Skip to contentSkip to conversation

remove CONFIG_SECURITY_FILE_CAPABILITIES compile option

PATCH In review Serge E. Hallyn 3 participants 5 messages Nov 23, 2009 – Nov 24, 2009

Subsystems (2)
lore
Share & tools
lore
#0 lore raw

PATCH 1/1 remove CONFIG_SECURITY_FILE_CAPABILITIES compile option

As far as I know, all distros currently ship kernels with default CONFIG_SECURITY_FILE_CAPABILITIES=y. Since having the option on leaves a 'no_file_caps' option to boot without file capabilities, the main reason to keep the option is that turning it off saves you (on my s390x partition) 5k. In particular, vmlinux sizes came to:

without patch fscaps=n: 53598392
without patch fscaps=y: 53603406
with this patch applied: 53603342

with the security-next tree.

Against this we must weigh the fact that there is no simple way for userspace to figure out whether file capabilities are supported, while things like per-process securebits, capability bounding sets, and adding bits to pI if CAP_SETPCAP is in pE are not supported with SECURITY_FILE_CAPABILITIES=n, leaving a bit of a problem for applications wanting to know whether they can use them and/or why something failed.

It also adds another subtly different set of semantics which we must maintain at the risk of severe security regressions.

So this patch removes the SECURITY_FILE_CAPABILITIES compile option. It drops the kernel size by about 50k over the stock SECURITY_FILE_CAPABILITIES=y kernel, by removing the cap_limit_ptraced_target() function.

Changelog:
Nov 20: remove cap_limit_ptraced_target() as it's logic
was ifndef'ed.

---
 include/linux/capability.h |    2 -
 include/linux/init_task.h  |    4 --
 kernel/capability.c        |    2 -
 security/Kconfig           |    9 -----
 security/commoncap.c       |   72 +------------------------------------------
 5 files changed, 2 insertions(+), 87 deletions(-)
include/linux/capability.h +0 -2
index c8f2a5f..39e5ff5 100644
@@ -92,9 +92,7 @@ struct vfs_cap_data {
 #define _KERNEL_CAPABILITY_VERSION _LINUX_CAPABILITY_VERSION_3
 #define _KERNEL_CAPABILITY_U32S    _LINUX_CAPABILITY_U32S_3
 
-#ifdef CONFIG_SECURITY_FILE_CAPABILITIES
 extern int file_caps_enabled;
-#endif
 
 typedef struct kernel_cap_struct {
 	__u32 cap[_KERNEL_CAPABILITY_U32S];
include/linux/init_task.h +0 -4
index 21a6f5d..8d10aa7 100644
@@ -83,16 +83,12 @@ extern struct group_info init_groups;
 #define INIT_IDS
 #endif
 
-#ifdef CONFIG_SECURITY_FILE_CAPABILITIES
 /*
  * Because of the reduced scope of CAP_SETPCAP when filesystem
  * capabilities are in effect, it is safe to allow CAP_SETPCAP to
  * be available in the default configuration.
  */
 # define CAP_INIT_BSET  CAP_FULL_SET
-#else
-# define CAP_INIT_BSET  CAP_INIT_EFF_SET
-#endif
 
 #ifdef CONFIG_TREE_PREEMPT_RCU
 #define INIT_TASK_RCU_PREEMPT(tsk)					\
kernel/capability.c +0 -2
index c2316d3..ab0aaf6 100644
@@ -29,7 +29,6 @@ EXPORT_SYMBOL(__cap_empty_set);
 EXPORT_SYMBOL(__cap_full_set);
 EXPORT_SYMBOL(__cap_init_eff_set);
 
-#ifdef CONFIG_SECURITY_FILE_CAPABILITIES
 int file_caps_enabled = 1;
 
 static int __init file_caps_disable(char *str)
@@ -38,7 +37,6 @@ static int __init file_caps_disable(char *str)
 	return 1;
 }
 __setup("no_file_caps", file_caps_disable);
-#endif
 
 /*
  * More recent versions of libcap are available from:
security/Kconfig +0 -9
index 95cc089..226b955 100644
@@ -91,15 +91,6 @@ config SECURITY_PATH
 	  implement pathname based access controls.
 	  If you are unsure how to answer this question, answer N.
 
-config SECURITY_FILE_CAPABILITIES
-	bool "File POSIX Capabilities"
-	default n
-	help
-	  This enables filesystem capabilities, allowing you to give
-	  binaries a subset of root's powers without using setuid 0.
-
-	  If in doubt, answer N.
-
 config INTEL_TXT
 	bool "Enable Intel(R) Trusted Execution Technology (Intel(R) TXT)"
 	depends on HAVE_INTEL_TXT
security/commoncap.c +2 -70

126 lines not on this page · raw message

Signature · 1 line
1.6.1
#1 lore raw
Quoted 21 lines · Serge E. Hallyn

As far as I know, all distros currently ship kernels with default CONFIG_SECURITY_FILE_CAPABILITIES=y. Since having the option on leaves a 'no_file_caps' option to boot without file capabilities, the main reason to keep the option is that turning it off saves you (on my s390x partition) 5k. In particular, vmlinux sizes came to:

without patch fscaps=n: 53598392
without patch fscaps=y: 53603406
with this patch applied: 53603342

with the security-next tree.

Against this we must weigh the fact that there is no simple way for userspace to figure out whether file capabilities are supported, while things like per-process securebits, capability bounding sets, and adding bits to pI if CAP_SETPCAP is in pE are not supported with SECURITY_FILE_CAPABILITIES=n, leaving a bit of a problem for applications wanting to know whether they can use them and/or why something failed.

Quoted 2 lines

It also adds another subtly different set of semantics which we must maintain at the risk of severe security regressions.

Hi,

This is or is not the same security option that is referred to in (new:) http://bugzilla.kernel.org/show_bug.cgi?id=14675 ?

Quoted 18 lines

So this patch removes the SECURITY_FILE_CAPABILITIES compile option. It drops the kernel size by about 50k over the stock SECURITY_FILE_CAPABILITIES=y kernel, by removing the cap_limit_ptraced_target() function.

Changelog:
Nov 20: remove cap_limit_ptraced_target() as it's logic
was ifndef'ed.

---
 include/linux/capability.h |    2 -
 include/linux/init_task.h  |    4 --
 kernel/capability.c        |    2 -
 security/Kconfig           |    9 -----
 security/commoncap.c       |   72 +------------------------------------------
 5 files changed, 2 insertions(+), 87 deletions(-)
Signature · 1 line
~Randy
#2 lore raw
Quoted 22 lines · Randy Dunlap
Quoted 21 lines · Serge E. Hallyn

As far as I know, all distros currently ship kernels with default CONFIG_SECURITY_FILE_CAPABILITIES=y. Since having the option on leaves a 'no_file_caps' option to boot without file capabilities, the main reason to keep the option is that turning it off saves you (on my s390x partition) 5k. In particular, vmlinux sizes came to:

without patch fscaps=n: 53598392
without patch fscaps=y: 53603406
with this patch applied: 53603342

with the security-next tree.

Against this we must weigh the fact that there is no simple way for userspace to figure out whether file capabilities are supported, while things like per-process securebits, capability bounding sets, and adding bits to pI if CAP_SETPCAP is in pE are not supported with SECURITY_FILE_CAPABILITIES=n, leaving a bit of a problem for applications wanting to know whether they can use them and/or why something failed.

Quoted 7 lines
Quoted 2 lines

It also adds another subtly different set of semantics which we must maintain at the risk of severe security regressions.

Hi,

This is or is not the same security option that is referred to in (new:) http://bugzilla.kernel.org/show_bug.cgi?id=14675 ?

I don't know. I would assume he meant 'CONFIG_SECURITY', but that wouldn't make sense since it should make no difference if no specific LSM is turned on.

I will subscribe to the bug so I can see his response. If it is in fact CONFIG_SECURITY_FILE_CAPABILITIES, then a binary which is setuid-root may be installed with some file capabilities, but without enough caps to do what it needs.

thanks,
-serge

#3 lore raw
In reply to Serge E. Hallyn
Quoted 21 lines · Serge E. Hallyn

As far as I know, all distros currently ship kernels with default CONFIG_SECURITY_FILE_CAPABILITIES=y. Since having the option on leaves a 'no_file_caps' option to boot without file capabilities, the main reason to keep the option is that turning it off saves you (on my s390x partition) 5k. In particular, vmlinux sizes came to:

without patch fscaps=n: 53598392
without patch fscaps=y: 53603406
with this patch applied: 53603342

with the security-next tree.

Against this we must weigh the fact that there is no simple way for userspace to figure out whether file capabilities are supported, while things like per-process securebits, capability bounding sets, and adding bits to pI if CAP_SETPCAP is in pE are not supported with SECURITY_FILE_CAPABILITIES=n, leaving a bit of a problem for applications wanting to know whether they can use them and/or why something failed.

Quoted 5 lines

It also adds another subtly different set of semantics which we must maintain at the risk of severe security regressions.

So this patch removes the SECURITY_FILE_CAPABILITIES compile option. It drops the kernel size by about 50k over the stock

As Andrew points out, not 50k! Try about 50 bytes :) Sorry about that.

-serge

maintainer
#4 lore raw
In reply to Serge E. Hallyn
Quoted 11 lines · Serge E. Hallyn

So this patch removes the SECURITY_FILE_CAPABILITIES compile option. It drops the kernel size by about 50k over the stock SECURITY_FILE_CAPABILITIES=y kernel, by removing the cap_limit_ptraced_target() function.

Changelog:
Nov 20: remove cap_limit_ptraced_target() as it's logic
was ifndef'ed.

Applied to
git://git.kernel.org/pub/scm/linux/kernel/git/jmorris/security-testing-2.6#next

Signature · 2 lines
James Morris
<[email protected]>