Ability to read the MKTME status from userspace (patch v2)
Subsystems (2)
PATCH 1/1 Ability to read the MKTME status from userspace (patch v2)
The intent of this patch is to provide visibility of the MKTME status to userspace. This is an important factor for firmware security related applilcations.
Changes since v1:
* Documentation/ABI/stable/securityfs-mktme-status (new file)
* include/uapi/misc/mktme_status.h (new file)
* Fixed MAINTAINER title.
* cpu.h: added values to the enumerands
* Renamed the function from get_mktme_status to mktme_get_status
* Improved Kconfig help
* Removed unnecessary license-related lines since there is a SPDX line
* Moved pr_fmt macro before the includes
* Turned global variables (mktme_dir, mktme_file) as static
* Removed BUFFER_SIZE macro
* No longer returning -1 for error, but using the previously error.
* No more pr_info for the normal behavior.
* Renamed this from a kernel driver to a kernel module.
--- .../ABI/stable/securityfs-mktme-status | 12 ++++ MAINTAINERS | 5 ++ arch/x86/include/asm/cpu.h | 8 +++ arch/x86/kernel/cpu/intel.c | 12 ++-- drivers/misc/Kconfig | 15 +++++ drivers/misc/Makefile | 1 + drivers/misc/mktme_status.c | 67 +++++++++++++++++++ include/uapi/misc/mktme_status.h | 12 ++++ 8 files changed, 127 insertions(+), 5 deletions(-) create mode 100644 Documentation/ABI/stable/securityfs-mktme-status create mode 100644 drivers/misc/mktme_status.c create mode 100644 include/uapi/misc/mktme_status.h
Documentation/ABI/stable/securityfs-mktme-status new file +12 -0
@@ -0,0 +1,12 @@ +What: /securityfs/mktme/status +Date: 24-Jun-2020 +KernelVersion: v5.7 +Contact: [email protected] +Description: The mktme securityfs interface exports the BIOS status + of the MKTME. +Values: For possible values see arch/x86/include/asm/cpu.h. + + Currently these values are: + 0 Enabled by BIOS + 1 Disabled by BIOS + 2 Uninitialized
MAINTAINERS +5 -0
@@ -11477,6 +11477,11 @@ W: https://linuxtv.org T: git git://linuxtv.org/media_tree.git F: drivers/media/radio/radio-miropcm20* +MKTME STATUS READING SUPPORT +M: Daniel Gutson <[email protected]> +S: Supported +F: drivers/misc/mktme_status.c + MMP SUPPORT R: Lubomir Rintel <[email protected]> L: [email protected] (moderated for non-subscribers)
arch/x86/include/asm/cpu.h +8 -0
@@ -26,6 +26,14 @@ struct x86_cpu {struct cpu cpu; }; +enum mktme_status_type { +MKTME_ENABLED =0 , +MKTME_DISABLED =1 , +MKTME_UNINITIALIZED =2 +}; + +extern enum mktme_status_type mktme_get_status (void ); +#ifdef CONFIG_HOTPLUG_CPU extern int arch_register_cpu (int num);extern void arch_unregister_cpu (int );
arch/x86/kernel/cpu/intel.c +7 -5
@@ -495,11 +495,7 @@ static void srat_detect_node(struct cpuinfo_x86 *c)#define TME_ACTIVATE_CRYPTO_ALGS (x) ((x >> 48) & 0xffff)/* Bits 63:48 */ #define TME_ACTIVATE_CRYPTO_AES_XTS_128 1 -/* Values for mktme_status (SW only construct) */ -#define MKTME_ENABLED 0 -#define MKTME_DISABLED 1 -#define MKTME_UNINITIALIZED 2 -static int mktme_status =MKTME_UNINITIALIZED ; +static enum mktme_status_type mktme_status =MKTME_UNINITIALIZED ;static void detect_tme (struct cpuinfo_x86 *c) {
@@ -1114,6 +1110,12 @@ bool handle_user_split_lock(struct pt_regs *regs, long error_code)return true; } +enum mktme_status_type mktme_get_status (void ) +{ +return mktme_status; +} +EXPORT_SYMBOL_GPL (mktme_get_status); +/* * This function is called only when switching between tasks with * different split-lock detection modes. It sets the MSR for the
drivers/misc/Kconfig +15 -0
@@ -456,6 +456,21 @@ config PVPANIC a paravirtualized device provided by QEMU; it lets a virtual machine (guest) communicate panic events to the host. +config MKTME_STATUS + tristate "MKTME status reading support" + depends on X86_64 && SECURITYFS + help + This module provides support for reading the MKTME status. + Enable this configuration option to enable reading the MKTME status + from the mktme virtual file in the securityfs filesystem, + under the mktme directory. + + The MKTME (Multi-Key Total Memory Encryption) status can be + 0 (enabled), 1 (disabled), or 3 (uninitialized). + + To compile this driver as a module, choose M here: the module will + be called mktme_status. + source "drivers/misc/c2port/Kconfig" source "drivers/misc/eeprom/Kconfig" source "drivers/misc/cb710/Kconfig"
drivers/misc/Makefile +1 -0
@@ -57,3 +57,4 @@ obj-$(CONFIG_PVPANIC) += pvpanic.o obj-$(CONFIG_HABANA_AI) += habanalabs/ obj-$(CONFIG_UACCE) += uacce/ obj-$(CONFIG_XILINX_SDFEC) += xilinx_sdfec.o +obj-$(CONFIG_MKTME_STATUS) += mktme_status.o
drivers/misc/mktme_status.c new file +67 -0
67 lines not on this page · raw message
include/uapi/misc/mktme_status.h new file +12 -0
@@ -0,0 +1,12 @@ +/* SPDX-License-Identifier: GPL-2.0 WITH Linux-syscall-note */ + +#ifndef __MKTME_STATUS_H__ +#define __MKTME_STATUS_H__ + +enum mktme_status_type { +MKTME_ENABLED =0 , +MKTME_DISABLED =1 , +MKTME_UNINITIALIZED =2 +}; + +#endif /* __MKTME_STATUS_H__ */
Signature · 1 line
2.25.1
Borislav Petkov
maintainerQuoted 4 lines · Daniel Gutson
The intent of this patch is to provide visibility of the MKTME status to userspace. This is an important factor for firmware security related applilcations.
Quoted 14 lines
Changes since v1:
*Documentation/ABI/stable/securityfs-mktme-status(new file)
*include/uapi/misc/mktme_status.h(new file)
* Fixed MAINTAINER title.
*cpu.h: added values to the enumerands
* Renamed the function fromget_mktme_statustomktme_get_status
* ImprovedKconfighelp
* Removed unnecessary license-related lines since there is a SPDX line
* Movedpr_fmtmacro before the includes
* Turned global variables (mktme_dir,mktme_file) as static
* RemovedBUFFER_SIZEmacro
* No longer returning -1 for error, but using the previously error.
* No morepr_infofor the normal behavior.
* Renamed this from a kernel driver to a kernel module.
No, we don't want a module driver which, as Dave explained, tells you exactly nothing whether encryption is actually enabled on the system. Didn't that become clear from the thread last time?
Signature · 4 lines
Regards/Gruss,
Boris.
https://people.kernel.org/tglx/notes-about-netiquetteA message that is not in the archive (sent off-list or not yet received): CAFmMkTGy5vOiPUpWw6HfQv-JM90JqLBcsKwMpbWdsjaLBw730Q@mail.gmail.com
Borislav Petkov
maintainerQuoted 2 lines · Daniel Gutson
What didn't become clear from the thread last time is the direction to proceed. Concrete suggestion?
Here are two:
https://lkml.kernel.org/r/[email protected]
https://lkml.kernel.org/r/[email protected]
but before that happens, I'd like to hear Dave confirm that when we expose all that information to userspace, it will actually be true and show the necessary bits which *actually* tell you that encryption is enabled.
If you're still unclear, go over the thread again pls.
Thx.
Signature · 4 lines
Regards/Gruss,
Boris.
https://people.kernel.org/tglx/notes-about-netiquetteQuoted 8 lines · Borislav Petkov
Quoted 2 lines · Daniel Gutson
What didn't become clear from the thread last time is the direction to proceed. Concrete suggestion?
Here are two:
https://lkml.kernel.org/r/[email protected]
https://lkml.kernel.org/r/[email protected]
Quoted 6 lines
but before that happens, I'd like to hear Dave confirm that when we expose all that information to userspace, it will actually be true and show the necessary bits which *actually* tell you that encryption is enabled.
If you're still unclear, go over the thread again pls.
It boils down to this: we shouldn't expose low-level, vendor-specific implementation details if we can avoid it. Let's expose something that app can actually use.
Something that will work for all of the TME, MKTME and SEV platforms that I know of and continue to work for a while would be to have a per-numa-node (/sys/devices/system/node[X]/file) that says: "user data on this node is protected by memory encryption".
SEV guests would always have a 1 in all nodes.
TME systems with no platform screwiness like PMEM would always have a 1.
Old systems would have a 0 in there.
TME systems which also have PMEM-only nodes would set 0 in PMEM nodes and 1 on DRAM nodes.
Systems with screwy EFI_MEMORY_CPU_CRYPTO mixing within NUMA nodes would turn it off for the screwy nodes.
Is that concrete enough?
Quoted 32 lines · Dave Hansen
Quoted 8 lines · Borislav Petkov
Quoted 2 lines · Daniel Gutson
What didn't become clear from the thread last time is the direction to proceed. Concrete suggestion?
Here are two:
https://lkml.kernel.org/r/[email protected]
https://lkml.kernel.org/r/[email protected]Quoted 6 lines
but before that happens, I'd like to hear Dave confirm that when we expose all that information to userspace, it will actually be true and show the necessary bits which *actually* tell you that encryption is enabled.
If you're still unclear, go over the thread again pls.
It boils down to this: we shouldn't expose low-level, vendor-specific implementation details if we can avoid it. Let's expose something that app can actually use.
Something that will work for all of the TME, MKTME and SEV platforms that I know of and continue to work for a while would be to have a per-numa-node (/sys/devices/system/node[X]/file) that says: "user data on this node is protected by memory encryption".
SEV guests would always have a 1 in all nodes.
TME systems with no platform screwiness like PMEM would always have a 1.
Old systems would have a 0 in there.
Quoted 7 lines
TME systems which also have PMEM-only nodes would set 0 in PMEM nodes and 1 on DRAM nodes.
Systems with screwy
EFI_MEMORY_CPU_CRYPTOmixing within NUMA nodes would turn it off for the screwy nodes.Is that concrete enough?
What about MKTME platforms that (using hypothetical future kernel support) have encryption enabled for a node but have disabled it for specific pages using madvise()? Or that have any other nontrivial policy like that?
Quoted 4 lines · Andy Lutomirski
What about MKTME platforms that (using hypothetical future kernel support) have encryption enabled for a node but have disabled it for specific pages using
madvise()? Or that have any other nontrivial policy like that?
I think it's fine if the magic new bit means "normal allocations get hardware encryption". If we have a way for users to opt out of that, that's fine with me because the default is to provide it and a user must have gone through _some_ hoop to undo the protection.
BTW, although the MKTME hardware and architecture support disabling encryption, we don't have any plans to expose that to applications.
Quoted 3 lines · Daniel Gutson
The intent of this patch is to provide visibility of the MKTME status to userspace. This is an important factor for firmware security related applilcations.
We need more specifics than this. It's an important factor for what, exactly? Who will consume this and what will they do with it?
I'm also not sure we want to have an Intel product name in the ABI. If we're meaning to tell folks if hardware memory encryption is available on the platform, let's say _that_, rather than talk about MKTME.
Also, MKTME enabling isn't all that interesting. TME is much more interesting and much more opaque.
Quoted 105 lines · Daniel Gutson
The intent of this patch is to provide visibility of the MKTME status to userspace. This is an important factor for firmware security related applilcations.
Changes since v1:
*Documentation/ABI/stable/securityfs-mktme-status(new file)
*include/uapi/misc/mktme_status.h(new file)
* Fixed MAINTAINER title.
*cpu.h: added values to the enumerands
* Renamed the function fromget_mktme_statustomktme_get_status
* ImprovedKconfighelp
* Removed unnecessary license-related lines since there is a SPDX line
* Movedpr_fmtmacro before the includes
* Turned global variables (mktme_dir,mktme_file) as static
* RemovedBUFFER_SIZEmacro
* No longer returning -1 for error, but using the previously error.
* No morepr_infofor the normal behavior.
* Renamed this from a kernel driver to a kernel module.Signed-off-by Daniel Gutson <[email protected]>--- .../ABI/stable/securityfs-mktme-status | 12 ++++ MAINTAINERS | 5 ++ arch/x86/include/asm/cpu.h | 8 +++ arch/x86/kernel/cpu/intel.c | 12 ++-- drivers/misc/Kconfig | 15 +++++ drivers/misc/Makefile | 1 + drivers/misc/mktme_status.c | 67 +++++++++++++++++++ include/uapi/misc/mktme_status.h | 12 ++++ 8 files changed, 127 insertions(+), 5 deletions(-) create mode 100644 Documentation/ABI/stable/securityfs-mktme-status create mode 100644 drivers/misc/mktme_status.c create mode 100644 include/uapi/misc/mktme_status.hDocumentation/ABI/stable/securityfs-mktme-status new file +12
-0@@ -0,0 +1,12 @@ +What: /securityfs/mktme/status +Date: 24-Jun-2020 +KernelVersion: v5.7 +Contact: [email protected] +Description: The mktme securityfs interface exports the BIOS status + of the MKTME. +Values: For possible values see arch/x86/include/asm/cpu.h. + + Currently these values are: + 0 Enabled by BIOS + 1 Disabled by BIOS + 2 UninitializedMAINTAINERS +5
-0@@ -11477,6 +11477,11 @@ W: https://linuxtv.org T: git git://linuxtv.org/media_tree.git F: drivers/media/radio/radio-miropcm20* +MKTME STATUS READING SUPPORT +M: Daniel Gutson <[email protected]> +S: Supported +F: drivers/misc/mktme_status.c + MMP SUPPORT R: Lubomir Rintel <[email protected]> L: [email protected] (moderated for non-subscribers)arch/x86/include/asm/cpu.h +8
-0@@ -26,6 +26,14 @@ struct x86_cpu {struct cpu cpu; }; +enum mktme_status_type { +MKTME_ENABLED =0 , +MKTME_DISABLED =1 , +MKTME_UNINITIALIZED =2 +}; + +extern enum mktme_status_type mktme_get_status (void ); +#ifdef CONFIG_HOTPLUG_CPU extern int arch_register_cpu (int num);extern void arch_unregister_cpu (int );arch/x86/kernel/cpu/intel.c +1
-5@@ -495,11 +495,7 @@ static void srat_detect_node(struct cpuinfo_x86 *c)#define TME_ACTIVATE_CRYPTO_ALGS (x) ((x >> 48) & 0xffff)/* Bits 63:48 */ #define TME_ACTIVATE_CRYPTO_AES_XTS_128 1 -/* Values for mktme_status (SW only construct) */ -#define MKTME_ENABLED 0 -#define MKTME_DISABLED 1 -#define MKTME_UNINITIALIZED 2 -static int mktme_status =MKTME_UNINITIALIZED ; +static enum mktme_status_type mktme_status =MKTME_UNINITIALIZED ;static void detect_tme (struct cpuinfo_x86 *c) {
arch/x86/kernel/cpu/intel.c bool handle_user_split_lock(struct pt_regs *regs, long error_code) · line 1118
Quoted 9 lines
arch/x86/kernel/cpu/intel.c +6
-0@@ -1114,6 +1110,12 @@ bool handle_user_split_lock(struct pt_regs *regs, long error_code)return true; } +enum mktme_status_type mktme_get_status (void ) +{ +return mktme_status; +} +EXPORT_SYMBOL_GPL (mktme_get_status); +
Hi Daniel,
It's not clear this code is getting what it says its getting. (putting aside for the moment the other feedback)
I think you said you want to know if TME is activated in BIOS. I get that. You don't want the system to be up and the customer can't tell if they actually turned on TME or not.
If you only look for MKTME enabled status, you may miss the TME enabled status. We can have TME on and MKTME off. (Can't have opposite)
Here are the boot message I observe based on BIOS selections:
Select TME-off MKTME-off
[] x86/tme: not enabled by BIOS
Select TME-on MKTME-off
[] x86/tme: enabled by BIOS
[] x86/mktme: No known encryption algorithm is supported: 0x0 [] x86/mktme: disabled by BIOS
Select TME-on MKTME-on
[] x86/tme: enabled by BIOS
[] x86/mktme: enabled by BIOS
[] x86/mktme: 63 KeyIDs available
It's that second case above, (TME_on MKTME-off) that this code seems to be miss. TME will be enabled, just not MKTME.
Alison