| Message ID | 20190624123540.20629-2-geert+renesas@glider.be (mailing list archive) |
|---|---|
| State | Not Applicable |
| Headers | show |
| Series | serial: sh-sci: Fix .flush_buffer() issues | expand |
Hi Geert, One bikeshedding below. On Mon, Jun 24, 2019 at 02:35:39PM +0200, Geert Uytterhoeven wrote: [..] > diff --git a/drivers/tty/serial/sh-sci.c b/drivers/tty/serial/sh-sci.c > index abc705716aa094fd..d4504daff99263f5 100644 > --- a/drivers/tty/serial/sh-sci.c > +++ b/drivers/tty/serial/sh-sci.c > @@ -1398,6 +1398,7 @@ static void sci_dma_tx_work_fn(struct work_struct *work) > struct circ_buf *xmit = &port->state->xmit; > unsigned long flags; > dma_addr_t buf; > + int head, tail; > > /* > * DMA is idle now. > @@ -1407,16 +1408,23 @@ static void sci_dma_tx_work_fn(struct work_struct *work) > * consistent xmit buffer state. > */ > spin_lock_irq(&port->lock); > - buf = s->tx_dma_addr + (xmit->tail & (UART_XMIT_SIZE - 1)); > + head = xmit->head; > + tail = xmit->tail; > + buf = s->tx_dma_addr + (tail & (UART_XMIT_SIZE - 1)); > s->tx_dma_len = min_t(unsigned int, > - CIRC_CNT(xmit->head, xmit->tail, UART_XMIT_SIZE), > - CIRC_CNT_TO_END(xmit->head, xmit->tail, UART_XMIT_SIZE)); > - spin_unlock_irq(&port->lock); > + CIRC_CNT(head, tail, UART_XMIT_SIZE), > + CIRC_CNT_TO_END(head, tail, UART_XMIT_SIZE)); > + if (!s->tx_dma_len) { > + /* Transmit buffer has been flushed */ > + spin_unlock_irq(&port->lock); > + return; Since we can now return before using the result stored in 'buf', we could relocate the 'buf' calculation next to the return statement, just before calling dmaengine_prep_slave_single(), as micro-optimization. > + } > > desc = dmaengine_prep_slave_single(chan, buf, s->tx_dma_len, > DMA_MEM_TO_DEV, > DMA_PREP_INTERRUPT | DMA_CTRL_ACK); Otherwise: Reviewed-by: Eugeniu Rosca <erosca@de.adit-jv.com> Tested-by: Eugeniu Rosca <erosca@de.adit-jv.com>
diff --git a/drivers/tty/serial/sh-sci.c b/drivers/tty/serial/sh-sci.c index abc705716aa094fd..d4504daff99263f5 100644 --- a/drivers/tty/serial/sh-sci.c +++ b/drivers/tty/serial/sh-sci.c @@ -1398,6 +1398,7 @@ static void sci_dma_tx_work_fn(struct work_struct *work) struct circ_buf *xmit = &port->state->xmit; unsigned long flags; dma_addr_t buf; + int head, tail; /* * DMA is idle now. @@ -1407,16 +1408,23 @@ static void sci_dma_tx_work_fn(struct work_struct *work) * consistent xmit buffer state. */ spin_lock_irq(&port->lock); - buf = s->tx_dma_addr + (xmit->tail & (UART_XMIT_SIZE - 1)); + head = xmit->head; + tail = xmit->tail; + buf = s->tx_dma_addr + (tail & (UART_XMIT_SIZE - 1)); s->tx_dma_len = min_t(unsigned int, - CIRC_CNT(xmit->head, xmit->tail, UART_XMIT_SIZE), - CIRC_CNT_TO_END(xmit->head, xmit->tail, UART_XMIT_SIZE)); - spin_unlock_irq(&port->lock); + CIRC_CNT(head, tail, UART_XMIT_SIZE), + CIRC_CNT_TO_END(head, tail, UART_XMIT_SIZE)); + if (!s->tx_dma_len) { + /* Transmit buffer has been flushed */ + spin_unlock_irq(&port->lock); + return; + } desc = dmaengine_prep_slave_single(chan, buf, s->tx_dma_len, DMA_MEM_TO_DEV, DMA_PREP_INTERRUPT | DMA_CTRL_ACK); if (!desc) { + spin_unlock_irq(&port->lock); dev_warn(port->dev, "Failed preparing Tx DMA descriptor\n"); goto switch_to_pio; } @@ -1424,18 +1432,18 @@ static void sci_dma_tx_work_fn(struct work_struct *work) dma_sync_single_for_device(chan->device->dev, buf, s->tx_dma_len, DMA_TO_DEVICE); - spin_lock_irq(&port->lock); desc->callback = sci_dma_tx_complete; desc->callback_param = s; - spin_unlock_irq(&port->lock); s->cookie_tx = dmaengine_submit(desc); if (dma_submit_error(s->cookie_tx)) { + spin_unlock_irq(&port->lock); dev_warn(port->dev, "Failed submitting Tx DMA descriptor\n"); goto switch_to_pio; } + spin_unlock_irq(&port->lock); dev_dbg(port->dev, "%s: %p: %d...%d, cookie %d\n", - __func__, xmit->buf, xmit->tail, xmit->head, s->cookie_tx); + __func__, xmit->buf, tail, head, s->cookie_tx); dma_async_issue_pending(chan); return;
When uart_flush_buffer() is called, the .flush_buffer() callback zeroes the tx_dma_len field. This may race with the work queue function handling transmit DMA requests: 1. If the buffer is flushed before the first DMA API call, dmaengine_prep_slave_single() may be called with a zero length, causing the DMA request to never complete, leading to messages like: rcar-dmac e7300000.dma-controller: Channel Address Error happen and, with debug enabled: sh-sci e6e88000.serial: sci_dma_tx_work_fn: ffff800639b55000: 0...0, cookie 126 and DMA timeouts. 2. If the buffer is flushed after the first DMA API call, but before the second, dma_sync_single_for_device() may be called with a zero length, causing the transmit data not to be flushed to RAM, and leading to stale data being output. Fix this by: 1. Letting sci_dma_tx_work_fn() return immediately if the transmit buffer is empty, 2. Extending the critical section to cover all DMA preparational work, so tx_dma_len stays consistent for all of it, 3. Using local copies of circ_buf.head and circ_buf.tail, to make sure they match the actual operation above. Reported-by: Eugeniu Rosca <erosca@de.adit-jv.com> Suggested-by: Yoshihiro Shimoda <yoshihiro.shimoda.uh@renesas.com> Signed-off-by: Geert Uytterhoeven <geert+renesas@glider.be> --- drivers/tty/serial/sh-sci.c | 22 +++++++++++++++------- 1 file changed, 15 insertions(+), 7 deletions(-)