; }; with a slightly complicated scheme to encode rights into the bitmask array. (The encoding scheme is taken from the FreeBSD implementation, which I've tried to stick to unless there's good reason to change.)How does extensibility work? For example, what happens when someone needs to add a new right for whatever reason and they fall off the end of the list? Linux so-called capabilities have done this a few times, resulting in a giant mess. --Andy
The rights encoding scheme is supposed to cope with extensions, so let me
have a go at explaining it.
The size of the array in the structure can potentially change in future,
so a less abbreviated version is:
#define CAP_RIGHTS_VERSION_00 0
#define CAP_RIGHTS_VERSION_01 1
#define CAP_RIGHTS_VERSION_02 2
#define CAP_RIGHTS_VERSION_03 3
#define CAP_RIGHTS_VERSION CAP_RIGHTS_VERSION_00
struct cap_rights {
uint64_t cr_rights[CAP_RIGHTS_VERSION + 2];
};
The encoding rules are then:
- There are between 2 and 5 entries in the array.
- The number of entries in the array is indicated by the top 2 bits of
cr_rights[0] (as array size minus 2); this allows for future
expansion (up to 285 distinct rights):
0b00 = 2 entries
0b01 = 3 entries
0b10 = 4 entries
0b11 = 5 entries
- The top 2 bits of cr_rights[i] are 0b00 for i>0.
- The next 5 bits of each array entry indicate its position in the
array:
0b00001 for cr_rights[0]
0b00010 for cr_rights[1]
0b00100 for cr_rights[2]
0b01000 for cr_rights[3]
0b10000 for cr_rights[4]
- The remaining 57 bits of each entry are used to hold rights values,
so the current structure can hold 114 rights, and the maximum is
285.
So a future kernel (with an expanded array) can cope with an old binary
(that uses a narrow array) by reading the first u64 from the structure,
and using the top 2 bits to figure out how much more memory to copy
from userspace. Slightly inefficient, but I wouldn't expect rights
setting to be a performance critical operation.
Of course, we can deviate from the FreeBSD implementation details if
we want to -- these details are deliberately hidden from userspace
programs in the rights-manipulation library functions, so a different
implementation under the covers wouldn't affect Capsicum-using
applications. But I figured it's best to stay close unless there's
a good reason to diverge.