linux-serial.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
From: Peter Hurley <peter@hurleysoftware.com>
To: One Thousand Gnomes <gnomes@lxorguk.ukuu.org.uk>
Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>,
	Jiri Slaby <jslaby@suse.cz>,
	linux-serial@vger.kernel.org, linux-kernel@vger.kernel.org
Subject: Re: [PATCH 21/26] tty: Convert tty_struct bitfield to bools
Date: Wed, 03 Sep 2014 08:14:59 -0400	[thread overview]
Message-ID: <54070643.9080903@hurleysoftware.com> (raw)
In-Reply-To: <20140903115852.5a720754@alan.etchedpixels.co.uk>

On 09/03/2014 06:58 AM, One Thousand Gnomes wrote:
> On Tue,  2 Sep 2014 17:39:30 -0400
> Peter Hurley <peter@hurleysoftware.com> wrote:
> 
>> The stopped, hw_stopped, flow_stopped and packet bits are smp-unsafe
>> and interrupt-unsafe. For example,
>>
>> CPU 0                         | CPU 1
>>                               |
>> tty->flow_stopped = 1         | tty->hw_stopped = 0
>>
>> One of these updates will be corrupted, as the bitwise operation
>> on the bitfield is non-atomic.
>>
>> Ensure each flag has a separate memory location, so concurrent
>> updates do not corrupt orthogonal states.
> 
> Ouch....
> 
> Alas the fix is not sufficient on some platforms. gcc will happily use
> 32bit operations on those fields if it thinks its a performance win.
> 
> It needs to use set_bit and friends.
> 
> x86 is generally ok, but ia64 gcc will do things like load all four
> bytes, mask and write the dword back. I believe ARM gcc may also
> sometimes generate similar results.

Ahh. Thanks for the insight, Alan.

But set_bit() et. al. will generate an incredible amount of churn;
what if I split the fields up to prevent false-sharing?

--- >% ---

diff --git a/include/linux/tty.h b/include/linux/tty.h
index 1c3316a..36d3cf7 100644
--- a/include/linux/tty.h
+++ b/include/linux/tty.h
@@ -252,6 +252,11 @@ struct tty_struct {
 	struct rw_semaphore termios_rwsem;
 	struct mutex winsize_mutex;
 	spinlock_t ctrl_lock;
+	unsigned char ctrl_status;	/* ctrl_lock fields */
+	bool packet;
 	spinlock_t flow_lock;
+	unsigned char stopped:1,	/* flow_lock fields */
+		      flow_stopped:1;
 	/* Termios values are protected by the termios rwsem */
 	struct ktermios termios, termios_locked;
 	struct termiox *termiox;	/* May be NULL for unsupported */
@@ -261,8 +266,7 @@ struct tty_struct {
 	unsigned long flags;
 	int count;
 	struct winsize winsize;		/* winsize_mutex */
-	unsigned char stopped:1, hw_stopped:1, flow_stopped:1, packet:1;
-	unsigned char ctrl_status;	/* ctrl_lock */
+	bool hw_stopped;
 	unsigned int receive_room;	/* Bytes free for queue */
 	int flow_change;


>>
>> Signed-off-by: Peter Hurley <peter@hurleysoftware.com>
>> ---
>>  include/linux/tty.h | 5 ++++-
>>  1 file changed, 4 insertions(+), 1 deletion(-)
>>
>> diff --git a/include/linux/tty.h b/include/linux/tty.h
>> index 1c3316a..7cf61cb 100644
>> --- a/include/linux/tty.h
>> +++ b/include/linux/tty.h
>> @@ -261,7 +261,10 @@ struct tty_struct {
>>  	unsigned long flags;
>>  	int count;
>>  	struct winsize winsize;		/* winsize_mutex */
>> -	unsigned char stopped:1, hw_stopped:1, flow_stopped:1, packet:1;
>> +	bool stopped;
>> +	bool hw_stopped;
>> +	bool flow_stopped;
>> +	bool packet;
>>  	unsigned char ctrl_status;	/* ctrl_lock */
>>  	unsigned int receive_room;	/* Bytes free for queue */
>>  	int flow_change;


  reply	other threads:[~2014-09-03 12:15 UTC|newest]

Thread overview: 68+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2014-09-02 21:39 [PATCH 00/26] tty/serial flow control fixes Peter Hurley
2014-09-02 21:39 ` [PATCH 01/26] Revert "serial: uart: add hw flow control support configuration" Peter Hurley
2014-09-02 21:50   ` Murali Karicheri
2014-09-02 21:58     ` Peter Hurley
2014-09-02 22:17       ` Murali Karicheri
2014-09-02 21:39 ` [PATCH 02/26] serial: Style fix Peter Hurley
2014-09-02 21:39 ` [PATCH 03/26] serial: imx: Fix x_char handling and tx flow control Peter Hurley
2014-09-02 21:39 ` [PATCH 04/26] serial: core: Fix x_char race Peter Hurley
2014-09-02 21:39 ` [PATCH 05/26] serial: core: Remove unsafe x_char optimization Peter Hurley
2014-09-02 21:39 ` [PATCH 06/26] serial: Fix send_xchar() handlers Peter Hurley
2014-09-02 21:39 ` [PATCH 07/26] serial: mpc52xx: Use default serial core x_char handler Peter Hurley
2014-09-02 21:39 ` [PATCH 08/26] serial: sunsab: Don't enable tx if tx stopped Peter Hurley
2014-09-02 21:39 ` [PATCH 09/26] serial: blackfin: Fix missing gpio.h Peter Hurley
2014-09-02 21:39 ` [PATCH 10/26] serial: core: Document lock requirement for UPF_* flags updates Peter Hurley
2014-09-02 21:39 ` [PATCH 11/26] serial: 8250: Document serial8250_modem_status() locking Peter Hurley
2014-09-02 21:39 ` [PATCH 12/26] serial: core: Unwrap tertiary assignment in uart_handle_dcd_change() Peter Hurley
2014-09-02 21:39 ` [PATCH 13/26] locking: Add non-fatal spin lock assert Peter Hurley
2014-09-03  9:27   ` Peter Zijlstra
2014-09-03 11:20     ` Peter Hurley
2014-09-03 14:40       ` Peter Zijlstra
2014-09-03 14:50         ` Peter Hurley
2014-09-03 15:07           ` Peter Zijlstra
2014-09-04  5:14             ` Ingo Molnar
2014-09-10 11:02               ` Peter Hurley
2014-09-10 13:08                 ` Peter Zijlstra
2014-09-10 14:45                   ` Peter Hurley
2014-09-10 18:34                     ` Greg Kroah-Hartman
2014-09-02 21:39 ` [PATCH 14/26] serial: core: Document and assert lock requirements for irq helpers Peter Hurley
2014-09-02 21:39 ` [PATCH 15/26] serial: core: Privatize modem status enable flags Peter Hurley
2014-09-02 21:39 ` [PATCH 16/26] isdn: i4l: Remove ASYNC_CTS_FLOW Peter Hurley
2014-09-02 21:39 ` [PATCH 17/26] serial: core: Privatize tty->hw_stopped Peter Hurley
2014-09-09  0:29   ` Peter Hurley
2014-09-02 21:39 ` [PATCH 18/26] usb: serial: Remove unused tty->hw_stopped Peter Hurley
2014-09-02 21:39 ` [PATCH 19/26] serial: bfin-uart: Fix auto CTS Peter Hurley
2014-09-02 21:39 ` [PATCH 20/26] serial: core: Use spin_lock_irq() in uart_set_termios() Peter Hurley
2014-09-02 21:39 ` [PATCH 21/26] tty: Convert tty_struct bitfield to bools Peter Hurley
2014-09-03 10:58   ` One Thousand Gnomes
2014-09-03 12:14     ` Peter Hurley [this message]
2014-09-03 12:19       ` One Thousand Gnomes
2014-09-03 15:10         ` Peter Hurley
2014-09-03 17:56           ` Greg Kroah-Hartman
2014-09-04 16:08             ` Peter Hurley
2014-09-02 21:39 ` [PATCH 22/26] tty: Serialize tty flow control changes with flow_lock Peter Hurley
2014-09-02 21:39 ` [PATCH 23/26] tty: Move packet mode flow control notifications to pty driver Peter Hurley
2014-09-02 21:39 ` [PATCH 24/26] tty: Serialize tcflow() with other tty flow control changes Peter Hurley
2014-09-02 21:39 ` [PATCH 25/26] tty: Move and rename send_prio_char() as tty_send_xchar() Peter Hurley
2014-09-02 21:39 ` [PATCH 26/26] tty: Hold termios_rwsem for tcflow(TCIxxx) Peter Hurley
2014-09-08 23:24 ` [PATCH 00/26] tty/serial flow control fixes Greg Kroah-Hartman
2014-09-08 23:33   ` Peter Hurley
2014-09-08 23:43     ` Greg Kroah-Hartman
2014-09-08 23:46       ` Peter Hurley
2014-09-09  0:01         ` Greg Kroah-Hartman
2014-09-10 19:06 ` [PATCH v2 00/14] " Peter Hurley
2014-09-10 19:06   ` [PATCH v2 01/14] serial: core: Document and assert lock requirements for irq helpers Peter Hurley
2014-09-10 19:06   ` [PATCH v2 02/14] serial: core: Privatize modem status enable flags Peter Hurley
2014-09-10 19:06   ` [PATCH v2 03/14] isdn: i4l: Remove ASYNC_CTS_FLOW Peter Hurley
2014-09-10 19:06   ` [PATCH v2 04/14] serial: core: Privatize tty->hw_stopped Peter Hurley
2014-09-10 19:06   ` [PATCH v2 05/14] usb: serial: Remove unused tty->hw_stopped Peter Hurley
2014-09-23 15:33     ` Johan Hovold
2014-09-10 19:06   ` [PATCH v2 06/14] serial: bfin-uart: Fix auto CTS Peter Hurley
2014-09-10 19:06   ` [PATCH v2 07/14] serial: core: Use spin_lock_irq() in uart_set_termios() Peter Hurley
2014-09-10 19:06   ` [PATCH v2 08/14] tty: Convert tty_struct bitfield to ints Peter Hurley
2014-09-10 19:06   ` [PATCH v2 09/14] tty: Serialize tty flow control changes with flow_lock Peter Hurley
2014-09-10 19:06   ` [PATCH v2 10/14] tty: Move packet mode flow control notifications to pty driver Peter Hurley
2014-09-10 19:06   ` [PATCH v2 11/14] tty: Serialize tcflow() with other tty flow control changes Peter Hurley
2014-09-10 19:06   ` [PATCH v2 12/14] tty: Move and rename send_prio_char() as tty_send_xchar() Peter Hurley
2014-09-10 19:06   ` [PATCH v2 13/14] tty: Hold termios_rwsem for tcflow(TCIxxx) Peter Hurley
2014-09-10 19:06   ` [PATCH v2 14/14] tty: Workaround Alpha non-atomic byte storage in tty_struct Peter Hurley

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=54070643.9080903@hurleysoftware.com \
    --to=peter@hurleysoftware.com \
    --cc=gnomes@lxorguk.ukuu.org.uk \
    --cc=gregkh@linuxfoundation.org \
    --cc=jslaby@suse.cz \
    --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).