From: Anatoly Burakov <anatoly.burakov@intel.com>
To: dev@dpdk.org
Cc: andras.kovacs@ericsson.com, laszlo.vadkeri@ericsson.com,
keith.wiles@intel.com, benjamin.walker@intel.com,
bruce.richardson@intel.com, thomas@monjalon.net
Subject: [dpdk-dev] [PATCH 20/23] eal: make memzones use rte_fbarray
Date: Tue, 19 Dec 2017 11:04:51 +0000 [thread overview]
Message-ID: <69a29e4ac2822d0c4b1f6c599b428977b2b25505.1513680516.git.anatoly.burakov@intel.com> (raw)
In-Reply-To: <cover.1513680516.git.anatoly.burakov@intel.com>
In-Reply-To: <cover.1513680516.git.anatoly.burakov@intel.com>
We greatly expand memzone list, and it makes some operations faster.
Plus, it's there, so we might as well use it.
As part of this commit, a potential memory leak is fixed (when we
allocate a memzone but there's no room in config, we don't free it
back), and there's a compile fix for ENA driver.
Signed-off-by: Anatoly Burakov <anatoly.burakov@intel.com>
---
config/common_base | 2 +-
drivers/net/ena/ena_ethdev.c | 10 +-
lib/librte_eal/common/eal_common_memzone.c | 168 ++++++++++++++++------
lib/librte_eal/common/include/rte_eal_memconfig.h | 4 +-
4 files changed, 137 insertions(+), 47 deletions(-)
diff --git a/config/common_base b/config/common_base
index 9730d4c..cce464d 100644
--- a/config/common_base
+++ b/config/common_base
@@ -92,7 +92,7 @@ CONFIG_RTE_MAX_LCORE=128
CONFIG_RTE_MAX_NUMA_NODES=8
CONFIG_RTE_MAX_MEMSEG_LISTS=16
CONFIG_RTE_MAX_MEMSEG_PER_LIST=32768
-CONFIG_RTE_MAX_MEMZONE=2560
+CONFIG_RTE_MAX_MEMZONE=32768
CONFIG_RTE_MAX_TAILQ=32
CONFIG_RTE_ENABLE_ASSERT=n
CONFIG_RTE_LOG_LEVEL=RTE_LOG_INFO
diff --git a/drivers/net/ena/ena_ethdev.c b/drivers/net/ena/ena_ethdev.c
index 22db895..aa37cad 100644
--- a/drivers/net/ena/ena_ethdev.c
+++ b/drivers/net/ena/ena_ethdev.c
@@ -249,11 +249,15 @@ static const struct eth_dev_ops ena_dev_ops = {
static inline int ena_cpu_to_node(int cpu)
{
struct rte_config *config = rte_eal_get_configuration();
+ const struct rte_fbarray *arr = &config->mem_config->memzones;
+ const struct rte_memzone *mz;
- if (likely(cpu < RTE_MAX_MEMZONE))
- return config->mem_config->memzone[cpu].socket_id;
+ if (unlikely(cpu >= RTE_MAX_MEMZONE))
+ return NUMA_NO_NODE;
- return NUMA_NO_NODE;
+ mz = rte_fbarray_get(arr, cpu);
+
+ return mz->socket_id;
}
static inline void ena_rx_mbuf_prepare(struct rte_mbuf *mbuf,
diff --git a/lib/librte_eal/common/eal_common_memzone.c b/lib/librte_eal/common/eal_common_memzone.c
index a9a4bef..58a4f25 100644
--- a/lib/librte_eal/common/eal_common_memzone.c
+++ b/lib/librte_eal/common/eal_common_memzone.c
@@ -58,20 +58,23 @@ static inline const struct rte_memzone *
memzone_lookup_thread_unsafe(const char *name)
{
const struct rte_mem_config *mcfg;
+ const struct rte_fbarray *arr;
const struct rte_memzone *mz;
- unsigned i = 0;
+ int i = 0;
/* get pointer to global configuration */
mcfg = rte_eal_get_configuration()->mem_config;
+ arr = &mcfg->memzones;
/*
* the algorithm is not optimal (linear), but there are few
* zones and this function should be called at init only
*/
- for (i = 0; i < RTE_MAX_MEMZONE; i++) {
- mz = &mcfg->memzone[i];
- if (mz->addr != NULL && !strncmp(name, mz->name, RTE_MEMZONE_NAMESIZE))
- return &mcfg->memzone[i];
+ while ((i = rte_fbarray_find_next_used(arr, i)) >= 0) {
+ mz = rte_fbarray_get(arr, i++);
+ if (mz->addr != NULL &&
+ !strncmp(name, mz->name, RTE_MEMZONE_NAMESIZE))
+ return mz;
}
return NULL;
@@ -81,17 +84,44 @@ static inline struct rte_memzone *
get_next_free_memzone(void)
{
struct rte_mem_config *mcfg;
- unsigned i = 0;
+ struct rte_fbarray *arr;
+ int i = 0;
/* get pointer to global configuration */
mcfg = rte_eal_get_configuration()->mem_config;
+ arr = &mcfg->memzones;
+
+ i = rte_fbarray_find_next_free(arr, 0);
+ if (i < 0) {
+ /* no space in config, so try expanding the list */
+ int old_len = arr->len;
+ int new_len = old_len * 2;
+ new_len = RTE_MIN(new_len, arr->capacity);
+
+ if (old_len == new_len) {
+ /* can't expand, the list is full */
+ RTE_LOG(ERR, EAL, "%s(): no space in memzone list\n",
+ __func__);
+ return NULL;
+ }
- for (i = 0; i < RTE_MAX_MEMZONE; i++) {
- if (mcfg->memzone[i].addr == NULL)
- return &mcfg->memzone[i];
- }
+ if (rte_fbarray_resize(arr, new_len)) {
+ RTE_LOG(ERR, EAL, "%s(): can't resize memzone list\n",
+ __func__);
+ return NULL;
+ }
- return NULL;
+ /* ensure we have free space */
+ i = rte_fbarray_find_next_free(arr, old_len);
+
+ if (i < 0) {
+ RTE_LOG(ERR, EAL, "%s(): Cannot find room in config!\n",
+ __func__);
+ return NULL;
+ }
+ }
+ rte_fbarray_set_used(arr, i, true);
+ return rte_fbarray_get(arr, i);
}
/* This function will return the greatest free block if a heap has been
@@ -132,14 +162,16 @@ memzone_reserve_aligned_thread_unsafe(const char *name, size_t len,
{
struct rte_memzone *mz;
struct rte_mem_config *mcfg;
+ struct rte_fbarray *arr;
size_t requested_len;
int socket;
/* get pointer to global configuration */
mcfg = rte_eal_get_configuration()->mem_config;
+ arr = &mcfg->memzones;
/* no more room in config */
- if (mcfg->memzone_cnt >= RTE_MAX_MEMZONE) {
+ if (arr->count >= arr->capacity) {
RTE_LOG(ERR, EAL, "%s(): No more room in config\n", __func__);
rte_errno = ENOSPC;
return NULL;
@@ -231,19 +263,19 @@ memzone_reserve_aligned_thread_unsafe(const char *name, size_t len,
return NULL;
}
- const struct malloc_elem *elem = malloc_elem_from_data(mz_addr);
+ struct malloc_elem *elem = malloc_elem_from_data(mz_addr);
/* fill the zone in config */
mz = get_next_free_memzone();
if (mz == NULL) {
- RTE_LOG(ERR, EAL, "%s(): Cannot find free memzone but there is room "
- "in config!\n", __func__);
+ RTE_LOG(ERR, EAL, "%s(): Cannot find free memzone but there is room in config!\n",
+ __func__);
rte_errno = ENOSPC;
+ malloc_heap_free(elem);
return NULL;
}
- mcfg->memzone_cnt++;
snprintf(mz->name, sizeof(mz->name), "%s", name);
mz->iova = rte_malloc_virt2iova(mz_addr);
mz->addr = mz_addr;
@@ -356,6 +388,8 @@ int
rte_memzone_free(const struct rte_memzone *mz)
{
struct rte_mem_config *mcfg;
+ struct rte_fbarray *arr;
+ struct rte_memzone *found_mz;
int ret = 0;
void *addr;
unsigned idx;
@@ -364,21 +398,22 @@ rte_memzone_free(const struct rte_memzone *mz)
return -EINVAL;
mcfg = rte_eal_get_configuration()->mem_config;
+ arr = &mcfg->memzones;
rte_rwlock_write_lock(&mcfg->mlock);
- idx = ((uintptr_t)mz - (uintptr_t)mcfg->memzone);
- idx = idx / sizeof(struct rte_memzone);
+ idx = rte_fbarray_find_idx(arr, mz);
+ found_mz = rte_fbarray_get(arr, idx);
- addr = mcfg->memzone[idx].addr;
+ addr = found_mz->addr;
if (addr == NULL)
ret = -EINVAL;
- else if (mcfg->memzone_cnt == 0) {
+ else if (arr->count == 0) {
rte_panic("%s(): memzone address not NULL but memzone_cnt is 0!\n",
__func__);
} else {
- memset(&mcfg->memzone[idx], 0, sizeof(mcfg->memzone[idx]));
- mcfg->memzone_cnt--;
+ memset(found_mz, 0, sizeof(*found_mz));
+ rte_fbarray_set_used(arr, idx, false);
}
rte_rwlock_write_unlock(&mcfg->mlock);
@@ -412,25 +447,71 @@ rte_memzone_lookup(const char *name)
void
rte_memzone_dump(FILE *f)
{
+ struct rte_fbarray *arr;
struct rte_mem_config *mcfg;
- unsigned i = 0;
+ int i = 0;
/* get pointer to global configuration */
mcfg = rte_eal_get_configuration()->mem_config;
+ arr = &mcfg->memzones;
rte_rwlock_read_lock(&mcfg->mlock);
/* dump all zones */
- for (i=0; i<RTE_MAX_MEMZONE; i++) {
- if (mcfg->memzone[i].addr == NULL)
- break;
- fprintf(f, "Zone %u: name:<%s>, IO:0x%"PRIx64", len:0x%zx"
+ while ((i = rte_fbarray_find_next_used(arr, i)) >= 0) {
+ void *cur_addr, *mz_end;
+ struct rte_memzone *mz;
+ struct rte_memseg_list *msl = NULL;
+ struct rte_memseg *ms;
+ int msl_idx, ms_idx;
+
+ mz = rte_fbarray_get(arr, i);
+
+ /*
+ * memzones can span multiple physical pages, so dump addresses
+ * of all physical pages this memzone spans.
+ */
+
+ fprintf(f, "Zone %u: name:<%s>, len:0x%zx"
", virt:%p, socket_id:%"PRId32", flags:%"PRIx32"\n", i,
- mcfg->memzone[i].name,
- mcfg->memzone[i].iova,
- mcfg->memzone[i].len,
- mcfg->memzone[i].addr,
- mcfg->memzone[i].socket_id,
- mcfg->memzone[i].flags);
+ mz->name,
+ mz->len,
+ mz->addr,
+ mz->socket_id,
+ mz->flags);
+
+ /* get pointer to appropriate memseg list */
+ for (msl_idx = 0; msl_idx < RTE_MAX_MEMSEG_LISTS; msl_idx++) {
+ if (mcfg->memsegs[msl_idx].hugepage_sz != mz->hugepage_sz)
+ continue;
+ if (mcfg->memsegs[msl_idx].socket_id != mz->socket_id)
+ continue;
+ msl = &mcfg->memsegs[msl_idx];
+ break;
+ }
+ if (!msl) {
+ RTE_LOG(DEBUG, EAL, "Skipping bad memzone\n");
+ continue;
+ }
+
+ cur_addr = RTE_PTR_ALIGN_FLOOR(mz->addr, mz->hugepage_sz);
+ mz_end = RTE_PTR_ADD(cur_addr, mz->len);
+
+ ms_idx = RTE_PTR_DIFF(mz->addr, msl->base_va) /
+ msl->hugepage_sz;
+ ms = rte_fbarray_get(&msl->memseg_arr, ms_idx);
+
+ fprintf(f, "physical pages used:\n");
+ do {
+ fprintf(f, " addr: %p iova: 0x%" PRIx64 " len: 0x%" PRIx64 " len: 0x%" PRIx64 "\n",
+ cur_addr, ms->iova, ms->len, ms->hugepage_sz);
+
+ /* advance VA to next page */
+ cur_addr = RTE_PTR_ADD(cur_addr, ms->hugepage_sz);
+
+ /* memzones occupy contiguous segments */
+ ++ms;
+ } while (cur_addr < mz_end);
+ i++;
}
rte_rwlock_read_unlock(&mcfg->mlock);
}
@@ -459,9 +540,11 @@ rte_eal_memzone_init(void)
rte_rwlock_write_lock(&mcfg->mlock);
- /* delete all zones */
- mcfg->memzone_cnt = 0;
- memset(mcfg->memzone, 0, sizeof(mcfg->memzone));
+ if (rte_fbarray_alloc(&mcfg->memzones, "memzone", 256,
+ RTE_MAX_MEMZONE, sizeof(struct rte_memzone))) {
+ RTE_LOG(ERR, EAL, "Cannot allocate memzone list\n");
+ return -1;
+ }
rte_rwlock_write_unlock(&mcfg->mlock);
@@ -473,14 +556,19 @@ void rte_memzone_walk(void (*func)(const struct rte_memzone *, void *),
void *arg)
{
struct rte_mem_config *mcfg;
- unsigned i;
+ struct rte_fbarray *arr;
+ int i;
mcfg = rte_eal_get_configuration()->mem_config;
+ arr = &mcfg->memzones;
+
+ i = 0;
rte_rwlock_read_lock(&mcfg->mlock);
- for (i=0; i<RTE_MAX_MEMZONE; i++) {
- if (mcfg->memzone[i].addr != NULL)
- (*func)(&mcfg->memzone[i], arg);
+ while ((i = rte_fbarray_find_next_used(arr, i)) > 0) {
+ struct rte_memzone *mz = rte_fbarray_get(arr, i);
+ (*func)(mz, arg);
+ i++;
}
rte_rwlock_read_unlock(&mcfg->mlock);
}
diff --git a/lib/librte_eal/common/include/rte_eal_memconfig.h b/lib/librte_eal/common/include/rte_eal_memconfig.h
index c9b57a4..8f4cc34 100644
--- a/lib/librte_eal/common/include/rte_eal_memconfig.h
+++ b/lib/librte_eal/common/include/rte_eal_memconfig.h
@@ -86,10 +86,8 @@ struct rte_mem_config {
rte_rwlock_t qlock; /**< used for tailq operation for thread safe. */
rte_rwlock_t mplock; /**< only used by mempool LIB for thread-safe. */
- uint32_t memzone_cnt; /**< Number of allocated memzones */
-
/* memory segments and zones */
- struct rte_memzone memzone[RTE_MAX_MEMZONE]; /**< Memzone descriptors. */
+ struct rte_fbarray memzones; /**< Memzone descriptors. */
struct rte_memseg_list memsegs[RTE_MAX_MEMSEG_LISTS];
/**< list of dynamic arrays holding memsegs */
--
2.7.4
next prev parent reply other threads:[~2017-12-19 11:05 UTC|newest]
Thread overview: 44+ messages / expand[flat|nested] mbox.gz Atom feed top
2017-12-19 11:04 [dpdk-dev] [PATCH 00/23] Dynamic memory allocation for DPDK Anatoly Burakov
2017-12-19 11:04 ` [dpdk-dev] [PATCH 01/23] eal: move get_virtual_area out of linuxapp eal_memory.c Anatoly Burakov
2017-12-19 11:04 ` [dpdk-dev] [PATCH 02/23] eal: add function to report number of detected sockets Anatoly Burakov
2017-12-19 11:04 ` [dpdk-dev] [PATCH 03/23] eal: add rte_fbarray Anatoly Burakov
2017-12-19 11:04 ` [dpdk-dev] [PATCH 04/23] eal: move all locking to heap Anatoly Burakov
2017-12-19 11:04 ` [dpdk-dev] [PATCH 05/23] eal: protect malloc heap stats with a lock Anatoly Burakov
2017-12-19 11:04 ` [dpdk-dev] [PATCH 06/23] eal: make malloc a doubly-linked list Anatoly Burakov
2017-12-19 11:04 ` [dpdk-dev] [PATCH 07/23] eal: make malloc_elem_join_adjacent_free public Anatoly Burakov
2017-12-19 11:04 ` [dpdk-dev] [PATCH 08/23] eal: add "single file segments" command-line option Anatoly Burakov
2017-12-19 11:04 ` [dpdk-dev] [PATCH 09/23] eal: add "legacy memory" option Anatoly Burakov
2017-12-19 11:04 ` [dpdk-dev] [PATCH 10/23] eal: read hugepage counts from node-specific sysfs path Anatoly Burakov
2017-12-19 11:04 ` [dpdk-dev] [PATCH 11/23] eal: replace memseg with memseg lists Anatoly Burakov
2017-12-19 11:04 ` [dpdk-dev] [PATCH 12/23] eal: add support for dynamic memory allocation Anatoly Burakov
2017-12-19 11:04 ` [dpdk-dev] [PATCH 13/23] eal: make use of dynamic memory allocation for init Anatoly Burakov
2017-12-19 11:04 ` [dpdk-dev] [PATCH 14/23] eal: add support for dynamic unmapping of pages Anatoly Burakov
2017-12-19 11:04 ` [dpdk-dev] [PATCH 15/23] eal: add API to check if memory is physically contiguous Anatoly Burakov
2017-12-19 11:04 ` [dpdk-dev] [PATCH 16/23] eal: enable dynamic memory allocation/free on malloc/free Anatoly Burakov
2017-12-19 11:04 ` [dpdk-dev] [PATCH 17/23] eal: add backend support for contiguous memory allocation Anatoly Burakov
2017-12-19 11:04 ` [dpdk-dev] [PATCH 18/23] eal: add rte_malloc support for allocating contiguous memory Anatoly Burakov
2017-12-19 11:04 ` [dpdk-dev] [PATCH 19/23] eal: enable reserving physically contiguous memzones Anatoly Burakov
[not found] ` <cover.1513594170.git.anatoly.burakov@intel.com>
2017-12-19 11:04 ` [dpdk-dev] [RFC 01/23] eal/memory: move get_virtual_area out of linuxapp eal_memory.c Anatoly Burakov
2017-12-19 11:15 ` Burakov, Anatoly
2017-12-19 11:04 ` [dpdk-dev] [RFC 02/23] eal/lcore: add function to report number of detected sockets Anatoly Burakov
2017-12-19 11:04 ` [dpdk-dev] [RFC 04/23] eal/malloc: move all locking to heap Anatoly Burakov
2017-12-19 11:04 ` [dpdk-dev] [RFC 05/23] eal/malloc: protect malloc heap stats with a lock Anatoly Burakov
2017-12-19 11:04 ` [dpdk-dev] [RFC 06/23] eal/malloc: make malloc a doubly-linked list Anatoly Burakov
2017-12-19 11:04 ` [dpdk-dev] [RFC 07/23] eal/malloc: make malloc_elem_join_adjacent_free public Anatoly Burakov
2017-12-19 11:04 ` [dpdk-dev] [RFC 10/23] eal: populate hugepage counts from socket-specific sysfs path Anatoly Burakov
2017-12-19 11:04 ` [dpdk-dev] [RFC 12/23] eal/memalloc: add support for dynamic memory allocation Anatoly Burakov
2017-12-19 11:04 ` [dpdk-dev] [RFC 13/23] eal/memory: make use of dynamic memory allocation for init Anatoly Burakov
2017-12-19 11:04 ` [dpdk-dev] [RFC 14/23] eal/memory: add support for dynamic unmapping of pages Anatoly Burakov
2017-12-19 11:04 ` [dpdk-dev] [RFC 15/23] eal/memalloc: add function to check if memory is physically contiguous Anatoly Burakov
2017-12-19 11:04 ` [dpdk-dev] [RFC 16/23] eal/malloc: enable dynamic memory allocation/free on malloc/free Anatoly Burakov
2017-12-19 11:04 ` [dpdk-dev] [RFC 17/23] eal/malloc: add backend support for contiguous memory allocation Anatoly Burakov
2017-12-19 11:04 ` [dpdk-dev] [RFC 18/23] eal/malloc: add rte_malloc support for allocating contiguous memory Anatoly Burakov
2017-12-19 11:04 ` [dpdk-dev] [RFC 19/23] eal/memzone: add support for reserving physically contiguous memzones Anatoly Burakov
2017-12-19 11:04 ` [dpdk-dev] [RFC 20/23] eal/memzone: make memzones use rte_fbarray Anatoly Burakov
2017-12-19 11:04 ` [dpdk-dev] [RFC 21/23] lib/mempool: add support for the new memory allocation methods Anatoly Burakov
2017-12-19 11:04 ` [dpdk-dev] [RFC 23/23] eal/memalloc: register/unregister memory with VFIO when alloc/free pages Anatoly Burakov
2017-12-19 11:04 ` Anatoly Burakov [this message]
2017-12-19 11:04 ` [dpdk-dev] [PATCH 21/23] mempool: add support for the new memory allocation methods Anatoly Burakov
2017-12-19 11:04 ` [dpdk-dev] [PATCH 22/23] vfio: allow to map other memory regions Anatoly Burakov
2017-12-19 11:04 ` [dpdk-dev] [PATCH 23/23] eal: map/unmap memory with VFIO when alloc/free pages Anatoly Burakov
2017-12-19 11:15 ` [dpdk-dev] [PATCH 00/23] Dynamic memory allocation for DPDK Burakov, Anatoly
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=69a29e4ac2822d0c4b1f6c599b428977b2b25505.1513680516.git.anatoly.burakov@intel.com \
--to=anatoly.burakov@intel.com \
--cc=andras.kovacs@ericsson.com \
--cc=benjamin.walker@intel.com \
--cc=bruce.richardson@intel.com \
--cc=dev@dpdk.org \
--cc=keith.wiles@intel.com \
--cc=laszlo.vadkeri@ericsson.com \
--cc=thomas@monjalon.net \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).