generic and PowerPC accessor functions for arch keystore
Subsystems (5)
cover generic and PowerPC accessor functions for arch keystore
From: Greg Joyce <[email protected]>
Changelog v3a:
- No code changes, but per reviewer requests, adding additional mailing lists(keyring, EFI) for wider review.
Architectural neutral functions have been defined for accessing architecture specific variable store. The neutral functions are defined as weak so that they may be superseded by platform specific versions. The functions have been desigined so that they can support a large range of platforms/architectures.
PowerPC/pseries versions of these functions provide read/write access to the non-volatile PLPKS data store.
This functionality allows kernel code such as the block SED opal driver to store authentication keys in a secure permanent store.
Greg Joyce (2):
lib: define generic accessor functions for arch specific keystore powerpc/pseries: Override lib/arch_vars.c functions
arch/powerpc/platforms/pseries/Makefile | 1 + .../platforms/pseries/plpks_arch_ops.c | 167 ++++++++++++++++++ include/linux/arch_vars.h | 23 +++ lib/Makefile | 2 +- lib/arch_vars.c | 25 +++ 5 files changed, 217 insertions(+), 1 deletion(-) create mode 100644 arch/powerpc/platforms/pseries/plpks_arch_ops.c create mode 100644 include/linux/arch_vars.h create mode 100644 lib/arch_vars.c
base-commit: ff6992735ade75aae3e35d16b17da1008d753d28
Signature · 1 line
2.27.0
PATCH 1/2 lib: generic accessor functions for arch keystore
From: Greg Joyce <[email protected]>
Generic kernel subsystems may rely on platform specific persistent KeyStore to store objects containing sensitive key material. In such case, they need to access architecture specific functions to perform read/write operations on these variables.
Define the generic variable read/write prototypes to be implemented by architecture specific versions. The default(weak) implementations of these prototypes return -EOPNOTSUPP unless overridden by architecture versions.
--- include/linux/arch_vars.h | 23 +++++++++++++++++++++++ lib/Makefile | 2 +- lib/arch_vars.c | 25 +++++++++++++++++++++++++ 3 files changed, 49 insertions(+), 1 deletion(-) create mode 100644 include/linux/arch_vars.h create mode 100644 lib/arch_vars.c
include/linux/arch_vars.h new file +23 -0
@@ -0,0 +1,23 @@ +/* SPDX-License-Identifier: GPL-2.0 */ +/* +* Platform variable opearations. +* +* Copyright (C) 2022 IBM Corporation +* +* These are the accessor functions (read/write) for architecture specific +* variables. Specific architectures can provide overrides. +* +*/ + +#include <linux/kernel.h> + +enum arch_variable_type { +ARCH_VAR_OPAL_KEY =0 ,/* SED Opal Authentication Key */ +ARCH_VAR_OTHER =1 ,/* Other type of variable */ +ARCH_VAR_MAX =1 ,/* Maximum type value */ +}; + +int arch_read_variable (enum arch_variable_type type,char *varname, +void *varbuf,u_int *varlen); +int arch_write_variable (enum arch_variable_type type,char *varname, +void *varbuf,u_int varlen);
lib/Makefile +1 -1
@@ -48,7 +48,7 @@ obj-y += bcd.o sort.o parser.o debug_locks.o random32.o \ bsearch.o find_bit.o llist.o memweight.o kfifo.o \ percpu-refcount.o rhashtable.o \ once.o refcount.o usercopy.o errseq.o bucket_locks.o \ - generic-radix-tree.o + generic-radix-tree.o arch_vars.o obj-$(CONFIG_STRING_SELFTEST) += test_string.o obj-y += string_helpers.o obj-$(CONFIG_TEST_STRING_HELPERS) += test-string_helpers.o
lib/arch_vars.c new file +25 -0
@@ -0,0 +1,25 @@ +// SPDX-License-Identifier: GPL-2.0-only +/* +* Platform variable operations. +* +* Copyright (C) 2022 IBM Corporation +* +* These are the accessor functions (read/write) for architecture specific +* variables. Specific architectures can provide overrides. +* +*/ + +#include <linux/kernel.h> +#include <linux/arch_vars.h> + +int __weak arch_read_variable (enum arch_variable_type type,char *varname, +void *varbuf,u_int *varlen) +{ +return -EOPNOTSUPP ; +} + +int __weak arch_write_variable (enum arch_variable_type type,char *varname, +void *varbuf,u_int varlen) +{ +return -EOPNOTSUPP ; +}
Signature · 1 line
2.27.0
Re: [PATCH v3a 1/2] lib: generic accessor functions for arch keystore
Le 08/08/2022 à 17:43, [email protected] a écrit :
Quoted 27 lines
From: Greg Joyce <[email protected]>
Generic kernel subsystems may rely on platform specific persistent KeyStore to store objects containing sensitive key material. In such case, they need to access architecture specific functions to perform read/write operations on these variables.
Define the generic variable read/write prototypes to be implemented by architecture specific versions. The default(weak) implementations of these prototypes return
-EOPNOTSUPPunless overridden by architecture versions.Signed-off-by Greg Joyce <[email protected]>--- include/linux/arch_vars.h | 23 +++++++++++++++++++++++ lib/Makefile | 2 +- lib/arch_vars.c | 25 +++++++++++++++++++++++++ 3 files changed, 49 insertions(+), 1 deletion(-)create mode 100644
include/linux/arch_vars.h
create mode 100644lib/arch_vars.cinclude/linux/arch_vars.h new file +0
-0
include/linux/arch_vars.h line 3
Quoted 3 lines
include/linux/arch_vars.h new file +3
-0@@ -0,0 +1,23 @@ +/* SPDX-License-Identifier: GPL-2.0 */ +/* +* Platform variable opearations.
Is it platform specific or architecture specific ?
include/linux/arch_vars.h line 8
Quoted 5 lines
include/linux/arch_vars.h new file +5
-0+* +* Copyright (C) 2022 IBM Corporation +* +* These are the accessor functions (read/write) for architecture specific +* variables. Specific architectures can provide overrides.
"variables" is a very generic word which I think doesn't match what you want to do.
For me "variables" are local variables and global variables in a C file. Here it seems to be something completely different hence the name is really meaningfull and misleading.
include/linux/arch_vars.h line 14
Quoted 6 lines
include/linux/arch_vars.h new file +6
-0+* +*/ + +#include <linux/kernel.h> + +enum arch_variable_type {
arch_variable_type ? What's that ? variable types are char, short, long, long long, etc ...
include/linux/arch_vars.h line 18
Quoted 4 lines
include/linux/arch_vars.h new file +4
-0+ARCH_VAR_OPAL_KEY =0 ,/* SED Opal Authentication Key */ +ARCH_VAR_OTHER =1 ,/* Other type of variable */ + ARCH_VAR_MAX = 1, /* Maximum type value */ +};
Why the hell do you need an enum for two values only ?
lib/arch_vars.c
Quoted 23 lines
include/linux/arch_vars.h new file +5
-0+ +int arch_read_variable (enum arch_variable_type type,char *varname, +void *varbuf,u_int *varlen); +int arch_write_variable (enum arch_variable_type type,char *varname, +void *varbuf,u_int varlen);lib/Makefile +1
-1@@ -48,7 +48,7 @@ obj-y += bcd.o sort.o parser.o debug_locks.o random32.o \ bsearch.o find_bit.o llist.o memweight.o kfifo.o \ percpu-refcount.o rhashtable.o \ once.o refcount.o usercopy.o errseq.o bucket_locks.o \ - generic-radix-tree.o + generic-radix-tree.o arch_vars.o obj-$(CONFIG_STRING_SELFTEST) += test_string.o obj-y += string_helpers.o obj-$(CONFIG_TEST_STRING_HELPERS) += test-string_helpers.olib/arch_vars.c new file +0
-0
The name is meaningless, too generic.
lib/arch_vars.c line 3
Quoted 4 lines
lib/arch_vars.c new file +3
-0@@ -0,0 +1,25 @@ +// SPDX-License-Identifier: GPL-2.0-only +/* +* Platform variable operations.
platform versus architecture ?
lib/arch_vars.c line 16
Quoted 13 lines
lib/arch_vars.c new file +13
-0+* +* Copyright (C) 2022 IBM Corporation +* +* These are the accessor functions (read/write) for architecture specific +* variables. Specific architectures can provide overrides. +* +*/ + +#include <linux/kernel.h> +#include <linux/arch_vars.h> + +int __weak arch_read_variable (enum arch_variable_type type,char *varname, +void *varbuf,u_int *varlen)
Sorry, to read a variable, I use READ_ONCE or I read it directly.
Quoted 9 lines
lib/arch_vars.c new file +9
-0+{ +return -EOPNOTSUPP ; +} + +int __weak arch_write_variable (enum arch_variable_type type,char *varname, +void *varbuf,u_int varlen) +{ +return -EOPNOTSUPP ; +}
Re: [PATCH v3a 1/2] lib: generic accessor functions for arch keystore
Quoted 100 lines · Christophe Leroy
Le 08/08/2022 à 17:43, [email protected] a écrit :
Quoted 27 lines
From: Greg Joyce <[email protected]>
Generic kernel subsystems may rely on platform specific persistent KeyStore to store objects containing sensitive key material. In such case, they need to access architecture specific functions to perform read/write operations on these variables.
Define the generic variable read/write prototypes to be implemented by architecture specific versions. The default(weak) implementations of these prototypes return
-EOPNOTSUPPunless overridden by architecture versions.Signed-off-by Greg Joyce <[email protected]>--- include/linux/arch_vars.h | 23 +++++++++++++++++++++++ lib/Makefile | 2 +- lib/arch_vars.c | 25 +++++++++++++++++++++++++ 3 files changed, 49 insertions(+), 1 deletion(-)create mode 100644
include/linux/arch_vars.h
create mode 100644lib/arch_vars.cinclude/linux/arch_vars.h new file +0
-0include/linux/arch_vars.h line 3
Quoted 3 lines
include/linux/arch_vars.h new file +3
-0@@ -0,0 +1,23 @@ +/* SPDX-License-Identifier: GPL-2.0 */ +/* +* Platform variable opearations. Is it platform specific or architecture specific ?
include/linux/arch_vars.h line 8
Quoted 5 lines
include/linux/arch_vars.h new file +5
-0+* +* Copyright (C) 2022 IBM Corporation +* +* These are the accessor functions (read/write) for architecture specific +* variables. Specific architectures can provide overrides. "variables" is a very generic word which I think doesn't match what you want to do.
For me "variables" are local variables and global variables in a C file. Here it seems to be something completely different hence the name is really meaningfull and misleading.
include/linux/arch_vars.h line 14
Quoted 6 lines
include/linux/arch_vars.h new file +6
-0+* +*/ + +#include <linux/kernel.h> + +enum arch_variable_type {
arch_variable_type? What's that ? variable types are char, short, long, long long, etc ...include/linux/arch_vars.h line 18
Quoted 4 lines
include/linux/arch_vars.h new file +4
-0+ARCH_VAR_OPAL_KEY =0 ,/* SED Opal Authentication Key */ +ARCH_VAR_OTHER =1 ,/* Other type of variable */ + ARCH_VAR_MAX = 1, /* Maximum type value */ +};Why the hell do you need an enum for two values only ?
lib/arch_vars.c
Quoted 23 lines
include/linux/arch_vars.h new file +5
-0+ +int arch_read_variable (enum arch_variable_type type,char *varname, +void *varbuf,u_int *varlen); +int arch_write_variable (enum arch_variable_type type,char *varname, +void *varbuf,u_int varlen);lib/Makefile +1
-1@@ -48,7 +48,7 @@ obj-y += bcd.o sort.o parser.o debug_locks.o random32.o \ bsearch.o find_bit.o llist.o memweight.o kfifo.o \ percpu-refcount.o rhashtable.o \ once.o refcount.o usercopy.o errseq.o bucket_locks.o \ - generic-radix-tree.o + generic-radix-tree.o arch_vars.o obj-$(CONFIG_STRING_SELFTEST) += test_string.o obj-y += string_helpers.o obj-$(CONFIG_TEST_STRING_HELPERS) += test-string_helpers.olib/arch_vars.c new file +0
-0The name is meaningless, too generic.
lib/arch_vars.c line 3
Quoted 4 lines
lib/arch_vars.c new file +3
-0@@ -0,0 +1,25 @@ +// SPDX-License-Identifier: GPL-2.0-only +/* +* Platform variable operations. platform versus architecture ?
Quoted 15 lines
lib/arch_vars.c line 16
Quoted 13 lines
lib/arch_vars.c new file +13
-0+* +* Copyright (C) 2022 IBM Corporation +* +* These are the accessor functions (read/write) for architecture specific +* variables. Specific architectures can provide overrides. +* +*/ + +#include <linux/kernel.h> +#include <linux/arch_vars.h> + +int __weak arch_read_variable (enum arch_variable_type type,char *varname, +void *varbuf,u_int *varlen)Sorry, to read a variable, I use
READ_ONCEor I read it directly.
This is supposed to be used for things like the EFI variables and the already existing powernv secure variables.
Nonetheless, without adding the plumbing for the existing implementations it is not clear what it's doing, and the interface is agruably meaningless.
Hence I would either suggest to provide the plumbing necessary for existing (secure) variable implementations to make use of the interface, or use private implementations like all the existing platforms do without exposing the values in any generic way, and leave that to somebody who is comfortable with designing a working general inteface for this.
Thanks
Michal
PATCH 2/2 powerpc/pseries: Override lib/arch_vars.c functions
In reply to [email protected]From: Greg Joyce <[email protected]>
Self Encrypting Drives(SED) make use of POWER LPAR Platform KeyStore for storing its variables. Thus the block subsystem needs to access PowerPC specific functions to read/write objects in PLPKS.
Override the default implementations in lib/arch_vars.c file with PowerPC specific versions.
--- arch/powerpc/platforms/pseries/Makefile | 1 + .../platforms/pseries/plpks_arch_ops.c | 167 ++++++++++++++++++ 2 files changed, 168 insertions(+) create mode 100644 arch/powerpc/platforms/pseries/plpks_arch_ops.c
arch/powerpc/platforms/pseries/Makefile +1 -0
@@ -29,6 +29,7 @@ obj-$(CONFIG_PPC_SPLPAR) += vphn.o obj-$(CONFIG_PPC_SVM) += svm.o obj-$(CONFIG_FA_DUMP) += rtas-fadump.o obj-$(CONFIG_PSERIES_PLPKS) += plpks.o +obj-$(CONFIG_PSERIES_PLPKS) += plpks_arch_ops.o obj-$(CONFIG_SUSPEND) += suspend.o obj-$(CONFIG_PPC_VAS) += vas.o vas-sysfs.o
arch/powerpc/platforms/pseries/plpks_arch_ops.c new file +167 -0
167 lines not on this page · raw message
Signature · 1 line
2.27.0
Re: [PATCH v3a 2/2] powerpc/pseries: Override lib/arch_vars.c functions
Le 08/08/2022 à 17:43, [email protected] a écrit :
Quoted 87 lines
From: Greg Joyce <[email protected]>
Self Encrypting Drives(SED) make use of POWER LPAR Platform KeyStore for storing its variables. Thus the block subsystem needs to access PowerPC specific functions to read/write objects in PLPKS.
Override the default implementations in
lib/arch_vars.cfile with PowerPC specific versions.Signed-off-by Greg Joyce <[email protected]>--- arch/powerpc/platforms/pseries/Makefile | 1 + .../platforms/pseries/plpks_arch_ops.c | 167 ++++++++++++++++++ 2 files changed, 168 insertions(+)create mode 100644
arch/powerpc/platforms/pseries/plpks_arch_ops.carch/powerpc/platforms/pseries/Makefile +1
-0@@ -29,6 +29,7 @@ obj-$(CONFIG_PPC_SPLPAR) += vphn.o obj-$(CONFIG_PPC_SVM) += svm.o obj-$(CONFIG_FA_DUMP) += rtas-fadump.o obj-$(CONFIG_PSERIES_PLPKS) += plpks.o +obj-$(CONFIG_PSERIES_PLPKS) += plpks_arch_ops.o obj-$(CONFIG_SUSPEND) += suspend.o obj-$(CONFIG_PPC_VAS) += vas.o vas-sysfs.oarch/powerpc/platforms/pseries/plpks_arch_ops.c new file +53
-0@@ -0,0 +1,167 @@ +// SPDX-License-Identifier: GPL-2.0-only +/* +* POWER Platform arch specific code for SED +* Copyright (C) 2022 IBM Corporation +* +* Define operations for generic kernel subsystems to read/write keys +* from POWER LPAR Platform KeyStore(PLPKS). +* +* List of subsystems/usecase using PLPKS: +* - Self Encrypting Drives(SED) +*/ + +#include <linux/kernel.h> +#include <linux/slab.h> +#include <linux/string.h> +#include <linux/ioctl.h> +#include <uapi/linux/sed-opal.h> +#include <linux/sed-opal.h> +#include <linux/arch_vars.h> +#include "plpks.h" + +/* +* variable structure that contains all SED data +*/ +struct plpks_sed_object_data { +u_char version; +u_char pad1[7 ]; +u_long authority; +u_long range; +u_int key_len; +u_char key[32 ]; +}; + +/* +* ext_type values +* 00 no extension exists +* 01-1F common +* 20-3F AIX +* 40-5F Linux +* 60-7F IBMi +*/ + +/* +* This extension is optional for version 1 sed_object_data +*/ +struct sed_object_extension { +u8 ext_type; +u8 rsvd[3 ]; +u8 ext_data[64 ]; +}; + +#define PKS_SED_OBJECT_DATA_V1 1 +#define PKS_SED_MANGLED_LABEL "/default/pri"
arch/powerpc/platforms/pseries/plpks_arch_ops.c line 63
Quoted 10 lines
arch/powerpc/platforms/pseries/plpks_arch_ops.c new file +10
-0+#define PLPKS_SED_COMPONENT "sed-opal" + +#define PLPKS_ARCHVAR_POLICY WORLDREADABLE +#define PLPKS_ARCHVAR_OS_COMMON 4 + +/* +* Read the variable data from PKS given the label +*/ +int arch_read_variable (enum arch_variable_type type,char *varname, +void *varbuf,u_int *varlen)
'enum arch_variable_type' is pointlessly long. And it only has two possible values, so should be a 'bool'
arch/powerpc/platforms/pseries/plpks_arch_ops.c line 67
Quoted 4 lines
arch/powerpc/platforms/pseries/plpks_arch_ops.c new file +4
-0+{ +struct plpks_var var; +struct plpks_sed_object_data *data; +u_int offset =0 ;
Would be better to set it to 0 in the code that handles ARCH_VAR_OTHER and leave it unitialised here.
arch/powerpc/platforms/pseries/plpks_arch_ops.c line 78
Quoted 11 lines
arch/powerpc/platforms/pseries/plpks_arch_ops.c new file +11
-0+char *buf = (char *)varbuf; +int ret; + + var.name = varname; + var.namelen =strlen (varname); + var.policy =PLPKS_ARCHVAR_POLICY ; + var.os =PLPKS_ARCHVAR_OS_COMMON ; + var.data =NULL ; + var.datalen =0 ; + +switch (type) {
A switch for a boolean value is pointless. Just do if/else, it will be a lot more readable.
arch/powerpc/platforms/pseries/plpks_arch_ops.c line 81
Quoted 3 lines
arch/powerpc/platforms/pseries/plpks_arch_ops.c new file +3
-0+case ARCH_VAR_OPAL_KEY : + var.component =PLPKS_SED_COMPONENT ; +#ifdef OPAL_AUTH_KEY
#ifdefs in C files should be avoided, refer
https://docs.kernel.org/process/coding-style.html#conditional-compilation
Quoted 68 lines
arch/powerpc/platforms/pseries/plpks_arch_ops.c new file +68
-0+if (strcmp (OPAL_AUTH_KEY , varname) ==0 ) { + var.name =PKS_SED_MANGLED_LABEL ; + var.namelen =strlen (varname); + } +#endif + offset = offsetof(struct plpks_sed_object_data , key); +break ; +case ARCH_VAR_OTHER : + var.component ="" ; +break ; + } + + ret =plpks_read_os_var (&var); +if (ret !=0 ) +return ret; + +if (offset > var.datalen) + offset =0 ; + +switch (type) { +case ARCH_VAR_OPAL_KEY : + data = (struct plpks_sed_object_data *)var.data; + *varlen = data->key_len; +break ; +case ARCH_VAR_OTHER : + *varlen = var.datalen; +break ; + } + +if (var.data) { +memcpy (varbuf, var.data + offset, var.datalen - offset); + buf[*varlen] ='\0' ; +kfree (var.data); + } + +return 0 ; +} + +/* +* Write the variable data to PKS given the label +*/ +int arch_write_variable (enum arch_variable_type type,char *varname, +void *varbuf,u_int varlen) +{ +struct plpks_var var; +struct plpks_sed_object_data data; +struct plpks_var_name vname; + + var.name = varname; + var.namelen =strlen (varname); + var.policy =PLPKS_ARCHVAR_POLICY ; + var.os =PLPKS_ARCHVAR_OS_COMMON ; + var.datalen = varlen; + var.data = varbuf; + +switch (type) { +case ARCH_VAR_OPAL_KEY : + var.component =PLPKS_SED_COMPONENT ; +#ifdef OPAL_AUTH_KEY +if (strcmp (OPAL_AUTH_KEY , varname) ==0 ) { + var.name =PKS_SED_MANGLED_LABEL ; + var.namelen =strlen (varname); + } +#endif + var.datalen =sizeof (struct plpks_sed_object_data ); + var.data = (u8 *)&data; + +/* initialize SED object */
arch/powerpc/platforms/pseries/plpks_arch_ops.c line 159
Quoted 10 lines
arch/powerpc/platforms/pseries/plpks_arch_ops.c new file +10
-0+ data.version =PKS_SED_OBJECT_DATA_V1 ; + data.authority =0 ; + data.range =0 ; + data.key_len = varlen; +memcpy (data.key, varbuf, varlen); +break ; +case ARCH_VAR_OTHER : + var.component ="" ; +break ; + }
That part of code seem to have a lot in common with arch_read_variable(), can you refactor ?
arch/powerpc/platforms/pseries/plpks_arch_ops.c line 164
Quoted 5 lines
arch/powerpc/platforms/pseries/plpks_arch_ops.c new file +5
-0+ +/* variable update requires delete first */ + vname.namelen = var.namelen; + vname.name = var.name; + (void )plpks_remove_var (var.component, var.os, vname);
Do you really need that cast to (void) ?
Quoted 3 lines
arch/powerpc/platforms/pseries/plpks_arch_ops.c new file +3
-0+ +return plpks_write_var (var); +}