[PATCH] eal: Support running as unprivileged user
From: Ben Walker <hidden>
Date: 2017-01-03 22:30:43
For Linux kernel 4.0 and newer, the ability to obtain physical page frame numbers for unprivileged users from /proc/self/pagemap was removed. Instead, when an IOMMU is present, simply choose our own DMA addresses instead. Signed-off-by: Ben Walker <redacted> --- lib/librte_eal/common/eal_private.h | 12 ++++++ lib/librte_eal/linuxapp/eal/eal_memory.c | 71 +++++++++++++++++++++++--------- lib/librte_eal/linuxapp/eal/eal_pci.c | 6 ++- 3 files changed, 68 insertions(+), 21 deletions(-)
diff --git a/lib/librte_eal/common/eal_private.h b/lib/librte_eal/common/eal_private.h
index 9e7d8f6..8b2d323 100644
--- a/lib/librte_eal/common/eal_private.h
+++ b/lib/librte_eal/common/eal_private.h@@ -34,6 +34,7 @@ #ifndef _EAL_PRIVATE_H_ #define _EAL_PRIVATE_H_ +#include <stdbool.h> #include <stdio.h> #include <rte_pci.h>
@@ -301,4 +302,15 @@ int rte_eal_hugepage_init(void); */ int rte_eal_hugepage_attach(void); +/** + * Returns true if the system is able to obtain + * physical addresses. Return false if using DMA + * addresses through an IOMMU. + * + * Drivers based on uio will not load unless physical + * addresses are obtainable. It is only possible to get + * physical addresses when running as a privileged user. + */ +bool rte_eal_using_phys_addrs(void); + #endif /* _EAL_PRIVATE_H_ */
diff --git a/lib/librte_eal/linuxapp/eal/eal_memory.c b/lib/librte_eal/linuxapp/eal/eal_memory.c
index a956bb2..33c66c1 100644
--- a/lib/librte_eal/linuxapp/eal/eal_memory.c
+++ b/lib/librte_eal/linuxapp/eal/eal_memory.c@@ -64,6 +64,7 @@ #define _FILE_OFFSET_BITS 64 #include <errno.h> #include <stdarg.h> +#include <stdbool.h> #include <stdlib.h> #include <stdio.h> #include <stdint.h>
@@ -122,26 +123,24 @@ int rte_xen_dom0_supported(void) static uint64_t baseaddr_offset; -static unsigned proc_pagemap_readable; +static bool phys_addrs_available = true; #define RANDOMIZE_VA_SPACE_FILE "/proc/sys/kernel/randomize_va_space" static void -test_proc_pagemap_readable(void) +test_phys_addrs_available(void) { - int fd = open("/proc/self/pagemap", O_RDONLY); + uint64_t tmp; + phys_addr_t physaddr; - if (fd < 0) { + physaddr = rte_mem_virt2phy(&tmp); + if (physaddr == RTE_BAD_PHYS_ADDR) { RTE_LOG(ERR, EAL, - "Cannot open /proc/self/pagemap: %s. " - "virt2phys address translation will not work\n", + "Cannot obtain physical addresses: %s. " + "Only vfio will function.\n", strerror(errno)); - return; + phys_addrs_available = false; } - - /* Is readable */ - close(fd); - proc_pagemap_readable = 1; } /* Lock page in physical memory and prevent from swapping. */
@@ -190,7 +189,7 @@ rte_mem_virt2phy(const void *virtaddr) } /* Cannot parse /proc/self/pagemap, no need to log errors everywhere */ - if (!proc_pagemap_readable) + if (!phys_addrs_available) return RTE_BAD_PHYS_ADDR; /* standard page size */
@@ -229,6 +228,9 @@ rte_mem_virt2phy(const void *virtaddr) * the pfn (page frame number) are bits 0-54 (see * pagemap.txt in linux Documentation) */ + if ((page & 0x7fffffffffffffULL) == 0) + return RTE_BAD_PHYS_ADDR; + physaddr = ((page & 0x7fffffffffffffULL) * page_size) + ((unsigned long)virtaddr % page_size);
@@ -255,6 +257,22 @@ find_physaddrs(struct hugepage_file *hugepg_tbl, struct hugepage_info *hpi) } /* + * For each hugepage in hugepg_tbl, fill the physaddr value sequentially. + */ +static int +set_physaddrs(struct hugepage_file *hugepg_tbl, struct hugepage_info *hpi) +{ + unsigned i; + phys_addr_t addr = 0; + + for (i = 0; i < hpi->num_pages[0]; i++) { + hugepg_tbl[i].physaddr = addr; + addr += hugepg_tbl[i].size; + } + return 0; +} + +/* * Check whether address-space layout randomization is enabled in * the kernel. This is important for multi-process as it can prevent * two processes mapping data to the same virtual address
@@ -951,7 +969,7 @@ rte_eal_hugepage_init(void) int nr_hugefiles, nr_hugepages = 0; void *addr; - test_proc_pagemap_readable(); + test_phys_addrs_available(); memset(used_hp, 0, sizeof(used_hp));
@@ -1043,11 +1061,20 @@ rte_eal_hugepage_init(void) continue; } - /* find physical addresses and sockets for each hugepage */ - if (find_physaddrs(&tmp_hp[hp_offset], hpi) < 0){ - RTE_LOG(DEBUG, EAL, "Failed to find phys addr for %u MB pages\n", - (unsigned)(hpi->hugepage_sz / 0x100000)); - goto fail; + if (phys_addrs_available) { + /* find physical addresses for each hugepage */ + if (find_physaddrs(&tmp_hp[hp_offset], hpi) < 0){ + RTE_LOG(DEBUG, EAL, "Failed to find phys addr for %u MB pages\n", + (unsigned)(hpi->hugepage_sz / 0x100000)); + goto fail; + } + } else { + /* set physical addresses for each hugepage */ + if (set_physaddrs(&tmp_hp[hp_offset], hpi) < 0) { + RTE_LOG(DEBUG, EAL, "Failed to set phys addr for %u MB pages\n", + (unsigned)(hpi->hugepage_sz / 0x100000)); + goto fail; + } } if (find_numasocket(&tmp_hp[hp_offset], hpi) < 0){
@@ -1289,7 +1316,7 @@ rte_eal_hugepage_attach(void) "into secondary processes\n"); } - test_proc_pagemap_readable(); + test_phys_addrs_available(); if (internal_config.xen_dom0_support) { #ifdef RTE_LIBRTE_XEN_DOM0
@@ -1426,3 +1453,9 @@ rte_eal_hugepage_attach(void) close(fd_hugepage); return -1; } + +bool +rte_eal_using_phys_addrs(void) +{ + return phys_addrs_available; +}
diff --git a/lib/librte_eal/linuxapp/eal/eal_pci.c b/lib/librte_eal/linuxapp/eal/eal_pci.c
index 4350134..cfbfedf 100644
--- a/lib/librte_eal/linuxapp/eal/eal_pci.c
+++ b/lib/librte_eal/linuxapp/eal/eal_pci.c@@ -99,8 +99,10 @@ rte_eal_pci_map_device(struct rte_pci_device *dev) break; case RTE_KDRV_IGB_UIO: case RTE_KDRV_UIO_GENERIC: - /* map resources for devices that use uio */ - ret = pci_uio_map_resource(dev); + if (rte_eal_using_phys_addrs()) { + /* map resources for devices that use uio */ + ret = pci_uio_map_resource(dev); + } break; default: RTE_LOG(DEBUG, EAL,
--
2.9.3