[dpdk-dev] [PATCH v9 02/21] mem: allow memseg lists to be marked as external

[Anatoly Burakov <anatoly.burakov@intel.com>]

When we allocate and use DPDK memory, we need to be able to
differentiate between DPDK hugepage segments and segments that
were made part of DPDK but are externally allocated. Add such
a property to memseg lists.

This breaks the ABI, so document the change in release notes.
This also breaks a few internal assumptions about memory
contiguousness, so adjust malloc code in a few places.

All current calls for memseg walk functions were adjusted to
ignore external segments where it made sense.

Mempools is a special case, because we may be asked to allocate
a mempool on a specific socket, and we need to ignore all page
sizes on other heaps or other sockets. Previously, this
assumption of knowing all page sizes was not a problem, but it
will be now, so we have to match socket ID with page size when
calculating minimum page size for a mempool.

Signed-off-by: Anatoly Burakov <anatoly.burakov@intel.com>
Acked-by: Andrew Rybchenko <arybchenko@solarflare.com>
Acked-by: Yongseok Koh <yskoh@mellanox.com>
---

Notes:
    v3:
    - Add comment to explain the process of picking up minimum
      page sizes for mempool
    
    v2:
    - Add documentation changes and ABI break
    
    v1:
    - Adjust all calls to memseg walk functions to ignore external
      segments where it made sense to do so

 doc/guides/rel_notes/deprecation.rst          | 15 --------
 doc/guides/rel_notes/release_18_11.rst        |  8 +++++
 drivers/bus/fslmc/fslmc_vfio.c                |  6 +++-
 drivers/net/mlx5/mlx5.c                       |  4 ++-
 drivers/net/virtio/virtio_user/vhost_kernel.c |  3 ++
 lib/librte_eal/bsdapp/eal/eal.c               |  3 ++
 lib/librte_eal/bsdapp/eal/eal_memory.c        |  7 ++--
 lib/librte_eal/common/eal_common_memory.c     |  3 ++
 .../common/include/rte_eal_memconfig.h        |  1 +
 lib/librte_eal/common/include/rte_memory.h    |  9 +++++
 lib/librte_eal/common/malloc_elem.c           | 10 ++++--
 lib/librte_eal/common/malloc_heap.c           |  9 +++--
 lib/librte_eal/common/rte_malloc.c            |  2 +-
 lib/librte_eal/linuxapp/eal/eal.c             | 10 +++++-
 lib/librte_eal/linuxapp/eal/eal_memalloc.c    |  9 +++++
 lib/librte_eal/linuxapp/eal/eal_vfio.c        | 17 ++++++---
 lib/librte_mempool/rte_mempool.c              | 35 ++++++++++++++-----
 test/test/test_malloc.c                       |  3 ++
 test/test/test_memzone.c                      |  3 ++
 19 files changed, 119 insertions(+), 38 deletions(-)

diff --git a/doc/guides/rel_notes/deprecation.rst b/doc/guides/rel_notes/deprecation.rst
index 138335dfb..d2aec64d1 100644
--- a/doc/guides/rel_notes/deprecation.rst
+++ b/doc/guides/rel_notes/deprecation.rst
@@ -11,21 +11,6 @@ API and ABI deprecation notices are to be posted here.
 Deprecation Notices
 -------------------
 
-* eal: certain structures will change in EAL on account of upcoming external
-  memory support. Aside from internal changes leading to an ABI break, the
-  following externally visible changes will also be implemented:
-
-  - ``rte_memseg_list`` will change to include a boolean flag indicating
-    whether a particular memseg list is externally allocated. This will have
-    implications for any users of memseg-walk-related functions, as they will
-    now have to skip externally allocated segments in most cases if the intent
-    is to only iterate over internal DPDK memory.
-  - ``socket_id`` parameter across the entire DPDK will gain additional meaning,
-    as some socket ID's will now be representing externally allocated memory. No
-    changes will be required for existing code as backwards compatibility will
-    be kept, and those who do not use this feature will not see these extra
-    socket ID's.
-
 * eal: both declaring and identifying devices will be streamlined in v18.11.
   New functions will appear to query a specific port from buses, classes of
   device and device drivers. Device declaration will be made coherent with the
diff --git a/doc/guides/rel_notes/release_18_11.rst b/doc/guides/rel_notes/release_18_11.rst
index 58bb79022..bc1d56130 100644
--- a/doc/guides/rel_notes/release_18_11.rst
+++ b/doc/guides/rel_notes/release_18_11.rst
@@ -118,6 +118,12 @@ API Changes
    Also, make sure to start the actual text at the margin.
    =========================================================
 
+* eal: ``rte_memseg_list`` structure now has an additional flag indicating
+  whether the memseg list is externally allocated. This will have implications
+  for any users of memseg-walk-related functions, as they will now have to skip
+  externally allocated segments in most cases if the intent is to only iterate
+  over internal DPDK memory.
+
 * mbuf: The ``__rte_mbuf_raw_free()`` and ``__rte_pktmbuf_prefree_seg()``
   functions were deprecated since 17.05 and are replaced by
   ``rte_mbuf_raw_free()`` and ``rte_pktmbuf_prefree_seg()``.
@@ -157,6 +163,8 @@ ABI Changes
        supporting external memory in DPDK:
          - structure ``rte_memseg_list`` now has a new field indicating length
            of memory addressed by the segment list
+         - structure ``rte_memseg_list`` now has a new flag indicating whether
+           the memseg list refers to external memory
 
 
 Removed Items
diff --git a/drivers/bus/fslmc/fslmc_vfio.c b/drivers/bus/fslmc/fslmc_vfio.c
index 4c2cd2a87..cb33dd891 100644
--- a/drivers/bus/fslmc/fslmc_vfio.c
+++ b/drivers/bus/fslmc/fslmc_vfio.c
@@ -318,11 +318,15 @@ fslmc_unmap_dma(uint64_t vaddr, uint64_t iovaddr __rte_unused, size_t len)
 
 static int
 fslmc_dmamap_seg(const struct rte_memseg_list *msl __rte_unused,
-		 const struct rte_memseg *ms, void *arg)
+		const struct rte_memseg *ms, void *arg)
 {
 	int *n_segs = arg;
 	int ret;
 
+	/* if IOVA address is invalid, skip */
+	if (ms->iova == RTE_BAD_IOVA)
+		return 0;
+
 	ret = fslmc_map_dma(ms->addr_64, ms->iova, ms->len);
 	if (ret)
 		DPAA2_BUS_ERR("Unable to VFIO map (addr=%p, len=%zu)",
diff --git a/drivers/net/mlx5/mlx5.c b/drivers/net/mlx5/mlx5.c
index fd89e2af3..af4a78ce9 100644
--- a/drivers/net/mlx5/mlx5.c
+++ b/drivers/net/mlx5/mlx5.c
@@ -568,11 +568,13 @@ static struct rte_pci_driver mlx5_driver;
 static void *uar_base;
 
 static int
-find_lower_va_bound(const struct rte_memseg_list *msl __rte_unused,
+find_lower_va_bound(const struct rte_memseg_list *msl,
 		const struct rte_memseg *ms, void *arg)
 {
 	void **addr = arg;
 
+	if (msl->external)
+		return 0;
 	if (*addr == NULL)
 		*addr = ms->addr;
 	else
diff --git a/drivers/net/virtio/virtio_user/vhost_kernel.c b/drivers/net/virtio/virtio_user/vhost_kernel.c
index b3bfcb76f..990ce80ce 100644
--- a/drivers/net/virtio/virtio_user/vhost_kernel.c
+++ b/drivers/net/virtio/virtio_user/vhost_kernel.c
@@ -78,6 +78,9 @@ add_memseg_list(const struct rte_memseg_list *msl, void *arg)
 	void *start_addr;
 	uint64_t len;
 
+	if (msl->external)
+		return 0;
+
 	if (vm->nregions >= max_regions)
 		return -1;
 
diff --git a/lib/librte_eal/bsdapp/eal/eal.c b/lib/librte_eal/bsdapp/eal/eal.c
index d7ae9d686..7735194a3 100644
--- a/lib/librte_eal/bsdapp/eal/eal.c
+++ b/lib/librte_eal/bsdapp/eal/eal.c
@@ -502,6 +502,9 @@ check_socket(const struct rte_memseg_list *msl, void *arg)
 {
 	int *socket_id = arg;
 
+	if (msl->external)
+		return 0;
+
 	if (msl->socket_id == *socket_id && msl->memseg_arr.count != 0)
 		return 1;
 
diff --git a/lib/librte_eal/bsdapp/eal/eal_memory.c b/lib/librte_eal/bsdapp/eal/eal_memory.c
index 65ea670f9..4b092e1f2 100644
--- a/lib/librte_eal/bsdapp/eal/eal_memory.c
+++ b/lib/librte_eal/bsdapp/eal/eal_memory.c
@@ -236,12 +236,15 @@ struct attach_walk_args {
 	int seg_idx;
 };
 static int
-attach_segment(const struct rte_memseg_list *msl __rte_unused,
-		const struct rte_memseg *ms, void *arg)
+attach_segment(const struct rte_memseg_list *msl, const struct rte_memseg *ms,
+		void *arg)
 {
 	struct attach_walk_args *wa = arg;
 	void *addr;
 
+	if (msl->external)
+		return 0;
+
 	addr = mmap(ms->addr, ms->len, PROT_READ | PROT_WRITE,
 			MAP_SHARED | MAP_FIXED, wa->fd_hugepage,
 			wa->seg_idx * EAL_PAGE_SIZE);
diff --git a/lib/librte_eal/common/eal_common_memory.c b/lib/librte_eal/common/eal_common_memory.c
index 30d018209..a2461ed79 100644
--- a/lib/librte_eal/common/eal_common_memory.c
+++ b/lib/librte_eal/common/eal_common_memory.c
@@ -272,6 +272,9 @@ physmem_size(const struct rte_memseg_list *msl, void *arg)
 {
 	uint64_t *total_len = arg;
 
+	if (msl->external)
+		return 0;
+
 	*total_len += msl->memseg_arr.count * msl->page_sz;
 
 	return 0;
diff --git a/lib/librte_eal/common/include/rte_eal_memconfig.h b/lib/librte_eal/common/include/rte_eal_memconfig.h
index 1d2362985..645288b02 100644
--- a/lib/librte_eal/common/include/rte_eal_memconfig.h
+++ b/lib/librte_eal/common/include/rte_eal_memconfig.h
@@ -34,6 +34,7 @@ struct rte_memseg_list {
 	int socket_id; /**< Socket ID for all memsegs in this list. */
 	volatile uint32_t version; /**< version number for multiprocess sync. */
 	size_t len; /**< Length of memory area covered by this memseg list. */
+	unsigned int external; /**< 1 if this list points to external memory */
 	struct rte_fbarray memseg_arr;
 };
 
diff --git a/lib/librte_eal/common/include/rte_memory.h b/lib/librte_eal/common/include/rte_memory.h
index 14bd277a4..ffdd56bfb 100644
--- a/lib/librte_eal/common/include/rte_memory.h
+++ b/lib/librte_eal/common/include/rte_memory.h
@@ -215,6 +215,9 @@ typedef int (*rte_memseg_list_walk_t)(const struct rte_memseg_list *msl,
  * @note This function read-locks the memory hotplug subsystem, and thus cannot
  *       be used within memory-related callback functions.
  *
+ * @note This function will also walk through externally allocated segments. It
+ *       is up to the user to decide whether to skip through these segments.
+ *
  * @param func
  *   Iterator function
  * @param arg
@@ -233,6 +236,9 @@ rte_memseg_walk(rte_memseg_walk_t func, void *arg);
  * @note This function read-locks the memory hotplug subsystem, and thus cannot
  *       be used within memory-related callback functions.
  *
+ * @note This function will also walk through externally allocated segments. It
+ *       is up to the user to decide whether to skip through these segments.
+ *
  * @param func
  *   Iterator function
  * @param arg
@@ -251,6 +257,9 @@ rte_memseg_contig_walk(rte_memseg_contig_walk_t func, void *arg);
  * @note This function read-locks the memory hotplug subsystem, and thus cannot
  *       be used within memory-related callback functions.
  *
+ * @note This function will also walk through externally allocated segments. It
+ *       is up to the user to decide whether to skip through these segments.
+ *
  * @param func
  *   Iterator function
  * @param arg
diff --git a/lib/librte_eal/common/malloc_elem.c b/lib/librte_eal/common/malloc_elem.c
index e0a8ed15b..1a74660de 100644
--- a/lib/librte_eal/common/malloc_elem.c
+++ b/lib/librte_eal/common/malloc_elem.c
@@ -39,10 +39,14 @@ malloc_elem_find_max_iova_contig(struct malloc_elem *elem, size_t align)
 	contig_seg_start = RTE_PTR_ALIGN_CEIL(data_start, align);
 
 	/* if we're in IOVA as VA mode, or if we're in legacy mode with
-	 * hugepages, all elements are IOVA-contiguous.
+	 * hugepages, all elements are IOVA-contiguous. however, we can only
+	 * make these assumptions about internal memory - externally allocated
+	 * segments have to be checked.
 	 */
-	if (rte_eal_iova_mode() == RTE_IOVA_VA ||
-			(internal_config.legacy_mem && rte_eal_has_hugepages()))
+	if (!elem->msl->external &&
+			(rte_eal_iova_mode() == RTE_IOVA_VA ||
+				(internal_config.legacy_mem &&
+					rte_eal_has_hugepages())))
 		return RTE_PTR_DIFF(data_end, contig_seg_start);
 
 	cur_page = RTE_PTR_ALIGN_FLOOR(contig_seg_start, page_sz);
diff --git a/lib/librte_eal/common/malloc_heap.c b/lib/librte_eal/common/malloc_heap.c
index ac7bbb3ba..3c8e2063b 100644
--- a/lib/librte_eal/common/malloc_heap.c
+++ b/lib/librte_eal/common/malloc_heap.c
@@ -95,6 +95,9 @@ malloc_add_seg(const struct rte_memseg_list *msl,
 	struct malloc_heap *heap;
 	int msl_idx;
 
+	if (msl->external)
+		return 0;
+
 	heap = &mcfg->malloc_heaps[msl->socket_id];
 
 	/* msl is const, so find it */
@@ -754,8 +757,10 @@ malloc_heap_free(struct malloc_elem *elem)
 	/* anything after this is a bonus */
 	ret = 0;
 
-	/* ...of which we can't avail if we are in legacy mode */
-	if (internal_config.legacy_mem)
+	/* ...of which we can't avail if we are in legacy mode, or if this is an
+	 * externally allocated segment.
+	 */
+	if (internal_config.legacy_mem || msl->external)
 		goto free_unlock;
 
 	/* check if we can free any memory back to the system */
diff --git a/lib/librte_eal/common/rte_malloc.c b/lib/librte_eal/common/rte_malloc.c
index b51a6d111..47ca5a742 100644
--- a/lib/librte_eal/common/rte_malloc.c
+++ b/lib/librte_eal/common/rte_malloc.c
@@ -223,7 +223,7 @@ rte_malloc_virt2iova(const void *addr)
 	if (elem == NULL)
 		return RTE_BAD_IOVA;
 
-	if (rte_eal_iova_mode() == RTE_IOVA_VA)
+	if (!elem->msl->external && rte_eal_iova_mode() == RTE_IOVA_VA)
 		return (uintptr_t) addr;
 
 	ms = rte_mem_virt2memseg(addr, elem->msl);
diff --git a/lib/librte_eal/linuxapp/eal/eal.c b/lib/librte_eal/linuxapp/eal/eal.c
index e59ac6577..253a6aece 100644
--- a/lib/librte_eal/linuxapp/eal/eal.c
+++ b/lib/librte_eal/linuxapp/eal/eal.c
@@ -725,6 +725,9 @@ check_socket(const struct rte_memseg_list *msl, void *arg)
 {
 	int *socket_id = arg;
 
+	if (msl->external)
+		return 0;
+
 	return *socket_id == msl->socket_id;
 }
 
@@ -1059,7 +1062,12 @@ mark_freeable(const struct rte_memseg_list *msl, const struct rte_memseg *ms,
 		void *arg __rte_unused)
 {
 	/* ms is const, so find this memseg */
-	struct rte_memseg *found = rte_mem_virt2memseg(ms->addr, msl);
+	struct rte_memseg *found;
+
+	if (msl->external)
+		return 0;
+
+	found = rte_mem_virt2memseg(ms->addr, msl);
 
 	found->flags &= ~RTE_MEMSEG_FLAG_DO_NOT_FREE;
 
diff --git a/lib/librte_eal/linuxapp/eal/eal_memalloc.c b/lib/librte_eal/linuxapp/eal/eal_memalloc.c
index 71a6e0fd9..f6a0098af 100644
--- a/lib/librte_eal/linuxapp/eal/eal_memalloc.c
+++ b/lib/librte_eal/linuxapp/eal/eal_memalloc.c
@@ -1408,6 +1408,9 @@ sync_walk(const struct rte_memseg_list *msl, void *arg __rte_unused)
 	unsigned int i;
 	int msl_idx;
 
+	if (msl->external)
+		return 0;
+
 	msl_idx = msl - mcfg->memsegs;
 	primary_msl = &mcfg->memsegs[msl_idx];
 	local_msl = &local_memsegs[msl_idx];
@@ -1456,6 +1459,9 @@ secondary_msl_create_walk(const struct rte_memseg_list *msl,
 	char name[PATH_MAX];
 	int msl_idx, ret;
 
+	if (msl->external)
+		return 0;
+
 	msl_idx = msl - mcfg->memsegs;
 	primary_msl = &mcfg->memsegs[msl_idx];
 	local_msl = &local_memsegs[msl_idx];
@@ -1509,6 +1515,9 @@ fd_list_create_walk(const struct rte_memseg_list *msl,
 	unsigned int len;
 	int msl_idx;
 
+	if (msl->external)
+		return 0;
+
 	msl_idx = msl - mcfg->memsegs;
 	len = msl->memseg_arr.len;
 
diff --git a/lib/librte_eal/linuxapp/eal/eal_vfio.c b/lib/librte_eal/linuxapp/eal/eal_vfio.c
index c68dc38e0..fddbc3b54 100644
--- a/lib/librte_eal/linuxapp/eal/eal_vfio.c
+++ b/lib/librte_eal/linuxapp/eal/eal_vfio.c
@@ -1082,11 +1082,14 @@ rte_vfio_get_group_num(const char *sysfs_base,
 }
 
 static int
-type1_map(const struct rte_memseg_list *msl __rte_unused,
-		const struct rte_memseg *ms, void *arg)
+type1_map(const struct rte_memseg_list *msl, const struct rte_memseg *ms,
+		void *arg)
 {
 	int *vfio_container_fd = arg;
 
+	if (msl->external)
+		return 0;
+
 	return vfio_type1_dma_mem_map(*vfio_container_fd, ms->addr_64, ms->iova,
 			ms->len, 1);
 }
@@ -1196,11 +1199,14 @@ vfio_spapr_dma_do_map(int vfio_container_fd, uint64_t vaddr, uint64_t iova,
 }
 
 static int
-vfio_spapr_map_walk(const struct rte_memseg_list *msl __rte_unused,
+vfio_spapr_map_walk(const struct rte_memseg_list *msl,
 		const struct rte_memseg *ms, void *arg)
 {
 	int *vfio_container_fd = arg;
 
+	if (msl->external)
+		return 0;
+
 	return vfio_spapr_dma_mem_map(*vfio_container_fd, ms->addr_64, ms->iova,
 			ms->len, 1);
 }
@@ -1210,12 +1216,15 @@ struct spapr_walk_param {
 	uint64_t hugepage_sz;
 };
 static int
-vfio_spapr_window_size_walk(const struct rte_memseg_list *msl __rte_unused,
+vfio_spapr_window_size_walk(const struct rte_memseg_list *msl,
 		const struct rte_memseg *ms, void *arg)
 {
 	struct spapr_walk_param *param = arg;
 	uint64_t max = ms->iova + ms->len;
 
+	if (msl->external)
+		return 0;
+
 	if (max > param->window_size) {
 		param->hugepage_sz = ms->hugepage_sz;
 		param->window_size = max;
diff --git a/lib/librte_mempool/rte_mempool.c b/lib/librte_mempool/rte_mempool.c
index 03e6b5f73..2ed539f01 100644
--- a/lib/librte_mempool/rte_mempool.c
+++ b/lib/librte_mempool/rte_mempool.c
@@ -99,25 +99,44 @@ static unsigned optimize_object_size(unsigned obj_size)
 	return new_obj_size * RTE_MEMPOOL_ALIGN;
 }
 
+struct pagesz_walk_arg {
+	int socket_id;
+	size_t min;
+};
+
 static int
 find_min_pagesz(const struct rte_memseg_list *msl, void *arg)
 {
-	size_t *min = arg;
+	struct pagesz_walk_arg *wa = arg;
+	bool valid;
 
-	if (msl->page_sz < *min)
-		*min = msl->page_sz;
+	/*
+	 * we need to only look at page sizes available for a particular socket
+	 * ID.  so, we either need an exact match on socket ID (can match both
+	 * native and external memory), or, if SOCKET_ID_ANY was specified as a
+	 * socket ID argument, we must only look at native memory and ignore any
+	 * page sizes associated with external memory.
+	 */
+	valid = msl->socket_id == wa->socket_id;
+	valid |= wa->socket_id == SOCKET_ID_ANY && msl->external == 0;
+
+	if (valid && msl->page_sz < wa->min)
+		wa->min = msl->page_sz;
 
 	return 0;
 }
 
 static size_t
-get_min_page_size(void)
+get_min_page_size(int socket_id)
 {
-	size_t min_pagesz = SIZE_MAX;
+	struct pagesz_walk_arg wa;
 
-	rte_memseg_list_walk(find_min_pagesz, &min_pagesz);
+	wa.min = SIZE_MAX;
+	wa.socket_id = socket_id;
 
-	return min_pagesz == SIZE_MAX ? (size_t) getpagesize() : min_pagesz;
+	rte_memseg_list_walk(find_min_pagesz, &wa);
+
+	return wa.min == SIZE_MAX ? (size_t) getpagesize() : wa.min;
 }
 
 
@@ -470,7 +489,7 @@ rte_mempool_populate_default(struct rte_mempool *mp)
 		pg_sz = 0;
 		pg_shift = 0;
 	} else if (try_contig) {
-		pg_sz = get_min_page_size();
+		pg_sz = get_min_page_size(mp->socket_id);
 		pg_shift = rte_bsf32(pg_sz);
 	} else {
 		pg_sz = getpagesize();
diff --git a/test/test/test_malloc.c b/test/test/test_malloc.c
index 4b5abb4e0..5e5272419 100644
--- a/test/test/test_malloc.c
+++ b/test/test/test_malloc.c
@@ -711,6 +711,9 @@ check_socket_mem(const struct rte_memseg_list *msl, void *arg)
 {
 	int32_t *socket = arg;
 
+	if (msl->external)
+		return 0;
+
 	return *socket == msl->socket_id;
 }
 
diff --git a/test/test/test_memzone.c b/test/test/test_memzone.c
index 452d7cc5e..9fe465e62 100644
--- a/test/test/test_memzone.c
+++ b/test/test/test_memzone.c
@@ -115,6 +115,9 @@ find_available_pagesz(const struct rte_memseg_list *msl, void *arg)
 {
 	struct walk_arg *wa = arg;
 
+	if (msl->external)
+		return 0;
+
 	if (msl->page_sz == RTE_PGSIZE_2M)
 		wa->hugepage_2MB_avail = 1;
 	if (msl->page_sz == RTE_PGSIZE_1G)
-- 
2.17.1