From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from mga09.intel.com (mga09.intel.com [134.134.136.24]) by dpdk.org (Postfix) with ESMTP id 25AEB7F2C for ; Tue, 11 Nov 2014 10:59:45 +0100 (CET) Received: from orsmga002.jf.intel.com ([10.7.209.21]) by orsmga102.jf.intel.com with ESMTP; 11 Nov 2014 02:07:42 -0800 X-ExtLoop1: 1 X-IronPort-AV: E=Sophos;i="5.07,360,1413270000"; d="scan'208";a="634968459" Received: from irvmail001.ir.intel.com ([163.33.26.43]) by orsmga002.jf.intel.com with ESMTP; 11 Nov 2014 02:09:33 -0800 Received: from sivswdev02.ir.intel.com (sivswdev02.ir.intel.com [10.237.217.46]) by irvmail001.ir.intel.com (8.14.3/8.13.6/MailSET/Hub) with ESMTP id sABA9WkY009594; Tue, 11 Nov 2014 10:09:32 GMT Received: from sivswdev02.ir.intel.com (localhost [127.0.0.1]) by sivswdev02.ir.intel.com with ESMTP id sABA9Vme019223; Tue, 11 Nov 2014 10:09:32 GMT Received: (from aburakov@localhost) by sivswdev02.ir.intel.com with id sABA9P6j019188; Tue, 11 Nov 2014 10:09:25 GMT From: Anatoly Burakov To: dev@dpdk.org Date: Tue, 11 Nov 2014 10:09:25 +0000 Message-Id: <1415700565-19157-1-git-send-email-anatoly.burakov@intel.com> X-Mailer: git-send-email 1.7.4.1 In-Reply-To: <1415619272-8281-1-git-send-email-anatoly.burakov@intel.com> References: <1415619272-8281-1-git-send-email-anatoly.burakov@intel.com> Subject: [dpdk-dev] [PATCH v8] eal: map PCI memory resources after hugepages 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: Tue, 11 Nov 2014 09:59:47 -0000 Multi-process DPDK application must mmap hugepages and PCI resources into the same virtual address space. By default the virtual addresses are chosen by the primary process automatically when calling the mmap. But sometimes the chosen virtual addresses aren't usable in secondary process - for example, secondary process is linked with more libraries than primary process, and the library occupies the same address space that the primary process has requested for PCI mappings. This patch makes EAL try and map PCI BARs right after the hugepages (instead of location chosen by mmap) in virtual memory, so that PCI BARs have less chance of ending up in random places in virtual memory. Signed-off-by: Liang Xu Signed-off-by: Anatoly Burakov --- lib/librte_eal/linuxapp/eal/eal_pci.c | 30 ++++++++++++++++------ lib/librte_eal/linuxapp/eal/eal_pci_uio.c | 13 ++++++++-- lib/librte_eal/linuxapp/eal/eal_pci_vfio.c | 19 +++++++++++--- lib/librte_eal/linuxapp/eal/include/eal_pci_init.h | 6 +++++ 4 files changed, 55 insertions(+), 13 deletions(-) diff --git a/lib/librte_eal/linuxapp/eal/eal_pci.c b/lib/librte_eal/linuxapp/eal/eal_pci.c index 5fe3961..79fbbb8 100644 --- a/lib/librte_eal/linuxapp/eal/eal_pci.c +++ b/lib/librte_eal/linuxapp/eal/eal_pci.c @@ -97,6 +97,25 @@ error: return -1; } +void * +pci_find_max_end_va(void) +{ + const struct rte_memseg *seg = rte_eal_get_physmem_layout(); + const struct rte_memseg *last = seg; + unsigned i = 0; + + for (i = 0; i < RTE_MAX_MEMSEG; i++, seg++) { + if (seg->addr == NULL) + break; + + if (seg->addr > last->addr) + last = seg; + + } + return RTE_PTR_ADD(last->addr, last->len); +} + + /* map a particular resource from a file */ void * pci_map_resource(void *requested_addr, int fd, off_t offset, size_t size) @@ -106,21 +125,16 @@ pci_map_resource(void *requested_addr, int fd, off_t offset, size_t size) /* Map the PCI memory resource of device */ mapaddr = mmap(requested_addr, size, PROT_READ | PROT_WRITE, MAP_SHARED, fd, offset); - if (mapaddr == MAP_FAILED || - (requested_addr != NULL && mapaddr != requested_addr)) { + if (mapaddr == MAP_FAILED) { RTE_LOG(ERR, EAL, "%s(): cannot mmap(%d, %p, 0x%lx, 0x%lx): %s (%p)\n", __func__, fd, requested_addr, (unsigned long)size, (unsigned long)offset, strerror(errno), mapaddr); - goto fail; + } else { + RTE_LOG(DEBUG, EAL, " PCI memory mapped at %p\n", mapaddr); } - RTE_LOG(DEBUG, EAL, " PCI memory mapped at %p\n", mapaddr); - return mapaddr; - -fail: - return NULL; } /* parse the "resource" sysfs file */ diff --git a/lib/librte_eal/linuxapp/eal/eal_pci_uio.c b/lib/librte_eal/linuxapp/eal/eal_pci_uio.c index 7e62266..e53f06b 100644 --- a/lib/librte_eal/linuxapp/eal/eal_pci_uio.c +++ b/lib/librte_eal/linuxapp/eal/eal_pci_uio.c @@ -35,6 +35,7 @@ #include #include #include +#include #include #include @@ -48,6 +49,8 @@ static int pci_parse_sysfs_value(const char *filename, uint64_t *val); +void *pci_map_addr = NULL; + #define OFF_MAX ((uint64_t)(off_t)-1) static int @@ -371,10 +374,16 @@ pci_uio_map_resource(struct rte_pci_device *dev) if (maps[j].addr != NULL) fail = 1; else { - mapaddr = pci_map_resource(NULL, fd, (off_t)offset, + /* try mapping somewhere close to the end of hugepages */ + if (pci_map_addr == NULL) + pci_map_addr = pci_find_max_end_va(); + + mapaddr = pci_map_resource(pci_map_addr, fd, (off_t)offset, (size_t)maps[j].size); - if (mapaddr == NULL) + if (mapaddr == MAP_FAILED) fail = 1; + + pci_map_addr = RTE_PTR_ADD(mapaddr, (size_t) maps[j].size); } if (fail) { diff --git a/lib/librte_eal/linuxapp/eal/eal_pci_vfio.c b/lib/librte_eal/linuxapp/eal/eal_pci_vfio.c index c776ddc..c1246e8 100644 --- a/lib/librte_eal/linuxapp/eal/eal_pci_vfio.c +++ b/lib/librte_eal/linuxapp/eal/eal_pci_vfio.c @@ -37,6 +37,7 @@ #include #include #include +#include #include #include @@ -720,10 +721,22 @@ pci_vfio_map_resource(struct rte_pci_device *dev) if (i == msix_bar) continue; - bar_addr = pci_map_resource(maps[i].addr, vfio_dev_fd, reg.offset, - reg.size); + if (internal_config.process_type == RTE_PROC_PRIMARY) { + /* try mapping somewhere close to the end of hugepages */ + if (pci_map_addr == NULL) + pci_map_addr = pci_find_max_end_va(); + + bar_addr = pci_map_resource(pci_map_addr, vfio_dev_fd, reg.offset, + reg.size); + pci_map_addr = RTE_PTR_ADD(bar_addr, (size_t) reg.size); + } else { + bar_addr = pci_map_resource(maps[i].addr, vfio_dev_fd, reg.offset, + reg.size); + } - if (bar_addr == NULL) { + if (bar_addr == MAP_FAILED || + (internal_config.process_type == RTE_PROC_SECONDARY && + bar_addr != maps[i].addr)) { RTE_LOG(ERR, EAL, " %s mapping BAR%i failed: %s\n", pci_addr, i, strerror(errno)); close(vfio_dev_fd); diff --git a/lib/librte_eal/linuxapp/eal/include/eal_pci_init.h b/lib/librte_eal/linuxapp/eal/include/eal_pci_init.h index d758bee..1070eb8 100644 --- a/lib/librte_eal/linuxapp/eal/include/eal_pci_init.h +++ b/lib/librte_eal/linuxapp/eal/include/eal_pci_init.h @@ -59,6 +59,12 @@ struct mapped_pci_resource { TAILQ_HEAD(mapped_pci_res_list, mapped_pci_resource); extern struct mapped_pci_res_list *pci_res_list; +/* + * Helper function to map PCI resources right after hugepages in virtual memory + */ +extern void *pci_map_addr; +void *pci_find_max_end_va(void); + void *pci_map_resource(void *requested_addr, int fd, off_t offset, size_t size); -- 1.8.1.4