From mboxrd@z Thu Jan 1 00:00:00 1970 Received: from mail-pj1-f73.google.com (mail-pj1-f73.google.com [209.85.216.73]) (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 A5EBA18C91F for ; Mon, 2 Jun 2025 23:30:14 +0000 (UTC) Authentication-Results: smtp.subspace.kernel.org; arc=none smtp.client-ip=209.85.216.73 ARC-Seal:i=1; a=rsa-sha256; d=subspace.kernel.org; s=arc-20240116; t=1748907016; cv=none; b=dR8FVEm1C969LiLFzH+RHImfyQsSjP8yMu7YGgs859A53+U6cKSglqQFx7Yw1F7bcbQWLA675fSSA1kwudIvteEK9CdXaf8tw0VEU+ETVeJmq9uJubkuPtZTKfjSnuE0x4ON947DroPtqwyNvn9iqDtCQXD3s8o7rXVfx0UYTGM= ARC-Message-Signature:i=1; a=rsa-sha256; d=subspace.kernel.org; s=arc-20240116; t=1748907016; c=relaxed/simple; bh=ERPVRkIbILIUcIjfV+EhBnHrMs9fZ3Q4qRlPwExSKOU=; h=Date:In-Reply-To:Mime-Version:References:Message-ID:Subject:From: To:Cc:Content-Type; b=UwBSWFw9qaw0iuyd5e6Ck8/4sPQ9rZo7n5LA3W1x0naAy1HuMEukIzPOuXZTDiIywkm4Ga9PAjF+IZkFKRue/D6k0pKgUrMcHeTgKm4zG4hncN2foigBjmf14lxewOvtAImuu5qqj5Dj6KfMdOJSYHOuZ3D35WDdF66Fam7WTu8= ARC-Authentication-Results:i=1; smtp.subspace.kernel.org; dmarc=pass (p=reject dis=none) header.from=google.com; spf=pass smtp.mailfrom=flex--seanjc.bounces.google.com; dkim=pass (2048-bit key) header.d=google.com header.i=@google.com header.b=gsebseMk; arc=none smtp.client-ip=209.85.216.73 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--seanjc.bounces.google.com Authentication-Results: smtp.subspace.kernel.org; dkim=pass (2048-bit key) header.d=google.com header.i=@google.com header.b="gsebseMk" Received: by mail-pj1-f73.google.com with SMTP id 98e67ed59e1d1-310efe825ccso4872247a91.3 for ; Mon, 02 Jun 2025 16:30:14 -0700 (PDT) DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=google.com; s=20230601; t=1748907014; x=1749511814; darn=lists.linux.dev; h=cc:to:from:subject:message-id:references:mime-version:in-reply-to :date:from:to:cc:subject:date:message-id:reply-to; bh=Yve1ZiL8T7g84JmBt/tikyI/6Bv4C2Bg0e+B1xxTdfo=; b=gsebseMkxi67MImod5O2Agr7XBZHfK+dEjq660ukYsh1g0hngTwFPhqNZfweUHrNxC s3WvJa9tUKB7vgAuKC1+vaGelv+kwHIZI23HFe0VSf6QVTKm9tsCiBkOE2p5O/umVQ1d EAaT0VBHI6yXLwM6ybd0pgKlmhM+Ce/Z5/IYqSS224OUnYj8k4u/4kETvFb0N4kmUza1 gw0JOSqGeCl0zVrN/jUlJqitzJbG4CHo/zY/ccz2JyS4xIc9h1XLfkOTyK2Ifr88ja8c 9c+a/ty3Jxf1wZQLIz164gsnFLMcW3B4kwKpOocz/36wO3ZKs+agaqSLKPS03aCxvyo0 KMbw== X-Google-DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=1e100.net; s=20230601; t=1748907014; x=1749511814; 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=Yve1ZiL8T7g84JmBt/tikyI/6Bv4C2Bg0e+B1xxTdfo=; b=gKiyjS/dhs8cHsePImO37JZVT53T2PfBGliQIKcGrwYIfh3Ga+fTWLWXsJmZORW9eI AemfLaqmMUvGyphoGuG9/JtUG3Mt6O5+ThoKeAY9qilAGq3pox+K+kTrlZUmxgFAux99 XchP5RWyB0qxjG3yACDSVky/Rj8jdmEzhP4+G8i9+BXtlJ5LcAUaoczHquVJyWlueWyH bwa56bDSiGz3FRkmfDZhobAWmNTkUEndFhd6RmhHvbACDHfjfTT/LqnQbJjmlxJ4Bz9p 1KNgHNb+8TMxadgL4g6DqW2nW+OZMGtIK6rQLb8xqrWUP2jtLIsKqceoZC9cn4HbTfsd ZcuA== X-Forwarded-Encrypted: i=1; AJvYcCXXPLGtwXuxJ9TIxuhSry9Jx4C4dmMj0UNfNnQKb5Xnyv3aAVpiItfgvQD+oY5MWMC3k1akVIvZInWqgK3pUg==@lists.linux.dev X-Gm-Message-State: AOJu0Yx5jpqgDwLym724GNdHhBw3laKAlR5JaDRPPWON/I/SNtXlbz15 w/H6wBW2/p4UNUmb21ry104zcF8QY1am/nSL3y9pl96Ie6yyeAur55KfqzZ4x9GsoXJCQFYu8Hx z9xno+g== X-Google-Smtp-Source: AGHT+IE14BFeGM0J7uhuk33V5fPG5r1MtBBSniKvSePrbW7/nzVp1S6VXU7zoEWKiW8wpnHDPojK7VzfMl4= X-Received: from pjur7.prod.google.com ([2002:a17:90a:d407:b0:311:ea2a:3919]) (user=seanjc job=prod-delivery.src-stubby-dispatcher) by 2002:a17:90b:4f4d:b0:312:1cd7:b337 with SMTP id 98e67ed59e1d1-3125034a47amr19482649a91.5.1748907013702; Mon, 02 Jun 2025 16:30:13 -0700 (PDT) Date: Mon, 2 Jun 2025 16:30:12 -0700 In-Reply-To: <20250602125442.19d41098.alex.williamson@redhat.com> Precedence: bulk X-Mailing-List: virtualization@lists.linux.dev List-Id: List-Subscribe: List-Unsubscribe: Mime-Version: 1.0 References: <20250516230734.2564775-1-seanjc@google.com> <20250602125442.19d41098.alex.williamson@redhat.com> Message-ID: Subject: Re: [PATCH v2 0/8] irqbypass: Cleanups and a perf improvement From: Sean Christopherson To: Alex Williamson Cc: Paolo Bonzini , "Michael S. Tsirkin" , Jason Wang , kvm@vger.kernel.org, virtualization@lists.linux.dev, netdev@vger.kernel.org, linux-kernel@vger.kernel.org, Kevin Tian , Oliver Upton , David Matlack , Like Xu , Binbin Wu , Yong He Content-Type: text/plain; charset="us-ascii" On Mon, Jun 02, 2025, Alex Williamson wrote: > On Fri, 16 May 2025 16:07:26 -0700 > Sean Christopherson wrote: > > > The two primary goals of this series are to make the irqbypass concept > > easier to understand, and to address the terrible performance that can > > result from using a list to track connections. > > > > For the first goal, track the producer/consumer "tokens" as eventfd context > > pointers instead of opaque "void *". Supporting arbitrary token types was > > dead infrastructure when it was added 10 years ago, and nothing has changed > > since. Taking an opaque token makes a very simple concept (device signals > > eventfd; KVM listens to eventfd) unnecessarily difficult to understand. > > > > Burying that simple behind a layer of obfuscation also makes the overall > > code more brittle, as callers can pass in literally anything. I.e. passing > > in a token that will never be paired would go unnoticed. > > > > For the performance issue, use an xarray. I'm definitely not wedded to an > > xarray, but IMO it doesn't add meaningful complexity (even requires less > > code), and pretty much Just Works. Like tried this a while back[1], but > > the implementation had undesirable behavior changes and stalled out. > > > > Note, I want to do more aggressive cleanups of irqbypass at some point, > > e.g. not reporting an error to userspace if connect() fails is awful > > behavior for environments that want/need irqbypass to always work. And > > KVM shold probably have a KVM_IRQFD_FLAG_NO_IRQBYPASS if a VM is never going > > to use device posted interrupts. But those are future problems. > > > > v2: > > - Collect reviews. [Kevin, Michael] > > - Track the pointer as "struct eventfd_ctx *eventfd" instead of "void *token". > > [Alex] > > - Fix typos and stale comments. [Kevin, Binbin] > > - Use "trigger" instead of the null token/eventfd pointer on failure in > > vfio_msi_set_vector_signal(). [Kevin] > > - Drop a redundant "tmp == consumer" check from patch 3. [Kevin] > > - Require producers to pass in the line IRQ number. > > > > v1: https://lore.kernel.org/all/20250404211449.1443336-1-seanjc@google.com > > > > [1] https://lore.kernel.org/all/20230801115646.33990-1-likexu@tencent.com > > [2] https://lore.kernel.org/all/20250401161804.842968-1-seanjc@google.com > > > > Sean Christopherson (8): > > irqbypass: Drop pointless and misleading THIS_MODULE get/put > > irqbypass: Drop superfluous might_sleep() annotations > > irqbypass: Take ownership of producer/consumer token tracking > > irqbypass: Explicitly track producer and consumer bindings > > irqbypass: Use paired consumer/producer to disconnect during > > unregister > > irqbypass: Use guard(mutex) in lieu of manual lock+unlock > > irqbypass: Use xarray to track producers and consumers > > irqbypass: Require producers to pass in Linux IRQ number during > > registration > > > > arch/x86/kvm/x86.c | 4 +- > > drivers/vfio/pci/vfio_pci_intrs.c | 10 +- > > drivers/vhost/vdpa.c | 10 +- > > include/linux/irqbypass.h | 46 ++++---- > > virt/kvm/eventfd.c | 7 +- > > virt/lib/irqbypass.c | 190 +++++++++++------------------- > > 6 files changed, 107 insertions(+), 160 deletions(-) > > > > > > base-commit: 7ef51a41466bc846ad794d505e2e34ff97157f7f > > Sorry for the delay. Heh, no worries. ~2 weeks is downright prompt by my standards ;-) > Do you intend to take this through your trees? Yes, ideally, it would go into Paolo's kvm/next sooner than later (I'll start poking him if necessary). The s/token/eventfd rename creates an annoying conflict in kvm/x86.c with an in-flight patch (significant code movement between files). It would be nice to be able to rebase the in-flight patch instead of having to resolve a merge confict (the conflict itself isn't difficult to resolve, I just find it hard to visually review/audit the resolution due to the code movement).