On Tue, Feb 20, 2018 at 10:23:41AM -0600, Christopher Lameter wrote:
> On Tue, 20 Feb 2018, Matthew Wilcox wrote:
>
> > I don't think it's fixable; there's just too much information per slab.
> > Anyway, I preferred the solution you & I were working on to limit the
> > length of names to 16 bytes, except for the cgroup slabs.
>
> So what do we do with the cgroup slab names? have a slabinfo per cgroup?
What I had in mind ...
struct kmem_cache_attr {
const char name[16];
unsigned int size;
unsigned int align;
unsigned int useroffset;
unsigned int usersize;
slab_flags_t flags;
kmem_cache_ctor ctor;
}
struct kmem_cache {
const struct kmem_cache_attr *a;
const char *name;
...
};
In kmem_cache_create_usercopy:
s->name = a->name;
In memcg_create_kmem_cache:
s->name = kasprintf(GFP_KERNEL, "%s(%llu:%s)", a->name,
css->serial_nr, memcg_name_buf);
In slab_kmem_cache_release:
if (s->name != s->a->name)
kfree_const(s->name);
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@xxxxxxxxx. For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@xxxxxxxxx";> email@xxxxxxxxx </a>
