From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from firstgate.proxmox.com (firstgate.proxmox.com [212.224.123.68]) (using TLSv1.3 with cipher TLS_AES_256_GCM_SHA384 (256/256 bits) key-exchange X25519 server-signature RSA-PSS (2048 bits)) (No client certificate requested) by lists.proxmox.com (Postfix) with ESMTPS id 23A45A25D0 for ; Mon, 19 Jun 2023 09:29:28 +0200 (CEST) Received: from firstgate.proxmox.com (localhost [127.0.0.1]) by firstgate.proxmox.com (Proxmox) with ESMTP id 0675225405 for ; Mon, 19 Jun 2023 09:28:58 +0200 (CEST) Received: from bastionodiso.odiso.net (bastionodiso.odiso.net [185.151.191.93]) (using TLSv1.3 with cipher TLS_AES_256_GCM_SHA384 (256/256 bits) key-exchange X25519 server-signature RSA-PSS (2048 bits)) (No client certificate requested) by firstgate.proxmox.com (Proxmox) with ESMTPS for ; Mon, 19 Jun 2023 09:28:54 +0200 (CEST) Received: from kvmformation3.odiso.net (formationkvm3.odiso.net [10.3.94.12]) by bastionodiso.odiso.net (Postfix) with ESMTP id 6556F7BB7; Mon, 19 Jun 2023 09:28:45 +0200 (CEST) Received: by kvmformation3.odiso.net (Postfix, from userid 0) id 6418724159F; Mon, 19 Jun 2023 09:28:45 +0200 (CEST) From: Alexandre Derumier To: pve-devel@lists.proxmox.com Date: Mon, 19 Jun 2023 09:28:39 +0200 Message-Id: <20230619072841.38531-11-aderumier@odiso.com> X-Mailer: git-send-email 2.39.2 In-Reply-To: <20230619072841.38531-1-aderumier@odiso.com> References: <20230619072841.38531-1-aderumier@odiso.com> MIME-Version: 1.0 Content-Transfer-Encoding: 8bit X-SPAM-LEVEL: Spam detection results: 0 AWL 0.036 Adjusted score from AWL reputation of From: address BAYES_00 -1.9 Bayes spam probability is 0 to 1% DMARC_MISSING 0.1 Missing DMARC policy HEADER_FROM_DIFFERENT_DOMAINS 0.25 From and EnvelopeFrom 2nd level mail domains are different KAM_DMARC_STATUS 0.01 Test Rule for DKIM or SPF Failure with Strict Alignment KAM_LAZY_DOMAIN_SECURITY 1 Sending domain does not have any anti-forgery methods SPF_HELO_NONE 0.001 SPF: HELO does not publish an SPF Record SPF_NONE 0.001 SPF: sender does not publish an SPF Record T_SCC_BODY_TEXT_LINE -0.01 - Subject: [pve-devel] [PATCH v6 qemu-server 08/10] memory: add virtio-mem support X-BeenThere: pve-devel@lists.proxmox.com X-Mailman-Version: 2.1.29 Precedence: list List-Id: Proxmox VE development discussion List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-List-Received-Date: Mon, 19 Jun 2023 07:29:28 -0000 a 4GiB static memory is needed for DMA+boot memory, as this memory is almost always un-unpluggeable. 1 virtio-mem pci device is setup for each numa node on pci.4 bridge virtio-mem use a fixed blocksize with 32000 blocks Blocksize is computed from the maxmemory-4096/32000 with a minimum of 2MiB to map THP. (lower blocksize = more chance to unplug memory). Note: Currently, linux only support 4MiB virtio blocksize, 2MiB support is currently is progress. For hotplug/unplug, we are try to allocate/unallocate same amount of memory aligned to the blocksize on each numa node if possible. If a node a not able to reach the target memory (could be an unmovable page on unplug for example), we try again to redispatch memory the remaining memory on others nodes. About hugepages: For ordinary memory devices, such as DIMMs, we preallocate memory via the memory backend for such use cases; however, with virtio-mem we're dealing with sparse memory backends; preallocating the whole memory backend destroys the whole purpose of virtio-mem. Instead, we want to preallocate memory when actually exposing memory to the VM dynamically, and fail plugging memory gracefully + warn the user in case preallocation fails. fixes: https://bugzilla.proxmox.com/show_bug.cgi?id=2949 Signed-off-by: Alexandre Derumier --- PVE/API2/Qemu.pm | 9 +- PVE/QemuServer.pm | 8 +- PVE/QemuServer/Memory.pm | 219 ++++++++++++++++++++++++++++++++++++--- PVE/QemuServer/PCI.pm | 8 ++ 4 files changed, 226 insertions(+), 18 deletions(-) diff --git a/PVE/API2/Qemu.pm b/PVE/API2/Qemu.pm index f0b0dda..456ee97 100644 --- a/PVE/API2/Qemu.pm +++ b/PVE/API2/Qemu.pm @@ -32,7 +32,7 @@ use PVE::QemuServer::Drive; use PVE::QemuServer::ImportDisk; use PVE::QemuServer::Monitor qw(mon_cmd); use PVE::QemuServer::Machine; -use PVE::QemuServer::Memory qw(get_current_memory parse_memory get_host_max_mem); +use PVE::QemuServer::Memory qw(get_current_memory parse_memory get_host_max_mem get_virtiomem_block_size); use PVE::QemuServer::PCI; use PVE::QemuServer::USB; use PVE::QemuMigrate; @@ -490,6 +490,13 @@ my $check_memory_param = sub { die "current memory value too large (must be smaller than max memory)\n" if $mem->{max} && $mem->{current} > $mem->{max}; + if ($mem->{virtio}) { + my $blocksize = get_virtiomem_block_size($conf); + + die "memory need to be a multiple of $blocksize MiB when virtiomem is enabled\n" + if $mem->{current} % $blocksize != 0; + } + if ($param->{memory} || defined($param->{balloon})) { my $memory = $param->{memory} || $conf->{pending}->{memory} || $conf->{memory}; diff --git a/PVE/QemuServer.pm b/PVE/QemuServer.pm index 1257902..7e0ade2 100644 --- a/PVE/QemuServer.pm +++ b/PVE/QemuServer.pm @@ -3887,8 +3887,14 @@ sub config_to_command { push @$cmd, get_cpu_options($conf, $arch, $kvm, $kvm_off, $machine_version, $winversion, $gpu_passthrough); } + my $mem_devices = {}; PVE::QemuServer::Memory::config( - $conf, $vmid, $sockets, $cores, $hotplug_features->{memory}, $cmd); + $conf, $vmid, $sockets, $cores, $hotplug_features->{memory}, $cmd, $mem_devices); + + foreach my $id (sort keys %$mem_devices) { + my $pciaddr = print_pci_addr($id, $bridges, $arch, $machine_type); + push @$devices, "-device", "$mem_devices->{$id}$pciaddr"; + } push @$cmd, '-S' if $conf->{freeze}; diff --git a/PVE/QemuServer/Memory.pm b/PVE/QemuServer/Memory.pm index 0d64a5f..dcdf318 100644 --- a/PVE/QemuServer/Memory.pm +++ b/PVE/QemuServer/Memory.pm @@ -3,8 +3,10 @@ package PVE::QemuServer::Memory; use strict; use warnings; +use POSIX qw(ceil); + use PVE::Exception qw(raise raise_param_exc); -use PVE::GuestHelpers qw(safe_num_ne); +use PVE::GuestHelpers qw(safe_boolean_ne safe_num_ne); use PVE::JSONSchema; use PVE::Tools qw(run_command lock_file lock_file_full file_read_firstline dir_glob_foreach); @@ -16,6 +18,7 @@ our @EXPORT_OK = qw( get_current_memory parse_memory get_host_max_mem +get_virtiomem_block_size ); my $MAX_NUMA = 8; @@ -38,6 +41,12 @@ our $memory_fmt = { maximum => 4194304, format => 'pve-qm-memory-max', }, + virtio => { + description => "Use virtio-mem devices for hotplug (Experimental: Only works with Linux guest with kernel >= 5.10)", + type => 'boolean', + optional => 1, + default => 0, + }, }; PVE::JSONSchema::register_format('pve-qm-memory-max', \&verify_qm_memory_max); @@ -74,7 +83,9 @@ my sub get_static_mem { my $static_memory = 0; my $memory = parse_memory($conf->{memory}); - if ($memory->{max}) { + if ($memory->{virtio}) { + $static_memory = 4096; + } elsif ($memory->{max}) { my $dimm_size = $memory->{max} / $MAX_SLOTS; #static mem can't be lower than 4G and lower than 1 dimmsize by socket $static_memory = $dimm_size * $sockets; @@ -163,6 +174,121 @@ sub get_current_memory { return $memory->{current}; } +sub get_virtiomem_block_size { + my ($conf) = @_; + + my $sockets = $conf->{sockets} || 1; + my $MAX_MEM = get_max_mem($conf); + my $static_memory = get_static_mem($conf, $sockets, 1); + my $memory = get_current_memory($conf->{memory}); + + #virtiomem can map 32000 block size. + #try to use lowest blocksize, lower = more chance to unplug memory. + my $blocksize = ($MAX_MEM - $static_memory) / 32000; + #2MB is the minimum to be aligned with THP + $blocksize = 2 if $blocksize < 2; + $blocksize = 2**(ceil(log($blocksize)/log(2))); + #Linux guest kernel only support 4MiB block currently (kernel <= 6.2) + $blocksize = 4 if $blocksize < 4; + + return $blocksize; +} + +my sub get_virtiomem_total_current_size { + my ($mems) = @_; + my $size = 0; + for my $mem (values %$mems) { + $size += $mem->{current}; + } + return $size; +} + +my sub balance_virtiomem { + my ($vmid, $virtiomems, $blocksize, $target_total) = @_; + + my $nb_virtiomem = scalar(keys %$virtiomems); + + print"try to balance memory on $nb_virtiomem virtiomems\n"; + + #if we can't share exactly the same amount, we add the remainder on last node + my $target_aligned = int( $target_total / $nb_virtiomem / $blocksize) * $blocksize; + my $target_remaining = $target_total - ($target_aligned * ($nb_virtiomem-1)); + + my $i = 0; + foreach my $id (sort keys %$virtiomems) { + my $virtiomem = $virtiomems->{$id}; + $i++; + my $virtiomem_target = $i == $nb_virtiomem ? $target_remaining : $target_aligned; + $virtiomem->{completed} = 0; + $virtiomem->{retry} = 0; + $virtiomem->{target} = $virtiomem_target; + + print "virtiomem$id: set-requested-size : $virtiomem_target\n"; + mon_cmd( + $vmid, + 'qom-set', + path => "/machine/peripheral/virtiomem$id", + property => "requested-size", + value => $virtiomem_target * 1024 * 1024 + ); + } + + my $total_finished = 0; + my $error = undef; + + while ($total_finished != $nb_virtiomem) { + + sleep 1; + + $total_finished = 0; + + foreach my $id (keys %$virtiomems) { + + my $virtiomem = $virtiomems->{$id}; + + if ($virtiomem->{error} || $virtiomem->{completed}) { + $total_finished++; + next; + } + + my $size = mon_cmd($vmid, 'qom-get', path => "/machine/peripheral/virtiomem$id", property => "size"); + $virtiomem->{current} = $size / 1024 / 1024; + print"virtiomem$id: last: $virtiomem->{last} current: $virtiomem->{current} target: $virtiomem->{target} retry: $virtiomem->{retry}\n"; + + if($virtiomem->{current} == $virtiomem->{target}) { + print"virtiomem$id: completed\n"; + $virtiomem->{completed} = 1; + next; + } + + if($virtiomem->{current} != $virtiomem->{last}) { + #if value has changed, but not yet completed + $virtiomem->{retry} = 0; + $virtiomem->{last} = $virtiomem->{current}; + next; + } + + if($virtiomem->{retry} >= 5) { + print "virtiomem$id: target memory still not reached, ignoring device from now on\n"; + $virtiomem->{error} = 1; + $error = 1; + #as change is async, we don't want that value change after the api call + eval { + mon_cmd( + $vmid, + 'qom-set', + path => "/machine/peripheral/virtiomem$id", + property => "requested-size", + value => $virtiomem->{current} * 1024 *1024 + ); + }; + } + $virtiomem->{retry}++; + } + } + die "No more available blocks in virtiomem to balance all requested memory\n" if $error; +} + sub get_numa_node_list { my ($conf) = @_; my @numa_map; @@ -249,7 +375,37 @@ sub qemu_memory_hotplug { my $MAX_MEM = get_max_mem($conf); die "you cannot add more memory than max mem $MAX_MEM MB!\n" if $value > $MAX_MEM; - if ($value > $memory) { + if ($oldmem->{virtio}) { + my $blocksize = get_virtiomem_block_size($conf); + + my $virtiomems = {}; + + for (my $i = 0; $i < $sockets; $i++) { + my $size = mon_cmd($vmid, 'qom-get', path => "/machine/peripheral/virtiomem$i", property => "size"); + $size = $size / 1024 /1024; + $virtiomems->{$i} = { + current => $size, + last => $size, + error => 0, + completed => 0, + retry => 0 + }; + } + + my $target_total = $value - $static_memory; + my $err; + eval { + balance_virtiomem($vmid, $virtiomems, $blocksize, $target_total); + }; + $err = $@ if $@; + + my $current_memory = $static_memory + get_virtiomem_total_current_size($virtiomems); + $newmem->{current} = $current_memory; + $conf->{memory} = print_memory($newmem); + PVE::QemuConfig->write_config($vmid, $conf); + die $err if $err; + + } elsif ($value > $memory) { my $numa_hostmap; @@ -342,8 +498,8 @@ sub can_hotplug { my $oldmem = parse_memory($conf->{memory}); my $newmem = parse_memory($value); - return if safe_num_ne($newmem->{max}, $oldmem->{max}); - + return if (safe_num_ne($newmem->{max}, $oldmem->{max}) || + safe_boolean_ne($newmem->{virtio}, $oldmem->{virtio})); return 1; } @@ -365,7 +521,7 @@ sub qemu_memdevices_list { } sub config { - my ($conf, $vmid, $sockets, $cores, $hotplug, $cmd) = @_; + my ($conf, $vmid, $sockets, $cores, $hotplug, $cmd, $mem_devices) = @_; my $memory = get_current_memory($conf->{memory}); my $static_memory = get_static_mem($conf, $sockets, $hotplug); @@ -383,7 +539,10 @@ sub config { die "minimum memory must be ${static_memory}MB\n" if($memory < $static_memory); my $confmem = parse_memory($conf->{memory}); my $slots = $confmem->{max} ? $MAX_SLOTS : 255; - push @$cmd, '-m', "size=${static_memory},slots=$slots,maxmem=${MAX_MEM}M"; + my $cmdstr = "size=${static_memory}"; + $cmdstr .= ",slots=$slots" if !$confmem->{'virtio'}; + $cmdstr .= ",maxmem=${MAX_MEM}M"; + push @$cmd, '-m', $cmdstr; } else { @@ -455,17 +614,42 @@ sub config { } if ($hotplug) { - foreach_dimm($conf, $vmid, $memory, $static_memory, sub { - my ($conf, $vmid, $name, $dimm_size, $numanode, $current_size, $memory) = @_; - my $mem_object = print_mem_object($conf, "mem-$name", $dimm_size); + my $confmem = parse_memory($conf->{memory}); - push @$cmd, "-object" , $mem_object; - push @$cmd, "-device", "pc-dimm,id=$name,memdev=mem-$name,node=$numanode"; + if ($confmem->{'virtio'}) { - die "memory size ($memory) must be aligned to $dimm_size for hotplugging\n" - if $current_size > $memory; - }); + my $MAX_MEM = get_max_mem($conf); + my $node_maxmem = ($MAX_MEM - $static_memory) / $sockets; + my $node_mem = ($memory - $static_memory) / $sockets; + my $blocksize = get_virtiomem_block_size($conf); + + die "memory need to be a multiple of $blocksize MiB with maxmemory $MAX_MEM MiB when virtiomem is enabled\n" + if $memory % $blocksize != 0; + + for (my $i = 0; $i < $sockets; $i++) { + + my $id = "virtiomem$i"; + my $mem_object = print_mem_object($conf, "mem-$id", $node_maxmem); + push @$cmd, "-object" , "$mem_object,reserve=off"; + + my $mem_device = "virtio-mem-pci,block-size=${blocksize}M,requested-size=${node_mem}M,id=$id,memdev=mem-$id,node=$i"; + $mem_device .= ",prealloc=on" if $conf->{hugepages}; + $mem_devices->{$id} = $mem_device; + } + } else { + foreach_dimm($conf, $vmid, $memory, $static_memory, sub { + my ($conf, $vmid, $name, $dimm_size, $numanode, $current_size, $memory) = @_; + + my $mem_object = print_mem_object($conf, "mem-$name", $dimm_size); + + push @$cmd, "-object" , $mem_object; + push @$cmd, "-device", "pc-dimm,id=$name,memdev=mem-$name,node=$numanode"; + + die "memory size ($memory) must be aligned to $dimm_size for hotplugging\n" + if $current_size > $memory; + }); + } } } @@ -476,8 +660,11 @@ sub print_mem_object { my $hugepages_size = hugepages_size($conf, $size); my $path = hugepages_mount_path($hugepages_size); + my $confmem = parse_memory($conf->{memory}); - return "memory-backend-file,id=$id,size=${size}M,mem-path=$path,share=on,prealloc=yes"; + my $object = "memory-backend-file,id=$id,size=${size}M,mem-path=$path,share=on"; + $object .= ",prealloc=yes" if !$confmem->{virtio}; + return $object; } else { return "memory-backend-ram,id=$id,size=${size}M"; } diff --git a/PVE/QemuServer/PCI.pm b/PVE/QemuServer/PCI.pm index 1673041..722c56f 100644 --- a/PVE/QemuServer/PCI.pm +++ b/PVE/QemuServer/PCI.pm @@ -261,6 +261,14 @@ sub get_pci_addr_map { 'scsihw2' => { bus => 4, addr => 1 }, 'scsihw3' => { bus => 4, addr => 2 }, 'scsihw4' => { bus => 4, addr => 3 }, + 'virtiomem0' => { bus => 4, addr => 4 }, + 'virtiomem1' => { bus => 4, addr => 5 }, + 'virtiomem2' => { bus => 4, addr => 6 }, + 'virtiomem3' => { bus => 4, addr => 7 }, + 'virtiomem4' => { bus => 4, addr => 8 }, + 'virtiomem5' => { bus => 4, addr => 9 }, + 'virtiomem6' => { bus => 4, addr => 10 }, + 'virtiomem7' => { bus => 4, addr => 11 }, } if !defined($pci_addr_map); return $pci_addr_map; } -- 2.39.2