This reverts commit dfadbbdb57.
Further upstream discussion between Marek and Thomas decided this wasn't
fully baked and needed further work, so revert it before it hits mainline.
Signed-off-by: Dave Airlie <airlied@redhat.com>
* 'drm-nouveau-next' of git://git.freedesktop.org/git/nouveau/linux-2.6: (353 commits)
drm/nouveau: remove allocations from gart populate() hook
drm/nvc0/fb: slightly improve PMFB intr handling, move out of nvc0_graph.c
drm/nvc0/fifo: avoid touching missing subfifos
drm/nvd9/disp: bail out of mode_set_base if no fb bound to crtc
drm/nvd9/disp: stub some more api hooks so we don't oops on resume
drm/nouveau: fix printk typo in ioremap failure path
drm/nvc0/pm: minor clock readback fixes
drm/nv40/pm: execute memory reset script from vbios
drm/nv50/gr: refactor initialisation
drm/nouveau: if requested, try harder at disabling sysmem pushbufs
drm/nv50/gr: enable ctxprog xfer only when we need it to save power
drm/nouveau/dp: add support for displayport table 0x30
drm/nouveau/dp: return master dp table pointer too when looking up encoder
drm/nouveau/bios: simplify U/d table hash matching func to just match
drm/nouveau/dp: preserve non-pattern bits in DP_TRAINING_PATTERN_SET
drm/nvc0/gr: remove MODULE_FIRMWARE() lines
drm/nouveau/dp: use alternate lane mask for nvaf
drm/nouveau/dp: link rate scripts are selected with a comparison table
drm/nv40/pm: write nv40-specific reclocking routines
drm/nv40/pm: parse geometric delta clock from vbios
...
Since some somewhat questionable changes a while back, TTM provides a
completely empty array of struct dma_address that stays around for the
entire lifetime of the TTM object.
Lets use this array, *always*, rather than wasting yet more memory on
another array who's purpose is identical, as well as yet another bool array
of the same size saying *which* of the previous two arrays to use...
This change will also solve the high order allocation failures seen by
some people while using nouveau.
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
This patch adds instructions to ctxprog and by doing, impacts context
switching performance. My testcase showed a 1% performance cost using
glxgears that is a context-switch bound application.
Please test and report bugs/performance/power/other.
Many thanks to Maxim Levitsky for his dedicated work on lowering power
consumption with nouveau.
More patches are coming thanks to his work:
https://bugs.freedesktop.org/show_bug.cgi?id=37922
Signed-off-by: Martin Peres <martin.peres@ensi-bourges.fr>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
Written from observations of my NVD9's vbios, completely untested due to
my NVD9 lacking actual DisplayPort connectors..
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
Will need to be able to distinguish 2.0/2.1 from 3.0 soon. Also, move
the vbios parsing to nouveau_dp where it belongs.
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
We don't use these by default anymore, and there's been complaints from a
number of places thinking that the firmware blobs are required still.
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
Not 100% perfect yet, but a good start towards what it'll look like in the
end.
Actually seems stable on a NV44 I have here, as much as running around OA
for a fair amount of time constantly switching between performance levels
can prove..
My NV49 isn't quite so happy, and semaphores mess up somehow (sometimes) as
a result of the memory reclocking.
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
This changes the meaning of what we reported as "core" clock previously.
The shader/rop units are allegedly supposed to be run at the base clock
listed in the perf table, while the geometric clock can be bumped from
this value on some boards.
So that we can report both, we'll report the base clock as "shader" (since
the shaders *do* run at it), and the geometric clock as "core".
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
Not used currently, but it will be used in preference to pre-determined
lane/bandwidth numbers at a later point.
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
I'm sure that out there somewhere, someone will need this. We currently
haven't seen an example of LVDS being on a non-0 SOR so far though.
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
The HW will only accept the DMA_FROM_MEMORY class for DMA_SEMAPHORE without
asking the driver to intervene.
It appears that semaphores will work correctly even without DMA_IN_MEMORY,
so lets avoid the large amount of interrupts generated by x-chan sync.
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
NV30: Create framework for memtm
NV50: Improve reg creation,
NV50: Use P.version instead of card codename/stepping,
NVC0: Initial memtiming code for Fermi,
Renamed regs for consistency,
Overall redesign to improve readability,
Avoid kfree on null-pointer
Signed-off-by: Roy Spliet <r.spliet@student.tudelft.nl>
If we support PGPIO interrupts, and know a hotplug GPIO tag for a
connector we use HPD, otherwise POLL_CONNECT.
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
The nouveau hwmon temperature support currently only functions when hwmon is
compiled into the kernel. There's no reason why this shouldn't also work when
both hwmon and nouveau are modularised (as is the case with Slackware's stock
kernels).
Signed-off-by: Ken Milmore <ken.milmore@googlemail.com>
Reviewed-by: Martin Peres <martin.peres@ensi-bourges.fr>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
Takes a gamble and presumes that we can safely store something random in
OR_MODE_CTRL+4, the hw doesn't seem to mind...
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
VBIOS does more than this, as does nv50/nvc0 driver in nouveau. Traces
of the NVIDIA binary driver however, show pretty much just this being
done... Seems to work for me, it'll be fine for the moment.
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
All the non-stubbed functions should be okay for this chipset, the rest
will be added back as they're figured out.
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
We previously assumed (incorrectly a lot of the time) that PTIMER would
be programmed at a frequency which'd give its 64-bit timestamps in
nanoseconds.
By programming PTIMER ourselves, we avoid this problem.
Reviewed-by: Martin Peres <martin.peres@ensi-bourges.fr>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
The set will be replaced with a wait on the same flag by a subsequent
commit in order to halt a ctxprog's execution temporarily.
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
This is probably better than having to tell the common code about all the
clocks that exist on every chipset.
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
This commit resolves a possible 'NULL pointer dereference'
It uses the same approach as radeon, intel and nouveau/nv50
Fixes bug 'Nouveau: Kernel oops when unplugging external monitor'
https://bugs.freedesktop.org/show_bug.cgi?id=40336
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
nv04_sgdma_bind binds the same page multiple times on
architectures where PAGE_SIZE != 4096.
Let's fix it.
Signed-off-by: Marcin Slusarz <marcin.slusarz@gmail.com>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
Not cleaning after alloc failure would result in crash on destroy,
because nouveau_sgdma_clear assumes "ttm_alloced" to be not null when
"pages" is not null.
Signed-off-by: Marcin Slusarz <marcin.slusarz@gmail.com>
Cc: stable@kernel.org
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
Sometimes we want to know whether a buffer is busy and wait for it (bo_wait).
However, sometimes it would be more useful to be able to query whether
a buffer is busy and being either read or written, and wait until it's stopped
being either read or written. The point of this is to be able to avoid
unnecessary waiting, e.g. if a GPU has written something to a buffer and is now
reading that buffer, and a CPU wants to map that buffer for read, it needs to
only wait for the last write. If there were no write, there wouldn't be any
waiting needed.
This, or course, requires user space drivers to send read/write flags
with each relocation (like we have read/write domains in radeon, so we can
actually use those for something useful now).
Now how this patch works:
The read/write flags should passed to ttm_validate_buffer. TTM maintains
separate sync objects of the last read and write for each buffer, in addition
to the sync object of the last use of a buffer. ttm_bo_wait then operates
with one the sync objects.
Signed-off-by: Marek Olšák <maraeo@gmail.com>
Reviewed-by: Jerome Glisse <jglisse@redhat.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
* 'drm-core-next' of git://git.kernel.org/pub/scm/linux/kernel/git/airlied/drm-2.6: (135 commits)
drm/radeon/kms: fix DP training for DPEncoderService revision bigger than 1.1
drm/radeon/kms: add missing vddci setting on NI+
drm/radeon: Add a rmb() in IH processing
drm/radeon: ATOM Endian fix for atombios_crtc_program_pll()
drm/radeon: Fix the definition of RADEON_BUF_SWAP_32BIT
drm/radeon: Do an MMIO read on interrupts when not uisng MSIs
drm/radeon: Writeback endian fixes
drm/radeon: Remove a bunch of useless _iomem casts
drm/gem: add support for private objects
DRM: clean up and document parsing of video= parameter
DRM: Radeon: Fix section mismatch.
drm: really make debug levels match in edid failure code
drm/radeon/kms: fix i2c map for rv250/280
drm/nouveau/gr: disable fifo access and idle before suspend ctx unload
drm/nouveau: pass flag to engine fini() method on suspend
drm/nouveau: replace nv04_graph_fifo_access() use with direct reg bashing
drm/nv40/gr: rewrite/split context takedown functions
drm/nouveau: detect disabled device in irq handler and return IRQ_NONE
drm/nouveau: ignore connector type when deciding digital/analog on DVI-I
drm/nouveau: Add a quirk for Gigabyte NX86T
...
* 'drm-nouveau-next' of git://anongit.freedesktop.org/git/nouveau/linux-2.6:
drm/nouveau/gr: disable fifo access and idle before suspend ctx unload
drm/nouveau: pass flag to engine fini() method on suspend
drm/nouveau: replace nv04_graph_fifo_access() use with direct reg bashing
drm/nv40/gr: rewrite/split context takedown functions
drm/nouveau: detect disabled device in irq handler and return IRQ_NONE
drm/nouveau: ignore connector type when deciding digital/analog on DVI-I
drm/nouveau: Add a quirk for Gigabyte NX86T
drm/nouveau: do not leak in nv20_graph_create
drm/nv50/dp: fix hack to work for macbooks booted via EFI
It may not be necessary to fail in certain cases (such as failing to idle)
on module unload, whereas on suspend it's important to ensure a consistent
state can be restored on resume.
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
It's completely pointless to save the PGRAPH context when destroying a
channel, so don't bother.
This commit should also fix kernel.org bug 39422, where the DRM channel
state was incorrectly being saved because we left PGRAPH FIFO access
enabled while running the ctxprog.
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
If the connector table is lying, which it often does on the boards of a
particular manufacturer, we may end up doing the wrong thing. Listen
to the encoder table instead, it's more reliable.
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
The connector table lies, the card has DVI-I not HDMI
Fixes bug https://bugs.freedesktop.org/show_bug.cgi?id=35675
v2: Mention the bugreport
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
If we return due to an unknown chipset in
drivers/gpu/drm/nouveau/nv20_graph.c:nv20_graph_create() we'll leak the
memory allocated to 'pgraph'.
This patch should fix the leak.
Signed-off-by: Jesper Juhl <jj@chaosbits.net>
Reviewed-by: Marcin Slusarz <marcin.slusarz@gmail.com>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
drm_pci_device_is_pcie duplicates the funcationality of pci_is_pcie.
Convert callers of the former to the latter. This has the side benefit
of removing an unnecessary search in the PCI configuration space due to
using a saved PCIe capability offset.
[airlied: update for new callsite]
Signed-off-by: Jon Mason <jdmason@kudzu.us>
Signed-off-by: Dave Airlie <airlied@redhat.com>
Two of these error paths returned without freeing "ctx".
Signed-off-by: Dan Carpenter <error27@gmail.com>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
'drm/nouveau: rework vram init/fini ordering a little' changed
the order of instmem.init() and nouveau_mem_vram_init() which
resulted in using ramin_rsvd_vram before it was calculated and
failing to init any accel on pre-NV50 cards.
Since it's only used on <NV50 just calculate it where it's needed
and leave it as default 0 for NV50.
Signed-off-by: Younes Manton <younes.m@gmail.com>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
It's a regression from "drm/nouveau: create temp vmas for both src and
dst of bo moves".
Signed-off-by: Marcin Slusarz <marcin.slusarz@gmail.com>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
Many (all?) of the coefficients related to calculating the
correct temperature are signed integers
This patch correcly parses and stores those values
It also ensures that the default offset is 0 (previously 1)
Affected cards - the original nv50 and the nv40 family
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
Commit "drm/nouveau: add some debug output if nouveau_mm busy at destroy time"
revealed an issue where vram mm takedown would actually fail due to there
still being nodes present, causing nouveau to leak a small amount of memory
on module unload.
This splits TTM/nouveau_mm a bit more cleanly and ensures nouveau_mm fini
isn't done until all gpuobjs are also destroyed.
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
GPU virtual addresses are constant now so this should never be getting hit
anyway and userspace shouldn't break from them being ignored.
This is being done in preference to teaching the code how to deal with BOs
that exist at different virtual addresses within separate VMs.
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
Greatly simplifies a number of things, particularly once per-client GPU
address spaces are involved.
May add this back later once I know what things'll look like.
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
Was previously assuming a page size of 4KiB unless a VMA was present to
override it. Eventually, a buffer won't necessarily have a VMA at all at
some stages of its life, so we need to store this info elsewhere.
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
Abuses existing gpuobj_new() chan argument for this, which in turn forces
all NVOBJ_FLAG_VM allocations to be done from the global heap, not
suballocated from the channel's private heap. Not a problem though in
practise.
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
Userspace hasn't passed us a channel_hint for a long long time now, and
there isn't actually a need to do so anymore anyway.
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
On chipsets using nouveau_vm, the virtual address stays constant, so
the value set at bo creation time is fine.
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
There's lots of boards (all recent ones) that don't have this anymore, so
punt the message to debug loglevel.
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
Not a clue what it is yet, but we get the same numbers as NVIDIA now.
My 465 didn't seem to care to greatly *what* I bashed into these registers..
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
The ability to use NVIDIA's fuc has been retained *temporarily* in order
to better debug any issues that may be lingering in our initial attempt
at writing this ucode. Once I'm fairly confident we're okay, it'll be
removed.
There's a number of things not implemented by this fuc currently, but
most of it is sets of state that our context setup would not have used
anyway. No doubt we'll find out what they're for at some point, and
implement it if required.
This has been tested on 0xc0/0xc4 thus far, and from what I could tell
it worked as well as NVIDIA's. It's also been tested on 0xc1, but even
with NVIDIA's fuc that chipset doesn't work correctly with nouveau yet.
0xc3/0xc8/0xce should in theory be supported too, but I don't have the
hardware to check that.
There's no doubt numerous bugs to squash yet, please report any!
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
We need the physical VRAM address in vinst, even for objects mapped into
a vm, as the gpuobj suspend/resume code uses PMEM to access the object.
Previously, vinst was overloaded to mean "VRAM address" for !VM objects,
and "VM address" for VM objects, causing the wrong data to be accessed
during suspend/resume.
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
Until we know these should work properly, would much rather default to
noaccel than risk giving people corruption/hangs out of the box..
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
There's issues with certain 3D apps still, unknown whether this is a kernel
issue or not.. It does appear that it may be in the 3D driver however.
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
Fixes DMAR faults during accel, more than likely a similar problem to what
was solved on nv50 previously.
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
Many stupid corrections of duplicated includes based on the output of
scripts/checkincludes.pl.
Signed-off-by: Vitaliy Ivanov <vitalivanov@gmail.com>
Signed-off-by: Jiri Kosina <jkosina@suse.cz>
