From nobody Sun Dec 14 11:32:40 2025 Received: from mail-vk1-f201.google.com (mail-vk1-f201.google.com [209.85.221.201]) (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 BF61A14A60F for ; Tue, 4 Feb 2025 00:41:03 +0000 (UTC) Authentication-Results: smtp.subspace.kernel.org; arc=none smtp.client-ip=209.85.221.201 ARC-Seal: i=1; a=rsa-sha256; d=subspace.kernel.org; s=arc-20240116; t=1738629666; cv=none; b=l5a7RgycscbVIoHnqR6pMmsEF+SzwXPJLj4jm4bHV0gh7T778d1TijbpgHa4np+rGi6rs6P9sIQcJ83EVxnsyfKYFvOwtQ2knPhTuAh6twnx4vOJaCm240vat1pIpDcV+qxfGH7PGyZGz7X44nffRjhWpuEAKevzMiGAThwKYlw= ARC-Message-Signature: i=1; a=rsa-sha256; d=subspace.kernel.org; s=arc-20240116; t=1738629666; c=relaxed/simple; bh=Zy4oye91lFdBLcv+ukgYWg+1Vp5yzDO7e3LTrEFoLk8=; h=Date:In-Reply-To:Mime-Version:References:Message-ID:Subject:From: To:Cc:Content-Type; b=XIb2J45yjB4Jna+7jzhGJBLCv8/UJ8bvlo8iEY3TNcfjePl6lAHPYti91PObgStz3U9+a3dz9ms76nlFyTjBJQDWf0Nh0RseFxZbQx5WXJ2T85VSZL3Rvh4G+PFZfHfqKHjBugjzVqPiDxq1mus5DXQsNCT/WxEGzy+RuAt3s3A= ARC-Authentication-Results: i=1; smtp.subspace.kernel.org; dmarc=pass (p=reject dis=none) header.from=google.com; spf=pass smtp.mailfrom=flex--jthoughton.bounces.google.com; dkim=pass (2048-bit key) header.d=google.com header.i=@google.com header.b=kA5Sn7zU; arc=none smtp.client-ip=209.85.221.201 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--jthoughton.bounces.google.com Authentication-Results: smtp.subspace.kernel.org; dkim=pass (2048-bit key) header.d=google.com header.i=@google.com header.b="kA5Sn7zU" Received: by mail-vk1-f201.google.com with SMTP id 71dfb90a1353d-51604d32337so1006767e0c.3 for ; Mon, 03 Feb 2025 16:41:03 -0800 (PST) DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=google.com; s=20230601; t=1738629662; x=1739234462; 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=eWoO+lMMXqWIzaVuugfET0/bTOkUIXS847FK2QmI/tI=; b=kA5Sn7zUMab4I3lAMkTkTosZQedg+WL1kLH87/ChT6Og4OaVKpMsEuvBZdCfnxRIOz cc6XrlcgM9HkLG5YN8h5bIJK7pKTuXw5iRKBhfhVGg+VgKhkSpqk5aisG0Ku8XXeRegN 15g2bFqyeMHtI0GzcLNGfGa4Af+qHi+n7WCykOKvv3sjVMY6T3J7rtr2b3yTTu1103QP GAXLgwlispCrjNoWHZDD3vQfh7nwiIIzs1+Chidc6LOPQLDJj44oMp2YfuPyNyO8Hpnv CAGS5cPbv/l5chByXW4dtGLdUHNg8W1Y/g8+TSPI1LTYULSr1NxZdjcmt4NtUIduLhxG xxRQ== X-Google-DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=1e100.net; s=20230601; t=1738629662; x=1739234462; 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=eWoO+lMMXqWIzaVuugfET0/bTOkUIXS847FK2QmI/tI=; b=dxGUNEl9a3wPBgAoepJxipLtJNBG63QBCH9lxXFjOhoPpzbQ680xrPk1gIRJ8Fxy8a 2UkXH4Cbk17PzzB1Z1UFafZ2F06XVsykFZ2Hm1C+J8oQQsTsDNhdZ4ab53V6Rjp/xiqz /ukh3m3Qe2aalVcHlYnVyxdEiNuSNNIXMpPaEpOXnqUsIHCttWd+w24p/7WdhbWkUptL pUe+DeQTutDTg099EAAk1tdcxuLQttRSwI6K23DDVIOBD80jqCvPtUq3lOnA8fH+3uuO wNY+Bf/kqYvJQrRKK8Hnp+uLcmXRWz+Q9iLdPf5MTAJoOEjSnxNeouLt2PsmmQ+I/WGP 1NOg== X-Forwarded-Encrypted: i=1; AJvYcCW0VZjZ+9PcNGTdJGA6XeqadKLkyXFFmW+aMka6o+HFioS0ypV/Dnmimq+Ym4Qe3mfOuxvcN544QAd9aSw=@vger.kernel.org X-Gm-Message-State: AOJu0YwzSpAsxCz7L/Q+9HD/fnQhaE0OMsmhwj0fiRpNm+ZfjlcNE0jb +xpOHTa1x3b/MfTuEVOVGe8FI+EYOM2GVJ3M5D3mCBR+PKd9dG73pMMrteezguF9oryKBIHMLUW shOCWtnbol2qek8JKSA== X-Google-Smtp-Source: AGHT+IGwZC+pQj5KVs1BoZvl9VTW+YoD2540d5PUuKxlYB7xTw8o5JqIwG6QPjFodj3FS42DjhlUFnaztBK8rSbO X-Received: from vkbej7.prod.google.com ([2002:a05:6122:2707:b0:51c:f313:dfc6]) (user=jthoughton job=prod-delivery.src-stubby-dispatcher) by 2002:a05:6122:240c:b0:518:859e:87c3 with SMTP id 71dfb90a1353d-51e9e3edb13mr19459670e0c.7.1738629662372; Mon, 03 Feb 2025 16:41:02 -0800 (PST) Date: Tue, 4 Feb 2025 00:40:36 +0000 In-Reply-To: <20250204004038.1680123-1-jthoughton@google.com> Precedence: bulk X-Mailing-List: linux-kernel@vger.kernel.org List-Id: List-Subscribe: List-Unsubscribe: Mime-Version: 1.0 References: <20250204004038.1680123-1-jthoughton@google.com> X-Mailer: git-send-email 2.48.1.362.g079036d154-goog Message-ID: <20250204004038.1680123-10-jthoughton@google.com> Subject: [PATCH v9 09/11] KVM: x86/mmu: Add infrastructure to allow walking rmaps outside of mmu_lock From: James Houghton To: Sean Christopherson , Paolo Bonzini Cc: David Matlack , David Rientjes , James Houghton , Marc Zyngier , Oliver Upton , Wei Xu , Yu Zhao , Axel Rasmussen , kvm@vger.kernel.org, linux-kernel@vger.kernel.org Content-Transfer-Encoding: quoted-printable Content-Type: text/plain; charset="utf-8" From: Sean Christopherson Steal another bit from rmap entries (which are word aligned pointers, i.e. have 2 free bits on 32-bit KVM, and 3 free bits on 64-bit KVM), and use the bit to implement a *very* rudimentary per-rmap spinlock. The only anticipated usage of the lock outside of mmu_lock is for aging gfns, and collisions between aging and other MMU rmap operations are quite rare, e.g. unless userspace is being silly and aging a tiny range over and over in a tight loop, time between contention when aging an actively running VM is O(seconds). In short, a more sophisticated locking scheme shouldn't be necessary. Note, the lock only protects the rmap structure itself, SPTEs that are pointed at by a locked rmap can still be modified and zapped by another task (KVM drops/zaps SPTEs before deleting the rmap entries) Signed-off-by: Sean Christopherson Co-developed-by: James Houghton Signed-off-by: James Houghton --- arch/x86/include/asm/kvm_host.h | 3 +- arch/x86/kvm/mmu/mmu.c | 107 ++++++++++++++++++++++++++++---- 2 files changed, 98 insertions(+), 12 deletions(-) diff --git a/arch/x86/include/asm/kvm_host.h b/arch/x86/include/asm/kvm_hos= t.h index 0e44fc1cec0d..bd18fde99116 100644 --- a/arch/x86/include/asm/kvm_host.h +++ b/arch/x86/include/asm/kvm_host.h @@ -28,6 +28,7 @@ #include #include #include +#include =20 #include #include @@ -406,7 +407,7 @@ union kvm_cpu_role { }; =20 struct kvm_rmap_head { - unsigned long val; + atomic_long_t val; }; =20 struct kvm_pio_request { diff --git a/arch/x86/kvm/mmu/mmu.c b/arch/x86/kvm/mmu/mmu.c index a24cf8ddca7f..267cf2d4c3e3 100644 --- a/arch/x86/kvm/mmu/mmu.c +++ b/arch/x86/kvm/mmu/mmu.c @@ -853,11 +853,95 @@ static struct kvm_memory_slot *gfn_to_memslot_dirty_b= itmap(struct kvm_vcpu *vcpu * About rmap_head encoding: * * If the bit zero of rmap_head->val is clear, then it points to the only = spte - * in this rmap chain. Otherwise, (rmap_head->val & ~1) points to a struct + * in this rmap chain. Otherwise, (rmap_head->val & ~3) points to a struct * pte_list_desc containing more mappings. */ #define KVM_RMAP_MANY BIT(0) =20 +/* + * rmaps and PTE lists are mostly protected by mmu_lock (the shadow MMU al= ways + * operates with mmu_lock held for write), but rmaps can be walked without + * holding mmu_lock so long as the caller can tolerate SPTEs in the rmap c= hain + * being zapped/dropped _while the rmap is locked_. + * + * Other than the KVM_RMAP_LOCKED flag, modifications to rmap entries must= be + * done while holding mmu_lock for write. This allows a task walking rmaps + * without holding mmu_lock to concurrently walk the same entries as a task + * that is holding mmu_lock but _not_ the rmap lock. Neither task will mo= dify + * the rmaps, thus the walks are stable. + * + * As alluded to above, SPTEs in rmaps are _not_ protected by KVM_RMAP_LOC= KED, + * only the rmap chains themselves are protected. E.g. holding an rmap's = lock + * ensures all "struct pte_list_desc" fields are stable. + */ +#define KVM_RMAP_LOCKED BIT(1) + +static unsigned long kvm_rmap_lock(struct kvm_rmap_head *rmap_head) +{ + unsigned long old_val, new_val; + + /* + * Elide the lock if the rmap is empty, as lockless walkers (read-only + * mode) don't need to (and can't) walk an empty rmap, nor can they add + * entries to the rmap. I.e. the only paths that process empty rmaps + * do so while holding mmu_lock for write, and are mutually exclusive. + */ + old_val =3D atomic_long_read(&rmap_head->val); + if (!old_val) + return 0; + + do { + /* + * If the rmap is locked, wait for it to be unlocked before + * trying acquire the lock, e.g. to bounce the cache line. + */ + while (old_val & KVM_RMAP_LOCKED) { + cpu_relax(); + old_val =3D atomic_long_read(&rmap_head->val); + } + + /* + * Recheck for an empty rmap, it may have been purged by the + * task that held the lock. + */ + if (!old_val) + return 0; + + new_val =3D old_val | KVM_RMAP_LOCKED; + /* + * Use try_cmpxchg_acquire to prevent reads and writes to the rmap + * from being reordered outside of the critical section created by + * kvm_rmap_lock. + * + * Pairs with smp_store_release in kvm_rmap_unlock. + * + * For the !old_val case, no ordering is needed, as there is no rmap + * to walk. + */ + } while (!atomic_long_try_cmpxchg_acquire(&rmap_head->val, &old_val, new_= val)); + + /* Return the old value, i.e. _without_ the LOCKED bit set. */ + return old_val; +} + +static void kvm_rmap_unlock(struct kvm_rmap_head *rmap_head, + unsigned long new_val) +{ + WARN_ON_ONCE(new_val & KVM_RMAP_LOCKED); + /* + * Ensure that all accesses to the rmap have completed + * before we actually unlock the rmap. + * + * Pairs with the atomic_long_try_cmpxchg_acquire in kvm_rmap_lock. + */ + atomic_long_set_release(&rmap_head->val, new_val); +} + +static unsigned long kvm_rmap_get(struct kvm_rmap_head *rmap_head) +{ + return atomic_long_read(&rmap_head->val) & ~KVM_RMAP_LOCKED; +} + /* * Returns the number of pointers in the rmap chain, not counting the new = one. */ @@ -868,7 +952,7 @@ static int pte_list_add(struct kvm_mmu_memory_cache *ca= che, u64 *spte, struct pte_list_desc *desc; int count =3D 0; =20 - old_val =3D rmap_head->val; + old_val =3D kvm_rmap_lock(rmap_head); =20 if (!old_val) { new_val =3D (unsigned long)spte; @@ -900,7 +984,7 @@ static int pte_list_add(struct kvm_mmu_memory_cache *ca= che, u64 *spte, desc->sptes[desc->spte_count++] =3D spte; } =20 - rmap_head->val =3D new_val; + kvm_rmap_unlock(rmap_head, new_val); =20 return count; } @@ -948,7 +1032,7 @@ static void pte_list_remove(struct kvm *kvm, u64 *spte, unsigned long rmap_val; int i; =20 - rmap_val =3D rmap_head->val; + rmap_val =3D kvm_rmap_lock(rmap_head); if (KVM_BUG_ON_DATA_CORRUPTION(!rmap_val, kvm)) goto out; =20 @@ -974,7 +1058,7 @@ static void pte_list_remove(struct kvm *kvm, u64 *spte, } =20 out: - rmap_head->val =3D rmap_val; + kvm_rmap_unlock(rmap_head, rmap_val); } =20 static void kvm_zap_one_rmap_spte(struct kvm *kvm, @@ -992,7 +1076,7 @@ static bool kvm_zap_all_rmap_sptes(struct kvm *kvm, unsigned long rmap_val; int i; =20 - rmap_val =3D rmap_head->val; + rmap_val =3D kvm_rmap_lock(rmap_head); if (!rmap_val) return false; =20 @@ -1011,13 +1095,13 @@ static bool kvm_zap_all_rmap_sptes(struct kvm *kvm, } out: /* rmap_head is meaningless now, remember to reset it */ - rmap_head->val =3D 0; + kvm_rmap_unlock(rmap_head, 0); return true; } =20 unsigned int pte_list_count(struct kvm_rmap_head *rmap_head) { - unsigned long rmap_val =3D rmap_head->val; + unsigned long rmap_val =3D kvm_rmap_get(rmap_head); struct pte_list_desc *desc; =20 if (!rmap_val) @@ -1083,7 +1167,7 @@ struct rmap_iterator { static u64 *rmap_get_first(struct kvm_rmap_head *rmap_head, struct rmap_iterator *iter) { - unsigned long rmap_val =3D rmap_head->val; + unsigned long rmap_val =3D kvm_rmap_get(rmap_head); u64 *sptep; =20 if (!rmap_val) @@ -1418,7 +1502,7 @@ static void slot_rmap_walk_next(struct slot_rmap_walk= _iterator *iterator) while (++iterator->rmap <=3D iterator->end_rmap) { iterator->gfn +=3D KVM_PAGES_PER_HPAGE(iterator->level); =20 - if (iterator->rmap->val) + if (atomic_long_read(&iterator->rmap->val)) return; } =20 @@ -2444,7 +2528,8 @@ static int mmu_page_zap_pte(struct kvm *kvm, struct k= vm_mmu_page *sp, * avoids retaining a large number of stale nested SPs. */ if (tdp_enabled && invalid_list && - child->role.guest_mode && !child->parent_ptes.val) + child->role.guest_mode && + !atomic_long_read(&child->parent_ptes.val)) return kvm_mmu_prepare_zap_page(kvm, child, invalid_list); } --=20 2.48.1.362.g079036d154-goog