Using the Fedora 20 targeted policy, running check_assertions requires
an avtab with around 22 million elements. With the default limit of 4096
buckets, performance is abysmal: it takes more than an hour to populate
the hash. Profiling shows most of that time under avtab_search_node.
This patch increases the hash from 13 to 20 bits and to a maximum of
1048576 buckets. The time for check_assertions on that policy is reduced
to about 3 minutes, which is enough to re-enable those checks as part of
the build process.
A full size table will allocate 4-8 MB of memory, up from 16-32 KB. In a
cursory review, these tables are usually short-lived and only 1-3 are
allocated together. Compared to the cost of entries in this table (up to
1 GB using the same policy), this isn't a significant increase.
Signed-off-by: John Brooks <john.brooks@jolla.com>
gcc puts literal strings lie in read-only memory. On x86_64, trying to
write to them triggers a segmentation fault.
To detect such issues at build time, variables holding a pointer to such
strings should be "const char*". "gcc -Wwrite-strings" warns when using
non-const pointers to literal strings.
Remove gcc warnings by adding const to local variables and argumens of
internal functions.
This does *not* fix this warning:
policydb_public.c:208:10: warning: passing argument 2 of 'hashtab_search' discards 'const' qualifier from pointer target type
return (hashtab_search(p->p.p_classes.table, PACKET_CLASS_NAME) ==
^
In file included from ../include/sepol/policydb/symtab.h:16:0,
from ../include/sepol/policydb/policydb.h:60,
from policydb_public.c:4:
../include/sepol/policydb/hashtab.h:98:24: note: expected 'hashtab_key_t' but argument is of type 'const char *'
extern hashtab_datum_t hashtab_search(hashtab_t h, const hashtab_key_t k);
^
Moreover the "const" word in hashtab_search prototype does not make the
second parameter "const char*" but "char* const".
Acked-by: Steve Lawrence <slawrence@tresys.com>
Set DISABLE_CIL=y to build libsepol without CIL support, e.g
make DISABLE_CIL=y
To enable CIL support in libsepol, set DISABLE_CIL=n. This is the default
if not specified.
Signed-off-by: Steve Lawrence <slawrence@tresys.com>
With pp modules, the target platform information comes form the base
module. However, CIL modules have no concept of target platform. So it
must come from somewhere else. This adds an API function that allows
setting the target platform.
Signed-off-by: Steve Lawrence <slawrence@tresys.com>
This will return mls/validatetrans constraint information for each
expression in a buffer. If POLICY_KERN version is >=
POLICYDB_VERSION_CONSTRAINT_NAMES then the policy defined types/attributes
will be returned.
Signed-off-by: Richard Haines <richard_c_haines@btinternet.com>
Adds policy source defined 'type' or 'typeattribute' names to
constraints by adding additional structures (->type_names->types) to a
binary policy.
Before this change all typeattributes were expanded to lists of types
and added to the constraint under ->names. This made it difficult for
system admins to determine from the policy source what attribute
needed to be updated. To facilitate analysis of constraint failures
a new function has also been added, see sepol_compute_av_reason_buffer.
As additional structures have been added to policy, the policy version
is also updated (POLICYDB_VERSION_CONSTRAINT_NAMES). There is also a
corresponding kernel patch to handle the additional structures.
sepol_compute_av_reason_buffer is an extended version of
sepol_compute_av_reason. This will return a buffer with constraint
expression information, containing the constrain type, class, perms,
keywords etc.. It will also contain which constraint expr failed plus
the final outcome. The buffer MUST be free'd with free(3).
The type information output by sepol_compute_av_reason_buffer depends on
the policy version:
If >= POLICYDB_VERSION_CONSTRAINT_NAMES, then the output will be
whatever was in the original policy (type or attribute names).
If < POLICYDB_VERSION_CONSTRAINT_NAMES, then the output will be
the types listed in the constraint (as no attribute information is
available in these versions).
For users and roles whatever policy version, only the names are listed
(as role attributes are not currently held in the constraint).
Also added are two functions that obtain the class and permissions
from a binary policy file that has been loaded for testing:
sepol_string_to_security_class
sepol_string_to_av_perm
Signed-off-by: Richard Haines <richard_c_haines@btinternet.com>
We currently have a mechanism in which the default user, role, and range
can be picked up from the source or the target object. This implements
the same thing for types. The kernel will override this with type
transition rules and similar. This is just the default if nothing
specific is given.
Signed-off-by: Eric Paris <eparis@redhat.com>
Acked-by: Dan Walsh <dwalsh@redhat.com>
Currently the packet class in SELinux is not checked if there are no
SECMARK rules in the security or mangle netfilter tables. Similarly, the
peer class is not checked if there is no NetLabel or labeled IPSEC. Some
systems prefer that these classes are always checked, for example, to
protect the system should the netfilter rules fail to load or if the
nefilter rules were maliciously flushed.
Add the always_check_network policy capability which, when enabled, treats
these mechanisms as enabled, even if there are no labeling rules.
Signed-off-by: Chris PeBenito <cpebenito@tresys.com>
Signed-off-by: Eric Paris <eparis@redhat.com>
We would like to be able to say that the user, role, or range of a newly
created object should be based on the user, role, or range of either the
source or the target of the creation operation. aka, for a new file
this could be the user of the creating process or the user or the parent
directory. This patch implements the new language and the policydb
support to give this information to the kernel.
Signed-off-by: Eric Paris <eparis@redhat.com>
Acked-by: Dan Walsh <dwalsh@redhat.com>
This patches moves some ebitmap functions (and, xor, not, etc.) from
mcstrans into libsepol, where they really belong and could be used by
other applications (e.g. CIL)
Signed-off-by: Eric Paris <eparis@redhat.com>
Acked-by: Dan Walsh <dwalsh@redhat.com>
Change the default "make" target for the libraries from "install" to
"all" in the makefiles.
Signed-off-by: Guido Trentalancia <guido@trentalancia.com>
Signed-off-by: Eric Paris <eparis@redhat.com>
Acked-by: Dan Walsh <dwalsh@redhat.com>
By default only the effective branch of a tunable conditional would be
expanded and written to raw policy, while all needless unused branches
would be discarded.
Add a new option '-P' or "--preserve_tunables" to the semodule program.
By default it is 0, if set to 1 then the above preserve_tunables flag
in the sepol_handle_t would be set to 1 accordingly.
Signed-off-by: Harry Ciao <qingtao.cao@windriver.com>
Signed-off-by: Eric Paris <eparis@redhat.com>
Acked-by: Dan Walsh <dwalsh@redhat.com>
Add flags to cond_bool_datum_t and cond_node_t structures to differentiate
the tunables' identifiers and conditionals from those of booleans.
Signed-off-by: Harry Ciao <qingtao.cao@windriver.com>
Signed-off-by: Eric Paris <eparis@redhat.com>
Acked-by: Dan Walsh <dwalsh@redhat.com>
When the rolemap and pointer to the base module are available, if
a non-zero bit in role_set_t.roles is a role attribute, expand it
before remap.
Note, during module compile the rolemap may not be available, the
potential duplicates of a regular role and the role attribute that
the regular role belongs to could be properly handled by
copy_role_allow() and copy_role_trans() during module expansion.
Take advantage of the role_val_to_struct[] of the base module, since
when role_set_expand() is invoked, the role_val_to_struct[] of the
out module may have not been established yet.
Also cleanup the error handling of role_set_expand().
Signed-off-by: Harry Ciao <qingtao.cao@windriver.com>
Signed-off-by: Steve Lawrence <slawrence@tresys.com>
Add support to read/write the flavor flag and roles ebitmap in the
role_datum_t structure from/to policy module, if its version is no less
than MOD_POLICYDB_VERSION_ROLEATTRIB.
Since the role ebitmap would be expanded and won't be written into
policy.X, neither is the flavor flag, kernel SELinux security server
needs no change, the maximum version number for policy.X needs no bump.
Signed-off-by: Harry Ciao <qingtao.cao@windriver.com>
Signed-off-by: Steve Lawrence <slawrence@tresys.com>
1. Add a uint32_t "flavor" field and an ebitmap "roles" to the
role_datum_t structure;
2. Add a new "attribute_role" statement and its handler to declare
a role attribute;
3. Modify declare_role() to setup role_datum_t.flavor according
to the isattr argument;
4. Add a new "roleattribute" rule and its handler, which will record
the regular role's (policy value - 1) into the role attribute's
role_datum_t.roles ebitmap;
5. Modify the syntax for the role-types rule only to define the
role-type associations;
6. Add a new role-attr rule to support the declaration of a single
role, and optionally the role attribute that the role belongs to;
7. Check if the new_role used in role-transition rule is a regular role;
8. Support to require a role attribute;
9. Modify symtab_insert() to allow multiple declarations only for
the regular role, while a role attribute can't be declared more than once
and can't share a same name with another regular role.
Signed-off-by: Harry Ciao <qingtao.cao@windriver.com>
Signed-off-by: Steve Lawrence <slawrence@tresys.com>
This patch adds libsepol support for filename_trans rules. These rules
allow one to make labeling decisions for new objects based partially on
the last path component. They are stored in a list. If we find that
the number of rules grows to an significant size I will likely choose to
store these in a hash, both in libsepol and in the kernel. But as long
as the number of such rules stays small, this should be good.
Signed-off-by: Eric Paris <eparis@redhat.com>
Signed-off-by: Steve Lawrence <slawrence@tresys.com>
Introduce the class support to role_trans and role_trans_rule
structures, which could be the subject class("process") or the
class that the newly created object belongs to.
Signed-off-by: Harry Ciao <qingtao.cao@windriver.com>
Signed-off-by: Steve Lawrence <slawrence@tresys.com>
On Wed, 2009-09-16 at 09:58 -0400, Joshua Brindle wrote:
> I'd rather have separate ocontext structs for each system. That way it
> is very easy to understand which ones apply to which system and you
> don't get a crazy out of context ocontext struct.
>
I looked into having separate ocontext structs but that would involve
changing a lot of files making the patch much larger and more intrusive.
> > } u;
> > union {
> > uint32_t sclass; /* security class for genfs */
> > @@ -313,6 +323,17 @@ typedef struct genfs {
> > #define OCON_NODE6 6 /* IPv6 nodes */
> > #define OCON_NUM 7
> >
> > +/* object context array indices for Xen */
> > +#define OCON_ISID 0 /* initial SIDs */
> > +#define OCON_PIRQ 1 /* physical irqs */
> > +#define OCON_IOPORT 2 /* io ports */
> > +#define OCON_IOMEM 3 /* io memory */
> > +#define OCON_DEVICE 4 /* pci devices */
> > +#define OCON_DUMMY1 5 /* reserved */
> > +#define OCON_DUMMY2 6 /* reserved */
> > +#define OCON_NUM 7
> > +
> > +
> >
> Should these be namespaced? What if <random other system> has io port
> objects? You'd have to align them with each other and you have a mess of
> keeping the numbers the same (you already do this with OCON_ISID)
Variables have been namespaced and there is no more overlap with
OCON_ISID.
> Also we are relying on having the same number of OCON's which isn't good
> I don't think. As much as I hate the policydb_compat_info (read: alot)
> why aren't we using that to say how many ocons a xen policy really has?
OCON_NUM is now dynamically read through policydb_compat_info.
> This is messy, why not an ocontext_selinux_free() and
> ocontext_xen_free() (note: I realize the xen_free() one won't do
> anything except freep the ocontext_t)
>
done.
> >
> > len = buf[1];
> > - if (len != strlen(target_str)&&
> > - (!alt_target_str || len != strlen(alt_target_str))) {
> > - ERR(fp->handle, "policydb string length %zu does not match "
> > - "expected length %zu", len, strlen(target_str));
> > + if (len> 32) {
> >
>
> magic number 32?
#defined.
Thanks for your input. Below is the updated patch for libsepol.
----
libsepol/include/sepol/policydb/policydb.h | 28 ++
libsepol/src/expand.c | 85 +++++++-
libsepol/src/policydb.c | 295
+++++++++++++++++++++++------
libsepol/src/policydb_internal.h | 1
libsepol/src/private.h | 4
libsepol/src/write.c | 93 ++++++++-
6 files changed, 443 insertions(+), 63 deletions(-)
Signed-off-by: Joshua Brindle <method@manicmethod.com>
This patch adds the ability to check on the value of the disable_dontaudit flag in the sepol handle. In the past the only way to know the value of this was to directly read the values from the handle. The get function provides a setter-getter symmetry similar to other functions found in libsepol.
Signed-off-by: Christopher Pardy <cpardy@redhat.com>
Signed-off-by: Stephen Smalley <sds@tycho.nsa.gov>
The boundry format mapped the primary field to a boolean in the
properties bitmap. This is appropriate for the kernel policy, but in
modular policy the primary field may be an integer that indicates the
primary type that is being aliased. In this case, the primary value cannot
be assumed to be boolean.
This patch creates a new module format that writes out the primary value
as was done before the boundry format.
Signed-off-by: Caleb Case <ccase@tresys.com>
Signed-off-by: Joshua Brindle <method@manicmethod.com>
Email: kaigai@ak.jp.nec.com
Subject: Thread/Child-Domain Assignment (rev.6)
Date: Tue, 07 Oct 2008 15:39:45 +0900
>> Hmm....
>> It seems to me what you pointed out is a bug of my patch. It prevents to deliver
>> actual number of type/attribute symbols to policy file, but it is unclear why does
>> it makes libsepol ignore the policyvers.
>> (I guess it may be a separated matter.)
>>
>>> Rather than trying to calculate the length without attributes I just removed
>>> the attribute check. This causes attributes to be written for all versions,
>>> but this should not cause any problems at all.
>> The reason why I injected such an ad-hoc code is that we cannot decide the policy
>> version written when type_attr_remove() is invoked.
>> Is it impossible to move it to policydb_write()?
>> It is invoked after the policyvers is fixed by caller.
>
> It isn't impossible. You are going to have to make it walk to type
> symbol table to calculate the length without attributes, then write
> that length instead of the total symtab length.
The attached patch enables to fixup the number of type/attribute entries
to be written. The type_attr_uncount() decrements the number of attribute
entries skipped at type_write().
At first, I had a plan to invoke type_attr_remove() with
hashtab_map_remove_on_error(), but it means the given policydb structure
is modified at policydb_write() and implicit changes to external interface.
Differences from the previous version are here:
Signed-off-by: Joshua Brindle <method@manicmethod.com>
This is the same off-by-one bug that was already fixed in the kernel.
(According to my understanding neither of these bugs has security
implications)
Signed-off-by: Vesa-Matti Kari <vmkari@cc.helsinki.fi>
Signed-off-by: Joshua Brindle <method@manicmethod.com>
John Brooks
Nicolas Iooss
Steve Lawrence
Stephen Smalley
Richard Haines
Dan Walsh
Alice Chu
Eric Paris
Chris PeBenito
Guido Trentalancia
Harry Ciao
Paul Nuzzi
Christopher Pardy
Caleb Case
Joshua Brindle
Vesa-Matti J Kari