seal system mappings
Subsystems (7)
v3 cover seal system mappings
From: Jeff Xu <[email protected]>
Seal vdso, vvar, sigpage, uprobes and vsyscall.
Those mappings are readonly or executable only, sealing can protect them from ever changing or unmapped during the life time of the process. For complete descriptions of memory sealing, please see mseal.rst [1].
System mappings such as vdso, vvar, and sigpage (for arm) are generated by the kernel during program initialization, and are sealed after creation.
Unlike the aforementioned mappings, the uprobe mapping is not established during program startup. However, its lifetime is the same as the process's lifetime [1]. It is sealed from creation.
The vdso, vvar, sigpage, and uprobe mappings all invoke the _install_special_mapping() function. As no other mappings utilize this function, it is logical to incorporate sealing logic within _install_special_mapping(). This approach avoids the necessity of modifying code across various architecture-specific implementations.
The vsyscall mapping, which has its own initialization function, is sealed in the XONLY case, it seems to be the most common and secure case of using vsyscall.
It is important to note that the CHECKPOINT_RESTORE feature (CRIU) may alter the mapping of vdso, vvar, and sigpage during restore operations. Consequently, this feature cannot be universally enabled across all systems. To address this, a kernel configuration option has been introduced to enable or disable this functionality.
[1] Documentation/userspace-api/mseal.rst
[2] https://lore.kernel.org/all/CABi2SkU9BRUnqf70-nksuMCQ+yyiWjo3fM4XkRkL-NrCZxYAyg@mail.gmail.com/
History:
V3:
Revert uprobe to v1 logic (Oleg Nesterov)
use CONFIG_SEAL_SYSTEM_MAPPINGS instead of _ALWAYS/_NEVER (Kees Cook) Move kernel cmd line from fs/exec.c to mm/mseal.c and misc. refactor (Liam R. Howlett)
V2:
https://lore.kernel.org/all/[email protected]/
Seal uprobe always (Oleg Nesterov)
Update comments and description (Randy Dunlap, Liam R.Howlett, Oleg Nesterov) Rebase to linux_main
V1:
https://lore.kernel.org/all/[email protected]/
Jeff Xu (1):
exec: seal system mappings
.../admin-guide/kernel-parameters.txt | 10 +++++ arch/x86/entry/vsyscall/vsyscall_64.c | 9 ++++- include/linux/mm.h | 12 ++++++ mm/mmap.c | 10 +++++ mm/mseal.c | 39 +++++++++++++++++++ security/Kconfig | 11 ++++++ 6 files changed, 89 insertions(+), 2 deletions(-)
Signature · 1 line
2.47.0.277.g8800431eea-goog
PATCH v3 1/1 exec: seal system mappings
From: Jeff Xu <[email protected]>
Seal vdso, vvar, sigpage, uprobes and vsyscall.
Those mappings are readonly or executable only, sealing can protect them from ever changing or unmapped during the life time of the process. For complete descriptions of memory sealing, please see mseal.rst [1].
System mappings such as vdso, vvar, and sigpage (for arm) are generated by the kernel during program initialization, and are sealed after creation.
Unlike the aforementioned mappings, the uprobe mapping is not established during program startup. However, its lifetime is the same as the process's lifetime [1]. It is sealed from creation.
The vdso, vvar, sigpage, and uprobe mappings all invoke the _install_special_mapping() function. As no other mappings utilize this function, it is logical to incorporate sealing logic within _install_special_mapping(). This approach avoids the necessity of modifying code across various architecture-specific implementations.
The vsyscall mapping, which has its own initialization function, is sealed in the XONLY case, it seems to be the most common and secure case of using vsyscall.
It is important to note that the CHECKPOINT_RESTORE feature (CRIU) may alter the mapping of vdso, vvar, and sigpage during restore operations. Consequently, this feature cannot be universally enabled across all systems. To address this, a kernel configuration option has been introduced to enable or disable this functionality.
[1] Documentation/userspace-api/mseal.rst
[2] https://lore.kernel.org/all/CABi2SkU9BRUnqf70-nksuMCQ+yyiWjo3fM4XkRkL-NrCZxYAyg@mail.gmail.com/
--- .../admin-guide/kernel-parameters.txt | 10 +++++ arch/x86/entry/vsyscall/vsyscall_64.c | 9 ++++- include/linux/mm.h | 12 ++++++ mm/mmap.c | 10 +++++ mm/mseal.c | 39 +++++++++++++++++++ security/Kconfig | 11 ++++++ 6 files changed, 89 insertions(+), 2 deletions(-)
Documentation/admin-guide/kernel-parameters.txt +10 -0
@@ -1538,6 +1538,16 @@ Permit 'security.evm' to be updated regardless of current integrity status. + exec.seal_system_mappings = [KNL] + Format: { no | yes } + Seal system mappings: vdso, vvar, sigpage, vsyscall, + uprobe. + This overwrites KCONFIG CONFIG_SEAL_SYSTEM_MAPPINGS + - 'no': do not seal system mappings. + - 'yes': seal system mappings. + If not specified or invalid, default is the KCONFIG value. + This option has no effect if CONFIG_64BIT=n + early_page_ext [KNL,EARLY] Enforces page_ext initialization to earlier stages so cover more early boot allocations. Please note that as side effect some optimizations
arch/x86/entry/vsyscall/vsyscall_64.c +7 -2
@@ -366,8 +366,13 @@ void __init map_vsyscall(void)set_vsyscall_pgtable_user_bits (swapper_pg_dir); } -if (vsyscall_mode ==XONLY ) -vm_flags_init (&gate_vma,VM_EXEC ); +if (vsyscall_mode ==XONLY ) { +unsigned long vm_flags =VM_EXEC ; + + vm_flags |=seal_system_mappings (); + +vm_flags_init (&gate_vma, vm_flags); + }BUILD_BUG_ON ((unsigned long )__fix_to_virt (VSYSCALL_PAGE ) != (unsigned long )VSYSCALL_ADDR );
include/linux/mm.h +12 -0
@@ -4238,4 +4238,16 @@ int arch_get_shadow_stack_status(struct task_struct *t, unsigned long __user *stint arch_set_shadow_stack_status (struct task_struct *t,unsigned long status);int arch_lock_shadow_stack_status (struct task_struct *t,unsigned long status); +#ifdef CONFIG_64BIT +/* +* return VM_SEALED if seal system mapping is enabled. +*/ +unsigned long seal_system_mappings (void ); +#else +static inline unsigned long seal_system_mappings (void ) +{ +return 0 ; +} +#endif +#endif /* _LINUX_MM_H */
mm/mmap.c +10 -0
@@ -2133,6 +2133,16 @@ struct vm_area_struct *_install_special_mapping( unsigned long addr, unsigned long len, unsigned long vm_flags, const struct vm_special_mapping *spec) { +/* +* At present, all mappings (vdso, vvar, sigpage, and uprobe) that +* invoke the _install_special_mapping function can be sealed. +* Therefore, it is logical to call the seal_system_mappings_enabled() +* function here. In the future, if this is not the case, i.e. if certain +* mappings cannot be sealed, then it would be necessary to move this +* check to the calling function. +*/ + vm_flags |= seal_system_mappings(); +return __install_special_mapping (mm, addr, len, vm_flags, (void *)spec, &special_mapping_vmops); }
mm/mseal.c +39 -0
@@ -7,6 +7,7 @@* Author: Jeff Xu <[email protected]> */ +#include <linux/fs_parser.h> #include <linux/mempolicy.h> #include <linux/mman.h> #include <linux/mm.h>
@@ -266,3 +267,41 @@ SYSCALL_DEFINE3(mseal, unsigned long, start, size_t, len, unsigned long, {return do_mseal (start, len, flags); } + +/* +* Kernel cmdline overwrite for CONFIG_SEAL_SYSTEM_MAPPINGS +*/ +enum seal_system_mappings_type { +SEAL_SYSTEM_MAPPINGS_DISABLED , +SEAL_SYSTEM_MAPPINGS_ENABLED +}; + +static enum seal_system_mappings_type seal_system_mappings_v__ro_after_init = +IS_ENABLED (CONFIG_SEAL_SYSTEM_MAPPINGS ) ?SEAL_SYSTEM_MAPPINGS_ENABLED : +SEAL_SYSTEM_MAPPINGS_DISABLED ; + +static const struct constant_table value_table_sys_mapping[]__initconst = { + {"no" ,SEAL_SYSTEM_MAPPINGS_DISABLED }, + {"yes" ,SEAL_SYSTEM_MAPPINGS_ENABLED }, + { } +}; + +static int __init early_seal_system_mappings_override (char *buf) +{ +if (!buf) +return -EINVAL ; + + seal_system_mappings_v =lookup_constant (value_table_sys_mapping, + buf, seal_system_mappings_v); +return 0 ; +} + +early_param ("exec.seal_system_mappings" , early_seal_system_mappings_override); + +unsigned long seal_system_mappings (void ) +{ +if (seal_system_mappings_v ==SEAL_SYSTEM_MAPPINGS_ENABLED ) +return VM_SEALED ; + +return 0 ; +}
security/Kconfig +11 -0
@@ -51,6 +51,17 @@ config PROC_MEM_NO_FORCE endchoice +config SEAL_SYSTEM_MAPPINGS + bool "seal system mappings" + default n + depends on 64BIT + depends on !CHECKPOINT_RESTORE + help + Seal system mappings such as vdso, vvar, sigpage, vsyscall, uprobes. + Note: CHECKPOINT_RESTORE might relocate vdso mapping during restore, + and remap will fail if the mapping is sealed, therefore + !CHECKPOINT_RESTORE is added as dependency. + config SECURITY bool "Enable different security models" depends on SYSFS
Signature · 1 line
2.47.0.277.g8800431eea-goog
Re: [PATCH v3 1/1] exec: seal system mappings
I'd prefer not to move forward with this until we have confirmation that adequate testing has been performed, given how invasive this change is, even if behind a flag (unless we explicitly mention it is untested in the Kconfig).
We are touching arch-specific stuff with VDSO, VVAR, etc. so we need to be cautious when we're in effect hooking an arch-specific function in mm.
Other than that, the actual patch isn't too crazy overall.
I think a sensible approach might be to only enable on known-good arches.
Quoted 26 lines · On Wed, Nov 13, 2024 at 07:16:02PM +0000, [email protected] wrote:
From: Jeff Xu <[email protected]>
Seal vdso, vvar, sigpage, uprobes and vsyscall.
Those mappings are readonly or executable only, sealing can protect them from ever changing or unmapped during the life time of the process. For complete descriptions of memory sealing, please see
mseal.rst[1].System mappings such as vdso, vvar, and sigpage (for arm) are generated by the kernel during program initialization, and are sealed after creation.
Unlike the aforementioned mappings, the uprobe mapping is not established during program startup. However, its lifetime is the same as the process's lifetime [1]. It is sealed from creation.
The vdso, vvar, sigpage, and uprobe mappings all invoke the
_install_special_mapping()function. As no other mappings utilize this function, it is logical to incorporate sealing logic within_install_special_mapping(). This approach avoids the necessity of modifying code across various architecture-specific implementations.The vsyscall mapping, which has its own initialization function, is sealed in the XONLY case, it seems to be the most common and secure case of using vsyscall.
Quoted 7 lines
It is important to note that the
CHECKPOINT_RESTOREfeature (CRIU) may alter the mapping of vdso, vvar, and sigpage during restore operations. Consequently, this feature cannot be universally enabled across all systems. To address this, a kernel configuration option has been introduced to enable or disable this functionality.[1]
Documentation/userspace-api/mseal.rst
It'd be nice to explicitly refer to this in the docs, it's not quite urgent though would be nice to be part of this series.
Quoted 18 lines
[2] https://lore.kernel.org/all/CABi2SkU9BRUnqf70-nksuMCQ+yyiWjo3fM4XkRkL-NrCZxYAyg@mail.gmail.com/
Signed-off-by Jeff Xu <[email protected]>--- .../admin-guide/kernel-parameters.txt | 10 +++++ arch/x86/entry/vsyscall/vsyscall_64.c | 9 ++++- include/linux/mm.h | 12 ++++++ mm/mmap.c | 10 +++++ mm/mseal.c | 39 +++++++++++++++++++ security/Kconfig | 11 ++++++ 6 files changed, 89 insertions(+), 2 deletions(-)Documentation/admin-guide/kernel-parameters.txt +0
-0@@ -1538,6 +1538,16 @@ Permit 'security.evm' to be updated regardless of current integrity status.
Documentation/admin-guide/kernel-parameters.txt line 1549
Quoted 10 lines
Documentation/admin-guide/kernel-parameters.txt +9
-0+ exec.seal_system_mappings = [KNL] + Format: { no | yes } + Seal system mappings: vdso, vvar, sigpage, vsyscall, + uprobe. + This overwrites KCONFIG CONFIG_SEAL_SYSTEM_MAPPINGS + - 'no': do not seal system mappings. + - 'yes': seal system mappings. + If not specified or invalid, default is the KCONFIG value. + This option has no effect if CONFIG_64BIT=n
Or if CONFIG_CHECKPOINT_RESTORE is not set. Please update to reference this also.
Quoted 10 lines
Documentation/admin-guide/kernel-parameters.txt +1
-0+ early_page_ext [KNL,EARLY] Enforces page_ext initialization to earlier stages so cover more early boot allocations. Please note that as side effect some optimizationsarch/x86/entry/vsyscall/vsyscall_64.c +0
-0@@ -366,8 +366,13 @@ void __init map_vsyscall(void)set_vsyscall_pgtable_user_bits (swapper_pg_dir);
arch/x86/entry/vsyscall/vsyscall_64.c void __init map_vsyscall(void) · line 374
Quoted 10 lines
arch/x86/entry/vsyscall/vsyscall_64.c +6
-2} -if (vsyscall_mode ==XONLY ) -vm_flags_init (&gate_vma,VM_EXEC ); +if (vsyscall_mode ==XONLY ) { +unsigned long vm_flags =VM_EXEC ; + + vm_flags |=seal_system_mappings (); + +vm_flags_init (&gate_vma, vm_flags);
Nit: remove weird whitespace above. Also might be worth adding a comment as to what we're doing here similar to the one in _install_special_mapping().
Quoted 31 lines
arch/x86/entry/vsyscall/vsyscall_64.c +1
-0+ }BUILD_BUG_ON ((unsigned long )__fix_to_virt (VSYSCALL_PAGE ) != (unsigned long )VSYSCALL_ADDR );include/linux/mm.h +12
-0@@ -4238,4 +4238,16 @@ int arch_get_shadow_stack_status(struct task_struct *t, unsigned long __user *stint arch_set_shadow_stack_status (struct task_struct *t,unsigned long status);int arch_lock_shadow_stack_status (struct task_struct *t,unsigned long status); +#ifdef CONFIG_64BIT +/* +* return VM_SEALED if seal system mapping is enabled. +*/ +unsigned long seal_system_mappings (void ); +#else +static inline unsigned long seal_system_mappings (void ) +{ +return 0 ; +} +#endif +#endif /* _LINUX_MM_H */ mm/mmap.c +0
-0@@ -2133,6 +2133,16 @@ struct vm_area_struct *_install_special_mapping( unsigned long addr, unsigned long len,
mm/mmap.c struct vm_area_struct *_install_special_mapping( · line 2143
Quoted 10 lines
mm/mmap.c +8
-0unsigned long vm_flags, const struct vm_special_mapping *spec) { +/* +* At present, all mappings (vdso, vvar, sigpage, and uprobe) that +* invoke the _install_special_mapping function can be sealed. +* Therefore, it is logical to call the seal_system_mappings_enabled() +* function here. In the future, if this is not the case, i.e. if certain +* mappings cannot be sealed, then it would be necessary to move this +* check to the calling function. +*/
Nice comment!
Quoted 64 lines
mm/mmap.c +2
-0+ vm_flags |=seal_system_mappings (); +return __install_special_mapping (mm, addr, len, vm_flags, (void *)spec, &special_mapping_vmops); }mm/mseal.c +39
-0@@ -7,6 +7,7 @@* Author: Jeff Xu <[email protected]> */ +#include <linux/fs_parser.h> #include <linux/mempolicy.h> #include <linux/mman.h> #include <linux/mm.h> @@ -266,3 +267,41 @@ SYSCALL_DEFINE3(mseal, unsigned long, start, size_t, len, unsigned long, {return do_mseal (start, len, flags); } + +/* +* Kernel cmdline overwrite for CONFIG_SEAL_SYSTEM_MAPPINGS +*/ +enum seal_system_mappings_type { +SEAL_SYSTEM_MAPPINGS_DISABLED , +SEAL_SYSTEM_MAPPINGS_ENABLED +}; + +static enum seal_system_mappings_type seal_system_mappings_v__ro_after_init = +IS_ENABLED (CONFIG_SEAL_SYSTEM_MAPPINGS ) ?SEAL_SYSTEM_MAPPINGS_ENABLED : +SEAL_SYSTEM_MAPPINGS_DISABLED ; + +static const struct constant_table value_table_sys_mapping[]__initconst = { + {"no" ,SEAL_SYSTEM_MAPPINGS_DISABLED }, + {"yes" ,SEAL_SYSTEM_MAPPINGS_ENABLED }, + { } +}; + +static int __init early_seal_system_mappings_override (char *buf) +{ +if (!buf) +return -EINVAL ; + + seal_system_mappings_v =lookup_constant (value_table_sys_mapping, + buf, seal_system_mappings_v); +return 0 ; +} + +early_param ("exec.seal_system_mappings" , early_seal_system_mappings_override); + +unsigned long seal_system_mappings (void ) +{ +if (seal_system_mappings_v ==SEAL_SYSTEM_MAPPINGS_ENABLED ) +return VM_SEALED ; + +return 0 ; +}security/Kconfig +0
-0
security/Kconfig config PROC_MEM_NO_FORCE · line 58
Quoted 8 lines
security/Kconfig +5
-0@@ -51,6 +51,17 @@ config PROC_MEM_NO_FORCE endchoice +config SEAL_SYSTEM_MAPPINGS + bool "seal system mappings" + default n + depends on 64BIT + depends on !CHECKPOINT_RESTORE
Would prefer to depend on actually tested architectures only.
Quoted 12 lines
security/Kconfig +6
-1+ help + Seal system mappings such as vdso, vvar, sigpage, vsyscall, uprobes. + Note: CHECKPOINT_RESTORE might relocate vdso mapping during restore, + and remap will fail if the mapping is sealed, therefore + !CHECKPOINT_RESTORE is added as dependency. + config SECURITY bool "Enable different security models" depends on SYSFS --2.47.0.277.g8800431eea-goog
Re: [PATCH v3 1/1] exec: seal system mappings
Hi Lorenzo
Quoted 6 lines · Lorenzo Stoakes
I'd prefer not to move forward with this until we have confirmation that adequate testing has been performed, given how invasive this change is, even if behind a flag (unless we explicitly mention it is untested in the
Kconfig).
Quoted 7 lines
We are touching arch-specific stuff with VDSO, VVAR, etc. so we need to be cautious when we're in effect hooking an arch-specific function in mm.
Other than that, the actual patch isn't too crazy overall.
I think a sensible approach might be to only enable on known-good arches.
I responded to this in the other email where you raised the same point.
Quoted 33 lines · On Wed, Nov 13, 2024 at 07:16:02PM +0000, [email protected] wrote:
Quoted 32 lines
From: Jeff Xu <[email protected]>
Seal vdso, vvar, sigpage, uprobes and vsyscall.
Those mappings are readonly or executable only, sealing can protect them from ever changing or unmapped during the life time of the process. For complete descriptions of memory sealing, please see
mseal.rst[1].System mappings such as vdso, vvar, and sigpage (for arm) are generated by the kernel during program initialization, and are sealed after creation.
Unlike the aforementioned mappings, the uprobe mapping is not established during program startup. However, its lifetime is the same as the process's lifetime [1]. It is sealed from creation.
The vdso, vvar, sigpage, and uprobe mappings all invoke the
_install_special_mapping()function. As no other mappings utilize this function, it is logical to incorporate sealing logic within_install_special_mapping(). This approach avoids the necessity of modifying code across various architecture-specific implementations.The vsyscall mapping, which has its own initialization function, is sealed in the XONLY case, it seems to be the most common and secure case of using vsyscall.
It is important to note that the
CHECKPOINT_RESTOREfeature (CRIU) may alter the mapping of vdso, vvar, and sigpage during restore operations. Consequently, this feature cannot be universally enabled across all systems. To address this, a kernel configuration option has been introduced to enable or disable this functionality.
Quoted 5 lines
Quoted 1 line
[1]
Documentation/userspace-api/mseal.rstIt'd be nice to explicitly refer to this in the docs, it's not quite urgent though would be nice to be part of this series.
will update mseal.rst next version.
Quoted 19 lines
Quoted 19 lines
[2] https://lore.kernel.org/all/CABi2SkU9BRUnqf70-nksuMCQ+yyiWjo3fM4XkRkL-NrCZxYAyg@mail.gmail.com/
Signed-off-by Jeff Xu <[email protected]>--- .../admin-guide/kernel-parameters.txt | 10 +++++ arch/x86/entry/vsyscall/vsyscall_64.c | 9 ++++- include/linux/mm.h | 12 ++++++ mm/mmap.c | 10 +++++ mm/mseal.c | 39 +++++++++++++++++++ security/Kconfig | 11 ++++++ 6 files changed, 89 insertions(+), 2 deletions(-)Documentation/admin-guide/kernel-parameters.txt +0
-0@@ -1538,6 +1538,16 @@ Permit 'security.evm' to be updated regardless of current integrity status.
Quoted 13 lines
Documentation/admin-guide/kernel-parameters.txt line 1549
Quoted 9 lines
Documentation/admin-guide/kernel-parameters.txt +9
-0+ exec.seal_system_mappings = [KNL] + Format: { no | yes } + Seal system mappings: vdso, vvar, sigpage, vsyscall, + uprobe. + This overwrites KCONFIG CONFIG_SEAL_SYSTEM_MAPPINGS + - 'no': do not seal system mappings. + - 'yes': seal system mappings. + If not specified or invalid, default is the KCONFIG value. + This option has no effect if CONFIG_64BIT=nOr if
CONFIG_CHECKPOINT_RESTOREis not set. Please update to reference this also.
I will update this part. Liam has a similar comment.
Quoted 12 lines
Quoted 12 lines
Documentation/admin-guide/kernel-parameters.txt +1
-0+ early_page_ext [KNL,EARLY] Enforces page_ext initialization to earlier stages so cover more early boot allocations. Please note that as side effect some optimizationsarch/x86/entry/vsyscall/vsyscall_64.c +0
-0@@ -366,8 +366,13 @@ void __init map_vsyscall(void)set_vsyscall_pgtable_user_bits (swapper_pg_dir); }
Quoted 12 lines
arch/x86/entry/vsyscall/vsyscall_64.c void __init map_vsyscall(void) · line 374
Quoted 8 lines
arch/x86/entry/vsyscall/vsyscall_64.c +6
-2-if (vsyscall_mode ==XONLY ) -vm_flags_init (&gate_vma,VM_EXEC ); +if (vsyscall_mode ==XONLY ) { +unsigned long vm_flags =VM_EXEC ; + + vm_flags |=seal_system_mappings (); + +vm_flags_init (&gate_vma, vm_flags);Nit: remove weird whitespace above. Also might be worth adding a comment as to what we're doing here similar to the one in
_install_special_mapping().
Done.
Quoted 111 lines
Quoted 31 lines
arch/x86/entry/vsyscall/vsyscall_64.c +1
-0+ }BUILD_BUG_ON ((unsigned long )__fix_to_virt (VSYSCALL_PAGE ) != (unsigned long )VSYSCALL_ADDR );include/linux/mm.h +12
-0@@ -4238,4 +4238,16 @@ int arch_get_shadow_stack_status(struct task_struct *t, unsigned long __user *stint arch_set_shadow_stack_status (struct task_struct *t,unsigned long status);int arch_lock_shadow_stack_status (struct task_struct *t,unsigned long status); +#ifdef CONFIG_64BIT +/* +* return VM_SEALED if seal system mapping is enabled. +*/ +unsigned long seal_system_mappings (void ); +#else +static inline unsigned long seal_system_mappings (void ) +{ +return 0 ; +} +#endif +#endif /* _LINUX_MM_H */ mm/mmap.c +0
-0@@ -2133,6 +2133,16 @@ struct vm_area_struct *_install_special_mapping( unsigned long addr, unsigned long len,mm/mmap.c struct vm_area_struct *_install_special_mapping( · line 2143
Quoted 10 lines
mm/mmap.c +8
-0unsigned long vm_flags, const struct vm_special_mapping *spec) { +/* +* At present, all mappings (vdso, vvar, sigpage, and uprobe) that +* invoke the _install_special_mapping function can be sealed. +* Therefore, it is logical to call the seal_system_mappings_enabled() +* function here. In the future, if this is not the case, i.e. if certain +* mappings cannot be sealed, then it would be necessary to move this +* check to the calling function. +*/ Nice comment!
Quoted 67 lines
mm/mmap.c +2
-0+ vm_flags |=seal_system_mappings (); +return __install_special_mapping (mm, addr, len, vm_flags, (void *)spec, &special_mapping_vmops); }mm/mseal.c +39
-0@@ -7,6 +7,7 @@* Author: Jeff Xu <[email protected]> */ +#include <linux/fs_parser.h> #include <linux/mempolicy.h> #include <linux/mman.h> #include <linux/mm.h> @@ -266,3 +267,41 @@ SYSCALL_DEFINE3(mseal, unsigned long, start, size_t, len, unsigned long, {return do_mseal (start, len, flags); } + +/* +* Kernel cmdline overwrite for CONFIG_SEAL_SYSTEM_MAPPINGS +*/ +enum seal_system_mappings_type { +SEAL_SYSTEM_MAPPINGS_DISABLED , +SEAL_SYSTEM_MAPPINGS_ENABLED +}; + +static enum seal_system_mappings_type seal_system_mappings_v__ro_after_init = +IS_ENABLED (CONFIG_SEAL_SYSTEM_MAPPINGS ) ?SEAL_SYSTEM_MAPPINGS_ENABLED : +SEAL_SYSTEM_MAPPINGS_DISABLED ; + +static const struct constant_table value_table_sys_mapping[]__initconst = { + {"no" ,SEAL_SYSTEM_MAPPINGS_DISABLED }, + {"yes" ,SEAL_SYSTEM_MAPPINGS_ENABLED }, + { } +}; + +static int __init early_seal_system_mappings_override (char *buf) +{ +if (!buf) +return -EINVAL ; + + seal_system_mappings_v =lookup_constant (value_table_sys_mapping, + buf, seal_system_mappings_v); +return 0 ; +} + +early_param ("exec.seal_system_mappings" , early_seal_system_mappings_override); + +unsigned long seal_system_mappings (void ) +{ +if (seal_system_mappings_v ==SEAL_SYSTEM_MAPPINGS_ENABLED ) +return VM_SEALED ; + +return 0 ; +}security/Kconfig +0
-0@@ -51,6 +51,17 @@ config PROC_MEM_NO_FORCE endchoice
Quoted 8 lines
security/Kconfig config PROC_MEM_NO_FORCE · line 58
Quoted 5 lines
security/Kconfig +5
-0+config SEAL_SYSTEM_MAPPINGS + bool "seal system mappings" + default n + depends on 64BIT + depends on !CHECKPOINT_RESTOREWould prefer to depend on actually tested architectures only.
I responded in the other email where you raised the same point.
Thanks for reviewing
-Jeff
Quoted 12 lines
Quoted 12 lines
security/Kconfig +6
-1+ help + Seal system mappings such as vdso, vvar, sigpage, vsyscall, uprobes. + Note: CHECKPOINT_RESTORE might relocate vdso mapping during restore, + and remap will fail if the mapping is sealed, therefore + !CHECKPOINT_RESTORE is added as dependency. + config SECURITY bool "Enable different security models" depends on SYSFS --2.47.0.277.g8800431eea-goog
Re: [PATCH v3 1/1] exec: seal system mappings
In reply to [email protected]Quoted 1 line · * [email protected] <[email protected]> [241113 14:16]:
From: Jeff Xu <[email protected]>
Thanks, this looks much better. Some minor things below.
Quoted 7 lines
Seal vdso, vvar, sigpage, uprobes and vsyscall.
Those mappings are readonly or executable only, sealing can protect them from ever changing or unmapped during the life time of the process. For complete descriptions of memory sealing, please see
mseal.rst[1].
Quoted 7 lines
System mappings such as vdso, vvar, and sigpage (for arm) are generated by the kernel during program initialization, and are sealed after creation.
Unlike the aforementioned mappings, the uprobe mapping is not established during program startup. However, its lifetime is the same as the process's lifetime [1]. It is sealed from creation.
Why are you referencing mseal.rst for the uprobe mapping lifetime? I can't find anything in there about uprobe.
Quoted 11 lines
The vdso, vvar, sigpage, and uprobe mappings all invoke the
_install_special_mapping()function. As no other mappings utilize this function, it is logical to incorporate sealing logic within_install_special_mapping(). This approach avoids the necessity of modifying code across various architecture-specific implementations.The vsyscall mapping, which has its own initialization function, is sealed in the XONLY case, it seems to be the most common and secure case of using vsyscall.
Quoted 5 lines
It is important to note that the
CHECKPOINT_RESTOREfeature (CRIU) may alter the mapping of vdso, vvar, and sigpage during restore operations. Consequently, this feature cannot be universally enabled across all systems. To address this, a kernel configuration option has been introduced to enable or disable this functionality.
It also can't be used on 32 bit systems, as per your kernel-parameters changes (and mseal specification). This is missing from the changelog.
Quoted 18 lines
[1]
Documentation/userspace-api/mseal.rst
[2] https://lore.kernel.org/all/CABi2SkU9BRUnqf70-nksuMCQ+yyiWjo3fM4XkRkL-NrCZxYAyg@mail.gmail.com/Signed-off-by Jeff Xu <[email protected]>--- .../admin-guide/kernel-parameters.txt | 10 +++++ arch/x86/entry/vsyscall/vsyscall_64.c | 9 ++++- include/linux/mm.h | 12 ++++++ mm/mmap.c | 10 +++++ mm/mseal.c | 39 +++++++++++++++++++ security/Kconfig | 11 ++++++ 6 files changed, 89 insertions(+), 2 deletions(-)Documentation/admin-guide/kernel-parameters.txt +0
-0
Documentation/admin-guide/kernel-parameters.txt line 1545
Quoted 8 lines
Documentation/admin-guide/kernel-parameters.txt +5
-0@@ -1538,6 +1538,16 @@ Permit 'security.evm' to be updated regardless of current integrity status. + exec.seal_system_mappings = [KNL] + Format: { no | yes } + Seal system mappings: vdso, vvar, sigpage, vsyscall, + uprobe. + This overwrites KCONFIG CONFIG_SEAL_SYSTEM_MAPPINGS
^^^^^^^^^ overrides ?
Quoted 13 lines
Documentation/admin-guide/kernel-parameters.txt +5
-0+ - 'no': do not seal system mappings. + - 'yes': seal system mappings. + If not specified or invalid, default is the KCONFIG value. + This option has no effect if CONFIG_64BIT=n + early_page_ext [KNL,EARLY] Enforces page_ext initialization to earlier stages so cover more early boot allocations. Please note that as side effect some optimizationsarch/x86/entry/vsyscall/vsyscall_64.c +0
-0
arch/x86/entry/vsyscall/vsyscall_64.c void __init map_vsyscall(void) · line 373
Quoted 10 lines
arch/x86/entry/vsyscall/vsyscall_64.c +5
-2@@ -366,8 +366,13 @@ void __init map_vsyscall(void)set_vsyscall_pgtable_user_bits (swapper_pg_dir); } -if (vsyscall_mode ==XONLY ) -vm_flags_init (&gate_vma,VM_EXEC ); +if (vsyscall_mode ==XONLY ) { +unsigned long vm_flags =VM_EXEC ; + + vm_flags |=seal_system_mappings (); +
nit, extra line here.
Quoted 34 lines
arch/x86/entry/vsyscall/vsyscall_64.c +2
-0+vm_flags_init (&gate_vma, vm_flags); + }BUILD_BUG_ON ((unsigned long )__fix_to_virt (VSYSCALL_PAGE ) != (unsigned long )VSYSCALL_ADDR );include/linux/mm.h +12
-0@@ -4238,4 +4238,16 @@ int arch_get_shadow_stack_status(struct task_struct *t, unsigned long __user *stint arch_set_shadow_stack_status (struct task_struct *t,unsigned long status);int arch_lock_shadow_stack_status (struct task_struct *t,unsigned long status); +#ifdef CONFIG_64BIT +/* +* return VM_SEALED if seal system mapping is enabled. +*/ +unsigned long seal_system_mappings (void ); +#else +static inline unsigned long seal_system_mappings (void ) +{ +return 0 ; +} +#endif +#endif /* _LINUX_MM_H */ mm/mmap.c +0
-0@@ -2133,6 +2133,16 @@ struct vm_area_struct *_install_special_mapping( unsigned long addr, unsigned long len, unsigned long vm_flags, const struct vm_special_mapping *spec) {
mm/mmap.c struct vm_area_struct *_install_special_mapping( · line 2145
Quoted 10 lines
mm/mmap.c +10
-0+/* +* At present, all mappings (vdso, vvar, sigpage, and uprobe) that +* invoke the _install_special_mapping function can be sealed. +* Therefore, it is logical to call the seal_system_mappings_enabled() +* function here. In the future, if this is not the case, i.e. if certain +* mappings cannot be sealed, then it would be necessary to move this +* check to the calling function. +*/ + vm_flags |=seal_system_mappings (); +
Thank you for making this change into an or operation with a single function. It is much easier to figure out what is going on.
But.. this will add the VM_SEALED flag on any 64bit architecture that enables the SEAL_SYSTEM_MAPPINGS config. That will happen by bots with random config builds. I don't know if they have test cases that specifically unmap the vmas you are sealing (ppc64 probably tries to unmap the vdso).
I do know that I've had syzbot bugs that unmap _all_ vmas. I'm guessing you will get bot notification on these failures for any 64bit architecture. You may want to look into it to avoid such fuzzing failures, but we still need this to be tested somehow.
mm/mseal.c
Quoted 7 lines
mm/mmap.c +0
-0return __install_special_mapping (mm, addr, len, vm_flags, (void *)spec, &special_mapping_vmops); }mm/mseal.c +0
-0
Thank you for moving this to mseal.c
Quoted 9 lines
mm/mseal.c +1
-0@@ -7,6 +7,7 @@* Author: Jeff Xu <[email protected]> */ +#include <linux/fs_parser.h> #include <linux/mempolicy.h> #include <linux/mman.h> #include <linux/mm.h>
mm/mseal.c SYSCALL_DEFINE3(mseal, unsigned long, start, size_t, len, unsigned long, · line 272
Quoted 6 lines
mm/mseal.c +3
-0@@ -266,3 +267,41 @@ SYSCALL_DEFINE3(mseal, unsigned long, start, size_t, len, unsigned long, {return do_mseal (start, len, flags); } + +/* +* Kernel cmdline overwrite for CONFIG_SEAL_SYSTEM_MAPPINGS
overwrite or override? I think the difference is that overwrite implies permanence where override doesn't. I'm fine with either, it just reads a bit odd to me.
Quoted 43 lines
mm/mseal.c +35
-0+*/ +enum seal_system_mappings_type { +SEAL_SYSTEM_MAPPINGS_DISABLED , +SEAL_SYSTEM_MAPPINGS_ENABLED +}; + +static enum seal_system_mappings_type seal_system_mappings_v__ro_after_init = +IS_ENABLED (CONFIG_SEAL_SYSTEM_MAPPINGS ) ?SEAL_SYSTEM_MAPPINGS_ENABLED : +SEAL_SYSTEM_MAPPINGS_DISABLED ; + +static const struct constant_table value_table_sys_mapping[]__initconst = { + {"no" ,SEAL_SYSTEM_MAPPINGS_DISABLED }, + {"yes" ,SEAL_SYSTEM_MAPPINGS_ENABLED }, + { } +}; + +static int __init early_seal_system_mappings_override (char *buf) +{ +if (!buf) +return -EINVAL ; + + seal_system_mappings_v =lookup_constant (value_table_sys_mapping, + buf, seal_system_mappings_v); +return 0 ; +} + +early_param ("exec.seal_system_mappings" , early_seal_system_mappings_override); + +unsigned long seal_system_mappings (void ) +{ +if (seal_system_mappings_v ==SEAL_SYSTEM_MAPPINGS_ENABLED ) +return VM_SEALED ; + +return 0 ; +}security/Kconfig +0
-0@@ -51,6 +51,17 @@ config PROC_MEM_NO_FORCE endchoice
security/Kconfig config PROC_MEM_NO_FORCE · line 63
Quoted 10 lines
security/Kconfig +10
-0+config SEAL_SYSTEM_MAPPINGS + bool "seal system mappings" + default n + depends on 64BIT + depends on !CHECKPOINT_RESTORE + help + Seal system mappings such as vdso, vvar, sigpage, vsyscall, uprobes. + Note: CHECKPOINT_RESTORE might relocate vdso mapping during restore, + and remap will fail if the mapping is sealed, therefore + !CHECKPOINT_RESTORE is added as dependency.
You could also add a portion here about your override functionality on command line. "this can be disabled or enabled by..."
I really think having something that can be found by searching for mseal is really desirable here. That is, make menuconfig, press / for search, type mseal -> find this feature. If this was MSEAL_SYSTEM_MAPPINGS, searching for seal or mseal would work and would serve to link the config option to the mseal document.
Right now there is no information in the help that will allow it to be found by 'mseal'. There is also nothing in the documentation that states this exists, which you should probably update with this feature?
Thanks,
Liam
Re: [PATCH v3 1/1] exec: seal system mappings
Hi Liam,
Quoted 7 lines · Liam R. Howlett
Quoted 3 lines
Unlike the aforementioned mappings, the uprobe mapping is not established during program startup. However, its lifetime is the same as the process's lifetime [1]. It is sealed from creation.
Why are you referencing
mseal.rstfor the uprobe mapping lifetime? I can't find anything in there about uprobe.
This should be [2], thanks for checking.
Quoted 4 lines
It also can't be used on 32 bit systems, as per your kernel-parameters changes (and mseal specification). This is missing from the changelog.
sure, I will add that to the commit msg.
Quoted 6 lines
Documentation/admin-guide/kernel-parameters.txt line 1545
Quoted 5 lines
Documentation/admin-guide/kernel-parameters.txt +5
-0+ exec.seal_system_mappings = [KNL] + Format: { no | yes } + Seal system mappings: vdso, vvar, sigpage, vsyscall, + uprobe. + This overwrites KCONFIG CONFIG_SEAL_SYSTEM_MAPPINGS^^^^^^^^^ overrides ?
sure.
Quoted 10 lines
arch/x86/entry/vsyscall/vsyscall_64.c void __init map_vsyscall(void) · line 373
Quoted 7 lines
arch/x86/entry/vsyscall/vsyscall_64.c +5
-2-if (vsyscall_mode ==XONLY ) -vm_flags_init (&gate_vma,VM_EXEC ); +if (vsyscall_mode ==XONLY ) { +unsigned long vm_flags =VM_EXEC ; + + vm_flags |=seal_system_mappings (); +nit, extra line here.
removed.
Quoted 11 lines
But.. this will add the
VM_SEALEDflag on any 64bit architecture that enables theSEAL_SYSTEM_MAPPINGSconfig. That will happen by bots with random config builds. I don't know if they have test cases that specifically unmap the vmas you are sealing (ppc64 probably tries to unmap the vdso).I do know that I've had syzbot bugs that unmap _all_ vmas. I'm guessing you will get bot notification on these failures for any 64bit architecture. You may want to look into it to avoid such fuzzing failures, but we still need this to be tested somehow.
test_mremap_vdso.c
I found one selftest that could fail:tools/testing/selftests/x86/test_mremap_vdso.c
I could add tools/testing/selftests/x86/config and add CONFIG_SYSTEM_MAPPINGS=n there.
as instructed in selftest documentation [1]
[1] https://docs.kernel.org/dev-tools/kselftest.html#contributing-new-tests-details
Quoted 5 lines
overwrite or override? I think the difference is that overwrite implies permanence where override doesn't. I'm fine with either, it just reads a bit odd to me.
sure, changed to override
Quoted 12 lines
security/Kconfig config PROC_MEM_NO_FORCE · line 63
Quoted 11 lines
security/Kconfig +10
-0+config SEAL_SYSTEM_MAPPINGS + bool "seal system mappings" + default n + depends on 64BIT + depends on !CHECKPOINT_RESTORE + help + Seal system mappings such as vdso, vvar, sigpage, vsyscall, uprobes. + Note: CHECKPOINT_RESTORE might relocate vdso mapping during restore, + and remap will fail if the mapping is sealed, therefore + !CHECKPOINT_RESTORE is added as dependency.
Quoted 3 lines
You could also add a portion here about your override functionality on command line. "this can be disabled or enabled by..."
sure.
Quoted 6 lines
I really think having something that can be found by searching for mseal is really desirable here. That is, make menuconfig, press / for search, type mseal -> find this feature. If this was
MSEAL_SYSTEM_MAPPINGS, searching for seal or mseal would work and would serve to link the config option to the mseal document.
using "seal" would work here. I will add a note here to mseal.rst for reference.
Quoted 4 lines
Right now there is no information in the help that will allow it to be found by 'mseal'. There is also nothing in the documentation that states this exists, which you should probably update with this feature?
I will update mseal.rst to include this feature.
Thanks for reviewing.
-Jeff
You left kernel test bots and review unanswered on v2, which makes it difficult to know whether you actually addressed everything.
Please respond to all outstanding comments in the v2 thread so we know, thanks, even if it is to say 'this is no longer an issue as v3 addresses'.
Quoted 16 lines · On Wed, Nov 13, 2024 at 07:16:01PM +0000, [email protected] wrote:
From: Jeff Xu <[email protected]>
Seal vdso, vvar, sigpage, uprobes and vsyscall.
Those mappings are readonly or executable only, sealing can protect them from ever changing or unmapped during the life time of the process. For complete descriptions of memory sealing, please see
mseal.rst[1].System mappings such as vdso, vvar, and sigpage (for arm) are generated by the kernel during program initialization, and are sealed after creation.
Unlike the aforementioned mappings, the uprobe mapping is not established during program startup. However, its lifetime is the same as the process's lifetime [1]. It is sealed from creation.
Quoted 5 lines
The vdso, vvar, sigpage, and uprobe mappings all invoke the
_install_special_mapping()function. As no other mappings utilize this function, it is logical to incorporate sealing logic within_install_special_mapping(). This approach avoids the necessity of modifying code across various architecture-specific implementations.
Some arches unmap VDSO's which mseal prevents, so are those broken now? Did you test this?
Quoted 5 lines
The vsyscall mapping, which has its own initialization function, is sealed in the XONLY case, it seems to be the most common and secure case of using vsyscall.
Quoted 8 lines
It is important to note that the
CHECKPOINT_RESTOREfeature (CRIU) may alter the mapping of vdso, vvar, and sigpage during restore operations. Consequently, this feature cannot be universally enabled across all systems. To address this, a kernel configuration option has been introduced to enable or disable this functionality.[1]
Documentation/userspace-api/mseal.rst
[2] https://lore.kernel.org/all/CABi2SkU9BRUnqf70-nksuMCQ+yyiWjo3fM4XkRkL-NrCZxYAyg@mail.gmail.com/
I don't see any mention to testing, and this affects multiple different architectures.
Please describe what testing you performed and on what architectures.
I suggest we allow this only for architectures you have explicitly tested, especially as this is 'hidden' from arch maintainers who might find this change surprising.
Quoted 30 lines
History:
V3:
Revert uprobe to v1 logic (Oleg Nesterov)
useCONFIG_SEAL_SYSTEM_MAPPINGSinstead of _ALWAYS/_NEVER (Kees Cook) Move kernel cmd line fromfs/exec.ctomm/mseal.cand misc. refactor (Liam R. Howlett)V2:
https://lore.kernel.org/all/[email protected]/
Seal uprobe always (Oleg Nesterov)
Update comments and description (Randy Dunlap, Liam R.Howlett, Oleg Nesterov) Rebase tolinux_mainV1:
https://lore.kernel.org/all/[email protected]/Jeff Xu (1):
exec: seal system mappings.../admin-guide/kernel-parameters.txt | 10 +++++ arch/x86/entry/vsyscall/vsyscall_64.c | 9 ++++- include/linux/mm.h | 12 ++++++ mm/mmap.c | 10 +++++ mm/mseal.c | 39 +++++++++++++++++++ security/Kconfig | 11 ++++++ 6 files changed, 89 insertions(+), 2 deletions(-)--
2.47.0.277.g8800431eea-goog
Hi Lorenzo
Quoted 4 lines · Lorenzo Stoakes
You left kernel test bots and review unanswered on v2, which makes it difficult to know whether you actually addressed everything.
Thanks for reminding me, I got distracted previously. I responded to the test bots.
Quoted 3 lines
Please respond to all outstanding comments in the v2 thread so we know, thanks, even if it is to say 'this is no longer an issue as v3 addresses'.
All comments of v2 were addressed in V3.
Quoted 17 lines · On Wed, Nov 13, 2024 at 07:16:01PM +0000, [email protected] wrote:
Quoted 16 lines
From: Jeff Xu <[email protected]>
Seal vdso, vvar, sigpage, uprobes and vsyscall.
Those mappings are readonly or executable only, sealing can protect them from ever changing or unmapped during the life time of the process. For complete descriptions of memory sealing, please see
mseal.rst[1].System mappings such as vdso, vvar, and sigpage (for arm) are generated by the kernel during program initialization, and are sealed after creation.
Unlike the aforementioned mappings, the uprobe mapping is not established during program startup. However, its lifetime is the same as the process's lifetime [1]. It is sealed from creation.
Quoted 9 lines
Quoted 5 lines
The vdso, vvar, sigpage, and uprobe mappings all invoke the
_install_special_mapping()function. As no other mappings utilize this function, it is logical to incorporate sealing logic within_install_special_mapping(). This approach avoids the necessity of modifying code across various architecture-specific implementations.Some arches unmap VDSO's which mseal prevents, so are those broken now? Did you test this?
Do you happen to know which arch might unmap vdso ?
The information I collected so far is only CHECKPOINT_RESTORE would remap/unmap vdso. And if CHECKPOINT_RESTORE is enabled, Kconfig will prevent this from being enabled.
Quoted 11 lines
Quoted 11 lines
The vsyscall mapping, which has its own initialization function, is sealed in the XONLY case, it seems to be the most common and secure case of using vsyscall.
It is important to note that the
CHECKPOINT_RESTOREfeature (CRIU) may alter the mapping of vdso, vvar, and sigpage during restore operations. Consequently, this feature cannot be universally enabled across all systems. To address this, a kernel configuration option has been introduced to enable or disable this functionality.
Quoted 8 lines
Quoted 2 lines
[1]
Documentation/userspace-api/mseal.rst
[2] https://lore.kernel.org/all/CABi2SkU9BRUnqf70-nksuMCQ+yyiWjo3fM4XkRkL-NrCZxYAyg@mail.gmail.com/I don't see any mention to testing, and this affects multiple different architectures.
Please describe what testing you performed and on what architectures.
The tests are done in ChromeOS and Android on ARM and INTEL.
Quoted 4 lines
I suggest we allow this only for architectures you have explicitly tested, especially as this is 'hidden' from arch maintainers who might find this change surprising.
I thought the current approach aligns with Linus's suggestion of unifying vdso/vvar code [1]. I honestly think this is not architecture dependent, i.e. this doesn't require any specific CPU feature.
I could add ARCH_HAS_SEAL_SYSTEM_MAPPINGS in KCONFIG and enable this for x86_64 and arm64 for now, this would allow other architecture maintainers to have opportunities to test this .
[1] https://lore.kernel.org/all/CAHk-=wgTXVMBRuya5J0peujSrtunehRtzk=WVrm6njPhHrpTJw@mail.gmail.com/
Thanks for reviewing.
-Jeff