ligang/vpp
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

vppinfra: remove linux/syscall.h

Browse Source 
For portabiliy reasons it is better to have all wrapped in clib code.
I.e. instead of using getcpu() we have clib_get_current_numa_node () and
clib_get_current_cpu_id().

Type: refactor
Change-Id: I29b52d7f29bc7f93873402c4070561f564b71c63
Signed-off-by: Damjan Marion <damarion@cisco.com>
This commit is contained in:
Damjan Marion
2020-10-09 17:16:55 +02:00
parent 30a819579c
commit f8cb70177f
14 changed files with 51 additions and 165 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
extras/scripts/build_static_vppctl.sh
View File

@ -11,8 +11,6 @@ ${CC:-cc} \
-static \
-I ${src} \
-I ${tmp} \
-DHAVE_MEMFD_CREATE \
-DHAVE_GETCPU \
${src}/vppinfra/backtrace.c \
${src}/vppinfra/dlmalloc.c \
${src}/vppinfra/elf.c \

23
src/cmake/syscall.cmake
View File

@ -11,29 +11,6 @@
# See the License for the specific language governing permissions and
# limitations under the License.
##############################################################################
# Check for memfd_create headers and libs
##############################################################################
check_c_source_compiles("
#define _GNU_SOURCE
#include <sys/mman.h>
int main() { return memfd_create (\"/dev/false\", 0); }
" HAVE_MEMFD_CREATE)
if (HAVE_MEMFD_CREATE)
add_definitions(-DHAVE_MEMFD_CREATE)
endif()
check_c_source_compiles("
#define _GNU_SOURCE
#include <sched.h>
int main() { return getcpu (0, 0); }
" HAVE_GETCPU)
if (HAVE_GETCPU)
add_definitions(-DHAVE_GETCPU)
endif()
check_c_source_compiles("
#define _GNU_SOURCE
#include <fcntl.h>

1
src/plugins/memif/memif.c
View File

@ -33,7 +33,6 @@
#include <vlib/vlib.h>
#include <vlib/unix/unix.h>
#include <vppinfra/linux/syscall.h>
#include <vnet/plugin/plugin.h>
#include <vnet/ethernet/ethernet.h>
#include <vnet/interface/rx_queue_funcs.h>

1
src/vlib/physmem.c
View File

@ -21,7 +21,6 @@
#include <sys/stat.h>
#include <unistd.h>
#include <vppinfra/linux/syscall.h>
#include <vppinfra/linux/sysfs.h>
#include <vlib/vlib.h>
#include <vlib/physmem.h>

2
src/vpp/stats/stat_segment.c
View File

@ -19,8 +19,6 @@
#include "stat_segment.h"
#include <vnet/vnet.h>
#include <vnet/devices/devices.h> /* vnet_get_aggregate_rx_packets */
#undef HAVE_MEMFD_CREATE
#include <vppinfra/linux/syscall.h>
#include <vpp-api/client/stat_client.h>
stat_segment_main_t stat_segment_main;

1
src/vppinfra/CMakeLists.txt
View File

@ -192,7 +192,6 @@ set(VPPINFRA_HEADERS
vector_sse42.h
warnings.h
xxhash.h
linux/syscall.h
linux/sysfs.h
)

2
src/vppinfra/bihash_template.c
View File

@ -272,7 +272,7 @@ void BV (clib_bihash_initiator_init_svm)
ASSERT (memory_size < (1ULL << 32));
/* Set up for memfd sharing */
if ((fd = memfd_create (name, MFD_ALLOW_SEALING)) == -1)
if ((fd = clib_mem_vm_create_fd (CLIB_MEM_PAGE_SZ_DEFAULT, name) == -1)
{
clib_unix_warning ("memfd_create");
return;

1
src/vppinfra/bihash_template.h
View File

@ -34,7 +34,6 @@
#endif
#ifdef BIHASH_32_64_SVM
#undef HAVE_MEMFD_CREATE
#include <vppinfra/linux/syscall.h>
#include <fcntl.h>
#define F_LINUX_SPECIFIC_BASE 1024

14
src/vppinfra/cpu.c
View File

@ -222,7 +222,21 @@ format_cpu_flags (u8 * s, va_list * args)
#endif
}
__clib_export u32
clib_get_current_cpu_id ()
{
unsigned cpu, node;
syscall (__NR_getcpu, &cpu, &node, 0);
return cpu;
}
__clib_export u32
clib_get_current_numa_node ()
{
unsigned cpu, node;
syscall (__NR_getcpu, &cpu, &node, 0);
return node;
}
/*
* fd.io coding-style-patch-verification: ON

17
src/vppinfra/cpu.h
View File

@ -167,21 +167,8 @@ _ (asimddp, 20) \
_ (sha512, 21) \
_ (sve, 22)
static inline u32
clib_get_current_cpu_id ()
{
unsigned cpu, node;
syscall (__NR_getcpu, &cpu, &node, 0);
return cpu;
}
static inline u32
clib_get_current_numa_node ()
{
unsigned cpu, node;
syscall (__NR_getcpu, &cpu, &node, 0);
return node;
}
u32 clib_get_current_cpu_id ();
u32 clib_get_current_numa_node ();
#if defined(__x86_64__)
#include "cpuid.h"

14
src/vppinfra/linux/mem.c
View File

@ -30,7 +30,6 @@
#include <vppinfra/time.h>
#include <vppinfra/format.h>
#include <vppinfra/clib_error.h>
#include <vppinfra/linux/syscall.h>
#include <vppinfra/linux/sysfs.h>
#ifndef F_LINUX_SPECIFIC_BASE
@ -149,7 +148,7 @@ clib_mem_main_init ()
mm->log2_page_sz = min_log2 (page_size);
/* default system hugeppage size */
if ((fd = memfd_create ("test", MFD_HUGETLB)) != -1)
if ((fd = syscall (__NR_memfd_create, "test", MFD_HUGETLB)) != -1)
{
mm->log2_default_hugepage_sz = clib_mem_get_fd_log2_page_size (fd);
close (fd);
@ -169,7 +168,7 @@ clib_mem_main_init ()
for (int i = 0; i < CLIB_MAX_NUMAS; i++)
{
int status;
if (move_pages (0, 1, &va, &i, &status, 0) == 0)
if (syscall (__NR_move_pages, 0, 1, &va, &i, &status, 0) == 0)
mm->numa_node_bitmap |= 1ULL << i;
}
@ -298,7 +297,7 @@ clib_mem_vm_create_fd (clib_mem_page_sz_t log2_page_size, char *fmt, ...)
vec_add1 (s, 0);
/* memfd_create introduced in kernel 3.17, we don't support older kernels */
fd = memfd_create ((char *) s, memfd_flags);
fd = syscall (__NR_memfd_create, (char *) s, memfd_flags);
/* kernel versions < 4.14 does not support memfd_create for huge pages */
if (fd == -1 && errno == EINVAL &&
@ -568,7 +567,7 @@ clib_mem_get_page_stats (void *start, clib_mem_page_sz_t log2_page_size,
stats->total = n_pages;
stats->log2_page_sz = log2_page_size;
if (move_pages (0, n_pages, ptr, 0, status, 0) != 0)
if (syscall (__NR_move_pages, 0, n_pages, ptr, 0, status, 0) != 0)
{
stats->unknown = n_pages;
goto done;
@ -658,7 +657,8 @@ clib_mem_set_numa_affinity (u8 numa_node, int force)
mask[0] = 1 << numa_node;
if (set_mempolicy (force ? MPOL_BIND : MPOL_PREFERRED, mask, mask_len))
if (syscall (__NR_set_mempolicy, force ? MPOL_BIND : MPOL_PREFERRED, mask,
mask_len))
goto error;
vec_reset_length (mm->error);
@ -675,7 +675,7 @@ clib_mem_set_default_numa_affinity ()
{
clib_mem_main_t *mm = &clib_mem_main;
if (set_mempolicy (MPOL_DEFAULT, 0, 0))
if (syscall (__NR_set_mempolicy, MPOL_DEFAULT, 0, 0))
{
vec_reset_length (mm->error);
mm->error = clib_error_return_unix (mm->error, (char *) __func__);

66
src/vppinfra/linux/syscall.h
View File

@ -1,66 +0,0 @@
/*
* Copyright (c) 2017 Cisco and/or its affiliates.
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at:
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
#ifndef included_linux_syscall_h
#define included_linux_syscall_h
#include <unistd.h>
#include <sys/syscall.h>
#ifndef HAVE_GETCPU
static inline int
getcpu (unsigned *cpu, unsigned *node)
{
return syscall (__NR_getcpu, cpu, node, 0);
}
#endif
static inline long
set_mempolicy (int mode, const unsigned long *nodemask, unsigned long maxnode)
{
return syscall (__NR_set_mempolicy, mode, nodemask, maxnode);
}
static inline int
get_mempolicy (int *mode, unsigned long *nodemask, unsigned long maxnode,
void *addr, unsigned long flags)
{
return syscall (__NR_get_mempolicy, mode, nodemask, maxnode, addr, flags);
}
static inline long
move_pages (int pid, unsigned long count, void **pages, const int *nodes,
int *status, int flags)
{
return syscall (__NR_move_pages, pid, count, pages, nodes, status, flags);
}
#ifndef HAVE_MEMFD_CREATE
static inline int
memfd_create (const char *name, unsigned int flags)
{
return syscall (__NR_memfd_create, name, flags);
}
#endif
#endif /* included_linux_syscall_h */
/*
* fd.io coding-style-patch-verification: ON
*
* Local Variables:
* eval: (c-set-style "gnu")
* End:
*/

70
src/vppinfra/pmalloc.c
View File

@ -19,16 +19,14 @@
#include <sys/stat.h>
#include <fcntl.h>
#include <unistd.h>
#include <linux/mempolicy.h>
#include <linux/memfd.h>
#include <sched.h>
#include <vppinfra/format.h>
#include <vppinfra/linux/syscall.h>
#include <vppinfra/linux/sysfs.h>
#include <vppinfra/mem.h>
#include <vppinfra/hash.h>
#include <vppinfra/pmalloc.h>
#include <vppinfra/cpu.h>
#if __SIZEOF_POINTER__ >= 8
#define DEFAULT_RESERVED_MB 16384
@ -48,18 +46,6 @@ pmalloc_size2pages (uword size, u32 log2_page_sz)
return round_pow2 (size, 1ULL << log2_page_sz) >> log2_page_sz;
}
static inline int
pmalloc_validate_numa_node (u32 * numa_node)
{
if (*numa_node == CLIB_PMALLOC_NUMA_LOCAL)
{
u32 cpu;
if (getcpu (&cpu, numa_node) != 0)
return 1;
}
return 0;
}
__clib_export int
clib_pmalloc_init (clib_pmalloc_main_t * pm, uword base_addr, uword size)
{
@ -241,12 +227,10 @@ static inline clib_pmalloc_page_t *
pmalloc_map_pages (clib_pmalloc_main_t * pm, clib_pmalloc_arena_t * a,
u32 numa_node, u32 n_pages)
{
clib_mem_page_stats_t stats = {};
clib_pmalloc_page_t *pp = 0;
int status, rv, i, mmap_flags;
int rv, i, mmap_flags;
void *va = MAP_FAILED;
int old_mpol = -1;
long unsigned int mask[16] = { 0 };
long unsigned int old_mask[16] = { 0 };
uword size = (uword) n_pages << pm->def_log2_page_sz;
clib_error_free (pm->error);
@ -266,17 +250,8 @@ pmalloc_map_pages (clib_pmalloc_main_t * pm, clib_pmalloc_arena_t * a,
return 0;
}
rv = get_mempolicy (&old_mpol, old_mask, sizeof (old_mask) * 8 + 1, 0, 0);
/* failure to get mempolicy means we can only proceed with numa 0 maps */
if (rv == -1 && numa_node != 0)
{
pm->error = clib_error_return_unix (0, "failed to get mempolicy");
return 0;
}
mask[0] = 1 << numa_node;
rv = set_mempolicy (MPOL_BIND, mask, sizeof (mask) * 8 + 1);
if (rv == -1 && numa_node != 0)
rv = clib_mem_set_numa_affinity (numa_node, /* force */ 1);
if (rv == CLIB_MEM_ERROR && numa_node != 0)
{
pm->error = clib_error_return_unix (0, "failed to set mempolicy for "
"numa node %u", numa_node);
@ -323,8 +298,8 @@ pmalloc_map_pages (clib_pmalloc_main_t * pm, clib_pmalloc_arena_t * a,
clib_memset (va, 0, size);
rv = set_mempolicy (old_mpol, old_mask, sizeof (old_mask) * 8 + 1);
if (rv == -1 && numa_node != 0)
rv = clib_mem_set_default_numa_affinity ();
if (rv == CLIB_MEM_ERROR && numa_node != 0)
{
pm->error = clib_error_return_unix (0, "failed to restore mempolicy");
goto error;
@ -332,14 +307,23 @@ pmalloc_map_pages (clib_pmalloc_main_t * pm, clib_pmalloc_arena_t * a,
/* we tolerate move_pages failure only if request os for numa node 0
to support non-numa kernels */
rv = move_pages (0, 1, &va, 0, &status, 0);
if ((rv == 0 && status != numa_node) || (rv != 0 && numa_node != 0))
clib_mem_get_page_stats (va, CLIB_MEM_PAGE_SZ_DEFAULT, 1, &stats);
if (stats.per_numa[numa_node] != 1)
{
pm->error = rv == -1 ?
clib_error_return_unix (0, "page allocated on wrong node, numa node "
"%u status %d", numa_node, status) :
clib_error_return (0, "page allocated on wrong node, numa node "
"%u status %d", numa_node, status);
u16 allocated_at = ~0;
if (stats.unknown)
clib_error_return (0,
"unable to get information about numa allocation");
for (u16 i = 0; i < CLIB_MAX_NUMAS; i++)
if (stats.per_numa[i] == 1)
allocated_at = i;
clib_error_return (0,
"page allocated on the wrong numa node (%u), "
"expected %u",
allocated_at, numa_node);
goto error;
}
@ -407,8 +391,8 @@ clib_pmalloc_create_shared_arena (clib_pmalloc_main_t * pm, char *name,
if (n_pages + vec_len (pm->pages) > pm->max_pages)
return 0;
if (pmalloc_validate_numa_node (&numa_node))
return 0;
if (numa_node == CLIB_PMALLOC_NUMA_LOCAL)
numa_node = clib_get_current_numa_node ();
pool_get (pm->arenas, a);
a->index = a - pm->arenas;
@ -438,8 +422,8 @@ clib_pmalloc_alloc_inline (clib_pmalloc_main_t * pm, clib_pmalloc_arena_t * a,
ASSERT (is_pow2 (align));
if (pmalloc_validate_numa_node (&numa_node))
return 0;
if (numa_node == CLIB_PMALLOC_NUMA_LOCAL)
numa_node = clib_get_current_numa_node ();
if (a == 0)
{

2
src/vppinfra/unix-formats.c
View File

@ -63,8 +63,6 @@
#include <vppinfra/time.h>
#if __linux__
#include <vppinfra/linux/syscall.h>
#ifdef AF_NETLINK
#include <linux/types.h>
#include <linux/netlink.h>