Commit f95499c3,
  n_tty: Don't wait for buffer work in read() loop
creates a race window which can cause a pty master read()
to miss the last pty slave write(s) and return -EIO instead,
thus signalling the pty slave is closed. This can happen when
the pty slave is written and immediately closed but before the
tty buffer i/o loop receives the new input; the pty master
read() is scheduled, sees its read buffer is empty and the
pty slave has been closed, and exits.

Because tty_flush_to_ldisc() has significant performance impact
for parallel i/o, rather than revert the commit, special case this
condition (ie., when the read buffer is empty and the 'other' pty
has been closed) and, only then, wait for buffer work to complete
before re-testing if the read buffer is still empty.

As before, subsequent pty master reads return any available data
until no more data is available, and then returns -EIO to
indicate the pty slave has closed.
Reported-by: NMikael Pettersson <mikpelinux@gmail.com>
Signed-off-by: NPeter Hurley <peter@hurleysoftware.com>
Tested-by: NMikael Pettersson <mikpelinux@gmail.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

Commit 40d5e090,
'n_tty: Fix EOF push handling' introduced a subtle state
change error wrt EOF push handling when the termios is
changed from non-canonical to canonical mode.

Reset line_start to the current read_tail index, not 0.
Signed-off-by: NPeter Hurley <peter@hurleysoftware.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

Lockdep reports a circular lock dependency between
atomic_read_lock and termios_rwsem [1]. However, a lock
order deadlock is not possible since CPU1 only holds a
read lock which cannot prevent CPU0 from also acquiring
a read lock on the same r/w semaphore.

Unfortunately, lockdep cannot currently distinguish whether
the locks are read or write for any particular lock graph,
merely that the locks _were_ previously read and/or write.

Until lockdep is fixed, re-order atomic_read_lock so
termios_rwsem can be dropped and reacquired without
triggering lockdep.

Patch based on original posted here https://lkml.org/lkml/2013/8/1/510
by Sergey Senozhatsky <sergey.senozhatsky@gmail.com>

[1] Initial lockdep report from Artem Savkov <artem.savkov@gmail.com>

 ======================================================
 [ INFO: possible circular locking dependency detected ]
 3.11.0-rc3-next-20130730+ #140 Tainted: G        W
 -------------------------------------------------------
 bash/1198 is trying to acquire lock:
  (&tty->termios_rwsem){++++..}, at: [<ffffffff816aa3bb>] n_tty_read+0x49b/0x660

 but task is already holding lock:
  (&ldata->atomic_read_lock){+.+...}, at: [<ffffffff816aa0f0>] n_tty_read+0x1d0/0x660

 which lock already depends on the new lock.

 the existing dependency chain (in reverse order) is:

 -> #1 (&ldata->atomic_read_lock){+.+...}:
        [<ffffffff811111cc>] validate_chain+0x73c/0x850
        [<ffffffff811117e0>] __lock_acquire+0x500/0x5d0
        [<ffffffff81111a29>] lock_acquire+0x179/0x1d0
        [<ffffffff81d34b9c>] mutex_lock_interruptible_nested+0x7c/0x540
        [<ffffffff816aa0f0>] n_tty_read+0x1d0/0x660
        [<ffffffff816a3bb6>] tty_read+0x86/0xf0
        [<ffffffff811f21d3>] vfs_read+0xc3/0x130
        [<ffffffff811f2702>] SyS_read+0x62/0xa0
        [<ffffffff81d45259>] system_call_fastpath+0x16/0x1b

 -> #0 (&tty->termios_rwsem){++++..}:
        [<ffffffff8111064f>] check_prev_add+0x14f/0x590
        [<ffffffff811111cc>] validate_chain+0x73c/0x850
        [<ffffffff811117e0>] __lock_acquire+0x500/0x5d0
        [<ffffffff81111a29>] lock_acquire+0x179/0x1d0
        [<ffffffff81d372c1>] down_read+0x51/0xa0
        [<ffffffff816aa3bb>] n_tty_read+0x49b/0x660
        [<ffffffff816a3bb6>] tty_read+0x86/0xf0
        [<ffffffff811f21d3>] vfs_read+0xc3/0x130
        [<ffffffff811f2702>] SyS_read+0x62/0xa0
        [<ffffffff81d45259>] system_call_fastpath+0x16/0x1b

 other info that might help us debug this:

  Possible unsafe locking scenario:

        CPU0                    CPU1
        ----                    ----
   lock(&ldata->atomic_read_lock);
                                lock(&tty->termios_rwsem);
                                lock(&ldata->atomic_read_lock);
   lock(&tty->termios_rwsem);

  *** DEADLOCK ***

 2 locks held by bash/1198:
  #0:  (&tty->ldisc_sem){.+.+.+}, at: [<ffffffff816ade04>] tty_ldisc_ref_wait+0x24/0x60
  #1:  (&ldata->atomic_read_lock){+.+...}, at: [<ffffffff816aa0f0>] n_tty_read+0x1d0/0x660

 stack backtrace:
 CPU: 1 PID: 1198 Comm: bash Tainted: G        W    3.11.0-rc3-next-20130730+ #140
 Hardware name: Bochs Bochs, BIOS Bochs 01/01/2007
  0000000000000000 ffff880019acdb28 ffffffff81d34074 0000000000000002
  0000000000000000 ffff880019acdb78 ffffffff8110ed75 ffff880019acdb98
  ffff880019fd0000 ffff880019acdb78 ffff880019fd0638 ffff880019fd0670
 Call Trace:
  [<ffffffff81d34074>] dump_stack+0x59/0x7d
  [<ffffffff8110ed75>] print_circular_bug+0x105/0x120
  [<ffffffff8111064f>] check_prev_add+0x14f/0x590
  [<ffffffff81d3ab5f>] ? _raw_spin_unlock_irq+0x4f/0x70
  [<ffffffff811111cc>] validate_chain+0x73c/0x850
  [<ffffffff8110ae0f>] ? trace_hardirqs_off_caller+0x1f/0x190
  [<ffffffff811117e0>] __lock_acquire+0x500/0x5d0
  [<ffffffff81111a29>] lock_acquire+0x179/0x1d0
  [<ffffffff816aa3bb>] ? n_tty_read+0x49b/0x660
  [<ffffffff81d372c1>] down_read+0x51/0xa0
  [<ffffffff816aa3bb>] ? n_tty_read+0x49b/0x660
  [<ffffffff816aa3bb>] n_tty_read+0x49b/0x660
  [<ffffffff810e4130>] ? try_to_wake_up+0x210/0x210
  [<ffffffff816a3bb6>] tty_read+0x86/0xf0
  [<ffffffff811f21d3>] vfs_read+0xc3/0x130
  [<ffffffff811f2702>] SyS_read+0x62/0xa0
  [<ffffffff815e24ee>] ? trace_hardirqs_on_thunk+0x3a/0x3f
  [<ffffffff81d45259>] system_call_fastpath+0x16/0x1b
Reported-by: NArtem Savkov <artem.savkov@gmail.com>
Reported-by: NSergey Senozhatsky <sergey.senozhatsky@gmail.com>
Signed-off-by: NPeter Hurley <peter@hurleysoftware.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

LNEXT processing accounts for ~15% of total cpu time in end-to-end
tty i/o; factor the lnext test/clear from the per-char i/o path.

Instead, attempt to immediately handle the literal next char if not
at the end of this received buffer; otherwise, handle the first char
of the next received buffer as the literal next char, then continue
with normal i/o.
Signed-off-by: NPeter Hurley <peter@hurleysoftware.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

gcc will likely inline these single-use functions anyway; remove
inline modifier.
Signed-off-by: NPeter Hurley <peter@hurleysoftware.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

Always pre-figure the space available in the read_buf and limit
the inbound receive request to that amount.

For compatibility reasons with the non-flow-controlled interface,
n_tty_receive_buf() will continue filling read_buf until all data
has been received or receive_room() returns 0.
Signed-off-by: NPeter Hurley <peter@hurleysoftware.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

Handle PARMRK processing on the slow per-char i/o path.
Signed-off-by: NPeter Hurley <peter@hurleysoftware.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

Convert to modal receive_buf processing; factor char receive
processing for unusual termios settings out of normal per-char
i/o path.
Signed-off-by: NPeter Hurley <peter@hurleysoftware.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

Factor 'special' per-char processing into standalone fn,
n_tty_receive_char_special(), which handles processing for chars
marked in the char_map.
Signed-off-by: NPeter Hurley <peter@hurleysoftware.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

Relocate the IXANY restart tty test to code paths where the
the received char is not START_CHAR, STOP_CHAR, INTR_CHAR,
QUIT_CHAR or SUSP_CHAR.

Fixes the condition when ISIG if off and one of INTR_CHAR,
QUIT_CHAR or SUSP_CHAR does not restart i/o.
Signed-off-by: NPeter Hurley <peter@hurleysoftware.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

Simplify __receive_buf() into a dispatch function; perform per-char
processing for all other modes not already handled.
Signed-off-by: NPeter Hurley <peter@hurleysoftware.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

Commit 20bafb3d
  'n_tty: Move buffers into n_tty_data'
broke the ppc64 build.

Include vmalloc.h for the required function declarations.
Reported-by: NStephen Rothwell <sfr@canb.auug.org.au>
Signed-off-by: NPeter Hurley <peter@hurleysoftware.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

Convert to modal receive_buf() processing; factor receive char
processing when tty->closing into n_tty_receive_buf_closing().

Note that EXTPROC when ISTRIP or IUCLC is set continues to be
handled by n_tty_receive_char().
Signed-off-by: NPeter Hurley <peter@hurleysoftware.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

When EXTPROC is set without ISTRIP or IUCLC, processing is
identical to raw mode; handle this receiving mode as a special-case
of raw mode.
Signed-off-by: NPeter Hurley <peter@hurleysoftware.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

Convert to modal receive_buf() processing; factor raw mode
per-char i/o into n_tty_receive_buf_raw().
Signed-off-by: NPeter Hurley <peter@hurleysoftware.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

Prepare for modal receive_buf() handling; factor handling for
TTY_BREAK, TTY_PARITY, TTY_FRAME and TTY_OVERRUN into
n_tty_receive_char_flagged().
Signed-off-by: NPeter Hurley <peter@hurleysoftware.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

Reduce the monolithic n_tty_receive_char() complexity; factor the
handling of INTR_CHAR, QUIT_CHAR and SUSP_CHAR into
n_tty_receive_signal_char().
Signed-off-by: NPeter Hurley <peter@hurleysoftware.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

Convert to modal receive_buf() processing; factor real_raw
receive_buf() into n_tty_receive_buf_real_raw().
Signed-off-by: NPeter Hurley <peter@hurleysoftware.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

Signed-off-by: NPeter Hurley <peter@hurleysoftware.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

Signed-off-by: NPeter Hurley <peter@hurleysoftware.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

Reduce pointer reloading and improve locality-of-reference;
allocate read_buf and echo_buf within struct n_tty_data.
Signed-off-by: NPeter Hurley <peter@hurleysoftware.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

The char and flag buffer local alias pointers, p and f, are
unnecessary; remove them.
Signed-off-by: NPeter Hurley <peter@hurleysoftware.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

In canonical mode, an EOF which is not the first character of the line
causes read() to complete and return the number of characters read so
far (commonly referred to as EOF push). However, if the previous read()
returned because the user buffer was full _and_ the next character
is an EOF not at the beginning of the line, read() must not return 0,
thus mistakenly indicating the end-of-file condition.

The TTY_PUSH flag is used to indicate an EOF was received which is not
at the beginning of the line. Because the EOF push condition is
evaluated by a thread other than the read(), multiple EOF pushes can
cause a premature end-of-file to be indicated.

Instead, discover the 'EOF push as first read character' condition
from the read() thread itself, and restart the i/o loop if detected.
Signed-off-by: NPeter Hurley <peter@hurleysoftware.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

Separate the head & commit indices from the tail index to avoid
cache-line contention (so called 'false-sharing') between concurrent
threads.
Signed-off-by: NPeter Hurley <peter@hurleysoftware.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

Since neither echo_commit nor echo_tail can change for the duration
of __process_echoes loop, substitute index comparison for the
snapshot counter.
Signed-off-by: NPeter Hurley <peter@hurleysoftware.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

Don't have the driver flush received echoes if no echoes were
actually output.
Signed-off-by: NPeter Hurley <peter@hurleysoftware.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

Byte-by-byte echo output is painfully slow, requiring a lock/unlock
cycle for every input byte.

Instead, perform the echo output in blocks of 256 characters, and
at least once per flip buffer receive. Enough space is reserved in
the echo buffer to guarantee a full block can be saved without
overrunning the echo output. Overrun is prevented by discarding
the oldest echoes until enough space exists in the echo buffer
to receive at least a full block of new echoes.
Signed-off-by: NPeter Hurley <peter@hurleysoftware.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

Use output_lock mutex as a memory barrier when storing echo_commit.
Signed-off-by: NPeter Hurley <peter@hurleysoftware.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

Adding data to echo_buf (via add_echo_byte()) is guaranteed to be
single-threaded, since all callers are from the n_tty_receive_buf()
path. Processing the echo_buf can be called from either the
n_tty_receive_buf() path or the n_tty_write() path; however, these
callers are already serialized by output_lock.

Publish cumulative echo_head changes to echo_commit; process echo_buf
from echo_tail to echo_commit; remove echo_lock.

On echo_buf overrun, claim output_lock to serialize changes to
echo_tail.
Signed-off-by: NPeter Hurley <peter@hurleysoftware.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

Prepare for lockless echo_buf handling; compute current byte count
of echo_buf from head and tail indices.
Signed-off-by: NPeter Hurley <peter@hurleysoftware.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

Instead of using a single index to track the current echo_buf position,
use a head index when adding to the buffer and a tail index when
consuming from the buffer. Allow these head and tail indices to wrap
at max representable value; perform modulo reduction via helper
functions when accessing the buffer.
Signed-off-by: NPeter Hurley <peter@hurleysoftware.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

The echo_overrun field is only assigned and never tested; remove it.
Signed-off-by: NPeter Hurley <peter@hurleysoftware.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

Scheduling buffer work on the same cpu as the read() thread
limits the parallelism now possible between the receive_buf path
and the n_tty_read() path.
Signed-off-by: NPeter Hurley <peter@hurleysoftware.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

The pty driver forces ldisc flow control on, regardless of available
receive buffer space, so the writer can be woken whenever unthrottle
is called. However, this 'forced throttle' has performance
consequences, as multiple atomic operations are necessary to
unthrottle and perform the write wakeup for every input line (in
canonical mode).

Instead, short-circuit the unthrottle if the tty is a pty and perform
the write wakeup directly.
Signed-off-by: NPeter Hurley <peter@hurleysoftware.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

Prepare to special case pty flow control; avoid forward declaration.
Signed-off-by: NPeter Hurley <peter@hurleysoftware.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

Prepare for special handling of pty throttle/unthrottle; factor
flow control into helper functions.
Signed-off-by: NPeter Hurley <peter@hurleysoftware.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

Prepare to factor throttle and unthrottle into helper functions;
relocate chars_in_buffer() to avoid forward declaration.
Signed-off-by: NPeter Hurley <peter@hurleysoftware.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

No tty driver modifies termios during throttle() or unthrottle().
Therefore, only read safety is required.

However, tty_throttle_safe and tty_unthrottle_safe must still be
mutually exclusive; introduce throttle_mutex for that purpose.
Signed-off-by: NPeter Hurley <peter@hurleysoftware.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

If the read buffer indices are in the same cache-line, cpus will
contended over the cache-line (so called 'false sharing').

Separate the producer-published fields from the consumer-published
fields; document the locks relevant to each field.
Signed-off-by: NPeter Hurley <peter@hurleysoftware.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

User-space read() can run concurrently with receiving from device;
waiting for receive_buf() to complete is not required.
Signed-off-by: NPeter Hurley <peter@hurleysoftware.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>