From: Alok Kumar <alkumar@nvidia.com>
To: <dakr@kernel.org>, <acourbot@nvidia.com>, <kvm@vger.kernel.org>,
<rust-for-linux@vger.kernel.org>, <nova-gpu@lists.linux.dev>,
<linux-pci@vger.kernel.org>
Cc: <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>,
<markus.probst@posteo.de>, <jhubbard@nvidia.com>,
<apopple@nvidia.com>, <joelagnelf@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>, <xinmengs@nvidia.com>
Subject: [PATCH v4 06/32] gpu: nova-core: gsp: decode the FIFO engine table
Date: Fri, 6 Nov 2026 11:09:39 +0530 [thread overview]
Message-ID: <340c9bb08c456c19d8768f19d4dc0bc4b3bf34b1.1793906667.git.alkumar@nvidia.com> (raw)
In-Reply-To: <cover.1793906667.git.alkumar@nvidia.com>
GSP_INIT reports the FIFO engine topology used to build the channel map
for vGPU instances.
Decode the count, GMC IDs and flags into GspStaticInfo and expose
host-driven engines in firmware FIFO order, retaining repeated IDs.
Reject counts and indexed entries beyond the driver's 64-entry table
capacity rather than silently accepting a prefix. An omitted count
remains zero.
Cc: Alexandre Courbot <acourbot@nvidia.com>
Co-developed-by: Zhi Wang <zhiw@nvidia.com>
Signed-off-by: Zhi Wang <zhiw@nvidia.com>
Signed-off-by: Alok Kumar <alkumar@nvidia.com>
---
drivers/gpu/nova-core/gsp/commands.rs | 10 +++
drivers/gpu/nova-core/gsp/commands/gmc.rs | 4 +-
drivers/gpu/nova-core/gsp/fw/commands.rs | 77 +++++++++++++++++++++++
3 files changed, 90 insertions(+), 1 deletion(-)
diff --git a/drivers/gpu/nova-core/gsp/commands.rs b/drivers/gpu/nova-core/gsp/commands.rs
index 8b047744c36d..f5ec27995627 100644
--- a/drivers/gpu/nova-core/gsp/commands.rs
+++ b/drivers/gpu/nova-core/gsp/commands.rs
@@ -11,6 +11,8 @@
use kernel::prelude::*;
+use super::fw::commands::FifoEngineList;
+
/// The static GPU configuration, which GSP-RM reports in reply to [`GspInit`](gmc::GspInit).
pub struct GspStaticInfo {
gpu_name: [u8; Self::MAX_GPU_NAME_LEN],
@@ -23,6 +25,8 @@ pub struct GspStaticInfo {
pub(crate) usable_fb_regions: KVec<Range<u64>>,
/// Exclusive end of the FB physical address space.
pub(crate) total_fb_end: u64,
+ /// Host-driven engines in firmware FIFO order.
+ fifo_engine_list: FifoEngineList,
}
/// Error type for [`GspStaticInfo::gpu_name`].
@@ -63,6 +67,12 @@ pub fn gpu_short_name(&self) -> core::result::Result<&str, GpuNameError> {
.map_err(GpuNameError::InvalidUtf8)
}
+ /// Returns the host-driven engines in their firmware FIFO order.
+ #[expect(dead_code)]
+ pub(crate) fn fifo_engine_list(&self) -> FifoEngineList {
+ self.fifo_engine_list
+ }
+
/// Returns the total usable VRAM size in bytes, i.e. the summed lengths of all usable FB
/// regions.
pub fn vram_size(&self) -> u64 {
diff --git a/drivers/gpu/nova-core/gsp/commands/gmc.rs b/drivers/gpu/nova-core/gsp/commands/gmc.rs
index 890462beb11f..c9258ee9e059 100644
--- a/drivers/gpu/nova-core/gsp/commands/gmc.rs
+++ b/drivers/gpu/nova-core/gsp/commands/gmc.rs
@@ -145,7 +145,8 @@ impl MessageFromGsp for GspStaticInfo {
/// # Errors
///
/// - `EINVAL` if the payload is not a whole number of NVKV words, if the stream is malformed
- /// or omits a required key, or if GSP-RM reported no framebuffer region.
+ /// or omits a required key, if GSP-RM reported no framebuffer region, or if the FIFO engine
+ /// count or indexed entries exceed the supported table capacity.
/// - `ENOMEM` if the words or the decoded regions cannot be allocated.
fn read(payload: &mut SBufferIter<core::array::IntoIter<&[u8], 2>>) -> Result<Self> {
let bytes = payload.flush_into_kvec(GFP_KERNEL)?;
@@ -184,6 +185,7 @@ fn read(payload: &mut SBufferIter<core::array::IntoIter<&[u8], 2>>) -> Result<Se
bar1_pde_base: reply.bar1_pde_base(),
usable_fb_regions,
total_fb_end: reply.total_fb_end().ok_or(EINVAL)?,
+ fifo_engine_list: reply.fifo_engine_list(),
})
}
}
diff --git a/drivers/gpu/nova-core/gsp/fw/commands.rs b/drivers/gpu/nova-core/gsp/fw/commands.rs
index bda6a7ef5ef7..07115ccf3e95 100644
--- a/drivers/gpu/nova-core/gsp/fw/commands.rs
+++ b/drivers/gpu/nova-core/gsp/fw/commands.rs
@@ -21,8 +21,10 @@
nvkv_decode,
nvkv_encode,
Accumulated,
+ DecoderValue,
Encodable,
Encoder,
+ Indexed,
Key,
Required,
TruncatingArray, //
@@ -307,6 +309,52 @@ pub(crate) fn new(ctx: &GspBootContext<'_, '_>) -> Result<Self> {
// Decode:
+const MAX_FIFO_ENGINES: usize = 64;
+
+/// Bit mask for `NVGMC_SC_ENGINE_FLAGS_IS_HOST_DRIVEN`.
+const ENGINE_FLAGS_IS_HOST_DRIVEN: u32 = 1 << 0;
+
+/// Host-driven GMC engine IDs in hardware FIFO order, including any repeated IDs.
+///
+/// # Invariants
+///
+/// `count` is at most [`MAX_FIFO_ENGINES`]. The first `count` slots are the retained engine IDs.
+#[derive(Copy, Clone)]
+pub(crate) struct FifoEngineList {
+ gmc_ids: [u32; MAX_FIFO_ENGINES],
+ count: usize,
+}
+
+impl FifoEngineList {
+ #[expect(dead_code)]
+ pub(crate) fn gmc_ids(&self) -> &[u32] {
+ // PANIC: The type invariant bounds `count` by the array capacity.
+ &self.gmc_ids[..self.count]
+ }
+}
+
+/// A FIFO engine count that fits the supported tables.
+///
+/// # Invariants
+///
+/// The count is at most [`MAX_FIFO_ENGINES`].
+#[derive(Default)]
+struct FifoEngineCount(usize);
+
+impl TryFrom<DecoderValue<'_>> for FifoEngineCount {
+ type Error = Error;
+
+ fn try_from(value: DecoderValue<'_>) -> Result<Self> {
+ let count = crate::num::u32_as_usize(u32::try_from(value)?);
+ if count > MAX_FIFO_ENGINES {
+ Err(EINVAL)
+ } else {
+ // INVARIANT: The count was checked against the table capacity above.
+ Ok(Self(count))
+ }
+ }
+}
+
// Should decode with UnknownKeyPolicy::Ignore.
nvkv_decode! {
/// Schema for the `GSP_INIT` response.
@@ -317,6 +365,9 @@ pub(crate) struct GspInitResponseSchema => GspInitResponse {
fb_regions: Accumulated<FbRegionSchema>,
bar1_pde_base: Required<Key<u64, 0x1020>>,
vmmu_segment_size: Key<u64, 0x1050>,
+ fifo_engine_count: Key<FifoEngineCount, 0x0500>,
+ fifo_engine_gmc_ids: Indexed<u32, MAX_FIFO_ENGINES, 0x0501>,
+ fifo_engine_flags: Indexed<u32, MAX_FIFO_ENGINES, 0x0502>,
}
}
@@ -328,6 +379,9 @@ pub(crate) struct GspInitResponse {
fb_regions: KVVec<FbRegion>,
bar1_pde_base: u64,
vmmu_segment_size: u64,
+ fifo_engine_count: FifoEngineCount,
+ fifo_engine_gmc_ids: [u32; MAX_FIFO_ENGINES],
+ fifo_engine_flags: [u32; MAX_FIFO_ENGINES],
}
impl GspInitResponse {
@@ -381,6 +435,29 @@ pub(crate) fn total_fb_end(&self) -> Option<u64> {
.checked_add(1)
}
+ /// Returns the host-driven engines in their firmware FIFO order.
+ pub(crate) fn fifo_engine_list(&self) -> FifoEngineList {
+ // INVARIANT: The list starts empty and appends at most one ID per supported input slot.
+ let mut fifo_engine_list = FifoEngineList {
+ gmc_ids: [0; MAX_FIFO_ENGINES],
+ count: 0,
+ };
+ for (&gmc_id, &flags) in self
+ .fifo_engine_gmc_ids
+ .iter()
+ .zip(&self.fifo_engine_flags)
+ .take(self.fifo_engine_count.0)
+ {
+ if flags & ENGINE_FLAGS_IS_HOST_DRIVEN != 0 {
+ // PANIC: At most one slot is filled per input, and the input has at most
+ // MAX_FIFO_ENGINES entries, so the next retained ID always fits.
+ fifo_engine_list.gmc_ids[fifo_engine_list.count] = gmc_id;
+ fifo_engine_list.count += 1;
+ }
+ }
+ fifo_engine_list
+ }
+
/// Returns the BAR1 page directory entry base address.
pub(crate) fn bar1_pde_base(&self) -> u64 {
self.bar1_pde_base
--
2.53.0
next prev parent reply other threads:[~2026-11-06 5:42 UTC|newest]
Thread overview: 33+ messages / expand[flat|nested] mbox.gz Atom feed top
2026-11-06 5:39 [PATCH v4 00/32] Introduce NVIDIA vGPU manager Alok Kumar
2026-11-06 5:39 ` [PATCH v4 01/32] gpu: nova-core: mm: retain BAR1 reservations after unmap failure Alok Kumar
2026-11-06 5:39 ` [PATCH v4 02/32] gpu: nova-core: gsp: pass boot context through setup helpers Alok Kumar
2026-11-06 5:39 ` [PATCH v4 03/32] gpu: nova-core: gsp: decouple boot context from VgpuManager Alok Kumar
2026-11-06 5:39 ` [PATCH v4 04/32] gpu: nova-core: vgpu: detect boot state independently Alok Kumar
2026-11-06 5:39 ` [PATCH v4 05/32] gpu: nova-core: gpu: add a channel ID pool for vGPU Alok Kumar
2026-11-06 5:39 ` Alok Kumar [this message]
2026-11-06 5:39 ` [PATCH v4 07/32] gpu: nova-core: vgpu: initialize runtime parameters after GSP boot Alok Kumar
2026-11-06 5:39 ` [PATCH v4 08/32] gpu: nova-core: vgpu: reserve the 48-VM WPR2 heap Alok Kumar
2026-11-06 5:39 ` [PATCH v4 09/32] gpu: nova-core: mm: borrow BarUser for temporary BAR1 access Alok Kumar
2026-11-06 5:39 ` [PATCH v4 10/32] gpu: nova-core: mm: add VramBlock Alok Kumar
2026-11-06 5:39 ` [PATCH v4 11/32] gpu: nova-core: mm: add VramRegion Alok Kumar
2026-11-06 5:39 ` [PATCH v4 12/32] gpu: nova-core: mm: add BarMapping Alok Kumar
2026-11-06 5:39 ` [PATCH v4 13/32] gpu: nova-core: gsp: add synchronous GMC transactions Alok Kumar
2026-11-06 5:39 ` [PATCH v4 14/32] gpu: nova-core: gsp: wait for GMC completion events Alok Kumar
2026-11-06 5:39 ` [PATCH v4 15/32] gpu: nova-core: vgpu: add r000 plugin bindings Alok Kumar
2026-11-06 5:39 ` [PATCH v4 16/32] gpu: nova-core: vgpu: add VRAM slot allocator Alok Kumar
2026-11-06 5:39 ` [PATCH v4 17/32] gpu: nova-core: gsp: factor out NVKV payload conversion Alok Kumar
2026-11-06 5:39 ` [PATCH v4 18/32] gpu: nova-core: vgpu: query VF assignments and properties Alok Kumar
2026-11-06 5:39 ` [PATCH v4 19/32] gpu: nova-core: vgpu: add instance create/destroy Alok Kumar
2026-11-06 5:39 ` [PATCH v4 20/32] gpu: nova-core: vgpu: encode vGPU boot requests Alok Kumar
2026-11-06 5:39 ` [PATCH v4 21/32] gpu: nova-core: vgpu: add GSP plugin communication buffers Alok Kumar
2026-11-06 5:39 ` [PATCH v4 22/32] gpu: nova-core: vgpu: add instance boot Alok Kumar
2026-11-06 5:39 ` [PATCH v4 23/32] gpu: nova-core: vgpu: add instance shutdown Alok Kumar
2026-11-06 5:39 ` [PATCH v4 24/32] gpu: nova-core: vgpu: initialize GSP plugin RPC buffers Alok Kumar
2026-11-06 5:39 ` [PATCH v4 25/32] gpu: nova-core: vgpu: add GSP plugin RPC transactions Alok Kumar
2026-11-06 5:39 ` [PATCH v4 26/32] gpu: nova-core: vgpu: negotiate the GSP plugin RPC version Alok Kumar
2026-11-06 5:40 ` [PATCH v4 27/32] gpu: nova-core: vgpu: send GSP plugin configuration parameters Alok Kumar
2026-11-06 5:40 ` [PATCH v4 28/32] gpu: nova-core: vgpu: update the GSP plugin BME state Alok Kumar
2026-11-06 5:40 ` [PATCH v4 29/32] gpu: nova-core: vgpu: add CeUtils commands Alok Kumar
2026-11-06 5:40 ` [PATCH v4 30/32] gpu: nova-core: vgpu: scrub guest VRAM with CeUtils Alok Kumar
2026-11-06 5:40 ` [PATCH v4 31/32] gpu: nova-core: vgpu: export plugin log buffers via debugfs Alok Kumar
2026-11-06 5:40 ` [PATCH v4 32/32] gpu: nova-core: vgpu: introduce SR-IOV PF APIs Alok Kumar
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=340c9bb08c456c19d8768f19d4dc0bc4b3bf34b1.1793906667.git.alkumar@nvidia.com \
--to=alkumar@nvidia.com \
--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=aniketa@nvidia.com \
--cc=ankita@nvidia.com \
--cc=apopple@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=dakr@kernel.org \
--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=joelagnelf@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=markus.probst@posteo.de \
--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®