From: Alex Williamson <alex.williamson@redhat.com>
To: Eric Auger <eric.auger@redhat.com>
Cc: eric.auger.pro@gmail.com, qemu-devel@nongnu.org,
qemu-arm@nongnu.org, clg@redhat.com, jean-philippe@linaro.org,
mst@redhat.com, pbonzini@redhat.com, peter.maydell@linaro.org,
peterx@redhat.com, david@redhat.com, philmd@linaro.org,
zhenzhong.duan@intel.com, yi.l.liu@intel.com
Subject: Re: [PATCH v3 03/13] vfio: Collect container iova range info
Date: Wed, 18 Oct 2023 13:07:37 -0600 [thread overview]
Message-ID: <20231018130737.3815d3c4.alex.williamson@redhat.com> (raw)
In-Reply-To: <20231011175516.541374-4-eric.auger@redhat.com>
On Wed, 11 Oct 2023 19:52:19 +0200
Eric Auger <eric.auger@redhat.com> wrote:
> Collect iova range information if VFIO_IOMMU_TYPE1_INFO_CAP_IOVA_RANGE
> capability is supported.
>
> This allows to propagate the information though the IOMMU MR
> set_iova_ranges() callback so that virtual IOMMUs
> get aware of those aperture constraints. This is only done if
> the info is available and the number of iova ranges is greater than
> 0.
>
> A new vfio_get_info_iova_range helper is introduced matching
> the coding style of existing vfio_get_info_dma_avail. The
> boolean returned value isn't used though. Code is aligned
> between both.
>
> Signed-off-by: Eric Auger <eric.auger@redhat.com>
>
> ---
>
> v2 -> v3:
> - Turn nr_iovas into a int initialized to -1
> - memory_region_iommu_set_iova_ranges only is called if nr_iovas > 0
> - vfio_get_info_iova_range returns a bool to match
> vfio_get_info_dma_avail. Uniformize both code by using !hdr in
> the check
> - rebase on top of vfio-next
> ---
> include/hw/vfio/vfio-common.h | 2 ++
> hw/vfio/common.c | 9 +++++++
> hw/vfio/container.c | 44 ++++++++++++++++++++++++++++++++---
> 3 files changed, 52 insertions(+), 3 deletions(-)
>
> diff --git a/include/hw/vfio/vfio-common.h b/include/hw/vfio/vfio-common.h
> index 7780b9073a..848ff47960 100644
> --- a/include/hw/vfio/vfio-common.h
> +++ b/include/hw/vfio/vfio-common.h
> @@ -99,6 +99,8 @@ typedef struct VFIOContainer {
> QLIST_HEAD(, VFIORamDiscardListener) vrdl_list;
> QLIST_ENTRY(VFIOContainer) next;
> QLIST_HEAD(, VFIODevice) device_list;
> + int nr_iovas;
> + GList *iova_ranges;
Nit, nr_iovas seems like it has a pretty weak use case here. We can
just test iova_ranges != NULL for calling set_iova_ranges. In patch 13
we can again test against NULL, which I think also negates the need to
assert nr_iovas since the NULL test automatically catches the zero
case. Otherwise
Reviewed-by: Alex Williamson <alex.williamson@redhat.com>
> } VFIOContainer;
>
> typedef struct VFIOGuestIOMMU {
> diff --git a/hw/vfio/common.c b/hw/vfio/common.c
> index 5ff5acf1d8..9d804152ba 100644
> --- a/hw/vfio/common.c
> +++ b/hw/vfio/common.c
> @@ -699,6 +699,15 @@ static void vfio_listener_region_add(MemoryListener *listener,
> goto fail;
> }
>
> + if (container->nr_iovas > 0) {
> + ret = memory_region_iommu_set_iova_ranges(giommu->iommu_mr,
> + container->iova_ranges, &err);
> + if (ret) {
> + g_free(giommu);
> + goto fail;
> + }
> + }
> +
> ret = memory_region_register_iommu_notifier(section->mr, &giommu->n,
> &err);
> if (ret) {
> diff --git a/hw/vfio/container.c b/hw/vfio/container.c
> index adc467210f..5122ff6d92 100644
> --- a/hw/vfio/container.c
> +++ b/hw/vfio/container.c
> @@ -382,7 +382,7 @@ bool vfio_get_info_dma_avail(struct vfio_iommu_type1_info *info,
> /* If the capability cannot be found, assume no DMA limiting */
> hdr = vfio_get_iommu_type1_info_cap(info,
> VFIO_IOMMU_TYPE1_INFO_DMA_AVAIL);
> - if (hdr == NULL) {
> + if (!hdr) {
> return false;
> }
>
> @@ -394,6 +394,33 @@ bool vfio_get_info_dma_avail(struct vfio_iommu_type1_info *info,
> return true;
> }
>
> +static bool vfio_get_info_iova_range(struct vfio_iommu_type1_info *info,
> + VFIOContainer *container)
> +{
> + struct vfio_info_cap_header *hdr;
> + struct vfio_iommu_type1_info_cap_iova_range *cap;
> +
> + hdr = vfio_get_iommu_type1_info_cap(info,
> + VFIO_IOMMU_TYPE1_INFO_CAP_IOVA_RANGE);
> + if (!hdr) {
> + return false;
> + }
> +
> + cap = (void *)hdr;
> +
> + container->nr_iovas = cap->nr_iovas;
> + for (int i = 0; i < cap->nr_iovas; i++) {
> + Range *range = g_new(Range, 1);
> +
> + range_set_bounds(range, cap->iova_ranges[i].start,
> + cap->iova_ranges[i].end);
> + container->iova_ranges =
> + range_list_insert(container->iova_ranges, range);
> + }
> +
> + return true;
> +}
> +
> static void vfio_kvm_device_add_group(VFIOGroup *group)
> {
> Error *err = NULL;
> @@ -535,6 +562,12 @@ static void vfio_get_iommu_info_migration(VFIOContainer *container,
> }
> }
>
> +static void vfio_free_container(VFIOContainer *container)
> +{
> + g_list_free_full(container->iova_ranges, g_free);
> + g_free(container);
> +}
> +
> static int vfio_connect_container(VFIOGroup *group, AddressSpace *as,
> Error **errp)
> {
> @@ -616,6 +649,8 @@ static int vfio_connect_container(VFIOGroup *group, AddressSpace *as,
> container->error = NULL;
> container->dirty_pages_supported = false;
> container->dma_max_mappings = 0;
> + container->nr_iovas = -1;
> + container->iova_ranges = NULL;
> QLIST_INIT(&container->giommu_list);
> QLIST_INIT(&container->hostwin_list);
> QLIST_INIT(&container->vrdl_list);
> @@ -652,6 +687,9 @@ static int vfio_connect_container(VFIOGroup *group, AddressSpace *as,
> if (!vfio_get_info_dma_avail(info, &container->dma_max_mappings)) {
> container->dma_max_mappings = 65535;
> }
> +
> + vfio_get_info_iova_range(info, container);
> +
> vfio_get_iommu_info_migration(container, info);
> g_free(info);
>
> @@ -765,7 +803,7 @@ enable_discards_exit:
> vfio_ram_block_discard_disable(container, false);
>
> free_container_exit:
> - g_free(container);
> + vfio_free_container(container);
>
> close_fd_exit:
> close(fd);
> @@ -819,7 +857,7 @@ static void vfio_disconnect_container(VFIOGroup *group)
>
> trace_vfio_disconnect_container(container->fd);
> close(container->fd);
> - g_free(container);
> + vfio_free_container(container);
>
> vfio_put_address_space(space);
> }
next prev parent reply other threads:[~2023-10-18 19:08 UTC|newest]
Thread overview: 27+ messages / expand[flat|nested] mbox.gz Atom feed top
2023-10-11 17:52 [PATCH v3 00/13] VIRTIO-IOMMU/VFIO: Don't assume 64b IOVA space Eric Auger
2023-10-11 17:52 ` [PATCH v3 01/13] memory: Let ReservedRegion use Range Eric Auger
2023-10-11 17:52 ` [PATCH v3 02/13] memory: Introduce memory_region_iommu_set_iova_ranges Eric Auger
2023-10-18 22:07 ` Peter Xu
2023-10-11 17:52 ` [PATCH v3 03/13] vfio: Collect container iova range info Eric Auger
2023-10-18 19:07 ` Alex Williamson [this message]
2023-10-19 6:39 ` Eric Auger
2023-10-11 17:52 ` [PATCH v3 04/13] virtio-iommu: Rename reserved_regions into prop_resv_regions Eric Auger
2023-10-11 17:52 ` [PATCH v3 05/13] range: Make range_compare() public Eric Auger
2023-10-11 17:52 ` [PATCH v3 06/13] util/reserved-region: Add new ReservedRegion helpers Eric Auger
2023-10-11 17:52 ` [PATCH v3 07/13] virtio-iommu: Introduce per IOMMUDevice reserved regions Eric Auger
2023-10-11 17:52 ` [PATCH v3 08/13] range: Introduce range_inverse_array() Eric Auger
2023-10-11 17:52 ` [PATCH v3 09/13] virtio-iommu: Record whether a probe request has been issued Eric Auger
2023-10-11 17:52 ` [PATCH v3 10/13] virtio-iommu: Implement set_iova_ranges() callback Eric Auger
2023-10-11 17:52 ` [PATCH v3 11/13] virtio-iommu: Consolidate host reserved regions and property set ones Eric Auger
2023-10-11 17:52 ` [PATCH v3 12/13] test: Add some tests for range and resv-mem helpers Eric Auger
2023-10-30 7:48 ` Cédric Le Goater
2023-10-11 17:52 ` [PATCH v3 13/13] vfio: Remove 64-bit IOVA address space assumption Eric Auger
2023-10-18 21:42 ` Alex Williamson
2023-10-19 6:37 ` Eric Auger
2023-10-18 13:37 ` [PATCH v3 00/13] VIRTIO-IOMMU/VFIO: Don't assume 64b IOVA space Michael S. Tsirkin
2023-10-19 9:07 ` YangHang Liu
2023-10-19 9:08 ` Eric Auger
2023-10-19 11:07 ` Cédric Le Goater
2023-10-19 11:20 ` Michael S. Tsirkin
2023-10-19 13:51 ` Eric Auger
2023-10-19 17:40 ` Cédric Le Goater
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=20231018130737.3815d3c4.alex.williamson@redhat.com \
--to=alex.williamson@redhat.com \
--cc=clg@redhat.com \
--cc=david@redhat.com \
--cc=eric.auger.pro@gmail.com \
--cc=eric.auger@redhat.com \
--cc=jean-philippe@linaro.org \
--cc=mst@redhat.com \
--cc=pbonzini@redhat.com \
--cc=peter.maydell@linaro.org \
--cc=peterx@redhat.com \
--cc=philmd@linaro.org \
--cc=qemu-arm@nongnu.org \
--cc=qemu-devel@nongnu.org \
--cc=yi.l.liu@intel.com \
--cc=zhenzhong.duan@intel.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).