f49e858580
This commit renames mempool_obj_ctor_t as mempool_obj_cb_t. In next commits, we will add the ability to populate the mempool and iterate through objects using the same function. We will use the same callback type for that. As the callback is not a constructor anymore, rename it into rte_mempool_obj_cb_t. The rte_mempool_obj_iter_t that was used to iterate over objects will be removed in next commits. No functional change. In this commit, the API is preserved through a compat typedef. Signed-off-by: Olivier Matz <olivier.matz@6wind.com>
296 lines
8.3 KiB
C
296 lines
8.3 KiB
C
/*-
|
|
* BSD LICENSE
|
|
*
|
|
* Copyright(c) 2010-2014 Intel Corporation. All rights reserved.
|
|
* All rights reserved.
|
|
*
|
|
* Redistribution and use in source and binary forms, with or without
|
|
* modification, are permitted provided that the following conditions
|
|
* are met:
|
|
*
|
|
* * Redistributions of source code must retain the above copyright
|
|
* notice, this list of conditions and the following disclaimer.
|
|
* * Redistributions in binary form must reproduce the above copyright
|
|
* notice, this list of conditions and the following disclaimer in
|
|
* the documentation and/or other materials provided with the
|
|
* distribution.
|
|
* * Neither the name of Intel Corporation nor the names of its
|
|
* contributors may be used to endorse or promote products derived
|
|
* from this software without specific prior written permission.
|
|
*
|
|
* THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
|
|
* "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
|
|
* LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
|
|
* A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
|
|
* OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
|
|
* SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
|
|
* LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
|
|
* DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
|
|
* THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
|
|
* (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
|
|
* OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
|
|
*/
|
|
|
|
#include <stdint.h>
|
|
#include <unistd.h>
|
|
#include <stdlib.h>
|
|
#include <sys/mman.h>
|
|
#include <sys/ioctl.h>
|
|
#include <string.h>
|
|
#include <xen/sys/gntalloc.h>
|
|
|
|
#include <rte_common.h>
|
|
#include <rte_mempool.h>
|
|
#include <rte_memory.h>
|
|
#include <rte_errno.h>
|
|
|
|
#include "rte_xen_lib.h"
|
|
#include "rte_eth_xenvirt.h"
|
|
|
|
struct _gntarr {
|
|
uint32_t gref;
|
|
phys_addr_t pa;
|
|
uint64_t index;
|
|
void *va;
|
|
};
|
|
|
|
struct _mempool_gntalloc_info {
|
|
struct rte_mempool *mp;
|
|
uint32_t pg_num;
|
|
uint32_t *gref_arr;
|
|
phys_addr_t *pa_arr;
|
|
void *va;
|
|
uint32_t mempool_idx;
|
|
uint64_t start_index;
|
|
};
|
|
|
|
|
|
static rte_atomic32_t global_xenvirt_mempool_idx = RTE_ATOMIC32_INIT(-1);
|
|
|
|
static int
|
|
compare(const void *p1, const void *p2)
|
|
{
|
|
return ((const struct _gntarr *)p1)->pa - ((const struct _gntarr *)p2)->pa;
|
|
}
|
|
|
|
|
|
static struct _mempool_gntalloc_info
|
|
_create_mempool(const char *name, unsigned elt_num, unsigned elt_size,
|
|
unsigned cache_size, unsigned private_data_size,
|
|
rte_mempool_ctor_t *mp_init, void *mp_init_arg,
|
|
rte_mempool_obj_cb_t *obj_init, void *obj_init_arg,
|
|
int socket_id, unsigned flags)
|
|
{
|
|
struct _mempool_gntalloc_info mgi;
|
|
struct rte_mempool *mp = NULL;
|
|
struct rte_mempool_objsz objsz;
|
|
uint32_t pg_num, rpg_num, pg_shift, pg_sz;
|
|
char *va, *orig_va, *uv; /* uv: from which, the pages could be freed */
|
|
ssize_t sz, usz; /* usz: unused size */
|
|
/*
|
|
* for each page allocated through xen_gntalloc driver,
|
|
* gref_arr:stores grant references,
|
|
* pa_arr: stores physical address,
|
|
* gnt_arr: stores all meta dat
|
|
*/
|
|
uint32_t *gref_arr = NULL;
|
|
phys_addr_t *pa_arr = NULL;
|
|
struct _gntarr *gnt_arr = NULL;
|
|
/* start index of the grant referances, used for dealloc*/
|
|
uint64_t start_index;
|
|
uint32_t i, j;
|
|
int rv = 0;
|
|
struct ioctl_gntalloc_dealloc_gref arg;
|
|
|
|
mgi.mp = NULL;
|
|
va = orig_va = uv = NULL;
|
|
pg_num = rpg_num = 0;
|
|
sz = 0;
|
|
|
|
pg_sz = getpagesize();
|
|
if (rte_is_power_of_2(pg_sz) == 0) {
|
|
goto out;
|
|
}
|
|
pg_shift = rte_bsf32(pg_sz);
|
|
|
|
rte_mempool_calc_obj_size(elt_size, flags, &objsz);
|
|
sz = rte_mempool_xmem_size(elt_num, objsz.total_size, pg_shift);
|
|
pg_num = sz >> pg_shift;
|
|
|
|
pa_arr = calloc(pg_num, sizeof(pa_arr[0]));
|
|
gref_arr = calloc(pg_num, sizeof(gref_arr[0]));
|
|
gnt_arr = calloc(pg_num, sizeof(gnt_arr[0]));
|
|
if ((gnt_arr == NULL) || (gref_arr == NULL) || (pa_arr == NULL))
|
|
goto out;
|
|
|
|
/* grant index is continuous in ascending order */
|
|
orig_va = gntalloc(sz, gref_arr, &start_index);
|
|
if (orig_va == NULL)
|
|
goto out;
|
|
|
|
get_phys_map(orig_va, pa_arr, pg_num, pg_sz);
|
|
for (i = 0; i < pg_num; i++) {
|
|
gnt_arr[i].index = start_index + i * pg_sz;
|
|
gnt_arr[i].gref = gref_arr[i];
|
|
gnt_arr[i].pa = pa_arr[i];
|
|
gnt_arr[i].va = RTE_PTR_ADD(orig_va, i * pg_sz);
|
|
}
|
|
qsort(gnt_arr, pg_num, sizeof(struct _gntarr), compare);
|
|
|
|
va = get_xen_virtual(sz, pg_sz);
|
|
if (va == NULL) {
|
|
goto out;
|
|
}
|
|
|
|
/*
|
|
* map one by one, as index isn't continuous now.
|
|
* pg_num VMAs, doesn't linux has a limitation on this?
|
|
*/
|
|
for (i = 0; i < pg_num; i++) {
|
|
/* update gref_arr and pa_arr after sort */
|
|
gref_arr[i] = gnt_arr[i].gref;
|
|
pa_arr[i] = gnt_arr[i].pa;
|
|
gnt_arr[i].va = mmap(va + i * pg_sz, pg_sz, PROT_READ | PROT_WRITE,
|
|
MAP_SHARED | MAP_FIXED, gntalloc_fd, gnt_arr[i].index);
|
|
if ((gnt_arr[i].va == MAP_FAILED) || (gnt_arr[i].va != (va + i * pg_sz))) {
|
|
RTE_LOG(ERR, PMD, "failed to map %d pages\n", i);
|
|
goto mmap_failed;
|
|
}
|
|
}
|
|
|
|
/*
|
|
* Check that allocated size is big enough to hold elt_num
|
|
* objects and a calcualte how many bytes are actually required.
|
|
*/
|
|
usz = rte_mempool_xmem_usage(va, elt_num, objsz.total_size, pa_arr, pg_num, pg_shift);
|
|
if (usz < 0) {
|
|
mp = NULL;
|
|
i = pg_num;
|
|
goto mmap_failed;
|
|
} else {
|
|
/* unmap unused pages if any */
|
|
uv = RTE_PTR_ADD(va, usz);
|
|
if ((usz = va + sz - uv) > 0) {
|
|
|
|
RTE_LOG(ERR, PMD,
|
|
"%s(%s): unmap unused %zu of %zu "
|
|
"mmaped bytes @%p orig:%p\n",
|
|
__func__, name, usz, sz, uv, va);
|
|
munmap(uv, usz);
|
|
i = (sz - usz) / pg_sz;
|
|
for (; i < pg_num; i++) {
|
|
arg.count = 1;
|
|
arg.index = gnt_arr[i].index;
|
|
rv = ioctl(gntalloc_fd, IOCTL_GNTALLOC_DEALLOC_GREF, &arg);
|
|
if (rv) {
|
|
/* shouldn't fail here */
|
|
RTE_LOG(ERR, PMD, "va=%p pa=%"PRIu64"x index=%"PRIu64" %s\n",
|
|
gnt_arr[i].va,
|
|
gnt_arr[i].pa,
|
|
arg.index, strerror(errno));
|
|
rte_panic("gntdealloc failed when freeing pages\n");
|
|
}
|
|
}
|
|
|
|
rpg_num = (sz - usz) >> pg_shift;
|
|
} else
|
|
rpg_num = pg_num;
|
|
|
|
mp = rte_mempool_xmem_create(name, elt_num, elt_size,
|
|
cache_size, private_data_size,
|
|
mp_init, mp_init_arg,
|
|
obj_init, obj_init_arg,
|
|
socket_id, flags, va, pa_arr, rpg_num, pg_shift);
|
|
|
|
RTE_ASSERT(elt_num == mp->size);
|
|
}
|
|
mgi.mp = mp;
|
|
mgi.pg_num = rpg_num;
|
|
mgi.gref_arr = gref_arr;
|
|
mgi.pa_arr = pa_arr;
|
|
if (mp)
|
|
mgi.mempool_idx = rte_atomic32_add_return(&global_xenvirt_mempool_idx, 1);
|
|
mgi.start_index = start_index;
|
|
mgi.va = va;
|
|
|
|
if (mp == NULL) {
|
|
i = pg_num;
|
|
goto mmap_failed;
|
|
}
|
|
|
|
/*
|
|
* unmap only, without deallocate grant reference.
|
|
* unused pages have already been unmaped,
|
|
* unmap twice will fail, but it is safe.
|
|
*/
|
|
mmap_failed:
|
|
for (j = 0; j < i; j++) {
|
|
if (gnt_arr[i].va)
|
|
munmap(gnt_arr[i].va, pg_sz);
|
|
}
|
|
out:
|
|
free(gnt_arr);
|
|
if (orig_va)
|
|
munmap(orig_va, sz);
|
|
if (mp == NULL) {
|
|
free(gref_arr);
|
|
free(pa_arr);
|
|
|
|
/* some gref has already been de-allocated from the list in the driver,
|
|
* so dealloc one by one, and it is safe to deallocate twice
|
|
*/
|
|
if (orig_va) {
|
|
for (i = 0; i < pg_num; i++) {
|
|
arg.index = start_index + i * pg_sz;
|
|
rv = ioctl(gntalloc_fd, IOCTL_GNTALLOC_DEALLOC_GREF, arg);
|
|
}
|
|
}
|
|
}
|
|
return mgi;
|
|
}
|
|
|
|
struct rte_mempool *
|
|
rte_mempool_gntalloc_create(const char *name, unsigned elt_num, unsigned elt_size,
|
|
unsigned cache_size, unsigned private_data_size,
|
|
rte_mempool_ctor_t *mp_init, void *mp_init_arg,
|
|
rte_mempool_obj_cb_t *obj_init, void *obj_init_arg,
|
|
int socket_id, unsigned flags)
|
|
{
|
|
int rv;
|
|
uint32_t i;
|
|
struct _mempool_gntalloc_info mgi;
|
|
struct ioctl_gntalloc_dealloc_gref arg;
|
|
int pg_sz = getpagesize();
|
|
|
|
mgi = _create_mempool(name, elt_num, elt_size,
|
|
cache_size, private_data_size,
|
|
mp_init, mp_init_arg,
|
|
obj_init, obj_init_arg,
|
|
socket_id, flags);
|
|
if (mgi.mp) {
|
|
rv = grant_gntalloc_mbuf_pool(mgi.mp,
|
|
mgi.pg_num,
|
|
mgi.gref_arr,
|
|
mgi.pa_arr,
|
|
mgi.mempool_idx);
|
|
free(mgi.gref_arr);
|
|
free(mgi.pa_arr);
|
|
if (rv == 0)
|
|
return mgi.mp;
|
|
/*
|
|
* in _create_mempool, unused pages have already been unmapped, deallocagted
|
|
* unmap and dealloc the remained ones here.
|
|
*/
|
|
munmap(mgi.va, pg_sz * mgi.pg_num);
|
|
for (i = 0; i < mgi.pg_num; i++) {
|
|
arg.index = mgi.start_index + i * pg_sz;
|
|
rv = ioctl(gntalloc_fd, IOCTL_GNTALLOC_DEALLOC_GREF, arg);
|
|
}
|
|
return NULL;
|
|
}
|
|
return NULL;
|
|
|
|
|
|
|
|
}
|