OpenSBI Domain Support
An OpenSBI domain is a system-level partition (subset) of underlying hardware having its own memory regions (RAM and MMIO devices) and HARTs. The OpenSBI will try to achieve secure isolation between domains using RISC-V platform features such as PMP, ePMP, IOPMP, SiFive Shield, etc.
Important entities which help implement OpenSBI domain support are:
- struct sbi_domain_memregion - Representation of a domain memory region
- struct sbi_hartmask - Representation of domain HART set
- struct sbi_domain - Representation of a domain instance
Each HART of a RISC-V platform must have an OpenSBI domain assigned to it. The OpenSBI platform support is responsible for populating domains and providing HART id to domain mapping. The OpenSBI domain support will by default assign the ROOT domain to all HARTs of a RISC-V platform, so it is not mandatory for the OpenSBI platform support to populate domains.
Domain Memory Region
A domain memory region is represented by struct sbi_domain_memregion in OpenSBI and has following details:
- order - The size of a memory region is 2 ^ order where order must be 3 <= order <= __riscv_xlen
- base - The base address of a memory region is 2 ^ order aligned start address
- flags - The flags of a memory region represent memory type (i.e. RAM or MMIO) and allowed accesses (i.e. READ, WRITE, EXECUTE, etc.)
Domain Instance
A domain instance is represented by struct sbi_domain in OpenSBI and has following details:
- index - Logical index of this domain
- name - Name of this domain
- assigned_harts - HARTs assigned to this domain
- possible_harts - HARTs possible in this domain
- regions - Array of memory regions terminated by a memory region with order zero
- boot_hartid - HART id of the HART booting this domain. The domain boot HART will be started at boot-time if boot HART is possible and assigned for this domain.
- next_addr - Address of the next booting stage for this domain
- next_arg1 - Arg1 (or ‘a1’ register) of the next booting stage for this domain
- next_mode - Privilege mode of the next booting stage for this domain. This can be either S-mode or U-mode.
- system_reset_allowed - Is domain allowed to reset the system?
- system_suspend_allowed - Is domain allowed to suspend the system?
The memory regions represented by regions in struct sbi_domain have following additional constraints to align with RISC-V PMP requirements:
- A memory region to protect OpenSBI firmware from S-mode and U-mode should always be present
- For two overlapping memory regions, one should be sub-region of another
- Two overlapping memory regions should not be of same size
- Two overlapping memory regions cannot have same flags
- Memory access checks on overlapping address should prefer smallest overlapping memory region flags.
ROOT Domain
The ROOT domain is the default OpenSBI domain which is assigned by default to all HARTs of a RISC-V platform. The OpenSBI domain support will hand-craft the ROOT domain very early at boot-time in the following manner:
- index - Logical index of the ROOT domain is always zero
- name - Name of the ROOT domain is “root”
- assigned_harts - At boot-time all valid HARTs of a RISC-V platform are assigned the ROOT domain which changes later based on OpenSBI platform support
- possible_harts - All valid HARTs of a RISC-V platform are possible HARTs of the ROOT domain
- regions - Two memory regions available to the ROOT domain: A) A memory region to protect OpenSBI firmware from S-mode and U-mode B) A memory region of order=__riscv_xlen allowing S-mode and U-mode access to full memory address space
- boot_hartid - Coldboot HART is the HART booting the ROOT domain
- next_addr - Next booting stage address in coldboot HART scratch space is the next address for the ROOT domain
- next_arg1 - Next booting stage arg1 in coldboot HART scratch space is the next arg1 for the ROOT domain
- next_mode - Next booting stage mode in coldboot HART scratch space is the next mode for the ROOT domain
- system_reset_allowed - The ROOT domain is allowed to reset the system
- system_suspend_allowed - The ROOT domain is allowed to suspend the system
Domain Effects
Few noteworthy effects of a system partitioned into domains are as follows:
- At any point in time, a HART is running in exactly one OpenSBI domain context
- The SBI IPI and RFENCE calls from HART A are restricted to the HARTs in domain assigned to HART A
- The SBI HSM calls which try to change/read state of HART B from HART A will only work if both HART A and HART B are assigned same domain
- A HART running in S-mode or U-mode can only access memory based on the memory regions of the domain assigned to the HART
Domain Device Tree Bindings
The OpenSBI domains can be described in the device tree (DT) blob (or flattened device tree) passed to the OpenSBI firmwares by the previous booting stage. This allows OpenSBI platform support to parse and populate OpenSBI domains from the device tree blob (or flattened device tree).
Domain Configuration Node
All OpenSBI domain description related DT nodes should be under the domain configuration DT node. The /chosen DT node is the preferred parent of the domain configuration DT node.
The DT properties of a domain configuration DT node are as follows:
- compatible (Mandatory) - The compatible string of the domain configuration. This DT property should have value “opensbi,domain,config”
Domain Memory Region Node
The domain memory region DT node describes details of a memory region and can be pointed by multiple domain instance DT nodes. The access permissions of the memory region are specified separately in domain instance node.
The DT properties of a domain memory region DT node are as follows:
- compatible (Mandatory) - The compatible string of the domain memory region. This DT property should have value “opensbi,domain,memregion”
- base (Mandatory) - The base address of the domain memory region. This DT property should have a 2 ^ order aligned 64 bit address (i.e. two DT cells).
- order (Mandatory) - The order of the domain memory region. This DT property should have a 32 bit value (i.e. one DT cell) in the range 3 <= order <= __riscv_xlen.
- mmio (Optional) - A boolean flag representing whether the domain memory region is a memory-mapped I/O (MMIO) region.
- devices (Optional) - The list of device DT node phandles for devices which fall under this domain memory region.
Domain Instance Node
The domain instance DT node describes set of possible HARTs, set of memory regions, and other details of a domain instance.
The DT properties of a domain instance DT node are as follows:
- compatible (Mandatory) - The compatible string of the domain instance. This DT property should have value “opensbi,domain,instance”
- possible-harts (Optional) - The list of CPU DT node phandles for the the domain instance. This list represents the possible HARTs of the domain instance.
- regions (Optional) - The list of domain memory region DT node phandle and access permissions for the domain instance. Each list entry is a pair of DT node phandle and access permissions. The access permissions are represented as a 32bit bitmask having bits: M readable (BIT[0]), M writeable (BIT[1]), M executable (BIT[2]), SU readable (BIT[3]), SU writable (BIT[4]), and SU executable (BIT[5]). The enforce permission bit (BIT[6]), if set, will lock the permissions in the PMP. This will enforce the permissions on M-mode as well which otherwise will have unrestricted access. This bit must be used with caution because no changes can be made to a PMP entry once its locked until the hart is reset. Any region of a domain defined in DT node cannot have only M-bits set in access permissions i.e. it cannot be an m-mode only accessible region.
- boot-hart (Optional) - The DT node phandle of the HART booting the domain instance. If coldboot HART is assigned to the domain instance then this DT property is ignored and the coldboot HART is assumed to be the boot HART of the domain instance.
- next-arg1 (Optional) - The 64 bit next booting stage arg1 for the domain instance. If this DT property is not available and coldboot HART is not assigned to the domain instance then 0x0 is used as default value. If this DT property is not available and coldboot HART is assigned to the domain instance then next booting stage arg1 of coldboot HART is used as default value.
- next-addr (Optional) - The 64 bit next booting stage address for the domain instance. If this DT property is not available and coldboot HART is not assigned to the domain instance then 0x0 is used as default value. If this DT property is not available and coldboot HART is assigned to the domain instance then next booting stage address of coldboot HART is used as default value.
- next-mode (Optional) - The 32 bit next booting stage mode for the domain instance. The possible values of this DT property are: 0x1 (S-mode), and 0x0 (U-mode). If this DT property is not available and coldboot HART is not assigned to the domain instance then 0x1 is used as default value. If this DT property is not available and coldboot HART is assigned to the domain instance then next booting stage mode of coldboot HART is used as default value.
- system-reset-allowed (Optional) - A boolean flag representing whether the domain instance is allowed to do system reset.
- system-suspend-allowed (Optional) - A boolean flag representing whether the domain instance is allowed to do system suspend.
Assigning HART To Domain Instance
By default, all HARTs are assigned to the ROOT domain. The OpenSBI platform support can provide the HART to domain instance assignment using platform specific callback.
The HART to domain instance assignment can be parsed from the device tree using optional DT property opensbi-domain in each CPU DT node. The value of DT property opensbi-domain is the DT phandle of the domain instance DT node. If opensbi-domain DT property is not specified then corresponding HART is assigned to the ROOT domain.
Domain Configuration Only Accessible to OpenSBI
The software running inside a domain instance should only be aware of devices and hardware resources accessible to itself.
To hide domain configuration from domain instances, the following should be done:
- The previous booting stage should preferably provide a separate device tree for each domain instance and mention location of device tree in respective domain instance DT nodes using next-arg1 DT property.
- If domain assigned to a HART does not have separate device tree then OpenSBI platform support should remove all domain configuration details from the device tree passed by previous booting stage before passing it to the next booting stage.
Example
chosen {
opensbi-domains {
compatible = "opensbi,domain,config";
tmem: tmem {
compatible = "opensbi,domain,memregion";
base = <0x0 0x80100000>;
order = <20>;
};
tuart: tuart {
compatible = "opensbi,domain,memregion";
base = <0x0 0x10011000>;
order = <12>;
mmio;
devices = <&uart1>;
};
allmem: allmem {
compatible = "opensbi,domain,memregion";
base = <0x0 0x0>;
order = <64>;
};
tdomain: trusted-domain {
compatible = "opensbi,domain,instance";
possible-harts = <&cpu0>;
regions = <&tmem 0x3f>, <&tuart 0x3f>;
boot-hart = <&cpu0>;
next-arg1 = <0x0 0x0>;
next-addr = <0x0 0x80100000>;
next-mode = <0x0>;
system-reset-allowed;
system-suspend-allowed;
};
udomain: untrusted-domain {
compatible = "opensbi,domain,instance";
possible-harts = <&cpu1 &cpu2 &cpu3 &cpu4>;
regions = <&tmem 0x0>, <&tuart 0x0>, <&allmem 0x3f>;
};
};
};
cpus {
#address-cells = <1>;
#size-cells = <0>;
timebase-frequency = <10000000>;
cpu0: cpu@0 {
device_type = "cpu";
reg = <0x00>;
compatible = "riscv";
opensbi-domain = <&tdomain>;
...
};
cpu1: cpu@1 {
device_type = "cpu";
reg = <0x01>;
compatible = "riscv";
opensbi-domain = <&udomain>;
...
};
cpu2: cpu@2 {
device_type = "cpu";
reg = <0x02>;
compatible = "riscv";
opensbi-domain = <&udomain>;
...
};
cpu3: cpu@3 {
device_type = "cpu";
reg = <0x03>;
compatible = "riscv";
opensbi-domain = <&udomain>;
...
};
cpu4: cpu@4 {
device_type = "cpu";
reg = <0x04>;
compatible = "riscv";
opensbi-domain = <&udomain>;
...
};
};
uart1: serial@10011000 {
...
};