From: Peter Hurley <peter@hurleysoftware.com>
To: california.l.sullivan@intel.com, linux-kernel@vger.kernel.org
Cc: linux-serial@vger.kernel.org, alan@linux.intel.com,
darren.hart@intel.com, gregkh@linuxfoundation.org,
jslaby@suse.com, heikki.krogerus@intel.com,
andriy.shevchenko@intel.com, arjan@linux.intel.com
Subject: Re: [PATCH RFC] serial: 8250: Handle invalid UART state
Date: Mon, 24 Aug 2015 09:31:38 -0400 [thread overview]
Message-ID: <55DB1CBA.3070504@hurleysoftware.com> (raw)
In-Reply-To: <1440015124-28393-1-git-send-email-california.l.sullivan@intel.com>
On 08/19/2015 04:12 PM, california.l.sullivan@intel.com wrote:
> From: California Sullivan <california.l.sullivan@intel.com>
>
> The debug UART on the Bay Trail is buggy and will sometimes enter a
> state where there is a Character Timeout interrupt, but the Data
> Ready bit in the Line Status Register is not set, despite there
> being data available in the FIFO. It stays in this state until the
> Receive Buffer is read. Because the 8250 driver does not read the
> Receive Buffer without the Data Ready bit being set, the UART is
> never read once we reach this point, and thus the Character Timeout
> interrupt is never cleared. This makes the driver spin in a loop
> causing multiple "too much work for irq" errors and eventually
> locking up entirely.
LSR doesn't reflect the actual state of the fifo?
So kgdb (CONSOLE_POLL) won't work on these platforms either, right?
Does this happen with only a single byte in the fifo or is it any
number of bytes under the fifo trigger?
> This patch handles the invalid state by setting the Data Ready bit
> in the 'status' variable when the invalid state occurs. This allows
> the receive buffer to be read, which clears the interrupt
> identification register in the UART and brings us back to a correct
> state.
>
> After this patch was submitted for internal comment, a similar bug on
> the HSUARTs of the Bay Trail and Braswell platforms was pointed out.
> On those UARTs, the invalid state mentioned previously is reached for
> some amount of time, cauing the same "too much work for irq" messages,
> but not permanently locking up the UART. This patch has been confirmed
> to solve that issue as well.
>
> We considered solving this by adding a new UART_BUG_ define and
> detecting the buggy UART by either identifying the processor or by
> adding a new PNP device ID to firmware. However, this solution
> would be more complex and have zero performance benefits, as the
> ISR would require a similar 'if' condition to detect and handle the
> bug.
>
> Our main concern with this fix is whether it having a Character
> Timeout with no Data Ready is an invalid state for all UARTs or
> just some. If other UARTs have a Character Timeout without
> immediately flipping the Data Ready bit in the Line Status Register
> for a good reason, setting the Data Ready bit in the 'status'
> variable could have unintended consequences.
I think there is every likelihood of spurious RX timeout interrupts
tripping this patch, sorry.
Unfortunately, I think UART_BUG_ is the only viable possibility.
Or perhaps fixing the port type as PORT_8250 (thus disabling the fifos).
Regards,
Peter Hurley
> Signed-off-by: California Sullivan <california.l.sullivan@intel.com>
> ---
> drivers/tty/serial/8250/8250_core.c | 8 ++++++++
> 1 file changed, 8 insertions(+)
>
> diff --git a/drivers/tty/serial/8250/8250_core.c b/drivers/tty/serial/8250/8250_core.c
> index ea1a8da..078b118 100644
> --- a/drivers/tty/serial/8250/8250_core.c
> +++ b/drivers/tty/serial/8250/8250_core.c
> @@ -1604,6 +1604,14 @@ int serial8250_handle_irq(struct uart_port *port, unsigned int iir)
>
> DEBUG_INTR("status = %x...", status);
>
> + /*
> + * Workaround for when there is a character timeout interrupt
> + * but the data ready bit is not set in the Line Status Register.
> + */
> + if ((iir & UART_IIR_RX_TIMEOUT) &&
> + !(status & (UART_LSR_DR | UART_LSR_BI)))
> + status |= UART_LSR_DR;
> +
> if (status & (UART_LSR_DR | UART_LSR_BI)) {
> if (up->dma)
> dma_err = up->dma->rx_dma(up, iir);
>
prev parent reply other threads:[~2015-08-24 13:31 UTC|newest]
Thread overview: 2+ messages / expand[flat|nested] mbox.gz Atom feed top
2015-08-19 20:12 [PATCH RFC] serial: 8250: Handle invalid UART state california.l.sullivan
2015-08-24 13:31 ` Peter Hurley [this message]
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=55DB1CBA.3070504@hurleysoftware.com \
--to=peter@hurleysoftware.com \
--cc=alan@linux.intel.com \
--cc=andriy.shevchenko@intel.com \
--cc=arjan@linux.intel.com \
--cc=california.l.sullivan@intel.com \
--cc=darren.hart@intel.com \
--cc=gregkh@linuxfoundation.org \
--cc=heikki.krogerus@intel.com \
--cc=jslaby@suse.com \
--cc=linux-kernel@vger.kernel.org \
--cc=linux-serial@vger.kernel.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).