linux/drivers/gpu/drm/nouveau/nvkm/engine/device/base.c

530 lines
13 KiB
C
Raw Normal View History

drm/nouveau/core: pull in most of the new core infrastructure This commit provides most of the infrastructure to support a major overhaul of Nouveau's internals coming in the following commits. This work aims to take all the things we've learned over the last several years, and turn that into a cleaner architecture that's more maintainable going forward. RAMHT and MM bits of the new core have been left out for the moment, and will be pulled in as I go through the process of porting the code to become either subdev or engine modules. There are several main goals I wanted to achieve through this work: -- Reduce complexity The goal here was to make each component of the driver as independent as possible, which will ease maintainability and readability, and provide a good base for resetting locked up GPU units in the future. -- Better tracking of GPU units that are required at any given time This is for future PM work, we'll be able to tell exactly what parts of the GPU we need powered at any given point (etc). -- Expose all available NVIDIA GPUs to the client In order to support things such as multi-GPU channels, we want to be able to expose all the NVIDIA GPUs to the client over a single file descriptor so it can send a single push buffer to multiple GPUs. -- Untangle the core hardware support code from the DRM implementation This happened initially as an unexpected side-effect of developing the initial core infrastructure in userspace, but it turned into a goal of the whole project. Initial benefits will be the availablility of a number of userspace tools and tests using the same code as the driver itself, but will also be important as I look into some virtualisation ideas. v2: Ben Skeggs <bskeggs@redhat.com> - fix duplicate assignments noticed by clang - implement some forgotten yelling in error path - ensure 64-bit engine mask is used everywhere v3: Marcin Slusarz <marcin.slusarz@gmail.com> - sparse fixes - inline nv_printk into nv_assert to prevent recursive inlining issues v4: Ben Skeggs <bskeggs@redhat.com> - fixed minor memory leak on gpuobj destruction Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
2012-07-05 21:36:43 +00:00
/*
* Copyright 2012 Red Hat Inc.
*
* Permission is hereby granted, free of charge, to any person obtaining a
* copy of this software and associated documentation files (the "Software"),
* to deal in the Software without restriction, including without limitation
* the rights to use, copy, modify, merge, publish, distribute, sublicense,
* and/or sell copies of the Software, and to permit persons to whom the
* Software is furnished to do so, subject to the following conditions:
*
* The above copyright notice and this permission notice shall be included in
* all copies or substantial portions of the Software.
*
* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
* IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
* FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
* THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
* OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
* ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
* OTHER DEALINGS IN THE SOFTWARE.
*
* Authors: Ben Skeggs
*/
#include "priv.h"
#include "acpi.h"
drm/nouveau/core: pull in most of the new core infrastructure This commit provides most of the infrastructure to support a major overhaul of Nouveau's internals coming in the following commits. This work aims to take all the things we've learned over the last several years, and turn that into a cleaner architecture that's more maintainable going forward. RAMHT and MM bits of the new core have been left out for the moment, and will be pulled in as I go through the process of porting the code to become either subdev or engine modules. There are several main goals I wanted to achieve through this work: -- Reduce complexity The goal here was to make each component of the driver as independent as possible, which will ease maintainability and readability, and provide a good base for resetting locked up GPU units in the future. -- Better tracking of GPU units that are required at any given time This is for future PM work, we'll be able to tell exactly what parts of the GPU we need powered at any given point (etc). -- Expose all available NVIDIA GPUs to the client In order to support things such as multi-GPU channels, we want to be able to expose all the NVIDIA GPUs to the client over a single file descriptor so it can send a single push buffer to multiple GPUs. -- Untangle the core hardware support code from the DRM implementation This happened initially as an unexpected side-effect of developing the initial core infrastructure in userspace, but it turned into a goal of the whole project. Initial benefits will be the availablility of a number of userspace tools and tests using the same code as the driver itself, but will also be important as I look into some virtualisation ideas. v2: Ben Skeggs <bskeggs@redhat.com> - fix duplicate assignments noticed by clang - implement some forgotten yelling in error path - ensure 64-bit engine mask is used everywhere v3: Marcin Slusarz <marcin.slusarz@gmail.com> - sparse fixes - inline nv_printk into nv_assert to prevent recursive inlining issues v4: Ben Skeggs <bskeggs@redhat.com> - fixed minor memory leak on gpuobj destruction Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
2012-07-05 21:36:43 +00:00
#include <core/notify.h>
#include <core/option.h>
#include <subdev/bios.h>
drm/nouveau/core: pull in most of the new core infrastructure This commit provides most of the infrastructure to support a major overhaul of Nouveau's internals coming in the following commits. This work aims to take all the things we've learned over the last several years, and turn that into a cleaner architecture that's more maintainable going forward. RAMHT and MM bits of the new core have been left out for the moment, and will be pulled in as I go through the process of porting the code to become either subdev or engine modules. There are several main goals I wanted to achieve through this work: -- Reduce complexity The goal here was to make each component of the driver as independent as possible, which will ease maintainability and readability, and provide a good base for resetting locked up GPU units in the future. -- Better tracking of GPU units that are required at any given time This is for future PM work, we'll be able to tell exactly what parts of the GPU we need powered at any given point (etc). -- Expose all available NVIDIA GPUs to the client In order to support things such as multi-GPU channels, we want to be able to expose all the NVIDIA GPUs to the client over a single file descriptor so it can send a single push buffer to multiple GPUs. -- Untangle the core hardware support code from the DRM implementation This happened initially as an unexpected side-effect of developing the initial core infrastructure in userspace, but it turned into a goal of the whole project. Initial benefits will be the availablility of a number of userspace tools and tests using the same code as the driver itself, but will also be important as I look into some virtualisation ideas. v2: Ben Skeggs <bskeggs@redhat.com> - fix duplicate assignments noticed by clang - implement some forgotten yelling in error path - ensure 64-bit engine mask is used everywhere v3: Marcin Slusarz <marcin.slusarz@gmail.com> - sparse fixes - inline nv_printk into nv_assert to prevent recursive inlining issues v4: Ben Skeggs <bskeggs@redhat.com> - fixed minor memory leak on gpuobj destruction Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
2012-07-05 21:36:43 +00:00
static DEFINE_MUTEX(nv_devices_mutex);
static LIST_HEAD(nv_devices);
static struct nvkm_device *
nvkm_device_find_locked(u64 handle)
drm/nouveau/core: pull in most of the new core infrastructure This commit provides most of the infrastructure to support a major overhaul of Nouveau's internals coming in the following commits. This work aims to take all the things we've learned over the last several years, and turn that into a cleaner architecture that's more maintainable going forward. RAMHT and MM bits of the new core have been left out for the moment, and will be pulled in as I go through the process of porting the code to become either subdev or engine modules. There are several main goals I wanted to achieve through this work: -- Reduce complexity The goal here was to make each component of the driver as independent as possible, which will ease maintainability and readability, and provide a good base for resetting locked up GPU units in the future. -- Better tracking of GPU units that are required at any given time This is for future PM work, we'll be able to tell exactly what parts of the GPU we need powered at any given point (etc). -- Expose all available NVIDIA GPUs to the client In order to support things such as multi-GPU channels, we want to be able to expose all the NVIDIA GPUs to the client over a single file descriptor so it can send a single push buffer to multiple GPUs. -- Untangle the core hardware support code from the DRM implementation This happened initially as an unexpected side-effect of developing the initial core infrastructure in userspace, but it turned into a goal of the whole project. Initial benefits will be the availablility of a number of userspace tools and tests using the same code as the driver itself, but will also be important as I look into some virtualisation ideas. v2: Ben Skeggs <bskeggs@redhat.com> - fix duplicate assignments noticed by clang - implement some forgotten yelling in error path - ensure 64-bit engine mask is used everywhere v3: Marcin Slusarz <marcin.slusarz@gmail.com> - sparse fixes - inline nv_printk into nv_assert to prevent recursive inlining issues v4: Ben Skeggs <bskeggs@redhat.com> - fixed minor memory leak on gpuobj destruction Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
2012-07-05 21:36:43 +00:00
{
struct nvkm_device *device;
drm/nouveau/core: pull in most of the new core infrastructure This commit provides most of the infrastructure to support a major overhaul of Nouveau's internals coming in the following commits. This work aims to take all the things we've learned over the last several years, and turn that into a cleaner architecture that's more maintainable going forward. RAMHT and MM bits of the new core have been left out for the moment, and will be pulled in as I go through the process of porting the code to become either subdev or engine modules. There are several main goals I wanted to achieve through this work: -- Reduce complexity The goal here was to make each component of the driver as independent as possible, which will ease maintainability and readability, and provide a good base for resetting locked up GPU units in the future. -- Better tracking of GPU units that are required at any given time This is for future PM work, we'll be able to tell exactly what parts of the GPU we need powered at any given point (etc). -- Expose all available NVIDIA GPUs to the client In order to support things such as multi-GPU channels, we want to be able to expose all the NVIDIA GPUs to the client over a single file descriptor so it can send a single push buffer to multiple GPUs. -- Untangle the core hardware support code from the DRM implementation This happened initially as an unexpected side-effect of developing the initial core infrastructure in userspace, but it turned into a goal of the whole project. Initial benefits will be the availablility of a number of userspace tools and tests using the same code as the driver itself, but will also be important as I look into some virtualisation ideas. v2: Ben Skeggs <bskeggs@redhat.com> - fix duplicate assignments noticed by clang - implement some forgotten yelling in error path - ensure 64-bit engine mask is used everywhere v3: Marcin Slusarz <marcin.slusarz@gmail.com> - sparse fixes - inline nv_printk into nv_assert to prevent recursive inlining issues v4: Ben Skeggs <bskeggs@redhat.com> - fixed minor memory leak on gpuobj destruction Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
2012-07-05 21:36:43 +00:00
list_for_each_entry(device, &nv_devices, head) {
if (device->handle == handle)
return device;
drm/nouveau/core: pull in most of the new core infrastructure This commit provides most of the infrastructure to support a major overhaul of Nouveau's internals coming in the following commits. This work aims to take all the things we've learned over the last several years, and turn that into a cleaner architecture that's more maintainable going forward. RAMHT and MM bits of the new core have been left out for the moment, and will be pulled in as I go through the process of porting the code to become either subdev or engine modules. There are several main goals I wanted to achieve through this work: -- Reduce complexity The goal here was to make each component of the driver as independent as possible, which will ease maintainability and readability, and provide a good base for resetting locked up GPU units in the future. -- Better tracking of GPU units that are required at any given time This is for future PM work, we'll be able to tell exactly what parts of the GPU we need powered at any given point (etc). -- Expose all available NVIDIA GPUs to the client In order to support things such as multi-GPU channels, we want to be able to expose all the NVIDIA GPUs to the client over a single file descriptor so it can send a single push buffer to multiple GPUs. -- Untangle the core hardware support code from the DRM implementation This happened initially as an unexpected side-effect of developing the initial core infrastructure in userspace, but it turned into a goal of the whole project. Initial benefits will be the availablility of a number of userspace tools and tests using the same code as the driver itself, but will also be important as I look into some virtualisation ideas. v2: Ben Skeggs <bskeggs@redhat.com> - fix duplicate assignments noticed by clang - implement some forgotten yelling in error path - ensure 64-bit engine mask is used everywhere v3: Marcin Slusarz <marcin.slusarz@gmail.com> - sparse fixes - inline nv_printk into nv_assert to prevent recursive inlining issues v4: Ben Skeggs <bskeggs@redhat.com> - fixed minor memory leak on gpuobj destruction Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
2012-07-05 21:36:43 +00:00
}
return NULL;
}
struct nvkm_device *
nvkm_device_find(u64 handle)
{
struct nvkm_device *device;
mutex_lock(&nv_devices_mutex);
device = nvkm_device_find_locked(handle);
drm/nouveau/core: pull in most of the new core infrastructure This commit provides most of the infrastructure to support a major overhaul of Nouveau's internals coming in the following commits. This work aims to take all the things we've learned over the last several years, and turn that into a cleaner architecture that's more maintainable going forward. RAMHT and MM bits of the new core have been left out for the moment, and will be pulled in as I go through the process of porting the code to become either subdev or engine modules. There are several main goals I wanted to achieve through this work: -- Reduce complexity The goal here was to make each component of the driver as independent as possible, which will ease maintainability and readability, and provide a good base for resetting locked up GPU units in the future. -- Better tracking of GPU units that are required at any given time This is for future PM work, we'll be able to tell exactly what parts of the GPU we need powered at any given point (etc). -- Expose all available NVIDIA GPUs to the client In order to support things such as multi-GPU channels, we want to be able to expose all the NVIDIA GPUs to the client over a single file descriptor so it can send a single push buffer to multiple GPUs. -- Untangle the core hardware support code from the DRM implementation This happened initially as an unexpected side-effect of developing the initial core infrastructure in userspace, but it turned into a goal of the whole project. Initial benefits will be the availablility of a number of userspace tools and tests using the same code as the driver itself, but will also be important as I look into some virtualisation ideas. v2: Ben Skeggs <bskeggs@redhat.com> - fix duplicate assignments noticed by clang - implement some forgotten yelling in error path - ensure 64-bit engine mask is used everywhere v3: Marcin Slusarz <marcin.slusarz@gmail.com> - sparse fixes - inline nv_printk into nv_assert to prevent recursive inlining issues v4: Ben Skeggs <bskeggs@redhat.com> - fixed minor memory leak on gpuobj destruction Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
2012-07-05 21:36:43 +00:00
mutex_unlock(&nv_devices_mutex);
return device;
drm/nouveau/core: pull in most of the new core infrastructure This commit provides most of the infrastructure to support a major overhaul of Nouveau's internals coming in the following commits. This work aims to take all the things we've learned over the last several years, and turn that into a cleaner architecture that's more maintainable going forward. RAMHT and MM bits of the new core have been left out for the moment, and will be pulled in as I go through the process of porting the code to become either subdev or engine modules. There are several main goals I wanted to achieve through this work: -- Reduce complexity The goal here was to make each component of the driver as independent as possible, which will ease maintainability and readability, and provide a good base for resetting locked up GPU units in the future. -- Better tracking of GPU units that are required at any given time This is for future PM work, we'll be able to tell exactly what parts of the GPU we need powered at any given point (etc). -- Expose all available NVIDIA GPUs to the client In order to support things such as multi-GPU channels, we want to be able to expose all the NVIDIA GPUs to the client over a single file descriptor so it can send a single push buffer to multiple GPUs. -- Untangle the core hardware support code from the DRM implementation This happened initially as an unexpected side-effect of developing the initial core infrastructure in userspace, but it turned into a goal of the whole project. Initial benefits will be the availablility of a number of userspace tools and tests using the same code as the driver itself, but will also be important as I look into some virtualisation ideas. v2: Ben Skeggs <bskeggs@redhat.com> - fix duplicate assignments noticed by clang - implement some forgotten yelling in error path - ensure 64-bit engine mask is used everywhere v3: Marcin Slusarz <marcin.slusarz@gmail.com> - sparse fixes - inline nv_printk into nv_assert to prevent recursive inlining issues v4: Ben Skeggs <bskeggs@redhat.com> - fixed minor memory leak on gpuobj destruction Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
2012-07-05 21:36:43 +00:00
}
int
nvkm_device_list(u64 *name, int size)
{
struct nvkm_device *device;
int nr = 0;
mutex_lock(&nv_devices_mutex);
list_for_each_entry(device, &nv_devices, head) {
if (nr++ < size)
name[nr - 1] = device->handle;
}
mutex_unlock(&nv_devices_mutex);
return nr;
}
#include <core/parent.h>
#include <core/client.h>
struct nvkm_device *
nv_device(void *obj)
{
struct nvkm_object *device = nv_object(obj);
if (device->engine == NULL) {
while (device && device->parent) {
if (!nv_iclass(device, NV_SUBDEV_CLASS) &&
device->parent == &nvkm_client(device)->namedb.parent.object) {
struct {
struct nvkm_parent base;
struct nvkm_device *device;
} *udevice = (void *)device;
return udevice->device;
}
device = device->parent;
}
} else {
device = &nv_object(obj)->engine->subdev.object;
if (device && device->parent)
device = device->parent;
}
#if CONFIG_NOUVEAU_DEBUG >= NV_DBG_PARANOIA
BUG_ON(!device);
#endif
return (void *)device;
}
static int
nvkm_device_event_ctor(struct nvkm_object *object, void *data, u32 size,
struct nvkm_notify *notify)
{
if (!WARN_ON(size != 0)) {
notify->size = 0;
notify->types = 1;
notify->index = 0;
return 0;
}
return -EINVAL;
}
static const struct nvkm_event_func
nvkm_device_event_func = {
.ctor = nvkm_device_event_ctor,
};
int
nvkm_device_fini(struct nvkm_device *device, bool suspend)
{
struct nvkm_object *subdev;
int ret, i;
for (i = NVDEV_SUBDEV_NR - 1; i >= 0; i--) {
if ((subdev = device->subdev[i])) {
if (!nv_iclass(subdev, NV_ENGINE_CLASS)) {
ret = nvkm_object_dec(subdev, suspend);
if (ret && suspend)
goto fail;
}
}
}
ret = nvkm_acpi_fini(device, suspend);
if (device->func->fini)
device->func->fini(device, suspend);
fail:
for (; ret && i < NVDEV_SUBDEV_NR; i++) {
if ((subdev = device->subdev[i])) {
if (!nv_iclass(subdev, NV_ENGINE_CLASS)) {
ret = nvkm_object_inc(subdev);
if (ret) {
/* XXX */
}
}
}
}
return ret;
}
int
nvkm_device_preinit(struct nvkm_device *device)
{
int ret;
s64 time;
nvdev_trace(device, "preinit running...\n");
time = ktime_to_us(ktime_get());
if (device->func->preinit) {
ret = device->func->preinit(device);
if (ret)
goto fail;
}
time = ktime_to_us(ktime_get()) - time;
nvdev_trace(device, "preinit completed in %lldus\n", time);
return 0;
fail:
nvdev_error(device, "preinit failed with %d\n", ret);
return ret;
}
int
nvkm_device_init(struct nvkm_device *device)
{
struct nvkm_object *subdev;
int ret, i = 0, c;
ret = nvkm_device_preinit(device);
if (ret)
return ret;
ret = nvkm_acpi_init(device);
if (ret)
goto fail;
for (i = 0, c = 0; i < NVDEV_SUBDEV_NR; i++) {
#define _(s,m) case s: if (device->oclass[s] && !device->subdev[s]) { \
ret = nvkm_object_old(nv_object(device), NULL, \
device->oclass[s], NULL, (s), \
(struct nvkm_object **)&device->m); \
if (ret == -ENODEV) { \
device->oclass[s] = NULL; \
continue; \
} \
if (ret) \
goto fail; \
device->subdev[s] = (struct nvkm_object *)device->m; \
} break
switch (i) {
_(NVDEV_SUBDEV_BAR , bar);
_(NVDEV_SUBDEV_VBIOS , bios);
_(NVDEV_SUBDEV_BUS , bus);
_(NVDEV_SUBDEV_CLK , clk);
_(NVDEV_SUBDEV_DEVINIT, devinit);
_(NVDEV_SUBDEV_FB , fb);
_(NVDEV_SUBDEV_FUSE , fuse);
_(NVDEV_SUBDEV_GPIO , gpio);
_(NVDEV_SUBDEV_I2C , i2c);
_(NVDEV_SUBDEV_IBUS , ibus);
_(NVDEV_SUBDEV_INSTMEM, imem);
_(NVDEV_SUBDEV_LTC , ltc);
_(NVDEV_SUBDEV_MC , mc);
_(NVDEV_SUBDEV_MMU , mmu);
_(NVDEV_SUBDEV_MXM , mxm);
_(NVDEV_SUBDEV_PMU , pmu);
_(NVDEV_SUBDEV_THERM , therm);
_(NVDEV_SUBDEV_TIMER , timer);
_(NVDEV_SUBDEV_VOLT , volt);
_(NVDEV_ENGINE_BSP , bsp);
_(NVDEV_ENGINE_CE0 , ce[0]);
_(NVDEV_ENGINE_CE1 , ce[1]);
_(NVDEV_ENGINE_CE2 , ce[2]);
_(NVDEV_ENGINE_CIPHER , cipher);
_(NVDEV_ENGINE_DISP , disp);
_(NVDEV_ENGINE_DMAOBJ , dma);
_(NVDEV_ENGINE_FIFO , fifo);
_(NVDEV_ENGINE_GR , gr);
_(NVDEV_ENGINE_IFB , ifb);
_(NVDEV_ENGINE_ME , me);
_(NVDEV_ENGINE_MPEG , mpeg);
_(NVDEV_ENGINE_MSENC , msenc);
_(NVDEV_ENGINE_MSPDEC , mspdec);
_(NVDEV_ENGINE_MSPPP , msppp);
_(NVDEV_ENGINE_MSVLD , msvld);
_(NVDEV_ENGINE_PM , pm);
_(NVDEV_ENGINE_SEC , sec);
_(NVDEV_ENGINE_SW , sw);
_(NVDEV_ENGINE_VIC , vic);
_(NVDEV_ENGINE_VP , vp);
default:
WARN_ON(1);
continue;
}
#undef _
/* note: can't init *any* subdevs until devinit has been run
* due to not knowing exactly what the vbios init tables will
* mess with. devinit also can't be run until all of its
* dependencies have been created.
*
* this code delays init of any subdev until all of devinit's
* dependencies have been created, and then initialises each
* subdev in turn as they're created.
*/
while (i >= NVDEV_SUBDEV_DEVINIT_LAST && c <= i) {
struct nvkm_object *subdev = device->subdev[c++];
if (subdev && !nv_iclass(subdev, NV_ENGINE_CLASS)) {
ret = nvkm_object_inc(subdev);
if (ret)
goto fail;
} else
if (subdev) {
nvkm_subdev_reset(subdev);
}
}
}
ret = 0;
fail:
for (--i; ret && i >= 0; i--) {
if ((subdev = device->subdev[i])) {
if (!nv_iclass(subdev, NV_ENGINE_CLASS))
nvkm_object_dec(subdev, false);
}
}
if (ret)
nvkm_acpi_fini(device, false);
return ret;
}
resource_size_t
nv_device_resource_start(struct nvkm_device *device, unsigned int bar)
{
if (nv_device_is_pci(device)) {
return pci_resource_start(device->pdev, bar);
} else {
struct resource *res;
res = platform_get_resource(device->platformdev,
IORESOURCE_MEM, bar);
if (!res)
return 0;
return res->start;
}
}
resource_size_t
nv_device_resource_len(struct nvkm_device *device, unsigned int bar)
{
if (nv_device_is_pci(device)) {
return pci_resource_len(device->pdev, bar);
} else {
struct resource *res;
res = platform_get_resource(device->platformdev,
IORESOURCE_MEM, bar);
if (!res)
return 0;
return resource_size(res);
}
}
int
nv_device_get_irq(struct nvkm_device *device, bool stall)
{
if (nv_device_is_pci(device)) {
return device->pdev->irq;
} else {
return platform_get_irq_byname(device->platformdev,
stall ? "stall" : "nonstall");
}
}
void
nvkm_device_del(struct nvkm_device **pdevice)
{
struct nvkm_device *device = *pdevice;
int i;
if (device) {
mutex_lock(&nv_devices_mutex);
for (i = NVDEV_SUBDEV_NR - 1; i >= 0; i--)
nvkm_object_ref(NULL, &device->subdev[i]);
nvkm_event_fini(&device->event);
if (device->pri)
iounmap(device->pri);
list_del(&device->head);
if (device->func->dtor)
*pdevice = device->func->dtor(device);
mutex_unlock(&nv_devices_mutex);
kfree(*pdevice);
*pdevice = NULL;
}
}
static const struct nvkm_engine_func
nvkm_device_func = {
};
drm/nouveau/core: pull in most of the new core infrastructure This commit provides most of the infrastructure to support a major overhaul of Nouveau's internals coming in the following commits. This work aims to take all the things we've learned over the last several years, and turn that into a cleaner architecture that's more maintainable going forward. RAMHT and MM bits of the new core have been left out for the moment, and will be pulled in as I go through the process of porting the code to become either subdev or engine modules. There are several main goals I wanted to achieve through this work: -- Reduce complexity The goal here was to make each component of the driver as independent as possible, which will ease maintainability and readability, and provide a good base for resetting locked up GPU units in the future. -- Better tracking of GPU units that are required at any given time This is for future PM work, we'll be able to tell exactly what parts of the GPU we need powered at any given point (etc). -- Expose all available NVIDIA GPUs to the client In order to support things such as multi-GPU channels, we want to be able to expose all the NVIDIA GPUs to the client over a single file descriptor so it can send a single push buffer to multiple GPUs. -- Untangle the core hardware support code from the DRM implementation This happened initially as an unexpected side-effect of developing the initial core infrastructure in userspace, but it turned into a goal of the whole project. Initial benefits will be the availablility of a number of userspace tools and tests using the same code as the driver itself, but will also be important as I look into some virtualisation ideas. v2: Ben Skeggs <bskeggs@redhat.com> - fix duplicate assignments noticed by clang - implement some forgotten yelling in error path - ensure 64-bit engine mask is used everywhere v3: Marcin Slusarz <marcin.slusarz@gmail.com> - sparse fixes - inline nv_printk into nv_assert to prevent recursive inlining issues v4: Ben Skeggs <bskeggs@redhat.com> - fixed minor memory leak on gpuobj destruction Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
2012-07-05 21:36:43 +00:00
int
nvkm_device_ctor(const struct nvkm_device_func *func,
const struct nvkm_device_quirk *quirk,
void *dev, enum nv_bus_type type, u64 handle,
const char *name, const char *cfg, const char *dbg,
bool detect, bool mmio, u64 subdev_mask,
struct nvkm_device *device)
drm/nouveau/core: pull in most of the new core infrastructure This commit provides most of the infrastructure to support a major overhaul of Nouveau's internals coming in the following commits. This work aims to take all the things we've learned over the last several years, and turn that into a cleaner architecture that's more maintainable going forward. RAMHT and MM bits of the new core have been left out for the moment, and will be pulled in as I go through the process of porting the code to become either subdev or engine modules. There are several main goals I wanted to achieve through this work: -- Reduce complexity The goal here was to make each component of the driver as independent as possible, which will ease maintainability and readability, and provide a good base for resetting locked up GPU units in the future. -- Better tracking of GPU units that are required at any given time This is for future PM work, we'll be able to tell exactly what parts of the GPU we need powered at any given point (etc). -- Expose all available NVIDIA GPUs to the client In order to support things such as multi-GPU channels, we want to be able to expose all the NVIDIA GPUs to the client over a single file descriptor so it can send a single push buffer to multiple GPUs. -- Untangle the core hardware support code from the DRM implementation This happened initially as an unexpected side-effect of developing the initial core infrastructure in userspace, but it turned into a goal of the whole project. Initial benefits will be the availablility of a number of userspace tools and tests using the same code as the driver itself, but will also be important as I look into some virtualisation ideas. v2: Ben Skeggs <bskeggs@redhat.com> - fix duplicate assignments noticed by clang - implement some forgotten yelling in error path - ensure 64-bit engine mask is used everywhere v3: Marcin Slusarz <marcin.slusarz@gmail.com> - sparse fixes - inline nv_printk into nv_assert to prevent recursive inlining issues v4: Ben Skeggs <bskeggs@redhat.com> - fixed minor memory leak on gpuobj destruction Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
2012-07-05 21:36:43 +00:00
{
u64 mmio_base, mmio_size;
u32 boot0, strap;
void __iomem *map;
drm/nouveau/core: pull in most of the new core infrastructure This commit provides most of the infrastructure to support a major overhaul of Nouveau's internals coming in the following commits. This work aims to take all the things we've learned over the last several years, and turn that into a cleaner architecture that's more maintainable going forward. RAMHT and MM bits of the new core have been left out for the moment, and will be pulled in as I go through the process of porting the code to become either subdev or engine modules. There are several main goals I wanted to achieve through this work: -- Reduce complexity The goal here was to make each component of the driver as independent as possible, which will ease maintainability and readability, and provide a good base for resetting locked up GPU units in the future. -- Better tracking of GPU units that are required at any given time This is for future PM work, we'll be able to tell exactly what parts of the GPU we need powered at any given point (etc). -- Expose all available NVIDIA GPUs to the client In order to support things such as multi-GPU channels, we want to be able to expose all the NVIDIA GPUs to the client over a single file descriptor so it can send a single push buffer to multiple GPUs. -- Untangle the core hardware support code from the DRM implementation This happened initially as an unexpected side-effect of developing the initial core infrastructure in userspace, but it turned into a goal of the whole project. Initial benefits will be the availablility of a number of userspace tools and tests using the same code as the driver itself, but will also be important as I look into some virtualisation ideas. v2: Ben Skeggs <bskeggs@redhat.com> - fix duplicate assignments noticed by clang - implement some forgotten yelling in error path - ensure 64-bit engine mask is used everywhere v3: Marcin Slusarz <marcin.slusarz@gmail.com> - sparse fixes - inline nv_printk into nv_assert to prevent recursive inlining issues v4: Ben Skeggs <bskeggs@redhat.com> - fixed minor memory leak on gpuobj destruction Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
2012-07-05 21:36:43 +00:00
int ret = -EEXIST;
int i;
drm/nouveau/core: pull in most of the new core infrastructure This commit provides most of the infrastructure to support a major overhaul of Nouveau's internals coming in the following commits. This work aims to take all the things we've learned over the last several years, and turn that into a cleaner architecture that's more maintainable going forward. RAMHT and MM bits of the new core have been left out for the moment, and will be pulled in as I go through the process of porting the code to become either subdev or engine modules. There are several main goals I wanted to achieve through this work: -- Reduce complexity The goal here was to make each component of the driver as independent as possible, which will ease maintainability and readability, and provide a good base for resetting locked up GPU units in the future. -- Better tracking of GPU units that are required at any given time This is for future PM work, we'll be able to tell exactly what parts of the GPU we need powered at any given point (etc). -- Expose all available NVIDIA GPUs to the client In order to support things such as multi-GPU channels, we want to be able to expose all the NVIDIA GPUs to the client over a single file descriptor so it can send a single push buffer to multiple GPUs. -- Untangle the core hardware support code from the DRM implementation This happened initially as an unexpected side-effect of developing the initial core infrastructure in userspace, but it turned into a goal of the whole project. Initial benefits will be the availablility of a number of userspace tools and tests using the same code as the driver itself, but will also be important as I look into some virtualisation ideas. v2: Ben Skeggs <bskeggs@redhat.com> - fix duplicate assignments noticed by clang - implement some forgotten yelling in error path - ensure 64-bit engine mask is used everywhere v3: Marcin Slusarz <marcin.slusarz@gmail.com> - sparse fixes - inline nv_printk into nv_assert to prevent recursive inlining issues v4: Ben Skeggs <bskeggs@redhat.com> - fixed minor memory leak on gpuobj destruction Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
2012-07-05 21:36:43 +00:00
mutex_lock(&nv_devices_mutex);
if (nvkm_device_find_locked(handle))
goto done;
drm/nouveau/core: pull in most of the new core infrastructure This commit provides most of the infrastructure to support a major overhaul of Nouveau's internals coming in the following commits. This work aims to take all the things we've learned over the last several years, and turn that into a cleaner architecture that's more maintainable going forward. RAMHT and MM bits of the new core have been left out for the moment, and will be pulled in as I go through the process of porting the code to become either subdev or engine modules. There are several main goals I wanted to achieve through this work: -- Reduce complexity The goal here was to make each component of the driver as independent as possible, which will ease maintainability and readability, and provide a good base for resetting locked up GPU units in the future. -- Better tracking of GPU units that are required at any given time This is for future PM work, we'll be able to tell exactly what parts of the GPU we need powered at any given point (etc). -- Expose all available NVIDIA GPUs to the client In order to support things such as multi-GPU channels, we want to be able to expose all the NVIDIA GPUs to the client over a single file descriptor so it can send a single push buffer to multiple GPUs. -- Untangle the core hardware support code from the DRM implementation This happened initially as an unexpected side-effect of developing the initial core infrastructure in userspace, but it turned into a goal of the whole project. Initial benefits will be the availablility of a number of userspace tools and tests using the same code as the driver itself, but will also be important as I look into some virtualisation ideas. v2: Ben Skeggs <bskeggs@redhat.com> - fix duplicate assignments noticed by clang - implement some forgotten yelling in error path - ensure 64-bit engine mask is used everywhere v3: Marcin Slusarz <marcin.slusarz@gmail.com> - sparse fixes - inline nv_printk into nv_assert to prevent recursive inlining issues v4: Ben Skeggs <bskeggs@redhat.com> - fixed minor memory leak on gpuobj destruction Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
2012-07-05 21:36:43 +00:00
ret = nvkm_engine_ctor(&nvkm_device_func, device, 0, 0,
true, &device->engine);
device->engine.subdev.object.parent = NULL;
device->func = func;
drm/nouveau/core: pull in most of the new core infrastructure This commit provides most of the infrastructure to support a major overhaul of Nouveau's internals coming in the following commits. This work aims to take all the things we've learned over the last several years, and turn that into a cleaner architecture that's more maintainable going forward. RAMHT and MM bits of the new core have been left out for the moment, and will be pulled in as I go through the process of porting the code to become either subdev or engine modules. There are several main goals I wanted to achieve through this work: -- Reduce complexity The goal here was to make each component of the driver as independent as possible, which will ease maintainability and readability, and provide a good base for resetting locked up GPU units in the future. -- Better tracking of GPU units that are required at any given time This is for future PM work, we'll be able to tell exactly what parts of the GPU we need powered at any given point (etc). -- Expose all available NVIDIA GPUs to the client In order to support things such as multi-GPU channels, we want to be able to expose all the NVIDIA GPUs to the client over a single file descriptor so it can send a single push buffer to multiple GPUs. -- Untangle the core hardware support code from the DRM implementation This happened initially as an unexpected side-effect of developing the initial core infrastructure in userspace, but it turned into a goal of the whole project. Initial benefits will be the availablility of a number of userspace tools and tests using the same code as the driver itself, but will also be important as I look into some virtualisation ideas. v2: Ben Skeggs <bskeggs@redhat.com> - fix duplicate assignments noticed by clang - implement some forgotten yelling in error path - ensure 64-bit engine mask is used everywhere v3: Marcin Slusarz <marcin.slusarz@gmail.com> - sparse fixes - inline nv_printk into nv_assert to prevent recursive inlining issues v4: Ben Skeggs <bskeggs@redhat.com> - fixed minor memory leak on gpuobj destruction Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
2012-07-05 21:36:43 +00:00
if (ret)
goto done;
device->quirk = quirk;
switch (type) {
case NVKM_BUS_PCI:
device->pdev = dev;
device->dev = &device->pdev->dev;
break;
case NVKM_BUS_PLATFORM:
device->platformdev = dev;
device->dev = &device->platformdev->dev;
break;
}
device->handle = handle;
drm/nouveau/core: pull in most of the new core infrastructure This commit provides most of the infrastructure to support a major overhaul of Nouveau's internals coming in the following commits. This work aims to take all the things we've learned over the last several years, and turn that into a cleaner architecture that's more maintainable going forward. RAMHT and MM bits of the new core have been left out for the moment, and will be pulled in as I go through the process of porting the code to become either subdev or engine modules. There are several main goals I wanted to achieve through this work: -- Reduce complexity The goal here was to make each component of the driver as independent as possible, which will ease maintainability and readability, and provide a good base for resetting locked up GPU units in the future. -- Better tracking of GPU units that are required at any given time This is for future PM work, we'll be able to tell exactly what parts of the GPU we need powered at any given point (etc). -- Expose all available NVIDIA GPUs to the client In order to support things such as multi-GPU channels, we want to be able to expose all the NVIDIA GPUs to the client over a single file descriptor so it can send a single push buffer to multiple GPUs. -- Untangle the core hardware support code from the DRM implementation This happened initially as an unexpected side-effect of developing the initial core infrastructure in userspace, but it turned into a goal of the whole project. Initial benefits will be the availablility of a number of userspace tools and tests using the same code as the driver itself, but will also be important as I look into some virtualisation ideas. v2: Ben Skeggs <bskeggs@redhat.com> - fix duplicate assignments noticed by clang - implement some forgotten yelling in error path - ensure 64-bit engine mask is used everywhere v3: Marcin Slusarz <marcin.slusarz@gmail.com> - sparse fixes - inline nv_printk into nv_assert to prevent recursive inlining issues v4: Ben Skeggs <bskeggs@redhat.com> - fixed minor memory leak on gpuobj destruction Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
2012-07-05 21:36:43 +00:00
device->cfgopt = cfg;
device->dbgopt = dbg;
device->name = name;
drm/nouveau/core: pull in most of the new core infrastructure This commit provides most of the infrastructure to support a major overhaul of Nouveau's internals coming in the following commits. This work aims to take all the things we've learned over the last several years, and turn that into a cleaner architecture that's more maintainable going forward. RAMHT and MM bits of the new core have been left out for the moment, and will be pulled in as I go through the process of porting the code to become either subdev or engine modules. There are several main goals I wanted to achieve through this work: -- Reduce complexity The goal here was to make each component of the driver as independent as possible, which will ease maintainability and readability, and provide a good base for resetting locked up GPU units in the future. -- Better tracking of GPU units that are required at any given time This is for future PM work, we'll be able to tell exactly what parts of the GPU we need powered at any given point (etc). -- Expose all available NVIDIA GPUs to the client In order to support things such as multi-GPU channels, we want to be able to expose all the NVIDIA GPUs to the client over a single file descriptor so it can send a single push buffer to multiple GPUs. -- Untangle the core hardware support code from the DRM implementation This happened initially as an unexpected side-effect of developing the initial core infrastructure in userspace, but it turned into a goal of the whole project. Initial benefits will be the availablility of a number of userspace tools and tests using the same code as the driver itself, but will also be important as I look into some virtualisation ideas. v2: Ben Skeggs <bskeggs@redhat.com> - fix duplicate assignments noticed by clang - implement some forgotten yelling in error path - ensure 64-bit engine mask is used everywhere v3: Marcin Slusarz <marcin.slusarz@gmail.com> - sparse fixes - inline nv_printk into nv_assert to prevent recursive inlining issues v4: Ben Skeggs <bskeggs@redhat.com> - fixed minor memory leak on gpuobj destruction Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
2012-07-05 21:36:43 +00:00
list_add_tail(&device->head, &nv_devices);
ret = nvkm_event_init(&nvkm_device_event_func, 1, 1, &device->event);
if (ret)
goto done;
mmio_base = nv_device_resource_start(device, 0);
mmio_size = nv_device_resource_len(device, 0);
/* identify the chipset, and determine classes of subdev/engines */
if (detect) {
map = ioremap(mmio_base, 0x102000);
if (ret = -ENOMEM, map == NULL)
goto done;
/* switch mmio to cpu's native endianness */
#ifndef __BIG_ENDIAN
if (ioread32_native(map + 0x000004) != 0x00000000) {
#else
if (ioread32_native(map + 0x000004) == 0x00000000) {
#endif
iowrite32_native(0x01000001, map + 0x000004);
ioread32_native(map);
}
/* read boot0 and strapping information */
boot0 = ioread32_native(map + 0x000000);
strap = ioread32_native(map + 0x101000);
iounmap(map);
/* determine chipset and derive architecture from it */
if ((boot0 & 0x1f000000) > 0) {
device->chipset = (boot0 & 0x1ff00000) >> 20;
device->chiprev = (boot0 & 0x000000ff);
switch (device->chipset & 0x1f0) {
case 0x010: {
if (0x461 & (1 << (device->chipset & 0xf)))
device->card_type = NV_10;
else
device->card_type = NV_11;
device->chiprev = 0x00;
break;
}
case 0x020: device->card_type = NV_20; break;
case 0x030: device->card_type = NV_30; break;
case 0x040:
case 0x060: device->card_type = NV_40; break;
case 0x050:
case 0x080:
case 0x090:
case 0x0a0: device->card_type = NV_50; break;
case 0x0c0:
case 0x0d0: device->card_type = NV_C0; break;
case 0x0e0:
case 0x0f0:
case 0x100: device->card_type = NV_E0; break;
case 0x110:
case 0x120: device->card_type = GM100; break;
default:
break;
}
} else
if ((boot0 & 0xff00fff0) == 0x20004000) {
if (boot0 & 0x00f00000)
device->chipset = 0x05;
else
device->chipset = 0x04;
device->card_type = NV_04;
}
switch (device->card_type) {
case NV_04: ret = nv04_identify(device); break;
case NV_10:
case NV_11: ret = nv10_identify(device); break;
case NV_20: ret = nv20_identify(device); break;
case NV_30: ret = nv30_identify(device); break;
case NV_40: ret = nv40_identify(device); break;
case NV_50: ret = nv50_identify(device); break;
case NV_C0: ret = gf100_identify(device); break;
case NV_E0: ret = gk104_identify(device); break;
case GM100: ret = gm100_identify(device); break;
default:
ret = -EINVAL;
break;
}
if (ret) {
nvdev_error(device, "unknown chipset (%08x)\n", boot0);
goto done;
}
nvdev_info(device, "NVIDIA %s (%08x)\n", device->cname, boot0);
/* determine frequency of timing crystal */
if ( device->card_type <= NV_10 || device->chipset < 0x17 ||
(device->chipset >= 0x20 && device->chipset < 0x25))
strap &= 0x00000040;
else
strap &= 0x00400040;
switch (strap) {
case 0x00000000: device->crystal = 13500; break;
case 0x00000040: device->crystal = 14318; break;
case 0x00400000: device->crystal = 27000; break;
case 0x00400040: device->crystal = 25000; break;
}
} else {
device->cname = "NULL";
device->oclass[NVDEV_SUBDEV_VBIOS] = &nvkm_bios_oclass;
}
if (mmio) {
device->pri = ioremap(mmio_base, mmio_size);
if (!device->pri) {
nvdev_error(device, "unable to map PRI\n");
return -ENOMEM;
}
}
/* disable subdevs that aren't required (used by tools) */
for (i = 0; i < NVDEV_SUBDEV_NR; i++) {
if (!(subdev_mask & (1ULL << i)))
device->oclass[i] = NULL;
}
atomic_set(&device->engine.subdev.object.usecount, 2);
mutex_init(&device->mutex);
drm/nouveau/core: pull in most of the new core infrastructure This commit provides most of the infrastructure to support a major overhaul of Nouveau's internals coming in the following commits. This work aims to take all the things we've learned over the last several years, and turn that into a cleaner architecture that's more maintainable going forward. RAMHT and MM bits of the new core have been left out for the moment, and will be pulled in as I go through the process of porting the code to become either subdev or engine modules. There are several main goals I wanted to achieve through this work: -- Reduce complexity The goal here was to make each component of the driver as independent as possible, which will ease maintainability and readability, and provide a good base for resetting locked up GPU units in the future. -- Better tracking of GPU units that are required at any given time This is for future PM work, we'll be able to tell exactly what parts of the GPU we need powered at any given point (etc). -- Expose all available NVIDIA GPUs to the client In order to support things such as multi-GPU channels, we want to be able to expose all the NVIDIA GPUs to the client over a single file descriptor so it can send a single push buffer to multiple GPUs. -- Untangle the core hardware support code from the DRM implementation This happened initially as an unexpected side-effect of developing the initial core infrastructure in userspace, but it turned into a goal of the whole project. Initial benefits will be the availablility of a number of userspace tools and tests using the same code as the driver itself, but will also be important as I look into some virtualisation ideas. v2: Ben Skeggs <bskeggs@redhat.com> - fix duplicate assignments noticed by clang - implement some forgotten yelling in error path - ensure 64-bit engine mask is used everywhere v3: Marcin Slusarz <marcin.slusarz@gmail.com> - sparse fixes - inline nv_printk into nv_assert to prevent recursive inlining issues v4: Ben Skeggs <bskeggs@redhat.com> - fixed minor memory leak on gpuobj destruction Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
2012-07-05 21:36:43 +00:00
done:
mutex_unlock(&nv_devices_mutex);
return ret;
}