[PATCH v2 0/7] 8250 DW UART fixes when under constant Rx pressure

[PATCH v2 0/7] 8250 DW UART fixes when under constant Rx pressure

[Ilpo Järvinen]

Hi all,

Here are fixes to 8250 DW UART conditions that mostly occur in scenarios
under constant Rx pressure which are made complicated by BUSY handling
of DW UARTs (used when !uart_16550_compatible).

A few of the changes touch also 8250_port but it's mostly moving existing
code around (except for the extra synchronize_irq() in shutdown).

I'll do UART_IIR_RX_TIMEOUT move to switch/case separately from this
fix series and a few other cleanups Andy brought up.


v2:
- Added Fixes tags pointing DW & INTC10EE introducing commits
- Added Cc stable to prerequisite patches
- Separate adding serial8250_handle_irq_locked() to own patch
- Put new exports to SERIAL_8250 namespace (eventually all 8250
  exports should move there, but out-of-scope for this series)
- Changes to no_int_count
	- Change type to u8
	- Use modulo in increment, add a define for it
	- Perform kick only on 4th NO_INT
- Use serial_port_in/out() throughout the series
- Add FIXME comments to ndelay(frame_time) about very low baud rates
- Add cleanup.h
- Tweak lockdep.h place among misordered includes
- Wording tweaks to changelogs and comments

Ilpo Järvinen (7):
  serial: 8250: Protect LCR write in shutdown
  serial: 8250_dw: Avoid unnecessary LCR writes
  serial: 8250: Add serial8250_handle_irq_locked()
  serial: 8250_dw: Rework dw8250_handle_irq() locking and IIR handling
  serial: 8250_dw: Rework IIR_NO_INT handling to stop interrupt storm
  serial: 8250: Add late synchronize_irq() to shutdown to handle DW UART
    BUSY
  serial: 8250_dw: Ensure BUSY is deasserted

 drivers/tty/serial/8250/8250.h      |  25 +++
 drivers/tty/serial/8250/8250_dw.c   | 309 ++++++++++++++++++++++------
 drivers/tty/serial/8250/8250_port.c |  69 ++++---
 include/linux/serial_8250.h         |   1 +
 4 files changed, 313 insertions(+), 91 deletions(-)


base-commit: 8f0b4cce4481fb22653697cced8d0d04027cb1e8
-- 
2.39.5

[PATCH v2 1/7] serial: 8250: Protect LCR write in shutdown

[Ilpo Järvinen]

The 8250_dw driver needs to potentially perform very complex operations
during LCR writes because its BUSY handling prevents updates to LCR
while UART is BUSY (which is not fully under our control without those
complex operations). Thus, LCR writes should occur under port's lock.

Move LCR write under port's lock in serial8250_do_shutdown(). Also
split the LCR RMW so that the logic is on a separate line for clarity.

Tested-by: "Bandal, Shankar" <shankar.bandal@intel.com>
Tested-by: "Murthy, Shanth" <shanth.murthy@intel.com>
Cc: stable@vger.kernel.org
Signed-off-by: Ilpo Järvinen <ilpo.jarvinen@linux.intel.com>
---
 drivers/tty/serial/8250/8250_port.c | 11 ++++++-----
 1 file changed, 6 insertions(+), 5 deletions(-)

diff --git a/drivers/tty/serial/8250/8250_port.c b/drivers/tty/serial/8250/8250_port.c
index 719faf92aa8a..f7a3c5555204 100644
--- a/drivers/tty/serial/8250/8250_port.c
+++ b/drivers/tty/serial/8250/8250_port.c
@@ -2350,6 +2350,7 @@ static int serial8250_startup(struct uart_port *port)
 void serial8250_do_shutdown(struct uart_port *port)
 {
 	struct uart_8250_port *up = up_to_u8250p(port);
+	u32 lcr;
 
 	serial8250_rpm_get(up);
 	/*
@@ -2376,13 +2377,13 @@ void serial8250_do_shutdown(struct uart_port *port)
 			port->mctrl &= ~TIOCM_OUT2;
 
 		serial8250_set_mctrl(port, port->mctrl);
+
+		/* Disable break condition */
+		lcr = serial_port_in(port, UART_LCR);
+		lcr &= ~UART_LCR_SBC;
+		serial_port_out(port, UART_LCR, lcr);
 	}
 
-	/*
-	 * Disable break condition and FIFOs
-	 */
-	serial_port_out(port, UART_LCR,
-			serial_port_in(port, UART_LCR) & ~UART_LCR_SBC);
 	serial8250_clear_fifos(up);
 
 	rsa_disable(up);
-- 
2.39.5

Re: [PATCH v2 1/7] serial: 8250: Protect LCR write in shutdown

[Andy Shevchenko]

On Wed, Jan 28, 2026 at 12:52:55PM +0200, Ilpo Järvinen wrote:
> The 8250_dw driver needs to potentially perform very complex operations
> during LCR writes because its BUSY handling prevents updates to LCR
> while UART is BUSY (which is not fully under our control without those
> complex operations). Thus, LCR writes should occur under port's lock.
> 
> Move LCR write under port's lock in serial8250_do_shutdown(). Also
> split the LCR RMW so that the logic is on a separate line for clarity.

Reviewed-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>

-- 
With Best Regards,
Andy Shevchenko

[PATCH v2 2/7] serial: 8250_dw: Avoid unnecessary LCR writes

[Ilpo Järvinen]

When DW UART is configured with BUSY flag, LCR writes may not always
succeed which can make any LCR write complex and very expensive.
Performing write directly can trigger IRQ and the driver has to perform
complex and distruptive sequence while retrying the write.

Therefore, it's better to avoid doing LCR write that would not change
the value of the LCR register. Add LCR write avoidance code into the
8250_dw driver's .serial_out() functions.

Reported-by: "Bandal, Shankar" <shankar.bandal@intel.com>
Tested-by: "Bandal, Shankar" <shankar.bandal@intel.com>
Tested-by: "Murthy, Shanth" <shanth.murthy@intel.com>
Cc: stable@vger.kernel.org
Signed-off-by: Ilpo Järvinen <ilpo.jarvinen@linux.intel.com>
---
 drivers/tty/serial/8250/8250_dw.c | 31 +++++++++++++++++++++++++++++++
 1 file changed, 31 insertions(+)

diff --git a/drivers/tty/serial/8250/8250_dw.c b/drivers/tty/serial/8250/8250_dw.c
index 27af83f0ff46..7500b1ff1ac1 100644
--- a/drivers/tty/serial/8250/8250_dw.c
+++ b/drivers/tty/serial/8250/8250_dw.c
@@ -181,6 +181,22 @@ static void dw8250_check_lcr(struct uart_port *p, unsigned int offset, u32 value
 	 */
 }
 
+/*
+ * With BUSY, LCR writes can be very expensive (IRQ + complex retry logic).
+ * If the write does not change the value of the LCR register, skip it entirely.
+ */
+static bool dw8250_can_skip_reg_write(struct uart_port *p, unsigned int offset, u32 value)
+{
+	struct dw8250_data *d = to_dw8250_data(p->private_data);
+	u32 lcr;
+
+	if (offset != UART_LCR || d->uart_16550_compatible)
+		return false;
+
+	lcr = serial_port_in(p, offset);
+	return lcr == value;
+}
+
 /* Returns once the transmitter is empty or we run out of retries */
 static void dw8250_tx_wait_empty(struct uart_port *p)
 {
@@ -207,12 +223,18 @@ static void dw8250_tx_wait_empty(struct uart_port *p)
 
 static void dw8250_serial_out(struct uart_port *p, unsigned int offset, u32 value)
 {
+	if (dw8250_can_skip_reg_write(p, offset, value))
+		return;
+
 	writeb(value, p->membase + (offset << p->regshift));
 	dw8250_check_lcr(p, offset, value);
 }
 
 static void dw8250_serial_out38x(struct uart_port *p, unsigned int offset, u32 value)
 {
+	if (dw8250_can_skip_reg_write(p, offset, value))
+		return;
+
 	/* Allow the TX to drain before we reconfigure */
 	if (offset == UART_LCR)
 		dw8250_tx_wait_empty(p);
@@ -237,6 +259,9 @@ static u32 dw8250_serial_inq(struct uart_port *p, unsigned int offset)
 
 static void dw8250_serial_outq(struct uart_port *p, unsigned int offset, u32 value)
 {
+	if (dw8250_can_skip_reg_write(p, offset, value))
+		return;
+
 	value &= 0xff;
 	__raw_writeq(value, p->membase + (offset << p->regshift));
 	/* Read back to ensure register write ordering. */
@@ -248,6 +273,9 @@ static void dw8250_serial_outq(struct uart_port *p, unsigned int offset, u32 val
 
 static void dw8250_serial_out32(struct uart_port *p, unsigned int offset, u32 value)
 {
+	if (dw8250_can_skip_reg_write(p, offset, value))
+		return;
+
 	writel(value, p->membase + (offset << p->regshift));
 	dw8250_check_lcr(p, offset, value);
 }
@@ -261,6 +289,9 @@ static u32 dw8250_serial_in32(struct uart_port *p, unsigned int offset)
 
 static void dw8250_serial_out32be(struct uart_port *p, unsigned int offset, u32 value)
 {
+	if (dw8250_can_skip_reg_write(p, offset, value))
+		return;
+
 	iowrite32be(value, p->membase + (offset << p->regshift));
 	dw8250_check_lcr(p, offset, value);
 }
-- 
2.39.5

Re: [PATCH v2 2/7] serial: 8250_dw: Avoid unnecessary LCR writes

[Andy Shevchenko]

On Wed, Jan 28, 2026 at 12:52:56PM +0200, Ilpo Järvinen wrote:
> When DW UART is configured with BUSY flag, LCR writes may not always
> succeed which can make any LCR write complex and very expensive.
> Performing write directly can trigger IRQ and the driver has to perform
> complex and distruptive sequence while retrying the write.
> 
> Therefore, it's better to avoid doing LCR write that would not change
> the value of the LCR register. Add LCR write avoidance code into the
> 8250_dw driver's .serial_out() functions.

Reviewed-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>

...

> +static bool dw8250_can_skip_reg_write(struct uart_port *p, unsigned int offset, u32 value)
> +{
> +	struct dw8250_data *d = to_dw8250_data(p->private_data);
> +	u32 lcr;
> +
> +	if (offset != UART_LCR || d->uart_16550_compatible)
> +		return false;
> +
> +	lcr = serial_port_in(p, offset);
> +	return lcr == value;

I don't expect the side-effects, but strictly speaking this compares
the 32-bit values, while in some callbacks the only part of that
is being used (often lowest 8 bits).

> +}

-- 
With Best Regards,
Andy Shevchenko

[PATCH v2 3/7] serial: 8250: Add serial8250_handle_irq_locked()

[Ilpo Järvinen]

8250_port exports serial8250_handle_irq() to HW specific 8250 drivers.
It takes port's lock within but a HW specific 8250 driver may want to
take port's lock itself, do something, and then call the generic
handler in 8250_port but to do that, the caller has to release port's
lock for no good reason.

Introduce serial8250_handle_irq_locked() which a HW specific driver can
call while already holding port's lock.

As this is new export, put it straight into a namespace (where all 8250
exports should eventually be moved).

Tested-by: "Bandal, Shankar" <shankar.bandal@intel.com>
Tested-by: "Murthy, Shanth" <shanth.murthy@intel.com>
Cc: stable@vger.kernel.org
Signed-off-by: Ilpo Järvinen <ilpo.jarvinen@linux.intel.com>
---
 drivers/tty/serial/8250/8250_port.c | 24 ++++++++++++++++--------
 include/linux/serial_8250.h         |  1 +
 2 files changed, 17 insertions(+), 8 deletions(-)

diff --git a/drivers/tty/serial/8250/8250_port.c b/drivers/tty/serial/8250/8250_port.c
index f7a3c5555204..bc223eb1f474 100644
--- a/drivers/tty/serial/8250/8250_port.c
+++ b/drivers/tty/serial/8250/8250_port.c
@@ -18,6 +18,7 @@
 #include <linux/irq.h>
 #include <linux/console.h>
 #include <linux/gpio/consumer.h>
+#include <linux/lockdep.h>
 #include <linux/sysrq.h>
 #include <linux/delay.h>
 #include <linux/platform_device.h>
@@ -1782,20 +1783,16 @@ static bool handle_rx_dma(struct uart_8250_port *up, unsigned int iir)
 }
 
 /*
- * This handles the interrupt from one port.
+ * Context: port's lock must be held by the caller.
  */
-int serial8250_handle_irq(struct uart_port *port, unsigned int iir)
+void serial8250_handle_irq_locked(struct uart_port *port, unsigned int iir)
 {
 	struct uart_8250_port *up = up_to_u8250p(port);
 	struct tty_port *tport = &port->state->port;
 	bool skip_rx = false;
-	unsigned long flags;
 	u16 status;
 
-	if (iir & UART_IIR_NO_INT)
-		return 0;
-
-	uart_port_lock_irqsave(port, &flags);
+	lockdep_assert_held_once(&port->lock);
 
 	status = serial_lsr_in(up);
 
@@ -1828,8 +1825,19 @@ int serial8250_handle_irq(struct uart_port *port, unsigned int iir)
 		else if (!up->dma->tx_running)
 			__stop_tx(up);
 	}
+}
+EXPORT_SYMBOL_NS_GPL(serial8250_handle_irq_locked, "SERIAL_8250");
 
-	uart_unlock_and_check_sysrq_irqrestore(port, flags);
+/*
+ * This handles the interrupt from one port.
+ */
+int serial8250_handle_irq(struct uart_port *port, unsigned int iir)
+{
+	if (iir & UART_IIR_NO_INT)
+		return 0;
+
+	guard(uart_port_lock_irqsave)(port);
+	serial8250_handle_irq_locked(port, iir);
 
 	return 1;
 }
diff --git a/include/linux/serial_8250.h b/include/linux/serial_8250.h
index 01efdce0fda0..a95b2d143d24 100644
--- a/include/linux/serial_8250.h
+++ b/include/linux/serial_8250.h
@@ -195,6 +195,7 @@ void serial8250_do_set_mctrl(struct uart_port *port, unsigned int mctrl);
 void serial8250_do_set_divisor(struct uart_port *port, unsigned int baud,
 			       unsigned int quot);
 int fsl8250_handle_irq(struct uart_port *port);
+void serial8250_handle_irq_locked(struct uart_port *port, unsigned int iir);
 int serial8250_handle_irq(struct uart_port *port, unsigned int iir);
 u16 serial8250_rx_chars(struct uart_8250_port *up, u16 lsr);
 void serial8250_read_char(struct uart_8250_port *up, u16 lsr);
-- 
2.39.5

Re: [PATCH v2 3/7] serial: 8250: Add serial8250_handle_irq_locked()

[Andy Shevchenko]

On Wed, Jan 28, 2026 at 12:52:57PM +0200, Ilpo Järvinen wrote:
> 8250_port exports serial8250_handle_irq() to HW specific 8250 drivers.
> It takes port's lock within but a HW specific 8250 driver may want to
> take port's lock itself, do something, and then call the generic
> handler in 8250_port but to do that, the caller has to release port's
> lock for no good reason.
> 
> Introduce serial8250_handle_irq_locked() which a HW specific driver can
> call while already holding port's lock.
> 
> As this is new export, put it straight into a namespace (where all 8250
> exports should eventually be moved).

Reviewed-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>

...

> +void serial8250_handle_irq_locked(struct uart_port *port, unsigned int iir);
>  int serial8250_handle_irq(struct uart_port *port, unsigned int iir);

Looking at these I think at some point we can move to 'u32 iir'.

>  u16 serial8250_rx_chars(struct uart_8250_port *up, u16 lsr);
>  void serial8250_read_char(struct uart_8250_port *up, u16 lsr);

-- 
With Best Regards,
Andy Shevchenko

Re: [PATCH v2 3/7] serial: 8250: Add serial8250_handle_irq_locked()

[Ilpo Järvinen]

[-- Attachment #1: Type: text/plain, Size: 1090 bytes --]

On Wed, 28 Jan 2026, Andy Shevchenko wrote:

> On Wed, Jan 28, 2026 at 12:52:57PM +0200, Ilpo Järvinen wrote:
> > 8250_port exports serial8250_handle_irq() to HW specific 8250 drivers.
> > It takes port's lock within but a HW specific 8250 driver may want to
> > take port's lock itself, do something, and then call the generic
> > handler in 8250_port but to do that, the caller has to release port's
> > lock for no good reason.
> > 
> > Introduce serial8250_handle_irq_locked() which a HW specific driver can
> > call while already holding port's lock.
> > 
> > As this is new export, put it straight into a namespace (where all 8250
> > exports should eventually be moved).
> 
> Reviewed-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
> 
> ...
> 
> > +void serial8250_handle_irq_locked(struct uart_port *port, unsigned int iir);
> >  int serial8250_handle_irq(struct uart_port *port, unsigned int iir);
> 
> Looking at these I think at some point we can move to 'u32 iir'.

Yes (though for common 8250 code even u8 would probably suffice).

-- 
 i.

[PATCH v2 4/7] serial: 8250_dw: Rework dw8250_handle_irq() locking and IIR handling

[Ilpo Järvinen]

dw8250_handle_irq() takes port's lock multiple times with no good
reason to release it in between and calls serial8250_handle_irq()
that also takes port's lock.

Take port's lock only once in dw8250_handle_irq() and use
serial8250_handle_irq_locked() to avoid releasing port's lock in
between.

As IIR_NO_INT check in serial8250_handle_irq() was outside of port's
lock, it has to be done already in dw8250_handle_irq().

DW UART can, in addition to IIR_NO_INT, report BUSY_DETECT (0x7) which
collided with the IIR_NO_INT (0x1) check in serial8250_handle_irq()
(because & is used instead of ==) meaning that no other work is done by
serial8250_handle_irq() during an BUSY_DETECT interrupt.

This allows reorganizing code in dw8250_handle_irq() to do both
IIR_NO_INT and BUSY_DETECT handling right at the start simplifying
the logic.

Tested-by: "Bandal, Shankar" <shankar.bandal@intel.com>
Tested-by: "Murthy, Shanth" <shanth.murthy@intel.com>
Cc: stable@vger.kernel.org
Signed-off-by: Ilpo Järvinen <ilpo.jarvinen@linux.intel.com>
---
 drivers/tty/serial/8250/8250_dw.c | 37 ++++++++++++++++++-------------
 1 file changed, 21 insertions(+), 16 deletions(-)

diff --git a/drivers/tty/serial/8250/8250_dw.c b/drivers/tty/serial/8250/8250_dw.c
index 7500b1ff1ac1..964750d17186 100644
--- a/drivers/tty/serial/8250/8250_dw.c
+++ b/drivers/tty/serial/8250/8250_dw.c
@@ -9,6 +9,9 @@
  * LCR is written whilst busy.  If it is, then a busy detect interrupt is
  * raised, the LCR needs to be rewritten and the uart status register read.
  */
+#include <linux/bitfield.h>
+#include <linux/bits.h>
+#include <linux/cleanup.h>
 #include <linux/clk.h>
 #include <linux/delay.h>
 #include <linux/device.h>
@@ -40,6 +43,8 @@
 #define RZN1_UART_RDMACR 0x110 /* DMA Control Register Receive Mode */
 
 /* DesignWare specific register fields */
+#define DW_UART_IIR_IID			GENMASK(3, 0)
+
 #define DW_UART_MCR_SIRE		BIT(6)
 
 /* Renesas specific register fields */
@@ -312,7 +317,19 @@ static int dw8250_handle_irq(struct uart_port *p)
 	bool rx_timeout = (iir & 0x3f) == UART_IIR_RX_TIMEOUT;
 	unsigned int quirks = d->pdata->quirks;
 	unsigned int status;
-	unsigned long flags;
+
+	switch (FIELD_GET(DW_UART_IIR_IID, iir)) {
+	case UART_IIR_NO_INT:
+		return 0;
+
+	case UART_IIR_BUSY:
+		/* Clear the USR */
+		serial_port_in(p, d->pdata->usr_reg);
+
+		return 1;
+	}
+
+	guard(uart_port_lock_irqsave)(p);
 
 	/*
 	 * There are ways to get Designware-based UARTs into a state where
@@ -325,20 +342,15 @@ static int dw8250_handle_irq(struct uart_port *p)
 	 * so we limit the workaround only to non-DMA mode.
 	 */
 	if (!up->dma && rx_timeout) {
-		uart_port_lock_irqsave(p, &flags);
 		status = serial_lsr_in(up);
 
 		if (!(status & (UART_LSR_DR | UART_LSR_BI)))
 			serial_port_in(p, UART_RX);
-
-		uart_port_unlock_irqrestore(p, flags);
 	}
 
 	/* Manually stop the Rx DMA transfer when acting as flow controller */
 	if (quirks & DW_UART_QUIRK_IS_DMA_FC && up->dma && up->dma->rx_running && rx_timeout) {
-		uart_port_lock_irqsave(p, &flags);
 		status = serial_lsr_in(up);
-		uart_port_unlock_irqrestore(p, flags);
 
 		if (status & (UART_LSR_DR | UART_LSR_BI)) {
 			dw8250_writel_ext(p, RZN1_UART_RDMACR, 0);
@@ -346,17 +358,9 @@ static int dw8250_handle_irq(struct uart_port *p)
 		}
 	}
 
-	if (serial8250_handle_irq(p, iir))
-		return 1;
-
-	if ((iir & UART_IIR_BUSY) == UART_IIR_BUSY) {
-		/* Clear the USR */
-		serial_port_in(p, d->pdata->usr_reg);
+	serial8250_handle_irq_locked(p, iir);
 
-		return 1;
-	}
-
-	return 0;
+	return 1;
 }
 
 static void dw8250_clk_work_cb(struct work_struct *work)
@@ -858,6 +862,7 @@ static struct platform_driver dw8250_platform_driver = {
 
 module_platform_driver(dw8250_platform_driver);
 
+MODULE_IMPORT_NS("SERIAL_8250");
 MODULE_AUTHOR("Jamie Iles");
 MODULE_LICENSE("GPL");
 MODULE_DESCRIPTION("Synopsys DesignWare 8250 serial port driver");
-- 
2.39.5

Re: [PATCH v2 4/7] serial: 8250_dw: Rework dw8250_handle_irq() locking and IIR handling

[Andy Shevchenko]

On Wed, Jan 28, 2026 at 12:52:58PM +0200, Ilpo Järvinen wrote:
> dw8250_handle_irq() takes port's lock multiple times with no good
> reason to release it in between and calls serial8250_handle_irq()
> that also takes port's lock.
> 
> Take port's lock only once in dw8250_handle_irq() and use
> serial8250_handle_irq_locked() to avoid releasing port's lock in
> between.
> 
> As IIR_NO_INT check in serial8250_handle_irq() was outside of port's
> lock, it has to be done already in dw8250_handle_irq().
> 
> DW UART can, in addition to IIR_NO_INT, report BUSY_DETECT (0x7) which
> collided with the IIR_NO_INT (0x1) check in serial8250_handle_irq()
> (because & is used instead of ==) meaning that no other work is done by
> serial8250_handle_irq() during an BUSY_DETECT interrupt.
> 
> This allows reorganizing code in dw8250_handle_irq() to do both
> IIR_NO_INT and BUSY_DETECT handling right at the start simplifying
> the logic.

Reviewed-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>

...

>  	if (!up->dma && rx_timeout) {
>  		status = serial_lsr_in(up);
>  
>  		if (!(status & (UART_LSR_DR | UART_LSR_BI)))
>  			serial_port_in(p, UART_RX);
>  	}
>  
>  	/* Manually stop the Rx DMA transfer when acting as flow controller */
>  	if (quirks & DW_UART_QUIRK_IS_DMA_FC && up->dma && up->dma->rx_running && rx_timeout) {
>  		status = serial_lsr_in(up);
>  
>  		if (status & (UART_LSR_DR | UART_LSR_BI)) {
>  			dw8250_writel_ext(p, RZN1_UART_RDMACR, 0);
			...
>  		}
>  	}

Looks like now (perhaps in a separate change) this may be refactored even more.

	if (rx_timeout) {
		status = serial_lsr_in(up);

// Although not sure about moving the above read out from the specific conditions.

		if (up->dma && (status & (UART_LSR_DR | UART_LSR_BI))) {
			/* Manually stop the Rx DMA transfer when acting as flow controller */
			if (quirks & DW_UART_QUIRK_IS_DMA_FC && up->dma->rx_running) {
				dw8250_writel_ext(p, RZN1_UART_RDMACR, 0);
				dw8250_writel_ext(p, DW_UART_DMASA, 1);
		} else if (!up->dma && !(status & (UART_LSR_DR | UART_LSR_BI)))
			/* ... */
			serial_port_in(p, UART_RX);
		}
	}

-- 
With Best Regards,
Andy Shevchenko

Re: [PATCH v2 4/7] serial: 8250_dw: Rework dw8250_handle_irq() locking and IIR handling

[Ilpo Järvinen]

[-- Attachment #1: Type: text/plain, Size: 2410 bytes --]

On Wed, 28 Jan 2026, Andy Shevchenko wrote:

> On Wed, Jan 28, 2026 at 12:52:58PM +0200, Ilpo Järvinen wrote:
> > dw8250_handle_irq() takes port's lock multiple times with no good
> > reason to release it in between and calls serial8250_handle_irq()
> > that also takes port's lock.
> > 
> > Take port's lock only once in dw8250_handle_irq() and use
> > serial8250_handle_irq_locked() to avoid releasing port's lock in
> > between.
> > 
> > As IIR_NO_INT check in serial8250_handle_irq() was outside of port's
> > lock, it has to be done already in dw8250_handle_irq().
> > 
> > DW UART can, in addition to IIR_NO_INT, report BUSY_DETECT (0x7) which
> > collided with the IIR_NO_INT (0x1) check in serial8250_handle_irq()
> > (because & is used instead of ==) meaning that no other work is done by
> > serial8250_handle_irq() during an BUSY_DETECT interrupt.
> > 
> > This allows reorganizing code in dw8250_handle_irq() to do both
> > IIR_NO_INT and BUSY_DETECT handling right at the start simplifying
> > the logic.
> 
> Reviewed-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
> 
> ...
> 
> >  	if (!up->dma && rx_timeout) {
> >  		status = serial_lsr_in(up);
> >  
> >  		if (!(status & (UART_LSR_DR | UART_LSR_BI)))
> >  			serial_port_in(p, UART_RX);
> >  	}
> >  
> >  	/* Manually stop the Rx DMA transfer when acting as flow controller */
> >  	if (quirks & DW_UART_QUIRK_IS_DMA_FC && up->dma && up->dma->rx_running && rx_timeout) {
> >  		status = serial_lsr_in(up);
> >  
> >  		if (status & (UART_LSR_DR | UART_LSR_BI)) {
> >  			dw8250_writel_ext(p, RZN1_UART_RDMACR, 0);
> 			...
> >  		}
> >  	}
> 
> Looks like now (perhaps in a separate change) this may be refactored even more.

Definitely, as was noted in the coverletter. ;-)

--
 i.

> 
> 	if (rx_timeout) {
> 		status = serial_lsr_in(up);
> 
> // Although not sure about moving the above read out from the specific conditions.
> 
> 		if (up->dma && (status & (UART_LSR_DR | UART_LSR_BI))) {
> 			/* Manually stop the Rx DMA transfer when acting as flow controller */
> 			if (quirks & DW_UART_QUIRK_IS_DMA_FC && up->dma->rx_running) {
> 				dw8250_writel_ext(p, RZN1_UART_RDMACR, 0);
> 				dw8250_writel_ext(p, DW_UART_DMASA, 1);
> 		} else if (!up->dma && !(status & (UART_LSR_DR | UART_LSR_BI)))
> 			/* ... */
> 			serial_port_in(p, UART_RX);
> 		}
> 	}
> 
> 

[PATCH v2 5/7] serial: 8250_dw: Rework IIR_NO_INT handling to stop interrupt storm

[Ilpo Järvinen]

INTC10EE UART can end up into an interrupt storm where it reports
IIR_NO_INT (0x1). If the storm happens during active UART operation, it
is promptly stopped by IIR value change due to Rx or Tx events.
However, when there is no activity, either due to idle serial line or
due to specific circumstances such as during shutdown that writes
IER=0, there is nothing to stop the storm.

During shutdown the storm is particularly problematic because
serial8250_do_shutdown() calls synchronize_irq() that will hang in
waiting for the storm to finish which never happens.

This problem can also result in triggering a warning:

  irq 45: nobody cared (try booting with the "irqpoll" option)
  [...snip...]
  handlers:
    serial8250_interrupt
  Disabling IRQ #45

Normal means to reset interrupt status by reading LSR, MSR, USR, or RX
register do not result in the UART deasserting the IRQ.

Add a quirk to INTC10EE UARTs to enable Tx interrupts if UART's Tx is
currently empty and inactive. Rework IIR_NO_INT to keep track of the
number of consecutive IIR_NO_INT, and on fourth one perform the quirk.
Enabling Tx interrupts should change IIR value from IIR_NO_INT to
IIR_THRI which has been observed to stop the storm.

Fixes: e92fad024929 ("serial: 8250_dw: Add ACPI ID for Granite Rapids-D UART")
Cc: stable@vger.kernel.org
Reported-by: "Bandal, Shankar" <shankar.bandal@intel.com>
Tested-by: "Bandal, Shankar" <shankar.bandal@intel.com>
Tested-by: "Murthy, Shanth" <shanth.murthy@intel.com>
Signed-off-by: Ilpo Järvinen <ilpo.jarvinen@linux.intel.com>
---
 drivers/tty/serial/8250/8250_dw.c | 67 +++++++++++++++++++++++++++++--
 1 file changed, 63 insertions(+), 4 deletions(-)

diff --git a/drivers/tty/serial/8250/8250_dw.c b/drivers/tty/serial/8250/8250_dw.c
index 964750d17186..edae359b1c3f 100644
--- a/drivers/tty/serial/8250/8250_dw.c
+++ b/drivers/tty/serial/8250/8250_dw.c
@@ -61,6 +61,13 @@
 #define DW_UART_QUIRK_IS_DMA_FC		BIT(3)
 #define DW_UART_QUIRK_APMC0D08		BIT(4)
 #define DW_UART_QUIRK_CPR_VALUE		BIT(5)
+#define DW_UART_QUIRK_IER_KICK		BIT(6)
+
+/*
+ * Number of consecutive IIR_NO_INT interrupts required to trigger interrupt
+ * storm prevention code.
+ */
+#define DW_UART_QUIRK_IER_KICK_THRES	4
 
 struct dw8250_platform_data {
 	u8 usr_reg;
@@ -82,6 +89,8 @@ struct dw8250_data {
 
 	unsigned int		skip_autocfg:1;
 	unsigned int		uart_16550_compatible:1;
+
+	u8			no_int_count;
 };
 
 static inline struct dw8250_data *to_dw8250_data(struct dw8250_port_data *data)
@@ -308,6 +317,29 @@ static u32 dw8250_serial_in32be(struct uart_port *p, unsigned int offset)
        return dw8250_modify_msr(p, offset, value);
 }
 
+/*
+ * INTC10EE UART can IRQ storm while reporting IIR_NO_INT. Inducing IIR value
+ * change has been observed to break the storm.
+ *
+ * If Tx is empty (THRE asserted), we use here IER_THRI to cause IIR_NO_INT ->
+ * IIR_THRI transition.
+ */
+static void dw8250_quirk_ier_kick(struct uart_port *p)
+{
+	struct uart_8250_port *up = up_to_u8250p(p);
+	u32 lsr;
+
+	if (up->ier & UART_IER_THRI)
+		return;
+
+	lsr = serial_lsr_in(up);
+	if (!(lsr & UART_LSR_THRE))
+		return;
+
+	serial_port_out(p, UART_IER, up->ier | UART_IER_THRI);
+	serial_port_in(p, UART_LCR);		/* safe, no side-effects */
+	serial_port_out(p, UART_IER, up->ier);
+}
 
 static int dw8250_handle_irq(struct uart_port *p)
 {
@@ -318,18 +350,30 @@ static int dw8250_handle_irq(struct uart_port *p)
 	unsigned int quirks = d->pdata->quirks;
 	unsigned int status;
 
+	guard(uart_port_lock_irqsave)(p);
+
 	switch (FIELD_GET(DW_UART_IIR_IID, iir)) {
 	case UART_IIR_NO_INT:
+		if (d->uart_16550_compatible || up->dma)
+			return 0;
+
+		if (quirks & DW_UART_QUIRK_IER_KICK &&
+		    d->no_int_count == (DW_UART_QUIRK_IER_KICK_THRES - 1))
+			dw8250_quirk_ier_kick(p);
+		d->no_int_count = (d->no_int_count + 1) % DW_UART_QUIRK_IER_KICK_THRES;
+
 		return 0;
 
 	case UART_IIR_BUSY:
 		/* Clear the USR */
 		serial_port_in(p, d->pdata->usr_reg);
 
+		d->no_int_count = 0;
+
 		return 1;
 	}
 
-	guard(uart_port_lock_irqsave)(p);
+	d->no_int_count = 0;
 
 	/*
 	 * There are ways to get Designware-based UARTs into a state where
@@ -562,6 +606,14 @@ static void dw8250_reset_control_assert(void *data)
 	reset_control_assert(data);
 }
 
+static void dw8250_shutdown(struct uart_port *port)
+{
+	struct dw8250_data *d = to_dw8250_data(port->private_data);
+
+	serial8250_do_shutdown(port);
+	d->no_int_count = 0;
+}
+
 static int dw8250_probe(struct platform_device *pdev)
 {
 	struct uart_8250_port uart = {}, *up = &uart;
@@ -685,10 +737,12 @@ static int dw8250_probe(struct platform_device *pdev)
 		dw8250_quirks(p, data);
 
 	/* If the Busy Functionality is not implemented, don't handle it */
-	if (data->uart_16550_compatible)
+	if (data->uart_16550_compatible) {
 		p->handle_irq = NULL;
-	else if (data->pdata)
+	} else if (data->pdata) {
 		p->handle_irq = dw8250_handle_irq;
+		p->shutdown = dw8250_shutdown;
+	}
 
 	dw8250_setup_dma_filter(p, data);
 
@@ -815,6 +869,11 @@ static const struct dw8250_platform_data dw8250_skip_set_rate_data = {
 	.quirks = DW_UART_QUIRK_SKIP_SET_RATE,
 };
 
+static const struct dw8250_platform_data dw8250_intc10ee = {
+	.usr_reg = DW_UART_USR,
+	.quirks = DW_UART_QUIRK_IER_KICK,
+};
+
 static const struct of_device_id dw8250_of_match[] = {
 	{ .compatible = "snps,dw-apb-uart", .data = &dw8250_dw_apb },
 	{ .compatible = "cavium,octeon-3860-uart", .data = &dw8250_octeon_3860_data },
@@ -844,7 +903,7 @@ static const struct acpi_device_id dw8250_acpi_match[] = {
 	{ "INT33C5", (kernel_ulong_t)&dw8250_dw_apb },
 	{ "INT3434", (kernel_ulong_t)&dw8250_dw_apb },
 	{ "INT3435", (kernel_ulong_t)&dw8250_dw_apb },
-	{ "INTC10EE", (kernel_ulong_t)&dw8250_dw_apb },
+	{ "INTC10EE", (kernel_ulong_t)&dw8250_intc10ee },
 	{ },
 };
 MODULE_DEVICE_TABLE(acpi, dw8250_acpi_match);
-- 
2.39.5

Re: [PATCH v2 5/7] serial: 8250_dw: Rework IIR_NO_INT handling to stop interrupt storm

[Andy Shevchenko]

On Wed, Jan 28, 2026 at 12:52:59PM +0200, Ilpo Järvinen wrote:
> INTC10EE UART can end up into an interrupt storm where it reports
> IIR_NO_INT (0x1). If the storm happens during active UART operation, it
> is promptly stopped by IIR value change due to Rx or Tx events.
> However, when there is no activity, either due to idle serial line or
> due to specific circumstances such as during shutdown that writes
> IER=0, there is nothing to stop the storm.
> 
> During shutdown the storm is particularly problematic because
> serial8250_do_shutdown() calls synchronize_irq() that will hang in
> waiting for the storm to finish which never happens.
> 
> This problem can also result in triggering a warning:
> 
>   irq 45: nobody cared (try booting with the "irqpoll" option)
>   [...snip...]
>   handlers:
>     serial8250_interrupt
>   Disabling IRQ #45
> 
> Normal means to reset interrupt status by reading LSR, MSR, USR, or RX
> register do not result in the UART deasserting the IRQ.
> 
> Add a quirk to INTC10EE UARTs to enable Tx interrupts if UART's Tx is
> currently empty and inactive. Rework IIR_NO_INT to keep track of the
> number of consecutive IIR_NO_INT, and on fourth one perform the quirk.
> Enabling Tx interrupts should change IIR value from IIR_NO_INT to
> IIR_THRI which has been observed to stop the storm.

Reviewed-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>

...

> +/*
> + * Number of consecutive IIR_NO_INT interrupts required to trigger interrupt
> + * storm prevention code.
> + */
> +#define DW_UART_QUIRK_IER_KICK_THRES	4

Actually if we...

>  struct dw8250_platform_data {

>  	u8 usr_reg;

>  	unsigned int		skip_autocfg:1;
>  	unsigned int		uart_16550_compatible:1;
> +
> +	u8			no_int_count;

...define this as

	unsigned int		no_int_count:2;

there will be no need to have a modulo at all. But I agree it might look a bit
weird.

>  };

-- 
With Best Regards,
Andy Shevchenko

[PATCH v2 6/7] serial: 8250: Add late synchronize_irq() to shutdown to handle DW UART BUSY

[Ilpo Järvinen]

When DW UART is !uart_16550_compatible, it can indicate BUSY at any
point (when under constant Rx pressure) unless a complex sequence of
steps is performed. Any LCR write can run a foul with the condition
that prevents writing LCR while the UART is BUSY, which triggers
BUSY_DETECT interrupt that seems unmaskable using IER bits.

Normal flow is that dw8250_handle_irq() handles BUSY_DETECT condition
by reading USR register. This BUSY feature, however, breaks the
assumptions made in serial8250_do_shutdown(), which runs
synchronize_irq() after clearing IER and assumes no interrupts can
occur after that point but then proceeds to update LCR, which on DW
UART can trigger an interrupt.

If serial8250_do_shutdown() releases the interrupt handler before the
handler has run and processed the BUSY_DETECT condition by read the USR
register, the IRQ is not deasserted resulting in interrupt storm that
triggers "irq x: nobody cared" warning leading to disabling the IRQ.

Add late synchronize_irq() into serial8250_do_shutdown() to ensure
BUSY_DETECT from DW UART is handled before port's interrupt handler is
released. Alternative would be to add DW UART specific shutdown
function but it would mostly duplicate the generic code and the extra
synchronize_irq() seems pretty harmless in serial8250_do_shutdown().

Fixes: 7d4008ebb1c9 ("tty: add a DesignWare 8250 driver")
Cc: stable@vger.kernel.org
Reported-by: "Bandal, Shankar" <shankar.bandal@intel.com>
Tested-by: "Bandal, Shankar" <shankar.bandal@intel.com>
Tested-by: "Murthy, Shanth" <shanth.murthy@intel.com>
Signed-off-by: Ilpo Järvinen <ilpo.jarvinen@linux.intel.com>
---
 drivers/tty/serial/8250/8250_port.c | 6 ++++++
 1 file changed, 6 insertions(+)

diff --git a/drivers/tty/serial/8250/8250_port.c b/drivers/tty/serial/8250/8250_port.c
index bc223eb1f474..fb0b8397cd4b 100644
--- a/drivers/tty/serial/8250/8250_port.c
+++ b/drivers/tty/serial/8250/8250_port.c
@@ -2401,6 +2401,12 @@ void serial8250_do_shutdown(struct uart_port *port)
 	 * the IRQ chain.
 	 */
 	serial_port_in(port, UART_RX);
+	/*
+	 * LCR writes on DW UART can trigger late (unmaskable) IRQs.
+	 * Handle them before releasing the handler.
+	 */
+	synchronize_irq(port->irq);
+
 	serial8250_rpm_put(up);
 
 	up->ops->release_irq(up);
-- 
2.39.5

Re: [PATCH v2 6/7] serial: 8250: Add late synchronize_irq() to shutdown to handle DW UART BUSY

[Andy Shevchenko]

On Wed, Jan 28, 2026 at 12:53:00PM +0200, Ilpo Järvinen wrote:
> When DW UART is !uart_16550_compatible, it can indicate BUSY at any
> point (when under constant Rx pressure) unless a complex sequence of
> steps is performed. Any LCR write can run a foul with the condition
> that prevents writing LCR while the UART is BUSY, which triggers
> BUSY_DETECT interrupt that seems unmaskable using IER bits.
> 
> Normal flow is that dw8250_handle_irq() handles BUSY_DETECT condition
> by reading USR register. This BUSY feature, however, breaks the
> assumptions made in serial8250_do_shutdown(), which runs
> synchronize_irq() after clearing IER and assumes no interrupts can
> occur after that point but then proceeds to update LCR, which on DW
> UART can trigger an interrupt.
> 
> If serial8250_do_shutdown() releases the interrupt handler before the
> handler has run and processed the BUSY_DETECT condition by read the USR
> register, the IRQ is not deasserted resulting in interrupt storm that
> triggers "irq x: nobody cared" warning leading to disabling the IRQ.
> 
> Add late synchronize_irq() into serial8250_do_shutdown() to ensure
> BUSY_DETECT from DW UART is handled before port's interrupt handler is
> released. Alternative would be to add DW UART specific shutdown
> function but it would mostly duplicate the generic code and the extra
> synchronize_irq() seems pretty harmless in serial8250_do_shutdown().

Dunno if the triggered interrupt may lead to a new DMA transfers (since
this is generic 8520 code...) in some cases. Anyway I've just sent a patch
that is Cc'ed to you to prevent that from happening. Not sure if it needs
to be incorporated into your series or should have a Fixes tag.

Reviewed-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>

-- 
With Best Regards,
Andy Shevchenko

Re: [PATCH v2 6/7] serial: 8250: Add late synchronize_irq() to shutdown to handle DW UART BUSY

[Ilpo Järvinen]

[-- Attachment #1: Type: text/plain, Size: 2062 bytes --]

On Wed, 28 Jan 2026, Andy Shevchenko wrote:

> On Wed, Jan 28, 2026 at 12:53:00PM +0200, Ilpo Järvinen wrote:
> > When DW UART is !uart_16550_compatible, it can indicate BUSY at any
> > point (when under constant Rx pressure) unless a complex sequence of
> > steps is performed. Any LCR write can run a foul with the condition
> > that prevents writing LCR while the UART is BUSY, which triggers
> > BUSY_DETECT interrupt that seems unmaskable using IER bits.
> > 
> > Normal flow is that dw8250_handle_irq() handles BUSY_DETECT condition
> > by reading USR register. This BUSY feature, however, breaks the
> > assumptions made in serial8250_do_shutdown(), which runs
> > synchronize_irq() after clearing IER and assumes no interrupts can
> > occur after that point but then proceeds to update LCR, which on DW
> > UART can trigger an interrupt.
> > 
> > If serial8250_do_shutdown() releases the interrupt handler before the
> > handler has run and processed the BUSY_DETECT condition by read the USR
> > register, the IRQ is not deasserted resulting in interrupt storm that
> > triggers "irq x: nobody cared" warning leading to disabling the IRQ.
> > 
> > Add late synchronize_irq() into serial8250_do_shutdown() to ensure
> > BUSY_DETECT from DW UART is handled before port's interrupt handler is
> > released. Alternative would be to add DW UART specific shutdown
> > function but it would mostly duplicate the generic code and the extra
> > synchronize_irq() seems pretty harmless in serial8250_do_shutdown().
> 
> Dunno if the triggered interrupt may lead to a new DMA transfers (since
> this is generic 8520 code...) in some cases.

If BUSY_DETECT interrupt fires (like I'd expect), the handler won't start 
anything but exits early from dw8250_handle_irq().

> Anyway I've just sent a patch
> that is Cc'ed to you to prevent that from happening. Not sure if it needs
> to be incorporated into your series or should have a Fixes tag.
> 
> Reviewed-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>


-- 
 i.

[PATCH v2 7/7] serial: 8250_dw: Ensure BUSY is deasserted

[Ilpo Järvinen]

DW UART cannot write to LCR, DLL, and DLH while BUSY is asserted.
Existance of BUSY depends on uart_16550_compatible, if UART HW is
configured with it those registers can always be written.

There currently is dw8250_force_idle() which attempts to achieve
non-BUSY state by disabling FIFO, however, the solution is unreliable
when Rx keeps getting more and more characters.

Create a sequence of operations that ensures UART cannot keep BUSY
asserted indefinitely. The new sequence relies on enabling loopback mode
temporarily to prevent incoming Rx characters keeping UART BUSY.

Ensure no Tx in ongoing while the UART is switches into the loopback
mode (requires exporting serial8250_fifo_wait_for_lsr_thre() and adding
DMA Tx pause/resume functions).

According to tests performed by Adriana Nicolae <adriana@arista.com>,
simply disabling FIFO or clearing FIFOs only once does not always
ensure BUSY is deasserted but up to two tries may be needed. This could
be related to ongoing Rx of a character (a guess, not known for sure).
Therefore, retry FIFO clearing a few times (retry limit 4 is arbitrary
number but using, e.g., p->fifosize seems overly large). Tests
performed by others did not exhibit similar challenge but it does not
seem harmful to leave the FIFO clearing loop in place for all DW UARTs
with BUSY functionality.

Use the new dw8250_idle_enter/exit() to do divisor writes and LCR
writes. In case of plain LCR writes, opportunistically try to update
LCR first and only invoke dw8250_idle_enter() if the write did not
succeed (it has been observed that in practice most LCR writes do
succeed without complications).

This issue was first reported by qianfan Zhao who put lots of debugging
effort into understanding the solution space.

Fixes: c49436b657d0 ("serial: 8250_dw: Improve unwritable LCR workaround")
Fixes: 7d4008ebb1c9 ("tty: add a DesignWare 8250 driver")
Cc: <stable@vger.kernel.org>
Reported-by: qianfan Zhao <qianfanguijin@163.com>
Link: https://lore.kernel.org/linux-serial/289bb78a-7509-1c5c-2923-a04ed3b6487d@163.com/
Reported-by: Adriana Nicolae <adriana@arista.com>
Link: https://lore.kernel.org/linux-serial/20250819182322.3451959-1-adriana@arista.com/
Reported-by: "Bandal, Shankar" <shankar.bandal@intel.com>
Tested-by: "Bandal, Shankar" <shankar.bandal@intel.com>
Tested-by: "Murthy, Shanth" <shanth.murthy@intel.com>
Signed-off-by: Ilpo Järvinen <ilpo.jarvinen@linux.intel.com>
---
 drivers/tty/serial/8250/8250.h      |  25 ++++
 drivers/tty/serial/8250/8250_dw.c   | 176 ++++++++++++++++++++--------
 drivers/tty/serial/8250/8250_port.c |  28 +++--
 3 files changed, 170 insertions(+), 59 deletions(-)

diff --git a/drivers/tty/serial/8250/8250.h b/drivers/tty/serial/8250/8250.h
index 8caecfc85d93..77fe0588fd6b 100644
--- a/drivers/tty/serial/8250/8250.h
+++ b/drivers/tty/serial/8250/8250.h
@@ -175,7 +175,9 @@ static unsigned int __maybe_unused serial_icr_read(struct uart_8250_port *up,
 	return value;
 }
 
+void serial8250_clear_fifos(struct uart_8250_port *p);
 void serial8250_clear_and_reinit_fifos(struct uart_8250_port *p);
+void serial8250_fifo_wait_for_lsr_thre(struct uart_8250_port *up, unsigned int count);
 
 void serial8250_rpm_get(struct uart_8250_port *p);
 void serial8250_rpm_put(struct uart_8250_port *p);
@@ -400,6 +402,26 @@ static inline bool serial8250_tx_dma_running(struct uart_8250_port *p)
 
 	return dma && dma->tx_running;
 }
+
+static inline void serial8250_tx_dma_pause(struct uart_8250_port *p)
+{
+	struct uart_8250_dma *dma = p->dma;
+
+	if (!dma->tx_running)
+		return;
+
+	dmaengine_pause(dma->txchan);
+}
+
+static inline void serial8250_tx_dma_resume(struct uart_8250_port *p)
+{
+	struct uart_8250_dma *dma = p->dma;
+
+	if (!dma->tx_running)
+		return;
+
+	dmaengine_resume(dma->txchan);
+}
 #else
 static inline int serial8250_tx_dma(struct uart_8250_port *p)
 {
@@ -421,6 +443,9 @@ static inline bool serial8250_tx_dma_running(struct uart_8250_port *p)
 {
 	return false;
 }
+
+static inline void serial8250_tx_dma_pause(struct uart_8250_port *p) { }
+static inline void serial8250_tx_dma_resume(struct uart_8250_port *p) { }
 #endif
 
 static inline int ns16550a_goto_highspeed(struct uart_8250_port *up)
diff --git a/drivers/tty/serial/8250/8250_dw.c b/drivers/tty/serial/8250/8250_dw.c
index edae359b1c3f..ca527be96179 100644
--- a/drivers/tty/serial/8250/8250_dw.c
+++ b/drivers/tty/serial/8250/8250_dw.c
@@ -16,6 +16,7 @@
 #include <linux/delay.h>
 #include <linux/device.h>
 #include <linux/io.h>
+#include <linux/lockdep.h>
 #include <linux/mod_devicetable.h>
 #include <linux/module.h>
 #include <linux/notifier.h>
@@ -47,6 +48,8 @@
 
 #define DW_UART_MCR_SIRE		BIT(6)
 
+#define DW_UART_USR_BUSY		BIT(0)
+
 /* Renesas specific register fields */
 #define RZN1_UART_xDMACR_DMA_EN		BIT(0)
 #define RZN1_UART_xDMACR_1_WORD_BURST	(0 << 1)
@@ -89,6 +92,7 @@ struct dw8250_data {
 
 	unsigned int		skip_autocfg:1;
 	unsigned int		uart_16550_compatible:1;
+	unsigned int		in_idle:1;
 
 	u8			no_int_count;
 };
@@ -122,77 +126,155 @@ static inline u32 dw8250_modify_msr(struct uart_port *p, unsigned int offset, u3
 }
 
 /*
- * This function is being called as part of the uart_port::serial_out()
- * routine. Hence, it must not call serial_port_out() or serial_out()
- * against the modified registers here, i.e. LCR.
+ * Ensure BUSY is not asserted. If DW UART is configured with
+ * !uart_16550_compatible, the writes to LCR, DLL, and DLH fail while
+ * BUSY is asserted.
+ *
+ * Context: port's lock must be held
  */
-static void dw8250_force_idle(struct uart_port *p)
+static int dw8250_idle_enter(struct uart_port *p)
 {
+	struct dw8250_data *d = to_dw8250_data(p->private_data);
 	struct uart_8250_port *up = up_to_u8250p(p);
-	unsigned int lsr;
+	unsigned int usr_reg = DW_UART_USR;
+	int retries;
+	u32 lsr;
 
-	/*
-	 * The following call currently performs serial_out()
-	 * against the FCR register. Because it differs to LCR
-	 * there will be no infinite loop, but if it ever gets
-	 * modified, we might need a new custom version of it
-	 * that avoids infinite recursion.
-	 */
-	serial8250_clear_and_reinit_fifos(up);
+	lockdep_assert_held_once(&p->lock);
+
+	if (d->uart_16550_compatible)
+		return 0;
+
+	if (d->pdata)
+		usr_reg = d->pdata->usr_reg;
+
+	d->in_idle = 1;
+
+	/* Prevent triggering interrupt from RBR filling */
+	serial_port_out(p, UART_IER, 0);
+
+	if (up->dma) {
+		serial8250_rx_dma_flush(up);
+		if (serial8250_tx_dma_running(up))
+			serial8250_tx_dma_pause(up);
+	}
 
 	/*
-	 * With PSLVERR_RESP_EN parameter set to 1, the device generates an
-	 * error response when an attempt to read an empty RBR with FIFO
-	 * enabled.
+	 * Wait until Tx becomes empty + one extra frame time to ensure all bits
+	 * have been sent on the wire.
+	 *
+	 * FIXME: frame_time delay is too long with very low baudrates.
 	 */
-	if (up->fcr & UART_FCR_ENABLE_FIFO) {
-		lsr = serial_port_in(p, UART_LSR);
-		if (!(lsr & UART_LSR_DR))
-			return;
+	serial8250_fifo_wait_for_lsr_thre(up, p->fifosize);
+	ndelay(p->frame_time);
+
+	serial_port_out(p, UART_MCR, up->mcr | UART_MCR_LOOP);
+
+	retries = 4;	/* Arbitrary limit, 2 was always enough in tests */
+	do {
+		serial8250_clear_fifos(up);
+		if (!(serial_port_in(p, usr_reg) & DW_UART_USR_BUSY))
+			break;
+		/* FIXME: frame_time delay is too long with very low baudrates. */
+		ndelay(p->frame_time);
+	} while (--retries);
+
+	lsr = serial_lsr_in(up);
+	if (lsr & UART_LSR_DR) {
+		serial_port_in(p, UART_RX);
+		up->lsr_saved_flags = 0;
 	}
 
-	serial_port_in(p, UART_RX);
+	/* Now guaranteed to have BUSY deasserted? Just sanity check */
+	if (serial_port_in(p, usr_reg) & DW_UART_USR_BUSY)
+		return -EBUSY;
+
+	return 0;
+}
+
+static void dw8250_idle_exit(struct uart_port *p)
+{
+	struct dw8250_data *d = to_dw8250_data(p->private_data);
+	struct uart_8250_port *up = up_to_u8250p(p);
+
+	if (d->uart_16550_compatible)
+		return;
+
+	if (up->capabilities & UART_CAP_FIFO)
+		serial_port_out(p, UART_FCR, up->fcr);
+	serial_port_out(p, UART_MCR, up->mcr);
+	serial_port_out(p, UART_IER, up->ier);
+
+	/* DMA Rx is restarted by IRQ handler as needed. */
+	if (up->dma)
+		serial8250_tx_dma_resume(up);
+
+	d->in_idle = 0;
+}
+
+static void dw8250_set_divisor(struct uart_port *p, unsigned int baud,
+			       unsigned int quot, unsigned int quot_frac)
+{
+	struct uart_8250_port *up = up_to_u8250p(p);
+	int ret;
+
+	ret = dw8250_idle_enter(p);
+	if (ret < 0)
+		goto idle_failed;
+
+	serial_port_out(p, UART_LCR, up->lcr | UART_LCR_DLAB);
+	if (!(serial_port_in(p, UART_LCR) & UART_LCR_DLAB))
+		goto idle_failed;
+
+	serial_dl_write(up, quot);
+	serial_port_out(p, UART_LCR, up->lcr);
+
+idle_failed:
+	dw8250_idle_exit(p);
 }
 
 /*
  * This function is being called as part of the uart_port::serial_out()
- * routine. Hence, it must not call serial_port_out() or serial_out()
- * against the modified registers here, i.e. LCR.
+ * routine. Hence, special care must be taken when serial_port_out() or
+ * serial_out() against the modified registers here, i.e. LCR (d->in_idle is
+ * used to break recursion loop).
  */
 static void dw8250_check_lcr(struct uart_port *p, unsigned int offset, u32 value)
 {
 	struct dw8250_data *d = to_dw8250_data(p->private_data);
-	void __iomem *addr = p->membase + (offset << p->regshift);
-	int tries = 1000;
+	u32 lcr;
+	int ret;
 
 	if (offset != UART_LCR || d->uart_16550_compatible)
 		return;
 
-	/* Make sure LCR write wasn't ignored */
-	while (tries--) {
-		u32 lcr = serial_port_in(p, offset);
+	lcr = serial_port_in(p, UART_LCR);
 
-		if ((value & ~UART_LCR_SPAR) == (lcr & ~UART_LCR_SPAR))
-			return;
+	/* Make sure LCR write wasn't ignored */
+	if ((value & ~UART_LCR_SPAR) == (lcr & ~UART_LCR_SPAR))
+		return;
 
-		dw8250_force_idle(p);
+	if (d->in_idle) {
+		/*
+		 * FIXME: this deadlocks if port->lock is already held
+		 * dev_err(p->dev, "Couldn't set LCR to %d\n", value);
+		 */
+		return;
+	}
 
-#ifdef CONFIG_64BIT
-		if (p->type == PORT_OCTEON)
-			__raw_writeq(value & 0xff, addr);
-		else
-#endif
-		if (p->iotype == UPIO_MEM32)
-			writel(value, addr);
-		else if (p->iotype == UPIO_MEM32BE)
-			iowrite32be(value, addr);
-		else
-			writeb(value, addr);
+	ret = dw8250_idle_enter(p);
+	if (ret < 0) {
+		/*
+		 * FIXME: this deadlocks if port->lock is already held
+		 * dev_err(p->dev, "Couldn't set LCR to %d\n", value);
+		 */
+		goto idle_failed;
 	}
-	/*
-	 * FIXME: this deadlocks if port->lock is already held
-	 * dev_err(p->dev, "Couldn't set LCR to %d\n", value);
-	 */
+
+	serial_port_out(p, UART_LCR, value);
+
+idle_failed:
+	dw8250_idle_exit(p);
 }
 
 /*
@@ -632,8 +714,10 @@ static int dw8250_probe(struct platform_device *pdev)
 	p->type		= PORT_8250;
 	p->flags	= UPF_FIXED_PORT;
 	p->dev		= dev;
+
 	p->set_ldisc	= dw8250_set_ldisc;
 	p->set_termios	= dw8250_set_termios;
+	p->set_divisor	= dw8250_set_divisor;
 
 	data = devm_kzalloc(dev, sizeof(*data), GFP_KERNEL);
 	if (!data)
diff --git a/drivers/tty/serial/8250/8250_port.c b/drivers/tty/serial/8250/8250_port.c
index fb0b8397cd4b..8b31b72e4676 100644
--- a/drivers/tty/serial/8250/8250_port.c
+++ b/drivers/tty/serial/8250/8250_port.c
@@ -489,7 +489,7 @@ serial_port_out_sync(struct uart_port *p, int offset, int value)
 /*
  * FIFO support.
  */
-static void serial8250_clear_fifos(struct uart_8250_port *p)
+void serial8250_clear_fifos(struct uart_8250_port *p)
 {
 	if (p->capabilities & UART_CAP_FIFO) {
 		serial_out(p, UART_FCR, UART_FCR_ENABLE_FIFO);
@@ -498,6 +498,7 @@ static void serial8250_clear_fifos(struct uart_8250_port *p)
 		serial_out(p, UART_FCR, 0);
 	}
 }
+EXPORT_SYMBOL_NS_GPL(serial8250_clear_fifos, "SERIAL_8250");
 
 static enum hrtimer_restart serial8250_em485_handle_start_tx(struct hrtimer *t);
 static enum hrtimer_restart serial8250_em485_handle_stop_tx(struct hrtimer *t);
@@ -3200,6 +3201,17 @@ void serial8250_set_defaults(struct uart_8250_port *up)
 }
 EXPORT_SYMBOL_GPL(serial8250_set_defaults);
 
+void serial8250_fifo_wait_for_lsr_thre(struct uart_8250_port *up, unsigned int count)
+{
+	unsigned int i;
+
+	for (i = 0; i < count; i++) {
+		if (wait_for_lsr(up, UART_LSR_THRE))
+			return;
+	}
+}
+EXPORT_SYMBOL_NS_GPL(serial8250_fifo_wait_for_lsr_thre, "SERIAL_8250");
+
 #ifdef CONFIG_SERIAL_8250_CONSOLE
 
 static void serial8250_console_putchar(struct uart_port *port, unsigned char ch)
@@ -3241,16 +3253,6 @@ static void serial8250_console_restore(struct uart_8250_port *up)
 	serial8250_out_MCR(up, up->mcr | UART_MCR_DTR | UART_MCR_RTS);
 }
 
-static void fifo_wait_for_lsr(struct uart_8250_port *up, unsigned int count)
-{
-	unsigned int i;
-
-	for (i = 0; i < count; i++) {
-		if (wait_for_lsr(up, UART_LSR_THRE))
-			return;
-	}
-}
-
 /*
  * Print a string to the serial port using the device FIFO
  *
@@ -3269,7 +3271,7 @@ static void serial8250_console_fifo_write(struct uart_8250_port *up,
 
 	while (s != end) {
 		/* Allow timeout for each byte of a possibly full FIFO */
-		fifo_wait_for_lsr(up, fifosize);
+		serial8250_fifo_wait_for_lsr_thre(up, fifosize);
 
 		for (i = 0; i < fifosize && s != end; ++i) {
 			if (*s == '\n' && !cr_sent) {
@@ -3287,7 +3289,7 @@ static void serial8250_console_fifo_write(struct uart_8250_port *up,
 	 * Allow timeout for each byte written since the caller will only wait
 	 * for UART_LSR_BOTH_EMPTY using the timeout of a single character
 	 */
-	fifo_wait_for_lsr(up, tx_count);
+	serial8250_fifo_wait_for_lsr_thre(up, tx_count);
 }
 
 /*
-- 
2.39.5

Re: [PATCH v2 7/7] serial: 8250_dw: Ensure BUSY is deasserted

[Andy Shevchenko]

On Wed, Jan 28, 2026 at 12:53:01PM +0200, Ilpo Järvinen wrote:
> DW UART cannot write to LCR, DLL, and DLH while BUSY is asserted.
> Existance of BUSY depends on uart_16550_compatible, if UART HW is
> configured with it those registers can always be written.
> 
> There currently is dw8250_force_idle() which attempts to achieve
> non-BUSY state by disabling FIFO, however, the solution is unreliable
> when Rx keeps getting more and more characters.
> 
> Create a sequence of operations that ensures UART cannot keep BUSY
> asserted indefinitely. The new sequence relies on enabling loopback mode
> temporarily to prevent incoming Rx characters keeping UART BUSY.
> 
> Ensure no Tx in ongoing while the UART is switches into the loopback
> mode (requires exporting serial8250_fifo_wait_for_lsr_thre() and adding
> DMA Tx pause/resume functions).
> 
> According to tests performed by Adriana Nicolae <adriana@arista.com>,
> simply disabling FIFO or clearing FIFOs only once does not always
> ensure BUSY is deasserted but up to two tries may be needed. This could
> be related to ongoing Rx of a character (a guess, not known for sure).
> Therefore, retry FIFO clearing a few times (retry limit 4 is arbitrary
> number but using, e.g., p->fifosize seems overly large). Tests
> performed by others did not exhibit similar challenge but it does not
> seem harmful to leave the FIFO clearing loop in place for all DW UARTs
> with BUSY functionality.
> 
> Use the new dw8250_idle_enter/exit() to do divisor writes and LCR
> writes. In case of plain LCR writes, opportunistically try to update
> LCR first and only invoke dw8250_idle_enter() if the write did not
> succeed (it has been observed that in practice most LCR writes do
> succeed without complications).
> 
> This issue was first reported by qianfan Zhao who put lots of debugging
> effort into understanding the solution space.

Some nit-picks below, otherwise seems good to go
Reviewed-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>

...

> Reported-by: qianfan Zhao <qianfanguijin@163.com>
> Link: https://lore.kernel.org/linux-serial/289bb78a-7509-1c5c-2923-a04ed3b6487d@163.com/
> Reported-by: Adriana Nicolae <adriana@arista.com>
> Link: https://lore.kernel.org/linux-serial/20250819182322.3451959-1-adriana@arista.com/

Shouldn't these Link:s be Closes: tags?

...

> +	struct dw8250_data *d = to_dw8250_data(p->private_data);
>  	struct uart_8250_port *up = up_to_u8250p(p);
> +	unsigned int usr_reg = DW_UART_USR;
> +	int retries;
> +	u32 lsr;


> +	if (d->pdata)
> +		usr_reg = d->pdata->usr_reg;

I would unite this with definition above:

	unsigned int usr_reg = d->pdata ? d->pdata->usr_reg : DW_UART_USR;

...

> +	lsr = serial_lsr_in(up);

> +	if (lsr & UART_LSR_DR) {
> +		serial_port_in(p, UART_RX);
> +		up->lsr_saved_flags = 0;
>  	}

This seems repeating a top of serial8250_read_char(). Perhaps we can do it
in a helper at some point?

...

> +	if (d->in_idle) {

> +		/*
> +		 * FIXME: this deadlocks if port->lock is already held
> +		 * dev_err(p->dev, "Couldn't set LCR to %d\n", value);
> +		 */

Does it make sense to print an error here (assuming it will work with nbcon)?
If so, maybe leave it at the end of the function, after dw8250_idle_exit()
and goto there?

> +		return;
> +	}

-- 
With Best Regards,
Andy Shevchenko

Re: [PATCH v2 7/7] serial: 8250_dw: Ensure BUSY is deasserted

[Ilpo Järvinen]

[-- Attachment #1: Type: text/plain, Size: 4083 bytes --]

On Wed, 28 Jan 2026, Andy Shevchenko wrote:

> On Wed, Jan 28, 2026 at 12:53:01PM +0200, Ilpo Järvinen wrote:
> > DW UART cannot write to LCR, DLL, and DLH while BUSY is asserted.
> > Existance of BUSY depends on uart_16550_compatible, if UART HW is
> > configured with it those registers can always be written.
> > 
> > There currently is dw8250_force_idle() which attempts to achieve
> > non-BUSY state by disabling FIFO, however, the solution is unreliable
> > when Rx keeps getting more and more characters.
> > 
> > Create a sequence of operations that ensures UART cannot keep BUSY
> > asserted indefinitely. The new sequence relies on enabling loopback mode
> > temporarily to prevent incoming Rx characters keeping UART BUSY.
> > 
> > Ensure no Tx in ongoing while the UART is switches into the loopback
> > mode (requires exporting serial8250_fifo_wait_for_lsr_thre() and adding
> > DMA Tx pause/resume functions).
> > 
> > According to tests performed by Adriana Nicolae <adriana@arista.com>,
> > simply disabling FIFO or clearing FIFOs only once does not always
> > ensure BUSY is deasserted but up to two tries may be needed. This could
> > be related to ongoing Rx of a character (a guess, not known for sure).
> > Therefore, retry FIFO clearing a few times (retry limit 4 is arbitrary
> > number but using, e.g., p->fifosize seems overly large). Tests
> > performed by others did not exhibit similar challenge but it does not
> > seem harmful to leave the FIFO clearing loop in place for all DW UARTs
> > with BUSY functionality.
> > 
> > Use the new dw8250_idle_enter/exit() to do divisor writes and LCR
> > writes. In case of plain LCR writes, opportunistically try to update
> > LCR first and only invoke dw8250_idle_enter() if the write did not
> > succeed (it has been observed that in practice most LCR writes do
> > succeed without complications).
> > 
> > This issue was first reported by qianfan Zhao who put lots of debugging
> > effort into understanding the solution space.
> 
> Some nit-picks below, otherwise seems good to go
> Reviewed-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
> 
> ...
> 
> > Reported-by: qianfan Zhao <qianfanguijin@163.com>
> > Link: https://lore.kernel.org/linux-serial/289bb78a-7509-1c5c-2923-a04ed3b6487d@163.com/
> > Reported-by: Adriana Nicolae <adriana@arista.com>
> > Link: https://lore.kernel.org/linux-serial/20250819182322.3451959-1-adriana@arista.com/
> 
> Shouldn't these Link:s be Closes: tags?

To not possibly give wrong signals, until they confirm their cases are 
indeed solved by this patch, I'd like to keep these as Link tag only.

> > +	struct dw8250_data *d = to_dw8250_data(p->private_data);
> >  	struct uart_8250_port *up = up_to_u8250p(p);
> > +	unsigned int usr_reg = DW_UART_USR;
> > +	int retries;
> > +	u32 lsr;
> 
> 
> > +	if (d->pdata)
> > +		usr_reg = d->pdata->usr_reg;
> 
> I would unite this with definition above:
> 
> 	unsigned int usr_reg = d->pdata ? d->pdata->usr_reg : DW_UART_USR;
>
> > +	lsr = serial_lsr_in(up);
> 
> > +	if (lsr & UART_LSR_DR) {
> > +		serial_port_in(p, UART_RX);
> > +		up->lsr_saved_flags = 0;
> >  	}
> 
> This seems repeating a top of serial8250_read_char(). Perhaps we can do it
> in a helper at some point?

I don't see enough similarity as I'd need to deal with lsr_saved_flags 
somehow still here.

> > +	if (d->in_idle) {
> 
> > +		/*
> > +		 * FIXME: this deadlocks if port->lock is already held
> > +		 * dev_err(p->dev, "Couldn't set LCR to %d\n", value);
> > +		 */
> 
> Does it make sense to print an error here (assuming it will work with nbcon)?
> If so, maybe leave it at the end of the function, after dw8250_idle_exit()
> and goto there?

I think the print would be useful. I'll leave the FIXME+commented out 
print into the end of the function in v3.

I also realized that on error, dw8250_idle_enter() should undo what it 
changed, that is, call dw8250_idle_exit() within which will simplify 
caller-side error handling slightly.

-- 
 i.