Message-Id: <1360095638-6624-1-git-send-email-peter@hurleysoftware.com>
Date:	Tue,  5 Feb 2013 15:20:15 -0500
From:	Peter Hurley <peter@...leysoftware.com>
To:	Greg Kroah-Hartman <gregkh@...uxfoundation.org>,
	Alan Cox <alan@...ux.intel.com>, Jiri Slaby <jslaby@...e.cz>,
	Sasha Levin <levinsasha928@...il.com>,
	Sebastian Andrzej Siewior <bigeasy@...utronix.de>
Cc:	linux-serial@...r.kernel.org, linux-kernel@...r.kernel.org,
	Ilya Zykov <ilya@...x.ru>, Dave Jones <davej@...hat.com>,
	Peter Hurley <peter@...leysoftware.com>
Subject: [PATCH v3 00/23] ldisc fixes

Sorry for the long delay. Yearly gnome desktop upgrade with lots of
new bugs, holidays, etc.

Alan, I included you in this series because you know more than most about
how the ldisc layer works. I apologize if you find this unwelcome.

Greg, I'd like special maintainer dispension for using printk(KERN_DEBUG...) in
  tty: Bracket ldisc release with TTY_DEBUG_HANGUP messages
  tty: Add ldisc hangup debug messages
I think this usage is in keeping with existing form (in any event, one of
messages is simply being relocated).

Changes in v3:

- Because of new changes to the way drivers interact with tty flip
  buffers, this series no longer purports to prevent the flush_to_ldisc()
  warning.

  That said, this series does fix the warnings generated by trinity and the
  various test jigs run without complaint. Ilya's latest stress_test_tty.c
  runs to completion (with the read and write threads enabled) with no
  WARNs (or BUGs with the series "tty: Fix parallel master and slave pty opens"
  -- which just got pushed to -tty-next)

- Since the lock correction initially reported on by
  Sebastian Andrzej Siewior <bigeasy@...utronix.de> here
  https://lkml.org/lkml/2012/11/21/267 was fixed by Sebastian in
  v3 of "tty: don't deadlock while flushing workqueue" and was
  pushed to -next, the patches
    "tty: Don't reenable already enable ldisc"
    "tty: Don't restart ldisc on hangup if racing ldisc kill"
    "tty: Make core responsible for synchronizing its work"
  fix the other races which may occur when async hangup races
  with the tty final close.

  Please re-test with your dummy_hcd/g_nokia testcase, although
  I'm not convinced that usb gadget is using tty_hangup() appropriately.
  tty drivers use this for async carrier loss coming from an IRQ
  which will be disabled if the tty has been shutdown. Does gserial
  prevent async hangup to a dead tty in a similar fashion?

- Jiri's debug patch was removed to avoid grumblings about using
  IS_ERR_OR_NULL() --
  see http://lkml.indiana.edu/hypermail/linux/kernel/1301.1/00913.html

- "tty: WARN if buffer work racing with tty free" was removed
  because this happens all the time now that drivers can push
  i/o to a dead tty.

- The false-positive diagnostic reported by Sasha Levin here
  http://lkml.indiana.edu/hypermail/linux/kernel/1212.2/01017.html
  was fixed in:
    "n_tty: Fully initialize ldisc before restarting buffer work"

- Fixed various comments which were either wrong or hadn't followed
  the code they belonged to (but only after endless hours of
  auditing various ldisc code paths)

- Added more debug log messages for use with debugging tty hangup

Changes in v2:

- Please review "tty: Don't flush buffer when closing ldisc".
  This patch replaces the earlier
  "tty: Don't reschedule buffer work while closing". The text of
  this commit details why not calling n_tty_flush_buffer() is the
  correct thing to do, so I won't repeat it here.

- The test jig has been included in the commit message for
  "tty: Don't flush buffer when closing ldisc" as Alan requested.

- Ilya Zykov was added as the Signed-off-by: for the test jig in
  that same commit message.

- Sasha Levin was added as the Reported-by: in that same patch.


This patch series addresses various causes of flip buffer work
being scheduled for a dead ldisc & tty.

The most common cause stems from the n_tty_close() path
flushing (which schedules buffer work), when the ldisc has already
been halted (meaning userspace has been locked out of further i/o
and existing users have finished). This is fixed in
'tty: Don't flush buffer when closing ldisc'

The other causes have a central theme: incorrect order-of-operations
when halting a line discipline. In general, to prevent buffer work
from being scheduled requires:
  1. Disallowing further ldisc references
  2. Waiting for all existing ldisc references to be released
  3. Cancelling existing buffer work
If the wait takes place _after_ cancellation, then new work
can be scheduled by existing holder(s) of ldisc reference(s). That's
bad.

Halting the line discipline is performed when,
 - hanging up the tty (tty_ldisc_hangup())
 - TIOCSETD ioctl (tty_set_ldisc())
 - finally closing the tty (pair) (tty_ldisc_release())

Concurrent halts are governed by the following requirements:
1. tty_ldisc_release is not concurrent with the other two and so
   does not need lock or state protection during the ldiscs halt.
2. Accesses through tty->ldisc must be protected by the ldisc_mutex.
   The wait operation checks the user count (ldisc references)
   in tty->ldisc->users.
3. tty_set_ldisc() reschedules buffer work that was pending when
   the ldiscs were halted. That must be an atomic operation in
   conjunction with re-enabling the ldisc -- which necessitates
   locking concurrent halts (tty_ldisc_release is exempt here)
4. The legacy mutex cannot be held while waiting for ldisc
   reference(s) release -or- for cancelling buffer work.
5. Because of #4, the legacy mutex must be dropped prior to or
   during the halt. Which means reacquiring after the halt. But
   to preserve lock order the ldisc_mutex must be dropped and
   reacquired after reacquiring the legacy mutex.
6. Because of #5, the ldisc state may have changed while the
   ldisc mutex was dropped.


Peter Hurley (23):
  tty: Add diagnostic for halted line discipline
  n_tty: Factor packet mode status change for reuse
  n_tty: Don't flush buffer when closing ldisc
  tty: Refactor wait for ldisc refs out of tty_ldisc_hangup()
  tty: Remove unnecessary re-test of ldisc ref count
  tty: Fix ldisc halt sequence on hangup
  tty: Strengthen no-subsequent-use guarantee of tty_ldisc_halt()
  tty: Halt both ldiscs concurrently
  tty: Remove unnecessary buffer work flush
  tty: Wait for SAK work before waiting for hangup work
  n_tty: Correct unthrottle-with-buffer-flush comments
  tty: Kick waiters _after_ the ldisc is locked
  n_tty: Fully initialize ldisc before restarting buffer work
  tty: Don't reenable already enabled ldisc
  tty: Don't restart ldisc on hangup if racing ldisc kill
  tty: Make core responsible for synchronizing its work
  tty: Document lock requirements to halt ldisc
  tty: Remove stale comment re: tty_ldisc_flush_works()
  tty: Fix 'deferred reopen' ldisc comment
  tty: Remove stale comment re: locking in tty_ldisc_release()
  tty: Re-parent orphaned tty_set_ldisc() comments
  tty: Bracket ldisc release with TTY_DEBUG_HANGUP messages
  tty: Add ldisc hangup debug messages

 drivers/tty/n_tty.c     |  62 +++++++-----
 drivers/tty/tty_io.c    |  23 ++++-
 drivers/tty/tty_ldisc.c | 248 ++++++++++++++++++++++++++++--------------------
 include/linux/tty.h     |   1 +
 4 files changed, 205 insertions(+), 129 deletions(-)

-- 
1.8.1.2

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/

Powered by blists - more mailing lists