\chapter{Machine-Level ISA, version \privrev} \label{machine} This chapter describes the machine-level operations available in machine-mode (M-mode), which is the highest privilege mode in a RISC-V system. M-mode is the only mandatory privilege mode in a RISC-V hardware implementation. M-mode is used for low-level access to a hardware platform and is the first mode entered at reset. M-mode can also be used to implement features that are too difficult or expensive to implement in hardware directly. The RISC-V machine-level ISA contains a common core that is extended depending on which other privilege levels are supported and other details of the hardware implementation. \section{Machine-Level CSRs} In addition to the machine-level CSRs described in this section, M-mode code can access all CSRs at lower privilege levels. \subsection{Machine ISA Register {\tt misa}} The {\tt misa} CSR is an XLEN-bit \warl\ read-write register reporting the ISA supported by the hart. This register must be readable in any implementation, but a value of zero can be returned to indicate the {\tt misa} register has not been implemented, requiring that CPU capabilities be determined through a separate non-standard mechanism. \begin{figure*}[h!] {\footnotesize \begin{center} \begin{tabular}{c@{}c@{}J} \instbitrange{XLEN-1}{XLEN-2} & \instbitrange{XLEN-3}{26} & \instbitrange{25}{0} \\ \hline \multicolumn{1}{|c|}{MXL[1:0] (\warl)} & \multicolumn{1}{c|}{\wiri} & \multicolumn{1}{c|}{Extensions[25:0] (\warl)} \\ \hline 2 & XLEN-28 & 26 \\ \end{tabular} \end{center} } \vspace{-0.1in} \caption{Machine ISA register ({\tt misa}).} \label{misareg} \end{figure*} The MXL (Machine XLEN) field encodes the native base integer ISA width as shown in Table~\ref{misabase}. The MXL field may be writable in implementations that support multiple base ISA widths. The effective XLEN in M-mode, {\em M-XLEN}, is given by the setting of MXL, or has a fixed value if {\tt misa} is zero. The MXL field is always set to the widest supported ISA variant at reset. \begin{table*}[h!] \begin{center} \begin{tabular}{|r|r|} \hline MXL & XLEN \\ \hline 1 & 32 \\ 2 & 64 \\ 3 & 128 \\ \hline \end{tabular} \end{center} \caption{Encoding of MXL field in {\tt misa}} \label{misabase} \end{table*} \begin{commentary} The base width can be quickly ascertained using branches on the sign of the returned {\tt misa} value, and possibly a shift left by one and a second branch on the sign. These checks can be written in assembly code without knowing the register width (XLEN) of the machine. The base width is given by $\mbox{XLEN}=2^{\mbox{MXL+4}}$. The base width can also be found if {\tt misa} is zero, by placing the immediate 4 in a register then shifting the register left by 31 bits at a time. If zero after one shift, then the machine is RV32. If zero after two shifts, then the machine is RV64, else RV128. \end{commentary} When MXL is set to a value less than the widest supported XLEN, all operations must ignore source operand register bits above the configured XLEN, and must sign-extend results to fill the entire widest supported XLEN in the destination register. \begin{commentary} We require that operations always fill the entire underlying hardware registers with defined values to avoid implementation-defined behavior. \end{commentary} The Extensions field encodes the presence of the standard extensions, with a single bit per letter of the alphabet (bit 0 encodes presence of extension ``A'' , bit 1 encodes presence of extension ``B'', through to bit 25 which encodes ``Z''). The ``I'' bit will be set for RV32I, RV64I, RV128I base ISAs, and the ``E'' bit will be set for RV32E. The Extension is a \warl\ field that can contain writable bits where the implementation allows the supported ISA to be modified. At reset, the Extension field should contain the maximal set of supported extensions, and I should be selected over E if both are available. The ``G'' bit is used as an escape to allow expansion to a larger space of standard extension names. \begin{commentary} G is used to indicate the combination IMAFD, so is redundant in the {\tt misa} CSR, hence we reserve the bit to indicate that additional standard extensions are present. \end{commentary} The ``U'' and ``S'' bits will be set if there is support for user and supervisor modes respectively. The ``X'' bit will be set if there are any non-standard extensions. \begin{table*} \begin{center} \begin{tabular}{|r|r|l|} \hline Bit & Character & Description \\ \hline 0 & A & Atomic extension \\ 1 & B & {\em Tentatively reserved for Bit operations extension} \\ 2 & C & Compressed extension \\ 3 & D & Double-precision floating-point extension \\ 4 & E & RV32E base ISA \\ 5 & F & Single-precision floating-point extension \\ 6 & G & Additional standard extensions present \\ 7 & H & {\em Reserved} \\ 8 & I & RV32I/64I/128I base ISA \\ 9 & J & {\em Tentatively reserved for Dynamically Translated Languages extension} \\ 10 & K & {\em Reserved} \\ 11 & L & {\em Tentatively reserved for Decimal Floating-Point extension} \\ 12 & M & Integer Multiply/Divide extension \\ 13 & N & User-level interrupts supported \\ 14 & O & {\em Reserved} \\ 15 & P & {\em Tentatively reserved for Packed-SIMD extension} \\ 16 & Q & Quad-precision floating-point extension \\ 17 & R & {\em Reserved} \\ 18 & S & Supervisor mode implemented \\ 19 & T & {\em Tentatively reserved for Transactional Memory extension} \\ 20 & U & User mode implemented \\ 21 & V & {\em Tentatively reserved for Vector extension} \\ 22 & W & {\em Reserved} \\ 23 & X & Non-standard extensions present \\ 24 & Y & {\em Reserved} \\ 25 & Z & {\em Reserved} \\ \hline \end{tabular} \end{center} \caption{Encoding of Extensions field in {\tt misa}. All bits that are reserved for future use must return zero when read.} \label{misaletters} \end{table*} \begin{commentary} The {\tt misa} CSR exposes a rudimentary catalog of CPU features to machine-mode code. More extensive information can be obtained in machine mode by probing other machine registers, and examining other ROM storage in the system as part of the boot process. We require that lower privilege levels execute environment calls instead of reading CPU registers to determine features available at each privilege level. This enables virtualization layers to alter the ISA observed at any level, and supports a much richer command interface without burdening hardware designs. \end{commentary} \clearpage \subsection{Machine Vendor ID Register {\tt mvendorid}} The {\tt mvendorid} CSR is an XLEN-bit read-only register providing the JEDEC manufacturer ID of the provider of the core. This register must be readable in any implementation, but a value of 0 can be returned to indicate the field is not implemented or that this is a non-commercial implementation. \begin{figure*}[h!] {\footnotesize \begin{center} \begin{tabular}{JS} \instbitrange{XLEN-1}{7} & \instbitrange{6}{0} \\ \hline \multicolumn{1}{|c|}{Bank} & \multicolumn{1}{c|}{Offset} \\ \hline XLEN-7 & 7 \\ \end{tabular} \end{center} } \vspace{-0.1in} \caption{Vendor ID register ({\tt mvendorid}).} \label{mvendorreg} \end{figure*} JEDEC manufacturer IDs are ordinarily encoded as a sequence of one-byte continuation codes {\tt 0x7f}, terminated by a one-byte ID not equal to {\tt 0x7f}, with an odd parity bit in the most-significant bit of each byte. {\tt mvendorid} encodes the number of one-byte continuation codes in the Bank field, and encodes the final byte in the Offset field, discarding the parity bit. For example, the JEDEC manufacturer ID {\tt 0x7f 0x7f 0x7f 0x7f 0x7f 0x7f 0x7f 0x7f 0x7f 0x7f 0x7f 0x7f 0x8a} (twelve continuation codes followed by {\tt 0x8a}) would be encoded in the {\tt mvendorid} field as {\tt 0x60a}. \begin{commentary} Previously the vendor ID was to be a number allocated by the RISC-V Foundation, but this duplicates the work of JEDEC in maintaining a manufacturer ID standard. At time of writing, registering a manufacturer ID with JEDEC has a one-time cost of \$500. \end{commentary} \subsection{Machine Architecture ID Register {\tt marchid}} The {\tt marchid} CSR is an XLEN-bit read-only register encoding the base microarchitecture of the hart. This register must be readable in any implementation, but a value of 0 can be returned to indicate the field is not implemented. The combination of {\tt mvendorid} and {\tt marchid} should uniquely identify the type of hart microarchitecture that is implemented. \begin{figure*}[h!] {\footnotesize \begin{center} \begin{tabular}{J} \instbitrange{XLEN-1}{0} \\ \hline \multicolumn{1}{|c|}{Architecture ID} \\ \hline XLEN \\ \end{tabular} \end{center} } \vspace{-0.1in} \caption{Machine Architecture ID register ({\tt marchid}).} \label{marchreg} \end{figure*} Open-source project architecture IDs are allocated globally by the RISC-V Foundation, and have non-zero architecture IDs with a zero most-significant-bit (MSB). Commercial architecture IDs are allocated by each commercial vendor independently, but must have the MSB set and cannot contain zero in the remaining XLEN-1 bits. \begin{commentary} The intent is for the architecture ID to represent the microarchitecture associated with the repo around which development occurs rather than a particular organization. Commercial fabrications of open-source designs should (and might be required by the license to) retain the original architecture ID. This will aid in reducing fragmentation and tool support costs, as well as provide attribution. Open-source architecture IDs should be administered by the Foundation and should only be allocated to released, functioning open-source projects. Commercial architecture IDs can be managed independently by any registered vendor but are required to have IDs disjoint from the open-source architecture IDs (MSB set) to prevent collisions if a vendor wishes to use both closed-source and open-source microarchitectures. The convention adopted within the following Implementation field can be used to segregate branches of the same architecture design, including by organization. The {\tt misa} register also helps distinguish different variants of a design, as does the configuration string if present. \end{commentary} \subsection{Machine Implementation ID Register {\tt mimpid}} The {\tt mimpid} CSR provides a unique encoding of the version of the processor implementation. This register must be readable in any implementation, but a value of 0 can be returned to indicate that the field is not implemented. The Implementation value should reflect the design of the RISC-V processor itself and not any surrounding system. \begin{figure*}[h!] {\footnotesize \begin{center} \begin{tabular}{J} \instbitrange{XLEN-1}{0} \\ \hline \multicolumn{1}{|c|}{Implementation} \\ \hline XLEN \\ \end{tabular} \end{center} } \vspace{-0.1in} \caption{Machine Implementation ID register ({\tt mimpid}).} \label{mimpidreg} \end{figure*} \begin{commentary} The format of this field is left to the provider of the architecture source code, but will be often be printed by standard tools as a hexadecimal string without any leading or trailing zeros, so the Implementation value can be left-justified (i.e., filled in from most-significant nibble down) with subfields aligned on nibble boundaries to ease human readability. \end{commentary} \subsection{Hart ID Register {\tt mhartid}} The {\tt mhartid} CSR is an XLEN-bit read-only register containing the integer ID of the hardware thread running the code. This register must be readable in any implementation. Hart IDs might not necessarily be numbered contiguously in a multiprocessor system, but at least one hart must have a hart ID of zero. \begin{figure*}[h!] {\footnotesize \begin{center} \begin{tabular}{J} \instbitrange{XLEN-1}{0} \\ \hline \multicolumn{1}{|c|}{Hart ID}\\ \hline XLEN \\ \end{tabular} \end{center} } \vspace{-0.1in} \caption{Hart ID register ({\tt mhartid}).} \label{mhartidreg} \end{figure*} \begin{commentary} In certain cases, we must ensure exactly one hart runs some code (e.g., at reset), and so require one hart to have a known hart ID of zero. For efficiency, system implementers should aim to reduce the magnitude of the largest hart ID used in a system. \end{commentary} \subsection{Machine Status Register ({\tt mstatus})} The {\tt mstatus} register is an XLEN-bit read/write register formatted as shown in Figure~\ref{mstatusreg-rv32} for RV32 and Figure~\ref{mstatusreg} for RV64 and RV128. The {\tt mstatus} register keeps track of and controls the hart's current operating state. Restricted views of the {\tt mstatus} register appear as the {\tt sstatus} and {\tt ustatus} registers in the S-level and U-level ISAs respectively. \begin{figure*}[h!] {\footnotesize \begin{center} \setlength{\tabcolsep}{4pt} \begin{tabular}{cKccccccc} \\ \instbit{31} & \instbitrange{30}{23} & \instbit{22} & \instbit{21} & \instbit{20} & \instbit{19} & \instbit{18} & \instbit{17} & \\ \hline \multicolumn{1}{|c|}{SD} & \multicolumn{1}{c|}{\wpri} & \multicolumn{1}{c|}{TSR} & \multicolumn{1}{c|}{TW} & \multicolumn{1}{c|}{TVM} & \multicolumn{1}{c|}{MXR} & \multicolumn{1}{c|}{SUM} & \multicolumn{1}{c|}{MPRV} & \\ \hline 1 & 8 & 1 & 1 & 1 & 1 & 1 & 1 & \\ \end{tabular} \begin{tabular}{cccccccccccccc} \\ & \instbitrange{16}{15} & \instbitrange{14}{13} & \instbitrange{12}{11} & \instbitrange{10}{9} & \instbit{8} & \instbit{7} & \instbit{6} & \instbit{5} & \instbit{4} & \instbit{3} & \instbit{2} & \instbit{1} & \instbit{0} \\ \hline & \multicolumn{1}{|c|}{XS[1:0]} & \multicolumn{1}{c|}{FS[1:0]} & \multicolumn{1}{c|}{MPP[1:0]} & \multicolumn{1}{c|}{\wpri} & \multicolumn{1}{c|}{SPP} & \multicolumn{1}{c|}{MPIE} & \multicolumn{1}{c|}{\wpri} & \multicolumn{1}{c|}{SPIE} & \multicolumn{1}{c|}{UPIE} & \multicolumn{1}{c|}{MIE} & \multicolumn{1}{c|}{\wpri} & \multicolumn{1}{c|}{SIE} & \multicolumn{1}{c|}{UIE} \\ \hline & 2 & 2 & 2 & 2 & 1 & 1 & 1 & 1 & 1 & 1 & 1 & 1 & 1 \\ \end{tabular} \end{center} } \vspace{-0.1in} \caption{Machine-mode status register ({\tt mstatus}) for RV32.} \label{mstatusreg-rv32} \end{figure*} \begin{figure*}[h!] {\footnotesize \begin{center} \setlength{\tabcolsep}{4pt} \begin{tabular}{cYccYccccccc} \\ \instbit{XLEN-1} & \instbitrange{XLEN-2}{36} & \instbitrange{35}{34} & \instbitrange{33}{32} & \instbitrange{31}{23} & \instbit{22} & \instbit{21} & \instbit{20} & \instbit{19} & \instbit{18} & \instbit{17} & \\ \hline \multicolumn{1}{|c|}{SD} & \multicolumn{1}{c|}{\wpri} & \multicolumn{1}{c|}{SXL[1:0]} & \multicolumn{1}{c|}{UXL[1:0]} & \multicolumn{1}{c|}{\wpri} & \multicolumn{1}{c|}{TSR} & \multicolumn{1}{c|}{TW} & \multicolumn{1}{c|}{TVM} & \multicolumn{1}{c|}{MXR} & \multicolumn{1}{c|}{SUM} & \multicolumn{1}{c|}{MPRV} & \\ \hline 1 & XLEN-37 & 2 & 2 & 9 & 1 & 1 & 1 & 1 & 1 & 1 & \\ \end{tabular} \begin{tabular}{cccccccccccccc} \\ & \instbitrange{16}{15} & \instbitrange{14}{13} & \instbitrange{12}{11} & \instbitrange{10}{9} & \instbit{8} & \instbit{7} & \instbit{6} & \instbit{5} & \instbit{4} & \instbit{3} & \instbit{2} & \instbit{1} & \instbit{0} \\ \hline & \multicolumn{1}{|c|}{XS[1:0]} & \multicolumn{1}{c|}{FS[1:0]} & \multicolumn{1}{c|}{MPP[1:0]} & \multicolumn{1}{c|}{\wpri} & \multicolumn{1}{c|}{SPP} & \multicolumn{1}{c|}{MPIE} & \multicolumn{1}{c|}{\wpri} & \multicolumn{1}{c|}{SPIE} & \multicolumn{1}{c|}{UPIE} & \multicolumn{1}{c|}{MIE} & \multicolumn{1}{c|}{\wpri} & \multicolumn{1}{c|}{SIE} & \multicolumn{1}{c|}{UIE} \\ \hline & 2 & 2 & 2 & 2 & 1 & 1 & 1 & 1 & 1 & 1 & 1 & 1 & 1 \\ \end{tabular} \end{center} } \vspace{-0.1in} \caption{Machine-mode status register ({\tt mstatus}) for RV64 and RV128.} \label{mstatusreg} \end{figure*} \subsection{Privilege and Global Interrupt-Enable Stack in {\tt mstatus} register} \label{privstack} Interrupt-enable bits, MIE, SIE, and UIE, are provided for each privilege mode. These bits are primarily used to guarantee atomicity with respect to interrupt handlers at the current privilege level. When a hart is executing in privilege mode {\em x}, interrupts are enabled when {\em x}\,IE=1. Interrupts for lower privilege modes are always disabled, whereas interrupts for higher privilege modes are always enabled. Higher-privilege-level code can use separate per-interrupt enable bits to disable selected interrupts before ceding control to a lower privilege level. \begin{commentary} The {\em x}IE bits are located in the low-order bits of {\tt mstatus}, allowing them to be atomically set or cleared with a single CSR instruction. \end{commentary} To support nested traps, each privilege mode {\em x} has a two-level stack of interrupt-enable bits and privilege modes. {\em x}\,PIE holds the value of the interrupt-enable bit active prior to the trap, and {\em x}\,PP holds the previous privilege mode. The {\em x}\,PP fields can only hold privilege modes up to {\em x}, so MPP is two bits wide, SPP is one bit wide, and UPP is implicitly zero. When a trap is taken from privilege mode {\em y} into privilege mode {\em x}, {\em x}\,PIE is set to the value of {\em x}\,IE; {\em x}\,IE is set to 0; and {\em x}\,PP is set to {\em y}. \begin{commentary} For lower privilege modes, any trap (synchronous or asynchronous) is usually taken at a higher privilege mode with interrupts disabled upon entry. The higher-level trap handler will either service the trap and return using the stacked information, or, if not returning immediately to the interrupted context, will save the privilege stack before re-enabling interrupts, so only one entry per stack is required. \end{commentary} The MRET, SRET, or URET instructions are used to return from traps in M-mode, S-mode, or U-mode respectively. When executing an {\em x}RET instruction, supposing {\em x}\,PP holds the value {\em y}, {\em x}\,IE is set to {\em x}\,PIE; the privilege mode is changed to {\em y}; {\em x}\,PIE is set to 1; and {\em x}\,PP is set to U (or M if user-mode is not supported). \begin{commentary} When the stack is popped, the lowest-supported privilege mode with interrupts enabled is added to the bottom of stack to help catch errors that cause invalid entries to be popped off the stack. \end{commentary} {\em x}\,PP fields are \wlrl\ fields that need only be able to store supported privilege modes, including {\em x} and any implemented privilege mode lower than {\em x}. \begin{commentary} If the machine provides only U and M modes, then only a single hardware storage bit is required to represent either 00 or 11 in MPP. \end{commentary} User-level interrupts are an optional extension and have been allocated the ISA extension letter N. If user-level interrupts are omitted, the UIE and UPIE bits are hardwired to zero. For all other supported privilege modes {\em x}, the {\em x}\,IE and {\em x}\,PIE must not be hardwired. \begin{commentary} User-level interrupts are primarily intended to support secure embedded systems with only M-mode and U-mode present, but can also be supported in systems running Unix-like operating systems to support user-level trap handling. \end{commentary} \begin{commentary} Fields that were previously allocated for H-mode support in {\tt mstatus} have now been reserved as \wpri\ fields. To reduce backwards incompatibility with existing implementations, we did not compact the register after removing these fields. \end{commentary} \subsection{Base ISA Control in {\tt mstatus} Register} For RV64 and RV128 systems, the SXL and UXL fields are \warl\ fields that control the value of XLEN for S-mode and U-mode, respectively. The encoding of these fields is the same as the MXL field of {\tt misa}, shown in Table~\ref{misabase}. The effective XLEN in S-mode and U-mode are termed {\em S-XLEN} and {\em U-XLEN}, respectively. For RV32 systems, the SXL and UXL fields do not exist, and S-XLEN~=~32 and U-XLEN~=~32. For RV64 and RV128 systems, if S-mode is not supported, then SXL is hardwired to zero. Otherwise, it is a \warl\ field that encodes the current value of S-XLEN. In particular, the implementation may hardwire SXL so that S-XLEN~=~M-XLEN. For RV64 and RV128 systems, if U-mode is not supported, then UXL is hardwired to zero. Otherwise, it is a \warl\ field that encodes the current value of U-XLEN. In particular, the implementation may hardwire UXL so that U-XLEN~=~M-XLEN. Whenever XLEN in any mode is set to a value less than the widest supported XLEN, all operations must ignore source operand register bits above the configured XLEN, and must sign-extend results to fill the entire widest supported XLEN in the destination register. \begin{commentary} To reduce hardware complexity, the architecture imposes no checks that lower-privilege modes have XLEN settings less than or equal to the next-higher privilege mode. In practice, such settings would almost always be an error, but machine operation is well-defined even in this case. \end{commentary} \subsection{Memory Privilege in {\tt mstatus} Register} The MPRV (Modify PRiVilege) bit modifies the privilege level at which loads and stores execute in all privilege modes. When MPRV=0, translation and protection behave as normal. When MPRV=1, load and store memory addresses are translated and protected as though the current privilege mode were set to MPP. Instruction address-translation and protection are unaffected. MPRV is hardwired to 0 if U-mode is not supported. The MXR (Make eXecutable Readable) bit modifies the privilege with which loads access virtual memory. When MXR=0, only loads from pages marked readable (R=1 in Figure~\ref{sv32pte}) will succeed. When MXR=1, loads from pages marked either readable or executable (R=1 or X=1) will succeed. MXR has no effect when page-based virtual memory is not in effect. MXR is hardwired to 0 if S-mode is not supported. \begin{commentary} The MPRV and MXR mechanisms were conceived to improve the efficiency of M-mode routines that emulate missing hardware features, e.g., misaligned loads and stores. MPRV obviates the need to perform address translation in software. MXR allows instruction words to be loaded from pages marked execute-only. For simplicity, MPRV and MXR are in effect regardless of privilege mode, but in normal use will only be enabled for short sequences in machine mode. \end{commentary} The SUM (permit Supervisor User Memory access) bit modifies the privilege with which S-mode loads, stores, and instruction fetches access virtual memory. When SUM=0, S-mode memory accesses to pages that are accessible by U-mode (U=1 in Figure~\ref{sv32pte}) will fault. When SUM=1, these accesses are permitted. SUM has no effect when page-based virtual memory is not in effect. Note that, while SUM is ordinarily ignored when not executing in S-mode, it {\em is} in effect when MPRV=1 and MPP=S. SUM is hardwired to 0 if S-mode is not supported. \subsection{Virtualization Support in {\tt mstatus} Register} The TVM (Trap Virtual Memory) bit supports intercepting supervisor virtual-memory management operations. When TVM=1, attempts to read or write the {\tt satp} CSR or execute the SFENCE.VMA instruction while executing in S-mode will raise an illegal instruction exception. When TVM=0, these operations are permitted in S-mode. TVM is hard-wired to 0 when S-mode is not supported. \begin{commentary} The TVM mechanism improves virtualization efficiency by permitting guest operating systems to execute in S-mode, rather than classically virtualizing them in U-mode. This approach obviates the need to trap accesses to most S-mode CSRs. Trapping {\tt satp} accesses and the SFENCE.VMA instruction provides the hooks necessary to lazily populate shadow page tables. \end{commentary} The TW (Timeout Wait) bit supports intercepting the WFI instruction (see Section~\ref{wfi}). When TW=0, the WFI instruction is permitted in S-mode. When TW=1, if WFI is executed in S-mode, and it does not complete within an implementation-specific, bounded time limit, the WFI instruction causes an illegal instruction trap. The time limit may always be 0, in which case WFI always causes an illegal instruction trap in S-mode when TW=1. TW is hard-wired to 0 when S-mode is not supported. \begin{commentary} Trapping the WFI instruction can trigger a world switch to another guest OS, rather than wastefully idling in the current guest. \end{commentary} The TSR (Trap SRET) bit supports intercepting the supervisor exception return instruction, SRET. When TSR=1, attempts to execute SRET while executing in S-mode will raise an illegal instruction exception. When TSR=0, this operation is permitted in S-mode. TSR is hard-wired to 0 when S-mode is not supported. \begin{commentary} Trapping SRET is necessary to emulate the Augmented Virtualization mechanism (see Chapter~\ref{hypervisor}) on implementations that do not provide it. \end{commentary} \subsection{Extension Context Status in {\tt mstatus} Register} Supporting substantial extensions is one of the primary goals of RISC-V, and hence we define a standard interface to allow unchanged privileged-mode code, particularly a supervisor-level OS, to support arbitrary user-mode state extensions. \begin{commentary} To date, there are no standard extensions that define additional state beyond the floating-point CSR and data registers. \end{commentary} The FS[1:0] read/write field and the XS[1:0] read-only field are used to reduce the cost of context save and restore by setting and tracking the current state of the floating-point unit and any other user-mode extensions respectively. The FS field encodes the status of the floating-point unit, including the CSR {\tt fcsr} and floating-point data registers {\tt f0}--{\tt f31}, while the XS field encodes the status of additional user-mode extensions and associated state. These fields can be checked by a context switch routine to quickly determine whether a state save or restore is required. If a save or restore is required, additional instructions and CSRs are typically required to effect and optimize the process. \begin{commentary} The design anticipates that most context switches will not need to save/restore state in either or both of the floating-point unit or other extensions, so provides a fast check via the SD bit. \end{commentary} The FS and XS fields use the same status encoding as shown in Table~\ref{fsxsencoding}, with the four possible status values being Off, Initial, Clean, and Dirty. \begin{table*}[h!] \begin{center} \begin{tabular}{|r|l|l|} \hline Status & FS Meaning & XS Meaning\\ \hline 0 & Off & All off \\ 1 & Initial & None dirty or clean, some on\\ 2 & Clean & None dirty, some clean \\ 3 & Dirty & Some dirty \\ \hline \end{tabular} \end{center} \caption{Encoding of FS[1:0] and XS[1:0] status fields.} \label{fsxsencoding} \end{table*} In systems that do not implement S-mode and do not have a floating-point unit, the FS field is hardwired to zero. In systems without additional user extensions requiring new state, the XS field is hardwired to zero. Every additional extension with state adds a local field to {\tt mstatus} encoding the equivalent of the XS states. The XS field represents a summary of all extensions' status as shown in Table~\ref{fsxsencoding}. \begin{commentary} The XS field effectively reports the maximum status value across all user-extension status fields, though individual extensions can use a different encoding than XS. \end{commentary} The SD bit is a read-only bit that summarizes whether either the FS field or XS field signals the presence of some dirty state that will require saving extended user context to memory. If both XS and FS are hardwired to zero, then SD is also always zero. When an extension's status is set to Off, any instruction that attempts to read or write the corresponding state will cause an exception. When the status is Initial, the corresponding state should have an initial constant value. When the status is Clean, the corresponding state is potentially different from the initial value, but matches the last value stored on a context swap. When the status is Dirty, the corresponding state has potentially been modified since the last context save. During a context save, the responsible privileged code need only write out the corresponding state if its status is Dirty, and can then reset the extension's status to Clean. During a context restore, the context need only be loaded from memory if the status is Clean (it should never be Dirty at restore). If the status is Initial, the context must be set to an initial constant value on context restore to avoid a security hole, but this can be done without accessing memory. For example, the floating-point registers can all be initialized to the immediate value 0. The FS and XS fields are read by the privileged code before saving the context. The FS field is set directly by privileged code when resuming a user context, while the XS field is set indirectly by writing to the status register of the individual extensions. The status fields will also be updated during execution of instructions, regardless of privilege mode. Extensions to the user-mode ISA often include additional user-mode state, and this state can be considerably larger than the base integer registers. The extensions might only be used for some applications, or might only be needed for short phases within a single application. To improve performance, the user-mode extension can define additional instructions to allow user-mode software to return the unit to an initial state or even to turn off the unit. For example, a coprocessor might require to be configured before use and can be ``unconfigured'' after use. The unconfigured state would be represented as the Initial state for context save. If the same application remains running between the unconfigure and the next configure (which would set status to Dirty), there is no need to actually reinitialize the state at the unconfigure instruction, as all state is local to the user process, i.e., the Initial state may only cause the coprocessor state to be initialized to a constant value at context restore, not at every unconfigure. Executing a user-mode instruction to disable a unit and place it into the Off state will cause an illegal instruction exception to be raised if any subsequent instruction tries to use the unit before it is turned back on. A user-mode instruction to turn a unit on must also ensure the unit's state is properly initialized, as the unit might have been used by another context meantime. Changing the setting of FS has no effect on the contents of the floating-point register state. In particular, setting FS=Off does not destroy the state, nor does setting FS=Initial clear the contents. Other extensions might not preserve state when set to Off. Table~\ref{fsxsstates} shows all the possible state transitions for the FS or XS status bits. Note that the standard floating-point extensions do not support user-mode unconfigure or disable/enable instructions. \begin{table*}[h!] \begin{center} \begin{tabular}{|l|l|l|l|l|} \hline \multicolumn{1}{|r|}{Current State} & Off & Initial & Clean & Dirty \\ Action & & & &\\ \hline \hline \multicolumn{5}{|c|}{At context save in privileged code}\\ \hline Save state? & No & No & No & Yes \\ Next state & Off & Initial & Clean & Clean \\ \hline \hline \multicolumn{5}{|c|}{At context restore in privileged code}\\ \hline Restore state? & No & Yes, to initial & Yes, from memory & N/A \\ Next state & Off & Initial & Clean & N/A \\ \hline \hline \multicolumn{5}{|c|}{Execute instruction to read state}\\ \hline Action? & Exception & Execute & Execute & Execute \\ Next state & Off & Initial & Clean & Dirty \\ \hline \hline \multicolumn{5}{|c|}{Execute instruction to modify state, including configuration}\\ \hline Action? & Exception & Execute & Execute & Execute \\ Next state & Off & Dirty & Dirty & Dirty \\ \hline \hline \multicolumn{5}{|c|}{Execute instruction to unconfigure unit}\\ \hline Action? & Exception & Execute & Execute & Execute \\ Next state & Off & Initial & Initial & Initial \\ \hline \hline \multicolumn{5}{|c|}{Execute instruction to disable unit}\\ \hline Action? & Execute & Execute & Execute & Execute \\ Next state & Off & Off & Off & Off \\ \hline \hline \multicolumn{5}{|c|}{Execute instruction to enable unit}\\ \hline Action? & Execute & Execute & Execute & Execute \\ Next state & Initial & Initial & Initial & Initial \\ \hline \end{tabular} \end{center} \caption{FS and XS state transitions.} \label{fsxsstates} \end{table*} Standard privileged instructions to initialize, save, and restore extension state are provided to insulate privileged code from details of the added extension state by treating the state as an opaque object. \begin{commentary} Many coprocessor extensions are only used in limited contexts that allows software to safely unconfigure or even disable units when done. This reduces the context-switch overhead of large stateful coprocessors. We separate out floating-point state from other extension state, as when a floating-point unit is present the floating-point registers are part of the standard calling convention, and so user-mode software cannot know when it is safe to disable the floating-point unit. \end{commentary} The XS field provides a summary of all added extension state, but additional microarchitectural bits might be maintained in the extension to further reduce context save and restore overhead. The SD bit is read-only and is set when either the FS or XS bits encode a Dirty state (i.e., SD=((FS==11) OR (XS==11))). This allows privileged code to quickly determine when no additional context save is required beyond the integer register set and PC. The floating-point unit state is always initialized, saved, and restored using standard instructions (F, D, and/or Q), and privileged code must be aware of FLEN to determine the appropriate space to reserve for each {\tt f} register. In a supervisor-level OS, any additional user-mode state should be initialized, saved, and restored using SBI calls that treats the additional context as an opaque object of a fixed maximum size. The implementation of the SBI initialize, save, and restore calls might require additional implementation-dependent privileged instructions to initialize, save, and restore microarchitectural state inside a coprocessor. All privileged modes share a single copy of the FS and XS bits. In a system with more than one privileged mode, supervisor mode would normally use the FS and XS bits directly to record the status with respect to the supervisor-level saved context. Other more-privileged active modes must be more conservative in saving and restoring the extension state in their corresponding version of the context. \begin{commentary} In any reasonable use case, the number of context switches between user and supervisor level should far outweigh the number of context switches to other privilege levels. Note that coprocessors should not require their context to be saved and restored to service asynchronous interrupts, unless the interrupt results in a user-level context swap. \end{commentary} \subsection{Machine Trap-Vector Base-Address Register ({\tt mtvec})} The {\tt mtvec} register is an XLEN-bit read/write register that holds trap vector configuration, consisting of a vector base address (BASE) and a vector mode (MODE). \begin{figure*}[h!] {\footnotesize \begin{center} \begin{tabular}{J@{}S} \instbitrange{XLEN-1}{2} & \instbitrange{1}{0} \\ \hline \multicolumn{1}{|c|}{BASE[XLEN-1:2] (\warl)} & \multicolumn{1}{c|}{MODE (\warl)} \\ \hline XLEN-2 & 2 \\ \end{tabular} \end{center} } \vspace{-0.1in} \caption{Machine trap-vector base-address register ({\tt mtvec}).} \label{mtvecreg} \end{figure*} The {\tt mtvec} register must always be implemented, but can contain a hardwired read-only value. If {\tt mtvec} is writable, the set of values the register may hold can vary by implementation. The value in the BASE field must always be aligned on a 4-byte boundary, and the MODE setting may impose additional alignment constraints on the value in the BASE field. \begin{commentary} We allow for considerable flexibility in implementation of the trap vector base address. On the one hand, we do not wish to burden low-end implementations with a large number of state bits, but on the other hand, we wish to allow flexibility for larger systems. \end{commentary} \begin{table*}[h!] \begin{center} \begin{tabular}{|r|c|l|} \hline Value & Name & Description \\ \hline 0 & Direct & All exceptions set {\tt pc} to BASE. \\ 1 & Vectored & Asynchronous interrupts set {\tt pc} to BASE+4$\times$cause. \\ $\ge$2 & --- & {\em Reserved} \\ \hline \end{tabular} \end{center} \caption{Encoding of {\tt mtvec} MODE field.} \label{mtvec-mode} \end{table*} The encoding of the MODE field is shown in Table~\ref{mtvec-mode}. When MODE=Direct, all traps into machine mode cause the {\tt pc} to be set to the address in the BASE field. When MODE=Vectored, all synchronous exceptions into machine mode cause the {\tt pc} to be set to the address in the BASE field, whereas interrupts cause the {\tt pc} to be set to the address in the BASE field plus four times the interrupt cause number. For example, a machine-mode timer interrupt (see Table~\ref{mcauses}) causes the {\tt pc} to be set to BASE+{\tt 0x1c}. Setting MODE=Vectored may impose an additional alignment constraint on BASE, requiring up to $4\times XLEN$-byte alignment. \begin{commentary} When vectored interrupts are enabled, interrupt cause 0, which corresponds to user-mode software interrupts, are vectored to the same location as synchronous exceptions. This ambiguity does not arise in practice, since user-mode software interrupts are either disabled or delegated to a less-privileged mode. \end{commentary} \begin{commentary} Reset and NMI vector locations are given in a platform specification. \end{commentary} \subsection{Machine Trap Delegation Registers ({\tt medeleg} and {\tt mideleg})} By default, all traps at any privilege level are handled in machine mode, though a machine-mode handler can redirect traps back to the appropriate level with the MRET instruction (Section~\ref{otherpriv}). To increase performance, implementations can provide individual read/write bits within {\tt medeleg} and {\tt mideleg} to indicate that certain exceptions and interrupts should be processed directly by a lower privilege level. The machine exception delegation register ({\tt medeleg}) and machine interrupt delegation register ({\tt mideleg}) are XLEN-bit read/write registers. In systems with all three privilege modes (M/S/U), setting a bit in {\tt medeleg} or {\tt mideleg} will delegate the corresponding trap in S-mode or U-mode to the S-mode trap handler. If U-mode traps are supported, S-mode may in turn set corresponding bits in the {\tt sedeleg} and {\tt sideleg} registers to delegate traps that occur in U-mode to the U-mode trap handler. In systems with two privilege modes (M/U) and support for U-mode traps, setting a bit in {\tt medeleg} or {\tt mideleg} will delegate the corresponding trap in U-mode to the U-mode trap handler. In systems with only M-mode, or with both M-mode and U-mode but without U-mode trap support, the {\tt medeleg} and {\tt mideleg} registers should not exist. \begin{commentary} In versions 1.9.1 and earlier , these registers existed but were hardwired to zero in M-mode only, or M/U without N systems. There is no reason to require they return zero in those cases, as the {\tt misa} register indicates whether they exist. \end{commentary} When a trap is delegated to a less-privileged mode {\em x}, the {\em x}\,{\tt cause} register is written with the trap cause; the {\em x}\,{\tt epc} register is written with the virtual address of the instruction that took the trap; the {\em x}\,PP field of {\tt mstatus} is written with the active privilege mode at the time of the trap; the {\em x}\,PIE field of {\tt mstatus} is written with the value of the active interrupt-enable bit at the time of the trap; and the {\em x}\,IE field of {\tt mstatus} is cleared. The {\tt mcause} and {\tt mepc} registers and the MPP and MPIE fields of {\tt mstatus} are not written. An implementation shall not hardwire any delegation bits to one, i.e., any trap that can be delegated must support not being delegated. An implementation can choose to subset the delegatable traps, with the supported delegatable bits found by writing one to every bit location, then reading back the value in {\tt medeleg} or {\tt mideleg} to see which bit positions hold a one. Traps never transition from a more-privileged mode to a less-privileged mode. For example, if M-mode has delegated illegal instruction traps to S-mode, and M-mode software later executes an illegal instruction, the trap is taken in M-mode, rather than being delegated to S-mode. By contrast, traps may be taken horizontally. Using the same example, if M-mode has delegated illegal instruction traps to S-mode, and S-mode software later executes an illegal instruction, the trap is taken in S-mode. \begin{figure}[h!] {\footnotesize \begin{center} \begin{tabular}{@{}U} \instbitrange{XLEN-1}{0} \\ \hline \multicolumn{1}{|c|}{Synchronous Exceptions (\warl)} \\ \hline XLEN \\ \end{tabular} \end{center} } \vspace{-0.1in} \caption{Machine Exception Delegation Register {\tt medeleg}.} \label{medelegreg} \end{figure} {\tt medeleg} has a bit position allocated for every synchronous exception shown in Table~\ref{mcauses}, with the index of the bit position equal to the value returned in the {\tt mcause} register (i.e., setting bit 8 allows user-mode environment calls to be delegated to a lower-privilege trap handler). \begin{figure}[h!] {\footnotesize \begin{center} \begin{tabular}{@{}U} \instbitrange{XLEN-1}{0} \\ \hline \multicolumn{1}{|c|}{Interrupts (\warl)} \\ \hline XLEN \\ \end{tabular} \end{center} } \vspace{-0.1in} \caption{Machine Exception Delegation Register {\tt mideleg}.} \label{midelegreg} \end{figure} {\tt mideleg} holds trap delegation bits for individual interrupts, with the layout of bits matching those in the {\tt mip} register (i.e., STIP interrupt delegation control is located in bit 5). Some exceptions cannot occur at less privileged modes, and corresponding {\em x}\,{\tt edeleg} bits should be hardwired to zero. In particular, {\tt medeleg}[11] and {\tt sedeleg}[11:9] are all hardwired to zero. \subsection{Machine Interrupt Registers ({\tt mip} and {\tt mie})} The {\tt mip} register is an XLEN-bit read/write register containing information on pending interrupts, while {\tt mie} is the corresponding XLEN-bit read/write register containing interrupt enable bits. Only the bits corresponding to lower-privilege software interrupts (USIP, SSIP), timer interrupts (UTIP, STIP), and external interrupts (UEIP, SEIP) in {\tt mip} are writable through this CSR address; the remaining bits are read-only. Restricted views of the {\tt mip} and {\tt mie} registers appear as the {\tt sip}/{\tt sie}, and {\tt uip}/{\tt uie} registers in S-mode and U-mode respectively. If an interrupt is delegated to privilege mode {\em x} by setting a bit in the {\tt mideleg} register, it becomes visible in the {\em x}\,{\tt ip} register and is maskable using the {\em x}\,{\tt ie} register. Otherwise, the corresponding bits in {\em x}\,{\tt ip} and {\em x}\,{\tt ie} appear to be hardwired to zero. \begin{figure*}[h!] {\footnotesize \begin{center} \setlength{\tabcolsep}{4pt} \begin{tabular}{Ycccccccccccc} \instbitrange{XLEN-1}{12} & \instbit{11} & \instbit{10} & \instbit{9} & \instbit{8} & \instbit{7} & \instbit{6} & \instbit{5} & \instbit{4} & \instbit{3} & \instbit{2} & \instbit{1} & \instbit{0} \\ \hline \multicolumn{1}{|c|}{\wiri} & \multicolumn{1}{c|}{MEIP} & \multicolumn{1}{c|}{\wiri} & \multicolumn{1}{c|}{SEIP} & \multicolumn{1}{c|}{UEIP} & \multicolumn{1}{c|}{MTIP} & \multicolumn{1}{c|}{\wiri} & \multicolumn{1}{c|}{STIP} & \multicolumn{1}{c|}{UTIP} & \multicolumn{1}{c|}{MSIP} & \multicolumn{1}{c|}{\wiri} & \multicolumn{1}{c|}{SSIP} & \multicolumn{1}{c|}{USIP} \\ \hline XLEN-12 & 1 & 1 & 1 & 1 & 1 & 1 & 1 & 1 & 1 & 1 & 1 & 1 \\ \end{tabular} \end{center} } \vspace{-0.1in} \caption{Machine interrupt-pending register ({\tt mip}).} \label{mipreg} \end{figure*} \begin{figure*}[h!] {\footnotesize \begin{center} \setlength{\tabcolsep}{4pt} \begin{tabular}{Ycccccccccccc} \instbitrange{XLEN-1}{12} & \instbit{11} & \instbit{10} & \instbit{9} & \instbit{8} & \instbit{7} & \instbit{6} & \instbit{5} & \instbit{4} & \instbit{3} & \instbit{2} & \instbit{1} & \instbit{0} \\ \hline \multicolumn{1}{|c|}{\wpri} & \multicolumn{1}{c|}{MEIE} & \multicolumn{1}{c|}{\wpri} & \multicolumn{1}{c|}{SEIE} & \multicolumn{1}{c|}{UEIE} & \multicolumn{1}{c|}{MTIE} & \multicolumn{1}{c|}{\wpri} & \multicolumn{1}{c|}{STIE} & \multicolumn{1}{c|}{UTIE} & \multicolumn{1}{c|}{MSIE} & \multicolumn{1}{c|}{\wpri} & \multicolumn{1}{c|}{SSIE} & \multicolumn{1}{c|}{USIE} \\ \hline XLEN-12 & 1 & 1 & 1 & 1 & 1 & 1 & 1 & 1 & 1 & 1 & 1 & 1 \\ \end{tabular} \end{center} } \vspace{-0.1in} \caption{Machine interrupt-enable register ({\tt mie}).} \label{miereg} \end{figure*} The MTIP, STIP, UTIP bits correspond to timer interrupt-pending bits for machine, supervisor, and user timer interrupts, respectively. The MTIP bit is read-only and is cleared by writing to the memory-mapped machine-mode timer compare register. The UTIP and STIP bits may be written by M-mode software to deliver timer interrupts to lower privilege levels. User and supervisor software may clear the UTIP and STIP bits with calls to the AEE and SEE respectively. There is a separate timer interrupt-enable bit, named MTIE, STIE, and UTIE for M-mode, S-mode, and U-mode timer interrupts respectively. Each lower privilege level has a separate software interrupt-pending bit (SSIP, USIP), which can be both read and written by CSR accesses from code running on the local hart at the associated or any higher privilege level. The machine-level MSIP bits are written by accesses to memory-mapped control registers, which are used by remote harts to provide machine-mode interprocessor interrupts. Interprocessor interrupts for lower privilege levels are implemented through ABI and SBI calls to the AEE or SEE respectively, which might ultimately result in a machine-mode write to the receiving hart's MSIP bit. A hart can write its own MSIP bit using the same memory-mapped control register. \begin{commentary} We only allow a hart to directly write its own SSIP or USIP bits when running in the appropriate mode, as other harts might be virtualized and possibly descheduled by higher privilege levels. We rely on ABI and SBI calls to provide interprocessor interrupts for this reason. Machine-mode harts are not virtualized and can directly interrupt other harts by setting their MSIP bits, typically using uncached I/O writes to memory-mapped control registers depending on the platform specification. \end{commentary} The MEIP field in {\tt mip} is a read-only bit that indicates a machine-mode external interrupt is pending. MEIP is set and cleared by a platform-specific interrupt controller, such as the standard platform-level interrupt controller specified in Chapter~\ref{plic}. The MEIE field in {\tt mie} enables machine external interrupts when set. The SEIP field in {\tt mip} contains a single read-write bit. SEIP may be written by M-mode software to indicate to S-mode that an external interrupt is pending. Additionally, the platform-level interrupt controller may generate supervisor-level external interrupts. The logical-OR of the software-writeable bit and the signal from the external interrupt controller is used to generate external interrupts to the supervisor. When the SEIP bit is read with a CSRRW, CSRRS, or CSRRC instruction, the value returned in the {\tt rd} destination register contains the logical-OR of the software-writable bit and the interrupt signal from the interrupt controller. However, the value used in the read-modify-write sequence of a CSRRS or CSRRC instruction is only the software-writable SEIP bit, ignoring the interrupt value from the external interrupt controller. \begin{commentary} The SEIP field behavior is designed to allow a higher privilege layer to mimic external interrupts cleanly, without losing any real external interrupts. The behavior of the CSR instructions is slightly modified from regular CSR accesses as a result. \end{commentary} The UEIP field in {\tt mip} provides user-mode external interrupts when the N extension for user-mode interrupts is implemented. It is defined analogously to SEIP. The MEIE, SEIE, and UEIE fields in the {\tt mie} CSR enable M-mode external interrupts, S-mode external interrupts, and U-mode external interrupts, respectively. \begin{commentary} The non-maskable interrupt is not made visible via the {\tt mip} register as its presence is implicitly known when executing the NMI trap handler. \end{commentary} For all the various interrupt types (software, timer, and external), if a privilege level is not supported, the associated pending and interrupt-enable bits are hardwired to zero in the {\tt mip} and {\tt mie} registers respectively. Hence, these are all effectively \warl\ fields. \begin{commentary} Implementations can add additional platform-specific machine-level interrupt sources to the high bits of these registers, though the expectation is that most external interrupts will be routed through the platform interrupt controller and be delivered via MEIP. \end{commentary} An interrupt {\em i} will be taken if bit {\em i} is set in both {\tt mip} and {\tt mie}, and if interrupts are globally enabled. By default, M-mode interrupts are globally enabled if the hart's current privilege mode is less than M, or if the current privilege mode is M and the MIE bit in the {\tt mstatus} register is set. If bit {\em i} in {\tt mideleg} is set, however, interrupts are considered to be globally enabled if the hart's current privilege mode equals the delegated privilege mode (S or U) and that mode's interrupt enable bit (SIE or UIE in {\tt mstatus}) is set, or if the current privilege mode is less than the delegated privilege mode. Multiple simultaneous interrupts and traps at the same privilege level are handled in the following decreasing priority order: external interrupts, software interrupts, timer interrupts, then finally any synchronous traps. \subsection{Machine Timer Registers ({\tt mtime} and {\tt mtimecmp})} Platforms provide a real-time counter, exposed as a memory-mapped machine-mode register, {\tt mtime}. {\tt mtime} must run at constant frequency, and the platform must provide a mechanism for determining the timebase of {\tt mtime}. The {\tt mtime} register has a 64-bit precision on all RV32, RV64, and RV128 systems. Platforms provide a 64-bit memory-mapped machine-mode timer compare register ({\tt mtimecmp}), which causes a timer interrupt to be posted when the {\tt mtime} register contains a value greater than or equal to the value in the {\tt mtimecmp} register. The interrupt remains posted until it is cleared by writing the {\tt mtimecmp} register. The interrupt will only be taken if interrupts are enabled and the MTIE bit is set in the {\tt mie} register. \begin{figure}[h!] {\footnotesize \begin{center} \begin{tabular}{@{}J} \instbitrange{63}{0} \\ \hline \multicolumn{1}{|c|}{\tt mtime} \\ \hline 64 \\ \end{tabular} \end{center} } \vspace{-0.1in} \caption{Machine time register (memory-mapped control register).} \end{figure} \begin{figure}[h!] {\footnotesize \begin{center} \begin{tabular}{@{}J} \instbitrange{63}{0} \\ \hline \multicolumn{1}{|c|}{\tt mtimecmp} \\ \hline 64 \\ \end{tabular} \end{center} } \vspace{-0.1in} \caption{Machine time compare register (memory-mapped control register).} \end{figure} \begin{commentary} The timer facility is defined to use wall-clock time rather than a cycle counter to support modern processors that run with a highly variable clock frequency to save energy through dynamic voltage and frequency scaling. Accurate real-time clocks (RTCs) are relatively expensive to provide (requiring a crystal or MEMS oscillator) and have to run even when the rest of system is powered down, and so there is usually only one in a system located in a different frequency/voltage domain from the processors. Hence, the RTC must be shared by all the harts in a system and accesses to the RTC will potentially incur the penalty of a voltage-level-shifter and clock-domain crossing. It is thus more natural to expose {\tt mtime} as a memory-mapped register than as a CSR. Lower privilege levels do not have their own {\tt timecmp} registers. Instead, machine-mode software can implement any number of virtual timers on a hart by multiplexing the next timer interrupt into the {\tt mtimecmp} register. Simple fixed-frequency systems can use a single clock for both cycle counting and wall-clock time. \end{commentary} In RV32, memory-mapped writes to {\tt mtimecmp} modify only one 32-bit part of the register. The following code sequence sets a 64-bit {\tt mtimecmp} value without spuriously generating a timer interrupt due to the intermediate value of the comparand: \begin{figure}[h!] \begin{center} \begin{verbatim} # New comparand is in a1:a0. li t0, -1 sw t0, mtimecmp # No smaller than old value. sw a1, mtimecmp+4 # No smaller than new value. sw a0, mtimecmp # New value. \end{verbatim} \end{center} \caption{Sample code for setting the 64-bit time comparand in RV32 assuming the registers live in a strongly ordered I/O region.} \label{mtimecmph} \end{figure} \subsection{Hardware Performance Monitor} M-mode includes a basic hardware performance monitoring facility. The {\tt mcycle} CSR holds a count of the number of cycles the hart has executed since some arbitrary time in the past. The {\tt minstret} CSR holds a count of the number of instructions the hart has retired since some arbitrary time in the past. The {\tt mcycle} and {\tt minstret} registers have 64-bit precision on all RV32, RV64, and RV128 systems. The hardware performance monitor includes 29 additional event counters, {\tt mhpmcounter3}--{\tt mhpmcounter31}. The event selector CSRs, {\tt mhpmevent3}--{\tt mhpmevent31}, are \warl\ registers that control which event causes the corresponding counter to increment. The meaning of these events is defined by the platform, but event 0 is reserved to mean ``no event.'' All counters should be implemented, but a legal implementation is to hard-wire both the counter and its corresponding event selector to 0. \begin{figure}[h!] {\footnotesize \begin{center} \begin{tabular}{@{}K@{}W@{}K} \instbitrange{63}{0} \\ \cline{1-1} \multicolumn{1}{|c|}{\tt mcycle} \\ \cline{1-1} \multicolumn{1}{|c|}{\tt minstret} \\ \cline{1-1} & & \instbitrange{XLEN-1}{0} \\ \cline{1-1}\cline{3-3} \multicolumn{1}{|c|}{\tt mhpmcounter3} & & \multicolumn{1}{|c|}{\tt mhpmevent3} \\ \cline{1-1}\cline{3-3} \multicolumn{1}{|c|}{\tt mhpmcounter4} & & \multicolumn{1}{|c|}{\tt mhpmevent4} \\ \cline{1-1}\cline{3-3} \multicolumn{1}{c}{\vdots} & & \multicolumn{1}{c}{\vdots} \\ \cline{1-1}\cline{3-3} \multicolumn{1}{|c|}{\tt mhpmcounter30} & & \multicolumn{1}{|c|}{\tt mhpmevent30} \\ \cline{1-1}\cline{3-3} \multicolumn{1}{|c|}{\tt mhpmcounter31} & & \multicolumn{1}{|c|}{\tt mhpmevent31} \\ \cline{1-1}\cline{3-3} 64 & & XLEN \\ \end{tabular} \end{center} } \vspace{-0.1in} \caption{Hardware performance monitor counters.} \end{figure} All of these counters have 64-bit precision on RV32, RV64, and RV128. On RV32 only, reads of the {\tt mcycle}, {\tt minstret}, and {\tt mhpmcounter{\em n}} CSRs return the low 32 bits, while reads of the {\tt mcycleh}, {\tt minstreth}, and {\tt mhpmcounter{\em n}h} CSRs return bits 63--32 of the corresponding counter. \begin{figure}[h!] {\footnotesize \begin{center} \begin{tabular}{@{}K} \instbitrange{31}{0} \\ \hline \multicolumn{1}{|c|}{\tt mcycleh} \\ \hline \multicolumn{1}{|c|}{\tt minstreth} \\ \hline \multicolumn{1}{|c|}{\tt mhpmcounter3h} \\ \hline \multicolumn{1}{|c|}{\tt mhpmcounter4h} \\ \hline \multicolumn{1}{c}{\vdots} \\ \hline \multicolumn{1}{|c|}{\tt mhpmcounter30h} \\ \hline \multicolumn{1}{|c|}{\tt mhpmcounter31h} \\ \hline 32 \\ \end{tabular} \end{center} } \vspace{-0.1in} \caption{Upper 32 bits of hardware performance monitor counters, RV32 only.} \end{figure} On RV128 systems, the 64-bit values in {\tt mcycle}, {\tt minstret}, and {\tt mhpmcounter{\em n}} are sign-extended to 128-bits when read. \begin{commentary} On RV128 systems, both signed and unsigned 64-bit values are held in a canonical form with bit 63 repeated in all higher bit positions. The counters are 64-bit values even in RV128, and so the counter CSR reads preserve the sign-extension invariant. Implementations may choose to implement fewer bits of the counters, provided software would be unlikely to experience wraparound (e.g., $2^{63}$ instructions executed) and thereby avoid having to actually implement the sign-extension circuitry. \end{commentary} \subsection{Counter-Enable Registers ({\tt [m|h|s]counteren})} \label{sec:mcounteren} \begin{figure*}[h!] {\footnotesize \begin{center} \setlength{\tabcolsep}{4pt} \begin{tabular}{cccMcccccc} \instbit{31} & \instbit{30} & \instbit{29} & \instbitrange{28}{6} & \instbit{5} & \instbit{4} & \instbit{3} & \instbit{2} & \instbit{1} & \instbit{0} \\ \hline \multicolumn{1}{|c|}{HPM31} & \multicolumn{1}{c|}{HPM30} & \multicolumn{1}{c|}{HPM29} & \multicolumn{1}{c|}{...} & \multicolumn{1}{c|}{HPM5} & \multicolumn{1}{c|}{HPM4} & \multicolumn{1}{c|}{HPM3} & \multicolumn{1}{c|}{IR} & \multicolumn{1}{c|}{TM} & \multicolumn{1}{c|}{CY} \\ \hline 1 & 1 & 1 & 23 & 1 & 1 & 1 & 1 & 1 & 1 \\ \end{tabular} \end{center} } \vspace{-0.1in} \caption{Counter-enable registers ({\tt mcounteren} and {\tt scounteren}).} \label{mcounteren} \end{figure*} The counter-enable registers {\tt mcounteren} and {\tt scounteren} control the availability of the hardware performance monitoring counters to the next-lowest privileged mode. When the CY, TM, IR, or HPM{\em n} bit in the {\tt mcounteren} register is clear, attempts to read the {\tt cycle}, {\tt time}, {\tt instret}, or {\tt hpmcounter{\em n}} register while executing in S-mode or U-mode will cause an illegal instruction exception. When one of these bits is set, access to the corresponding register is permitted in the next implemented privilege mode (S-mode if implemented, otherwise U-mode). If S-mode is implemented, the same bit positions in the {\tt scounteren} register analogously control access to these registers while executing in U-mode. If S-mode is permitted to access a counter register and the corresponding bit is set in {\tt scounteren}, then U-mode is also permitted to access that register. Registers {\tt mcounteren} and {\tt scounteren} are \warl\ registers that must be implemented if U-mode and S-mode are implemented, Any of the bits may contain a hardwired value of zero, indicating reads to the corresponding counter will cause an exception when executing in a less-privileged mode. \begin{commentary} The counter-enable bits support two common use cases with minimal hardware. For systems that do not need high-performance timers and counters, machine-mode software can trap accesses and implement all features in software. For systems that need high-performance timers and counters but are not concerned with obfuscating the underlying hardware counters, the counters can be directly exposed to lower privilege modes. \end{commentary} The {\tt cycle}, {\tt instret}, and {\tt hpmcounter{\em n}} CSRs are read-only shadows of {\tt mcycle}, {\tt minstret}, and {\tt mhpmcounter{\em n}}, respectively. The {\tt time} CSR is a read-only shadow of the memory-mapped {\tt mtime} register. \begin{commentary} Implementations can convert reads of the {\tt time} CSR into loads to the memory-mapped {\tt mtime} register, or hard-wire the TM bits in {\tt m{\em x}counteren} to 0 and emulate this functionality in M-mode software. \end{commentary} \subsection{Machine Scratch Register ({\tt mscratch})} The {\tt mscratch} register is an XLEN-bit read/write register dedicated for use by machine mode. Typically, it is used to hold a pointer to a machine-mode hart-local context space and swapped with a user register upon entry to an M-mode trap handler. \begin{figure}[h!] {\footnotesize \begin{center} \begin{tabular}{@{}J} \instbitrange{XLEN-1}{0} \\ \hline \multicolumn{1}{|c|}{\tt mscratch} \\ \hline XLEN \\ \end{tabular} \end{center} } \vspace{-0.1in} \caption{Machine-mode scratch register.} \label{mscratchreg} \end{figure} \begin{commentary} The MIPS ISA allocated two user registers ({\tt k0}/{\tt k1}) for use by the operating system. Although the MIPS scheme provides a fast and simple implementation, it also reduces available user registers, and does not scale to further privilege levels, or nested traps. It can also require both registers are cleared before returning to user level to avoid a potential security hole and to provide deterministic debugging behavior. The RISC-V user ISA was designed to support many possible privileged system environments and so we did not want to infect the user-level ISA with any OS-dependent features. The RISC-V CSR swap instructions can quickly save/restore values to the {\tt mscratch} register. Unlike the MIPS design, the OS can rely on holding a value in the {\tt mscratch} register while the user context is running. \end{commentary} \subsection{Machine Exception Program Counter ({\tt mepc})} {\tt mepc} is an XLEN-bit read/write register formatted as shown in Figure~\ref{mepcreg}. The low bit of {\tt mepc} ({\tt mepc[0]}) is always zero. On implementations that do not support instruction-set extensions with 16-bit instruction alignment, the two low bits ({\tt mepc[1:0]}) are always zero. \begin{commentary} The {\tt mepc} register can never hold a PC value that would cause an instruction-address-misaligned exception. \end{commentary} {\tt mepc} is a \warl\ register that must be able to hold all valid physical and virtual addresses. It need not be capable of holding all possible invalid addresses. Implementations may convert some invalid address patterns into other invalid addresses prior to writing them to {\tt mepc}. When a trap is taken into M-mode, {\tt mepc} is written with the virtual address of the instruction that encountered the exception. Otherwise, {\tt mepc} is never written by the implementation, though it may be explicitly written by software. \begin{figure}[h!] {\footnotesize \begin{center} \begin{tabular}{@{}J} \instbitrange{XLEN-1}{0} \\ \hline \multicolumn{1}{|c|}{\tt mepc} \\ \hline XLEN \\ \end{tabular} \end{center} } \vspace{-0.1in} \caption{Machine exception program counter register.} \label{mepcreg} \end{figure} \subsection{Machine Cause Register ({\tt mcause})} The {\tt mcause} register is an XLEN-bit read-write register formatted as shown in Figure~\ref{mcausereg}. When a trap is taken into M-mode, {\tt mcause} is written with a code indicating the event that caused the trap. Otherwise, {\tt mcause} is never written by the implementation, though it may be explicitly written by software. The Interrupt bit in the {\tt mcause} register is set if the trap was caused by an interrupt. The Exception Code field contains a code identifying the last exception. Table~\ref{mcauses} lists the possible machine-level exception codes. The Exception Code is an \wlrl\ field, so is only guaranteed to hold supported exception codes. \begin{figure*}[h!] {\footnotesize \begin{center} \begin{tabular}{c@{}U} \instbit{XLEN-1} & \instbitrange{XLEN-2}{0} \\ \hline \multicolumn{1}{|c|}{Interrupt} & \multicolumn{1}{c|}{Exception Code (\wlrl)} \\ \hline 1 & XLEN-1 \\ \end{tabular} \end{center} } \vspace{-0.1in} \caption{Machine Cause register {\tt mcause}.} \label{mcausereg} \end{figure*} \begin{table*}[h!] \begin{center} \begin{tabular}{|r|r|l|l|} \hline Interrupt & Exception Code & Description \\ \hline 1 & 0 & User software interrupt \\ 1 & 1 & Supervisor software interrupt \\ 1 & 2 & {\em Reserved} \\ 1 & 3 & Machine software interrupt \\ \hline 1 & 4 & User timer interrupt \\ 1 & 5 & Supervisor timer interrupt \\ 1 & 6 & {\em Reserved} \\ 1 & 7 & Machine timer interrupt \\ \hline 1 & 8 & User external interrupt \\ 1 & 9 & Supervisor external interrupt \\ 1 & 10 & {\em Reserved} \\ 1 & 11 & Machine external interrupt \\ \hline 1 & $\ge$12 & {\em Reserved} \\ \hline 0 & 0 & Instruction address misaligned \\ 0 & 1 & Instruction access fault \\ 0 & 2 & Illegal instruction \\ 0 & 3 & Breakpoint \\ 0 & 4 & Load address misaligned \\ 0 & 5 & Load access fault \\ 0 & 6 & Store/AMO address misaligned \\ 0 & 7 & Store/AMO access fault \\ 0 & 8 & Environment call from U-mode\\ 0 & 9 & Environment call from S-mode \\ 0 & 10 & {\em Reserved} \\ 0 & 11 & Environment call from M-mode \\ 0 & 12 & Instruction page fault \\ 0 & 13 & Load page fault \\ 0 & 14 & {\em Reserved} \\ 0 & 15 & Store/AMO page fault \\ 0 & $\ge$16 & {\em Reserved} \\ \hline \end{tabular} \end{center} \caption{Machine cause register ({\tt mcause}) values after trap.} \label{mcauses} \end{table*} \begin{commentary} We do not distinguish privileged instruction exceptions from illegal opcode exceptions. This simplifies the architecture and also hides details of which higher-privilege instructions are supported by an implementation. The privilege level servicing the trap can implement a policy on whether these need to be distinguished, and if so, whether a given opcode should be treated as illegal or privileged. \end{commentary} \begin{commentary} Interrupts can be separated from other traps with a single branch on the sign of the {\tt mcause} register value. A shift left can remove the interrupt bit and scale the exception codes to index into a trap vector table. \end{commentary} \subsection{Machine Trap Value ({\tt mtval}) Register} The {\tt mtval} register is an XLEN-bit read-write register formatted as shown in Figure~\ref{mtvalreg}. When a trap is taken into M-mode, {\tt mtval} is written with exception-specific information to assist software in handling the trap. Otherwise, {\tt mtval} is never written by the implementation, though it may be explicitly written by software. When a hardware breakpoint is triggered, or an instruction-fetch, load, or store address-misaligned, access, or page-fault exception occurs, {\tt mtval} is written with the faulting effective address. On an illegal instruction trap, {\tt mtval} is written with the first XLEN bits of the faulting instruction as described below. For other exceptions, {\tt mtval} is set to zero, but a future standard may redefine {\tt mtval}'s setting for other exceptions. \begin{commentary} The {\tt mtval} register replaces the {\tt mbadaddr} register in the previous specification. In addition to providing bad addresses, the register can now provide the bad instruction that triggered an illegal instruction trap (and may in future be used to return other information). Returning the instruction bits accelerates instruction emulation and also removes some races that might be present when trying to emulate illegal instructions. \end{commentary} \begin{commentary} When page-based virtual memory is enabled, {\tt mtval} is written with the faulting virtual address, even for physical-memory access exceptions. This design reduces datapath cost for most implementations, particularly those with hardware page-table walkers. \end{commentary} \begin{figure}[h!] {\footnotesize \begin{center} \begin{tabular}{@{}J} \instbitrange{XLEN-1}{0} \\ \hline \multicolumn{1}{|c|}{\tt mtval} \\ \hline XLEN \\ \end{tabular} \end{center} } \vspace{-0.1in} \caption{Machine Trap Value register.} \label{mtvalreg} \end{figure} For instruction-fetch access faults on RISC-V systems with variable-length instructions, {\tt mtval} will contain a pointer to the portion of the instruction that caused the fault while {\tt mepc} will point to the beginning of the instruction. The {\tt mtval} register can optionally also be used to return the faulting instruction bits on an illegal instruction exception ({\tt mepc} points to the faulting instruction in memory). If this feature is not provided, then {\tt mtval} is set to zero on an illegal instruction fault. If the feature is provided, after an illegal instruction trap, {\tt mtval} will contain the entire faulting instruction provided the instruction is no longer than XLEN bits. If the instruction is less than XLEN bits long, the upper bits of {\tt mtval} are cleared to zero. If the instruction is more than XLEN bits long, {\tt mtval} will contain the first XLEN bits of the instruction. \begin{commentary} Capturing the faulting instruction in {\tt mtval} reduces the overhead of instruction emulation, potentially avoiding several partial instruction loads if the instruction is misaligned, and likely data cache misses or slow uncached accesses when loads are used to fetch the instruction into a data register. There is also a problem of atomicity if another agent is manipulating the instruction memory, as might occur in a dynamic translation system. A requirement is that the entire instruction (or at least the first XLEN bits) are fetched into {\tt mtval} before taking the trap. This should not constrain implementations, which would typically fetch the entire instruction before attempting to decode the instruction, and avoids complicating software handlers. A value of zero in {\tt mtval} signifies either that the feature is not supported, or an illegal zero instruction was fetched. A load from the instruction memory pointed to by {\tt mepc} can be used to distinguish these two cases (or alternatively, the system configuration information can be interrogated to install the appropriate trap handling before runtime). \end{commentary} {\tt mtval} is a \warl\ register that must be able to hold all valid physical and virtual addresses and the value 0. It need not be capable of holding all possible invalid addresses. Implementations may convert some invalid address patterns into other invalid addresses prior to writing them to {\tt mtval}. If the feature to return the faulting instruction bits is implemented, {\tt mtval} must also be able to hold all values less than $2^N$, where $N$ is the smaller of XLEN and the width of the longest supported instruction. \section{Machine-Mode Privileged Instructions} \subsection{Environment Call and Breakpoint} \vspace{-0.2in} \begin{center} \begin{tabular}{M@{}R@{}F@{}R@{}S} \\ \instbitrange{31}{20} & \instbitrange{19}{15} & \instbitrange{14}{12} & \instbitrange{11}{7} & \instbitrange{6}{0} \\ \hline \multicolumn{1}{|c|}{funct12} & \multicolumn{1}{c|}{rs1} & \multicolumn{1}{c|}{funct3} & \multicolumn{1}{c|}{rd} & \multicolumn{1}{c|}{opcode} \\ \hline 12 & 5 & 3 & 5 & 7 \\ ECALL & 0 & PRIV & 0 & SYSTEM \\ EBREAK & 0 & PRIV & 0 & SYSTEM \\ \end{tabular} \end{center} The ECALL instruction is used to make a request to the supporting execution environment. When executed in U-mode, S-mode, or M-mode, it generates an environment-call-from-U-mode exception, environment-call-from-S-mode exception, or environment-call-from-M-mode exception, respectively, and performs no other operation. \begin{commentary} ECALL generates a different exception for each originating privilege mode so that environment call exceptions can be selectively delegated. A typical use case for Unix-like operating systems is to delegate to S-mode the environment-call-from-U-mode exception but not the others. \end{commentary} The EBREAK instruction is used by debuggers to cause control to be transferred back to a debugging environment. It generates a breakpoint exception and performs no other operation. \begin{commentary} As described in the ``C'' Standard Extension for Compressed Instructions in Volume I of this manual, the C.EBREAK instruction performs the same operation as the EBREAK instruction. \end{commentary} ECALL and EBREAK cause the receiving privilege mode's {\tt epc} register to be set to the address of the ECALL or EBREAK instruction itself, {\em not} the address of the following instruction. \subsection{Trap-Return Instructions} \label{otherpriv} Instructions to return from trap are encoded under the PRIV minor opcode. \vspace{-0.2in} \begin{center} \begin{tabular}{M@{}R@{}F@{}R@{}S} \\ \instbitrange{31}{20} & \instbitrange{19}{15} & \instbitrange{14}{12} & \instbitrange{11}{7} & \instbitrange{6}{0} \\ \hline \multicolumn{1}{|c|}{funct12} & \multicolumn{1}{c|}{rs1} & \multicolumn{1}{c|}{funct3} & \multicolumn{1}{c|}{rd} & \multicolumn{1}{c|}{opcode} \\ \hline 12 & 5 & 3 & 5 & 7 \\ MRET/SRET/URET & 0 & PRIV & 0 & SYSTEM \\ \end{tabular} \end{center} To return after handling a trap, there are separate trap return instructions per privilege level: MRET, SRET, and URET. MRET is always provided, while SRET must be provided if supervisor mode is supported. URET is only provided if user-mode traps are supported. An {\em x}\,RET instruction can be executed in privilege mode {\em x} or higher, where executing a lower-privilege {\em x}\,RET instruction will pop the relevant lower-privilege interrupt enable and privilege mode stack. In addition to manipulating the privilege stack as described in Section~\ref{privstack}, {\em x}\,RET sets the {\tt pc} to the value stored in the {\em x}\,{\tt epc} register. \begin{commentary} Previously, there was only a single ERET instruction (which was also earlier known as SRET). To support the addition of user-level interrupts, we needed to add a separate URET instruction to continue to allow classic virtualization of OS code using the ERET instruction. It then became more orthogonal to support a different {\em x}RET instruction per privilege level. \end{commentary} \subsection{Wait for Interrupt} \label{wfi} The Wait for Interrupt instruction (WFI) provides a hint to the implementation that the current hart can be stalled until an interrupt might need servicing. Execution of the WFI instruction can also be used to inform the hardware platform that suitable interrupts should preferentially be routed to this hart. WFI is available in all of the supported S and M privilege modes, and optionally available to U-mode for implementations that support U-mode interrupts. \vspace{-0.2in} \begin{center} \begin{tabular}{M@{}R@{}F@{}R@{}S} \\ \instbitrange{31}{20} & \instbitrange{19}{15} & \instbitrange{14}{12} & \instbitrange{11}{7} & \instbitrange{6}{0} \\ \hline \multicolumn{1}{|c|}{funct12} & \multicolumn{1}{c|}{rs1} & \multicolumn{1}{c|}{funct3} & \multicolumn{1}{c|}{rd} & \multicolumn{1}{c|}{opcode} \\ \hline 12 & 5 & 3 & 5 & 7 \\ WFI & 0 & PRIV & 0 & SYSTEM \\ \end{tabular} \end{center} If an enabled interrupt is present or later becomes present while the hart is stalled, the interrupt exception will be taken on the following instruction, i.e., execution resumes in the trap handler and {\tt mepc} = {\tt pc} + 4. \begin{commentary} The following instruction takes the interrupt exception and trap, so that a simple return from the trap handler will execute code after the WFI instruction. \end{commentary} The WFI instruction is just a hint, and a legal implementation is to implement WFI as a NOP. \begin{commentary} If the implementation does not stall the hart on execution of the instruction, then the interrupt will be taken on some instruction in the idle loop containing the WFI, and on a simple return from the handler, the idle loop will resume execution. \end{commentary} \begin{commentary} We have removed the earlier requirement that implementations ignore the {\em rs1} and {\em rd} fields, so non-zero values in these fields should now raise illegal instruction exceptions. \end{commentary} The WFI instruction can also be executed when interrupts are disabled. The operation of WFI must be unaffected by the global interrupt bits in {\tt mstatus} (MIE/SIE/UIE) and the delegation registers {\tt [m|s|u]ideleg} (i.e., the hart must resume if a locally enabled interrupt becomes pending, even if it has been delegated to a less-privileged mode), but should honor the individual interrupt enables (e.g, MTIE) (i.e., implementations should avoid resuming the hart if the interrupt is pending but not individually enabled). WFI is also required to resume execution for locally enabled interrupts pending at any privilege level, regardless of the global interrupt enable at each privilege level. If the event that causes the hart to resume execution does not cause an interrupt to be taken, execution will resume at {\tt pc} + 4, and software must determine what action to take, including looping back to repeat the WFI if there was no actionable event. \begin{commentary} By allowing wakeup when interrupts are disabled, an alternate entry point to an interrupt handler can be called that does not require saving the current context, as the current context can be saved or discarded before the WFI is executed. As implementations are free to implement WFI as a NOP, software must explicitly check for any relevant pending but disabled interrupts in the code following an WFI, and should loop back to the WFI if no suitable interrupt was detected. The {\tt mip}, {\tt sip}, or {\tt uip} registers can be interrogated to determine the presence of any interrupt in machine, supervisor, or user mode respectively. The operation of WFI is unaffected by the delegation register settings. WFI is defined so that an implementation can trap into a higher privilege mode, either immediately on encountering the WFI or after some interval to initiate a machine-mode transition to a lower power state, for example. \end{commentary} \begin{commentary} The same ``wait-for-event'' template might be used for possible future extensions that wait on memory locations changing, or message arrival. \end{commentary} \section{Reset} \label{sec:reset} Upon reset, a hart's privilege mode is set to M. The {\tt mstatus} fields MIE and MPRV are reset to 0. The {\tt pc} is set to an implementation-defined reset vector. The {\tt mcause} register is set to a value indicating the cause of the reset. All other hart state is undefined. The {\tt mcause} values after reset have implementation-specific interpretation, but the value 0 should be returned on implementations that do not distinguish different reset conditions. Implementations that distinguish different reset conditions should only use 0 to indicate the most complete reset (e.g., hard reset). \begin{commentary} Some designs may have multiple causes of reset (e.g., power-on reset, external hard reset, brownout detected, watchdog timer elapse, sleep-mode wakeup), which machine-mode software and debuggers may wish to distinguish. {\tt mcause} reset values may alias {\tt mcause} values following synchronous exceptions. There is no ambiguity in this overlap, since on reset the {\tt pc} is set to a different value than on other traps. \end{commentary} \section{Non-Maskable Interrupts} \label{sec:nmi} Non-maskable interrupts (NMIs) are only used for hardware error conditions, and cause an immediate jump to an implementation-defined NMI vector running in M-mode regardless of the state of a hart's interrupt enable bits. The {\tt mepc} register is written with the address of the next instruction to be executed at the time the NMI was taken, and {\tt mcause} is set to a value indicating the source of the NMI. The NMI can thus overwrite state in an active machine-mode interrupt handler. The values written to {\tt mcause} on an NMI are implementation-defined, but a value of 0 is reserved to mean ``unknown cause'' and implementations that do not distinguish sources of NMIs via the {\tt mcause} register should return 0. Unlike resets, NMIs do not reset processor state, enabling diagnosis, reporting, and possible containment of the hardware error. \section{Physical Memory Attributes} \label{sec:pma} The physical memory map for a complete system includes various address ranges, some corresponding to memory regions, some to memory-mapped control registers, and some to empty holes in the address space. Some memory regions might not support reads, writes, or execution; some might not support subword or subblock accesses; some might not support atomic operations; and some might not support cache coherence or might have different memory models. Similarly, memory-mapped control registers vary in their supported access widths, support for atomic operations, and whether read and write accesses have associated side effects. In RISC-V systems, these properties and capabilities of each region of the machine's physical address space are termed {\em physical memory attributes} (PMAs). This section describes RISC-V PMA terminology and how RISC-V systems implement and check PMAs. PMAs are inherent properties of the underlying hardware and rarely change during system operation. Unlike physical memory protection values described in Section~\ref{sec:pmp}, PMAs do not vary by execution context. The PMAs of some memory regions are fixed at chip design time---for example, for an on-chip ROM. Others are fixed at board design time, depending, for example, on which other chips are connected to off-chip buses. Off-chip buses might also support devices that could be changed on every power cycle (cold pluggable) or dynamically while the system is running (hot pluggable). Some devices might be configurable at run time to support different uses that imply different PMAs---for example, an on-chip scratchpad RAM might be cached privately by one core in one end-application, or accessed as a shared non-cached memory in another end-application. Most systems will require that at least some PMAs are dynamically checked in hardware later in the execution pipeline after the physical address is known, as some operations will not be supported at all physical memory addresses, and some operations require knowing the current setting of a configurable PMA attribute. While many other systems specify some PMAs in the virtual memory page tables and use the TLB to inform the pipeline of these properties, this approach injects platform-specific information into a virtualized layer and can cause system errors unless attributes are correctly initialized in each page-table entry for each physical memory region. In addition, the available page sizes might not be optimal for specifying attributes in the physical memory space, leading to address-space fragmentation and inefficient use of expensive TLB entries. For RISC-V, we separate out specification and checking of PMAs into a separate hardware structure, the {\em PMA checker}. In many cases, the attributes are known at system design time for each physical address region, and can be hardwired into the PMA checker. Where the attributes are run-time configurable, platform-specific memory-mapped control registers can be provided to specify these attributes at a granularity appropriate to each region on the platform (e.g., for an on-chip SRAM that can be flexibly divided between cacheable and uncacheable uses). PMAs are checked for any access to physical memory, including accesses that have undergone virtual to physical memory translation. To aid in system debugging, we strongly recommend that, where possible, RISC-V processors precisely trap physical memory accesses that fail PMA checks. Precisely trapped PMA violations manifest as load, store, or instruction-fetch access exceptions, distinct from virtual-memory page-fault exceptions. Precise PMA traps might not always be possible, for example, when probing a legacy bus architecture that uses access failures as part of the discovery mechanism. In this case, error responses from slave devices will be reported as imprecise bus-error interrupts. PMAs must also be readable by software to correctly access certain devices or to correctly configure other hardware components that access memory, such as DMA engines. As PMAs are tightly tied to a given physical platform's organization, many details are inherently platform-specific, as is the means by which software can learn the PMA values for a platform. The configuration string (Chapter~\ref{cfgstr}) can encode PMAs for on-chip devices and might also describe on-chip controllers for off-chip buses that can be dynamically interrogated to discover attached device PMAs. Some devices, particularly legacy buses, do not support discovery of PMAs and so will give error responses or time out if an unsupported access is attempted. Typically, platform-specific machine-mode code will extract PMAs and ultimately present this information to higher-level less-privileged software using some standard representation. Where platforms support dynamic reconfiguration of PMAs, an interface will be provided to set the attributes by passing requests to a machine-mode driver that can correctly reconfigure the platform. For example, switching cacheability attributes on some memory regions might involve platform-specific operations, such as cache flushes, that are available only to machine-mode. \subsection{Main Memory versus I/O versus Empty Regions} The most important characterization of a given memory address range is whether it holds regular main memory, or I/O devices, or is empty. Regular main memory is required to have a number of properties, specified below, whereas I/O devices can have a much broader range of attributes. Memory regions that do not fit into regular main memory, for example, device scratchpad RAMs, are categorized as I/O regions. Empty regions are also classified as I/O regions but with attributes specifying that no accesses are supported. \subsection{Supported Access Type PMAs} Access types specify which access widths, from 8-bit byte to long multi-word burst, are supported, and also whether misaligned accesses are supported for each access width. \begin{commentary} Although software running on a RISC-V hart cannot directly generate bursts to memory, software might have to program DMA engines to access I/O devices and might therefore need to know which access sizes are supported. \end{commentary} Main memory regions always support read, write, and execute of all access widths required by the attached devices. \begin{commentary} In some cases, the design of a processor or device accessing main memory might support other widths, but must be able to function with the types supported by the main memory. \end{commentary} I/O regions can specify which combinations of read, write, or execute accesses to which data widths are supported. \subsection{Atomicity PMAs} Atomicity PMAs describes which atomic instructions are supported in this address region. Main memory regions must support the atomic operations required by the processors attached. I/O regions may only support a subset or none of the processor-supported atomic operations. Support for atomic instructions is divided into two categories: {\em LR/SC} and {\em AMOs}. Within AMOs, there are four levels of support: {\em AMONone}, {\em AMOSwap}, {\em AMOLogical}, and {\em AMOArithmetic}. AMONone indicates that no AMO operations are supported. AMOSwap indicates that only {\tt amoswap} instructions are supported in this address range. AMOLogical indicates that swap instructions plus all the logical AMOs ({\tt amoand}, {\tt amoor}, {\tt amoxor}) are supported. AMOArithmetic indicates that all RISC-V AMOs are supported. For each level of support, naturally aligned AMOs of a given width are supported if the underlying memory region supports reads and writes of that width. \begin{table*}[h!] \begin{center} \begin{tabular}{|l|l|} \hline AMO Class & Supported Operations \\ \hline AMONone & {\em None} \\ AMOSwap & {\tt amoswap} \\ AMOLogical & above + {\tt amoand}, {\tt amoor}, {\tt amoxor} \\ AMOArithmetic & above + {\tt amoadd}, {\tt amomin}, {\tt amomax}, {\tt amominu}, {\tt amomaxu} \\ \hline \end{tabular} \end{center} \caption{Classes of AMOs supported by I/O regions. Main memory regions must always support all AMOs required by the processor.} \label{amoclasses} \end{table*} \begin{commentary} We recommend providing at least AMOLogical support for I/O regions where possible. Most I/O regions will not support LR/SC accesses, as these are most conveniently built on top of a cache-coherence scheme. \end{commentary} \subsection{Memory-Ordering PMAs} Regions of the address space are classified as either {\em main memory} or {\em I/O} for the purposes of ordering by the FENCE instruction and atomic-instruction ordering bits. Accesses by one hart to main memory regions are observable not only by other harts but also by other devices with the capability to initiate requests in the main memory system (e.g., DMA engines). Main memory regions always have the standard RISC-V relaxed memory model. Accesses by one hart to the I/O space are observable not only by other harts and bus mastering devices, but also by targeted slave I/O devices. Within I/O, regions may further be classified as implementing either {\em relaxed} or {\em strong} ordering. A relaxed I/O region has no ordering guarantees on how memory accesses made by one hart are observable by different harts or I/O devices beyond those enforced by FENCE and AMO instructions. A strongly ordered I/O region ensures that all accesses made by a hart to that region are only observable in program order by all other harts or I/O devices. Each strongly ordered I/O region specifies a numbered ordering channel, which is a mechanism by which ordering guarantees can be provided between different I/O regions. Channel 0 is used to indicate point-to-point strong ordering only, where only accesses by the hart to the single associated I/O region are strongly ordered. Channel 1 is used to provide global strong ordering across all I/O regions. Any accesses by a hart to any I/O region associated with channel 1 can only be observed to have occurred in program order by all other harts and I/O devices, including relative to accesses made by that hart to relaxed I/O regions or strongly ordered I/O regions with different channel numbers. In other words, any access to a region in channel 1 is equivalent to executing a {\tt fence io,io} instruction before and after the instruction. Other larger channel numbers provide program ordering to accesses by that hart across any regions with the same channel number. Systems might support dynamic configuration of ordering properties on each memory region. \begin{commentary} Strong ordering can be used to improve compatibility with legacy device driver code, or to enable increased performance compared to insertion of explicit ordering instructions when the implementation is known to not reorder accesses. Local strong ordering (channel 0) is the default form of strong ordering as it is often straightforward to provide if there is only a single in-order communication path between the hart and the I/O device. Generally, different strongly ordered I/O regions can share the same ordering channel without additional ordering hardware if they share the same interconnect path and the path does not reorder requests. \end{commentary} \subsection{Coherence and Cacheability PMAs} Coherence is a property defined for a single physical address, and indicates that writes to that address by one agent will eventually be made visible to other agents in the system. Coherence is not to be confused with the memory consistency model of a system, which defines what values a memory read can return given the previous history of reads and writes to the entire memory system. In RISC-V platforms, the use of hardware-incoherent regions is discouraged due to software complexity, performance, and energy impacts. The cacheability of a memory region should not affect the software view of the region except for differences reflected in other PMAs, such as main memory versus I/O classification, memory ordering, supported accesses and atomic operations, and coherence. For this reason, we treat cacheability as a platform-level setting managed by machine-mode software only. Where a platform supports configurable cacheability settings for a memory region, a platform-specific machine-mode routine will change the settings and flush caches if necessary, so the system is only incoherent during the transition between cacheability settings. This transitory state should not be visible to lower privilege levels. \begin{commentary} We categorize RISC-V caches into three types: {\em master-private}, {\em shared}, and {\em slave-private}. Master-private caches are attached to a single master agent, i.e., one that issues read/write requests to the memory system. Shared caches are located inbetween masters and slaves and may be hierarchically organized. Slave-private caches do not impact coherence, as they are local to a single slave and do not affect other PMAs at a master, so are not considered further here. We use {\em private cache} to mean a master-private cache in the following section, unless explicitly stated otherwise. Coherence is straightforward to provide for a shared memory region that is not cached by any agent. The PMA for such a region would simply indicate it should not be cached in a private or shared cache. Coherence is also straightforward for read-only regions, which can be safely cached by multiple agents without requiring a cache-coherence scheme. The PMA for this region would indicate that it can be cached, but that writes are not supported. Some read-write regions might only be accessed by a single agent, in which case they can be cached privately by that agent without requiring a coherence scheme. The PMA for such regions would indicate they can be cached. The data can also be cached in a shared cache, as other agents should not access the region. If an agent can cache a read-write region that is accessible by other agents, whether caching or non-caching, a cache-coherence scheme is required to avoid use of stale values. In regions lacking hardware cache coherence (hardware-incoherent regions), cache coherence can be implemented entirely in software, but software coherence schemes are notoriously difficult to implement correctly and often have severe performance impacts due to the need for conservative software-directed cache-flushing. Hardware cache-coherence schemes require more complex hardware and can impact performance due to the cache-coherence probes, but are otherwise invisible to software. For each hardware cache-coherent region, the PMA would indicate that the region is coherent and which hardware coherence controller to use if the system has multiple coherence controllers. For some systems, the coherence controller might be an outer-level shared cache, which might itself access further outer-level cache-coherence controllers hierarchically. Most memory regions within a platform will be coherent to software, because they will be fixed as either uncached, read-only, hardware cache-coherent, or only accessed by one agent. \end{commentary} \subsection{Idempotency PMAs} Idempotency PMAs describe whether reads and writes to an address region are idempotent. Main memory regions are assumed to be idempotent. For I/O regions, idempotency on reads and writes can be specified separately (e.g., reads are idempotent but writes are not). If accesses are non-idempotent, i.e., there is potentially a side effect on any read or write access, then speculative or redundant accesses must be avoided. For the purposes of defining the idempotency PMAs, changes in observed memory ordering created by redundant accesses are not considered a side effect. \begin{commentary} While hardware should always be designed to avoid speculative or redundant accesses to memory regions marked as non-idempotent, it is also necessary to ensure software or compiler optimizations do not generate spurious accesses to non-idempotent memory regions. \end{commentary} \begin{commentary} Non-idempotent regions might not support misaligned accesses, in which case software might emulate misaligned accesses as a sequence of aligned accesses, each possibly causing side effects. Therefore, portable software should not issue misaligned accesses to non-idempotent regions. \end{commentary} \section{Physical Memory Protection} \label{sec:pmp} To support secure processing and contain faults, it is desirable to limit the physical addresses accessible by software running on a hart. An optional physical memory protection (PMP) unit provides per-hart machine-mode control registers to allow physical memory access privileges (read, write, execute) to be specified for each physical memory region. The PMP values are checked in parallel with the PMA checks described in Section~\ref{sec:pma}. The granularity of PMP access control settings are platform-specific and within a platform may vary by physical memory region, but the standard PMP encoding supports regions as small as four bytes. Certain regions' privileges can be hardwired---for example, some regions might only ever be visible in machine mode but in no lower-privilege layers. \begin{commentary} Platforms vary widely in demands for physical memory protection, and some platforms may provide other PMP structures in addition to or instead of the scheme described in this section. \end{commentary} PMP checks are applied to all accesses when the hart is running in S or U modes, and for loads and stores when the MPRV bit is set in the {\tt mstatus} register and the MPP field in the {\tt mstatus} register contains S or U. PMP checks are also applied to page-table accesses for virtual-address translation, for which the effective privilege mode is S. Optionally, PMP checks may additionally apply to M-mode accesses, in which case the PMP registers themselves are locked, so that even M-mode software cannot change them without a system reset. In effect, PMP can {\em grant} permissions to S and U modes, which by default have none, and can {\em revoke} permissions from M-mode, which by default has full permissions. PMP violations are always trapped precisely at the processor. \subsection{Physical Memory Protection CSRs} PMP entries are described by an 8-bit configuration register and one XLEN-bit address register. Some PMP settings additionally use the address register associated with the preceding PMP entry. Up to 16 PMP entries are supported. If any PMP entries are implemented, then all PMP CSRs must be implemented, but all PMP CSR fields are \warl\ and may be hardwired to zero. PMP CSRs are only accessible to M-mode. The PMP configuration registers are densely packed into CSRs to minimize context-switch time. For RV32, four CSRs, {\tt pmpcfg0}--{\tt pmpcfg3}, hold the configurations {\tt pmp0cfg}--{\tt pmp15cfg} for the 16 PMP entries, as shown in Figure~\ref{pmpcfg-rv32}. For RV64, {\tt pmpcfg0} and {\tt pmpcfg2} hold the configurations for the 16 PMP entries, as shown in Figure~\ref{pmpcfg-rv64}; {\tt pmpcfg1} and {\tt pmpcfg3} are illegal. \begin{commentary} RV64 systems use {\tt pmpcfg2}, rather than {\tt pmpcfg1}, to hold configurations for PMP entries 8--15. This design reduces the cost of supporting multiple M-XLEN values, since the configurations for PMP entries 8--11 appear in {\tt pmpcfg2}[31:0] for both RV32 and RV64. \end{commentary} \begin{figure}[h!] {\footnotesize \begin{center} \begin{tabular}{@{}Y@{}Y@{}Y@{}Yl} \instbitrange{31}{24} & \instbitrange{23}{16} & \instbitrange{15}{8} & \instbitrange{7}{0} & \\ \cline{1-4} \multicolumn{1}{|c|}{pmp3cfg} & \multicolumn{1}{c|}{pmp2cfg} & \multicolumn{1}{c|}{pmp1cfg} & \multicolumn{1}{c|}{pmp0cfg} & \tt pmpcfg0 \\ \cline{1-4} 8 & 8 & 8 & 8 & \\ \instbitrange{31}{24} & \instbitrange{23}{16} & \instbitrange{15}{8} & \instbitrange{7}{0} & \\ \cline{1-4} \multicolumn{1}{|c|}{pmp7cfg} & \multicolumn{1}{c|}{pmp6cfg} & \multicolumn{1}{c|}{pmp5cfg} & \multicolumn{1}{c|}{pmp4cfg} & \tt pmpcfg1 \\ \cline{1-4} 8 & 8 & 8 & 8 & \\ \instbitrange{31}{24} & \instbitrange{23}{16} & \instbitrange{15}{8} & \instbitrange{7}{0} & \\ \cline{1-4} \multicolumn{1}{|c|}{pmp11cfg} & \multicolumn{1}{c|}{pmp10cfg} & \multicolumn{1}{c|}{pmp9cfg} & \multicolumn{1}{c|}{pmp8cfg} & \tt pmpcfg2 \\ \cline{1-4} 8 & 8 & 8 & 8 & \\ \instbitrange{31}{24} & \instbitrange{23}{16} & \instbitrange{15}{8} & \instbitrange{7}{0} & \\ \cline{1-4} \multicolumn{1}{|c|}{pmp15cfg} & \multicolumn{1}{c|}{pmp14cfg} & \multicolumn{1}{c|}{pmp13cfg} & \multicolumn{1}{c|}{pmp12cfg} & \tt pmpcfg3 \\ \cline{1-4} 8 & 8 & 8 & 8 & \\ \end{tabular} \end{center} } \vspace{-0.1in} \caption{RV32 PMP configuration CSR layout.} \label{pmpcfg-rv32} \end{figure} \begin{figure}[h!] {\footnotesize \begin{center} \begin{tabular}{@{}Y@{}Y@{}Y@{}Y@{}Y@{}Y@{}Y@{}Yl} \instbitrange{63}{56} & \instbitrange{55}{48} & \instbitrange{47}{40} & \instbitrange{39}{32} & \instbitrange{31}{24} & \instbitrange{23}{16} & \instbitrange{15}{8} & \instbitrange{7}{0} & \\ \cline{1-8} \multicolumn{1}{|c|}{pmp7cfg} & \multicolumn{1}{c|}{pmp6cfg} & \multicolumn{1}{c|}{pmp5cfg} & \multicolumn{1}{c|}{pmp4cfg} & \multicolumn{1}{c|}{pmp3cfg} & \multicolumn{1}{c|}{pmp2cfg} & \multicolumn{1}{c|}{pmp1cfg} & \multicolumn{1}{c|}{pmp0cfg} & \tt pmpcfg0 \\ \cline{1-8} 8 & 8 & 8 & 8 & 8 & 8 & 8 & 8 & \\ \instbitrange{63}{56} & \instbitrange{55}{48} & \instbitrange{47}{40} & \instbitrange{39}{32} & \instbitrange{31}{24} & \instbitrange{23}{16} & \instbitrange{15}{8} & \instbitrange{7}{0} & \\ \cline{1-8} \multicolumn{1}{|c|}{pmp15cfg} & \multicolumn{1}{c|}{pmp14cfg} & \multicolumn{1}{c|}{pmp13cfg} & \multicolumn{1}{c|}{pmp12cfg} & \multicolumn{1}{c|}{pmp11cfg} & \multicolumn{1}{c|}{pmp10cfg} & \multicolumn{1}{c|}{pmp9cfg} & \multicolumn{1}{c|}{pmp8cfg} & \tt pmpcfg2 \\ \cline{1-8} 8 & 8 & 8 & 8 & 8 & 8 & 8 & 8 & \\ \end{tabular} \end{center} } \vspace{-0.1in} \caption{RV64 PMP configuration CSR layout.} \label{pmpcfg-rv64} \end{figure} The PMP address registers are CSRs named {\tt pmpaddr0}--{\tt pmpaddr15}. Each PMP address register encodes bits 33--2 of a 34-bit physical address for RV32, as shown in Figure~\ref{pmpaddr-rv32}. For RV64, each PMP address register encodes bits 55--2 of a 56-bit physical address, as shown in Figure~\ref{pmpaddr-rv64}. Not all physical address bits may be implemented, and so the {\tt pmpaddr} registers are \warl. \begin{commentary} The Sv32 page-based virtual-memory scheme described in Section~\ref{sec:sv32} supports 34-bit physical addresses for RV32, so the PMP scheme must support addresses wider than XLEN for RV32. The Sv39 and Sv48 page-based virtual-memory schemes described in Sections~\ref{sec:sv39} and~\ref{sec:sv48} support a 56-bit physical address space, so the RV64 PMP address registers impose the same limit. \end{commentary} \begin{figure}[h!] {\footnotesize \begin{center} \begin{tabular}{@{}J} \instbitrange{31}{0} \\ \hline \multicolumn{1}{|c|}{address[33:2] (\warl)} \\ \hline 32 \\ \end{tabular} \end{center} } \vspace{-0.1in} \caption{PMP address register format, RV32.} \label{pmpaddr-rv32} \end{figure} \begin{figure}[h!] {\footnotesize \begin{center} \begin{tabular}{@{}F@{}J} \instbitrange{63}{54} & \instbitrange{53}{0} \\ \hline \multicolumn{1}{|c|}{\wiri} & \multicolumn{1}{c|}{address[55:2] (\warl)} \\ \hline 32 \\ \end{tabular} \end{center} } \vspace{-0.1in} \caption{PMP address register format, RV64.} \label{pmpaddr-rv64} \end{figure} Figure~\ref{pmpcfg} shows the layout of a PMP configuration register. The R, W, and X bits, when set, indicate that the PMP entry permits read, write, and instruction execution, respectively. When one of these bits is clear, the corresponding access type is denied. The remaining two fields, A and L, are described in the following sections. \begin{figure}[h!] {\footnotesize \begin{center} \begin{tabular}{YSSYYY} \instbit{7} & \instbitrange{6}{5} & \instbitrange{4}{3} & \instbit{2} & \instbit{1} & \instbit{0} \\ \hline \multicolumn{1}{|c|}{L (\warl)} & \multicolumn{1}{c|}{\wiri} & \multicolumn{1}{c|}{A (\warl)} & \multicolumn{1}{c|}{X (\warl)} & \multicolumn{1}{c|}{W (\warl)} & \multicolumn{1}{c|}{R (\warl)} \\ \hline 1 & 2 & 2 & 1 & 1 & 1 \\ \end{tabular} \end{center} } \vspace{-0.1in} \caption{PMP configuration register format.} \label{pmpcfg} \end{figure} \subsubsection*{Address Matching} The A field in a PMP entry's configuration register encodes the address-matching mode of the associated PMP address register. The encoding of this field is shown in Table~\ref{pmpcfg-a}. When A=0, this PMP entry is disabled and matches no addresses. Two other address-matching modes are supported: naturally aligned power-of-2 regions (NAPOT), including the special case of naturally aligned four-byte regions (NA4); and the top boundary of an arbitrary range (TOR). These modes support four-byte granularity. \begin{table*}[h!] \begin{center} \begin{tabular}{|r|c|l|} \hline A & Name & Description \\ \hline 0 & OFF & Null region (disabled) \\ 1 & TOR & Top of range \\ 2 & NA4 & Naturally aligned four-byte region \\ 3 & NAPOT & Naturally aligned power-of-two region, $\ge$8 bytes \\ \hline \end{tabular} \end{center} \caption{Encoding of A field in PMP configuration registers.} \label{pmpcfg-a} \end{table*} NAPOT ranges make use of the low-order bits of the associated address register to encode the size of the range, as shown in Table~\ref{pmpcfg-napot}. \begin{table*}[h!] \begin{center} \begin{tabular}{|c|c|l|} \hline \tt pmpaddr & {\tt pmpcfg}.A & Match type and size \\ \hline \tt aaaa...aaaa & NA4 & 4-byte NAPOT range \\ \tt aaaa...aaa0 & NAPOT & 8-byte NAPOT range \\ \tt aaaa...aa01 & NAPOT & 16-byte NAPOT range \\ \tt aaaa...a011 & NAPOT & 32-byte NAPOT range \\ \multicolumn{1}{|c|}{\ldots} & \ldots & \multicolumn{1}{|c|}{\ldots} \\ \tt aa01...1111 & NAPOT & $2^{XLEN}$-byte NAPOT range \\ \tt a011...1111 & NAPOT & $2^{XLEN+1}$-byte NAPOT range \\ \tt 0111...1111 & NAPOT & $2^{XLEN+2}$-byte NAPOT range \\ \hline \end{tabular} \end{center} \caption{NAPOT range encoding in PMP address and configuration registers.} \label{pmpcfg-napot} \end{table*} If TOR is selected, the associated address register forms the top of the address range, and the preceding PMP address register forms the bottom of the address range. If PMP entry $i$'s A field is set to TOR, the entry matches any address $a$ such that ${\tt pmpaddr}_{i-1}\leq a < {\tt pmpaddr}_i$. If PMP entry 0's A field is set to TOR, zero is used for the lower bound, and so it matches any address $a < {\tt pmpaddr}_0$. \subsubsection*{Locking and Privilege Mode} The L bit indicates that the PMP entry is locked, i.e., writes to the configuration register and associated address registers are ignored. Locked PMP entries may only be unlocked with a system reset. If PMP entry $i$ is locked, writes to {\tt pmp}$i${\tt cfg} and {\tt pmpaddr}$i$ are ignored. Additionally, if {\tt pmp}$i${\tt cfg}.A is set to TOR, writes to {\tt pmpaddr}$i$-1 are ignored. In addition to locking the PMP entry, the L bit indicates whether the R/W/X permissions are enforced on M-mode accesses. When the L bit is set, these permissions are enforced for all privilege modes. When the L bit is clear, any M-mode access matching the PMP entry will succeed; the R/W/X permissions apply only to S and U modes. \subsubsection*{Priority and Matching Logic} PMP entries are statically prioritized. The lowest-numbered PMP entry that matches any byte of an access determines whether that access succeeds or fails. The matching PMP entry must match all bytes of an access, or the access fails, irrespective of the L, R, W, and X bits. For example, if a PMP entry is configured to match the four-byte range {\tt 0xC}--{\tt 0xF}, then an 8-byte access to the range {\tt 0x8}--{\tt 0xF} will fail, assuming that PMP entry is the highest-priority entry that matches those addresses. If a PMP entry matches all bytes of an access, then the L, R, W, and X bits determine whether the access succeeds or fails. If the L bit is clear and the privilege mode of the access is M, the access succeeds. Otherwise, if the L bit is set or the privilege mode of the access is S or U, then the access succeeds only if the R, W, or X bit corresponding to the access type is set. If no PMP entry matches an M-mode access, the access succeeds. If no PMP entry matches an S-mode or U-mode access, but at least one PMP entry is implemented, the access fails. Failed accesses generate a load, store, or instruction access exception. Note that a single instruction may generate multiple accesses, which may not be mutually atomic. An access exception is generated if at least one access generated by an instruction fails, though other accesses generated by that instruction may succeed with visible side effects. Notably, instructions that reference virtual memory are decomposed into multiple accesses. On some implementations, misaligned loads, stores, and instruction fetches may also be decomposed into multiple accesses, some of which may succeed before an access exception occurs. In particular, a portion of a misaligned store that passes the PMP check may become visible, even if another portion fails the PMP check. The same behavior may manifest for floating-point stores wider than XLEN bits (e.g., the FSD instruction in RV32D), even when the store address is naturally aligned.