[PATCH v2 0/2] serial/amba-pl011: Activate TX IRQ passively

[PATCH v2 0/2] serial/amba-pl011: Activate TX IRQ passively

[Dave Martin]

Changes since v1:

This is a cleanup of v1 which significantly simplifies some things and
fixes a couple of bugs.

Due to these changes, I've stripped the Tested-bys.  If the guys who
already gave a Tested-by (thanks) have a chance, a re-test would be
much appreciated.

 * Refactored pl011_tx_char() and associated control flow to have
   "try to transmit a char" semantics, instead of "transmit, then
   tell me whether I can transmit another".

   This makes the code clearer, and ensures by construction that TXFF is
   always checked before sending a char, if character transmission was
   not triggered by an interrupt.

   As a side effect, this change exposed and also fixes a bug which can
   cause the last character to be sent twice when the FIFO becomes full
   (thanks to Dom Cobley for reporting this).

 * Removal of some redundant changes, and verbose comments that are no
   longer needed.

[Russell: I didn't do all the restructuring you suggested, because I
wanted to clean up and stabilise the existing series.  The result seems
a fair bit less convoluted, but I'm happy to further restructuring if
you feel it's needed.]


Original cover letter:

The current PL011 driver transmits a dummy character when the UART
is opened, to assert the TX IRQ for the first time
(see pl011_startup()).  The UART is put in loopback mode temporarily,
so the receiver presumably shouldn't see anything.

However...

At least some platforms containing a PL011 send characters down the
wire even when loopback mode is enabled.  This means that a
spurious NUL character may be seen at the receiver when the PL011 is
opened through the TTY layer.

The current code also temporarily sets the baud rate to maximum and
the character width to the minimum, to that the dummy TX completes
as quickly as possible.  If this is seen by the receiver it will
result in a framing error and can knock the receiver out of sync --
turning subsequent output into garbage until synchronisation
is reestablished.  (Particularly problematic during boot with systemd.)

To avoid spurious transmissions, this patch removes assumptions about
whether the TX IRQ will fire until at least one TX IRQ has been seen.

Instead, the UART will unmask the TX IRQ and then slow-start via
polling and timer-based soft IRQs initially.  If the TTY layer writes
enough data to fill the FIFO to the interrupt threshold in one go,
the TX IRQ should assert, at which point the driver changes to
fully interrupt-driven TX.

In this way, the TX IRQ is activated as a side-effect instead of
being done deliberately.

This should also mean that the driver works on the SBSA Generic
UART[1] (a cut-down PL011) without invasive changes.  The Generic
UART lacks some features needed for the dummy TX approach to work
(FIFO disabling and loopback).

[1] Server Base System Architecture (ARM-DEN-0029-v2.3)
    http://infocenter.arm.com/
    (click-thru required :/)

Dave Martin (2):
  serial/amba-pl011: Activate TX IRQ passively
  serial/amba-pl011: Leave the TX IRQ alone when the UART is not open

 drivers/tty/serial/amba-pl011.c |  120 +++++++++++++++++++++++++--------------
 1 file changed, 76 insertions(+), 44 deletions(-)

-- 
1.7.10.4

[PATCH v2 1/2] serial/amba-pl011: Activate TX IRQ passively

[Dave Martin]

The current PL011 driver transmits a dummy character when the UART
is opened, to assert the TX IRQ for the first time
(see pl011_startup()).  The UART is put in loopback mode temporarily,
so the receiver presumably shouldn't see anything.

However...

At least some platforms containing a PL011 send characters down the
wire even when loopback mode is enabled.  This means that a
spurious NUL character may be seen at the receiver when the PL011 is
opened through the TTY layer.

The current code also temporarily sets the baud rate to maximum and
the character width to the minimum, to that the dummy TX completes
as quickly as possible.  If this is seen by the receiver it will
result in a framing error and can knock the receiver out of sync --
turning subsequent output into garbage until synchronisation
is reestablished.  (Particularly problematic during boot with systemd.)

To avoid spurious transmissions, this patch removes assumptions about
whether the TX IRQ will fire until at least one TX IRQ has been seen.

Instead, the UART will unmask the TX IRQ and then slow-start via
polling and timer-based soft IRQs initially.  If the TTY layer writes
enough data to fill the FIFO to the interrupt threshold in one go,
the TX IRQ should assert, at which point the driver changes to
fully interrupt-driven TX.

This should also mean that the driver works on the SBSA Generic
UART[1] (a cut-down PL011) without invasive changes.  The Generic
UART lacks some features needed for the dummy TX approach to work
(FIFO disabling and loopback).

[1] Server Base System Architecture (ARM-DEN-0029-v2.3)
    http://infocenter.arm.com/
    (click-thru required :/)

Signed-off-by: Dave Martin <Dave.Martin@arm.com>
---
 drivers/tty/serial/amba-pl011.c |  121 +++++++++++++++++++++++++--------------
 1 file changed, 78 insertions(+), 43 deletions(-)

diff --git a/drivers/tty/serial/amba-pl011.c b/drivers/tty/serial/amba-pl011.c
index 8d94c19..bdb4831 100644
--- a/drivers/tty/serial/amba-pl011.c
+++ b/drivers/tty/serial/amba-pl011.c
@@ -58,6 +58,7 @@
 #include <linux/pinctrl/consumer.h>
 #include <linux/sizes.h>
 #include <linux/io.h>
+#include <linux/workqueue.h>
 
 #define UART_NR			14
 
@@ -156,7 +157,9 @@ struct uart_amba_port {
 	unsigned int		lcrh_tx;	/* vendor-specific */
 	unsigned int		lcrh_rx;	/* vendor-specific */
 	unsigned int		old_cr;		/* state during shutdown */
+	struct delayed_work	tx_softirq_work;
 	bool			autorts;
+	bool			tx_irq_seen;
 	char			type[12];
 #ifdef CONFIG_DMA_ENGINE
 	/* DMA stuff */
@@ -440,8 +443,9 @@ static void pl011_dma_remove(struct uart_amba_port *uap)
 		dma_release_channel(uap->dmarx.chan);
 }
 
-/* Forward declare this for the refill routine */
+/* Forward declare these for the refill routine */
 static int pl011_dma_tx_refill(struct uart_amba_port *uap);
+static void pl011_start_tx_pio(struct uart_amba_port *uap);
 
 /*
  * The current DMA TX buffer has been sent.
@@ -479,14 +483,13 @@ static void pl011_dma_tx_callback(void *data)
 		return;
 	}
 
-	if (pl011_dma_tx_refill(uap) <= 0) {
+	if (pl011_dma_tx_refill(uap) <= 0)
 		/*
 		 * We didn't queue a DMA buffer for some reason, but we
 		 * have data pending to be sent.  Re-enable the TX IRQ.
 		 */
-		uap->im |= UART011_TXIM;
-		writew(uap->im, uap->port.membase + UART011_IMSC);
-	}
+		pl011_start_tx_pio(uap);
+
 	spin_unlock_irqrestore(&uap->port.lock, flags);
 }
 
@@ -1208,15 +1211,24 @@ static void pl011_stop_tx(struct uart_port *port)
 	pl011_dma_tx_stop(uap);
 }
 
+static void pl011_tx_chars(struct uart_amba_port *uap);
+
+/* Start TX with programmed I/O only (no DMA) */
+static void pl011_start_tx_pio(struct uart_amba_port *uap)
+{
+	uap->im |= UART011_TXIM;
+	writew(uap->im, uap->port.membase + UART011_IMSC);
+	if (!uap->tx_irq_seen)
+		pl011_tx_chars(uap);
+}
+
 static void pl011_start_tx(struct uart_port *port)
 {
 	struct uart_amba_port *uap =
 	    container_of(port, struct uart_amba_port, port);
 
-	if (!pl011_dma_tx_start(uap)) {
-		uap->im |= UART011_TXIM;
-		writew(uap->im, uap->port.membase + UART011_IMSC);
-	}
+	if (!pl011_dma_tx_start(uap))
+		pl011_start_tx_pio(uap);
 }
 
 static void pl011_stop_rx(struct uart_port *port)
@@ -1274,16 +1286,31 @@ __acquires(&uap->port.lock)
 	spin_lock(&uap->port.lock);
 }
 
+static bool pl011_tx_char(struct uart_amba_port *uap, unsigned char c)
+{
+	if (unlikely(!uap->tx_irq_seen) &&
+	    readw(uap->port.membase + UART01x_FR) & UART01x_FR_TXFF)
+		return false; /* unable to transmit character */
+
+	writew(c, uap->port.membase + UART01x_DR);
+	uap->port.icount.tx++;
+
+	return true;
+}
+
 static void pl011_tx_chars(struct uart_amba_port *uap)
 {
 	struct circ_buf *xmit = &uap->port.state->xmit;
-	int count;
+	int count = uap->fifosize >> 1;
+
+	if (unlikely(!uap->tx_irq_seen))
+		count = uap->fifosize;
 
 	if (uap->port.x_char) {
-		writew(uap->port.x_char, uap->port.membase + UART01x_DR);
-		uap->port.icount.tx++;
+		if (!pl011_tx_char(uap, uap->port.x_char))
+			goto out;
 		uap->port.x_char = 0;
-		return;
+		--count;
 	}
 	if (uart_circ_empty(xmit) || uart_tx_stopped(&uap->port)) {
 		pl011_stop_tx(&uap->port);
@@ -1294,20 +1321,23 @@ static void pl011_tx_chars(struct uart_amba_port *uap)
 	if (pl011_dma_tx_irq(uap))
 		return;
 
-	count = uap->fifosize >> 1;
-	do {
-		writew(xmit->buf[xmit->tail], uap->port.membase + UART01x_DR);
+	while (count-- > 0 && pl011_tx_char(uap, xmit->buf[xmit->tail])) {
 		xmit->tail = (xmit->tail + 1) & (UART_XMIT_SIZE - 1);
-		uap->port.icount.tx++;
 		if (uart_circ_empty(xmit))
 			break;
-	} while (--count > 0);
+	}
 
 	if (uart_circ_chars_pending(xmit) < WAKEUP_CHARS)
 		uart_write_wakeup(&uap->port);
 
-	if (uart_circ_empty(xmit))
+	if (uart_circ_empty(xmit)) {
 		pl011_stop_tx(&uap->port);
+		return;
+	}
+
+out:
+	if (unlikely(!uap->tx_irq_seen))
+		schedule_delayed_work(&uap->tx_softirq_work, uap->port.timeout);
 }
 
 static void pl011_modem_status(struct uart_amba_port *uap)
@@ -1334,6 +1364,25 @@ static void pl011_modem_status(struct uart_amba_port *uap)
 	wake_up_interruptible(&uap->port.state->port.delta_msr_wait);
 }
 
+static void pl011_tx_softirq(struct work_struct *work)
+{
+	struct delayed_work *dwork = to_delayed_work(work);
+	struct uart_amba_port *uap =
+		container_of(dwork, struct uart_amba_port, tx_softirq_work);
+
+	spin_lock_irq(&uap->port.lock);
+	pl011_tx_chars(uap);
+	spin_unlock_irq(&uap->port.lock);
+}
+
+static void pl011_tx_irq_seen(struct uart_amba_port *uap)
+{
+	if (!likely(uap->tx_irq_seen)) {
+		uap->tx_irq_seen = true;
+		cancel_delayed_work(&uap->tx_softirq_work);
+	}
+}
+
 static irqreturn_t pl011_int(int irq, void *dev_id)
 {
 	struct uart_amba_port *uap = dev_id;
@@ -1372,8 +1421,10 @@ static irqreturn_t pl011_int(int irq, void *dev_id)
 			if (status & (UART011_DSRMIS|UART011_DCDMIS|
 				      UART011_CTSMIS|UART011_RIMIS))
 				pl011_modem_status(uap);
-			if (status & UART011_TXIS)
+			if (status & UART011_TXIS) {
+				pl011_tx_irq_seen(uap);
 				pl011_tx_chars(uap);
+			}
 
 			if (pass_counter-- == 0)
 				break;
@@ -1577,7 +1628,7 @@ static int pl011_startup(struct uart_port *port)
 {
 	struct uart_amba_port *uap =
 	    container_of(port, struct uart_amba_port, port);
-	unsigned int cr, lcr_h, fbrd, ibrd;
+	unsigned int cr;
 	int retval;
 
 	retval = pl011_hwinit(port);
@@ -1595,29 +1646,10 @@ static int pl011_startup(struct uart_port *port)
 
 	writew(uap->vendor->ifls, uap->port.membase + UART011_IFLS);
 
-	/*
-	 * Provoke TX FIFO interrupt into asserting. Taking care to preserve
-	 * baud rate and data format specified by FBRD, IBRD and LCRH as the
-	 * UART may already be in use as a console.
-	 */
-	spin_lock_irq(&uap->port.lock);
-
-	fbrd = readw(uap->port.membase + UART011_FBRD);
-	ibrd = readw(uap->port.membase + UART011_IBRD);
-	lcr_h = readw(uap->port.membase + uap->lcrh_rx);
-
-	cr = UART01x_CR_UARTEN | UART011_CR_TXE | UART011_CR_LBE;
-	writew(cr, uap->port.membase + UART011_CR);
-	writew(0, uap->port.membase + UART011_FBRD);
-	writew(1, uap->port.membase + UART011_IBRD);
-	pl011_write_lcr_h(uap, 0);
-	writew(0, uap->port.membase + UART01x_DR);
-	while (readw(uap->port.membase + UART01x_FR) & UART01x_FR_BUSY)
-		barrier();
+	/* Assume that TX IRQ doesn't work until we see one: */
+	uap->tx_irq_seen = false;
 
-	writew(fbrd, uap->port.membase + UART011_FBRD);
-	writew(ibrd, uap->port.membase + UART011_IBRD);
-	pl011_write_lcr_h(uap, lcr_h);
+	spin_lock_irq(&uap->port.lock);
 
 	/* restore RTS and DTR */
 	cr = uap->old_cr & (UART011_CR_RTS | UART011_CR_DTR);
@@ -1672,6 +1704,8 @@ static void pl011_shutdown(struct uart_port *port)
 	    container_of(port, struct uart_amba_port, port);
 	unsigned int cr;
 
+	cancel_delayed_work_sync(&uap->tx_softirq_work);
+
 	/*
 	 * disable all interrupts
 	 */
@@ -2218,6 +2252,7 @@ static int pl011_probe(struct amba_device *dev, const struct amba_id *id)
 	uap->port.ops = &amba_pl011_pops;
 	uap->port.flags = UPF_BOOT_AUTOCONF;
 	uap->port.line = i;
+	INIT_DELAYED_WORK(&uap->tx_softirq_work, pl011_tx_softirq);
 	pl011_dma_probe(&dev->dev, uap);
 
 	/* Ensure interrupts from this UART are masked and cleared */
-- 
1.7.10.4

[PATCH v2 2/2] serial/amba-pl011: Leave the TX IRQ alone when the UART is not open

[Dave Martin]

Getting the TX IRQ re-asserted from scratch can be inefficient in
some setups.

This patch avoids clearing the TX IRQ across pl011_shutdown()...
pl011_startup(), so that if the port is closed and reopened, the
IRQ will still work afterwards without having to bootstrap it again.

The TX IRQ continues to be masked in IMSC when the UART is not in
use.

Signed-off-by: Dave Martin <Dave.Martin@arm.com>
---
 drivers/tty/serial/amba-pl011.c |    5 +----
 1 file changed, 1 insertion(+), 4 deletions(-)

diff --git a/drivers/tty/serial/amba-pl011.c b/drivers/tty/serial/amba-pl011.c
index bdb4831..c3d631d 100644
--- a/drivers/tty/serial/amba-pl011.c
+++ b/drivers/tty/serial/amba-pl011.c
@@ -1646,9 +1646,6 @@ static int pl011_startup(struct uart_port *port)
 
 	writew(uap->vendor->ifls, uap->port.membase + UART011_IFLS);
 
-	/* Assume that TX IRQ doesn't work until we see one: */
-	uap->tx_irq_seen = false;
-
 	spin_lock_irq(&uap->port.lock);
 
 	/* restore RTS and DTR */
@@ -1712,7 +1709,7 @@ static void pl011_shutdown(struct uart_port *port)
 	spin_lock_irq(&uap->port.lock);
 	uap->im = 0;
 	writew(uap->im, uap->port.membase + UART011_IMSC);
-	writew(0xffff, uap->port.membase + UART011_ICR);
+	writew(0xffff & ~UART011_TXIS, uap->port.membase + UART011_ICR);
 	spin_unlock_irq(&uap->port.lock);
 
 	pl011_dma_shutdown(uap);
-- 
1.7.10.4

Re: [PATCH v2 2/2] serial/amba-pl011: Leave the TX IRQ alone when the UART is not open

[Greg Kroah-Hartman]

On Fri, Mar 13, 2015 at 03:56:22PM +0000, Dave Martin wrote:
> Getting the TX IRQ re-asserted from scratch can be inefficient in
> some setups.
> 
> This patch avoids clearing the TX IRQ across pl011_shutdown()...
> pl011_startup(), so that if the port is closed and reopened, the
> IRQ will still work afterwards without having to bootstrap it again.
> 
> The TX IRQ continues to be masked in IMSC when the UART is not in
> use.
> 
> Signed-off-by: Dave Martin <Dave.Martin@arm.com>
> ---
>  drivers/tty/serial/amba-pl011.c |    5 +----
>  1 file changed, 1 insertion(+), 4 deletions(-)

This doesn't apply to my tty-testing branch of tty.git on git.kernel.org
:(

Can you please rebase it and resend?

thanks,

greg k-h

Re: [PATCH v2 1/2] serial/amba-pl011: Activate TX IRQ passively

[Greg Kroah-Hartman]

On Fri, Mar 13, 2015 at 03:56:21PM +0000, Dave Martin wrote:
> The current PL011 driver transmits a dummy character when the UART
> is opened, to assert the TX IRQ for the first time
> (see pl011_startup()).  The UART is put in loopback mode temporarily,
> so the receiver presumably shouldn't see anything.
> 
> However...
> 
> At least some platforms containing a PL011 send characters down the
> wire even when loopback mode is enabled.  This means that a
> spurious NUL character may be seen at the receiver when the PL011 is
> opened through the TTY layer.
> 
> The current code also temporarily sets the baud rate to maximum and
> the character width to the minimum, to that the dummy TX completes
> as quickly as possible.  If this is seen by the receiver it will
> result in a framing error and can knock the receiver out of sync --
> turning subsequent output into garbage until synchronisation
> is reestablished.  (Particularly problematic during boot with systemd.)
> 
> To avoid spurious transmissions, this patch removes assumptions about
> whether the TX IRQ will fire until at least one TX IRQ has been seen.
> 
> Instead, the UART will unmask the TX IRQ and then slow-start via
> polling and timer-based soft IRQs initially.  If the TTY layer writes
> enough data to fill the FIFO to the interrupt threshold in one go,
> the TX IRQ should assert, at which point the driver changes to
> fully interrupt-driven TX.
> 
> This should also mean that the driver works on the SBSA Generic
> UART[1] (a cut-down PL011) without invasive changes.  The Generic
> UART lacks some features needed for the dummy TX approach to work
> (FIFO disabling and loopback).
> 
> [1] Server Base System Architecture (ARM-DEN-0029-v2.3)
>     http://infocenter.arm.com/
>     (click-thru required :/)
> 
> Signed-off-by: Dave Martin <Dave.Martin@arm.com>
> ---
>  drivers/tty/serial/amba-pl011.c |  121 +++++++++++++++++++++++++--------------
>  1 file changed, 78 insertions(+), 43 deletions(-)

Wait, I already took this patch, right?

Was there some difference here with this version?

confused,

greg k-h