From: "Michael S. Tsirkin" <mst@redhat.com>
To: qemu-devel@nongnu.org
Cc: "Peter Maydell" <peter.maydell@linaro.org>,
"Ilya Maximets" <i.maximets@ovn.org>,
"Stefan Hajnoczi" <stefanha@redhat.com>,
"Jason Wang" <jasowang@redhat.com>,
"Paolo Bonzini" <pbonzini@redhat.com>,
"Peter Xu" <peterx@redhat.com>,
"David Hildenbrand" <david@redhat.com>,
"Philippe Mathieu-Daudé" <philmd@linaro.org>
Subject: [PULL v2 14/53] virtio: don't zero out memory region cache for indirect descriptors
Date: Wed, 4 Oct 2023 23:43:16 -0400 [thread overview]
Message-ID: <43d6376980d5567f2a6d00cfb30d10c0961671e6.1696477105.git.mst@redhat.com> (raw)
In-Reply-To: <cover.1696477105.git.mst@redhat.com>
From: Ilya Maximets <i.maximets@ovn.org>
Lots of virtio functions that are on a hot path in data transmission
are initializing indirect descriptor cache at the point of stack
allocation. It's a 112 byte structure that is getting zeroed out on
each call adding unnecessary overhead. It's going to be correctly
initialized later via special init function. The only reason to
actually initialize right away is the ability to safely destruct it.
Replacing a designated initializer with a function to only initialize
what is necessary.
Removal of the unnecessary stack initializations improves throughput
of virtio-net devices in terms of 64B packets per second by 6-14 %
depending on the case. Tested with a proposed af-xdp network backend
and a dpdk testpmd application in the guest, but should be beneficial
for other virtio devices as well.
Signed-off-by: Ilya Maximets <i.maximets@ovn.org>
Message-Id: <20230811143423.3258788-1-i.maximets@ovn.org>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Acked-by: Jason Wang <jasowang@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
---
include/exec/memory.h | 16 +++++++++++++---
hw/virtio/virtio.c | 20 +++++++++++++++-----
2 files changed, 28 insertions(+), 8 deletions(-)
diff --git a/include/exec/memory.h b/include/exec/memory.h
index ef23d65afc..c99842d2fc 100644
--- a/include/exec/memory.h
+++ b/include/exec/memory.h
@@ -2671,9 +2671,6 @@ struct MemoryRegionCache {
bool is_write;
};
-#define MEMORY_REGION_CACHE_INVALID ((MemoryRegionCache) { .mrs.mr = NULL })
-
-
/* address_space_ld*_cached: load from a cached #MemoryRegion
* address_space_st*_cached: store into a cached #MemoryRegion
*
@@ -2762,6 +2759,19 @@ int64_t address_space_cache_init(MemoryRegionCache *cache,
hwaddr len,
bool is_write);
+/**
+ * address_space_cache_init_empty: Initialize empty #MemoryRegionCache
+ *
+ * @cache: The #MemoryRegionCache to operate on.
+ *
+ * Initializes #MemoryRegionCache structure without memory region attached.
+ * Cache initialized this way can only be safely destroyed, but not used.
+ */
+static inline void address_space_cache_init_empty(MemoryRegionCache *cache)
+{
+ cache->mrs.mr = NULL;
+}
+
/**
* address_space_cache_invalidate: complete a write to a #MemoryRegionCache
*
diff --git a/hw/virtio/virtio.c b/hw/virtio/virtio.c
index 4577f3f5b3..d3a22e3d36 100644
--- a/hw/virtio/virtio.c
+++ b/hw/virtio/virtio.c
@@ -1071,10 +1071,12 @@ static void virtqueue_split_get_avail_bytes(VirtQueue *vq,
VirtIODevice *vdev = vq->vdev;
unsigned int idx;
unsigned int total_bufs, in_total, out_total;
- MemoryRegionCache indirect_desc_cache = MEMORY_REGION_CACHE_INVALID;
+ MemoryRegionCache indirect_desc_cache;
int64_t len = 0;
int rc;
+ address_space_cache_init_empty(&indirect_desc_cache);
+
idx = vq->last_avail_idx;
total_bufs = in_total = out_total = 0;
@@ -1207,12 +1209,14 @@ static void virtqueue_packed_get_avail_bytes(VirtQueue *vq,
VirtIODevice *vdev = vq->vdev;
unsigned int idx;
unsigned int total_bufs, in_total, out_total;
+ MemoryRegionCache indirect_desc_cache;
MemoryRegionCache *desc_cache;
- MemoryRegionCache indirect_desc_cache = MEMORY_REGION_CACHE_INVALID;
int64_t len = 0;
VRingPackedDesc desc;
bool wrap_counter;
+ address_space_cache_init_empty(&indirect_desc_cache);
+
idx = vq->last_avail_idx;
wrap_counter = vq->last_avail_wrap_counter;
total_bufs = in_total = out_total = 0;
@@ -1487,7 +1491,7 @@ static void *virtqueue_split_pop(VirtQueue *vq, size_t sz)
{
unsigned int i, head, max;
VRingMemoryRegionCaches *caches;
- MemoryRegionCache indirect_desc_cache = MEMORY_REGION_CACHE_INVALID;
+ MemoryRegionCache indirect_desc_cache;
MemoryRegionCache *desc_cache;
int64_t len;
VirtIODevice *vdev = vq->vdev;
@@ -1498,6 +1502,8 @@ static void *virtqueue_split_pop(VirtQueue *vq, size_t sz)
VRingDesc desc;
int rc;
+ address_space_cache_init_empty(&indirect_desc_cache);
+
RCU_READ_LOCK_GUARD();
if (virtio_queue_empty_rcu(vq)) {
goto done;
@@ -1624,7 +1630,7 @@ static void *virtqueue_packed_pop(VirtQueue *vq, size_t sz)
{
unsigned int i, max;
VRingMemoryRegionCaches *caches;
- MemoryRegionCache indirect_desc_cache = MEMORY_REGION_CACHE_INVALID;
+ MemoryRegionCache indirect_desc_cache;
MemoryRegionCache *desc_cache;
int64_t len;
VirtIODevice *vdev = vq->vdev;
@@ -1636,6 +1642,8 @@ static void *virtqueue_packed_pop(VirtQueue *vq, size_t sz)
uint16_t id;
int rc;
+ address_space_cache_init_empty(&indirect_desc_cache);
+
RCU_READ_LOCK_GUARD();
if (virtio_queue_packed_empty_rcu(vq)) {
goto done;
@@ -3970,13 +3978,15 @@ VirtioQueueElement *qmp_x_query_virtio_queue_element(const char *path,
} else {
unsigned int head, i, max;
VRingMemoryRegionCaches *caches;
- MemoryRegionCache indirect_desc_cache = MEMORY_REGION_CACHE_INVALID;
+ MemoryRegionCache indirect_desc_cache;
MemoryRegionCache *desc_cache;
VRingDesc desc;
VirtioRingDescList *list = NULL;
VirtioRingDescList *node;
int rc; int ndescs;
+ address_space_cache_init_empty(&indirect_desc_cache);
+
RCU_READ_LOCK_GUARD();
max = vq->vring.num;
--
MST
next prev parent reply other threads:[~2023-10-05 3:48 UTC|newest]
Thread overview: 64+ messages / expand[flat|nested] mbox.gz Atom feed top
2023-10-05 3:42 [PULL v2 00/53] virtio,pci: features, cleanups Michael S. Tsirkin
2023-10-05 3:42 ` [PULL v2 01/53] pci: SLT must be RO Michael S. Tsirkin
2023-10-05 3:42 ` [PULL v2 02/53] hw/virtio: Propagate page_mask to vhost_vdpa_listener_skipped_section() Michael S. Tsirkin
2023-10-05 3:42 ` [PULL v2 03/53] hw/virtio: Propagate page_mask to vhost_vdpa_section_end() Michael S. Tsirkin
2023-10-05 3:42 ` [PULL v2 04/53] hw/virtio/vhost-vdpa: Inline TARGET_PAGE_ALIGN() macro Michael S. Tsirkin
2023-10-05 3:42 ` [PULL v2 05/53] hw/virtio/vhost-vdpa: Use target-agnostic qemu_target_page_mask() Michael S. Tsirkin
2023-10-05 3:42 ` [PULL v2 06/53] hw/virtio: Build vhost-vdpa.o once Michael S. Tsirkin
2023-10-05 3:42 ` [PULL v2 07/53] hw/virtio/meson: Rename softmmu_virtio_ss[] -> system_virtio_ss[] Michael S. Tsirkin
2023-10-05 3:42 ` [PULL v2 08/53] virtio: add vhost-user-base and a generic vhost-user-device Michael S. Tsirkin
2023-10-05 3:42 ` [PULL v2 09/53] hw/virtio: add config support to vhost-user-device Michael S. Tsirkin
2023-10-05 3:42 ` [PULL v2 10/53] virtio-net: do not reset vlan filtering at set_features Michael S. Tsirkin
2023-10-05 3:43 ` [PULL v2 11/53] virtio-net: Expose MAX_VLAN Michael S. Tsirkin
2023-10-05 3:43 ` [PULL v2 12/53] vdpa: Restore vlan filtering state Michael S. Tsirkin
2023-10-05 3:43 ` [PULL v2 13/53] vdpa: Allow VIRTIO_NET_F_CTRL_VLAN in SVQ Michael S. Tsirkin
2023-10-05 3:43 ` Michael S. Tsirkin [this message]
2023-10-05 3:43 ` [PULL v2 15/53] vdpa: use first queue SVQ state for CVQ default Michael S. Tsirkin
2023-10-05 3:43 ` [PULL v2 16/53] vdpa: export vhost_vdpa_set_vring_ready Michael S. Tsirkin
2023-10-05 3:43 ` [PULL v2 17/53] vdpa: rename vhost_vdpa_net_load to vhost_vdpa_net_cvq_load Michael S. Tsirkin
2023-10-05 3:43 ` [PULL v2 18/53] vdpa: move vhost_vdpa_set_vring_ready to the caller Michael S. Tsirkin
2023-10-05 3:43 ` [PULL v2 19/53] vdpa: remove net cvq migration blocker Michael S. Tsirkin
2023-10-05 3:43 ` [PULL v2 20/53] vhost: Add count argument to vhost_svq_poll() Michael S. Tsirkin
2023-10-05 3:43 ` [PULL v2 21/53] qmp: remove virtio_list, search QOM tree instead Michael S. Tsirkin
2023-10-05 3:43 ` [PULL v2 22/53] qmp: update virtio feature maps, vhost-user-gpio introspection Michael S. Tsirkin
2023-10-05 3:43 ` [PULL v2 23/53] vhost-user: move VhostUserProtocolFeature definition to header file Michael S. Tsirkin
2023-10-05 3:44 ` [PULL v2 24/53] hw/isa/ich9: Add comment on imperfect emulation of PIC vs. I/O APIC routing Michael S. Tsirkin
2023-10-05 3:44 ` [PULL v2 25/53] hw/i386/acpi-build: Use pc_madt_cpu_entry() directly Michael S. Tsirkin
2023-10-05 3:44 ` [PULL v2 26/53] hw/acpi/cpu: Have build_cpus_aml() take a build_madt_cpu_fn callback Michael S. Tsirkin
2023-10-05 3:44 ` [PULL v2 27/53] hw/acpi/acpi_dev_interface: Remove now unused madt_cpu virtual method Michael S. Tsirkin
2023-10-05 3:44 ` [PULL v2 28/53] hw/acpi/acpi_dev_interface: Remove now unused #include "hw/boards.h" Michael S. Tsirkin
2023-10-05 3:44 ` [PULL v2 29/53] hw/i386: Remove now redundant TYPE_ACPI_GED_X86 Michael S. Tsirkin
2023-10-18 17:38 ` Salil Mehta
2023-10-19 10:33 ` Bernhard Beschow
2023-10-19 18:15 ` Michael S. Tsirkin
2023-10-20 23:54 ` Salil Mehta
2023-10-27 11:22 ` Igor Mammedov
2023-10-27 11:50 ` Igor Mammedov
2023-10-05 3:44 ` [PULL v2 30/53] hw/i386/acpi-build: Determine SMI command port just once Michael S. Tsirkin
2023-10-05 3:44 ` [PULL v2 31/53] hw/acpi: Trace GPE access in all device models, not just PIIX4 Michael S. Tsirkin
2023-10-05 3:44 ` [PULL v2 32/53] hw/acpi/core: Trace enable and status registers of GPE separately Michael S. Tsirkin
2023-10-05 3:44 ` [PULL v2 33/53] vdpa: fix gcc cvq_isolated uninitialized variable warning Michael S. Tsirkin
2023-10-05 3:44 ` [PULL v2 34/53] vdpa net: zero vhost_vdpa iova_tree pointer at cleanup Michael S. Tsirkin
2023-10-05 3:44 ` [PULL v2 35/53] hw/cxl: Push cxl_decoder_count_enc() and cxl_decode_ig() into .c Michael S. Tsirkin
2023-10-05 3:44 ` [PULL v2 36/53] hw/cxl: Add utility functions decoder interleave ways and target count Michael S. Tsirkin
2023-10-05 3:44 ` [PULL v2 37/53] hw/cxl: Fix and use same calculation for HDM decoder block size everywhere Michael S. Tsirkin
2023-10-05 3:45 ` [PULL v2 38/53] hw/cxl: Support 4 HDM decoders at all levels of topology Michael S. Tsirkin
2023-10-19 12:31 ` Peter Maydell
2023-10-19 14:04 ` Jonathan Cameron via
2023-10-19 14:04 ` Jonathan Cameron
2023-10-05 3:45 ` [PULL v2 39/53] hw/pci-bridge/cxl-upstream: Add serial number extended capability support Michael S. Tsirkin
2023-10-05 3:45 ` [PULL v2 40/53] vdpa net: fix error message setting virtio status Michael S. Tsirkin
2023-10-05 3:45 ` [PULL v2 41/53] vdpa net: stop probing if cannot set features Michael S. Tsirkin
2023-10-05 3:45 ` [PULL v2 42/53] vdpa net: follow VirtIO initialization properly at cvq isolation probing Michael S. Tsirkin
2023-10-05 3:45 ` [PULL v2 43/53] amd_iommu: Fix APIC address check Michael S. Tsirkin
2023-10-05 3:45 ` [PULL v2 44/53] hw/i386/pc: improve physical address space bound check for 32-bit x86 systems Michael S. Tsirkin
2023-10-05 3:45 ` [PULL v2 45/53] pcie_sriov: unregister_vfs(): fix error path Michael S. Tsirkin
2023-10-05 3:45 ` [PULL v2 46/53] libvhost-user.c: add assertion to vu_message_read_default Michael S. Tsirkin
2023-10-05 3:45 ` [PULL v2 47/53] virtio: use shadow_avail_idx while checking number of heads Michael S. Tsirkin
2023-10-05 3:45 ` [PULL v2 48/53] virtio: remove unnecessary thread fence while reading next descriptor Michael S. Tsirkin
2023-10-05 3:45 ` [PULL v2 49/53] virtio: remove unused next argument from virtqueue_split_read_next_desc() Michael S. Tsirkin
2023-10-05 3:45 ` [PULL v2 50/53] util/uuid: add a hash function Michael S. Tsirkin
2023-10-05 3:45 ` [PULL v2 51/53] hw/display: introduce virtio-dmabuf Michael S. Tsirkin
2023-10-05 3:45 ` [PULL v2 52/53] vhost-user: add shared_object msg Michael S. Tsirkin
2023-10-05 3:45 ` [PULL v2 53/53] libvhost-user: handle " Michael S. Tsirkin
2023-10-05 15:56 ` [PULL v2 00/53] virtio,pci: features, cleanups Stefan Hajnoczi
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=43d6376980d5567f2a6d00cfb30d10c0961671e6.1696477105.git.mst@redhat.com \
--to=mst@redhat.com \
--cc=david@redhat.com \
--cc=i.maximets@ovn.org \
--cc=jasowang@redhat.com \
--cc=pbonzini@redhat.com \
--cc=peter.maydell@linaro.org \
--cc=peterx@redhat.com \
--cc=philmd@linaro.org \
--cc=qemu-devel@nongnu.org \
--cc=stefanha@redhat.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;
as well as URLs for NNTP newsgroup(s).