From: "Cédric Le Goater" <clg@redhat.com>
To: Zhenzhong Duan <zhenzhong.duan@intel.com>, qemu-devel@nongnu.org
Cc: alex.williamson@redhat.com, jgg@nvidia.com, nicolinc@nvidia.com,
joao.m.martins@oracle.com, eric.auger@redhat.com,
peterx@redhat.com, jasowang@redhat.com, kevin.tian@intel.com,
yi.l.liu@intel.com, yi.y.sun@intel.com, chao.p.peng@intel.com
Subject: Re: [PATCH v3 27/37] vfio/iommufd: Switch to manual hwpt allocation
Date: Mon, 30 Oct 2023 14:52:21 +0100 [thread overview]
Message-ID: <4adcd2be-98fb-45bc-9023-a08b035b8b71@redhat.com> (raw)
In-Reply-To: <20231026103104.1686921-28-zhenzhong.duan@intel.com>
On 10/26/23 12:30, Zhenzhong Duan wrote:
> IOMMUFD supports auto allocated hwpt and manually allocated one.
> Manually allocated hwpt has benefit that its life cycle is under
> user's control, so it could be used as stage 2 page table by nested
> feature in the future.
Would an option be useful to switch from one mode to another ?
>
> Introduce two helpers __vfio_device_attach/detach_hwpt to facilitate
> this change.
I think merging this change with the previous patch makes sense.
It doesn't add much to keep it as a standalone patch unless we
want a feature toggle.
Thanks,
C.
> Signed-off-by: Zhenzhong Duan <zhenzhong.duan@intel.com>
> ---
> hw/vfio/iommufd.c | 89 +++++++++++++++++++++++++++++++++++++----------
> 1 file changed, 70 insertions(+), 19 deletions(-)
>
> diff --git a/hw/vfio/iommufd.c b/hw/vfio/iommufd.c
> index aee64d63f3..c1daaf1c39 100644
> --- a/hw/vfio/iommufd.c
> +++ b/hw/vfio/iommufd.c
> @@ -217,38 +217,91 @@ static VFIOIOASHwpt *vfio_container_get_hwpt(VFIOIOMMUFDContainer *container,
> static void vfio_container_put_hwpt(IOMMUFDBackend *be, VFIOIOASHwpt *hwpt)
> {
> QLIST_REMOVE(hwpt, next);
> + iommufd_backend_free_id(be->fd, hwpt->hwpt_id);
> g_free(hwpt);
> }
>
> -static int vfio_device_attach_container(VFIODevice *vbasedev,
> - VFIOIOMMUFDContainer *container,
> - Error **errp)
> +static int __vfio_device_attach_hwpt(VFIODevice *vbasedev, uint32_t hwpt_id,
> + Error **errp)
> {
> - int ret, iommufd = vbasedev->iommufd->fd;
> - VFIOIOASHwpt *hwpt;
> struct vfio_device_attach_iommufd_pt attach_data = {
> .argsz = sizeof(attach_data),
> .flags = 0,
> - .pt_id = container->ioas_id,
> + .pt_id = hwpt_id,
> };
> + int ret;
>
> - /* Attach device to an ioas within iommufd */
> ret = ioctl(vbasedev->fd, VFIO_DEVICE_ATTACH_IOMMUFD_PT, &attach_data);
> if (ret) {
> error_setg_errno(errp, errno,
> - "[iommufd=%d] error attach %s (%d) to ioasid=%d",
> - container->be->fd, vbasedev->name, vbasedev->fd,
> - attach_data.pt_id);
> + "[iommufd=%d] error attach %s (%d) to hwpt_id=%d",
> + vbasedev->iommufd->fd, vbasedev->name, vbasedev->fd,
> + hwpt_id);
> + }
> + return ret;
> +}
> +
> +static int __vfio_device_detach_hwpt(VFIODevice *vbasedev, Error **errp)
> +{
> + struct vfio_device_detach_iommufd_pt detach_data = {
> + .argsz = sizeof(detach_data),
> + .flags = 0,
> + };
> + int ret;
> +
> + ret = ioctl(vbasedev->fd, VFIO_DEVICE_DETACH_IOMMUFD_PT, &detach_data);
> + if (ret) {
> + error_setg_errno(errp, errno, "detach %s from ioas failed",
> + vbasedev->name);
> + }
> + return ret;
> +}
> +
> +static int vfio_device_attach_container(VFIODevice *vbasedev,
> + VFIOIOMMUFDContainer *container,
> + Error **errp)
> +{
> + int ret, iommufd = vbasedev->iommufd->fd;
> + VFIOIOASHwpt *hwpt;
> + uint32_t hwpt_id;
> + Error *err = NULL;
> +
> + /* try to attach to an existing hwpt in this container */
> + QLIST_FOREACH(hwpt, &container->hwpt_list, next) {
> + ret = __vfio_device_attach_hwpt(vbasedev, hwpt->hwpt_id, &err);
> + if (ret) {
> + const char *msg = error_get_pretty(err);
> +
> + trace_vfio_iommufd_fail_attach_existing_hwpt(msg);
> + error_free(err);
> + err = NULL;
> + } else {
> + goto found_hwpt;
> + }
> + }
> +
> + ret = iommufd_backend_alloc_hwpt(iommufd, vbasedev->devid,
> + container->ioas_id, &hwpt_id);
> +
> + if (ret) {
> + error_setg_errno(errp, errno, "error alloc shadow hwpt");
> return ret;
> + }
>
> + /* Attach device to an hwpt within iommufd */
> + ret = __vfio_device_attach_hwpt(vbasedev, hwpt_id, errp);
> + if (ret) {
> + iommufd_backend_free_id(iommufd, hwpt_id);
> + return ret;
> }
> - hwpt = vfio_container_get_hwpt(container, attach_data.pt_id);
>
> + hwpt = vfio_container_get_hwpt(container, hwpt_id);
> +found_hwpt:
> QLIST_INSERT_HEAD(&hwpt->device_list, vbasedev, next);
> vbasedev->hwpt = hwpt;
>
> trace_vfio_iommufd_attach_device(iommufd, vbasedev->name, vbasedev->fd,
> - container->ioas_id, attach_data.pt_id);
> + container->ioas_id, hwpt->hwpt_id);
> return ret;
> }
>
> @@ -256,14 +309,12 @@ static void vfio_device_detach_container(VFIODevice *vbasedev,
> VFIOIOMMUFDContainer *container)
> {
> VFIOIOASHwpt *hwpt = vbasedev->hwpt;
> - struct vfio_device_attach_iommufd_pt detach_data = {
> - .argsz = sizeof(detach_data),
> - .flags = 0,
> - };
> + Error *err = NULL;
> + int ret;
>
> - if (ioctl(vbasedev->fd, VFIO_DEVICE_DETACH_IOMMUFD_PT, &detach_data)) {
> - error_report("detach %s from ioas id=%d failed: %m", vbasedev->name,
> - container->ioas_id);
> + ret = __vfio_device_detach_hwpt(vbasedev, &err);
> + if (ret) {
> + error_report_err(err);
> }
>
> QLIST_REMOVE(vbasedev, next);
next prev parent reply other threads:[~2023-10-30 13:52 UTC|newest]
Thread overview: 95+ messages / expand[flat|nested] mbox.gz Atom feed top
2023-10-26 10:30 [PATCH v3 00/37] vfio: Adopt iommufd Zhenzhong Duan
2023-10-26 10:30 ` [PATCH v3 01/37] vfio/container: Move IBM EEH related functions into spapr_pci_vfio.c Zhenzhong Duan
2023-10-26 14:23 ` Eric Farman
2023-10-27 9:19 ` Cédric Le Goater
2023-10-26 10:30 ` [PATCH v3 02/37] vfio/container: Move vfio_container_add/del_section_window into spapr.c Zhenzhong Duan
2023-10-27 9:19 ` Cédric Le Goater
2023-10-26 10:30 ` [PATCH v3 03/37] vfio/container: Move spapr specific init/deinit " Zhenzhong Duan
2023-10-27 9:27 ` Cédric Le Goater
2023-10-26 10:30 ` [PATCH v3 04/37] vfio/spapr: Make vfio_spapr_create/remove_window static Zhenzhong Duan
2023-10-27 9:27 ` Cédric Le Goater
2023-10-26 10:30 ` [PATCH v3 05/37] vfio/common: Move vfio_host_win_add/del into spapr.c Zhenzhong Duan
2023-10-27 9:30 ` Cédric Le Goater
2023-10-26 10:30 ` [PATCH v3 06/37] vfio: Introduce base object for VFIOContainer and targetted interface Zhenzhong Duan
2023-10-27 14:02 ` Cédric Le Goater
2023-10-30 2:40 ` Duan, Zhenzhong
2023-10-31 7:57 ` Cédric Le Goater
2023-10-31 8:31 ` Duan, Zhenzhong
2023-10-26 10:30 ` [PATCH v3 07/37] vfio/container: Introduce a empty VFIOIOMMUOps Zhenzhong Duan
2023-10-27 14:20 ` Cédric Le Goater
2023-10-30 2:43 ` Duan, Zhenzhong
2023-10-31 8:21 ` Cédric Le Goater
2023-10-26 10:30 ` [PATCH v3 08/37] vfio/container: Switch to dma_map|unmap API Zhenzhong Duan
2023-10-27 14:26 ` Cédric Le Goater
2023-10-30 2:48 ` Duan, Zhenzhong
2023-10-26 10:30 ` [PATCH v3 09/37] vfio/common: Move giommu_list in base container Zhenzhong Duan
2023-10-27 14:38 ` Cédric Le Goater
2023-10-30 2:48 ` Duan, Zhenzhong
2023-10-26 10:30 ` [PATCH v3 10/37] vfio/container: Move space field to " Zhenzhong Duan
2023-10-27 14:46 ` Cédric Le Goater
2023-10-30 2:51 ` Duan, Zhenzhong
2023-10-31 8:24 ` Cédric Le Goater
2023-10-26 10:30 ` [PATCH v3 11/37] vfio/container: Switch to IOMMU BE set_dirty_page_tracking/query_dirty_bitmap API Zhenzhong Duan
2023-10-27 14:53 ` Cédric Le Goater
2023-10-30 2:53 ` Duan, Zhenzhong
2023-10-26 10:30 ` [PATCH v3 12/37] vfio/container: Move per container device list in base container Zhenzhong Duan
2023-10-27 15:03 ` Cédric Le Goater
2023-10-26 10:30 ` [PATCH v3 13/37] vfio/container: Convert functions to " Zhenzhong Duan
2023-10-27 15:03 ` Cédric Le Goater
2023-10-26 10:30 ` [PATCH v3 14/37] vfio/container: Move vrdl_list, pgsizes and dma_max_mappings " Zhenzhong Duan
2023-10-27 15:52 ` Cédric Le Goater
2023-10-30 3:14 ` Duan, Zhenzhong
2023-10-31 8:26 ` Cédric Le Goater
2023-10-26 10:30 ` [PATCH v3 15/37] vfio/container: Move listener " Zhenzhong Duan
2023-10-26 10:30 ` [PATCH v3 16/37] vfio/container: Move dirty_pgsizes and max_dirty_bitmap_size " Zhenzhong Duan
2023-10-27 16:01 ` Cédric Le Goater
2023-10-26 10:30 ` [PATCH v3 17/37] vfio/container: Move iova_ranges " Zhenzhong Duan
2023-10-26 10:30 ` [PATCH v3 18/37] vfio/container: Implement attach/detach_device Zhenzhong Duan
2023-10-27 16:04 ` Cédric Le Goater
2023-10-27 16:06 ` Cédric Le Goater
2023-10-30 3:20 ` Duan, Zhenzhong
2023-10-26 10:30 ` [PATCH v3 19/37] vfio/spapr: Introduce spapr backend and target interface Zhenzhong Duan
2023-10-27 16:04 ` Cédric Le Goater
2023-10-30 3:15 ` Duan, Zhenzhong
2023-10-26 10:30 ` [PATCH v3 20/37] vfio/spapr: switch to spapr IOMMU BE add/del_section_window Zhenzhong Duan
2023-10-26 10:30 ` [PATCH v3 21/37] vfio/spapr: Move prereg_listener into spapr container Zhenzhong Duan
2023-10-26 10:30 ` [PATCH v3 22/37] vfio/spapr: Move hostwin_list " Zhenzhong Duan
2023-10-26 10:30 ` [PATCH v3 23/37] Add iommufd configure option Zhenzhong Duan
2023-10-31 8:58 ` Cédric Le Goater
2023-10-31 10:52 ` Duan, Zhenzhong
2023-10-31 11:25 ` Cédric Le Goater
2023-10-26 10:30 ` [PATCH v3 24/37] backends/iommufd: Introduce the iommufd object Zhenzhong Duan
2023-10-26 13:27 ` Markus Armbruster
2023-10-27 7:50 ` Duan, Zhenzhong
2023-10-27 8:30 ` Markus Armbruster
2023-10-27 9:41 ` Duan, Zhenzhong
2023-10-26 10:30 ` [PATCH v3 25/37] util/char_dev: Add open_cdev() Zhenzhong Duan
2023-10-30 14:53 ` Cédric Le Goater
2023-10-31 1:59 ` Duan, Zhenzhong
2023-10-26 10:30 ` [PATCH v3 26/37] vfio/iommufd: Implement the iommufd backend Zhenzhong Duan
2023-10-26 10:30 ` [PATCH v3 27/37] vfio/iommufd: Switch to manual hwpt allocation Zhenzhong Duan
2023-10-30 13:52 ` Cédric Le Goater [this message]
2023-10-31 2:02 ` Duan, Zhenzhong
2023-10-26 10:30 ` [PATCH v3 28/37] vfio/iommufd: Add support for iova_ranges Zhenzhong Duan
2023-10-26 10:30 ` [PATCH v3 29/37] vfio/iommufd: Bypass EEH if iommufd backend Zhenzhong Duan
2023-10-30 13:56 ` Cédric Le Goater
2023-10-31 2:26 ` Duan, Zhenzhong
2023-10-31 9:01 ` Cédric Le Goater
2023-10-31 9:06 ` Duan, Zhenzhong
2023-10-26 10:30 ` [PATCH v3 30/37] vfio/pci: Extract out a helper vfio_pci_get_pci_hot_reset_info Zhenzhong Duan
2023-10-30 13:59 ` Cédric Le Goater
2023-10-31 2:30 ` Duan, Zhenzhong
2023-10-26 10:30 ` [PATCH v3 31/37] vfio/pci: Adapt vfio pci hot reset support with iommufd BE Zhenzhong Duan
2023-10-30 14:04 ` Cédric Le Goater
2023-10-31 2:27 ` Duan, Zhenzhong
2023-10-26 10:30 ` [PATCH v3 32/37] vfio/pci: Allow the selection of a given iommu backend Zhenzhong Duan
2023-10-26 10:31 ` [PATCH v3 33/37] vfio/pci: Make vfio cdev pre-openable by passing a file handle Zhenzhong Duan
2023-10-26 10:31 ` [PATCH v3 34/37] vfio: Allow the selection of a given iommu backend for platform ap and ccw Zhenzhong Duan
2023-10-26 10:31 ` [PATCH v3 35/37] vfio/platform: Make vfio cdev pre-openable by passing a file handle Zhenzhong Duan
2023-10-26 10:31 ` [PATCH v3 36/37] vfio/ap: " Zhenzhong Duan
2023-10-26 10:31 ` [PATCH v3 37/37] vfio/ccw: " Zhenzhong Duan
2023-10-26 13:27 ` [PATCH v3 00/37] vfio: Adopt iommufd Markus Armbruster
2023-10-27 6:17 ` Duan, Zhenzhong
2023-10-27 7:45 ` Cédric Le Goater
2023-10-27 8:16 ` Duan, Zhenzhong
2023-10-27 13:43 ` 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=4adcd2be-98fb-45bc-9023-a08b035b8b71@redhat.com \
--to=clg@redhat.com \
--cc=alex.williamson@redhat.com \
--cc=chao.p.peng@intel.com \
--cc=eric.auger@redhat.com \
--cc=jasowang@redhat.com \
--cc=jgg@nvidia.com \
--cc=joao.m.martins@oracle.com \
--cc=kevin.tian@intel.com \
--cc=nicolinc@nvidia.com \
--cc=peterx@redhat.com \
--cc=qemu-devel@nongnu.org \
--cc=yi.l.liu@intel.com \
--cc=yi.y.sun@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).