Skip to contentSkip to conversation

procfs: introduce hidepid=, hidenet=, gid= mount options

RFC In review Vasiliy Kulikov 5 participants 11 messages Jun 15, 2011 – Oct 30, 2011

lore
Share & tools
lore
#0 lore raw

v4 cover procfs: introduce hidepid=, hidenet=, gid= mount options

This patch series adds support of procfs mount options and adds mount options to restrict /proc/<pid>/ directories to owners and /proc/<pid>/net/* to root. Additional group may be defined via gid=, and this group will be privileged to study others /proc/<pid>/ and networking information.

Similar features are implemented for old kernels in -ow patches (for Linux 2.2 and 2.4) and for Linux 2.6 in -grsecurity, but both of them are implemented as configure options, not cofigurable in runtime, with changes of gid of /proc/<pid>/, and without backward-compatible /proc/<pid>/net/* handling.

The first patch introduces mount option parsing without any new options. The second patch adds options of /proc/<pid>/ restrictions. The third patch makes net_create() globally visible.
The fourth patch adds options of /proc/<pid>/net/ restrictions. The fifth patch adds documentations for options above.

Vasiliy Kulikov (5):
procfs: parse mount options
procfs: add hidepid= and gid= mount options
net: restore net_create and make it globally visible
procfs: Add hidenet/nohidenet procfs mount options.
procfs: add documentation for procfs mount options

 Documentation/filesystems/proc.txt |   51 ++++++++++++++++++++++
 fs/proc/base.c                     |   62 ++++++++++++++++++++++++++-
 fs/proc/inode.c                    |   20 +++++++++
 fs/proc/internal.h                 |    1 +
 fs/proc/proc_net.c                 |   26 +++++++++++
 fs/proc/root.c                     |   83 +++++++++++++++++++++++++++++++++++-
 include/linux/pid_namespace.h      |    3 +
 include/net/net_namespace.h        |    2 +
 net/core/net_namespace.c           |   12 +++--
 9 files changed, 253 insertions(+), 7 deletions(-)
#1 lore raw
Quoted 7 lines · Vasiliy Kulikov

This patch series adds support of procfs mount options and adds mount options to restrict /proc/<pid>/ directories to owners and /proc/<pid>/net/* to root. Additional group may be defined via gid=, and this group will be privileged to study others /proc/<pid>/ and networking information.

Quoted 5 lines

Similar features are implemented for old kernels in -ow patches (for Linux 2.2 and 2.4) and for Linux 2.6 in -grsecurity, but both of them are implemented as configure options, not cofigurable in runtime, with changes of gid of /proc/<pid>/, and without backward-compatible /proc/<pid>/net/* handling.

The patches all look good to me implementation-wise.

I have no opinion on whether it's a good idea to include the feature or not.

Arnd

#2 lore raw

Hi Arnd,

Quoted 1 line · Arnd Bergmann

The patches all look good to me implementation-wise.

Thanks for the review!

Quoted 1 line

I have no opinion on whether it's a good idea to include the feature or not.

Why not? Have you some specific complains where it can be perhaps too strong/insufficient/non-configurable?

Thanks,

Signature · 2 lines
Vasiliy Kulikov
http://www.openwall.com - bringing security into open computing environments
#3 lore raw
Quoted 4 lines · Vasiliy Kulikov
Quoted 1 line

I have no opinion on whether it's a good idea to include the feature or not.

Why not? Have you some specific complains where it can be perhaps too strong/insufficient/non-configurable?

No, not at all. I just haven't had the need for this myself, and I'm not enough of a security person to judge whether the vulnerability addressed by the patch is a relevant one. E.g. if all the sensitive information you are hiding in procfs is still available through netlink, your patch is pointless. Similarly if there is no recorded case of an attack that relies on any of the information in procfs.

Arnd

#4 lore raw

Arnd,

Quoted 3 lines · Arnd Bergmann

E.g. if all the sensitive information
you are hiding in procfs is still available through netlink, your patch is pointless.

Ah, I've complitely missed this piece of a puzzle! :( With procfs, proc connector and taskstats (probably, something else) should be restricted too.

Thank you very much for this notice!

Signature · 2 lines
Vasiliy Kulikov
http://www.openwall.com - bringing security into open computing environments
#5 lore raw
In reply to Arnd Bergmann
Quoted 6 lines · Arnd Bergmann
Quoted 4 lines · Vasiliy Kulikov
Quoted 1 line

I have no opinion on whether it's a good idea to include the feature or not.

Why not?  Have you some specific complains where it can be perhaps too strong/insufficient/non-configurable?

Quoted 7 lines

No, not at all. I just haven't had the need for this myself, and I'm not enough of a security person to judge whether the vulnerability addressed by the patch is a relevant one. E.g. if all the sensitive information you are hiding in procfs is still available through netlink, your patch is pointless. Similarly if there is no recorded case of an attack that relies on any of the information in procfs.

Is this interface somewhere documented?
IOW how is it possible to get all processes via netlink?

Signature · 2 lines
Thanks,
//richard
#6 lore raw
In reply to Vasiliy Kulikov

Plese cc Alexey on procfs things.

Quoted 11 lines · On Wed, 15 Jun 2011 22:51:35 +0400 Vasiliy Kulikov <[email protected]> wrote:

This patch series adds support of procfs mount options and adds mount options to restrict /proc/<pid>/ directories to owners and /proc/<pid>/net/* to root. Additional group may be defined via gid=, and this group will be privileged to study others /proc/<pid>/ and networking information.

Similar features are implemented for old kernels in -ow patches (for Linux 2.2 and 2.4) and for Linux 2.6 in -grsecurity, but both of them are implemented as configure options, not cofigurable in runtime, with changes of gid of /proc/<pid>/, and without backward-compatible /proc/<pid>/net/* handling.

This all seems highly specific to one particular set of requirements. We have one set of access permission rules and then dive into procfs and hard-wire those rules into the implementation? What happens if someone else has a similar but slightly different set of requirements? More kernel patches?

IOW is there some more general way of doing all this? <handwaving>Like better permissions/chmod support in procfs and an inherited-across-fork per-process procfs permissions mask.</handwaving>

Does all this code support `mount -o remount' as expected?

#7 lore raw

Hi,

First of all, to make it clear, this specific patch is not proposed anymore because it doesn't restrict taskstats which can be used to gather similar information. The patch working with taskstats (and without hidepid=2) was not yet posted on LKML, but is available here:

http://www.openwall.com/lists/kernel-hardening/2011/06/19/3

Quoted 6 lines · Andrew Morton
Quoted 6 lines

This patch series adds support of procfs mount options and adds mount options to restrict /proc/<pid>/ directories to owners and /proc/<pid>/net/* to root. Additional group may be defined via gid=, and this group will be privileged to study others /proc/<pid>/ and networking information.

Quoted 7 lines
Quoted 5 lines

Similar features are implemented for old kernels in -ow patches (for Linux 2.2 and 2.4) and for Linux 2.6 in -grsecurity, but both of them are implemented as configure options, not cofigurable in runtime, with changes of gid of /proc/<pid>/, and without backward-compatible /proc/<pid>/net/* handling.

This all seems highly specific to one particular set of requirements.

Yes, I admit this. The problem with procfs is that it's possible to chmod/chown some procfs files, but not /proc/PID/*. Even if make it possible to chmod/chown them (and introducing an inodes revalidation on execve() setuid and similar binaries) it is still racy - new processes would have /proc/PID/ and some files inside with perms=0555. So, for more generic mechanism something like umask is needed. The patch in question implements 2 border cases:

1) relaxed. umask=0555.

2) restricted. umask=0550 (with tricky gid) and files are still not chmod'able.

More generic solution (I'm not suggesting it, but merely discussing) would use some user-supplied set of files to restrict access to (or, better, the set of allowed files because white list is almost always better than black list). Maybe this one:

mount -t proc -o "pid_allow=exe,status,comm,oom_*" proc /proc

And without pid_allow it would behave like pid_allow=*. "pid_allow=." would deny access to the whole /proc/PID.

This would be a bit inconsistent with current permissions because e.g. if use pid_allow=environ then environ file would not be accessible because of posix permissions. Hierarchical mode (pid_allow=fd/1) is not allowed too.

But it wouldn't work with taskstats. It needs its own set of allowed fields or field sets like delayacct,csw,bacct,xacct.

Quoted 3 lines

IOW is there some more general way of doing all this? <handwaving>Like better permissions/chmod support in procfs and an inherited-across-fork per-process procfs permissions mask.</handwaving>

I don't know such way, but it would ease procfs logic.

Quoted 1 line

Does all this code support `mount -o remount' as expected?

Yes.

Thanks,

Signature · 2 lines
Vasiliy Kulikov
http://www.openwall.com - bringing security into open computing environments
#8 lore raw
Quoted 11 lines · Vasiliy Kulikov

More generic solution (I'm not suggesting it, but merely discussing) would use some user-supplied set of files to restrict access to (or, better, the set of allowed files because white list is almost always better than black list). Maybe this one:

mount -t proc -o "pid_allow=exe,status,comm,oom_*" proc /proc

And without pid_allow it would behave like pid_allow=*.

"pid_allow=."
would deny access to the whole /proc/PID.

I mean "pid_allow=", of course.

Quoted 4 lines

This would be a bit inconsistent with current permissions because e.g. if use pid_allow=environ then environ file would not be accessible because of posix permissions. Hierarchical mode (pid_allow=fd/1) is not allowed too.

For hierarchical mode: attr_allowed, tgid_allowed, tid_allowed.

Thanks,

Signature · 2 lines
Vasiliy Kulikov
http://www.openwall.com - bringing security into open computing environments
#9 lore raw
In reply to Vasiliy Kulikov
Quoted 16 lines · Vasiliy Kulikov
Quoted 1 line

This all seems highly specific to one particular set of requirements.

Yes, I admit this. The problem with procfs is that it's possible to chmod/chown some procfs files, but not /proc/PID/*. Even if make it possible to chmod/chown them (and introducing an inodes revalidation on execve() setuid and similar binaries) it is still racy - new processes would have /proc/PID/ and some files inside with perms=0555. So, for more generic mechanism something like umask is needed. The patch in question implements 2 border cases:

1) relaxed. umask=0555.

2) restricted. umask=0550 (with tricky gid) and files are still not chmod'able.

Quoted 6 lines

More generic solution (I'm not suggesting it, but merely discussing) would use some user-supplied set of files to restrict access to (or, better, the set of allowed files because white list is almost always better than black list). Maybe this one:

mount -t proc -o "pid_allow=exe,status,comm,oom_*" proc /proc

Does this scheme make sense? Should I rensend the patch with these architecture?

pid_allow=, tid_allow=, attr_allow= and watch_gid= or smth like that.

Thanks,

Signature · 2 lines
Vasiliy Kulikov
http://www.openwall.com - bringing security into open computing environments
#10 lore raw
Quoted 6 lines · Vasiliy Kulikov
Quoted 1 line

    mount -t proc -o "pid_allow=exe,status,comm,oom_*" proc /proc

Does this scheme make sense?  Should I rensend the patch with these architecture?

pid_allow=, tid_allow=, attr_allow= and watch_gid= or smth like that.

Wildcards are scary, can we please not do them from the beginning.