Re: [PATCH v10] mm: slub: move sysfs slab alloc/free interfaces to debugfs

[Vlastimil Babka]

On 6/6/21 6:14 PM, Faiyaz Mohammed wrote:
> alloc_calls and free_calls implementation in sysfs have two issues,
> one is PAGE_SIZE limitiation of sysfs and other is it does not adhere
> to "one value per file" rule.
> 
> To overcome this issues, move the alloc_calls and free_calls implemeation
> to debugfs.
> 
> Debugfs cache will be created if SLAB_STORE_USER flag is set.
> 
> Rename the alloc_calls/free_calls to alloc_traces/free_traces,
> to be inline with what it does.
> 
> Signed-off-by: Faiyaz Mohammed <faiyazm@xxxxxxxxxxxxxx>
> ---
>  mm/slab.h        |   8 ++
>  mm/slab_common.c |   2 +
>  mm/slub.c        | 292 +++++++++++++++++++++++++++++++++++++------------------
>  3 files changed, 209 insertions(+), 93 deletions(-)
> 

...

> +static int slab_debug_trace_open(struct inode *inode, struct file *filep)
> +{
> +
> +	struct kmem_cache_node *n;
> +	enum track_item alloc;
> +	int node;
> +	struct loc_track *t = __seq_open_private(filep, &slab_debugfs_sops,
> +						sizeof(struct loc_track));
> +	struct kmem_cache *s = file_inode(filep)->i_private;
> +
> +	if (strcmp(filep->f_path.dentry->d_name.name, "alloc_traces") == 0)
> +		alloc =  TRACK_ALLOC;

			^ extra space here?

> +	else
> +		alloc =  TRACK_FREE;

same here

> +
> +	if (!alloc_loc_track(t, PAGE_SIZE / sizeof(struct location), GFP_KERNEL)) {
> +		pr_err("Out of memory\n");

Hm I would remove this. It doesn't print any context, so it's not useful to let
users know where/why we ran out of memory. Also if a GFP_KERNEL allocation
fails, there will be a big warning including stacktrace from the page allocator
anyway.

> +		return -ENOMEM;
> +	}
> +
> +	/* Push back cpu slabs */
> +	flush_all(s);
> +
> +	for_each_kmem_cache_node(s, node, n) {
> +		unsigned long flags;
> +		struct page *page;
> +
> +		if (!atomic_long_read(&n->nr_slabs))
> +			continue;
> +
> +		spin_lock_irqsave(&n->list_lock, flags);
> +		list_for_each_entry(page, &n->partial, slab_list)
> +			process_slab(t, s, page, alloc);
> +		list_for_each_entry(page, &n->full, slab_list)
> +			process_slab(t, s, page, alloc);
> +			spin_unlock_irqrestore(&n->list_lock, flags);

At least this is not Python, so it's just a visual flaw :)

> +	}
> +
> +	return 0;
> +}
> +
> +static int slab_debug_trace_release(struct inode *inode, struct file *file)
> +{
> +	struct seq_file *seq = file->private_data;
> +	struct loc_track *t = seq->private;
> +
> +	free_loc_track(t);
> +	kfree(seq->private);
> +	seq->private = NULL;
> +	return seq_release(inode, file);

You can call seq_release_private() instead and deal just with free_loc_track here.

Thanks!
Vlastimil