Serial: Two concurrency fixes & one related cleanup
Subsystems (3)
Ilpo Järvinen
maintainercover Serial: Two concurrency fixes & one related cleanup
Here are two fixes to problem that have been introduced by recent changes. In addition, one cleanup patch to the related code (to keep the fix change strictly a fix).
Ilpo Järvinen (3):
serial: 8250: fix __stop_tx() & DMA Tx restart races
serial: 8250_dw: Take port lock while accessing LSR
serial: 8250_dma: No need for if (dma->tx_err)
drivers/tty/serial/8250/8250_dma.c | 7 +++---- drivers/tty/serial/8250/8250_dw.c | 3 +++ drivers/tty/serial/8250/8250_port.c | 2 +- 3 files changed, 7 insertions(+), 5 deletions(-)
Signature · 1 line
2.30.2
Ilpo Järvinen
maintainerPATCH 1/3 serial: 8250: Fix __stop_tx() & DMA Tx restart races
Commit e8ffbb71f783 ("serial: 8250: use THRE & __stop_tx also with DMA") changed __dma_tx_complete() to enable THRI that is cleared in __stop_tx() once THRE is asserted as UART runs out bits to transmit. It is possible, however, that more data arrives in between in which case serial8250_tx_dma() resumes Tx. THRI is not supposed to be on during DMA Tx because DMA is based on completion handler, therefore THRI must be cleared unconditionally in serial8250_tx_dma().
When Tx is about to start, another race window exists with serial8250_handle_irq() leading to a call into __stop_tx() while the Tx has already been resumed:
__tx_complete():
-> spin_lock(port->lock)
-> dma->tx_running = 0
-> serial8250_set_THRI()
-> spin_unlock(port->lock)
uart_start():
serial8250_handle_irq():
-> spin_lock(port->lock)
-> serial8250_tx_dma():
-> dma->tx_running = 1
-> spin_unlock(port->lock)
-> spin_lock(port->lock)
-> __stop_tx()Close this race by checking !dma->tx_running before calling into __stop_tx().
--- drivers/tty/serial/8250/8250_dma.c | 6 +++--- drivers/tty/serial/8250/8250_port.c | 2 +- 2 files changed, 4 insertions(+), 4 deletions(-)
drivers/tty/serial/8250/8250_dma.c +3 -3
@@ -106,10 +106,10 @@ int serial8250_tx_dma(struct uart_8250_port *p) UART_XMIT_SIZE, DMA_TO_DEVICE);dma_async_issue_pending (dma->txchan); -if (dma->tx_err) { +serial8250_clear_THRI (p); +if (dma->tx_err) dma->tx_err =0 ; -serial8250_clear_THRI (p); - } +return 0 ; err: dma->tx_err =1 ;
drivers/tty/serial/8250/8250_port.c +1 -1
@@ -1936,7 +1936,7 @@ int serial8250_handle_irq(struct uart_port *port, unsigned int iir)if ((status &UART_LSR_THRE ) && (up->ier &UART_IER_THRI )) {if (!up->dma || up->dma->tx_err)serial8250_tx_chars (up); -else +else if (!up->dma->tx_running)__stop_tx (up); }
Signature · 1 line
2.30.2
Ilpo Järvinen
maintainerPATCH 2/3 serial: 8250_dw: Take port lock while accessing LSR
In reply to Ilpo JärvinenAccessing LSR requires port lock because it mutates lsr_saved_flags in serial_lsr_in().
--- drivers/tty/serial/8250/8250_dw.c | 3 +++ 1 file changed, 3 insertions(+)
drivers/tty/serial/8250/8250_dw.c +3 -0
@@ -266,7 +266,10 @@ static int dw8250_handle_irq(struct uart_port *p)/* 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) { +spin_lock_irqsave (&p->lock, flags); status =serial_lsr_in (up); +spin_unlock_irqrestore (&p->lock, flags); +if (status & (UART_LSR_DR |UART_LSR_BI )) {dw8250_writel_ext (p,RZN1_UART_RDMACR ,0 );dw8250_writel_ext (p,DW_UART_DMASA ,1 );
Signature · 1 line
2.30.2
Andy Shevchenko
reviewerRe: [PATCH 2/3] serial: 8250_dw: Take port lock while accessing LSR
Quoted 2 lines · Ilpo Järvinen
Accessing LSR requires port lock because it mutates
lsr_saved_flagsinserial_lsr_in().
I got this as patch 2/3, where are the 1/3 and 3/3?
...
drivers/tty/serial/8250/8250_dw.c static int dw8250_handle_irq(struct uart_port *p) · line 271
Quoted 7 lines
drivers/tty/serial/8250/8250_dw.c +2
-0@@ -266,7 +266,10 @@ static int dw8250_handle_irq(struct uart_port *p)/* 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) { +spin_lock_irqsave (&p->lock, flags); status =serial_lsr_in (up); +spin_unlock_irqrestore (&p->lock, flags);
This reminds me the question, why do we need to save flags here? Aren't we in IRQ context already? (Perhaps another patch might be issued.)
Signature · 2 lines
With Best Regards, Andy Shevchenko
Ilpo Järvinen
maintainerRe: [PATCH 2/3] serial: 8250_dw: Take port lock while accessing LSR
Quoted 5 lines · Andy Shevchenko
Quoted 2 lines · Ilpo Järvinen
Accessing LSR requires port lock because it mutates
lsr_saved_flagsinserial_lsr_in().I got this as patch 2/3, where are the 1/3 and 3/3?
That's probably because get_maintainer.pl didn't pick you up for the other patches and I didn't explicitly send them to you. Here they are:
https://lore.kernel.org/linux-serial/[email protected]/T/#t
Quoted 10 lines
drivers/tty/serial/8250/8250_dw.c static int dw8250_handle_irq(struct uart_port *p) · line 271
Quoted 7 lines
drivers/tty/serial/8250/8250_dw.c +2
-0@@ -266,7 +266,10 @@ static int dw8250_handle_irq(struct uart_port *p)/* 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) { +spin_lock_irqsave (&p->lock, flags); status =serial_lsr_in (up); +spin_unlock_irqrestore (&p->lock, flags);This reminds me the question, why do we need to save flags here? Aren't we in IRQ context already? (Perhaps another patch might be issued.)
Currently serial8250_handle_irq() reads from LSR again. I guess it would be possible to read LSR only once and pass it as a parameter. It would require some rework though.
Signature · 1 line
i.
Greg KH
maintainerRe: [PATCH 2/3] serial: 8250_dw: Take port lock while accessing LSR
In reply to Ilpo JärvinenQuoted 8 lines · Ilpo Järvinen
Accessing LSR requires port lock because it mutates
lsr_saved_flagsinserial_lsr_in().Fixes 197eb5c416ff ("serial: 8250_dw: Use serial_lsr_in() in dw8250_handle_irq()")Signed-off-by Ilpo Järvinen <[email protected]>--- drivers/tty/serial/8250/8250_dw.c | 3 +++ 1 file changed, 3 insertions(+)
Does not apply against my tty-linus branch, you mixed patches up in this series which is confusing for anyone wanting to apply them.
I've applied the first, but the next two you should resend.
thanks,
greg k-h
Ilpo Järvinen
maintainerPATCH 3/3 serial: 8250_dma: No need for if (dma->tx_err)
In reply to Ilpo JärvinenThe code can just clear dma->tx_err unconditionally.
--- drivers/tty/serial/8250/8250_dma.c | 3 +-- 1 file changed, 1 insertion(+), 2 deletions(-)
drivers/tty/serial/8250/8250_dma.c +1 -2
@@ -107,8 +107,7 @@ int serial8250_tx_dma(struct uart_8250_port *p)dma_async_issue_pending (dma->txchan);serial8250_clear_THRI (p); -if (dma->tx_err) - dma->tx_err =0 ; + dma->tx_err =0 ;return 0 ; err:
Signature · 1 line
2.30.2