Re: [PATCH 12/15] stackdepot: add refcount for records

[Marco Elver]

On Tue, Aug 29, 2023 at 07:11PM +0200, andrey.konovalov@xxxxxxxxx wrote:
> From: Andrey Konovalov <andreyknvl@xxxxxxxxxx>
> 
> Add a reference counter for how many times a stack records has been added
> to stack depot.
> 
> Do no yet decrement the refcount, this is implemented in one of the
> following patches.
> 
> This is preparatory patch for implementing the eviction of stack records
> from the stack depot.
> 
> Signed-off-by: Andrey Konovalov <andreyknvl@xxxxxxxxxx>
> ---
>  lib/stackdepot.c | 4 ++++
>  1 file changed, 4 insertions(+)
> 
> diff --git a/lib/stackdepot.c b/lib/stackdepot.c
> index 5ad454367379..a84c0debbb9e 100644
> --- a/lib/stackdepot.c
> +++ b/lib/stackdepot.c
> @@ -22,6 +22,7 @@
>  #include <linux/mutex.h>
>  #include <linux/percpu.h>
>  #include <linux/printk.h>
> +#include <linux/refcount.h>
>  #include <linux/slab.h>
>  #include <linux/spinlock.h>
>  #include <linux/stacktrace.h>
> @@ -60,6 +61,7 @@ struct stack_record {
>  	u32 hash;			/* Hash in hash table */
>  	u32 size;			/* Number of stored frames */
>  	union handle_parts handle;
> +	refcount_t count;
>  	unsigned long entries[DEPOT_STACK_MAX_FRAMES];	/* Frames */
>  };
>  
> @@ -348,6 +350,7 @@ depot_alloc_stack(unsigned long *entries, int size, u32 hash, void **prealloc)
>  	stack->hash = hash;
>  	stack->size = size;
>  	/* stack->handle is already filled in by depot_init_pool. */
> +	refcount_set(&stack->count, 1);
>  	memcpy(stack->entries, entries, flex_array_size(stack, entries, size));
>  
>  	/*
> @@ -452,6 +455,7 @@ depot_stack_handle_t __stack_depot_save(unsigned long *entries,
>  	/* Fast path: look the stack trace up without full locking. */
>  	found = find_stack(*bucket, entries, nr_entries, hash);
>  	if (found) {
> +		refcount_inc(&found->count);

If someone doesn't use stack_depot_evict(), and the refcount eventually
overflows, it'll do a WARN (per refcount_warn_saturate()).

I think the interface needs to be different:

	stack_depot_get(): increments refcount (could be inline if just
	wrapper around refcount_inc())

	stack_depot_put(): what stack_depot_evict() currently does

Then it's clear that if someone uses either stack_depot_get() or _put()
that these need to be balanced. Not using either will result in the old
behaviour of never evicting an entry.

>  		read_unlock_irqrestore(&pool_rwlock, flags);
>  		goto exit;
>  	}
> -- 
> 2.25.1
>