From patchwork Wed Apr 24 15:49:08 2024 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Zhao Liu X-Patchwork-Id: 1927246 Return-Path: X-Original-To: incoming@patchwork.ozlabs.org Delivered-To: patchwork-incoming@legolas.ozlabs.org Authentication-Results: legolas.ozlabs.org; dkim=pass (2048-bit key; unprotected) header.d=intel.com header.i=@intel.com header.a=rsa-sha256 header.s=Intel header.b=I75o2wBs; dkim-atps=neutral Authentication-Results: legolas.ozlabs.org; spf=pass (sender SPF authorized) smtp.mailfrom=nongnu.org (client-ip=209.51.188.17; helo=lists.gnu.org; envelope-from=qemu-devel-bounces+incoming=patchwork.ozlabs.org@nongnu.org; receiver=patchwork.ozlabs.org) Received: from lists.gnu.org (lists.gnu.org [209.51.188.17]) (using TLSv1.2 with cipher ECDHE-ECDSA-AES256-GCM-SHA384 (256/256 bits)) (No client certificate requested) by legolas.ozlabs.org (Postfix) with ESMTPS id 4VPjlK3yKQz1yZy for ; Thu, 25 Apr 2024 01:36:45 +1000 (AEST) Received: from localhost ([::1] helo=lists1p.gnu.org) by lists.gnu.org with esmtp (Exim 4.90_1) (envelope-from ) id 1rzefH-0004Nc-Sk; Wed, 24 Apr 2024 11:36:07 -0400 Received: from eggs.gnu.org ([2001:470:142:3::10]) by lists.gnu.org with esmtps (TLS1.2:ECDHE_RSA_AES_256_GCM_SHA384:256) (Exim 4.90_1) (envelope-from ) id 1rzefF-0004NE-TN for qemu-devel@nongnu.org; Wed, 24 Apr 2024 11:36:05 -0400 Received: from mgamail.intel.com ([192.198.163.13]) by eggs.gnu.org with esmtps (TLS1.2:ECDHE_RSA_AES_256_GCM_SHA384:256) (Exim 4.90_1) (envelope-from ) id 1rzefC-0005Zl-2t for qemu-devel@nongnu.org; Wed, 24 Apr 2024 11:36:05 -0400 DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/simple; d=intel.com; i=@intel.com; q=dns/txt; s=Intel; t=1713972962; x=1745508962; h=from:to:cc:subject:date:message-id:mime-version: content-transfer-encoding; bh=XWfkZ9VA2R2HJERCd5//qZedMkLwW/aqjQy3eLL3++s=; b=I75o2wBsdqSQxMs7Xw/uANfNfPCQJa88SjOdMq7xrhjLPg5Y30mr5FZb oiyRVCrEVaMzeGP41zxiM+MA5USBRbAbTi1WVdWybhx/FAs6ZXYd4TSn1 ptufgulQKmQX6zmAv1UaNABJ42Udsb5uXnncQfhD2eSupzQa3pTOLJEPN UcVXPRiIMN3/u3ViSP7Wyjxu9v15qu0aHK1va6TgRhz/y5uaX21aO5iKm VKQcQvQhd3ddaIeEWzmhAjGLHOJh7pmtMXDg25KjaclIS+QTjQVcJNyg9 ZFkcuL+mhoYzyqiqwJkj3vd5XoxvnFWwvIrv1O+iyj4WdDeL0BULd8hNK A==; X-CSE-ConnectionGUID: aAv7ynMIQxqjw1IyQ5MSfA== X-CSE-MsgGUID: 9Kuk6c09R6ej8/MQ1AQd/g== X-IronPort-AV: E=McAfee;i="6600,9927,11054"; a="12545546" X-IronPort-AV: E=Sophos;i="6.07,226,1708416000"; d="scan'208";a="12545546" Received: from orviesa008.jf.intel.com ([10.64.159.148]) by fmvoesa107.fm.intel.com with ESMTP/TLS/ECDHE-RSA-AES256-GCM-SHA384; 24 Apr 2024 08:35:58 -0700 X-CSE-ConnectionGUID: vnfp1TfESwCemLa7d6mhUw== X-CSE-MsgGUID: mb7SF0WLRuGe4PFSdNoCXg== X-ExtLoop1: 1 X-IronPort-AV: E=Sophos;i="6.07,226,1708416000"; d="scan'208";a="25363034" Received: from liuzhao-optiplex-7080.sh.intel.com ([10.239.160.36]) by orviesa008.jf.intel.com with ESMTP; 24 Apr 2024 08:35:53 -0700 From: Zhao Liu To: Eduardo Habkost , Marcel Apfelbaum , =?utf-8?q?Philippe_Mathieu-D?= =?utf-8?q?aud=C3=A9?= , Yanan Wang , "Michael S . Tsirkin" , Richard Henderson , Paolo Bonzini , Eric Blake , Markus Armbruster , Marcelo Tosatti , =?utf-8?q?Daniel_P_=2E_Berrang=C3=A9?= Cc: qemu-devel@nongnu.org, kvm@vger.kernel.org, Zhenyu Wang , Zhuocheng Ding , Babu Moger , Xiaoyao Li , Dapeng Mi , Yongwei Ma , Zhao Liu Subject: [PATCH v11 00/21] i386: Introduce smp.modules and clean up cache topology Date: Wed, 24 Apr 2024 23:49:08 +0800 Message-Id: <20240424154929.1487382-1-zhao1.liu@intel.com> X-Mailer: git-send-email 2.34.1 MIME-Version: 1.0 Received-SPF: pass client-ip=192.198.163.13; envelope-from=zhao1.liu@intel.com; helo=mgamail.intel.com X-Spam_score_int: -50 X-Spam_score: -5.1 X-Spam_bar: ----- X-Spam_report: (-5.1 / 5.0 requ) BAYES_00=-1.9, DKIMWL_WL_HIGH=-0.668, DKIM_SIGNED=0.1, DKIM_VALID=-0.1, DKIM_VALID_AU=-0.1, DKIM_VALID_EF=-0.1, RCVD_IN_DNSWL_MED=-2.3, SPF_HELO_NONE=0.001, SPF_PASS=-0.001 autolearn=ham autolearn_force=no X-Spam_action: no action X-BeenThere: qemu-devel@nongnu.org X-Mailman-Version: 2.1.29 Precedence: list List-Id: List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Errors-To: qemu-devel-bounces+incoming=patchwork.ozlabs.org@nongnu.org Sender: qemu-devel-bounces+incoming=patchwork.ozlabs.org@nongnu.org Hi maintainers, This is the my v11 patch series, rebased on the master branch at the commit 88daa112d4ed ("Merge tag 'migration-20240423-pull-request' of https://gitlab.com/peterx/qemu into staging"). Since v10 didn't receive comment so v11 just rebase on the latest v9.0 to resolve minor conflicts to facilitate maintainer. ;-) This is a larger series, and almost all patches (containing all the key changes) has gotten Reviewed-by and Acked-by, and tested from both Intel and AMD. Thank you all for your patience and help! Abstract ======== Intel's hybrid Client platform and E core server platform introduce module level and share L2 cache on the module level, in order to configure the CPU/cache topology for the Guest to be consistent with Host's, this series did the following work: * Add now "module" CPU topology level for x86 CPU. * Refacter cache topology encoding for x86 CPU (This is base to support the L2 per module). So, this series is also necessary to support subsequent user configurations of cache topology (via -smp, [2]) and Intel heterogeneous CPU topology ([3] and [4]). Background ========== At present, x86 defaults L2 cache is shared in one core, but this is not enough. There're some platforms that multiple cores share the same L2 cache, e.g., Alder Lake-P shares L2 cache for one module of Atom cores, that is, every four Atom cores shares one L2 cache. On E core server platform, there's the similar L2 per module topology. Therefore, we need the new CPU topology level. Another reason is that Intel client hybrid architectures organize P cores and E cores via module, so a new CPU topology level is necessary to support hybrid CPU topology! Why We Introduce Module Instead of Reusing Cluster -------------------------------------------------- For the discussion in v7 about whether we should reuse current smp.clusters for x86 module, the core point is what's the essential differences between x86 module and general cluster. Since, cluster (for ARM/riscv) lacks a comprehensive and rigorous hardware definition, and judging from the description of smp.clusters [5] when it was introduced by QEMU, x86 module is very similar to general smp.clusters: they are all a layer above existing core level to organize the physical cores and share L2 cache. But there are following reasons that drive us to introduce the new smp.modules: * As the CPU topology abstraction in device tree [6], cluster supports nesting (though currently QEMU hasn't support that). In contrast, (x86) module does not support nesting. * Due to nesting, there is great flexibility in sharing resources on cluster, rather than narrowing cluster down to sharing L2 (and L3 tags) as the lowest topology level that contains cores. * Flexible nesting of cluster allows it to correspond to any level between the x86 package and core. * In Linux kernel, x86's cluster only represents the L2 cache domain but QEMU's smp.clusters is the CPU topology level. Linux kernel will also expose module level topology information in sysfs for x86. To avoid cluster ambiguity and keep a consistent CPU topology naming style with the Linux kernel, we introduce module level for x86. Based on the above considerations, and in order to eliminate the naming confusion caused by the mapping between general cluster and x86 module, we now formally introduce smp.modules as the new topology level. Where to Place Module in Existing Topology Levels ------------------------------------------------- The module is, in existing hardware practice, the lowest layer that contains the core, while the cluster is able to have a higher topological scope than the module due to its nesting. Therefore, we place the module between the cluster and the core: drawer/book/socket/die/cluster/module/core/thread Patch Series Overview ===================== Introduction of Module Level in -smp ------------------------------------ First, a new module level is introduced in the -smp related code to support the module topology in subsequent x86 parts. Users can specify the number of modules (in one die) for a PC machine with "-smp modules=*". Why not Share L2 Cache in Module Directly ----------------------------------------- Though one of module's goals is to implement L2 cache per module, directly using module to define x86's L2 cache topology will cause the compatibility problem: Currently, x86 defaults that the L2 cache is shared in one core, which actually implies a default setting "cores per L2 cache is 1" and therefore implicitly defaults to having as many L2 caches as cores. For example (i386 PC machine): -smp 16,sockets=2,dies=2,cores=2,threads=2,maxcpus=16 (*) Considering the topology of the L2 cache, this (*) implicitly means "1 core per L2 cache" and "2 L2 caches per die". If we use module to configure L2 cache topology with the new default setting "modules per L2 cache is 1", the above semantics will change to "2 cores per module" and "1 module per L2 cache", that is, "2 cores per L2 cache". So the same command (*) will cause changes in the L2 cache topology, further affecting the performance of the virtual machine. Therefore, x86 should only treat module as a cpu topology level and avoid using it to change L2 cache by default for compatibility. Thereby, we need another way to allow user to configure cache topology, this is anther RFC [2]. Module Level in CPUID --------------------- Linux kernel (from v6.4, with commit edc0a2b595765 ("x86/topology: Fix erroneous smp_num_siblings on Intel Hybrid platforms") is able to handle platforms with Module level enumerated via CPUID.1F. Expose the module level in CPUID[0x1F] (for Intel CPUs) if the machine has more than 1 modules since v3. New Cache Topology Info in CPUCacheInfo --------------------------------------- (This is in preparation for users being able to configure cache topology from the command line later on.) Currently, by default, the cache topology is encoded as: 1. i/d cache is shared in one core. 2. L2 cache is shared in one core. 3. L3 cache is shared in one die. This default general setting has caused a misunderstanding, that is, the cache topology is completely equated with a specific CPU topology, such as the connection between L2 cache and core level, and the connection between L3 cache and die level. In fact, the settings of these topologies depend on the specific platform and are not static. For example, on Alder Lake-P, every four Atom cores share the same L2 cache [3]. Thus, in this patch set, we explicitly define the corresponding cache topology for different cache models and this has two benefits: 1. Easy to expand to new cache models with different topology in the future. 2. It can easily support custom cache topology by some command. Patch Description ================= Patch 01-04: Add module support in -smp. Patch 05: Fix Intel L1 cache topology. Patch 06-08: Clean up cache topology related CPUID encoding and QEMU topology variables. Patch 09-11: Refactor CPUID[0x1F] (CPU topology) encoding to prepare to introduce module level. Patch 12-18: Add the module as the new CPU topology level in x86. Patch 19-21: Refactor cache topology encoding for Intel and AMD. Reference ========= [1]: https://lore.kernel.org/qemu-devel/20240321144048.3699388-1-zhao1.liu@linux.intel.com/ [2]: https://lore.kernel.org/qemu-devel/20240220092504.726064-1-zhao1.liu@linux.intel.com/ [3]: https://lore.kernel.org/qemu-devel/20230213095035.158240-1-zhao1.liu@linux.intel.com/ [4]: https://lore.kernel.org/qemu-devel/20231130144203.2307629-1-zhao1.liu@linux.intel.com/ Thanks and Best Regards, Zhao --- Changelog: Changes since v10: * Rebased on commit 88daa11. Changes since v9: * Collected a/b, t/b and r/b tags. * Fixed typos. * Minor cleanup of code. * Added more comments and polished commit message. Changes since v8: * Added the reason of why a new module level is needed in commit message. (Markus). * Added the description about how Linux kernel supports x86 module level in commit message. (Daniel) * Added module description in qemu_smp_opts. * Added missing "modules" parameter of -smp example in documentation. * Added Philippe's reviewed-by tag. Changes since v7 (main changes): * Introduced smp.modules as a new CPU topology level. (Xiaoyao) * Fixed calculations of cache_info_passthrough case in the patch "i386/cpu: Use APIC ID info to encode cache topo in CPUID[4]". (Xiaoyao) * Moved the patch "i386/cpu: Use APIC ID info get NumSharingCache for CPUID[0x8000001D].EAX[bits 25:14]" after CPUID[4]'s similar change ("i386/cpu: Use APIC ID offset to encode cache topo in CPUID[4]"). (Xiaoyao) * Introduced a bitmap in CPUX86State to cache available CPU topology levels. * Refactored the encode_topo_cpuid1f() to use traversal to search the encoded level and avoid using static variables. * Mapped x86 module to smp module instead of cluster. * Dropped Michael/Babu's ACKed/Tested tags for most patches since the code change. Changes since v6: * Updated the comment when check cluster-id. Since there's no v8.2, the cluster-id support should at least start from v9.0. * Rebased on commit d328fef93ae7 ("Merge tag 'pull-20231230' of https://gitlab.com/rth7680/qemu into staging"). Changes since v5: * The first four patches of v5 [1] have been merged, v6 contains the remaining patches. * Reabsed on the latest master. * Updated the comment when check cluster-id. Since current QEMU is v8.2, the cluster-id support should at least start from v8.3. Changes since v4: * Dropped the "x-l2-cache-topo" option. (Michael) * Added A/R/T tags. Changes since v3 (main changes): * Exposed module level in CPUID[0x1F]. * Fixed compile warnings. (Babu) * Fixed cache topology uninitialization bugs for some AMD CPUs. (Babu) Changes since v2: * Added "Tested-by", "Reviewed-by" and "ACKed-by" tags. * Used newly added wrapped helper to get cores per socket in qemu_init_vcpu(). Changes since v1: * Reordered patches. (Yanan) * Deprecated the patch to fix comment of machine_parse_smp_config(). (Yanan) * Renamed test-x86-cpuid.c to test-x86-topo.c. (Yanan) * Split the intel's l1 cache topology fix into a new separate patch. (Yanan) * Combined module_id and APIC ID for module level support into one patch. (Yanan) * Made cache_into_passthrough case of cpuid 0x04 leaf also encode via APICID way. (Yanan) * cpu_x86_cpuid() used max_processor_ids_for_cache() and max_core_ids_in_package() to encode CPUID[4]. (Yanan) * Added the prefix "CPU_TOPO_LEVEL_*" for CPU topology level names. (Yanan) --- Zhao Liu (20): hw/core/machine: Introduce the module as a CPU topology level hw/core/machine: Support modules in -smp hw/core: Introduce module-id as the topology subindex hw/core: Support module-id in numa configuration i386/cpu: Fix i/d-cache topology to core level for Intel CPU i386/cpu: Use APIC ID info to encode cache topo in CPUID[4] i386/cpu: Use APIC ID info get NumSharingCache for CPUID[0x8000001D].EAX[bits 25:14] i386/cpu: Consolidate the use of topo_info in cpu_x86_cpuid() i386/cpu: Introduce bitmap to cache available CPU topology levels i386: Split topology types of CPUID[0x1F] from the definitions of CPUID[0xB] i386/cpu: Decouple CPUID[0x1F] subleaf with specific topology level i386: Introduce module level cpu topology to CPUX86State i386: Support modules_per_die in X86CPUTopoInfo i386: Expose module level in CPUID[0x1F] i386: Support module_id in X86CPUTopoIDs i386/cpu: Introduce module-id to X86CPU hw/i386/pc: Support smp.modules for x86 PC machine i386: Add cache topology info in CPUCacheInfo i386/cpu: Use CPUCacheInfo.share_level to encode CPUID[4] i386/cpu: Use CPUCacheInfo.share_level to encode CPUID[0x8000001D].EAX[bits 25:14] Zhuocheng Ding (1): tests: Add test case of APIC ID for module level parsing hw/core/machine-hmp-cmds.c | 4 + hw/core/machine-smp.c | 41 ++++- hw/core/machine.c | 18 +++ hw/i386/pc.c | 1 + hw/i386/x86.c | 67 +++++++-- include/hw/boards.h | 4 + include/hw/i386/topology.h | 60 +++++++- qapi/machine.json | 7 + qemu-options.hx | 18 ++- system/vl.c | 3 + target/i386/cpu.c | 301 +++++++++++++++++++++++++++++-------- target/i386/cpu.h | 29 +++- target/i386/kvm/kvm.c | 3 +- tests/unit/test-x86-topo.c | 56 ++++--- 14 files changed, 490 insertions(+), 122 deletions(-)