From nobody Fri Dec 19 04:58:33 2025 Received: from mail-pl1-f202.google.com (mail-pl1-f202.google.com [209.85.214.202]) (using TLSv1.2 with cipher ECDHE-RSA-AES128-GCM-SHA256 (128/128 bits)) (No client certificate requested) by smtp.subspace.kernel.org (Postfix) with ESMTPS id C3397270EB2 for ; Thu, 13 Feb 2025 22:47:04 +0000 (UTC) Authentication-Results: smtp.subspace.kernel.org; arc=none smtp.client-ip=209.85.214.202 ARC-Seal: i=1; a=rsa-sha256; d=subspace.kernel.org; s=arc-20240116; t=1739486826; cv=none; b=T0JrXfBBAOcMzhRZhTQc03/1/a/CAlIGEY4HnziSkB2st4VIEm5Ys44XuJbrRlhaZeHtxEMoET1APP/xNe1H06qCoolyfqrrGOLaMFmtLSa9zZHmZjVuOuUVgU9UFoUI+lxd3PwZfz3sqdGgaLDwDmBTBwzYnGHyWSCcqbVuHFU= ARC-Message-Signature: i=1; a=rsa-sha256; d=subspace.kernel.org; s=arc-20240116; t=1739486826; c=relaxed/simple; bh=KABNpsb25nd8x3O/oXZpCeD64dxAEJiUS2WGVDITMD4=; h=Date:In-Reply-To:Mime-Version:References:Message-ID:Subject:From: To:Cc:Content-Type; b=G7tM8zXP8KKukLEXaKb1DpsOIGAok4l4PmAQHkmpRSY2ANFKOzP9jZVVijlDYS0xODwVp2P4soygU40pssxKQop9+eUmlyxi6s8pJRNXu/t8lbpvH+peL2MQ8NjkrfLdi3DZke1dxhj+30P83cx9q+IG6X2FcaoCEjNps11rKPk= ARC-Authentication-Results: i=1; smtp.subspace.kernel.org; dmarc=pass (p=reject dis=none) header.from=google.com; spf=pass smtp.mailfrom=flex--surenb.bounces.google.com; dkim=pass (2048-bit key) header.d=google.com header.i=@google.com header.b=mfuOxG4q; arc=none smtp.client-ip=209.85.214.202 Authentication-Results: smtp.subspace.kernel.org; dmarc=pass (p=reject dis=none) header.from=google.com Authentication-Results: smtp.subspace.kernel.org; spf=pass smtp.mailfrom=flex--surenb.bounces.google.com Authentication-Results: smtp.subspace.kernel.org; dkim=pass (2048-bit key) header.d=google.com header.i=@google.com header.b="mfuOxG4q" Received: by mail-pl1-f202.google.com with SMTP id d9443c01a7336-21f379b7bb3so26395825ad.2 for ; Thu, 13 Feb 2025 14:47:04 -0800 (PST) DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=google.com; s=20230601; t=1739486824; x=1740091624; darn=vger.kernel.org; h=cc:to:from:subject:message-id:references:mime-version:in-reply-to :date:from:to:cc:subject:date:message-id:reply-to; bh=nixUczrejX4FKBcz+HGm8JxV/4Rz6UEjTIeiIS536Tg=; b=mfuOxG4qHIiNgbAI52ssBmz1Y9eamsYLh7Nq9JRHwOBAS166WRb/oNfXk+i8CpNMm+ kbK07TYg2BghHPo1+5S9+0NsETT3tzwS0SsTn4dZaFEAxh5HiILSzJgiO9TxITuhHX0z A8Rmj4yURrkesWynjVdFZMke8i+IVxJr+t6VwI6ngjDvLXjWQTnTurjpE4bz8RWAgc8M NUnz0FHzOyHgt++snPFYgLlwLH/5JLFn7JEY82q0IGM0NU0IBP4OKdreFZfyqvf9Ylu6 Sj3Bx2aj77KcHFdYr0ExaYoeC9Ghp6Rjug2s9CMwKp2nVUfsxDTXihg/uiTUSk32RryU Y1rg== X-Google-DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=1e100.net; s=20230601; t=1739486824; x=1740091624; h=cc:to:from:subject:message-id:references:mime-version:in-reply-to :date:x-gm-message-state:from:to:cc:subject:date:message-id:reply-to; bh=nixUczrejX4FKBcz+HGm8JxV/4Rz6UEjTIeiIS536Tg=; b=XjWif7bzqhfYTEQcUPXf4roYmCbv32AOw6KqNFEQ5kPA3GfQh/Uawpvv8ZV1lHNs4r HiASgPU3YC75ablrqDcs8lFisDgsJpYbMxQsUAgN0XElI3CDKQlKNKth3WLvuPAqk4BP JlozJfIY17QcIyYEm0v1WNVHEN+FAnhGEqD6tkLy8x5AgzCt/HtbJ3lKTkkVhZJ+mY5l kmbW5xKMk9Uo1sU6xxiZDhdMvUyWhgs+Jrv+hpmkRV/iEYofBjY5EFnbG7q1K6MV23FZ ZvAti4eaVT21kT2lLXGNXoeeQQXohyxBWIXNVGFOoKYziCv48UqzP4AZFydwVohAXzlD oKgA== X-Forwarded-Encrypted: i=1; AJvYcCUnyqFP0om6QyM9cm26fmtu+bjuuAzkKgxIDWZ3uQWs7+AaeetksLyN5VJJGEaa4jg8cZWZLCqc+gr/HC4=@vger.kernel.org X-Gm-Message-State: AOJu0YygV2UpK/1TDQ/7K+CDKXaxstnbSMj+IvOTkTnqkR+J2/0MHwpD G9nXMvrD0BH5Jxb8LI08TaqYT4E98z6+lBcxkD+7aZRnm6CrkFq8VxYD7D8pFw0MY7VBZH/F6UA MPQ== X-Google-Smtp-Source: AGHT+IHexWPTyEo/ZnzJkqYLOzvXPZYi2c9eLV49bS4E26ucZsJK7lHyl46qKggig4vIRb/FsPaNzCiwY30= X-Received: from pgjp13.prod.google.com ([2002:a63:e64d:0:b0:ad8:73cf:4390]) (user=surenb job=prod-delivery.src-stubby-dispatcher) by 2002:a05:6a20:c998:b0:1ee:1e05:206d with SMTP id adf61e73a8af0-1ee5e5b7c8amr15452640637.21.1739486823384; Thu, 13 Feb 2025 14:47:03 -0800 (PST) Date: Thu, 13 Feb 2025 14:46:39 -0800 In-Reply-To: <20250213224655.1680278-1-surenb@google.com> Precedence: bulk X-Mailing-List: linux-kernel@vger.kernel.org List-Id: List-Subscribe: List-Unsubscribe: Mime-Version: 1.0 References: <20250213224655.1680278-1-surenb@google.com> X-Mailer: git-send-email 2.48.1.601.g30ceb7b040-goog Message-ID: <20250213224655.1680278-3-surenb@google.com> Subject: [PATCH v10 02/18] mm: move per-vma lock into vm_area_struct From: Suren Baghdasaryan To: akpm@linux-foundation.org Cc: peterz@infradead.org, willy@infradead.org, liam.howlett@oracle.com, lorenzo.stoakes@oracle.com, david.laight.linux@gmail.com, mhocko@suse.com, vbabka@suse.cz, hannes@cmpxchg.org, mjguzik@gmail.com, oliver.sang@intel.com, mgorman@techsingularity.net, david@redhat.com, peterx@redhat.com, oleg@redhat.com, dave@stgolabs.net, paulmck@kernel.org, brauner@kernel.org, dhowells@redhat.com, hdanton@sina.com, hughd@google.com, lokeshgidra@google.com, minchan@google.com, jannh@google.com, shakeel.butt@linux.dev, souravpanda@google.com, pasha.tatashin@soleen.com, klarasmodin@gmail.com, richard.weiyang@gmail.com, corbet@lwn.net, linux-doc@vger.kernel.org, linux-mm@kvack.org, linux-kernel@vger.kernel.org, kernel-team@android.com, surenb@google.com, "Liam R. Howlett" Content-Transfer-Encoding: quoted-printable Content-Type: text/plain; charset="utf-8" Back when per-vma locks were introduces, vm_lock was moved out of vm_area_struct in [1] because of the performance regression caused by false cacheline sharing. Recent investigation [2] revealed that the regressions is limited to a rather old Broadwell microarchitecture and even there it can be mitigated by disabling adjacent cacheline prefetching, see [3]. Splitting single logical structure into multiple ones leads to more complicated management, extra pointer dereferences and overall less maintainable code. When that split-away part is a lock, it complicates things even further. With no performance benefits, there are no reasons for this split. Merging the vm_lock back into vm_area_struct also allows vm_area_struct to use SLAB_TYPESAFE_BY_RCU later in this patchset. Move vm_lock back into vm_area_struct, aligning it at the cacheline boundary and changing the cache to be cacheline-aligned as well. With kernel compiled using defconfig, this causes VMA memory consumption to grow from 160 (vm_area_struct) + 40 (vm_lock) bytes to 256 bytes: slabinfo before: ... : ... vma_lock ... 40 102 1 : ... vm_area_struct ... 160 51 2 : ... slabinfo after moving vm_lock: ... : ... vm_area_struct ... 256 32 2 : ... Aggregate VMA memory consumption per 1000 VMAs grows from 50 to 64 pages, which is 5.5MB per 100000 VMAs. Note that the size of this structure is dependent on the kernel configuration and typically the original size is higher than 160 bytes. Therefore these calculations are close to the worst case scenario. A more realistic vm_area_struct usage before this change is: ... : ... vma_lock ... 40 102 1 : ... vm_area_struct ... 176 46 2 : ... Aggregate VMA memory consumption per 1000 VMAs grows from 54 to 64 pages, which is 3.9MB per 100000 VMAs. This memory consumption growth can be addressed later by optimizing the vm_lock. [1] https://lore.kernel.org/all/20230227173632.3292573-34-surenb@google.com/ [2] https://lore.kernel.org/all/ZsQyI%2F087V34JoIt@xsang-OptiPlex-9020/ [3] https://lore.kernel.org/all/CAJuCfpEisU8Lfe96AYJDZ+OM4NoPmnw9bP53cT_kbf= P_pR+-2g@mail.gmail.com/ Signed-off-by: Suren Baghdasaryan Reviewed-by: Lorenzo Stoakes Reviewed-by: Shakeel Butt Reviewed-by: Vlastimil Babka Reviewed-by: Liam R. Howlett --- include/linux/mm.h | 28 ++++++++++-------- include/linux/mm_types.h | 6 ++-- kernel/fork.c | 49 ++++---------------------------- tools/testing/vma/vma_internal.h | 33 +++++---------------- 4 files changed, 32 insertions(+), 84 deletions(-) diff --git a/include/linux/mm.h b/include/linux/mm.h index 16b3cd3de29a..e75fae95b48d 100644 --- a/include/linux/mm.h +++ b/include/linux/mm.h @@ -697,6 +697,12 @@ static inline void vma_numab_state_free(struct vm_area= _struct *vma) {} #endif /* CONFIG_NUMA_BALANCING */ =20 #ifdef CONFIG_PER_VMA_LOCK +static inline void vma_lock_init(struct vm_area_struct *vma) +{ + init_rwsem(&vma->vm_lock.lock); + vma->vm_lock_seq =3D UINT_MAX; +} + /* * Try to read-lock a vma. The function is allowed to occasionally yield f= alse * locked result to avoid performance overhead, in which case we fall back= to @@ -714,7 +720,7 @@ static inline bool vma_start_read(struct vm_area_struct= *vma) if (READ_ONCE(vma->vm_lock_seq) =3D=3D READ_ONCE(vma->vm_mm->mm_lock_seq.= sequence)) return false; =20 - if (unlikely(down_read_trylock(&vma->vm_lock->lock) =3D=3D 0)) + if (unlikely(down_read_trylock(&vma->vm_lock.lock) =3D=3D 0)) return false; =20 /* @@ -729,7 +735,7 @@ static inline bool vma_start_read(struct vm_area_struct= *vma) * This pairs with RELEASE semantics in vma_end_write_all(). */ if (unlikely(vma->vm_lock_seq =3D=3D raw_read_seqcount(&vma->vm_mm->mm_lo= ck_seq))) { - up_read(&vma->vm_lock->lock); + up_read(&vma->vm_lock.lock); return false; } return true; @@ -744,7 +750,7 @@ static inline bool vma_start_read(struct vm_area_struct= *vma) static inline void vma_start_read_locked_nested(struct vm_area_struct *vma= , int subclass) { mmap_assert_locked(vma->vm_mm); - down_read_nested(&vma->vm_lock->lock, subclass); + down_read_nested(&vma->vm_lock.lock, subclass); } =20 /* @@ -756,13 +762,13 @@ static inline void vma_start_read_locked_nested(struc= t vm_area_struct *vma, int static inline void vma_start_read_locked(struct vm_area_struct *vma) { mmap_assert_locked(vma->vm_mm); - down_read(&vma->vm_lock->lock); + down_read(&vma->vm_lock.lock); } =20 static inline void vma_end_read(struct vm_area_struct *vma) { rcu_read_lock(); /* keeps vma alive till the end of up_read */ - up_read(&vma->vm_lock->lock); + up_read(&vma->vm_lock.lock); rcu_read_unlock(); } =20 @@ -791,7 +797,7 @@ static inline void vma_start_write(struct vm_area_struc= t *vma) if (__is_vma_write_locked(vma, &mm_lock_seq)) return; =20 - down_write(&vma->vm_lock->lock); + down_write(&vma->vm_lock.lock); /* * We should use WRITE_ONCE() here because we can have concurrent reads * from the early lockless pessimistic check in vma_start_read(). @@ -799,7 +805,7 @@ static inline void vma_start_write(struct vm_area_struc= t *vma) * we should use WRITE_ONCE() for cleanliness and to keep KCSAN happy. */ WRITE_ONCE(vma->vm_lock_seq, mm_lock_seq); - up_write(&vma->vm_lock->lock); + up_write(&vma->vm_lock.lock); } =20 static inline void vma_assert_write_locked(struct vm_area_struct *vma) @@ -811,7 +817,7 @@ static inline void vma_assert_write_locked(struct vm_ar= ea_struct *vma) =20 static inline void vma_assert_locked(struct vm_area_struct *vma) { - if (!rwsem_is_locked(&vma->vm_lock->lock)) + if (!rwsem_is_locked(&vma->vm_lock.lock)) vma_assert_write_locked(vma); } =20 @@ -844,6 +850,7 @@ struct vm_area_struct *lock_vma_under_rcu(struct mm_str= uct *mm, =20 #else /* CONFIG_PER_VMA_LOCK */ =20 +static inline void vma_lock_init(struct vm_area_struct *vma) {} static inline bool vma_start_read(struct vm_area_struct *vma) { return false; } static inline void vma_end_read(struct vm_area_struct *vma) {} @@ -878,10 +885,6 @@ static inline void assert_fault_locked(struct vm_fault= *vmf) =20 extern const struct vm_operations_struct vma_dummy_vm_ops; =20 -/* - * WARNING: vma_init does not initialize vma->vm_lock. - * Use vm_area_alloc()/vm_area_free() if vma needs locking. - */ static inline void vma_init(struct vm_area_struct *vma, struct mm_struct *= mm) { memset(vma, 0, sizeof(*vma)); @@ -890,6 +893,7 @@ static inline void vma_init(struct vm_area_struct *vma,= struct mm_struct *mm) INIT_LIST_HEAD(&vma->anon_vma_chain); vma_mark_detached(vma, false); vma_numab_state_init(vma); + vma_lock_init(vma); } =20 /* Use when VMA is not part of the VMA tree and needs no locking */ diff --git a/include/linux/mm_types.h b/include/linux/mm_types.h index 8efafef4637e..8a645bcb2b31 100644 --- a/include/linux/mm_types.h +++ b/include/linux/mm_types.h @@ -740,8 +740,6 @@ struct vm_area_struct { * slowpath. */ unsigned int vm_lock_seq; - /* Unstable RCU readers are allowed to read this. */ - struct vma_lock *vm_lock; #endif =20 /* @@ -794,6 +792,10 @@ struct vm_area_struct { struct vma_numab_state *numab_state; /* NUMA Balancing state */ #endif struct vm_userfaultfd_ctx vm_userfaultfd_ctx; +#ifdef CONFIG_PER_VMA_LOCK + /* Unstable RCU readers are allowed to read this. */ + struct vma_lock vm_lock ____cacheline_aligned_in_smp; +#endif } __randomize_layout; =20 #ifdef CONFIG_NUMA diff --git a/kernel/fork.c b/kernel/fork.c index 735405a9c5f3..bdbabe73fb29 100644 --- a/kernel/fork.c +++ b/kernel/fork.c @@ -436,35 +436,6 @@ static struct kmem_cache *vm_area_cachep; /* SLAB cache for mm_struct structures (tsk->mm) */ static struct kmem_cache *mm_cachep; =20 -#ifdef CONFIG_PER_VMA_LOCK - -/* SLAB cache for vm_area_struct.lock */ -static struct kmem_cache *vma_lock_cachep; - -static bool vma_lock_alloc(struct vm_area_struct *vma) -{ - vma->vm_lock =3D kmem_cache_alloc(vma_lock_cachep, GFP_KERNEL); - if (!vma->vm_lock) - return false; - - init_rwsem(&vma->vm_lock->lock); - vma->vm_lock_seq =3D UINT_MAX; - - return true; -} - -static inline void vma_lock_free(struct vm_area_struct *vma) -{ - kmem_cache_free(vma_lock_cachep, vma->vm_lock); -} - -#else /* CONFIG_PER_VMA_LOCK */ - -static inline bool vma_lock_alloc(struct vm_area_struct *vma) { return tru= e; } -static inline void vma_lock_free(struct vm_area_struct *vma) {} - -#endif /* CONFIG_PER_VMA_LOCK */ - struct vm_area_struct *vm_area_alloc(struct mm_struct *mm) { struct vm_area_struct *vma; @@ -474,10 +445,6 @@ struct vm_area_struct *vm_area_alloc(struct mm_struct = *mm) return NULL; =20 vma_init(vma, mm); - if (!vma_lock_alloc(vma)) { - kmem_cache_free(vm_area_cachep, vma); - return NULL; - } =20 return vma; } @@ -496,10 +463,7 @@ struct vm_area_struct *vm_area_dup(struct vm_area_stru= ct *orig) * will be reinitialized. */ data_race(memcpy(new, orig, sizeof(*new))); - if (!vma_lock_alloc(new)) { - kmem_cache_free(vm_area_cachep, new); - return NULL; - } + vma_lock_init(new); INIT_LIST_HEAD(&new->anon_vma_chain); vma_numab_state_init(new); dup_anon_vma_name(orig, new); @@ -511,7 +475,6 @@ void __vm_area_free(struct vm_area_struct *vma) { vma_numab_state_free(vma); free_anon_vma_name(vma); - vma_lock_free(vma); kmem_cache_free(vm_area_cachep, vma); } =20 @@ -522,7 +485,7 @@ static void vm_area_free_rcu_cb(struct rcu_head *head) vm_rcu); =20 /* The vma should not be locked while being destroyed. */ - VM_BUG_ON_VMA(rwsem_is_locked(&vma->vm_lock->lock), vma); + VM_BUG_ON_VMA(rwsem_is_locked(&vma->vm_lock.lock), vma); __vm_area_free(vma); } #endif @@ -3200,11 +3163,9 @@ void __init proc_caches_init(void) sizeof(struct fs_struct), 0, SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_ACCOUNT, NULL); - - vm_area_cachep =3D KMEM_CACHE(vm_area_struct, SLAB_PANIC|SLAB_ACCOUNT); -#ifdef CONFIG_PER_VMA_LOCK - vma_lock_cachep =3D KMEM_CACHE(vma_lock, SLAB_PANIC|SLAB_ACCOUNT); -#endif + vm_area_cachep =3D KMEM_CACHE(vm_area_struct, + SLAB_HWCACHE_ALIGN|SLAB_NO_MERGE|SLAB_PANIC| + SLAB_ACCOUNT); mmap_init(); nsproxy_cache_init(); } diff --git a/tools/testing/vma/vma_internal.h b/tools/testing/vma/vma_inter= nal.h index bb273927af0f..4506e6fb3c6f 100644 --- a/tools/testing/vma/vma_internal.h +++ b/tools/testing/vma/vma_internal.h @@ -275,10 +275,10 @@ struct vm_area_struct { /* * Can only be written (using WRITE_ONCE()) while holding both: * - mmap_lock (in write mode) - * - vm_lock->lock (in write mode) + * - vm_lock.lock (in write mode) * Can be read reliably while holding one of: * - mmap_lock (in read or write mode) - * - vm_lock->lock (in read or write mode) + * - vm_lock.lock (in read or write mode) * Can be read unreliably (using READ_ONCE()) for pessimistic bailout * while holding nothing (except RCU to keep the VMA struct allocated). * @@ -287,7 +287,7 @@ struct vm_area_struct { * slowpath. */ unsigned int vm_lock_seq; - struct vma_lock *vm_lock; + struct vma_lock vm_lock; #endif =20 /* @@ -464,17 +464,10 @@ static inline struct vm_area_struct *vma_next(struct = vma_iterator *vmi) return mas_find(&vmi->mas, ULONG_MAX); } =20 -static inline bool vma_lock_alloc(struct vm_area_struct *vma) +static inline void vma_lock_init(struct vm_area_struct *vma) { - vma->vm_lock =3D calloc(1, sizeof(struct vma_lock)); - - if (!vma->vm_lock) - return false; - - init_rwsem(&vma->vm_lock->lock); + init_rwsem(&vma->vm_lock.lock); vma->vm_lock_seq =3D UINT_MAX; - - return true; } =20 static inline void vma_assert_write_locked(struct vm_area_struct *); @@ -497,6 +490,7 @@ static inline void vma_init(struct vm_area_struct *vma,= struct mm_struct *mm) vma->vm_ops =3D &vma_dummy_vm_ops; INIT_LIST_HEAD(&vma->anon_vma_chain); vma_mark_detached(vma, false); + vma_lock_init(vma); } =20 static inline struct vm_area_struct *vm_area_alloc(struct mm_struct *mm) @@ -507,10 +501,6 @@ static inline struct vm_area_struct *vm_area_alloc(str= uct mm_struct *mm) return NULL; =20 vma_init(vma, mm); - if (!vma_lock_alloc(vma)) { - free(vma); - return NULL; - } =20 return vma; } @@ -523,10 +513,7 @@ static inline struct vm_area_struct *vm_area_dup(struc= t vm_area_struct *orig) return NULL; =20 memcpy(new, orig, sizeof(*new)); - if (!vma_lock_alloc(new)) { - free(new); - return NULL; - } + vma_lock_init(new); INIT_LIST_HEAD(&new->anon_vma_chain); =20 return new; @@ -696,14 +683,8 @@ static inline void mpol_put(struct mempolicy *) { } =20 -static inline void vma_lock_free(struct vm_area_struct *vma) -{ - free(vma->vm_lock); -} - static inline void __vm_area_free(struct vm_area_struct *vma) { - vma_lock_free(vma); free(vma); } =20 --=20 2.48.1.601.g30ceb7b040-goog