2022-09-04 04:27:01

by Rebecca Mckeever

Subject: [PATCH v4 0/4] memblock tests: add NUMA tests for memblock_alloc_try_nid*

These patches add additional tests for memblock_alloc_try_nid() and
memblock_alloc_try_nid_raw() that use a simulated physical
memory that is set up with multiple NUMA nodes. Additionally, most of
these tests set nid != NUMA_NO_NODE.

To set up a simulated physical memory with multiple NUMA nodes, patch 1
introduces setup_numa_memblock(). This function uses a previously
allocated dummy physical memory. It can be used in place of
setup_memblock() in tests that need to simulate a NUMA system.

These tests are run twice, once for memblock_alloc_try_nid() and once
for memblock_alloc_try_nid_raw(), so that both functions are tested with
the same set of tests. When the tests run memblock_alloc_try_nid(), they
test that the entire memory region is zero. When the tests run
memblock_alloc_try_nid_raw(), they test that the entire memory region is
nonzero.

---
Changelog

v3 -> v4
Based on feedback from David Hildenbrand:
PATCH 1:
- tests/common.h:
+ Remove #define MEM_FACTOR
PATCH 2:
- tests/alloc_nid_api.c:
+ Add missing blank line after alloc_nid_test_flags = flags;
PATCH 2, PATCH 3, PATCH 4:
- tests/alloc_nid_api.c:
+ Define variables in new tests in a single block
+ Change "cleared memory region" to "memory region" in comment blocks
for new tests

Based on feedback from Mike Rapoport:
PATCH 1:
- tests/common.c:
+ Remove node_sizes[] array
- tests/common.c, tests/common.h:
+ Combine setup_numa_memblock_generic() and setup_numa_memblock():
* Remove setup_numa_memblock(void)
* Change setup_numa_memblock_generic() to setup_numa_memblock()
and remove the factor and node_cnt parameters
* Change nodes[] parameter to an array containing the denominators
of the fractions of MEM_SIZE contained in each node
e.g., if nodes[0] = SZ_8, node 0 will contain 1/8th of MEM_SIZE
PATCH 2:
- tests/alloc_nid_api.c:
+ Add node_fractions[] array:
* Based on values from node_sizes[] array previously in PATCH 1
* Update to fit the new nodes[] parameter of setup_numa_memblock()

v2 -> v3
PATCH 2, PATCH 3, PATCH 4:
- tests/alloc_nid_api.c:
+ Edits for consistency between tests and test naming styles:
* Change "numa_top_down" to "top_down_numa"
* Change "numa_bottom_up" to "bottom_up_numa"
* Move "generic" to immediately before "check"
* alloc_try_nid_bottom_up_numa_small_node_check() and
alloc_try_nid_bottom_up_numa_node_reserved_check():
- Separate size definition from its declaration

Updates based on changes to the dependent patch set noted above:
PATCH 2, PATCH 3, PATCH 4:
- tests/alloc_nid_api.c:
+ get_func_testing():
* Rename to get_memblock_alloc_try_nid_name().
+ Update calls to verify_mem_content() to assert_mem_content().

v1 -> v2
Updates based on changes to the dependent patch set noted above:
PATCH 2, PATCH 3, PATCH 4:
- tests/alloc_nid_api.c:
+ Update calls to verify_mem_content() to include the flags parameter
required by the new, common verify_mem_content() function.
---

Rebecca Mckeever (4):
memblock tests: add simulation of physical memory with multiple NUMA
nodes
memblock tests: add top-down NUMA tests for memblock_alloc_try_nid*
memblock tests: add bottom-up NUMA tests for memblock_alloc_try_nid*
memblock tests: add generic NUMA tests for memblock_alloc_try_nid*

.../testing/memblock/scripts/Makefile.include | 2 +-
tools/testing/memblock/tests/alloc_nid_api.c | 1459 ++++++++++++++++-
tools/testing/memblock/tests/alloc_nid_api.h | 16 +
tools/testing/memblock/tests/common.c | 29 +
tools/testing/memblock/tests/common.h | 22 +-
5 files changed, 1515 insertions(+), 13 deletions(-)

--
2.25.1

2022-09-04 04:27:22

by Rebecca Mckeever

Subject: [PATCH v4 2/4] memblock tests: add top-down NUMA tests for memblock_alloc_try_nid*

Add tests for memblock_alloc_try_nid() and memblock_alloc_try_nid_raw()
where the simulated physical memory is set up with multiple NUMA nodes.
Additionally, all of these tests set nid != NUMA_NO_NODE. These tests are
run with a top-down allocation direction.

The tested scenarios are:

Range unrestricted:
- region can be allocated in the specific node requested:
+ there are no previously reserved regions
+ the requested node is partially reserved but has enough space
- the specific node requested cannot accommodate the request, but the
region can be allocated in a different node:
+ there are no previously reserved regions, but node is too small
+ the requested node is fully reserved
+ the requested node is partially reserved and does not have
enough space

Range restricted:
- region can be allocated in the specific node requested after dropping
min_addr:
+ range partially overlaps with two different nodes, where the first
node is the requested node
+ range partially overlaps with two different nodes, where the
requested node ends before min_addr
- region cannot be allocated in the specific node requested, but it can be
allocated in the requested range:
+ range overlaps with multiple nodes along node boundaries, and the
requested node ends before min_addr
+ range overlaps with multiple nodes along node boundaries, and the
requested node starts after max_addr
- region cannot be allocated in the specific node requested, but it can be
allocated after dropping min_addr:
+ range partially overlaps with two different nodes, where the
second node is the requested node

Signed-off-by: Rebecca Mckeever <[email protected]>
---
tools/testing/memblock/tests/alloc_nid_api.c | 694 ++++++++++++++++++-
tools/testing/memblock/tests/alloc_nid_api.h | 16 +
tools/testing/memblock/tests/common.h | 18 +
3 files changed, 717 insertions(+), 11 deletions(-)

diff --git a/tools/testing/memblock/tests/alloc_nid_api.c b/tools/testing/memblock/tests/alloc_nid_api.c
index 32b3c1594fdd..a639e9d3e098 100644
--- a/tools/testing/memblock/tests/alloc_nid_api.c
+++ b/tools/testing/memblock/tests/alloc_nid_api.c
@@ -3,6 +3,14 @@

static int alloc_nid_test_flags = TEST_F_NONE;

+/*
+ * contains the denominators of the fractions of MEM_SIZE contained in each node
+ * (e.g., if node_fractions[0] = SZ_8, node 0 will contain 1/8th of MEM_SIZE)
+ */
+static const phys_addr_t node_fractions[] = {
+ SZ_4, SZ_16, SZ_8, SZ_8, SZ_16, SZ_16, SZ_4, SZ_16
+};
+
static inline const char * const get_memblock_alloc_try_nid_name(int flags)
{
if (flags & TEST_F_RAW)
@@ -1094,7 +1102,7 @@ static int alloc_try_nid_bottom_up_cap_min_check(void)
return 0;
}

-/* Test case wrappers */
+/* Test case wrappers for range tests */
static int alloc_try_nid_simple_check(void)
{
test_print("\tRunning %s...\n", __func__);
@@ -1226,17 +1234,10 @@ static int alloc_try_nid_low_max_check(void)
return 0;
}

-static int memblock_alloc_nid_checks_internal(int flags)
+static int memblock_alloc_nid_range_checks(void)
{
- const char *func = get_memblock_alloc_try_nid_name(flags);
-
- alloc_nid_test_flags = flags;
- prefix_reset();
- prefix_push(func);
- test_print("Running %s tests...\n", func);
-
- reset_memblock_attributes();
- dummy_physical_memory_init();
+ test_print("Running %s range tests...\n",
+ get_memblock_alloc_try_nid_name(alloc_nid_test_flags));

alloc_try_nid_simple_check();
alloc_try_nid_misaligned_check();
@@ -1253,6 +1254,677 @@ static int memblock_alloc_nid_checks_internal(int flags)
alloc_try_nid_reserved_all_check();
alloc_try_nid_low_max_check();

+ return 0;
+}
+
+/*
+ * A test that tries to allocate a memory region in a specific NUMA node that
+ * has enough memory to allocate a region of the requested size.
+ * Expect to allocate an aligned region at the end of the requested node.
+ */
+static int alloc_try_nid_top_down_numa_simple_check(void)
+{
+ int nid_req = 3;
+ struct memblock_region *new_rgn = &memblock.reserved.regions[0];
+ struct memblock_region *req_node = &memblock.memory.regions[nid_req];
+ void *allocated_ptr = NULL;
+ phys_addr_t size;
+ phys_addr_t min_addr;
+ phys_addr_t max_addr;
+
+ PREFIX_PUSH();
+ setup_numa_memblock(node_fractions);
+
+ ASSERT_LE(SZ_4, req_node->size);
+ size = req_node->size / SZ_4;
+ min_addr = memblock_start_of_DRAM();
+ max_addr = memblock_end_of_DRAM();
+
+ allocated_ptr = run_memblock_alloc_try_nid(size, SMP_CACHE_BYTES,
+ min_addr, max_addr, nid_req);
+
+ ASSERT_NE(allocated_ptr, NULL);
+ assert_mem_content(allocated_ptr, size, alloc_nid_test_flags);
+
+ ASSERT_EQ(new_rgn->size, size);
+ ASSERT_EQ(new_rgn->base, region_end(req_node) - size);
+ ASSERT_LE(req_node->base, new_rgn->base);
+
+ ASSERT_EQ(memblock.reserved.cnt, 1);
+ ASSERT_EQ(memblock.reserved.total_size, size);
+
+ test_pass_pop();
+
+ return 0;
+}
+
+/*
+ * A test that tries to allocate a memory region in a specific NUMA node that
+ * does not have enough memory to allocate a region of the requested size:
+ *
+ * | +-----+ +------------------+ |
+ * | | req | | expected | |
+ * +---+-----+----------+------------------+-----+
+ *
+ * | +---------+ |
+ * | | rgn | |
+ * +-----------------------------+---------+-----+
+ *
+ * Expect to allocate an aligned region at the end of the last node that has
+ * enough memory (in this case, nid = 6) after falling back to NUMA_NO_NODE.
+ */
+static int alloc_try_nid_top_down_numa_small_node_check(void)
+{
+ int nid_req = 1;
+ int nid_exp = 6;
+ struct memblock_region *new_rgn = &memblock.reserved.regions[0];
+ struct memblock_region *req_node = &memblock.memory.regions[nid_req];
+ struct memblock_region *exp_node = &memblock.memory.regions[nid_exp];
+ void *allocated_ptr = NULL;
+ phys_addr_t size;
+ phys_addr_t min_addr;
+ phys_addr_t max_addr;
+
+ PREFIX_PUSH();
+ setup_numa_memblock(node_fractions);
+
+ size = SZ_2 * req_node->size;
+ min_addr = memblock_start_of_DRAM();
+ max_addr = memblock_end_of_DRAM();
+
+ allocated_ptr = run_memblock_alloc_try_nid(size, SMP_CACHE_BYTES,
+ min_addr, max_addr, nid_req);
+
+ ASSERT_NE(allocated_ptr, NULL);
+ assert_mem_content(allocated_ptr, size, alloc_nid_test_flags);
+
+ ASSERT_EQ(new_rgn->size, size);
+ ASSERT_EQ(new_rgn->base, region_end(exp_node) - size);
+ ASSERT_LE(exp_node->base, new_rgn->base);
+
+ ASSERT_EQ(memblock.reserved.cnt, 1);
+ ASSERT_EQ(memblock.reserved.total_size, size);
+
+ test_pass_pop();
+
+ return 0;
+}
+
+/*
+ * A test that tries to allocate a memory region in a specific NUMA node that
+ * is fully reserved:
+ *
+ * | +---------+ +------------------+ |
+ * | |requested| | expected | |
+ * +--------------+---------+------------+------------------+-----+
+ *
+ * | +---------+ +---------+ |
+ * | | reserved| | new | |
+ * +--------------+---------+---------------------+---------+-----+
+ *
+ * Expect to allocate an aligned region at the end of the last node that is
+ * large enough and has enough unreserved memory (in this case, nid = 6) after
+ * falling back to NUMA_NO_NODE. The region count and total size get updated.
+ */
+static int alloc_try_nid_top_down_numa_node_reserved_check(void)
+{
+ int nid_req = 2;
+ int nid_exp = 6;
+ struct memblock_region *new_rgn = &memblock.reserved.regions[1];
+ struct memblock_region *req_node = &memblock.memory.regions[nid_req];
+ struct memblock_region *exp_node = &memblock.memory.regions[nid_exp];
+ void *allocated_ptr = NULL;
+ phys_addr_t size;
+ phys_addr_t min_addr;
+ phys_addr_t max_addr;
+
+ PREFIX_PUSH();
+ setup_numa_memblock(node_fractions);
+
+ size = req_node->size;
+ min_addr = memblock_start_of_DRAM();
+ max_addr = memblock_end_of_DRAM();
+
+ memblock_reserve(req_node->base, req_node->size);
+ allocated_ptr = run_memblock_alloc_try_nid(size, SMP_CACHE_BYTES,
+ min_addr, max_addr, nid_req);
+
+ ASSERT_NE(allocated_ptr, NULL);
+ assert_mem_content(allocated_ptr, size, alloc_nid_test_flags);
+
+ ASSERT_EQ(new_rgn->size, size);
+ ASSERT_EQ(new_rgn->base, region_end(exp_node) - size);
+ ASSERT_LE(exp_node->base, new_rgn->base);
+
+ ASSERT_EQ(memblock.reserved.cnt, 2);
+ ASSERT_EQ(memblock.reserved.total_size, size + req_node->size);
+
+ test_pass_pop();
+
+ return 0;
+}
+
+/*
+ * A test that tries to allocate a memory region in a specific NUMA node that
+ * is partially reserved but has enough memory for the allocated region:
+ *
+ * | +---------------------------------------+ |
+ * | | requested | |
+ * +-----------+---------------------------------------+----------+
+ *
+ * | +------------------+ +-----+ |
+ * | | reserved | | new | |
+ * +-----------+------------------+--------------+-----+----------+
+ *
+ * Expect to allocate an aligned region at the end of the requested node. The
+ * region count and total size get updated.
+ */
+static int alloc_try_nid_top_down_numa_part_reserved_check(void)
+{
+ int nid_req = 4;
+ struct memblock_region *new_rgn = &memblock.reserved.regions[1];
+ struct memblock_region *req_node = &memblock.memory.regions[nid_req];
+ void *allocated_ptr = NULL;
+ struct region r1;
+ phys_addr_t size;
+ phys_addr_t min_addr;
+ phys_addr_t max_addr;
+
+ PREFIX_PUSH();
+ setup_numa_memblock(node_fractions);
+
+ ASSERT_LE(SZ_8, req_node->size);
+ r1.base = req_node->base;
+ r1.size = req_node->size / SZ_2;
+ size = r1.size / SZ_4;
+ min_addr = memblock_start_of_DRAM();
+ max_addr = memblock_end_of_DRAM();
+
+ memblock_reserve(r1.base, r1.size);
+ allocated_ptr = run_memblock_alloc_try_nid(size, SMP_CACHE_BYTES,
+ min_addr, max_addr, nid_req);
+
+ ASSERT_NE(allocated_ptr, NULL);
+ assert_mem_content(allocated_ptr, size, alloc_nid_test_flags);
+
+ ASSERT_EQ(new_rgn->size, size);
+ ASSERT_EQ(new_rgn->base, region_end(req_node) - size);
+ ASSERT_LE(req_node->base, new_rgn->base);
+
+ ASSERT_EQ(memblock.reserved.cnt, 2);
+ ASSERT_EQ(memblock.reserved.total_size, size + r1.size);
+
+ test_pass_pop();
+
+ return 0;
+}
+
+/*
+ * A test that tries to allocate a memory region in a specific NUMA node that
+ * is partially reserved and does not have enough contiguous memory for the
+ * allocated region:
+ *
+ * | +-----------------------+ +----------------------|
+ * | | requested | | expected |
+ * +-----------+-----------------------+---------+----------------------+
+ *
+ * | +----------+ +-----------|
+ * | | reserved | | new |
+ * +-----------------+----------+---------------------------+-----------+
+ *
+ * Expect to allocate an aligned region at the end of the last node that is
+ * large enough and has enough unreserved memory (in this case,
+ * nid = NUMA_NODES - 1) after falling back to NUMA_NO_NODE. The region count
+ * and total size get updated.
+ */
+static int alloc_try_nid_top_down_numa_part_reserved_fallback_check(void)
+{
+ int nid_req = 4;
+ int nid_exp = NUMA_NODES - 1;
+ struct memblock_region *new_rgn = &memblock.reserved.regions[1];
+ struct memblock_region *req_node = &memblock.memory.regions[nid_req];
+ struct memblock_region *exp_node = &memblock.memory.regions[nid_exp];
+ void *allocated_ptr = NULL;
+ struct region r1;
+ phys_addr_t size;
+ phys_addr_t min_addr;
+ phys_addr_t max_addr;
+
+ PREFIX_PUSH();
+ setup_numa_memblock(node_fractions);
+
+ ASSERT_LE(SZ_4, req_node->size);
+ size = req_node->size / SZ_2;
+ r1.base = req_node->base + (size / SZ_2);
+ r1.size = size;
+
+ min_addr = memblock_start_of_DRAM();
+ max_addr = memblock_end_of_DRAM();
+
+ memblock_reserve(r1.base, r1.size);
+ allocated_ptr = run_memblock_alloc_try_nid(size, SMP_CACHE_BYTES,
+ min_addr, max_addr, nid_req);
+
+ ASSERT_NE(allocated_ptr, NULL);
+ assert_mem_content(allocated_ptr, size, alloc_nid_test_flags);
+
+ ASSERT_EQ(new_rgn->size, size);
+ ASSERT_EQ(new_rgn->base, region_end(exp_node) - size);
+ ASSERT_LE(exp_node->base, new_rgn->base);
+
+ ASSERT_EQ(memblock.reserved.cnt, 2);
+ ASSERT_EQ(memblock.reserved.total_size, size + r1.size);
+
+ test_pass_pop();
+
+ return 0;
+}
+
+/*
+ * A test that tries to allocate a memory region that spans over the min_addr
+ * and max_addr range and overlaps with two different nodes, where the first
+ * node is the requested node:
+ *
+ * min_addr
+ * | max_addr
+ * | |
+ * v v
+ * | +-----------------------+-----------+ |
+ * | | requested | node3 | |
+ * +-----------+-----------------------+-----------+--------------+
+ * + +
+ * | +-----------+ |
+ * | | rgn | |
+ * +-----------------------+-----------+--------------------------+
+ *
+ * Expect to drop the lower limit and allocate a memory region that ends at
+ * the end of the requested node.
+ */
+static int alloc_try_nid_top_down_numa_split_range_low_check(void)
+{
+ int nid_req = 2;
+ struct memblock_region *new_rgn = &memblock.reserved.regions[0];
+ struct memblock_region *req_node = &memblock.memory.regions[nid_req];
+ void *allocated_ptr = NULL;
+ phys_addr_t size = SZ_512;
+ phys_addr_t min_addr;
+ phys_addr_t max_addr;
+ phys_addr_t req_node_end;
+
+ PREFIX_PUSH();
+ setup_numa_memblock(node_fractions);
+
+ req_node_end = region_end(req_node);
+ min_addr = req_node_end - SZ_256;
+ max_addr = min_addr + size;
+
+ allocated_ptr = run_memblock_alloc_try_nid(size, SMP_CACHE_BYTES,
+ min_addr, max_addr, nid_req);
+
+ ASSERT_NE(allocated_ptr, NULL);
+ assert_mem_content(allocated_ptr, size, alloc_nid_test_flags);
+
+ ASSERT_EQ(new_rgn->size, size);
+ ASSERT_EQ(new_rgn->base, req_node_end - size);
+ ASSERT_LE(req_node->base, new_rgn->base);
+
+ ASSERT_EQ(memblock.reserved.cnt, 1);
+ ASSERT_EQ(memblock.reserved.total_size, size);
+
+ test_pass_pop();
+
+ return 0;
+}
+
+/*
+ * A test that tries to allocate a memory region that spans over the min_addr
+ * and max_addr range and overlaps with two different nodes, where the second
+ * node is the requested node:
+ *
+ * min_addr
+ * | max_addr
+ * | |
+ * v v
+ * | +--------------------------+---------+ |
+ * | | expected |requested| |
+ * +------+--------------------------+---------+----------------+
+ * + +
+ * | +---------+ |
+ * | | rgn | |
+ * +-----------------------+---------+--------------------------+
+ *
+ * Expect to drop the lower limit and allocate a memory region that
+ * ends at the end of the first node that overlaps with the range.
+ */
+static int alloc_try_nid_top_down_numa_split_range_high_check(void)
+{
+ int nid_req = 3;
+ int nid_exp = nid_req - 1;
+ struct memblock_region *new_rgn = &memblock.reserved.regions[0];
+ struct memblock_region *exp_node = &memblock.memory.regions[nid_exp];
+ void *allocated_ptr = NULL;
+ phys_addr_t size = SZ_512;
+ phys_addr_t min_addr;
+ phys_addr_t max_addr;
+ phys_addr_t exp_node_end;
+
+ PREFIX_PUSH();
+ setup_numa_memblock(node_fractions);
+
+ exp_node_end = region_end(exp_node);
+ min_addr = exp_node_end - SZ_256;
+ max_addr = min_addr + size;
+
+ allocated_ptr = run_memblock_alloc_try_nid(size, SMP_CACHE_BYTES,
+ min_addr, max_addr, nid_req);
+
+ ASSERT_NE(allocated_ptr, NULL);
+ assert_mem_content(allocated_ptr, size, alloc_nid_test_flags);
+
+ ASSERT_EQ(new_rgn->size, size);
+ ASSERT_EQ(new_rgn->base, exp_node_end - size);
+ ASSERT_LE(exp_node->base, new_rgn->base);
+
+ ASSERT_EQ(memblock.reserved.cnt, 1);
+ ASSERT_EQ(memblock.reserved.total_size, size);
+
+ test_pass_pop();
+
+ return 0;
+}
+
+/*
+ * A test that tries to allocate a memory region that spans over the min_addr
+ * and max_addr range and overlaps with two different nodes, where the requested
+ * node ends before min_addr:
+ *
+ * min_addr
+ * | max_addr
+ * | |
+ * v v
+ * | +---------------+ +-------------+---------+ |
+ * | | requested | | node1 | node2 | |
+ * +----+---------------+--------+-------------+---------+----------+
+ * + +
+ * | +---------+ |
+ * | | rgn | |
+ * +----------+---------+-------------------------------------------+
+ *
+ * Expect to drop the lower limit and allocate a memory region that ends at
+ * the end of the requested node.
+ */
+static int alloc_try_nid_top_down_numa_no_overlap_split_check(void)
+{
+ int nid_req = 2;
+ struct memblock_region *new_rgn = &memblock.reserved.regions[0];
+ struct memblock_region *req_node = &memblock.memory.regions[nid_req];
+ struct memblock_region *node2 = &memblock.memory.regions[6];
+ void *allocated_ptr = NULL;
+ phys_addr_t size;
+ phys_addr_t min_addr;
+ phys_addr_t max_addr;
+
+ PREFIX_PUSH();
+ setup_numa_memblock(node_fractions);
+
+ size = SZ_512;
+ min_addr = node2->base - SZ_256;
+ max_addr = min_addr + size;
+
+ allocated_ptr = run_memblock_alloc_try_nid(size, SMP_CACHE_BYTES,
+ min_addr, max_addr, nid_req);
+
+ ASSERT_NE(allocated_ptr, NULL);
+ assert_mem_content(allocated_ptr, size, alloc_nid_test_flags);
+
+ ASSERT_EQ(new_rgn->size, size);
+ ASSERT_EQ(new_rgn->base, region_end(req_node) - size);
+ ASSERT_LE(req_node->base, new_rgn->base);
+
+ ASSERT_EQ(memblock.reserved.cnt, 1);
+ ASSERT_EQ(memblock.reserved.total_size, size);
+
+ test_pass_pop();
+
+ return 0;
+}
+
+/*
+ * A test that tries to allocate memory within min_addr and max_add range when
+ * the requested node and the range do not overlap, and requested node ends
+ * before min_addr. The range overlaps with multiple nodes along node
+ * boundaries:
+ *
+ * min_addr
+ * | max_addr
+ * | |
+ * v v
+ * |-----------+ +----------+----...----+----------+ |
+ * | requested | | min node | ... | max node | |
+ * +-----------+-----------+----------+----...----+----------+------+
+ * + +
+ * | +-----+ |
+ * | | rgn | |
+ * +---------------------------------------------------+-----+------+
+ *
+ * Expect to allocate a memory region at the end of the final node in
+ * the range after falling back to NUMA_NO_NODE.
+ */
+static int alloc_try_nid_top_down_numa_no_overlap_low_check(void)
+{
+ int nid_req = 0;
+ struct memblock_region *new_rgn = &memblock.reserved.regions[0];
+ struct memblock_region *min_node = &memblock.memory.regions[2];
+ struct memblock_region *max_node = &memblock.memory.regions[5];
+ void *allocated_ptr = NULL;
+ phys_addr_t size = SZ_64;
+ phys_addr_t max_addr;
+ phys_addr_t min_addr;
+
+ PREFIX_PUSH();
+ setup_numa_memblock(node_fractions);
+
+ min_addr = min_node->base;
+ max_addr = region_end(max_node);
+
+ allocated_ptr = run_memblock_alloc_try_nid(size, SMP_CACHE_BYTES,
+ min_addr, max_addr, nid_req);
+
+ ASSERT_NE(allocated_ptr, NULL);
+ assert_mem_content(allocated_ptr, size, alloc_nid_test_flags);
+
+ ASSERT_EQ(new_rgn->size, size);
+ ASSERT_EQ(new_rgn->base, max_addr - size);
+ ASSERT_LE(max_node->base, new_rgn->base);
+
+ ASSERT_EQ(memblock.reserved.cnt, 1);
+ ASSERT_EQ(memblock.reserved.total_size, size);
+
+ test_pass_pop();
+
+ return 0;
+}
+
+/*
+ * A test that tries to allocate memory within min_addr and max_add range when
+ * the requested node and the range do not overlap, and requested node starts
+ * after max_addr. The range overlaps with multiple nodes along node
+ * boundaries:
+ *
+ * min_addr
+ * | max_addr
+ * | |
+ * v v
+ * | +----------+----...----+----------+ +-----------+ |
+ * | | min node | ... | max node | | requested | |
+ * +-----+----------+----...----+----------+--------+-----------+---+
+ * + +
+ * | +-----+ |
+ * | | rgn | |
+ * +---------------------------------+-----+------------------------+
+ *
+ * Expect to allocate a memory region at the end of the final node in
+ * the range after falling back to NUMA_NO_NODE.
+ */
+static int alloc_try_nid_top_down_numa_no_overlap_high_check(void)
+{
+ int nid_req = 7;
+ struct memblock_region *new_rgn = &memblock.reserved.regions[0];
+ struct memblock_region *min_node = &memblock.memory.regions[2];
+ struct memblock_region *max_node = &memblock.memory.regions[5];
+ void *allocated_ptr = NULL;
+ phys_addr_t size = SZ_64;
+ phys_addr_t max_addr;
+ phys_addr_t min_addr;
+
+ PREFIX_PUSH();
+ setup_numa_memblock(node_fractions);
+
+ min_addr = min_node->base;
+ max_addr = region_end(max_node);
+
+ allocated_ptr = run_memblock_alloc_try_nid(size, SMP_CACHE_BYTES,
+ min_addr, max_addr, nid_req);
+
+ ASSERT_NE(allocated_ptr, NULL);
+ assert_mem_content(allocated_ptr, size, alloc_nid_test_flags);
+
+ ASSERT_EQ(new_rgn->size, size);
+ ASSERT_EQ(new_rgn->base, max_addr - size);
+ ASSERT_LE(max_node->base, new_rgn->base);
+
+ ASSERT_EQ(memblock.reserved.cnt, 1);
+ ASSERT_EQ(memblock.reserved.total_size, size);
+
+ test_pass_pop();
+
+ return 0;
+}
+
+/* Test case wrappers for NUMA tests */
+static int alloc_try_nid_numa_simple_check(void)
+{
+ test_print("\tRunning %s...\n", __func__);
+ memblock_set_bottom_up(false);
+ alloc_try_nid_top_down_numa_simple_check();
+
+ return 0;
+}
+
+static int alloc_try_nid_numa_small_node_check(void)
+{
+ test_print("\tRunning %s...\n", __func__);
+ memblock_set_bottom_up(false);
+ alloc_try_nid_top_down_numa_small_node_check();
+
+ return 0;
+}
+
+static int alloc_try_nid_numa_node_reserved_check(void)
+{
+ test_print("\tRunning %s...\n", __func__);
+ memblock_set_bottom_up(false);
+ alloc_try_nid_top_down_numa_node_reserved_check();
+
+ return 0;
+}
+
+static int alloc_try_nid_numa_part_reserved_check(void)
+{
+ test_print("\tRunning %s...\n", __func__);
+ memblock_set_bottom_up(false);
+ alloc_try_nid_top_down_numa_part_reserved_check();
+
+ return 0;
+}
+
+static int alloc_try_nid_numa_part_reserved_fallback_check(void)
+{
+ test_print("\tRunning %s...\n", __func__);
+ memblock_set_bottom_up(false);
+ alloc_try_nid_top_down_numa_part_reserved_fallback_check();
+
+ return 0;
+}
+
+static int alloc_try_nid_numa_split_range_low_check(void)
+{
+ test_print("\tRunning %s...\n", __func__);
+ memblock_set_bottom_up(false);
+ alloc_try_nid_top_down_numa_split_range_low_check();
+
+ return 0;
+}
+
+static int alloc_try_nid_numa_split_range_high_check(void)
+{
+ test_print("\tRunning %s...\n", __func__);
+ memblock_set_bottom_up(false);
+ alloc_try_nid_top_down_numa_split_range_high_check();
+
+ return 0;
+}
+
+static int alloc_try_nid_numa_no_overlap_split_check(void)
+{
+ test_print("\tRunning %s...\n", __func__);
+ memblock_set_bottom_up(false);
+ alloc_try_nid_top_down_numa_no_overlap_split_check();
+
+ return 0;
+}
+
+static int alloc_try_nid_numa_no_overlap_low_check(void)
+{
+ test_print("\tRunning %s...\n", __func__);
+ memblock_set_bottom_up(false);
+ alloc_try_nid_top_down_numa_no_overlap_low_check();
+
+ return 0;
+}
+
+static int alloc_try_nid_numa_no_overlap_high_check(void)
+{
+ test_print("\tRunning %s...\n", __func__);
+ memblock_set_bottom_up(false);
+ alloc_try_nid_top_down_numa_no_overlap_high_check();
+
+ return 0;
+}
+
+int __memblock_alloc_nid_numa_checks(void)
+{
+ test_print("Running %s NUMA tests...\n",
+ get_memblock_alloc_try_nid_name(alloc_nid_test_flags));
+
+ alloc_try_nid_numa_simple_check();
+ alloc_try_nid_numa_small_node_check();
+ alloc_try_nid_numa_node_reserved_check();
+ alloc_try_nid_numa_part_reserved_check();
+ alloc_try_nid_numa_part_reserved_fallback_check();
+ alloc_try_nid_numa_split_range_low_check();
+ alloc_try_nid_numa_split_range_high_check();
+
+ alloc_try_nid_numa_no_overlap_split_check();
+ alloc_try_nid_numa_no_overlap_low_check();
+ alloc_try_nid_numa_no_overlap_high_check();
+
+ return 0;
+}
+
+static int memblock_alloc_nid_checks_internal(int flags)
+{
+ alloc_nid_test_flags = flags;
+
+ prefix_reset();
+ prefix_push(get_memblock_alloc_try_nid_name(flags));
+
+ reset_memblock_attributes();
+ dummy_physical_memory_init();
+
+ memblock_alloc_nid_range_checks();
+ memblock_alloc_nid_numa_checks();
+
dummy_physical_memory_cleanup();

prefix_pop();
diff --git a/tools/testing/memblock/tests/alloc_nid_api.h b/tools/testing/memblock/tests/alloc_nid_api.h
index b35cf3c3f489..92d07d230e18 100644
--- a/tools/testing/memblock/tests/alloc_nid_api.h
+++ b/tools/testing/memblock/tests/alloc_nid_api.h
@@ -5,5 +5,21 @@
#include "common.h"

int memblock_alloc_nid_checks(void);
+int __memblock_alloc_nid_numa_checks(void);
+
+#ifdef CONFIG_NUMA
+static inline int memblock_alloc_nid_numa_checks(void)
+{
+ __memblock_alloc_nid_numa_checks();
+ return 0;
+}
+
+#else
+static inline int memblock_alloc_nid_numa_checks(void)
+{
+ return 0;
+}
+
+#endif /* CONFIG_NUMA */

#endif
diff --git a/tools/testing/memblock/tests/common.h b/tools/testing/memblock/tests/common.h
index caf01543a315..da9a2d4afd54 100644
--- a/tools/testing/memblock/tests/common.h
+++ b/tools/testing/memblock/tests/common.h
@@ -59,6 +59,19 @@ enum test_flags {
assert((_expected) < (_seen)); \
} while (0)

+/**
+ * ASSERT_LE():
+ * Check the condition
+ * @_expected <= @_seen
+ * If false, print failed test message (if running with --verbose) and then
+ * assert.
+ */
+#define ASSERT_LE(_expected, _seen) do { \
+ if ((_expected) > (_seen)) \
+ test_fail(); \
+ assert((_expected) <= (_seen)); \
+} while (0)
+
/**
* ASSERT_MEM_EQ():
* Check that the first @_size bytes of @_seen are all equal to @_expected.
@@ -100,6 +113,11 @@ struct region {
phys_addr_t size;
};

+static inline phys_addr_t __maybe_unused region_end(struct memblock_region *rgn)
+{
+ return rgn->base + rgn->size;
+}
+
void reset_memblock_regions(void);
void reset_memblock_attributes(void);
void setup_memblock(void);
--
2.25.1

2022-09-04 04:39:20

by Rebecca Mckeever

Subject: [PATCH v4 4/4] memblock tests: add generic NUMA tests for memblock_alloc_try_nid*

Add tests for memblock_alloc_try_nid() and memblock_alloc_try_nid_raw()
where the simulated physical memory is set up with multiple NUMA nodes.
Additionally, two of these tests set nid != NUMA_NO_NODE. All tests are
run for both top-down and bottom-up allocation directions.

The tested scenarios are:

Range unrestricted:
- region cannot be allocated:
+ none of the nodes have enough memory to allocate the region

Range restricted:
- region can be allocated in the specific node requested without dropping
min_addr:
+ the range fully overlaps with the node, and there are adjacent
reserved regions
- region cannot be allocated:
+ nid is set to NUMA_NO_NODE and the total range can fit the region,
but the range is split between two nodes and everything else is
reserved

Signed-off-by: Rebecca Mckeever <[email protected]>
---
tools/testing/memblock/tests/alloc_nid_api.c | 197 +++++++++++++++++++
1 file changed, 197 insertions(+)

diff --git a/tools/testing/memblock/tests/alloc_nid_api.c b/tools/testing/memblock/tests/alloc_nid_api.c
index 133d77e83e0e..8f25c90c2300 100644
--- a/tools/testing/memblock/tests/alloc_nid_api.c
+++ b/tools/testing/memblock/tests/alloc_nid_api.c
@@ -2349,6 +2349,173 @@ static int alloc_try_nid_bottom_up_numa_no_overlap_high_check(void)
return 0;
}

+/*
+ * A test that tries to allocate a memory region in a specific NUMA node that
+ * does not have enough memory to allocate a region of the requested size.
+ * Additionally, none of the nodes have enough memory to allocate the region:
+ *
+ * +-----------------------------------+
+ * | new |
+ * +-----------------------------------+
+ * |-------+-------+-------+-------+-------+-------+-------+-------|
+ * | node0 | node1 | node2 | node3 | node4 | node5 | node6 | node7 |
+ * +-------+-------+-------+-------+-------+-------+-------+-------+
+ *
+ * Expect no allocation to happen.
+ */
+static int alloc_try_nid_numa_large_region_generic_check(void)
+{
+ int nid_req = 3;
+ void *allocated_ptr = NULL;
+ phys_addr_t size = MEM_SIZE / SZ_2;
+ phys_addr_t min_addr;
+ phys_addr_t max_addr;
+
+ PREFIX_PUSH();
+ setup_numa_memblock(node_fractions);
+
+ min_addr = memblock_start_of_DRAM();
+ max_addr = memblock_end_of_DRAM();
+
+ allocated_ptr = run_memblock_alloc_try_nid(size, SMP_CACHE_BYTES,
+ min_addr, max_addr, nid_req);
+ ASSERT_EQ(allocated_ptr, NULL);
+
+ test_pass_pop();
+
+ return 0;
+}
+
+/*
+ * A test that tries to allocate memory within min_addr and max_addr range when
+ * there are two reserved regions at the borders. The requested node starts at
+ * min_addr and ends at max_addr and is the same size as the region to be
+ * allocated:
+ *
+ * min_addr
+ * | max_addr
+ * | |
+ * v v
+ * | +-----------+-----------------------+-----------------------|
+ * | | node5 | requested | node7 |
+ * +------+-----------+-----------------------+-----------------------+
+ * + +
+ * | +----+-----------------------+----+ |
+ * | | r2 | new | r1 | |
+ * +-------------+----+-----------------------+----+------------------+
+ *
+ * Expect to merge all of the regions into one. The region counter and total
+ * size fields get updated.
+ */
+static int alloc_try_nid_numa_reserved_full_merge_generic_check(void)
+{
+ int nid_req = 6;
+ int nid_next = nid_req + 1;
+ struct memblock_region *new_rgn = &memblock.reserved.regions[0];
+ struct memblock_region *req_node = &memblock.memory.regions[nid_req];
+ struct memblock_region *next_node = &memblock.memory.regions[nid_next];
+ void *allocated_ptr = NULL;
+ struct region r1, r2;
+ phys_addr_t size = req_node->size;
+ phys_addr_t total_size;
+ phys_addr_t max_addr;
+ phys_addr_t min_addr;
+
+ PREFIX_PUSH();
+ setup_numa_memblock(node_fractions);
+
+ r1.base = next_node->base;
+ r1.size = SZ_128;
+
+ r2.size = SZ_128;
+ r2.base = r1.base - (size + r2.size);
+
+ total_size = r1.size + r2.size + size;
+ min_addr = r2.base + r2.size;
+ max_addr = r1.base;
+
+ memblock_reserve(r1.base, r1.size);
+ memblock_reserve(r2.base, r2.size);
+
+ allocated_ptr = run_memblock_alloc_try_nid(size, SMP_CACHE_BYTES,
+ min_addr, max_addr, nid_req);
+
+ ASSERT_NE(allocated_ptr, NULL);
+ assert_mem_content(allocated_ptr, size, alloc_nid_test_flags);
+
+ ASSERT_EQ(new_rgn->size, total_size);
+ ASSERT_EQ(new_rgn->base, r2.base);
+
+ ASSERT_LE(new_rgn->base, req_node->base);
+ ASSERT_LE(region_end(req_node), region_end(new_rgn));
+
+ ASSERT_EQ(memblock.reserved.cnt, 1);
+ ASSERT_EQ(memblock.reserved.total_size, total_size);
+
+ test_pass_pop();
+
+ return 0;
+}
+
+/*
+ * A test that tries to allocate memory within min_addr and max_add range,
+ * where the total range can fit the region, but it is split between two nodes
+ * and everything else is reserved. Additionally, nid is set to NUMA_NO_NODE
+ * instead of requesting a specific node:
+ *
+ * +-----------+
+ * | new |
+ * +-----------+
+ * | +---------------------+-----------|
+ * | | prev node | next node |
+ * +------+---------------------+-----------+
+ * + +
+ * |----------------------+ +-----|
+ * | r1 | | r2 |
+ * +----------------------+-----------+-----+
+ * ^ ^
+ * | |
+ * | max_addr
+ * |
+ * min_addr
+ *
+ * Expect no allocation to happen.
+ */
+static int alloc_try_nid_numa_split_all_reserved_generic_check(void)
+{
+ void *allocated_ptr = NULL;
+ struct memblock_region *next_node = &memblock.memory.regions[7];
+ struct region r1, r2;
+ phys_addr_t size = SZ_256;
+ phys_addr_t max_addr;
+ phys_addr_t min_addr;
+
+ PREFIX_PUSH();
+ setup_numa_memblock(node_fractions);
+
+ r2.base = next_node->base + SZ_128;
+ r2.size = memblock_end_of_DRAM() - r2.base;
+
+ r1.size = MEM_SIZE - (r2.size + size);
+ r1.base = memblock_start_of_DRAM();
+
+ min_addr = r1.base + r1.size;
+ max_addr = r2.base;
+
+ memblock_reserve(r1.base, r1.size);
+ memblock_reserve(r2.base, r2.size);
+
+ allocated_ptr = run_memblock_alloc_try_nid(size, SMP_CACHE_BYTES,
+ min_addr, max_addr,
+ NUMA_NO_NODE);
+
+ ASSERT_EQ(allocated_ptr, NULL);
+
+ test_pass_pop();
+
+ return 0;
+}
+
/* Test case wrappers for NUMA tests */
static int alloc_try_nid_numa_simple_check(void)
{
@@ -2460,6 +2627,33 @@ static int alloc_try_nid_numa_no_overlap_high_check(void)
return 0;
}

+static int alloc_try_nid_numa_large_region_check(void)
+{
+ test_print("\tRunning %s...\n", __func__);
+ run_top_down(alloc_try_nid_numa_large_region_generic_check);
+ run_bottom_up(alloc_try_nid_numa_large_region_generic_check);
+
+ return 0;
+}
+
+static int alloc_try_nid_numa_reserved_full_merge_check(void)
+{
+ test_print("\tRunning %s...\n", __func__);
+ run_top_down(alloc_try_nid_numa_reserved_full_merge_generic_check);
+ run_bottom_up(alloc_try_nid_numa_reserved_full_merge_generic_check);
+
+ return 0;
+}
+
+static int alloc_try_nid_numa_split_all_reserved_check(void)
+{
+ test_print("\tRunning %s...\n", __func__);
+ run_top_down(alloc_try_nid_numa_split_all_reserved_generic_check);
+ run_bottom_up(alloc_try_nid_numa_split_all_reserved_generic_check);
+
+ return 0;
+}
+
int __memblock_alloc_nid_numa_checks(void)
{
test_print("Running %s NUMA tests...\n",
@@ -2476,6 +2670,9 @@ int __memblock_alloc_nid_numa_checks(void)
alloc_try_nid_numa_no_overlap_split_check();
alloc_try_nid_numa_no_overlap_low_check();
alloc_try_nid_numa_no_overlap_high_check();
+ alloc_try_nid_numa_large_region_check();
+ alloc_try_nid_numa_reserved_full_merge_check();
+ alloc_try_nid_numa_split_all_reserved_check();

return 0;
}
--
2.25.1

2022-09-04 04:40:03

by Rebecca Mckeever

Subject: [PATCH v4 1/4] memblock tests: add simulation of physical memory with multiple NUMA nodes

Add function setup_numa_memblock() for setting up a memory layout with
multiple NUMA nodes in a previously allocated dummy physical memory.
This function can be used in place of setup_memblock() in tests that need
to simulate a NUMA system.

setup_numa_memblock():
- allows for setting up a memory layout by specifying the fraction of
MEM_SIZE in each node

Set CONFIG_NODES_SHIFT to 4 when building with NUMA=1 to allow for up to
16 NUMA nodes.

Signed-off-by: Rebecca Mckeever <[email protected]>
---
.../testing/memblock/scripts/Makefile.include | 2 +-
tools/testing/memblock/tests/common.c | 29 +++++++++++++++++++
tools/testing/memblock/tests/common.h | 4 ++-
3 files changed, 33 insertions(+), 2 deletions(-)

diff --git a/tools/testing/memblock/scripts/Makefile.include b/tools/testing/memblock/scripts/Makefile.include
index aa6d82d56a23..998281723590 100644
--- a/tools/testing/memblock/scripts/Makefile.include
+++ b/tools/testing/memblock/scripts/Makefile.include
@@ -3,7 +3,7 @@

# Simulate CONFIG_NUMA=y
ifeq ($(NUMA), 1)
- CFLAGS += -D CONFIG_NUMA
+ CFLAGS += -D CONFIG_NUMA -D CONFIG_NODES_SHIFT=4
endif

# Use 32 bit physical addresses.
diff --git a/tools/testing/memblock/tests/common.c b/tools/testing/memblock/tests/common.c
index eec6901081af..b6110df21b2a 100644
--- a/tools/testing/memblock/tests/common.c
+++ b/tools/testing/memblock/tests/common.c
@@ -72,6 +72,35 @@ void setup_memblock(void)
fill_memblock();
}

+/**
+ * setup_numa_memblock:
+ * Set up a memory layout with multiple NUMA nodes in a previously allocated
+ * dummy physical memory.
+ * @nodes: an array containing the denominators of the fractions of MEM_SIZE
+ * contained in each node (e.g., if nodes[0] = SZ_8, node 0 will
+ * contain 1/8th of MEM_SIZE)
+ *
+ * The nids will be set to 0 through NUMA_NODES - 1.
+ */
+void setup_numa_memblock(const phys_addr_t nodes[])
+{
+ phys_addr_t base;
+ int flags;
+
+ reset_memblock_regions();
+ base = (phys_addr_t)memory_block.base;
+ flags = (movable_node_is_enabled()) ? MEMBLOCK_NONE : MEMBLOCK_HOTPLUG;
+
+ for (int i = 0; i < NUMA_NODES; i++) {
+ assert(nodes[i] <= MEM_SIZE && nodes[i] > 0);
+ phys_addr_t size = MEM_SIZE / nodes[i];
+
+ memblock_add_node(base, size, i, flags);
+ base += size;
+ }
+ fill_memblock();
+}
+
void dummy_physical_memory_init(void)
{
memory_block.base = malloc(MEM_SIZE);
diff --git a/tools/testing/memblock/tests/common.h b/tools/testing/memblock/tests/common.h
index 78128e109a95..caf01543a315 100644
--- a/tools/testing/memblock/tests/common.h
+++ b/tools/testing/memblock/tests/common.h
@@ -10,7 +10,8 @@
#include <linux/printk.h>
#include <../selftests/kselftest.h>

-#define MEM_SIZE SZ_16K
+#define MEM_SIZE SZ_16K
+#define NUMA_NODES 8

enum test_flags {
/* No special request. */
@@ -102,6 +103,7 @@ struct region {
void reset_memblock_regions(void);
void reset_memblock_attributes(void);
void setup_memblock(void);
+void setup_numa_memblock(const phys_addr_t nodes[]);
void dummy_physical_memory_init(void);
void dummy_physical_memory_cleanup(void);
void parse_args(int argc, char **argv);
--
2.25.1

2022-09-04 04:41:29

by Rebecca Mckeever

Subject: [PATCH v4 3/4] memblock tests: add bottom-up NUMA tests for memblock_alloc_try_nid*

Add tests for memblock_alloc_try_nid() and memblock_alloc_try_nid_raw()
where the simulated physical memory is set up with multiple NUMA nodes.
Additionally, all of these tests set nid != NUMA_NO_NODE. These tests are
run with a bottom-up allocation direction.

The tested scenarios are:

Range unrestricted:
- region can be allocated in the specific node requested:
+ there are no previously reserved regions
+ the requested node is partially reserved but has enough space
- the specific node requested cannot accommodate the request, but the
region can be allocated in a different node:
+ there are no previously reserved regions, but node is too small
+ the requested node is fully reserved
+ the requested node is partially reserved and does not have
enough space

Range restricted:
- region can be allocated in the specific node requested after dropping
min_addr:
+ range partially overlaps with two different nodes, where the first
node is the requested node
+ range partially overlaps with two different nodes, where the
requested node ends before min_addr
- region cannot be allocated in the specific node requested, but it can be
allocated in the requested range:
+ range overlaps with multiple nodes along node boundaries, and the
requested node ends before min_addr
+ range overlaps with multiple nodes along node boundaries, and the
requested node starts after max_addr
- region cannot be allocated in the specific node requested, but it can be
allocated after dropping min_addr:
+ range partially overlaps with two different nodes, where the
second node is the requested node

Signed-off-by: Rebecca Mckeever <[email protected]>
---
tools/testing/memblock/tests/alloc_nid_api.c | 568 +++++++++++++++++++
1 file changed, 568 insertions(+)

diff --git a/tools/testing/memblock/tests/alloc_nid_api.c b/tools/testing/memblock/tests/alloc_nid_api.c
index a639e9d3e098..133d77e83e0e 100644
--- a/tools/testing/memblock/tests/alloc_nid_api.c
+++ b/tools/testing/memblock/tests/alloc_nid_api.c
@@ -1801,12 +1801,562 @@ static int alloc_try_nid_top_down_numa_no_overlap_high_check(void)
return 0;
}

+/*
+ * A test that tries to allocate a memory region in a specific NUMA node that
+ * has enough memory to allocate a region of the requested size.
+ * Expect to allocate an aligned region at the beginning of the requested node.
+ */
+static int alloc_try_nid_bottom_up_numa_simple_check(void)
+{
+ int nid_req = 3;
+ struct memblock_region *new_rgn = &memblock.reserved.regions[0];
+ struct memblock_region *req_node = &memblock.memory.regions[nid_req];
+ void *allocated_ptr = NULL;
+ phys_addr_t size;
+ phys_addr_t min_addr;
+ phys_addr_t max_addr;
+
+ PREFIX_PUSH();
+ setup_numa_memblock(node_fractions);
+
+ ASSERT_LE(SZ_4, req_node->size);
+ size = req_node->size / SZ_4;
+ min_addr = memblock_start_of_DRAM();
+ max_addr = memblock_end_of_DRAM();
+
+ allocated_ptr = run_memblock_alloc_try_nid(size, SMP_CACHE_BYTES,
+ min_addr, max_addr, nid_req);
+
+ ASSERT_NE(allocated_ptr, NULL);
+ assert_mem_content(allocated_ptr, size, alloc_nid_test_flags);
+
+ ASSERT_EQ(new_rgn->size, size);
+ ASSERT_EQ(new_rgn->base, req_node->base);
+ ASSERT_LE(region_end(new_rgn), region_end(req_node));
+
+ ASSERT_EQ(memblock.reserved.cnt, 1);
+ ASSERT_EQ(memblock.reserved.total_size, size);
+
+ test_pass_pop();
+
+ return 0;
+}
+
+/*
+ * A test that tries to allocate a memory region in a specific NUMA node that
+ * does not have enough memory to allocate a region of the requested size:
+ *
+ * |----------------------+-----+ |
+ * | expected | req | |
+ * +----------------------+-----+----------------+
+ *
+ * |---------+ |
+ * | rgn | |
+ * +---------+-----------------------------------+
+ *
+ * Expect to allocate an aligned region at the beginning of the first node that
+ * has enough memory (in this case, nid = 0) after falling back to NUMA_NO_NODE.
+ */
+static int alloc_try_nid_bottom_up_numa_small_node_check(void)
+{
+ int nid_req = 1;
+ int nid_exp = 0;
+ struct memblock_region *new_rgn = &memblock.reserved.regions[0];
+ struct memblock_region *req_node = &memblock.memory.regions[nid_req];
+ struct memblock_region *exp_node = &memblock.memory.regions[nid_exp];
+ void *allocated_ptr = NULL;
+ phys_addr_t size;
+ phys_addr_t min_addr;
+ phys_addr_t max_addr;
+
+ PREFIX_PUSH();
+ setup_numa_memblock(node_fractions);
+
+ size = SZ_2 * req_node->size;
+ min_addr = memblock_start_of_DRAM();
+ max_addr = memblock_end_of_DRAM();
+
+ allocated_ptr = run_memblock_alloc_try_nid(size, SMP_CACHE_BYTES,
+ min_addr, max_addr, nid_req);
+
+ ASSERT_NE(allocated_ptr, NULL);
+ assert_mem_content(allocated_ptr, size, alloc_nid_test_flags);
+
+ ASSERT_EQ(new_rgn->size, size);
+ ASSERT_EQ(new_rgn->base, exp_node->base);
+ ASSERT_LE(region_end(new_rgn), region_end(exp_node));
+
+ ASSERT_EQ(memblock.reserved.cnt, 1);
+ ASSERT_EQ(memblock.reserved.total_size, size);
+
+ test_pass_pop();
+
+ return 0;
+}
+
+/*
+ * A test that tries to allocate a memory region in a specific NUMA node that
+ * is fully reserved:
+ *
+ * |----------------------+ +-----------+ |
+ * | expected | | requested | |
+ * +----------------------+-----+-----------+--------------------+
+ *
+ * |-----------+ +-----------+ |
+ * | new | | reserved | |
+ * +-----------+----------------+-----------+--------------------+
+ *
+ * Expect to allocate an aligned region at the beginning of the first node that
+ * is large enough and has enough unreserved memory (in this case, nid = 0)
+ * after falling back to NUMA_NO_NODE. The region count and total size get
+ * updated.
+ */
+static int alloc_try_nid_bottom_up_numa_node_reserved_check(void)
+{
+ int nid_req = 2;
+ int nid_exp = 0;
+ struct memblock_region *new_rgn = &memblock.reserved.regions[0];
+ struct memblock_region *req_node = &memblock.memory.regions[nid_req];
+ struct memblock_region *exp_node = &memblock.memory.regions[nid_exp];
+ void *allocated_ptr = NULL;
+ phys_addr_t size;
+ phys_addr_t min_addr;
+ phys_addr_t max_addr;
+
+ PREFIX_PUSH();
+ setup_numa_memblock(node_fractions);
+
+ size = req_node->size;
+ min_addr = memblock_start_of_DRAM();
+ max_addr = memblock_end_of_DRAM();
+
+ memblock_reserve(req_node->base, req_node->size);
+ allocated_ptr = run_memblock_alloc_try_nid(size, SMP_CACHE_BYTES,
+ min_addr, max_addr, nid_req);
+
+ ASSERT_NE(allocated_ptr, NULL);
+ assert_mem_content(allocated_ptr, size, alloc_nid_test_flags);
+
+ ASSERT_EQ(new_rgn->size, size);
+ ASSERT_EQ(new_rgn->base, exp_node->base);
+ ASSERT_LE(region_end(new_rgn), region_end(exp_node));
+
+ ASSERT_EQ(memblock.reserved.cnt, 2);
+ ASSERT_EQ(memblock.reserved.total_size, size + req_node->size);
+
+ test_pass_pop();
+
+ return 0;
+}
+
+/*
+ * A test that tries to allocate a memory region in a specific NUMA node that
+ * is partially reserved but has enough memory for the allocated region:
+ *
+ * | +---------------------------------------+ |
+ * | | requested | |
+ * +-----------+---------------------------------------+---------+
+ *
+ * | +------------------+-----+ |
+ * | | reserved | new | |
+ * +-----------+------------------+-----+------------------------+
+ *
+ * Expect to allocate an aligned region in the requested node that merges with
+ * the existing reserved region. The total size gets updated.
+ */
+static int alloc_try_nid_bottom_up_numa_part_reserved_check(void)
+{
+ int nid_req = 4;
+ struct memblock_region *new_rgn = &memblock.reserved.regions[0];
+ struct memblock_region *req_node = &memblock.memory.regions[nid_req];
+ void *allocated_ptr = NULL;
+ struct region r1;
+ phys_addr_t size;
+ phys_addr_t min_addr;
+ phys_addr_t max_addr;
+ phys_addr_t total_size;
+
+ PREFIX_PUSH();
+ setup_numa_memblock(node_fractions);
+
+ ASSERT_LE(SZ_8, req_node->size);
+ r1.base = req_node->base;
+ r1.size = req_node->size / SZ_2;
+ size = r1.size / SZ_4;
+ min_addr = memblock_start_of_DRAM();
+ max_addr = memblock_end_of_DRAM();
+ total_size = size + r1.size;
+
+ memblock_reserve(r1.base, r1.size);
+ allocated_ptr = run_memblock_alloc_try_nid(size, SMP_CACHE_BYTES,
+ min_addr, max_addr, nid_req);
+
+ ASSERT_NE(allocated_ptr, NULL);
+ assert_mem_content(allocated_ptr, size, alloc_nid_test_flags);
+
+ ASSERT_EQ(new_rgn->size, total_size);
+ ASSERT_EQ(new_rgn->base, req_node->base);
+ ASSERT_LE(region_end(new_rgn), region_end(req_node));
+
+ ASSERT_EQ(memblock.reserved.cnt, 1);
+ ASSERT_EQ(memblock.reserved.total_size, total_size);
+
+ test_pass_pop();
+
+ return 0;
+}
+
+/*
+ * A test that tries to allocate a memory region in a specific NUMA node that
+ * is partially reserved and does not have enough contiguous memory for the
+ * allocated region:
+ *
+ * |----------------------+ +-----------------------+ |
+ * | expected | | requested | |
+ * +----------------------+-------+-----------------------+---------+
+ *
+ * |-----------+ +----------+ |
+ * | new | | reserved | |
+ * +-----------+------------------------+----------+----------------+
+ *
+ * Expect to allocate an aligned region at the beginning of the first
+ * node that is large enough and has enough unreserved memory (in this case,
+ * nid = 0) after falling back to NUMA_NO_NODE. The region count and total size
+ * get updated.
+ */
+static int alloc_try_nid_bottom_up_numa_part_reserved_fallback_check(void)
+{
+ int nid_req = 4;
+ int nid_exp = 0;
+ struct memblock_region *new_rgn = &memblock.reserved.regions[0];
+ struct memblock_region *req_node = &memblock.memory.regions[nid_req];
+ struct memblock_region *exp_node = &memblock.memory.regions[nid_exp];
+ void *allocated_ptr = NULL;
+ struct region r1;
+ phys_addr_t size;
+ phys_addr_t min_addr;
+ phys_addr_t max_addr;
+
+ PREFIX_PUSH();
+ setup_numa_memblock(node_fractions);
+
+ ASSERT_LE(SZ_4, req_node->size);
+ size = req_node->size / SZ_2;
+ r1.base = req_node->base + (size / SZ_2);
+ r1.size = size;
+
+ min_addr = memblock_start_of_DRAM();
+ max_addr = memblock_end_of_DRAM();
+
+ memblock_reserve(r1.base, r1.size);
+ allocated_ptr = run_memblock_alloc_try_nid(size, SMP_CACHE_BYTES,
+ min_addr, max_addr, nid_req);
+
+ ASSERT_NE(allocated_ptr, NULL);
+ assert_mem_content(allocated_ptr, size, alloc_nid_test_flags);
+
+ ASSERT_EQ(new_rgn->size, size);
+ ASSERT_EQ(new_rgn->base, exp_node->base);
+ ASSERT_LE(region_end(new_rgn), region_end(exp_node));
+
+ ASSERT_EQ(memblock.reserved.cnt, 2);
+ ASSERT_EQ(memblock.reserved.total_size, size + r1.size);
+
+ test_pass_pop();
+
+ return 0;
+}
+
+/*
+ * A test that tries to allocate a memory region that spans over the min_addr
+ * and max_addr range and overlaps with two different nodes, where the first
+ * node is the requested node:
+ *
+ * min_addr
+ * | max_addr
+ * | |
+ * v v
+ * | +-----------------------+-----------+ |
+ * | | requested | node3 | |
+ * +-----------+-----------------------+-----------+--------------+
+ * + +
+ * | +-----------+ |
+ * | | rgn | |
+ * +-----------+-----------+--------------------------------------+
+ *
+ * Expect to drop the lower limit and allocate a memory region at the beginning
+ * of the requested node.
+ */
+static int alloc_try_nid_bottom_up_numa_split_range_low_check(void)
+{
+ int nid_req = 2;
+ struct memblock_region *new_rgn = &memblock.reserved.regions[0];
+ struct memblock_region *req_node = &memblock.memory.regions[nid_req];
+ void *allocated_ptr = NULL;
+ phys_addr_t size = SZ_512;
+ phys_addr_t min_addr;
+ phys_addr_t max_addr;
+ phys_addr_t req_node_end;
+
+ PREFIX_PUSH();
+ setup_numa_memblock(node_fractions);
+
+ req_node_end = region_end(req_node);
+ min_addr = req_node_end - SZ_256;
+ max_addr = min_addr + size;
+
+ allocated_ptr = run_memblock_alloc_try_nid(size, SMP_CACHE_BYTES,
+ min_addr, max_addr, nid_req);
+
+ ASSERT_NE(allocated_ptr, NULL);
+ assert_mem_content(allocated_ptr, size, alloc_nid_test_flags);
+
+ ASSERT_EQ(new_rgn->size, size);
+ ASSERT_EQ(new_rgn->base, req_node->base);
+ ASSERT_LE(region_end(new_rgn), req_node_end);
+
+ ASSERT_EQ(memblock.reserved.cnt, 1);
+ ASSERT_EQ(memblock.reserved.total_size, size);
+
+ test_pass_pop();
+
+ return 0;
+}
+
+/*
+ * A test that tries to allocate a memory region that spans over the min_addr
+ * and max_addr range and overlaps with two different nodes, where the second
+ * node is the requested node:
+ *
+ * min_addr
+ * | max_addr
+ * | |
+ * v v
+ * |------------------+ +----------------------+---------+ |
+ * | expected | | previous |requested| |
+ * +------------------+--------+----------------------+---------+------+
+ * + +
+ * |---------+ |
+ * | rgn | |
+ * +---------+---------------------------------------------------------+
+ *
+ * Expect to drop the lower limit and allocate a memory region at the beginning
+ * of the first node that has enough memory.
+ */
+static int alloc_try_nid_bottom_up_numa_split_range_high_check(void)
+{
+ int nid_req = 3;
+ int nid_exp = 0;
+ struct memblock_region *new_rgn = &memblock.reserved.regions[0];
+ struct memblock_region *req_node = &memblock.memory.regions[nid_req];
+ struct memblock_region *exp_node = &memblock.memory.regions[nid_exp];
+ void *allocated_ptr = NULL;
+ phys_addr_t size = SZ_512;
+ phys_addr_t min_addr;
+ phys_addr_t max_addr;
+ phys_addr_t exp_node_end;
+
+ PREFIX_PUSH();
+ setup_numa_memblock(node_fractions);
+
+ exp_node_end = region_end(req_node);
+ min_addr = req_node->base - SZ_256;
+ max_addr = min_addr + size;
+
+ allocated_ptr = run_memblock_alloc_try_nid(size, SMP_CACHE_BYTES,
+ min_addr, max_addr, nid_req);
+
+ ASSERT_NE(allocated_ptr, NULL);
+ assert_mem_content(allocated_ptr, size, alloc_nid_test_flags);
+
+ ASSERT_EQ(new_rgn->size, size);
+ ASSERT_EQ(new_rgn->base, exp_node->base);
+ ASSERT_LE(region_end(new_rgn), exp_node_end);
+
+ ASSERT_EQ(memblock.reserved.cnt, 1);
+ ASSERT_EQ(memblock.reserved.total_size, size);
+
+ test_pass_pop();
+
+ return 0;
+}
+
+/*
+ * A test that tries to allocate a memory region that spans over the min_addr
+ * and max_addr range and overlaps with two different nodes, where the requested
+ * node ends before min_addr:
+ *
+ * min_addr
+ * | max_addr
+ * | |
+ * v v
+ * | +---------------+ +-------------+---------+ |
+ * | | requested | | node1 | node2 | |
+ * +----+---------------+--------+-------------+---------+---------+
+ * + +
+ * | +---------+ |
+ * | | rgn | |
+ * +----+---------+------------------------------------------------+
+ *
+ * Expect to drop the lower limit and allocate a memory region that starts at
+ * the beginning of the requested node.
+ */
+static int alloc_try_nid_bottom_up_numa_no_overlap_split_check(void)
+{
+ int nid_req = 2;
+ struct memblock_region *new_rgn = &memblock.reserved.regions[0];
+ struct memblock_region *req_node = &memblock.memory.regions[nid_req];
+ struct memblock_region *node2 = &memblock.memory.regions[6];
+ void *allocated_ptr = NULL;
+ phys_addr_t size;
+ phys_addr_t min_addr;
+ phys_addr_t max_addr;
+
+ PREFIX_PUSH();
+ setup_numa_memblock(node_fractions);
+
+ size = SZ_512;
+ min_addr = node2->base - SZ_256;
+ max_addr = min_addr + size;
+
+ allocated_ptr = run_memblock_alloc_try_nid(size, SMP_CACHE_BYTES,
+ min_addr, max_addr, nid_req);
+
+ ASSERT_NE(allocated_ptr, NULL);
+ assert_mem_content(allocated_ptr, size, alloc_nid_test_flags);
+
+ ASSERT_EQ(new_rgn->size, size);
+ ASSERT_EQ(new_rgn->base, req_node->base);
+ ASSERT_LE(region_end(new_rgn), region_end(req_node));
+
+ ASSERT_EQ(memblock.reserved.cnt, 1);
+ ASSERT_EQ(memblock.reserved.total_size, size);
+
+ test_pass_pop();
+
+ return 0;
+}
+
+/*
+ * A test that tries to allocate memory within min_addr and max_add range when
+ * the requested node and the range do not overlap, and requested node ends
+ * before min_addr. The range overlaps with multiple nodes along node
+ * boundaries:
+ *
+ * min_addr
+ * | max_addr
+ * | |
+ * v v
+ * |-----------+ +----------+----...----+----------+ |
+ * | requested | | min node | ... | max node | |
+ * +-----------+-----------+----------+----...----+----------+------+
+ * + +
+ * | +-----+ |
+ * | | rgn | |
+ * +-----------------------+-----+----------------------------------+
+ *
+ * Expect to allocate a memory region at the beginning of the first node
+ * in the range after falling back to NUMA_NO_NODE.
+ */
+static int alloc_try_nid_bottom_up_numa_no_overlap_low_check(void)
+{
+ int nid_req = 0;
+ struct memblock_region *new_rgn = &memblock.reserved.regions[0];
+ struct memblock_region *min_node = &memblock.memory.regions[2];
+ struct memblock_region *max_node = &memblock.memory.regions[5];
+ void *allocated_ptr = NULL;
+ phys_addr_t size = SZ_64;
+ phys_addr_t max_addr;
+ phys_addr_t min_addr;
+
+ PREFIX_PUSH();
+ setup_numa_memblock(node_fractions);
+
+ min_addr = min_node->base;
+ max_addr = region_end(max_node);
+
+ allocated_ptr = run_memblock_alloc_try_nid(size, SMP_CACHE_BYTES,
+ min_addr, max_addr, nid_req);
+
+ ASSERT_NE(allocated_ptr, NULL);
+ assert_mem_content(allocated_ptr, size, alloc_nid_test_flags);
+
+ ASSERT_EQ(new_rgn->size, size);
+ ASSERT_EQ(new_rgn->base, min_addr);
+ ASSERT_LE(region_end(new_rgn), region_end(min_node));
+
+ ASSERT_EQ(memblock.reserved.cnt, 1);
+ ASSERT_EQ(memblock.reserved.total_size, size);
+
+ test_pass_pop();
+
+ return 0;
+}
+
+/*
+ * A test that tries to allocate memory within min_addr and max_add range when
+ * the requested node and the range do not overlap, and requested node starts
+ * after max_addr. The range overlaps with multiple nodes along node
+ * boundaries:
+ *
+ * min_addr
+ * | max_addr
+ * | |
+ * v v
+ * | +----------+----...----+----------+ +---------+ |
+ * | | min node | ... | max node | |requested| |
+ * +-----+----------+----...----+----------+---------+---------+---+
+ * + +
+ * | +-----+ |
+ * | | rgn | |
+ * +-----+-----+---------------------------------------------------+
+ *
+ * Expect to allocate a memory region at the beginning of the first node
+ * in the range after falling back to NUMA_NO_NODE.
+ */
+static int alloc_try_nid_bottom_up_numa_no_overlap_high_check(void)
+{
+ int nid_req = 7;
+ struct memblock_region *new_rgn = &memblock.reserved.regions[0];
+ struct memblock_region *min_node = &memblock.memory.regions[2];
+ struct memblock_region *max_node = &memblock.memory.regions[5];
+ void *allocated_ptr = NULL;
+ phys_addr_t size = SZ_64;
+ phys_addr_t max_addr;
+ phys_addr_t min_addr;
+
+ PREFIX_PUSH();
+ setup_numa_memblock(node_fractions);
+
+ min_addr = min_node->base;
+ max_addr = region_end(max_node);
+
+ allocated_ptr = run_memblock_alloc_try_nid(size, SMP_CACHE_BYTES,
+ min_addr, max_addr, nid_req);
+
+ ASSERT_NE(allocated_ptr, NULL);
+ assert_mem_content(allocated_ptr, size, alloc_nid_test_flags);
+
+ ASSERT_EQ(new_rgn->size, size);
+ ASSERT_EQ(new_rgn->base, min_addr);
+ ASSERT_LE(region_end(new_rgn), region_end(min_node));
+
+ ASSERT_EQ(memblock.reserved.cnt, 1);
+ ASSERT_EQ(memblock.reserved.total_size, size);
+
+ test_pass_pop();
+
+ return 0;
+}
+
/* Test case wrappers for NUMA tests */
static int alloc_try_nid_numa_simple_check(void)
{
test_print("\tRunning %s...\n", __func__);
memblock_set_bottom_up(false);
alloc_try_nid_top_down_numa_simple_check();
+ memblock_set_bottom_up(true);
+ alloc_try_nid_bottom_up_numa_simple_check();

return 0;
}
@@ -1816,6 +2366,8 @@ static int alloc_try_nid_numa_small_node_check(void)
test_print("\tRunning %s...\n", __func__);
memblock_set_bottom_up(false);
alloc_try_nid_top_down_numa_small_node_check();
+ memblock_set_bottom_up(true);
+ alloc_try_nid_bottom_up_numa_small_node_check();

return 0;
}
@@ -1825,6 +2377,8 @@ static int alloc_try_nid_numa_node_reserved_check(void)
test_print("\tRunning %s...\n", __func__);
memblock_set_bottom_up(false);
alloc_try_nid_top_down_numa_node_reserved_check();
+ memblock_set_bottom_up(true);
+ alloc_try_nid_bottom_up_numa_node_reserved_check();

return 0;
}
@@ -1834,6 +2388,8 @@ static int alloc_try_nid_numa_part_reserved_check(void)
test_print("\tRunning %s...\n", __func__);
memblock_set_bottom_up(false);
alloc_try_nid_top_down_numa_part_reserved_check();
+ memblock_set_bottom_up(true);
+ alloc_try_nid_bottom_up_numa_part_reserved_check();

return 0;
}
@@ -1843,6 +2399,8 @@ static int alloc_try_nid_numa_part_reserved_fallback_check(void)
test_print("\tRunning %s...\n", __func__);
memblock_set_bottom_up(false);
alloc_try_nid_top_down_numa_part_reserved_fallback_check();
+ memblock_set_bottom_up(true);
+ alloc_try_nid_bottom_up_numa_part_reserved_fallback_check();

return 0;
}
@@ -1852,6 +2410,8 @@ static int alloc_try_nid_numa_split_range_low_check(void)
test_print("\tRunning %s...\n", __func__);
memblock_set_bottom_up(false);
alloc_try_nid_top_down_numa_split_range_low_check();
+ memblock_set_bottom_up(true);
+ alloc_try_nid_bottom_up_numa_split_range_low_check();

return 0;
}
@@ -1861,6 +2421,8 @@ static int alloc_try_nid_numa_split_range_high_check(void)
test_print("\tRunning %s...\n", __func__);
memblock_set_bottom_up(false);
alloc_try_nid_top_down_numa_split_range_high_check();
+ memblock_set_bottom_up(true);
+ alloc_try_nid_bottom_up_numa_split_range_high_check();

return 0;
}
@@ -1870,6 +2432,8 @@ static int alloc_try_nid_numa_no_overlap_split_check(void)
test_print("\tRunning %s...\n", __func__);
memblock_set_bottom_up(false);
alloc_try_nid_top_down_numa_no_overlap_split_check();
+ memblock_set_bottom_up(true);
+ alloc_try_nid_bottom_up_numa_no_overlap_split_check();

return 0;
}
@@ -1879,6 +2443,8 @@ static int alloc_try_nid_numa_no_overlap_low_check(void)
test_print("\tRunning %s...\n", __func__);
memblock_set_bottom_up(false);
alloc_try_nid_top_down_numa_no_overlap_low_check();
+ memblock_set_bottom_up(true);
+ alloc_try_nid_bottom_up_numa_no_overlap_low_check();

return 0;
}
@@ -1888,6 +2454,8 @@ static int alloc_try_nid_numa_no_overlap_high_check(void)
test_print("\tRunning %s...\n", __func__);
memblock_set_bottom_up(false);
alloc_try_nid_top_down_numa_no_overlap_high_check();
+ memblock_set_bottom_up(true);
+ alloc_try_nid_bottom_up_numa_no_overlap_high_check();

return 0;
}
--
2.25.1

2022-09-06 13:34:44

by David Hildenbrand

Subject: Re: [PATCH v4 1/4] memblock tests: add simulation of physical memory with multiple NUMA nodes

On 04.09.22 06:21, Rebecca Mckeever wrote:
> Add function setup_numa_memblock() for setting up a memory layout with
> multiple NUMA nodes in a previously allocated dummy physical memory.
> This function can be used in place of setup_memblock() in tests that need
> to simulate a NUMA system.
>
> setup_numa_memblock():
> - allows for setting up a memory layout by specifying the fraction of
> MEM_SIZE in each node
>
> Set CONFIG_NODES_SHIFT to 4 when building with NUMA=1 to allow for up to
> 16 NUMA nodes.
>
> Signed-off-by: Rebecca Mckeever <[email protected]>
> ---
> .../testing/memblock/scripts/Makefile.include | 2 +-
> tools/testing/memblock/tests/common.c | 29 +++++++++++++++++++
> tools/testing/memblock/tests/common.h | 4 ++-
> 3 files changed, 33 insertions(+), 2 deletions(-)
>
> diff --git a/tools/testing/memblock/scripts/Makefile.include b/tools/testing/memblock/scripts/Makefile.include
> index aa6d82d56a23..998281723590 100644
> --- a/tools/testing/memblock/scripts/Makefile.include
> +++ b/tools/testing/memblock/scripts/Makefile.include
> @@ -3,7 +3,7 @@
>
> # Simulate CONFIG_NUMA=y
> ifeq ($(NUMA), 1)
> - CFLAGS += -D CONFIG_NUMA
> + CFLAGS += -D CONFIG_NUMA -D CONFIG_NODES_SHIFT=4
> endif
>
> # Use 32 bit physical addresses.
> diff --git a/tools/testing/memblock/tests/common.c b/tools/testing/memblock/tests/common.c
> index eec6901081af..b6110df21b2a 100644
> --- a/tools/testing/memblock/tests/common.c
> +++ b/tools/testing/memblock/tests/common.c
> @@ -72,6 +72,35 @@ void setup_memblock(void)
> fill_memblock();
> }
>
> +/**
> + * setup_numa_memblock:
> + * Set up a memory layout with multiple NUMA nodes in a previously allocated
> + * dummy physical memory.
> + * @nodes: an array containing the denominators of the fractions of MEM_SIZE
> + * contained in each node (e.g., if nodes[0] = SZ_8, node 0 will
> + * contain 1/8th of MEM_SIZE)
> + *
> + * The nids will be set to 0 through NUMA_NODES - 1.
> + */
> +void setup_numa_memblock(const phys_addr_t nodes[])
> +{
> + phys_addr_t base;
> + int flags;
> +
> + reset_memblock_regions();
> + base = (phys_addr_t)memory_block.base;
> + flags = (movable_node_is_enabled()) ? MEMBLOCK_NONE : MEMBLOCK_HOTPLUG;
> +
> + for (int i = 0; i < NUMA_NODES; i++) {
> + assert(nodes[i] <= MEM_SIZE && nodes[i] > 0);

I think it would be even easier to get if this would just be a fraction.
E.g., instead of "1/8 * MEM_SIZE" just "1/8". All values have to add up
to 1.

... but then we'd have to mess with floats eventually, so I guess this
makes it easier to handle these fractions.


We could use "int" and simply specify the fraction in percent, like

nodes[0] = 50;
nodes[1] = 25;
nodes[2] = 25;

and everything has to add up to 100.


> + phys_addr_t size = MEM_SIZE / nodes[i];


Hmmm, assuming a single node with "MEM_SIZE", we would get size=1.

Shouldn't this be "size = nodes[i]"

?

--
Thanks,

David / dhildenb

2022-09-07 00:00:31

by Rebecca Mckeever

Subject: Re: [PATCH v4 1/4] memblock tests: add simulation of physical memory with multiple NUMA nodes

On Tue, Sep 06, 2022 at 03:17:46PM +0200, David Hildenbrand wrote:
> On 04.09.22 06:21, Rebecca Mckeever wrote:
> > Add function setup_numa_memblock() for setting up a memory layout with
> > multiple NUMA nodes in a previously allocated dummy physical memory.
> > This function can be used in place of setup_memblock() in tests that need
> > to simulate a NUMA system.
> >
> > setup_numa_memblock():
> > - allows for setting up a memory layout by specifying the fraction of
> > MEM_SIZE in each node
> >
> > Set CONFIG_NODES_SHIFT to 4 when building with NUMA=1 to allow for up to
> > 16 NUMA nodes.
> >
> > Signed-off-by: Rebecca Mckeever <[email protected]>
> > ---
> > .../testing/memblock/scripts/Makefile.include | 2 +-
> > tools/testing/memblock/tests/common.c | 29 +++++++++++++++++++
> > tools/testing/memblock/tests/common.h | 4 ++-
> > 3 files changed, 33 insertions(+), 2 deletions(-)
> >
> > diff --git a/tools/testing/memblock/scripts/Makefile.include b/tools/testing/memblock/scripts/Makefile.include
> > index aa6d82d56a23..998281723590 100644
> > --- a/tools/testing/memblock/scripts/Makefile.include
> > +++ b/tools/testing/memblock/scripts/Makefile.include
> > @@ -3,7 +3,7 @@
> > # Simulate CONFIG_NUMA=y
> > ifeq ($(NUMA), 1)
> > - CFLAGS += -D CONFIG_NUMA
> > + CFLAGS += -D CONFIG_NUMA -D CONFIG_NODES_SHIFT=4
> > endif
> > # Use 32 bit physical addresses.
> > diff --git a/tools/testing/memblock/tests/common.c b/tools/testing/memblock/tests/common.c
> > index eec6901081af..b6110df21b2a 100644
> > --- a/tools/testing/memblock/tests/common.c
> > +++ b/tools/testing/memblock/tests/common.c
> > @@ -72,6 +72,35 @@ void setup_memblock(void)
> > fill_memblock();
> > }
> > +/**
> > + * setup_numa_memblock:
> > + * Set up a memory layout with multiple NUMA nodes in a previously allocated
> > + * dummy physical memory.
> > + * @nodes: an array containing the denominators of the fractions of MEM_SIZE
> > + * contained in each node (e.g., if nodes[0] = SZ_8, node 0 will
> > + * contain 1/8th of MEM_SIZE)
> > + *
> > + * The nids will be set to 0 through NUMA_NODES - 1.
> > + */
> > +void setup_numa_memblock(const phys_addr_t nodes[])
> > +{
> > + phys_addr_t base;
> > + int flags;
> > +
> > + reset_memblock_regions();
> > + base = (phys_addr_t)memory_block.base;
> > + flags = (movable_node_is_enabled()) ? MEMBLOCK_NONE : MEMBLOCK_HOTPLUG;
> > +
> > + for (int i = 0; i < NUMA_NODES; i++) {
> > + assert(nodes[i] <= MEM_SIZE && nodes[i] > 0);
>
> I think it would be even easier to get if this would just be a fraction.
> E.g., instead of "1/8 * MEM_SIZE" just "1/8". All values have to add up to
> 1.
>
> ... but then we'd have to mess with floats eventually, so I guess this makes
> it easier to handle these fractions.
>
>
> We could use "int" and simply specify the fraction in percent, like
>
> nodes[0] = 50;
> nodes[1] = 25;
> nodes[2] = 25;
>
> and everything has to add up to 100.
>
This would still be a float for 1/8th (12.5) and 1/16th (6.25). What if
it was the "percent" of 256 (i.e., 0x100)?
>
> > + phys_addr_t size = MEM_SIZE / nodes[i];
>
>
> Hmmm, assuming a single node with "MEM_SIZE", we would get size=1.
>
For a single node of MEM_SIZE, nodes[0] would be 1.

> Shouldn't this be "size = nodes[i]"
>
> ?
No, not with the current implementation. The nodes array stores the
denominator of the fraction that will be multiplied by MEM_SIZE to
determine the size of that node (the numerator is always 1). So if the
size of the node should be 1/8 * MEM_SIZE, the nodes array just stores
the 8. I think the name of the array is misleading. Do you have any
suggestions for a better name?
>
> --
> Thanks,
>
> David / dhildenb
>
Thanks,
Rebecca

2022-09-07 09:40:24

by David Hildenbrand

Subject: Re: [PATCH v4 1/4] memblock tests: add simulation of physical memory with multiple NUMA nodes

On 07.09.22 01:43, Rebecca Mckeever wrote:
> On Tue, Sep 06, 2022 at 03:17:46PM +0200, David Hildenbrand wrote:
>> On 04.09.22 06:21, Rebecca Mckeever wrote:
>>> Add function setup_numa_memblock() for setting up a memory layout with
>>> multiple NUMA nodes in a previously allocated dummy physical memory.
>>> This function can be used in place of setup_memblock() in tests that need
>>> to simulate a NUMA system.
>>>
>>> setup_numa_memblock():
>>> - allows for setting up a memory layout by specifying the fraction of
>>> MEM_SIZE in each node
>>>
>>> Set CONFIG_NODES_SHIFT to 4 when building with NUMA=1 to allow for up to
>>> 16 NUMA nodes.
>>>
>>> Signed-off-by: Rebecca Mckeever <[email protected]>
>>> ---
>>> .../testing/memblock/scripts/Makefile.include | 2 +-
>>> tools/testing/memblock/tests/common.c | 29 +++++++++++++++++++
>>> tools/testing/memblock/tests/common.h | 4 ++-
>>> 3 files changed, 33 insertions(+), 2 deletions(-)
>>>
>>> diff --git a/tools/testing/memblock/scripts/Makefile.include b/tools/testing/memblock/scripts/Makefile.include
>>> index aa6d82d56a23..998281723590 100644
>>> --- a/tools/testing/memblock/scripts/Makefile.include
>>> +++ b/tools/testing/memblock/scripts/Makefile.include
>>> @@ -3,7 +3,7 @@
>>> # Simulate CONFIG_NUMA=y
>>> ifeq ($(NUMA), 1)
>>> - CFLAGS += -D CONFIG_NUMA
>>> + CFLAGS += -D CONFIG_NUMA -D CONFIG_NODES_SHIFT=4
>>> endif
>>> # Use 32 bit physical addresses.
>>> diff --git a/tools/testing/memblock/tests/common.c b/tools/testing/memblock/tests/common.c
>>> index eec6901081af..b6110df21b2a 100644
>>> --- a/tools/testing/memblock/tests/common.c
>>> +++ b/tools/testing/memblock/tests/common.c
>>> @@ -72,6 +72,35 @@ void setup_memblock(void)
>>> fill_memblock();
>>> }
>>> +/**
>>> + * setup_numa_memblock:
>>> + * Set up a memory layout with multiple NUMA nodes in a previously allocated
>>> + * dummy physical memory.
>>> + * @nodes: an array containing the denominators of the fractions of MEM_SIZE
>>> + * contained in each node (e.g., if nodes[0] = SZ_8, node 0 will
>>> + * contain 1/8th of MEM_SIZE)
>>> + *
>>> + * The nids will be set to 0 through NUMA_NODES - 1.
>>> + */
>>> +void setup_numa_memblock(const phys_addr_t nodes[])
>>> +{
>>> + phys_addr_t base;
>>> + int flags;
>>> +
>>> + reset_memblock_regions();
>>> + base = (phys_addr_t)memory_block.base;
>>> + flags = (movable_node_is_enabled()) ? MEMBLOCK_NONE : MEMBLOCK_HOTPLUG;
>>> +
>>> + for (int i = 0; i < NUMA_NODES; i++) {
>>> + assert(nodes[i] <= MEM_SIZE && nodes[i] > 0);
>>
>> I think it would be even easier to get if this would just be a fraction.
>> E.g., instead of "1/8 * MEM_SIZE" just "1/8". All values have to add up to
>> 1.
>>
>> ... but then we'd have to mess with floats eventually, so I guess this makes
>> it easier to handle these fractions.
>>
>>
>> We could use "int" and simply specify the fraction in percent, like
>>
>> nodes[0] = 50;
>> nodes[1] = 25;
>> nodes[2] = 25;
>>
>> and everything has to add up to 100.
>>
> This would still be a float for 1/8th (12.5) and 1/16th (6.25). What if
> it was the "percent" of 256 (i.e., 0x100)?

Right, or in something "smaller" like 1/32 th. I don't think we go below
that?

If we don't need more digits, why not in "basis points" (per ten thousand)
-> https://en.wikipedia.org/wiki/Basis_point

nodes[0] = 5000; /* 1/2 */
nodes[1] = 2500; /* 1/4 */
nodes[2] = 1250; /* 1/8 */
nodes[4] = 0625; /* 1/32 */
nodes[5] = 0625;


>>
>>> + phys_addr_t size = MEM_SIZE / nodes[i];
>>
>>
>> Hmmm, assuming a single node with "MEM_SIZE", we would get size=1.
>>
> For a single node of MEM_SIZE, nodes[0] would be 1.
>
>> Shouldn't this be "size = nodes[i]"
>>
>> ?
> No, not with the current implementation. The nodes array stores the
> denominator of the fraction that will be multiplied by MEM_SIZE to
> determine the size of that node (the numerator is always 1). So if the
> size of the node should be 1/8 * MEM_SIZE, the nodes array just stores
> the 8. I think the name of the array is misleading. Do you have any
> suggestions for a better name?

Then I am confused about the
assert(nodes[i] <= MEM_SIZE && nodes[i] > 0);

assertion :)

I think it would really be best to just store the actual fraction
somehow. But maybe just I am confused :)

--
Thanks,

David / dhildenb

2022-09-08 00:05:52

by Rebecca Mckeever

Subject: Re: [PATCH v4 1/4] memblock tests: add simulation of physical memory with multiple NUMA nodes

On Wed, Sep 07, 2022 at 10:44:44AM +0200, David Hildenbrand wrote:
> On 07.09.22 01:43, Rebecca Mckeever wrote:
> > On Tue, Sep 06, 2022 at 03:17:46PM +0200, David Hildenbrand wrote:
> > > On 04.09.22 06:21, Rebecca Mckeever wrote:
> > > > Add function setup_numa_memblock() for setting up a memory layout with
> > > > multiple NUMA nodes in a previously allocated dummy physical memory.
> > > > This function can be used in place of setup_memblock() in tests that need
> > > > to simulate a NUMA system.
> > > >
> > > > setup_numa_memblock():
> > > > - allows for setting up a memory layout by specifying the fraction of
> > > > MEM_SIZE in each node
> > > >
> > > > Set CONFIG_NODES_SHIFT to 4 when building with NUMA=1 to allow for up to
> > > > 16 NUMA nodes.
> > > >
> > > > Signed-off-by: Rebecca Mckeever <[email protected]>
> > > > ---
> > > > .../testing/memblock/scripts/Makefile.include | 2 +-
> > > > tools/testing/memblock/tests/common.c | 29 +++++++++++++++++++
> > > > tools/testing/memblock/tests/common.h | 4 ++-
> > > > 3 files changed, 33 insertions(+), 2 deletions(-)
> > > >
> > > > diff --git a/tools/testing/memblock/scripts/Makefile.include b/tools/testing/memblock/scripts/Makefile.include
> > > > index aa6d82d56a23..998281723590 100644
> > > > --- a/tools/testing/memblock/scripts/Makefile.include
> > > > +++ b/tools/testing/memblock/scripts/Makefile.include
> > > > @@ -3,7 +3,7 @@
> > > > # Simulate CONFIG_NUMA=y
> > > > ifeq ($(NUMA), 1)
> > > > - CFLAGS += -D CONFIG_NUMA
> > > > + CFLAGS += -D CONFIG_NUMA -D CONFIG_NODES_SHIFT=4
> > > > endif
> > > > # Use 32 bit physical addresses.
> > > > diff --git a/tools/testing/memblock/tests/common.c b/tools/testing/memblock/tests/common.c
> > > > index eec6901081af..b6110df21b2a 100644
> > > > --- a/tools/testing/memblock/tests/common.c
> > > > +++ b/tools/testing/memblock/tests/common.c
> > > > @@ -72,6 +72,35 @@ void setup_memblock(void)
> > > > fill_memblock();
> > > > }
> > > > +/**
> > > > + * setup_numa_memblock:
> > > > + * Set up a memory layout with multiple NUMA nodes in a previously allocated
> > > > + * dummy physical memory.
> > > > + * @nodes: an array containing the denominators of the fractions of MEM_SIZE
> > > > + * contained in each node (e.g., if nodes[0] = SZ_8, node 0 will
> > > > + * contain 1/8th of MEM_SIZE)
> > > > + *
> > > > + * The nids will be set to 0 through NUMA_NODES - 1.
> > > > + */
> > > > +void setup_numa_memblock(const phys_addr_t nodes[])
> > > > +{
> > > > + phys_addr_t base;
> > > > + int flags;
> > > > +
> > > > + reset_memblock_regions();
> > > > + base = (phys_addr_t)memory_block.base;
> > > > + flags = (movable_node_is_enabled()) ? MEMBLOCK_NONE : MEMBLOCK_HOTPLUG;
> > > > +
> > > > + for (int i = 0; i < NUMA_NODES; i++) {
> > > > + assert(nodes[i] <= MEM_SIZE && nodes[i] > 0);
> > >
> > > I think it would be even easier to get if this would just be a fraction.
> > > E.g., instead of "1/8 * MEM_SIZE" just "1/8". All values have to add up to
> > > 1.
> > >
> > > ... but then we'd have to mess with floats eventually, so I guess this makes
> > > it easier to handle these fractions.
> > >
> > >
> > > We could use "int" and simply specify the fraction in percent, like
> > >
> > > nodes[0] = 50;
> > > nodes[1] = 25;
> > > nodes[2] = 25;
> > >
> > > and everything has to add up to 100.
> > >
> > This would still be a float for 1/8th (12.5) and 1/16th (6.25). What if
> > it was the "percent" of 256 (i.e., 0x100)?
>
> Right, or in something "smaller" like 1/32 th. I don't think we go below
> that?
>
> If we don't need more digits, why not in "basis points" (per ten thousand)
> -> https://en.wikipedia.org/wiki/Basis_point
>
Basis points should work.

> nodes[0] = 5000; /* 1/2 */
> nodes[1] = 2500; /* 1/4 */
> nodes[2] = 1250; /* 1/8 */
> nodes[4] = 0625; /* 1/32 */
> nodes[5] = 0625;
>
>
> > >
> > > > + phys_addr_t size = MEM_SIZE / nodes[i];
> > >
> > >
> > > Hmmm, assuming a single node with "MEM_SIZE", we would get size=1.
> > >
> > For a single node of MEM_SIZE, nodes[0] would be 1.
> >
> > > Shouldn't this be "size = nodes[i]"
> > >
> > > ?
> > No, not with the current implementation. The nodes array stores the
> > denominator of the fraction that will be multiplied by MEM_SIZE to
> > determine the size of that node (the numerator is always 1). So if the
> > size of the node should be 1/8 * MEM_SIZE, the nodes array just stores
> > the 8. I think the name of the array is misleading. Do you have any
> > suggestions for a better name?
>
> Then I am confused about the
> assert(nodes[i] <= MEM_SIZE && nodes[i] > 0);
>
> assertion :)
>
The first part of the assert ensures that size doesn't become less than
1, and the second part prevents a divide by 0. I see how this is
confusing now.

> I think it would really be best to just store the actual fraction somehow.
> But maybe just I am confused :)
>
> --
> Thanks,
>
> David / dhildenb
>
Thanks,
Rebecca

2022-09-08 12:34:11

by David Hildenbrand

Subject: Re: [PATCH v4 4/4] memblock tests: add generic NUMA tests for memblock_alloc_try_nid*

On 04.09.22 06:21, Rebecca Mckeever wrote:
> Add tests for memblock_alloc_try_nid() and memblock_alloc_try_nid_raw()
> where the simulated physical memory is set up with multiple NUMA nodes.
> Additionally, two of these tests set nid != NUMA_NO_NODE. All tests are
> run for both top-down and bottom-up allocation directions.
>
> The tested scenarios are:
>
> Range unrestricted:
> - region cannot be allocated:
> + none of the nodes have enough memory to allocate the region
>
> Range restricted:
> - region can be allocated in the specific node requested without dropping
> min_addr:
> + the range fully overlaps with the node, and there are adjacent
> reserved regions
> - region cannot be allocated:
> + nid is set to NUMA_NO_NODE and the total range can fit the region,
> but the range is split between two nodes and everything else is
> reserved
>
> Signed-off-by: Rebecca Mckeever <[email protected]>
> ---
> tools/testing/memblock/tests/alloc_nid_api.c | 197 +++++++++++++++++++
> 1 file changed, 197 insertions(+)
>

Acked-by: David Hildenbrand <[email protected]>

--
Thanks,

David / dhildenb

2022-09-08 12:50:23

by David Hildenbrand

Subject: Re: [PATCH v4 2/4] memblock tests: add top-down NUMA tests for memblock_alloc_try_nid*

On 04.09.22 06:21, Rebecca Mckeever wrote:
> Add tests for memblock_alloc_try_nid() and memblock_alloc_try_nid_raw()
> where the simulated physical memory is set up with multiple NUMA nodes.
> Additionally, all of these tests set nid != NUMA_NO_NODE. These tests are
> run with a top-down allocation direction.
>
> The tested scenarios are:
>
> Range unrestricted:
> - region can be allocated in the specific node requested:
> + there are no previously reserved regions
> + the requested node is partially reserved but has enough space
> - the specific node requested cannot accommodate the request, but the
> region can be allocated in a different node:
> + there are no previously reserved regions, but node is too small
> + the requested node is fully reserved
> + the requested node is partially reserved and does not have
> enough space
>
> Range restricted:
> - region can be allocated in the specific node requested after dropping
> min_addr:
> + range partially overlaps with two different nodes, where the first
> node is the requested node
> + range partially overlaps with two different nodes, where the
> requested node ends before min_addr
> - region cannot be allocated in the specific node requested, but it can be
> allocated in the requested range:
> + range overlaps with multiple nodes along node boundaries, and the
> requested node ends before min_addr
> + range overlaps with multiple nodes along node boundaries, and the
> requested node starts after max_addr
> - region cannot be allocated in the specific node requested, but it can be
> allocated after dropping min_addr:
> + range partially overlaps with two different nodes, where the
> second node is the requested node
>
> Signed-off-by: Rebecca Mckeever <[email protected]>
> ---
> tools/testing/memblock/tests/alloc_nid_api.c | 694 ++++++++++++++++++-
> tools/testing/memblock/tests/alloc_nid_api.h | 16 +
> tools/testing/memblock/tests/common.h | 18 +
> 3 files changed, 717 insertions(+), 11 deletions(-)
>
> diff --git a/tools/testing/memblock/tests/alloc_nid_api.c b/tools/testing/memblock/tests/alloc_nid_api.c
> index 32b3c1594fdd..a639e9d3e098 100644
> --- a/tools/testing/memblock/tests/alloc_nid_api.c
> +++ b/tools/testing/memblock/tests/alloc_nid_api.c
> @@ -3,6 +3,14 @@
>
> static int alloc_nid_test_flags = TEST_F_NONE;
>
> +/*
> + * contains the denominators of the fractions of MEM_SIZE contained in each node
> + * (e.g., if node_fractions[0] = SZ_8, node 0 will contain 1/8th of MEM_SIZE)
> + */
> +static const phys_addr_t node_fractions[] = {
> + SZ_4, SZ_16, SZ_8, SZ_8, SZ_16, SZ_16, SZ_4, SZ_16
> +};
> +

Besides that part, that I consider confusing :)

Acked-by: David Hildenbrand <[email protected]>

--
Thanks,

David / dhildenb

2022-09-08 13:06:46

by David Hildenbrand

Subject: Re: [PATCH v4 3/4] memblock tests: add bottom-up NUMA tests for memblock_alloc_try_nid*

On 04.09.22 06:21, Rebecca Mckeever wrote:
> Add tests for memblock_alloc_try_nid() and memblock_alloc_try_nid_raw()
> where the simulated physical memory is set up with multiple NUMA nodes.
> Additionally, all of these tests set nid != NUMA_NO_NODE. These tests are
> run with a bottom-up allocation direction.
>
> The tested scenarios are:
>
> Range unrestricted:
> - region can be allocated in the specific node requested:
> + there are no previously reserved regions
> + the requested node is partially reserved but has enough space
> - the specific node requested cannot accommodate the request, but the
> region can be allocated in a different node:
> + there are no previously reserved regions, but node is too small
> + the requested node is fully reserved
> + the requested node is partially reserved and does not have
> enough space
>
> Range restricted:
> - region can be allocated in the specific node requested after dropping
> min_addr:
> + range partially overlaps with two different nodes, where the first
> node is the requested node
> + range partially overlaps with two different nodes, where the
> requested node ends before min_addr
> - region cannot be allocated in the specific node requested, but it can be
> allocated in the requested range:
> + range overlaps with multiple nodes along node boundaries, and the
> requested node ends before min_addr
> + range overlaps with multiple nodes along node boundaries, and the
> requested node starts after max_addr
> - region cannot be allocated in the specific node requested, but it can be
> allocated after dropping min_addr:
> + range partially overlaps with two different nodes, where the
> second node is the requested node
>
> Signed-off-by: Rebecca Mckeever <[email protected]>
> ---
> tools/testing/memblock/tests/alloc_nid_api.c | 568 +++++++++++++++++++
> 1 file changed, 568 insertions(+)

Acked-by: David Hildenbrand <[email protected]>

--
Thanks,

David / dhildenb