mm/slub: enable debugging memory wasting of kmalloc

kmalloc's API family is critical for mm, with one nature that it will
round up the request size to a fixed one (mostly power of 2). Say
when user requests memory for '2^n + 1' bytes, actually 2^(n+1) bytes
could be allocated, so in worst case, there is around 50% memory
space waste.

The wastage is not a big issue for requests that get allocated/freed
quickly, but may cause problems with objects that have longer life
time.

We've met a kernel boot OOM panic (v5.10), and from the dumped slab
info:

    [   26.062145] kmalloc-2k            814056KB     814056KB

From debug we found there are huge number of 'struct iova_magazine',
whose size is 1032 bytes (1024 + 8), so each allocation will waste
1016 bytes. Though the issue was solved by giving the right (bigger)
size of RAM, it is still nice to optimize the size (either use a
kmalloc friendly size or create a dedicated slab for it).

And from lkml archive, there was another crash kernel OOM case [1]
back in 2019, which seems to be related with the similar slab waste
situation, as the log is similar:

    [    4.332648] iommu: Adding device 0000:20:02.0 to group 16
    [    4.338946] swapper/0 invoked oom-killer: gfp_mask=0x6040c0(GFP_KERNEL|__GFP_COMP), nodemask=(null), order=0, oom_score_adj=0
    ...
    [    4.857565] kmalloc-2048           59164KB      59164KB

The crash kernel only has 256M memory, and 59M is pretty big here.
(Note: the related code has been changed and optimised in recent
kernel [2], these logs are just picked to demo the problem, also
a patch changing its size to 1024 bytes has been merged)

So add an way to track each kmalloc's memory waste info, and
leverage the existing SLUB debug framework (specifically
SLUB_STORE_USER) to show its call stack of original allocation,
so that user can evaluate the waste situation, identify some hot
spots and optimize accordingly, for a better utilization of memory.

The waste info is integrated into existing interface:
'/sys/kernel/debug/slab/kmalloc-xx/alloc_traces', one example of
'kmalloc-4k' after boot is:

 126 ixgbe_alloc_q_vector+0xbe/0x830 [ixgbe] waste=233856/1856 age=280763/281414/282065 pid=1330 cpus=32 nodes=1
     __kmem_cache_alloc_node+0x11f/0x4e0
     __kmalloc_node+0x4e/0x140
     ixgbe_alloc_q_vector+0xbe/0x830 [ixgbe]
     ixgbe_init_interrupt_scheme+0x2ae/0xc90 [ixgbe]
     ixgbe_probe+0x165f/0x1d20 [ixgbe]
     local_pci_probe+0x78/0xc0
     work_for_cpu_fn+0x26/0x40
     ...

which means in 'kmalloc-4k' slab, there are 126 requests of
2240 bytes which got a 4KB space (wasting 1856 bytes each
and 233856 bytes in total), from ixgbe_alloc_q_vector().

And when system starts some real workload like multiple docker
instances, there could are more severe waste.

[1]. https://lkml.org/lkml/2019/8/12/266
[2]. https://lore.kernel.org/lkml/2920df89-9975-5785-f79b-257d3052dfaf@huawei.com/

[Thanks Hyeonggon for pointing out several bugs about sorting/format]
[Thanks Vlastimil for suggesting way to reduce memory usage of
 orig_size and keep it only for kmalloc objects]

Signed-off-by: Feng Tang <feng.tang@intel.com>
Reviewed-by: Hyeonggon Yoo <42.hyeyoo@gmail.com>
Cc: Robin Murphy <robin.murphy@arm.com>
Cc: John Garry <john.garry@huawei.com>
Cc: Kefeng Wang <wangkefeng.wang@huawei.com>
Signed-off-by: Vlastimil Babka <vbabka@suse.cz>
This commit is contained in:
Feng Tang 2022-09-13 14:54:20 +08:00 committed by Vlastimil Babka
parent 1f04b07d97
commit 6edf2576a6
4 changed files with 142 additions and 50 deletions

View File

@ -400,21 +400,30 @@ information:
allocated objects. The output is sorted by frequency of each trace.
Information in the output:
Number of objects, allocating function, minimal/average/maximal jiffies since alloc,
pid range of the allocating processes, cpu mask of allocating cpus, and stack trace.
Number of objects, allocating function, possible memory wastage of
kmalloc objects(total/per-object), minimal/average/maximal jiffies
since alloc, pid range of the allocating processes, cpu mask of
allocating cpus, numa node mask of origins of memory, and stack trace.
Example:::
1085 populate_error_injection_list+0x97/0x110 age=166678/166680/166682 pid=1 cpus=1::
__slab_alloc+0x6d/0x90
kmem_cache_alloc_trace+0x2eb/0x300
populate_error_injection_list+0x97/0x110
init_error_injection+0x1b/0x71
do_one_initcall+0x5f/0x2d0
kernel_init_freeable+0x26f/0x2d7
kernel_init+0xe/0x118
ret_from_fork+0x22/0x30
338 pci_alloc_dev+0x2c/0xa0 waste=521872/1544 age=290837/291891/293509 pid=1 cpus=106 nodes=0-1
__kmem_cache_alloc_node+0x11f/0x4e0
kmalloc_trace+0x26/0xa0
pci_alloc_dev+0x2c/0xa0
pci_scan_single_device+0xd2/0x150
pci_scan_slot+0xf7/0x2d0
pci_scan_child_bus_extend+0x4e/0x360
acpi_pci_root_create+0x32e/0x3b0
pci_acpi_scan_root+0x2b9/0x2d0
acpi_pci_root_add.cold.11+0x110/0xb0a
acpi_bus_attach+0x262/0x3f0
device_for_each_child+0xb7/0x110
acpi_dev_for_each_child+0x77/0xa0
acpi_bus_attach+0x108/0x3f0
device_for_each_child+0xb7/0x110
acpi_dev_for_each_child+0x77/0xa0
acpi_bus_attach+0x108/0x3f0
2. free_traces::

View File

@ -29,6 +29,8 @@
#define SLAB_RED_ZONE ((slab_flags_t __force)0x00000400U)
/* DEBUG: Poison objects */
#define SLAB_POISON ((slab_flags_t __force)0x00000800U)
/* Indicate a kmalloc slab */
#define SLAB_KMALLOC ((slab_flags_t __force)0x00001000U)
/* Align objs on cache lines */
#define SLAB_HWCACHE_ALIGN ((slab_flags_t __force)0x00002000U)
/* Use GFP_DMA memory */

View File

@ -649,7 +649,8 @@ struct kmem_cache *__init create_kmalloc_cache(const char *name,
if (!s)
panic("Out of memory when creating slab %s\n", name);
create_boot_cache(s, name, size, flags, useroffset, usersize);
create_boot_cache(s, name, size, flags | SLAB_KMALLOC, useroffset,
usersize);
kasan_cache_create_kmalloc(s);
list_add(&s->list, &slab_caches);
s->refcount = 1;

154
mm/slub.c
View File

@ -194,11 +194,24 @@ DEFINE_STATIC_KEY_FALSE(slub_debug_enabled);
#endif
#endif /* CONFIG_SLUB_DEBUG */
/* Structure holding parameters for get_partial() call chain */
struct partial_context {
struct slab **slab;
gfp_t flags;
unsigned int orig_size;
};
static inline bool kmem_cache_debug(struct kmem_cache *s)
{
return kmem_cache_debug_flags(s, SLAB_DEBUG_FLAGS);
}
static inline bool slub_debug_orig_size(struct kmem_cache *s)
{
return (kmem_cache_debug_flags(s, SLAB_STORE_USER) &&
(s->flags & SLAB_KMALLOC));
}
void *fixup_red_left(struct kmem_cache *s, void *p)
{
if (kmem_cache_debug_flags(s, SLAB_RED_ZONE))
@ -785,6 +798,39 @@ static void print_slab_info(const struct slab *slab)
folio_flags(folio, 0));
}
/*
* kmalloc caches has fixed sizes (mostly power of 2), and kmalloc() API
* family will round up the real request size to these fixed ones, so
* there could be an extra area than what is requested. Save the original
* request size in the meta data area, for better debug and sanity check.
*/
static inline void set_orig_size(struct kmem_cache *s,
void *object, unsigned int orig_size)
{
void *p = kasan_reset_tag(object);
if (!slub_debug_orig_size(s))
return;
p += get_info_end(s);
p += sizeof(struct track) * 2;
*(unsigned int *)p = orig_size;
}
static inline unsigned int get_orig_size(struct kmem_cache *s, void *object)
{
void *p = kasan_reset_tag(object);
if (!slub_debug_orig_size(s))
return s->object_size;
p += get_info_end(s);
p += sizeof(struct track) * 2;
return *(unsigned int *)p;
}
static void slab_bug(struct kmem_cache *s, char *fmt, ...)
{
struct va_format vaf;
@ -844,6 +890,9 @@ static void print_trailer(struct kmem_cache *s, struct slab *slab, u8 *p)
if (s->flags & SLAB_STORE_USER)
off += 2 * sizeof(struct track);
if (slub_debug_orig_size(s))
off += sizeof(unsigned int);
off += kasan_metadata_size(s);
if (off != size_from_object(s))
@ -977,7 +1026,8 @@ skip_bug_print:
*
* A. Free pointer (if we cannot overwrite object on free)
* B. Tracking data for SLAB_STORE_USER
* C. Padding to reach required alignment boundary or at minimum
* C. Original request size for kmalloc object (SLAB_STORE_USER enabled)
* D. Padding to reach required alignment boundary or at minimum
* one word if debugging is on to be able to detect writes
* before the word boundary.
*
@ -995,10 +1045,14 @@ static int check_pad_bytes(struct kmem_cache *s, struct slab *slab, u8 *p)
{
unsigned long off = get_info_end(s); /* The end of info */
if (s->flags & SLAB_STORE_USER)
if (s->flags & SLAB_STORE_USER) {
/* We also have user information there */
off += 2 * sizeof(struct track);
if (s->flags & SLAB_KMALLOC)
off += sizeof(unsigned int);
}
off += kasan_metadata_size(s);
if (size_from_object(s) == off)
@ -1293,7 +1347,7 @@ static inline int alloc_consistency_checks(struct kmem_cache *s,
}
static noinline int alloc_debug_processing(struct kmem_cache *s,
struct slab *slab, void *object)
struct slab *slab, void *object, int orig_size)
{
if (s->flags & SLAB_CONSISTENCY_CHECKS) {
if (!alloc_consistency_checks(s, slab, object))
@ -1302,6 +1356,7 @@ static noinline int alloc_debug_processing(struct kmem_cache *s,
/* Success. Perform special debug activities for allocs */
trace(s, slab, object, 1);
set_orig_size(s, object, orig_size);
init_object(s, object, SLUB_RED_ACTIVE);
return 1;
@ -1570,7 +1625,7 @@ static inline
void setup_slab_debug(struct kmem_cache *s, struct slab *slab, void *addr) {}
static inline int alloc_debug_processing(struct kmem_cache *s,
struct slab *slab, void *object) { return 0; }
struct slab *slab, void *object, int orig_size) { return 0; }
static inline void free_debug_processing(
struct kmem_cache *s, struct slab *slab,
@ -2013,7 +2068,7 @@ static inline void remove_partial(struct kmem_cache_node *n,
* it to full list if it was the last free object.
*/
static void *alloc_single_from_partial(struct kmem_cache *s,
struct kmem_cache_node *n, struct slab *slab)
struct kmem_cache_node *n, struct slab *slab, int orig_size)
{
void *object;
@ -2023,7 +2078,7 @@ static void *alloc_single_from_partial(struct kmem_cache *s,
slab->freelist = get_freepointer(s, object);
slab->inuse++;
if (!alloc_debug_processing(s, slab, object)) {
if (!alloc_debug_processing(s, slab, object, orig_size)) {
remove_partial(n, slab);
return NULL;
}
@ -2042,7 +2097,7 @@ static void *alloc_single_from_partial(struct kmem_cache *s,
* and put the slab to the partial (or full) list.
*/
static void *alloc_single_from_new_slab(struct kmem_cache *s,
struct slab *slab)
struct slab *slab, int orig_size)
{
int nid = slab_nid(slab);
struct kmem_cache_node *n = get_node(s, nid);
@ -2054,7 +2109,7 @@ static void *alloc_single_from_new_slab(struct kmem_cache *s,
slab->freelist = get_freepointer(s, object);
slab->inuse = 1;
if (!alloc_debug_processing(s, slab, object))
if (!alloc_debug_processing(s, slab, object, orig_size))
/*
* It's not really expected that this would fail on a
* freshly allocated slab, but a concurrent memory
@ -2132,7 +2187,7 @@ static inline bool pfmemalloc_match(struct slab *slab, gfp_t gfpflags);
* Try to allocate a partial slab from a specific node.
*/
static void *get_partial_node(struct kmem_cache *s, struct kmem_cache_node *n,
struct slab **ret_slab, gfp_t gfpflags)
struct partial_context *pc)
{
struct slab *slab, *slab2;
void *object = NULL;
@ -2152,11 +2207,12 @@ static void *get_partial_node(struct kmem_cache *s, struct kmem_cache_node *n,
list_for_each_entry_safe(slab, slab2, &n->partial, slab_list) {
void *t;
if (!pfmemalloc_match(slab, gfpflags))
if (!pfmemalloc_match(slab, pc->flags))
continue;
if (kmem_cache_debug(s)) {
object = alloc_single_from_partial(s, n, slab);
object = alloc_single_from_partial(s, n, slab,
pc->orig_size);
if (object)
break;
continue;
@ -2167,7 +2223,7 @@ static void *get_partial_node(struct kmem_cache *s, struct kmem_cache_node *n,
break;
if (!object) {
*ret_slab = slab;
*pc->slab = slab;
stat(s, ALLOC_FROM_PARTIAL);
object = t;
} else {
@ -2191,14 +2247,13 @@ static void *get_partial_node(struct kmem_cache *s, struct kmem_cache_node *n,
/*
* Get a slab from somewhere. Search in increasing NUMA distances.
*/
static void *get_any_partial(struct kmem_cache *s, gfp_t flags,
struct slab **ret_slab)
static void *get_any_partial(struct kmem_cache *s, struct partial_context *pc)
{
#ifdef CONFIG_NUMA
struct zonelist *zonelist;
struct zoneref *z;
struct zone *zone;
enum zone_type highest_zoneidx = gfp_zone(flags);
enum zone_type highest_zoneidx = gfp_zone(pc->flags);
void *object;
unsigned int cpuset_mems_cookie;
@ -2226,15 +2281,15 @@ static void *get_any_partial(struct kmem_cache *s, gfp_t flags,
do {
cpuset_mems_cookie = read_mems_allowed_begin();
zonelist = node_zonelist(mempolicy_slab_node(), flags);
zonelist = node_zonelist(mempolicy_slab_node(), pc->flags);
for_each_zone_zonelist(zone, z, zonelist, highest_zoneidx) {
struct kmem_cache_node *n;
n = get_node(s, zone_to_nid(zone));
if (n && cpuset_zone_allowed(zone, flags) &&
if (n && cpuset_zone_allowed(zone, pc->flags) &&
n->nr_partial > s->min_partial) {
object = get_partial_node(s, n, ret_slab, flags);
object = get_partial_node(s, n, pc);
if (object) {
/*
* Don't check read_mems_allowed_retry()
@ -2255,8 +2310,7 @@ static void *get_any_partial(struct kmem_cache *s, gfp_t flags,
/*
* Get a partial slab, lock it and return it.
*/
static void *get_partial(struct kmem_cache *s, gfp_t flags, int node,
struct slab **ret_slab)
static void *get_partial(struct kmem_cache *s, int node, struct partial_context *pc)
{
void *object;
int searchnode = node;
@ -2264,11 +2318,11 @@ static void *get_partial(struct kmem_cache *s, gfp_t flags, int node,
if (node == NUMA_NO_NODE)
searchnode = numa_mem_id();
object = get_partial_node(s, get_node(s, searchnode), ret_slab, flags);
object = get_partial_node(s, get_node(s, searchnode), pc);
if (object || node != NUMA_NO_NODE)
return object;
return get_any_partial(s, flags, ret_slab);
return get_any_partial(s, pc);
}
#ifdef CONFIG_PREEMPTION
@ -2989,11 +3043,12 @@ static inline void *get_freelist(struct kmem_cache *s, struct slab *slab)
* already disabled (which is the case for bulk allocation).
*/
static void *___slab_alloc(struct kmem_cache *s, gfp_t gfpflags, int node,
unsigned long addr, struct kmem_cache_cpu *c)
unsigned long addr, struct kmem_cache_cpu *c, unsigned int orig_size)
{
void *freelist;
struct slab *slab;
unsigned long flags;
struct partial_context pc;
stat(s, ALLOC_SLOWPATH);
@ -3107,7 +3162,10 @@ new_slab:
new_objects:
freelist = get_partial(s, gfpflags, node, &slab);
pc.flags = gfpflags;
pc.slab = &slab;
pc.orig_size = orig_size;
freelist = get_partial(s, node, &pc);
if (freelist)
goto check_new_slab;
@ -3123,7 +3181,7 @@ new_objects:
stat(s, ALLOC_SLAB);
if (kmem_cache_debug(s)) {
freelist = alloc_single_from_new_slab(s, slab);
freelist = alloc_single_from_new_slab(s, slab, orig_size);
if (unlikely(!freelist))
goto new_objects;
@ -3155,6 +3213,7 @@ check_new_slab:
*/
if (s->flags & SLAB_STORE_USER)
set_track(s, freelist, TRACK_ALLOC, addr);
return freelist;
}
@ -3197,7 +3256,7 @@ retry_load_slab:
* pointer.
*/
static void *__slab_alloc(struct kmem_cache *s, gfp_t gfpflags, int node,
unsigned long addr, struct kmem_cache_cpu *c)
unsigned long addr, struct kmem_cache_cpu *c, unsigned int orig_size)
{
void *p;
@ -3210,7 +3269,7 @@ static void *__slab_alloc(struct kmem_cache *s, gfp_t gfpflags, int node,
c = slub_get_cpu_ptr(s->cpu_slab);
#endif
p = ___slab_alloc(s, gfpflags, node, addr, c);
p = ___slab_alloc(s, gfpflags, node, addr, c, orig_size);
#ifdef CONFIG_PREEMPT_COUNT
slub_put_cpu_ptr(s->cpu_slab);
#endif
@ -3295,7 +3354,7 @@ redo:
if (!USE_LOCKLESS_FAST_PATH() ||
unlikely(!object || !slab || !node_match(slab, node))) {
object = __slab_alloc(s, gfpflags, node, addr, c);
object = __slab_alloc(s, gfpflags, node, addr, c, orig_size);
} else {
void *next_object = get_freepointer_safe(s, object);
@ -3793,7 +3852,7 @@ int kmem_cache_alloc_bulk(struct kmem_cache *s, gfp_t flags, size_t size,
* of re-populating per CPU c->freelist
*/
p[i] = ___slab_alloc(s, flags, NUMA_NO_NODE,
_RET_IP_, c);
_RET_IP_, c, s->object_size);
if (unlikely(!p[i]))
goto error;
@ -4196,12 +4255,17 @@ static int calculate_sizes(struct kmem_cache *s)
}
#ifdef CONFIG_SLUB_DEBUG
if (flags & SLAB_STORE_USER)
if (flags & SLAB_STORE_USER) {
/*
* Need to store information about allocs and frees after
* the object.
*/
size += 2 * sizeof(struct track);
/* Save the original kmalloc request size */
if (flags & SLAB_KMALLOC)
size += sizeof(unsigned int);
}
#endif
kasan_cache_create(s, &size, &s->flags);
@ -5146,6 +5210,7 @@ struct location {
depot_stack_handle_t handle;
unsigned long count;
unsigned long addr;
unsigned long waste;
long long sum_time;
long min_time;
long max_time;
@ -5192,13 +5257,15 @@ static int alloc_loc_track(struct loc_track *t, unsigned long max, gfp_t flags)
}
static int add_location(struct loc_track *t, struct kmem_cache *s,
const struct track *track)
const struct track *track,
unsigned int orig_size)
{
long start, end, pos;
struct location *l;
unsigned long caddr, chandle;
unsigned long caddr, chandle, cwaste;
unsigned long age = jiffies - track->when;
depot_stack_handle_t handle = 0;
unsigned int waste = s->object_size - orig_size;
#ifdef CONFIG_STACKDEPOT
handle = READ_ONCE(track->handle);
@ -5216,11 +5283,13 @@ static int add_location(struct loc_track *t, struct kmem_cache *s,
if (pos == end)
break;
caddr = t->loc[pos].addr;
chandle = t->loc[pos].handle;
if ((track->addr == caddr) && (handle == chandle)) {
l = &t->loc[pos];
caddr = l->addr;
chandle = l->handle;
cwaste = l->waste;
if ((track->addr == caddr) && (handle == chandle) &&
(waste == cwaste)) {
l = &t->loc[pos];
l->count++;
if (track->when) {
l->sum_time += age;
@ -5245,6 +5314,9 @@ static int add_location(struct loc_track *t, struct kmem_cache *s,
end = pos;
else if (track->addr == caddr && handle < chandle)
end = pos;
else if (track->addr == caddr && handle == chandle &&
waste < cwaste)
end = pos;
else
start = pos;
}
@ -5268,6 +5340,7 @@ static int add_location(struct loc_track *t, struct kmem_cache *s,
l->min_pid = track->pid;
l->max_pid = track->pid;
l->handle = handle;
l->waste = waste;
cpumask_clear(to_cpumask(l->cpus));
cpumask_set_cpu(track->cpu, to_cpumask(l->cpus));
nodes_clear(l->nodes);
@ -5280,13 +5353,16 @@ static void process_slab(struct loc_track *t, struct kmem_cache *s,
unsigned long *obj_map)
{
void *addr = slab_address(slab);
bool is_alloc = (alloc == TRACK_ALLOC);
void *p;
__fill_map(obj_map, s, slab);
for_each_object(p, s, addr, slab->objects)
if (!test_bit(__obj_to_index(s, addr, p), obj_map))
add_location(t, s, get_track(s, p, alloc));
add_location(t, s, get_track(s, p, alloc),
is_alloc ? get_orig_size(s, p) :
s->object_size);
}
#endif /* CONFIG_DEBUG_FS */
#endif /* CONFIG_SLUB_DEBUG */
@ -6156,6 +6232,10 @@ static int slab_debugfs_show(struct seq_file *seq, void *v)
else
seq_puts(seq, "<not-available>");
if (l->waste)
seq_printf(seq, " waste=%lu/%lu",
l->count * l->waste, l->waste);
if (l->sum_time != l->min_time) {
seq_printf(seq, " age=%ld/%llu/%ld",
l->min_time, div_u64(l->sum_time, l->count),