| Message ID | 1411360048-3388-1-git-send-email-agross@codeaurora.org (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
Hi Andy, On Sun, 2014-09-21 at 23:27 -0500, Andy Gross wrote: > This patch fixes a number of errors with the QUP block transfer mode. Errors > manifested themselves as input underruns, output overruns, and timed out > transactions. At what speeds are you seeing those errors? > > The block mode does not require the priming that occurs in FIFO mode. At the > moment that the QUP is placed into the RUN state, the QUP may immediately raise > an interrupt if the request is a write. Therefore, there is no need to prime > the pump. > > In addition, the block transfers require that whole blocks of data are > read/written at a time. The last block of data that completes a transaction may > contain less than a full blocks worth of data. Does this mean that block transfer will start only if the required bytes from block is written into buffer? > > Each block of data results in an input/output service interrupt accompanied with > a input/output block flag set. Additional block reads/writes require clearing > of the service flag. It is ok to check for additional blocks of data in the > ISR, but you have to ack every block you transfer. Imbalanced acks result in > early return from complete transactions with pending interrupts that still have > to be ack'd. The next transaction can be affected by these interrupts. > > Signed-off-by: Andy Gross <agross@codeaurora.org> > --- > drivers/spi/spi-qup.c | 194 +++++++++++++++++++++++++++++++++++-------------- > 1 file changed, 141 insertions(+), 53 deletions(-) > <snip> > +static void qup_fill_read_buffer(struct spi_qup *controller, > + struct spi_transfer *xfer, u32 data) Please, could prefix this whit spi_ to be consistent with the rest of the code. > { > u8 *rx_buf = xfer->rx_buf; > - u32 word, state; > - int idx, shift, w_size; > - > - w_size = controller->w_size; > - > - while (controller->rx_bytes < xfer->len) { > - > - state = readl_relaxed(controller->base + QUP_OPERATIONAL); > - if (0 == (state & QUP_OP_IN_FIFO_NOT_EMPTY)) > - break; > + int idx, shift; > + int read_len = min_t(int, xfer->len - controller->rx_bytes, > + controller->w_size); You should not need this check here. xfer->len is multiple of controller->w_size and you always read one word at time. > > - word = readl_relaxed(controller->base + QUP_INPUT_FIFO); > - > - if (!rx_buf) { > - controller->rx_bytes += w_size; > - continue; > - } > - > - for (idx = 0; idx < w_size; idx++, controller->rx_bytes++) { > + if (rx_buf) > + for (idx = 0; idx < read_len; idx++) { > /* > * The data format depends on bytes per SPI word: > * 4 bytes: 0x12345678 > @@ -229,40 +218,129 @@ static void spi_qup_fifo_read(struct spi_qup *controller, > * 1 byte : 0x00000012 > */ > shift = BITS_PER_BYTE; > - shift *= (w_size - idx - 1); > - rx_buf[controller->rx_bytes] = word >> shift; > + shift *= (controller->w_size - idx - 1); > + rx_buf[controller->rx_bytes + idx] = data >> shift; > } > - } > + > + controller->rx_bytes += read_len; > } > > -static void spi_qup_fifo_write(struct spi_qup *controller, > - struct spi_transfer *xfer) > +static void qup_prepare_write_data(struct spi_qup *controller, > + struct spi_transfer *xfer, u32 *data) > { > const u8 *tx_buf = xfer->tx_buf; > - u32 word, state, data; > - int idx, w_size; > + u32 val; > + int idx; > + int write_len = min_t(int, xfer->len - controller->tx_bytes, > + controller->w_size); > Same here. > - w_size = controller->w_size; > + *data = 0; > > - while (controller->tx_bytes < xfer->len) { > + if (tx_buf) > + for (idx = 0; idx < write_len; idx++) { > + val = tx_buf[controller->tx_bytes + idx]; > + *data |= val << (BITS_PER_BYTE * (3 - idx)); > + } > > - state = readl_relaxed(controller->base + QUP_OPERATIONAL); > - if (state & QUP_OP_OUT_FIFO_FULL) > - break; > + controller->tx_bytes += write_len; > +} > > - word = 0; > - for (idx = 0; idx < w_size; idx++, controller->tx_bytes++) { > +static void spi_qup_service_block(struct spi_qup *controller, > + struct spi_transfer *xfer, bool is_read) > +{ Please, could you split this function to read and write, so we can use: spi_qup_fifo_read() and spi_qup_fifo_write() in FIFO modes and spi_qup_block_read() and spi_qup_block_write() for BLOCK mode. > + u32 data, words_per_blk, num_words, ack_flag, op_flag; > + int i; > + > + if (is_read) { > + op_flag = QUP_OP_IN_BLOCK_READ_REQ; > + ack_flag = QUP_OP_IN_SERVICE_FLAG; > + num_words = DIV_ROUND_UP(xfer->len - controller->rx_bytes, > + controller->w_size); Same here and below. > + words_per_blk = controller->in_blk_sz >> 2; > + } else { > + op_flag = QUP_OP_OUT_BLOCK_WRITE_REQ; > + ack_flag = QUP_OP_OUT_SERVICE_FLAG; > + num_words = DIV_ROUND_UP(xfer->len - controller->tx_bytes, > + controller->w_size); > + words_per_blk = controller->out_blk_sz >> 2; > + } Thank you. Ivan
On Tue, Sep 23, 2014 at 12:24:27PM +0300, Ivan T. Ivanov wrote: > > Hi Andy, > > On Sun, 2014-09-21 at 23:27 -0500, Andy Gross wrote: > > This patch fixes a number of errors with the QUP block transfer mode. Errors > > manifested themselves as input underruns, output overruns, and timed out > > transactions. > > At what speeds are you seeing those errors? We've tried 25MHz and 50MHz. Both fail in the same way. Keep in mind this is definitely a timing / race issue and it probably also dependent on the latency of the attached device. I cannot reproduce this at all on my IPQ8064 based board, but others can. This problem manifested itself while using spidev and a usermode flash programming application (flashrom). > > > > > The block mode does not require the priming that occurs in FIFO mode. At the > > moment that the QUP is placed into the RUN state, the QUP may immediately raise > > an interrupt if the request is a write. Therefore, there is no need to prime > > the pump. > > > > In addition, the block transfers require that whole blocks of data are > > read/written at a time. The last block of data that completes a transaction may > > contain less than a full blocks worth of data. > > Does this mean that block transfer will start only if the required > bytes from block is written into buffer? No, a better way of putting this is that immediately on setting RUN state, you'll get a service interrupt to fill the FIFO in block mode. So there is no need to prime the FIFO from the non-isr context due to this behavior. <snip> > > > +static void qup_fill_read_buffer(struct spi_qup *controller, > > + struct spi_transfer *xfer, u32 data) > > Please, could prefix this whit spi_ to be consistent with the > rest of the code. Good point. I need to be consistent. > > { > > u8 *rx_buf = xfer->rx_buf; > > - u32 word, state; > > - int idx, shift, w_size; > > - > > - w_size = controller->w_size; > > - > > - while (controller->rx_bytes < xfer->len) { > > - > > - state = readl_relaxed(controller->base + QUP_OPERATIONAL); > > - if (0 == (state & QUP_OP_IN_FIFO_NOT_EMPTY)) > > - break; > > + int idx, shift; > > + int read_len = min_t(int, xfer->len - controller->rx_bytes, > > + controller->w_size); > > You should not need this check here. xfer->len is multiple of controller->w_size > and you always read one word at time. Ah I missed the __spi_validate where this is done. I'll remove this. Good catch. <snip> > > const u8 *tx_buf = xfer->tx_buf; > > - u32 word, state, data; > > - int idx, w_size; > > + u32 val; > > + int idx; > > + int write_len = min_t(int, xfer->len - controller->tx_bytes, > > + controller->w_size); > > > > Same here. Agreed. <snip> > > - word = 0; > > - for (idx = 0; idx < w_size; idx++, controller->tx_bytes++) { > > +static void spi_qup_service_block(struct spi_qup *controller, > > + struct spi_transfer *xfer, bool is_read) > > +{ > > Please, could you split this function to read and write, so we can use: > > spi_qup_fifo_read() and spi_qup_fifo_write() in FIFO modes and > spi_qup_block_read() and spi_qup_block_write() for BLOCK mode. Well I had it collapsed and the functions are identical except for the read/write specific pieces, which amount to 2 lines. I can resplit it out. It makes it symmetric. > > > + u32 data, words_per_blk, num_words, ack_flag, op_flag; > > + int i; > > + > > + if (is_read) { > > + op_flag = QUP_OP_IN_BLOCK_READ_REQ; > > + ack_flag = QUP_OP_IN_SERVICE_FLAG; > > + num_words = DIV_ROUND_UP(xfer->len - controller->rx_bytes, > > + controller->w_size); > > Same here and below. Agreed. > > + words_per_blk = controller->in_blk_sz >> 2; > > + } else { > > + op_flag = QUP_OP_OUT_BLOCK_WRITE_REQ; > > + ack_flag = QUP_OP_OUT_SERVICE_FLAG; > > + num_words = DIV_ROUND_UP(xfer->len - controller->tx_bytes, > > + controller->w_size); > > + words_per_blk = controller->out_blk_sz >> 2; > > + }
On Tue, Sep 23, 2014 at 02:26:13PM -0500, Andy Gross wrote: > On Tue, Sep 23, 2014 at 12:24:27PM +0300, Ivan T. Ivanov wrote: > > On Sun, 2014-09-21 at 23:27 -0500, Andy Gross wrote: > > > This patch fixes a number of errors with the QUP block transfer mode. Errors > > > manifested themselves as input underruns, output overruns, and timed out > > > transactions. > > At what speeds are you seeing those errors? > We've tried 25MHz and 50MHz. Both fail in the same way. Keep in mind this is > definitely a timing / race issue and it probably also dependent on the latency > of the attached device. I cannot reproduce this at all on my IPQ8064 based > board, but others can. With SPI everything is entirely clocked from the master - the attached device can't cause any of those issues unless there's an electrical problem that *really* upsets the master.
diff --git a/drivers/spi/spi-qup.c b/drivers/spi/spi-qup.c index 9f83d29..9c4c745 100644 --- a/drivers/spi/spi-qup.c +++ b/drivers/spi/spi-qup.c @@ -80,6 +80,8 @@ #define QUP_IO_M_MODE_BAM 3 /* QUP_OPERATIONAL fields */ +#define QUP_OP_IN_BLOCK_READ_REQ BIT(13) +#define QUP_OP_OUT_BLOCK_WRITE_REQ BIT(12) #define QUP_OP_MAX_INPUT_DONE_FLAG BIT(11) #define QUP_OP_MAX_OUTPUT_DONE_FLAG BIT(10) #define QUP_OP_IN_SERVICE_FLAG BIT(9) @@ -143,6 +145,7 @@ struct spi_qup { int tx_bytes; int rx_bytes; int qup_v1; + int mode; }; @@ -198,30 +201,16 @@ static int spi_qup_set_state(struct spi_qup *controller, u32 state) return 0; } - -static void spi_qup_fifo_read(struct spi_qup *controller, - struct spi_transfer *xfer) +static void qup_fill_read_buffer(struct spi_qup *controller, + struct spi_transfer *xfer, u32 data) { u8 *rx_buf = xfer->rx_buf; - u32 word, state; - int idx, shift, w_size; - - w_size = controller->w_size; - - while (controller->rx_bytes < xfer->len) { - - state = readl_relaxed(controller->base + QUP_OPERATIONAL); - if (0 == (state & QUP_OP_IN_FIFO_NOT_EMPTY)) - break; + int idx, shift; + int read_len = min_t(int, xfer->len - controller->rx_bytes, + controller->w_size); - word = readl_relaxed(controller->base + QUP_INPUT_FIFO); - - if (!rx_buf) { - controller->rx_bytes += w_size; - continue; - } - - for (idx = 0; idx < w_size; idx++, controller->rx_bytes++) { + if (rx_buf) + for (idx = 0; idx < read_len; idx++) { /* * The data format depends on bytes per SPI word: * 4 bytes: 0x12345678 @@ -229,40 +218,129 @@ static void spi_qup_fifo_read(struct spi_qup *controller, * 1 byte : 0x00000012 */ shift = BITS_PER_BYTE; - shift *= (w_size - idx - 1); - rx_buf[controller->rx_bytes] = word >> shift; + shift *= (controller->w_size - idx - 1); + rx_buf[controller->rx_bytes + idx] = data >> shift; } - } + + controller->rx_bytes += read_len; } -static void spi_qup_fifo_write(struct spi_qup *controller, - struct spi_transfer *xfer) +static void qup_prepare_write_data(struct spi_qup *controller, + struct spi_transfer *xfer, u32 *data) { const u8 *tx_buf = xfer->tx_buf; - u32 word, state, data; - int idx, w_size; + u32 val; + int idx; + int write_len = min_t(int, xfer->len - controller->tx_bytes, + controller->w_size); - w_size = controller->w_size; + *data = 0; - while (controller->tx_bytes < xfer->len) { + if (tx_buf) + for (idx = 0; idx < write_len; idx++) { + val = tx_buf[controller->tx_bytes + idx]; + *data |= val << (BITS_PER_BYTE * (3 - idx)); + } - state = readl_relaxed(controller->base + QUP_OPERATIONAL); - if (state & QUP_OP_OUT_FIFO_FULL) - break; + controller->tx_bytes += write_len; +} - word = 0; - for (idx = 0; idx < w_size; idx++, controller->tx_bytes++) { +static void spi_qup_service_block(struct spi_qup *controller, + struct spi_transfer *xfer, bool is_read) +{ + u32 data, words_per_blk, num_words, ack_flag, op_flag; + int i; + + if (is_read) { + op_flag = QUP_OP_IN_BLOCK_READ_REQ; + ack_flag = QUP_OP_IN_SERVICE_FLAG; + num_words = DIV_ROUND_UP(xfer->len - controller->rx_bytes, + controller->w_size); + words_per_blk = controller->in_blk_sz >> 2; + } else { + op_flag = QUP_OP_OUT_BLOCK_WRITE_REQ; + ack_flag = QUP_OP_OUT_SERVICE_FLAG; + num_words = DIV_ROUND_UP(xfer->len - controller->tx_bytes, + controller->w_size); + words_per_blk = controller->out_blk_sz >> 2; + } - if (!tx_buf) { - controller->tx_bytes += w_size; - break; + do { + /* ACK by clearing service flag */ + writel_relaxed(ack_flag, controller->base + QUP_OPERATIONAL); + + /* transfer up to a block size of data in a single pass */ + for (i = 0; num_words && i < words_per_blk; i++, num_words--) { + + if (is_read) { + /* read data and fill up rx buffer */ + data = readl_relaxed(controller->base + + QUP_INPUT_FIFO); + qup_fill_read_buffer(controller, xfer, data); + } else { + /* swizzle the bytes for output and write out */ + qup_prepare_write_data(controller, xfer, &data); + writel_relaxed(data, + controller->base + QUP_OUTPUT_FIFO); } - - data = tx_buf[controller->tx_bytes]; - word |= data << (BITS_PER_BYTE * (3 - idx)); } - writel_relaxed(word, controller->base + QUP_OUTPUT_FIFO); + /* check to see if next block is ready */ + if (!(readl_relaxed(controller->base + QUP_OPERATIONAL) & + op_flag)) + break; + + } while (num_words); + + /* + * Due to extra stickiness of the QUP_OP_IN_SERVICE_FLAG during block + * reads, it has to be cleared again at the very end + */ + if (is_read && (readl_relaxed(controller->base + QUP_OPERATIONAL) & + QUP_OP_MAX_INPUT_DONE_FLAG)) + writel_relaxed(ack_flag, controller->base + QUP_OPERATIONAL); + +} + + +static void spi_qup_fifo_read(struct spi_qup *controller, + struct spi_transfer *xfer) +{ + u32 data; + + /* clear service request */ + writel_relaxed(QUP_OP_IN_SERVICE_FLAG, + controller->base + QUP_OPERATIONAL); + + while (controller->rx_bytes < xfer->len) { + if (!(readl_relaxed(controller->base + QUP_OPERATIONAL) & + QUP_OP_IN_FIFO_NOT_EMPTY)) + break; + + data = readl_relaxed(controller->base + QUP_INPUT_FIFO); + + qup_fill_read_buffer(controller, xfer, data); + } + +} + +static void spi_qup_fifo_write(struct spi_qup *controller, + struct spi_transfer *xfer) +{ + u32 data; + + /* clear service request */ + writel_relaxed(QUP_OP_OUT_SERVICE_FLAG, + controller->base + QUP_OPERATIONAL); + + while (controller->tx_bytes < xfer->len) { + + if (readl_relaxed(controller->base + QUP_OPERATIONAL) & + QUP_OP_OUT_FIFO_FULL) + break; + + qup_prepare_write_data(controller, xfer, &data); + writel_relaxed(data, controller->base + QUP_OUTPUT_FIFO); } } @@ -285,9 +363,9 @@ static irqreturn_t spi_qup_qup_irq(int irq, void *dev_id) writel_relaxed(qup_err, controller->base + QUP_ERROR_FLAGS); writel_relaxed(spi_err, controller->base + SPI_ERROR_FLAGS); - writel_relaxed(opflags, controller->base + QUP_OPERATIONAL); if (!xfer) { + writel_relaxed(opflags, controller->base + QUP_OPERATIONAL); dev_err_ratelimited(controller->dev, "unexpected irq %08x %08x %08x\n", qup_err, spi_err, opflags); return IRQ_HANDLED; @@ -315,11 +393,19 @@ static irqreturn_t spi_qup_qup_irq(int irq, void *dev_id) error = -EIO; } - if (opflags & QUP_OP_IN_SERVICE_FLAG) - spi_qup_fifo_read(controller, xfer); + if (opflags & QUP_OP_IN_SERVICE_FLAG) { + if (opflags & QUP_OP_IN_BLOCK_READ_REQ) + spi_qup_service_block(controller, xfer, 1); + else + spi_qup_fifo_read(controller, xfer); + } - if (opflags & QUP_OP_OUT_SERVICE_FLAG) - spi_qup_fifo_write(controller, xfer); + if (opflags & QUP_OP_OUT_SERVICE_FLAG) { + if (opflags & QUP_OP_OUT_BLOCK_WRITE_REQ) + spi_qup_service_block(controller, xfer, 0); + else + spi_qup_fifo_write(controller, xfer); + } spin_lock_irqsave(&controller->lock, flags); controller->error = error; @@ -337,7 +423,7 @@ static irqreturn_t spi_qup_qup_irq(int irq, void *dev_id) static int spi_qup_io_config(struct spi_device *spi, struct spi_transfer *xfer) { struct spi_qup *controller = spi_master_get_devdata(spi->master); - u32 config, iomode, mode; + u32 config, iomode; int ret, n_words, w_size; if (spi->mode & SPI_LOOP && xfer->len > controller->in_fifo_sz) { @@ -368,14 +454,14 @@ static int spi_qup_io_config(struct spi_device *spi, struct spi_transfer *xfer) controller->w_size = w_size; if (n_words <= (controller->in_fifo_sz / sizeof(u32))) { - mode = QUP_IO_M_MODE_FIFO; + controller->mode = QUP_IO_M_MODE_FIFO; writel_relaxed(n_words, controller->base + QUP_MX_READ_CNT); writel_relaxed(n_words, controller->base + QUP_MX_WRITE_CNT); /* must be zero for FIFO */ writel_relaxed(0, controller->base + QUP_MX_INPUT_CNT); writel_relaxed(0, controller->base + QUP_MX_OUTPUT_CNT); } else { - mode = QUP_IO_M_MODE_BLOCK; + controller->mode = QUP_IO_M_MODE_BLOCK; writel_relaxed(n_words, controller->base + QUP_MX_INPUT_CNT); writel_relaxed(n_words, controller->base + QUP_MX_OUTPUT_CNT); /* must be zero for BLOCK and BAM */ @@ -387,8 +473,8 @@ static int spi_qup_io_config(struct spi_device *spi, struct spi_transfer *xfer) /* Set input and output transfer mode */ iomode &= ~(QUP_IO_M_INPUT_MODE_MASK | QUP_IO_M_OUTPUT_MODE_MASK); iomode &= ~(QUP_IO_M_PACK_EN | QUP_IO_M_UNPACK_EN); - iomode |= (mode << QUP_IO_M_OUTPUT_MODE_MASK_SHIFT); - iomode |= (mode << QUP_IO_M_INPUT_MODE_MASK_SHIFT); + iomode |= (controller->mode << QUP_IO_M_OUTPUT_MODE_MASK_SHIFT); + iomode |= (controller->mode << QUP_IO_M_INPUT_MODE_MASK_SHIFT); writel_relaxed(iomode, controller->base + QUP_IO_M_MODES); @@ -462,7 +548,8 @@ static int spi_qup_transfer_one(struct spi_master *master, goto exit; } - spi_qup_fifo_write(controller, xfer); + if (controller->mode == QUP_IO_M_MODE_FIFO) + spi_qup_fifo_write(controller, xfer); if (spi_qup_set_state(controller, QUP_STATE_RUN)) { dev_warn(controller->dev, "cannot set EXECUTE state\n"); @@ -478,6 +565,7 @@ exit: if (!ret) ret = controller->error; spin_unlock_irqrestore(&controller->lock, flags); + return ret; }
This patch fixes a number of errors with the QUP block transfer mode. Errors manifested themselves as input underruns, output overruns, and timed out transactions. The block mode does not require the priming that occurs in FIFO mode. At the moment that the QUP is placed into the RUN state, the QUP may immediately raise an interrupt if the request is a write. Therefore, there is no need to prime the pump. In addition, the block transfers require that whole blocks of data are read/written at a time. The last block of data that completes a transaction may contain less than a full blocks worth of data. Each block of data results in an input/output service interrupt accompanied with a input/output block flag set. Additional block reads/writes require clearing of the service flag. It is ok to check for additional blocks of data in the ISR, but you have to ack every block you transfer. Imbalanced acks result in early return from complete transactions with pending interrupts that still have to be ack'd. The next transaction can be affected by these interrupts. Signed-off-by: Andy Gross <agross@codeaurora.org> --- drivers/spi/spi-qup.c | 194 +++++++++++++++++++++++++++++++++++-------------- 1 file changed, 141 insertions(+), 53 deletions(-)