At the moment - the memory allocation for fwsec-sb is created as-needed and
is released after being used. Typically this is at some point well after
driver load, which can cause runtime suspend/resume to initially work on
driver load but then later fail on a machine that has been running for long
enough with sufficiently high enough memory pressure:
kworker/7:1: page allocation failure: order:5, mode:0xcc0(GFP_KERNEL),
nodemask=(null),cpuset=/,mems_allowed=0
CPU: 7 UID: 0 PID: 875159 Comm: kworker/7:1 Not tainted
6.17.8-300.fc43.x86_64 #1 PREEMPT(lazy)
Hardware name: SLIMBOOK Executive/Executive, BIOS N.1.10GRU06 02/02/2024
Workqueue: pm pm_runtime_work
Call Trace:
<TASK>
dump_stack_lvl+0x5d/0x80
warn_alloc+0x163/0x190
? __alloc_pages_direct_compact+0x1b3/0x220
__alloc_pages_slowpath.constprop.0+0x57a/0xb10
__alloc_frozen_pages_noprof+0x334/0x350
__alloc_pages_noprof+0xe/0x20
__dma_direct_alloc_pages.isra.0+0x1eb/0x330
dma_direct_alloc_pages+0x3c/0x190
dma_alloc_pages+0x29/0x130
nvkm_firmware_ctor+0x1ae/0x280 [nouveau]
nvkm_falcon_fw_ctor+0x3e/0x60 [nouveau]
nvkm_gsp_fwsec+0x10e/0x2c0 [nouveau]
? sysvec_apic_timer_interrupt+0xe/0x90
nvkm_gsp_fwsec_sb+0x27/0x70 [nouveau]
tu102_gsp_fini+0x65/0x110 [nouveau]
? ktime_get+0x3c/0xf0
nvkm_subdev_fini+0x67/0xc0 [nouveau]
nvkm_device_fini+0x94/0x140 [nouveau]
nvkm_udevice_fini+0x50/0x70 [nouveau]
nvkm_object_fini+0xb1/0x140 [nouveau]
nvkm_object_fini+0x70/0x140 [nouveau]
? __pfx_pci_pm_runtime_suspend+0x10/0x10
nouveau_do_suspend+0xe4/0x170 [nouveau]
nouveau_pmops_runtime_suspend+0x3e/0xb0 [nouveau]
pci_pm_runtime_suspend+0x67/0x1a0
? __pfx_pci_pm_runtime_suspend+0x10/0x10
__rpm_callback+0x45/0x1f0
? __pfx_pci_pm_runtime_suspend+0x10/0x10
rpm_callback+0x6d/0x80
rpm_suspend+0xe5/0x5e0
? finish_task_switch.isra.0+0x99/0x2c0
pm_runtime_work+0x98/0xb0
process_one_work+0x18f/0x350
worker_thread+0x25a/0x3a0
? __pfx_worker_thread+0x10/0x10
kthread+0xf9/0x240
? __pfx_kthread+0x10/0x10
? __pfx_kthread+0x10/0x10
ret_from_fork+0xf1/0x110
? __pfx_kthread+0x10/0x10
ret_from_fork_asm+0x1a/0x30
</TASK>
The reason this happens is because the fwsec-sb firmware image only
supports being booted from a contiguous coherent sysmem allocation. If a
system runs into enough memory fragmentation from memory pressure, such as
what can happen on systems with low amounts of memory, this can lead to a
situation where it later becomes impossible to find space for a large
enough contiguous allocation to hold fwsec-sb. This causes us to fail to
boot the firmware image, causing the GPU to fail booting and causing the
driver to fail.
Since this firmware can't use non-contiguous allocations, the best solution
to avoid this issue is to simply allocate the memory for fwsec-sb during
initial driver-load, and reuse the memory allocation when fwsec-sb needs to
be used. We then release the memory allocations on driver unload.
Signed-off-by: Lyude Paul <lyude@redhat.com>
Fixes: 594766ca3e ("drm/nouveau/gsp: move booter handling to GPU-specific code")
Cc: <stable@vger.kernel.org> # v6.16+
Reviewed-by: Timur Tabi <ttabi@nvidia.com>
Link: https://patch.msgid.link/20251202175918.63533-1-lyude@redhat.com
This commit enables basic support for the GB100/GB102 Blackwell GPUs.
Beyond HW class ID plumbing there's very little change here vs GH100.
Signed-off-by: Ben Skeggs <bskeggs@nvidia.com>
Reviewed-by: Dave Airlie <airlied@redhat.com>
Reviewed-by: Timur Tabi <ttabi@nvidia.com>
Tested-by: Timur Tabi <ttabi@nvidia.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
This commit enables basic support for Hopper GPUs, and is intended
primarily as a base supporting Blackwell GPUs, which reuse most of
the code added here.
Advanced features such as Confidential Compute are not supported.
Beyond a few miscellaneous register moves and HW class ID plumbing,
the bulk of the changes implemented here are to support the GSP-RM
boot sequence used on Hopper/Blackwell GPUs, as well as a new page
table layout.
There should be no changes here that impact prior GPUs.
Signed-off-by: Ben Skeggs <bskeggs@nvidia.com>
Co-developed-by: Timur Tabi <ttabi@nvidia.com>
Signed-off-by: Timur Tabi <ttabi@nvidia.com>
Reviewed-by: Dave Airlie <airlied@redhat.com>
Reviewed-by: Timur Tabi <ttabi@nvidia.com>
Tested-by: Timur Tabi <ttabi@nvidia.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
With GSP-RM handling the majority of the HW programming, NVKM's usual
HALs are more elaborate than necessary, resulting in a fair amount of
duplicated boilerplate.
Adds 'nvkm_rm_gpu' which serves to provide GPU-specific constants and
functions in a more streamlined manner.
This is initially used in subsequent commits to store engine class IDs,
and replace the per-engine/engobj boilerplate with common code for all
GSP-RM supported engines - and is further extended when adding GH100,
GB10x and GB20x support.
Signed-off-by: Ben Skeggs <bskeggs@nvidia.com>
Reviewed-by: Dave Airlie <airlied@redhat.com>
Reviewed-by: Timur Tabi <ttabi@nvidia.com>
Tested-by: Timur Tabi <ttabi@nvidia.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
This commit adds the initial code needed to boot the GSP-RM firmware
provided by NVIDIA, bringing with it the beginnings of Ada support.
Until it's had more testing and time to bake, support is disabled by
default (except on Ada). GSP-RM usage can be enabled by passing the
"config=NvGspRm=1" module option.
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20230918202149.4343-33-skeggsb@gmail.com
This will allow us to register the falcon with ACR, and further customise
its behaviour by providing the nvkm_falcon_func structure directly.
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
