linux/drivers/gpu/host1x/cdma.c
Mikko Perttunen c78f837ae3 gpu: host1x: Add no-recovery mode
Add a new property for jobs to enable or disable recovery i.e.
CPU increments of syncpoints to max value on job timeout. This
allows for a more solid model for hanged jobs, where userspace
doesn't need to guess if a syncpoint increment happened because
the job completed, or because job timeout was triggered.

On job timeout, we stop the channel, NOP all future jobs on the
channel using the same syncpoint, mark the syncpoint as locked
and resume the channel from the next job, if any.

The future jobs are NOPed, since because we don't do the CPU
increments, the value of the syncpoint is no longer synchronized,
and any waiters would become confused if a future job incremented
the syncpoint. The syncpoint is marked locked to ensure that any
future jobs cannot increment the syncpoint either, until the
application has recognized the situation and reallocated the
syncpoint.

Signed-off-by: Mikko Perttunen <mperttunen@nvidia.com>
Signed-off-by: Thierry Reding <treding@nvidia.com>
2021-08-10 14:40:23 +02:00

677 lines
16 KiB
C

// SPDX-License-Identifier: GPL-2.0-only
/*
* Tegra host1x Command DMA
*
* Copyright (c) 2010-2013, NVIDIA Corporation.
*/
#include <asm/cacheflush.h>
#include <linux/device.h>
#include <linux/dma-mapping.h>
#include <linux/host1x.h>
#include <linux/interrupt.h>
#include <linux/kernel.h>
#include <linux/kfifo.h>
#include <linux/slab.h>
#include <trace/events/host1x.h>
#include "cdma.h"
#include "channel.h"
#include "dev.h"
#include "debug.h"
#include "job.h"
/*
* push_buffer
*
* The push buffer is a circular array of words to be fetched by command DMA.
* Note that it works slightly differently to the sync queue; fence == pos
* means that the push buffer is full, not empty.
*/
/*
* Typically the commands written into the push buffer are a pair of words. We
* use slots to represent each of these pairs and to simplify things. Note the
* strange number of slots allocated here. 512 slots will fit exactly within a
* single memory page. We also need one additional word at the end of the push
* buffer for the RESTART opcode that will instruct the CDMA to jump back to
* the beginning of the push buffer. With 512 slots, this means that we'll use
* 2 memory pages and waste 4092 bytes of the second page that will never be
* used.
*/
#define HOST1X_PUSHBUFFER_SLOTS 511
/*
* Clean up push buffer resources
*/
static void host1x_pushbuffer_destroy(struct push_buffer *pb)
{
struct host1x_cdma *cdma = pb_to_cdma(pb);
struct host1x *host1x = cdma_to_host1x(cdma);
if (!pb->mapped)
return;
if (host1x->domain) {
iommu_unmap(host1x->domain, pb->dma, pb->alloc_size);
free_iova(&host1x->iova, iova_pfn(&host1x->iova, pb->dma));
}
dma_free_wc(host1x->dev, pb->alloc_size, pb->mapped, pb->phys);
pb->mapped = NULL;
pb->phys = 0;
}
/*
* Init push buffer resources
*/
static int host1x_pushbuffer_init(struct push_buffer *pb)
{
struct host1x_cdma *cdma = pb_to_cdma(pb);
struct host1x *host1x = cdma_to_host1x(cdma);
struct iova *alloc;
u32 size;
int err;
pb->mapped = NULL;
pb->phys = 0;
pb->size = HOST1X_PUSHBUFFER_SLOTS * 8;
size = pb->size + 4;
/* initialize buffer pointers */
pb->fence = pb->size - 8;
pb->pos = 0;
if (host1x->domain) {
unsigned long shift;
size = iova_align(&host1x->iova, size);
pb->mapped = dma_alloc_wc(host1x->dev, size, &pb->phys,
GFP_KERNEL);
if (!pb->mapped)
return -ENOMEM;
shift = iova_shift(&host1x->iova);
alloc = alloc_iova(&host1x->iova, size >> shift,
host1x->iova_end >> shift, true);
if (!alloc) {
err = -ENOMEM;
goto iommu_free_mem;
}
pb->dma = iova_dma_addr(&host1x->iova, alloc);
err = iommu_map(host1x->domain, pb->dma, pb->phys, size,
IOMMU_READ);
if (err)
goto iommu_free_iova;
} else {
pb->mapped = dma_alloc_wc(host1x->dev, size, &pb->phys,
GFP_KERNEL);
if (!pb->mapped)
return -ENOMEM;
pb->dma = pb->phys;
}
pb->alloc_size = size;
host1x_hw_pushbuffer_init(host1x, pb);
return 0;
iommu_free_iova:
__free_iova(&host1x->iova, alloc);
iommu_free_mem:
dma_free_wc(host1x->dev, size, pb->mapped, pb->phys);
return err;
}
/*
* Push two words to the push buffer
* Caller must ensure push buffer is not full
*/
static void host1x_pushbuffer_push(struct push_buffer *pb, u32 op1, u32 op2)
{
u32 *p = (u32 *)((void *)pb->mapped + pb->pos);
WARN_ON(pb->pos == pb->fence);
*(p++) = op1;
*(p++) = op2;
pb->pos += 8;
if (pb->pos >= pb->size)
pb->pos -= pb->size;
}
/*
* Pop a number of two word slots from the push buffer
* Caller must ensure push buffer is not empty
*/
static void host1x_pushbuffer_pop(struct push_buffer *pb, unsigned int slots)
{
/* Advance the next write position */
pb->fence += slots * 8;
if (pb->fence >= pb->size)
pb->fence -= pb->size;
}
/*
* Return the number of two word slots free in the push buffer
*/
static u32 host1x_pushbuffer_space(struct push_buffer *pb)
{
unsigned int fence = pb->fence;
if (pb->fence < pb->pos)
fence += pb->size;
return (fence - pb->pos) / 8;
}
/*
* Sleep (if necessary) until the requested event happens
* - CDMA_EVENT_SYNC_QUEUE_EMPTY : sync queue is completely empty.
* - Returns 1
* - CDMA_EVENT_PUSH_BUFFER_SPACE : there is space in the push buffer
* - Return the amount of space (> 0)
* Must be called with the cdma lock held.
*/
unsigned int host1x_cdma_wait_locked(struct host1x_cdma *cdma,
enum cdma_event event)
{
for (;;) {
struct push_buffer *pb = &cdma->push_buffer;
unsigned int space;
switch (event) {
case CDMA_EVENT_SYNC_QUEUE_EMPTY:
space = list_empty(&cdma->sync_queue) ? 1 : 0;
break;
case CDMA_EVENT_PUSH_BUFFER_SPACE:
space = host1x_pushbuffer_space(pb);
break;
default:
WARN_ON(1);
return -EINVAL;
}
if (space)
return space;
trace_host1x_wait_cdma(dev_name(cdma_to_channel(cdma)->dev),
event);
/* If somebody has managed to already start waiting, yield */
if (cdma->event != CDMA_EVENT_NONE) {
mutex_unlock(&cdma->lock);
schedule();
mutex_lock(&cdma->lock);
continue;
}
cdma->event = event;
mutex_unlock(&cdma->lock);
wait_for_completion(&cdma->complete);
mutex_lock(&cdma->lock);
}
return 0;
}
/*
* Sleep (if necessary) until the push buffer has enough free space.
*
* Must be called with the cdma lock held.
*/
static int host1x_cdma_wait_pushbuffer_space(struct host1x *host1x,
struct host1x_cdma *cdma,
unsigned int needed)
{
while (true) {
struct push_buffer *pb = &cdma->push_buffer;
unsigned int space;
space = host1x_pushbuffer_space(pb);
if (space >= needed)
break;
trace_host1x_wait_cdma(dev_name(cdma_to_channel(cdma)->dev),
CDMA_EVENT_PUSH_BUFFER_SPACE);
host1x_hw_cdma_flush(host1x, cdma);
/* If somebody has managed to already start waiting, yield */
if (cdma->event != CDMA_EVENT_NONE) {
mutex_unlock(&cdma->lock);
schedule();
mutex_lock(&cdma->lock);
continue;
}
cdma->event = CDMA_EVENT_PUSH_BUFFER_SPACE;
mutex_unlock(&cdma->lock);
wait_for_completion(&cdma->complete);
mutex_lock(&cdma->lock);
}
return 0;
}
/*
* Start timer that tracks the time spent by the job.
* Must be called with the cdma lock held.
*/
static void cdma_start_timer_locked(struct host1x_cdma *cdma,
struct host1x_job *job)
{
if (cdma->timeout.client) {
/* timer already started */
return;
}
cdma->timeout.client = job->client;
cdma->timeout.syncpt = job->syncpt;
cdma->timeout.syncpt_val = job->syncpt_end;
cdma->timeout.start_ktime = ktime_get();
schedule_delayed_work(&cdma->timeout.wq,
msecs_to_jiffies(job->timeout));
}
/*
* Stop timer when a buffer submission completes.
* Must be called with the cdma lock held.
*/
static void stop_cdma_timer_locked(struct host1x_cdma *cdma)
{
cancel_delayed_work(&cdma->timeout.wq);
cdma->timeout.client = NULL;
}
/*
* For all sync queue entries that have already finished according to the
* current sync point registers:
* - unpin & unref their mems
* - pop their push buffer slots
* - remove them from the sync queue
* This is normally called from the host code's worker thread, but can be
* called manually if necessary.
* Must be called with the cdma lock held.
*/
static void update_cdma_locked(struct host1x_cdma *cdma)
{
bool signal = false;
struct host1x_job *job, *n;
/*
* Walk the sync queue, reading the sync point registers as necessary,
* to consume as many sync queue entries as possible without blocking
*/
list_for_each_entry_safe(job, n, &cdma->sync_queue, list) {
struct host1x_syncpt *sp = job->syncpt;
/* Check whether this syncpt has completed, and bail if not */
if (!host1x_syncpt_is_expired(sp, job->syncpt_end) &&
!job->cancelled) {
/* Start timer on next pending syncpt */
if (job->timeout)
cdma_start_timer_locked(cdma, job);
break;
}
/* Cancel timeout, when a buffer completes */
if (cdma->timeout.client)
stop_cdma_timer_locked(cdma);
/* Unpin the memory */
host1x_job_unpin(job);
/* Pop push buffer slots */
if (job->num_slots) {
struct push_buffer *pb = &cdma->push_buffer;
host1x_pushbuffer_pop(pb, job->num_slots);
if (cdma->event == CDMA_EVENT_PUSH_BUFFER_SPACE)
signal = true;
}
list_del(&job->list);
host1x_job_put(job);
}
if (cdma->event == CDMA_EVENT_SYNC_QUEUE_EMPTY &&
list_empty(&cdma->sync_queue))
signal = true;
if (signal) {
cdma->event = CDMA_EVENT_NONE;
complete(&cdma->complete);
}
}
void host1x_cdma_update_sync_queue(struct host1x_cdma *cdma,
struct device *dev)
{
struct host1x *host1x = cdma_to_host1x(cdma);
u32 restart_addr, syncpt_incrs, syncpt_val;
struct host1x_job *job, *next_job = NULL;
syncpt_val = host1x_syncpt_load(cdma->timeout.syncpt);
dev_dbg(dev, "%s: starting cleanup (thresh %d)\n",
__func__, syncpt_val);
/*
* Move the sync_queue read pointer to the first entry that hasn't
* completed based on the current HW syncpt value. It's likely there
* won't be any (i.e. we're still at the head), but covers the case
* where a syncpt incr happens just prior/during the teardown.
*/
dev_dbg(dev, "%s: skip completed buffers still in sync_queue\n",
__func__);
list_for_each_entry(job, &cdma->sync_queue, list) {
if (syncpt_val < job->syncpt_end) {
if (!list_is_last(&job->list, &cdma->sync_queue))
next_job = list_next_entry(job, list);
goto syncpt_incr;
}
host1x_job_dump(dev, job);
}
/* all jobs have been completed */
job = NULL;
syncpt_incr:
/*
* Increment with CPU the remaining syncpts of a partially executed job.
*
* CDMA will continue execution starting with the next job or will get
* into idle state.
*/
if (next_job)
restart_addr = next_job->first_get;
else
restart_addr = cdma->last_pos;
if (!job)
goto resume;
/* do CPU increments for the remaining syncpts */
if (job->syncpt_recovery) {
dev_dbg(dev, "%s: perform CPU incr on pending buffers\n",
__func__);
/* won't need a timeout when replayed */
job->timeout = 0;
syncpt_incrs = job->syncpt_end - syncpt_val;
dev_dbg(dev, "%s: CPU incr (%d)\n", __func__, syncpt_incrs);
host1x_job_dump(dev, job);
/* safe to use CPU to incr syncpts */
host1x_hw_cdma_timeout_cpu_incr(host1x, cdma, job->first_get,
syncpt_incrs, job->syncpt_end,
job->num_slots);
dev_dbg(dev, "%s: finished sync_queue modification\n",
__func__);
} else {
struct host1x_job *failed_job = job;
host1x_job_dump(dev, job);
host1x_syncpt_set_locked(job->syncpt);
failed_job->cancelled = true;
list_for_each_entry_continue(job, &cdma->sync_queue, list) {
unsigned int i;
if (job->syncpt != failed_job->syncpt)
continue;
for (i = 0; i < job->num_slots; i++) {
unsigned int slot = (job->first_get/8 + i) %
HOST1X_PUSHBUFFER_SLOTS;
u32 *mapped = cdma->push_buffer.mapped;
/*
* Overwrite opcodes with 0 word writes
* to offset 0xbad. This does nothing but
* has a easily detected signature in debug
* traces.
*/
mapped[2*slot+0] = 0x1bad0000;
mapped[2*slot+1] = 0x1bad0000;
}
job->cancelled = true;
}
wmb();
update_cdma_locked(cdma);
}
resume:
/* roll back DMAGET and start up channel again */
host1x_hw_cdma_resume(host1x, cdma, restart_addr);
}
/*
* Create a cdma
*/
int host1x_cdma_init(struct host1x_cdma *cdma)
{
int err;
mutex_init(&cdma->lock);
init_completion(&cdma->complete);
INIT_LIST_HEAD(&cdma->sync_queue);
cdma->event = CDMA_EVENT_NONE;
cdma->running = false;
cdma->torndown = false;
err = host1x_pushbuffer_init(&cdma->push_buffer);
if (err)
return err;
return 0;
}
/*
* Destroy a cdma
*/
int host1x_cdma_deinit(struct host1x_cdma *cdma)
{
struct push_buffer *pb = &cdma->push_buffer;
struct host1x *host1x = cdma_to_host1x(cdma);
if (cdma->running) {
pr_warn("%s: CDMA still running\n", __func__);
return -EBUSY;
}
host1x_pushbuffer_destroy(pb);
host1x_hw_cdma_timeout_destroy(host1x, cdma);
return 0;
}
/*
* Begin a cdma submit
*/
int host1x_cdma_begin(struct host1x_cdma *cdma, struct host1x_job *job)
{
struct host1x *host1x = cdma_to_host1x(cdma);
mutex_lock(&cdma->lock);
/*
* Check if syncpoint was locked due to previous job timeout.
* This needs to be done within the cdma lock to avoid a race
* with the timeout handler.
*/
if (job->syncpt->locked) {
mutex_unlock(&cdma->lock);
return -EPERM;
}
if (job->timeout) {
/* init state on first submit with timeout value */
if (!cdma->timeout.initialized) {
int err;
err = host1x_hw_cdma_timeout_init(host1x, cdma);
if (err) {
mutex_unlock(&cdma->lock);
return err;
}
}
}
if (!cdma->running)
host1x_hw_cdma_start(host1x, cdma);
cdma->slots_free = 0;
cdma->slots_used = 0;
cdma->first_get = cdma->push_buffer.pos;
trace_host1x_cdma_begin(dev_name(job->channel->dev));
return 0;
}
/*
* Push two words into a push buffer slot
* Blocks as necessary if the push buffer is full.
*/
void host1x_cdma_push(struct host1x_cdma *cdma, u32 op1, u32 op2)
{
struct host1x *host1x = cdma_to_host1x(cdma);
struct push_buffer *pb = &cdma->push_buffer;
u32 slots_free = cdma->slots_free;
if (host1x_debug_trace_cmdbuf)
trace_host1x_cdma_push(dev_name(cdma_to_channel(cdma)->dev),
op1, op2);
if (slots_free == 0) {
host1x_hw_cdma_flush(host1x, cdma);
slots_free = host1x_cdma_wait_locked(cdma,
CDMA_EVENT_PUSH_BUFFER_SPACE);
}
cdma->slots_free = slots_free - 1;
cdma->slots_used++;
host1x_pushbuffer_push(pb, op1, op2);
}
/*
* Push four words into two consecutive push buffer slots. Note that extra
* care needs to be taken not to split the two slots across the end of the
* push buffer. Otherwise the RESTART opcode at the end of the push buffer
* that ensures processing will restart at the beginning will break up the
* four words.
*
* Blocks as necessary if the push buffer is full.
*/
void host1x_cdma_push_wide(struct host1x_cdma *cdma, u32 op1, u32 op2,
u32 op3, u32 op4)
{
struct host1x_channel *channel = cdma_to_channel(cdma);
struct host1x *host1x = cdma_to_host1x(cdma);
struct push_buffer *pb = &cdma->push_buffer;
unsigned int needed = 2, extra = 0, i;
unsigned int space = cdma->slots_free;
if (host1x_debug_trace_cmdbuf)
trace_host1x_cdma_push_wide(dev_name(channel->dev), op1, op2,
op3, op4);
/* compute number of extra slots needed for padding */
if (pb->pos + 16 > pb->size) {
extra = (pb->size - pb->pos) / 8;
needed += extra;
}
host1x_cdma_wait_pushbuffer_space(host1x, cdma, needed);
space = host1x_pushbuffer_space(pb);
cdma->slots_free = space - needed;
cdma->slots_used += needed;
/*
* Note that we rely on the fact that this is only used to submit wide
* gather opcodes, which consist of 3 words, and they are padded with
* a NOP to avoid having to deal with fractional slots (a slot always
* represents 2 words). The fourth opcode passed to this function will
* therefore always be a NOP.
*
* This works around a slight ambiguity when it comes to opcodes. For
* all current host1x incarnations the NOP opcode uses the exact same
* encoding (0x20000000), so we could hard-code the value here, but a
* new incarnation may change it and break that assumption.
*/
for (i = 0; i < extra; i++)
host1x_pushbuffer_push(pb, op4, op4);
host1x_pushbuffer_push(pb, op1, op2);
host1x_pushbuffer_push(pb, op3, op4);
}
/*
* End a cdma submit
* Kick off DMA, add job to the sync queue, and a number of slots to be freed
* from the pushbuffer. The handles for a submit must all be pinned at the same
* time, but they can be unpinned in smaller chunks.
*/
void host1x_cdma_end(struct host1x_cdma *cdma,
struct host1x_job *job)
{
struct host1x *host1x = cdma_to_host1x(cdma);
bool idle = list_empty(&cdma->sync_queue);
host1x_hw_cdma_flush(host1x, cdma);
job->first_get = cdma->first_get;
job->num_slots = cdma->slots_used;
host1x_job_get(job);
list_add_tail(&job->list, &cdma->sync_queue);
/* start timer on idle -> active transitions */
if (job->timeout && idle)
cdma_start_timer_locked(cdma, job);
trace_host1x_cdma_end(dev_name(job->channel->dev));
mutex_unlock(&cdma->lock);
}
/*
* Update cdma state according to current sync point values
*/
void host1x_cdma_update(struct host1x_cdma *cdma)
{
mutex_lock(&cdma->lock);
update_cdma_locked(cdma);
mutex_unlock(&cdma->lock);
}