mirror of
https://github.com/AsahiLinux/u-boot
synced 2024-11-15 01:17:39 +00:00
047d7bb712
Import cvmx-fpa.c from 2013 U-Boot. It will be used by the later added drivers to support networking on the MIPS Octeon II / III platforms. Signed-off-by: Aaron Williams <awilliams@marvell.com> Signed-off-by: Stefan Roese <sr@denx.de>
1127 lines
32 KiB
C
1127 lines
32 KiB
C
// SPDX-License-Identifier: GPL-2.0
|
|
/*
|
|
* Copyright (C) 2018-2022 Marvell International Ltd.
|
|
*
|
|
* Support library for the hardware Free Pool Allocator.
|
|
*/
|
|
|
|
#include <errno.h>
|
|
#include <log.h>
|
|
#include <time.h>
|
|
#include <linux/delay.h>
|
|
|
|
#include <mach/cvmx-regs.h>
|
|
#include <mach/cvmx-csr.h>
|
|
#include <mach/cvmx-bootmem.h>
|
|
#include <mach/octeon-model.h>
|
|
#include <mach/cvmx-fuse.h>
|
|
#include <mach/octeon-feature.h>
|
|
#include <mach/cvmx-qlm.h>
|
|
#include <mach/octeon_qlm.h>
|
|
#include <mach/cvmx-pcie.h>
|
|
#include <mach/cvmx-coremask.h>
|
|
#include <mach/cvmx-range.h>
|
|
#include <mach/cvmx-global-resources.h>
|
|
|
|
#include <mach/cvmx-agl-defs.h>
|
|
#include <mach/cvmx-bgxx-defs.h>
|
|
#include <mach/cvmx-ciu-defs.h>
|
|
#include <mach/cvmx-gmxx-defs.h>
|
|
#include <mach/cvmx-gserx-defs.h>
|
|
#include <mach/cvmx-ilk-defs.h>
|
|
#include <mach/cvmx-ipd-defs.h>
|
|
#include <mach/cvmx-pcsx-defs.h>
|
|
#include <mach/cvmx-pcsxx-defs.h>
|
|
#include <mach/cvmx-pki-defs.h>
|
|
#include <mach/cvmx-pko-defs.h>
|
|
#include <mach/cvmx-xcv-defs.h>
|
|
|
|
#include <mach/cvmx-hwpko.h>
|
|
#include <mach/cvmx-ilk.h>
|
|
#include <mach/cvmx-pki.h>
|
|
#include <mach/cvmx-pko3.h>
|
|
#include <mach/cvmx-pko3-queue.h>
|
|
#include <mach/cvmx-pko3-resources.h>
|
|
|
|
#include <mach/cvmx-helper.h>
|
|
#include <mach/cvmx-helper-board.h>
|
|
#include <mach/cvmx-helper-cfg.h>
|
|
|
|
#include <mach/cvmx-helper-bgx.h>
|
|
#include <mach/cvmx-helper-cfg.h>
|
|
#include <mach/cvmx-helper-util.h>
|
|
#include <mach/cvmx-helper-pki.h>
|
|
|
|
static const int debug;
|
|
|
|
/* Due to suspected errata, we may not be able to let the FPA_AURAX_CNT
|
|
* get too close to 0, to avoid a spurious wrap-around error
|
|
*/
|
|
const unsigned int __cvmx_fpa3_cnt_offset = 32;
|
|
|
|
/* For advanced checks, a guard-band is created around the internal
|
|
* stack, to make sure the stack is not overwritten.
|
|
*/
|
|
const u64 magic_pattern = 0xbab4faced095f00d;
|
|
const unsigned int guard_band_size = 0 << 10; /* 1KiB default*/
|
|
|
|
#define CVMX_CACHE_LINE_SHIFT (7)
|
|
|
|
#define CVMX_FPA3_NAME_LEN (16)
|
|
|
|
typedef struct {
|
|
char name[CVMX_FPA3_NAME_LEN];
|
|
u64 stack_paddr; /* Internal stack storage */
|
|
u64 bufs_paddr; /* Buffer pool base address */
|
|
u64 stack_psize; /* Internal stack storage size */
|
|
u64 bufs_psize; /* Buffer pool raw size */
|
|
u64 buf_count; /* Number of buffer filled */
|
|
u64 buf_size; /* Buffer size */
|
|
} cvmx_fpa3_poolx_info_t;
|
|
|
|
typedef struct {
|
|
char name[CVMX_FPA3_NAME_LEN];
|
|
unsigned int buf_size; /* Buffer size */
|
|
} cvmx_fpa3_aurax_info_t;
|
|
|
|
typedef struct {
|
|
char name[CVMX_FPA1_NAME_SIZE];
|
|
u64 size; /* Block size of pool buffers */
|
|
u64 buffer_count;
|
|
u64 base_paddr; /* Base physical addr */
|
|
/* if buffer is allocated at initialization */
|
|
} cvmx_fpa1_pool_info_t;
|
|
|
|
/**
|
|
* FPA1/FPA3 info structure is stored in a named block
|
|
* that is allocated once and shared among applications.
|
|
*/
|
|
static cvmx_fpa1_pool_info_t *cvmx_fpa1_pool_info;
|
|
static cvmx_fpa3_poolx_info_t *cvmx_fpa3_pool_info[CVMX_MAX_NODES];
|
|
static cvmx_fpa3_aurax_info_t *cvmx_fpa3_aura_info[CVMX_MAX_NODES];
|
|
|
|
/**
|
|
* Return the size of buffers held in a POOL
|
|
*
|
|
* @param pool is the POOL handle
|
|
* @return buffer size in bytes
|
|
*
|
|
*/
|
|
int cvmx_fpa3_get_pool_buf_size(cvmx_fpa3_pool_t pool)
|
|
{
|
|
cvmx_fpa_poolx_cfg_t pool_cfg;
|
|
|
|
if (!__cvmx_fpa3_pool_valid(pool))
|
|
return -1;
|
|
|
|
pool_cfg.u64 = csr_rd_node(pool.node, CVMX_FPA_POOLX_CFG(pool.lpool));
|
|
return pool_cfg.cn78xx.buf_size << CVMX_CACHE_LINE_SHIFT;
|
|
}
|
|
|
|
/**
|
|
* Return the size of buffers held in a buffer pool
|
|
*
|
|
* @param pool is the pool number
|
|
*
|
|
* This function will work with CN78XX models in backward-compatible mode
|
|
*/
|
|
unsigned int cvmx_fpa_get_block_size(int pool)
|
|
{
|
|
if (octeon_has_feature(OCTEON_FEATURE_FPA3)) {
|
|
return cvmx_fpa3_get_pool_buf_size(cvmx_fpa3_aura_to_pool(
|
|
cvmx_fpa1_pool_to_fpa3_aura(pool)));
|
|
} else {
|
|
if ((unsigned int)pool >= CVMX_FPA1_NUM_POOLS)
|
|
return 0;
|
|
if (!cvmx_fpa1_pool_info)
|
|
cvmx_fpa_global_init_node(0);
|
|
return cvmx_fpa1_pool_info[pool].size;
|
|
}
|
|
}
|
|
|
|
static void cvmx_fpa3_set_aura_name(cvmx_fpa3_gaura_t aura, const char *name)
|
|
{
|
|
cvmx_fpa3_aurax_info_t *pinfo;
|
|
|
|
pinfo = cvmx_fpa3_aura_info[aura.node];
|
|
if (!pinfo)
|
|
return;
|
|
pinfo += aura.laura;
|
|
memset(pinfo->name, 0, sizeof(pinfo->name));
|
|
if (name)
|
|
strlcpy(pinfo->name, name, sizeof(pinfo->name));
|
|
}
|
|
|
|
static void cvmx_fpa3_set_pool_name(cvmx_fpa3_pool_t pool, const char *name)
|
|
{
|
|
cvmx_fpa3_poolx_info_t *pinfo;
|
|
|
|
pinfo = cvmx_fpa3_pool_info[pool.node];
|
|
if (!pinfo)
|
|
return;
|
|
pinfo += pool.lpool;
|
|
memset(pinfo->name, 0, sizeof(pinfo->name));
|
|
if (name)
|
|
strlcpy(pinfo->name, name, sizeof(pinfo->name));
|
|
}
|
|
|
|
static void cvmx_fpa_set_name(int pool_num, const char *name)
|
|
{
|
|
if (octeon_has_feature(OCTEON_FEATURE_FPA3)) {
|
|
cvmx_fpa3_set_aura_name(cvmx_fpa1_pool_to_fpa3_aura(pool_num),
|
|
name);
|
|
} else {
|
|
cvmx_fpa1_pool_info_t *pinfo;
|
|
|
|
if ((unsigned int)pool_num >= CVMX_FPA1_NUM_POOLS)
|
|
return;
|
|
if (!cvmx_fpa1_pool_info)
|
|
cvmx_fpa_global_init_node(0);
|
|
pinfo = &cvmx_fpa1_pool_info[pool_num];
|
|
memset(pinfo->name, 0, sizeof(pinfo->name));
|
|
if (name)
|
|
strlcpy(pinfo->name, name, sizeof(pinfo->name));
|
|
}
|
|
}
|
|
|
|
static int cvmx_fpa3_aura_cfg(cvmx_fpa3_gaura_t aura, cvmx_fpa3_pool_t pool,
|
|
u64 limit, u64 threshold, int ptr_dis)
|
|
{
|
|
cvmx_fpa3_aurax_info_t *pinfo;
|
|
cvmx_fpa_aurax_cfg_t aura_cfg;
|
|
cvmx_fpa_poolx_cfg_t pool_cfg;
|
|
cvmx_fpa_aurax_cnt_t cnt_reg;
|
|
cvmx_fpa_aurax_cnt_limit_t limit_reg;
|
|
cvmx_fpa_aurax_cnt_threshold_t thresh_reg;
|
|
cvmx_fpa_aurax_int_t int_reg;
|
|
unsigned int block_size;
|
|
|
|
if (debug)
|
|
debug("%s: AURA %u:%u POOL %u:%u\n", __func__, aura.node,
|
|
aura.laura, pool.node, pool.lpool);
|
|
|
|
if (aura.node != pool.node) {
|
|
printf("ERROR: %s: AURA/POOL node mismatch\n", __func__);
|
|
return -1;
|
|
}
|
|
|
|
if (!__cvmx_fpa3_aura_valid(aura)) {
|
|
printf("ERROR: %s: AURA invalid\n", __func__);
|
|
return -1;
|
|
}
|
|
|
|
if (!__cvmx_fpa3_pool_valid(pool)) {
|
|
printf("ERROR: %s: POOL invalid\n", __func__);
|
|
return -1;
|
|
}
|
|
|
|
/* Record POOL block size in AURA info entry */
|
|
pool_cfg.u64 = csr_rd_node(pool.node, CVMX_FPA_POOLX_CFG(pool.lpool));
|
|
|
|
block_size = pool_cfg.cn78xx.buf_size << 7;
|
|
pinfo = cvmx_fpa3_aura_info[aura.node];
|
|
if (!pinfo)
|
|
return -1;
|
|
pinfo += aura.laura;
|
|
|
|
pinfo->buf_size = block_size;
|
|
|
|
/* block_size should be >0 except for POOL=0 which is never enabled*/
|
|
if (pool_cfg.cn78xx.ena && block_size == 0) {
|
|
printf("ERROR; %s: POOL buf_size invalid\n", __func__);
|
|
return -1;
|
|
}
|
|
|
|
/* Initialize AURA count, limit and threshold registers */
|
|
cnt_reg.u64 = 0;
|
|
cnt_reg.cn78xx.cnt = 0 + __cvmx_fpa3_cnt_offset;
|
|
|
|
limit_reg.u64 = 0;
|
|
limit_reg.cn78xx.limit = limit;
|
|
/* Apply count offset, unless it cases a wrap-around */
|
|
if ((limit + __cvmx_fpa3_cnt_offset) < CVMX_FPA3_AURAX_LIMIT_MAX)
|
|
limit_reg.cn78xx.limit += __cvmx_fpa3_cnt_offset;
|
|
|
|
thresh_reg.u64 = 0;
|
|
thresh_reg.cn78xx.thresh = threshold;
|
|
/* Apply count offset, unless it cases a wrap-around */
|
|
if ((threshold + __cvmx_fpa3_cnt_offset) < CVMX_FPA3_AURAX_LIMIT_MAX)
|
|
thresh_reg.cn78xx.thresh += __cvmx_fpa3_cnt_offset;
|
|
|
|
csr_wr_node(aura.node, CVMX_FPA_AURAX_CNT(aura.laura), cnt_reg.u64);
|
|
csr_wr_node(aura.node, CVMX_FPA_AURAX_CNT_LIMIT(aura.laura),
|
|
limit_reg.u64);
|
|
csr_wr_node(aura.node, CVMX_FPA_AURAX_CNT_THRESHOLD(aura.laura),
|
|
thresh_reg.u64);
|
|
|
|
/* Clear any pending error interrupts */
|
|
int_reg.u64 = 0;
|
|
int_reg.cn78xx.thresh = 1;
|
|
|
|
/* Follow a write to clear FPA_AURAX_INT[THRESH] with a read as
|
|
* a workaround to Errata FPA-23410. If FPA_AURAX_INT[THRESH]
|
|
* isn't clear, try again.
|
|
*/
|
|
do {
|
|
csr_wr_node(aura.node, CVMX_FPA_AURAX_INT(aura.laura),
|
|
int_reg.u64);
|
|
int_reg.u64 =
|
|
csr_rd_node(aura.node, CVMX_FPA_AURAX_INT(aura.laura));
|
|
} while (int_reg.s.thresh);
|
|
|
|
/* Disable backpressure etc.*/
|
|
csr_wr_node(aura.node, CVMX_FPA_AURAX_CNT_LEVELS(aura.laura), 0);
|
|
csr_wr_node(aura.node, CVMX_FPA_AURAX_POOL_LEVELS(aura.laura), 0);
|
|
|
|
aura_cfg.u64 = 0;
|
|
aura_cfg.s.ptr_dis = ptr_dis;
|
|
csr_wr_node(aura.node, CVMX_FPA_AURAX_CFG(aura.laura), aura_cfg.u64);
|
|
csr_wr_node(aura.node, CVMX_FPA_AURAX_POOL(aura.laura), pool.lpool);
|
|
|
|
return 0;
|
|
}
|
|
|
|
/**
|
|
* @INTERNAL
|
|
*
|
|
* Fill a newly created FPA3 POOL with buffers
|
|
* using a temporary AURA.
|
|
*/
|
|
static int cvmx_fpa3_pool_populate(cvmx_fpa3_pool_t pool, unsigned int buf_cnt,
|
|
unsigned int buf_sz, void *mem_ptr,
|
|
unsigned int mem_node)
|
|
{
|
|
cvmx_fpa3_poolx_info_t *pinfo;
|
|
cvmx_fpa3_gaura_t aura;
|
|
cvmx_fpa3_pool_t zero_pool;
|
|
cvmx_fpa_poolx_cfg_t pool_cfg;
|
|
cvmx_fpa_poolx_start_addr_t pool_start_reg;
|
|
cvmx_fpa_poolx_end_addr_t pool_end_reg;
|
|
cvmx_fpa_poolx_available_t avail_reg;
|
|
cvmx_fpa_poolx_threshold_t thresh_reg;
|
|
cvmx_fpa_poolx_int_t int_reg;
|
|
unsigned int block_size, align;
|
|
unsigned long long mem_size;
|
|
u64 paddr;
|
|
unsigned int i;
|
|
|
|
if (debug)
|
|
debug("%s: POOL %u:%u buf_sz=%u count=%d\n", __func__,
|
|
pool.node, pool.lpool, buf_sz, buf_cnt);
|
|
|
|
if (!__cvmx_fpa3_pool_valid(pool))
|
|
return -1;
|
|
|
|
zero_pool = __cvmx_fpa3_pool(pool.node, 0);
|
|
|
|
pool_cfg.u64 = csr_rd_node(pool.node, CVMX_FPA_POOLX_CFG(pool.lpool));
|
|
|
|
block_size = pool_cfg.cn78xx.buf_size << 7;
|
|
|
|
if (pool_cfg.cn78xx.nat_align) {
|
|
/* Assure block_size is legit */
|
|
if (block_size > (1 << 17)) {
|
|
printf("ERROR: %s: POOL %u:%u block size %u is not valid\n",
|
|
__func__, pool.node, pool.lpool, block_size);
|
|
return -1;
|
|
}
|
|
}
|
|
align = CVMX_CACHE_LINE_SIZE;
|
|
|
|
pinfo = cvmx_fpa3_pool_info[pool.node];
|
|
if (!pinfo)
|
|
return -1;
|
|
pinfo += pool.lpool;
|
|
|
|
if (pinfo->buf_size != block_size || block_size != buf_sz) {
|
|
printf("ERROR: %s: POOL %u:%u buffer size mismatch\n", __func__,
|
|
pool.node, pool.lpool);
|
|
return -1;
|
|
}
|
|
|
|
if (!mem_ptr) {
|
|
/* When allocating our own memory
|
|
* make sure at least 'buf_cnt' blocks
|
|
* will fit into it.
|
|
*/
|
|
mem_size = (long long)buf_cnt * block_size + (block_size - 128);
|
|
|
|
mem_ptr = cvmx_helper_mem_alloc(mem_node, mem_size, align);
|
|
|
|
if (!mem_ptr) {
|
|
printf("ERROR: %s: POOL %u:%u out of memory, could not allocate %llu bytes\n",
|
|
__func__, pool.node, pool.lpool, mem_size);
|
|
return -1;
|
|
}
|
|
|
|
/* Record memory base for use in shutdown */
|
|
pinfo->bufs_paddr = cvmx_ptr_to_phys(mem_ptr);
|
|
} else {
|
|
/* caller-allocated memory is sized simply, may reduce count */
|
|
mem_size = (long long)buf_cnt * block_size;
|
|
/* caller responsable to free this memory too */
|
|
}
|
|
|
|
/* Recalculate buf_cnt after possible alignment adjustment */
|
|
buf_cnt = mem_size / block_size;
|
|
|
|
/* Get temporary AURA */
|
|
aura = cvmx_fpa3_reserve_aura(pool.node, -1);
|
|
if (!__cvmx_fpa3_aura_valid(aura))
|
|
return -1;
|
|
|
|
/* Attach the temporary AURA to the POOL */
|
|
(void)cvmx_fpa3_aura_cfg(aura, pool, buf_cnt, buf_cnt + 1, 0);
|
|
|
|
/* Set AURA count to buffer count to avoid wrap-around */
|
|
csr_wr_node(aura.node, CVMX_FPA_AURAX_CNT(aura.laura), buf_cnt);
|
|
|
|
/* Set POOL threshold just above buf count so it does not misfire */
|
|
thresh_reg.u64 = 0;
|
|
thresh_reg.cn78xx.thresh = buf_cnt + 1;
|
|
csr_wr_node(pool.node, CVMX_FPA_POOLX_THRESHOLD(pool.lpool),
|
|
thresh_reg.u64);
|
|
|
|
/* Set buffer memory region bounds checking */
|
|
paddr = (cvmx_ptr_to_phys(mem_ptr) >> 7) << 7;
|
|
pool_start_reg.u64 = 0;
|
|
pool_end_reg.u64 = 0;
|
|
pool_start_reg.cn78xx.addr = paddr >> 7;
|
|
pool_end_reg.cn78xx.addr = (paddr + mem_size + 127) >> 7;
|
|
|
|
csr_wr_node(pool.node, CVMX_FPA_POOLX_START_ADDR(pool.lpool),
|
|
pool_start_reg.u64);
|
|
csr_wr_node(pool.node, CVMX_FPA_POOLX_END_ADDR(pool.lpool),
|
|
pool_end_reg.u64);
|
|
|
|
/* Make sure 'paddr' is divisible by 'block_size' */
|
|
i = (paddr % block_size);
|
|
if (i > 0) {
|
|
i = block_size - i;
|
|
paddr += i;
|
|
mem_size -= i;
|
|
}
|
|
|
|
/* The above alignment mimics how the FPA3 hardware
|
|
* aligns pointers to the buffer size, which only
|
|
* needs to be multiple of cache line size
|
|
*/
|
|
|
|
if (debug && paddr != cvmx_ptr_to_phys(mem_ptr))
|
|
debug("%s: pool mem paddr %#llx adjusted to %#llx for block size %#x\n",
|
|
__func__, CAST_ULL(cvmx_ptr_to_phys(mem_ptr)),
|
|
CAST_ULL(paddr), block_size);
|
|
|
|
for (i = 0; i < buf_cnt; i++) {
|
|
void *ptr = cvmx_phys_to_ptr(paddr);
|
|
|
|
cvmx_fpa3_free_nosync(ptr, aura, 0);
|
|
|
|
paddr += block_size;
|
|
|
|
if ((paddr + block_size - 1) >= (paddr + mem_size))
|
|
break;
|
|
}
|
|
|
|
if (debug && i < buf_cnt) {
|
|
debug("%s: buffer count reduced from %u to %u\n", __func__,
|
|
buf_cnt, i);
|
|
buf_cnt = i;
|
|
}
|
|
|
|
/* Wait for all buffers to reach the POOL before removing temp AURA */
|
|
do {
|
|
CVMX_SYNC;
|
|
avail_reg.u64 = csr_rd_node(
|
|
pool.node, CVMX_FPA_POOLX_AVAILABLE(pool.lpool));
|
|
} while (avail_reg.cn78xx.count < buf_cnt);
|
|
|
|
/* Detach the temporary AURA */
|
|
(void)cvmx_fpa3_aura_cfg(aura, zero_pool, 0, 0, 0);
|
|
|
|
/* Release temporary AURA */
|
|
(void)cvmx_fpa3_release_aura(aura);
|
|
|
|
/* Clear all POOL interrupts */
|
|
int_reg.u64 = 0;
|
|
int_reg.cn78xx.ovfls = 1;
|
|
int_reg.cn78xx.crcerr = 1;
|
|
int_reg.cn78xx.range = 1;
|
|
int_reg.cn78xx.thresh = 1;
|
|
csr_wr_node(pool.node, CVMX_FPA_POOLX_INT(pool.lpool), int_reg.u64);
|
|
|
|
/* Record buffer count for shutdown */
|
|
pinfo->buf_count = buf_cnt;
|
|
|
|
return buf_cnt;
|
|
}
|
|
|
|
/**
|
|
* @INTERNAL
|
|
*
|
|
* Fill a legacy FPA pool with buffers
|
|
*/
|
|
static int cvmx_fpa1_fill_pool(cvmx_fpa1_pool_t pool, int num_blocks,
|
|
void *buffer)
|
|
{
|
|
cvmx_fpa_poolx_start_addr_t pool_start_reg;
|
|
cvmx_fpa_poolx_end_addr_t pool_end_reg;
|
|
unsigned int block_size = cvmx_fpa_get_block_size(pool);
|
|
unsigned int mem_size;
|
|
char *bufp;
|
|
|
|
if ((unsigned int)pool >= CVMX_FPA1_NUM_POOLS)
|
|
return -1;
|
|
|
|
mem_size = block_size * num_blocks;
|
|
|
|
if (!buffer) {
|
|
buffer = cvmx_helper_mem_alloc(0, mem_size,
|
|
CVMX_CACHE_LINE_SIZE);
|
|
|
|
cvmx_fpa1_pool_info[pool].base_paddr = cvmx_ptr_to_phys(buffer);
|
|
} else {
|
|
/* Align user-supplied buffer to cache line size */
|
|
unsigned int off =
|
|
(CVMX_CACHE_LINE_SIZE - 1) & cvmx_ptr_to_phys(buffer);
|
|
if (off > 0) {
|
|
// buffer += CVMX_CACHE_LINE_SIZE - off;
|
|
buffer = (char *)buffer + CVMX_CACHE_LINE_SIZE - off;
|
|
mem_size -= CVMX_CACHE_LINE_SIZE - off;
|
|
num_blocks = mem_size / block_size;
|
|
}
|
|
}
|
|
|
|
if (debug)
|
|
debug("%s: memory at %p size %#x\n", __func__, buffer,
|
|
mem_size);
|
|
|
|
pool_start_reg.u64 = 0;
|
|
pool_end_reg.u64 = 0;
|
|
|
|
/* buffer pointer range checks are highly recommended, but optional */
|
|
pool_start_reg.cn61xx.addr = 1; /* catch NULL pointers */
|
|
pool_end_reg.cn61xx.addr = (1ull << (40 - 7)) - 1; /* max paddr */
|
|
if (!OCTEON_IS_MODEL(OCTEON_CN63XX)) {
|
|
csr_wr(CVMX_FPA_POOLX_START_ADDR(pool), pool_start_reg.u64);
|
|
csr_wr(CVMX_FPA_POOLX_END_ADDR(pool), pool_end_reg.u64);
|
|
}
|
|
|
|
bufp = (char *)buffer;
|
|
while (num_blocks--) {
|
|
cvmx_fpa1_free(bufp, pool, 0);
|
|
cvmx_fpa1_pool_info[pool].buffer_count++;
|
|
bufp += block_size;
|
|
}
|
|
return 0;
|
|
}
|
|
|
|
/**
|
|
* @INTERNAL
|
|
*
|
|
* Setup a legacy FPA pool
|
|
*/
|
|
static int cvmx_fpa1_pool_init(cvmx_fpa1_pool_t pool_id, int num_blocks,
|
|
int block_size, void *buffer)
|
|
{
|
|
int max_pool = cvmx_fpa_get_max_pools();
|
|
|
|
if (pool_id < 0 || pool_id >= max_pool) {
|
|
printf("ERROR: %s pool %d invalid\n", __func__, pool_id);
|
|
return -1;
|
|
}
|
|
|
|
if (!cvmx_fpa1_pool_info)
|
|
cvmx_fpa_global_init_node(0);
|
|
|
|
if (debug)
|
|
debug("%s: initializing info pool %d\n", __func__, pool_id);
|
|
|
|
cvmx_fpa1_pool_info[pool_id].size = block_size;
|
|
cvmx_fpa1_pool_info[pool_id].buffer_count = 0;
|
|
|
|
if (debug)
|
|
debug("%s: enabling unit for pool %d\n", __func__, pool_id);
|
|
|
|
return 0;
|
|
}
|
|
|
|
/**
|
|
* Initialize global configuration for FPA block for specified node.
|
|
*
|
|
* @param node is the node number
|
|
*
|
|
* @note this function sets the initial QoS averaging timing parameters,
|
|
* for the entire FPA unit (per node), which may be overridden on a
|
|
* per AURA basis.
|
|
*/
|
|
int cvmx_fpa_global_init_node(int node)
|
|
{
|
|
/* There are just the initial parameter values */
|
|
#define FPA_RED_AVG_DLY 1
|
|
#define FPA_RED_LVL_DLY 3
|
|
#define FPA_QOS_AVRG 0
|
|
/* Setting up avg_dly and prb_dly, enable bits */
|
|
if (octeon_has_feature(OCTEON_FEATURE_FPA3)) {
|
|
char pool_info_name[32] = "cvmx_fpa3_pools_";
|
|
char aura_info_name[32] = "cvmx_fpa3_auras_";
|
|
char ns[2] = "0";
|
|
|
|
ns[0] += node;
|
|
strcat(pool_info_name, ns);
|
|
strcat(aura_info_name, ns);
|
|
|
|
cvmx_fpa3_config_red_params(node, FPA_QOS_AVRG, FPA_RED_LVL_DLY,
|
|
FPA_RED_AVG_DLY);
|
|
|
|
/* Allocate the pinfo named block */
|
|
cvmx_fpa3_pool_info[node] = (cvmx_fpa3_poolx_info_t *)
|
|
cvmx_bootmem_alloc_named_range_once(
|
|
sizeof(cvmx_fpa3_pool_info[0][0]) *
|
|
cvmx_fpa3_num_pools(),
|
|
0, 0, 0, pool_info_name, NULL);
|
|
|
|
cvmx_fpa3_aura_info[node] = (cvmx_fpa3_aurax_info_t *)
|
|
cvmx_bootmem_alloc_named_range_once(
|
|
sizeof(cvmx_fpa3_aura_info[0][0]) *
|
|
cvmx_fpa3_num_auras(),
|
|
0, 0, 0, aura_info_name, NULL);
|
|
|
|
//XXX add allocation error check
|
|
|
|
/* Setup zero_pool on this node */
|
|
cvmx_fpa3_reserve_pool(node, 0);
|
|
cvmx_fpa3_pool_info[node][0].buf_count = 0;
|
|
} else {
|
|
char pool_info_name[32] = "cvmx_fpa_pool";
|
|
|
|
/* Allocate the pinfo named block */
|
|
cvmx_fpa1_pool_info = (cvmx_fpa1_pool_info_t *)
|
|
cvmx_bootmem_alloc_named_range_once(
|
|
sizeof(cvmx_fpa1_pool_info[0]) *
|
|
CVMX_FPA1_NUM_POOLS,
|
|
0, 0, 0, pool_info_name, NULL);
|
|
|
|
cvmx_fpa1_enable();
|
|
}
|
|
|
|
return 0;
|
|
}
|
|
|
|
static void __memset_u64(u64 *ptr, u64 pattern, unsigned int words)
|
|
{
|
|
while (words--)
|
|
*ptr++ = pattern;
|
|
}
|
|
|
|
/**
|
|
* @INTERNAL
|
|
* Initialize pool pointer-storage memory
|
|
*
|
|
* Unlike legacy FPA, which used free buffers to store pointers that
|
|
* exceed on-chip memory, FPA3 requires a dedicated memory buffer for
|
|
* free pointer stack back-store.
|
|
*
|
|
* @param pool - pool to initialize
|
|
* @param mem_node - if memory should be allocated from a different node
|
|
* @param max_buffer_cnt - maximum block capacity of pool
|
|
* @param align - buffer alignment mode,
|
|
* current FPA_NATURAL_ALIGNMENT is supported
|
|
* @param buffer_sz - size of buffers in pool
|
|
*/
|
|
static int cvmx_fpa3_pool_stack_init(cvmx_fpa3_pool_t pool,
|
|
unsigned int mem_node,
|
|
unsigned int max_buffer_cnt,
|
|
enum cvmx_fpa3_pool_alignment_e align,
|
|
unsigned int buffer_sz)
|
|
{
|
|
cvmx_fpa3_poolx_info_t *pinfo;
|
|
u64 stack_paddr;
|
|
void *mem_ptr;
|
|
unsigned int stack_memory_size;
|
|
cvmx_fpa_poolx_cfg_t pool_cfg;
|
|
cvmx_fpa_poolx_fpf_marks_t pool_fpf_marks;
|
|
|
|
if (debug)
|
|
debug("%s: POOL %u:%u bufsz=%u maxbuf=%u\n", __func__,
|
|
pool.node, pool.lpool, buffer_sz, max_buffer_cnt);
|
|
|
|
if (!__cvmx_fpa3_pool_valid(pool)) {
|
|
printf("ERROR: %s: POOL invalid\n", __func__);
|
|
return -1;
|
|
}
|
|
|
|
pinfo = cvmx_fpa3_pool_info[pool.node];
|
|
if (!pinfo) {
|
|
printf("ERROR: %s: FPA on node#%u is not initialized\n",
|
|
__func__, pool.node);
|
|
return -1;
|
|
}
|
|
pinfo += pool.lpool;
|
|
|
|
/* Calculate stack size based on buffer count with one line to spare */
|
|
stack_memory_size = (max_buffer_cnt * 128) / 29 + 128 + 127;
|
|
|
|
/* Increase stack size by band guard */
|
|
stack_memory_size += guard_band_size << 1;
|
|
|
|
/* Align size to cache line */
|
|
stack_memory_size = (stack_memory_size >> 7) << 7;
|
|
|
|
/* Allocate internal stack */
|
|
mem_ptr = cvmx_helper_mem_alloc(mem_node, stack_memory_size,
|
|
CVMX_CACHE_LINE_SIZE);
|
|
|
|
if (debug)
|
|
debug("%s: stack_mem=%u ptr=%p\n", __func__, stack_memory_size,
|
|
mem_ptr);
|
|
|
|
if (!mem_ptr) {
|
|
debug("ERROR: %sFailed to allocate stack for POOL %u:%u\n",
|
|
__func__, pool.node, pool.lpool);
|
|
return -1;
|
|
}
|
|
|
|
/* Initialize guard bands */
|
|
if (guard_band_size > 0) {
|
|
__memset_u64((u64 *)mem_ptr, magic_pattern,
|
|
guard_band_size >> 3);
|
|
__memset_u64((u64 *)((char *)mem_ptr + stack_memory_size -
|
|
guard_band_size),
|
|
magic_pattern, guard_band_size >> 3);
|
|
}
|
|
|
|
pinfo->stack_paddr = cvmx_ptr_to_phys(mem_ptr);
|
|
pinfo->stack_psize = stack_memory_size;
|
|
|
|
/* Calculate usable stack start */
|
|
stack_paddr = cvmx_ptr_to_phys((char *)mem_ptr + guard_band_size);
|
|
|
|
csr_wr_node(pool.node, CVMX_FPA_POOLX_STACK_BASE(pool.lpool),
|
|
stack_paddr);
|
|
csr_wr_node(pool.node, CVMX_FPA_POOLX_STACK_ADDR(pool.lpool),
|
|
stack_paddr);
|
|
|
|
/* Calculate usable stack end - start of last cache line */
|
|
stack_paddr = stack_paddr + stack_memory_size - (guard_band_size << 1);
|
|
|
|
csr_wr_node(pool.node, CVMX_FPA_POOLX_STACK_END(pool.lpool),
|
|
stack_paddr);
|
|
|
|
if (debug)
|
|
debug("%s: Stack paddr %#llx - %#llx\n", __func__,
|
|
CAST_ULL(csr_rd_node(pool.node, CVMX_FPA_POOLX_STACK_BASE(
|
|
pool.lpool))),
|
|
CAST_ULL(csr_rd_node(pool.node, CVMX_FPA_POOLX_STACK_END(
|
|
pool.lpool))));
|
|
|
|
/* Setup buffer size for this pool until it is shutdown */
|
|
pinfo->buf_size = buffer_sz;
|
|
|
|
pool_cfg.u64 = 0;
|
|
pool_cfg.cn78xx.buf_size = buffer_sz >> 7;
|
|
pool_cfg.cn78xx.l_type = 0x2;
|
|
pool_cfg.cn78xx.ena = 0;
|
|
if (align == FPA_NATURAL_ALIGNMENT)
|
|
pool_cfg.cn78xx.nat_align = 1;
|
|
|
|
/* FPA-26117, FPA-22443 */
|
|
pool_fpf_marks.u64 =
|
|
csr_rd_node(pool.node, CVMX_FPA_POOLX_FPF_MARKS(pool.lpool));
|
|
pool_fpf_marks.s.fpf_rd = 0x80;
|
|
csr_wr_node(pool.node, CVMX_FPA_POOLX_FPF_MARKS(pool.lpool),
|
|
pool_fpf_marks.u64);
|
|
|
|
csr_wr_node(pool.node, CVMX_FPA_POOLX_CFG(pool.lpool), pool_cfg.u64);
|
|
pool_cfg.cn78xx.ena = 1;
|
|
csr_wr_node(pool.node, CVMX_FPA_POOLX_CFG(pool.lpool), pool_cfg.u64);
|
|
|
|
/* Pool is now ready to be filled up */
|
|
return 0;
|
|
}
|
|
|
|
/**
|
|
* Create an FPA POOL and fill it up with buffers
|
|
*
|
|
* @param node is the node number for the pool and memory location
|
|
* @param desired_pool is the local pool number desired
|
|
* or -1 for first available
|
|
* @param name is the symbolic name to assign the POOL
|
|
* @param block_size is the size of all buffers held in this POOL
|
|
* @param num_blocks is the number of free buffers to fill into the POOL
|
|
* @param buffer is an optionally caller-supplied memory for the buffers
|
|
* or NULL to cause the buffer memory to be allocated automatically.
|
|
* @return the POOL handle
|
|
*
|
|
* Note: if the buffer memory is supplied by caller, the application
|
|
* will be responsable to free this memory.
|
|
*
|
|
* Only supported on CN78XX.
|
|
*/
|
|
cvmx_fpa3_pool_t cvmx_fpa3_setup_fill_pool(int node, int desired_pool,
|
|
const char *name,
|
|
unsigned int block_size,
|
|
unsigned int num_blocks,
|
|
void *buffer)
|
|
{
|
|
cvmx_fpa3_pool_t pool;
|
|
unsigned int mem_node;
|
|
int rc;
|
|
|
|
if (node < 0)
|
|
node = cvmx_get_node_num();
|
|
|
|
if (debug)
|
|
debug("%s: desired pool=%d bufsize=%u cnt=%u '%s'\n", __func__,
|
|
desired_pool, block_size, num_blocks, name);
|
|
|
|
/* Use memory from the node local to the AURA/POOL */
|
|
mem_node = node;
|
|
|
|
if (num_blocks == 0 || num_blocks > 1 << 30) {
|
|
printf("ERROR: %s: invalid block count %u\n", __func__,
|
|
num_blocks);
|
|
return CVMX_FPA3_INVALID_POOL;
|
|
}
|
|
|
|
/*
|
|
* Check for block size validity:
|
|
* With user-supplied buffer, can't increase block size,
|
|
* so make sure it is at least 128, and is aligned to 128
|
|
* For all cases make sure it is not too big
|
|
*/
|
|
if ((buffer && (block_size < CVMX_CACHE_LINE_SIZE ||
|
|
(block_size & (CVMX_CACHE_LINE_SIZE - 1)))) ||
|
|
(block_size > (1 << 17))) {
|
|
printf("ERROR: %s: invalid block size %u\n", __func__,
|
|
block_size);
|
|
return CVMX_FPA3_INVALID_POOL;
|
|
}
|
|
|
|
if (block_size < CVMX_CACHE_LINE_SIZE)
|
|
block_size = CVMX_CACHE_LINE_SIZE;
|
|
|
|
/* Reserve POOL */
|
|
pool = cvmx_fpa3_reserve_pool(node, desired_pool);
|
|
|
|
if (!__cvmx_fpa3_pool_valid(pool)) {
|
|
printf("ERROR: %s: POOL %u:%d not available\n", __func__, node,
|
|
desired_pool);
|
|
return CVMX_FPA3_INVALID_POOL;
|
|
}
|
|
|
|
/* Initialize POOL with stack storage */
|
|
rc = cvmx_fpa3_pool_stack_init(pool, mem_node, num_blocks,
|
|
FPA_NATURAL_ALIGNMENT, block_size);
|
|
if (rc < 0) {
|
|
printf("ERROR: %s: POOL %u:%u stack setup failed\n", __func__,
|
|
pool.node, pool.lpool);
|
|
cvmx_fpa3_release_pool(pool);
|
|
return CVMX_FPA3_INVALID_POOL;
|
|
}
|
|
|
|
/* Populate the POOL with buffers */
|
|
rc = cvmx_fpa3_pool_populate(pool, num_blocks, block_size, buffer,
|
|
mem_node);
|
|
if (rc < 0) {
|
|
printf("ERROR: %s: POOL %u:%u memory fill failed\n", __func__,
|
|
pool.node, pool.lpool);
|
|
cvmx_fpa3_release_pool(pool);
|
|
return CVMX_FPA3_INVALID_POOL;
|
|
}
|
|
|
|
cvmx_fpa3_set_pool_name(pool, name);
|
|
|
|
return pool;
|
|
}
|
|
|
|
/**
|
|
* Attach an AURA to an existing POOL
|
|
*
|
|
* @param pool is the handle of the POOL to be attached
|
|
* @param desired_aura is the number of the AURA resired
|
|
* or -1 for the AURA to be automatically assigned
|
|
* @param name is a symbolic name for the new AURA
|
|
* @param block_size is the size of all buffers that will be handed
|
|
* out by this AURA
|
|
* @param num_blocks is the maximum number of buffers that can be
|
|
* handed out by this AURA, and can not exceed the number
|
|
* of buffers filled into the attached POOL
|
|
* @return the AURA handle
|
|
*
|
|
* Only supported on CN78XX.
|
|
*/
|
|
cvmx_fpa3_gaura_t cvmx_fpa3_set_aura_for_pool(cvmx_fpa3_pool_t pool,
|
|
int desired_aura,
|
|
const char *name,
|
|
unsigned int block_size,
|
|
unsigned int num_blocks)
|
|
{
|
|
cvmx_fpa3_gaura_t aura;
|
|
cvmx_fpa_poolx_available_t avail_reg;
|
|
const char *emsg;
|
|
int rc;
|
|
|
|
if (debug)
|
|
debug("%s: aura=%d bufsize=%u cnt=%u '%s'\n", __func__,
|
|
desired_aura, block_size, num_blocks, name);
|
|
|
|
if (!__cvmx_fpa3_pool_valid(pool)) {
|
|
printf("ERROR: %s: POOL argument invalid\n", __func__);
|
|
return CVMX_FPA3_INVALID_GAURA;
|
|
}
|
|
|
|
/* Verify the AURA buffer count limit is not above POOL buffer count */
|
|
avail_reg.u64 =
|
|
csr_rd_node(pool.node, CVMX_FPA_POOLX_AVAILABLE(pool.lpool));
|
|
if (avail_reg.cn78xx.count < num_blocks) {
|
|
printf("WARNING: %s: AURA %u:%u buffer count limit %u reduced to POOL available count %u\n",
|
|
__func__, aura.node, aura.laura, num_blocks,
|
|
(unsigned int)avail_reg.cn78xx.count);
|
|
num_blocks = avail_reg.cn78xx.count;
|
|
}
|
|
|
|
/* Reserve an AURA number, follow desired number */
|
|
aura = cvmx_fpa3_reserve_aura(pool.node, desired_aura);
|
|
|
|
if (!__cvmx_fpa3_aura_valid(aura)) {
|
|
printf("ERROR: %s: AURA %u:%d not available\n", __func__,
|
|
pool.node, desired_aura);
|
|
return CVMX_FPA3_INVALID_GAURA;
|
|
}
|
|
|
|
/* Initialize AURA attached to the above POOL */
|
|
rc = cvmx_fpa3_aura_cfg(aura, pool, num_blocks, num_blocks + 1, 0);
|
|
if (rc < 0) {
|
|
emsg = "AURA configuration";
|
|
goto _fail;
|
|
}
|
|
|
|
cvmx_fpa3_set_aura_name(aura, name);
|
|
|
|
return aura;
|
|
|
|
_fail:
|
|
printf("ERROR: %s: %s\n", __func__, emsg);
|
|
cvmx_fpa3_release_aura(aura);
|
|
return CVMX_FPA3_INVALID_GAURA;
|
|
}
|
|
|
|
/**
|
|
* Create a combination of an AURA and a POOL
|
|
*
|
|
* @param node is the node number for the pool and memory location
|
|
* @param desired_aura is the number of the AURA resired
|
|
* or -1 for the AURA to be automatically assigned
|
|
* @param name is a symbolic name for the new AURA
|
|
* @param block_size is the size of all buffers that will be handed
|
|
* out by this AURA
|
|
* @param num_blocks is the maximum number of buffers that can be
|
|
* handed out by this AURA, and can not exceed the number
|
|
* of buffers filled into the attached POOL
|
|
* @param buffer is an optionally caller-supplied memory for the buffers
|
|
* or NULL to cause the buffer memory to be allocated automatically.
|
|
*
|
|
* @return the AURA handle
|
|
*
|
|
* Note: if the buffer memory is supplied by caller, the application
|
|
* will be responsable to free this memory.
|
|
* The POOL number is always automatically assigned.
|
|
*
|
|
* Only supported on CN78XX.
|
|
*/
|
|
cvmx_fpa3_gaura_t cvmx_fpa3_setup_aura_and_pool(int node, int desired_aura,
|
|
const char *name, void *buffer,
|
|
unsigned int block_size,
|
|
unsigned int num_blocks)
|
|
{
|
|
cvmx_fpa3_gaura_t aura = CVMX_FPA3_INVALID_GAURA;
|
|
cvmx_fpa3_pool_t pool = CVMX_FPA3_INVALID_POOL;
|
|
const char *emsg = "";
|
|
unsigned int mem_node;
|
|
int rc;
|
|
|
|
if (debug)
|
|
debug("%s: aura=%d size=%u cnt=%u '%s'\n", __func__,
|
|
desired_aura, block_size, num_blocks, name);
|
|
|
|
if (node < 0)
|
|
node = cvmx_get_node_num();
|
|
|
|
if (num_blocks == 0 || num_blocks > 1 << 30) {
|
|
printf("ERROR: %s: invalid block count %u\n", __func__,
|
|
num_blocks);
|
|
return CVMX_FPA3_INVALID_GAURA;
|
|
}
|
|
|
|
/* Use memory from the node local to the AURA/POOL */
|
|
mem_node = node;
|
|
|
|
/* Reserve an AURA number, follow desired number */
|
|
aura = cvmx_fpa3_reserve_aura(node, desired_aura);
|
|
|
|
if (!__cvmx_fpa3_aura_valid(aura)) {
|
|
emsg = "AURA not available";
|
|
goto _fail;
|
|
}
|
|
|
|
/* Reserve POOL dynamically to underpin this AURA */
|
|
pool = cvmx_fpa3_reserve_pool(node, -1);
|
|
|
|
if (!__cvmx_fpa3_pool_valid(pool)) {
|
|
emsg = "POOL not available";
|
|
goto _fail;
|
|
}
|
|
|
|
/*
|
|
* Check for block size validity:
|
|
* With user-supplied buffer, can't increase block size,
|
|
* so make sure it is at least 128, and is aligned to 128
|
|
* For all cases make sure it is not too big
|
|
*/
|
|
if ((buffer && (block_size < CVMX_CACHE_LINE_SIZE ||
|
|
(block_size & (CVMX_CACHE_LINE_SIZE - 1)))) ||
|
|
block_size > (1 << 17)) {
|
|
printf("ERROR: %s: invalid block size %u\n", __func__,
|
|
block_size);
|
|
emsg = "invalid block size";
|
|
goto _fail;
|
|
}
|
|
|
|
if (block_size < CVMX_CACHE_LINE_SIZE)
|
|
block_size = CVMX_CACHE_LINE_SIZE;
|
|
|
|
/* Initialize POOL with stack storage */
|
|
rc = cvmx_fpa3_pool_stack_init(pool, mem_node, num_blocks,
|
|
FPA_NATURAL_ALIGNMENT, block_size);
|
|
if (rc < 0) {
|
|
emsg = "POOL Stack setup";
|
|
goto _fail;
|
|
}
|
|
|
|
/* Populate the AURA/POOL with buffers */
|
|
rc = cvmx_fpa3_pool_populate(pool, num_blocks, block_size, buffer,
|
|
mem_node);
|
|
if (rc < 0) {
|
|
emsg = "POOL buffer memory";
|
|
goto _fail;
|
|
}
|
|
|
|
/* Initialize AURA attached to the above POOL */
|
|
rc = cvmx_fpa3_aura_cfg(aura, pool, num_blocks, num_blocks + 1, 0);
|
|
if (rc < 0) {
|
|
emsg = "AURA configuration";
|
|
goto _fail;
|
|
}
|
|
|
|
cvmx_fpa3_set_aura_name(aura, name);
|
|
cvmx_fpa3_set_pool_name(pool, name);
|
|
|
|
if (debug)
|
|
debug("%s: AURA %u:%u ready, avail=%lld\n", __func__, aura.node,
|
|
aura.laura, cvmx_fpa3_get_available(aura));
|
|
|
|
return aura;
|
|
|
|
_fail:
|
|
printf("ERROR: %s: Failed in %s\n", __func__, emsg);
|
|
/* These will silently fail if POOL/AURA is not valid */
|
|
cvmx_fpa3_release_aura(aura);
|
|
cvmx_fpa3_release_pool(pool);
|
|
return CVMX_FPA3_INVALID_GAURA;
|
|
}
|
|
|
|
/**
|
|
* Setup a legacy FPA pool
|
|
*
|
|
* @param desired_pool is the POOL number desired or -1 for automatic
|
|
* assignment
|
|
* @param name is the symbolic POOL name
|
|
* @param block_size is the size of all buffers held in this POOL
|
|
* @param num_blocks is the number of free buffers to fill into the POOL
|
|
* @param buffer is an optionally caller-supplied memory for the buffers
|
|
* or NULL to cause the buffer memory to be allocated automatically.
|
|
* @return pool number or -1 on error.
|
|
*
|
|
* Note: if the buffer memory is supplied by caller, the application
|
|
* will be responsable to free this memory.
|
|
*/
|
|
int cvmx_fpa1_setup_pool(int desired_pool, const char *name, void *buffer,
|
|
unsigned int block_size, unsigned int num_blocks)
|
|
{
|
|
cvmx_fpa1_pool_t pool = CVMX_FPA1_INVALID_POOL;
|
|
int rc;
|
|
|
|
if (debug)
|
|
debug("%s: desired pool %d, name '%s', mem %p size %u count %u\n",
|
|
__func__, desired_pool, name, buffer, block_size,
|
|
num_blocks);
|
|
|
|
/* Reserve desired pool or get one dynamically */
|
|
pool = cvmx_fpa1_reserve_pool(desired_pool);
|
|
|
|
/* Validate reserved pool, if successful */
|
|
if (pool < 0 || pool >= cvmx_fpa_get_max_pools()) {
|
|
/* global resources would have printed an error message here */
|
|
return CVMX_FPA1_INVALID_POOL;
|
|
}
|
|
|
|
/* Initialize the pool */
|
|
rc = cvmx_fpa1_pool_init(pool, num_blocks, block_size, buffer);
|
|
if (rc < 0) {
|
|
printf("ERROR: %s: failed pool %u init\n", __func__, pool);
|
|
cvmx_fpa1_release_pool(pool);
|
|
return CVMX_FPA1_INVALID_POOL;
|
|
}
|
|
|
|
rc = cvmx_fpa1_fill_pool(pool, num_blocks, buffer);
|
|
if (rc < 0) {
|
|
printf("ERROR: %s: failed pool %u memory\n", __func__, pool);
|
|
cvmx_fpa1_release_pool(pool);
|
|
return CVMX_FPA1_INVALID_POOL;
|
|
}
|
|
|
|
if (debug)
|
|
debug("%s: pool %d filled up\b", __func__, pool);
|
|
|
|
cvmx_fpa_set_name(pool, name);
|
|
return pool;
|
|
}
|
|
|
|
/**
|
|
* Setup an FPA pool with buffers
|
|
*
|
|
* @param pool is the POOL number desired or -1 for automatic assignment
|
|
* @param name is the symbolic POOL name
|
|
* @param buffer is an optionally caller-supplied memory for the buffers
|
|
* or NULL to cause the buffer memory to be allocated automatically.
|
|
* @param block_size is the size of all buffers held in this POOL
|
|
* @param num_blocks is the number of free buffers to fill into the POOL
|
|
* @param buffer is an optionally caller-supplied memory for the buffers
|
|
* or NULL to cause the buffer memory to be allocated automatically.
|
|
*
|
|
* @return pool number or -1 on error.
|
|
*
|
|
* Note: if the buffer memory is supplied by caller, the application
|
|
* will be responsable to free this memory.
|
|
* This function will work with CN78XX models in backward-compatible mode
|
|
*/
|
|
int cvmx_fpa_setup_pool(int pool, const char *name, void *buffer,
|
|
u64 block_size, u64 num_blocks)
|
|
{
|
|
if (octeon_has_feature(OCTEON_FEATURE_FPA3)) {
|
|
cvmx_fpa3_gaura_t aura;
|
|
|
|
aura = cvmx_fpa3_setup_aura_and_pool(-1, pool, name, buffer,
|
|
block_size, num_blocks);
|
|
if (!__cvmx_fpa3_aura_valid(aura))
|
|
return -1;
|
|
if (aura.laura >= CVMX_FPA1_NUM_POOLS && pool >= 0)
|
|
printf("WARNING: %s: AURA %u is out of range for backward-compatible operation\n",
|
|
__func__, aura.laura);
|
|
return aura.laura;
|
|
} else {
|
|
return cvmx_fpa1_setup_pool(pool, name, buffer, block_size,
|
|
num_blocks);
|
|
}
|
|
}
|