// SPDX-License-Identifier: (GPL-2.0-only OR BSD-2-Clause) /* Copyright (C) 2015-2018 Netronome Systems, Inc. */ /* * nfp_target.c * CPP Access Width Decoder * Authors: Jakub Kicinski * Jason McMullan * Francois H. Theron */ #define pr_fmt(fmt) "NFP target: " fmt #include #include #include #include "nfp_cpp.h" #include "nfp6000/nfp6000.h" #define P32 1 #define P64 2 /* This structure ONLY includes items that can be done with a read or write of * 32-bit or 64-bit words. All others are not listed. */ #define AT(_action, _token, _pull, _push) \ case NFP_CPP_ID(0, (_action), (_token)): \ return PUSHPULL((_pull), (_push)) static int target_rw(u32 cpp_id, int pp, int start, int len) { switch (cpp_id & NFP_CPP_ID(0, ~0, ~0)) { AT(0, 0, 0, pp); AT(1, 0, pp, 0); AT(NFP_CPP_ACTION_RW, 0, pp, pp); default: return -EINVAL; } } static int nfp6000_nbi_dma(u32 cpp_id) { switch (cpp_id & NFP_CPP_ID(0, ~0, ~0)) { AT(0, 0, 0, P64); /* ReadNbiDma */ AT(1, 0, P64, 0); /* WriteNbiDma */ AT(NFP_CPP_ACTION_RW, 0, P64, P64); default: return -EINVAL; } } static int nfp6000_nbi_stats(u32 cpp_id) { switch (cpp_id & NFP_CPP_ID(0, ~0, ~0)) { AT(0, 0, 0, P32); /* ReadNbiStats */ AT(1, 0, P32, 0); /* WriteNbiStats */ AT(NFP_CPP_ACTION_RW, 0, P32, P32); default: return -EINVAL; } } static int nfp6000_nbi_tm(u32 cpp_id) { switch (cpp_id & NFP_CPP_ID(0, ~0, ~0)) { AT(0, 0, 0, P64); /* ReadNbiTM */ AT(1, 0, P64, 0); /* WriteNbiTM */ AT(NFP_CPP_ACTION_RW, 0, P64, P64); default: return -EINVAL; } } static int nfp6000_nbi_ppc(u32 cpp_id) { switch (cpp_id & NFP_CPP_ID(0, ~0, ~0)) { AT(0, 0, 0, P64); /* ReadNbiPreclassifier */ AT(1, 0, P64, 0); /* WriteNbiPreclassifier */ AT(NFP_CPP_ACTION_RW, 0, P64, P64); default: return -EINVAL; } } static int nfp6000_nbi(u32 cpp_id, u64 address) { u64 rel_addr = address & 0x3fFFFF; if (rel_addr < (1 << 20)) return nfp6000_nbi_dma(cpp_id); if (rel_addr < (2 << 20)) return nfp6000_nbi_stats(cpp_id); if (rel_addr < (3 << 20)) return nfp6000_nbi_tm(cpp_id); return nfp6000_nbi_ppc(cpp_id); } /* This structure ONLY includes items that can be done with a read or write of * 32-bit or 64-bit words. All others are not listed. */ static int nfp6000_mu_common(u32 cpp_id) { switch (cpp_id & NFP_CPP_ID(0, ~0, ~0)) { AT(NFP_CPP_ACTION_RW, 0, P64, P64); /* read_be/write_be */ AT(NFP_CPP_ACTION_RW, 1, P64, P64); /* read_le/write_le */ AT(NFP_CPP_ACTION_RW, 2, P64, P64); /* read_swap_be/write_swap_be */ AT(NFP_CPP_ACTION_RW, 3, P64, P64); /* read_swap_le/write_swap_le */ AT(0, 0, 0, P64); /* read_be */ AT(0, 1, 0, P64); /* read_le */ AT(0, 2, 0, P64); /* read_swap_be */ AT(0, 3, 0, P64); /* read_swap_le */ AT(1, 0, P64, 0); /* write_be */ AT(1, 1, P64, 0); /* write_le */ AT(1, 2, P64, 0); /* write_swap_be */ AT(1, 3, P64, 0); /* write_swap_le */ AT(3, 0, 0, P32); /* atomic_read */ AT(3, 2, P32, 0); /* mask_compare_write */ AT(4, 0, P32, 0); /* atomic_write */ AT(4, 2, 0, 0); /* atomic_write_imm */ AT(4, 3, 0, P32); /* swap_imm */ AT(5, 0, P32, 0); /* set */ AT(5, 3, 0, P32); /* test_set_imm */ AT(6, 0, P32, 0); /* clr */ AT(6, 3, 0, P32); /* test_clr_imm */ AT(7, 0, P32, 0); /* add */ AT(7, 3, 0, P32); /* test_add_imm */ AT(8, 0, P32, 0); /* addsat */ AT(8, 3, 0, P32); /* test_subsat_imm */ AT(9, 0, P32, 0); /* sub */ AT(9, 3, 0, P32); /* test_sub_imm */ AT(10, 0, P32, 0); /* subsat */ AT(10, 3, 0, P32); /* test_subsat_imm */ AT(13, 0, 0, P32); /* microq128_get */ AT(13, 1, 0, P32); /* microq128_pop */ AT(13, 2, P32, 0); /* microq128_put */ AT(15, 0, P32, 0); /* xor */ AT(15, 3, 0, P32); /* test_xor_imm */ AT(28, 0, 0, P32); /* read32_be */ AT(28, 1, 0, P32); /* read32_le */ AT(28, 2, 0, P32); /* read32_swap_be */ AT(28, 3, 0, P32); /* read32_swap_le */ AT(31, 0, P32, 0); /* write32_be */ AT(31, 1, P32, 0); /* write32_le */ AT(31, 2, P32, 0); /* write32_swap_be */ AT(31, 3, P32, 0); /* write32_swap_le */ default: return -EINVAL; } } static int nfp6000_mu_ctm(u32 cpp_id) { switch (cpp_id & NFP_CPP_ID(0, ~0, ~0)) { AT(16, 1, 0, P32); /* packet_read_packet_status */ AT(17, 1, 0, P32); /* packet_credit_get */ AT(17, 3, 0, P64); /* packet_add_thread */ AT(18, 2, 0, P64); /* packet_free_and_return_pointer */ AT(18, 3, 0, P64); /* packet_return_pointer */ AT(21, 0, 0, P64); /* pe_dma_to_memory_indirect */ AT(21, 1, 0, P64); /* pe_dma_to_memory_indirect_swap */ AT(21, 2, 0, P64); /* pe_dma_to_memory_indirect_free */ AT(21, 3, 0, P64); /* pe_dma_to_memory_indirect_free_swap */ default: return nfp6000_mu_common(cpp_id); } } static int nfp6000_mu_emu(u32 cpp_id) { switch (cpp_id & NFP_CPP_ID(0, ~0, ~0)) { AT(18, 0, 0, P32); /* read_queue */ AT(18, 1, 0, P32); /* read_queue_ring */ AT(18, 2, P32, 0); /* write_queue */ AT(18, 3, P32, 0); /* write_queue_ring */ AT(20, 2, P32, 0); /* journal */ AT(21, 0, 0, P32); /* get */ AT(21, 1, 0, P32); /* get_eop */ AT(21, 2, 0, P32); /* get_freely */ AT(22, 0, 0, P32); /* pop */ AT(22, 1, 0, P32); /* pop_eop */ AT(22, 2, 0, P32); /* pop_freely */ default: return nfp6000_mu_common(cpp_id); } } static int nfp6000_mu_imu(u32 cpp_id) { return nfp6000_mu_common(cpp_id); } static int nfp6000_mu(u32 cpp_id, u64 address) { int pp; if (address < 0x2000000000ULL) pp = nfp6000_mu_ctm(cpp_id); else if (address < 0x8000000000ULL) pp = nfp6000_mu_emu(cpp_id); else if (address < 0x9800000000ULL) pp = nfp6000_mu_ctm(cpp_id); else if (address < 0x9C00000000ULL) pp = nfp6000_mu_emu(cpp_id); else if (address < 0xA000000000ULL) pp = nfp6000_mu_imu(cpp_id); else pp = nfp6000_mu_ctm(cpp_id); return pp; } static int nfp6000_ila(u32 cpp_id) { switch (cpp_id & NFP_CPP_ID(0, ~0, ~0)) { AT(0, 1, 0, P32); /* read_check_error */ AT(2, 0, 0, P32); /* read_int */ AT(3, 0, P32, 0); /* write_int */ default: return target_rw(cpp_id, P32, 48, 4); } } static int nfp6000_pci(u32 cpp_id) { switch (cpp_id & NFP_CPP_ID(0, ~0, ~0)) { AT(2, 0, 0, P32); AT(3, 0, P32, 0); default: return target_rw(cpp_id, P32, 4, 4); } } static int nfp6000_crypto(u32 cpp_id) { switch (cpp_id & NFP_CPP_ID(0, ~0, ~0)) { AT(2, 0, P64, 0); default: return target_rw(cpp_id, P64, 12, 4); } } static int nfp6000_cap_xpb(u32 cpp_id) { switch (cpp_id & NFP_CPP_ID(0, ~0, ~0)) { AT(0, 1, 0, P32); /* RingGet */ AT(0, 2, P32, 0); /* Interthread Signal */ AT(1, 1, P32, 0); /* RingPut */ AT(1, 2, P32, 0); /* CTNNWr */ AT(2, 0, 0, P32); /* ReflectRd, signal none */ AT(2, 1, 0, P32); /* ReflectRd, signal self */ AT(2, 2, 0, P32); /* ReflectRd, signal remote */ AT(2, 3, 0, P32); /* ReflectRd, signal both */ AT(3, 0, P32, 0); /* ReflectWr, signal none */ AT(3, 1, P32, 0); /* ReflectWr, signal self */ AT(3, 2, P32, 0); /* ReflectWr, signal remote */ AT(3, 3, P32, 0); /* ReflectWr, signal both */ AT(NFP_CPP_ACTION_RW, 1, P32, P32); default: return target_rw(cpp_id, P32, 1, 63); } } static int nfp6000_cls(u32 cpp_id) { switch (cpp_id & NFP_CPP_ID(0, ~0, ~0)) { AT(0, 3, P32, 0); /* xor */ AT(2, 0, P32, 0); /* set */ AT(2, 1, P32, 0); /* clr */ AT(4, 0, P32, 0); /* add */ AT(4, 1, P32, 0); /* add64 */ AT(6, 0, P32, 0); /* sub */ AT(6, 1, P32, 0); /* sub64 */ AT(6, 2, P32, 0); /* subsat */ AT(8, 2, P32, 0); /* hash_mask */ AT(8, 3, P32, 0); /* hash_clear */ AT(9, 0, 0, P32); /* ring_get */ AT(9, 1, 0, P32); /* ring_pop */ AT(9, 2, 0, P32); /* ring_get_freely */ AT(9, 3, 0, P32); /* ring_pop_freely */ AT(10, 0, P32, 0); /* ring_put */ AT(10, 2, P32, 0); /* ring_journal */ AT(14, 0, P32, 0); /* reflect_write_sig_local */ AT(15, 1, 0, P32); /* reflect_read_sig_local */ AT(17, 2, P32, 0); /* statisic */ AT(24, 0, 0, P32); /* ring_read */ AT(24, 1, P32, 0); /* ring_write */ AT(25, 0, 0, P32); /* ring_workq_add_thread */ AT(25, 1, P32, 0); /* ring_workq_add_work */ default: return target_rw(cpp_id, P32, 0, 64); } } int nfp_target_pushpull(u32 cpp_id, u64 address) { switch (NFP_CPP_ID_TARGET_of(cpp_id)) { case NFP_CPP_TARGET_NBI: return nfp6000_nbi(cpp_id, address); case NFP_CPP_TARGET_QDR: return target_rw(cpp_id, P32, 24, 4); case NFP_CPP_TARGET_ILA: return nfp6000_ila(cpp_id); case NFP_CPP_TARGET_MU: return nfp6000_mu(cpp_id, address); case NFP_CPP_TARGET_PCIE: return nfp6000_pci(cpp_id); case NFP_CPP_TARGET_ARM: if (address < 0x10000) return target_rw(cpp_id, P64, 1, 1); else return target_rw(cpp_id, P32, 1, 1); case NFP_CPP_TARGET_CRYPTO: return nfp6000_crypto(cpp_id); case NFP_CPP_TARGET_CT_XPB: return nfp6000_cap_xpb(cpp_id); case NFP_CPP_TARGET_CLS: return nfp6000_cls(cpp_id); case 0: return target_rw(cpp_id, P32, 4, 4); default: return -EINVAL; } } #undef AT #undef P32 #undef P64 /* All magic NFP-6xxx IMB 'mode' numbers here are from: * Databook (1 August 2013) * - System Overview and Connectivity * -- Internal Connectivity * --- Distributed Switch Fabric - Command Push/Pull (DSF-CPP) Bus * ---- CPP addressing * ----- Table 3.6. CPP Address Translation Mode Commands */ #define _NIC_NFP6000_MU_LOCALITY_DIRECT 2 static int nfp_decode_basic(u64 addr, int *dest_island, int cpp_tgt, int mode, bool addr40, int isld1, int isld0) { int iid_lsb, idx_lsb; /* This function doesn't handle MU or CTXBP */ if (cpp_tgt == NFP_CPP_TARGET_MU || cpp_tgt == NFP_CPP_TARGET_CT_XPB) return -EINVAL; switch (mode) { case 0: /* For VQDR, in this mode for 32-bit addressing * it would be islands 0, 16, 32 and 48 depending on channel * and upper address bits. * Since those are not all valid islands, most decode * cases would result in bad island IDs, but we do them * anyway since this is decoding an address that is already * assumed to be used as-is to get to sram. */ iid_lsb = addr40 ? 34 : 26; *dest_island = (addr >> iid_lsb) & 0x3F; return 0; case 1: /* For VQDR 32-bit, this would decode as: * Channel 0: island#0 * Channel 1: island#0 * Channel 2: island#1 * Channel 3: island#1 * That would be valid as long as both islands * have VQDR. Let's allow this. */ idx_lsb = addr40 ? 39 : 31; if (addr & BIT_ULL(idx_lsb)) *dest_island = isld1; else *dest_island = isld0; return 0; case 2: /* For VQDR 32-bit: * Channel 0: (island#0 | 0) * Channel 1: (island#0 | 1) * Channel 2: (island#1 | 0) * Channel 3: (island#1 | 1) * * Make sure we compare against isldN values * by clearing the LSB. * This is what the silicon does. */ isld0 &= ~1; isld1 &= ~1; idx_lsb = addr40 ? 39 : 31; iid_lsb = idx_lsb - 1; if (addr & BIT_ULL(idx_lsb)) *dest_island = isld1 | (int)((addr >> iid_lsb) & 1); else *dest_island = isld0 | (int)((addr >> iid_lsb) & 1); return 0; case 3: /* In this mode the data address starts to affect the island ID * so rather not allow it. In some really specific case * one could use this to send the upper half of the * VQDR channel to another MU, but this is getting very * specific. * However, as above for mode 0, this is the decoder * and the caller should validate the resulting IID. * This blindly does what the silicon would do. */ isld0 &= ~3; isld1 &= ~3; idx_lsb = addr40 ? 39 : 31; iid_lsb = idx_lsb - 2; if (addr & BIT_ULL(idx_lsb)) *dest_island = isld1 | (int)((addr >> iid_lsb) & 3); else *dest_island = isld0 | (int)((addr >> iid_lsb) & 3); return 0; default: return -EINVAL; } } static int nfp_encode_basic_qdr(u64 addr, int dest_island, int cpp_tgt, int mode, bool addr40, int isld1, int isld0) { int v, ret; /* Full Island ID and channel bits overlap? */ ret = nfp_decode_basic(addr, &v, cpp_tgt, mode, addr40, isld1, isld0); if (ret) return ret; /* The current address won't go where expected? */ if (dest_island != -1 && dest_island != v) return -EINVAL; /* If dest_island was -1, we don't care where it goes. */ return 0; } /* Try each option, take first one that fits. * Not sure if we would want to do some smarter * searching and prefer 0 or non-0 island IDs. */ static int nfp_encode_basic_search(u64 *addr, int dest_island, int *isld, int iid_lsb, int idx_lsb, int v_max) { int i, v; for (i = 0; i < 2; i++) for (v = 0; v < v_max; v++) { if (dest_island != (isld[i] | v)) continue; *addr &= ~GENMASK_ULL(idx_lsb, iid_lsb); *addr |= ((u64)i << idx_lsb); *addr |= ((u64)v << iid_lsb); return 0; } return -ENODEV; } /* For VQDR, we may not modify the Channel bits, which might overlap * with the Index bit. When it does, we need to ensure that isld0 == isld1. */ static int nfp_encode_basic(u64 *addr, int dest_island, int cpp_tgt, int mode, bool addr40, int isld1, int isld0) { int iid_lsb, idx_lsb; int isld[2]; u64 v64; isld[0] = isld0; isld[1] = isld1; /* This function doesn't handle MU or CTXBP */ if (cpp_tgt == NFP_CPP_TARGET_MU || cpp_tgt == NFP_CPP_TARGET_CT_XPB) return -EINVAL; switch (mode) { case 0: if (cpp_tgt == NFP_CPP_TARGET_QDR && !addr40) /* In this specific mode we'd rather not modify * the address but we can verify if the existing * contents will point to a valid island. */ return nfp_encode_basic_qdr(*addr, cpp_tgt, dest_island, mode, addr40, isld1, isld0); iid_lsb = addr40 ? 34 : 26; /* <39:34> or <31:26> */ v64 = GENMASK_ULL(iid_lsb + 5, iid_lsb); *addr &= ~v64; *addr |= ((u64)dest_island << iid_lsb) & v64; return 0; case 1: if (cpp_tgt == NFP_CPP_TARGET_QDR && !addr40) return nfp_encode_basic_qdr(*addr, cpp_tgt, dest_island, mode, addr40, isld1, isld0); idx_lsb = addr40 ? 39 : 31; if (dest_island == isld0) { /* Only need to clear the Index bit */ *addr &= ~BIT_ULL(idx_lsb); return 0; } if (dest_island == isld1) { /* Only need to set the Index bit */ *addr |= BIT_ULL(idx_lsb); return 0; } return -ENODEV; case 2: /* iid<0> = addr<30> = channel<0> * channel<1> = addr<31> = Index */ if (cpp_tgt == NFP_CPP_TARGET_QDR && !addr40) /* Special case where we allow channel bits to * be set before hand and with them select an island. * So we need to confirm that it's at least plausible. */ return nfp_encode_basic_qdr(*addr, cpp_tgt, dest_island, mode, addr40, isld1, isld0); /* Make sure we compare against isldN values * by clearing the LSB. * This is what the silicon does. */ isld[0] &= ~1; isld[1] &= ~1; idx_lsb = addr40 ? 39 : 31; iid_lsb = idx_lsb - 1; return nfp_encode_basic_search(addr, dest_island, isld, iid_lsb, idx_lsb, 2); case 3: if (cpp_tgt == NFP_CPP_TARGET_QDR && !addr40) /* iid<0> = addr<29> = data * iid<1> = addr<30> = channel<0> * channel<1> = addr<31> = Index */ return nfp_encode_basic_qdr(*addr, cpp_tgt, dest_island, mode, addr40, isld1, isld0); isld[0] &= ~3; isld[1] &= ~3; idx_lsb = addr40 ? 39 : 31; iid_lsb = idx_lsb - 2; return nfp_encode_basic_search(addr, dest_island, isld, iid_lsb, idx_lsb, 4); default: return -EINVAL; } } static int nfp_encode_mu(u64 *addr, int dest_island, int mode, bool addr40, int isld1, int isld0) { int iid_lsb, idx_lsb, locality_lsb; int isld[2]; u64 v64; int da; isld[0] = isld0; isld[1] = isld1; locality_lsb = nfp_cppat_mu_locality_lsb(mode, addr40); if (((*addr >> locality_lsb) & 3) == _NIC_NFP6000_MU_LOCALITY_DIRECT) da = 1; else da = 0; switch (mode) { case 0: iid_lsb = addr40 ? 32 : 24; v64 = GENMASK_ULL(iid_lsb + 5, iid_lsb); *addr &= ~v64; *addr |= (((u64)dest_island) << iid_lsb) & v64; return 0; case 1: if (da) { iid_lsb = addr40 ? 32 : 24; v64 = GENMASK_ULL(iid_lsb + 5, iid_lsb); *addr &= ~v64; *addr |= (((u64)dest_island) << iid_lsb) & v64; return 0; } idx_lsb = addr40 ? 37 : 29; if (dest_island == isld0) { *addr &= ~BIT_ULL(idx_lsb); return 0; } if (dest_island == isld1) { *addr |= BIT_ULL(idx_lsb); return 0; } return -ENODEV; case 2: if (da) { iid_lsb = addr40 ? 32 : 24; v64 = GENMASK_ULL(iid_lsb + 5, iid_lsb); *addr &= ~v64; *addr |= (((u64)dest_island) << iid_lsb) & v64; return 0; } /* Make sure we compare against isldN values * by clearing the LSB. * This is what the silicon does. */ isld[0] &= ~1; isld[1] &= ~1; idx_lsb = addr40 ? 37 : 29; iid_lsb = idx_lsb - 1; return nfp_encode_basic_search(addr, dest_island, isld, iid_lsb, idx_lsb, 2); case 3: /* Only the EMU will use 40 bit addressing. Silently * set the direct locality bit for everyone else. * The SDK toolchain uses dest_island <= 0 to test * for atypical address encodings to support access * to local-island CTM with a 32-but address (high-locality * is effewctively ignored and just used for * routing to island #0). */ if (dest_island > 0 && (dest_island < 24 || dest_island > 26)) { *addr |= ((u64)_NIC_NFP6000_MU_LOCALITY_DIRECT) << locality_lsb; da = 1; } if (da) { iid_lsb = addr40 ? 32 : 24; v64 = GENMASK_ULL(iid_lsb + 5, iid_lsb); *addr &= ~v64; *addr |= (((u64)dest_island) << iid_lsb) & v64; return 0; } isld[0] &= ~3; isld[1] &= ~3; idx_lsb = addr40 ? 37 : 29; iid_lsb = idx_lsb - 2; return nfp_encode_basic_search(addr, dest_island, isld, iid_lsb, idx_lsb, 4); default: return -EINVAL; } } static int nfp_cppat_addr_encode(u64 *addr, int dest_island, int cpp_tgt, int mode, bool addr40, int isld1, int isld0) { switch (cpp_tgt) { case NFP_CPP_TARGET_NBI: case NFP_CPP_TARGET_QDR: case NFP_CPP_TARGET_ILA: case NFP_CPP_TARGET_PCIE: case NFP_CPP_TARGET_ARM: case NFP_CPP_TARGET_CRYPTO: case NFP_CPP_TARGET_CLS: return nfp_encode_basic(addr, dest_island, cpp_tgt, mode, addr40, isld1, isld0); case NFP_CPP_TARGET_MU: return nfp_encode_mu(addr, dest_island, mode, addr40, isld1, isld0); case NFP_CPP_TARGET_CT_XPB: if (mode != 1 || addr40) return -EINVAL; *addr &= ~GENMASK_ULL(29, 24); *addr |= ((u64)dest_island << 24) & GENMASK_ULL(29, 24); return 0; default: return -EINVAL; } } int nfp_target_cpp(u32 cpp_island_id, u64 cpp_island_address, u32 *cpp_target_id, u64 *cpp_target_address, const u32 *imb_table) { const int island = NFP_CPP_ID_ISLAND_of(cpp_island_id); const int target = NFP_CPP_ID_TARGET_of(cpp_island_id); u32 imb; int err; if (target < 0 || target >= 16) { pr_err("Invalid CPP target: %d\n", target); return -EINVAL; } if (island == 0) { /* Already translated */ *cpp_target_id = cpp_island_id; *cpp_target_address = cpp_island_address; return 0; } /* CPP + Island only allowed on systems with IMB tables */ if (!imb_table) return -EINVAL; imb = imb_table[target]; *cpp_target_address = cpp_island_address; err = nfp_cppat_addr_encode(cpp_target_address, island, target, ((imb >> 13) & 7), ((imb >> 12) & 1), ((imb >> 6) & 0x3f), ((imb >> 0) & 0x3f)); if (err) { pr_err("Can't encode CPP address: %d\n", err); return err; } *cpp_target_id = NFP_CPP_ID(target, NFP_CPP_ID_ACTION_of(cpp_island_id), NFP_CPP_ID_TOKEN_of(cpp_island_id)); return 0; }