summaryrefslogtreecommitdiff
path: root/drivers/mxc/gpu-viv/hal/os/linux/kernel/gc_hal_kernel_driver.c
diff options
context:
space:
mode:
Diffstat (limited to 'drivers/mxc/gpu-viv/hal/os/linux/kernel/gc_hal_kernel_driver.c')
-rw-r--r--drivers/mxc/gpu-viv/hal/os/linux/kernel/gc_hal_kernel_driver.c1297
1 files changed, 1297 insertions, 0 deletions
diff --git a/drivers/mxc/gpu-viv/hal/os/linux/kernel/gc_hal_kernel_driver.c b/drivers/mxc/gpu-viv/hal/os/linux/kernel/gc_hal_kernel_driver.c
new file mode 100644
index 000000000000..724e77100205
--- /dev/null
+++ b/drivers/mxc/gpu-viv/hal/os/linux/kernel/gc_hal_kernel_driver.c
@@ -0,0 +1,1297 @@
+/****************************************************************************
+*
+* The MIT License (MIT)
+*
+* Copyright (c) 2014 - 2018 Vivante Corporation
+*
+* 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
+* AUTHORS OR COPYRIGHT HOLDERS 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.
+*
+*****************************************************************************
+*
+* The GPL License (GPL)
+*
+* Copyright (C) 2014 - 2018 Vivante Corporation
+*
+* This program is free software; you can redistribute it and/or
+* modify it under the terms of the GNU General Public License
+* as published by the Free Software Foundation; either version 2
+* of the License, or (at your option) any later version.
+*
+* This program is distributed in the hope that it will be useful,
+* but WITHOUT ANY WARRANTY; without even the implied warranty of
+* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
+* GNU General Public License for more details.
+*
+* You should have received a copy of the GNU General Public License
+* along with this program; if not, write to the Free Software Foundation,
+* Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA.
+*
+*****************************************************************************
+*
+* Note: This software is released under dual MIT and GPL licenses. A
+* recipient may use this file under the terms of either the MIT license or
+* GPL License. If you wish to use only one license not the other, you can
+* indicate your decision by deleting one of the above license notices in your
+* version of this file.
+*
+*****************************************************************************/
+
+
+#include <linux/device.h>
+#include <linux/slab.h>
+#include <linux/miscdevice.h>
+#include <linux/uaccess.h>
+
+#include "gc_hal_kernel_linux.h"
+#include "gc_hal_driver.h"
+
+#include <linux/platform_device.h>
+
+/* Zone used for header/footer. */
+#define _GC_OBJ_ZONE gcvZONE_DRIVER
+
+MODULE_DESCRIPTION("Vivante Graphics Driver");
+MODULE_LICENSE("Dual MIT/GPL");
+
+/* Disable MSI for internal FPGA build except PPC */
+#if gcdFPGA_BUILD && !defined(CONFIG_PPC)
+#define USE_MSI 0
+#else
+#define USE_MSI 1
+#endif
+
+static struct class* gpuClass;
+
+static gcsPLATFORM *platform;
+
+static gckGALDEVICE galDevice;
+
+static uint major = 199;
+module_param(major, uint, 0644);
+MODULE_PARM_DESC(major, "major device number for GC device");
+
+static int irqLine = -1;
+module_param(irqLine, int, 0644);
+MODULE_PARM_DESC(irqLine, "IRQ number of GC core");
+
+static ulong registerMemBase = 0x80000000;
+module_param(registerMemBase, ulong, 0644);
+MODULE_PARM_DESC(registerMemBase, "Base of bus address of GC core AHB register");
+
+static ulong registerMemSize = 2 << 10;
+module_param(registerMemSize, ulong, 0644);
+MODULE_PARM_DESC(registerMemSize, "Size of bus address range of GC core AHB register");
+
+static int irqLine2D = -1;
+module_param(irqLine2D, int, 0644);
+MODULE_PARM_DESC(irqLine2D, "IRQ number of G2D core if irqLine is used for a G3D core");
+
+static ulong registerMemBase2D = 0x00000000;
+module_param(registerMemBase2D, ulong, 0644);
+MODULE_PARM_DESC(registerMemBase2D, "Base of bus address of G2D core if registerMemBase2D is used for a G3D core");
+
+static ulong registerMemSize2D = 2 << 10;
+module_param(registerMemSize2D, ulong, 0644);
+MODULE_PARM_DESC(registerMemSize2D, "Size of bus address range of G2D core if registerMemSize is used for a G3D core");
+
+static int irqLineVG = -1;
+module_param(irqLineVG, int, 0644);
+MODULE_PARM_DESC(irqLineVG, "IRQ number of VG core");
+
+static ulong registerMemBaseVG = 0x00000000;
+module_param(registerMemBaseVG, ulong, 0644);
+MODULE_PARM_DESC(registerMemBaseVG, "Base of bus address of VG core");
+
+static ulong registerMemSizeVG = 2 << 10;
+module_param(registerMemSizeVG, ulong, 0644);
+MODULE_PARM_DESC(registerMemSizeVG, "Size of bus address range of VG core");
+
+#if gcdDEC_ENABLE_AHB
+static ulong registerMemBaseDEC300 = 0x00000000;
+module_param(registerMemBaseDEC300, ulong, 0644);
+
+static ulong registerMemSizeDEC300 = 2 << 10;
+module_param(registerMemSizeDEC300, ulong, 0644);
+#endif
+
+#ifndef gcdDEFAULT_CONTIGUOUS_SIZE
+#define gcdDEFAULT_CONTIGUOUS_SIZE (4 << 20)
+#endif
+static ulong contiguousSize = gcdDEFAULT_CONTIGUOUS_SIZE;
+module_param(contiguousSize, ulong, 0644);
+MODULE_PARM_DESC(contiguousSize, "Size of memory reserved for GC");
+
+static ulong contiguousBase = 0;
+module_param(contiguousBase, ulong, 0644);
+MODULE_PARM_DESC(contiguousBase, "Base address of memory reserved for GC, if it is 0, GC driver will try to allocate a buffer whose size defined by contiguousSize");
+
+static ulong externalSize = 0;
+module_param(externalSize, ulong, 0644);
+MODULE_PARM_DESC(externalSize, "Size of external memory, if it is 0, means there is no external pool");
+
+static ulong externalBase = 0;
+module_param(externalBase, ulong, 0644);
+MODULE_PARM_DESC(externalBase, "Base address of external memory");
+
+static int fastClear = -1;
+module_param(fastClear, int, 0644);
+MODULE_PARM_DESC(fastClear, "Disable fast clear if set it to 0, enabled by default");
+
+static int compression = -1;
+module_param(compression, int, 0644);
+MODULE_PARM_DESC(compression, "Disable compression if set it to 0, enabled by default");
+
+static int powerManagement = 1;
+module_param(powerManagement, int, 0644);
+MODULE_PARM_DESC(powerManagement, "Disable auto power saving if set it to 1, enabled by default");
+
+static int gpuProfiler = 0;
+module_param(gpuProfiler, int, 0644);
+MODULE_PARM_DESC(gpuProfiler, "Enable profiling support, disabled by default");
+
+static ulong baseAddress = 0;
+module_param(baseAddress, ulong, 0644);
+MODULE_PARM_DESC(baseAddress, "Only used for old MMU, set it to 0 if memory which can be accessed by GPU falls into 0 - 2G, otherwise set it to 0x80000000");
+
+static ulong physSize = 0;
+module_param(physSize, ulong, 0644);
+MODULE_PARM_DESC(physSize, "Obsolete");
+
+static uint logFileSize = 0;
+module_param(logFileSize,uint, 0644);
+MODULE_PARM_DESC(logFileSize, "Size of buffer to store GC driver output messsage, if it is not 0, message is read from /sys/kernel/debug/gc/galcore_trace, default value is 0");
+
+static uint recovery = 0;
+module_param(recovery, uint, 0644);
+MODULE_PARM_DESC(recovery, "Recover GPU from stuck (1: Enable, 0: Disable)");
+
+/* Middle needs about 40KB buffer, Maximal may need more than 200KB buffer. */
+static uint stuckDump = 0;
+module_param(stuckDump, uint, 0644);
+MODULE_PARM_DESC(stuckDump, "Level of stuck dump content (1: Minimal, 2: Middle, 3: Maximal)");
+
+static int showArgs = 0;
+module_param(showArgs, int, 0644);
+MODULE_PARM_DESC(showArgs, "Display parameters value when driver loaded");
+
+static int mmu = 1;
+module_param(mmu, int, 0644);
+MODULE_PARM_DESC(mmu, "Disable MMU if set it to 0, enabled by default");
+
+static int irqs[gcvCORE_COUNT] = {[0 ... gcvCORE_COUNT - 1] = -1};
+module_param_array(irqs, int, NULL, 0644);
+MODULE_PARM_DESC(irqs, "Array of IRQ numbers of multi-GPU");
+
+static uint registerBases[gcvCORE_COUNT];
+module_param_array(registerBases, uint, NULL, 0644);
+MODULE_PARM_DESC(registerBases, "Array of bases of bus address of register of multi-GPU");
+
+static uint registerSizes[gcvCORE_COUNT] = {[0 ... gcvCORE_COUNT - 1] = 2 << 10};
+module_param_array(registerSizes, uint, NULL, 0644);
+MODULE_PARM_DESC(registerSizes, "Array of sizes of bus address range of register of multi-GPU");
+
+static uint chipIDs[gcvCORE_COUNT] = {[0 ... gcvCORE_COUNT - 1] = gcvCHIP_ID_DEFAULT};
+module_param_array(chipIDs, uint, NULL, 0644);
+MODULE_PARM_DESC(chipIDs, "Array of chipIDs of multi-GPU");
+
+static uint type = 0;
+module_param(type, uint, 0664);
+MODULE_PARM_DESC(type, "0 - Char Driver (Default), 1 - Misc Driver");
+
+static int gpu3DMinClock = 1;
+
+static int contiguousRequested = 0;
+
+static gctBOOL registerMemMapped = gcvFALSE;
+static gctPOINTER registerMemAddress = gcvNULL;
+static ulong bankSize = 0;
+static int signal = 48;
+
+void
+_UpdateModuleParam(
+ gcsMODULE_PARAMETERS *Param
+ )
+{
+ irqLine = Param->irqLine ;
+ registerMemBase = Param->registerMemBase;
+ registerMemSize = Param->registerMemSize;
+ irqLine2D = Param->irqLine2D ;
+ registerMemBase2D = Param->registerMemBase2D;
+ registerMemSize2D = Param->registerMemSize2D;
+#if gcdENABLE_VG
+ irqLineVG = Param->irqLineVG;
+ registerMemBaseVG = Param->registerMemBaseVG;
+ registerMemSizeVG = Param->registerMemSizeVG;
+#endif
+ contiguousSize = Param->contiguousSize;
+ contiguousBase = Param->contiguousBase;
+ externalSize = Param->externalSize;
+ externalBase = Param->externalBase;
+ bankSize = Param->bankSize;
+ fastClear = Param->fastClear;
+ compression = (gctINT)Param->compression;
+ powerManagement = Param->powerManagement;
+ gpuProfiler = Param->gpuProfiler;
+ signal = Param->signal;
+ baseAddress = Param->baseAddress;
+ physSize = Param->physSize;
+ logFileSize = Param->logFileSize;
+ recovery = Param->recovery;
+ stuckDump = Param->stuckDump;
+ showArgs = Param->showArgs;
+ contiguousRequested = Param->contiguousRequested;
+ gpu3DMinClock = Param->gpu3DMinClock;
+ registerMemMapped = Param->registerMemMapped;
+ registerMemAddress = Param->registerMemAddress;
+
+ memcpy(irqs, Param->irqs, gcmSIZEOF(gctINT) * gcvCORE_COUNT);
+ memcpy(registerBases, Param->registerBases, gcmSIZEOF(gctUINT) * gcvCORE_COUNT);
+ memcpy(registerSizes, Param->registerSizes, gcmSIZEOF(gctUINT) * gcvCORE_COUNT);
+ memcpy(chipIDs, Param->chipIDs, gcmSIZEOF(gctUINT) * gcvCORE_COUNT);
+}
+
+void
+gckOS_DumpParam(
+ void
+ )
+{
+ gctINT i;
+
+ printk("Galcore options:\n");
+ if (irqLine != -1)
+ {
+ printk(" irqLine = %d\n", irqLine);
+ printk(" registerMemBase = 0x%08lX\n", registerMemBase);
+ printk(" registerMemSize = 0x%08lX\n", registerMemSize);
+ }
+
+ if (irqLine2D != -1)
+ {
+ printk(" irqLine2D = %d\n", irqLine2D);
+ printk(" registerMemBase2D = 0x%08lX\n", registerMemBase2D);
+ printk(" registerMemSize2D = 0x%08lX\n", registerMemSize2D);
+ }
+
+ if (irqLineVG != -1)
+ {
+ printk(" irqLineVG = %d\n", irqLineVG);
+ printk(" registerMemBaseVG = 0x%08lX\n", registerMemBaseVG);
+ printk(" registerMemSizeVG = 0x%08lX\n", registerMemSizeVG);
+ }
+
+#if gcdDEC_ENABLE_AHB
+ printk(" registerMemBaseDEC300 = 0x%08lX\n", registerMemBaseDEC300);
+ printk(" registerMemSizeDEC300 = 0x%08lX\n", registerMemSizeDEC300);
+#endif
+
+ printk(" contiguousSize = 0x%08lX\n", contiguousSize);
+ printk(" contiguousBase = 0x%08lX\n", contiguousBase);
+ printk(" externalSize = 0x%08lX\n", externalSize);
+ printk(" externalBase = 0x%08lX\n", externalBase);
+ printk(" bankSize = 0x%08lX\n", bankSize);
+ printk(" fastClear = %d\n", fastClear);
+ printk(" compression = %d\n", compression);
+ printk(" signal = %d\n", signal);
+ printk(" powerManagement = %d\n", powerManagement);
+ printk(" baseAddress = 0x%08lX\n", baseAddress);
+ printk(" physSize = 0x%08lX\n", physSize);
+ printk(" logFileSize = %d KB \n", logFileSize);
+ printk(" recovery = %d\n", recovery);
+ printk(" stuckDump = %d\n", stuckDump);
+ printk(" gpuProfiler = %d\n", gpuProfiler);
+
+ printk(" irqs = ");
+ for (i = 0; i < gcvCORE_COUNT; i++)
+ {
+ printk("%d, ", irqs[i]);
+ }
+ printk("\n");
+
+ printk(" registerBases = ");
+ for (i = 0; i < gcvCORE_COUNT; i++)
+ {
+ printk("0x%08X, ", registerBases[i]);
+ }
+ printk("\n");
+
+ printk(" registerSizes = ");
+ for (i = 0; i < gcvCORE_COUNT; i++)
+ {
+ printk("0x%08X, ", registerSizes[i]);
+ }
+ printk("\n");
+
+ printk(" chipIDs = ");
+ for (i = 0; i < gcvCORE_COUNT; i++)
+ {
+ printk("0x%08X, ", chipIDs[i]);
+ }
+ printk("\n");
+
+ printk("Build options:\n");
+ printk(" gcdGPU_TIMEOUT = %d\n", gcdGPU_TIMEOUT);
+ printk(" gcdGPU_2D_TIMEOUT = %d\n", gcdGPU_2D_TIMEOUT);
+ printk(" gcdINTERRUPT_STATISTIC = %d\n", gcdINTERRUPT_STATISTIC);
+}
+
+static int drv_open(
+ struct inode* inode,
+ struct file* filp
+ )
+{
+ gceSTATUS status;
+ gctBOOL attached = gcvFALSE;
+ gcsHAL_PRIVATE_DATA_PTR data = gcvNULL;
+ gctINT i;
+
+ gcmkHEADER_ARG("inode=0x%08X filp=0x%08X", inode, filp);
+
+ if (filp == gcvNULL)
+ {
+ gcmkTRACE_ZONE(
+ gcvLEVEL_ERROR, gcvZONE_DRIVER,
+ "%s(%d): filp is NULL\n",
+ __FUNCTION__, __LINE__
+ );
+
+ gcmkONERROR(gcvSTATUS_INVALID_ARGUMENT);
+ }
+
+ data = kmalloc(sizeof(gcsHAL_PRIVATE_DATA), GFP_KERNEL | __GFP_NOWARN);
+
+ if (data == gcvNULL)
+ {
+ gcmkTRACE_ZONE(
+ gcvLEVEL_ERROR, gcvZONE_DRIVER,
+ "%s(%d): private_data is NULL\n",
+ __FUNCTION__, __LINE__
+ );
+
+ gcmkONERROR(gcvSTATUS_OUT_OF_MEMORY);
+ }
+
+ data->device = galDevice;
+ data->pidOpen = _GetProcessID();
+
+ /* Attached the process. */
+ for (i = 0; i < gcdMAX_GPU_COUNT; i++)
+ {
+ if (galDevice->kernels[i] != gcvNULL)
+ {
+ gcmkONERROR(gckKERNEL_AttachProcess(galDevice->kernels[i], gcvTRUE));
+ }
+ }
+ attached = gcvTRUE;
+
+ filp->private_data = data;
+
+ /* Success. */
+ gcmkFOOTER_NO();
+ return 0;
+
+OnError:
+ if (data != gcvNULL)
+ {
+ kfree(data);
+ }
+
+ if (attached)
+ {
+ for (i = 0; i < gcdMAX_GPU_COUNT; i++)
+ {
+ if (galDevice->kernels[i] != gcvNULL)
+ {
+ gcmkVERIFY_OK(gckKERNEL_AttachProcess(galDevice->kernels[i], gcvFALSE));
+ }
+ }
+ }
+
+ gcmkFOOTER();
+ return -ENOTTY;
+}
+
+static int drv_release(
+ struct inode* inode,
+ struct file* filp
+ )
+{
+ gceSTATUS status;
+ gcsHAL_PRIVATE_DATA_PTR data;
+ gckGALDEVICE device;
+ gctINT i;
+
+ gcmkHEADER_ARG("inode=0x%08X filp=0x%08X", inode, filp);
+
+ if (filp == gcvNULL)
+ {
+ gcmkTRACE_ZONE(
+ gcvLEVEL_ERROR, gcvZONE_DRIVER,
+ "%s(%d): filp is NULL\n",
+ __FUNCTION__, __LINE__
+ );
+
+ gcmkONERROR(gcvSTATUS_INVALID_ARGUMENT);
+ }
+
+ data = filp->private_data;
+
+ if (data == gcvNULL)
+ {
+ gcmkTRACE_ZONE(
+ gcvLEVEL_ERROR, gcvZONE_DRIVER,
+ "%s(%d): private_data is NULL\n",
+ __FUNCTION__, __LINE__
+ );
+
+ gcmkONERROR(gcvSTATUS_INVALID_ARGUMENT);
+ }
+
+ device = data->device;
+
+ if (device == gcvNULL)
+ {
+ gcmkTRACE_ZONE(
+ gcvLEVEL_ERROR, gcvZONE_DRIVER,
+ "%s(%d): device is NULL\n",
+ __FUNCTION__, __LINE__
+ );
+
+ gcmkONERROR(gcvSTATUS_INVALID_ARGUMENT);
+ }
+
+ /* A process gets detached. */
+ for (i = 0; i < gcdMAX_GPU_COUNT; i++)
+ {
+ if (galDevice->kernels[i] != gcvNULL)
+ {
+ gcmkONERROR(gckKERNEL_AttachProcessEx(galDevice->kernels[i], gcvFALSE, data->pidOpen));
+ }
+ }
+
+ kfree(data);
+ filp->private_data = NULL;
+
+ /* Success. */
+ gcmkFOOTER_NO();
+ return 0;
+
+OnError:
+ gcmkFOOTER();
+ return -ENOTTY;
+}
+
+static long drv_ioctl(
+ struct file* filp,
+ unsigned int ioctlCode,
+ unsigned long arg
+ )
+{
+ gceSTATUS status;
+ gcsHAL_INTERFACE iface;
+ gctUINT32 copyLen;
+ DRIVER_ARGS drvArgs;
+ gckGALDEVICE device;
+ gcsHAL_PRIVATE_DATA_PTR data;
+
+ gcmkHEADER_ARG(
+ "filp=0x%08X ioctlCode=0x%08X arg=0x%08X",
+ filp, ioctlCode, arg
+ );
+
+ if (filp == gcvNULL)
+ {
+ gcmkTRACE_ZONE(
+ gcvLEVEL_ERROR, gcvZONE_DRIVER,
+ "%s(%d): filp is NULL\n",
+ __FUNCTION__, __LINE__
+ );
+
+ gcmkONERROR(gcvSTATUS_INVALID_ARGUMENT);
+ }
+
+ data = filp->private_data;
+
+ if (data == gcvNULL)
+ {
+ gcmkTRACE_ZONE(
+ gcvLEVEL_ERROR, gcvZONE_DRIVER,
+ "%s(%d): private_data is NULL\n",
+ __FUNCTION__, __LINE__
+ );
+
+ gcmkONERROR(gcvSTATUS_INVALID_ARGUMENT);
+ }
+
+ device = data->device;
+
+ if (device == gcvNULL)
+ {
+ gcmkTRACE_ZONE(
+ gcvLEVEL_ERROR, gcvZONE_DRIVER,
+ "%s(%d): device is NULL\n",
+ __FUNCTION__, __LINE__
+ );
+
+ gcmkONERROR(gcvSTATUS_INVALID_ARGUMENT);
+ }
+
+ if ((ioctlCode != IOCTL_GCHAL_INTERFACE)
+ && (ioctlCode != IOCTL_GCHAL_KERNEL_INTERFACE)
+ )
+ {
+ gcmkTRACE_ZONE(
+ gcvLEVEL_ERROR, gcvZONE_DRIVER,
+ "%s(%d): unknown command %d\n",
+ __FUNCTION__, __LINE__,
+ ioctlCode
+ );
+
+ gcmkONERROR(gcvSTATUS_INVALID_ARGUMENT);
+ }
+
+ /* Get the drvArgs. */
+ copyLen = copy_from_user(
+ &drvArgs, (void *) arg, sizeof(DRIVER_ARGS)
+ );
+
+ if (copyLen != 0)
+ {
+ gcmkTRACE_ZONE(
+ gcvLEVEL_ERROR, gcvZONE_DRIVER,
+ "%s(%d): error copying of the input arguments.\n",
+ __FUNCTION__, __LINE__
+ );
+
+ gcmkONERROR(gcvSTATUS_INVALID_ARGUMENT);
+ }
+
+ /* Now bring in the gcsHAL_INTERFACE structure. */
+ if ((drvArgs.InputBufferSize != sizeof(gcsHAL_INTERFACE))
+ || (drvArgs.OutputBufferSize != sizeof(gcsHAL_INTERFACE))
+ )
+ {
+ gcmkTRACE_ZONE(
+ gcvLEVEL_ERROR, gcvZONE_DRIVER,
+ "%s(%d): input or/and output structures are invalid.\n",
+ __FUNCTION__, __LINE__
+ );
+
+ gcmkONERROR(gcvSTATUS_INVALID_ARGUMENT);
+ }
+
+ copyLen = copy_from_user(
+ &iface, gcmUINT64_TO_PTR(drvArgs.InputBuffer), sizeof(gcsHAL_INTERFACE)
+ );
+
+ if (copyLen != 0)
+ {
+ gcmkTRACE_ZONE(
+ gcvLEVEL_ERROR, gcvZONE_DRIVER,
+ "%s(%d): error copying of input HAL interface.\n",
+ __FUNCTION__, __LINE__
+ );
+
+ gcmkONERROR(gcvSTATUS_INVALID_ARGUMENT);
+ }
+
+ status = gckDEVICE_Dispatch(device->device, &iface);
+
+ /* Redo system call after pending signal is handled. */
+ if (status == gcvSTATUS_INTERRUPTED)
+ {
+ gcmkFOOTER();
+ return -ERESTARTSYS;
+ }
+
+ /* Copy data back to the user. */
+ copyLen = copy_to_user(
+ gcmUINT64_TO_PTR(drvArgs.OutputBuffer), &iface, sizeof(gcsHAL_INTERFACE)
+ );
+
+ if (copyLen != 0)
+ {
+ gcmkTRACE_ZONE(
+ gcvLEVEL_ERROR, gcvZONE_DRIVER,
+ "%s(%d): error copying of output HAL interface.\n",
+ __FUNCTION__, __LINE__
+ );
+
+ gcmkONERROR(gcvSTATUS_INVALID_ARGUMENT);
+ }
+
+ /* Success. */
+ gcmkFOOTER_NO();
+ return 0;
+
+OnError:
+ gcmkFOOTER();
+ return -ENOTTY;
+}
+
+static struct file_operations driver_fops =
+{
+ .owner = THIS_MODULE,
+ .open = drv_open,
+ .release = drv_release,
+ .unlocked_ioctl = drv_ioctl,
+#ifdef HAVE_COMPAT_IOCTL
+ .compat_ioctl = drv_ioctl,
+#endif
+};
+
+static struct miscdevice gal_device = {
+ .minor = MISC_DYNAMIC_MINOR,
+ .name = DEVICE_NAME,
+ .fops = &driver_fops,
+};
+
+static int drv_init(void)
+{
+ int ret;
+ int result = -EINVAL;
+ gceSTATUS status;
+ gckGALDEVICE device = gcvNULL;
+ struct class* device_class = gcvNULL;
+
+ gcsDEVICE_CONSTRUCT_ARGS args = {
+ .recovery = recovery,
+ .stuckDump = stuckDump,
+ .gpu3DMinClock = gpu3DMinClock,
+ .contiguousRequested = contiguousRequested,
+ .platform = platform,
+ .mmu = mmu,
+ .registerMemMapped = registerMemMapped,
+ .registerMemAddress = registerMemAddress,
+#if gcdDEC_ENABLE_AHB
+ .registerMemBaseDEC300 = registerMemBaseDEC300,
+ .registerMemSizeDEC300 = registerMemSizeDEC300,
+#endif
+ };
+
+ gcmkHEADER();
+
+ memcpy(args.irqs, irqs, gcmSIZEOF(gctINT) * gcvCORE_COUNT);
+ memcpy(args.registerBases, registerBases, gcmSIZEOF(gctUINT) * gcvCORE_COUNT);
+ memcpy(args.registerSizes, registerSizes, gcmSIZEOF(gctUINT) * gcvCORE_COUNT);
+ memcpy(args.chipIDs, chipIDs, gcmSIZEOF(gctUINT) * gcvCORE_COUNT);
+
+ printk(KERN_INFO "Galcore version %d.%d.%d.%d\n",
+ gcvVERSION_MAJOR, gcvVERSION_MINOR, gcvVERSION_PATCH, gcvVERSION_BUILD);
+
+ args.powerManagement = powerManagement;
+ args.gpuProfiler = gpuProfiler;
+
+ if (showArgs)
+ {
+ gckOS_DumpParam();
+ }
+
+ if (logFileSize != 0)
+ {
+ gckDEBUGFS_Initialize();
+ }
+
+ /* Create the GAL device. */
+ status = gckGALDEVICE_Construct(
+ irqLine,
+ registerMemBase, registerMemSize,
+ irqLine2D,
+ registerMemBase2D, registerMemSize2D,
+ irqLineVG,
+ registerMemBaseVG, registerMemSizeVG,
+ contiguousBase, contiguousSize,
+ externalBase, externalSize,
+ bankSize, fastClear, compression, baseAddress, physSize, signal,
+ logFileSize,
+ powerManagement,
+ gpuProfiler,
+ &args,
+ &device
+ );
+
+ if (gcmIS_ERROR(status))
+ {
+ gcmkTRACE_ZONE(gcvLEVEL_ERROR, gcvZONE_DRIVER,
+ "%s(%d): Failed to create the GAL device: status=%d\n",
+ __FUNCTION__, __LINE__, status);
+
+ goto OnError;
+ }
+
+ /* Start the GAL device. */
+ gcmkONERROR(gckGALDEVICE_Start(device));
+
+ if ((physSize != 0)
+ && (device->kernels[gcvCORE_MAJOR] != gcvNULL)
+ && (device->kernels[gcvCORE_MAJOR]->hardware->mmuVersion != 0))
+ {
+ /* Reset the base address */
+ device->baseAddress = 0;
+ }
+
+ /* Set global galDevice pointer. */
+ galDevice = device;
+
+ if (type == 1)
+ {
+ /* Register as misc driver. */
+ ret = misc_register(&gal_device);
+
+ if (ret < 0)
+ {
+ gcmkTRACE_ZONE(
+ gcvLEVEL_ERROR, gcvZONE_DRIVER,
+ "%s(%d): misc_register fails.\n",
+ __FUNCTION__, __LINE__
+ );
+
+ gcmkONERROR(gcvSTATUS_OUT_OF_MEMORY);
+ }
+ }
+ else
+ {
+ /* Register the character device. */
+ ret = register_chrdev(major, DEVICE_NAME, &driver_fops);
+
+ if (ret < 0)
+ {
+ gcmkTRACE_ZONE(
+ gcvLEVEL_ERROR, gcvZONE_DRIVER,
+ "%s(%d): Could not allocate major number for mmap.\n",
+ __FUNCTION__, __LINE__
+ );
+
+ gcmkONERROR(gcvSTATUS_OUT_OF_MEMORY);
+ }
+
+ if (major == 0)
+ {
+ major = ret;
+ }
+
+ /* Create the device class. */
+ device_class = class_create(THIS_MODULE, CLASS_NAME);
+
+ if (IS_ERR(device_class))
+ {
+ gcmkTRACE_ZONE(
+ gcvLEVEL_ERROR, gcvZONE_DRIVER,
+ "%s(%d): Failed to create the class.\n",
+ __FUNCTION__, __LINE__
+ );
+
+ gcmkONERROR(gcvSTATUS_OUT_OF_RESOURCES);
+ }
+
+#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,27)
+ device_create(device_class, NULL, MKDEV(major, 0), NULL, DEVICE_NAME);
+#else
+ device_create(device_class, NULL, MKDEV(major, 0), DEVICE_NAME);
+#endif
+
+ gpuClass = device_class;
+ }
+
+ gcmkTRACE_ZONE(
+ gcvLEVEL_INFO, gcvZONE_DRIVER,
+ "%s(%d): irqLine=%d, contiguousSize=%lu, memBase=0x%lX\n",
+ __FUNCTION__, __LINE__,
+ irqLine, contiguousSize, registerMemBase
+ );
+
+ /* Success. */
+ gcmkFOOTER_NO();
+ return 0;
+
+OnError:
+ /* Roll back. */
+ if (device_class != gcvNULL)
+ {
+ device_destroy(device_class, MKDEV(major, 0));
+ class_destroy(device_class);
+ }
+
+ if (device != gcvNULL)
+ {
+ gcmkVERIFY_OK(gckGALDEVICE_Stop(device));
+ gcmkVERIFY_OK(gckGALDEVICE_Destroy(device));
+ }
+
+ gcmkFOOTER();
+ return result;
+}
+
+static void drv_exit(void)
+{
+ gcmkHEADER();
+
+ if (type == 1)
+ {
+ misc_deregister(&gal_device);
+ }
+ else
+ {
+ gcmkASSERT(gpuClass != gcvNULL);
+ device_destroy(gpuClass, MKDEV(major, 0));
+ class_destroy(gpuClass);
+
+ unregister_chrdev(major, DEVICE_NAME);
+ }
+
+ gcmkVERIFY_OK(gckGALDEVICE_Stop(galDevice));
+ gcmkVERIFY_OK(gckGALDEVICE_Destroy(galDevice));
+
+ if(gckDEBUGFS_IsEnabled())
+ {
+ gckDEBUGFS_Terminate();
+ }
+
+ gcmkFOOTER_NO();
+}
+
+#if gcdENABLE_DRM
+int viv_drm_probe(struct device *dev);
+int viv_drm_remove(struct device *dev);
+#endif
+
+#if USE_LINUX_PCIE
+static int gpu_probe(struct pci_dev *pdev, const struct pci_device_id *ent)
+#else /* USE_LINUX_PCIE */
+#if LINUX_VERSION_CODE >= KERNEL_VERSION(3, 8, 0)
+ static int gpu_probe(struct platform_device *pdev)
+#else
+ static int __devinit gpu_probe(struct platform_device *pdev)
+#endif
+#endif /* USE_LINUX_PCIE */
+{
+ int ret = -ENODEV;
+ gcsMODULE_PARAMETERS moduleParam = {
+ .irqLine = irqLine,
+ .registerMemBase = registerMemBase,
+ .registerMemSize = registerMemSize,
+ .irqLine2D = irqLine2D,
+ .registerMemBase2D = registerMemBase2D,
+ .registerMemSize2D = registerMemSize2D,
+ .irqLineVG = irqLineVG,
+ .registerMemBaseVG = registerMemBaseVG,
+ .registerMemSizeVG = registerMemSizeVG,
+ .contiguousSize = contiguousSize,
+ .contiguousBase = contiguousBase,
+ .externalSize = externalSize,
+ .externalBase = externalBase,
+ .bankSize = bankSize,
+ .fastClear = fastClear,
+ .powerManagement = powerManagement,
+ .gpuProfiler = gpuProfiler,
+ .signal = signal,
+ .baseAddress = baseAddress,
+ .physSize = physSize,
+ .logFileSize = logFileSize,
+ .recovery = recovery,
+ .stuckDump = stuckDump,
+ .showArgs = showArgs,
+ .gpu3DMinClock = gpu3DMinClock,
+ .registerMemMapped = registerMemMapped,
+ };
+
+ gcmkHEADER();
+
+ memcpy(moduleParam.irqs, irqs, gcmSIZEOF(gctINT) * gcvCORE_COUNT);
+ memcpy(moduleParam.registerBases, registerBases, gcmSIZEOF(gctUINT) * gcvCORE_COUNT);
+ memcpy(moduleParam.registerSizes, registerSizes, gcmSIZEOF(gctUINT) * gcvCORE_COUNT);
+ memcpy(moduleParam.chipIDs, chipIDs, gcmSIZEOF(gctUINT) * gcvCORE_COUNT);
+ moduleParam.compression = compression;
+ platform->device = pdev;
+#if USE_LINUX_PCIE
+ if (pci_enable_device(pdev)) {
+ printk(KERN_ERR "galcore: pci_enable_device() failed.\n");
+ }
+
+ if (pci_set_dma_mask(pdev, DMA_BIT_MASK(32))) {
+ printk(KERN_ERR "galcore: Failed to set DMA mask.\n");
+ }
+
+ pci_set_master(pdev);
+
+ if (pci_request_regions(pdev, "galcore")) {
+ printk(KERN_ERR "galcore: Failed to get ownership of BAR region.\n");
+ }
+
+#if USE_MSI
+ if (pci_enable_msi(pdev)) {
+ printk(KERN_ERR "galcore: Failed to enable MSI.\n");
+ }
+#endif
+#endif
+
+ if (platform->ops->getPower)
+ {
+ if (gcmIS_ERROR(platform->ops->getPower(platform)))
+ {
+ gcmkFOOTER_NO();
+ return ret;
+ }
+ }
+
+ if (platform->ops->adjustParam)
+ {
+ /* Override default module param. */
+ platform->ops->adjustParam(platform, &moduleParam);
+
+ /* Update module param because drv_init() uses them directly. */
+ _UpdateModuleParam(&moduleParam);
+ }
+
+ ret = drv_init();
+
+ if (!ret)
+ {
+#if USE_LINUX_PCIE
+ pci_set_drvdata(pdev, galDevice);
+#else
+ platform_set_drvdata(pdev, galDevice);
+#endif
+
+#if gcdENABLE_DRM
+ ret = viv_drm_probe(&pdev->dev);
+#endif
+ }
+
+ if (ret < 0)
+ {
+ gcmkFOOTER_ARG(KERN_INFO "Failed to register gpu driver: %d\n", ret);
+ }
+ else
+ {
+ gcmkFOOTER_NO();
+ }
+ return ret;
+}
+
+#if USE_LINUX_PCIE
+static void gpu_remove(struct pci_dev *pdev)
+#else /* USE_LINUX_PCIE */
+#if LINUX_VERSION_CODE >= KERNEL_VERSION(3, 8, 0)
+ static int gpu_remove(struct platform_device *pdev)
+#else
+ static int __devexit gpu_remove(struct platform_device *pdev)
+#endif
+#endif /* USE_LINUX_PCIE */
+{
+ gcmkHEADER();
+
+#if gcdENABLE_DRM
+ viv_drm_remove(&pdev->dev);
+#endif
+
+ drv_exit();
+
+ if (platform->ops->putPower)
+ {
+ platform->ops->putPower(platform);
+ }
+
+#if USE_LINUX_PCIE
+ pci_set_drvdata(pdev, NULL);
+#if USE_MSI
+ pci_disable_msi(pdev);
+#endif
+ pci_clear_master(pdev);
+ pci_release_regions(pdev);
+ pci_disable_device(pdev);
+ gcmkFOOTER_NO();
+ return;
+#else
+ gcmkFOOTER_NO();
+ return 0;
+#endif
+}
+
+static int gpu_suspend(struct platform_device *dev, pm_message_t state)
+{
+ gceSTATUS status;
+ gckGALDEVICE device;
+ gctINT i;
+
+ device = platform_get_drvdata(dev);
+
+ if (!device)
+ {
+ return -1;
+ }
+
+ for (i = 0; i < gcdMAX_GPU_COUNT; i++)
+ {
+ if (device->kernels[i] != gcvNULL)
+ {
+ /* Store states. */
+#if gcdENABLE_VG
+ if (i == gcvCORE_VG)
+ {
+ status = gckVGHARDWARE_QueryPowerManagementState(device->kernels[i]->vg->hardware, &device->statesStored[i]);
+ }
+ else
+#endif
+ {
+ status = gckHARDWARE_QueryPowerManagementState(device->kernels[i]->hardware, &device->statesStored[i]);
+ }
+
+ if (gcmIS_ERROR(status))
+ {
+ return -1;
+ }
+
+ /* need pull up power to flush gpu command buffer before suspend */
+#if gcdENABLE_VG
+ if (i == gcvCORE_VG)
+ {
+ status = gckVGHARDWARE_SetPowerManagementState(device->kernels[i]->vg->hardware, gcvPOWER_ON);
+ }
+ else
+#endif
+ {
+ status = gckHARDWARE_SetPowerManagementState(device->kernels[i]->hardware, gcvPOWER_ON);
+ }
+
+ if (gcmIS_ERROR(status))
+ {
+ return -1;
+ }
+
+#if gcdENABLE_VG
+ if (i == gcvCORE_VG)
+ {
+ status = gckVGHARDWARE_SetPowerManagementState(device->kernels[i]->vg->hardware, gcvPOWER_OFF);
+ }
+ else
+#endif
+ {
+ status = gckHARDWARE_SetPowerManagementState(device->kernels[i]->hardware, gcvPOWER_OFF);
+ }
+
+ if (gcmIS_ERROR(status))
+ {
+ return -1;
+ }
+
+ }
+ }
+
+ return 0;
+}
+
+static int gpu_resume(struct platform_device *dev)
+{
+ gceSTATUS status;
+ gckGALDEVICE device;
+ gctINT i;
+ gceCHIPPOWERSTATE statesStored;
+
+ device = platform_get_drvdata(dev);
+
+ if (!device)
+ {
+ return -1;
+ }
+
+ for (i = 0; i < gcdMAX_GPU_COUNT; i++)
+ {
+ if (device->kernels[i] != gcvNULL)
+ {
+#if gcdENABLE_VG
+ if (i == gcvCORE_VG)
+ {
+ status = gckVGHARDWARE_SetPowerManagementState(device->kernels[i]->vg->hardware, gcvPOWER_ON);
+ }
+ else
+#endif
+ {
+ status = gckHARDWARE_SetPowerManagementState(device->kernels[i]->hardware, gcvPOWER_ON);
+ }
+
+ if (gcmIS_ERROR(status))
+ {
+ return -1;
+ }
+
+ /* Convert global state to crossponding internal state. */
+ switch(device->statesStored[i])
+ {
+ case gcvPOWER_OFF:
+ statesStored = gcvPOWER_OFF_BROADCAST;
+ break;
+ case gcvPOWER_IDLE:
+ statesStored = gcvPOWER_IDLE_BROADCAST;
+ break;
+ case gcvPOWER_SUSPEND:
+ statesStored = gcvPOWER_SUSPEND_BROADCAST;
+ break;
+ case gcvPOWER_ON:
+ statesStored = gcvPOWER_ON_AUTO;
+ break;
+ default:
+ statesStored = device->statesStored[i];
+ break;
+ }
+
+ /* Restore states. */
+#if gcdENABLE_VG
+ if (i == gcvCORE_VG)
+ {
+ status = gckVGHARDWARE_SetPowerManagementState(device->kernels[i]->vg->hardware, statesStored);
+ }
+ else
+#endif
+ {
+ gctINT j = 0;
+
+ for (; j < 100; j++)
+ {
+ status = gckHARDWARE_SetPowerManagementState(device->kernels[i]->hardware, statesStored);
+
+ if (( statesStored != gcvPOWER_OFF_BROADCAST
+ && statesStored != gcvPOWER_SUSPEND_BROADCAST)
+ || status != gcvSTATUS_CHIP_NOT_READY)
+ {
+ break;
+ }
+
+ gcmkVERIFY_OK(gckOS_Delay(device->kernels[i]->os, 10));
+ };
+ }
+
+ if (gcmIS_ERROR(status))
+ {
+ return -1;
+ }
+ }
+ }
+
+ return 0;
+}
+
+#if defined(CONFIG_PM) && LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 30)
+#ifdef CONFIG_PM_SLEEP
+static int gpu_system_suspend(struct device *dev)
+{
+ pm_message_t state={0};
+ return gpu_suspend(to_platform_device(dev), state);
+}
+
+static int gpu_system_resume(struct device *dev)
+{
+ return gpu_resume(to_platform_device(dev));
+}
+#endif
+
+static const struct dev_pm_ops gpu_pm_ops = {
+ SET_SYSTEM_SLEEP_PM_OPS(gpu_system_suspend, gpu_system_resume)
+};
+#endif
+
+#if USE_LINUX_PCIE
+static const struct pci_device_id vivpci_ids[] = {
+ {
+ .class = 0x000000,
+ .class_mask = 0x000000,
+ .vendor = 0x10ee,
+ .device = 0x7012,
+ .subvendor = PCI_ANY_ID,
+ .subdevice = PCI_ANY_ID,
+ .driver_data = 0
+ }, { /* End: all zeroes */ }
+};
+
+MODULE_DEVICE_TABLE(pci, vivpci_ids);
+
+static struct pci_driver gpu_driver = {
+ .name = DEVICE_NAME,
+ .id_table = vivpci_ids,
+ .probe = gpu_probe,
+ .remove = gpu_remove
+};
+
+#else /* USE_LINUX_PCIE */
+
+static struct platform_driver gpu_driver = {
+ .probe = gpu_probe,
+#if LINUX_VERSION_CODE >= KERNEL_VERSION(3, 8, 0)
+ .remove = gpu_remove,
+#else
+ .remove = __devexit_p(gpu_remove),
+#endif
+
+ .suspend = gpu_suspend,
+ .resume = gpu_resume,
+
+ .driver = {
+ .owner = THIS_MODULE,
+ .name = DEVICE_NAME,
+#if defined(CONFIG_PM) && LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 30)
+ .pm = &gpu_pm_ops,
+#endif
+ }
+};
+#endif /* USE_LINUX_PCIE */
+
+static int __init gpu_init(void)
+{
+ int ret = 0;
+
+ ret = soc_platform_init(&gpu_driver, &platform);
+
+ if (ret || !platform)
+ {
+ printk(KERN_ERR "galcore: Soc platform init failed.\n");
+ return -ENODEV;
+ }
+
+#if USE_LINUX_PCIE
+ ret = pci_register_driver(&gpu_driver);
+#else /* USE_LINUX_PCIE */
+ ret = platform_driver_register(&gpu_driver);
+#endif /* USE_LINUX_PCIE */
+
+ if (ret)
+ {
+ printk(KERN_ERR "galcore: gpu_init() failed to register driver!\n");
+ soc_platform_terminate(platform);
+ platform = NULL;
+ return ret;
+ }
+
+ platform->driver = &gpu_driver;
+
+ return 0;
+}
+
+static void __exit gpu_exit(void)
+{
+#if USE_LINUX_PCIE
+ pci_unregister_driver(&gpu_driver);
+#else
+ platform_driver_unregister(&gpu_driver);
+#endif /* USE_LINUX_PCIE */
+
+ soc_platform_terminate(platform);
+ platform = NULL;
+}
+
+module_init(gpu_init);
+module_exit(gpu_exit);