From: Marcel Apfelbaum <marcel@redhat.com>
To: "Michael S. Tsirkin" <mst@redhat.com>
Cc: qemu-devel@nongnu.org
Subject: Re: [Qemu-devel] [PATCH V2] hw/virtio-pci: fix virtio behaviour on modern (PCIe) machines
Date: Wed, 20 Jul 2016 11:27:19 +0300 [thread overview]
Message-ID: <578F35E7.6020005@redhat.com> (raw)
In-Reply-To: <20160720013838-mutt-send-email-mst@kernel.org>
On 07/20/2016 01:56 AM, Michael S. Tsirkin wrote:
> On Tue, Jul 19, 2016 at 09:42:58PM +0300, Marcel Apfelbaum wrote:
>> Modern machines are expected to be used by newer setups with
>> modern guests aiming the use of the latest features.
>>
>> Enable modern and disable legacy for virtio devices
>> plugged into PCIe ports (Root ports or Downstream ports).
>> Using the Virtio 1 mode will remove the limitation
>> of the number of devices that can be attached to a machine
>> by removing the need for the IO BAR.
>>
>> Convert 'disable-modern' and 'disable-legacy' properties to OnOffAuto
>> with default Auto.
>>
>> Signed-off-by: Marcel Apfelbaum <marcel@redhat.com>
>> ---
>>
>> Hi,
>>
>> v1 -> v2:
>> - Stick to existing defaults for old machine types (Michael S. Tsirkin)
>>
>> If everyone agrees, I am thinking about getting it into 2.7
>> to avoid the ~15 virtio devices limitation per machine.
>>
>> Notes:
>> - The non PCIe machines behaviour should remain the same.
>> - I hope is OK to make the disable-* properties OnOffAuto. Previous setups
>> using them can be affected,
>
Hi Michael,
> why? could you explain pls? isn't onoffauto compatible with bit
> properties?
>
Yes, indeed. the value of 'Off' is replaced by 'Auto', but the semantics
for the names "on/off" remain.
>
>> but libvirt is not using them yet (as far as I know)
>> - My tests were limited to checking all possible disable-* configurations (and make check for all archs)
>>
>> Thanks,
>> Marcel
>>
>> hw/virtio/virtio-pci.c | 31 ++++++++++++++++++++++++-------
>> hw/virtio/virtio-pci.h | 2 ++
>> include/hw/compat.h | 8 ++++++++
>> 3 files changed, 34 insertions(+), 7 deletions(-)
>>
>> diff --git a/hw/virtio/virtio-pci.c b/hw/virtio/virtio-pci.c
>> index 2b34b43..ec9e84f 100644
>> --- a/hw/virtio/virtio-pci.c
>> +++ b/hw/virtio/virtio-pci.c
>> @@ -1716,6 +1716,8 @@ static void virtio_pci_realize(PCIDevice *pci_dev, Error **errp)
>> {
>> VirtIOPCIProxy *proxy = VIRTIO_PCI(pci_dev);
>> VirtioPCIClass *k = VIRTIO_PCI_GET_CLASS(pci_dev);
>> + bool pcie_port = (pci_bus_is_express(pci_dev->bus) &&
>> + !pci_bus_is_root(pci_dev->bus));
>
>
> drop the extra outside () please.
>
OK
>>
>> /*
>> * virtio pci bar layout used by default.
>> @@ -1766,8 +1768,23 @@ static void virtio_pci_realize(PCIDevice *pci_dev, Error **errp)
>>
>> address_space_init(&proxy->modern_as, &proxy->modern_cfg, "virtio-pci-cfg-as");
>>
>> - if (pci_is_express(pci_dev) && pci_bus_is_express(pci_dev->bus) &&
>> - !pci_bus_is_root(pci_dev->bus)) {
>> + if ((pcie_port && (proxy->disable_modern == ON_OFF_AUTO_AUTO))
>> + || (proxy->disable_modern == ON_OFF_AUTO_OFF)) {
>
> drop the () around == - logic and math mix naturally in C.
>
OK
> also, pls put || at end of line, not at the beginning of
> continuation. this way you can see there is
> continuation as you read the code naturally.
>
The line length would be more than 80, so I needed to chose a way to "lose".
I'll add || at the end of the line, sure.
>
>
>> + proxy->flags &= ~VIRTIO_PCI_FLAG_DISABLE_MODERN;
>> + pci_dev->cap_present |= QEMU_PCI_CAP_EXPRESS;
>> + } else {
>> + proxy->flags |= VIRTIO_PCI_FLAG_DISABLE_MODERN;
>> + pci_dev->cap_present &= ~QEMU_PCI_CAP_EXPRESS;
>> + }
>> +
>> + if ((pcie_port && (proxy->disable_legacy == ON_OFF_AUTO_AUTO))
>> + || (proxy->disable_legacy == ON_OFF_AUTO_ON)) {
>> + proxy->flags |= VIRTIO_PCI_FLAG_DISABLE_LEGACY;
>> + } else {
>> + proxy->flags &= ~VIRTIO_PCI_FLAG_DISABLE_LEGACY;
>> + }
>> +
>
> So this is a bit messy.
>
> Can we do:
>
> if (pcie_port)
> default_disable_legacy = false;
> else
> default_disable_legacy = true;
>
> now
>
> if (proxy->disable_legacy == ON_OFF_AUTO_ON ||
> (proxy->disable_legacy == ON_OFF_AUTO_AUTO && default_disable_legacy))
>
> proxy->flags |= VIRTIO_PCI_FLAG_DISABLE_LEGACY;
> else
> proxy->flags &= ~VIRTIO_PCI_FLAG_DISABLE_LEGACY;
>
> ?
>
We have 2 *different* logic:
-disable-modern: make it *false* if plugged into a pcie_port or the user requested it.
In this case update QEMU_PCI_CAP_EXPRESS flag accordingly.
-disble-legacy: make it *true* if plugged into a pcie_port or the user requested it.
If the code seems complicated I'll try to make it more readable.
Maybe I'll add a switch for each flag.
>
> Also, I wonder how does this interact with devices that play with
> these flags themselves, like virtio gpu?
> I guess we could just set VIRTIO_PCI_FLAG_DISABLE_LEGACY, avoid clearing it.
>
I thought this method is called before any other init call sites. (like virtio gpu or serial)
I'll double-check and update accordingly.
> Setting disable modern might be problematic for same reason, not sure what
> to do about that one.
>
> Did I miss anything?
I'll check this code runs before the other specific devices init code.
If it runs before, we have no problem here, otherwise I'll need to re-think
this patch.
Thank you for the review and I'll provide the answers in the next version.
Marcel
>
>> + if (pcie_port && pci_is_express(pci_dev)) {
>> int pos;
>>
>> pos = pcie_endpoint_cap_init(pci_dev, 0);
>> @@ -1821,10 +1838,10 @@ static void virtio_pci_reset(DeviceState *qdev)
>> static Property virtio_pci_properties[] = {
>> DEFINE_PROP_BIT("virtio-pci-bus-master-bug-migration", VirtIOPCIProxy, flags,
>> VIRTIO_PCI_FLAG_BUS_MASTER_BUG_MIGRATION_BIT, false),
>> - DEFINE_PROP_BIT("disable-legacy", VirtIOPCIProxy, flags,
>> - VIRTIO_PCI_FLAG_DISABLE_LEGACY_BIT, false),
>> - DEFINE_PROP_BIT("disable-modern", VirtIOPCIProxy, flags,
>> - VIRTIO_PCI_FLAG_DISABLE_MODERN_BIT, true),
>> + DEFINE_PROP_ON_OFF_AUTO("disable-legacy", VirtIOPCIProxy, disable_legacy,
>> + ON_OFF_AUTO_AUTO),
>> + DEFINE_PROP_ON_OFF_AUTO("disable-modern", VirtIOPCIProxy, disable_modern,
>> + ON_OFF_AUTO_AUTO),
>> DEFINE_PROP_BIT("migrate-extra", VirtIOPCIProxy, flags,
>> VIRTIO_PCI_FLAG_MIGRATE_EXTRA_BIT, true),
>> DEFINE_PROP_BIT("modern-pio-notify", VirtIOPCIProxy, flags,
>> @@ -1841,7 +1858,7 @@ static void virtio_pci_dc_realize(DeviceState *qdev, Error **errp)
>> PCIDevice *pci_dev = &proxy->pci_dev;
>>
>> if (!(proxy->flags & VIRTIO_PCI_FLAG_DISABLE_PCIE) &&
>> - !(proxy->flags & VIRTIO_PCI_FLAG_DISABLE_MODERN)) {
>> + !(proxy->disable_modern == ON_OFF_AUTO_ON)) {
>> pci_dev->cap_present |= QEMU_PCI_CAP_EXPRESS;
>> }
>>
>> diff --git a/hw/virtio/virtio-pci.h b/hw/virtio/virtio-pci.h
>> index e4548c2..4f219d4 100644
>> --- a/hw/virtio/virtio-pci.h
>> +++ b/hw/virtio/virtio-pci.h
>> @@ -144,6 +144,8 @@ struct VirtIOPCIProxy {
>> uint32_t modern_mem_bar;
>> int config_cap;
>> uint32_t flags;
>> + OnOffAuto disable_modern;
>> + OnOffAuto disable_legacy;
>> uint32_t class_code;
>> uint32_t nvectors;
>> uint32_t dfselect;
>> diff --git a/include/hw/compat.h b/include/hw/compat.h
>> index 9914e7a..1531399 100644
>> --- a/include/hw/compat.h
>> +++ b/include/hw/compat.h
>> @@ -6,6 +6,14 @@
>> .driver = "virtio-mmio",\
>> .property = "format_transport_address",\
>> .value = "off",\
>> + },{\
>> + .driver = "virtio-pci",\
>> + .property = "disable-modern",\
>> + .value = "on",\
>> + },{\
>> + .driver = "virtio-pci",\
>> + .property = "disable-legacy",\
>> + .value = "off",\
>> },
>>
>> #define HW_COMPAT_2_5 \
>> --
>> 2.4.3
next prev parent reply other threads:[~2016-07-20 8:27 UTC|newest]
Thread overview: 9+ messages / expand[flat|nested] mbox.gz Atom feed top
2016-07-19 18:42 [Qemu-devel] [PATCH V2] hw/virtio-pci: fix virtio behaviour on modern (PCIe) machines Marcel Apfelbaum
2016-07-19 22:56 ` Michael S. Tsirkin
2016-07-20 8:27 ` Marcel Apfelbaum [this message]
2016-07-20 8:41 ` Gerd Hoffmann
2016-07-20 9:01 ` Marcel Apfelbaum
2016-07-20 9:43 ` Gerd Hoffmann
2016-07-20 9:16 ` Cornelia Huck
2016-07-20 9:37 ` Marcel Apfelbaum
2016-07-20 11:23 ` Cornelia Huck
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=578F35E7.6020005@redhat.com \
--to=marcel@redhat.com \
--cc=mst@redhat.com \
--cc=qemu-devel@nongnu.org \
/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;
as well as URLs for NNTP newsgroup(s).