search for: gpuobj

Signed-off-by: Marcin Slusarz <marcin.slusarz at gmail.com> --- drivers/gpu/drm/nouveau/core/core/gpuobj.c | 6 +++++- drivers/gpu/drm/nouveau/core/include/core/gpuobj.h | 3 +++ 2 files changed, 8 insertions(+), 1 deletion(-) diff --git a/drivers/gpu/drm/nouveau/core/core/gpuobj.c b/drivers/gpu/drm/nouveau/core/core/gpuobj.c index c2a7608..48121d2 100644 --- a/drivers/gpu/drm/nouveau/core/co...

...those code paths and uploaded "patch" here: http://people.freedesktop.org/~mslusarz/0001-codepaths-without-error-handling.patch (Please let me know if you are going to fix those to not duplicate work) --- drivers/gpu/drm/nouveau/core/core/engine.c | 1 + drivers/gpu/drm/nouveau/core/core/gpuobj.c | 9 ++++++--- drivers/gpu/drm/nouveau/core/core/parent.c | 4 +++- 3 files changed, 10 insertions(+), 4 deletions(-) diff --git a/drivers/gpu/drm/nouveau/core/core/engine.c b/drivers/gpu/drm/nouveau/core/core/engine.c index 09b3bd5..4319854 100644 --- a/drivers/gpu/drm/nouveau/core/core/engine....

...t [drm:drm_sg_alloc] drm_sg_alloc Aug 6 21:11:46 localhost [drm:drm_sg_alloc] sg size=16777216 pages=4096 Aug 6 21:11:46 localhost [drm:drm_sg_alloc] sg alloc handle = 15f7d200 Aug 6 21:11:46 localhost [drm:drm_sg_alloc] sg alloc virtual = ffffc20015f81000 Aug 6 21:11:46 localhost [drm:nouveau_gpuobj_new_fake] offset=0x00010000 size=0x00000200 flags=0x00000003 Aug 6 21:11:46 localhost [drm:nouveau_gpuobj_new_fake] gpuobj ffff81003a5a19c0 Aug 6 21:11:46 localhost [drm:nv40_graph_init] Loading context-switch voodoo Aug 6 21:11:46 localhost [drm:nouveau_fifo_init] Setting defaults for remaining...

Hi, Please find attached the patches which I currently use on my desktop machine for dual head with the randr branch to fix the issues which I found. They may help others as well but may e.g. also disable the Xv blitter which might be working for some (but didn't on my card) - more information is found in the text comments in the patches. I have to hurry so this is short, will be back.

There are a lot of places that allocate multiples of 1000, but do not set alignment correctly and still require this alignment implicitly or explicitly. --- drivers/gpu/drm/nouveau/core/core/gpuobj.c | 8 ++++++++ 1 file changed, 8 insertions(+) diff --git a/drivers/gpu/drm/nouveau/core/core/gpuobj.c b/drivers/gpu/drm/nouveau/core/core/gpuobj.c index 7595506..7bcae1d 100644 --- a/drivers/gpu/drm/nouveau/core/core/gpuobj.c +++ b/drivers/gpu/drm/nouveau/core/core/gpuobj.c @@ -65,6 +65,14 @@ no...

...uveau/nouveau_object.c @@ -33,6 +33,7 @@ #include "drmP.h" #include "drm.h" #include "nouveau_drv.h" +#include "nouveau_dma.h" #include "nouveau_drm.h" /* NVidia uses context objects to drive drawing operations. @@ -1099,6 +1100,41 @@ nouveau_gpuobj_channel_init(struct nouveau_channel *chan, return ret; } + /* Two objects for kernel consumption, on nv50 they allow direct access + * to vram. This is needed because we don't know the tiling layout. + */ + vram = NULL; + ret = nouveau_gpuobj_dma_new(chan, NV_CLASS_DMA_IN_MEMORY, +...

...e = &dev_priv->engine; dev_priv->init_state = NOUVEAU_CARD_INIT_FAILED; @@ -325,12 +325,12 @@ nouveau_card_init(struct drm_device *dev) if (drm_core_check_feature(dev, DRIVER_MODESET)) { ret = nouveau_bios_init(dev); if (ret) - return ret; + goto out; } ret = nouveau_gpuobj_early_init(dev); if (ret) - return ret; + goto out; /* Initialise instance memory, must happen before mem_init so we * know exactly how much VRAM we're able to use for "normal" @@ -338,52 +338,52 @@ nouveau_card_init(struct drm_device *dev) */ ret = engine->instmem....

...e = &dev_priv->engine; dev_priv->init_state = NOUVEAU_CARD_INIT_FAILED; @@ -325,12 +325,12 @@ nouveau_card_init(struct drm_device *dev) if (drm_core_check_feature(dev, DRIVER_MODESET)) { ret = nouveau_bios_init(dev); if (ret) - return ret; + goto out; } ret = nouveau_gpuobj_early_init(dev); if (ret) - return ret; + goto out_bios; /* Initialise instance memory, must happen before mem_init so we * know exactly how much VRAM we're able to use for "normal" @@ -338,52 +338,52 @@ nouveau_card_init(struct drm_device *dev) */ ret = engine->ins...

Signed-off-by: Marcin Ko?cielnicki <koriakin at 0x04.net> --- drivers/gpu/drm/nouveau/nv04_graph.c | 2 +- 1 files changed, 1 insertions(+), 1 deletions(-) diff --git a/drivers/gpu/drm/nouveau/nv04_graph.c b/drivers/gpu/drm/nouveau/nv04_graph.c index 396ee92..d561d77 100644 --- a/drivers/gpu/drm/nouveau/nv04_graph.c +++ b/drivers/gpu/drm/nouveau/nv04_graph.c @@ -543,7 +543,7 @@

...command not found: 0x61 [drm] nouveau 0000:01:00.0: 0xB5E2: parsing output script 0 [drm] nouveau 0000:01:00.0: 0xB758: parsing output script 0 [drm] nouveau 0000:01:00.0: 0xADC8: parsing output script 0 [drm] nouveau 0000:01:00.0: Error reading back PRAMIN at 0x000004d0 [drm] nouveau 0000:01:00.0: gpuobj dbb3fc00 still exists at takedown, refs=1 [drm] nouveau 0000:01:00.0: gpuobj dbb3fa80 still exists at takedown, refs=1 [drm] nouveau 0000:01:00.0: gpuobj dbb3f940 still exists at takedown, refs=0 [drm] nouveau 0000:01:00.0: gpuobj dbb3f8c0 still exists at takedown, refs=1 [drm] nouveau 0000:01:00.0...

...om> wrote: > There are a lot of places that allocate multiples of 1000, > but do not set alignment correctly and still require this > alignment implicitly or explicitly. This is wrong. Where are the places you think you need this? Ben. > --- > drivers/gpu/drm/nouveau/core/core/gpuobj.c | 8 ++++++++ > 1 file changed, 8 insertions(+) > > diff --git a/drivers/gpu/drm/nouveau/core/core/gpuobj.c b/drivers/gpu/drm/nouveau/core/core/gpuobj.c > index 7595506..7bcae1d 100644 > --- a/drivers/gpu/drm/nouveau/core/core/gpuobj.c > +++ b/drivers/gpu/drm/nouveau/core/core/g...

...hash_handle:75 - ch0 handle=0x80000001 [ 13.271499] [drm] nouveau 0000:01:00.0: nouveau_ramht_hash_handle:86 - hash=0x00000088 [ 13.271499] [drm] nouveau 0000:01:00.0: nouveau_ramht_insert:141 - insert ch0 0x00000088: h=0x80000001, c=0x80011639 [ 13.271499] [drm] nouveau 0000:01:00.0: nouveau_gpuobj_ref_add:457 - ch0 h=0x8000000e gpuobj=ffff88000219ac00 [ 13.271499] [drm] nouveau 0000:01:00.0: nouveau_ramht_hash_handle:75 - ch0 handle=0x8000000e [ 13.271499] [drm] nouveau 0000:01:00.0: nouveau_ramht_hash_handle:86 - hash=0x000000f0 [ 13.271499] [drm] nouveau 0000:01:00.0: nouveau_ramht_i...

....4.45-server-1.mga2. At the time of both lockups, I had a couple dozen Firefox windows plus several other miscellaneous windows open. Here are the some snippets from /var/log/messages a few days before the first lock-up: May 27 10:24:25 one kernel: [779598.847459] [drm] nouveau 0000:06:00.0: gpuobj -28 ... May 27 10:28:37 one kernel: [779850.989024] [drm] nouveau 0000:06:00.0: gpuobj -28 ... May 27 10:28:45 one kernel: [779858.365759] [drm] nouveau 0000:06:00.0: Error creating object: -28 (3/0xbeef9039) These happened at approximately the time of the first lockup. Keyboard L...

...gt;id = channel; chan->file_priv = file_priv; diff --git a/drivers/gpu/drm/nouveau/nouveau_object.c b/drivers/gpu/drm/nouveau/nouveau_object.c index 65d8d25..efcce79 100644 --- a/drivers/gpu/drm/nouveau/nouveau_object.c +++ b/drivers/gpu/drm/nouveau/nouveau_object.c @@ -468,6 +468,7 @@ nouveau_gpuobj_ref_add(struct drm_device *dev, struct nouveau_channel *chan, ref = kzalloc(sizeof(*ref), GFP_KERNEL); if (!ref) return -ENOMEM; + INIT_LIST_HEAD(&ref->list); ref->gpuobj = gpuobj; ref->channel = chan; ref->instance = instance; -- 1.6.4

...er got redefined. > > Sometimes the fixes are not obvious. > > This is the output of: /path/to/smatch_scripts/filter_kernel_deref_check.sh warns.txt > > regards, > dan carpenter > > (...) > drivers/gpu/drm/nouveau/nouveau_object.c +891 'chan': if (!chan || !gpuobj_ret || *gpuobj_ret != NULL) > drivers/gpu/drm/nouveau/nouveau_sgdma.c +61 'nvbe': if (nvbe && nvbe->pages) { > drivers/gpu/drm/nouveau/nouveau_connector.c +91 'connector': if (!connector) > drivers/gpu/drm/nouveau/nv50_crtc.c +306 'crtc': if (!crtc) >...

...fffffa016c323>] nouveau_instobj_create_+0x43/0x90 [nouveau] [<ffffffffa016cf8c>] nv50_instobj_ctor+0x4c/0xf0 [nouveau] [<ffffffffa0152163>] nouveau_object_ctor+0x33/0xc0 [nouveau] [<ffffffffa016cd51>] nv50_instmem_alloc+0x21/0x30 [nouveau] [<ffffffffa0150917>] nouveau_gpuobj_create_+0x247/0x2f0 [nouveau] [<ffffffff8155b35a>] ? _raw_spin_unlock_irqrestore+0x3a/0x70 [<ffffffff810921fd>] ? trace_hardirqs_on_caller+0x10d/0x1a0 [<ffffffffa014f4bc>] nouveau_engctx_create_+0x25c/0x2a0 [nouveau] [<ffffffffa0176791>] nv50_disp_data_ctor+0xc1/0xd0 [no...

nouveau_bo_wait will make the GPU channel wait for fence if possible, otherwise falling back to waiting with the CPU using ttm_bo_wait. The nouveau_fence_sync function currently returns -ENOSYS, and is the focus of the next patch. Signed-off-by: Luca Barbieri <luca at luca-barbieri.com> --- drivers/gpu/drm/nouveau/nouveau_bo.c | 68 ++++++++++++++++++++++++++++++-

.... Please review and comment --- drivers/gpu/drm/nouveau/nouveau_acpi.c | 3 ++- drivers/gpu/drm/nouveau/nouveau_device.c | 26 +++++++++++++++----------- drivers/gpu/drm/nouveau/nouveau_fb.h | 1 - drivers/gpu/drm/nouveau/nouveau_gpio.h | 3 --- drivers/gpu/drm/nouveau/nouveau_gpuobj.c | 12 ++++-------- drivers/gpu/drm/nouveau/nouveau_perf.c | 2 +- drivers/gpu/drm/nouveau/nouveau_pm.c | 6 +++--- drivers/gpu/drm/nouveau/nouveau_state.c | 2 ++ drivers/gpu/drm/nouveau/nouveau_therm.c | 63 +++++++++++++++++++++++++++++++++++++++++++++++++-------------- dri...

Signed-off-by: Marcin Slusarz <marcin.slusarz at gmail.com> --- drivers/gpu/drm/nouveau/core/core/gpuobj.c | 2 +- drivers/gpu/drm/nouveau/core/core/object.c | 8 ++++---- drivers/gpu/drm/nouveau/core/core/option.c | 2 +- .../drm/nouveau/core/engine/copy/fuc/nva3.fuc.h | 4 ++-- .../drm/nouveau/core/engine/copy/fuc/nvc0.fuc.h | 4 ++-- .../drm/nouveau/core...

...e is released as well. It may or may not happen at the same time as the signed GR firmware, hence this currently limited implementation to allow GR to be used. The current abstraction for reseting securely-managed falcons should be suitable for both models. Alexandre Courbot (5): core: add gpuobj memcpy helper functions core: add support for secure boot gr: support for securely-booted FECS firmware secboot/gm200: add secure-boot support secboot/gm20b: add secure boot support drm/nouveau/include/nvkm/core/device.h | 3 + drm/nouveau/include/nvkm/core/gpuobj.h | 4 + drm...