From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from e23smtp07.au.ibm.com (e23smtp07.au.ibm.com [202.81.31.140]) by dpdk.org (Postfix) with ESMTP id 714707FC9 for ; Sun, 23 Nov 2014 14:11:40 +0100 (CET) Received: from /spool/local by e23smtp07.au.ibm.com with IBM ESMTP SMTP Gateway: Authorized Use Only! Violators will be prosecuted for from ; Sun, 23 Nov 2014 23:22:22 +1000 Received: from d23dlp01.au.ibm.com (202.81.31.203) by e23smtp07.au.ibm.com (202.81.31.204) with IBM ESMTP SMTP Gateway: Authorized Use Only! Violators will be prosecuted; Sun, 23 Nov 2014 23:22:19 +1000 Received: from d23relay10.au.ibm.com (d23relay10.au.ibm.com [9.190.26.77]) by d23dlp01.au.ibm.com (Postfix) with ESMTP id D869D2CE8050 for ; Mon, 24 Nov 2014 00:22:18 +1100 (EST) Received: from d23av03.au.ibm.com (d23av03.au.ibm.com [9.190.234.97]) by d23relay10.au.ibm.com (8.14.9/8.14.9/NCO v10.0) with ESMTP id sANDMApZ21233754 for ; Mon, 24 Nov 2014 00:22:18 +1100 Received: from d23av03.au.ibm.com (localhost [127.0.0.1]) by d23av03.au.ibm.com (8.14.4/8.14.4/NCO v10.0 AVout) with ESMTP id sANDLk0q017477 for ; Mon, 24 Nov 2014 00:21:46 +1100 Received: from os_controller.crl.ibm.com ([9.186.57.97]) by d23av03.au.ibm.com (8.14.4/8.14.4/NCO v10.0 AVin) with ESMTP id sANDLYg3017172 for ; Mon, 24 Nov 2014 00:21:45 +1100 From: Chao Zhu To: dev@dpdk.org Date: Sun, 23 Nov 2014 20:22:20 -0500 Message-Id: <1416792142-23132-13-git-send-email-chaozhu@linux.vnet.ibm.com> X-Mailer: git-send-email 1.7.1 In-Reply-To: <1416792142-23132-1-git-send-email-chaozhu@linux.vnet.ibm.com> References: <1416792142-23132-1-git-send-email-chaozhu@linux.vnet.ibm.com> X-TM-AS-MML: disable X-Content-Scanned: Fidelis XPS MAILER x-cbid: 14112313-0025-0000-0000-0000008F8AB7 Subject: [dpdk-dev] [PATCH v3 12/14] Add eal memory support for IBM Power Architecture X-BeenThere: dev@dpdk.org X-Mailman-Version: 2.1.15 Precedence: list List-Id: patches and discussions about DPDK List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-List-Received-Date: Sun, 23 Nov 2014 13:11:41 -0000 The mmap of hugepage files on IBM Power starts from high address to low address. This is different from x86. This patch modified the memory segment detection code to get the correct memory segment layout on Power architecture. This patch also added a commond ARCH_PPC_64 defination for 64 bit systems. Signed-off-by: Chao Zhu --- config/defconfig_ppc_64-power8-linuxapp-gcc | 1 + config/defconfig_x86_64-native-linuxapp-clang | 1 + config/defconfig_x86_64-native-linuxapp-gcc | 1 + config/defconfig_x86_64-native-linuxapp-icc | 1 + lib/librte_eal/linuxapp/eal/eal_memory.c | 75 ++++++++++++++++++------- 5 files changed, 59 insertions(+), 20 deletions(-) diff --git a/config/defconfig_ppc_64-power8-linuxapp-gcc b/config/defconfig_ppc_64-power8-linuxapp-gcc index b10f60c..23a5591 100644 --- a/config/defconfig_ppc_64-power8-linuxapp-gcc +++ b/config/defconfig_ppc_64-power8-linuxapp-gcc @@ -35,6 +35,7 @@ CONFIG_RTE_MACHINE="power8" CONFIG_RTE_ARCH="ppc_64" CONFIG_RTE_ARCH_PPC_64=y CONFIG_RTE_ARCH_BIG_ENDIAN=y +CONFIG_RTE_ARCH_64=y CONFIG_RTE_TOOLCHAIN="gcc" CONFIG_RTE_TOOLCHAIN_GCC=y diff --git a/config/defconfig_x86_64-native-linuxapp-clang b/config/defconfig_x86_64-native-linuxapp-clang index bbda080..5f3074e 100644 --- a/config/defconfig_x86_64-native-linuxapp-clang +++ b/config/defconfig_x86_64-native-linuxapp-clang @@ -36,6 +36,7 @@ CONFIG_RTE_MACHINE="native" CONFIG_RTE_ARCH="x86_64" CONFIG_RTE_ARCH_X86_64=y +CONFIG_RTE_ARCH_64=y CONFIG_RTE_TOOLCHAIN="clang" CONFIG_RTE_TOOLCHAIN_CLANG=y diff --git a/config/defconfig_x86_64-native-linuxapp-gcc b/config/defconfig_x86_64-native-linuxapp-gcc index 3de818a..60baf5b 100644 --- a/config/defconfig_x86_64-native-linuxapp-gcc +++ b/config/defconfig_x86_64-native-linuxapp-gcc @@ -36,6 +36,7 @@ CONFIG_RTE_MACHINE="native" CONFIG_RTE_ARCH="x86_64" CONFIG_RTE_ARCH_X86_64=y +CONFIG_RTE_ARCH_64=y CONFIG_RTE_TOOLCHAIN="gcc" CONFIG_RTE_TOOLCHAIN_GCC=y diff --git a/config/defconfig_x86_64-native-linuxapp-icc b/config/defconfig_x86_64-native-linuxapp-icc index 795333b..71d1e28 100644 --- a/config/defconfig_x86_64-native-linuxapp-icc +++ b/config/defconfig_x86_64-native-linuxapp-icc @@ -36,6 +36,7 @@ CONFIG_RTE_MACHINE="native" CONFIG_RTE_ARCH="x86_64" CONFIG_RTE_ARCH_X86_64=y +CONFIG_RTE_ARCH_64=y CONFIG_RTE_TOOLCHAIN="icc" CONFIG_RTE_TOOLCHAIN_ICC=y diff --git a/lib/librte_eal/linuxapp/eal/eal_memory.c b/lib/librte_eal/linuxapp/eal/eal_memory.c index f2454f4..a8e7421 100644 --- a/lib/librte_eal/linuxapp/eal/eal_memory.c +++ b/lib/librte_eal/linuxapp/eal/eal_memory.c @@ -316,11 +316,11 @@ map_all_hugepages(struct hugepage_file *hugepg_tbl, #endif hugepg_tbl[i].filepath[sizeof(hugepg_tbl[i].filepath) - 1] = '\0'; } -#ifndef RTE_ARCH_X86_64 - /* for 32-bit systems, don't remap 1G pages, just reuse original +#ifndef RTE_ARCH_64 + /* for 32-bit systems, don't remap 1G and 16G pages, just reuse original * map address as final map address. */ - else if (hugepage_sz == RTE_PGSIZE_1G){ + else if ((hugepage_sz == RTE_PGSIZE_1G) || (hugepage_sz == RTE_PGSIZE_16G)){ hugepg_tbl[i].final_va = hugepg_tbl[i].orig_va; hugepg_tbl[i].orig_va = NULL; continue; @@ -335,9 +335,16 @@ map_all_hugepages(struct hugepage_file *hugepg_tbl, * physical block: count the number of * contiguous physical pages. */ for (j = i+1; j < hpi->num_pages[0] ; j++) { +#ifdef RTE_ARCH_PPC_64 + /* The physical addresses are sorted in descending order on PPC64 */ + if (hugepg_tbl[j].physaddr != + hugepg_tbl[j-1].physaddr - hugepage_sz) + break; +#else if (hugepg_tbl[j].physaddr != hugepg_tbl[j-1].physaddr + hugepage_sz) break; +#endif } num_pages = j - i; vma_len = num_pages * hugepage_sz; @@ -412,11 +419,11 @@ remap_all_hugepages(struct hugepage_file *hugepg_tbl, struct hugepage_info *hpi) while (i < hpi->num_pages[0]) { -#ifndef RTE_ARCH_X86_64 - /* for 32-bit systems, don't remap 1G pages, just reuse original +#ifndef RTE_ARCH_64 + /* for 32-bit systems, don't remap 1G pages and 16G pages, just reuse original * map address as final map address. */ - if (hugepage_sz == RTE_PGSIZE_1G){ + if ((hugepage_sz == RTE_PGSIZE_1G) || (hugepage_sz == RTE_PGSIZE_16G)){ hugepg_tbl[i].final_va = hugepg_tbl[i].orig_va; hugepg_tbl[i].orig_va = NULL; i++; @@ -428,9 +435,15 @@ remap_all_hugepages(struct hugepage_file *hugepg_tbl, struct hugepage_info *hpi) * physical block: count the number of * contiguous physical pages. */ for (j = i+1; j < hpi->num_pages[0] ; j++) { +#ifdef RTE_ARCH_PPC_64 + /* The physical addresses are sorted in descending order on PPC64 */ + if (hugepg_tbl[j].physaddr != hugepg_tbl[j-1].physaddr - hugepage_sz) + break; +#else if (hugepg_tbl[j].physaddr != hugepg_tbl[j-1].physaddr + hugepage_sz) break; - } +#endif + } num_pages = j - i; vma_len = num_pages * hugepage_sz; @@ -652,21 +665,21 @@ error: } /* - * Sort the hugepg_tbl by physical address (lower addresses first). We - * use a slow algorithm, but we won't have millions of pages, and this + * Sort the hugepg_tbl by physical address (lower addresses first on x86, higher address first + * on powerpc). We use a slow algorithm, but we won't have millions of pages, and this * is only done at init time. */ static int sort_by_physaddr(struct hugepage_file *hugepg_tbl, struct hugepage_info *hpi) { unsigned i, j; - int smallest_idx; - uint64_t smallest_addr; + int compare_idx; + uint64_t compare_addr; struct hugepage_file tmp; for (i = 0; i < hpi->num_pages[0]; i++) { - smallest_addr = 0; - smallest_idx = -1; + compare_addr = 0; + compare_idx = -1; /* * browse all entries starting at 'i', and find the @@ -674,22 +687,26 @@ sort_by_physaddr(struct hugepage_file *hugepg_tbl, struct hugepage_info *hpi) */ for (j=i; j< hpi->num_pages[0]; j++) { - if (smallest_addr == 0 || - hugepg_tbl[j].physaddr < smallest_addr) { - smallest_addr = hugepg_tbl[j].physaddr; - smallest_idx = j; + if (compare_addr == 0 || +#ifdef RTE_ARCH_PPC_64 + hugepg_tbl[j].physaddr > compare_addr) { +#else + hugepg_tbl[j].physaddr < compare_addr) { +#endif + compare_addr = hugepg_tbl[j].physaddr; + compare_idx = j; } } /* should not happen */ - if (smallest_idx == -1) { + if (compare_idx == -1) { RTE_LOG(ERR, EAL, "%s(): error in physaddr sorting\n", __func__); return -1; } /* swap the 2 entries in the table */ - memcpy(&tmp, &hugepg_tbl[smallest_idx], sizeof(struct hugepage_file)); - memcpy(&hugepg_tbl[smallest_idx], &hugepg_tbl[i], + memcpy(&tmp, &hugepg_tbl[compare_idx], sizeof(struct hugepage_file)); + memcpy(&hugepg_tbl[compare_idx], &hugepg_tbl[i], sizeof(struct hugepage_file)); memcpy(&hugepg_tbl[i], &tmp, sizeof(struct hugepage_file)); } @@ -1260,12 +1277,24 @@ rte_eal_hugepage_init(void) new_memseg = 1; else if (hugepage[i].size != hugepage[i-1].size) new_memseg = 1; + +#ifdef RTE_ARCH_PPC_64 + /* On PPC64 architecture, the mmap always start from higher virtual address to lower address. + * Here, both the physical address and virtual address are in descending order */ + else if ((hugepage[i-1].physaddr - hugepage[i].physaddr) != + hugepage[i].size) + new_memseg = 1; + else if (((unsigned long)hugepage[i-1].final_va - + (unsigned long)hugepage[i].final_va) != hugepage[i].size) + new_memseg = 1; +#else else if ((hugepage[i].physaddr - hugepage[i-1].physaddr) != hugepage[i].size) new_memseg = 1; else if (((unsigned long)hugepage[i].final_va - (unsigned long)hugepage[i-1].final_va) != hugepage[i].size) new_memseg = 1; +#endif if (new_memseg) { j += 1; @@ -1284,6 +1313,12 @@ rte_eal_hugepage_init(void) } /* continuation of previous memseg */ else { +#ifdef RTE_ARCH_PPC_64 + /* Use the phy and virt address of the last page as segment address + * for IBM Power architecture */ + mcfg->memseg[j].phys_addr = hugepage[i].physaddr; + mcfg->memseg[j].addr = hugepage[i].final_va; +#endif mcfg->memseg[j].len += mcfg->memseg[j].hugepage_sz; } hugepage[i].memseg_id = j; -- 1.7.1