- replaces the hacked-up version that existed solely to support TTM
v2. remove earlier hack preventing use of non-stall intr for fences
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
Exposes a bunch of the new features that became possible as a result
of the earlier commits. DRM will build on this in the future to add
support for features such as SCG ("async compute") and multi-device
rendering, as part of the work necessary to be able to write a half-
decent vulkan driver - finally.
For the moment, this just crudely ports DRM to the API changes.
- channel class interfaces now the same for all HW classes
- channel group class exposed (SCG)
- channel runqueue selector exposed (SCG)
- channel sub-device id control exposed (multi-device rendering)
- channel names in logging will reflect creating process, not fd owner
- explicit USERD allocation required by VOLTA_CHANNEL_GPFIFO_A and newer
- drm is smarter about determining the appropriate channel class to use
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
Reviewed-by: Lyude Paul <lyude@redhat.com>
Builds on the context tracking that was added earlier.
- marks engine context PTEs as 'priv' where possible
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
Reviewed-by: Lyude Paul <lyude@redhat.com>
And use it to cleanup multiple implementations of almost the same thing.
- prepares for non-polled / client-provided USERD
- only zeroes relevant "registers", rather than entire USERD
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
Reviewed-by: Lyude Paul <lyude@redhat.com>
- less dependence on waiting for runlist updates, on GPUs that allow it
- supports runqueue selector in RAMRL entries
- completes switch to common runl/cgrp/chan topology info
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
Reviewed-by: Lyude Paul <lyude@redhat.com>
That sure was fun to untangle.
- handled per-runlist, rather than globally
- more straight-forward process in general
- various potential SW/HW races have been fixed
- fixes lockdep issues that were present in >=gk104's prior implementation
- volta recovery now actually stands a chance of working
- volta/turing waiting for PBDMA idle before engine reset
- turing using hw-provided TSG info for CTXSW_TIMEOUT
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
Reviewed-by: Lyude Paul <lyude@redhat.com>
- nvkm_chan_error() built on top, stops channel and sends 'killed' event
- removes an odd double-bashing of channel enable regs on kepler and up
- pokes doorbell on turing and up, after enabling channel
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
Reviewed-by: Lyude Paul <lyude@redhat.com>
- adds g8x/turing registers, which were missing before
- switches fermi to polled wait, like later hw (see: 4f2fc25c0f8bc...)
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
Reviewed-by: Lyude Paul <lyude@redhat.com>
- supports per-runlist CHIDs
- channel group lock held across reference, rather than global lock
v2:
- remove unnecessary parenthesis
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
Reviewed-by: Lyude Paul <lyude@redhat.com>
- bumps pbdma timeout to value RM uses on newer HW
- bumps fb timeout to max from boot default
- one/both of these greatly improves stability on // piglit runs
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
Reviewed-by: Lyude Paul <lyude@redhat.com>
- removes a layer of indirection in the intr handling
- prevents non-stall ctrl racing with unknown intrs
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
Reviewed-by: Lyude Paul <lyude@redhat.com>
Engine context tracking will move to nvkm_cgrp in later commits, so we
create SW-only channel groups on HW without support for them.
- switches to nvkm_chid for TSG/channel ID allocation
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
Reviewed-by: Lyude Paul <lyude@redhat.com>
DRM uses this to setup fence-related items.
- nouveau_chan.runlist will always be "0" for the moment, not an issue
as GPUs prior to ampere have system-wide channel IDs,
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
Reviewed-by: Lyude Paul <lyude@redhat.com>
Previously only available from Kepler onwards.
- also fixes the info() queries causing fifo init()/fini() unnecessarily
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
Reviewed-by: Lyude Paul <lyude@redhat.com>
Creates an nvkm_runl for each runlist on the GPU, and an nvkm_engn for
each engine that is reachable from a runlist.
- basically what gk104- already does, but extended to all chips
- adds per-runlist CHID allocators (Ampere)
- splits g98/gt2xx out from g84 (different target engines)
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
Reviewed-by: Lyude Paul <lyude@redhat.com>
Creates an nvkm_runq for each PBDMA, these will be associated with the
relevant runlist(s) later.
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
Reviewed-by: Lyude Paul <lyude@redhat.com>
We need to be able to allocate TSG IDs as well as channel IDs, also,
Ampere has per-runlist channel IDs.
- holds per-ID private data, which will be used for/to protect lookup
- holds an nvkm_event which will be used for events tied to IDs
- not used yet beyond setup, and switching use of "fifo->nr - 1" for
channel ID mask to "chid->mask"
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
Reviewed-by: Lyude Paul <lyude@redhat.com>
This makes it easier to transition everything.
- a couple of function renames for collisions
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
Reviewed-by: Lyude Paul <lyude@redhat.com>
- will make subsequent patches more obvious
- no code changes
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
Reviewed-by: Lyude Paul <lyude@redhat.com>
Adds the basic skeleton for common channel (group) interfaces.
- common behaviour between <gk104 and >=gk104 impl's
- separates priv/user channel objects
- passthrough to existing object for now, kludges removed later
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
Reviewed-by: Lyude Paul <lyude@redhat.com>
- reads channel count from GPU from gm200 onwards
- removes gm20b/gp10b (they become identical to gm200/gp100)
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
Reviewed-by: Lyude Paul <lyude@redhat.com>
This replaces the twisty, confusing, relationship between nvkm_event and
nvkm_notify with something much simpler, and less racey. It also places
events in the object tree hierarchy, which will allow a heap of the code
tracking events across allocation/teardown/suspend to be removed.
This commit just adds the new interfaces, and passes the owning subdev to
the event constructor to enable debug-tracing in the new code.
v2:
- use ?: (lyude)
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
Reviewed-by: Lyude Paul <lyude@redhat.com>
RM appears to do this really early in its initialisation, before DEVINIT.
We currently do this before BAR2 initialisation for some reason.
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
There are instances (such as non-recoverable GPU page faults) where
NVKM decides that a channel's context is no longer viable, and will
be removed from the runlist.
This commit notifies the owner of the channel when this happens, so
it has the opportunity to take some kind of recovery action instead
of hanging.
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
