From: tony@atomide.com (Tony Lindgren)
To: linux-arm-kernel@lists.infradead.org
Subject: [PATCH 03/16] tty: serial: 8250_core: read only RX if there is something in the FIFO
Date: Wed, 11 Feb 2015 12:03:14 -0800 [thread overview]
Message-ID: <20150211200313.GE2531@atomide.com> (raw)
In-Reply-To: <54DBB531.2030504@hurleysoftware.com>
* Peter Hurley <peter@hurleysoftware.com> [150211 12:05]:
> On 02/10/2015 12:46 PM, Peter Hurley wrote:
> > On 02/10/2015 07:04 AM, Nicolas Schichan wrote:
> >> On 02/10/2015 12:34 AM, Peter Hurley wrote:
> >>> Hi Nicolas,
> >>>
> >>> Thanks for the report.
> >>>
> >> [...]
> >>>> When a caracter is received on the UART while the kernel is printing
> >>>> the boot messages, as soon as the kernel configures the UART for
> >>>> receiving (after root filesystem mount), it gets stuck printing the
> >>>> following message repeatedly:
> >>>>
> >>>> serial8250: too much work for irq29
> >>>>
> >>>> Once stuck, the reception of another character allows the boot process
> >>>> to finish.
> >>>>
> >>>> From what I can gather, when we hit that, the UART_IIR_NO_INT is 0 (so the
> >>>> interrupt is raised), but the UART_LSR_DR bit is 0 as well so the UART_RX
> >>>> register is never read to clear the interrupt.
> >>>
> >>> The "too much work" message means serial8250_handle_irq() is returning 0,
> >>> ie., not handled. Which in turn means IIR indicates no interrupt is pending
> >>> (UART_IIR_NO_INT == 1).
> >>>
> >>> Can you log the register values for LSR and IIR at both patch locations
> >>> in serial8250_do_startup()?
> >>>
> >>> (I can get you a debug patch, if necessary. Let me know)
> >>
> >> Hi Peter,
> >>
> >> Thanks for your reply.
> >>
> >> Here is what I have when the issue is triggered:
> >>
> >> [ 12.154877] lsr 0x60 / iir 0x01
> >> [ 12.158071] lsr 0x60 / iir 0x01
> >> [ 12.161438] serial8250: too much work for irq29
> >> [ 12.165982] lsr 0x60 / iir 0x0c
> >> [ 12.169354] serial8250: too much work for irq29
> >> [ 12.173900] lsr 0x60 / iir 0x0c
> >> (previous two messages are repeated and printk_ratelimited())
> >
> > Thanks for this information; I see I was wrong about the cause of message.
> >
> > I think what happens during startup is that on this silicon clearing
> > the rx fifo (by serial8250_clear_fifos()) clears data ready but not
> > the rx timeout condition which causes a spurious rx interrupt when
> > interrupts are enabled.
> >
> > So caught between two broken UARTs: one that underflows its rx fifo because
> > of unsolicited rx reads and the other that generates spurious interrupt
> > without unsolicited rx reads.
> >
> >
> >> When the issue is not triggered:
> >>
> >> [ 10.784871] lsr 0x60 / iir 0x01
> >> [ 10.788066] lsr 0x60 / iir 0x01
> >> [ 10.794734] VFS: Mounted root (nfs filesystem) readonly on device 0:13.
> >> [ 10.801654] devtmpfs: mounted
> >> [ 10.805169] Freeing unused kernel memory: 184K (807be000 - 807ec000)
> >> (userland takes over after that)
> >>
> >> I have also displayed the IIR and LSR registers when the "too much fork for
> >> IRQ" condition is triggered.
> >>
> >> In the serial8250_do_startup(), before the interrupt are unmasked at the end,
> >> the IIR looks sane and UART_IIR_NO_INT bit is set. When stuck
> >> serial8250_interrupt(), UART_IIR_NO_INT is cleared and the interrupt ID is set
> >> to 0xc which is not handled by the kernel at this time (the Kirkwood datasheet
> >> indicates that it is some kind of timeout condition from what I can gather).
> >
> > Yes, IIR == UART_IIR_RX_TIMEOUT is to used indicate that data is in the rx fifo
> > but has not reached the rx trigger level yet.
> >
> > ATM, I'm not exactly sure if there is a safe way to clear the spurious interrupt
> > from the interrupt handler.
> >
> > I'm fairly certain the only way to clear the rx timeout interrupt is to read
> > the rx fifo, but I think this would race with actual data arrival. IOW, there
> > might not be a way to determine if the data read is spurious or not.
>
> Yep, I see no safe way to clear the spurious interrupt [1] and no idea how to
> keep it from happening (other than via the unsolicited RX reads in
> serial8250_do_startup).
>
> Unfortunately, I think this means we'll have to revert Sebastian's commit:
>
> commit 0aa525d11859c1a4d5b78fdc704148e2ae03ae13
> Author: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
> Date: Wed Sep 10 21:29:58 2014 +0200
>
> tty: serial: 8250_core: read only RX if there is something in the FIFO
>
> which just means OMAP3630 will be limited to using the omap_serial driver.
Reverting makes sense to me if it has caused a regression. Maybe Sebastian
can update his patch to do this based on some quirk flag instead?
Regards,
Tony
> [1] To clear the RX timeout interrupt requires reading the rx fifo even though
> LSR[data ready] indicates no data. However, this could result in dropped data
> if the data became available just before clearing the RX timeout. For example,
>
> CPU | Device
> |
> irq handler (simplified) |
> |
> read IIR |
> is interrupt? yes |
> read LSR |
> is data ready? no |
> is IIR == Rx timeout? yes | new data arrives
> | rx_fifo[0] = new data
> | lsr[data ready] = 1
> read RX and discard |
> |
>
next prev parent reply other threads:[~2015-02-11 20:03 UTC|newest]
Thread overview: 97+ messages / expand[flat|nested] mbox.gz Atom feed top
2014-09-10 19:29 [PATCH 00/16 v9] omap 8250 based uart + DMA Sebastian Andrzej Siewior
2014-09-10 19:29 ` [PATCH 01/16] tty: serial: 8250_core: allow to set ->throttle / ->unthrottle callbacks Sebastian Andrzej Siewior
2014-09-10 19:29 ` [PATCH 02/16] tty: serial: 8250_core: add run time pm Sebastian Andrzej Siewior
2014-09-29 9:46 ` Frans Klaver
2014-09-29 13:39 ` Sebastian Andrzej Siewior
2014-09-10 19:29 ` [PATCH 03/16] tty: serial: 8250_core: read only RX if there is something in the FIFO Sebastian Andrzej Siewior
2015-02-09 13:34 ` Nicolas Schichan
2015-02-09 23:34 ` Peter Hurley
2015-02-10 9:32 ` Sebastian Andrzej Siewior
2015-02-10 12:04 ` Nicolas Schichan
2015-02-10 17:46 ` Peter Hurley
2015-02-11 20:01 ` Peter Hurley
2015-02-11 20:03 ` Tony Lindgren [this message]
2015-02-11 20:42 ` Peter Hurley
2015-02-12 8:45 ` Sebastian Andrzej Siewior
2015-02-12 9:40 ` Russell King - ARM Linux
2015-02-12 16:32 ` Peter Hurley
2015-02-12 19:23 ` Sebastian Andrzej Siewior
2015-02-12 19:55 ` Peter Hurley
2015-02-12 20:34 ` Sebastian Andrzej Siewior
2015-02-13 18:51 ` Sebastian Andrzej Siewior
2015-02-13 23:15 ` Russell King - ARM Linux
2015-02-15 17:32 ` [PATCH] serial: 8250: Revert "tty: serial: 8250_core: read only RX if there is something in the FIFO" Sebastian Andrzej Siewior
2015-05-12 20:25 ` Tony Lindgren
2014-09-10 19:29 ` [PATCH 04/16] tty: serial: 8250_core: use the ->line argument as a hint in serial8250_find_match_or_unused() Sebastian Andrzej Siewior
2014-09-10 19:30 ` [PATCH 05/16] tty: serial: 8250_core: remove UART_IER_RDI in serial8250_stop_rx() Sebastian Andrzej Siewior
2014-09-11 11:19 ` Heikki Krogerus
2014-09-10 19:30 ` [PATCH 06/16] tty: serial: Add 8250-core based omap driver Sebastian Andrzej Siewior
2014-09-11 11:57 ` Peter Hurley
2014-09-16 17:01 ` Sebastian Andrzej Siewior
2014-09-29 9:38 ` Frans Klaver
2014-09-29 13:27 ` Sebastian Andrzej Siewior
2014-09-29 13:34 ` Frans Klaver
2014-09-10 19:30 ` [PATCH 07/16] tty: serial: 8250_dma: handle error on TX submit Sebastian Andrzej Siewior
2014-09-10 19:30 ` [PATCH 08/16] tty: serial: 8250_dma: enqueue RX dma again on completion Sebastian Andrzej Siewior
2014-09-10 19:30 ` [PATCH 09/16] tty: serial: 8250_dma: Add a TX trigger workaround for AM33xx Sebastian Andrzej Siewior
2014-09-11 11:17 ` Heikki Krogerus
2014-09-11 11:42 ` Sebastian Andrzej Siewior
2014-09-11 12:32 ` Peter Hurley
2014-09-11 12:50 ` Sebastian Andrzej Siewior
2014-09-11 14:35 ` Peter Hurley
2014-09-15 17:01 ` Sebastian Andrzej Siewior
2014-09-16 16:55 ` Sebastian Andrzej Siewior
2014-09-17 12:20 ` Peter Hurley
2014-09-17 16:25 ` Sebastian Andrzej Siewior
2014-09-29 16:15 ` Sebastian Andrzej Siewior
2014-09-11 15:11 ` Frans Klaver
2014-09-11 16:04 ` Sebastian Andrzej Siewior
2014-09-11 17:04 ` Frans Klaver
2014-09-12 7:23 ` Sebastian Andrzej Siewior
[not found] ` <20140912094010.GE28458@ci00147.xsens-tech.local>
2014-09-12 9:51 ` Sebastian Andrzej Siewior
2014-09-12 10:28 ` Frans Klaver
2014-09-15 16:42 ` Sebastian Andrzej Siewior
2014-09-16 9:05 ` Frans Klaver
2014-09-16 12:42 ` Frans Klaver
2014-09-16 14:23 ` Frans Klaver
2014-09-17 10:28 ` Frans Klaver
2014-09-21 20:41 ` Sebastian Andrzej Siewior
2014-09-22 9:28 ` Frans Klaver
2014-09-24 7:56 ` Sebastian Andrzej Siewior
2014-09-25 15:14 ` Sebastian Andrzej Siewior
2014-09-25 15:18 ` Frans Klaver
2014-09-29 8:50 ` Frans Klaver
2014-09-29 9:54 ` Sebastian Andrzej Siewior
2014-09-29 10:30 ` Frans Klaver
2014-09-30 8:44 ` Frans Klaver
2014-10-02 10:27 ` Sebastian Andrzej Siewior
2014-10-13 14:55 ` Frans Klaver
2014-09-23 17:03 ` Peter Hurley
2014-09-24 7:53 ` Sebastian Andrzej Siewior
2014-09-25 10:42 ` Sebastian Andrzej Siewior
2014-09-25 11:31 ` Peter Hurley
2014-09-25 13:11 ` Sebastian Andrzej Siewior
2014-09-17 16:34 ` Sebastian Andrzej Siewior
2014-09-19 10:22 ` Heikki Krogerus
2014-09-19 10:58 ` Sebastian Andrzej Siewior
2014-09-19 11:25 ` Peter Hurley
2014-09-22 7:46 ` Heikki Krogerus
2014-09-25 9:24 ` Sebastian Andrzej Siewior
2014-09-10 19:30 ` [PATCH 10/16] tty: serial: 8250_dma: optimize the xmit path due to UART_BUG_DMA_TX Sebastian Andrzej Siewior
2014-09-10 19:30 ` [PATCH 11/16] tty: serial: 8250_dma: keep own book keeping about RX transfers Sebastian Andrzej Siewior
2014-09-10 19:30 ` [PATCH 12/16] tty: serial: 8250_dma: handle the UART RDI event while DMA remains idle Sebastian Andrzej Siewior
2014-09-29 9:23 ` Frans Klaver
2014-09-10 19:30 ` [PATCH 13/16] tty: serial: 8250_dma: add pm runtime Sebastian Andrzej Siewior
2014-09-29 9:26 ` Frans Klaver
2014-09-29 9:56 ` Sebastian Andrzej Siewior
2014-09-10 19:30 ` [PATCH 14/16] arm: dts: am33xx: add DMA properties for UART Sebastian Andrzej Siewior
2014-09-10 19:30 ` [PATCH 15/16] arm: dts: dra7: " Sebastian Andrzej Siewior
2014-09-10 19:30 ` [PATCH 16/16] tty: serial: 8250: omap: add dma support Sebastian Andrzej Siewior
2014-09-12 22:43 ` [PATCH 00/16 v9] omap 8250 based uart + DMA Tony Lindgren
2014-09-15 11:50 ` Sebastian Andrzej Siewior
2014-09-16 12:57 ` Sebastian Andrzej Siewior
2014-09-16 16:48 ` Tony Lindgren
2014-09-16 21:30 ` Tony Lindgren
2014-09-17 8:38 ` Sebastian Andrzej Siewior
2014-09-17 9:05 ` Sebastian Andrzej Siewior
2014-09-26 16:02 ` Greg KH
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=20150211200313.GE2531@atomide.com \
--to=tony@atomide.com \
--cc=linux-arm-kernel@lists.infradead.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).