From: Jason Wang <jasowang@redhat.com>
To: "Michael S. Tsirkin" <mst@redhat.com>
Cc: Alexander Graf <agraf@suse.de>,
qemu-devel@nongnu.org, qemu-ppc@nongnu.org,
cornelia.huck@de.ibm.com, Paolo Bonzini <pbonzini@redhat.com>,
Richard Henderson <rth@twiddle.net>
Subject: Re: [Qemu-devel] [PATCH V4 19/19] virtio-pci: introduce auto_msix_bar_size property
Date: Fri, 20 Mar 2015 13:35:55 +0800 [thread overview]
Message-ID: <1426829755.5879.1@smtp.corp.redhat.com> (raw)
In-Reply-To: <20150319105643-mutt-send-email-mst@redhat.com>
On Thu, Mar 19, 2015 at 6:01 PM, Michael S. Tsirkin <mst@redhat.com>
wrote:
> On Thu, Mar 19, 2015 at 01:23:12PM +0800, Jason Wang wrote:
>>
>>
>> On Wed, Mar 18, 2015 at 8:57 PM, Michael S. Tsirkin
>> <mst@redhat.com> wrote:
>> >On Wed, Mar 18, 2015 at 05:35:09PM +0800, Jason Wang wrote:
>> >> Currently we don't support more than 128 MSI-X vectors for a pci
>> >> devices, trying to use vector=129 for a virtio-net-pci device
>> may get:
>> >> qemu-system-x86_64: -device
>> virtio-net-pci,netdev=hn0,vectors=129:
>> >> unable to init msix vectors to 129
>> >> This this because the MSI-X bar size were hard-coded as 4096.
>> So this
>> >> patch introduces boolean auto_msix_bar_size property for
>> >> virito-pci devices. Enable this will let the device calculate
>> the msix
>> >> bar size based on the number of MSI-X entries instead of
>> previous 4096
>> >> hard-coded limit.
>> >> This is a must to let virtio-net can up to 256 queues and each
>> queue
>> >> were associated with a specific MSI-X entry.
>> >> Cc: Paolo Bonzini <pbonzini@redhat.com>
>> >> Cc: Richard Henderson <rth@twiddle.net>
>> >> Cc: Michael S. Tsirkin <mst@redhat.com>
>> >> Cc: Alexander Graf <agraf@suse.de>
>> >> Cc: qemu-ppc@nongnu.org
>> >> Signed-off-by: Jason Wang <jasowang@redhat.com>
>> >
>> >I don't understand what this property does.
>> >What if I *don't* set auto_msix_bar_size?
>> >vectors<=128 works exactly like it dif previously, vectors=129
>> fails?
>>
>> Yes, but there looks like a bug in the code which can lead a bar
>> size less
>> than 4K.
>>
>> >
>> >Does not seem like useful behaviour, to me.
>>
>> This property allows to have more than 128 vectors to be used. We
>> disable
>> this for legacy machine types to stick bar size to 4K to keep the
>> migration
>> compatibility.
>
> Compatibility between which two configurations?
I mean migration from 2.4 to 2.3 with e.g vectors=129.
>
> qemu 2.3 can not run when > 128 vectors are requested.
Unfortunately not, qemu can run with just a warning like:
qemu-system-x86_64: -device virtio-net-pci,vectors=129: unable to init
msix vectors to 129
vectors will reset to zero in this case.
>
> So there is no need to worry about what happens when you
> request > 128 vectors and an old machine type.
> qemu exiting is not guest visible behaviour.
>
> I can imagine two reasonable solutions:
> 1. increase bar size for everyone. make it 8k for compat machine types
> simple but wastes memory
> 2. make bar size depend on # of vectors
> as # of vectors is user-specified, there's no problem
> with compatibility, so no need to tweak machine types,
> but layout is dynamic so more complex.
This is what I want to do. And since vectors=129 won't crash for 2.3.
Still need to keep the 4k for compat machine types.
>
>
>
>>
>> >
>> >
>> >> ---
>> >> hw/i386/pc_piix.c | 8 ++++++++
>> >> hw/i386/pc_q35.c | 8 ++++++++
>> >> hw/ppc/spapr.c | 11 ++++++++++-
>> >> hw/virtio/virtio-pci.c | 17 +++++++++++++++--
>> >> hw/virtio/virtio-pci.h | 3 +++
>> >> include/hw/compat.h | 11 +++++++++++
>> >> 6 files changed, 55 insertions(+), 3 deletions(-)
>> >> diff --git a/hw/i386/pc_piix.c b/hw/i386/pc_piix.c
>> >> index 0796719..8808500 100644
>> >> --- a/hw/i386/pc_piix.c
>> >> +++ b/hw/i386/pc_piix.c
>> >> @@ -552,6 +552,10 @@ static QEMUMachine pc_i440fx_machine_v2_3 =
>> {
>> >> PC_I440FX_2_3_MACHINE_OPTIONS,
>> >> .name = "pc-i440fx-2.3",
>> >> .init = pc_init_pci_2_3,
>> >> + .compat_props = (GlobalProperty[]) {
>> >> + HW_COMPAT_2_3,
>> >> + { /* end of list */ }
>> >> + },
>> >> };
>> >> #define PC_I440FX_2_2_MACHINE_OPTIONS
>> PC_I440FX_2_3_MACHINE_OPTIONS
>> >> @@ -560,6 +564,10 @@ static QEMUMachine pc_i440fx_machine_v2_2 =
>> {
>> >> PC_I440FX_2_2_MACHINE_OPTIONS,
>> >> .name = "pc-i440fx-2.2",
>> >> .init = pc_init_pci_2_2,
>> >> + .compat_props = (GlobalProperty[]) {
>> >> + HW_COMPAT_2_2,
>> >> + { /* end of list */ }
>> >> + },
>> >> };
>> >> #define PC_I440FX_2_1_MACHINE_OPTIONS
>> \
>> >> diff --git a/hw/i386/pc_q35.c b/hw/i386/pc_q35.c
>> >> index a8a34a4..4a34349 100644
>> >> --- a/hw/i386/pc_q35.c
>> >> +++ b/hw/i386/pc_q35.c
>> >> @@ -448,6 +448,10 @@ static QEMUMachine pc_q35_machine_v2_3 = {
>> >> PC_Q35_2_3_MACHINE_OPTIONS,
>> >> .name = "pc-q35-2.3",
>> >> .init = pc_q35_init_2_3,
>> >> + .compat_props = (GlobalProperty[]) {
>> >> + HW_COMPAT_2_3,
>> >> + { /* end of list */ }
>> >> + },
>> >> };
>> >> #define PC_Q35_2_2_MACHINE_OPTIONS PC_Q35_2_3_MACHINE_OPTIONS
>> >> @@ -456,6 +460,10 @@ static QEMUMachine pc_q35_machine_v2_2 = {
>> >> PC_Q35_2_2_MACHINE_OPTIONS,
>> >> .name = "pc-q35-2.2",
>> >> .init = pc_q35_init_2_2,
>> >> + .compat_props = (GlobalProperty[]) {
>> >> + HW_COMPAT_2_2,
>> >> + { /* end of list */ }
>> >> + },
>> >> };
>> >> #define PC_Q35_2_1_MACHINE_OPTIONS \
>> >> diff --git a/hw/ppc/spapr.c b/hw/ppc/spapr.c
>> >> index 5f25dd3..853a5cc 100644
>> >> --- a/hw/ppc/spapr.c
>> >> +++ b/hw/ppc/spapr.c
>> >> @@ -1794,12 +1794,16 @@ static const TypeInfo spapr_machine_info
>> = {
>> >> },
>> >> };
>> >> +#define SPAPR_COMPAT_2_3 \
>> >> + HW_COMPAT_2_3
>> >> +
>> >> #define SPAPR_COMPAT_2_2 \
>> >> + SPAPR_COMPAT_2_3, \
>> >> {\
>> >> .driver = TYPE_SPAPR_PCI_HOST_BRIDGE,\
>> >> .property = "mem_win_size",\
>> >> .value = "0x20000000",\
>> >> - }
>> >> + } \
>> >> #define SPAPR_COMPAT_2_1 \
>> >> SPAPR_COMPAT_2_2
>> >> @@ -1883,10 +1887,15 @@ static const TypeInfo
>> spapr_machine_2_2_info =
>> >>{
>> >> static void spapr_machine_2_3_class_init(ObjectClass *oc,
>> void
>> >>*data)
>> >> {
>> >> + static GlobalProperty compat_props[] = {
>> >> + SPAPR_COMPAT_2_3,
>> >> + { /* end of list */ }
>> >> + };
>> >> MachineClass *mc = MACHINE_CLASS(oc);
>> >> mc->name = "pseries-2.3";
>> >> mc->desc = "pSeries Logical Partition (PAPR compliant)
>> v2.3";
>> >> + mc->compat_props = compat_props;
>> >> }
>> >> static const TypeInfo spapr_machine_2_3_info = {
>> >> diff --git a/hw/virtio/virtio-pci.c b/hw/virtio/virtio-pci.c
>> >> index 4a5febb..f4cd405 100644
>> >> --- a/hw/virtio/virtio-pci.c
>> >> +++ b/hw/virtio/virtio-pci.c
>> >> @@ -925,7 +925,7 @@ static void
>> virtio_pci_device_plugged(DeviceState
>> >>*d)
>> >> VirtIOPCIProxy *proxy = VIRTIO_PCI(d);
>> >> VirtioBusState *bus = &proxy->bus;
>> >> uint8_t *config;
>> >> - uint32_t size;
>> >> + uint32_t size, bar_size;
>> >> config = proxy->pci_dev.config;
>> >> if (proxy->class_code) {
>> >> @@ -936,8 +936,19 @@ static void
>> virtio_pci_device_plugged(DeviceState
>> >>*d)
>> >> pci_set_word(config + PCI_SUBSYSTEM_ID,
>> >>virtio_bus_get_vdev_id(bus));
>> >> config[PCI_INTERRUPT_PIN] = 1;
>> >> + if (proxy->flags & VIRTIO_PCI_FLAG_AUTO_MSIX_SIZE) {
>> >> + bar_size = proxy->nvectors * PCI_MSIX_ENTRY_SIZE * 2;
>> >> + if (bar_size & (bar_size - 1)) {
>> >> + bar_size = 1 << qemu_fls(bar_size);
>> >> + }
>> >> + } else {
>> >> + /* For migration compatibility */
>> >> + bar_size = 4096;
>> >> + }
>> >> +
>> >> if (proxy->nvectors &&
>> >> - msix_init_exclusive_bar(&proxy->pci_dev,
>> proxy->nvectors, 1,
>> >>4096)) {
>> >> + msix_init_exclusive_bar(&proxy->pci_dev,
>> proxy->nvectors, 1,
>> >> + bar_size)) {
>> >> error_report("unable to init msix vectors to %" PRIu32,
>> >> proxy->nvectors);
>> >> proxy->nvectors = 0;
>> >
>> >
>> >As I expected, msix format stuff spreads out to virtio.
>> >Consider "vectors * PCI_MSIX_ENTRY_SIZE * 2"
>> >That's because you use half the BAR for BIR in msix.c
>> >So any change will have to be done in two places,
>> >that's bad.
>> >
>> >
>> >> @@ -1370,6 +1381,8 @@ static const TypeInfo
>> virtio_serial_pci_info = {
>> >> static Property virtio_net_properties[] = {
>> >> DEFINE_PROP_BIT("ioeventfd", VirtIOPCIProxy, flags,
>> >> VIRTIO_PCI_FLAG_USE_IOEVENTFD_BIT, false),
>> >> + DEFINE_PROP_BIT("auto_msix_bar_size", VirtIOPCIProxy, flags,
>> >> + VIRTIO_PCI_FLAG_AUTO_MSIX_SIZE_BIT, true),
>> >> DEFINE_PROP_UINT32("vectors", VirtIOPCIProxy, nvectors, 3),
>> >> DEFINE_VIRTIO_NET_FEATURES(VirtIOPCIProxy, host_features),
>> >> DEFINE_PROP_END_OF_LIST(),
>> >> diff --git a/hw/virtio/virtio-pci.h b/hw/virtio/virtio-pci.h
>> >> index 3bac016..82a6782 100644
>> >> --- a/hw/virtio/virtio-pci.h
>> >> +++ b/hw/virtio/virtio-pci.h
>> >> @@ -62,6 +62,9 @@ typedef struct VirtioBusClass
>> VirtioPCIBusClass;
>> >> * vcpu thread using ioeventfd for some devices. */
>> >> #define VIRTIO_PCI_FLAG_USE_IOEVENTFD_BIT 1
>> >> #define VIRTIO_PCI_FLAG_USE_IOEVENTFD (1 <<
>> >>VIRTIO_PCI_FLAG_USE_IOEVENTFD_BIT)
>> >> +#define VIRTIO_PCI_FLAG_AUTO_MSIX_SIZE_BIT 2
>> >> +#define VIRTIO_PCI_FLAG_AUTO_MSIX_SIZE \
>> >> + (1 << VIRTIO_PCI_FLAG_AUTO_MSIX_SIZE_BIT)
>> >> typedef struct {
>> >> MSIMessage msg;
>> >> diff --git a/include/hw/compat.h b/include/hw/compat.h
>> >> index 313682a..3186275 100644
>> >> --- a/include/hw/compat.h
>> >> +++ b/include/hw/compat.h
>> >> @@ -1,7 +1,18 @@
>> >> #ifndef HW_COMPAT_H
>> >> #define HW_COMPAT_H
>> >> +#define HW_COMPAT_2_3 \
>> >> + {\
>> >> + .driver = "virtio-net-pci",\
>> >> + .property = "auto_msix_bar_size",\
>> >> + .value = "off",\
>> >> + }
>> >> +
>> >> +#define HW_COMPAT_2_2 \
>> >> + HW_COMPAT_2_3
>> >> +
>> >> #define HW_COMPAT_2_1 \
>> >> + HW_COMPAT_2_2, \
>> >> {\
>> >> .driver = "intel-hda",\
>> >> .property = "old_msi_addr",\
>> >> -- 2.1.0
next prev parent reply other threads:[~2015-03-20 5:36 UTC|newest]
Thread overview: 52+ messages / expand[flat|nested] mbox.gz Atom feed top
2015-03-18 9:34 [Qemu-devel] [PATCH V4 00/19] Support more virtio queues Jason Wang
2015-03-18 9:34 ` [Qemu-devel] [PATCH V4 01/19] pc: add 2.4 machine types Jason Wang
2015-03-18 9:34 ` [Qemu-devel] [PATCH V4 02/19] spapr: add machine type specific instance init function Jason Wang
2015-03-18 9:34 ` [Qemu-devel] [PATCH V4 03/19] ppc: spapr: add 2.4 machine type Jason Wang
2015-03-18 9:34 ` [Qemu-devel] [PATCH V4 04/19] monitor: replace the magic number 255 with MAX_QUEUE_NUM Jason Wang
2015-03-18 9:34 ` [Qemu-devel] [PATCH V4 05/19] monitor: check return value of qemu_find_net_clients_except() Jason Wang
2015-03-18 9:34 ` [Qemu-devel] [PATCH V4 06/19] virtio-ccw: using VIRTIO_NO_VECTOR instead of 0 for invalid virtqueue Jason Wang
2015-03-18 13:08 ` Michael S. Tsirkin
2015-03-20 7:39 ` Cornelia Huck
2015-03-21 18:27 ` Michael S. Tsirkin
2015-03-23 9:02 ` Cornelia Huck
2015-03-18 9:34 ` [Qemu-devel] [PATCH V4 07/19] virtio-net: validate backend queue numbers against bus limitation Jason Wang
2015-03-18 13:05 ` Michael S. Tsirkin
2015-03-19 5:26 ` Jason Wang
2015-03-18 9:34 ` [Qemu-devel] [PATCH V4 08/19] virtio-net: fix the upper bound when trying to delete queues Jason Wang
2015-03-18 13:06 ` Michael S. Tsirkin
2015-03-19 5:28 ` Jason Wang
2015-03-18 9:34 ` [Qemu-devel] [PATCH V4 09/19] virito: introduce bus specific queue limit Jason Wang
2015-03-20 10:20 ` Cornelia Huck
2015-03-31 2:34 ` Jason Wang
2015-03-18 9:35 ` [Qemu-devel] [PATCH V4 10/19] virtio-ccw: introduce ccw " Jason Wang
2015-03-20 11:33 ` Cornelia Huck
2015-03-31 2:36 ` Jason Wang
2015-03-18 9:35 ` [Qemu-devel] [PATCH V4 11/19] virtio-s390: switch to bus " Jason Wang
2015-03-20 11:34 ` Cornelia Huck
2015-03-18 9:35 ` [Qemu-devel] [PATCH V4 12/19] virtio-mmio: " Jason Wang
2015-03-18 9:35 ` [Qemu-devel] [PATCH V4 13/19] virtio-pci: switch to use " Jason Wang
2015-03-18 9:35 ` [Qemu-devel] [PATCH V4 14/19] virtio: introduce vector to virtqueues mapping Jason Wang
2015-03-20 11:39 ` Cornelia Huck
2015-03-31 2:37 ` Jason Wang
2015-03-18 9:35 ` [Qemu-devel] [PATCH V4 15/19] virtio: introduce virtio_queue_get_index() Jason Wang
2015-03-18 9:35 ` [Qemu-devel] [PATCH V4 16/19] virtio-pci: speedup MSI-X masking and unmasking Jason Wang
2015-03-18 9:35 ` [Qemu-devel] [PATCH V4 17/19] virtio-pci: increase the maximum number of virtqueues to 513 Jason Wang
2015-03-18 9:35 ` [Qemu-devel] [PATCH V4 18/19] pci: remove hard-coded bar size in msix_init_exclusive_bar() Jason Wang
2015-03-18 12:52 ` Michael S. Tsirkin
2015-03-19 5:19 ` Jason Wang
2015-03-19 10:09 ` Michael S. Tsirkin
2015-03-20 5:43 ` Jason Wang
2015-03-18 9:35 ` [Qemu-devel] [PATCH V4 19/19] virtio-pci: introduce auto_msix_bar_size property Jason Wang
2015-03-18 12:57 ` Michael S. Tsirkin
2015-03-19 5:23 ` Jason Wang
2015-03-19 10:01 ` Michael S. Tsirkin
2015-03-20 5:35 ` Jason Wang [this message]
2015-03-19 5:23 ` Jason Wang
2015-03-19 10:02 ` Michael S. Tsirkin
2015-03-20 5:38 ` Jason Wang
2015-03-18 12:58 ` [Qemu-devel] [PATCH V4 00/19] Support more virtio queues Michael S. Tsirkin
2015-03-19 5:24 ` Jason Wang
2015-03-19 7:32 ` Michael S. Tsirkin
2015-03-19 7:42 ` Jason Wang
2015-03-19 9:23 ` Michael S. Tsirkin
2015-03-20 5:11 ` Jason Wang
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=1426829755.5879.1@smtp.corp.redhat.com \
--to=jasowang@redhat.com \
--cc=agraf@suse.de \
--cc=cornelia.huck@de.ibm.com \
--cc=mst@redhat.com \
--cc=pbonzini@redhat.com \
--cc=qemu-devel@nongnu.org \
--cc=qemu-ppc@nongnu.org \
--cc=rth@twiddle.net \
/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).