Wasn't aware they even existed, apparently they do!  They're actually
AGP chips with a bridge as far as I can tell, which puts them in the
same boat as nv40/nv45.
Signed-off-by: NBen Skeggs <bskeggs@redhat.com>

In line with envytools, verified on 4 or 5 BIOS'es.
Signed-off-by: NBen Skeggs <bskeggs@redhat.com>
Signed-off-by: NRoy Spliet <r.spliet@student.tudelft.nl>

This also makes the fact we're giving 512MiB of GART space to all PCIE
boards explicit, although the vast majority (if not all) of them will
now have a ramin_rsvd_vram larger than 2MiB anyway.
Signed-off-by: NBen Skeggs <bskeggs@redhat.com>

TTM assumes an error condition from man->func->get_node() means that
something went horribly wrong, and causes it to bail.

The driver is supposed to return 0, and leave mm_node == NULL to
signal that it couldn't allocate any memory.
Signed-off-by: NBen Skeggs <bskeggs@redhat.com>
Signed-off-by: NDave Airlie <airlied@redhat.com>

Signed-off-by: NBen Skeggs <bskeggs@redhat.com>

The immediate benefit of doing this is that on NV50 and up, the GPU
virtual address of any buffer is now constant, regardless of what
memtype they're placed in.
Signed-off-by: NBen Skeggs <bskeggs@redhat.com>

This structure will also be used for GART in the near future.
Signed-off-by: NBen Skeggs <bskeggs@redhat.com>

Signed-off-by: NBen Skeggs <bskeggs@redhat.com>

'mappable' isn't really used at all, nor is it necessary anymore as the
bo code is capable of moving buffers to mappable vram as required.

'no_vm' isn't necessary anymore either, any places that don't want to be
mapped into a GPU address space should allocate the VRAM directly instead.
Signed-off-by: NBen Skeggs <bskeggs@redhat.com>

Signed-off-by: NBen Skeggs <bskeggs@redhat.com>

This reverts commit 5a893fc2.

This causes a use after free in the ttm free alloc pages path,
when it tries to get the be after the be has been destroyed.
Signed-off-by: NDave Airlie <airlied@redhat.com>

This makes the accounting when using 'debug_dma_dump_mappings()'
and CONFIG_DMA_API_DEBUG=y be assigned to the correct device
instead of 'fallback'.

No functional change - just cosmetic.
Signed-off-by: NKonrad Rzeszutek Wilk <konrad.wilk@oracle.com>

This abstracts the pci/platform interface out a step further,
we can go further but this is far enough for now to allow USB
to be plugged in.

The drivers now just call the init code directly for their
device type.
Signed-off-by: NDave Airlie <airlied@redhat.com>

Reviewed-by: NFrancisco Jerez <currojerez@riseup.net>
Signed-off-by: NBen Skeggs <bskeggs@redhat.com>

Signed-off-by: NBen Skeggs <bskeggs@redhat.com>

Signed-off-by: NBen Skeggs <bskeggs@redhat.com>

NVC0 will be able to share some of nv50's paths this way.  This also makes
it the card-specific vram code responsible for deciding if a given set
of tile_flags is valid, rather than duplicating the allowed types in
nv50_vram.c and nouveau_gem.c
Signed-off-by: NBen Skeggs <bskeggs@redhat.com>

Signed-off-by: NBen Skeggs <bskeggs@redhat.com>

Signed-off-by: NBen Skeggs <bskeggs@redhat.com>

As of this commit, it's guaranteed that if an object is in VRAM that its
GPU virtual address will be constant.
Signed-off-by: NBen Skeggs <bskeggs@redhat.com>

Signed-off-by: NBen Skeggs <bskeggs@redhat.com>

This is required on nv50 as we need to be able to have more precise control
over physical VRAM allocations to avoid buffer corruption when using
buffers of mixed memory types.

This removes some nasty overallocation/alignment that we were previously
using to "control" this problem.
Signed-off-by: NBen Skeggs <bskeggs@redhat.com>

The point is to share more code between the PFB/PGRAPH tile region
hooks, and give the hardware specific functions a chance to allocate
per-region resources.
Signed-off-by: NFrancisco Jerez <currojerez@riseup.net>
Signed-off-by: NBen Skeggs <bskeggs@redhat.com>

nouveau_fence_* functions are not type safe, which could lead to bugs.
Additionally every use of nouveau_fence_unref had to cast struct
nouveau_fence to void **.
Fix it by renaming old functions and creating static inline functions with
new prototypes. We still need old functions, because we pass function
pointers to ttm.
As we are wrapping functions, drop unused "void *arg" parameter where possible.
Signed-off-by: NMarcin Slusarz <marcin.slusarz@gmail.com>
Signed-off-by: NFrancisco Jerez <currojerez@riseup.net>
Signed-off-by: NBen Skeggs <bskeggs@redhat.com>

Improvements:
 - Fix bug in switch statement
 - Add parts of 0x10022c, 0x10023c
 - Clean up 0x100234
 - Comment out assumption in 0x100228 until verified
Signed-off-by: NRoy Spliet <r.spliet@student.tudelft.nl>
Signed-off-by: NBen Skeggs <bskeggs@redhat.com>

Signed-off-by: NBen Skeggs <bskeggs@redhat.com>

Remove some unused/duplicated definitions and make sparse happy again.
Signed-off-by: NFrancisco Jerez <currojerez@riseup.net>
Signed-off-by: NBen Skeggs <bskeggs@redhat.com>

Signed-off-by: NBen Skeggs <bskeggs@redhat.com>

This isn't correct everywhere yet, but since we don't use the data yet
it's perfectly safe to push in, and the information we gain from logs
will help to fix the remaining issues.

v2 (Ben Skeggs <bskeggs@redhat.com>):
- fixed up formatting
- free parsed timing info on takedown
- switched timing table printout to debug loglevel
Signed-off-by: NRoy Spliet <r.spliet@student.tudelft.nl>
Signed-off-by: NBen Skeggs <bskeggs@redhat.com>

Instead of emptying the caches to avoid a race with the PFIFO puller,
go straight ahead and try to recover from it when it happens. Also,
kill pfifo->cache_flush and tile->lock, we don't need them anymore.
Signed-off-by: NFrancisco Jerez <currojerez@riseup.net>
Signed-off-by: NBen Skeggs <bskeggs@redhat.com>

Signed-off-by: NFrancisco Jerez <currojerez@riseup.net>
Signed-off-by: NBen Skeggs <bskeggs@redhat.com>

FW seems to be broken on nv18, it causes random lockups and breaks
suspend/resume even with the blob.
Signed-off-by: NFrancisco Jerez <currojerez@riseup.net>
Signed-off-by: NBen Skeggs <bskeggs@redhat.com>

Reviewed-by: NFrancisco Jerez <currojerez@riseup.net>
Signed-off-by: NBen Skeggs <bskeggs@redhat.com>

Will be used at a later point when we plug in an alternative VRAM memory
manager for GeForce 8+ boards.

Based on pscnv code to do the same.
Signed-off-by: NBen Skeggs <bskeggs@redhat.com>
Signed-off-by: NMarcin Kościelnicki <koriakin@0x04.net>
Signed-off-by: NBen Skeggs <bskeggs@redhat.com>

Reviewed-by: NFrancisco Jerez <currojerez@riseup.net>
Signed-off-by: NBen Skeggs <bskeggs@redhat.com>

On some nv4x cards (specifically, the ones that use an internal
PCIE->AGP bridge) the AGP controller state isn't preserved after a
suspend/resume cycle, and the AGP control registers have moved from
0x18xx to 0x100xx, so the FW check in nouveau_mem_reset_agp() doesn't
quite work. Check "dev->agp->mode" instead.
Signed-off-by: NFrancisco Jerez <currojerez@riseup.net>
Signed-off-by: NBen Skeggs <bskeggs@redhat.com>

Signed-off-by: NBen Skeggs <bskeggs@redhat.com>

Previously nouveau_mem_reset_agp() was only disabling AGP fast writes
when coming back from suspend. However, the "locked out of the card
because of FW" problem can also be reproduced on init if you
unload/reload nouveau.ko several times. This patch makes the AGP code
reset FW on init.
Signed-off-by: NFrancisco Jerez <currojerez@riseup.net>
Signed-off-by: NBen Skeggs <bskeggs@redhat.com>

Signed-off-by: NBen Skeggs <bskeggs@redhat.com>

BIOS scripts usually make an attempt to reset the AGP controller,
however on some nv4x cards doing it properly involves switching FW off
and on: if we do that without updating the AGP bridge settings
accordingly (e.g. with the corresponding calls to agp_enable()) we
will be locking ourselves out of the card MMIO space. Do it from
nouveau_mem_reset_agp() before the init scripts are executed.
Signed-off-by: NFrancisco Jerez <currojerez@riseup.net>
Signed-off-by: NBen Skeggs <bskeggs@redhat.com>