| Message ID | 20250115021746.34691-7-alexei.starovoitov@gmail.com (mailing list archive) |
|---|---|
| State | New |
| Headers | show |
| Series | bpf, mm: Introduce try_alloc_pages() | expand |
On 1/15/25 03:17, Alexei Starovoitov wrote: > From: Alexei Starovoitov <ast@kernel.org> > > When gfpflags_allow_spinning() == false spin_locks cannot be taken. > Make failslab, kfence, kmemleak compliant. > > Signed-off-by: Alexei Starovoitov <ast@kernel.org> All these are related to slab, so this would rather belong to a followup series that expands the support from page allocator to slab, no? > --- > mm/failslab.c | 3 +++ > mm/kfence/core.c | 4 ++++ > mm/kmemleak.c | 3 +++ > 3 files changed, 10 insertions(+) > > diff --git a/mm/failslab.c b/mm/failslab.c > index c3901b136498..86c7304ef25a 100644 > --- a/mm/failslab.c > +++ b/mm/failslab.c > @@ -27,6 +27,9 @@ int should_failslab(struct kmem_cache *s, gfp_t gfpflags) > if (gfpflags & __GFP_NOFAIL) > return 0; > > + if (!gfpflags_allow_spinning(gfpflags)) > + return 0; > + > if (failslab.ignore_gfp_reclaim && > (gfpflags & __GFP_DIRECT_RECLAIM)) > return 0; > diff --git a/mm/kfence/core.c b/mm/kfence/core.c > index 67fc321db79b..e5f2d63f3220 100644 > --- a/mm/kfence/core.c > +++ b/mm/kfence/core.c > @@ -1096,6 +1096,10 @@ void *__kfence_alloc(struct kmem_cache *s, size_t size, gfp_t flags) > if (s->flags & SLAB_SKIP_KFENCE) > return NULL; > > + /* Bailout, since kfence_guarded_alloc() needs to take a lock */ > + if (!gfpflags_allow_spinning(flags)) > + return NULL; > + > allocation_gate = atomic_inc_return(&kfence_allocation_gate); > if (allocation_gate > 1) > return NULL; > diff --git a/mm/kmemleak.c b/mm/kmemleak.c > index 2a945c07ae99..64cb44948e9e 100644 > --- a/mm/kmemleak.c > +++ b/mm/kmemleak.c > @@ -648,6 +648,9 @@ static struct kmemleak_object *__alloc_object(gfp_t gfp) > { > struct kmemleak_object *object; > > + if (!gfpflags_allow_spinning(gfp)) > + return NULL; > + > object = mem_pool_alloc(gfp); > if (!object) { > pr_warn("Cannot allocate a kmemleak_object structure\n");
On Wed, Jan 15, 2025 at 9:57 AM Vlastimil Babka <vbabka@suse.cz> wrote: > > On 1/15/25 03:17, Alexei Starovoitov wrote: > > From: Alexei Starovoitov <ast@kernel.org> > > > > When gfpflags_allow_spinning() == false spin_locks cannot be taken. > > Make failslab, kfence, kmemleak compliant. > > > > Signed-off-by: Alexei Starovoitov <ast@kernel.org> > > All these are related to slab, so this would rather belong to a followup > series that expands the support from page allocator to slab, no? Sure. I can drop it for now. It was more of the preview of things to come. And how gfpflags_allow_spinning() fits in other places.
diff --git a/mm/failslab.c b/mm/failslab.c index c3901b136498..86c7304ef25a 100644 --- a/mm/failslab.c +++ b/mm/failslab.c @@ -27,6 +27,9 @@ int should_failslab(struct kmem_cache *s, gfp_t gfpflags) if (gfpflags & __GFP_NOFAIL) return 0; + if (!gfpflags_allow_spinning(gfpflags)) + return 0; + if (failslab.ignore_gfp_reclaim && (gfpflags & __GFP_DIRECT_RECLAIM)) return 0; diff --git a/mm/kfence/core.c b/mm/kfence/core.c index 67fc321db79b..e5f2d63f3220 100644 --- a/mm/kfence/core.c +++ b/mm/kfence/core.c @@ -1096,6 +1096,10 @@ void *__kfence_alloc(struct kmem_cache *s, size_t size, gfp_t flags) if (s->flags & SLAB_SKIP_KFENCE) return NULL; + /* Bailout, since kfence_guarded_alloc() needs to take a lock */ + if (!gfpflags_allow_spinning(flags)) + return NULL; + allocation_gate = atomic_inc_return(&kfence_allocation_gate); if (allocation_gate > 1) return NULL; diff --git a/mm/kmemleak.c b/mm/kmemleak.c index 2a945c07ae99..64cb44948e9e 100644 --- a/mm/kmemleak.c +++ b/mm/kmemleak.c @@ -648,6 +648,9 @@ static struct kmemleak_object *__alloc_object(gfp_t gfp) { struct kmemleak_object *object; + if (!gfpflags_allow_spinning(gfp)) + return NULL; + object = mem_pool_alloc(gfp); if (!object) { pr_warn("Cannot allocate a kmemleak_object structure\n");