From: Paolo Bonzini <pbonzini@redhat.com>
To: Alexey Kardashevskiy <aik@ozlabs.ru>, qemu-devel@nongnu.org
Cc: "Michael S. Tsirkin" <mst@redhat.com>
Subject: Re: [Qemu-devel] [RFC PATCH qemu] virtio-pci: Replace modern_as with direct access to modern_bar
Date: Fri, 22 Sep 2017 17:45:28 +0200 [thread overview]
Message-ID: <df2f7ce1-a868-4a99-52ec-d4f934e1b7b7@redhat.com> (raw)
In-Reply-To: <20170922151208.24590-1-aik@ozlabs.ru>
On 22/09/2017 17:12, Alexey Kardashevskiy wrote:
> The modern bar is accessed now via yet another address space created just
> for that purpose and it does not really need FlatView and dispatch tree
> as it has a single memory region so it is just a waste of memory. Things
> get even worse when there are dozens or hundreds of virtio-pci devices -
> since these address spaces are global, changing any of them triggers
> rebuilding all address spaces.
>
> This replaces indirect accesses to the modern BAR with a simple lookup
> and direct calls to memory_region_dispatch_read/write.
>
> This is expected to save lots of memory at boot time after applying:
> [Qemu-devel] [PULL 00/32] Misc changes for 2017-09-22
>
> Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru>
> ---
There used to be this patch:
https://marc.info/?l=linux-virtualization&m=142363659531278&q=raw
> What is the easiest way to test it? Git history suggests that only
> SeaBIOS supports this, is this still correct?
>
> What is best to do if MR is not found (and is it possible in practice)?
> assert or redirect to unassigned memory?
Do nothing at all, I think.
I haven't thought much about the endian switches, but it looks good
apart from that.
Paolo
> ---
> hw/virtio/virtio-pci.h | 17 +++++++-----
> hw/virtio/virtio-pci.c | 71 ++++++++++++++++++++++++++++----------------------
> 2 files changed, 50 insertions(+), 38 deletions(-)
>
> diff --git a/hw/virtio/virtio-pci.h b/hw/virtio/virtio-pci.h
> index 69f5959623..12d3a90686 100644
> --- a/hw/virtio/virtio-pci.h
> +++ b/hw/virtio/virtio-pci.h
> @@ -155,15 +155,18 @@ typedef struct VirtIOPCIQueue {
> struct VirtIOPCIProxy {
> PCIDevice pci_dev;
> MemoryRegion bar;
> - VirtIOPCIRegion common;
> - VirtIOPCIRegion isr;
> - VirtIOPCIRegion device;
> - VirtIOPCIRegion notify;
> - VirtIOPCIRegion notify_pio;
> + union {
> + struct {
> + VirtIOPCIRegion common;
> + VirtIOPCIRegion isr;
> + VirtIOPCIRegion device;
> + VirtIOPCIRegion notify;
> + VirtIOPCIRegion notify_pio;
> + };
> + VirtIOPCIRegion regs[5];
> + };
> MemoryRegion modern_bar;
> MemoryRegion io_bar;
> - MemoryRegion modern_cfg;
> - AddressSpace modern_as;
> uint32_t legacy_io_bar_idx;
> uint32_t msix_bar_idx;
> uint32_t modern_io_bar_idx;
> diff --git a/hw/virtio/virtio-pci.c b/hw/virtio/virtio-pci.c
> index 8b0d6b69cd..0fcb8589f7 100644
> --- a/hw/virtio/virtio-pci.c
> +++ b/hw/virtio/virtio-pci.c
> @@ -545,6 +545,24 @@ static const MemoryRegionOps virtio_pci_config_ops = {
> .endianness = DEVICE_LITTLE_ENDIAN,
> };
>
> +static MemoryRegion *virtio_address_space_lookup(VirtIOPCIProxy *proxy,
> + hwaddr *off, int len)
> +{
> + int i;
> + VirtIOPCIRegion *reg;
> +
> + for (i = 0; i < ARRAY_SIZE(proxy->regs); ++i) {
> + reg = &proxy->regs[i];
> + if (*off >= reg->offset &&
> + *off + len <= reg->offset + reg->size) {
> + *off -= reg->offset;
> + return ®->mr;
> + }
> + }
> +
> + return NULL;
> +}
> +
> /* Below are generic functions to do memcpy from/to an address space,
> * without byteswaps, with input validation.
> *
> @@ -558,63 +576,68 @@ static const MemoryRegionOps virtio_pci_config_ops = {
> * Note: host pointer must be aligned.
> */
> static
> -void virtio_address_space_write(AddressSpace *as, hwaddr addr,
> +void virtio_address_space_write(VirtIOPCIProxy *proxy, hwaddr addr,
> const uint8_t *buf, int len)
> {
> - uint32_t val;
> + uint64_t val;
> + MemoryRegion *mr;
>
> /* address_space_* APIs assume an aligned address.
> * As address is under guest control, handle illegal values.
> */
> addr &= ~(len - 1);
>
> + mr = virtio_address_space_lookup(proxy, &addr, len);
> + assert(mr);
> +
> /* Make sure caller aligned buf properly */
> assert(!(((uintptr_t)buf) & (len - 1)));
>
> switch (len) {
> case 1:
> val = pci_get_byte(buf);
> - address_space_stb(as, addr, val, MEMTXATTRS_UNSPECIFIED, NULL);
> break;
> case 2:
> - val = pci_get_word(buf);
> - address_space_stw_le(as, addr, val, MEMTXATTRS_UNSPECIFIED, NULL);
> + val = cpu_to_le16(pci_get_word(buf));
> break;
> case 4:
> - val = pci_get_long(buf);
> - address_space_stl_le(as, addr, val, MEMTXATTRS_UNSPECIFIED, NULL);
> + val = cpu_to_le32(pci_get_long(buf));
> break;
> default:
> /* As length is under guest control, handle illegal values. */
> - break;
> + return;
> }
> + memory_region_dispatch_write(mr, addr, val, len, MEMTXATTRS_UNSPECIFIED);
> }
>
> static void
> -virtio_address_space_read(AddressSpace *as, hwaddr addr, uint8_t *buf, int len)
> +virtio_address_space_read(VirtIOPCIProxy *proxy, hwaddr addr,
> + uint8_t *buf, int len)
> {
> - uint32_t val;
> + uint64_t val;
> + MemoryRegion *mr;
>
> /* address_space_* APIs assume an aligned address.
> * As address is under guest control, handle illegal values.
> */
> addr &= ~(len - 1);
>
> + mr = virtio_address_space_lookup(proxy, &addr, len);
> + assert(mr);
> +
> /* Make sure caller aligned buf properly */
> assert(!(((uintptr_t)buf) & (len - 1)));
>
> + memory_region_dispatch_read(mr, addr, &val, len, MEMTXATTRS_UNSPECIFIED);
> switch (len) {
> case 1:
> - val = address_space_ldub(as, addr, MEMTXATTRS_UNSPECIFIED, NULL);
> pci_set_byte(buf, val);
> break;
> case 2:
> - val = address_space_lduw_le(as, addr, MEMTXATTRS_UNSPECIFIED, NULL);
> - pci_set_word(buf, val);
> + pci_set_word(buf, le16_to_cpu(val));
> break;
> case 4:
> - val = address_space_ldl_le(as, addr, MEMTXATTRS_UNSPECIFIED, NULL);
> - pci_set_long(buf, val);
> + pci_set_long(buf, le32_to_cpu(val));
> break;
> default:
> /* As length is under guest control, handle illegal values. */
> @@ -650,8 +673,7 @@ static void virtio_write_config(PCIDevice *pci_dev, uint32_t address,
>
> if (len == 1 || len == 2 || len == 4) {
> assert(len <= sizeof cfg->pci_cfg_data);
> - virtio_address_space_write(&proxy->modern_as, off,
> - cfg->pci_cfg_data, len);
> + virtio_address_space_write(proxy, off, cfg->pci_cfg_data, len);
> }
> }
> }
> @@ -675,8 +697,7 @@ static uint32_t virtio_read_config(PCIDevice *pci_dev,
>
> if (len == 1 || len == 2 || len == 4) {
> assert(len <= sizeof cfg->pci_cfg_data);
> - virtio_address_space_read(&proxy->modern_as, off,
> - cfg->pci_cfg_data, len);
> + virtio_address_space_read(proxy, off, cfg->pci_cfg_data, len);
> }
> }
>
> @@ -1783,15 +1804,6 @@ static void virtio_pci_realize(PCIDevice *pci_dev, Error **errp)
> /* PCI BAR regions must be powers of 2 */
> pow2ceil(proxy->notify.offset + proxy->notify.size));
>
> - memory_region_init_alias(&proxy->modern_cfg,
> - OBJECT(proxy),
> - "virtio-pci-cfg",
> - &proxy->modern_bar,
> - 0,
> - memory_region_size(&proxy->modern_bar));
> -
> - address_space_init(&proxy->modern_as, &proxy->modern_cfg, "virtio-pci-cfg-as");
> -
> if (proxy->disable_legacy == ON_OFF_AUTO_AUTO) {
> proxy->disable_legacy = pcie_port ? ON_OFF_AUTO_ON : ON_OFF_AUTO_OFF;
> }
> @@ -1860,10 +1872,7 @@ static void virtio_pci_realize(PCIDevice *pci_dev, Error **errp)
>
> static void virtio_pci_exit(PCIDevice *pci_dev)
> {
> - VirtIOPCIProxy *proxy = VIRTIO_PCI(pci_dev);
> -
> msix_uninit_exclusive_bar(pci_dev);
> - address_space_destroy(&proxy->modern_as);
> }
>
> static void virtio_pci_reset(DeviceState *qdev)
>
next prev parent reply other threads:[~2017-09-22 15:45 UTC|newest]
Thread overview: 5+ messages / expand[flat|nested] mbox.gz Atom feed top
2017-09-22 15:12 [Qemu-devel] [RFC PATCH qemu] virtio-pci: Replace modern_as with direct access to modern_bar Alexey Kardashevskiy
2017-09-22 15:45 ` Paolo Bonzini [this message]
2017-10-03 2:11 ` Alexey Kardashevskiy
2017-10-06 12:58 ` Paolo Bonzini
2017-10-09 3:20 ` Alexey Kardashevskiy
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=df2f7ce1-a868-4a99-52ec-d4f934e1b7b7@redhat.com \
--to=pbonzini@redhat.com \
--cc=aik@ozlabs.ru \
--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).