linux/drivers/media/pci/ivtv/ivtv-irq.c

1089 lines
33 KiB
C
Raw Normal View History

/* interrupt handling
Copyright (C) 2003-2004 Kevin Thayer <nufan_wfk at yahoo.com>
Copyright (C) 2004 Chris Kennedy <c@groovy.org>
Copyright (C) 2005-2007 Hans Verkuil <hverkuil@xs4all.nl>
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., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
*/
#include "ivtv-driver.h"
#include "ivtv-queue.h"
#include "ivtv-udma.h"
#include "ivtv-irq.h"
#include "ivtv-mailbox.h"
#include "ivtv-vbi.h"
#include "ivtv-yuv.h"
#include <media/v4l2-event.h>
#define DMA_MAGIC_COOKIE 0x000001fe
static void ivtv_dma_dec_start(struct ivtv_stream *s);
static const int ivtv_stream_map[] = {
IVTV_ENC_STREAM_TYPE_MPG,
IVTV_ENC_STREAM_TYPE_YUV,
IVTV_ENC_STREAM_TYPE_PCM,
IVTV_ENC_STREAM_TYPE_VBI,
};
static void ivtv_pcm_work_handler(struct ivtv *itv)
{
struct ivtv_stream *s = &itv->streams[IVTV_ENC_STREAM_TYPE_PCM];
struct ivtv_buffer *buf;
/* Pass the PCM data to ivtv-alsa */
while (1) {
/*
* Users should not be using both the ALSA and V4L2 PCM audio
* capture interfaces at the same time. If the user is doing
* this, there maybe a buffer in q_io to grab, use, and put
* back in rotation.
*/
buf = ivtv_dequeue(s, &s->q_io);
if (buf == NULL)
buf = ivtv_dequeue(s, &s->q_full);
if (buf == NULL)
break;
if (buf->readpos < buf->bytesused)
itv->pcm_announce_callback(itv->alsa,
(u8 *)(buf->buf + buf->readpos),
(size_t)(buf->bytesused - buf->readpos));
ivtv_enqueue(s, buf, &s->q_free);
}
}
static void ivtv_pio_work_handler(struct ivtv *itv)
{
struct ivtv_stream *s = &itv->streams[itv->cur_pio_stream];
struct ivtv_buffer *buf;
int i = 0;
IVTV_DEBUG_HI_DMA("ivtv_pio_work_handler\n");
if (itv->cur_pio_stream < 0 || itv->cur_pio_stream >= IVTV_MAX_STREAMS ||
s->vdev == NULL || !ivtv_use_pio(s)) {
itv->cur_pio_stream = -1;
/* trigger PIO complete user interrupt */
write_reg(IVTV_IRQ_ENC_PIO_COMPLETE, 0x44);
return;
}
IVTV_DEBUG_HI_DMA("Process PIO %s\n", s->name);
list_for_each_entry(buf, &s->q_dma.list, list) {
u32 size = s->sg_processing[i].size & 0x3ffff;
/* Copy the data from the card to the buffer */
if (s->type == IVTV_DEC_STREAM_TYPE_VBI) {
memcpy_fromio(buf->buf, itv->dec_mem + s->sg_processing[i].src - IVTV_DECODER_OFFSET, size);
}
else {
memcpy_fromio(buf->buf, itv->enc_mem + s->sg_processing[i].src, size);
}
i++;
if (i == s->sg_processing_size)
break;
}
write_reg(IVTV_IRQ_ENC_PIO_COMPLETE, 0x44);
}
void ivtv_irq_work_handler(struct kthread_work *work)
{
struct ivtv *itv = container_of(work, struct ivtv, irq_work);
if (test_and_clear_bit(IVTV_F_I_WORK_HANDLER_PIO, &itv->i_flags))
ivtv_pio_work_handler(itv);
if (test_and_clear_bit(IVTV_F_I_WORK_HANDLER_VBI, &itv->i_flags))
ivtv_vbi_work_handler(itv);
if (test_and_clear_bit(IVTV_F_I_WORK_HANDLER_YUV, &itv->i_flags))
ivtv_yuv_work_handler(itv);
if (test_and_clear_bit(IVTV_F_I_WORK_HANDLER_PCM, &itv->i_flags))
ivtv_pcm_work_handler(itv);
}
/* Determine the required DMA size, setup enough buffers in the predma queue and
actually copy the data from the card to the buffers in case a PIO transfer is
required for this stream.
*/
static int stream_enc_dma_append(struct ivtv_stream *s, u32 data[CX2341X_MBOX_MAX_DATA])
{
struct ivtv *itv = s->itv;
struct ivtv_buffer *buf;
u32 bytes_needed = 0;
u32 offset, size;
u32 UVoffset = 0, UVsize = 0;
int skip_bufs = s->q_predma.buffers;
int idx = s->sg_pending_size;
int rc;
/* sanity checks */
if (s->vdev == NULL) {
IVTV_DEBUG_WARN("Stream %s not started\n", s->name);
return -1;
}
if (!test_bit(IVTV_F_S_CLAIMED, &s->s_flags)) {
IVTV_DEBUG_WARN("Stream %s not open\n", s->name);
return -1;
}
/* determine offset, size and PTS for the various streams */
switch (s->type) {
case IVTV_ENC_STREAM_TYPE_MPG:
offset = data[1];
size = data[2];
s->pending_pts = 0;
break;
case IVTV_ENC_STREAM_TYPE_YUV:
offset = data[1];
size = data[2];
UVoffset = data[3];
UVsize = data[4];
s->pending_pts = ((u64) data[5] << 32) | data[6];
break;
case IVTV_ENC_STREAM_TYPE_PCM:
offset = data[1] + 12;
size = data[2] - 12;
s->pending_pts = read_dec(offset - 8) |
((u64)(read_dec(offset - 12)) << 32);
if (itv->has_cx23415)
offset += IVTV_DECODER_OFFSET;
break;
case IVTV_ENC_STREAM_TYPE_VBI:
size = itv->vbi.enc_size * itv->vbi.fpi;
offset = read_enc(itv->vbi.enc_start - 4) + 12;
if (offset == 12) {
IVTV_DEBUG_INFO("VBI offset == 0\n");
return -1;
}
s->pending_pts = read_enc(offset - 4) | ((u64)read_enc(offset - 8) << 32);
break;
case IVTV_DEC_STREAM_TYPE_VBI:
size = read_dec(itv->vbi.dec_start + 4) + 8;
offset = read_dec(itv->vbi.dec_start) + itv->vbi.dec_start;
s->pending_pts = 0;
offset += IVTV_DECODER_OFFSET;
break;
default:
/* shouldn't happen */
return -1;
}
/* if this is the start of the DMA then fill in the magic cookie */
if (s->sg_pending_size == 0 && ivtv_use_dma(s)) {
if (itv->has_cx23415 && (s->type == IVTV_ENC_STREAM_TYPE_PCM ||
s->type == IVTV_DEC_STREAM_TYPE_VBI)) {
s->pending_backup = read_dec(offset - IVTV_DECODER_OFFSET);
write_dec_sync(cpu_to_le32(DMA_MAGIC_COOKIE), offset - IVTV_DECODER_OFFSET);
}
else {
s->pending_backup = read_enc(offset);
write_enc_sync(cpu_to_le32(DMA_MAGIC_COOKIE), offset);
}
s->pending_offset = offset;
}
bytes_needed = size;
if (s->type == IVTV_ENC_STREAM_TYPE_YUV) {
/* The size for the Y samples needs to be rounded upwards to a
multiple of the buf_size. The UV samples then start in the
next buffer. */
bytes_needed = s->buf_size * ((bytes_needed + s->buf_size - 1) / s->buf_size);
bytes_needed += UVsize;
}
IVTV_DEBUG_HI_DMA("%s %s: 0x%08x bytes at 0x%08x\n",
ivtv_use_pio(s) ? "PIO" : "DMA", s->name, bytes_needed, offset);
rc = ivtv_queue_move(s, &s->q_free, &s->q_full, &s->q_predma, bytes_needed);
if (rc < 0) { /* Insufficient buffers */
IVTV_DEBUG_WARN("Cannot obtain %d bytes for %s data transfer\n",
bytes_needed, s->name);
return -1;
}
if (rc && !s->buffers_stolen && test_bit(IVTV_F_S_APPL_IO, &s->s_flags)) {
IVTV_WARN("All %s stream buffers are full. Dropping data.\n", s->name);
IVTV_WARN("Cause: the application is not reading fast enough.\n");
}
s->buffers_stolen = rc;
/* got the buffers, now fill in sg_pending */
buf = list_entry(s->q_predma.list.next, struct ivtv_buffer, list);
memset(buf->buf, 0, 128);
list_for_each_entry(buf, &s->q_predma.list, list) {
if (skip_bufs-- > 0)
continue;
s->sg_pending[idx].dst = buf->dma_handle;
s->sg_pending[idx].src = offset;
s->sg_pending[idx].size = s->buf_size;
buf->bytesused = min(size, s->buf_size);
buf->dma_xfer_cnt = s->dma_xfer_cnt;
s->q_predma.bytesused += buf->bytesused;
size -= buf->bytesused;
offset += s->buf_size;
/* Sync SG buffers */
ivtv_buf_sync_for_device(s, buf);
if (size == 0) { /* YUV */
/* process the UV section */
offset = UVoffset;
size = UVsize;
}
idx++;
}
s->sg_pending_size = idx;
return 0;
}
static void dma_post(struct ivtv_stream *s)
{
struct ivtv *itv = s->itv;
struct ivtv_buffer *buf = NULL;
struct list_head *p;
u32 offset;
__le32 *u32buf;
int x = 0;
IVTV_DEBUG_HI_DMA("%s %s completed (%x)\n", ivtv_use_pio(s) ? "PIO" : "DMA",
s->name, s->dma_offset);
list_for_each(p, &s->q_dma.list) {
buf = list_entry(p, struct ivtv_buffer, list);
u32buf = (__le32 *)buf->buf;
/* Sync Buffer */
ivtv_buf_sync_for_cpu(s, buf);
if (x == 0 && ivtv_use_dma(s)) {
offset = s->dma_last_offset;
if (u32buf[offset / 4] != DMA_MAGIC_COOKIE)
{
for (offset = 0; offset < 64; offset++) {
if (u32buf[offset] == DMA_MAGIC_COOKIE) {
break;
}
}
offset *= 4;
if (offset == 256) {
IVTV_DEBUG_WARN("%s: Couldn't find start of buffer within the first 256 bytes\n", s->name);
offset = s->dma_last_offset;
}
if (s->dma_last_offset != offset)
IVTV_DEBUG_WARN("%s: offset %d -> %d\n", s->name, s->dma_last_offset, offset);
s->dma_last_offset = offset;
}
if (itv->has_cx23415 && (s->type == IVTV_ENC_STREAM_TYPE_PCM ||
s->type == IVTV_DEC_STREAM_TYPE_VBI)) {
write_dec_sync(0, s->dma_offset - IVTV_DECODER_OFFSET);
}
else {
write_enc_sync(0, s->dma_offset);
}
if (offset) {
buf->bytesused -= offset;
memcpy(buf->buf, buf->buf + offset, buf->bytesused + offset);
}
*u32buf = cpu_to_le32(s->dma_backup);
}
x++;
/* flag byteswap ABCD -> DCBA for MPG & VBI data outside irq */
if (s->type == IVTV_ENC_STREAM_TYPE_MPG ||
s->type == IVTV_ENC_STREAM_TYPE_VBI)
buf->b_flags |= IVTV_F_B_NEED_BUF_SWAP;
}
if (buf)
buf->bytesused += s->dma_last_offset;
if (buf && s->type == IVTV_DEC_STREAM_TYPE_VBI) {
list_for_each_entry(buf, &s->q_dma.list, list) {
/* Parse and Groom VBI Data */
s->q_dma.bytesused -= buf->bytesused;
ivtv_process_vbi_data(itv, buf, 0, s->type);
s->q_dma.bytesused += buf->bytesused;
}
if (s->fh == NULL) {
ivtv_queue_move(s, &s->q_dma, NULL, &s->q_free, 0);
return;
}
}
ivtv_queue_move(s, &s->q_dma, NULL, &s->q_full, s->q_dma.bytesused);
if (s->type == IVTV_ENC_STREAM_TYPE_PCM &&
itv->pcm_announce_callback != NULL) {
/*
* Set up the work handler to pass the data to ivtv-alsa.
*
* We just use q_full and let the work handler race with users
* making ivtv-fileops.c calls on the PCM device node.
*
* Users should not be using both the ALSA and V4L2 PCM audio
* capture interfaces at the same time. If the user does this,
* fragments of data will just go out each interface as they
* race for PCM data.
*/
set_bit(IVTV_F_I_WORK_HANDLER_PCM, &itv->i_flags);
set_bit(IVTV_F_I_HAVE_WORK, &itv->i_flags);
}
if (s->fh)
wake_up(&s->waitq);
}
void ivtv_dma_stream_dec_prepare(struct ivtv_stream *s, u32 offset, int lock)
{
struct ivtv *itv = s->itv;
V4L/DVB (6717): ivtv: Initial merge of video48 yuv handling into the IVTV_IOC_DMA_FRAME framework Previously, all yuv data written to /dev/video48 had only basic support with no double buffering to avoid display tearing. With this patch, yuv frames written to video48 are now handled by the existing IVTV_IOC_DMA_FRAME framework. As such, the frames are hardware buffered to avoid tearing, and honour scaling mode & field order options. Unlike the proprietary IVTV_IOC_DMA_FRAME ioctl, all parameters are controlled by the V4L2 API. Due to mpeg & yuv output restrictions being different, their V4L2 output controls have been separated. To control the yuv output, the V4L2 calls must be done via video48. If the ivtvfb module is loaded, there will be one side effect to this merge. The yuv output window will be constrained to the visible framebuffer area. In the event that a virtual framebuffer size is being used, the limit to the output size will be the virtual dimensions, but only the portion that falls within the currently visible area of the framebuffer will be shown. Like the IVTV_IOC_DMA_FRAME ioctl, the supplied frames must be padded to 720 pixels wide. However the height must only be padded up the nearest multiple of 32. This would mean an image of 102 lines must be padded to 128. As long as the true source image size is given, the padding will not be visible in the final output. Signed-off-by: Ian Armstrong <ian@iarmst.demon.co.uk> Signed-off-by: Hans Verkuil <hverkuil@xs4all.nl> Signed-off-by: Mauro Carvalho Chehab <mchehab@infradead.org>
2007-11-05 17:27:09 +00:00
struct yuv_playback_info *yi = &itv->yuv_info;
u8 frame = yi->draw_frame;
struct yuv_frame_info *f = &yi->new_frame_info[frame];
struct ivtv_buffer *buf;
V4L/DVB (6717): ivtv: Initial merge of video48 yuv handling into the IVTV_IOC_DMA_FRAME framework Previously, all yuv data written to /dev/video48 had only basic support with no double buffering to avoid display tearing. With this patch, yuv frames written to video48 are now handled by the existing IVTV_IOC_DMA_FRAME framework. As such, the frames are hardware buffered to avoid tearing, and honour scaling mode & field order options. Unlike the proprietary IVTV_IOC_DMA_FRAME ioctl, all parameters are controlled by the V4L2 API. Due to mpeg & yuv output restrictions being different, their V4L2 output controls have been separated. To control the yuv output, the V4L2 calls must be done via video48. If the ivtvfb module is loaded, there will be one side effect to this merge. The yuv output window will be constrained to the visible framebuffer area. In the event that a virtual framebuffer size is being used, the limit to the output size will be the virtual dimensions, but only the portion that falls within the currently visible area of the framebuffer will be shown. Like the IVTV_IOC_DMA_FRAME ioctl, the supplied frames must be padded to 720 pixels wide. However the height must only be padded up the nearest multiple of 32. This would mean an image of 102 lines must be padded to 128. As long as the true source image size is given, the padding will not be visible in the final output. Signed-off-by: Ian Armstrong <ian@iarmst.demon.co.uk> Signed-off-by: Hans Verkuil <hverkuil@xs4all.nl> Signed-off-by: Mauro Carvalho Chehab <mchehab@infradead.org>
2007-11-05 17:27:09 +00:00
u32 y_size = 720 * ((f->src_h + 31) & ~31);
u32 uv_offset = offset + IVTV_YUV_BUFFER_UV_OFFSET;
int y_done = 0;
int bytes_written = 0;
unsigned long flags = 0;
int idx = 0;
IVTV_DEBUG_HI_DMA("DEC PREPARE DMA %s: %08x %08x\n", s->name, s->q_predma.bytesused, offset);
V4L/DVB (6717): ivtv: Initial merge of video48 yuv handling into the IVTV_IOC_DMA_FRAME framework Previously, all yuv data written to /dev/video48 had only basic support with no double buffering to avoid display tearing. With this patch, yuv frames written to video48 are now handled by the existing IVTV_IOC_DMA_FRAME framework. As such, the frames are hardware buffered to avoid tearing, and honour scaling mode & field order options. Unlike the proprietary IVTV_IOC_DMA_FRAME ioctl, all parameters are controlled by the V4L2 API. Due to mpeg & yuv output restrictions being different, their V4L2 output controls have been separated. To control the yuv output, the V4L2 calls must be done via video48. If the ivtvfb module is loaded, there will be one side effect to this merge. The yuv output window will be constrained to the visible framebuffer area. In the event that a virtual framebuffer size is being used, the limit to the output size will be the virtual dimensions, but only the portion that falls within the currently visible area of the framebuffer will be shown. Like the IVTV_IOC_DMA_FRAME ioctl, the supplied frames must be padded to 720 pixels wide. However the height must only be padded up the nearest multiple of 32. This would mean an image of 102 lines must be padded to 128. As long as the true source image size is given, the padding will not be visible in the final output. Signed-off-by: Ian Armstrong <ian@iarmst.demon.co.uk> Signed-off-by: Hans Verkuil <hverkuil@xs4all.nl> Signed-off-by: Mauro Carvalho Chehab <mchehab@infradead.org>
2007-11-05 17:27:09 +00:00
/* Insert buffer block for YUV if needed */
if (s->type == IVTV_DEC_STREAM_TYPE_YUV && f->offset_y) {
if (yi->blanking_dmaptr) {
s->sg_pending[idx].src = yi->blanking_dmaptr;
s->sg_pending[idx].dst = offset;
s->sg_pending[idx].size = 720 * 16;
}
offset += 720 * 16;
idx++;
}
list_for_each_entry(buf, &s->q_predma.list, list) {
/* YUV UV Offset from Y Buffer */
if (s->type == IVTV_DEC_STREAM_TYPE_YUV && !y_done &&
(bytes_written + buf->bytesused) >= y_size) {
s->sg_pending[idx].src = buf->dma_handle;
s->sg_pending[idx].dst = offset;
s->sg_pending[idx].size = y_size - bytes_written;
offset = uv_offset;
if (s->sg_pending[idx].size != buf->bytesused) {
idx++;
s->sg_pending[idx].src =
buf->dma_handle + s->sg_pending[idx - 1].size;
s->sg_pending[idx].dst = offset;
s->sg_pending[idx].size =
buf->bytesused - s->sg_pending[idx - 1].size;
offset += s->sg_pending[idx].size;
}
y_done = 1;
} else {
s->sg_pending[idx].src = buf->dma_handle;
s->sg_pending[idx].dst = offset;
s->sg_pending[idx].size = buf->bytesused;
offset += buf->bytesused;
}
bytes_written += buf->bytesused;
/* Sync SG buffers */
ivtv_buf_sync_for_device(s, buf);
idx++;
}
s->sg_pending_size = idx;
/* Sync Hardware SG List of buffers */
ivtv_stream_sync_for_device(s);
if (lock)
spin_lock_irqsave(&itv->dma_reg_lock, flags);
if (!test_bit(IVTV_F_I_DMA, &itv->i_flags)) {
ivtv_dma_dec_start(s);
}
else {
set_bit(IVTV_F_S_DMA_PENDING, &s->s_flags);
}
if (lock)
spin_unlock_irqrestore(&itv->dma_reg_lock, flags);
}
static void ivtv_dma_enc_start_xfer(struct ivtv_stream *s)
{
struct ivtv *itv = s->itv;
s->sg_dma->src = cpu_to_le32(s->sg_processing[s->sg_processed].src);
s->sg_dma->dst = cpu_to_le32(s->sg_processing[s->sg_processed].dst);
s->sg_dma->size = cpu_to_le32(s->sg_processing[s->sg_processed].size | 0x80000000);
s->sg_processed++;
/* Sync Hardware SG List of buffers */
ivtv_stream_sync_for_device(s);
write_reg(s->sg_handle, IVTV_REG_ENCDMAADDR);
write_reg_sync(read_reg(IVTV_REG_DMAXFER) | 0x02, IVTV_REG_DMAXFER);
itv->dma_timer.expires = jiffies + msecs_to_jiffies(300);
add_timer(&itv->dma_timer);
}
static void ivtv_dma_dec_start_xfer(struct ivtv_stream *s)
{
struct ivtv *itv = s->itv;
s->sg_dma->src = cpu_to_le32(s->sg_processing[s->sg_processed].src);
s->sg_dma->dst = cpu_to_le32(s->sg_processing[s->sg_processed].dst);
s->sg_dma->size = cpu_to_le32(s->sg_processing[s->sg_processed].size | 0x80000000);
s->sg_processed++;
/* Sync Hardware SG List of buffers */
ivtv_stream_sync_for_device(s);
write_reg(s->sg_handle, IVTV_REG_DECDMAADDR);
write_reg_sync(read_reg(IVTV_REG_DMAXFER) | 0x01, IVTV_REG_DMAXFER);
itv->dma_timer.expires = jiffies + msecs_to_jiffies(300);
add_timer(&itv->dma_timer);
}
/* start the encoder DMA */
static void ivtv_dma_enc_start(struct ivtv_stream *s)
{
struct ivtv *itv = s->itv;
struct ivtv_stream *s_vbi = &itv->streams[IVTV_ENC_STREAM_TYPE_VBI];
int i;
IVTV_DEBUG_HI_DMA("start %s for %s\n", ivtv_use_dma(s) ? "DMA" : "PIO", s->name);
if (s->q_predma.bytesused)
ivtv_queue_move(s, &s->q_predma, NULL, &s->q_dma, s->q_predma.bytesused);
if (ivtv_use_dma(s))
s->sg_pending[s->sg_pending_size - 1].size += 256;
/* If this is an MPEG stream, and VBI data is also pending, then append the
VBI DMA to the MPEG DMA and transfer both sets of data at once.
VBI DMA is a second class citizen compared to MPEG and mixing them together
will confuse the firmware (the end of a VBI DMA is seen as the end of a
MPEG DMA, thus effectively dropping an MPEG frame). So instead we make
sure we only use the MPEG DMA to transfer the VBI DMA if both are in
use. This way no conflicts occur. */
clear_bit(IVTV_F_S_DMA_HAS_VBI, &s->s_flags);
if (s->type == IVTV_ENC_STREAM_TYPE_MPG && s_vbi->sg_pending_size &&
s->sg_pending_size + s_vbi->sg_pending_size <= s->buffers) {
ivtv_queue_move(s_vbi, &s_vbi->q_predma, NULL, &s_vbi->q_dma, s_vbi->q_predma.bytesused);
if (ivtv_use_dma(s_vbi))
s_vbi->sg_pending[s_vbi->sg_pending_size - 1].size += 256;
for (i = 0; i < s_vbi->sg_pending_size; i++) {
s->sg_pending[s->sg_pending_size++] = s_vbi->sg_pending[i];
}
s_vbi->dma_offset = s_vbi->pending_offset;
s_vbi->sg_pending_size = 0;
s_vbi->dma_xfer_cnt++;
set_bit(IVTV_F_S_DMA_HAS_VBI, &s->s_flags);
IVTV_DEBUG_HI_DMA("include DMA for %s\n", s_vbi->name);
}
s->dma_xfer_cnt++;
memcpy(s->sg_processing, s->sg_pending, sizeof(struct ivtv_sg_host_element) * s->sg_pending_size);
s->sg_processing_size = s->sg_pending_size;
s->sg_pending_size = 0;
s->sg_processed = 0;
s->dma_offset = s->pending_offset;
s->dma_backup = s->pending_backup;
s->dma_pts = s->pending_pts;
if (ivtv_use_pio(s)) {
set_bit(IVTV_F_I_WORK_HANDLER_PIO, &itv->i_flags);
set_bit(IVTV_F_I_HAVE_WORK, &itv->i_flags);
set_bit(IVTV_F_I_PIO, &itv->i_flags);
itv->cur_pio_stream = s->type;
}
else {
itv->dma_retries = 0;
ivtv_dma_enc_start_xfer(s);
set_bit(IVTV_F_I_DMA, &itv->i_flags);
itv->cur_dma_stream = s->type;
}
}
static void ivtv_dma_dec_start(struct ivtv_stream *s)
{
struct ivtv *itv = s->itv;
if (s->q_predma.bytesused)
ivtv_queue_move(s, &s->q_predma, NULL, &s->q_dma, s->q_predma.bytesused);
s->dma_xfer_cnt++;
memcpy(s->sg_processing, s->sg_pending, sizeof(struct ivtv_sg_host_element) * s->sg_pending_size);
s->sg_processing_size = s->sg_pending_size;
s->sg_pending_size = 0;
s->sg_processed = 0;
IVTV_DEBUG_HI_DMA("start DMA for %s\n", s->name);
itv->dma_retries = 0;
ivtv_dma_dec_start_xfer(s);
set_bit(IVTV_F_I_DMA, &itv->i_flags);
itv->cur_dma_stream = s->type;
}
static void ivtv_irq_dma_read(struct ivtv *itv)
{
struct ivtv_stream *s = NULL;
struct ivtv_buffer *buf;
int hw_stream_type = 0;
IVTV_DEBUG_HI_IRQ("DEC DMA READ\n");
del_timer(&itv->dma_timer);
if (!test_bit(IVTV_F_I_UDMA, &itv->i_flags) && itv->cur_dma_stream < 0)
return;
if (!test_bit(IVTV_F_I_UDMA, &itv->i_flags)) {
s = &itv->streams[itv->cur_dma_stream];
ivtv_stream_sync_for_cpu(s);
if (read_reg(IVTV_REG_DMASTATUS) & 0x14) {
IVTV_DEBUG_WARN("DEC DMA ERROR %x (xfer %d of %d, retry %d)\n",
read_reg(IVTV_REG_DMASTATUS),
s->sg_processed, s->sg_processing_size, itv->dma_retries);
write_reg(read_reg(IVTV_REG_DMASTATUS) & 3, IVTV_REG_DMASTATUS);
if (itv->dma_retries == 3) {
/* Too many retries, give up on this frame */
itv->dma_retries = 0;
s->sg_processed = s->sg_processing_size;
}
else {
/* Retry, starting with the first xfer segment.
Just retrying the current segment is not sufficient. */
s->sg_processed = 0;
itv->dma_retries++;
}
}
if (s->sg_processed < s->sg_processing_size) {
/* DMA next buffer */
ivtv_dma_dec_start_xfer(s);
return;
}
if (s->type == IVTV_DEC_STREAM_TYPE_YUV)
hw_stream_type = 2;
IVTV_DEBUG_HI_DMA("DEC DATA READ %s: %d\n", s->name, s->q_dma.bytesused);
/* For some reason must kick the firmware, like PIO mode,
I think this tells the firmware we are done and the size
of the xfer so it can calculate what we need next.
I think we can do this part ourselves but would have to
fully calculate xfer info ourselves and not use interrupts
*/
ivtv_vapi(itv, CX2341X_DEC_SCHED_DMA_FROM_HOST, 3, 0, s->q_dma.bytesused,
hw_stream_type);
/* Free last DMA call */
while ((buf = ivtv_dequeue(s, &s->q_dma)) != NULL) {
ivtv_buf_sync_for_cpu(s, buf);
ivtv_enqueue(s, buf, &s->q_free);
}
wake_up(&s->waitq);
}
clear_bit(IVTV_F_I_UDMA, &itv->i_flags);
clear_bit(IVTV_F_I_DMA, &itv->i_flags);
itv->cur_dma_stream = -1;
wake_up(&itv->dma_waitq);
}
static void ivtv_irq_enc_dma_complete(struct ivtv *itv)
{
u32 data[CX2341X_MBOX_MAX_DATA];
struct ivtv_stream *s;
ivtv_api_get_data(&itv->enc_mbox, IVTV_MBOX_DMA_END, 2, data);
IVTV_DEBUG_HI_IRQ("ENC DMA COMPLETE %x %d (%d)\n", data[0], data[1], itv->cur_dma_stream);
del_timer(&itv->dma_timer);
if (itv->cur_dma_stream < 0)
return;
s = &itv->streams[itv->cur_dma_stream];
ivtv_stream_sync_for_cpu(s);
if (data[0] & 0x18) {
IVTV_DEBUG_WARN("ENC DMA ERROR %x (offset %08x, xfer %d of %d, retry %d)\n", data[0],
s->dma_offset, s->sg_processed, s->sg_processing_size, itv->dma_retries);
write_reg(read_reg(IVTV_REG_DMASTATUS) & 3, IVTV_REG_DMASTATUS);
if (itv->dma_retries == 3) {
/* Too many retries, give up on this frame */
itv->dma_retries = 0;
s->sg_processed = s->sg_processing_size;
}
else {
/* Retry, starting with the first xfer segment.
Just retrying the current segment is not sufficient. */
s->sg_processed = 0;
itv->dma_retries++;
}
}
if (s->sg_processed < s->sg_processing_size) {
/* DMA next buffer */
ivtv_dma_enc_start_xfer(s);
return;
}
clear_bit(IVTV_F_I_DMA, &itv->i_flags);
itv->cur_dma_stream = -1;
dma_post(s);
if (test_and_clear_bit(IVTV_F_S_DMA_HAS_VBI, &s->s_flags)) {
s = &itv->streams[IVTV_ENC_STREAM_TYPE_VBI];
dma_post(s);
}
s->sg_processing_size = 0;
s->sg_processed = 0;
wake_up(&itv->dma_waitq);
}
static void ivtv_irq_enc_pio_complete(struct ivtv *itv)
{
struct ivtv_stream *s;
if (itv->cur_pio_stream < 0 || itv->cur_pio_stream >= IVTV_MAX_STREAMS) {
itv->cur_pio_stream = -1;
return;
}
s = &itv->streams[itv->cur_pio_stream];
IVTV_DEBUG_HI_IRQ("ENC PIO COMPLETE %s\n", s->name);
clear_bit(IVTV_F_I_PIO, &itv->i_flags);
itv->cur_pio_stream = -1;
dma_post(s);
if (s->type == IVTV_ENC_STREAM_TYPE_MPG)
ivtv_vapi(itv, CX2341X_ENC_SCHED_DMA_TO_HOST, 3, 0, 0, 0);
else if (s->type == IVTV_ENC_STREAM_TYPE_YUV)
ivtv_vapi(itv, CX2341X_ENC_SCHED_DMA_TO_HOST, 3, 0, 0, 1);
else if (s->type == IVTV_ENC_STREAM_TYPE_PCM)
ivtv_vapi(itv, CX2341X_ENC_SCHED_DMA_TO_HOST, 3, 0, 0, 2);
clear_bit(IVTV_F_I_PIO, &itv->i_flags);
if (test_and_clear_bit(IVTV_F_S_DMA_HAS_VBI, &s->s_flags)) {
s = &itv->streams[IVTV_ENC_STREAM_TYPE_VBI];
dma_post(s);
}
wake_up(&itv->dma_waitq);
}
static void ivtv_irq_dma_err(struct ivtv *itv)
{
u32 data[CX2341X_MBOX_MAX_DATA];
[media] ivtv: Fix corrective action taken upon DMA ERR interrupt to avoid hang After upgrading the kernel from stock Ubuntu 7.10 to 10.04, with no hardware changes, I started getting the dreaded DMA TIMEOUT errors, followed by inability to encode until the machine was rebooted. I came across a post from Andy in March (http://www.gossamer-threads.com/lists/ivtv/users/40943#40943) where he speculates that perhaps the corrective actions being taken after a DMA ERROR are not sufficient to recover the situation. After some testing I suspect that this is indeed the case, and that in fact the corrective action may be what hangs the card's DMA engine, rather than the original error. Specifically these DMA ERROR IRQs seem to present with two different values in the IVTV_REG_DMASTATUS register: 0x11 and 0x13. The current corrective action is to clear that status register back to 0x01 or 0x03, and then issue the next DMA request. In the case of a 0x13 this seems to result in a minor glitch in the encoded stream due to the failed transfer that was not retried, but otherwise things continue OK. In the case of a 0x11 the card's DMA write engine is never heard from again, and a DMA TIMEOUT follows shortly after. 0x11 is the killer. I suspect that the two cases need to be handled differently. The difference is in bit 1 (0x02), which is set when the error is about to be successfully recovered, and clear when things are about to go bad. Bit 1 of DMASTATUS is described differently in different places either as a positive "write finished", or an inverted "write busy". If we take the first definition, then when an error arises with state 0x11, it means that the write did not complete. It makes sense to start a new transfer, as in the current code. But if we take the second definition, then 0x11 means "an error but the write engine is still busy". Trying to feed it a new transfer in this situation might not be a good idea. As an experiment, I added code to ignore the DMA ERROR IRQ if DMASTATUS is 0x11. I.e., don't start a new transfer, don't clear our flags, etc. The hope was that the card would complete the transfer and issue a ENC DMA COMPLETE, either successfully or with an error condition there. However the card still hung. The only remaining corrective action being taken with a 0x11 status was then the write back to the status register to clear the error, i.e. DMASTATUS = DMASTATUS & ~3. This would have the effect of clearing the error bit 4, while leaving the lower bits indicating DMA write busy. Strangely enough, removing this write to the status register solved the problem! If the DMA ERROR IRQ with DMASTATUS=0x11 is completely ignored, with no corrective action at all, then the card will complete the transfer and issue a new IRQ. If the status register is written to when it has the value 0x11, then the DMA engine hangs. Perhaps it's illegal to write to DMASTATUS while the read or write busy bit is set? At any rate, it appears that the current corrective action is indeed making things worse rather than better. I put together a patch that modifies ivtv_irq_dma_err to do the following: - Don't write back to IVTV_REG_DMASTATUS. - If write-busy is asserted, leave the card alone. Just extend the timeout slightly. - If write-busy is de-asserted, retry the current transfer. This has completely fixed my DMA TIMEOUT woes. DMA ERR events still occur, but now they seem to be correctly handled. 0x11 events no longer hang the card, and 0x13 events no longer result in a glitch in the stream, as the failed transfer is retried. I'm happy. I've inlined the patch below in case it is of interest. As described above, I have a theory about why it works (based on a different interpretation of bit 1 of DMASTATUS), but I can't guarantee that my theory is correct. There may be another explanation, or it may be a fluke. Maybe ignoring that IRQ entirely would be equally effective? Maybe the status register read/writeback sequence is race condition if the card changes it in the mean time? Also as I am using a PVR-150 only, I have not been able to test it on other cards, which may be especially relevant for 350s that support concurrent decoding. Hopefully the patch does not break the DMA READ path. Mike [awalls@md.metrocast.net: Modified patch to add a verbose comment, make minor brace reformats, and clear the error flags in the IVTV_REG_DMASTATUS iff both read and write DMA were not in progress. Mike's conjecture about a race condition with the writeback is correct; it can confuse the DMA engine.] [Comment and analysis from the ML post by Michael <mike@rsy.com>] Signed-off-by: Andy Walls <awalls@md.metrocast.net> Cc: stable@kernel.org Signed-off-by: Mauro Carvalho Chehab <mchehab@redhat.com>
2011-02-26 04:56:34 +00:00
u32 status;
del_timer(&itv->dma_timer);
[media] ivtv: Fix corrective action taken upon DMA ERR interrupt to avoid hang After upgrading the kernel from stock Ubuntu 7.10 to 10.04, with no hardware changes, I started getting the dreaded DMA TIMEOUT errors, followed by inability to encode until the machine was rebooted. I came across a post from Andy in March (http://www.gossamer-threads.com/lists/ivtv/users/40943#40943) where he speculates that perhaps the corrective actions being taken after a DMA ERROR are not sufficient to recover the situation. After some testing I suspect that this is indeed the case, and that in fact the corrective action may be what hangs the card's DMA engine, rather than the original error. Specifically these DMA ERROR IRQs seem to present with two different values in the IVTV_REG_DMASTATUS register: 0x11 and 0x13. The current corrective action is to clear that status register back to 0x01 or 0x03, and then issue the next DMA request. In the case of a 0x13 this seems to result in a minor glitch in the encoded stream due to the failed transfer that was not retried, but otherwise things continue OK. In the case of a 0x11 the card's DMA write engine is never heard from again, and a DMA TIMEOUT follows shortly after. 0x11 is the killer. I suspect that the two cases need to be handled differently. The difference is in bit 1 (0x02), which is set when the error is about to be successfully recovered, and clear when things are about to go bad. Bit 1 of DMASTATUS is described differently in different places either as a positive "write finished", or an inverted "write busy". If we take the first definition, then when an error arises with state 0x11, it means that the write did not complete. It makes sense to start a new transfer, as in the current code. But if we take the second definition, then 0x11 means "an error but the write engine is still busy". Trying to feed it a new transfer in this situation might not be a good idea. As an experiment, I added code to ignore the DMA ERROR IRQ if DMASTATUS is 0x11. I.e., don't start a new transfer, don't clear our flags, etc. The hope was that the card would complete the transfer and issue a ENC DMA COMPLETE, either successfully or with an error condition there. However the card still hung. The only remaining corrective action being taken with a 0x11 status was then the write back to the status register to clear the error, i.e. DMASTATUS = DMASTATUS & ~3. This would have the effect of clearing the error bit 4, while leaving the lower bits indicating DMA write busy. Strangely enough, removing this write to the status register solved the problem! If the DMA ERROR IRQ with DMASTATUS=0x11 is completely ignored, with no corrective action at all, then the card will complete the transfer and issue a new IRQ. If the status register is written to when it has the value 0x11, then the DMA engine hangs. Perhaps it's illegal to write to DMASTATUS while the read or write busy bit is set? At any rate, it appears that the current corrective action is indeed making things worse rather than better. I put together a patch that modifies ivtv_irq_dma_err to do the following: - Don't write back to IVTV_REG_DMASTATUS. - If write-busy is asserted, leave the card alone. Just extend the timeout slightly. - If write-busy is de-asserted, retry the current transfer. This has completely fixed my DMA TIMEOUT woes. DMA ERR events still occur, but now they seem to be correctly handled. 0x11 events no longer hang the card, and 0x13 events no longer result in a glitch in the stream, as the failed transfer is retried. I'm happy. I've inlined the patch below in case it is of interest. As described above, I have a theory about why it works (based on a different interpretation of bit 1 of DMASTATUS), but I can't guarantee that my theory is correct. There may be another explanation, or it may be a fluke. Maybe ignoring that IRQ entirely would be equally effective? Maybe the status register read/writeback sequence is race condition if the card changes it in the mean time? Also as I am using a PVR-150 only, I have not been able to test it on other cards, which may be especially relevant for 350s that support concurrent decoding. Hopefully the patch does not break the DMA READ path. Mike [awalls@md.metrocast.net: Modified patch to add a verbose comment, make minor brace reformats, and clear the error flags in the IVTV_REG_DMASTATUS iff both read and write DMA were not in progress. Mike's conjecture about a race condition with the writeback is correct; it can confuse the DMA engine.] [Comment and analysis from the ML post by Michael <mike@rsy.com>] Signed-off-by: Andy Walls <awalls@md.metrocast.net> Cc: stable@kernel.org Signed-off-by: Mauro Carvalho Chehab <mchehab@redhat.com>
2011-02-26 04:56:34 +00:00
ivtv_api_get_data(&itv->enc_mbox, IVTV_MBOX_DMA_END, 2, data);
[media] ivtv: Fix corrective action taken upon DMA ERR interrupt to avoid hang After upgrading the kernel from stock Ubuntu 7.10 to 10.04, with no hardware changes, I started getting the dreaded DMA TIMEOUT errors, followed by inability to encode until the machine was rebooted. I came across a post from Andy in March (http://www.gossamer-threads.com/lists/ivtv/users/40943#40943) where he speculates that perhaps the corrective actions being taken after a DMA ERROR are not sufficient to recover the situation. After some testing I suspect that this is indeed the case, and that in fact the corrective action may be what hangs the card's DMA engine, rather than the original error. Specifically these DMA ERROR IRQs seem to present with two different values in the IVTV_REG_DMASTATUS register: 0x11 and 0x13. The current corrective action is to clear that status register back to 0x01 or 0x03, and then issue the next DMA request. In the case of a 0x13 this seems to result in a minor glitch in the encoded stream due to the failed transfer that was not retried, but otherwise things continue OK. In the case of a 0x11 the card's DMA write engine is never heard from again, and a DMA TIMEOUT follows shortly after. 0x11 is the killer. I suspect that the two cases need to be handled differently. The difference is in bit 1 (0x02), which is set when the error is about to be successfully recovered, and clear when things are about to go bad. Bit 1 of DMASTATUS is described differently in different places either as a positive "write finished", or an inverted "write busy". If we take the first definition, then when an error arises with state 0x11, it means that the write did not complete. It makes sense to start a new transfer, as in the current code. But if we take the second definition, then 0x11 means "an error but the write engine is still busy". Trying to feed it a new transfer in this situation might not be a good idea. As an experiment, I added code to ignore the DMA ERROR IRQ if DMASTATUS is 0x11. I.e., don't start a new transfer, don't clear our flags, etc. The hope was that the card would complete the transfer and issue a ENC DMA COMPLETE, either successfully or with an error condition there. However the card still hung. The only remaining corrective action being taken with a 0x11 status was then the write back to the status register to clear the error, i.e. DMASTATUS = DMASTATUS & ~3. This would have the effect of clearing the error bit 4, while leaving the lower bits indicating DMA write busy. Strangely enough, removing this write to the status register solved the problem! If the DMA ERROR IRQ with DMASTATUS=0x11 is completely ignored, with no corrective action at all, then the card will complete the transfer and issue a new IRQ. If the status register is written to when it has the value 0x11, then the DMA engine hangs. Perhaps it's illegal to write to DMASTATUS while the read or write busy bit is set? At any rate, it appears that the current corrective action is indeed making things worse rather than better. I put together a patch that modifies ivtv_irq_dma_err to do the following: - Don't write back to IVTV_REG_DMASTATUS. - If write-busy is asserted, leave the card alone. Just extend the timeout slightly. - If write-busy is de-asserted, retry the current transfer. This has completely fixed my DMA TIMEOUT woes. DMA ERR events still occur, but now they seem to be correctly handled. 0x11 events no longer hang the card, and 0x13 events no longer result in a glitch in the stream, as the failed transfer is retried. I'm happy. I've inlined the patch below in case it is of interest. As described above, I have a theory about why it works (based on a different interpretation of bit 1 of DMASTATUS), but I can't guarantee that my theory is correct. There may be another explanation, or it may be a fluke. Maybe ignoring that IRQ entirely would be equally effective? Maybe the status register read/writeback sequence is race condition if the card changes it in the mean time? Also as I am using a PVR-150 only, I have not been able to test it on other cards, which may be especially relevant for 350s that support concurrent decoding. Hopefully the patch does not break the DMA READ path. Mike [awalls@md.metrocast.net: Modified patch to add a verbose comment, make minor brace reformats, and clear the error flags in the IVTV_REG_DMASTATUS iff both read and write DMA were not in progress. Mike's conjecture about a race condition with the writeback is correct; it can confuse the DMA engine.] [Comment and analysis from the ML post by Michael <mike@rsy.com>] Signed-off-by: Andy Walls <awalls@md.metrocast.net> Cc: stable@kernel.org Signed-off-by: Mauro Carvalho Chehab <mchehab@redhat.com>
2011-02-26 04:56:34 +00:00
status = read_reg(IVTV_REG_DMASTATUS);
IVTV_DEBUG_WARN("DMA ERROR %08x %08x %08x %d\n", data[0], data[1],
[media] ivtv: Fix corrective action taken upon DMA ERR interrupt to avoid hang After upgrading the kernel from stock Ubuntu 7.10 to 10.04, with no hardware changes, I started getting the dreaded DMA TIMEOUT errors, followed by inability to encode until the machine was rebooted. I came across a post from Andy in March (http://www.gossamer-threads.com/lists/ivtv/users/40943#40943) where he speculates that perhaps the corrective actions being taken after a DMA ERROR are not sufficient to recover the situation. After some testing I suspect that this is indeed the case, and that in fact the corrective action may be what hangs the card's DMA engine, rather than the original error. Specifically these DMA ERROR IRQs seem to present with two different values in the IVTV_REG_DMASTATUS register: 0x11 and 0x13. The current corrective action is to clear that status register back to 0x01 or 0x03, and then issue the next DMA request. In the case of a 0x13 this seems to result in a minor glitch in the encoded stream due to the failed transfer that was not retried, but otherwise things continue OK. In the case of a 0x11 the card's DMA write engine is never heard from again, and a DMA TIMEOUT follows shortly after. 0x11 is the killer. I suspect that the two cases need to be handled differently. The difference is in bit 1 (0x02), which is set when the error is about to be successfully recovered, and clear when things are about to go bad. Bit 1 of DMASTATUS is described differently in different places either as a positive "write finished", or an inverted "write busy". If we take the first definition, then when an error arises with state 0x11, it means that the write did not complete. It makes sense to start a new transfer, as in the current code. But if we take the second definition, then 0x11 means "an error but the write engine is still busy". Trying to feed it a new transfer in this situation might not be a good idea. As an experiment, I added code to ignore the DMA ERROR IRQ if DMASTATUS is 0x11. I.e., don't start a new transfer, don't clear our flags, etc. The hope was that the card would complete the transfer and issue a ENC DMA COMPLETE, either successfully or with an error condition there. However the card still hung. The only remaining corrective action being taken with a 0x11 status was then the write back to the status register to clear the error, i.e. DMASTATUS = DMASTATUS & ~3. This would have the effect of clearing the error bit 4, while leaving the lower bits indicating DMA write busy. Strangely enough, removing this write to the status register solved the problem! If the DMA ERROR IRQ with DMASTATUS=0x11 is completely ignored, with no corrective action at all, then the card will complete the transfer and issue a new IRQ. If the status register is written to when it has the value 0x11, then the DMA engine hangs. Perhaps it's illegal to write to DMASTATUS while the read or write busy bit is set? At any rate, it appears that the current corrective action is indeed making things worse rather than better. I put together a patch that modifies ivtv_irq_dma_err to do the following: - Don't write back to IVTV_REG_DMASTATUS. - If write-busy is asserted, leave the card alone. Just extend the timeout slightly. - If write-busy is de-asserted, retry the current transfer. This has completely fixed my DMA TIMEOUT woes. DMA ERR events still occur, but now they seem to be correctly handled. 0x11 events no longer hang the card, and 0x13 events no longer result in a glitch in the stream, as the failed transfer is retried. I'm happy. I've inlined the patch below in case it is of interest. As described above, I have a theory about why it works (based on a different interpretation of bit 1 of DMASTATUS), but I can't guarantee that my theory is correct. There may be another explanation, or it may be a fluke. Maybe ignoring that IRQ entirely would be equally effective? Maybe the status register read/writeback sequence is race condition if the card changes it in the mean time? Also as I am using a PVR-150 only, I have not been able to test it on other cards, which may be especially relevant for 350s that support concurrent decoding. Hopefully the patch does not break the DMA READ path. Mike [awalls@md.metrocast.net: Modified patch to add a verbose comment, make minor brace reformats, and clear the error flags in the IVTV_REG_DMASTATUS iff both read and write DMA were not in progress. Mike's conjecture about a race condition with the writeback is correct; it can confuse the DMA engine.] [Comment and analysis from the ML post by Michael <mike@rsy.com>] Signed-off-by: Andy Walls <awalls@md.metrocast.net> Cc: stable@kernel.org Signed-off-by: Mauro Carvalho Chehab <mchehab@redhat.com>
2011-02-26 04:56:34 +00:00
status, itv->cur_dma_stream);
/*
* We do *not* write back to the IVTV_REG_DMASTATUS register to
* clear the error status, if either the encoder write (0x02) or
* decoder read (0x01) bus master DMA operation do not indicate
* completed. We can race with the DMA engine, which may have
* transitioned to completed status *after* we read the register.
* Setting a IVTV_REG_DMASTATUS flag back to "busy" status, after the
* DMA engine has completed, will cause the DMA engine to stop working.
*/
status &= 0x3;
if (status == 0x3)
write_reg(status, IVTV_REG_DMASTATUS);
if (!test_bit(IVTV_F_I_UDMA, &itv->i_flags) &&
itv->cur_dma_stream >= 0 && itv->cur_dma_stream < IVTV_MAX_STREAMS) {
struct ivtv_stream *s = &itv->streams[itv->cur_dma_stream];
[media] ivtv: Fix corrective action taken upon DMA ERR interrupt to avoid hang After upgrading the kernel from stock Ubuntu 7.10 to 10.04, with no hardware changes, I started getting the dreaded DMA TIMEOUT errors, followed by inability to encode until the machine was rebooted. I came across a post from Andy in March (http://www.gossamer-threads.com/lists/ivtv/users/40943#40943) where he speculates that perhaps the corrective actions being taken after a DMA ERROR are not sufficient to recover the situation. After some testing I suspect that this is indeed the case, and that in fact the corrective action may be what hangs the card's DMA engine, rather than the original error. Specifically these DMA ERROR IRQs seem to present with two different values in the IVTV_REG_DMASTATUS register: 0x11 and 0x13. The current corrective action is to clear that status register back to 0x01 or 0x03, and then issue the next DMA request. In the case of a 0x13 this seems to result in a minor glitch in the encoded stream due to the failed transfer that was not retried, but otherwise things continue OK. In the case of a 0x11 the card's DMA write engine is never heard from again, and a DMA TIMEOUT follows shortly after. 0x11 is the killer. I suspect that the two cases need to be handled differently. The difference is in bit 1 (0x02), which is set when the error is about to be successfully recovered, and clear when things are about to go bad. Bit 1 of DMASTATUS is described differently in different places either as a positive "write finished", or an inverted "write busy". If we take the first definition, then when an error arises with state 0x11, it means that the write did not complete. It makes sense to start a new transfer, as in the current code. But if we take the second definition, then 0x11 means "an error but the write engine is still busy". Trying to feed it a new transfer in this situation might not be a good idea. As an experiment, I added code to ignore the DMA ERROR IRQ if DMASTATUS is 0x11. I.e., don't start a new transfer, don't clear our flags, etc. The hope was that the card would complete the transfer and issue a ENC DMA COMPLETE, either successfully or with an error condition there. However the card still hung. The only remaining corrective action being taken with a 0x11 status was then the write back to the status register to clear the error, i.e. DMASTATUS = DMASTATUS & ~3. This would have the effect of clearing the error bit 4, while leaving the lower bits indicating DMA write busy. Strangely enough, removing this write to the status register solved the problem! If the DMA ERROR IRQ with DMASTATUS=0x11 is completely ignored, with no corrective action at all, then the card will complete the transfer and issue a new IRQ. If the status register is written to when it has the value 0x11, then the DMA engine hangs. Perhaps it's illegal to write to DMASTATUS while the read or write busy bit is set? At any rate, it appears that the current corrective action is indeed making things worse rather than better. I put together a patch that modifies ivtv_irq_dma_err to do the following: - Don't write back to IVTV_REG_DMASTATUS. - If write-busy is asserted, leave the card alone. Just extend the timeout slightly. - If write-busy is de-asserted, retry the current transfer. This has completely fixed my DMA TIMEOUT woes. DMA ERR events still occur, but now they seem to be correctly handled. 0x11 events no longer hang the card, and 0x13 events no longer result in a glitch in the stream, as the failed transfer is retried. I'm happy. I've inlined the patch below in case it is of interest. As described above, I have a theory about why it works (based on a different interpretation of bit 1 of DMASTATUS), but I can't guarantee that my theory is correct. There may be another explanation, or it may be a fluke. Maybe ignoring that IRQ entirely would be equally effective? Maybe the status register read/writeback sequence is race condition if the card changes it in the mean time? Also as I am using a PVR-150 only, I have not been able to test it on other cards, which may be especially relevant for 350s that support concurrent decoding. Hopefully the patch does not break the DMA READ path. Mike [awalls@md.metrocast.net: Modified patch to add a verbose comment, make minor brace reformats, and clear the error flags in the IVTV_REG_DMASTATUS iff both read and write DMA were not in progress. Mike's conjecture about a race condition with the writeback is correct; it can confuse the DMA engine.] [Comment and analysis from the ML post by Michael <mike@rsy.com>] Signed-off-by: Andy Walls <awalls@md.metrocast.net> Cc: stable@kernel.org Signed-off-by: Mauro Carvalho Chehab <mchehab@redhat.com>
2011-02-26 04:56:34 +00:00
if (s->type >= IVTV_DEC_STREAM_TYPE_MPG) {
/* retry */
/*
* FIXME - handle cases of DMA error similar to
* encoder below, except conditioned on status & 0x1
*/
ivtv_dma_dec_start(s);
[media] ivtv: Fix corrective action taken upon DMA ERR interrupt to avoid hang After upgrading the kernel from stock Ubuntu 7.10 to 10.04, with no hardware changes, I started getting the dreaded DMA TIMEOUT errors, followed by inability to encode until the machine was rebooted. I came across a post from Andy in March (http://www.gossamer-threads.com/lists/ivtv/users/40943#40943) where he speculates that perhaps the corrective actions being taken after a DMA ERROR are not sufficient to recover the situation. After some testing I suspect that this is indeed the case, and that in fact the corrective action may be what hangs the card's DMA engine, rather than the original error. Specifically these DMA ERROR IRQs seem to present with two different values in the IVTV_REG_DMASTATUS register: 0x11 and 0x13. The current corrective action is to clear that status register back to 0x01 or 0x03, and then issue the next DMA request. In the case of a 0x13 this seems to result in a minor glitch in the encoded stream due to the failed transfer that was not retried, but otherwise things continue OK. In the case of a 0x11 the card's DMA write engine is never heard from again, and a DMA TIMEOUT follows shortly after. 0x11 is the killer. I suspect that the two cases need to be handled differently. The difference is in bit 1 (0x02), which is set when the error is about to be successfully recovered, and clear when things are about to go bad. Bit 1 of DMASTATUS is described differently in different places either as a positive "write finished", or an inverted "write busy". If we take the first definition, then when an error arises with state 0x11, it means that the write did not complete. It makes sense to start a new transfer, as in the current code. But if we take the second definition, then 0x11 means "an error but the write engine is still busy". Trying to feed it a new transfer in this situation might not be a good idea. As an experiment, I added code to ignore the DMA ERROR IRQ if DMASTATUS is 0x11. I.e., don't start a new transfer, don't clear our flags, etc. The hope was that the card would complete the transfer and issue a ENC DMA COMPLETE, either successfully or with an error condition there. However the card still hung. The only remaining corrective action being taken with a 0x11 status was then the write back to the status register to clear the error, i.e. DMASTATUS = DMASTATUS & ~3. This would have the effect of clearing the error bit 4, while leaving the lower bits indicating DMA write busy. Strangely enough, removing this write to the status register solved the problem! If the DMA ERROR IRQ with DMASTATUS=0x11 is completely ignored, with no corrective action at all, then the card will complete the transfer and issue a new IRQ. If the status register is written to when it has the value 0x11, then the DMA engine hangs. Perhaps it's illegal to write to DMASTATUS while the read or write busy bit is set? At any rate, it appears that the current corrective action is indeed making things worse rather than better. I put together a patch that modifies ivtv_irq_dma_err to do the following: - Don't write back to IVTV_REG_DMASTATUS. - If write-busy is asserted, leave the card alone. Just extend the timeout slightly. - If write-busy is de-asserted, retry the current transfer. This has completely fixed my DMA TIMEOUT woes. DMA ERR events still occur, but now they seem to be correctly handled. 0x11 events no longer hang the card, and 0x13 events no longer result in a glitch in the stream, as the failed transfer is retried. I'm happy. I've inlined the patch below in case it is of interest. As described above, I have a theory about why it works (based on a different interpretation of bit 1 of DMASTATUS), but I can't guarantee that my theory is correct. There may be another explanation, or it may be a fluke. Maybe ignoring that IRQ entirely would be equally effective? Maybe the status register read/writeback sequence is race condition if the card changes it in the mean time? Also as I am using a PVR-150 only, I have not been able to test it on other cards, which may be especially relevant for 350s that support concurrent decoding. Hopefully the patch does not break the DMA READ path. Mike [awalls@md.metrocast.net: Modified patch to add a verbose comment, make minor brace reformats, and clear the error flags in the IVTV_REG_DMASTATUS iff both read and write DMA were not in progress. Mike's conjecture about a race condition with the writeback is correct; it can confuse the DMA engine.] [Comment and analysis from the ML post by Michael <mike@rsy.com>] Signed-off-by: Andy Walls <awalls@md.metrocast.net> Cc: stable@kernel.org Signed-off-by: Mauro Carvalho Chehab <mchehab@redhat.com>
2011-02-26 04:56:34 +00:00
return;
} else {
if ((status & 0x2) == 0) {
/*
* CX2341x Bus Master DMA write is ongoing.
* Reset the timer and let it complete.
*/
itv->dma_timer.expires =
jiffies + msecs_to_jiffies(600);
add_timer(&itv->dma_timer);
return;
}
if (itv->dma_retries < 3) {
/*
* CX2341x Bus Master DMA write has ended.
* Retry the write, starting with the first
* xfer segment. Just retrying the current
* segment is not sufficient.
*/
s->sg_processed = 0;
itv->dma_retries++;
ivtv_dma_enc_start_xfer(s);
return;
}
/* Too many retries, give up on this one */
}
}
if (test_bit(IVTV_F_I_UDMA, &itv->i_flags)) {
ivtv_udma_start(itv);
return;
}
clear_bit(IVTV_F_I_UDMA, &itv->i_flags);
clear_bit(IVTV_F_I_DMA, &itv->i_flags);
itv->cur_dma_stream = -1;
wake_up(&itv->dma_waitq);
}
static void ivtv_irq_enc_start_cap(struct ivtv *itv)
{
u32 data[CX2341X_MBOX_MAX_DATA];
struct ivtv_stream *s;
/* Get DMA destination and size arguments from card */
ivtv_api_get_data(&itv->enc_mbox, IVTV_MBOX_DMA, 7, data);
IVTV_DEBUG_HI_IRQ("ENC START CAP %d: %08x %08x\n", data[0], data[1], data[2]);
if (data[0] > 2 || data[1] == 0 || data[2] == 0) {
IVTV_DEBUG_WARN("Unknown input: %08x %08x %08x\n",
data[0], data[1], data[2]);
return;
}
s = &itv->streams[ivtv_stream_map[data[0]]];
if (!stream_enc_dma_append(s, data)) {
set_bit(ivtv_use_pio(s) ? IVTV_F_S_PIO_PENDING : IVTV_F_S_DMA_PENDING, &s->s_flags);
}
}
static void ivtv_irq_enc_vbi_cap(struct ivtv *itv)
{
u32 data[CX2341X_MBOX_MAX_DATA];
struct ivtv_stream *s;
IVTV_DEBUG_HI_IRQ("ENC START VBI CAP\n");
s = &itv->streams[IVTV_ENC_STREAM_TYPE_VBI];
if (!stream_enc_dma_append(s, data))
set_bit(ivtv_use_pio(s) ? IVTV_F_S_PIO_PENDING : IVTV_F_S_DMA_PENDING, &s->s_flags);
}
static void ivtv_irq_dec_vbi_reinsert(struct ivtv *itv)
{
u32 data[CX2341X_MBOX_MAX_DATA];
struct ivtv_stream *s = &itv->streams[IVTV_DEC_STREAM_TYPE_VBI];
IVTV_DEBUG_HI_IRQ("DEC VBI REINSERT\n");
if (test_bit(IVTV_F_S_CLAIMED, &s->s_flags) &&
!stream_enc_dma_append(s, data)) {
set_bit(IVTV_F_S_PIO_PENDING, &s->s_flags);
}
}
static void ivtv_irq_dec_data_req(struct ivtv *itv)
{
u32 data[CX2341X_MBOX_MAX_DATA];
struct ivtv_stream *s;
/* YUV or MPG */
if (test_bit(IVTV_F_I_DEC_YUV, &itv->i_flags)) {
ivtv_api_get_data(&itv->dec_mbox, IVTV_MBOX_DMA, 2, data);
V4L/DVB (6717): ivtv: Initial merge of video48 yuv handling into the IVTV_IOC_DMA_FRAME framework Previously, all yuv data written to /dev/video48 had only basic support with no double buffering to avoid display tearing. With this patch, yuv frames written to video48 are now handled by the existing IVTV_IOC_DMA_FRAME framework. As such, the frames are hardware buffered to avoid tearing, and honour scaling mode & field order options. Unlike the proprietary IVTV_IOC_DMA_FRAME ioctl, all parameters are controlled by the V4L2 API. Due to mpeg & yuv output restrictions being different, their V4L2 output controls have been separated. To control the yuv output, the V4L2 calls must be done via video48. If the ivtvfb module is loaded, there will be one side effect to this merge. The yuv output window will be constrained to the visible framebuffer area. In the event that a virtual framebuffer size is being used, the limit to the output size will be the virtual dimensions, but only the portion that falls within the currently visible area of the framebuffer will be shown. Like the IVTV_IOC_DMA_FRAME ioctl, the supplied frames must be padded to 720 pixels wide. However the height must only be padded up the nearest multiple of 32. This would mean an image of 102 lines must be padded to 128. As long as the true source image size is given, the padding will not be visible in the final output. Signed-off-by: Ian Armstrong <ian@iarmst.demon.co.uk> Signed-off-by: Hans Verkuil <hverkuil@xs4all.nl> Signed-off-by: Mauro Carvalho Chehab <mchehab@infradead.org>
2007-11-05 17:27:09 +00:00
itv->dma_data_req_size =
1080 * ((itv->yuv_info.v4l2_src_h + 31) & ~31);
itv->dma_data_req_offset = data[1];
if (atomic_read(&itv->yuv_info.next_dma_frame) >= 0)
ivtv_yuv_frame_complete(itv);
s = &itv->streams[IVTV_DEC_STREAM_TYPE_YUV];
}
else {
ivtv_api_get_data(&itv->dec_mbox, IVTV_MBOX_DMA, 3, data);
itv->dma_data_req_size = min_t(u32, data[2], 0x10000);
itv->dma_data_req_offset = data[1];
s = &itv->streams[IVTV_DEC_STREAM_TYPE_MPG];
}
IVTV_DEBUG_HI_IRQ("DEC DATA REQ %s: %d %08x %u\n", s->name, s->q_full.bytesused,
itv->dma_data_req_offset, itv->dma_data_req_size);
if (itv->dma_data_req_size == 0 || s->q_full.bytesused < itv->dma_data_req_size) {
set_bit(IVTV_F_S_NEEDS_DATA, &s->s_flags);
}
else {
V4L/DVB (6717): ivtv: Initial merge of video48 yuv handling into the IVTV_IOC_DMA_FRAME framework Previously, all yuv data written to /dev/video48 had only basic support with no double buffering to avoid display tearing. With this patch, yuv frames written to video48 are now handled by the existing IVTV_IOC_DMA_FRAME framework. As such, the frames are hardware buffered to avoid tearing, and honour scaling mode & field order options. Unlike the proprietary IVTV_IOC_DMA_FRAME ioctl, all parameters are controlled by the V4L2 API. Due to mpeg & yuv output restrictions being different, their V4L2 output controls have been separated. To control the yuv output, the V4L2 calls must be done via video48. If the ivtvfb module is loaded, there will be one side effect to this merge. The yuv output window will be constrained to the visible framebuffer area. In the event that a virtual framebuffer size is being used, the limit to the output size will be the virtual dimensions, but only the portion that falls within the currently visible area of the framebuffer will be shown. Like the IVTV_IOC_DMA_FRAME ioctl, the supplied frames must be padded to 720 pixels wide. However the height must only be padded up the nearest multiple of 32. This would mean an image of 102 lines must be padded to 128. As long as the true source image size is given, the padding will not be visible in the final output. Signed-off-by: Ian Armstrong <ian@iarmst.demon.co.uk> Signed-off-by: Hans Verkuil <hverkuil@xs4all.nl> Signed-off-by: Mauro Carvalho Chehab <mchehab@infradead.org>
2007-11-05 17:27:09 +00:00
if (test_bit(IVTV_F_I_DEC_YUV, &itv->i_flags))
ivtv_yuv_setup_stream_frame(itv);
clear_bit(IVTV_F_S_NEEDS_DATA, &s->s_flags);
ivtv_queue_move(s, &s->q_full, NULL, &s->q_predma, itv->dma_data_req_size);
ivtv_dma_stream_dec_prepare(s, itv->dma_data_req_offset + IVTV_DECODER_OFFSET, 0);
}
}
static void ivtv_irq_vsync(struct ivtv *itv)
{
/* The vsync interrupt is unusual in that it won't clear until
* the end of the first line for the current field, at which
* point it clears itself. This can result in repeated vsync
* interrupts, or a missed vsync. Read some of the registers
* to determine the line being displayed and ensure we handle
* one vsync per frame.
*/
unsigned int frame = read_reg(IVTV_REG_DEC_LINE_FIELD) & 1;
struct yuv_playback_info *yi = &itv->yuv_info;
int last_dma_frame = atomic_read(&yi->next_dma_frame);
struct yuv_frame_info *f = &yi->new_frame_info[last_dma_frame];
if (0) IVTV_DEBUG_IRQ("DEC VSYNC\n");
if (((frame ^ f->sync_field) == 0 &&
((itv->last_vsync_field & 1) ^ f->sync_field)) ||
(frame != (itv->last_vsync_field & 1) && !f->interlaced)) {
int next_dma_frame = last_dma_frame;
if (!(f->interlaced && f->delay && yi->fields_lapsed < 1)) {
if (next_dma_frame >= 0 && next_dma_frame != atomic_read(&yi->next_fill_frame)) {
write_reg(yuv_offset[next_dma_frame] >> 4, 0x82c);
write_reg((yuv_offset[next_dma_frame] + IVTV_YUV_BUFFER_UV_OFFSET) >> 4, 0x830);
write_reg(yuv_offset[next_dma_frame] >> 4, 0x834);
write_reg((yuv_offset[next_dma_frame] + IVTV_YUV_BUFFER_UV_OFFSET) >> 4, 0x838);
next_dma_frame = (next_dma_frame + 1) % IVTV_YUV_BUFFERS;
atomic_set(&yi->next_dma_frame, next_dma_frame);
yi->fields_lapsed = -1;
yi->running = 1;
}
}
}
if (frame != (itv->last_vsync_field & 1)) {
static const struct v4l2_event evtop = {
.type = V4L2_EVENT_VSYNC,
.u.vsync.field = V4L2_FIELD_TOP,
};
static const struct v4l2_event evbottom = {
.type = V4L2_EVENT_VSYNC,
.u.vsync.field = V4L2_FIELD_BOTTOM,
};
struct ivtv_stream *s = ivtv_get_output_stream(itv);
itv->last_vsync_field += 1;
if (frame == 0) {
clear_bit(IVTV_F_I_VALID_DEC_TIMINGS, &itv->i_flags);
clear_bit(IVTV_F_I_EV_VSYNC_FIELD, &itv->i_flags);
}
else {
set_bit(IVTV_F_I_EV_VSYNC_FIELD, &itv->i_flags);
}
if (test_bit(IVTV_F_I_EV_VSYNC_ENABLED, &itv->i_flags)) {
set_bit(IVTV_F_I_EV_VSYNC, &itv->i_flags);
wake_up(&itv->event_waitq);
if (s)
wake_up(&s->waitq);
}
if (s && s->vdev)
v4l2_event_queue(s->vdev, frame ? &evtop : &evbottom);
wake_up(&itv->vsync_waitq);
/* Send VBI to saa7127 */
if (frame && (itv->output_mode == OUT_PASSTHROUGH ||
test_bit(IVTV_F_I_UPDATE_WSS, &itv->i_flags) ||
test_bit(IVTV_F_I_UPDATE_VPS, &itv->i_flags) ||
test_bit(IVTV_F_I_UPDATE_CC, &itv->i_flags))) {
set_bit(IVTV_F_I_WORK_HANDLER_VBI, &itv->i_flags);
set_bit(IVTV_F_I_HAVE_WORK, &itv->i_flags);
}
/* Check if we need to update the yuv registers */
if (yi->running && (yi->yuv_forced_update || f->update)) {
if (!f->update) {
last_dma_frame =
(u8)(atomic_read(&yi->next_dma_frame) -
1) % IVTV_YUV_BUFFERS;
f = &yi->new_frame_info[last_dma_frame];
}
if (f->src_w) {
yi->update_frame = last_dma_frame;
f->update = 0;
yi->yuv_forced_update = 0;
set_bit(IVTV_F_I_WORK_HANDLER_YUV, &itv->i_flags);
set_bit(IVTV_F_I_HAVE_WORK, &itv->i_flags);
}
}
yi->fields_lapsed++;
}
}
#define IVTV_IRQ_DMA (IVTV_IRQ_DMA_READ | IVTV_IRQ_ENC_DMA_COMPLETE | IVTV_IRQ_DMA_ERR | IVTV_IRQ_ENC_START_CAP | IVTV_IRQ_ENC_VBI_CAP | IVTV_IRQ_DEC_DATA_REQ | IVTV_IRQ_DEC_VBI_RE_INSERT)
irqreturn_t ivtv_irq_handler(int irq, void *dev_id)
{
struct ivtv *itv = (struct ivtv *)dev_id;
u32 combo;
u32 stat;
int i;
u8 vsync_force = 0;
spin_lock(&itv->dma_reg_lock);
/* get contents of irq status register */
stat = read_reg(IVTV_REG_IRQSTATUS);
combo = ~itv->irqmask & stat;
/* Clear out IRQ */
if (combo) write_reg(combo, IVTV_REG_IRQSTATUS);
if (0 == combo) {
/* The vsync interrupt is unusual and clears itself. If we
* took too long, we may have missed it. Do some checks
*/
if (~itv->irqmask & IVTV_IRQ_DEC_VSYNC) {
/* vsync is enabled, see if we're in a new field */
if ((itv->last_vsync_field & 1) !=
(read_reg(IVTV_REG_DEC_LINE_FIELD) & 1)) {
/* New field, looks like we missed it */
IVTV_DEBUG_YUV("VSync interrupt missed %d\n",
read_reg(IVTV_REG_DEC_LINE_FIELD) >> 16);
vsync_force = 1;
}
}
if (!vsync_force) {
/* No Vsync expected, wasn't for us */
spin_unlock(&itv->dma_reg_lock);
return IRQ_NONE;
}
}
/* Exclude interrupts noted below from the output, otherwise the log is flooded with
these messages */
if (combo & ~0xff6d0400)
IVTV_DEBUG_HI_IRQ("======= valid IRQ bits: 0x%08x ======\n", combo);
if (combo & IVTV_IRQ_DEC_DMA_COMPLETE) {
IVTV_DEBUG_HI_IRQ("DEC DMA COMPLETE\n");
}
if (combo & IVTV_IRQ_DMA_READ) {
ivtv_irq_dma_read(itv);
}
if (combo & IVTV_IRQ_ENC_DMA_COMPLETE) {
ivtv_irq_enc_dma_complete(itv);
}
if (combo & IVTV_IRQ_ENC_PIO_COMPLETE) {
ivtv_irq_enc_pio_complete(itv);
}
if (combo & IVTV_IRQ_DMA_ERR) {
ivtv_irq_dma_err(itv);
}
if (combo & IVTV_IRQ_ENC_START_CAP) {
ivtv_irq_enc_start_cap(itv);
}
if (combo & IVTV_IRQ_ENC_VBI_CAP) {
ivtv_irq_enc_vbi_cap(itv);
}
if (combo & IVTV_IRQ_DEC_VBI_RE_INSERT) {
ivtv_irq_dec_vbi_reinsert(itv);
}
if (combo & IVTV_IRQ_ENC_EOS) {
IVTV_DEBUG_IRQ("ENC EOS\n");
set_bit(IVTV_F_I_EOS, &itv->i_flags);
wake_up(&itv->eos_waitq);
}
if (combo & IVTV_IRQ_DEC_DATA_REQ) {
ivtv_irq_dec_data_req(itv);
}
/* Decoder Vertical Sync - We can't rely on 'combo', so check if vsync enabled */
if (~itv->irqmask & IVTV_IRQ_DEC_VSYNC) {
ivtv_irq_vsync(itv);
}
if (combo & IVTV_IRQ_ENC_VIM_RST) {
IVTV_DEBUG_IRQ("VIM RST\n");
/*ivtv_vapi(itv, CX2341X_ENC_REFRESH_INPUT, 0); */
}
if (combo & IVTV_IRQ_DEC_AUD_MODE_CHG) {
IVTV_DEBUG_INFO("Stereo mode changed\n");
}
if ((combo & IVTV_IRQ_DMA) && !test_bit(IVTV_F_I_DMA, &itv->i_flags)) {
itv->irq_rr_idx++;
for (i = 0; i < IVTV_MAX_STREAMS; i++) {
int idx = (i + itv->irq_rr_idx) % IVTV_MAX_STREAMS;
struct ivtv_stream *s = &itv->streams[idx];
if (!test_and_clear_bit(IVTV_F_S_DMA_PENDING, &s->s_flags))
continue;
if (s->type >= IVTV_DEC_STREAM_TYPE_MPG)
ivtv_dma_dec_start(s);
else
ivtv_dma_enc_start(s);
break;
}
if (i == IVTV_MAX_STREAMS &&
test_bit(IVTV_F_I_UDMA_PENDING, &itv->i_flags))
ivtv_udma_start(itv);
}
if ((combo & IVTV_IRQ_DMA) && !test_bit(IVTV_F_I_PIO, &itv->i_flags)) {
itv->irq_rr_idx++;
for (i = 0; i < IVTV_MAX_STREAMS; i++) {
int idx = (i + itv->irq_rr_idx) % IVTV_MAX_STREAMS;
struct ivtv_stream *s = &itv->streams[idx];
if (!test_and_clear_bit(IVTV_F_S_PIO_PENDING, &s->s_flags))
continue;
if (s->type == IVTV_DEC_STREAM_TYPE_VBI || s->type < IVTV_DEC_STREAM_TYPE_MPG)
ivtv_dma_enc_start(s);
break;
}
}
if (test_and_clear_bit(IVTV_F_I_HAVE_WORK, &itv->i_flags)) {
queue_kthread_work(&itv->irq_worker, &itv->irq_work);
}
spin_unlock(&itv->dma_reg_lock);
/* If we've just handled a 'forced' vsync, it's safest to say it
* wasn't ours. Another device may have triggered it at just
* the right time.
*/
return vsync_force ? IRQ_NONE : IRQ_HANDLED;
}
void ivtv_unfinished_dma(unsigned long arg)
{
struct ivtv *itv = (struct ivtv *)arg;
if (!test_bit(IVTV_F_I_DMA, &itv->i_flags))
return;
IVTV_ERR("DMA TIMEOUT %08x %d\n", read_reg(IVTV_REG_DMASTATUS), itv->cur_dma_stream);
write_reg(read_reg(IVTV_REG_DMASTATUS) & 3, IVTV_REG_DMASTATUS);
clear_bit(IVTV_F_I_UDMA, &itv->i_flags);
clear_bit(IVTV_F_I_DMA, &itv->i_flags);
itv->cur_dma_stream = -1;
wake_up(&itv->dma_waitq);
}