mirror of https://lore.kernel.org/lkml/
 help / color / mirror / Atom feed
From: "Danilo Krummrich" <dakr@kernel.org>
To: "Simon Song" <xinmengs@nvidia.com>
Cc: <acourbot@nvidia.com>, <kvm@vger.kernel.org>,
	<rust-for-linux@vger.kernel.org>, <nova-gpu@lists.linux.dev>,
	<linux-pci@vger.kernel.org>, <alex@shazbot.org>, <jgg@nvidia.com>,
	<yishaih@nvidia.com>, <skolothumtho@nvidia.com>,
	<kevin.tian@intel.com>, <airlied@gmail.com>, <simona@ffwll.ch>,
	<ojeda@kernel.org>, <alex.gaynor@gmail.com>,
	<boqun.feng@gmail.com>, <gary@garyguo.net>,
	<bjorn3_gh@protonmail.com>, <lossin@kernel.org>,
	<a.hindborg@kernel.org>, <aliceryhl@google.com>,
	<tmgross@umich.edu>, <jhubbard@nvidia.com>,
	<ecourtney@nvidia.com>, <cjia@nvidia.com>, <smitra@nvidia.com>,
	<kjaju@nvidia.com>, <alkumar@nvidia.com>, <ankita@nvidia.com>,
	<aniketa@nvidia.com>, <kwankhede@nvidia.com>,
	<targupta@nvidia.com>, <linux-kernel@vger.kernel.org>,
	<zhiwang@kernel.org>, <zhiw@nvidia.com>, <boqun@kernel.org>,
	<daniel.almeida@collabora.com>, <tamird@kernel.org>,
	<work@onurozkan.dev>, <jgg@ziepe.ca>, <bhelgaas@google.com>,
	<kwilczynski@kernel.org>, <dri-devel@lists.freedesktop.org>,
	<peter@colberg.org>
Subject: Re: [PATCH v2 11/12] gpu: nova-core: publish typed SR-IOV PF APIs
Date: Tue, 10 Nov 2026 16:50:31 +0100	[thread overview]
Message-ID: <DMRNF7IZQ1YI.1MRPH47QJC9XC@kernel.org> (raw)
In-Reply-To: <20261023154009.2246627-12-xinmengs@nvidia.com>

On Fri Oct 23, 2026 at 5:40 PM CEST, Simon Song wrote:
> diff --git a/drivers/gpu/nova-core/driver.rs b/drivers/gpu/nova-core/driver.rs
> index dc6febeee65..99beab18987 100644
> --- a/drivers/gpu/nova-core/driver.rs
> +++ b/drivers/gpu/nova-core/driver.rs
> @@ -35,6 +35,12 @@
>      }, //
>  };
>  
> +#[cfg(CONFIG_PCI_IOV)]
> +use kernel::types::ForLt;
> +
> +#[cfg(CONFIG_PCI_IOV)]
> +use crate::vgpu::vgpu_api::NovaCoreVfApi;
> +
>  /// Counter for generating unique auxiliary device IDs.
>  static AUXILIARY_ID_COUNTER: Atomic<u32> = Atomic::new(0);
>  
> @@ -42,6 +48,10 @@
>  pub(crate) struct NovaCore<'bound> {
>      /// Firmware-control registration.
>      _fwctl: fwctl::Registration<'bound, NovaCoreFwCtl>,
> +    #[cfg(CONFIG_PCI_IOV)]
> +    #[allow(clippy::type_complexity)]
> +    #[pin]
> +    _vf_registration: pci::VfRegistration<'bound, ForLt!(NovaCoreVfApi<'_>)>,
>      #[pin]
>      pub(crate) gpu: Gpu<'bound>,
>      bar: pci::Bar<'bound, BAR0_SIZE>,
> @@ -114,6 +124,11 @@ fn probe<'bound>(
>          pin_init::pin_init_scope(move || {
>              dev_dbg!(pdev, "Probe Nova Core GPU driver.\n");
>  
> +            #[cfg(CONFIG_PCI_IOV)]
> +            if pdev.is_virtfn() {
> +                return Err(ENODEV);
> +            }

There's a lot of config guards, but we should actually stub the corresponding
functions, so they work regardless.

I suggest something like the diff in [2], that should be worked into the IOV
series.

> +
>              pdev.enable_device_mem()?;
>              pdev.set_master();
>  
> @@ -144,6 +159,18 @@ fn probe<'bound>(
>                  // Run optional GPU selftests.
>                  #[cfg(CONFIG_NOVA_CORE_SELFTESTS)]
>                  _: { gpu.run_selftests(pdev) },
> +                #[cfg(CONFIG_PCI_IOV)]
> +                _vf_registration <- {
> +                    // SAFETY: `gpu` is initialized at its pinned address and
> +                    // outlives the registration and its borrowed API data.
> +                    let gpu = unsafe { &(*this.as_ptr()).gpu };

You can use pin-init self-reference for this, see also [1].

> +                    let api = NovaCoreVfApi::new(gpu, pdev);
> +                    // SAFETY: Probe has exclusive access to the registration
> +                    // slot and no VFs are enabled before successful probe. The
> +                    // registration keeps PF services available until VF removal
> +                    // completes and is dropped before its borrowed GPU resources.

VFs not being enabled before successful probe is not a safety requirement, the
API does ensure this.

The only requirement is to store it in the driver's bus device private data.

> +                    unsafe { pci::VfRegistration::new(pdev, Ok(api)) }
> +                },
>                  _reg: auxiliary::Registration::new(
>                      pdev.as_ref(),
>                      c"nova-drm",
> @@ -176,4 +203,24 @@ fn probe<'bound>(
>              }))
>          })
>      }
> +
> +    #[cfg(CONFIG_PCI_IOV)]
> +    fn sriov_enable<'bound, 'callback>(
> +        _dev: &'bound pci::Device<Core<'_>>,
> +        _this: Pin<&Self::Data<'bound>>,
> +        token: pci::SriovEnable<'callback>,
> +    ) -> Result<pci::SriovEnabled<'callback>> {
> +        let num_vfs = token.num_vfs();
> +        token.enable(num_vfs)

NIT: Maybe provide a pci::SriovEnable::enable_all() helper for this in the core code.

> +    }
> +
> +    #[cfg(CONFIG_PCI_IOV)]
> +    fn sriov_disable<'bound>(
> +        _dev: &'bound pci::Device<Core<'_>>,
> +        _this: Pin<&Self::Data<'bound>>,
> +        token: pci::SriovDisable<'_>,
> +    ) -> Result {
> +        token.disable();
> +        Ok(())
> +    }
>  }

[1] https://lore.kernel.org/nova-gpu/20261110152746.140351-1-dakr@kernel.org/

[2] Diff for IOV series

diff --git a/rust/helpers/pci.c b/rust/helpers/pci.c
index a8540f90b13c..0185baa7aacc 100644
--- a/rust/helpers/pci.c
+++ b/rust/helpers/pci.c
@@ -41,6 +41,27 @@ __rust_helper u32 rust_helper_pci_ext_cap_next(u32 header)
 }

 #ifndef CONFIG_PCI_IOV
+__rust_helper int rust_helper_pci_iov_vf_id(struct pci_dev *pdev)
+{
+	return pci_iov_vf_id(pdev);
+}
+
+__rust_helper int rust_helper_pci_enable_sriov(struct pci_dev *pdev,
+					    int nr_virtfn)
+{
+	return pci_enable_sriov(pdev, nr_virtfn);
+}
+
+__rust_helper void rust_helper_pci_disable_sriov(struct pci_dev *pdev)
+{
+	pci_disable_sriov(pdev);
+}
+
+__rust_helper int rust_helper_pci_num_vf(struct pci_dev *pdev)
+{
+	return pci_num_vf(pdev);
+}
+
 __rust_helper unsigned int
 rust_helper_pci_sriov_get_totalvfs(struct pci_dev *pdev)
 {
diff --git a/rust/kernel/pci.rs b/rust/kernel/pci.rs
index 0bcbf73bb363..dcdef0eb1ae2 100644
--- a/rust/kernel/pci.rs
+++ b/rust/kernel/pci.rs
@@ -36,7 +36,6 @@
 mod cap;
 mod id;
 mod io;
-#[cfg(CONFIG_PCI_IOV)]
 mod iov;
 mod irq;

@@ -60,7 +59,6 @@
     Extended,
     Normal, //
 };
-#[cfg(CONFIG_PCI_IOV)]
 pub use self::iov::{
     SriovDisable,
     SriovEnable,
@@ -96,7 +94,6 @@ unsafe fn register(
         name: &'static CStr,
         module: &'static ThisModule,
     ) -> Result {
-        #[cfg(CONFIG_PCI_IOV)]
         build_assert!(
             T::HAS_SRIOV_ENABLE == T::HAS_SRIOV_DISABLE,
             "PCI drivers must implement both sriov_enable and sriov_disable"
@@ -116,7 +113,6 @@ unsafe fn register(
             if let Some(ops) = T::PM_OPS {
                 (*pdrv.get()).driver.pm = ops.raw;
             }
-            #[cfg(CONFIG_PCI_IOV)]
             if T::HAS_SRIOV_ENABLE {
                 (*pdrv.get()).sriov_configure = Some(Self::sriov_configure_callback);
             }
@@ -592,6 +588,8 @@ fn runtime_resume<'bound>(dev: &Device<device::Bound>, this: Pin<&Self::Data<'bo

     /// Enables Single Root I/O Virtualization (SR-IOV) for a PF.
     ///
+    /// This callback is only invoked when `CONFIG_PCI_IOV` is enabled.
+    ///
     /// Called when userspace writes a nonzero VF count to `sriov_numvfs`. The token carries the
     /// requested count and permission to enable VFs on this PF. Return the guard obtained from
     /// [`SriovEnable::enable()`] after completing setup; dropping it rolls back the enable.
@@ -624,7 +622,6 @@ fn runtime_resume<'bound>(dev: &Device<device::Bound>, this: Pin<&Self::Data<'bo
     ///     Ok(enabled)
     /// }
     /// ```
-    #[cfg(CONFIG_PCI_IOV)]
     fn sriov_enable<'a>(
         dev: &'a Device<device::Core<'_>>,
         data: Pin<&Self::Data<'a>>,
@@ -636,6 +633,8 @@ fn sriov_enable<'a>(

     /// Disables all VFs of a PF in response to a userspace request.
     ///
+    /// This callback is only invoked when `CONFIG_PCI_IOV` is enabled.
+    ///
     /// Called when userspace writes zero to `sriov_numvfs`. Call [`SriovDisable::disable()`]
     /// while resources used by VF drivers are still available. Returning an error before calling
     /// it leaves the VFs enabled. Returning success requires that all VFs have been disabled.
@@ -661,7 +660,6 @@ fn sriov_enable<'a>(
     ///     Ok(())
     /// }
     /// ```
-    #[cfg(CONFIG_PCI_IOV)]
     fn sriov_disable<'a>(
         dev: &'a Device<device::Core<'_>>,
         data: Pin<&Self::Data<'a>>,
diff --git a/rust/kernel/pci/iov.rs b/rust/kernel/pci/iov.rs
index c5b6fd31d1e4..abba5eeaa49b 100644
--- a/rust/kernel/pci/iov.rs
+++ b/rust/kernel/pci/iov.rs
@@ -23,12 +23,50 @@
 };
 use core::{
     any::TypeId,
+    ffi::c_void,
     marker::PhantomPinned,
     mem::ManuallyDrop, //
 };

+/// Returns the PF's registration pointer, or null when SR-IOV is disabled.
+///
+/// # Safety
+///
+/// `pdev` must point to a live PCI device. The caller must exclude concurrent changes to its
+/// registration pointer.
+unsafe fn vf_registration_data_ptr(pdev: *mut bindings::pci_dev) -> *mut c_void {
+    #[cfg(CONFIG_PCI_IOV)]
+    {
+        // SAFETY: The caller guarantees the device's validity and excludes concurrent changes.
+        unsafe { (*pdev).vf_registration_data_rust }
+    }
+    #[cfg(not(CONFIG_PCI_IOV))]
+    {
+        let _ = pdev;
+        core::ptr::null_mut()
+    }
+}
+
+/// Updates the PF's registration pointer, doing nothing when SR-IOV is disabled.
+///
+/// # Safety
+///
+/// `pdev` must point to a live PCI device. The caller must exclude concurrent access to its
+/// registration pointer and uphold the invariants of any published registration data.
+unsafe fn set_vf_registration_data_ptr(pdev: *mut bindings::pci_dev, data: *mut c_void) {
+    #[cfg(CONFIG_PCI_IOV)]
+    // SAFETY: The caller guarantees the device's validity and exclusive access to its slot.
+    unsafe {
+        (*pdev).vf_registration_data_rust = data;
+    }
+    #[cfg(not(CONFIG_PCI_IOV))]
+    let _ = (pdev, data);
+}
+
 impl Device<device::Core<'_>> {
     /// Returns the number of Virtual Functions (VF) enabled for a Physical Function (PF).
+    ///
+    /// Returns zero when `CONFIG_PCI_IOV` is disabled.
     #[inline]
     pub fn num_vf(&self) -> u16 {
         // SAFETY: `self.as_raw()` is valid and this call runs in the PCI core callback context.
@@ -110,7 +148,7 @@ pub fn enable(self, num_vfs: u16) -> Result<SriovEnabled<'a>> {

         // SAFETY: The PF's driver data and registration remain installed throughout this callback.
         // The registration owns final VF teardown after the enable guard is disarmed.
-        if unsafe { (*self.pdev.as_raw()).vf_registration_data_rust }.is_null() {
+        if unsafe { vf_registration_data_ptr(self.pdev.as_raw()) }.is_null() {
             return Err(ENODEV);
         }

@@ -225,6 +263,9 @@ fn new<E>(data: impl PinInit<F::Of<'a>, E>) -> impl PinInit<Self, E> {
 /// [`Device::vf_registration_data_with()`] and [`Device::vf_registration_data()`].
 ///
 /// Drop disables SR-IOV before clearing the registration pointer and releasing the data.
+///
+/// When `CONFIG_PCI_IOV` is disabled, the registration still initializes, pins and owns its data,
+/// but does not publish it to VF drivers.
 #[pin_data(PinnedDrop)]
 pub struct VfRegistration<'a, F: ForLt + 'static> {
     pdev: &'a Device<device::Bound>,
@@ -262,7 +303,7 @@ pub unsafe fn new<'core, I>(
                     return Err(EBUSY);
                 }
                 // SAFETY: The core callback serializes changes to this device's registration.
-                if !unsafe { (*pdev.as_raw()).vf_registration_data_rust }.is_null() {
+                if !unsafe { vf_registration_data_ptr(pdev.as_raw()) }.is_null() {
                     return Err(EBUSY);
                 }
             },
@@ -272,15 +313,17 @@ pub unsafe fn new<'core, I>(
             _: {
                 // The data initializer may have registered another object on this device.
                 // SAFETY: The captured core context still serializes registration changes.
-                if !unsafe { (*pdev.as_raw()).vf_registration_data_rust }.is_null() {
+                if !unsafe { vf_registration_data_ptr(pdev.as_raw()) }.is_null() {
                     return Err(EBUSY);
                 }

                 // SAFETY: The data is initialized and pinned, the slot is unoccupied, and
                 // no VF can access it yet. No fallible work follows publication.
                 unsafe {
-                    (*pdev.as_raw()).vf_registration_data_rust =
-                        core::ptr::from_ref(inner.as_ref().get_ref()).cast_mut().cast();
+                    set_vf_registration_data_ptr(
+                        pdev.as_raw(),
+                        core::ptr::from_ref(inner.as_ref().get_ref()).cast_mut().cast(),
+                    );
                 }
             },
         })
@@ -297,7 +340,7 @@ fn drop(self: Pin<&mut Self>) {

         // SAFETY: The device is valid and all VF remove callbacks have completed, so no VF
         // can access the registration pointer. The data remains alive until this returns.
-        unsafe { (*self.pdev.as_raw()).vf_registration_data_rust = core::ptr::null_mut() };
+        unsafe { set_vf_registration_data_ptr(self.pdev.as_raw(), core::ptr::null_mut()) };
     }
 }

@@ -310,6 +353,8 @@ unsafe impl<F: ForLt> Sync for VfRegistration<'_, F> where for<'a> F::Of<'a>: Se

 impl Device<device::Bound> {
     /// Return the zero-based VF index within its PF, or an error for a non-VF device.
+    ///
+    /// Returns [`ENOSYS`] when `CONFIG_PCI_IOV` is disabled.
     pub fn vf_id(&self) -> Result<u32> {
         // SAFETY: `self.as_raw()` points to a live PCI device; the helper checks VF membership.
         let id = unsafe { bindings::pci_iov_vf_id(self.as_raw()) };
@@ -324,6 +369,7 @@ pub fn vf_id(&self) -> Result<u32> {
     ///
     /// The data lifetime must be hidden behind a higher-ranked closure independently of the
     /// reference lifetime, or `F` must be covariant in its encoded lifetime.
+    #[cfg(CONFIG_PCI_IOV)]
     unsafe fn vf_registration_data_pinned<F: ForLt + 'static>(&self) -> Result<Pin<&F::Of<'_>>> {
         if !self.is_virtfn() {
             return Err(ENODEV);
@@ -333,7 +379,7 @@ unsafe fn vf_registration_data_pinned<F: ForLt + 'static>(&self) -> Result<Pin<&
         // VF removal completes, and the PF keeps the registration installed until then.
         let ptr = unsafe {
             let pf = (*self.as_raw()).__bindgen_anon_1.physfn;
-            (*pf).vf_registration_data_rust
+            vf_registration_data_ptr(pf)
         };

         if ptr.is_null() {
@@ -360,11 +406,23 @@ unsafe fn vf_registration_data_pinned<F: ForLt + 'static>(&self) -> Result<Pin<&
         Ok(unsafe { Pin::new_unchecked(&*data_ptr) })
     }

+    /// Returns [`ENODEV`] because no VF registration can be published.
+    ///
+    /// # Safety
+    ///
+    /// The data lifetime must be hidden behind a higher-ranked closure independently of the
+    /// reference lifetime, or `F` must be covariant in its encoded lifetime.
+    #[cfg(not(CONFIG_PCI_IOV))]
+    unsafe fn vf_registration_data_pinned<F: ForLt + 'static>(&self) -> Result<Pin<&F::Of<'_>>> {
+        Err(ENODEV)
+    }
+
     /// Access the VF registration data through a closure with an HRTB lifetime.
     ///
     /// `F` is the [`ForLt`](trait@ForLt) encoding of the data type. Returns
     /// [`ENODEV`] if this is not a VF or no data was registered,
     /// or [`EINVAL`] if `F` does not match the type registered by the PF.
+    /// Returns [`ENODEV`] when `CONFIG_PCI_IOV` is disabled.
     ///
     /// The reference is borrowed from this VF, while the registration data's lifetime remains
     /// abstract so the closure cannot store shorter-lived references in invariant data.

  reply	other threads:[~2026-11-10 15:50 UTC|newest]

Thread overview: 25+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2026-10-23 15:39 [PATCH v2 00/12] Rust VFIO PCI support and NVIDIA vGPU driver Simon Song
2026-10-23 15:39 ` [PATCH v2 01/12] rust: pci: add VFIO override device IDs Simon Song
2026-10-23 20:07   ` Gary Guo
2026-10-30 20:34     ` Simon Song
2026-10-23 15:39 ` [PATCH v2 02/12] rust: pci: expose the VF index within its PF Simon Song
2026-11-09 22:17   ` Danilo Krummrich
2026-10-23 15:39 ` [PATCH v2 03/12] rust: pci: allow drivers to manage DMA ownership Simon Song
2026-11-09 22:18   ` Danilo Krummrich
2026-10-23 15:39 ` [PATCH v2 04/12] rust: vfio: add PCI variant driver abstractions Simon Song
2026-10-23 15:40 ` [PATCH v2 05/12] rust: pci: add reset callbacks Simon Song
2026-11-09 23:19   ` Danilo Krummrich
2026-10-23 15:40 ` [PATCH v2 06/12] rust: pci: add a bind callback after installing driver data Simon Song
2026-11-10  0:10   ` Danilo Krummrich
2026-10-23 15:40 ` [PATCH v2 07/12] rust: pci: add an error_detected callback Simon Song
2026-11-10  0:47   ` Danilo Krummrich
2026-10-23 15:40 ` [PATCH v2 08/12] rust: pci: add runtime PM callbacks Simon Song
2026-11-10  0:23   ` Danilo Krummrich
2026-10-23 15:40 ` [PATCH v2 09/12] rust: vfio: use the PCI adapter for variant drivers Simon Song
2026-10-23 15:40 ` [PATCH v2 10/12] rust: vfio: separate common device handling from PCI Simon Song
2026-10-23 15:40 ` [PATCH v2 11/12] gpu: nova-core: publish typed SR-IOV PF APIs Simon Song
2026-11-10 15:50   ` Danilo Krummrich [this message]
2026-10-23 15:40 ` [PATCH v2 12/12] vfio/nvidia-vgpu: add the Rust VFIO variant driver Simon Song
2026-10-30 20:40 ` [PATCH v2 00/12] Rust VFIO PCI support and NVIDIA vGPU driver Simon Song
2026-10-30 20:49   ` Danilo Krummrich
2026-11-10  1:04 ` Danilo Krummrich

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=DMRNF7IZQ1YI.1MRPH47QJC9XC@kernel.org \
    --to=dakr@kernel.org \
    --cc=a.hindborg@kernel.org \
    --cc=acourbot@nvidia.com \
    --cc=airlied@gmail.com \
    --cc=alex.gaynor@gmail.com \
    --cc=alex@shazbot.org \
    --cc=aliceryhl@google.com \
    --cc=alkumar@nvidia.com \
    --cc=aniketa@nvidia.com \
    --cc=ankita@nvidia.com \
    --cc=bhelgaas@google.com \
    --cc=bjorn3_gh@protonmail.com \
    --cc=boqun.feng@gmail.com \
    --cc=boqun@kernel.org \
    --cc=cjia@nvidia.com \
    --cc=daniel.almeida@collabora.com \
    --cc=dri-devel@lists.freedesktop.org \
    --cc=ecourtney@nvidia.com \
    --cc=gary@garyguo.net \
    --cc=jgg@nvidia.com \
    --cc=jgg@ziepe.ca \
    --cc=jhubbard@nvidia.com \
    --cc=kevin.tian@intel.com \
    --cc=kjaju@nvidia.com \
    --cc=kvm@vger.kernel.org \
    --cc=kwankhede@nvidia.com \
    --cc=kwilczynski@kernel.org \
    --cc=linux-kernel@vger.kernel.org \
    --cc=linux-pci@vger.kernel.org \
    --cc=lossin@kernel.org \
    --cc=nova-gpu@lists.linux.dev \
    --cc=ojeda@kernel.org \
    --cc=peter@colberg.org \
    --cc=rust-for-linux@vger.kernel.org \
    --cc=simona@ffwll.ch \
    --cc=skolothumtho@nvidia.com \
    --cc=smitra@nvidia.com \
    --cc=tamird@kernel.org \
    --cc=targupta@nvidia.com \
    --cc=tmgross@umich.edu \
    --cc=work@onurozkan.dev \
    --cc=xinmengs@nvidia.com \
    --cc=yishaih@nvidia.com \
    --cc=zhiw@nvidia.com \
    --cc=zhiwang@kernel.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

all inboxes | Powered by JetHome®