Commit eb63cfcd authored by Johan Almbladh's avatar Johan Almbladh Committed by Andrii Nakryiko
Browse files

mips, bpf: Add eBPF JIT for 32-bit MIPS 



This is an implementation of an eBPF JIT for 32-bit MIPS I-V and MIPS32.
The implementation supports all 32-bit and 64-bit ALU and JMP operations,
including the recently-added atomics. 64-bit div/mod and 64-bit atomics
are implemented using function calls to math64 and atomic64 functions,
respectively. All 32-bit operations are implemented natively by the JIT,
except if the CPU lacks ll/sc instructions.

Register mapping
================
All 64-bit eBPF registers are mapped to native 32-bit MIPS register pairs,
and does not use any stack scratch space for register swapping. This means
that all eBPF register data is kept in CPU registers all the time, and
this simplifies the register management a lot. It also reduces the JIT's
pressure on temporary registers since we do not have to move data around.

Native register pairs are ordered according to CPU endiannes, following
the O32 calling convention for passing 64-bit arguments and return values.
The eBPF return value, arguments and callee-saved registers are mapped to
their native MIPS equivalents.

Since the 32 highest bits in the eBPF FP (frame pointer) register are
always zero, only one general-purpose register is actually needed for the
mapping. The MIPS fp register is used for this purpose. The high bits are
mapped to MIPS register r0. This saves us one CPU register, which is much
needed for temporaries, while still allowing us to treat the R10 (FP)
register just like any other eBPF register in the JIT.

The MIPS gp (global pointer) and at (assembler temporary) registers are
used as internal temporary registers for constant blinding. CPU registers
t6-t9 are used internally by the JIT when constructing more complex 64-bit
operations. This is precisely what is needed - two registers to store an
operand value, and two more as scratch registers when performing the
operation.

The register mapping is shown below.

    R0 - $v1, $v0   return value
    R1 - $a1, $a0   argument 1, passed in registers
    R2 - $a3, $a2   argument 2, passed in registers
    R3 - $t1, $t0   argument 3, passed on stack
    R4 - $t3, $t2   argument 4, passed on stack
    R5 - $t4, $t3   argument 5, passed on stack
    R6 - $s1, $s0   callee-saved
    R7 - $s3, $s2   callee-saved
    R8 - $s5, $s4   callee-saved
    R9 - $s7, $s6   callee-saved
    FP - $r0, $fp   32-bit frame pointer
    AX - $gp, $at   constant-blinding
         $t6 - $t9  unallocated, JIT temporaries

Jump offsets
============
The JIT tries to map all conditional JMP operations to MIPS conditional
PC-relative branches. The MIPS branch offset field is 18 bits, in bytes,
which is equivalent to the eBPF 16-bit instruction offset. However, since
the JIT may emit more than one CPU instruction per eBPF instruction, the
field width may overflow. If that happens, the JIT converts the long
conditional jump to a short PC-relative branch with the condition
inverted, jumping over a long unconditional absolute jmp (j).

This conversion will change the instruction offset mapping used for jumps,
and may in turn result in more branch offset overflows. The JIT therefore
dry-runs the translation until no more branches are converted and the
offsets do not change anymore. There is an upper bound on this of course,
and if the JIT hits that limit, the last two iterations are run with all
branches being converted.

Tail call count
===============
The current tail call count is stored in the 16-byte area of the caller's
stack frame that is reserved for the callee in the o32 ABI. The value is
initialized in the prologue, and propagated to the tail-callee by skipping
the initialization instructions when emitting the tail call.

Signed-off-by: default avatarJohan Almbladh <johan.almbladh@anyfinetworks.com>
Signed-off-by: default avatarDaniel Borkmann <daniel@iogearbox.net>
Signed-off-by: default avatarAndrii Nakryiko <andrii@kernel.org>
Link: https://lore.kernel.org/bpf/20211005165408.2305108-4-johan.almbladh@anyfinetworks.com
parent f7c036c1

arch/mips/net/Makefile

+6 −1
Original line number Diff line number Diff line
@@ -2,4 +2,9 @@ 
# MIPS networking code



obj-$(CONFIG_MIPS_CBPF_JIT) += bpf_jit.o bpf_jit_asm.o



ifeq ($(CONFIG_32BIT),y)

        obj-$(CONFIG_MIPS_EBPF_JIT) += bpf_jit_comp.o bpf_jit_comp32.o

else

        obj-$(CONFIG_MIPS_EBPF_JIT) += ebpf_jit.o

endif

arch/mips/net/bpf_jit_comp.c

0 → 100644
+1032 −0

File added.

Preview size limit exceeded, changes collapsed.

arch/mips/net/bpf_jit_comp.h

0 → 100644
+211 −0
Original line number Diff line number Diff line 
/* SPDX-License-Identifier: GPL-2.0-only */

/*

 * Just-In-Time compiler for eBPF bytecode on 32-bit and 64-bit MIPS.

 *

 * Copyright (c) 2021 Anyfi Networks AB.

 * Author: Johan Almbladh <johan.almbladh@gmail.com>

 *

 * Based on code and ideas from

 * Copyright (c) 2017 Cavium, Inc.

 * Copyright (c) 2017 Shubham Bansal <illusionist.neo@gmail.com>

 * Copyright (c) 2011 Mircea Gherzan <mgherzan@gmail.com>

 */



#ifndef _BPF_JIT_COMP_H

#define _BPF_JIT_COMP_H



/* MIPS registers */

#define MIPS_R_ZERO	0   /* Const zero */

#define MIPS_R_AT	1   /* Asm temp   */

#define MIPS_R_V0	2   /* Result     */

#define MIPS_R_V1	3   /* Result     */

#define MIPS_R_A0	4   /* Argument   */

#define MIPS_R_A1	5   /* Argument   */

#define MIPS_R_A2	6   /* Argument   */

#define MIPS_R_A3	7   /* Argument   */

#define MIPS_R_A4	8   /* Arg (n64)  */

#define MIPS_R_A5	9   /* Arg (n64)  */

#define MIPS_R_A6	10  /* Arg (n64)  */

#define MIPS_R_A7	11  /* Arg (n64)  */

#define MIPS_R_T0	8   /* Temp (o32) */

#define MIPS_R_T1	9   /* Temp (o32) */

#define MIPS_R_T2	10  /* Temp (o32) */

#define MIPS_R_T3	11  /* Temp (o32) */

#define MIPS_R_T4	12  /* Temporary  */

#define MIPS_R_T5	13  /* Temporary  */

#define MIPS_R_T6	14  /* Temporary  */

#define MIPS_R_T7	15  /* Temporary  */

#define MIPS_R_S0	16  /* Saved      */

#define MIPS_R_S1	17  /* Saved      */

#define MIPS_R_S2	18  /* Saved      */

#define MIPS_R_S3	19  /* Saved      */

#define MIPS_R_S4	20  /* Saved      */

#define MIPS_R_S5	21  /* Saved      */

#define MIPS_R_S6	22  /* Saved      */

#define MIPS_R_S7	23  /* Saved      */

#define MIPS_R_T8	24  /* Temporary  */

#define MIPS_R_T9	25  /* Temporary  */

/*      MIPS_R_K0	26     Reserved   */

/*      MIPS_R_K1	27     Reserved   */

#define MIPS_R_GP	28  /* Global ptr */

#define MIPS_R_SP	29  /* Stack ptr  */

#define MIPS_R_FP	30  /* Frame ptr  */

#define MIPS_R_RA	31  /* Return     */



/*

 * Jump address mask for immediate jumps. The four most significant bits

 * must be equal to PC.

 */

#define MIPS_JMP_MASK	0x0fffffffUL



/* Maximum number of iterations in offset table computation */

#define JIT_MAX_ITERATIONS	8



/*

 * Jump pseudo-instructions used internally

 * for branch conversion and branch optimization.

 */

#define JIT_JNSET	0xe0

#define JIT_JNOP	0xf0



/* Descriptor flag for PC-relative branch conversion */

#define JIT_DESC_CONVERT	BIT(31)



/* JIT context for an eBPF program */

struct jit_context {

	struct bpf_prog *program;     /* The eBPF program being JITed        */

	u32 *descriptors;             /* eBPF to JITed CPU insn descriptors  */

	u32 *target;                  /* JITed code buffer                   */

	u32 bpf_index;                /* Index of current BPF program insn   */

	u32 jit_index;                /* Index of current JIT target insn    */

	u32 changes;                  /* Number of PC-relative branch conv   */

	u32 accessed;                 /* Bit mask of read eBPF registers     */

	u32 clobbered;                /* Bit mask of modified CPU registers  */

	u32 stack_size;               /* Total allocated stack size in bytes */

	u32 saved_size;               /* Size of callee-saved registers      */

	u32 stack_used;               /* Stack size used for function calls  */

};



/* Emit the instruction if the JIT memory space has been allocated */

#define emit(ctx, func, ...)					\

do {								\

	if ((ctx)->target != NULL) {				\

		u32 *p = &(ctx)->target[ctx->jit_index];	\

		uasm_i_##func(&p, ##__VA_ARGS__);		\

	}							\

	(ctx)->jit_index++;					\

} while (0)



/*

 * Mark a BPF register as accessed, it needs to be

 * initialized by the program if expected, e.g. FP.

 */

static inline void access_reg(struct jit_context *ctx, u8 reg)

{

	ctx->accessed |= BIT(reg);

}



/*

 * Mark a CPU register as clobbered, it needs to be

 * saved/restored by the program if callee-saved.

 */

static inline void clobber_reg(struct jit_context *ctx, u8 reg)

{

	ctx->clobbered |= BIT(reg);

}



/*

 * Push registers on the stack, starting at a given depth from the stack

 * pointer and increasing. The next depth to be written is returned.

 */

int push_regs(struct jit_context *ctx, u32 mask, u32 excl, int depth);



/*

 * Pop registers from the stack, starting at a given depth from the stack

 * pointer and increasing. The next depth to be read is returned.

 */

int pop_regs(struct jit_context *ctx, u32 mask, u32 excl, int depth);



/* Compute the 28-bit jump target address from a BPF program location */

int get_target(struct jit_context *ctx, u32 loc);



/* Compute the PC-relative offset to relative BPF program offset */

int get_offset(const struct jit_context *ctx, int off);



/* dst = imm (32-bit) */

void emit_mov_i(struct jit_context *ctx, u8 dst, s32 imm);



/* dst = src (32-bit) */

void emit_mov_r(struct jit_context *ctx, u8 dst, u8 src);



/* Validate ALU/ALU64 immediate range */

bool valid_alu_i(u8 op, s32 imm);



/* Rewrite ALU/ALU64 immediate operation */

bool rewrite_alu_i(u8 op, s32 imm, u8 *alu, s32 *val);



/* ALU immediate operation (32-bit) */

void emit_alu_i(struct jit_context *ctx, u8 dst, s32 imm, u8 op);



/* ALU register operation (32-bit) */

void emit_alu_r(struct jit_context *ctx, u8 dst, u8 src, u8 op);



/* Atomic read-modify-write (32-bit) */

void emit_atomic_r(struct jit_context *ctx, u8 dst, u8 src, s16 off, u8 code);



/* Atomic compare-and-exchange (32-bit) */

void emit_cmpxchg_r(struct jit_context *ctx, u8 dst, u8 src, u8 res, s16 off);



/* Swap bytes and truncate a register word or half word */

void emit_bswap_r(struct jit_context *ctx, u8 dst, u32 width);



/* Validate JMP/JMP32 immediate range */

bool valid_jmp_i(u8 op, s32 imm);



/* Prepare a PC-relative jump operation with immediate conditional */

void setup_jmp_i(struct jit_context *ctx, s32 imm, u8 width,

		 u8 bpf_op, s16 bpf_off, u8 *jit_op, s32 *jit_off);



/* Prepare a PC-relative jump operation with register conditional */

void setup_jmp_r(struct jit_context *ctx, bool same_reg,

		 u8 bpf_op, s16 bpf_off, u8 *jit_op, s32 *jit_off);



/* Finish a PC-relative jump operation */

int finish_jmp(struct jit_context *ctx, u8 jit_op, s16 bpf_off);



/* Conditional JMP/JMP32 immediate */

void emit_jmp_i(struct jit_context *ctx, u8 dst, s32 imm, s32 off, u8 op);



/* Conditional JMP/JMP32 register */

void emit_jmp_r(struct jit_context *ctx, u8 dst, u8 src, s32 off, u8 op);



/* Jump always */

int emit_ja(struct jit_context *ctx, s16 off);



/* Jump to epilogue */

int emit_exit(struct jit_context *ctx);



/*

 * Build program prologue to set up the stack and registers.

 * This function is implemented separately for 32-bit and 64-bit JITs.

 */

void build_prologue(struct jit_context *ctx);



/*

 * Build the program epilogue to restore the stack and registers.

 * This function is implemented separately for 32-bit and 64-bit JITs.

 */

void build_epilogue(struct jit_context *ctx, int dest_reg);



/*

 * Convert an eBPF instruction to native instruction, i.e

 * JITs an eBPF instruction.

 * Returns :

 *	0  - Successfully JITed an 8-byte eBPF instruction

 *	>0 - Successfully JITed a 16-byte eBPF instruction

 *	<0 - Failed to JIT.

 * This function is implemented separately for 32-bit and 64-bit JITs.

 */

int build_insn(const struct bpf_insn *insn, struct jit_context *ctx);



#endif /* _BPF_JIT_COMP_H */

arch/mips/net/bpf_jit_comp32.c

0 → 100644
+1899 −0

File added.

Preview size limit exceeded, changes collapsed.