blob: 16448610afec5d3881d81d5d1e9bdcf77bdd2ac0 [file] [log] [blame]
/* Copyright (c) 2002,2007-2014, The Linux Foundation. All rights reserved.
*
* This program is free software; you can redistribute it and/or modify
* it under the terms of the GNU General Public License version 2 and
* only version 2 as published by the Free Software Foundation.
*
* 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.
*
*/
#include <linux/slab.h>
#include <linux/msm_kgsl.h>
#include <linux/sched.h>
#include "kgsl.h"
#include "kgsl_sharedmem.h"
#include "adreno.h"
#include "adreno_trace.h"
#define KGSL_INIT_REFTIMESTAMP 0x7FFFFFFF
/* quad for copying GMEM to context shadow */
#define QUAD_LEN 12
#define QUAD_RESTORE_LEN 14
static unsigned int gmem_copy_quad[QUAD_LEN] = {
0x00000000, 0x00000000, 0x3f800000,
0x00000000, 0x00000000, 0x3f800000,
0x00000000, 0x00000000, 0x3f800000,
0x00000000, 0x00000000, 0x3f800000
};
static unsigned int gmem_restore_quad[QUAD_RESTORE_LEN] = {
0x00000000, 0x3f800000, 0x3f800000,
0x00000000, 0x00000000, 0x00000000,
0x3f800000, 0x00000000, 0x00000000,
0x3f800000, 0x00000000, 0x00000000,
0x3f800000, 0x3f800000,
};
#define TEXCOORD_LEN 8
static unsigned int gmem_copy_texcoord[TEXCOORD_LEN] = {
0x00000000, 0x3f800000,
0x3f800000, 0x3f800000,
0x00000000, 0x00000000,
0x3f800000, 0x00000000
};
/*
* Helper functions
* These are global helper functions used by the GPUs during context switch
*/
/**
* uint2float - convert a uint to IEEE754 single precision float
* @ uintval - value to convert
*/
unsigned int uint2float(unsigned int uintval)
{
unsigned int exp, frac = 0;
if (uintval == 0)
return 0;
exp = ilog2(uintval);
/* Calculate fraction */
if (23 > exp)
frac = (uintval & (~(1 << exp))) << (23 - exp);
/* Exp is biased by 127 and shifted 23 bits */
exp = (exp + 127) << 23;
return exp | frac;
}
static void set_gmem_copy_quad(struct gmem_shadow_t *shadow)
{
/* set vertex buffer values */
gmem_copy_quad[1] = uint2float(shadow->height);
gmem_copy_quad[3] = uint2float(shadow->width);
gmem_copy_quad[4] = uint2float(shadow->height);
gmem_copy_quad[9] = uint2float(shadow->width);
gmem_restore_quad[5] = uint2float(shadow->height);
gmem_restore_quad[7] = uint2float(shadow->width);
memcpy(shadow->quad_vertices.hostptr, gmem_copy_quad, QUAD_LEN << 2);
memcpy(shadow->quad_vertices_restore.hostptr, gmem_restore_quad,
QUAD_RESTORE_LEN << 2);
memcpy(shadow->quad_texcoords.hostptr, gmem_copy_texcoord,
TEXCOORD_LEN << 2);
}
/**
* build_quad_vtxbuff - Create a quad for saving/restoring GMEM
* @ context - Pointer to the context being created
* @ shadow - Pointer to the GMEM shadow structure
* @ incmd - Pointer to pointer to the temporary command buffer
*/
/* quad for saving/restoring gmem */
void build_quad_vtxbuff(struct adreno_context *drawctxt,
struct gmem_shadow_t *shadow, unsigned int **incmd)
{
unsigned int *cmd = *incmd;
/* quad vertex buffer location (in GPU space) */
shadow->quad_vertices.hostptr = cmd;
shadow->quad_vertices.gpuaddr = virt2gpu(cmd, &drawctxt->gpustate);
cmd += QUAD_LEN;
/* Used by A3XX, but define for both to make the code easier */
shadow->quad_vertices_restore.hostptr = cmd;
shadow->quad_vertices_restore.gpuaddr =
virt2gpu(cmd, &drawctxt->gpustate);
cmd += QUAD_RESTORE_LEN;
/* tex coord buffer location (in GPU space) */
shadow->quad_texcoords.hostptr = cmd;
shadow->quad_texcoords.gpuaddr = virt2gpu(cmd, &drawctxt->gpustate);
cmd += TEXCOORD_LEN;
set_gmem_copy_quad(shadow);
*incmd = cmd;
}
static void wait_callback(struct kgsl_device *device, void *priv, u32 id,
u32 timestamp, u32 type)
{
struct adreno_context *drawctxt = priv;
wake_up_all(&drawctxt->waiting);
}
#define adreno_wait_event_interruptible_timeout(wq, condition, timeout, io) \
({ \
long __ret = timeout; \
if (io) \
__wait_io_event_interruptible_timeout(wq, condition, __ret); \
else \
__wait_event_interruptible_timeout(wq, condition, __ret); \
__ret; \
})
#define adreno_wait_event_interruptible(wq, condition, io) \
({ \
long __ret; \
if (io) \
__wait_io_event_interruptible(wq, condition, __ret); \
else \
__wait_event_interruptible(wq, condition, __ret); \
__ret; \
})
static int _check_context_timestamp(struct kgsl_device *device,
struct adreno_context *drawctxt, unsigned int timestamp)
{
int ret = 0;
/* Bail if the drawctxt has been invalidated or destroyed */
if (kgsl_context_detached(&drawctxt->base) ||
drawctxt->state != ADRENO_CONTEXT_STATE_ACTIVE)
return 1;
mutex_lock(&device->mutex);
ret = kgsl_check_timestamp(device, &drawctxt->base, timestamp);
mutex_unlock(&device->mutex);
return ret;
}
/**
* adreno_drawctxt_wait() - sleep until a timestamp expires
* @adreno_dev: pointer to the adreno_device struct
* @drawctxt: Pointer to the draw context to sleep for
* @timetamp: Timestamp to wait on
* @timeout: Number of jiffies to wait (0 for infinite)
*
* Register an event to wait for a timestamp on a context and sleep until it
* has past. Returns < 0 on error, -ETIMEDOUT if the timeout expires or 0
* on success
*/
int adreno_drawctxt_wait(struct adreno_device *adreno_dev,
struct kgsl_context *context,
uint32_t timestamp, unsigned int timeout)
{
static unsigned int io_cnt;
struct kgsl_device *device = &adreno_dev->dev;
struct kgsl_pwrctrl *pwr = &device->pwrctrl;
struct adreno_context *drawctxt = ADRENO_CONTEXT(context);
int ret, io;
if (kgsl_context_detached(context))
return -EINVAL;
if (drawctxt->state == ADRENO_CONTEXT_STATE_INVALID)
return -EDEADLK;
/* Needs to hold the device mutex */
BUG_ON(!mutex_is_locked(&device->mutex));
trace_adreno_drawctxt_wait_start(context->id, timestamp);
ret = kgsl_add_event(device, context->id, timestamp,
wait_callback, drawctxt, NULL);
if (ret)
goto done;
/*
* For proper power accounting sometimes we need to call
* io_wait_interruptible_timeout and sometimes we need to call
* plain old wait_interruptible_timeout. We call the regular
* timeout N times out of 100, where N is a number specified by
* the current power level
*/
io_cnt = (io_cnt + 1) % 100;
io = (io_cnt < pwr->pwrlevels[pwr->active_pwrlevel].io_fraction)
? 0 : 1;
mutex_unlock(&device->mutex);
if (timeout) {
ret = (int) adreno_wait_event_interruptible_timeout(
drawctxt->waiting,
_check_context_timestamp(device, drawctxt, timestamp),
msecs_to_jiffies(timeout), io);
if (ret == 0)
ret = -ETIMEDOUT;
else if (ret > 0)
ret = 0;
} else {
ret = (int) adreno_wait_event_interruptible(drawctxt->waiting,
_check_context_timestamp(device, drawctxt, timestamp),
io);
}
mutex_lock(&device->mutex);
/* -EDEADLK if the context was invalidated while we were waiting */
if (drawctxt->state == ADRENO_CONTEXT_STATE_INVALID)
ret = -EDEADLK;
/* Return -EINVAL if the context was detached while we were waiting */
if (kgsl_context_detached(context))
ret = -EINVAL;
done:
trace_adreno_drawctxt_wait_done(context->id, timestamp, ret);
return ret;
}
static void global_wait_callback(struct kgsl_device *device, void *priv, u32 id,
u32 timestamp, u32 type)
{
struct adreno_context *drawctxt = priv;
wake_up_all(&drawctxt->waiting);
kgsl_context_put(&drawctxt->base);
}
static int _check_global_timestamp(struct kgsl_device *device,
struct adreno_context *drawctxt, unsigned int timestamp)
{
/* Stop waiting if the context is invalidated */
if (drawctxt->state == ADRENO_CONTEXT_STATE_INVALID)
return 1;
return kgsl_check_timestamp(device, NULL, timestamp);
}
int adreno_drawctxt_wait_global(struct adreno_device *adreno_dev,
struct kgsl_context *context,
uint32_t timestamp, unsigned int timeout)
{
struct kgsl_device *device = &adreno_dev->dev;
struct adreno_context *drawctxt = ADRENO_CONTEXT(context);
int ret = 0;
/* Needs to hold the device mutex */
BUG_ON(!mutex_is_locked(&device->mutex));
if (!_kgsl_context_get(context)) {
ret = -EINVAL;
goto done;
}
/*
* If the context is invalid then return immediately - we may end up
* waiting for a timestamp that will never come
*/
if (drawctxt->state == ADRENO_CONTEXT_STATE_INVALID) {
kgsl_context_put(context);
goto done;
}
trace_adreno_drawctxt_wait_start(KGSL_MEMSTORE_GLOBAL, timestamp);
ret = kgsl_add_event(device, KGSL_MEMSTORE_GLOBAL, timestamp,
global_wait_callback, drawctxt, NULL);
if (ret) {
kgsl_context_put(context);
goto done;
}
mutex_unlock(&device->mutex);
if (timeout) {
ret = (int) wait_event_timeout(drawctxt->waiting,
_check_global_timestamp(device, drawctxt, timestamp),
msecs_to_jiffies(timeout));
if (ret == 0)
ret = -ETIMEDOUT;
else if (ret > 0)
ret = 0;
} else {
wait_event(drawctxt->waiting,
_check_global_timestamp(device, drawctxt, timestamp));
}
mutex_lock(&device->mutex);
if (ret)
kgsl_cancel_events_timestamp(device, NULL, timestamp);
done:
trace_adreno_drawctxt_wait_done(KGSL_MEMSTORE_GLOBAL, timestamp, ret);
return ret;
}
/**
* adreno_drawctxt_invalidate() - Invalidate an adreno draw context
* @device: Pointer to the KGSL device structure for the GPU
* @context: Pointer to the KGSL context structure
*
* Invalidate the context and remove all queued commands and cancel any pending
* waiters
*/
void adreno_drawctxt_invalidate(struct kgsl_device *device,
struct kgsl_context *context)
{
struct adreno_context *drawctxt = ADRENO_CONTEXT(context);
drawctxt->state = ADRENO_CONTEXT_STATE_INVALID;
/* Clear the pending queue */
mutex_lock(&drawctxt->mutex);
/*
* set the timestamp to the last value since the context is invalidated
* and we want the pending events for this context to go away
*/
kgsl_sharedmem_writel(device, &device->memstore,
KGSL_MEMSTORE_OFFSET(context->id, soptimestamp),
drawctxt->timestamp);
kgsl_sharedmem_writel(device, &device->memstore,
KGSL_MEMSTORE_OFFSET(context->id, eoptimestamp),
drawctxt->timestamp);
while (drawctxt->cmdqueue_head != drawctxt->cmdqueue_tail) {
struct kgsl_cmdbatch *cmdbatch =
drawctxt->cmdqueue[drawctxt->cmdqueue_head];
drawctxt->cmdqueue_head = (drawctxt->cmdqueue_head + 1) %
ADRENO_CONTEXT_CMDQUEUE_SIZE;
mutex_unlock(&drawctxt->mutex);
mutex_lock(&device->mutex);
kgsl_cancel_events_timestamp(device, context,
cmdbatch->timestamp);
mutex_unlock(&device->mutex);
kgsl_cmdbatch_destroy(cmdbatch);
mutex_lock(&drawctxt->mutex);
}
mutex_unlock(&drawctxt->mutex);
/* Give the bad news to everybody waiting around */
wake_up_all(&drawctxt->waiting);
wake_up_all(&drawctxt->wq);
}
/**
* adreno_drawctxt_create - create a new adreno draw context
* @dev_priv: the owner of the context
* @flags: flags for the context (passed from user space)
*
* Create and return a new draw context for the 3D core.
*/
struct kgsl_context *
adreno_drawctxt_create(struct kgsl_device_private *dev_priv,
uint32_t *flags)
{
struct adreno_context *drawctxt;
struct kgsl_device *device = dev_priv->device;
struct adreno_device *adreno_dev = ADRENO_DEVICE(device);
int ret;
drawctxt = kzalloc(sizeof(struct adreno_context), GFP_KERNEL);
if (drawctxt == NULL)
return ERR_PTR(-ENOMEM);
ret = kgsl_context_init(dev_priv, &drawctxt->base);
if (ret != 0) {
kfree(drawctxt);
return ERR_PTR(ret);
}
drawctxt->bin_base_offset = 0;
drawctxt->timestamp = 0;
drawctxt->base.flags = *flags & (KGSL_CONTEXT_PREAMBLE |
KGSL_CONTEXT_NO_GMEM_ALLOC |
KGSL_CONTEXT_PER_CONTEXT_TS |
KGSL_CONTEXT_USER_GENERATED_TS |
KGSL_CONTEXT_NO_FAULT_TOLERANCE |
KGSL_CONTEXT_TYPE_MASK |
KGSL_CONTEXT_PWR_CONSTRAINT);
/* Always enable per-context timestamps */
drawctxt->base.flags |= KGSL_CONTEXT_PER_CONTEXT_TS;
mutex_init(&drawctxt->mutex);
init_waitqueue_head(&drawctxt->wq);
init_waitqueue_head(&drawctxt->waiting);
/*
* Set up the plist node for the dispatcher. For now all contexts have
* the same priority, but later the priority will be set at create time
* by the user
*/
plist_node_init(&drawctxt->pending, ADRENO_CONTEXT_DEFAULT_PRIORITY);
if (adreno_dev->gpudev->ctxt_create) {
ret = adreno_dev->gpudev->ctxt_create(adreno_dev, drawctxt);
if (ret)
goto err;
} else if ((drawctxt->base.flags & KGSL_CONTEXT_PREAMBLE) == 0 ||
(drawctxt->base.flags & KGSL_CONTEXT_NO_GMEM_ALLOC) == 0) {
KGSL_DEV_ERR_ONCE(device,
"legacy context switch not supported\n");
ret = -EINVAL;
goto err;
} else {
drawctxt->ops = &adreno_preamble_ctx_ops;
}
kgsl_sharedmem_writel(device, &device->memstore,
KGSL_MEMSTORE_OFFSET(drawctxt->base.id, soptimestamp),
0);
kgsl_sharedmem_writel(device, &device->memstore,
KGSL_MEMSTORE_OFFSET(drawctxt->base.id, eoptimestamp),
0);
/* copy back whatever flags we dediced were valid */
*flags = drawctxt->base.flags;
return &drawctxt->base;
err:
kgsl_context_detach(&drawctxt->base);
return ERR_PTR(ret);
}
/**
* adreno_drawctxt_sched() - Schedule a previously blocked context
* @device: pointer to a KGSL device
* @drawctxt: drawctxt to rechedule
*
* This function is called by the core when it knows that a previously blocked
* context has been unblocked. The default adreno response is to reschedule the
* context on the dispatcher
*/
void adreno_drawctxt_sched(struct kgsl_device *device,
struct kgsl_context *context)
{
adreno_dispatcher_queue_context(device, ADRENO_CONTEXT(context));
}
/**
* adreno_drawctxt_detach(): detach a context from the GPU
* @context: Generic KGSL context container for the context
*
*/
int adreno_drawctxt_detach(struct kgsl_context *context)
{
struct kgsl_device *device;
struct adreno_device *adreno_dev;
struct adreno_context *drawctxt;
int ret;
if (context == NULL)
return 0;
device = context->device;
adreno_dev = ADRENO_DEVICE(device);
drawctxt = ADRENO_CONTEXT(context);
/* deactivate context */
if (adreno_dev->drawctxt_active == drawctxt)
adreno_drawctxt_switch(adreno_dev, NULL, 0);
mutex_lock(&drawctxt->mutex);
while (drawctxt->cmdqueue_head != drawctxt->cmdqueue_tail) {
struct kgsl_cmdbatch *cmdbatch =
drawctxt->cmdqueue[drawctxt->cmdqueue_head];
drawctxt->cmdqueue_head = (drawctxt->cmdqueue_head + 1) %
ADRENO_CONTEXT_CMDQUEUE_SIZE;
mutex_unlock(&drawctxt->mutex);
/*
* Don't hold the drawctxt mutex while the cmdbatch is being
* destroyed because the cmdbatch destroy takes the device
* mutex and the world falls in on itself
*/
kgsl_cmdbatch_destroy(cmdbatch);
mutex_lock(&drawctxt->mutex);
}
mutex_unlock(&drawctxt->mutex);
/*
* internal_timestamp is set in adreno_ringbuffer_addcmds,
* which holds the device mutex. The entire context destroy
* process requires the device mutex as well. But lets
* make sure we notice if the locking changes.
*/
BUG_ON(!mutex_is_locked(&device->mutex));
/* Wait for the last global timestamp to pass before continuing.
* The maxumum wait time is 30s, some large IB's can take longer
* than 10s and if hang happens then the time for the context's
* commands to retire will be greater than 10s. 30s should be sufficient
* time to wait for the commands even if a hang happens.
*/
ret = adreno_drawctxt_wait_global(adreno_dev, context,
drawctxt->internal_timestamp, 30 * 1000);
/*
* If the wait for global fails then nothing after this point is likely
* to work very well - BUG_ON() so we can take advantage of the debug
* tools to figure out what the h - e - double hockey sticks happened
*/
BUG_ON(ret);
kgsl_sharedmem_writel(device, &device->memstore,
KGSL_MEMSTORE_OFFSET(context->id, soptimestamp),
drawctxt->timestamp);
kgsl_sharedmem_writel(device, &device->memstore,
KGSL_MEMSTORE_OFFSET(context->id, eoptimestamp),
drawctxt->timestamp);
kgsl_sharedmem_free(&drawctxt->gpustate);
kgsl_sharedmem_free(&drawctxt->context_gmem_shadow.gmemshadow);
if (drawctxt->ops->detach)
drawctxt->ops->detach(drawctxt);
/* wake threads waiting to submit commands from this context */
wake_up_all(&drawctxt->waiting);
wake_up_all(&drawctxt->wq);
return ret;
}
void adreno_drawctxt_destroy(struct kgsl_context *context)
{
struct adreno_context *drawctxt;
if (context == NULL)
return;
drawctxt = ADRENO_CONTEXT(context);
kfree(drawctxt);
}
/**
* adreno_context_restore() - generic context restore handler
* @adreno_dev: the device
* @context: the context
*
* Basic context restore handler that writes the context identifier
* to the ringbuffer and issues pagetable switch commands if necessary.
* May be called directly from the adreno_context_ops.restore function
* pointer or as the first action in a hardware specific restore
* function.
*/
int adreno_context_restore(struct adreno_device *adreno_dev,
struct adreno_context *context)
{
struct kgsl_device *device;
unsigned int cmds[8];
if (adreno_dev == NULL || context == NULL)
return -EINVAL;
device = &adreno_dev->dev;
/* write the context identifier to the ringbuffer */
cmds[0] = cp_nop_packet(1);
cmds[1] = KGSL_CONTEXT_TO_MEM_IDENTIFIER;
cmds[2] = cp_type3_packet(CP_MEM_WRITE, 2);
cmds[3] = device->memstore.gpuaddr +
KGSL_MEMSTORE_OFFSET(KGSL_MEMSTORE_GLOBAL, current_context);
cmds[4] = context->base.id;
/* Flush the UCHE for new context */
cmds[5] = cp_type0_packet(
adreno_getreg(adreno_dev, ADRENO_REG_UCHE_INVALIDATE0), 2);
cmds[6] = 0;
cmds[7] = 0x90000000;
return adreno_ringbuffer_issuecmds(device, context,
KGSL_CMD_FLAGS_NONE, cmds, 8);
}
const struct adreno_context_ops adreno_preamble_ctx_ops = {
.restore = adreno_context_restore,
};
/**
* context_save() - save old context when necessary
* @drawctxt - the old context
*
* For legacy context switching, we need to issue save
* commands unless the context is being destroyed.
*/
static inline int context_save(struct adreno_device *adreno_dev,
struct adreno_context *context)
{
if (context->ops->save == NULL
|| kgsl_context_detached(&context->base)
|| context->state == ADRENO_CONTEXT_STATE_INVALID)
return 0;
return context->ops->save(adreno_dev, context);
}
/**
* adreno_drawctxt_set_bin_base_offset - set bin base offset for the context
* @device - KGSL device that owns the context
* @context- Generic KGSL context container for the context
* @offset - Offset to set
*
* Set the bin base offset for A2XX devices. Not valid for A3XX devices.
*/
void adreno_drawctxt_set_bin_base_offset(struct kgsl_device *device,
struct kgsl_context *context,
unsigned int offset)
{
struct adreno_context *drawctxt;
if (context == NULL)
return;
drawctxt = ADRENO_CONTEXT(context);
drawctxt->bin_base_offset = offset;
}
/**
* adreno_drawctxt_switch - switch the current draw context
* @adreno_dev - The 3D device that owns the context
* @drawctxt - the 3D context to switch to
* @flags - Flags to accompany the switch (from user space)
*
* Switch the current draw context
*/
int adreno_drawctxt_switch(struct adreno_device *adreno_dev,
struct adreno_context *drawctxt,
unsigned int flags)
{
struct kgsl_device *device = &adreno_dev->dev;
int ret = 0;
if (drawctxt) {
/*
* Handle legacy gmem / save restore flag on each IB.
* Userspace sets to guard IB sequences that require
* gmem to be saved and clears it at the end of the
* sequence.
*/
if (flags & KGSL_CONTEXT_SAVE_GMEM)
/* Set the flag in context so that the save is done
* when this context is switched out. */
set_bit(ADRENO_CONTEXT_GMEM_SAVE, &drawctxt->priv);
else
/* Remove GMEM saving flag from the context */
clear_bit(ADRENO_CONTEXT_GMEM_SAVE, &drawctxt->priv);
}
/* already current? */
if (adreno_dev->drawctxt_active == drawctxt) {
if (drawctxt && drawctxt->ops->draw_workaround)
ret = drawctxt->ops->draw_workaround(adreno_dev,
drawctxt);
return ret;
}
trace_adreno_drawctxt_switch(adreno_dev->drawctxt_active,
drawctxt, flags);
/* Get a refcount to the new instance */
if (drawctxt) {
if (!_kgsl_context_get(&drawctxt->base))
return -EINVAL;
ret = kgsl_mmu_setstate(&device->mmu,
drawctxt->base.proc_priv->pagetable,
adreno_dev->drawctxt_active ?
adreno_dev->drawctxt_active->base.id :
KGSL_CONTEXT_INVALID);
/* Set the new context */
ret = drawctxt->ops->restore(adreno_dev, drawctxt);
if (ret) {
KGSL_DRV_ERR(device,
"Error in GPU context %d restore: %d\n",
drawctxt->base.id, ret);
return ret;
}
} else {
/*
* No context - set the default pagetable and thats it.
* If there isn't a current context, the kgsl_mmu_setstate
* will use the CPU path so we don't need to give
* it a valid context id.
*/
ret = kgsl_mmu_setstate(&device->mmu,
device->mmu.defaultpagetable,
adreno_dev->drawctxt_active->base.id);
}
/* Put the old instance of the active drawctxt */
if (adreno_dev->drawctxt_active)
kgsl_context_put(&adreno_dev->drawctxt_active->base);
adreno_dev->drawctxt_active = drawctxt;
return 0;
}