Re: [PATCH v2 2/4] powerpc: expose secure variables to userspace via sysfs

[Nayna <nayna@linux.vnet.ibm.com>]

[-- Attachment #1: Type: text/plain, Size: 8538 bytes --]



On 08/22/2019 01:18 AM, Oliver O'Halloran wrote:
> On Wed, 2019-08-21 at 11:08 -0400, Nayna Jain wrote:
>> PowerNV secure variables, which store the keys used for OS kernel
>> verification, are managed by the firmware. These secure variables need to
>> be accessed by the userspace for addition/deletion of the certificates.
>>
>> This patch adds the sysfs interface to expose secure variables for PowerNV
>> secureboot. The users shall use this interface for manipulating
>> the keys stored in the secure variables.
>>
>> Signed-off-by: Nayna Jain<nayna@linux.ibm.com>
>> ---
>>   Documentation/ABI/testing/sysfs-secvar |  27 ++++
>>   arch/powerpc/Kconfig                   |   9 ++
>>   arch/powerpc/kernel/Makefile           |   1 +
>>   arch/powerpc/kernel/secvar-sysfs.c     | 210 +++++++++++++++++++++++++
>>   4 files changed, 247 insertions(+)
>>   create mode 100644 Documentation/ABI/testing/sysfs-secvar
>>   create mode 100644 arch/powerpc/kernel/secvar-sysfs.c
>>
>> diff --git a/Documentation/ABI/testing/sysfs-secvar b/Documentation/ABI/testing/sysfs-secvar
>> new file mode 100644
>> index 000000000000..68f0e03d873d
>> --- /dev/null
>> +++ b/Documentation/ABI/testing/sysfs-secvar
>> @@ -0,0 +1,27 @@
>> +What:		/sys/firmware/secvar
>> +Date:		August 2019
>> +Contact:	Nayna Jain<nayna@linux.ibm.com>
>> +Description:
>> +		This directory exposes interfaces for interacting with
>> +		the secure variables managed by OPAL firmware.
>> +
>> +		This is only for the powerpc/powernv platform.
>> +
>> +		Directory:
>> +		vars:		This directory lists all the variables that
>> +				are supported by the OPAL. The variables are
>> +				represented in the form of directories with
>> +				their variable names. The variable name is
>> +				unique and is in ASCII representation. The data
>> +				and size can be determined by reading their
>> +				respective attribute files.
>> +
>> +		Each variable directory has the following files:
>> +		name:		An ASCII representation of the variable name
>> +		data:		A read-only file containing the value of the
>> +				variable
>> +		size:		An integer representation of the size of the
>> +				content of the variable. In other works, it
>> +				represents the size of the data
>> +		update:		A write-only file that is used to submit the new
>> +				value for the variable.
>> diff --git a/arch/powerpc/Kconfig b/arch/powerpc/Kconfig
>> index 42109682b727..b4bdf77837b2 100644
>> --- a/arch/powerpc/Kconfig
>> +++ b/arch/powerpc/Kconfig
>> @@ -925,6 +925,15 @@ config PPC_SECURE_BOOT
>>   	  allows user to enable OS Secure Boot on PowerPC systems that
>>   	  have firmware secure boot support.
>>   
>> +config SECVAR_SYSFS
>> +        tristate "Enable sysfs interface for POWER secure variables"
>> +        depends on PPC_SECURE_BOOT
>> +        help
>> +          POWER secure variables are managed and controlled by firmware.
>> +          These variables are exposed to userspace via sysfs to enable
>> +          read/write operations on these variables. Say Y if you have
>> +	  secure boot enabled and want to expose variables to userspace.
>> +
>>   endmenu
>>   
>>   config ISA_DMA_API
>> diff --git a/arch/powerpc/kernel/Makefile b/arch/powerpc/kernel/Makefile
>> index 9041563f1c74..4ea7b738c3a3 100644
>> --- a/arch/powerpc/kernel/Makefile
>> +++ b/arch/powerpc/kernel/Makefile
>> @@ -158,6 +158,7 @@ obj-$(CONFIG_EPAPR_PARAVIRT)	+= epapr_paravirt.o epapr_hcalls.o
>>   obj-$(CONFIG_KVM_GUEST)		+= kvm.o kvm_emul.o
>>   
>>   obj-$(CONFIG_PPC_SECURE_BOOT)	+= secboot.o ima_arch.o secvar-ops.o
>> +obj-$(CONFIG_SECVAR_SYSFS)     += secvar-sysfs.o
>>   
>>   # Disable GCOV, KCOV & sanitizers in odd or sensitive code
>>   GCOV_PROFILE_prom_init.o := n
>> diff --git a/arch/powerpc/kernel/secvar-sysfs.c b/arch/powerpc/kernel/secvar-sysfs.c
>> new file mode 100644
>> index 000000000000..e46986bb29a0
>> --- /dev/null
>> +++ b/arch/powerpc/kernel/secvar-sysfs.c
>> @@ -0,0 +1,210 @@
>> +// SPDX-License-Identifier: GPL-2.0+
>> +/*
>> + * Copyright (C) 2019 IBM Corporation<nayna@linux.ibm.com>
>> + *
>> + * This code exposes secure variables to user via sysfs
>> + */
>> +
>> +#include <linux/module.h>
>> +#include <linux/slab.h>
>> +#include <linux/compat.h>
>> +#include <linux/string.h>
>> +#include <asm/opal.h>
>> +#include <asm/secvar.h>
>> +
>> +//Approximating it for now, it is bound to change.
>> +#define VARIABLE_MAX_SIZE  32000
> this needs to be communicated from the secvar backend, maybe via a
> field in the ops structure?

Thanks Oliver, I have just posted v3 version which includes yours and 
Greg's feedbacks.
And giving some of the responses here.
Yes for this one,  thinking of doing it via device-tree as they will be 
fixed values for a particular platform

>> +
>> +static struct kobject *powerpc_kobj;
> Call it secvar_kobj or something.
>
>> +static struct secvar_operations *secvarops;
> Ah, the old I-can't-believe-it's-not-global trick.
>
>> +struct kset *secvar_kset;
> shouldn't that be static too?
>
>> +
>> +static ssize_t name_show(struct kobject *kobj, struct kobj_attribute *attr,
>> +			 char *buf)
>> +{
>> +	return sprintf(buf, "%s", kobj->name);
>> +}
>> +
>> +static ssize_t size_show(struct kobject *kobj, struct kobj_attribute *attr,
>> +			 char *buf)
>> +{
>> +	unsigned long dsize;
>> +	int rc;
>> +
>> +	rc = secvarops->get_variable(kobj->name, strlen(kobj->name) + 1, NULL,
>> +				     &dsize);
>> +	if (rc) {
>> +		pr_err("Error retrieving variable size %d\n", rc);
>> +		return rc;
>> +	}
>> +
>> +	rc = sprintf(buf, "%ld", dsize);
>> +
>> +	return rc;
>> +}
>> +
>> +static ssize_t data_read(struct file *filep, struct kobject *kobj,
>> +			 struct bin_attribute *attr, char *buf, loff_t off,
>> +			 size_t count)
>> +{
>> +	unsigned long dsize;
>> +	int rc;
>> +	char *data;
>> +
>> +	rc = secvarops->get_variable(kobj->name, strlen(kobj->name) + 1, NULL,
>> +				     &dsize);
>> +	if (rc) {
>> +		pr_err("Error getting variable size %d\n", rc);
>> +		return rc;
>> +	}
>> +	pr_debug("dsize is %ld\n", dsize);
>> +
>> +	data = kzalloc(dsize, GFP_KERNEL);
>> +	if (!data)
>> +		return -ENOMEM;
>> +
>> +	rc = secvarops->get_variable(kobj->name, strlen(kobj->name)+1, data,
>> +				     &dsize);
>> +	if (rc) {
>> +		pr_err("Error getting variable %d\n", rc);
>> +		goto data_fail;
>> +	}
>> +
>> +	rc = memory_read_from_buffer(buf, count, &off, data, dsize);
>> +
>> +data_fail:
>> +	kfree(data);
>> +	return rc;
>> +}
>> +
>> +static ssize_t update_write(struct file *filep, struct kobject *kobj,
>> +			    struct bin_attribute *attr, char *buf, loff_t off,
>> +			    size_t count)
>> +{
>> +	int rc;
>> +
>> +	pr_debug("count is %ld\n", count);
>> +	rc = secvarops->set_variable(kobj->name, strlen(kobj->name)+1, buf,
>> +				     count);
>> +	if (rc) {
>> +		pr_err("Error setting the variable %s\n", kobj->name);
>> +		return rc;
>> +	}
>> +
>> +	return count;
>> +}
>> +
>> +static struct kobj_attribute name_attr =
>> +__ATTR(name, 0444, name_show, NULL);
>> +
>> +static struct kobj_attribute size_attr =
>> +__ATTR(size, 0444, size_show, NULL);
>> +
>> +static struct bin_attribute data_attr = {
>> +	.attr = {.name = "data", .mode = 0444},
>> +	.size = VARIABLE_MAX_SIZE,
>> +	.read = data_read,
>> +};
> Should they be globally readable? If efivars is globally readable I'm
> happy to follow that example, but mpe might have opinions.

efivars are globally readable.


>> +
>> +
>> +static struct bin_attribute update_attr = {
>> +	.attr = {.name = "update", .mode = 0200},
>> +	.size = VARIABLE_MAX_SIZE,
>> +	.write = update_write,
>> +};
>> +
>> +static struct bin_attribute  *secvar_bin_attrs[] = {
>> +	&data_attr,
>> +	&update_attr,
>> +	NULL,
>> +};
>> +
>> +static struct attribute *secvar_attrs[] = {
>> +	&name_attr.attr,
>> +	&size_attr.attr,
>> +	NULL,
>> +};
>> +
>> +const struct attribute_group secvar_attr_group = {
>> +	.attrs = secvar_attrs,
>> +	.bin_attrs = secvar_bin_attrs,
>> +};
>> +
>> +int secvar_sysfs_load(void)
>> +{
>> +
>> +	char *name;
>> +	unsigned long namesize;
>> +	struct kobject *kobj;
>> +	int status;
>> +	int rc = 0;
>> +
>> +	name = kzalloc(1024, GFP_KERNEL);
>> +	if (!name)
>> +		return -ENOMEM;
> Where'd the 1024 restriction on the length of the variable name come
> from? is that enforced by firmware? If so, how does firmware
> communicate the limited key length?

It is not enforced by the firmware. Currently, it is sort of agreed upon 
value between the firmware and the kernel, and taken from the examples 
of efivars. Probably it can be reduced.


[-- Attachment #2: Type: text/html, Size: 9617 bytes --]