This patch adds a new directory called 'page_owner_stacks' under
/sys/kernel/debug/, with a file called 'show_stacks' in it.
Reading from that file will show all stacks that were added by page_owner
followed by their counting, giving us a clear overview of stack <-> count
relationship.
E.g:
prep_new_page+0xa9/0x120
get_page_from_freelist+0x801/0x2210
__alloc_pages+0x18b/0x350
alloc_pages_mpol+0x91/0x1f0
folio_alloc+0x14/0x50
filemap_alloc_folio+0xb2/0x100
__filemap_get_folio+0x14a/0x490
ext4_write_begin+0xbd/0x4b0 [ext4]
generic_perform_write+0xc1/0x1e0
ext4_buffered_write_iter+0x68/0xe0 [ext4]
ext4_file_write_iter+0x70/0x740 [ext4]
vfs_write+0x33d/0x420
ksys_write+0xa5/0xe0
do_syscall_64+0x80/0x160
entry_SYSCALL_64_after_hwframe+0x6e/0x76
stack_count: 4578
The seq stack_{start,next} functions will iterate through the list
stack_list in order to print all stacks.
Signed-off-by: Oscar Salvador <osalvador@suse.de>
---
mm/page_owner.c | 99 ++++++++++++++++++++++++++++++++++++++++++++++++-
1 file changed, 98 insertions(+), 1 deletion(-)
diff --git a/mm/page_owner.c b/mm/page_owner.c
index 7d1b3f75cef3..3e4b7cd7c8f8 100644
--- a/mm/page_owner.c
+++ b/mm/page_owner.c
@@ -84,7 +84,12 @@ static void add_stack_record_to_list(struct stack_record *stack_record)
stack_list = stack;
} else {
stack->next = stack_list;
- stack_list = stack;
+ /* This pairs with smp_load_acquire() from function
+ * stack_start(). This guarantees that stack_start()
+ * will see an updated stack_list before starting to
+ * traverse the list.
+ */
+ smp_store_release(&stack_list, stack);
}
spin_unlock_irqrestore(&stack_list_lock, flags);
}
@@ -792,8 +797,97 @@ static const struct file_operations proc_page_owner_operations = {
.llseek = lseek_page_owner,
};
+static void *stack_start(struct seq_file *m, loff_t *ppos)
+{
+ struct stack *stack;
+
+ if (*ppos == -1UL)
+ return NULL;
+
+ if (!*ppos) {
+ /*
+ * This pairs with smp_store_release() from function
+ * add_stack_record_to_list(), so we get a consistent
+ * value of stack_list.
+ */
+ stack = smp_load_acquire(&stack_list);
+ } else {
+ stack = m->private;
+ stack = stack->next;
+ }
+
+ m->private = stack;
+
+ return stack;
+}
+
+static void *stack_next(struct seq_file *m, void *v, loff_t *ppos)
+{
+ struct stack *stack = v;
+
+ stack = stack->next;
+ *ppos = stack ? *ppos + 1 : -1UL;
+ m->private = stack;
+
+ return stack;
+}
+
+static int stack_print(struct seq_file *m, void *v)
+{
+ char *buf;
+ int ret = 0;
+ struct stack *stack = v;
+ struct stack_record *stack_record = stack->stack_record;
+
+ if (!stack_record->size || stack_record->size < 0 ||
+ refcount_read(&stack_record->count) < 2)
+ return 0;
+
+ buf = kzalloc(PAGE_SIZE, GFP_KERNEL);
+
+ ret += stack_trace_snprint(buf, PAGE_SIZE, stack_record->entries,
+ stack_record->size, 0);
+ if (!ret)
+ goto out;
+
+ scnprintf(buf + ret, PAGE_SIZE - ret, "stack_count: %d\n\n",
+ refcount_read(&stack_record->count));
+
+ seq_printf(m, buf);
+ seq_puts(m, "\n\n");
+out:
+ kfree(buf);
+
+ return 0;
+}
+
+static void stack_stop(struct seq_file *m, void *v)
+{
+}
+
+static const struct seq_operations page_owner_stack_op = {
+ .start = stack_start,
+ .next = stack_next,
+ .stop = stack_stop,
+ .show = stack_print
+};
+
+static int page_owner_stack_open(struct inode *inode, struct file *file)
+{
+ return seq_open_private(file, &page_owner_stack_op, 0);
+}
+
+static const struct file_operations page_owner_stack_operations = {
+ .open = page_owner_stack_open,
+ .read = seq_read,
+ .llseek = seq_lseek,
+ .release = seq_release,
+};
+
static int __init pageowner_init(void)
{
+ struct dentry *dir;
+
if (!static_branch_unlikely(&page_owner_inited)) {
pr_info("page_owner is disabled\n");
return 0;
@@ -801,6 +895,9 @@ static int __init pageowner_init(void)
debugfs_create_file("page_owner", 0400, NULL, NULL,
&proc_page_owner_operations);
+ dir = debugfs_create_dir("page_owner_stacks", NULL);
+ debugfs_create_file("show_stacks", 0400, dir, NULL,
+ &page_owner_stack_operations);
return 0;
}
--
2.43.0On 2/12/24 23:30, Oscar Salvador wrote:
> This patch adds a new directory called 'page_owner_stacks' under
> /sys/kernel/debug/, with a file called 'show_stacks' in it.
> Reading from that file will show all stacks that were added by page_owner
> followed by their counting, giving us a clear overview of stack <-> count
> relationship.
>
> E.g:
>
> prep_new_page+0xa9/0x120
> get_page_from_freelist+0x801/0x2210
> __alloc_pages+0x18b/0x350
> alloc_pages_mpol+0x91/0x1f0
> folio_alloc+0x14/0x50
> filemap_alloc_folio+0xb2/0x100
> __filemap_get_folio+0x14a/0x490
> ext4_write_begin+0xbd/0x4b0 [ext4]
> generic_perform_write+0xc1/0x1e0
> ext4_buffered_write_iter+0x68/0xe0 [ext4]
> ext4_file_write_iter+0x70/0x740 [ext4]
> vfs_write+0x33d/0x420
> ksys_write+0xa5/0xe0
> do_syscall_64+0x80/0x160
> entry_SYSCALL_64_after_hwframe+0x6e/0x76
> stack_count: 4578
>
> The seq stack_{start,next} functions will iterate through the list
> stack_list in order to print all stacks.
>
> Signed-off-by: Oscar Salvador <osalvador@suse.de>
...
> +static int stack_print(struct seq_file *m, void *v)
> +{
> + char *buf;
> + int ret = 0;
> + struct stack *stack = v;
> + struct stack_record *stack_record = stack->stack_record;
> +
> + if (!stack_record->size || stack_record->size < 0 ||
> + refcount_read(&stack_record->count) < 2)
> + return 0;
> +
> + buf = kzalloc(PAGE_SIZE, GFP_KERNEL);
> +
> + ret += stack_trace_snprint(buf, PAGE_SIZE, stack_record->entries,
> + stack_record->size, 0);
> + if (!ret)
> + goto out;
> +
> + scnprintf(buf + ret, PAGE_SIZE - ret, "stack_count: %d\n\n",
> + refcount_read(&stack_record->count));
> +
> + seq_printf(m, buf);
> + seq_puts(m, "\n\n");
> +out:
> + kfree(buf);
Seems rather wasteful to do kzalloc/kfree so you can print into that buffer
first and then print/copy it again using seq_printf. If you give up on using
stack_trace_snprintf() it's not much harder to print the stack directly with
a loop of seq_printf. See e.g. slab_debugfs_show().
> +
> + return 0;
> +}
> +
On Tue, Feb 13, 2024 at 03:25:26PM +0100, Vlastimil Babka wrote:
> On 2/12/24 23:30, Oscar Salvador wrote:
> > +static int stack_print(struct seq_file *m, void *v)
> > +{
> > + char *buf;
> > + int ret = 0;
> > + struct stack *stack = v;
> > + struct stack_record *stack_record = stack->stack_record;
> > +
> > + if (!stack_record->size || stack_record->size < 0 ||
> > + refcount_read(&stack_record->count) < 2)
> > + return 0;
> > +
> > + buf = kzalloc(PAGE_SIZE, GFP_KERNEL);
> > +
> > + ret += stack_trace_snprint(buf, PAGE_SIZE, stack_record->entries,
> > + stack_record->size, 0);
> > + if (!ret)
> > + goto out;
> > +
> > + scnprintf(buf + ret, PAGE_SIZE - ret, "stack_count: %d\n\n",
> > + refcount_read(&stack_record->count));
> > +
> > + seq_printf(m, buf);
> > + seq_puts(m, "\n\n");
> > +out:
> > + kfree(buf);
>
> Seems rather wasteful to do kzalloc/kfree so you can print into that buffer
> first and then print/copy it again using seq_printf. If you give up on using
> stack_trace_snprintf() it's not much harder to print the stack directly with
> a loop of seq_printf. See e.g. slab_debugfs_show().
Well, I thought about not reinventing the wheel there, but fair enough
than performing a kmalloc/free op on every print might be suboptimal.
I will try to do ir with seq_printf alone.
Thanks
--
Oscar Salvador
SUSE Labs
On 2/13/24 16:33, Oscar Salvador wrote:
> On Tue, Feb 13, 2024 at 03:25:26PM +0100, Vlastimil Babka wrote:
>> On 2/12/24 23:30, Oscar Salvador wrote:
>> > +static int stack_print(struct seq_file *m, void *v)
>> > +{
>> > + char *buf;
>> > + int ret = 0;
>> > + struct stack *stack = v;
>> > + struct stack_record *stack_record = stack->stack_record;
>> > +
>> > + if (!stack_record->size || stack_record->size < 0 ||
>> > + refcount_read(&stack_record->count) < 2)
>> > + return 0;
>> > +
>> > + buf = kzalloc(PAGE_SIZE, GFP_KERNEL);
>> > +
>> > + ret += stack_trace_snprint(buf, PAGE_SIZE, stack_record->entries,
>> > + stack_record->size, 0);
>> > + if (!ret)
>> > + goto out;
>> > +
>> > + scnprintf(buf + ret, PAGE_SIZE - ret, "stack_count: %d\n\n",
>> > + refcount_read(&stack_record->count));
>> > +
>> > + seq_printf(m, buf);
>> > + seq_puts(m, "\n\n");
>> > +out:
>> > + kfree(buf);
>>
>> Seems rather wasteful to do kzalloc/kfree so you can print into that buffer
>> first and then print/copy it again using seq_printf. If you give up on using
>> stack_trace_snprintf() it's not much harder to print the stack directly with
>> a loop of seq_printf. See e.g. slab_debugfs_show().
>
> Well, I thought about not reinventing the wheel there, but fair enough
> than performing a kmalloc/free op on every print might be suboptimal.
> I will try to do ir with seq_printf alone.
Of course once there's more than one stackdepot user printing into a
seq_file, creating a common seq_file helper analogy of
stack_trace_snprintf() and using it from all places, would also be an option :)
> Thanks
>
>
On Mon, 12 Feb 2024 at 23:29, Oscar Salvador <osalvador@suse.de> wrote:
>
> This patch adds a new directory called 'page_owner_stacks' under
> /sys/kernel/debug/, with a file called 'show_stacks' in it.
> Reading from that file will show all stacks that were added by page_owner
> followed by their counting, giving us a clear overview of stack <-> count
> relationship.
>
> E.g:
>
> prep_new_page+0xa9/0x120
> get_page_from_freelist+0x801/0x2210
> __alloc_pages+0x18b/0x350
> alloc_pages_mpol+0x91/0x1f0
> folio_alloc+0x14/0x50
> filemap_alloc_folio+0xb2/0x100
> __filemap_get_folio+0x14a/0x490
> ext4_write_begin+0xbd/0x4b0 [ext4]
> generic_perform_write+0xc1/0x1e0
> ext4_buffered_write_iter+0x68/0xe0 [ext4]
> ext4_file_write_iter+0x70/0x740 [ext4]
> vfs_write+0x33d/0x420
> ksys_write+0xa5/0xe0
> do_syscall_64+0x80/0x160
> entry_SYSCALL_64_after_hwframe+0x6e/0x76
> stack_count: 4578
>
> The seq stack_{start,next} functions will iterate through the list
> stack_list in order to print all stacks.
>
> Signed-off-by: Oscar Salvador <osalvador@suse.de>
Acked-by: Marco Elver <elver@google.com>
Minor comments below.
> ---
> mm/page_owner.c | 99 ++++++++++++++++++++++++++++++++++++++++++++++++-
> 1 file changed, 98 insertions(+), 1 deletion(-)
>
> diff --git a/mm/page_owner.c b/mm/page_owner.c
> index 7d1b3f75cef3..3e4b7cd7c8f8 100644
> --- a/mm/page_owner.c
> +++ b/mm/page_owner.c
> @@ -84,7 +84,12 @@ static void add_stack_record_to_list(struct stack_record *stack_record)
> stack_list = stack;
> } else {
> stack->next = stack_list;
> - stack_list = stack;
> + /* This pairs with smp_load_acquire() from function
Comment should be
/*
*
...
*/
(Unless in networking or other special subsystems with their own comment style.)
> + * stack_start(). This guarantees that stack_start()
> + * will see an updated stack_list before starting to
> + * traverse the list.
> + */
> + smp_store_release(&stack_list, stack);
> }
> spin_unlock_irqrestore(&stack_list_lock, flags);
> }
> @@ -792,8 +797,97 @@ static const struct file_operations proc_page_owner_operations = {
> .llseek = lseek_page_owner,
> };
>
> +static void *stack_start(struct seq_file *m, loff_t *ppos)
> +{
> + struct stack *stack;
> +
> + if (*ppos == -1UL)
> + return NULL;
> +
> + if (!*ppos) {
> + /*
> + * This pairs with smp_store_release() from function
> + * add_stack_record_to_list(), so we get a consistent
> + * value of stack_list.
> + */
> + stack = smp_load_acquire(&stack_list);
I'm not sure if it'd make your code simpler or not: there is
<linux/llist.h> for singly-linked linked lists, although the code to
manage the list is simple enough I'm indifferent here. Only consider
it if it helps you make the code simpler.
> + } else {
> + stack = m->private;
> + stack = stack->next;
> + }
> +
> + m->private = stack;
> +
> + return stack;
> +}
> +
> +static void *stack_next(struct seq_file *m, void *v, loff_t *ppos)
> +{
> + struct stack *stack = v;
> +
> + stack = stack->next;
> + *ppos = stack ? *ppos + 1 : -1UL;
> + m->private = stack;
> +
> + return stack;
> +}
> +
> +static int stack_print(struct seq_file *m, void *v)
> +{
> + char *buf;
> + int ret = 0;
> + struct stack *stack = v;
> + struct stack_record *stack_record = stack->stack_record;
> +
> + if (!stack_record->size || stack_record->size < 0 ||
> + refcount_read(&stack_record->count) < 2)
> + return 0;
> +
> + buf = kzalloc(PAGE_SIZE, GFP_KERNEL);
> +
> + ret += stack_trace_snprint(buf, PAGE_SIZE, stack_record->entries,
> + stack_record->size, 0);
> + if (!ret)
> + goto out;
> +
> + scnprintf(buf + ret, PAGE_SIZE - ret, "stack_count: %d\n\n",
> + refcount_read(&stack_record->count));
> +
> + seq_printf(m, buf);
> + seq_puts(m, "\n\n");
> +out:
> + kfree(buf);
> +
> + return 0;
> +}
> +
> +static void stack_stop(struct seq_file *m, void *v)
> +{
> +}
Is this function even needed if it's empty? I recall there were some
boilerplate "nop" functions that could be used.
> +static const struct seq_operations page_owner_stack_op = {
> + .start = stack_start,
> + .next = stack_next,
> + .stop = stack_stop,
> + .show = stack_print
> +};
> +
> +static int page_owner_stack_open(struct inode *inode, struct file *file)
> +{
> + return seq_open_private(file, &page_owner_stack_op, 0);
> +}
> +
> +static const struct file_operations page_owner_stack_operations = {
> + .open = page_owner_stack_open,
> + .read = seq_read,
> + .llseek = seq_lseek,
> + .release = seq_release,
> +};
> +
> static int __init pageowner_init(void)
> {
> + struct dentry *dir;
> +
> if (!static_branch_unlikely(&page_owner_inited)) {
> pr_info("page_owner is disabled\n");
> return 0;
> @@ -801,6 +895,9 @@ static int __init pageowner_init(void)
>
> debugfs_create_file("page_owner", 0400, NULL, NULL,
> &proc_page_owner_operations);
> + dir = debugfs_create_dir("page_owner_stacks", NULL);
> + debugfs_create_file("show_stacks", 0400, dir, NULL,
> + &page_owner_stack_operations);
>
> return 0;
> }
> --
> 2.43.0
>
On Tue, Feb 13, 2024 at 09:38:43AM +0100, Marco Elver wrote:
> On Mon, 12 Feb 2024 at 23:29, Oscar Salvador <osalvador@suse.de> wrote:
> > Signed-off-by: Oscar Salvador <osalvador@suse.de>
>
> Acked-by: Marco Elver <elver@google.com>
Thanks!
> > + /* This pairs with smp_load_acquire() from function
>
> Comment should be
>
> /*
> *
> ...
> */
Yap, fat fingers here.
> > + if (!*ppos) {
> > + /*
> > + * This pairs with smp_store_release() from function
> > + * add_stack_record_to_list(), so we get a consistent
> > + * value of stack_list.
> > + */
> > + stack = smp_load_acquire(&stack_list);
>
> I'm not sure if it'd make your code simpler or not: there is
> <linux/llist.h> for singly-linked linked lists, although the code to
> manage the list is simple enough I'm indifferent here. Only consider
> it if it helps you make the code simpler.
I will check if it eases the code somehow.
> > +static void stack_stop(struct seq_file *m, void *v)
> > +{
> > +}
>
> Is this function even needed if it's empty? I recall there were some
> boilerplate "nop" functions that could be used.
I will check if seq already provides a dummy function for these cases.
--
Oscar Salvador
SUSE Labs
© 2016 - 2026 Red Hat, Inc.