From patchwork Thu Aug 17 21:45:00 2023 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Helge Deller X-Patchwork-Id: 13357017 Return-Path: X-Spam-Checker-Version: SpamAssassin 3.4.0 (2014-02-07) on aws-us-west-2-korg-lkml-1.web.codeaurora.org Received: from vger.kernel.org (vger.kernel.org [23.128.96.18]) by smtp.lore.kernel.org (Postfix) with ESMTP id 9D884C678DC for ; Thu, 17 Aug 2023 21:46:05 +0000 (UTC) Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1355413AbjHQVpd (ORCPT ); Thu, 17 Aug 2023 17:45:33 -0400 Received: from lindbergh.monkeyblade.net ([23.128.96.19]:53668 "EHLO lindbergh.monkeyblade.net" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1355436AbjHQVpQ (ORCPT ); Thu, 17 Aug 2023 17:45:16 -0400 Received: from dfw.source.kernel.org (dfw.source.kernel.org [139.178.84.217]) by lindbergh.monkeyblade.net (Postfix) with ESMTPS id DF3F61BF9 for ; Thu, 17 Aug 2023 14:45:11 -0700 (PDT) Received: from smtp.kernel.org (relay.kernel.org [52.25.139.140]) (using TLSv1.3 with cipher TLS_AES_256_GCM_SHA384 (256/256 bits) key-exchange X25519 server-signature RSA-PSS (2048 bits)) (No client certificate requested) by dfw.source.kernel.org (Postfix) with ESMTPS id 6EA9B63309 for ; Thu, 17 Aug 2023 21:45:11 +0000 (UTC) Received: by smtp.kernel.org (Postfix) with ESMTPSA id 218FFC433C7; Thu, 17 Aug 2023 21:45:09 +0000 (UTC) DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/simple; d=kernel.org; s=k20201202; t=1692308710; bh=PmjiP6RqbQ4c0/2RUBUdyuSGEP2kDQbhU68Bar93eqs=; h=From:To:Cc:Subject:Date:In-Reply-To:References:From; b=Nljf4fba/Q1hKBq80cSfORlqDOEtxaUPSd6li0EWUgWerz6hN8Pt6ATOSnrTrwZfe bngYqaiL2WrbF/Jv99urHJoQ5K44OPEXDGj6cuR9A+/OmYDWMHXWU8JKprPOJFUftU aLWo5gZgqIq8poOuBuXo7ZDKj5+XUPx5rEeW7wEeCqNJggW0adwEnhie5opu4wS7Mx deBJooPRYS1YWQJEcjBPK9nV2X++F9YBBA1Dh1Y9JMlv4xm6y5VR+qZk8Z+d0uZmgq iYmU6UjhTu0pXcaiEl6FLR44z6/Vs7DmvZq1MfQALVQAa4j2gwbEikKaYOrOb6X8fp RZKEJEhXj3AHg== From: deller@kernel.org To: linux-parisc@vger.kernel.org Cc: Helge Deller Subject: [PATCH 3/5] parisc: Add eBPF JIT compiler glue code and Makefile Date: Thu, 17 Aug 2023 23:45:00 +0200 Message-ID: <20230817214503.108660-4-deller@kernel.org> X-Mailer: git-send-email 2.41.0 In-Reply-To: <20230817214503.108660-1-deller@kernel.org> References: <20230817214503.108660-1-deller@kernel.org> MIME-Version: 1.0 Precedence: bulk List-ID: X-Mailing-List: linux-parisc@vger.kernel.org From: Helge Deller Signed-off-by: Helge Deller --- arch/parisc/net/Makefile | 9 + arch/parisc/net/bpf_jit.h | 479 +++++++++++++++++++++++++++++++++ arch/parisc/net/bpf_jit_core.c | 207 ++++++++++++++ 3 files changed, 695 insertions(+) create mode 100644 arch/parisc/net/Makefile create mode 100644 arch/parisc/net/bpf_jit.h create mode 100644 arch/parisc/net/bpf_jit_core.c diff --git a/arch/parisc/net/Makefile b/arch/parisc/net/Makefile new file mode 100644 index 000000000000..22b12024d4c3 --- /dev/null +++ b/arch/parisc/net/Makefile @@ -0,0 +1,9 @@ +# SPDX-License-Identifier: GPL-2.0-only + +obj-$(CONFIG_BPF_JIT) += bpf_jit_core.o + +ifeq ($(CONFIG_64BIT),y) + obj-$(CONFIG_BPF_JIT) += bpf_jit_comp64.o +else + obj-$(CONFIG_BPF_JIT) += bpf_jit_comp32.o +endif diff --git a/arch/parisc/net/bpf_jit.h b/arch/parisc/net/bpf_jit.h new file mode 100644 index 000000000000..8b8896959f04 --- /dev/null +++ b/arch/parisc/net/bpf_jit.h @@ -0,0 +1,479 @@ +/* SPDX-License-Identifier: GPL-2.0 */ +/* + * Common functionality for PARISC32 and PARISC64 BPF JIT compilers + * + * Copyright (c) 2023 Helge Deller + * + */ + +#ifndef _BPF_JIT_H +#define _BPF_JIT_H + +#include +#include +#include + +#define HPPA_JIT_DEBUG 0 +#define HPPA_JIT_REBOOT 0 +#define HPPA_JIT_DUMP 0 + +#define OPTIMIZE_HPPA 1 /* enable some asm optimizations */ +// echo 1 > /proc/sys/net/core/bpf_jit_enable + +#define HPPA_R(nr) nr /* use HPPA register #nr */ + +enum { + HPPA_REG_ZERO = 0, /* The constant value 0 */ + HPPA_REG_R1 = 1, /* used for addil */ + HPPA_REG_RP = 2, /* Return address */ + + HPPA_REG_ARG7 = 19, /* ARG4-7 used in 64-bit ABI */ + HPPA_REG_ARG6 = 20, + HPPA_REG_ARG5 = 21, + HPPA_REG_ARG4 = 22, + + HPPA_REG_ARG3 = 23, /* ARG0-3 in 32- and 64-bit ABI */ + HPPA_REG_ARG2 = 24, + HPPA_REG_ARG1 = 25, + HPPA_REG_ARG0 = 26, + HPPA_REG_GP = 27, /* Global pointer */ + HPPA_REG_RET0 = 28, /* Return value, HI in 32-bit */ + HPPA_REG_RET1 = 29, /* Return value, LOW in 32-bit */ + HPPA_REG_SP = 30, /* Stack pointer */ + HPPA_REG_R31 = 31, + +#ifdef CONFIG_64BIT + HPPA_REG_TCC = 3, + HPPA_REG_TCC_SAVED = 4, + HPPA_REG_TCC_IN_INIT = HPPA_REG_R31, +#else + HPPA_REG_TCC = 18, + HPPA_REG_TCC_SAVED = 17, + HPPA_REG_TCC_IN_INIT = HPPA_REG_R31, +#endif + + HPPA_REG_T0 = HPPA_REG_R1, /* Temporaries */ + HPPA_REG_T1 = HPPA_REG_R31, + HPPA_REG_T2 = HPPA_REG_ARG4, +#ifndef CONFIG_64BIT + HPPA_REG_T3 = HPPA_REG_ARG5, /* not used in 64-bit */ + HPPA_REG_T4 = HPPA_REG_ARG6, + HPPA_REG_T5 = HPPA_REG_ARG7, +#endif +}; + +struct hppa_jit_context { + struct bpf_prog *prog; + u32 *insns; /* HPPA insns */ + int ninsns; + int reg_seen_collect; + int reg_seen; + int body_len; + int epilogue_offset; + int prologue_len; + int *offset; /* BPF to HPPA */ +}; + +#define REG_SET_SEEN(ctx, nr) { if (ctx->reg_seen_collect) ctx->reg_seen |= BIT(nr); } +#define REG_SET_SEEN_ALL(ctx) { if (ctx->reg_seen_collect) ctx->reg_seen = -1; } +#define REG_FORCE_SEEN(ctx, nr) { ctx->reg_seen |= BIT(nr); } +#define REG_WAS_SEEN(ctx, nr) (ctx->reg_seen & BIT(nr)) +#define REG_ALL_SEEN(ctx) (ctx->reg_seen == -1) + +#define HPPA_INSN_SIZE 4 /* bytes per HPPA asm instruction */ +#define REG_SIZE REG_SZ /* bytes per native "long" word */ + +/* subtract hppa displacement on branches which is .+8 */ +#define HPPA_BRANCH_DISPLACEMENT 2 /* instructions */ + +/* asm statement indicator to execute delay slot */ +#define EXEC_NEXT_INSTR 0 +#define NOP_NEXT_INSTR 1 + +#define im11(val) (((u32)(val)) & 0x07ff) + +#define hppa_ldil(addr, reg) \ + hppa_t5_insn(0x08, reg, ((u32)(addr)) >> 11) /* ldil im21,reg */ +#define hppa_addil(addr, reg) \ + hppa_t5_insn(0x0a, reg, ((u32)(addr)) >> 11) /* addil im21,reg -> result in gr1 */ +#define hppa_ldo(im14, reg, target) \ + hppa_t1_insn(0x0d, reg, target, im14) /* ldo val14(reg),target */ +#define hppa_ldi(im14, reg) \ + hppa_ldo(im14, HPPA_REG_ZERO, reg) /* ldi val14,reg */ +#define hppa_or(reg1, reg2, target) \ + hppa_t6_insn(0x02, reg2, reg1, 0, 0, 0x09, target) /* or reg1,reg2,target */ +#define hppa_or_cond(reg1, reg2, cond, f, target) \ + hppa_t6_insn(0x02, reg2, reg1, cond, f, 0x09, target) +#define hppa_and(reg1, reg2, target) \ + hppa_t6_insn(0x02, reg2, reg1, 0, 0, 0x08, target) /* and reg1,reg2,target */ +#define hppa_and_cond(reg1, reg2, cond, f, target) \ + hppa_t6_insn(0x02, reg2, reg1, cond, f, 0x08, target) +#define hppa_xor(reg1, reg2, target) \ + hppa_t6_insn(0x02, reg2, reg1, 0, 0, 0x0a, target) /* xor reg1,reg2,target */ +#define hppa_add(reg1, reg2, target) \ + hppa_t6_insn(0x02, reg2, reg1, 0, 0, 0x18, target) /* add reg1,reg2,target */ +#define hppa_addc(reg1, reg2, target) \ + hppa_t6_insn(0x02, reg2, reg1, 0, 0, 0x1c, target) /* add,c reg1,reg2,target */ +#define hppa_sub(reg1, reg2, target) \ + hppa_t6_insn(0x02, reg2, reg1, 0, 0, 0x10, target) /* sub reg1,reg2,target */ +#define hppa_subb(reg1, reg2, target) \ + hppa_t6_insn(0x02, reg2, reg1, 0, 0, 0x14, target) /* sub,b reg1,reg2,target */ +#define hppa_nop() \ + hppa_or(0,0,0) /* nop: or 0,0,0 */ +#define hppa_addi(val11, reg, target) \ + hppa_t7_insn(0x2d, reg, target, val11) /* addi im11,reg,target */ +#define hppa_subi(val11, reg, target) \ + hppa_t7_insn(0x25, reg, target, val11) /* subi im11,reg,target */ +#define hppa_copy(reg, target) \ + hppa_or(reg, HPPA_REG_ZERO, target) /* copy reg,target */ +#define hppa_ldw(val14, reg, target) \ + hppa_t1_insn(0x12, reg, target, val14) /* ldw im14(reg),target */ +#define hppa_ldb(val14, reg, target) \ + hppa_t1_insn(0x10, reg, target, val14) /* ldb im14(reg),target */ +#define hppa_ldh(val14, reg, target) \ + hppa_t1_insn(0x11, reg, target, val14) /* ldh im14(reg),target */ +#define hppa_stw(reg, val14, base) \ + hppa_t1_insn(0x1a, base, reg, val14) /* stw reg,im14(base) */ +#define hppa_stb(reg, val14, base) \ + hppa_t1_insn(0x18, base, reg, val14) /* stb reg,im14(base) */ +#define hppa_sth(reg, val14, base) \ + hppa_t1_insn(0x19, base, reg, val14) /* sth reg,im14(base) */ +#define hppa_stwma(reg, val14, base) \ + hppa_t1_insn(0x1b, base, reg, val14) /* stw,ma reg,im14(base) */ +#define hppa_bv(reg, base, nop) \ + hppa_t11_insn(0x3a, base, reg, 0x06, 0, nop) /* bv(,n) reg(base) */ +#define hppa_be(offset, base) \ + hppa_t12_insn(0x38, base, offset, 0x00, 1) /* be,n offset(0,base) */ +#define hppa_be_l(offset, base, nop) \ + hppa_t12_insn(0x39, base, offset, 0x00, nop) /* ble(,nop) offset(0,base) */ +#define hppa_mtctl(reg, cr) \ + hppa_t21_insn(0x00, cr, reg, 0xc2, 0) /* mtctl reg,cr */ +#define hppa_mtsar(reg) \ + hppa_mtctl(reg, 11) /* mtsar reg */ +#define hppa_zdep(r, p, len, target) \ + hppa_t10_insn(0x35, target, r, 0, 2, p, len) /* zdep r,a,b,t */ +#define hppa_shl(r, len, target) \ + hppa_zdep(r, len, len, lo(rd)) +#define hppa_depwz(r, p, len, target) \ + hppa_t10_insn(0x35, target, r, 0, 3, 31-(p), 32-(len)) /* depw,z r,p,len,ret1 */ +#define hppa_depwz_sar(reg, target) \ + hppa_t1_insn(0x35, target, reg, 0) /* depw,z reg,sar,32,target */ +#define hppa_shrpw_sar(reg, target) \ + hppa_t10_insn(0x34, reg, 0, 0, 0, 0, target) /* shrpw r0,reg,sar,target */ +#define hppa_shrpw(r1, r2, p, target) \ + hppa_t10_insn(0x34, r2, r1, 0, 2, 31-(p), target) /* shrpw r1,r2,p,target */ +#define hppa_shd(r1, r2, p, target) \ + hppa_t10_insn(0x34, r2, r1, 0, 2, 31-(p), target) /* shrpw r1,r2,p,tarfer */ +#define hppa_extrws_sar(reg, target) \ + hppa_t10_insn(0x34, reg, target, 0, 5, 0, 0) /* extrw,s reg,sar,32,ret0 */ +#define hppa_extrws(reg, p, len, target) \ + hppa_t10_insn(0x34, reg, target, 0, 7, p, len) /* extrw,s reg,p,len,target */ +#define hppa_extru(r, p, len, target) \ + hppa_t10_insn(0x34, r, target, 0, 6, p, 32-(len)) +#define hppa_shr(r, len, target) \ + hppa_extru(r, 31-(len), 32-(len), target) +#define hppa_bl(imm17, rp) \ + hppa_t12_insn(0x3a, rp, imm17, 0x00, 1) /* bl,n target_addr,rp */ +#define hppa_sh2add(r1, r2, target) \ + hppa_t6_insn(0x02, r2, r1, 0, 0, 0x1a, target) /* sh2add r1,r2,target */ + +#define hppa_combt(r1, r2, target_addr, condition, nop) \ + hppa_t11_insn(IS_ENABLED(CONFIG_64BIT) ? 0x27 : 0x20, \ + r2, r1, condition, target_addr, nop) /* combt,cond,n r1,r2,addr */ +#define hppa_beq(r1, r2, target_addr) \ + hppa_combt(r1, r2, target_addr, 1, NOP_NEXT_INSTR) +#define hppa_blt(r1, r2, target_addr) \ + hppa_combt(r1, r2, target_addr, 2, NOP_NEXT_INSTR) +#define hppa_ble(r1, r2, target_addr) \ + hppa_combt(r1, r2, target_addr, 3, NOP_NEXT_INSTR) +#define hppa_bltu(r1, r2, target_addr) \ + hppa_combt(r1, r2, target_addr, 4, NOP_NEXT_INSTR) +#define hppa_bleu(r1, r2, target_addr) \ + hppa_combt(r1, r2, target_addr, 5, NOP_NEXT_INSTR) + +#define hppa_combf(r1, r2, target_addr, condition, nop) \ + hppa_t11_insn(IS_ENABLED(CONFIG_64BIT) ? 0x2f : 0x22, \ + r2, r1, condition, target_addr, nop) /* combf,cond,n r1,r2,addr */ +#define hppa_bne(r1, r2, target_addr) \ + hppa_combf(r1, r2, target_addr, 1, NOP_NEXT_INSTR) +#define hppa_bge(r1, r2, target_addr) \ + hppa_combf(r1, r2, target_addr, 2, NOP_NEXT_INSTR) +#define hppa_bgt(r1, r2, target_addr) \ + hppa_combf(r1, r2, target_addr, 3, NOP_NEXT_INSTR) +#define hppa_bgeu(r1, r2, target_addr) \ + hppa_combf(r1, r2, target_addr, 4, NOP_NEXT_INSTR) +#define hppa_bgtu(r1, r2, target_addr) \ + hppa_combf(r1, r2, target_addr, 5, NOP_NEXT_INSTR) + +/* 64-bit instructions */ +#ifdef CONFIG_64BIT +#define hppa64_ldd_reg(reg, b, target) \ + hppa_t10_insn(0x03, b, reg, 0, 0, 3<<1, target) +#define hppa64_ldd_im5(im5, b, target) \ + hppa_t10_insn(0x03, b, low_sign_unext(im5,5), 0, 1<<2, 3<<1, target) +#define hppa64_ldd_im16(im16, b, target) \ + hppa_t10_insn(0x14, b, target, 0, 0, 0, 0) | re_assemble_16(im16) +#define hppa64_std_im5(src, im5, b) \ + hppa_t10_insn(0x03, b, src, 0, 1<<2, 0xB<<1, low_sign_unext(im5,5)) +#define hppa64_std_im16(src, im16, b) \ + hppa_t10_insn(0x1c, b, src, 0, 0, 0, 0) | re_assemble_16(im16) +#define hppa64_bl_long(offs22) \ + hppa_t12_L_insn(0x3a, offs22, 1) +#define hppa64_mtsarcm(reg) \ + hppa_t21_insn(0x00, 11, reg, 0xc6, 0) +#define hppa64_shrpd_sar(reg, target) \ + hppa_t10_insn(0x34, reg, 0, 0, 0, 1<<4, target) +#define hppa64_shladd(r1, sa, r2, target) \ + hppa_t6_insn(0x02, r2, r1, 0, 0, 1<<4|1<<3|sa, target) +#define hppa64_depdz_sar(reg, target) \ + hppa_t21_insn(0x35, target, reg, 3<<3, 0) +#define hppa_extrd_sar(reg, target, se) \ + hppa_t10_insn(0x34, reg, target, 0, 0, 0, 0) | 2<<11 | (se&1)<<10 | 1<<9 | 1<<8 +#define hppa64_bve_l_rp(base) \ + (0x3a << 26) | (base << 21) | 0xf000 +#define hppa64_permh_3210(r, target) \ + (0x3e << 26) | (r << 21) | (r << 16) | (target) | 0x00006900 +#define hppa64_hshl(r, sa, target) \ + (0x3e << 26) | (0 << 21) | (r << 16) | (sa << 6) | (target) | 0x00008800 +#define hppa64_hshr_u(r, sa, target) \ + (0x3e << 26) | (r << 21) | (0 << 16) | (sa << 6) | (target) | 0x0000c800 +#endif + +struct hppa_jit_data { + struct bpf_binary_header *header; + u8 *image; + struct hppa_jit_context ctx; +}; + +static inline void bpf_fill_ill_insns(void *area, unsigned int size) +{ + memset(area, 0, size); +} + +static inline void bpf_flush_icache(void *start, void *end) +{ + flush_icache_range((unsigned long)start, (unsigned long)end); +} + +/* Emit a 4-byte HPPA instruction. */ +static inline void emit(const u32 insn, struct hppa_jit_context *ctx) +{ + if (ctx->insns) { + ctx->insns[ctx->ninsns] = insn; + } + + ctx->ninsns++; +} + +static inline int epilogue_offset(struct hppa_jit_context *ctx) +{ + int to = ctx->epilogue_offset, from = ctx->ninsns; + + return (to - from); +} + +/* Return -1 or inverted cond. */ +static inline int invert_bpf_cond(u8 cond) +{ + switch (cond) { + case BPF_JEQ: + return BPF_JNE; + case BPF_JGT: + return BPF_JLE; + case BPF_JLT: + return BPF_JGE; + case BPF_JGE: + return BPF_JLT; + case BPF_JLE: + return BPF_JGT; + case BPF_JNE: + return BPF_JEQ; + case BPF_JSGT: + return BPF_JSLE; + case BPF_JSLT: + return BPF_JSGE; + case BPF_JSGE: + return BPF_JSLT; + case BPF_JSLE: + return BPF_JSGT; + } + return -1; +} + + +static inline signed long hppa_offset(int insn, int off, struct hppa_jit_context *ctx) +{ + signed long from, to; + + off++; /* BPF branch is from PC+1 */ + from = (insn > 0) ? ctx->offset[insn - 1] : 0; + to = (insn + off > 0) ? ctx->offset[insn + off - 1] : 0; + return (to - from); +} + +/* does the signed value fits into a given number of bits ? */ +static inline int check_bits_int(signed long val, int bits) +{ + return ((val >= 0) && ((val >> bits) == 0)) || + ((val < 0) && (((~((u32)val)) >> (bits-1)) == 0)); +} + +/* can the signed value be used in relative code ? */ +static inline int relative_bits_ok(signed long val, int bits) +{ + return ((val >= 0) && (val < (1UL << (bits-1)))) || /* XXX */ + ((val < 0) && (((~((unsigned long)val)) >> (bits-1)) == 0) + && (val & (1UL << (bits-1)))); +} + +/* can the signed value be used in relative branches ? */ +static inline int relative_branch_ok(signed long val, int bits) +{ + return ((val >= 0) && (val < (1UL << (bits-2)))) || /* XXX */ + ((val < 0) && (((~((unsigned long)val)) < (1UL << (bits-2)))) + && (val & (1UL << (bits-1)))); +} + + +#define is_5b_int(val) check_bits_int(val, 5) + +static inline unsigned sign_unext(unsigned x, unsigned len) +{ + unsigned len_ones; + + len_ones = (1 << len) - 1; + return x & len_ones; +} + +static inline unsigned low_sign_unext(unsigned x, unsigned len) +{ + unsigned temp; + unsigned sign; + + sign = (x >> (len-1)) & 1; + temp = sign_unext (x, len-1); + return (temp << 1) | sign; +} + +static inline unsigned re_assemble_12(unsigned as12) +{ + return (( (as12 & 0x800) >> 11) + | ((as12 & 0x400) >> (10 - 2)) + | ((as12 & 0x3ff) << (1 + 2))); +} + +static inline unsigned re_assemble_14(unsigned as14) +{ + return (( (as14 & 0x1fff) << 1) + | ((as14 & 0x2000) >> 13)); +} + +#ifdef CONFIG_64BIT +static inline unsigned re_assemble_16(unsigned as16) +{ + unsigned s, t; + + /* Unusual 16-bit encoding, for wide mode only. */ + t = (as16 << 1) & 0xffff; + s = (as16 & 0x8000); + return (t ^ s ^ (s >> 1)) | (s >> 15); +} +#endif + +static inline unsigned re_assemble_17(unsigned as17) +{ + return (( (as17 & 0x10000) >> 16) + | ((as17 & 0x0f800) << (16 - 11)) + | ((as17 & 0x00400) >> (10 - 2)) + | ((as17 & 0x003ff) << (1 + 2))); +} + +static inline unsigned re_assemble_21(unsigned as21) +{ + return (( (as21 & 0x100000) >> 20) + | ((as21 & 0x0ffe00) >> 8) + | ((as21 & 0x000180) << 7) + | ((as21 & 0x00007c) << 14) + | ((as21 & 0x000003) << 12)); +} + +static inline unsigned re_assemble_22(unsigned as22) +{ + return (( (as22 & 0x200000) >> 21) + | ((as22 & 0x1f0000) << (21 - 16)) + | ((as22 & 0x00f800) << (16 - 11)) + | ((as22 & 0x000400) >> (10 - 2)) + | ((as22 & 0x0003ff) << (1 + 2))); +} + +/* Various HPPA instruction formats. */ +/* see https://parisc.wiki.kernel.org/images-parisc/6/68/Pa11_acd.pdf, appendix C */ + +static inline u32 hppa_t1_insn(u8 opcode, u8 b, u8 r, s16 im14) +{ + return ((opcode << 26) | (b << 21) | (r << 16) | re_assemble_14(im14)); +} + +static inline u32 hppa_t5_insn(u8 opcode, u8 tr, u32 val21) +{ + return ((opcode << 26) | (tr << 21) | re_assemble_21(val21)); +} + +static inline u32 hppa_t6_insn(u8 opcode, u8 r2, u8 r1, u8 c, u8 f, u8 ext6, u16 t) +{ + return ((opcode << 26) | (r2 << 21) | (r1 << 16) | (c << 13) | (f << 12) | + (ext6 << 6) | t); +} + +/* 7. Arithmetic immediate */ +static inline u32 hppa_t7_insn(u8 opcode, u8 r, u8 t, u32 im11) +{ + return ((opcode << 26) | (r << 21) | (t << 16) | low_sign_unext(im11, 11)); +} + +/* 10. Shift instructions */ +static inline u32 hppa_t10_insn(u8 opcode, u8 r2, u8 r1, u8 c, u8 ext3, u8 cp, u8 t) +{ + return ((opcode << 26) | (r2 << 21) | (r1 << 16) | (c << 13) | + (ext3 << 10) | (cp << 5) | t); +} + +/* 11. Conditional branch instructions */ +static inline u32 hppa_t11_insn(u8 opcode, u8 r2, u8 r1, u8 c, u32 w, u8 nop) +{ + u32 ra = re_assemble_12(w); + // ra = low_sign_unext(w,11) | (w & (1<<10) + return ((opcode << 26) | (r2 << 21) | (r1 << 16) | (c << 13) | (nop << 1) | ra); +} + +/* 12. Branch instructions */ +static inline u32 hppa_t12_insn(u8 opcode, u8 rp, u32 w, u8 ext3, u8 nop) +{ + return ((opcode << 26) | (rp << 21) | (ext3 << 13) | (nop << 1) | re_assemble_17(w)); +} + +static inline u32 hppa_t12_L_insn(u8 opcode, u32 w, u8 nop) +{ + return ((opcode << 26) | (0x05 << 13) | (nop << 1) | re_assemble_22(w)); +} + +/* 21. Move to control register */ +static inline u32 hppa_t21_insn(u8 opcode, u8 r2, u8 r1, u8 ext8, u8 t) +{ + return ((opcode << 26) | (r2 << 21) | (r1 << 16) | (ext8 << 5) | t); +} + +/* Helper functions called by jit code on HPPA32 and HPPA64. */ + +u64 hppa_div64(u64 div, u64 divisor); +u64 hppa_div64_rem(u64 div, u64 divisor); + +/* Helper functions that emit HPPA instructions when possible. */ + +void bpf_jit_build_prologue(struct hppa_jit_context *ctx); +void bpf_jit_build_epilogue(struct hppa_jit_context *ctx); + +int bpf_jit_emit_insn(const struct bpf_insn *insn, struct hppa_jit_context *ctx, + bool extra_pass); + +#endif /* _BPF_JIT_H */ diff --git a/arch/parisc/net/bpf_jit_core.c b/arch/parisc/net/bpf_jit_core.c new file mode 100644 index 000000000000..bda8d1f7bec3 --- /dev/null +++ b/arch/parisc/net/bpf_jit_core.c @@ -0,0 +1,207 @@ +// SPDX-License-Identifier: GPL-2.0 +/* + * Common functionality for HPPA32 and HPPA64 BPF JIT compilers + * + * Copyright (c) 2023 Helge Deller + * + */ + +#include +#include +#include "bpf_jit.h" + +/* Number of iterations to try until offsets converge. */ +#define NR_JIT_ITERATIONS 35 + +static int build_body(struct hppa_jit_context *ctx, bool extra_pass, int *offset) +{ + const struct bpf_prog *prog = ctx->prog; + int i; + bool again = false; + + ctx->reg_seen_collect = true; + for (i = 0; i < prog->len; i++) { + const struct bpf_insn *insn = &prog->insnsi[i]; + int ret; + + ret = bpf_jit_emit_insn(insn, ctx, extra_pass); +// printk("instruction #%d -> RET %d\n", i, ret); + /* BPF_LD | BPF_IMM | BPF_DW: skip the next instruction. */ + if (ret > 0) + i++; + if (offset) { + if (offset[i] != ctx->ninsns) + again = 1; + offset[i] = ctx->ninsns; + } +// printk("INSTR %d offset at %d\n", i, offset ? offset[i]: 0); + if (ret < 0) + return ret; + } + ctx->reg_seen_collect = false; + return 0; +} + +bool bpf_jit_needs_zext(void) +{ + return true; +} + +struct bpf_prog *bpf_int_jit_compile(struct bpf_prog *prog) +{ + unsigned int prog_size = 0, extable_size = 0; + bool tmp_blinded = false, extra_pass = false; + struct bpf_prog *tmp, *orig_prog = prog; + int pass = 0, prev_ninsns = 0, prologue_len, i; + struct hppa_jit_data *jit_data; + struct hppa_jit_context *ctx; + + if (!prog->jit_requested) + return orig_prog; + + tmp = bpf_jit_blind_constants(prog); + if (IS_ERR(tmp)) + return orig_prog; + if (tmp != prog) { + tmp_blinded = true; + prog = tmp; + } + + jit_data = prog->aux->jit_data; + if (!jit_data) { + jit_data = kzalloc(sizeof(*jit_data), GFP_KERNEL); + if (!jit_data) { + prog = orig_prog; + goto out; + } + prog->aux->jit_data = jit_data; + } + + ctx = &jit_data->ctx; + + if (ctx->offset) { + extra_pass = true; + prog_size = sizeof(*ctx->insns) * ctx->ninsns; + goto skip_init_ctx; + } + + ctx->prog = prog; + ctx->offset = kcalloc(prog->len, sizeof(int), GFP_KERNEL); + if (!ctx->offset) { + prog = orig_prog; + goto out_offset; + } + for (i = 0; i < prog->len; i++) { + prev_ninsns += 20; + ctx->offset[i] = prev_ninsns; + } + + for (i = 0; i < NR_JIT_ITERATIONS; i++) { + pass++; + ctx->ninsns = 0; + if (build_body(ctx, extra_pass, ctx->offset)) { + prog = orig_prog; + goto out_offset; + } + ctx->body_len = ctx->ninsns; + bpf_jit_build_prologue(ctx); + ctx->prologue_len = ctx->ninsns - ctx->body_len; + ctx->epilogue_offset = ctx->ninsns; + bpf_jit_build_epilogue(ctx); + + if (ctx->ninsns == prev_ninsns) { + if (jit_data->header) + break; + /* obtain the actual image size */ + extable_size = prog->aux->num_exentries * + sizeof(struct exception_table_entry); + prog_size = sizeof(*ctx->insns) * ctx->ninsns; + + jit_data->header = + bpf_jit_binary_alloc(prog_size + extable_size, + &jit_data->image, + sizeof(u32), + bpf_fill_ill_insns); + if (!jit_data->header) { + prog = orig_prog; + goto out_offset; + } + + ctx->insns = (u32 *)jit_data->image; + /* + * Now, when the image is allocated, the image can + * potentially shrink more (auipc/jalr -> jal). + */ + } + prev_ninsns = ctx->ninsns; + } + + if (i == NR_JIT_ITERATIONS) { + pr_err("bpf-jit: image did not converge in <%d passes!\n", i); + if (jit_data->header) + bpf_jit_binary_free(jit_data->header); + prog = orig_prog; + goto out_offset; + } + + if (extable_size) + prog->aux->extable = (void *)ctx->insns + prog_size; + +skip_init_ctx: + pass++; + ctx->ninsns = 0; + + bpf_jit_build_prologue(ctx); + if (build_body(ctx, extra_pass, NULL)) { + bpf_jit_binary_free(jit_data->header); + prog = orig_prog; + goto out_offset; + } + bpf_jit_build_epilogue(ctx); + + if (HPPA_JIT_DEBUG || bpf_jit_enable > 1) { + if (HPPA_JIT_DUMP) + bpf_jit_dump(prog->len, prog_size, pass, ctx->insns); + if (HPPA_JIT_REBOOT) + { extern int machine_restart(char *); machine_restart(""); } + } + + prog->bpf_func = (void *)ctx->insns; + prog->jited = 1; + prog->jited_len = prog_size; + + bpf_flush_icache(jit_data->header, ctx->insns + ctx->ninsns); + + if (!prog->is_func || extra_pass) { + bpf_jit_binary_lock_ro(jit_data->header); + prologue_len = ctx->epilogue_offset - ctx->body_len; + for (i = 0; i < prog->len; i++) + ctx->offset[i] += prologue_len; + bpf_prog_fill_jited_linfo(prog, ctx->offset); +out_offset: + kfree(ctx->offset); + kfree(jit_data); + prog->aux->jit_data = NULL; + } +out: + if (HPPA_JIT_REBOOT) + { extern int machine_restart(char *); machine_restart(""); } + + if (tmp_blinded) + bpf_jit_prog_release_other(prog, prog == orig_prog ? + tmp : orig_prog); + return prog; +} + +u64 hppa_div64(u64 div, u64 divisor) +{ + div = div64_u64(div, divisor); + return div; +} + +u64 hppa_div64_rem(u64 div, u64 divisor) +{ + u64 rem; + div64_u64_rem(div, divisor, &rem); + return rem; +}