From nobody Mon Feb 9 17:56:26 2026 Return-Path: X-Spam-Checker-Version: SpamAssassin 3.4.0 (2014-02-07) on aws-us-west-2-korg-lkml-1.web.codeaurora.org Received: from vger.kernel.org (vger.kernel.org [23.128.96.18]) by smtp.lore.kernel.org (Postfix) with ESMTP id 5BB99EB64DD for ; Fri, 30 Jun 2023 02:05:43 +0000 (UTC) Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S230125AbjF3CFk (ORCPT ); Thu, 29 Jun 2023 22:05:40 -0400 Received: from lindbergh.monkeyblade.net ([23.128.96.19]:48704 "EHLO lindbergh.monkeyblade.net" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S232088AbjF3CFK (ORCPT ); Thu, 29 Jun 2023 22:05:10 -0400 Received: from mail-yb1-xb4a.google.com (mail-yb1-xb4a.google.com [IPv6:2607:f8b0:4864:20::b4a]) by lindbergh.monkeyblade.net (Postfix) with ESMTPS id 99AA23593 for ; Thu, 29 Jun 2023 19:04:54 -0700 (PDT) Received: by mail-yb1-xb4a.google.com with SMTP id 3f1490d57ef6-c0d62f4487cso1199363276.0 for ; Thu, 29 Jun 2023 19:04:54 -0700 (PDT) DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=google.com; s=20221208; t=1688090693; x=1690682693; h=cc:to:from:subject:message-id:references:mime-version:in-reply-to :date:from:to:cc:subject:date:message-id:reply-to; bh=CYZzVVGdFW3BZ7BX39EfN49IRv4dv2Sl4PMSdQqxDi4=; b=OfSNNJfPU4LwQ+4p9lDzr7d1rFDuCbznmXwZGpFoQOb9UUudIpyItFW/YnfzN2Uh76 17QZyg1dnJZ0RJOiYlJksJqf+ObRNyfur60YXCCTmzLOEpNbJptma3iOBWfggxbFypeQ SuvH+35aKNV8vVQgv8xFp3l8bhmps0ul+sZaLrf42oWdSvTkXAXDfwilgWGr0jerAG4l sSY5toJRTgjNO5CA+/+11sgpLNowGp96hGOd7L/hLE+sdwuDraoAAgdGjVBoc+inBYNF /ADidYiQiiUKM6I0fDPL0+DkQgUrxvrePdheW3EmveHdBrFadg1h+kL02l0hzoDZg27p 8OYQ== X-Google-DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=1e100.net; s=20221208; t=1688090693; x=1690682693; 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=CYZzVVGdFW3BZ7BX39EfN49IRv4dv2Sl4PMSdQqxDi4=; b=V7xxJo9lqvhXuVDl9LCkrzsWL0fFtUWzfiRDFW279R8LkKAriCToSAG3YHnqW1cgcu FQ8nhyljlrIjQ8WlsoDkL8trw7k3HDdFVEL1lTk4NfRkfdNHqjNYlDinMU8j2CNGorWg s66W3LLRZYh2fxd794vAgO/Xa9N1h35VvnfL0H7862fYzQSTbDPKNu22sDL20MJrCSRa 5sdldK+dZNWwZvO4grfY8xte58Iop0Zeg5QFSBHwarCgpZSmbuRvtNNtZzSspf06H8+s A1urAGqyU3MA7GagH9mJR3lgGBhYrKtU4w5iCiLEKs8Nq1MQJBkwDB2wtMz7m7Q3i5Ns oSzw== X-Gm-Message-State: ABy/qLYeQA6n2o5pU/g/ygh7dfiJwX4McCZwhdatP+SrRTypi97q93BQ KR5Pmq0zwE/w9Ew1623KAE6ZovhLMtE= X-Google-Smtp-Source: APBJJlHphA7u0fNQhe1Hsg8YKxbjFOsjCu3koFUTN0LEhcBBqL04HvYnu+QQsZhyytzMLSCp+6sbkkOtcng= X-Received: from surenb-desktop.mtv.corp.google.com ([2620:15c:211:201:1f11:a3d0:19a9:16e5]) (user=surenb job=sendgmr) by 2002:a25:4183:0:b0:c1e:f91c:2691 with SMTP id o125-20020a254183000000b00c1ef91c2691mr12237yba.10.1688090693727; Thu, 29 Jun 2023 19:04:53 -0700 (PDT) Date: Thu, 29 Jun 2023 19:04:35 -0700 In-Reply-To: <20230630020436.1066016-1-surenb@google.com> Mime-Version: 1.0 References: <20230630020436.1066016-1-surenb@google.com> X-Mailer: git-send-email 2.41.0.255.g8b1d071c50-goog Message-ID: <20230630020436.1066016-7-surenb@google.com> Subject: [PATCH v6 6/6] mm: handle userfaults under VMA lock From: Suren Baghdasaryan To: akpm@linux-foundation.org Cc: willy@infradead.org, hannes@cmpxchg.org, mhocko@suse.com, josef@toxicpanda.com, jack@suse.cz, ldufour@linux.ibm.com, laurent.dufour@fr.ibm.com, michel@lespinasse.org, liam.howlett@oracle.com, jglisse@google.com, vbabka@suse.cz, minchan@google.com, dave@stgolabs.net, punit.agrawal@bytedance.com, lstoakes@gmail.com, hdanton@sina.com, apopple@nvidia.com, peterx@redhat.com, ying.huang@intel.com, david@redhat.com, yuzhao@google.com, dhowells@redhat.com, hughd@google.com, viro@zeniv.linux.org.uk, brauner@kernel.org, pasha.tatashin@soleen.com, surenb@google.com, linux-mm@kvack.org, linux-fsdevel@vger.kernel.org, linux-kernel@vger.kernel.org, kernel-team@android.com Precedence: bulk List-ID: X-Mailing-List: linux-kernel@vger.kernel.org Content-Transfer-Encoding: quoted-printable Content-Type: text/plain; charset="utf-8" Enable handle_userfault to operate under VMA lock by releasing VMA lock instead of mmap_lock and retrying. Note that FAULT_FLAG_RETRY_NOWAIT should never be used when handling faults under per-VMA lock protection because that would break the assumption that lock is dropped on retry. Signed-off-by: Suren Baghdasaryan Acked-by: Peter Xu --- fs/userfaultfd.c | 34 ++++++++++++++-------------------- include/linux/mm.h | 24 ++++++++++++++++++++++++ mm/memory.c | 9 --------- 3 files changed, 38 insertions(+), 29 deletions(-) diff --git a/fs/userfaultfd.c b/fs/userfaultfd.c index 7cecd49e078b..21a546eaf9f7 100644 --- a/fs/userfaultfd.c +++ b/fs/userfaultfd.c @@ -277,17 +277,16 @@ static inline struct uffd_msg userfault_msg(unsigned = long address, * hugepmd ranges. */ static inline bool userfaultfd_huge_must_wait(struct userfaultfd_ctx *ctx, - struct vm_area_struct *vma, - unsigned long address, - unsigned long flags, - unsigned long reason) + struct vm_fault *vmf, + unsigned long reason) { + struct vm_area_struct *vma =3D vmf->vma; pte_t *ptep, pte; bool ret =3D true; =20 - mmap_assert_locked(ctx->mm); + assert_fault_locked(vmf); =20 - ptep =3D hugetlb_walk(vma, address, vma_mmu_pagesize(vma)); + ptep =3D hugetlb_walk(vma, vmf->address, vma_mmu_pagesize(vma)); if (!ptep) goto out; =20 @@ -308,10 +307,8 @@ static inline bool userfaultfd_huge_must_wait(struct u= serfaultfd_ctx *ctx, } #else static inline bool userfaultfd_huge_must_wait(struct userfaultfd_ctx *ctx, - struct vm_area_struct *vma, - unsigned long address, - unsigned long flags, - unsigned long reason) + struct vm_fault *vmf, + unsigned long reason) { return false; /* should never get here */ } @@ -325,11 +322,11 @@ static inline bool userfaultfd_huge_must_wait(struct = userfaultfd_ctx *ctx, * threads. */ static inline bool userfaultfd_must_wait(struct userfaultfd_ctx *ctx, - unsigned long address, - unsigned long flags, + struct vm_fault *vmf, unsigned long reason) { struct mm_struct *mm =3D ctx->mm; + unsigned long address =3D vmf->address; pgd_t *pgd; p4d_t *p4d; pud_t *pud; @@ -338,7 +335,7 @@ static inline bool userfaultfd_must_wait(struct userfau= ltfd_ctx *ctx, pte_t ptent; bool ret =3D true; =20 - mmap_assert_locked(mm); + assert_fault_locked(vmf); =20 pgd =3D pgd_offset(mm, address); if (!pgd_present(*pgd)) @@ -440,7 +437,7 @@ vm_fault_t handle_userfault(struct vm_fault *vmf, unsig= ned long reason) * Coredumping runs without mmap_lock so we can only check that * the mmap_lock is held, if PF_DUMPCORE was not set. */ - mmap_assert_locked(mm); + assert_fault_locked(vmf); =20 ctx =3D vma->vm_userfaultfd_ctx.ctx; if (!ctx) @@ -556,15 +553,12 @@ vm_fault_t handle_userfault(struct vm_fault *vmf, uns= igned long reason) spin_unlock_irq(&ctx->fault_pending_wqh.lock); =20 if (!is_vm_hugetlb_page(vma)) - must_wait =3D userfaultfd_must_wait(ctx, vmf->address, vmf->flags, - reason); + must_wait =3D userfaultfd_must_wait(ctx, vmf, reason); else - must_wait =3D userfaultfd_huge_must_wait(ctx, vma, - vmf->address, - vmf->flags, reason); + must_wait =3D userfaultfd_huge_must_wait(ctx, vmf, reason); if (is_vm_hugetlb_page(vma)) hugetlb_vma_unlock_read(vma); - mmap_read_unlock(mm); + release_fault_lock(vmf); =20 if (likely(must_wait && !READ_ONCE(ctx->released))) { wake_up_poll(&ctx->fd_wqh, EPOLLIN); diff --git a/include/linux/mm.h b/include/linux/mm.h index 54ab11214f4f..2794225b2d42 100644 --- a/include/linux/mm.h +++ b/include/linux/mm.h @@ -705,6 +705,17 @@ static inline bool vma_try_start_write(struct vm_area_= struct *vma) return true; } =20 +static inline void vma_assert_locked(struct vm_area_struct *vma) +{ + int mm_lock_seq; + + if (__is_vma_write_locked(vma, &mm_lock_seq)) + return; + + lockdep_assert_held(&vma->vm_lock->lock); + VM_BUG_ON_VMA(!rwsem_is_locked(&vma->vm_lock->lock), vma); +} + static inline void vma_assert_write_locked(struct vm_area_struct *vma) { int mm_lock_seq; @@ -728,6 +739,14 @@ static inline void release_fault_lock(struct vm_fault = *vmf) mmap_read_unlock(vmf->vma->vm_mm); } =20 +static inline void assert_fault_locked(struct vm_fault *vmf) +{ + if (vmf->flags & FAULT_FLAG_VMA_LOCK) + vma_assert_locked(vmf->vma); + else + mmap_assert_locked(vmf->vma->vm_mm); +} + struct vm_area_struct *lock_vma_under_rcu(struct mm_struct *mm, unsigned long address); =20 @@ -748,6 +767,11 @@ static inline void release_fault_lock(struct vm_fault = *vmf) mmap_read_unlock(vmf->vma->vm_mm); } =20 +static inline void assert_fault_locked(struct vm_fault *vmf) +{ + mmap_assert_locked(vmf->vma->vm_mm); +} + #endif /* CONFIG_PER_VMA_LOCK */ =20 /* diff --git a/mm/memory.c b/mm/memory.c index bb0f68a73b0c..d9f36f9392a9 100644 --- a/mm/memory.c +++ b/mm/memory.c @@ -5407,15 +5407,6 @@ struct vm_area_struct *lock_vma_under_rcu(struct mm_= struct *mm, if (!vma_start_read(vma)) goto inval; =20 - /* - * Due to the possibility of userfault handler dropping mmap_lock, avoid - * it for now and fall back to page fault handling under mmap_lock. - */ - if (userfaultfd_armed(vma)) { - vma_end_read(vma); - goto inval; - } - /* Check since vm_start/vm_end might change before we lock the VMA */ if (unlikely(address < vma->vm_start || address >=3D vma->vm_end)) { vma_end_read(vma); --=20 2.41.0.255.g8b1d071c50-goog