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>
---
Documentation/mm/slub.rst | 33 +++++---
include/linux/slab.h | 2 +
mm/slub.c | 156 ++++++++++++++++++++++++++++----------
3 files changed, 141 insertions(+), 50 deletions(-)
diff --git a/Documentation/mm/slub.rst b/Documentation/mm/slub.rst
index 43063ade737a..4e1578186b4f 100644
--- a/Documentation/mm/slub.rst
+++ b/Documentation/mm/slub.rst
@@ -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::
diff --git a/include/linux/slab.h b/include/linux/slab.h
index 9b592e611cb1..6dc495f76644 100644
--- a/include/linux/slab.h
+++ b/include/linux/slab.h
@@ -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 */
diff --git a/mm/slub.c b/mm/slub.c
index fe4fe0e72daf..c8ba16b3a4db 100644
--- a/mm/slub.c
+++ b/mm/slub.c
@@ -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 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 @@ static int check_bytes_and_report(struct kmem_cache *s, struct slab *slab,
*
* 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 +1625,10 @@ 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 set_orig_size(struct kmem_cache *s,
+ void *object, unsigned int orig_size) {}
static inline void free_debug_processing(
struct kmem_cache *s, struct slab *slab,
@@ -1999,7 +2057,7 @@ 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;
@@ -2009,7 +2067,7 @@ 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;
}
@@ -2028,7 +2086,7 @@ 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);
@@ -2040,7 +2098,7 @@ 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
@@ -2118,7 +2176,7 @@ 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;
@@ -2138,11 +2196,12 @@ 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;
@@ -2153,7 +2212,7 @@ 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 {
@@ -2177,14 +2236,13 @@ 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;
@@ -2212,15 +2270,15 @@ 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()
@@ -2241,8 +2299,7 @@ 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;
@@ -2250,11 +2307,11 @@ 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
@@ -2974,11 +3031,12 @@ 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);
@@ -3092,7 +3150,10 @@ static void *___slab_alloc(struct kmem_cache *s, gfp_t gfpflags, int node,
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;
@@ -3108,7 +3169,7 @@ static void *___slab_alloc(struct kmem_cache *s, gfp_t gfpflags, int node,
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;
@@ -3140,6 +3201,7 @@ static void *___slab_alloc(struct kmem_cache *s, gfp_t gfpflags, int node,
*/
if (s->flags & SLAB_STORE_USER)
set_track(s, freelist, TRACK_ALLOC, addr);
+
return freelist;
}
@@ -3182,7 +3244,7 @@ static void *___slab_alloc(struct kmem_cache *s, gfp_t gfpflags, int node,
* 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;
@@ -3195,7 +3257,7 @@ 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
@@ -3280,7 +3342,7 @@ static __always_inline void *slab_alloc_node(struct kmem_cache *s, struct list_l
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);
@@ -3747,7 +3809,7 @@ 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;
@@ -4150,12 +4212,17 @@ 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);
@@ -4770,7 +4837,7 @@ void __init kmem_cache_init(void)
/* Now we can use the kmem_cache to allocate kmalloc slabs */
setup_kmalloc_cache_index_table();
- create_kmalloc_caches(0);
+ create_kmalloc_caches(SLAB_KMALLOC);
/* Setup random freelists for each cache */
init_freelist_randomization();
@@ -4937,6 +5004,7 @@ 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;
@@ -4983,13 +5051,15 @@ 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);
@@ -5007,11 +5077,13 @@ 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;
@@ -5036,6 +5108,9 @@ 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;
}
@@ -5059,6 +5134,7 @@ 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);
@@ -5077,7 +5153,7 @@ static void process_slab(struct loc_track *t, struct kmem_cache *s,
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), get_orig_size(s, p));
}
#endif /* CONFIG_DEBUG_FS */
#endif /* CONFIG_SLUB_DEBUG */
@@ -5942,6 +6018,10 @@ 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),
--
2.34.1On Tue, 13 Sep 2022, at 06:54, Feng Tang wrote:
> 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.
I have a ralink mt7621 router running Openwrt, using the mips ZBOOT kernel, and appear to have bisected
a very-nearly-clean kernel v6.1rc-2 boot issue to this commit.
I have 3 commits atop 6.1-rc2: fix a ZBOOT compile error, use the Openwrt LZMA options,
and enable DEBUG_ZBOOT for my platform. I am compiling my kernel within the Openwrt build system.
No guarantees this is not due to something I am doing wrong, but any insight would be greatly appreciated.
On UART, No indication of the (once extracted) kernel booting:
transfer started ......................................... transfer ok, time=2.01s
setting up elf image... OK
jumping to kernel code
zimage at: 80BA4100 810D4720
Uncompressing Linux at load address 80001000
Copy device tree to address 80B96EE0
Now, booting the kernel...
Nothing follows
6edf2576a6cc ("mm/slub: enable debugging memory wasting of kmalloc") reverted, normal boot:
transfer started ......................................... transfer ok, time=2.01s
setting up elf image... OK
jumping to kernel code
zimage at: 80BA4100 810D47A4
Uncompressing Linux at load address 80001000
Copy device tree to address 80B96EE0
Now, booting the kernel...
[ 0.000000] Linux version 6.1.0-rc2 (john@john) (mipsel-openwrt-linux-musl-gcc (OpenWrt GCC 11.3.0 r19724+16-1521d5f453) 11.3.0, GNU ld (GNU Binutils) 2.37) #0 SMP Fri Oct 28 03:48:10 2022
[ 0.000000] SoC Type: MediaTek MT7621 ver:1 eco:3
[ 0.000000] printk: bootconsole [early0] enabled
[ 0.000000] CPU0 revision is: 0001992f (MIPS 1004Kc)
[ 0.000000] MIPS: machine is MikroTik RouterBOARD 760iGS
[ 0.000000] Initrd not found or empty - disabling initrd
[ 0.000000] VPE topology {2,2} total 4
[ 0.000000] Primary instruction cache 32kB, VIPT, 4-way, linesize 32 bytes.
[ 0.000000] Primary data cache 32kB, 4-way, PIPT, no aliases, linesize 32 bytes
[ 0.000000] MIPS secondary cache 256kB, 8-way, linesize 32 bytes.
[ 0.000000] Zone ranges:
[ 0.000000] Normal [mem 0x0000000000000000-0x000000000fffffff]
[ 0.000000] HighMem empty
[ 0.000000] Movable zone start for each node
[ 0.000000] Early memory node ranges
[ 0.000000] node 0: [mem 0x0000000000000000-0x000000000fffffff]
[ 0.000000] Initmem setup node 0 [mem 0x0000000000000000-0x000000000fffffff]
[ 0.000000] percpu: Embedded 11 pages/cpu s16064 r8192 d20800 u45056
[ 0.000000] Built 1 zonelists, mobility grouping on. Total pages: 64960
[ 0.000000] Kernel command line: console=ttyS0,115200 rootfstype=squashfs,jffs2
[ 0.000000] Dentry cache hash table entries: 32768 (order: 5, 131072 bytes, linear)
[ 0.000000] Inode-cache hash table entries: 16384 (order: 4, 65536 bytes, linear)
[ 0.000000] Writing ErrCtl register=00019146
[ 0.000000] Readback ErrCtl register=00019146
[ 0.000000] mem auto-init: stack:off, heap alloc:off, heap free:off
[ 0.000000] Memory: 246220K/262144K available (7455K kernel code, 628K rwdata, 1308K rodata, 3524K init, 245K bss, 15924K reserved, 0K cma-reserved, 0K highmem)
[ 0.000000] SLUB: HWalign=32, Order=0-3, MinObjects=0, CPUs=4, Nodes=1
[ 0.000000] rcu: Hierarchical RCU implementation.
boot continues as expected
possibly relevant config options:
grep -E '(SLUB|SLAB)' .config
# SLAB allocator options
# CONFIG_SLAB is not set
CONFIG_SLUB=y
CONFIG_SLAB_MERGE_DEFAULT=y
# CONFIG_SLAB_FREELIST_RANDOM is not set
# CONFIG_SLAB_FREELIST_HARDENED is not set
# CONFIG_SLUB_STATS is not set
CONFIG_SLUB_CPU_PARTIAL=y
# end of SLAB allocator options
# CONFIG_SLUB_DEBUG is not set
With this commit reverted: cpuinfo and meminfo
system type : MediaTek MT7621 ver:1 eco:3
machine : MikroTik RouterBOARD 760iGS
processor : 0
cpu model : MIPS 1004Kc V2.15
BogoMIPS : 586.13
wait instruction : yes
microsecond timers : yes
tlb_entries : 32
extra interrupt vector : yes
hardware watchpoint : yes, count: 4, address/irw mask: [0x0ffc, 0x0ffc, 0x0ffb, 0x0ffb]
isa : mips1 mips2 mips32r1 mips32r2
ASEs implemented : mips16 dsp mt
Options implemented : tlb 4kex 4k_cache prefetch mcheck ejtag llsc pindexed_dcache userlocal vint perf_cntr_intr_bit cdmm perf
shadow register sets : 1
kscratch registers : 0
package : 0
core : 0
VPE : 0
VCED exceptions : not available
VCEI exceptions : not available
processor : 1
cpu model : MIPS 1004Kc V2.15
BogoMIPS : 586.13
wait instruction : yes
microsecond timers : yes
tlb_entries : 32
extra interrupt vector : yes
hardware watchpoint : yes, count: 4, address/irw mask: [0x0ffc, 0x0ffc, 0x0ffb, 0x0ffb]
isa : mips1 mips2 mips32r1 mips32r2
ASEs implemented : mips16 dsp mt
Options implemented : tlb 4kex 4k_cache prefetch mcheck ejtag llsc pindexed_dcache userlocal vint perf_cntr_intr_bit cdmm perf
shadow register sets : 1
kscratch registers : 0
package : 0
core : 0
VPE : 1
VCED exceptions : not available
VCEI exceptions : not available
processor : 2
cpu model : MIPS 1004Kc V2.15
BogoMIPS : 586.13
wait instruction : yes
microsecond timers : yes
tlb_entries : 32
extra interrupt vector : yes
hardware watchpoint : yes, count: 4, address/irw mask: [0x0ffc, 0x0ffc, 0x0ffb, 0x0ffb]
isa : mips1 mips2 mips32r1 mips32r2
ASEs implemented : mips16 dsp mt
Options implemented : tlb 4kex 4k_cache prefetch mcheck ejtag llsc pindexed_dcache userlocal vint perf_cntr_intr_bit cdmm perf
shadow register sets : 1
kscratch registers : 0
package : 0
core : 1
VPE : 0
VCED exceptions : not available
VCEI exceptions : not available
processor : 3
cpu model : MIPS 1004Kc V2.15
BogoMIPS : 586.13
wait instruction : yes
microsecond timers : yes
tlb_entries : 32
extra interrupt vector : yes
hardware watchpoint : yes, count: 4, address/irw mask: [0x0ffc, 0x0ffc, 0x0ffb, 0x0ffb]
isa : mips1 mips2 mips32r1 mips32r2
ASEs implemented : mips16 dsp mt
Options implemented : tlb 4kex 4k_cache prefetch mcheck ejtag llsc pindexed_dcache userlocal vint perf_cntr_intr_bit cdmm perf
shadow register sets : 1
kscratch registers : 0
package : 0
core : 1
VPE : 1
VCED exceptions : not available
VCEI exceptions : not available
MemTotal: 249744 kB
MemFree: 211088 kB
MemAvailable: 187364 kB
Buffers: 0 kB
Cached: 8824 kB
SwapCached: 0 kB
Active: 1104 kB
Inactive: 8860 kB
Active(anon): 1104 kB
Inactive(anon): 8860 kB
Active(file): 0 kB
Inactive(file): 0 kB
Unevictable: 0 kB
Mlocked: 0 kB
HighTotal: 0 kB
HighFree: 0 kB
LowTotal: 249744 kB
LowFree: 211088 kB
SwapTotal: 0 kB
SwapFree: 0 kB
Dirty: 0 kB
Writeback: 0 kB
AnonPages: 1192 kB
Mapped: 2092 kB
Shmem: 8824 kB
KReclaimable: 1704 kB
Slab: 9372 kB
SReclaimable: 1704 kB
SUnreclaim: 7668 kB
KernelStack: 592 kB
PageTables: 264 kB
SecPageTables: 0 kB
NFS_Unstable: 0 kB
Bounce: 0 kB
WritebackTmp: 0 kB
CommitLimit: 124872 kB
Committed_AS: 14676 kB
VmallocTotal: 1040376 kB
VmallocUsed: 2652 kB
VmallocChunk: 0 kB
Percpu: 272 kB
Cheers,
--
John Thomson
On 10/30/22 20:23, John Thomson wrote:
> On Tue, 13 Sep 2022, at 06:54, Feng Tang wrote:
>> 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.
>
>
> I have a ralink mt7621 router running Openwrt, using the mips ZBOOT kernel, and appear to have bisected
> a very-nearly-clean kernel v6.1rc-2 boot issue to this commit.
> I have 3 commits atop 6.1-rc2: fix a ZBOOT compile error, use the Openwrt LZMA options,
> and enable DEBUG_ZBOOT for my platform. I am compiling my kernel within the Openwrt build system.
> No guarantees this is not due to something I am doing wrong, but any insight would be greatly appreciated.
>
>
> On UART, No indication of the (once extracted) kernel booting:
>
> transfer started ......................................... transfer ok, time=2.01s
> setting up elf image... OK
> jumping to kernel code
> zimage at: 80BA4100 810D4720
> Uncompressing Linux at load address 80001000
> Copy device tree to address 80B96EE0
> Now, booting the kernel...
It's weird that the commit would cause no output so early, SLUB code is
run only later.
> Nothing follows
>
> 6edf2576a6cc ("mm/slub: enable debugging memory wasting of kmalloc") reverted, normal boot:
> transfer started ......................................... transfer ok, time=2.01s
> setting up elf image... OK
> jumping to kernel code
> zimage at: 80BA4100 810D47A4
> Uncompressing Linux at load address 80001000
> Copy device tree to address 80B96EE0
> Now, booting the kernel...
>
> [ 0.000000] Linux version 6.1.0-rc2 (john@john) (mipsel-openwrt-linux-musl-gcc (OpenWrt GCC 11.3.0 r19724+16-1521d5f453) 11.3.0, GNU ld (GNU Binutils) 2.37) #0 SMP Fri Oct 28 03:48:10 2022
> [ 0.000000] SoC Type: MediaTek MT7621 ver:1 eco:3
> [ 0.000000] printk: bootconsole [early0] enabled
> [ 0.000000] CPU0 revision is: 0001992f (MIPS 1004Kc)
> [ 0.000000] MIPS: machine is MikroTik RouterBOARD 760iGS
> [ 0.000000] Initrd not found or empty - disabling initrd
> [ 0.000000] VPE topology {2,2} total 4
> [ 0.000000] Primary instruction cache 32kB, VIPT, 4-way, linesize 32 bytes.
> [ 0.000000] Primary data cache 32kB, 4-way, PIPT, no aliases, linesize 32 bytes
> [ 0.000000] MIPS secondary cache 256kB, 8-way, linesize 32 bytes.
> [ 0.000000] Zone ranges:
> [ 0.000000] Normal [mem 0x0000000000000000-0x000000000fffffff]
> [ 0.000000] HighMem empty
> [ 0.000000] Movable zone start for each node
> [ 0.000000] Early memory node ranges
> [ 0.000000] node 0: [mem 0x0000000000000000-0x000000000fffffff]
> [ 0.000000] Initmem setup node 0 [mem 0x0000000000000000-0x000000000fffffff]
> [ 0.000000] percpu: Embedded 11 pages/cpu s16064 r8192 d20800 u45056
> [ 0.000000] Built 1 zonelists, mobility grouping on. Total pages: 64960
> [ 0.000000] Kernel command line: console=ttyS0,115200 rootfstype=squashfs,jffs2
> [ 0.000000] Dentry cache hash table entries: 32768 (order: 5, 131072 bytes, linear)
> [ 0.000000] Inode-cache hash table entries: 16384 (order: 4, 65536 bytes, linear)
> [ 0.000000] Writing ErrCtl register=00019146
> [ 0.000000] Readback ErrCtl register=00019146
> [ 0.000000] mem auto-init: stack:off, heap alloc:off, heap free:off
> [ 0.000000] Memory: 246220K/262144K available (7455K kernel code, 628K rwdata, 1308K rodata, 3524K init, 245K bss, 15924K reserved, 0K cma-reserved, 0K highmem)
> [ 0.000000] SLUB: HWalign=32, Order=0-3, MinObjects=0, CPUs=4, Nodes=1
> [ 0.000000] rcu: Hierarchical RCU implementation.
>
>
> boot continues as expected
>
>
> possibly relevant config options:
> grep -E '(SLUB|SLAB)' .config
> # SLAB allocator options
> # CONFIG_SLAB is not set
> CONFIG_SLUB=y
> CONFIG_SLAB_MERGE_DEFAULT=y
> # CONFIG_SLAB_FREELIST_RANDOM is not set
> # CONFIG_SLAB_FREELIST_HARDENED is not set
> # CONFIG_SLUB_STATS is not set
> CONFIG_SLUB_CPU_PARTIAL=y
> # end of SLAB allocator options
> # CONFIG_SLUB_DEBUG is not set
Also not having CONFIG_SLUB_DEBUG enabled means most of the code the
patch/commit touches is not even active.
Could this be some miscompile or code layout change exposing some
different bug, hmm.
Is it any different if you do enable CONFIG_SLUB_DEBUG ?
Or change to CONFIG_SLAB? (that would be really weird if not)
>
> With this commit reverted: cpuinfo and meminfo
>
> system type : MediaTek MT7621 ver:1 eco:3
> machine : MikroTik RouterBOARD 760iGS
> processor : 0
> cpu model : MIPS 1004Kc V2.15
> BogoMIPS : 586.13
> wait instruction : yes
> microsecond timers : yes
> tlb_entries : 32
> extra interrupt vector : yes
> hardware watchpoint : yes, count: 4, address/irw mask: [0x0ffc, 0x0ffc, 0x0ffb, 0x0ffb]
> isa : mips1 mips2 mips32r1 mips32r2
> ASEs implemented : mips16 dsp mt
> Options implemented : tlb 4kex 4k_cache prefetch mcheck ejtag llsc pindexed_dcache userlocal vint perf_cntr_intr_bit cdmm perf
> shadow register sets : 1
> kscratch registers : 0
> package : 0
> core : 0
> VPE : 0
> VCED exceptions : not available
> VCEI exceptions : not available
>
> processor : 1
> cpu model : MIPS 1004Kc V2.15
> BogoMIPS : 586.13
> wait instruction : yes
> microsecond timers : yes
> tlb_entries : 32
> extra interrupt vector : yes
> hardware watchpoint : yes, count: 4, address/irw mask: [0x0ffc, 0x0ffc, 0x0ffb, 0x0ffb]
> isa : mips1 mips2 mips32r1 mips32r2
> ASEs implemented : mips16 dsp mt
> Options implemented : tlb 4kex 4k_cache prefetch mcheck ejtag llsc pindexed_dcache userlocal vint perf_cntr_intr_bit cdmm perf
> shadow register sets : 1
> kscratch registers : 0
> package : 0
> core : 0
> VPE : 1
> VCED exceptions : not available
> VCEI exceptions : not available
>
> processor : 2
> cpu model : MIPS 1004Kc V2.15
> BogoMIPS : 586.13
> wait instruction : yes
> microsecond timers : yes
> tlb_entries : 32
> extra interrupt vector : yes
> hardware watchpoint : yes, count: 4, address/irw mask: [0x0ffc, 0x0ffc, 0x0ffb, 0x0ffb]
> isa : mips1 mips2 mips32r1 mips32r2
> ASEs implemented : mips16 dsp mt
> Options implemented : tlb 4kex 4k_cache prefetch mcheck ejtag llsc pindexed_dcache userlocal vint perf_cntr_intr_bit cdmm perf
> shadow register sets : 1
> kscratch registers : 0
> package : 0
> core : 1
> VPE : 0
> VCED exceptions : not available
> VCEI exceptions : not available
>
> processor : 3
> cpu model : MIPS 1004Kc V2.15
> BogoMIPS : 586.13
> wait instruction : yes
> microsecond timers : yes
> tlb_entries : 32
> extra interrupt vector : yes
> hardware watchpoint : yes, count: 4, address/irw mask: [0x0ffc, 0x0ffc, 0x0ffb, 0x0ffb]
> isa : mips1 mips2 mips32r1 mips32r2
> ASEs implemented : mips16 dsp mt
> Options implemented : tlb 4kex 4k_cache prefetch mcheck ejtag llsc pindexed_dcache userlocal vint perf_cntr_intr_bit cdmm perf
> shadow register sets : 1
> kscratch registers : 0
> package : 0
> core : 1
> VPE : 1
> VCED exceptions : not available
> VCEI exceptions : not available
>
> MemTotal: 249744 kB
> MemFree: 211088 kB
> MemAvailable: 187364 kB
> Buffers: 0 kB
> Cached: 8824 kB
> SwapCached: 0 kB
> Active: 1104 kB
> Inactive: 8860 kB
> Active(anon): 1104 kB
> Inactive(anon): 8860 kB
> Active(file): 0 kB
> Inactive(file): 0 kB
> Unevictable: 0 kB
> Mlocked: 0 kB
> HighTotal: 0 kB
> HighFree: 0 kB
> LowTotal: 249744 kB
> LowFree: 211088 kB
> SwapTotal: 0 kB
> SwapFree: 0 kB
> Dirty: 0 kB
> Writeback: 0 kB
> AnonPages: 1192 kB
> Mapped: 2092 kB
> Shmem: 8824 kB
> KReclaimable: 1704 kB
> Slab: 9372 kB
> SReclaimable: 1704 kB
> SUnreclaim: 7668 kB
> KernelStack: 592 kB
> PageTables: 264 kB
> SecPageTables: 0 kB
> NFS_Unstable: 0 kB
> Bounce: 0 kB
> WritebackTmp: 0 kB
> CommitLimit: 124872 kB
> Committed_AS: 14676 kB
> VmallocTotal: 1040376 kB
> VmallocUsed: 2652 kB
> VmallocChunk: 0 kB
> Percpu: 272 kB
>
>
> Cheers,
>
Hi John,
Thanks for the bisecting and reporting!
On Mon, Oct 31, 2022 at 05:30:24AM +0800, Vlastimil Babka wrote:
> On 10/30/22 20:23, John Thomson wrote:
> > On Tue, 13 Sep 2022, at 06:54, Feng Tang wrote:
> >> 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.
> >
> >
> > I have a ralink mt7621 router running Openwrt, using the mips ZBOOT kernel, and appear to have bisected
> > a very-nearly-clean kernel v6.1rc-2 boot issue to this commit.
> > I have 3 commits atop 6.1-rc2: fix a ZBOOT compile error, use the Openwrt LZMA options,
> > and enable DEBUG_ZBOOT for my platform. I am compiling my kernel within the Openwrt build system.
> > No guarantees this is not due to something I am doing wrong, but any insight would be greatly appreciated.
> >
> >
> > On UART, No indication of the (once extracted) kernel booting:
> >
> > transfer started ......................................... transfer ok, time=2.01s
> > setting up elf image... OK
> > jumping to kernel code
> > zimage at: 80BA4100 810D4720
> > Uncompressing Linux at load address 80001000
> > Copy device tree to address 80B96EE0
> > Now, booting the kernel...
>
> It's weird that the commit would cause no output so early, SLUB code is
> run only later.
I noticed your cmdline has console setting, could you enable the
earlyprintk in cmdline like "earlyprintk=ttyS0,115200" etc to see
if there is more message printed out.
Also I want to confirm this is a boot failure and not only a boot
message missing.
> > Nothing follows
> >
> > 6edf2576a6cc ("mm/slub: enable debugging memory wasting of kmalloc") reverted, normal boot:
> > transfer started ......................................... transfer ok, time=2.01s
> > setting up elf image... OK
> > jumping to kernel code
> > zimage at: 80BA4100 810D47A4
> > Uncompressing Linux at load address 80001000
> > Copy device tree to address 80B96EE0
> > Now, booting the kernel...
> >
> > [ 0.000000] Linux version 6.1.0-rc2 (john@john) (mipsel-openwrt-linux-musl-gcc (OpenWrt GCC 11.3.0 r19724+16-1521d5f453) 11.3.0, GNU ld (GNU Binutils) 2.37) #0 SMP Fri Oct 28 03:48:10 2022
> > [ 0.000000] SoC Type: MediaTek MT7621 ver:1 eco:3
> > [ 0.000000] printk: bootconsole [early0] enabled
> > [ 0.000000] CPU0 revision is: 0001992f (MIPS 1004Kc)
> > [ 0.000000] MIPS: machine is MikroTik RouterBOARD 760iGS
> > [ 0.000000] Initrd not found or empty - disabling initrd
> > [ 0.000000] VPE topology {2,2} total 4
> > [ 0.000000] Primary instruction cache 32kB, VIPT, 4-way, linesize 32 bytes.
> > [ 0.000000] Primary data cache 32kB, 4-way, PIPT, no aliases, linesize 32 bytes
> > [ 0.000000] MIPS secondary cache 256kB, 8-way, linesize 32 bytes.
> > [ 0.000000] Zone ranges:
> > [ 0.000000] Normal [mem 0x0000000000000000-0x000000000fffffff]
> > [ 0.000000] HighMem empty
> > [ 0.000000] Movable zone start for each node
> > [ 0.000000] Early memory node ranges
> > [ 0.000000] node 0: [mem 0x0000000000000000-0x000000000fffffff]
> > [ 0.000000] Initmem setup node 0 [mem 0x0000000000000000-0x000000000fffffff]
> > [ 0.000000] percpu: Embedded 11 pages/cpu s16064 r8192 d20800 u45056
> > [ 0.000000] Built 1 zonelists, mobility grouping on. Total pages: 64960
> > [ 0.000000] Kernel command line: console=ttyS0,115200 rootfstype=squashfs,jffs2
> > [ 0.000000] Dentry cache hash table entries: 32768 (order: 5, 131072 bytes, linear)
> > [ 0.000000] Inode-cache hash table entries: 16384 (order: 4, 65536 bytes, linear)
> > [ 0.000000] Writing ErrCtl register=00019146
> > [ 0.000000] Readback ErrCtl register=00019146
> > [ 0.000000] mem auto-init: stack:off, heap alloc:off, heap free:off
> > [ 0.000000] Memory: 246220K/262144K available (7455K kernel code, 628K rwdata, 1308K rodata, 3524K init, 245K bss, 15924K reserved, 0K cma-reserved, 0K highmem)
> > [ 0.000000] SLUB: HWalign=32, Order=0-3, MinObjects=0, CPUs=4, Nodes=1
> > [ 0.000000] rcu: Hierarchical RCU implementation.
> >
> >
> > boot continues as expected
> >
> >
> > possibly relevant config options:
> > grep -E '(SLUB|SLAB)' .config
> > # SLAB allocator options
> > # CONFIG_SLAB is not set
> > CONFIG_SLUB=y
> > CONFIG_SLAB_MERGE_DEFAULT=y
> > # CONFIG_SLAB_FREELIST_RANDOM is not set
> > # CONFIG_SLAB_FREELIST_HARDENED is not set
> > # CONFIG_SLUB_STATS is not set
> > CONFIG_SLUB_CPU_PARTIAL=y
> > # end of SLAB allocator options
> > # CONFIG_SLUB_DEBUG is not set
>
> Also not having CONFIG_SLUB_DEBUG enabled means most of the code the
> patch/commit touches is not even active.
> Could this be some miscompile or code layout change exposing some
> different bug, hmm.
> Is it any different if you do enable CONFIG_SLUB_DEBUG ?
> Or change to CONFIG_SLAB? (that would be really weird if not)
I haven't found any clue from the code either, and I compiled
kernel with the config above and tested booting on an Alder-lake
desktop and a QEMU, which boot fine.
Could you provide the full kernel config and demsg (in compressed
format if you think it's too big), so we can check more?
Thanks,
Feng
On Mon, 31 Oct 2022, at 02:36, Feng Tang wrote:
> Hi John,
>
> Thanks for the bisecting and reporting!
>
> On Mon, Oct 31, 2022 at 05:30:24AM +0800, Vlastimil Babka wrote:
>> On 10/30/22 20:23, John Thomson wrote:
>> > On Tue, 13 Sep 2022, at 06:54, Feng Tang wrote:
>> >> 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.
>> >
>> >
>> > I have a ralink mt7621 router running Openwrt, using the mips ZBOOT kernel, and appear to have bisected
>> > a very-nearly-clean kernel v6.1rc-2 boot issue to this commit.
>> > I have 3 commits atop 6.1-rc2: fix a ZBOOT compile error, use the Openwrt LZMA options,
>> > and enable DEBUG_ZBOOT for my platform. I am compiling my kernel within the Openwrt build system.
>> > No guarantees this is not due to something I am doing wrong, but any insight would be greatly appreciated.
>> >
>> >
>> > On UART, No indication of the (once extracted) kernel booting:
>> >
>> > transfer started ......................................... transfer ok, time=2.01s
>> > setting up elf image... OK
>> > jumping to kernel code
>> > zimage at: 80BA4100 810D4720
>> > Uncompressing Linux at load address 80001000
>> > Copy device tree to address 80B96EE0
>> > Now, booting the kernel...
>>
>> It's weird that the commit would cause no output so early, SLUB code is
>> run only later.
>
> I noticed your cmdline has console setting, could you enable the
> earlyprintk in cmdline like "earlyprintk=ttyS0,115200" etc to see
> if there is more message printed out.
Still nothing from vmlinux with earlykprint on UART unless revert.
>
> Also I want to confirm this is a boot failure and not only a boot
> message missing.
Yes, boot failure.
Network comes up automatically on successful boot. Not happening when no kernel UART
>
>> > Nothing follows
>> >
>> > 6edf2576a6cc ("mm/slub: enable debugging memory wasting of kmalloc") reverted, normal boot:
>> > transfer started ......................................... transfer ok, time=2.01s
>> > setting up elf image... OK
>> > jumping to kernel code
>> > zimage at: 80BA4100 810D47A4
>> > Uncompressing Linux at load address 80001000
>> > Copy device tree to address 80B96EE0
>> > Now, booting the kernel...
>> >
>> > [ 0.000000] Linux version 6.1.0-rc2 (john@john) (mipsel-openwrt-linux-musl-gcc (OpenWrt GCC 11.3.0 r19724+16-1521d5f453) 11.3.0, GNU ld (GNU Binutils) 2.37) #0 SMP Fri Oct 28 03:48:10 2022
>> > [ 0.000000] SoC Type: MediaTek MT7621 ver:1 eco:3
>> > [ 0.000000] printk: bootconsole [early0] enabled
>> > [ 0.000000] CPU0 revision is: 0001992f (MIPS 1004Kc)
>> > [ 0.000000] MIPS: machine is MikroTik RouterBOARD 760iGS
>> > [ 0.000000] Initrd not found or empty - disabling initrd
>> > [ 0.000000] VPE topology {2,2} total 4
>> > [ 0.000000] Primary instruction cache 32kB, VIPT, 4-way, linesize 32 bytes.
>> > [ 0.000000] Primary data cache 32kB, 4-way, PIPT, no aliases, linesize 32 bytes
>> > [ 0.000000] MIPS secondary cache 256kB, 8-way, linesize 32 bytes.
>> > [ 0.000000] Zone ranges:
>> > [ 0.000000] Normal [mem 0x0000000000000000-0x000000000fffffff]
>> > [ 0.000000] HighMem empty
>> > [ 0.000000] Movable zone start for each node
>> > [ 0.000000] Early memory node ranges
>> > [ 0.000000] node 0: [mem 0x0000000000000000-0x000000000fffffff]
>> > [ 0.000000] Initmem setup node 0 [mem 0x0000000000000000-0x000000000fffffff]
>> > [ 0.000000] percpu: Embedded 11 pages/cpu s16064 r8192 d20800 u45056
>> > [ 0.000000] Built 1 zonelists, mobility grouping on. Total pages: 64960
>> > [ 0.000000] Kernel command line: console=ttyS0,115200 rootfstype=squashfs,jffs2
>> > [ 0.000000] Dentry cache hash table entries: 32768 (order: 5, 131072 bytes, linear)
>> > [ 0.000000] Inode-cache hash table entries: 16384 (order: 4, 65536 bytes, linear)
>> > [ 0.000000] Writing ErrCtl register=00019146
>> > [ 0.000000] Readback ErrCtl register=00019146
>> > [ 0.000000] mem auto-init: stack:off, heap alloc:off, heap free:off
>> > [ 0.000000] Memory: 246220K/262144K available (7455K kernel code, 628K rwdata, 1308K rodata, 3524K init, 245K bss, 15924K reserved, 0K cma-reserved, 0K highmem)
>> > [ 0.000000] SLUB: HWalign=32, Order=0-3, MinObjects=0, CPUs=4, Nodes=1
>> > [ 0.000000] rcu: Hierarchical RCU implementation.
>> >
>> >
>> > boot continues as expected
>> >
>> >
>> > possibly relevant config options:
>> > grep -E '(SLUB|SLAB)' .config
>> > # SLAB allocator options
>> > # CONFIG_SLAB is not set
>> > CONFIG_SLUB=y
>> > CONFIG_SLAB_MERGE_DEFAULT=y
>> > # CONFIG_SLAB_FREELIST_RANDOM is not set
>> > # CONFIG_SLAB_FREELIST_HARDENED is not set
>> > # CONFIG_SLUB_STATS is not set
>> > CONFIG_SLUB_CPU_PARTIAL=y
>> > # end of SLAB allocator options
>> > # CONFIG_SLUB_DEBUG is not set
>>
>> Also not having CONFIG_SLUB_DEBUG enabled means most of the code the
>> patch/commit touches is not even active.
>> Could this be some miscompile or code layout change exposing some
>> different bug, hmm.
Yes, it could be.
>> Is it any different if you do enable CONFIG_SLUB_DEBUG ?
No change
>> Or change to CONFIG_SLAB? (that would be really weird if not)
This boots fine
> I haven't found any clue from the code either, and I compiled
> kernel with the config above and tested booting on an Alder-lake
> desktop and a QEMU, which boot fine.
>
> Could you provide the full kernel config and demsg (in compressed
> format if you think it's too big), so we can check more?
Attached
> Thanks,
> Feng
vmlinux is bigger, and entry point is larger (0x8074081c vs 0x807407dc revert vs 0x8073fcbc),
so that may be it? Or not?
revert + SLUB_DEBUG + SLUB_DEBUG_ON is bigger still, but does successfully boot.
vmlinux entry point is 0x8074705c
transfer started ......................................... transfer ok, time=2.01s
setting up elf image... OK
jumping to kernel code
zimage at: 80BA4100 810D6FA0
Uncompressing Linux at load address 80001000
Copy device tree to address 80B9EEE0
Now, booting the kernel...
[ 0.000000] Linux version 6.1.0-rc2 (john@john) (mipsel-openwrt-linux-musl-gc
c (OpenWrt GCC 11.3.0 r19724+16-1521d5f453) 11.3.0, GNU ld (GNU Binutils) 2.37)
#0 SMP Fri Oct 28 03:48:10 2022
I will keep looking.
Thank you,
--
John Thomson(�/�X���D� < Gd� B@ ����>��z6�إ誑��.������W�.��y��������ؑ�� J\F�� *< \'�"'A�: DX�Hr�l� Q��&.�� fa��I�� | 7=m���z��O�8 ��.����=`ƀ
�7���,�#��\��� &
JBTP"��L�j��$4��w�dҡ��x �0K�@��<M#1P8
< T@�E�EZE�t�D�(�cFR�J
� � ő���6 ��0x�ȼ ��d[|����8 �i1p"Fbo�0Mt:�� ���
"�����@(�
�ă
���|�1� >8
�4�I��&.��E@x�́EtٷipA��G�M�C�8Q\p���q �Ǘ �����Ħ=0�q�I �'z$(�=& � "<N\�L����M�� )ń�Y��z% X�X[��|[ٸNS�e���F��%0ߖE& �Ef���u�a��6� �& >�Ƒb��8���$�l�˸̺���>�i��
h`&�E�G�M�,�(��D
�D
̲�
�S��H�"yn���I�#�oˠ2�eN�8��8�0P�4K��,��<Q��e�
f��,<4X*�� ��(
`������ ܰБe�|tp���P< D!�\��`�M�16O��L�i�i��; �iO=
�@��(!*�_�I$��� �%�
����9��$0�)��D�� ���Xj8�4a����=
����`�q��Ç(IN�$0�����`%����c�E��H�d�бhPAA���E�q��'6��8N���E�C�0ia$��"T@��
�0�m�Ė�,�u0A�a�#�@e�ˠ�4�")H��B��Gˠ�e���)1�p�
�m]6�e Y�_� ��|��H��8 d��"P�M8�2�@��'� ��P!"*xN{D��8
��'��N��M$)K���q @��AbRH �qpɺ �Q h���i :� `�� � ��;�(�� D�I@ �>� *P>@M�4�CS�= �n��8�Ӿ� <2Z\��(
��� T��h�
p�PAIP4�;nk�(Z�e�
Ⱥ��5�&�@�$ |�"(.N��f� 1R�'J�AO� (��,�� �1� ~�a� �p*�� @
�c�a�KP���8�$EF� �� ����:Nx�e
���y��Y"� ��.��!ZG��)��D �I$��@�MS�4�D��<��dr�D��&J�$D�f���Y�q�(ʂe�x � 4p���
�h�� LX��(�e�B��h.�Q9 )�b * �|`)>_��1��O\h��i_�>/�h T�2Ob�t�� A��8 T`�*ع����q���~j h�|A�"}�EX ea�M/� �Y�FY �G�M$���s@ς
��l��H1�(�#�� ���ó0������|]&A� ����<R�D�w��u��o��i�/q|�!Z�(�-{hp[�)����$E����O[�!���ƃL"T(F�e_�̃
>m�""$�6 �a��� ��2/��c�" FI\���ۺL'%6͛@� ��&|h Ea@�!����N��4��>�����?��#�P�� �&2K�>��1� �4�)d 瀐�sl�>(����a�}��5��4��(�<M�<K�R�M$IL��%�MHL��,�8qy���0N�@
�@�0��@
�0��@��(Q\��Dq�$ �KIP\6�"��� �M8��m\7��t�9��x� �&>K�J|�Eʳ�Gy���]{J&Pb�$6����$���Ez�%B� �\�_]S�Z,s���������0�M�B ���� $�Y��x�c�>o��
f 6!� 0 'FH���͢��F��~ڄ'I�#M�� �#8�e���M�� Z�m���<
���D��I�h� Ii���(�6�0|X %B
"���i�2�"�T`�E�C����Y�'� ����H 8 n�LLx"&$$H8�u��j��Qb���(��4�A�)Ƞ'�������yY`a%���A�� `�@���x�$ jG����
��3�C;h �j"0*x����`�@�
<,���B��&����A [�m�n |d �l�����������<kZ����${�o�"��A� �� �G6��`� S هA&���^�0��߆i � Jl��| 6N��o�"D��59�`m��a]�YPa��ₗ=� ��� ( ��� �7
�AY�G����,lcXPa�n�d� m�8�B��LA�yY�"9 x T8���0\��Dx�}�cA�����p�2 �0T4M\��BY��Y��������`�<���8�σ
��i�7�`$�$p`,�M�QQ��.�<M"� �]W�Ħ}Z�x�y�A�D�������Dx�@�.{t�=��@/Ӻo�>8�aߖ�́I<�e�r |d��I`!m<H �uٗE���p[�m �ˋ�H��1�ax ����E| �AEY�������c�"�!��e
呚*x6Jl�d�u���A4.��� ~m2Af�
|P )
b\�
�&��q��@E���)F l � �H �L���#{<2�H�
���$�d�e �d�O۬�!l�q�,@:�fn�,� ��<�m�`��pF2<!�Y �"�B�H�e�Lb�:�m�C���4���92��<L�`��6�a�/�@����4/�4<h�-@h���v� � ��}�I�y�tXD�v!��}���� �@��p�u/4(�i�� 6�$� v |l.~� ݧm�W��@�r��.�& � 4�u 4 =
f�6��"P�o�p(E��o�
��E���(P��4Y�k�G��� >(
�! f��8��qP!��i��Q� �
]� ��<� 2�i� C�L����u�EY��F�Q`��ϡ��Qp�����<��M6m�W� �^�� 0N�<��D�'�Y )vPI`b�sV��[&���!� �&�M� |݂-�p�̓hp�m�K�� .���������Ӽ.S L6�B�@ ,E �)�5�Q�i EQ ��u���C��� �� @�a�7��D��
�a"-L�8PT �)�� ���"{ d��-@�D��@_$AW)�۷#�͏~s=� qQ��o��+<��ѫZ���Z��=��0����w�}����.�ֹ��+��r����m��S��?�>��<��(��S�o������W�R���b�_INU�)�m���mL�2�}џ��Tnh��d)�K��HEP���b�*G0����%���L�E��З�~~�k���}��T�z1t�+����e�2{�e�;�ߵ�|͓ۅ�(�%���M/?��_�KWl=55�Y;H�+}{C�r��S]뫵��l�K*�P�}-��2�:n֊Z>��7��7X[�Y�Wp��5o���ws�-��OU����OW��"��w���~���Y��y|��O�;�>�m ���
b�H`�O��Z����6亸w�Wξ�]Vu!��Ч��f�^v����S�m{���5[M��Uh=n��������W^����]O�.���Z��d�����V���oڨ��kƜj\�5%�e�%-�_��l���f,i�}��a�y��!!�a�S��K���:O���>Wί9��]q�W��]y�vQ�����ۻ���/l��l�Z��]R
�e�uY���OW�\Z��[+��$k�+��Է_ �a��]2�C�z[>��n��{Le;}0�hGƠ�*۽��u�֙�i���s�3�ݿ��45(d��~O[;�g�_q��Y�Na.���߫�ކ��5~�\^ݧ����+�����kڧĖK/ee���nX�����,TZ���$
<ܻ��\��Nr�Ok��5Y���v����vb0���{�-������>۪O�3xÔU��
]�i�W�>��g\'yp��� H`��rl��[��^�{����_ꛯ�p�ĽQ���綺��sV]Z =+��u�����\���b�k}��璡���y3Cӭ��,YZ��^o�`��Yf��mT��[{�Ք�uw�q]H�?e��������k����������Y2.ʫ�j���,�*���(���}�_b�@Y�5�q��UY��Bk���b�g_���%�\[��S�&y���c�$�������s�� �ݗ�מߣ0�<l���\1��/t�!� x�\w���]���d- l�"��V��ߚY��댩��ˍ�t�BCf��K�1�ua9�v�����
��xnx`�6 �7���~��j+���R��7[�5Q���z��7�<ŚS�f{)� ɩ[=���\��_g�k��gSa��%��Q�s���RQ��,K���ĨY���,l�?��O+�Vox��I��q�^^��'�mqe�h�{;Q�en�K�.�
˫��KK�i���s�����=����T��{E���禮�_��9}�*�מ�Ǹ`��Q<qR4�:��@Y�rJ�RߟI|\2(�ֵ����,,*����kť�R��BV�5��sy}=9D���ө֔��+����\��g�����
��rY{CϕWKϬ�^�{�������/��W,9���Z�J��O��iU�2��D��0Q�x���YW?�Ǘ]��L�+��\!5
�m��Եϵ_�^���\2��8�2[��_}�d���e�o���WTe�R�\}�y���)����m �. �u�����,���\�l $�Y鷦�WNz2�Sj�& r�C\�j�+�C2[<�c�
*� 鸈.�J�{�{%^�گO�)�� ��ba�����j9���v���__[+�~e�l�F�uQ�v9Q�^S����Z�~Y/<%���F��O.?Kks�-���J��TV��u���,1���sM��`��[���ǿ.x��ﵞ(whPN)��{�-_T�W�Z����J�z����ފY-_3���~�m����S�^��^�?u�xNy�9��Rj����گ=�C���o�疧�A��~���VRSN�[�P� W��WL5�
�s������J�}i^�|��7H����/q��Y�� ,�� D� ,Xե�����vڻ��szY7��,v�ԷQi�L�d�
�-�KCb}���)k�O]ӯx��Z
k�4��?�5�+}���Sr�]��yݍ��ɯ���wxMn�����i���+��s��{���O����ϗ�S�kO�Ms`����K�a���'���
\�|]��Y��7��o*��|jO���|}jOY�� W�O�ړE�=UjO���U�C��{�ߪ�5-Ֆ�� ��1��;X_gm����e��K�WL����곞^�}��q��Jͧ���[���s]�+go�J�:���b�;#8��2�@��l������5����6X�K^����X����5�ğ�� �g����V�4���ke�%'{k��)����k\��]����%��
���~����]�@�q��*��wm�{I�R� �+��m�@}wh��uhH�p�m �m �k�U��BZ ,�+����?���2��Ӡ���z��z*���,x�������6[H@@h����I�_����k���e�X�����˟�_��4�_��͖V�����������ߥ��^���ޟ��s�K��w�酘��+wYey��0J� U��9��m ,�F�?k�y]�9P ��I��H�V�u��Bd-��9��p�DBP�Y�&* Pb�<�0 "�@�Ih,�c��0�(MB�HR$1
-��D
�� !5��h�L��0 #=�%(̆-�"FZ�H��H��@\�,�"m�%�O� Gr��i"�Q"�q R��iIz����(1Q�0L��Hj��q %b�E�$fa(J �,� -��H��}��A�MK�� ,1�A:�=M�@ ����zY�sd
�*Kk�E��fT֓S*��C�0�r����Q(Z"GR�'Z�%!*�
����Hr"�y"&� CI���p�D4XD!J"<(�����D ��DL��'2N=�*<@ #9PB� ���<���C���y kvf�P3;� �a�2q��2P����x�3`�սսŧ%'���sg � ?3�������@�� ��H�#��,�$a|!'� _|'���Y����̞Xb����A����
� �
�cC�fY��L%�i��z7A�m|,kQ��@�E�V^���� �q�%�"D9J!
˯pl���t7|�Է�ٙ�-gyb�/E�Q"^����22�
��YcS]�$Q%�y9N#y�(�疢WǦV��f½�HN�8P�0L��, Sl� � �i�'6�a���}��!�ņ���0����t�,��x�^N��l*?���t����}u\�"S�+�WU��L}����� ��U�U%!����ԯ&J���F 3cl4��Lʫ�_[j��;��p�L��Q�#ɹjH�!�ݟQQ��0 ������P� /�⊶���6>2��,�d��S::��[�]y��m�
� o(Nl�Oa߫l�5��v�H�̲'v�5��}�3�}�l�����m��i�G�y��`�"'����{[�7��^���)�������o̐ �l|���e���ۭ<�Fqv��ll��� ������v�z���aw�
iw�0 ��v��7S])o�M|�7d��\�����]���z��Y������s��m]�1�X���SGP� #���E����*KUse{�.�_��p �
��E �-��. ��K��m�s�՟��
k�������F��i0���8�N&��ܓIέ�+uu埋������ع�l�b��X�/�ý��5�ƋJ�ޒa}����s������K;�\����~�}�����g٨Y�F�T����)�e����G����R�#�\^
�!1T�%BظLi^7�_�<�Y�l�,-���,Ѽ��J��?�6/��ρu�ĬXV��f ���a����H6K�Ŵ[_s�
�`dɀ��A�Nc�J &ym���ڶsLe��ѯ.�$�CN�zdY��<��㬾w��rm1n4.�d��A&X���0l����9���V��n-�������(]�� J[z5�ۑ��+,��b-��C��@O51���6��h�i}͒W�Pb�������1J�;���q�����}���e��6�eYQT)qƨD_���A2�7n��i'�(j�F�)Ml <'E�1[*��,l��ڰ�絪,��ֻ�����֢���^ې�����2��u�]|i�F��.b�0�O���wF��[Y�W��'�6��s�ͽ�Y�Wܰ���)1��7j�v��Divܧ�U��w��ԟT��R\^-�n��m��"Mym��NU�_ ˮ���^�#����Hbb�<�ṟ� ��@r2`=���45�~�
_�e`�e��e�Y�\-����kJ,�l�\-_��EV,5KP5��D:k�t˫�u/k��6�|e��hyo�t�|�9;Ƶ���HY^�����fi�\D�̺�"���?���O��N���q���ފ�����k���ʫ��EW�ŭ .��1�l��A�ʒQ)M�hT��F�Y"ː�e�|Q��k%s��V�.u��Y�|���{�i��n
���ZS�<-�<e]Y?�Vx6S�^^/���W�����Y^ ��c��;H�}���+�����-��F&�Z+�K��[R����=� =��5�u�}���u��h0��_{��.f� /�Gr2\����ٻ�okMͧ}��F��)�?��3I�Ⱦ ��}�X�|�|yݳ���>%�Xҫr9�\��b�)q����P��rK��l��[F�Z �4F�i��HL��Vk�E�N��ۧ��7���w�{��YNҚ�����-���?��֬[��%���o}�i����6
�mX��G�a^�Bc�y%B�Tb��nK����i4�1 ���Ď� k�k]����?�M���^ې^��g���Y��{y�����'v�0�ĖAњ��v�;�=(�r���I�ny��ִ ^������\��/�ԧŔ�>Kks�0�覿�,���E�x�/_��2ӉS����l��W;���2��ҟ~J#���ҕ{J{�y�
֧ O����.e�����V����2��*?��-��ѵo��xn� rJ��p�K�3����9�ۚY+��zZn;�2;<
W�K.5���
�'�&�W�J# $�@�� X�q��2i��f��jj*�&&¦x� ����Y|�J�ܨQ0˶�����:�ďlӰo� ��{yw�|�ݕ���z�y]H��ݨwc�I0\��?���K�Nq����b��f����r�[b���[j�\�O=��V��p�� ���,9��z�5�T�R�f*�Nק �,�0�u]�u�i�N�2s�,�yp���M|0����Tq�O}�Bo�: �İ˼ c=y���O=��s���c��O��a�q� �Id��J�
��� b.sֳ~kaL}C���O��ɧ� W�O�~��\h\���z�����7�5}ʠ �����4ˍ��R� D�q�V�_:4�%�r�ۘ��' �������Vk\ ^���̕������ [��ff�-S�Ap8�0�&�y�v�h��Էs?��vn.��u/��N�_����_%� m������b�������d�Wkoy}���ދ��WLݫ�n��ίy�}�6�}S�[�N+��}ͻ��~��xu�]�����>�=�l��UT��r˸��,V|�7��$� K�V^�?/����3L��_�opۻ�s7�op��<�ʯ=�� �W�%�����E��[+�� ��\���g���W���tRMY�[���}����~�&�,:�~�������; Z��yd& � ��+ "�MYj�m�e��@Q=ښ��3�o���f���N+�)_��/C��Ꜳ�o�>�o��}�$6-�d'.JTk�������+���^��~���(P�k%��c�7�����/'�XqYJ1�� o�k��������ue9��+f�s���W.���s��iai��*n����O��u�g�}���k��X�)���r^:��iJ��)kK^�|Q��� U���5����J+��K|q����W�d�V�Jk���\2�����v��kg�� ���*�*!!��e���d�$6M�� ܭUQ�����/���=.������g�%6̒L8� < ��յ�kF�]*(^���ח���l�5+�Igݭޙ����Z�ݿĦ��p��f��}�.� [I,!1��農�x����ڳ���ʷ�/J�h始V����S\2�����8��}�9�h��h B�8�ګ'�@��S%{��ڋa���z�5������K�ս�:�|�(5ً-e�y���˛'}��@������f������T�Y闘NX�/1�]b�=�[�ߵ�]Zv������ߡA}�W;�l�docN��s�91�l��0 �0���O>��O>��v�?�ի�������B���������5�z������ro��ytX�]��l�~��u��$4/��l6�i��7ti9� � ��.�o��>W^�����=��[�Í.�yY'�pT�~��L�%pd���a�m��:�� �Л|Z��f,�Z�� ^�#-���UZ�b����T��Ӕ>6,�� ���&<��=[i1�9���Vu�� ���<l�-shߖM|� ����� ~9��*��֬~�IZSZ������4 ��'�eW3|�7����w9�XN\NH4��8K]�k��g�բ����$ �A� ��A &{� GK�*���
�T�[]Z
�m��WW~ $�� �OҠ�~^�w�%��_�^9��աr*k����a 4��N"�6��}��E`���I 9�"X��m�H�� ,�~%��j<����WL��[�7T��_�W��\��P��
U��Ӻ�#� ������oV��_��:m�ս�7X/5�J|U����ό5�S��GV+~�Y� �a�8���KN�n%F���i�%B�AM�9"HNq����YJ�焻���Y�z-z��O�� ��z�T�S�;-*��Fo�g����b�SN�s��%���.~酤,ז�J5��� ����ŧ�����BP��S�U)Z�,���R�W����x��+��Z�O�m�X15� ����g/)�xI5�/�]S�^��H�[������& ��zJ_���/�S�_R_ ����:S������ um�S���N��m�
�u�ک_����E�(�n{���M�2��/$���l�O��4ˇ�u�MN��W�n�v�� L�Y[
֦�.S�p��/��ڋ+y��S���Z��$ Qg����`m�o����v��r�k�l�%fl��S���A�ƌp6axt��M���n'h��0�`� � 3��F�p��u���͑I<��D��p��Խ��E$|W�q���d �8����Z�����{C���ɻ��n�u���A&�r`(���|s�w�k�(6�,W5'סC�2������� ��Jq�
�EY��I��8�7RC� 5�_y� Y���k��Z�ۆ��6�!L�_���/R�����Yy�M$���$
,�Ep �94컶6]䊹�W㒲Z���Ib*1�8o�:�ւ�瞺�D�8�YLg�aZ ܱ����� �
ߛr��Yb�x������0Ȏ�-L�v��'� 5��L�7
I bb�K�c�!��=�f��9Q�3�b �/ ��c���Q��x�x�e��y���4$�p�=L�4j8�ګ�3Mrz���^Q��������<O�-�=W _���ӊb�4q�?=��O6[*�k'�/����g���It�i��マD[X?�u�\V�KHh��& ��c��owo\���N����Z� ��!���$ v�c�@O�*�@4�`� ��Td�u�(9�:���a�]�)�Y��q���p ��iQeq⭱��§�, �Y&KZ/��������wZ5[�qŸ^��ԯ��+*+j ����O��u6��^���ym��+��O�0S����jDg������S� *��U�~�����ݲ�K���ע�,��S�(�w��e��z�Ҙچ 5̆�uUo_��/jq�S�,��ۅA�?�+ө_��?� �X8����4�u\D�,x����o�qЃ��\�����S��״ �y ��~ťa+q���O���m)q��*<�s۹��N�1��(
��x���S���2��k{�����W>[l���3*�y��ׄ�6 �5���ԯ~M+g=���~M���e����/ݽU٫�r��� �
���M]1���� sJ ڄ�Y�s�D�Y�����}��)1j�Nu;�0x��qax<���� �ϥ(�W#�I��< K���v�h�����\"R�%��'��ؐ��P�H̾Z �v�0N�T\�$��84���Վ"^���J����+�g�^Z�/܀ )Bt���\���L�M�2k�e%�5����Qc�� R�0�; � ������\�g��2|��$E��w�@ѶP\��o�ħo����[ AѾ�������0X~�?�6�M�n<Pc���wr�p�3y��<PC���PsJa_y�g^�]�O��áRj� �+���GR$E�1M=�9Ml�lj���O)���S
k<�C�>�����s�?�����[�v���V\��Z˳l�`�~Wpͨ^V�Ffq(�IR*�^v��#�HT� <�O)<�u+�P�x�L�7�t��"?��&��s�S�<e�m�~t�\��SƤ��Vi���ԫ�0�e��D�+
�iH̠ ɭu��f�)ݫ_�o�e�˺ͨ|n-0tƼZ�B��8t���H1h� &bT 1c��"��=�.��,7�[ �~Y�-��l�U��=G(�:�1��'
� ��1D�`��!��@C X�d(�Ȅ�*� �M>6ZH0.4:0H�0 'A�)�P
���:2m"��S�m�
��g�V ����3C��b =��2 �Akk�h �\(W?
���^�ECp�@���я�4�ӐZ ���B�Bw(�d [Eը70F&�W��pv3�GZ�E�����o�ۉ|\�a�y����9|���]r�0�i6D@�AG����9�� w� ئ�Sn%� ma�X�q��o/�-,3,+�8�zI�u��B�#K��ĄI�lb�}�6�R1Ɛ�V�X
�������V��x!1l77��'%"4�Sj'��U�â��Z��-��D��m �I�P!� �SČ�F3e�N�5 Ht�� �[ <�)��n��%��@s���-��v��j� $��b�!� ,q�$m Y\N�gk��mP�_>\gt����;����a�~� ��~�1�����=gh/�X�U�H \v����ve ��>�4���u;��V0R>� &O,t���%R U���z��t�B�s�`�OD���%�ѮM���֍� �Y��>�J�8�F^�ZЍ[��4�M>�#� ��W��`]���"� )�wb�S�qZ0� nHw ���9����"�_<~�oW�*�k�ۣ�Sh�� ��k2w��d ��"�A��ꃵG� ����8�=�x#;bU P!7�o��{C!��/ U���aE�
���"A�$�3�9�;�j���1����hP�0Th�� �/�<9�t����"켓�Z��8H�Ѕ�لo���U8"ܸ��u����� ������>�H����~��\X� ��a&��tg�Q��ؼ�
Q+Ѝ���JU�Et#R��j cDs�A>b�g"yUIa��t�ێ�^;/���,4Q���4A�F���樂 7�U \Zy�B�A���\�&Wrw�P��*��16�\$��͗�8���V�^N�D������������+a ��?*B�0�e�F�
��T�:�5kpL1[�w� %�S�k���q�Qc�wQϪ+s��V�´Dd�ɇp�� Z�
�������Z 4k�t�5D�SU�ZZ���M���O>���(y� ��F0�+��[I�)l9����l�|��jWeLrE����d#
~����`�a�
v�ѵ��m�g˹�-�C1�MJ��F�"� �����*��ϔ�l������
aٹ��mL��E��L{@.z?im����}��N1��� v�
G�:�>!z�R +iB(�T��)���Vd�dj� �7�P 8C� �H�r[��%Y=����Z��'fᚑe ���6wnү4(毖v�3� -��\�8���
��뱤O[R�Pe�l�\O��.�XM�˭S����?�<]O�q���~]� � S&�I�@Y ��"��};L��*袎�̰��!.� ,�)�O>��z�b�p�:9)=�a7�I��Rj"�'����Q2��d�h �Â�X-�GR�� �����U?M��;����A6���J��QQ��yu�&C�f^X +�4^�xҠ:��7����p�U��.�[��MR�f3��[�~�|�KM��H(�fQU[�X���9�<���1�WW �-z
� wsO� �|�N eC$�oeo�u��cI�[��M�#R2Ϣ���F�~��I� � ��_�Rl��>)Ӆd f|>�{fM���%�B�Sf��ɩ4����+,�.�O�a�͓�A{�As R� ®¤����@��r�Y��R(F՚����`z��3��a� vr���+1�T�{� ���u� %% ]v�� �<��&�o���?{����U� ����2��%(�?��^�#ǯ���hP�@�����NJE����{��� $D��#�J][���,JO�sɑ�cL ����
�"ɥ���L��[ ���4������{DK�$/�<�����|'^�[�� �5붊*T� �z�<��,
�<X��Z�h���0�T.A���>�K� �DH _'];cȄ5&������$6+1��w#N��QT��<.�\�L�ܑv���%�< X�>��̪ t�
����M�9�e���odO���1G�R�c��_儝�e�̯�̭�_7��a���t�S��
p�v̙?���?g���I��f��X(�_sx�Ձ���ń�)�4��l�赸� �L'
Ї/�n���n��T�4�� Q�*�7.��T>�07[�
��R��9)/V0)��v����G�S�T҆_-�E����R�}�k�>%�@�/���;6 ���ͼGg�����ƣp��Z�)�絁�a�]!�@)U���R0w���Ҁ���uU!���n�������/���H�S��I���eV�
�B^$i(��)�"50:X ��V�o��0¸��H]�d ��y�t�� �~���E���D��ɏ� (j�`мP�n��� ���R���T�Q��2܈�u1�1Au((c�B � �npn���R�T��b�C��oJҖ�k}&a� ����^�Q3]u�QI�-���u��-W�K��ŷXͣSU��� ����E��r�@t�^��� ��)O�Re��%�ל�
�퇊%��<���� Sv`�H�Ār��_L��q�[M9'Rn�ܖ� �z^���G D(@z�Fj����?�r����Lg�g&{�v�J=2(��)��
��.^5�覢��f�2ć+ZV�8�
�=o�%O��R9�8}(�I��gϲߴ�,p �}�+N��8T�RF�r"o�e�;�$hq�H��/��� ����a��
���cs3�n]ѕ|:y�ˌU�`�����B��ך������|� nK�A\� ~fQ���� �0�@�LU��6��� �_*���q�x,%}�X6��5>�����'��Ql(i� L��0��� }r�� $�Qu��-�&����g�O�m܄�����TY�I�7����� ��饒�Y�2
�N��`ۃl/�r���,E�9}=���st 5���"�m,G�jW�^F�� ��a���EG}�vQ�<�O���r�8z�n��J�523�6IT�O,'�J�)�I�K�B��Gר� �70�B(�������j�*s�4 ���:^� df��#��� ��SHʼnJ���9/L���f�� �<�1�ʝ����Q�<���],�� (y�d;��A%�V.��*2�L� AX���_�R�6�$K�M�ו8�$5 ��[z�R%<����ǁ (�;��� $�b�$��R���5�����r�'v���vpJ�n<�Z�=|_�z�W��<�ҕPT�`*���n��e
Og�0 ~n,Xq��8�S���\��?�5s"*��Ҡ�p���|�H�`� �[��)���}�
t�u&@��7�7�{&�ę�H��b�M��B�}�e���bt�Ӻ�d>��H>?�"�� >�aᘀo�����j�?.�d |�����1�5N�%e��hm��0 ��?&������OYBv̛|Ö��i]�����µ>Q�q�I1��g�� :�\��:Ф�" �A�_����j��!T12� 6 ���`�1�dT2j�H�� �f D��#>��m�O坺��'���#=0�Cn�Xs���g��-!�QY(%46��Sb�mf���=ov�I�� 5�r0[�A���ǚ�g�I �%�mK��C� ?�j�+��$@ɋR)�K��?`��S8��iی��}L�bpD����ÁՃ �!�}��^�w���W����v��!��c��� � ixE��5Ff�3rM" ��x ���3' y8�����B���:�d�e�V+��A�b4�Fx ��+�j6�)���Gi��y�iV� ��I���tLvg���|0�[q����>�d�
��M^�n=kw�$7������t����� ����j���v�v�E�?}}�����3���W�.c�o�R� ��TX���dH@V"�%CFC�QPO(�k�h"�A�� L�k��ӽ�!��a�B��`�n*p16e��͏�*׳�`�߉6�
=�`��������O7�:���ٝ�������7�g� >�?Udv
=c�{���u +�dc&�Ͻ�0�y�b�l�4m|ͧ�;ɉ.����u-Q*�~D.�0��0���<��$#�R`�zn�8[�e�"�tX{y]� ��RX�;�ɡ�酪(D���\D��O�".ׂ�3��%D�!0�i��+��ٞXu��׀�.�hmK���rb�o��x���"C�l� �dz2�E �O"JS` ���Vk�%BٗTL�>G�TǙ�+�#�!á1��q_N.�3��; ��2�TJ�4���=0h!R{��)�D�(�>��je���{��=@v�G�K
�5� ��E�D��8&�B���jT�
�ӒGι�{ܑ� nhâ|xH�,�[�E���#�Z. � �E��HW'n��
�� #��o� �S\t6��[p��!Qff� �JC���
օ�
�C���r�명�I�}�?�ld<�^���'a|l'��_��wSiZd��n�G;�L���4 n ���0i���T�W� [}rI ��H��/ߐerFؔ�503Я�~�T�[qRWɢ!fd�
aǾ���13b�sVJË�r�ͦ���>k�d�
� � � )e���NJm-K$�
H��P��I�Ω���Q� ��)�0� ��b�<~VT] zva�72��ج!�p���� B��*T�O��7�[����ck�Y��T�r5�@{Ř}�,���9�f�n����dE�*Q>S�OƐ�1&}O/�J ���*�,�ʸF�a���٤�3��j� 8b�\R{�G.��j5���4rɍ����ǡ���"6tj�'E�o.1T+T
ڂrV�,=dɼM4�W���?�O�j&(A����r=,WYe�go4@&r�N
�a�h �f�.�*�UG]�.2�h�������_�I��%�V ���e�Y�����PW
;Y�l�
6M���D�I.�aT.�J~I�������Q�x���
�B�;��D� b���K�=*V=諑C6B����������:q��
܍WFb�*�@��,og1�Xh�Ǘ��N�ff9� ��/f`~�\�r@�f.!� �u�%�������"6�>tL2C#� WK+_H�v�؇] ��z�(�z��
8��R��`�W)E�$��!rFF ν�r�x�6�E��WnW.61 ��C<�H���N�C�[N=��� ϵ��8��q�0�g�M��Hހ��/,�)Y`�9��{�7� �Y�
(���R��PΪ��Z�<z����_����ә�N�������L x�}�*O���F ���� ���DM��L�ܖAE��@Dio�opt�R Ly
Q�O}aH܌P�Me��n� �@? �`>�̬��V[w%̞�����u�\�Ə|>7g��@����r'�!^e5���U �� ���Vd l!q���L�S
����JLe�D��#?Fl(�=d - �&}o�t1�!+r<���8���Yޜ����7k�
��FY��i47\\{~�[u�m�G��h ����)��b�l�vX���|8 ��� ��|| EN�� ��j����6�cЦ"U�s!� �R�z���s��}�}σi��0E;�� �P#��,���n2�p�` ���(�H�?X/U�-��Aݶy���t�f���'�m �L�ɘx�`�P��8�������N �^�.u̯���mx˛H��:�
7�+� T������E��� ��G<����D����� l�(�&$���}*@��T��e�iu�<�/� 6��b����T���đ�#/�����A
����p�(>�H9Yds���u�H����b� r����N� �/q��������ق����b�Z�Rї��1!}�{?+���8�9���g� �^�
�:�K?�U$F�
Ϣ`�]��B�}� Y��; ��&�ﶜ�*1 d ��P�T藥O��
�S��*�,i���0Ao3+���U7nO]Lo&���$�2��=�v�{�e�g�i$�g�W{Ѡ@M�
�:O �۬���"Vb
���T.�= s�̦}ӆPh��]%n��]��Q�2Ni���^H�����,n�(N�S��E4���>$���� ���)�r��兖4E��>֕:"z��c�
����)�K�%�� 7�$?C!䉻����5,䞄Dx!ɞ3�SX�a�gp�:28�8�a\�h(�|���s�If� �Ow� Z�O�i�j���I��B{�w�M e`�|� a�uJ�PKW�쉄�I�D ������ �)� �vܱ��4�N��!I���4���]�5H������r���R�����k���r� �G-
�4����q%X.��S8�D~�~�H-�Xj DGdg�bctx�`SQh�
�m N��\į7 � RCV��3�!�X�� P�ȡK^!�Jo�5dm���)m�]��������I� �8[���6�����ʞ�����
��w`��t�� ���L S��<����ifG `�(8=A��Rs$�:��żw3��iD��ffB��lP
X����)���
�++��$PH� �p0��6����̘�����6��.��!�Z�먰J����h��H����XR�r�J��lJ���N��$�9�?#@7,bHq��@gsi!x����/���ۇW1٤����0h�i#��=b ����d!�� =. �X�7Ɨ�e2�]��\����v�C�j�Ώ���$`? v=mO���U�(^Dhim`Fg�� �u���&w*��qw�
͝#jP�)�> �X�hd9�q�����jI��A����1ց�$5�Z��L�CnJ���NǏ�&���8l�������Ğ����,6�4v���9�e���F��L32 �(��?@�$��s
�{�"���DƏdN(�s�5G9��. ��(5E�ů,tJ��W��9��)D� �Mu쯰$$�I��h� E���Ț��\������ɼ��EH�U�̄�D!n9
7~�㘈�<�X�"��Xx�+~��P�ԭ�����c�Kr}�)�ߺQ��X���TIC�̱� �: ��"*�,�Gw���V���R����8x��F��)�'�:O�a#$l ]�ښ0�����B' ��5�ʝz�G��:�
iS�Ǔ�@�rTi M���
9�a��9�-ǫ�A;:<�� s
K�M�hH(:.`�7����Dj� �tf�"]�b]���� �.K+Lo ��2Z�����x�Rv�/@*p����s���H|����=�1>:|����5�뉵�e7��,+iu��F�q���ҝ�P��iv̽5I��9A��ZW�0ٵl4��i*�&�S�`ۨ>���CzWot���ƾ�p{I�m�o�d�1T �$+���w �Zr�6��J�<wR�xV|1���cС$aM��Vo( ���.�¹s��v��,�J5��ih�� M��8��P�y,�2D3 �{��BB� ^,S�>uH�0&��" ����"<z$3D���Y.�W���d!) �#�N y����TlEL��{U�V���j `ߐ�#1ՠ�p�} |��
���� |ya3�YE��x�$B��V���0O�_��<��V�R�ER�ĄG ���6_�@�a�) ��m�h@U�?��(8v� J \l5Fcp��������,����I��$E��O��nt�;46�| �(d�Ʈ) �M�W�E��ݺ�C+P�e�� ��i���w(� ��o��m����`OwK�Z^=u���u ~}�⺍��*y1�p��;��鸃qT �,�rsVf�jʜ�
X� Ij� �W� �y�D8䢄�� �F���W;FL�S�\����8�]
�
Q �/ɂ7�ťb��0����j��*XX��� �;�T47q,vVK�f
�R������f��Щ ���6��lJB�GsK�LL��"�x�u�8"s�n��r�0�}�YT:\�E�R\Q2d�Ğ�D[�S+���T��AP�I���� ���
��G`� �$/�_�ZW| a����/���Y��� %� W��� �[q�^�ہ��I�ɸz�yB��DH��Z)��-�R.��1�*:ձ���1 � j�<��� F��vy�*�'�]�υ�К(�6�*�c��P�
�n��vn��� oNX|k�D��u:P�����p�syS幰g�|-f�� ��8'��ْ������eqg!��a�Dq��$��S�)���_|5h������s� �M)��&c4�l`9 � ����O��� iJ������j��g-,��2�����&����h��S'�=��W�u|o�T$�
� �� pEd J<%zn93B�*w�ZJ��ء���j�&o9&�1M ����z5�w .v����c�:��9
�͘z�TȽ�݅b U!���"8�� F(���F��k������AgӢ�)� B @��� ��RR��G�q�fP�(����!.V0 ��e$��8NjD��� u���3��e
�*��\� �w4��m�.��u��,���j��7ѽ�%�M28w"�����)�B�@�$L�_��Rp�Q���Z Q��!��b��=1��+��T�k�Jv1�67����UD v�Pt�3^�j���$H^�$�Y����V��d�]=^n� ��\ z�.J#o�@�-��_�S�Zz�� �Q�.4 !C��6{ e�Ml1�{>�xx��B�c<#J���sX�9��q���w�d H��N�]���� %P��D(��M�_
�&r��S&��m?��"��F�"�ċّ�]h��qϢ���daJ����s2hl��� )�H�i1���OR ��$C �wQC�-�T� d8I5~ڷ@�� �$��G*�)���T/@�8��XeC�;<���
y��1'�&h�"�9��|�j� �2�d�f=J�lg
) �P\�����l(t�Ě�.�=�%kX�� %�hT�?���ó$�9�ҵ��S$ġ����w���em������R �`����H���JR�`�H�SڄQ-��%�ʸ|d;�B��c Q�&��9��{*��ț����5o�S���*�J��9���c���R�����EI|�Fj����j
��?���P��U��+!�H>�}Fǩ[�$���fnؠ >����=�QK�� ��AK��3"�j5�a+���maN|-�br%/D
\�^�����3H��G9�� h�#�-��*����#}D��I��� �d��ix<�E*Iݝ�FVJ(W�<�4�TT�I,P��<�.���Z=0��� ��A�>�8&�yfߏ�9�9 潭��7�m �i�hm!z�B��9)~��:d'\ �F�=�#VG�L8���
��wF4˓{�3eD,� BB;��� ��wV�<K�r��#�աa��[�*����
*ę$�j�/����l�X�
d\1NĎL� � �d���9Ӹ�l `K�UM�OS%�g���4��/�N/(�.����ĽQE�/��<�}���m��\K : �T�r���Xj�x@'�1��Em�~c? �� ��9K
P�<^M:.Kb�����m�Gy���>�n�^����N
K�W���(��;�t���23gP��Sի��t��=V���s����}���E��'�K�z�Z� AN���zb%@�cZ�K[�PT��+hF>��Z5���u�;�'�y��\0 ���^�Q̡���m��k�ï;K�-�L�N$�~_�71V&�zաٱJW�]�]F�nlή�ԇ�C�S#(iLp^�� �ٿX�I@6��P�o,=�M|�W� ��iL��!9!ve[Ә�O�kA�Z��Р
���:�f���w
ch���e"N�x4�k�*,A �17�n�:��
.
(���g���#��R�L��P4X�{�=
�oH>:qǻ�[p~
�s$�V�J��3�j@�� ���Ӿfin+{y���2:Q�UR�}m�d߉��/
2�[ �q���_�P]Klr "3������,��g�>�+@7���v�����lo`�c��$����.[�bCHni�@_^�� = ���ܽtXF�g���T-�U��#v�� Z�>�;� ��6M� ���tӶΉ8�{N��a}��?d�ͦ&����
�rH�W�0:�>��y�6(�L$L)��
z�*CK��� k]��ye�@v��y�\Tf
6i�C���bn�@���,]��W̓Q ���Ęl���6/�p�� �T���͍[H1L/+*�G� *�ۨ�1-��]�ڦA��B�:k�����V��K�B�mb�OxD1IC�֮�H��'��2gt�,�O��w:�iR2QV�i|_��C s��s�Z|�_��-��dv��/� �N<����o�m~�#���g��ы�)! nC�������o� Tf%����de��XHe���j��J�JW�� \���gwA���6 ���� x�ג���*z ����#� ���!vC�@&$@=�^�)�\�X�@6�����\.�I;�Ж��+[��G�j��D����ļ �QA���UA!7g2�h
�1��ڂD��:����P��?��DPx�S�J ^:3�n�`;�4�f� �hl�5���鎩F��y�7��<�ɔ�F����&t>{A
� �Ml��I�o�ZWE�u�Pߝ�L�Cd +�f�0�(����Q����kuѼw��ד<4�1���i���n �X�5TcV�ʅT1JI��j��'a y&�����Q�ӔNx����qO�6�rĪ����.��ܱ<�4�z�d2DI�W�������n�7]7`0���(Su��hF��A��[Rp�RWm�MӰ�̞ie8S��Ѝm����嶃 ���K�JA^�W�<��"�� )|�q����֮����U�e����6'4lYǵ�,�#���l�ux�� �5ƌ�Y�u�g��+%���f]،A��N|`����� 怇D�@�4� f����̺w:ظ��-IA���St3_ͷ|K�Q� ���*-\ �^�E|o�c!{u��L�>��Q���[�n�W�`=����fEP��ok_����F�V*` �40�0�R*��^��a���k�-]<M���IDz��eH"�+�A�B.��t� �n��Ok Y\)� k�s �V�P��2Ī�īP�H����������0�G|pRť�mn9 �
ВU/r9�$�3���i����1BMЋ�LN�#�r"��ޔ�ҔXkl��E'�O�6@����v�~ �pk����;]}�z� �=��us�, �|��;W���=|�Y>���[�&$��A�jn���- -J~�co��k�*�Ǹߍ���J7=�����/%����L����� ������wɥ�V<p ���`8!�)t
֠��#V���˺��;
�ۼce ���+Y$�sj_ ,�*���$ \B���͘���L��_E�͐@9�1G6���pq��3QTZ\B��I����n�Ma!�Ddjx���U�} �0�*�� ��(\��R8��.��6 @�)�%��cq{�y yg ��&T�=�Uq�&D�쐆e �[x���
ABUs�d�L���>�}3�(z�s11 ��?Q�<3�C���kn�G3��=1�q�p�]{bGj��{���a6��YJL�*w�ЎX�E�eM0�:.{E$Ҋ� ��0� �" R���MT/�T]�*�Jp��|�z)C��L~0�#�� �>�v�'oG��s�E�h<#x�D� `~�q �C#�9��'357Ի_�C� ԫ� ���N�'( �ެ8̌rW� x( ����}0�����R��f��!�S�f��'KV�����E? i�D�jc�'� ����^�!,��J˶�>�d� x�x��FW�S9]�� '�i��4 Mꩊ�(thXK�Kʯf�/}|��=�{�Y=��Ow�}�LYM=���y 3�=�xY��&����]�ʠ ��}� f�9�Y��ݞuv��ٴ`��ajr
.� �{�잟���"�u\Cyt20x�%5�("}r��Z|��w<�����r--/����SKGn
�V��N�lM�����<�ϡ! �)9 �P�z�Y ��)�80���#�T���g�n
s�;e� � v��
��vMXvFu)�
D���Wb�sɉ �� �ngq6U� )�M�d �"��"1�`�����3]+ �mz1��%�B,.A�b��a�G���" '���.�[�M��l���������g[��o��@g�2�E�?�^��I�R�YD!~��X�����i�6:VgنZ%������ť_� ���:�t�A�U����Fp/��V��@�5�.F�P}�H�Ǝ��vR@
��n�u��"frJʦ� 3U���# �#0+��S$���(�ܦC5�s�\�0y9 x@�<��� ���gp���D�f�6�88�
Pҟ��m�B�݀�J�:X��F#s��!}��x��=iu^-I�@ K��;*�v�J1�M #oP�APU@kE����V�){� F�_ �+^$�F�W�� ��aN(��L<] ���d���..q�5F�)�
* 4^���j"��9�u�~�u�@���/�dU$���8d}ЕxNt����a\��Z��� J��X�R��f{���O
{X;$��$r��W�� ��+��& !t�>�QˠK�C�`:;�p �s8&�y��R b��#Y �'(��+#)|����� 0r��?$'mF�݅���>1r�?��� gN �T�ˆx����Q��OH�f��r�Ex�U�>/���ɢ�c�Ei��y��h��J�\6��a2+7�j����aF��}rE`O4�&��#�0147z�>��t���� �y^5?� �� p�~l~op�ӳȰ6E�o4�=LD�?bk��wtzD�+;�l+hg� �&�#LG�jÑH���,wt2�f� <J�Q�ioJ!� � ��% @� �"�a�O�Eg�/w�9 puJp�φ�E:� ��n`�9k�"��U=a��+ҥ
��ý�Ҷc�sP�)�A �Z���7�/�`�&¡B�H͡�]�#`��f�`�Ć�m�钋� J,N��XRG�ExT���1Ʊ��O
fLiU��q�]cPE�
>��T�IUσw���"��� |���^�P������P�BkU�V�RS0��n����l��yo�7za>��S�!@ZA4sk�y��M�%7�) �$E�zً�gi� �I��}�ȗu�mSO�Ѣ ?'��_D���</}�� �L2g��q'y.�#���s�\������
%c̵Ƥ+N�]���w�yhauA��L�ed ��
^bʚ ���8�6ٓ�e �D��n�i).�2 ������^E ` V�<�q��;>#AA�lj��:�Y֊i<������P�q1�%ui�9&����d)��WyT%xak����Bm�������%}m�n\�ԫ4��;[�����c�A�����=6f� �Қ
z>+$Y j|�LcC��'G�d^��Yb/+�K `��q�hR��6���O}/o�UQ�"����a_7�p���ƟvZ��Nl�wu1�XV�l��"p5�<O����6�G_����:R�(١RN�y�����YFϿ��XL�����X hdtw��lGD2� �ܺ��Ѩ���K*Կ� 9q �j��,��XUh���E��S�����UTr��U��1k
�ǵ����'\���©�֟V9���� H�e����KL i���˸ S�-��<0���s����?H�*�E�?��V�q���f{ݕSK֕'�F�U��'ݗX�E��C�!j�4!�`K�)A��̠!�~}�x%�Vh��j � � co& ~�U�$����pG� '�@J����p�8���-ܯ.s��za��rkD����ke0F�5_ �~zU�X 6fNFˌ�k�X�a��'v4-�lKr�x����F��q�U19���dd�
;��ܵg
x�9�-C B7i�$kIE?J�Q���� ��lfq�OxР@
����Ή[�
ې+\Q� �*��0����� C�Y`�ėv���G�fcM zF�p ��� Z�m"Ќ|ddn6�R��tr�� &\C���#'����� `�h�a�3��qL�I�}TRsF�bNZ �`D� ��H^ B�E�����B S �� �}#;����U�m���@�ǻ�p�fP2ޒ>�m���K�<�1,6['�<4���'%)"�����$`�" v!w?��� �a%Lpw˺��o`�������U4���p�B@�5~'m��� �FE�Lcm��j#�J�dNg C �s|���0��0J���*�+�T��304�˨2D�$�1�6�U�,:Āƍ�u�GFXH4]E�[D�����.t �|j�GS �aa�:��]�A�>�a378Sȇjk��� v,� n�-��h���@���Iy`gƢ� ��HU�e�J�Av������H3��s� �4���H�� \ =oM�B�� D� �>�į6�t�qNs��;g \��
��� 2�$|9�,���-φi~sZ-j��P�e�IZVC�� 3U��:L���u
ܒx T-�b�-�D9���]{���P�� �7��� Z�)�P��D�7��%m#��pX�_<Iٷ �iI'8�{��աL�2"Q.f(�3$8t���d�>���A�6餱�8�O�8�� �y��?����J/w
�
��
���#�������aK�����^L�H~��L��ʖV�ԑFArtQ�j~��K_�H@�C$�ҦPtb�D�W�9M�SJ�[���N�z����C���Q\ː�������h�D� �1[�~�ٮ�ymR�,1 �� �� ݈�q���>PC�����F���M�A�"������ 6B�h���d��'��P ��Ϛ�6��~�����������OH'y����m�42�- �^$����ڋ ��6N��ɑ3��5��l������*-���b�%����-� ���v� �lT�3a��=�<�/�R�ШNe�J���B� ���7�i�-�m�H�j���*ƣ O��0��|m�0
�5���o^X�(�|��L�d��o��q�+W8D���� �-:������愼?�P�ެ��%
*u9�c�@��"�^�r~�\�hA����x�͢'�
�$g�^� �Q��R46�p�b��{h{
�)�r���;8�� ��+��5|sh4P X/�E�sf�,���?}��� l�W
1�eb���]0X,�ɓ����xe���)�m��g�bSx8�6I}u�PNm�>T��/�o����uMSL���ĝ��`�0�>�X�ŗ�,�J�JD�ʅX�D���N(�V����nA�y��*�_�Q�ś��hؑON](삂Ӳ��+ZmYH�z�屻��eF)�IMb�"�-�Q������F6c�A���b �h�\�"��RbV��Wm4�����x��ճ m'2P|�$��p�4 o��� 1zG㔕@�*�6WD:
�C��=--Z�/�j�'� k:�ܷ�>{�b�GT�x�הB��J��������nb���]c�D_p�B�1:� ��w�ov��� ;�4��ӂ�E>�NA�����YQ��¨�ܶ��)s� ��P���E�0��"s�
*I�>#k�([Z�%�)B��Y�����y�p
y{��,&�,�3e�5�^�wzT����,�{ooю������"ط#�3]����k��{OO��(�8���S���b���%
gB�ȱ�9@~�4D���?��s$'얞�5���fa� ��� w^��8͂̄(1�MWI-S����t#�H�!�����6���p���9�����>̫h*� �m����ᔹ�9����x���高�8���<�|ܫ2��o� ��/3wᰙ��2 ʍsj���<��
n�`@n�1�BpD�R��L�w8��oRȾI ���5�S���!N~���e��# ������Y��q��<L�hC �HT���M�͈z�� c�ڶ�o���F�U�|D��/�¾C=��������e�u�!�tFعV�5� ��'�E�y��e�"�ܮ�f_ÁgZ `>ꚙ�>�������)@ y�o!Y�� � �Mv�& .?�
��|�C˃�bm��=�.i�YT���QO��;@ ��h]GL�%������a� �i8CkX^�M@a�{,7i3`��@�.x��.��@'���CCQP)��؏�ߡ狭/W���%P�ܓ�j��"{$?lUb���`�0
k ).�⽜�G�� �o��~ݑЄ�Ӷ��ْ���� =�HE/M�L�}m�3�� dp��#
��L�f� jf�
��/����"%�̠,�#\֬�=�آ ~��^�m�[k��H���M�g�Q�҂"� @��,z6AV܆���[2�n�����j�9��#�.�X6�Y�����x;�K\*�{� d s� ���f�� �'�݉�i�����r�jy�m��1iΤh���S� �Uf�@����y鋁�fͷ����
>�]�
$|�o����уQ.�� � �� J2�����?��?����oCPb���+�`�����Kw�TO|�����XYJ�.K���1E�)1�Ib~5����`�cBV�ϋV��Ꮽ 5�@ +��.\>�!�0��k��2J!N '_�mʥ �^� � �J㼆�-�J�V�TԔv c�Kd�#CЀ&j��2�I���&d�]7Lp���+WO�@� � jx5[���y�7�E�FhA7��V��?� ��MT)һ����S����qP��%�g^�cH�i�k��[}/Y��$��Kco��L|�?�N��qj:��G?-�Il�r4��K�� ��ғ&� l�[��r Y�Z��RJ���½��!zvD��5W��?W���o�a�]$�� � v&�?z^��������r hK�HS�TK�V���$�U&�˷Kʰ����v�^sI�l�-%%�}8�j��WZB_��1��gm�m\{�\�B��F�� ���Z�*�#�O�?=Bm��|��{m���#� K�����{x�a�_]�W�d�����~)��� ��t�� i��Ӷ �º�
��D>C�<�O��ͩ��0Ww������0��� ���xݼ~LZ.��-�O�T"4��V� h2���V�(���݅h�$a��o^2T�T�8�! ��[���~&��9"�i(^Є�_o@W�� O��7�o8R�[�j�7����1�7l� �79� �T-�^�`j��b�Q��V�Q������/(�؉f��/?/ë0t :��J��[���t�wt�~����'�^mW��HX��"�W=�����v/8�y�
�Y֔qcۛ[' � Dz@�&�i �� Z���{;s{TH�i�;X�s�����!��H.�@pC��F���e��
>u ��������.#�b�R�〩�i/��>��z3 �!����!gA���fb~/�M&?J�2��4�M :��2R�]\X[D4�0�0g����iQ����� /�N��r<= ��CH��;Ȍ���#�<!�!����1$�%���Z�?���9� ����b+�Q<���"y^���ݫ��S�a!^3����4d� <�TN����Q}���u���~V ��Z \�
��]�R�TC�6�|:����~�Yr|��*���ñ~���e`v݀���#Q���r� B�6�C��.6@&��s��k���Gl�
On Mon, Oct 31, 2022 at 10:05:58AM +0000, John Thomson wrote: > On Mon, 31 Oct 2022, at 02:36, Feng Tang wrote: > >> > > >> > possibly relevant config options: > >> > grep -E '(SLUB|SLAB)' .config > >> > # SLAB allocator options > >> > # CONFIG_SLAB is not set > >> > CONFIG_SLUB=y > >> > CONFIG_SLAB_MERGE_DEFAULT=y > >> > # CONFIG_SLAB_FREELIST_RANDOM is not set > >> > # CONFIG_SLAB_FREELIST_HARDENED is not set > >> > # CONFIG_SLUB_STATS is not set > >> > CONFIG_SLUB_CPU_PARTIAL=y > >> > # end of SLAB allocator options > >> > # CONFIG_SLUB_DEBUG is not set > >> > >> Also not having CONFIG_SLUB_DEBUG enabled means most of the code the > >> patch/commit touches is not even active. > >> Could this be some miscompile or code layout change exposing some > >> different bug, hmm. > > Yes, it could be. > > >> Is it any different if you do enable CONFIG_SLUB_DEBUG ? > > No change > > >> Or change to CONFIG_SLAB? (that would be really weird if not) > > This boots fine > > > I haven't found any clue from the code either, and I compiled > > kernel with the config above and tested booting on an Alder-lake > > desktop and a QEMU, which boot fine. > > > > Could you provide the full kernel config and demsg (in compressed > > format if you think it's too big), so we can check more? > > Attached > > > Thanks, > > Feng > > vmlinux is bigger, and entry point is larger (0x8074081c vs 0x807407dc revert vs 0x8073fcbc), > so that may be it? Or not? > revert + SLUB_DEBUG + SLUB_DEBUG_ON is bigger still, but does successfully boot. > vmlinux entry point is 0x8074705c Thanks for prompt info! As I can't reproduce it locally yet, could you help try 3 tests separately: * change the O2/O3 compile option to O1 * try the attached 0001 patch (which cut part of commit) * try attached 0001+0002 patch Thanks! - Feng
On Mon, 31 Oct 2022, at 11:42, Feng Tang wrote:
> On Mon, Oct 31, 2022 at 10:05:58AM +0000, John Thomson wrote:
>> On Mon, 31 Oct 2022, at 02:36, Feng Tang wrote:
>> >> >
>> >> > possibly relevant config options:
>> >> > grep -E '(SLUB|SLAB)' .config
>> >> > # SLAB allocator options
>> >> > # CONFIG_SLAB is not set
>> >> > CONFIG_SLUB=y
>> >> > CONFIG_SLAB_MERGE_DEFAULT=y
>> >> > # CONFIG_SLAB_FREELIST_RANDOM is not set
>> >> > # CONFIG_SLAB_FREELIST_HARDENED is not set
>> >> > # CONFIG_SLUB_STATS is not set
>> >> > CONFIG_SLUB_CPU_PARTIAL=y
>> >> > # end of SLAB allocator options
>> >> > # CONFIG_SLUB_DEBUG is not set
>> >>
>> >> Also not having CONFIG_SLUB_DEBUG enabled means most of the code the
>> >> patch/commit touches is not even active.
>> >> Could this be some miscompile or code layout change exposing some
>> >> different bug, hmm.
>>
>> Yes, it could be.
>>
>> >> Is it any different if you do enable CONFIG_SLUB_DEBUG ?
>>
>> No change
>>
>> >> Or change to CONFIG_SLAB? (that would be really weird if not)
>>
>> This boots fine
>>
>> > I haven't found any clue from the code either, and I compiled
>> > kernel with the config above and tested booting on an Alder-lake
>> > desktop and a QEMU, which boot fine.
>> >
>> > Could you provide the full kernel config and demsg (in compressed
>> > format if you think it's too big), so we can check more?
>>
>> Attached
>>
>> > Thanks,
>> > Feng
>>
>> vmlinux is bigger, and entry point is larger (0x8074081c vs 0x807407dc revert vs 0x8073fcbc),
>> so that may be it? Or not?
>> revert + SLUB_DEBUG + SLUB_DEBUG_ON is bigger still, but does successfully boot.
>> vmlinux entry point is 0x8074705c
>
> Thanks for prompt info!
>
> As I can't reproduce it locally yet, could you help try 3 tests separately:
> * change the O2/O3 compile option to O1
> * try the attached 0001 patch (which cut part of commit)
> * try attached 0001+0002 patch
None of these changed my outcome.
I also tried compiling the same linux tree & config with the Bootlin toolchain
(mipsel-buildroot-linux-gnu-gcc.br_real (Buildroot 2021.11-4428-g6b6741b) 12.2.0)
with the same results.
I will look into finding or building a mips clang toolchain.
No JTAG capability to debug, sorry.
I get the same outcome with either the ZBOOT vmlinuz, or vmlinux
Same happening with 6.1-rc3
After some blind poking around changing (how much of the commit affected) mm/slub.c,
I may have got lucky. it appears as though this is all I need to boot:
(against 6.1-rc3), and with the Bootlin toolchain. Will test my other build system as well.
--- a/mm/slub.c
+++ b/mm/slub.c
@@ -3276,7 +3276,7 @@ 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, orig_size);
+ p = ___slab_alloc(s, gfpflags, node, addr, c, 0);
#ifdef CONFIG_PREEMPT_COUNT
slub_put_cpu_ptr(s->cpu_slab);
#endif
Would like to hear your thoughts, but I will keep digging.
>
> Thanks!
>
>
> - Feng
>
>
>
>
> Attachments:
> * 0001-reduced-slub-patch.patch
> * 0002-reorder-the-partial_context-initialization.patch
--
John Thomson
Hi Thomson, Thanks for testing! + mips maintainer and mail list. The original report is here https://lore.kernel.org/lkml/becf2ac3-2a90-4f3a-96d9-a70f67c66e4a@app.fastmail.com/ On Tue, Nov 01, 2022 at 12:18:19AM +0000, John Thomson wrote: > > As I can't reproduce it locally yet, could you help try 3 tests separately: > > * change the O2/O3 compile option to O1 > > * try the attached 0001 patch (which cut part of commit) > > * try attached 0001+0002 patch > > None of these changed my outcome. 0001 patch chops off most of the functional changes, and it still fails. I'm wondering if it is related to the function arguments' passing, I noticed this is a 32 bits MIPS platform (from your config) > I also tried compiling the same linux tree & config with the Bootlin toolchain > (mipsel-buildroot-linux-gnu-gcc.br_real (Buildroot 2021.11-4428-g6b6741b) 12.2.0) > with the same results. > I will look into finding or building a mips clang toolchain. Yes, Hyeonggon's suggestion makes sense, different compiler may show some difference. > No JTAG capability to debug, sorry. > > I get the same outcome with either the ZBOOT vmlinuz, or vmlinux > > Same happening with 6.1-rc3 > > > After some blind poking around changing (how much of the commit affected) mm/slub.c, > I may have got lucky. it appears as though this is all I need to boot: > (against 6.1-rc3), and with the Bootlin toolchain. Will test my other build system as well. > > --- a/mm/slub.c > +++ b/mm/slub.c > @@ -3276,7 +3276,7 @@ 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, orig_size); > + p = ___slab_alloc(s, gfpflags, node, addr, c, 0); ___slab_alloc()'s argument number has been changed from 5 to 6, and some others from 4 to 5. Thanks, Feng > #ifdef CONFIG_PREEMPT_COUNT > slub_put_cpu_ptr(s->cpu_slab); > #endif
On Tue, 1 Nov 2022, at 07:57, Feng Tang wrote:
> Hi Thomson,
>
> Thanks for testing!
>
> + mips maintainer and mail list. The original report is here
>
> https://lore.kernel.org/lkml/becf2ac3-2a90-4f3a-96d9-a70f67c66e4a@app.fastmail.com/
I am guessing my issue comes from __kmem_cache_alloc_lru accessing s->object_size when (kmem_cache) s is NULL?
If that is the case, this change is not to blame, it only exposes the issue?
I get the following dmesg (note very early NULL kmem_cache) with the below change atop v6.1-rc3:
transfer started ......................................... transfer ok, time=2.02s
setting up elf image... OK
jumping to kernel code
zimage at: 80B842A0 810B4EFC
Uncompressing Linux at load address 80001000
Copy device tree to address 80B80EE0
Now, booting the kernel...
[ 0.000000] Linux version 6.1.0-rc3+ (john@john) (mipsel-buildroot-linux-gnu-gcc.br_real (Buildroot 2021.11-4428-g6b6741b) 12.2.0, GNU ld (GNU Binutils) 2.39) #61 SMP Tue Nov 1 18:04:13 AEST 2022
[ 0.000000] slub: kmem_cache_alloc called with kmem_cache: 0x0
[ 0.000000] slub: __kmem_cache_alloc_lru called with kmem_cache: 0x0
[ 0.000000] SoC Type: MediaTek MT7621 ver:1 eco:3
[ 0.000000] printk: bootconsole [early0] enabled
[ 0.000000] CPU0 revision is: 0001992f (MIPS 1004Kc)
[ 0.000000] MIPS: machine is MikroTik RouterBOARD 760iGS
normal boot
diff --git a/mm/slub.c b/mm/slub.c
index 157527d7101b..10fcdf2520d2 100644
--- a/mm/slub.c
+++ b/mm/slub.c
@@ -3410,7 +3410,13 @@ static __always_inline
void *__kmem_cache_alloc_lru(struct kmem_cache *s, struct list_lru *lru,
gfp_t gfpflags)
{
- void *ret = slab_alloc(s, lru, gfpflags, _RET_IP_, s->object_size);
+ void *ret;
+ if (IS_ERR_OR_NULL(s)) {
+ pr_warn("slub: __kmem_cache_alloc_lru called with kmem_cache: %pSR\n", s);
+ ret = slab_alloc(s, lru, gfpflags, _RET_IP_, 0);
+ } else {
+ ret = slab_alloc(s, lru, gfpflags, _RET_IP_, s->object_size);
+ }
trace_kmem_cache_alloc(_RET_IP_, ret, s, gfpflags, NUMA_NO_NODE);
@@ -3419,6 +3425,8 @@ void *__kmem_cache_alloc_lru(struct kmem_cache *s, struct list_lru *lru,
void *kmem_cache_alloc(struct kmem_cache *s, gfp_t gfpflags)
{
+ if (IS_ERR_OR_NULL(s))
+ pr_warn("slub: kmem_cache_alloc called with kmem_cache: %pSR\n", s);
return __kmem_cache_alloc_lru(s, NULL, gfpflags);
}
EXPORT_SYMBOL(kmem_cache_alloc);
@@ -3426,6 +3434,8 @@ EXPORT_SYMBOL(kmem_cache_alloc);
void *kmem_cache_alloc_lru(struct kmem_cache *s, struct list_lru *lru,
gfp_t gfpflags)
{
+ if (IS_ERR_OR_NULL(s))
+ pr_warn("slub: __kmem_cache_alloc_lru called with kmem_cache: %pSR\n", s);
return __kmem_cache_alloc_lru(s, lru, gfpflags);
}
EXPORT_SYMBOL(kmem_cache_alloc_lru);
Any hints on where kmem_cache_alloc would be being called from this early?
I will start looking from /init/main.c around pr_notice("%s", linux_banner);
Thank you for your help.
Let me know if you want me to stop replying to this mm/slub debug memory wasting email,
and take this to a new email?
Cheers,
--
John Thomson
On Tue, Nov 01, 2022 at 09:20:21AM +0000, John Thomson wrote:
> On Tue, 1 Nov 2022, at 07:57, Feng Tang wrote:
> > Hi Thomson,
> >
> > Thanks for testing!
> >
> > + mips maintainer and mail list. The original report is here
> >
> > https://lore.kernel.org/lkml/becf2ac3-2a90-4f3a-96d9-a70f67c66e4a@app.fastmail.com/
>
> I am guessing my issue comes from __kmem_cache_alloc_lru accessing s->object_size when (kmem_cache) s is NULL?
> If that is the case, this change is not to blame, it only exposes the issue?
>
> I get the following dmesg (note very early NULL kmem_cache) with the below change atop v6.1-rc3:
>
> transfer started ......................................... transfer ok, time=2.02s
> setting up elf image... OK
> jumping to kernel code
> zimage at: 80B842A0 810B4EFC
>
> Uncompressing Linux at load address 80001000
>
> Copy device tree to address 80B80EE0
>
> Now, booting the kernel...
>
> [ 0.000000] Linux version 6.1.0-rc3+ (john@john) (mipsel-buildroot-linux-gnu-gcc.br_real (Buildroot 2021.11-4428-g6b6741b) 12.2.0, GNU ld (GNU Binutils) 2.39) #61 SMP Tue Nov 1 18:04:13 AEST 2022
> [ 0.000000] slub: kmem_cache_alloc called with kmem_cache: 0x0
> [ 0.000000] slub: __kmem_cache_alloc_lru called with kmem_cache: 0x0
> [ 0.000000] SoC Type: MediaTek MT7621 ver:1 eco:3
> [ 0.000000] printk: bootconsole [early0] enabled
> [ 0.000000] CPU0 revision is: 0001992f (MIPS 1004Kc)
> [ 0.000000] MIPS: machine is MikroTik RouterBOARD 760iGS
>
> normal boot
>
>
> diff --git a/mm/slub.c b/mm/slub.c
> index 157527d7101b..10fcdf2520d2 100644
> --- a/mm/slub.c
> +++ b/mm/slub.c
> @@ -3410,7 +3410,13 @@ static __always_inline
> void *__kmem_cache_alloc_lru(struct kmem_cache *s, struct list_lru *lru,
> gfp_t gfpflags)
> {
> - void *ret = slab_alloc(s, lru, gfpflags, _RET_IP_, s->object_size);
> + void *ret;
> + if (IS_ERR_OR_NULL(s)) {
> + pr_warn("slub: __kmem_cache_alloc_lru called with kmem_cache: %pSR\n", s);
> + ret = slab_alloc(s, lru, gfpflags, _RET_IP_, 0);
> + } else {
> + ret = slab_alloc(s, lru, gfpflags, _RET_IP_, s->object_size);
> + }
>
> trace_kmem_cache_alloc(_RET_IP_, ret, s, gfpflags, NUMA_NO_NODE);
>
> @@ -3419,6 +3425,8 @@ void *__kmem_cache_alloc_lru(struct kmem_cache *s, struct list_lru *lru,
>
> void *kmem_cache_alloc(struct kmem_cache *s, gfp_t gfpflags)
> {
> + if (IS_ERR_OR_NULL(s))
> + pr_warn("slub: kmem_cache_alloc called with kmem_cache: %pSR\n", s);
> return __kmem_cache_alloc_lru(s, NULL, gfpflags);
> }
> EXPORT_SYMBOL(kmem_cache_alloc);
> @@ -3426,6 +3434,8 @@ EXPORT_SYMBOL(kmem_cache_alloc);
> void *kmem_cache_alloc_lru(struct kmem_cache *s, struct list_lru *lru,
> gfp_t gfpflags)
> {
> + if (IS_ERR_OR_NULL(s))
> + pr_warn("slub: __kmem_cache_alloc_lru called with kmem_cache: %pSR\n", s);
> return __kmem_cache_alloc_lru(s, lru, gfpflags);
> }
> EXPORT_SYMBOL(kmem_cache_alloc_lru);
>
>
> Any hints on where kmem_cache_alloc would be being called from this early?
> I will start looking from /init/main.c around pr_notice("%s", linux_banner);
Great. Would you try calling dump_stack(); when we observed s == NULL?
That would give more information about who passed s == NULL to these
functions.
--
Thanks,
Hyeonggon
On Tue, 1 Nov 2022, at 09:31, Hyeonggon Yoo wrote:
> On Tue, Nov 01, 2022 at 09:20:21AM +0000, John Thomson wrote:
>> On Tue, 1 Nov 2022, at 07:57, Feng Tang wrote:
>> > Hi Thomson,
>> >
>> > Thanks for testing!
>> >
>> > + mips maintainer and mail list. The original report is here
>> >
>> > https://lore.kernel.org/lkml/becf2ac3-2a90-4f3a-96d9-a70f67c66e4a@app.fastmail.com/
>>
>> I am guessing my issue comes from __kmem_cache_alloc_lru accessing s->object_size when (kmem_cache) s is NULL?
>> If that is the case, this change is not to blame, it only exposes the issue?
>>
>> I get the following dmesg (note very early NULL kmem_cache) with the below change atop v6.1-rc3:
>>
>> transfer started ......................................... transfer ok, time=2.02s
>> setting up elf image... OK
>> jumping to kernel code
>> zimage at: 80B842A0 810B4EFC
>>
>> Uncompressing Linux at load address 80001000
>>
>> Copy device tree to address 80B80EE0
>>
>> Now, booting the kernel...
>>
>> [ 0.000000] Linux version 6.1.0-rc3+ (john@john) (mipsel-buildroot-linux-gnu-gcc.br_real (Buildroot 2021.11-4428-g6b6741b) 12.2.0, GNU ld (GNU Binutils) 2.39) #61 SMP Tue Nov 1 18:04:13 AEST 2022
>> [ 0.000000] slub: kmem_cache_alloc called with kmem_cache: 0x0
>> [ 0.000000] slub: __kmem_cache_alloc_lru called with kmem_cache: 0x0
>> [ 0.000000] SoC Type: MediaTek MT7621 ver:1 eco:3
>> [ 0.000000] printk: bootconsole [early0] enabled
>> [ 0.000000] CPU0 revision is: 0001992f (MIPS 1004Kc)
>> [ 0.000000] MIPS: machine is MikroTik RouterBOARD 760iGS
>>
>> normal boot
>>
>>
>> diff --git a/mm/slub.c b/mm/slub.c
>> index 157527d7101b..10fcdf2520d2 100644
>> --- a/mm/slub.c
>> +++ b/mm/slub.c
>> @@ -3410,7 +3410,13 @@ static __always_inline
>> void *__kmem_cache_alloc_lru(struct kmem_cache *s, struct list_lru *lru,
>> gfp_t gfpflags)
>> {
>> - void *ret = slab_alloc(s, lru, gfpflags, _RET_IP_, s->object_size);
>> + void *ret;
>> + if (IS_ERR_OR_NULL(s)) {
>> + pr_warn("slub: __kmem_cache_alloc_lru called with kmem_cache: %pSR\n", s);
>> + ret = slab_alloc(s, lru, gfpflags, _RET_IP_, 0);
>> + } else {
>> + ret = slab_alloc(s, lru, gfpflags, _RET_IP_, s->object_size);
>> + }
>>
>> trace_kmem_cache_alloc(_RET_IP_, ret, s, gfpflags, NUMA_NO_NODE);
>>
>> @@ -3419,6 +3425,8 @@ void *__kmem_cache_alloc_lru(struct kmem_cache *s, struct list_lru *lru,
>>
>> void *kmem_cache_alloc(struct kmem_cache *s, gfp_t gfpflags)
>> {
>> + if (IS_ERR_OR_NULL(s))
>> + pr_warn("slub: kmem_cache_alloc called with kmem_cache: %pSR\n", s);
>> return __kmem_cache_alloc_lru(s, NULL, gfpflags);
>> }
>> EXPORT_SYMBOL(kmem_cache_alloc);
>> @@ -3426,6 +3434,8 @@ EXPORT_SYMBOL(kmem_cache_alloc);
>> void *kmem_cache_alloc_lru(struct kmem_cache *s, struct list_lru *lru,
>> gfp_t gfpflags)
>> {
>> + if (IS_ERR_OR_NULL(s))
>> + pr_warn("slub: __kmem_cache_alloc_lru called with kmem_cache: %pSR\n", s);
>> return __kmem_cache_alloc_lru(s, lru, gfpflags);
>> }
>> EXPORT_SYMBOL(kmem_cache_alloc_lru);
>>
>>
>> Any hints on where kmem_cache_alloc would be being called from this early?
>> I will start looking from /init/main.c around pr_notice("%s", linux_banner);
>
> Great. Would you try calling dump_stack(); when we observed s == NULL?
> That would give more information about who passed s == NULL to these
> functions.
>
With the dump_stack() in place:
Now, booting the kernel...
[ 0.000000] Linux version 6.1.0-rc3+ (john@john) (mipsel-buildroot-linux-gnu-gcc.br_real (Buildroot 2021.11-4428-g6b6741b) 12.2.0, GNU ld (GNU Binutils) 2.39) #62 SMP Tue Nov 1 19:49:52 AEST 2022
[ 0.000000] slub: __kmem_cache_alloc_lru called with kmem_cache ptr: 0x0
[ 0.000000] CPU: 0 PID: 0 Comm: swapper Not tainted 6.1.0-rc3+ #62
[ 0.000000] Stack : 810fff78 80084d98 80889d00 00000004 00000000 00000000 80889d5c 80c90000
[ 0.000000] 80920000 807bd380 8089d368 80923bd3 00000000 00000001 80889d08 00000000
[ 0.000000] 00000000 00000000 807bd380 8084bd51 00000002 00000002 00000001 6d6f4320
[ 0.000000] 00000000 80c97ce9 80c97d14 fffffffc 807bd380 00000000 00000003 00000dc0
[ 0.000000] 00000000 a0000000 80910000 8110a0b4 00000000 00000020 80010000 80010000
[ 0.000000] ...
[ 0.000000] Call Trace:
[ 0.000000] [<80008260>] show_stack+0x28/0xf0
[ 0.000000] [<8070cdc0>] dump_stack_lvl+0x60/0x80
[ 0.000000] [<801c1428>] kmem_cache_alloc+0x5c0/0x740
[ 0.000000] [<8092856c>] prom_soc_init+0x1fc/0x2b4
[ 0.000000] [<80928060>] prom_init+0x44/0xf0
[ 0.000000] [<80929214>] setup_arch+0x4c/0x6a8
[ 0.000000] [<809257e0>] start_kernel+0x88/0x7c0
[ 0.000000]
[ 0.000000] SoC Type: MediaTek MT7621 ver:1 eco:3
Now, booting the kernel...
[ 0.000000] Linux version 6.1.0-rc3+ (john@john) (mipsel-buildroot-linux-gnu-gcc.br_real (Buildroot 2021.11-4428-g6b6741b) 12.2.0, GNU ld (GNU Binutils) 2.39) #62 SMP Tue Nov 1 19:49:52 AEST 2022
[ 0.000000] slub: __kmem_cache_alloc_lru called with kmem_cache ptr: 0x0
[ 0.000000] CPU: 0 PID: 0 Comm: swapper Not tainted 6.1.0-rc3+ #62
[ 0.000000] Stack : 810fff78 80084d98 80889d00 00000004 00000000 00000000 80889d5c 80c90000
[ 0.000000] 80920000 807bd380 8089d368 80923bd3 00000000 00000001 80889d08 00000000
[ 0.000000] 00000000 00000000 807bd380 8084bd51 00000002 00000002 00000001 6d6f4320
[ 0.000000] 00000000 80c97ce9 80c97d14 fffffffc 807bd380 00000000 00000003 00000dc0
[ 0.000000] 00000000 a0000000 80910000 8110a0b4 00000000 00000020 80010000 80010000
[ 0.000000] ...
[ 0.000000] Call Trace:
[ 0.000000] show_stack (/mnt/pool_ssd/code/linux/linux-stable-mt7621/./arch/mips/include/asm/stacktrace.h:43 /mnt/pool_ssd/code/linux/linux-stable-mt7621/arch/mips/kernel/traps.c:223)
[ 0.000000] dump_stack_lvl (/mnt/pool_ssd/code/linux/linux-stable-mt7621/lib/dump_stack.c:107 (discriminator 1))
[ 0.000000] kmem_cache_alloc (/mnt/pool_ssd/code/linux/linux-stable-mt7621/mm/slub.c:3318 /mnt/pool_ssd/code/linux/linux-stable-mt7621/mm/slub.c:3406 /mnt/pool_ssd/code/linux/linux-stable-mt7621/mm/slub.c:3418 /mnt/pool_ssd/code/linux/linux-stable-mt7621/mm/slub.c:3430)
[ 0.000000] prom_soc_init (/mnt/pool_ssd/code/linux/linux-stable-mt7621/arch/mips/ralink/mt7621.c:106 /mnt/pool_ssd/code/linux/linux-stable-mt7621/arch/mips/ralink/mt7621.c:177)
[ 0.000000] prom_init (/mnt/pool_ssd/code/linux/linux-stable-mt7621/arch/mips/ralink/prom.c:64)
[ 0.000000] setup_arch (/mnt/pool_ssd/code/linux/linux-stable-mt7621/arch/mips/kernel/setup.c:786)
[ 0.000000] start_kernel (/mnt/pool_ssd/code/linux/linux-stable-mt7621/init/main.c:279 /mnt/pool_ssd/code/linux/linux-stable-mt7621/init/main.c:477 /mnt/pool_ssd/code/linux/linux-stable-mt7621/init/main.c:960)
[ 0.000000]
[ 0.000000] SoC Type: MediaTek MT7621 ver:1 eco:3
I have not found it yet.
Cheers,
--
John Thomson
On 11/1/22 11:33, John Thomson wrote:
> On Tue, 1 Nov 2022, at 09:31, Hyeonggon Yoo wrote:
>> On Tue, Nov 01, 2022 at 09:20:21AM +0000, John Thomson wrote:
>>> On Tue, 1 Nov 2022, at 07:57, Feng Tang wrote:
>>> > Hi Thomson,
>>> >
>>> > Thanks for testing!
>>> >
>>> > + mips maintainer and mail list. The original report is here
>>> >
>>> > https://lore.kernel.org/lkml/becf2ac3-2a90-4f3a-96d9-a70f67c66e4a@app.fastmail.com/
>>>
>>> I am guessing my issue comes from __kmem_cache_alloc_lru accessing s->object_size when (kmem_cache) s is NULL?
>>> If that is the case, this change is not to blame, it only exposes the issue?
>>>
>>> I get the following dmesg (note very early NULL kmem_cache) with the below change atop v6.1-rc3:
>>>
>>> transfer started ......................................... transfer ok, time=2.02s
>>> setting up elf image... OK
>>> jumping to kernel code
>>> zimage at: 80B842A0 810B4EFC
>>>
>>> Uncompressing Linux at load address 80001000
>>>
>>> Copy device tree to address 80B80EE0
>>>
>>> Now, booting the kernel...
>>>
>>> [ 0.000000] Linux version 6.1.0-rc3+ (john@john) (mipsel-buildroot-linux-gnu-gcc.br_real (Buildroot 2021.11-4428-g6b6741b) 12.2.0, GNU ld (GNU Binutils) 2.39) #61 SMP Tue Nov 1 18:04:13 AEST 2022
>>> [ 0.000000] slub: kmem_cache_alloc called with kmem_cache: 0x0
>>> [ 0.000000] slub: __kmem_cache_alloc_lru called with kmem_cache: 0x0
>>> [ 0.000000] SoC Type: MediaTek MT7621 ver:1 eco:3
>>> [ 0.000000] printk: bootconsole [early0] enabled
>>> [ 0.000000] CPU0 revision is: 0001992f (MIPS 1004Kc)
>>> [ 0.000000] MIPS: machine is MikroTik RouterBOARD 760iGS
>>>
>>> normal boot
>>>
>>>
>>> diff --git a/mm/slub.c b/mm/slub.c
>>> index 157527d7101b..10fcdf2520d2 100644
>>> --- a/mm/slub.c
>>> +++ b/mm/slub.c
>>> @@ -3410,7 +3410,13 @@ static __always_inline
>>> void *__kmem_cache_alloc_lru(struct kmem_cache *s, struct list_lru *lru,
>>> gfp_t gfpflags)
>>> {
>>> - void *ret = slab_alloc(s, lru, gfpflags, _RET_IP_, s->object_size);
>>> + void *ret;
>>> + if (IS_ERR_OR_NULL(s)) {
>>> + pr_warn("slub: __kmem_cache_alloc_lru called with kmem_cache: %pSR\n", s);
>>> + ret = slab_alloc(s, lru, gfpflags, _RET_IP_, 0);
>>> + } else {
>>> + ret = slab_alloc(s, lru, gfpflags, _RET_IP_, s->object_size);
>>> + }
>>>
>>> trace_kmem_cache_alloc(_RET_IP_, ret, s, gfpflags, NUMA_NO_NODE);
>>>
>>> @@ -3419,6 +3425,8 @@ void *__kmem_cache_alloc_lru(struct kmem_cache *s, struct list_lru *lru,
>>>
>>> void *kmem_cache_alloc(struct kmem_cache *s, gfp_t gfpflags)
>>> {
>>> + if (IS_ERR_OR_NULL(s))
>>> + pr_warn("slub: kmem_cache_alloc called with kmem_cache: %pSR\n", s);
>>> return __kmem_cache_alloc_lru(s, NULL, gfpflags);
>>> }
>>> EXPORT_SYMBOL(kmem_cache_alloc);
>>> @@ -3426,6 +3434,8 @@ EXPORT_SYMBOL(kmem_cache_alloc);
>>> void *kmem_cache_alloc_lru(struct kmem_cache *s, struct list_lru *lru,
>>> gfp_t gfpflags)
>>> {
>>> + if (IS_ERR_OR_NULL(s))
>>> + pr_warn("slub: __kmem_cache_alloc_lru called with kmem_cache: %pSR\n", s);
>>> return __kmem_cache_alloc_lru(s, lru, gfpflags);
>>> }
>>> EXPORT_SYMBOL(kmem_cache_alloc_lru);
>>>
>>>
>>> Any hints on where kmem_cache_alloc would be being called from this early?
>>> I will start looking from /init/main.c around pr_notice("%s", linux_banner);
>>
>> Great. Would you try calling dump_stack(); when we observed s == NULL?
>> That would give more information about who passed s == NULL to these
>> functions.
>>
>
> With the dump_stack() in place:
>
> Now, booting the kernel...
>
> [ 0.000000] Linux version 6.1.0-rc3+ (john@john) (mipsel-buildroot-linux-gnu-gcc.br_real (Buildroot 2021.11-4428-g6b6741b) 12.2.0, GNU ld (GNU Binutils) 2.39) #62 SMP Tue Nov 1 19:49:52 AEST 2022
> [ 0.000000] slub: __kmem_cache_alloc_lru called with kmem_cache ptr: 0x0
> [ 0.000000] CPU: 0 PID: 0 Comm: swapper Not tainted 6.1.0-rc3+ #62
> [ 0.000000] Stack : 810fff78 80084d98 80889d00 00000004 00000000 00000000 80889d5c 80c90000
> [ 0.000000] 80920000 807bd380 8089d368 80923bd3 00000000 00000001 80889d08 00000000
> [ 0.000000] 00000000 00000000 807bd380 8084bd51 00000002 00000002 00000001 6d6f4320
> [ 0.000000] 00000000 80c97ce9 80c97d14 fffffffc 807bd380 00000000 00000003 00000dc0
> [ 0.000000] 00000000 a0000000 80910000 8110a0b4 00000000 00000020 80010000 80010000
> [ 0.000000] ...
> [ 0.000000] Call Trace:
> [ 0.000000] [<80008260>] show_stack+0x28/0xf0
> [ 0.000000] [<8070cdc0>] dump_stack_lvl+0x60/0x80
> [ 0.000000] [<801c1428>] kmem_cache_alloc+0x5c0/0x740
> [ 0.000000] [<8092856c>] prom_soc_init+0x1fc/0x2b4
> [ 0.000000] [<80928060>] prom_init+0x44/0xf0
> [ 0.000000] [<80929214>] setup_arch+0x4c/0x6a8
> [ 0.000000] [<809257e0>] start_kernel+0x88/0x7c0
> [ 0.000000]
> [ 0.000000] SoC Type: MediaTek MT7621 ver:1 eco:3
The stack means CONFIG_TRACING=n, is that right?
That would mean
prom_soc_init()
soc_dev_init()
kzalloc() -> kmalloc()
kmalloc_trace() // after #else /* CONFIG_TRACING */
kmem_cache_alloc(s, flags);
Looks like this path is a small bug in the wasting detection patch, as we
throw away size there.
AFAICS before this patch, we "survive" "kmem_cache *s" being NULL as
slab_pre_alloc_hook() will happen to return NULL and we bail out from
slab_alloc_node(). But this is a side-effect, not an intended protection.
Also the CONFIG_TRACING variant of kmalloc_trace() would have called
trace_kmalloc dereferencing s->size anyway even before this patch.
I don't think we should add WARNS in the slab hot paths just to prevent this
rare error of using slab too early. At most VM_WARN... would be acceptable
but still not necessary as crashing immediately from a NULL pointer is
sufficient.
So IMHO mips should fix their soc init, and we should look into the
CONFIG_TRACING=n variant of kmalloc_trace(), to pass orig_size properly.
On Wed, Nov 02, 2022 at 04:22:37PM +0800, Vlastimil Babka wrote:
> On 11/1/22 11:33, John Thomson wrote:
[...]
> >
> > [ 0.000000] Linux version 6.1.0-rc3+ (john@john) (mipsel-buildroot-linux-gnu-gcc.br_real (Buildroot 2021.11-4428-g6b6741b) 12.2.0, GNU ld (GNU Binutils) 2.39) #62 SMP Tue Nov 1 19:49:52 AEST 2022
> > [ 0.000000] slub: __kmem_cache_alloc_lru called with kmem_cache ptr: 0x0
> > [ 0.000000] CPU: 0 PID: 0 Comm: swapper Not tainted 6.1.0-rc3+ #62
> > [ 0.000000] Stack : 810fff78 80084d98 80889d00 00000004 00000000 00000000 80889d5c 80c90000
> > [ 0.000000] 80920000 807bd380 8089d368 80923bd3 00000000 00000001 80889d08 00000000
> > [ 0.000000] 00000000 00000000 807bd380 8084bd51 00000002 00000002 00000001 6d6f4320
> > [ 0.000000] 00000000 80c97ce9 80c97d14 fffffffc 807bd380 00000000 00000003 00000dc0
> > [ 0.000000] 00000000 a0000000 80910000 8110a0b4 00000000 00000020 80010000 80010000
> > [ 0.000000] ...
> > [ 0.000000] Call Trace:
> > [ 0.000000] [<80008260>] show_stack+0x28/0xf0
> > [ 0.000000] [<8070cdc0>] dump_stack_lvl+0x60/0x80
> > [ 0.000000] [<801c1428>] kmem_cache_alloc+0x5c0/0x740
> > [ 0.000000] [<8092856c>] prom_soc_init+0x1fc/0x2b4
> > [ 0.000000] [<80928060>] prom_init+0x44/0xf0
> > [ 0.000000] [<80929214>] setup_arch+0x4c/0x6a8
> > [ 0.000000] [<809257e0>] start_kernel+0x88/0x7c0
> > [ 0.000000]
> > [ 0.000000] SoC Type: MediaTek MT7621 ver:1 eco:3
>
> The stack means CONFIG_TRACING=n, is that right?
Yes, from the kconfig, CONFIG_TRACING is not set.
> That would mean
> prom_soc_init()
> soc_dev_init()
> kzalloc() -> kmalloc()
> kmalloc_trace() // after #else /* CONFIG_TRACING */
> kmem_cache_alloc(s, flags);
>
> Looks like this path is a small bug in the wasting detection patch, as we
> throw away size there.
Yes, from the code reading and log from John, it is.
One strange thing is, I reset the code to v6.0, and found that
__kmem_cache_alloc_lru() also access the 's->object_size'
void *__kmem_cache_alloc_lru(struct kmem_cache *s, struct list_lru *lru,
gfp_t gfpflags)
{
void *ret = slab_alloc(s, lru, gfpflags, _RET_IP_, s->object_size);
...
}
And from John's dump_stack() info, this call is also where the NULL pointer
happens, which I still can't figue out.
> AFAICS before this patch, we "survive" "kmem_cache *s" being NULL as
> slab_pre_alloc_hook() will happen to return NULL and we bail out from
> slab_alloc_node(). But this is a side-effect, not an intended protection.
> Also the CONFIG_TRACING variant of kmalloc_trace() would have called
> trace_kmalloc dereferencing s->size anyway even before this patch.
>
> I don't think we should add WARNS in the slab hot paths just to prevent this
> rare error of using slab too early. At most VM_WARN... would be acceptable
> but still not necessary as crashing immediately from a NULL pointer is
> sufficient.
>
> So IMHO mips should fix their soc init,
Yes, for the mips fix, John has proposed to defer the calling of prom_soc_init(),
which looks reasonable.
> and we should look into the
> CONFIG_TRACING=n variant of kmalloc_trace(), to pass orig_size properly.
You mean check if the pointer is NULL and bail out early.
Thanks,
Feng
On 11/3/22 06:54, Feng Tang wrote:
> On Wed, Nov 02, 2022 at 04:22:37PM +0800, Vlastimil Babka wrote:
>> On 11/1/22 11:33, John Thomson wrote:
> [...]
>> >
>> > [ 0.000000] Linux version 6.1.0-rc3+ (john@john) (mipsel-buildroot-linux-gnu-gcc.br_real (Buildroot 2021.11-4428-g6b6741b) 12.2.0, GNU ld (GNU Binutils) 2.39) #62 SMP Tue Nov 1 19:49:52 AEST 2022
>> > [ 0.000000] slub: __kmem_cache_alloc_lru called with kmem_cache ptr: 0x0
>> > [ 0.000000] CPU: 0 PID: 0 Comm: swapper Not tainted 6.1.0-rc3+ #62
>> > [ 0.000000] Stack : 810fff78 80084d98 80889d00 00000004 00000000 00000000 80889d5c 80c90000
>> > [ 0.000000] 80920000 807bd380 8089d368 80923bd3 00000000 00000001 80889d08 00000000
>> > [ 0.000000] 00000000 00000000 807bd380 8084bd51 00000002 00000002 00000001 6d6f4320
>> > [ 0.000000] 00000000 80c97ce9 80c97d14 fffffffc 807bd380 00000000 00000003 00000dc0
>> > [ 0.000000] 00000000 a0000000 80910000 8110a0b4 00000000 00000020 80010000 80010000
>> > [ 0.000000] ...
>> > [ 0.000000] Call Trace:
>> > [ 0.000000] [<80008260>] show_stack+0x28/0xf0
>> > [ 0.000000] [<8070cdc0>] dump_stack_lvl+0x60/0x80
>> > [ 0.000000] [<801c1428>] kmem_cache_alloc+0x5c0/0x740
>> > [ 0.000000] [<8092856c>] prom_soc_init+0x1fc/0x2b4
>> > [ 0.000000] [<80928060>] prom_init+0x44/0xf0
>> > [ 0.000000] [<80929214>] setup_arch+0x4c/0x6a8
>> > [ 0.000000] [<809257e0>] start_kernel+0x88/0x7c0
>> > [ 0.000000]
>> > [ 0.000000] SoC Type: MediaTek MT7621 ver:1 eco:3
>>
>> The stack means CONFIG_TRACING=n, is that right?
>
> Yes, from the kconfig, CONFIG_TRACING is not set.
>
>> That would mean
>> prom_soc_init()
>> soc_dev_init()
>> kzalloc() -> kmalloc()
>> kmalloc_trace() // after #else /* CONFIG_TRACING */
>> kmem_cache_alloc(s, flags);
>>
>> Looks like this path is a small bug in the wasting detection patch, as we
>> throw away size there.
>
> Yes, from the code reading and log from John, it is.
>
> One strange thing is, I reset the code to v6.0, and found that
> __kmem_cache_alloc_lru() also access the 's->object_size'
>
> void *__kmem_cache_alloc_lru(struct kmem_cache *s, struct list_lru *lru,
> gfp_t gfpflags)
> {
> void *ret = slab_alloc(s, lru, gfpflags, _RET_IP_, s->object_size);
> ...
> }
>
> And from John's dump_stack() info, this call is also where the NULL pointer
> happens, which I still can't figue out.
>
>> AFAICS before this patch, we "survive" "kmem_cache *s" being NULL as
>> slab_pre_alloc_hook() will happen to return NULL and we bail out from
>> slab_alloc_node(). But this is a side-effect, not an intended protection.
>> Also the CONFIG_TRACING variant of kmalloc_trace() would have called
>> trace_kmalloc dereferencing s->size anyway even before this patch.
>>
>> I don't think we should add WARNS in the slab hot paths just to prevent this
>> rare error of using slab too early. At most VM_WARN... would be acceptable
>> but still not necessary as crashing immediately from a NULL pointer is
>> sufficient.
>>
>> So IMHO mips should fix their soc init,
>
> Yes, for the mips fix, John has proposed to defer the calling of prom_soc_init(),
> which looks reasonable.
>
>> and we should look into the
>> CONFIG_TRACING=n variant of kmalloc_trace(), to pass orig_size properly.
>
> You mean check if the pointer is NULL and bail out early.
No I mean here:
#else /* CONFIG_TRACING */
/* Save a function call when CONFIG_TRACING=n */
static __always_inline __alloc_size(3)
void *kmalloc_trace(struct kmem_cache *s, gfp_t flags, size_t size)
{
void *ret = kmem_cache_alloc(s, flags);
ret = kasan_kmalloc(s, ret, size, flags);
return ret;
}
we call kmem_cache_alloc() and discard the size parameter, so it will assume
s->object_size (and as the side-effect, crash if s is NULL). We shouldn't
add "s is NULL?" checks, but fix passing the size - probably switch to
__kmem_cache_alloc_node()? and in the following kmalloc_node_trace() analogically.
On Thu, Nov 03, 2022 at 09:33:28AM +0100, Vlastimil Babka wrote:
[...]
> >> AFAICS before this patch, we "survive" "kmem_cache *s" being NULL as
> >> slab_pre_alloc_hook() will happen to return NULL and we bail out from
> >> slab_alloc_node(). But this is a side-effect, not an intended protection.
> >> Also the CONFIG_TRACING variant of kmalloc_trace() would have called
> >> trace_kmalloc dereferencing s->size anyway even before this patch.
> >>
> >> I don't think we should add WARNS in the slab hot paths just to prevent this
> >> rare error of using slab too early. At most VM_WARN... would be acceptable
> >> but still not necessary as crashing immediately from a NULL pointer is
> >> sufficient.
> >>
> >> So IMHO mips should fix their soc init,
> >
> > Yes, for the mips fix, John has proposed to defer the calling of prom_soc_init(),
> > which looks reasonable.
> >
> >> and we should look into the
> >> CONFIG_TRACING=n variant of kmalloc_trace(), to pass orig_size properly.
> >
> > You mean check if the pointer is NULL and bail out early.
>
> No I mean here:
>
> #else /* CONFIG_TRACING */
> /* Save a function call when CONFIG_TRACING=n */
> static __always_inline __alloc_size(3)
> void *kmalloc_trace(struct kmem_cache *s, gfp_t flags, size_t size)
> {
> void *ret = kmem_cache_alloc(s, flags);
>
> ret = kasan_kmalloc(s, ret, size, flags);
> return ret;
> }
>
> we call kmem_cache_alloc() and discard the size parameter, so it will assume
> s->object_size (and as the side-effect, crash if s is NULL). We shouldn't
> add "s is NULL?" checks, but fix passing the size - probably switch to
> __kmem_cache_alloc_node()? and in the following kmalloc_node_trace() analogically.
Got it, thanks! I might have missed it during some rebasing for the
kmalloc wastage debug patch.
How about the following fix?
Thanks,
Feng
---
From 9f9fa9da8946fd44625f873c0f51167357075be1 Mon Sep 17 00:00:00 2001
From: Feng Tang <feng.tang@intel.com>
Date: Thu, 3 Nov 2022 21:32:10 +0800
Subject: [PATCH] mm/slub: Add missing orig_size parameter for wastage debug
commit 6edf2576a6cc ("mm/slub: enable debugging memory wasting of
kmalloc") was introduced for debugging kmalloc memory wastage,
and it missed to pass the original request size for kmalloc_trace()
and kmalloc_node_trace() in CONFIG_TRACING=n path.
Fix it by using __kmem_cache_alloc_node() with correct original
request size.
Fixes: 6edf2576a6cc ("mm/slub: enable debugging memory wasting of kmalloc")
Suggested-by: Vlastimil Babka <vbabka@suse.cz>
Signed-off-by: Feng Tang <feng.tang@intel.com>
---
include/linux/slab.h | 9 +++++++--
1 file changed, 7 insertions(+), 2 deletions(-)
diff --git a/include/linux/slab.h b/include/linux/slab.h
index 90877fcde70b..9691afa569e1 100644
--- a/include/linux/slab.h
+++ b/include/linux/slab.h
@@ -469,6 +469,9 @@ void *__kmalloc_node(size_t size, gfp_t flags, int node) __assume_kmalloc_alignm
__alloc_size(1);
void *kmem_cache_alloc_node(struct kmem_cache *s, gfp_t flags, int node) __assume_slab_alignment
__malloc;
+void *__kmem_cache_alloc_node(struct kmem_cache *s, gfp_t flags, int node,
+ size_t orig_size, unsigned long caller) __assume_slab_alignment
+ __malloc;
#ifdef CONFIG_TRACING
void *kmalloc_trace(struct kmem_cache *s, gfp_t flags, size_t size)
@@ -482,7 +485,8 @@ void *kmalloc_node_trace(struct kmem_cache *s, gfp_t gfpflags,
static __always_inline __alloc_size(3)
void *kmalloc_trace(struct kmem_cache *s, gfp_t flags, size_t size)
{
- void *ret = kmem_cache_alloc(s, flags);
+ void *ret = __kmem_cache_alloc_node(s, flags, NUMA_NO_NODE,
+ size, _RET_IP_);
ret = kasan_kmalloc(s, ret, size, flags);
return ret;
@@ -492,7 +496,8 @@ static __always_inline __alloc_size(4)
void *kmalloc_node_trace(struct kmem_cache *s, gfp_t gfpflags,
int node, size_t size)
{
- void *ret = kmem_cache_alloc_node(s, gfpflags, node);
+ void *ret = __kmem_cache_alloc_node(s, gfpflags, node,
+ size, _RET_IP_);
ret = kasan_kmalloc(s, ret, size, gfpflags);
return ret;
--
2.34.1
On Thu, Nov 03, 2022 at 10:16:12PM +0800, Feng Tang wrote:
> On Thu, Nov 03, 2022 at 09:33:28AM +0100, Vlastimil Babka wrote:
> [...]
> > >> AFAICS before this patch, we "survive" "kmem_cache *s" being NULL as
> > >> slab_pre_alloc_hook() will happen to return NULL and we bail out from
> > >> slab_alloc_node(). But this is a side-effect, not an intended protection.
> > >> Also the CONFIG_TRACING variant of kmalloc_trace() would have called
> > >> trace_kmalloc dereferencing s->size anyway even before this patch.
> > >>
> > >> I don't think we should add WARNS in the slab hot paths just to prevent this
> > >> rare error of using slab too early. At most VM_WARN... would be acceptable
> > >> but still not necessary as crashing immediately from a NULL pointer is
> > >> sufficient.
> > >>
> > >> So IMHO mips should fix their soc init,
> > >
> > > Yes, for the mips fix, John has proposed to defer the calling of prom_soc_init(),
> > > which looks reasonable.
> > >
> > >> and we should look into the
> > >> CONFIG_TRACING=n variant of kmalloc_trace(), to pass orig_size properly.
> > >
> > > You mean check if the pointer is NULL and bail out early.
> >
> > No I mean here:
> >
> > #else /* CONFIG_TRACING */
> > /* Save a function call when CONFIG_TRACING=n */
> > static __always_inline __alloc_size(3)
> > void *kmalloc_trace(struct kmem_cache *s, gfp_t flags, size_t size)
> > {
> > void *ret = kmem_cache_alloc(s, flags);
> >
> > ret = kasan_kmalloc(s, ret, size, flags);
> > return ret;
> > }
> >
> > we call kmem_cache_alloc() and discard the size parameter, so it will assume
> > s->object_size (and as the side-effect, crash if s is NULL). We shouldn't
> > add "s is NULL?" checks, but fix passing the size - probably switch to
> > __kmem_cache_alloc_node()? and in the following kmalloc_node_trace() analogically.
>
> Got it, thanks! I might have missed it during some rebasing for the
> kmalloc wastage debug patch.
That was good catch and I missed too!
But FYI I'm suggesting to drop CONFIG_TRACING=n variant:
https://lore.kernel.org/linux-mm/20221101222520.never.109-kees@kernel.org/T/#m20ecf14390e406247bde0ea9cce368f469c539ed
Any thoughts?
>
> How about the following fix?
>
> Thanks,
> Feng
>
> ---
> From 9f9fa9da8946fd44625f873c0f51167357075be1 Mon Sep 17 00:00:00 2001
> From: Feng Tang <feng.tang@intel.com>
> Date: Thu, 3 Nov 2022 21:32:10 +0800
> Subject: [PATCH] mm/slub: Add missing orig_size parameter for wastage debug
>
> commit 6edf2576a6cc ("mm/slub: enable debugging memory wasting of
> kmalloc") was introduced for debugging kmalloc memory wastage,
> and it missed to pass the original request size for kmalloc_trace()
> and kmalloc_node_trace() in CONFIG_TRACING=n path.
>
> Fix it by using __kmem_cache_alloc_node() with correct original
> request size.
>
> Fixes: 6edf2576a6cc ("mm/slub: enable debugging memory wasting of kmalloc")
> Suggested-by: Vlastimil Babka <vbabka@suse.cz>
> Signed-off-by: Feng Tang <feng.tang@intel.com>
> ---
> include/linux/slab.h | 9 +++++++--
> 1 file changed, 7 insertions(+), 2 deletions(-)
>
> diff --git a/include/linux/slab.h b/include/linux/slab.h
> index 90877fcde70b..9691afa569e1 100644
> --- a/include/linux/slab.h
> +++ b/include/linux/slab.h
> @@ -469,6 +469,9 @@ void *__kmalloc_node(size_t size, gfp_t flags, int node) __assume_kmalloc_alignm
> __alloc_size(1);
> void *kmem_cache_alloc_node(struct kmem_cache *s, gfp_t flags, int node) __assume_slab_alignment
> __malloc;
> +void *__kmem_cache_alloc_node(struct kmem_cache *s, gfp_t flags, int node,
> + size_t orig_size, unsigned long caller) __assume_slab_alignment
> + __malloc;
>
> #ifdef CONFIG_TRACING
> void *kmalloc_trace(struct kmem_cache *s, gfp_t flags, size_t size)
> @@ -482,7 +485,8 @@ void *kmalloc_node_trace(struct kmem_cache *s, gfp_t gfpflags,
> static __always_inline __alloc_size(3)
> void *kmalloc_trace(struct kmem_cache *s, gfp_t flags, size_t size)
> {
> - void *ret = kmem_cache_alloc(s, flags);
> + void *ret = __kmem_cache_alloc_node(s, flags, NUMA_NO_NODE,
> + size, _RET_IP_);
>
> ret = kasan_kmalloc(s, ret, size, flags);
> return ret;
> @@ -492,7 +496,8 @@ static __always_inline __alloc_size(4)
> void *kmalloc_node_trace(struct kmem_cache *s, gfp_t gfpflags,
> int node, size_t size)
> {
> - void *ret = kmem_cache_alloc_node(s, gfpflags, node);
> + void *ret = __kmem_cache_alloc_node(s, gfpflags, node,
> + size, _RET_IP_);
>
> ret = kasan_kmalloc(s, ret, size, gfpflags);
> return ret;
> --
> 2.34.1
>
>
>
--
Thanks,
Hyeonggon
On 11/3/22 15:36, Hyeonggon Yoo wrote:
> On Thu, Nov 03, 2022 at 10:16:12PM +0800, Feng Tang wrote:
>> On Thu, Nov 03, 2022 at 09:33:28AM +0100, Vlastimil Babka wrote:
>> [...]
>> > >> AFAICS before this patch, we "survive" "kmem_cache *s" being NULL as
>> > >> slab_pre_alloc_hook() will happen to return NULL and we bail out from
>> > >> slab_alloc_node(). But this is a side-effect, not an intended protection.
>> > >> Also the CONFIG_TRACING variant of kmalloc_trace() would have called
>> > >> trace_kmalloc dereferencing s->size anyway even before this patch.
>> > >>
>> > >> I don't think we should add WARNS in the slab hot paths just to prevent this
>> > >> rare error of using slab too early. At most VM_WARN... would be acceptable
>> > >> but still not necessary as crashing immediately from a NULL pointer is
>> > >> sufficient.
>> > >>
>> > >> So IMHO mips should fix their soc init,
>> > >
>> > > Yes, for the mips fix, John has proposed to defer the calling of prom_soc_init(),
>> > > which looks reasonable.
>> > >
>> > >> and we should look into the
>> > >> CONFIG_TRACING=n variant of kmalloc_trace(), to pass orig_size properly.
>> > >
>> > > You mean check if the pointer is NULL and bail out early.
>> >
>> > No I mean here:
>> >
>> > #else /* CONFIG_TRACING */
>> > /* Save a function call when CONFIG_TRACING=n */
>> > static __always_inline __alloc_size(3)
>> > void *kmalloc_trace(struct kmem_cache *s, gfp_t flags, size_t size)
>> > {
>> > void *ret = kmem_cache_alloc(s, flags);
>> >
>> > ret = kasan_kmalloc(s, ret, size, flags);
>> > return ret;
>> > }
>> >
>> > we call kmem_cache_alloc() and discard the size parameter, so it will assume
>> > s->object_size (and as the side-effect, crash if s is NULL). We shouldn't
>> > add "s is NULL?" checks, but fix passing the size - probably switch to
>> > __kmem_cache_alloc_node()? and in the following kmalloc_node_trace() analogically.
>>
>> Got it, thanks! I might have missed it during some rebasing for the
>> kmalloc wastage debug patch.
>
> That was good catch and I missed too!
> But FYI I'm suggesting to drop CONFIG_TRACING=n variant:
>
> https://lore.kernel.org/linux-mm/20221101222520.never.109-kees@kernel.org/T/#m20ecf14390e406247bde0ea9cce368f469c539ed
>
> Any thoughts?
I'll get to it, also I think we were pondering that within your series too,
but I wanted to postpone in case somebody objects to the extra function call
it creates.
But that would be for 6.2 anyway while I'll collect the fix here for 6.1.
>>
>> How about the following fix?
>>
>> Thanks,
>> Feng
>>
>> ---
>> From 9f9fa9da8946fd44625f873c0f51167357075be1 Mon Sep 17 00:00:00 2001
>> From: Feng Tang <feng.tang@intel.com>
>> Date: Thu, 3 Nov 2022 21:32:10 +0800
>> Subject: [PATCH] mm/slub: Add missing orig_size parameter for wastage debug
>>
>> commit 6edf2576a6cc ("mm/slub: enable debugging memory wasting of
>> kmalloc") was introduced for debugging kmalloc memory wastage,
>> and it missed to pass the original request size for kmalloc_trace()
>> and kmalloc_node_trace() in CONFIG_TRACING=n path.
>>
>> Fix it by using __kmem_cache_alloc_node() with correct original
>> request size.
>>
>> Fixes: 6edf2576a6cc ("mm/slub: enable debugging memory wasting of kmalloc")
>> Suggested-by: Vlastimil Babka <vbabka@suse.cz>
>> Signed-off-by: Feng Tang <feng.tang@intel.com>
>> ---
>> include/linux/slab.h | 9 +++++++--
>> 1 file changed, 7 insertions(+), 2 deletions(-)
>>
>> diff --git a/include/linux/slab.h b/include/linux/slab.h
>> index 90877fcde70b..9691afa569e1 100644
>> --- a/include/linux/slab.h
>> +++ b/include/linux/slab.h
>> @@ -469,6 +469,9 @@ void *__kmalloc_node(size_t size, gfp_t flags, int node) __assume_kmalloc_alignm
>> __alloc_size(1);
>> void *kmem_cache_alloc_node(struct kmem_cache *s, gfp_t flags, int node) __assume_slab_alignment
>> __malloc;
>> +void *__kmem_cache_alloc_node(struct kmem_cache *s, gfp_t flags, int node,
>> + size_t orig_size, unsigned long caller) __assume_slab_alignment
>> + __malloc;
>>
>> #ifdef CONFIG_TRACING
>> void *kmalloc_trace(struct kmem_cache *s, gfp_t flags, size_t size)
>> @@ -482,7 +485,8 @@ void *kmalloc_node_trace(struct kmem_cache *s, gfp_t gfpflags,
>> static __always_inline __alloc_size(3)
>> void *kmalloc_trace(struct kmem_cache *s, gfp_t flags, size_t size)
>> {
>> - void *ret = kmem_cache_alloc(s, flags);
>> + void *ret = __kmem_cache_alloc_node(s, flags, NUMA_NO_NODE,
>> + size, _RET_IP_);
>>
>> ret = kasan_kmalloc(s, ret, size, flags);
>> return ret;
>> @@ -492,7 +496,8 @@ static __always_inline __alloc_size(4)
>> void *kmalloc_node_trace(struct kmem_cache *s, gfp_t gfpflags,
>> int node, size_t size)
>> {
>> - void *ret = kmem_cache_alloc_node(s, gfpflags, node);
>> + void *ret = __kmem_cache_alloc_node(s, gfpflags, node,
>> + size, _RET_IP_);
>>
>> ret = kasan_kmalloc(s, ret, size, gfpflags);
>> return ret;
>> --
>> 2.34.1
>>
>>
>>
>
On 11/3/22 17:57, Vlastimil Babka wrote:
> On 11/3/22 15:36, Hyeonggon Yoo wrote:
>> On Thu, Nov 03, 2022 at 10:16:12PM +0800, Feng Tang wrote:
>>> On Thu, Nov 03, 2022 at 09:33:28AM +0100, Vlastimil Babka wrote:
>>> [...]
>>> > >> AFAICS before this patch, we "survive" "kmem_cache *s" being NULL as
>>> > >> slab_pre_alloc_hook() will happen to return NULL and we bail out from
>>> > >> slab_alloc_node(). But this is a side-effect, not an intended protection.
>>> > >> Also the CONFIG_TRACING variant of kmalloc_trace() would have called
>>> > >> trace_kmalloc dereferencing s->size anyway even before this patch.
>>> > >>
>>> > >> I don't think we should add WARNS in the slab hot paths just to prevent this
>>> > >> rare error of using slab too early. At most VM_WARN... would be acceptable
>>> > >> but still not necessary as crashing immediately from a NULL pointer is
>>> > >> sufficient.
>>> > >>
>>> > >> So IMHO mips should fix their soc init,
>>> > >
>>> > > Yes, for the mips fix, John has proposed to defer the calling of prom_soc_init(),
>>> > > which looks reasonable.
>>> > >
>>> > >> and we should look into the
>>> > >> CONFIG_TRACING=n variant of kmalloc_trace(), to pass orig_size properly.
>>> > >
>>> > > You mean check if the pointer is NULL and bail out early.
>>> >
>>> > No I mean here:
>>> >
>>> > #else /* CONFIG_TRACING */
>>> > /* Save a function call when CONFIG_TRACING=n */
>>> > static __always_inline __alloc_size(3)
>>> > void *kmalloc_trace(struct kmem_cache *s, gfp_t flags, size_t size)
>>> > {
>>> > void *ret = kmem_cache_alloc(s, flags);
>>> >
>>> > ret = kasan_kmalloc(s, ret, size, flags);
>>> > return ret;
>>> > }
>>> >
>>> > we call kmem_cache_alloc() and discard the size parameter, so it will assume
>>> > s->object_size (and as the side-effect, crash if s is NULL). We shouldn't
>>> > add "s is NULL?" checks, but fix passing the size - probably switch to
>>> > __kmem_cache_alloc_node()? and in the following kmalloc_node_trace() analogically.
>>>
>>> Got it, thanks! I might have missed it during some rebasing for the
>>> kmalloc wastage debug patch.
>>
>> That was good catch and I missed too!
>> But FYI I'm suggesting to drop CONFIG_TRACING=n variant:
>>
>> https://lore.kernel.org/linux-mm/20221101222520.never.109-kees@kernel.org/T/#m20ecf14390e406247bde0ea9cce368f469c539ed
>>
>> Any thoughts?
>
> I'll get to it, also I think we were pondering that within your series too,
> but I wanted to postpone in case somebody objects to the extra function call
> it creates.
> But that would be for 6.2 anyway while I'll collect the fix here for 6.1.
On second thought, the fix is making the inlined kmalloc_trace() expand to a
call that had 2 parameters and now it has 5, which seems to me like a worse
thing (code bloat) than the function call. With the other reasons to ditch
the CONFIG_TRACING=n variant I'm inclined to just do it right now.
>>>
>>> How about the following fix?
>>>
>>> Thanks,
>>> Feng
>>>
>>> ---
>>> From 9f9fa9da8946fd44625f873c0f51167357075be1 Mon Sep 17 00:00:00 2001
>>> From: Feng Tang <feng.tang@intel.com>
>>> Date: Thu, 3 Nov 2022 21:32:10 +0800
>>> Subject: [PATCH] mm/slub: Add missing orig_size parameter for wastage debug
>>>
>>> commit 6edf2576a6cc ("mm/slub: enable debugging memory wasting of
>>> kmalloc") was introduced for debugging kmalloc memory wastage,
>>> and it missed to pass the original request size for kmalloc_trace()
>>> and kmalloc_node_trace() in CONFIG_TRACING=n path.
>>>
>>> Fix it by using __kmem_cache_alloc_node() with correct original
>>> request size.
>>>
>>> Fixes: 6edf2576a6cc ("mm/slub: enable debugging memory wasting of kmalloc")
>>> Suggested-by: Vlastimil Babka <vbabka@suse.cz>
>>> Signed-off-by: Feng Tang <feng.tang@intel.com>
>>> ---
>>> include/linux/slab.h | 9 +++++++--
>>> 1 file changed, 7 insertions(+), 2 deletions(-)
>>>
>>> diff --git a/include/linux/slab.h b/include/linux/slab.h
>>> index 90877fcde70b..9691afa569e1 100644
>>> --- a/include/linux/slab.h
>>> +++ b/include/linux/slab.h
>>> @@ -469,6 +469,9 @@ void *__kmalloc_node(size_t size, gfp_t flags, int node) __assume_kmalloc_alignm
>>> __alloc_size(1);
>>> void *kmem_cache_alloc_node(struct kmem_cache *s, gfp_t flags, int node) __assume_slab_alignment
>>> __malloc;
>>> +void *__kmem_cache_alloc_node(struct kmem_cache *s, gfp_t flags, int node,
>>> + size_t orig_size, unsigned long caller) __assume_slab_alignment
>>> + __malloc;
>>>
>>> #ifdef CONFIG_TRACING
>>> void *kmalloc_trace(struct kmem_cache *s, gfp_t flags, size_t size)
>>> @@ -482,7 +485,8 @@ void *kmalloc_node_trace(struct kmem_cache *s, gfp_t gfpflags,
>>> static __always_inline __alloc_size(3)
>>> void *kmalloc_trace(struct kmem_cache *s, gfp_t flags, size_t size)
>>> {
>>> - void *ret = kmem_cache_alloc(s, flags);
>>> + void *ret = __kmem_cache_alloc_node(s, flags, NUMA_NO_NODE,
>>> + size, _RET_IP_);
>>>
>>> ret = kasan_kmalloc(s, ret, size, flags);
>>> return ret;
>>> @@ -492,7 +496,8 @@ static __always_inline __alloc_size(4)
>>> void *kmalloc_node_trace(struct kmem_cache *s, gfp_t gfpflags,
>>> int node, size_t size)
>>> {
>>> - void *ret = kmem_cache_alloc_node(s, gfpflags, node);
>>> + void *ret = __kmem_cache_alloc_node(s, gfpflags, node,
>>> + size, _RET_IP_);
>>>
>>> ret = kasan_kmalloc(s, ret, size, gfpflags);
>>> return ret;
>>> --
>>> 2.34.1
>>>
>>>
>>>
>>
>
On Thu, Nov 03, 2022 at 06:35:53PM +0100, Vlastimil Babka wrote:
[...]
> >> But FYI I'm suggesting to drop CONFIG_TRACING=n variant:
> >>
> >> https://lore.kernel.org/linux-mm/20221101222520.never.109-kees@kernel.org/T/#m20ecf14390e406247bde0ea9cce368f469c539ed
> >>
> >> Any thoughts?
> >
> > I'll get to it, also I think we were pondering that within your series too,
> > but I wanted to postpone in case somebody objects to the extra function call
> > it creates.
> > But that would be for 6.2 anyway while I'll collect the fix here for 6.1.
>
> On second thought, the fix is making the inlined kmalloc_trace() expand to a
> call that had 2 parameters and now it has 5, which seems to me like a worse
> thing (code bloat) than the function call. With the other reasons to ditch
> the CONFIG_TRACING=n variant I'm inclined to just do it right now.
That's great! It will save much trouble, and reduce code complexity.
Btw, the patch below also has some compiling issue for some kconfig
(thanks to 0Day's kbuild bot).
Thanks,
Feng
> >>>
> >>> How about the following fix?
> >>>
> >>> Thanks,
> >>> Feng
> >>>
> >>> ---
> >>> From 9f9fa9da8946fd44625f873c0f51167357075be1 Mon Sep 17 00:00:00 2001
> >>> From: Feng Tang <feng.tang@intel.com>
> >>> Date: Thu, 3 Nov 2022 21:32:10 +0800
> >>> Subject: [PATCH] mm/slub: Add missing orig_size parameter for wastage debug
> >>>
> >>> commit 6edf2576a6cc ("mm/slub: enable debugging memory wasting of
> >>> kmalloc") was introduced for debugging kmalloc memory wastage,
> >>> and it missed to pass the original request size for kmalloc_trace()
> >>> and kmalloc_node_trace() in CONFIG_TRACING=n path.
> >>>
> >>> Fix it by using __kmem_cache_alloc_node() with correct original
> >>> request size.
> >>>
> >>> Fixes: 6edf2576a6cc ("mm/slub: enable debugging memory wasting of kmalloc")
> >>> Suggested-by: Vlastimil Babka <vbabka@suse.cz>
> >>> Signed-off-by: Feng Tang <feng.tang@intel.com>
> >>> ---
> >>> include/linux/slab.h | 9 +++++++--
> >>> 1 file changed, 7 insertions(+), 2 deletions(-)
> >>>
> >>> diff --git a/include/linux/slab.h b/include/linux/slab.h
> >>> index 90877fcde70b..9691afa569e1 100644
> >>> --- a/include/linux/slab.h
> >>> +++ b/include/linux/slab.h
> >>> @@ -469,6 +469,9 @@ void *__kmalloc_node(size_t size, gfp_t flags, int node) __assume_kmalloc_alignm
> >>> __alloc_size(1);
> >>> void *kmem_cache_alloc_node(struct kmem_cache *s, gfp_t flags, int node) __assume_slab_alignment
> >>> __malloc;
> >>> +void *__kmem_cache_alloc_node(struct kmem_cache *s, gfp_t flags, int node,
> >>> + size_t orig_size, unsigned long caller) __assume_slab_alignment
> >>> + __malloc;
> >>>
> >>> #ifdef CONFIG_TRACING
> >>> void *kmalloc_trace(struct kmem_cache *s, gfp_t flags, size_t size)
> >>> @@ -482,7 +485,8 @@ void *kmalloc_node_trace(struct kmem_cache *s, gfp_t gfpflags,
> >>> static __always_inline __alloc_size(3)
> >>> void *kmalloc_trace(struct kmem_cache *s, gfp_t flags, size_t size)
> >>> {
> >>> - void *ret = kmem_cache_alloc(s, flags);
> >>> + void *ret = __kmem_cache_alloc_node(s, flags, NUMA_NO_NODE,
> >>> + size, _RET_IP_);
> >>>
> >>> ret = kasan_kmalloc(s, ret, size, flags);
> >>> return ret;
> >>> @@ -492,7 +496,8 @@ static __always_inline __alloc_size(4)
> >>> void *kmalloc_node_trace(struct kmem_cache *s, gfp_t gfpflags,
> >>> int node, size_t size)
> >>> {
> >>> - void *ret = kmem_cache_alloc_node(s, gfpflags, node);
> >>> + void *ret = __kmem_cache_alloc_node(s, gfpflags, node,
> >>> + size, _RET_IP_);
> >>>
> >>> ret = kasan_kmalloc(s, ret, size, gfpflags);
> >>> return ret;
> >>> --
> >>> 2.34.1
> >>>
> >>>
> >>>
> >>
> >
>
>
On Tue, Nov 01, 2022 at 10:33:32AM +0000, John Thomson wrote:
> On Tue, 1 Nov 2022, at 09:31, Hyeonggon Yoo wrote:
> > On Tue, Nov 01, 2022 at 09:20:21AM +0000, John Thomson wrote:
> >> On Tue, 1 Nov 2022, at 07:57, Feng Tang wrote:
> >> > Hi Thomson,
> >> >
> >> > Thanks for testing!
> >> >
> >> > + mips maintainer and mail list. The original report is here
> >> >
> >> > https://lore.kernel.org/lkml/becf2ac3-2a90-4f3a-96d9-a70f67c66e4a@app.fastmail.com/
> >>
> >> I am guessing my issue comes from __kmem_cache_alloc_lru accessing s->object_size when (kmem_cache) s is NULL?
> >> If that is the case, this change is not to blame, it only exposes the issue?
> >>
> >> I get the following dmesg (note very early NULL kmem_cache) with the below change atop v6.1-rc3:
> >>
> >> transfer started ......................................... transfer ok, time=2.02s
> >> setting up elf image... OK
> >> jumping to kernel code
> >> zimage at: 80B842A0 810B4EFC
> >>
> >> Uncompressing Linux at load address 80001000
> >>
> >> Copy device tree to address 80B80EE0
> >>
> >> Now, booting the kernel...
> >>
> >> [ 0.000000] Linux version 6.1.0-rc3+ (john@john) (mipsel-buildroot-linux-gnu-gcc.br_real (Buildroot 2021.11-4428-g6b6741b) 12.2.0, GNU ld (GNU Binutils) 2.39) #61 SMP Tue Nov 1 18:04:13 AEST 2022
> >> [ 0.000000] slub: kmem_cache_alloc called with kmem_cache: 0x0
> >> [ 0.000000] slub: __kmem_cache_alloc_lru called with kmem_cache: 0x0
> >> [ 0.000000] SoC Type: MediaTek MT7621 ver:1 eco:3
> >> [ 0.000000] printk: bootconsole [early0] enabled
> >> [ 0.000000] CPU0 revision is: 0001992f (MIPS 1004Kc)
> >> [ 0.000000] MIPS: machine is MikroTik RouterBOARD 760iGS
> >>
> >> normal boot
> >>
> >>
> >> diff --git a/mm/slub.c b/mm/slub.c
> >> index 157527d7101b..10fcdf2520d2 100644
> >> --- a/mm/slub.c
> >> +++ b/mm/slub.c
> >> @@ -3410,7 +3410,13 @@ static __always_inline
> >> void *__kmem_cache_alloc_lru(struct kmem_cache *s, struct list_lru *lru,
> >> gfp_t gfpflags)
> >> {
> >> - void *ret = slab_alloc(s, lru, gfpflags, _RET_IP_, s->object_size);
> >> + void *ret;
> >> + if (IS_ERR_OR_NULL(s)) {
> >> + pr_warn("slub: __kmem_cache_alloc_lru called with kmem_cache: %pSR\n", s);
> >> + ret = slab_alloc(s, lru, gfpflags, _RET_IP_, 0);
> >> + } else {
> >> + ret = slab_alloc(s, lru, gfpflags, _RET_IP_, s->object_size);
> >> + }
> >>
> >> trace_kmem_cache_alloc(_RET_IP_, ret, s, gfpflags, NUMA_NO_NODE);
> >>
> >> @@ -3419,6 +3425,8 @@ void *__kmem_cache_alloc_lru(struct kmem_cache *s, struct list_lru *lru,
> >>
> >> void *kmem_cache_alloc(struct kmem_cache *s, gfp_t gfpflags)
> >> {
> >> + if (IS_ERR_OR_NULL(s))
> >> + pr_warn("slub: kmem_cache_alloc called with kmem_cache: %pSR\n", s);
> >> return __kmem_cache_alloc_lru(s, NULL, gfpflags);
> >> }
> >> EXPORT_SYMBOL(kmem_cache_alloc);
> >> @@ -3426,6 +3434,8 @@ EXPORT_SYMBOL(kmem_cache_alloc);
> >> void *kmem_cache_alloc_lru(struct kmem_cache *s, struct list_lru *lru,
> >> gfp_t gfpflags)
> >> {
> >> + if (IS_ERR_OR_NULL(s))
> >> + pr_warn("slub: __kmem_cache_alloc_lru called with kmem_cache: %pSR\n", s);
> >> return __kmem_cache_alloc_lru(s, lru, gfpflags);
> >> }
> >> EXPORT_SYMBOL(kmem_cache_alloc_lru);
> >>
> >>
> >> Any hints on where kmem_cache_alloc would be being called from this early?
> >> I will start looking from /init/main.c around pr_notice("%s", linux_banner);
> >
> > Great. Would you try calling dump_stack(); when we observed s == NULL?
> > That would give more information about who passed s == NULL to these
> > functions.
> >
>
> With the dump_stack() in place:
>
> Now, booting the kernel...
>
> [ 0.000000] Linux version 6.1.0-rc3+ (john@john) (mipsel-buildroot-linux-gnu-gcc.br_real (Buildroot 2021.11-4428-g6b6741b) 12.2.0, GNU ld (GNU Binutils) 2.39) #62 SMP Tue Nov 1 19:49:52 AEST 2022
> [ 0.000000] slub: __kmem_cache_alloc_lru called with kmem_cache ptr: 0x0
> [ 0.000000] CPU: 0 PID: 0 Comm: swapper Not tainted 6.1.0-rc3+ #62
> [ 0.000000] Stack : 810fff78 80084d98 80889d00 00000004 00000000 00000000 80889d5c 80c90000
> [ 0.000000] 80920000 807bd380 8089d368 80923bd3 00000000 00000001 80889d08 00000000
> [ 0.000000] 00000000 00000000 807bd380 8084bd51 00000002 00000002 00000001 6d6f4320
> [ 0.000000] 00000000 80c97ce9 80c97d14 fffffffc 807bd380 00000000 00000003 00000dc0
> [ 0.000000] 00000000 a0000000 80910000 8110a0b4 00000000 00000020 80010000 80010000
> [ 0.000000] ...
> [ 0.000000] Call Trace:
> [ 0.000000] [<80008260>] show_stack+0x28/0xf0
> [ 0.000000] [<8070cdc0>] dump_stack_lvl+0x60/0x80
> [ 0.000000] [<801c1428>] kmem_cache_alloc+0x5c0/0x740
> [ 0.000000] [<8092856c>] prom_soc_init+0x1fc/0x2b4
> [ 0.000000] [<80928060>] prom_init+0x44/0xf0
> [ 0.000000] [<80929214>] setup_arch+0x4c/0x6a8
> [ 0.000000] [<809257e0>] start_kernel+0x88/0x7c0
> [ 0.000000]
> [ 0.000000] SoC Type: MediaTek MT7621 ver:1 eco:3
setup_arch() is too early to use slab allocators.
I think slab received NULL pointer because kmalloc is not initialized.
It seems arch/mips/ralink/mt7621.c is using slab too early.
>
>
> Now, booting the kernel...
>
> [ 0.000000] Linux version 6.1.0-rc3+ (john@john) (mipsel-buildroot-linux-gnu-gcc.br_real (Buildroot 2021.11-4428-g6b6741b) 12.2.0, GNU ld (GNU Binutils) 2.39) #62 SMP Tue Nov 1 19:49:52 AEST 2022
> [ 0.000000] slub: __kmem_cache_alloc_lru called with kmem_cache ptr: 0x0
> [ 0.000000] CPU: 0 PID: 0 Comm: swapper Not tainted 6.1.0-rc3+ #62
> [ 0.000000] Stack : 810fff78 80084d98 80889d00 00000004 00000000 00000000 80889d5c 80c90000
> [ 0.000000] 80920000 807bd380 8089d368 80923bd3 00000000 00000001 80889d08 00000000
> [ 0.000000] 00000000 00000000 807bd380 8084bd51 00000002 00000002 00000001 6d6f4320
> [ 0.000000] 00000000 80c97ce9 80c97d14 fffffffc 807bd380 00000000 00000003 00000dc0
> [ 0.000000] 00000000 a0000000 80910000 8110a0b4 00000000 00000020 80010000 80010000
> [ 0.000000] ...
> [ 0.000000] Call Trace:
> [ 0.000000] show_stack (/mnt/pool_ssd/code/linux/linux-stable-mt7621/./arch/mips/include/asm/stacktrace.h:43 /mnt/pool_ssd/code/linux/linux-stable-mt7621/arch/mips/kernel/traps.c:223)
> [ 0.000000] dump_stack_lvl (/mnt/pool_ssd/code/linux/linux-stable-mt7621/lib/dump_stack.c:107 (discriminator 1))
> [ 0.000000] kmem_cache_alloc (/mnt/pool_ssd/code/linux/linux-stable-mt7621/mm/slub.c:3318 /mnt/pool_ssd/code/linux/linux-stable-mt7621/mm/slub.c:3406 /mnt/pool_ssd/code/linux/linux-stable-mt7621/mm/slub.c:3418 /mnt/pool_ssd/code/linux/linux-stable-mt7621/mm/slub.c:3430)
> [ 0.000000] prom_soc_init (/mnt/pool_ssd/code/linux/linux-stable-mt7621/arch/mips/ralink/mt7621.c:106 /mnt/pool_ssd/code/linux/linux-stable-mt7621/arch/mips/ralink/mt7621.c:177)
> [ 0.000000] prom_init (/mnt/pool_ssd/code/linux/linux-stable-mt7621/arch/mips/ralink/prom.c:64)
> [ 0.000000] setup_arch (/mnt/pool_ssd/code/linux/linux-stable-mt7621/arch/mips/kernel/setup.c:786)
> [ 0.000000] start_kernel (/mnt/pool_ssd/code/linux/linux-stable-mt7621/init/main.c:279 /mnt/pool_ssd/code/linux/linux-stable-mt7621/init/main.c:477 /mnt/pool_ssd/code/linux/linux-stable-mt7621/init/main.c:960)
> [ 0.000000]
> [ 0.000000] SoC Type: MediaTek MT7621 ver:1 eco:3
>
>
> I have not found it yet.
>
>
> Cheers,
> --
> John Thomson
--
Thanks,
Hyeonggon
On Tue, Nov 01, 2022 at 06:42:23PM +0800, Hyeonggon Yoo wrote:
> On Tue, Nov 01, 2022 at 10:33:32AM +0000, John Thomson wrote:
> > On Tue, 1 Nov 2022, at 09:31, Hyeonggon Yoo wrote:
> > > On Tue, Nov 01, 2022 at 09:20:21AM +0000, John Thomson wrote:
> > >> On Tue, 1 Nov 2022, at 07:57, Feng Tang wrote:
> > >> > Hi Thomson,
> > >> >
> > >> > Thanks for testing!
> > >> >
> > >> > + mips maintainer and mail list. The original report is here
> > >> >
> > >> > https://lore.kernel.org/lkml/becf2ac3-2a90-4f3a-96d9-a70f67c66e4a@app.fastmail.com/
> > >>
> > >> I am guessing my issue comes from __kmem_cache_alloc_lru accessing s->object_size when (kmem_cache) s is NULL?
> > >> If that is the case, this change is not to blame, it only exposes the issue?
> > >>
> > >> I get the following dmesg (note very early NULL kmem_cache) with the below change atop v6.1-rc3:
> > >>
> > >> transfer started ......................................... transfer ok, time=2.02s
> > >> setting up elf image... OK
> > >> jumping to kernel code
> > >> zimage at: 80B842A0 810B4EFC
> > >>
> > >> Uncompressing Linux at load address 80001000
> > >>
> > >> Copy device tree to address 80B80EE0
> > >>
> > >> Now, booting the kernel...
> > >>
> > >> [ 0.000000] Linux version 6.1.0-rc3+ (john@john) (mipsel-buildroot-linux-gnu-gcc.br_real (Buildroot 2021.11-4428-g6b6741b) 12.2.0, GNU ld (GNU Binutils) 2.39) #61 SMP Tue Nov 1 18:04:13 AEST 2022
> > >> [ 0.000000] slub: kmem_cache_alloc called with kmem_cache: 0x0
> > >> [ 0.000000] slub: __kmem_cache_alloc_lru called with kmem_cache: 0x0
> > >> [ 0.000000] SoC Type: MediaTek MT7621 ver:1 eco:3
> > >> [ 0.000000] printk: bootconsole [early0] enabled
> > >> [ 0.000000] CPU0 revision is: 0001992f (MIPS 1004Kc)
> > >> [ 0.000000] MIPS: machine is MikroTik RouterBOARD 760iGS
> > >>
> > >> normal boot
> > >>
> > >>
> > >> diff --git a/mm/slub.c b/mm/slub.c
> > >> index 157527d7101b..10fcdf2520d2 100644
> > >> --- a/mm/slub.c
> > >> +++ b/mm/slub.c
> > >> @@ -3410,7 +3410,13 @@ static __always_inline
> > >> void *__kmem_cache_alloc_lru(struct kmem_cache *s, struct list_lru *lru,
> > >> gfp_t gfpflags)
> > >> {
> > >> - void *ret = slab_alloc(s, lru, gfpflags, _RET_IP_, s->object_size);
> > >> + void *ret;
> > >> + if (IS_ERR_OR_NULL(s)) {
> > >> + pr_warn("slub: __kmem_cache_alloc_lru called with kmem_cache: %pSR\n", s);
> > >> + ret = slab_alloc(s, lru, gfpflags, _RET_IP_, 0);
> > >> + } else {
> > >> + ret = slab_alloc(s, lru, gfpflags, _RET_IP_, s->object_size);
> > >> + }
> > >>
> > >> trace_kmem_cache_alloc(_RET_IP_, ret, s, gfpflags, NUMA_NO_NODE);
> > >>
> > >> @@ -3419,6 +3425,8 @@ void *__kmem_cache_alloc_lru(struct kmem_cache *s, struct list_lru *lru,
> > >>
> > >> void *kmem_cache_alloc(struct kmem_cache *s, gfp_t gfpflags)
> > >> {
> > >> + if (IS_ERR_OR_NULL(s))
> > >> + pr_warn("slub: kmem_cache_alloc called with kmem_cache: %pSR\n", s);
> > >> return __kmem_cache_alloc_lru(s, NULL, gfpflags);
> > >> }
> > >> EXPORT_SYMBOL(kmem_cache_alloc);
> > >> @@ -3426,6 +3434,8 @@ EXPORT_SYMBOL(kmem_cache_alloc);
> > >> void *kmem_cache_alloc_lru(struct kmem_cache *s, struct list_lru *lru,
> > >> gfp_t gfpflags)
> > >> {
> > >> + if (IS_ERR_OR_NULL(s))
> > >> + pr_warn("slub: __kmem_cache_alloc_lru called with kmem_cache: %pSR\n", s);
> > >> return __kmem_cache_alloc_lru(s, lru, gfpflags);
> > >> }
> > >> EXPORT_SYMBOL(kmem_cache_alloc_lru);
> > >>
> > >>
> > >> Any hints on where kmem_cache_alloc would be being called from this early?
> > >> I will start looking from /init/main.c around pr_notice("%s", linux_banner);
> > >
> > > Great. Would you try calling dump_stack(); when we observed s == NULL?
> > > That would give more information about who passed s == NULL to these
> > > functions.
> > >
> >
> > With the dump_stack() in place:
> >
> > Now, booting the kernel...
> >
> > [ 0.000000] Linux version 6.1.0-rc3+ (john@john) (mipsel-buildroot-linux-gnu-gcc.br_real (Buildroot 2021.11-4428-g6b6741b) 12.2.0, GNU ld (GNU Binutils) 2.39) #62 SMP Tue Nov 1 19:49:52 AEST 2022
> > [ 0.000000] slub: __kmem_cache_alloc_lru called with kmem_cache ptr: 0x0
> > [ 0.000000] CPU: 0 PID: 0 Comm: swapper Not tainted 6.1.0-rc3+ #62
> > [ 0.000000] Stack : 810fff78 80084d98 80889d00 00000004 00000000 00000000 80889d5c 80c90000
> > [ 0.000000] 80920000 807bd380 8089d368 80923bd3 00000000 00000001 80889d08 00000000
> > [ 0.000000] 00000000 00000000 807bd380 8084bd51 00000002 00000002 00000001 6d6f4320
> > [ 0.000000] 00000000 80c97ce9 80c97d14 fffffffc 807bd380 00000000 00000003 00000dc0
> > [ 0.000000] 00000000 a0000000 80910000 8110a0b4 00000000 00000020 80010000 80010000
> > [ 0.000000] ...
> > [ 0.000000] Call Trace:
> > [ 0.000000] [<80008260>] show_stack+0x28/0xf0
> > [ 0.000000] [<8070cdc0>] dump_stack_lvl+0x60/0x80
> > [ 0.000000] [<801c1428>] kmem_cache_alloc+0x5c0/0x740
> > [ 0.000000] [<8092856c>] prom_soc_init+0x1fc/0x2b4
> > [ 0.000000] [<80928060>] prom_init+0x44/0xf0
> > [ 0.000000] [<80929214>] setup_arch+0x4c/0x6a8
> > [ 0.000000] [<809257e0>] start_kernel+0x88/0x7c0
> > [ 0.000000]
> > [ 0.000000] SoC Type: MediaTek MT7621 ver:1 eco:3
>
> setup_arch() is too early to use slab allocators.
> I think slab received NULL pointer because kmalloc is not initialized.
>
> It seems arch/mips/ralink/mt7621.c is using slab too early.
Cool! it is finally root caused :) Thanks!
The following patch should solve it and give it a warning message, though
I'm not sure if there is other holes.
Thanks,
Feng
---
diff --git a/mm/slab_common.c b/mm/slab_common.c
index 33b1886b06eb..429c21b7ecbc 100644
--- a/mm/slab_common.c
+++ b/mm/slab_common.c
@@ -1043,7 +1043,14 @@ size_t __ksize(const void *object)
#ifdef CONFIG_TRACING
void *kmalloc_trace(struct kmem_cache *s, gfp_t gfpflags, size_t size)
{
- void *ret = __kmem_cache_alloc_node(s, gfpflags, NUMA_NO_NODE,
+ void *ret;
+
+ if (unlikely(ZERO_OR_NULL_PTR(s))) {
+ WARN_ON_ONCE(1);
+ return s;
+ }
+
+ ret = __kmem_cache_alloc_node(s, gfpflags, NUMA_NO_NODE,
size, _RET_IP_);
trace_kmalloc(_RET_IP_, ret, size, s->size, gfpflags, NUMA_NO_NODE);
diff --git a/mm/slub.c b/mm/slub.c
index 157527d7101b..85d24bb6eda7 100644
--- a/mm/slub.c
+++ b/mm/slub.c
@@ -3410,8 +3410,14 @@ static __always_inline
void *__kmem_cache_alloc_lru(struct kmem_cache *s, struct list_lru *lru,
gfp_t gfpflags)
{
- void *ret = slab_alloc(s, lru, gfpflags, _RET_IP_, s->object_size);
+ void *ret;
+ if (unlikely(ZERO_OR_NULL_PTR(s))) {
+ WARN_ON_ONCE(1);
+ return s;
+ }
+
+ ret = slab_alloc(s, lru, gfpflags, _RET_IP_, s->object_size);
trace_kmem_cache_alloc(_RET_IP_, ret, s, gfpflags, NUMA_NO_NODE);
return ret;
On Tue, 1 Nov 2022, at 13:55, Feng Tang wrote:
> On Tue, Nov 01, 2022 at 06:42:23PM +0800, Hyeonggon Yoo wrote:
>> setup_arch() is too early to use slab allocators.
>> I think slab received NULL pointer because kmalloc is not initialized.
>>
>> It seems arch/mips/ralink/mt7621.c is using slab too early.
>
> Cool! it is finally root caused :) Thanks!
>
> The following patch should solve it and give it a warning message, though
> I'm not sure if there is other holes.
>
> Thanks,
> Feng
>
> ---
> diff --git a/mm/slab_common.c b/mm/slab_common.c
> index 33b1886b06eb..429c21b7ecbc 100644
> --- a/mm/slab_common.c
> +++ b/mm/slab_common.c
> @@ -1043,7 +1043,14 @@ size_t __ksize(const void *object)
> #ifdef CONFIG_TRACING
> void *kmalloc_trace(struct kmem_cache *s, gfp_t gfpflags, size_t size)
> {
> - void *ret = __kmem_cache_alloc_node(s, gfpflags, NUMA_NO_NODE,
> + void *ret;
> +
> + if (unlikely(ZERO_OR_NULL_PTR(s))) {
> + WARN_ON_ONCE(1);
> + return s;
> + }
> +
> + ret = __kmem_cache_alloc_node(s, gfpflags, NUMA_NO_NODE,
> size, _RET_IP_);
>
> trace_kmalloc(_RET_IP_, ret, size, s->size, gfpflags, NUMA_NO_NODE);
> diff --git a/mm/slub.c b/mm/slub.c
> index 157527d7101b..85d24bb6eda7 100644
> --- a/mm/slub.c
> +++ b/mm/slub.c
> @@ -3410,8 +3410,14 @@ static __always_inline
> void *__kmem_cache_alloc_lru(struct kmem_cache *s, struct list_lru *lru,
> gfp_t gfpflags)
> {
> - void *ret = slab_alloc(s, lru, gfpflags, _RET_IP_, s->object_size);
> + void *ret;
>
> + if (unlikely(ZERO_OR_NULL_PTR(s))) {
> + WARN_ON_ONCE(1);
> + return s;
> + }
> +
> + ret = slab_alloc(s, lru, gfpflags, _RET_IP_, s->object_size);
> trace_kmem_cache_alloc(_RET_IP_, ret, s, gfpflags, NUMA_NO_NODE);
>
> return ret;
Yes, thank you, that patch atop v6.1-rc3 lets me boot, and shows the warning and stack dump.
Will you submit that, or how do we want to proceed?
transfer started ......................................... transfer ok, time=2.11s
setting up elf image... OK
jumping to kernel code
zimage at: 80B842A0 810B4BC0
Uncompressing Linux at load address 80001000
Copy device tree to address 80B80EE0
Now, booting the kernel...
[ 0.000000] Linux version 6.1.0-rc3+ (john@john) (mipsel-buildroot-linux-gnu-gcc.br_real (Buildroot 2021.11-4428-g6b6741b) 12.2.0, GNU ld (GNU Binutils) 2.39) #73 SMP Wed Nov 2 05:10:01 AEST 2022
[ 0.000000] ------------[ cut here ]------------
[ 0.000000] WARNING: CPU: 0 PID: 0 at mm/slub.c:3416 kmem_cache_alloc+0x5a4/0x5e8
[ 0.000000] Modules linked in:
[ 0.000000] CPU: 0 PID: 0 Comm: swapper Not tainted 6.1.0-rc3+ #73
[ 0.000000] Stack : 810fff78 80084d98 00000000 00000004 00000000 00000000 80889d04 80c90000
[ 0.000000] 80920000 807bd328 8089d368 80923bd3 00000000 00000001 80889cb0 00000000
[ 0.000000] 00000000 00000000 807bd328 8084bcb1 00000002 00000002 00000001 6d6f4320
[ 0.000000] 00000000 80c97d3d 80c97d68 fffffffc 807bd328 00000000 00000000 00000000
[ 0.000000] 00000000 a0000000 80910000 8110a0b4 00000000 00000020 80010000 80010000
[ 0.000000] ...
[ 0.000000] Call Trace:
[ 0.000000] [<80008260>] show_stack+0x28/0xf0
[ 0.000000] [<8070c958>] dump_stack_lvl+0x60/0x80
[ 0.000000] [<8002e184>] __warn+0xc4/0xf8
[ 0.000000] [<8002e210>] warn_slowpath_fmt+0x58/0xa4
[ 0.000000] [<801c0fac>] kmem_cache_alloc+0x5a4/0x5e8
[ 0.000000] [<8092856c>] prom_soc_init+0x1fc/0x2b4
[ 0.000000] [<80928060>] prom_init+0x44/0xf0
[ 0.000000] [<80929214>] setup_arch+0x4c/0x6a8
[ 0.000000] [<809257e0>] start_kernel+0x88/0x7c0
[ 0.000000]
[ 0.000000] ---[ end trace 0000000000000000 ]---
[ 0.000000] SoC Type: MediaTek MT7621 ver:1 eco:3
[ 0.000000] printk: bootconsole [early0] enabled
Thank you for working through this with me.
I will try to address the root cause in mt7621.c.
It looks like other arch/** soc_device_register users use postcore_initcall, device_initcall,
or the ARM DT_MACHINE_START .init_machine. A quick hack to use postcore_initcall in mt7621
avoided this zero ptr kmem_cache passed to kmem_cache_alloc_lru.
Thanks,
--
John Thomson
On Tue, Nov 01, 2022 at 07:39:13PM +0000, John Thomson wrote:
>
>
> On Tue, 1 Nov 2022, at 13:55, Feng Tang wrote:
> > On Tue, Nov 01, 2022 at 06:42:23PM +0800, Hyeonggon Yoo wrote:
> >> setup_arch() is too early to use slab allocators.
> >> I think slab received NULL pointer because kmalloc is not initialized.
> >>
> >> It seems arch/mips/ralink/mt7621.c is using slab too early.
> >
> > Cool! it is finally root caused :) Thanks!
> >
> > The following patch should solve it and give it a warning message, though
> > I'm not sure if there is other holes.
> >
> > Thanks,
> > Feng
> >
> > ---
> > diff --git a/mm/slab_common.c b/mm/slab_common.c
> > index 33b1886b06eb..429c21b7ecbc 100644
> > --- a/mm/slab_common.c
> > +++ b/mm/slab_common.c
> > @@ -1043,7 +1043,14 @@ size_t __ksize(const void *object)
> > #ifdef CONFIG_TRACING
> > void *kmalloc_trace(struct kmem_cache *s, gfp_t gfpflags, size_t size)
> > {
> > - void *ret = __kmem_cache_alloc_node(s, gfpflags, NUMA_NO_NODE,
> > + void *ret;
> > +
> > + if (unlikely(ZERO_OR_NULL_PTR(s))) {
> > + WARN_ON_ONCE(1);
> > + return s;
> > + }
> > +
> > + ret = __kmem_cache_alloc_node(s, gfpflags, NUMA_NO_NODE,
> > size, _RET_IP_);
> >
> > trace_kmalloc(_RET_IP_, ret, size, s->size, gfpflags, NUMA_NO_NODE);
> > diff --git a/mm/slub.c b/mm/slub.c
> > index 157527d7101b..85d24bb6eda7 100644
> > --- a/mm/slub.c
> > +++ b/mm/slub.c
> > @@ -3410,8 +3410,14 @@ static __always_inline
> > void *__kmem_cache_alloc_lru(struct kmem_cache *s, struct list_lru *lru,
> > gfp_t gfpflags)
> > {
> > - void *ret = slab_alloc(s, lru, gfpflags, _RET_IP_, s->object_size);
> > + void *ret;
> >
> > + if (unlikely(ZERO_OR_NULL_PTR(s))) {
> > + WARN_ON_ONCE(1);
> > + return s;
> > + }
> > +
> > + ret = slab_alloc(s, lru, gfpflags, _RET_IP_, s->object_size);
> > trace_kmem_cache_alloc(_RET_IP_, ret, s, gfpflags, NUMA_NO_NODE);
> >
> > return ret;
>
> Yes, thank you, that patch atop v6.1-rc3 lets me boot, and shows the warning and stack dump.
> Will you submit that, or how do we want to proceed?
Thanks for confirming. I wanted to wait for Vlastimil, Hyeonggon and
other developer's opinion. And yes, I can also post a more formal one.
> transfer started ......................................... transfer ok, time=2.11s
> setting up elf image... OK
> jumping to kernel code
> zimage at: 80B842A0 810B4BC0
>
> Uncompressing Linux at load address 80001000
>
> Copy device tree to address 80B80EE0
>
> Now, booting the kernel...
>
> [ 0.000000] Linux version 6.1.0-rc3+ (john@john) (mipsel-buildroot-linux-gnu-gcc.br_real (Buildroot 2021.11-4428-g6b6741b) 12.2.0, GNU ld (GNU Binutils) 2.39) #73 SMP Wed Nov 2 05:10:01 AEST 2022
> [ 0.000000] ------------[ cut here ]------------
> [ 0.000000] WARNING: CPU: 0 PID: 0 at mm/slub.c:3416 kmem_cache_alloc+0x5a4/0x5e8
> [ 0.000000] Modules linked in:
> [ 0.000000] CPU: 0 PID: 0 Comm: swapper Not tainted 6.1.0-rc3+ #73
> [ 0.000000] Stack : 810fff78 80084d98 00000000 00000004 00000000 00000000 80889d04 80c90000
> [ 0.000000] 80920000 807bd328 8089d368 80923bd3 00000000 00000001 80889cb0 00000000
> [ 0.000000] 00000000 00000000 807bd328 8084bcb1 00000002 00000002 00000001 6d6f4320
> [ 0.000000] 00000000 80c97d3d 80c97d68 fffffffc 807bd328 00000000 00000000 00000000
> [ 0.000000] 00000000 a0000000 80910000 8110a0b4 00000000 00000020 80010000 80010000
> [ 0.000000] ...
> [ 0.000000] Call Trace:
> [ 0.000000] [<80008260>] show_stack+0x28/0xf0
> [ 0.000000] [<8070c958>] dump_stack_lvl+0x60/0x80
> [ 0.000000] [<8002e184>] __warn+0xc4/0xf8
> [ 0.000000] [<8002e210>] warn_slowpath_fmt+0x58/0xa4
> [ 0.000000] [<801c0fac>] kmem_cache_alloc+0x5a4/0x5e8
> [ 0.000000] [<8092856c>] prom_soc_init+0x1fc/0x2b4
> [ 0.000000] [<80928060>] prom_init+0x44/0xf0
> [ 0.000000] [<80929214>] setup_arch+0x4c/0x6a8
> [ 0.000000] [<809257e0>] start_kernel+0x88/0x7c0
> [ 0.000000]
> [ 0.000000] ---[ end trace 0000000000000000 ]---
> [ 0.000000] SoC Type: MediaTek MT7621 ver:1 eco:3
> [ 0.000000] printk: bootconsole [early0] enabled
>
> Thank you for working through this with me.
> I will try to address the root cause in mt7621.c.
> It looks like other arch/** soc_device_register users use postcore_initcall, device_initcall,
> or the ARM DT_MACHINE_START .init_machine. A quick hack to use postcore_initcall in mt7621
> avoided this zero ptr kmem_cache passed to kmem_cache_alloc_lru.
If IIUC, the prom_soc_init() is only called once in kernel, can the
'soc_dev_attr' just be defined as a global data structure instead
of calling kzalloc(), as its size is small only containing 7 pointers.
Thanks,
Feng
On Wed, Nov 02, 2022 at 02:08:09PM +0800, Feng Tang wrote:
> On Tue, Nov 01, 2022 at 07:39:13PM +0000, John Thomson wrote:
> >
> >
> > On Tue, 1 Nov 2022, at 13:55, Feng Tang wrote:
> > > On Tue, Nov 01, 2022 at 06:42:23PM +0800, Hyeonggon Yoo wrote:
> > >> setup_arch() is too early to use slab allocators.
> > >> I think slab received NULL pointer because kmalloc is not initialized.
> > >>
> > >> It seems arch/mips/ralink/mt7621.c is using slab too early.
> > >
> > > Cool! it is finally root caused :) Thanks!
> > >
> > > The following patch should solve it and give it a warning message, though
> > > I'm not sure if there is other holes.
> > >
> > > Thanks,
> > > Feng
> > >
> > > ---
> > > diff --git a/mm/slab_common.c b/mm/slab_common.c
> > > index 33b1886b06eb..429c21b7ecbc 100644
> > > --- a/mm/slab_common.c
> > > +++ b/mm/slab_common.c
> > > @@ -1043,7 +1043,14 @@ size_t __ksize(const void *object)
> > > #ifdef CONFIG_TRACING
> > > void *kmalloc_trace(struct kmem_cache *s, gfp_t gfpflags, size_t size)
> > > {
> > > - void *ret = __kmem_cache_alloc_node(s, gfpflags, NUMA_NO_NODE,
> > > + void *ret;
> > > +
> > > + if (unlikely(ZERO_OR_NULL_PTR(s))) {
> > > + WARN_ON_ONCE(1);
> > > + return s;
> > > + }
> > > +
> > > + ret = __kmem_cache_alloc_node(s, gfpflags, NUMA_NO_NODE,
> > > size, _RET_IP_);
> > >
> > > trace_kmalloc(_RET_IP_, ret, size, s->size, gfpflags, NUMA_NO_NODE);
> > > diff --git a/mm/slub.c b/mm/slub.c
> > > index 157527d7101b..85d24bb6eda7 100644
> > > --- a/mm/slub.c
> > > +++ b/mm/slub.c
> > > @@ -3410,8 +3410,14 @@ static __always_inline
> > > void *__kmem_cache_alloc_lru(struct kmem_cache *s, struct list_lru *lru,
> > > gfp_t gfpflags)
> > > {
> > > - void *ret = slab_alloc(s, lru, gfpflags, _RET_IP_, s->object_size);
> > > + void *ret;
> > >
> > > + if (unlikely(ZERO_OR_NULL_PTR(s))) {
> > > + WARN_ON_ONCE(1);
> > > + return s;
> > > + }
> > > +
Thank you for suggestion!
I think the holes are:
kmalloc_node_trace(), kmem_cache_alloc_node(), __do_kmalloc_node()
And want to suggest:
What about using VM_WARN_ON_ONCE() instead?
> > > + ret = slab_alloc(s, lru, gfpflags, _RET_IP_, s->object_size);
> > > trace_kmem_cache_alloc(_RET_IP_, ret, s, gfpflags, NUMA_NO_NODE);
> > >
> > > return ret;
> >
> > Yes, thank you, that patch atop v6.1-rc3 lets me boot, and shows the warning and stack dump.
> > Will you submit that, or how do we want to proceed?
>
> Thanks for confirming. I wanted to wait for Vlastimil, Hyeonggon and
> other developer's opinion. And yes, I can also post a more formal one.
>
> > transfer started ......................................... transfer ok, time=2.11s
> > setting up elf image... OK
> > jumping to kernel code
> > zimage at: 80B842A0 810B4BC0
> >
> > Uncompressing Linux at load address 80001000
> >
> > Copy device tree to address 80B80EE0
> >
> > Now, booting the kernel...
> >
> > [ 0.000000] Linux version 6.1.0-rc3+ (john@john) (mipsel-buildroot-linux-gnu-gcc.br_real (Buildroot 2021.11-4428-g6b6741b) 12.2.0, GNU ld (GNU Binutils) 2.39) #73 SMP Wed Nov 2 05:10:01 AEST 2022
> > [ 0.000000] ------------[ cut here ]------------
> > [ 0.000000] WARNING: CPU: 0 PID: 0 at mm/slub.c:3416 kmem_cache_alloc+0x5a4/0x5e8
> > [ 0.000000] Modules linked in:
> > [ 0.000000] CPU: 0 PID: 0 Comm: swapper Not tainted 6.1.0-rc3+ #73
> > [ 0.000000] Stack : 810fff78 80084d98 00000000 00000004 00000000 00000000 80889d04 80c90000
> > [ 0.000000] 80920000 807bd328 8089d368 80923bd3 00000000 00000001 80889cb0 00000000
> > [ 0.000000] 00000000 00000000 807bd328 8084bcb1 00000002 00000002 00000001 6d6f4320
> > [ 0.000000] 00000000 80c97d3d 80c97d68 fffffffc 807bd328 00000000 00000000 00000000
> > [ 0.000000] 00000000 a0000000 80910000 8110a0b4 00000000 00000020 80010000 80010000
> > [ 0.000000] ...
> > [ 0.000000] Call Trace:
> > [ 0.000000] [<80008260>] show_stack+0x28/0xf0
> > [ 0.000000] [<8070c958>] dump_stack_lvl+0x60/0x80
> > [ 0.000000] [<8002e184>] __warn+0xc4/0xf8
> > [ 0.000000] [<8002e210>] warn_slowpath_fmt+0x58/0xa4
> > [ 0.000000] [<801c0fac>] kmem_cache_alloc+0x5a4/0x5e8
> > [ 0.000000] [<8092856c>] prom_soc_init+0x1fc/0x2b4
> > [ 0.000000] [<80928060>] prom_init+0x44/0xf0
> > [ 0.000000] [<80929214>] setup_arch+0x4c/0x6a8
> > [ 0.000000] [<809257e0>] start_kernel+0x88/0x7c0
> > [ 0.000000]
> > [ 0.000000] ---[ end trace 0000000000000000 ]---
> > [ 0.000000] SoC Type: MediaTek MT7621 ver:1 eco:3
> > [ 0.000000] printk: bootconsole [early0] enabled
> >
> > Thank you for working through this with me.
> > I will try to address the root cause in mt7621.c.
> > It looks like other arch/** soc_device_register users use postcore_initcall, device_initcall,
> > or the ARM DT_MACHINE_START .init_machine. A quick hack to use postcore_initcall in mt7621
> > avoided this zero ptr kmem_cache passed to kmem_cache_alloc_lru.
>
> If IIUC, the prom_soc_init() is only called once in kernel, can the
> 'soc_dev_attr' just be defined as a global data structure instead
> of calling kzalloc(), as its size is small only containing 7 pointers.
But soc_device_registers() too uses kmalloc. I think calling it
after slab initialization will be best solution - if that is correct.
>
> Thanks,
> Feng
>
--
Thanks,
Hyeonggon
On Wed, Nov 02, 2022 at 04:16:06PM +0900, Hyeonggon Yoo wrote: > On Wed, Nov 02, 2022 at 02:08:09PM +0800, Feng Tang wrote: [...] > > > transfer started ......................................... transfer ok, time=2.11s > > > setting up elf image... OK > > > jumping to kernel code > > > zimage at: 80B842A0 810B4BC0 > > > > > > Uncompressing Linux at load address 80001000 > > > > > > Copy device tree to address 80B80EE0 > > > > > > Now, booting the kernel... > > > > > > [ 0.000000] Linux version 6.1.0-rc3+ (john@john) (mipsel-buildroot-linux-gnu-gcc.br_real (Buildroot 2021.11-4428-g6b6741b) 12.2.0, GNU ld (GNU Binutils) 2.39) #73 SMP Wed Nov 2 05:10:01 AEST 2022 > > > [ 0.000000] ------------[ cut here ]------------ > > > [ 0.000000] WARNING: CPU: 0 PID: 0 at mm/slub.c:3416 kmem_cache_alloc+0x5a4/0x5e8 > > > [ 0.000000] Modules linked in: > > > [ 0.000000] CPU: 0 PID: 0 Comm: swapper Not tainted 6.1.0-rc3+ #73 > > > [ 0.000000] Stack : 810fff78 80084d98 00000000 00000004 00000000 00000000 80889d04 80c90000 > > > [ 0.000000] 80920000 807bd328 8089d368 80923bd3 00000000 00000001 80889cb0 00000000 > > > [ 0.000000] 00000000 00000000 807bd328 8084bcb1 00000002 00000002 00000001 6d6f4320 > > > [ 0.000000] 00000000 80c97d3d 80c97d68 fffffffc 807bd328 00000000 00000000 00000000 > > > [ 0.000000] 00000000 a0000000 80910000 8110a0b4 00000000 00000020 80010000 80010000 > > > [ 0.000000] ... > > > [ 0.000000] Call Trace: > > > [ 0.000000] [<80008260>] show_stack+0x28/0xf0 > > > [ 0.000000] [<8070c958>] dump_stack_lvl+0x60/0x80 > > > [ 0.000000] [<8002e184>] __warn+0xc4/0xf8 > > > [ 0.000000] [<8002e210>] warn_slowpath_fmt+0x58/0xa4 > > > [ 0.000000] [<801c0fac>] kmem_cache_alloc+0x5a4/0x5e8 > > > [ 0.000000] [<8092856c>] prom_soc_init+0x1fc/0x2b4 > > > [ 0.000000] [<80928060>] prom_init+0x44/0xf0 > > > [ 0.000000] [<80929214>] setup_arch+0x4c/0x6a8 > > > [ 0.000000] [<809257e0>] start_kernel+0x88/0x7c0 > > > [ 0.000000] > > > [ 0.000000] ---[ end trace 0000000000000000 ]--- > > > [ 0.000000] SoC Type: MediaTek MT7621 ver:1 eco:3 > > > [ 0.000000] printk: bootconsole [early0] enabled > > > > > > Thank you for working through this with me. > > > I will try to address the root cause in mt7621.c. > > > It looks like other arch/** soc_device_register users use postcore_initcall, device_initcall, > > > or the ARM DT_MACHINE_START .init_machine. A quick hack to use postcore_initcall in mt7621 > > > avoided this zero ptr kmem_cache passed to kmem_cache_alloc_lru. > > > > If IIUC, the prom_soc_init() is only called once in kernel, can the > > 'soc_dev_attr' just be defined as a global data structure instead > > of calling kzalloc(), as its size is small only containing 7 pointers. > > But soc_device_registers() too uses kmalloc. I think calling it > after slab initialization will be best solution - if that is correct. Yes, you are right, there is other kmalloc() down the call chain. Hi John, Will you verify and submit a patch for your proposal of deferring calling prom_soc_init()? thanks - Feng
On Thu, 3 Nov 2022, at 07:18, Feng Tang wrote: > On Wed, Nov 02, 2022 at 04:16:06PM +0900, Hyeonggon Yoo wrote: >> On Wed, Nov 02, 2022 at 02:08:09PM +0800, Feng Tang wrote: > [...] >> > > transfer started ......................................... transfer ok, time=2.11s >> > > setting up elf image... OK >> > > jumping to kernel code >> > > zimage at: 80B842A0 810B4BC0 >> > > >> > > Uncompressing Linux at load address 80001000 >> > > >> > > Copy device tree to address 80B80EE0 >> > > >> > > Now, booting the kernel... >> > > >> > > [ 0.000000] Linux version 6.1.0-rc3+ (john@john) (mipsel-buildroot-linux-gnu-gcc.br_real (Buildroot 2021.11-4428-g6b6741b) 12.2.0, GNU ld (GNU Binutils) 2.39) #73 SMP Wed Nov 2 05:10:01 AEST 2022 >> > > [ 0.000000] ------------[ cut here ]------------ >> > > [ 0.000000] WARNING: CPU: 0 PID: 0 at mm/slub.c:3416 kmem_cache_alloc+0x5a4/0x5e8 >> > > [ 0.000000] Modules linked in: >> > > [ 0.000000] CPU: 0 PID: 0 Comm: swapper Not tainted 6.1.0-rc3+ #73 >> > > [ 0.000000] Stack : 810fff78 80084d98 00000000 00000004 00000000 00000000 80889d04 80c90000 >> > > [ 0.000000] 80920000 807bd328 8089d368 80923bd3 00000000 00000001 80889cb0 00000000 >> > > [ 0.000000] 00000000 00000000 807bd328 8084bcb1 00000002 00000002 00000001 6d6f4320 >> > > [ 0.000000] 00000000 80c97d3d 80c97d68 fffffffc 807bd328 00000000 00000000 00000000 >> > > [ 0.000000] 00000000 a0000000 80910000 8110a0b4 00000000 00000020 80010000 80010000 >> > > [ 0.000000] ... >> > > [ 0.000000] Call Trace: >> > > [ 0.000000] [<80008260>] show_stack+0x28/0xf0 >> > > [ 0.000000] [<8070c958>] dump_stack_lvl+0x60/0x80 >> > > [ 0.000000] [<8002e184>] __warn+0xc4/0xf8 >> > > [ 0.000000] [<8002e210>] warn_slowpath_fmt+0x58/0xa4 >> > > [ 0.000000] [<801c0fac>] kmem_cache_alloc+0x5a4/0x5e8 >> > > [ 0.000000] [<8092856c>] prom_soc_init+0x1fc/0x2b4 >> > > [ 0.000000] [<80928060>] prom_init+0x44/0xf0 >> > > [ 0.000000] [<80929214>] setup_arch+0x4c/0x6a8 >> > > [ 0.000000] [<809257e0>] start_kernel+0x88/0x7c0 >> > > [ 0.000000] >> > > [ 0.000000] ---[ end trace 0000000000000000 ]--- >> > > [ 0.000000] SoC Type: MediaTek MT7621 ver:1 eco:3 >> > > [ 0.000000] printk: bootconsole [early0] enabled >> > > >> > > Thank you for working through this with me. >> > > I will try to address the root cause in mt7621.c. >> > > It looks like other arch/** soc_device_register users use postcore_initcall, device_initcall, >> > > or the ARM DT_MACHINE_START .init_machine. A quick hack to use postcore_initcall in mt7621 >> > > avoided this zero ptr kmem_cache passed to kmem_cache_alloc_lru. >> > >> > If IIUC, the prom_soc_init() is only called once in kernel, can the >> > 'soc_dev_attr' just be defined as a global data structure instead >> > of calling kzalloc(), as its size is small only containing 7 pointers. >> >> But soc_device_registers() too uses kmalloc. I think calling it >> after slab initialization will be best solution - if that is correct. > > Yes, you are right, there is other kmalloc() down the call chain. > > Hi John, > > Will you verify and submit a patch for your proposal of deferring > calling prom_soc_init()? thanks > > - Feng Hi Feng, My proposed mt7621.c changes are RFC here: https://lore.kernel.org/lkml/20221103050538.1930758-1-git@johnthomson.fastmail.com.au/ That series lets me boot the v6.1-rc3 kernel. I have only tried it with my config (as sent earlier). If there are other suspect config settings that I should test, please let me know? I used device_initcall, but postcore_initcall also works fine. I rephrased Vlastimil's explanation and used it in patch 3 description. I have not referenced a Fixes tag yet (unsure which/if any I should use) Cheers, -- John Thomson
On Thu, Nov 03, 2022 at 07:45:49AM +0000, John Thomson wrote: > On Thu, 3 Nov 2022, at 07:18, Feng Tang wrote: > > On Wed, Nov 02, 2022 at 04:16:06PM +0900, Hyeonggon Yoo wrote: > >> On Wed, Nov 02, 2022 at 02:08:09PM +0800, Feng Tang wrote: > > [...] > >> > > transfer started ......................................... transfer ok, time=2.11s > >> > > setting up elf image... OK > >> > > jumping to kernel code > >> > > zimage at: 80B842A0 810B4BC0 > >> > > > >> > > Uncompressing Linux at load address 80001000 > >> > > > >> > > Copy device tree to address 80B80EE0 > >> > > > >> > > Now, booting the kernel... > >> > > > >> > > [ 0.000000] Linux version 6.1.0-rc3+ (john@john) (mipsel-buildroot-linux-gnu-gcc.br_real (Buildroot 2021.11-4428-g6b6741b) 12.2.0, GNU ld (GNU Binutils) 2.39) #73 SMP Wed Nov 2 05:10:01 AEST 2022 > >> > > [ 0.000000] ------------[ cut here ]------------ > >> > > [ 0.000000] WARNING: CPU: 0 PID: 0 at mm/slub.c:3416 kmem_cache_alloc+0x5a4/0x5e8 > >> > > [ 0.000000] Modules linked in: > >> > > [ 0.000000] CPU: 0 PID: 0 Comm: swapper Not tainted 6.1.0-rc3+ #73 > >> > > [ 0.000000] Stack : 810fff78 80084d98 00000000 00000004 00000000 00000000 80889d04 80c90000 > >> > > [ 0.000000] 80920000 807bd328 8089d368 80923bd3 00000000 00000001 80889cb0 00000000 > >> > > [ 0.000000] 00000000 00000000 807bd328 8084bcb1 00000002 00000002 00000001 6d6f4320 > >> > > [ 0.000000] 00000000 80c97d3d 80c97d68 fffffffc 807bd328 00000000 00000000 00000000 > >> > > [ 0.000000] 00000000 a0000000 80910000 8110a0b4 00000000 00000020 80010000 80010000 > >> > > [ 0.000000] ... > >> > > [ 0.000000] Call Trace: > >> > > [ 0.000000] [<80008260>] show_stack+0x28/0xf0 > >> > > [ 0.000000] [<8070c958>] dump_stack_lvl+0x60/0x80 > >> > > [ 0.000000] [<8002e184>] __warn+0xc4/0xf8 > >> > > [ 0.000000] [<8002e210>] warn_slowpath_fmt+0x58/0xa4 > >> > > [ 0.000000] [<801c0fac>] kmem_cache_alloc+0x5a4/0x5e8 > >> > > [ 0.000000] [<8092856c>] prom_soc_init+0x1fc/0x2b4 > >> > > [ 0.000000] [<80928060>] prom_init+0x44/0xf0 > >> > > [ 0.000000] [<80929214>] setup_arch+0x4c/0x6a8 > >> > > [ 0.000000] [<809257e0>] start_kernel+0x88/0x7c0 > >> > > [ 0.000000] > >> > > [ 0.000000] ---[ end trace 0000000000000000 ]--- > >> > > [ 0.000000] SoC Type: MediaTek MT7621 ver:1 eco:3 > >> > > [ 0.000000] printk: bootconsole [early0] enabled > >> > > > >> > > Thank you for working through this with me. > >> > > I will try to address the root cause in mt7621.c. > >> > > It looks like other arch/** soc_device_register users use postcore_initcall, device_initcall, > >> > > or the ARM DT_MACHINE_START .init_machine. A quick hack to use postcore_initcall in mt7621 > >> > > avoided this zero ptr kmem_cache passed to kmem_cache_alloc_lru. > >> > > >> > If IIUC, the prom_soc_init() is only called once in kernel, can the > >> > 'soc_dev_attr' just be defined as a global data structure instead > >> > of calling kzalloc(), as its size is small only containing 7 pointers. > >> > >> But soc_device_registers() too uses kmalloc. I think calling it > >> after slab initialization will be best solution - if that is correct. > > > > Yes, you are right, there is other kmalloc() down the call chain. > > > > Hi John, > > > > Will you verify and submit a patch for your proposal of deferring > > calling prom_soc_init()? thanks > > > > - Feng > > Hi Feng, > > My proposed mt7621.c changes are RFC here: > https://lore.kernel.org/lkml/20221103050538.1930758-1-git@johnthomson.fastmail.com.au/ Great! > That series lets me boot the v6.1-rc3 kernel. I have only tried it with my config (as sent earlier). If there are other suspect config settings that I should test, please let me know? > I used device_initcall, but postcore_initcall also works fine. I'm not sure which order is better, due to lack of mips platform knowledge. > I rephrased Vlastimil's explanation and used it in patch 3 description. > I have not referenced a Fixes tag yet (unsure which/if any I should use) With older version, the kernel boots fine with soc_dev_init() not being actually called, and I don't know if they also need to get this called. Thanks, Feng > Cheers, > -- > John Thomson >
On Tue, 1 Nov 2022, at 00:18, John Thomson wrote:
> I may have got lucky. it appears as though this is all I need to boot:
> (against 6.1-rc3), and with the Bootlin toolchain. Will test my other
> build system as well.
>
> --- a/mm/slub.c
> +++ b/mm/slub.c
> @@ -3276,7 +3276,7 @@ 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, orig_size);
> + p = ___slab_alloc(s, gfpflags, node, addr, c, 0);
> #ifdef CONFIG_PREEMPT_COUNT
> slub_put_cpu_ptr(s->cpu_slab);
> #endif
Tested that with and without SLUB_DEBUG
Testing without SLUB_DEBUG below:
With this change on 6.1-rc3:
diff --git a/mm/slub.c b/mm/slub.c
index 157527d7101b..5fdb7609bb9e 100644
--- a/mm/slub.c
+++ b/mm/slub.c
@@ -3410,6 +3410,8 @@ static __always_inline
void *__kmem_cache_alloc_lru(struct kmem_cache *s, struct list_lru *lru,
gfp_t gfpflags)
{
+
+ pr_warn("SLUB: __slab_alloc from slab_alloc s->object_size=%d\n", s->object_size);
void *ret = slab_alloc(s, lru, gfpflags, _RET_IP_, s->object_size);
trace_kmem_cache_alloc(_RET_IP_, ret, s, gfpflags, NUMA_NO_NODE);
UART & boot stops here:
transfer started ......................................... transfer ok, time=2.00s
setting up elf image... OK
jumping to kernel code
zimage at: 80B842A0 810B4BE4
Uncompressing Linux at load address 80001000
Copy device tree to address 80B80EE0
Now, booting the kernel...
[ 0.000000] Linux version 6.1.0-rc3+ (john@john) (mipsel-buildroot-linux-gnu-gcc.br_real (Buildroot 2021.11-4428-g6b6741b) 12.2.0, GNU ld (GNU Binutils) 2.39) #34 SMP Tue Nov 1 12:33:10 AEST 2022
[ 0.000000] Overriding previously set SMP ops
[ 0.000000] SoC Type: MediaTek MT7621 ver:1 eco:3
[ 0.000000] printk: bootconsole [early0] enabled
[ 0.000000] CPU0 revision is: 0001992f (MIPS 1004Kc)
[ 0.000000] MIPS: machine is MikroTik RouterBOARD 760iGS
[ 0.000000] Initrd not found or empty - disabling initrd
[ 0.000000] VPE topology {2,2} total 4
[ 0.000000] Primary instruction cache 32kB, VIPT, 4-way, linesize 32 bytes.
[ 0.000000] Primary data cache 32kB, 4-way, PIPT, no aliases, linesize 32 bytes
[ 0.000000] MIPS secondary cache 256kB, 8-way, linesize 32 bytes.
[ 0.000000] Zone ranges:
[ 0.000000] Normal [mem 0x0000000000000000-0x000000000fffffff]
[ 0.000000] HighMem empty
[ 0.000000] Movable zone start for each node
[ 0.000000] Early memory node ranges
[ 0.000000] node 0: [mem 0x0000000000000000-0x000000000fffffff]
[ 0.000000] Initmem setup node 0 [mem 0x0000000000000000-0x000000000fffffff]
[ 0.000000] percpu: Embedded 11 pages/cpu s16064 r8192 d20800 u45056
[ 0.000000] Built 1 zonelists, mobility grouping on. Total pages: 64960
[ 0.000000] Kernel command line: earlyprintk=ttyS0,115200 console=ttyS0,115200 rootfstype=squashfs,jffs2
[ 0.000000] Unknown kernel command line parameters "earlyprintk=ttyS0,115200", will be passed to user space.
[ 0.000000] Dentry cache hash table entries: 32768 (order: 5, 131072 bytes, linear)
[ 0.000000] Inode-cache hash table entries: 16384 (order: 4, 65536 bytes, linear)
[ 0.000000] Writing ErrCtl register=00011146
[ 0.000000] Readback ErrCtl register=00011146
[ 0.000000] mem auto-init: stack:off, heap alloc:off, heap free:off
[ 0.000000] Memory: 246284K/262144K available (7417K kernel code, 630K rwdata, 1304K rodata, 3500K init, 245K bss, 15860K reserved, 0K cma-reserved, 0K highmem)
[ 0.000000] SLUB: __slab_alloc from slab_alloc s->object_size=132
[ 0.000000] SLUB: __slab_alloc from slab_alloc s->object_size=132
[ 0.000000] SLUB: __slab_alloc from slab_alloc s->object_size=132
[ 0.000000] SLUB: __slab_alloc from slab_alloc s->object_size=132
[ 0.000000] SLUB: __slab_alloc from slab_alloc s->object_size=132
[ 0.000000] SLUB: __slab_alloc from slab_alloc s->object_size=132
[ 0.000000] SLUB: __slab_alloc from slab_alloc s->object_size=132
[ 0.000000] SLUB: __slab_alloc from slab_alloc s->object_size=132
[ 0.000000] SLUB: __slab_alloc from slab_alloc s->object_size=132
[ 0.000000] SLUB: __slab_alloc from slab_alloc s->object_size=132
[ 0.000000] SLUB: __slab_alloc from slab_alloc s->object_size=132
[ 0.000000] SLUB: __slab_alloc from slab_alloc s->object_size=132
[ 0.000000] SLUB: __slab_alloc from slab_alloc s->object_size=132
[ 0.000000] SLUB: __slab_alloc from slab_alloc s->object_size=132
[ 0.000000] SLUB: __slab_alloc from slab_alloc s->object_size=132
[ 0.000000] SLUB: __slab_alloc from slab_alloc s->object_size=132
[ 0.000000] SLUB: __slab_alloc from slab_alloc s->object_size=132
[ 0.000000] SLUB: __slab_alloc from slab_alloc s->object_size=132
[ 0.000000] SLUB: __slab_alloc from slab_alloc s->object_size=132
[ 0.000000] SLUB: __slab_alloc from slab_alloc s->object_size=132
[ 0.000000] SLUB: __slab_alloc from slab_alloc s->object_size=132
[ 0.000000] SLUB: __slab_alloc from slab_alloc s->object_size=132
[ 0.000000] SLUB: __slab_alloc from slab_alloc s->object_size=132
[ 0.000000] SLUB: HWalign=32, Order=0-3, MinObjects=0, CPUs=4, Nodes=1
[ 0.000000] SLUB: __slab_alloc from slab_alloc s->object_size=132
[ 0.000000] SLUB: __slab_alloc from slab_alloc s->object_size=32
[ 0.000000] SLUB: __slab_alloc from slab_alloc s->object_size=132
[ 0.000000] SLUB: __slab_alloc from slab_alloc s->object_size=128
[ 0.000000] SLUB: __slab_alloc from slab_alloc s->object_size=132
[ 0.000000] SLUB: __slab_alloc from slab_alloc s->object_size=132
[ 0.000000] SLUB: __slab_alloc from slab_alloc s->object_size=128
[ 0.000000] SLUB: __slab_alloc from slab_alloc s->object_size=128
[ 0.000000] SLUB: __slab_alloc from slab_alloc s->object_size=300
[ 0.000000] SLUB: __slab_alloc from slab_alloc s->object_size=128
[ 0.000000] SLUB: __slab_alloc from slab_alloc s->object_size=128
[ 0.000000] SLUB: __slab_alloc from slab_alloc s->object_size=128
[ 0.000000] SLUB: __slab_alloc from slab_alloc s->object_size=128
[ 0.000000] SLUB: __slab_alloc from slab_alloc s->object_size=128
[ 0.000000] SLUB: __slab_alloc from slab_alloc s->object_size=128
[ 0.000000] SLUB: __slab_alloc from slab_alloc s->object_size=128
[ 0.000000] SLUB: __slab_alloc from slab_alloc s->object_size=128
[ 0.000000] SLUB: __slab_alloc from slab_alloc s->object_size=128
[ 0.000000] SLUB: __slab_alloc from slab_alloc s->object_size=128
[ 0.000000] SLUB: __slab_alloc from slab_alloc s->object_size=128
[ 0.000000] SLUB: __slab_alloc from slab_alloc s->object_size=128
[ 0.000000] SLUB: __slab_alloc from slab_alloc s->object_size=128
[ 0.000000] SLUB: __slab_alloc from slab_alloc s->object_size=128
[ 0.000000] SLUB: __slab_alloc from slab_alloc s->object_size=128
[ 0.000000] SLUB: __slab_alloc from slab_alloc s->object_size=128
[ 0.000000] SLUB: __slab_alloc from slab_alloc s->object_size=128
[ 0.000000] rcu: Hierarchical RCU implementation.
[ 0.000000] Tracing variant of Tasks RCU enabled.
[ 0.000000] rcu: RCU calculated value of scheduler-enlistment delay is 10 jiffies.
[ 0.000000] NR_IRQS: 256
[ 0.000000] SLUB: __slab_alloc from slab_alloc s->object_size=128
[ 0.000000] SLUB: __slab_alloc from slab_alloc s->object_size=128
[ 0.000000] SLUB: __slab_alloc from slab_alloc s->object_size=128
[ 0.000000] rcu: srcu_init: Setting srcu_struct sizes based on contention.
[ 0.000000] SLUB: __slab_alloc from slab_alloc s->object_size=128
[ 0.000000] SLUB: __slab_alloc from slab_alloc s->object_size=128
[ 0.000000] SLUB: __slab_alloc from slab_alloc s->object_size=128
[ 0.000000] SLUB: __slab_alloc from slab_alloc s->object_size=128
[ 0.000000] SLUB: __slab_alloc from slab_alloc s->object_size=128
[ 0.000000] SLUB: __slab_alloc from slab_alloc s->object_size=128
[ 0.000000] SLUB: __slab_alloc from slab_alloc s->object_size=512
[ 0.000000] SLUB: __slab_alloc from slab_alloc s->object_size=128
[ 0.000000] SLUB: __slab_alloc from slab_alloc s->object_size=128
[ 0.000000] SLUB: __slab_alloc from slab_alloc s->object_size=128
[ 0.000000] SLUB: __slab_alloc from slab_alloc s->object_size=512
[ 0.000000] SLUB: __slab_alloc from slab_alloc s->object_size=128
[ 0.000000] SLUB: __slab_alloc from slab_alloc s->object_size=128
[ 0.000000] SLUB: __slab_alloc from slab_alloc s->object_size=256
[ 0.000000] SLUB: __slab_alloc from slab_alloc s->object_size=128
[ 0.000000] SLUB: __slab_alloc from slab_alloc s->object_size=256
[ 0.000000] SLUB: __slab_alloc from slab_alloc s->object_size=128
[ 0.000000] SLUB: __slab_alloc from slab_alloc s->object_size=256
[ 0.000000] SLUB: __slab_alloc from slab_alloc s->object_size=128
[ 0.000000] SLUB: __slab_alloc from slab_alloc s->object_size=128
[ 0.000000] SLUB: __slab_alloc from slab_alloc s->object_size=128
[ 0.000000] SLUB: __slab_alloc from slab_alloc s->object_size=128
[ 0.000000] clocksource: GIC: mask: 0xffffffffffffffff max_cycles: 0xcaf478abb4, max_idle_ns: 440795247997 ns
[ 0.000000] SLUB: __slab_alloc from slab_alloc s->object_size=256
[ 0.000004] sched_clock: 64 bits at 880MHz, resolution 1ns, wraps every 4398046511103ns
This change, and kernel boots fine:
diff --git a/mm/slub.c b/mm/slub.c
index 157527d7101b..e9677c04d19c 100644
--- a/mm/slub.c
+++ b/mm/slub.c
@@ -3410,7 +3410,11 @@ static __always_inline
void *__kmem_cache_alloc_lru(struct kmem_cache *s, struct list_lru *lru,
gfp_t gfpflags)
{
- void *ret = slab_alloc(s, lru, gfpflags, _RET_IP_, s->object_size);
+
+ /*
+ pr_warn("SLUB: __slab_alloc from slab_alloc s->object_size=%d\n", s->object_size);
+ void *ret = slab_alloc(s, lru, gfpflags, _RET_IP_, s->object_size);*/
+ void *ret = slab_alloc(s, lru, gfpflags, _RET_IP_, 0);
trace_kmem_cache_alloc(_RET_IP_, ret, s, gfpflags, NUMA_NO_NODE);
Cheers,
--
John Thomson
On Mon, Oct 31, 2022 at 10:05:58AM +0000, John Thomson wrote: > On Mon, 31 Oct 2022, at 02:36, Feng Tang wrote: > > Hi John, > > > > Thanks for the bisecting and reporting! > > > > On Mon, Oct 31, 2022 at 05:30:24AM +0800, Vlastimil Babka wrote: > >> On 10/30/22 20:23, John Thomson wrote: > >> > On Tue, 13 Sep 2022, at 06:54, Feng Tang wrote: > >> >> 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. > >> > > >> > > >> > I have a ralink mt7621 router running Openwrt, using the mips ZBOOT kernel, and appear to have bisected > >> > a very-nearly-clean kernel v6.1rc-2 boot issue to this commit. > >> > I have 3 commits atop 6.1-rc2: fix a ZBOOT compile error, use the Openwrt LZMA options, > >> > and enable DEBUG_ZBOOT for my platform. I am compiling my kernel within the Openwrt build system. > >> > No guarantees this is not due to something I am doing wrong, but any insight would be greatly appreciated. > >> > > >> > > >> > On UART, No indication of the (once extracted) kernel booting: > >> > > >> > transfer started ......................................... transfer ok, time=2.01s > >> > setting up elf image... OK > >> > jumping to kernel code > >> > zimage at: 80BA4100 810D4720 > >> > Uncompressing Linux at load address 80001000 > >> > Copy device tree to address 80B96EE0 > >> > Now, booting the kernel... > >> > >> It's weird that the commit would cause no output so early, SLUB code is > >> run only later. > > > > I noticed your cmdline has console setting, could you enable the > > earlyprintk in cmdline like "earlyprintk=ttyS0,115200" etc to see > > if there is more message printed out. > > Still nothing from vmlinux with earlykprint on UART unless revert. > > > > > Also I want to confirm this is a boot failure and not only a boot > > message missing. > > Yes, boot failure. > Network comes up automatically on successful boot. Not happening when no kernel UART It is really weird that I see no boot issue on my MIPS emulation with almost same config, with different target - Malta board that QEMU supports. it just boot fine. Can you attach debugger to the board? (Which I hadn't tried. I had tried it only to QEMU) [...] > >> > > >> > > >> > possibly relevant config options: > >> > grep -E '(SLUB|SLAB)' .config > >> > # SLAB allocator options > >> > # CONFIG_SLAB is not set > >> > CONFIG_SLUB=y > >> > CONFIG_SLAB_MERGE_DEFAULT=y > >> > # CONFIG_SLAB_FREELIST_RANDOM is not set > >> > # CONFIG_SLAB_FREELIST_HARDENED is not set > >> > # CONFIG_SLUB_STATS is not set > >> > CONFIG_SLUB_CPU_PARTIAL=y > >> > # end of SLAB allocator options > >> > # CONFIG_SLUB_DEBUG is not set > >> > >> Also not having CONFIG_SLUB_DEBUG enabled means most of the code the > >> patch/commit touches is not even active. > >> Could this be some miscompile or code layout change exposing some > >> different bug, hmm. > > Yes, it could be. What happens with clang? > > >> Is it any different if you do enable CONFIG_SLUB_DEBUG ? > > No change > > >> Or change to CONFIG_SLAB? (that would be really weird if not) > > This boots fine > > > I haven't found any clue from the code either, and I compiled > > kernel with the config above and tested booting on an Alder-lake > > desktop and a QEMU, which boot fine. > > > > Could you provide the full kernel config and demsg (in compressed > > format if you think it's too big), so we can check more? > > Attached > > > Thanks, > > Feng > > vmlinux is bigger, and entry point is larger (0x8074081c vs 0x807407dc revert vs 0x8073fcbc), > so that may be it? Or not? > revert + SLUB_DEBUG + SLUB_DEBUG_ON is bigger still, but does successfully boot. > vmlinux entry point is 0x8074705c > > > transfer started ......................................... transfer ok, time=2.01s > setting up elf image... OK > jumping to kernel code > zimage at: 80BA4100 810D6FA0 > Uncompressing Linux at load address 80001000 > Copy device tree to address 80B9EEE0 > Now, booting the kernel... > [ 0.000000] Linux version 6.1.0-rc2 (john@john) (mipsel-openwrt-linux-musl-gc > c (OpenWrt GCC 11.3.0 r19724+16-1521d5f453) 11.3.0, GNU ld (GNU Binutils) 2.37) > #0 SMP Fri Oct 28 03:48:10 2022 > > > I will keep looking. > > Thank you, > -- > John Thomson -- Thanks, Hyeonggon
On 9/13/22 08:54, Feng Tang wrote:
> 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>
Thanks.
Given that the merge window is nearing, and the rest of the series a) has
some changes suggested and b) could be hopefully done in a simpler way with
the proposed ksize() cleanup, I am picking just this patch now to slab.git
(and thus -next), with some small modifications:
...
> +
> +static unsigned int get_orig_size(struct kmem_cache *s, void *object)
Made this inline for consistency.
> +{
> + 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 @@ static int check_bytes_and_report(struct kmem_cache *s, struct slab *slab,
> *
> * 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 +1625,10 @@ 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 set_orig_size(struct kmem_cache *s,
> + void *object, unsigned int orig_size) {}
There's no caller (in this patch alone) for the !SLUB_DEBUG version, so removed.
> static inline void free_debug_processing(
> struct kmem_cache *s, struct slab *slab,
> @@ -1999,7 +2057,7 @@ 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;
>
> @@ -2009,7 +2067,7 @@ 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;
> }
> @@ -2028,7 +2086,7 @@ 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);
> @@ -2040,7 +2098,7 @@ 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
> @@ -2118,7 +2176,7 @@ 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;
> @@ -2138,11 +2196,12 @@ 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;
> @@ -2153,7 +2212,7 @@ 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 {
> @@ -2177,14 +2236,13 @@ 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;
>
> @@ -2212,15 +2270,15 @@ 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()
> @@ -2241,8 +2299,7 @@ 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;
> @@ -2250,11 +2307,11 @@ 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
> @@ -2974,11 +3031,12 @@ 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);
>
> @@ -3092,7 +3150,10 @@ static void *___slab_alloc(struct kmem_cache *s, gfp_t gfpflags, int node,
>
> 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;
>
> @@ -3108,7 +3169,7 @@ static void *___slab_alloc(struct kmem_cache *s, gfp_t gfpflags, int node,
> 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;
> @@ -3140,6 +3201,7 @@ static void *___slab_alloc(struct kmem_cache *s, gfp_t gfpflags, int node,
> */
> if (s->flags & SLAB_STORE_USER)
> set_track(s, freelist, TRACK_ALLOC, addr);
> +
> return freelist;
> }
>
> @@ -3182,7 +3244,7 @@ static void *___slab_alloc(struct kmem_cache *s, gfp_t gfpflags, int node,
> * 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;
>
> @@ -3195,7 +3257,7 @@ 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
> @@ -3280,7 +3342,7 @@ static __always_inline void *slab_alloc_node(struct kmem_cache *s, struct list_l
>
> 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);
>
> @@ -3747,7 +3809,7 @@ 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;
>
> @@ -4150,12 +4212,17 @@ 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);
> @@ -4770,7 +4837,7 @@ void __init kmem_cache_init(void)
>
> /* Now we can use the kmem_cache to allocate kmalloc slabs */
> setup_kmalloc_cache_index_table();
> - create_kmalloc_caches(0);
> + create_kmalloc_caches(SLAB_KMALLOC);
Instead of this, add the flag in the common creation function, so SLAB kmalloc caches are also marked even if there's no use for it there now.
--- a/mm/slab_common.c
+++ b/mm/slab_common.c
@@ -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;
> /* Setup random freelists for each cache */
> init_freelist_randomization();
> @@ -4937,6 +5004,7 @@ 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;
> @@ -4983,13 +5051,15 @@ 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);
> @@ -5007,11 +5077,13 @@ 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;
> @@ -5036,6 +5108,9 @@ 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;
> }
> @@ -5059,6 +5134,7 @@ 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);
> @@ -5077,7 +5153,7 @@ static void process_slab(struct loc_track *t, struct kmem_cache *s,
>
> 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), get_orig_size(s, p));
I think it makes little sense to report waste in the 'free_traces' file?
So adjusted like this to make sure nothing is reported there:
@@ -5356,13 +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), get_orig_size(s, p));
+ 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 */
> @@ -5942,6 +6018,10 @@ 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),
On Fri, Sep 23, 2022 at 07:43:22PM +0800, Vlastimil Babka wrote:
> On 9/13/22 08:54, Feng Tang wrote:
[...]
> > 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>
>
> Thanks.
> Given that the merge window is nearing, and the rest of the series a) has
> some changes suggested and b) could be hopefully done in a simpler way with
> the proposed ksize() cleanup, I am picking just this patch now to slab.git
> (and thus -next), with some small modifications:
OK, and all the cleanup/improvments from you look good to me. Many thanks!
For kasan and ksize() related patches, I'll keep monitoring and working on
them.
- Feng
> ...
>
> > +
> > +static unsigned int get_orig_size(struct kmem_cache *s, void *object)
>
> Made this inline for consistency.
>
> > +{
> > + 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 @@ static int check_bytes_and_report(struct kmem_cache *s, struct slab *slab,
> > *
> > * 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 +1625,10 @@ 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 set_orig_size(struct kmem_cache *s,
> > + void *object, unsigned int orig_size) {}
>
> There's no caller (in this patch alone) for the !SLUB_DEBUG version, so removed.
>
> > static inline void free_debug_processing(
> > struct kmem_cache *s, struct slab *slab,
> > @@ -1999,7 +2057,7 @@ 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;
> >
> > @@ -2009,7 +2067,7 @@ 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;
> > }
> > @@ -2028,7 +2086,7 @@ 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);
> > @@ -2040,7 +2098,7 @@ 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
> > @@ -2118,7 +2176,7 @@ 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;
> > @@ -2138,11 +2196,12 @@ 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;
> > @@ -2153,7 +2212,7 @@ 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 {
> > @@ -2177,14 +2236,13 @@ 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;
> >
> > @@ -2212,15 +2270,15 @@ 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()
> > @@ -2241,8 +2299,7 @@ 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;
> > @@ -2250,11 +2307,11 @@ 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
> > @@ -2974,11 +3031,12 @@ 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);
> >
> > @@ -3092,7 +3150,10 @@ static void *___slab_alloc(struct kmem_cache *s, gfp_t gfpflags, int node,
> >
> > 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;
> >
> > @@ -3108,7 +3169,7 @@ static void *___slab_alloc(struct kmem_cache *s, gfp_t gfpflags, int node,
> > 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;
> > @@ -3140,6 +3201,7 @@ static void *___slab_alloc(struct kmem_cache *s, gfp_t gfpflags, int node,
> > */
> > if (s->flags & SLAB_STORE_USER)
> > set_track(s, freelist, TRACK_ALLOC, addr);
> > +
> > return freelist;
> > }
> >
> > @@ -3182,7 +3244,7 @@ static void *___slab_alloc(struct kmem_cache *s, gfp_t gfpflags, int node,
> > * 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;
> >
> > @@ -3195,7 +3257,7 @@ 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
> > @@ -3280,7 +3342,7 @@ static __always_inline void *slab_alloc_node(struct kmem_cache *s, struct list_l
> >
> > 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);
> >
> > @@ -3747,7 +3809,7 @@ 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;
> >
> > @@ -4150,12 +4212,17 @@ 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);
> > @@ -4770,7 +4837,7 @@ void __init kmem_cache_init(void)
> >
> > /* Now we can use the kmem_cache to allocate kmalloc slabs */
> > setup_kmalloc_cache_index_table();
> > - create_kmalloc_caches(0);
> > + create_kmalloc_caches(SLAB_KMALLOC);
>
> Instead of this, add the flag in the common creation function, so SLAB kmalloc caches are also marked even if there's no use for it there now.
>
> --- a/mm/slab_common.c
> +++ b/mm/slab_common.c
> @@ -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;
>
>
> > /* Setup random freelists for each cache */
> > init_freelist_randomization();
> > @@ -4937,6 +5004,7 @@ 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;
> > @@ -4983,13 +5051,15 @@ 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);
> > @@ -5007,11 +5077,13 @@ 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;
> > @@ -5036,6 +5108,9 @@ 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;
> > }
> > @@ -5059,6 +5134,7 @@ 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);
> > @@ -5077,7 +5153,7 @@ static void process_slab(struct loc_track *t, struct kmem_cache *s,
> >
> > 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), get_orig_size(s, p));
>
> I think it makes little sense to report waste in the 'free_traces' file?
> So adjusted like this to make sure nothing is reported there:
>
> @@ -5356,13 +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), get_orig_size(s, p));
> + 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 */
> > @@ -5942,6 +6018,10 @@ 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),
>
>
© 2016 - 2026 Red Hat, Inc.