public inbox for pve-devel@lists.proxmox.com
 help / color / mirror / Atom feed
From: Dominik Csapak <d.csapak@proxmox.com>
To: pve-devel@lists.proxmox.com
Subject: [pve-devel] [PATCH common v3 3/3] HardwareMap: add support for multiple pci device paths per mapping
Date: Tue, 20 Sep 2022 14:50:15 +0200	[thread overview]
Message-ID: <20220920125041.3636561-10-d.csapak@proxmox.com> (raw)
In-Reply-To: <20220920125041.3636561-1-d.csapak@proxmox.com>

With this, we can now tell qemu-server to choose the first avaiable
devices, which makes using vGPUs and SR-IOV capable devices much easier
to use, since the user does not have to hardcode the device, but can
give a list of identical ones, and qemu-server chooses dynamically.

note that we require the devices all to be the same vendor/device,
because we don't want to group unrelated devices, but we only check
the iommugroup for the first device, but there is a high chance
that this also changes when somethings off since e.g. SRIOV devices
are most often created at the same time, so when the any has a different
group, the first one will too

Signed-off-by: Dominik Csapak <d.csapak@proxmox.com>
---
 src/PVE/HardwareMap.pm | 71 ++++++++++++++++++++++++------------------
 1 file changed, 41 insertions(+), 30 deletions(-)

diff --git a/src/PVE/HardwareMap.pm b/src/PVE/HardwareMap.pm
index 31841b4..52c45e1 100644
--- a/src/PVE/HardwareMap.pm
+++ b/src/PVE/HardwareMap.pm
@@ -84,9 +84,11 @@ my $format = {
 	},
 	path => {
 	    description => "The path to the device. If the function is omitted, the whole device is"
-		." mapped. In that case use the attributes of the first device.",
+		." mapped. In that case use the attributes of the first device. You can give"
+		." multiple paths as a semicolon seperated list, the first available will then"
+		." be chosen on guest start.",
 	    type => 'string',
-	    pattern => "${PCI_RE}",
+	    pattern => "(?:${PCI_RE};)*${PCI_RE}",
 	},
 	mdev => {
 	    description => "The Device supports mediated devices.",
@@ -213,39 +215,48 @@ sub write_hardware_map {
 my $pci_valid = sub {
     my ($cfg) = @_;
 
-    my $multifunction = 0;
-    if ($path !~ m/\.[a-f0-9]/i) {
-	# whole device, add .0 (must exist)
-	$path = "$path.0";
-	$multifunction = 1;
-    }
+    my @paths = split(';', $cfg->{path} // '');
 
-    my $info = PVE::SysFSTools::pci_device_info($path, 1);
-    die "pci device '$path' not found\n" if !defined($info);
+    my $idx = 0;
+    for my $path (@paths) {
 
-    my $correct_props = {
-	vendor => $info->{vendor},
-	device => $info->{device},
-	'subsystem-vendor' => $info->{'subsystem_vendor'},
-	'subsystem-device' => $info->{'subsystem_device'},
-	mdev => $multifunction ? undef : $info->{mdev}, # don't allow mdev for multifunction
-	iommugroup => $info->{iommugroup},
-    };
+	my $multifunction = 0;
+	if ($path !~ m/\.[a-f0-9]/i) {
+	    # whole device, add .0 (must exist)
+	    $path = "$path.0";
+	    $multifunction = 1;
+	}
+
+	my $info = PVE::SysFSTools::pci_device_info($path, 1);
+	die "pci device '$path' not found\n" if !defined($info);
 
-    for my $prop (sort keys %$correct_props) {
-	next if !defined($correct_props->{$prop}) && !defined($cfg->{$prop});
-	die "no '$prop' for device '$path'\n"
-	    if defined($correct_props->{$prop}) && !defined($cfg->{$prop});
-	die "'$prop' configured but should not be\n"
-	    if !defined($correct_props->{$prop}) && defined($cfg->{$prop});
+	my $correct_props = {
+	    vendor => $info->{vendor},
+	    device => $info->{device},
+	    'subsystem-vendor' => $info->{'subsystem_vendor'},
+	    'subsystem-device' => $info->{'subsystem_device'},
+	    mdev => $multifunction ? undef : $info->{mdev}, # don't allow mdev for multifunction
+	    iommugroup => $info->{iommugroup},
+	};
+
+	for my $prop (sort keys %$correct_props) {
+	    next if $prop eq 'iommugroup' && $idx > 0; # check iommu only on the first device
 
-	my $correct_prop = $correct_props->{$prop};
-	$correct_prop =~ s/^0x//;
-	my $configured_prop = $cfg->{$prop};
-	$configured_prop =~ s/^0x//;
+	    next if !defined($correct_props->{$prop}) && !defined($cfg->{$prop});
+	    die "no '$prop' for device '$path'\n"
+		if defined($correct_props->{$prop}) && !defined($cfg->{$prop});
+	    die "'$prop' configured but should not be\n"
+		if !defined($correct_props->{$prop}) && defined($cfg->{$prop});
 
-	die "'$prop' does not match for '$cfg->{name}' ($correct_prop != $configured_prop)\n"
-	    if $correct_prop ne $configured_prop;
+	    my $correct_prop = $correct_props->{$prop};
+	    $correct_prop =~ s/^0x//;
+	    my $configured_prop = $cfg->{$prop};
+	    $configured_prop =~ s/^0x//;
+
+	    die "'$prop' does not match for '$cfg->{name}' ($correct_prop != $configured_prop)\n"
+		if $correct_prop ne $configured_prop;
+	}
+	$idx++;
     }
 
     return 1;
-- 
2.30.2





  parent reply	other threads:[~2022-09-20 12:51 UTC|newest]

Thread overview: 54+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2022-09-20 12:50 [pve-devel] [PATCH many v3] add cluster-wide hardware device mapping Dominik Csapak
2022-09-20 12:50 ` [pve-devel] [PATCH cluster v3 1/1] add nodes/hardware-map.conf Dominik Csapak
2022-11-08 18:03   ` [pve-devel] applied: " Thomas Lamprecht
2022-09-20 12:50 ` [pve-devel] [PATCH qemu-server v3 10/13] PVE/API2/Qemu: migrate preconditions: use new check_local_resources info Dominik Csapak
2022-09-20 12:50 ` [pve-devel] [PATCH qemu-server v3 11/13] PVE/QemuMigrate: check for mapped resources on migration Dominik Csapak
2022-09-20 12:50 ` [pve-devel] [PATCH qemu-server v3 12/13] fix #3574: enable multi pci device mapping from config Dominik Csapak
2022-09-20 12:50 ` [pve-devel] [PATCH qemu-server v3 13/13] add tests for mapped pci devices Dominik Csapak
2022-09-20 12:50 ` [pve-devel] [PATCH access-control v3 1/1] PVE/AccessControl: add Hardware.* privileges and /hardware/ paths Dominik Csapak
2022-11-09 12:05   ` Fabian Grünbichler
2022-11-09 12:39     ` Dominik Csapak
2022-11-09 13:06       ` Fabian Grünbichler
2022-11-09 13:23         ` Dominik Csapak
2022-11-09 12:52     ` Thomas Lamprecht
2022-09-20 12:50 ` [pve-devel] [PATCH common v3 1/3] SysFSTools: make mdev cleanup independent of pciid Dominik Csapak
2022-11-09  8:38   ` Thomas Lamprecht
2022-09-20 12:50 ` [pve-devel] [PATCH common v3 2/3] add PVE/HardwareMap Dominik Csapak
2022-11-09  8:46   ` Thomas Lamprecht
2022-09-20 12:50 ` Dominik Csapak [this message]
2022-09-20 12:50 ` [pve-devel] [PATCH qemu-server v3 01/13] cleanup pci devices in more situations Dominik Csapak
2022-11-09  8:00   ` [pve-devel] applied: " Thomas Lamprecht
2022-09-20 12:50 ` [pve-devel] [PATCH qemu-server v3 02/13] PCI: make mediated device path independent of pci id Dominik Csapak
2022-11-09  8:08   ` [pve-devel] applied: " Thomas Lamprecht
2022-09-20 12:50 ` [pve-devel] [PATCH qemu-server v3 03/13] PCI: refactor print_pci_device Dominik Csapak
2022-11-09  7:49   ` Thomas Lamprecht
2022-09-20 12:50 ` [pve-devel] [PATCH qemu-server v3 04/13] PCI: reuse parsed info from print_hostpci_devices Dominik Csapak
2022-11-09  8:23   ` Thomas Lamprecht
2022-09-20 12:50 ` [pve-devel] [PATCH qemu-server v3 05/13] PVE/QemuServer: allow mapped usb devices in config Dominik Csapak
2022-09-20 12:50 ` [pve-devel] [PATCH qemu-server v3 06/13] PVE/QemuServer: allow mapped pci deviced " Dominik Csapak
2022-09-20 12:50 ` [pve-devel] [PATCH qemu-server v3 07/13] PVE/API2/Qemu: add permission checks for mapped usb devices Dominik Csapak
2022-09-20 12:50 ` [pve-devel] [PATCH qemu-server v3 08/13] PVE/API2/Qemu: add permission checks for mapped pci devices Dominik Csapak
2022-11-09 12:14   ` Fabian Grünbichler
2022-11-09 12:51     ` Dominik Csapak
2022-11-09 13:28       ` Fabian Grünbichler
2022-09-20 12:50 ` [pve-devel] [PATCH qemu-server v3 09/13] PVE/QemuServer: extend 'check_local_resources' for mapped resources Dominik Csapak
2022-09-20 12:50 ` [pve-devel] [PATCH qemu-server v3 10/13] PVE/API2/Qemu: migrate preconditions: use new check_local_resources info Dominik Csapak
2022-09-20 12:50 ` [pve-devel] [PATCH qemu-server v3 11/13] PVE/QemuMigrate: check for mapped resources on migration Dominik Csapak
2022-09-20 12:50 ` [pve-devel] [PATCH qemu-server v3 12/13] fix #3574: enable multi pci device mapping from config Dominik Csapak
2022-09-20 12:50 ` [pve-devel] [PATCH qemu-server v3 13/13] add tests for mapped pci devices Dominik Csapak
2022-09-20 12:50 ` [pve-devel] [PATCH manager v3 01/13] PVE/API2/Hardware: add Mapping.pm Dominik Csapak
2022-09-20 12:50 ` [pve-devel] [PATCH manager v3 02/13] PVE/API2/Cluster: add Hardware mapping list api call Dominik Csapak
2022-09-20 12:50 ` [pve-devel] [PATCH manager v3 03/13] ui: form/USBSelector: make it more flexible with nodename Dominik Csapak
2022-09-20 12:50 ` [pve-devel] [PATCH manager v3 04/13] ui: form: add PCIMapSelector Dominik Csapak
2022-09-20 12:50 ` [pve-devel] [PATCH manager v3 05/13] ui: form: add USBMapSelector Dominik Csapak
2022-09-20 12:50 ` [pve-devel] [PATCH manager v3 06/13] ui: qemu/PCIEdit: rework panel to add a mapped configuration Dominik Csapak
2022-09-20 12:50 ` [pve-devel] [PATCH manager v3 07/13] ui: qemu/USBEdit: add 'mapped' device case Dominik Csapak
2022-09-20 12:50 ` [pve-devel] [PATCH manager v3 08/13] ui: form: add MultiPCISelector Dominik Csapak
2022-09-20 12:50 ` [pve-devel] [PATCH manager v3 09/13] ui: add window/PCIEdit: edit window for pci mappings Dominik Csapak
2022-09-20 12:50 ` [pve-devel] [PATCH manager v3 10/13] ui: add window/USBEdit: edit window for usb mappings Dominik Csapak
2022-09-20 12:50 ` [pve-devel] [PATCH manager v3 11/13] ui: add dc/HardwareView: a CRUD interface for hardware mapping Dominik Csapak
2022-09-20 12:50 ` [pve-devel] [PATCH manager v3 12/13] ui: window/Migrate: allow mapped devices Dominik Csapak
2022-09-20 12:50 ` [pve-devel] [PATCH manager v3 13/13] ui: improve permission handling for hardware Dominik Csapak
2022-09-20 16:12 ` [pve-devel] [PATCH many v3] add cluster-wide hardware device mapping DERUMIER, Alexandre
2022-09-23 16:13 ` DERUMIER, Alexandre
2022-11-08 18:03 ` Thomas Lamprecht

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20220920125041.3636561-10-d.csapak@proxmox.com \
    --to=d.csapak@proxmox.com \
    --cc=pve-devel@lists.proxmox.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox
Service provided by Proxmox Server Solutions GmbH | Privacy | Legal