According to fcntl(2), "a SIGIO signal is sent whenever input
or output becomes possible on that file descriptor", i.e.
after the output buffer was full and now has space for new data.
But in fact SIGIO is sent after every write.

n_tty_write() should set TTY_DO_WRITE_WAKEUP only when
not all data could be written to the buffer.

[pjh: Also fixes missed SIGIO if amt written just happens to be
[     amount still to write
Signed-off-by: NJohannes Stezenbach <js@sig21.net>
[pjh: minor patch edits and re-submit]
Signed-off-by: NPeter Hurley <peter@hurleysoftware.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

Since n_tty_check_unthrottle() is only called from n_tty_read()
which only originates from a userspace read(), the tty count cannot
be 0; the read() guarantees the file descriptor has not yet been
released.
Signed-off-by: NPeter Hurley <peter@hurleysoftware.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

If signal-driven i/o is disabled while write wakeup is pending (ie.,
n_tty_write() has set TTY_DO_WRITE_WAKEUP but then signal-driven i/o
is disabled), the TTY_DO_WRITE_WAKEUP bit will never be cleared and
will cause tty_wakeup() to always call n_tty_write_wakeup.

Unconditionally clear the write wakeup, and since kill_fasync()
already checks if the fasync ptr is null, call kill_fasync()
unconditionally as well.
Signed-off-by: NPeter Hurley <peter@hurleysoftware.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

A small race window exists which allows signal-driven async i/o to be
enabled for the tty when the file ptr has already been hungup and
signal-driven i/o has been disabled:

CPU 0                                CPU 1
-----                                ------
ioctl_fioasync(on)
  filp->f_op->fasync(on)             __tty_hangup()
    tty_fasync(on)                     tty_lock()
      tty_lock()                       ...
        .                              filp->f_op = &hung_up_tty_fops;
      (waiting)                       __tty_fasync(off)
        .                              tty_unlock()
      /* gets tty lock  */
      /* enables FASYNC */

Check the tty has not been hungup while holding tty_lock.
Signed-off-by: NPeter Hurley <peter@hurleysoftware.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

VFS uses a two-stage check-and-call method for invoking file_operations
methods, without explicitly snapshotting either the file_operations ptr
or the function ptr. Since the tty core is one of the few VFS users that
changes the f_op file_operations ptr of the file descriptor (when the
tty has been hung up), and since the likelihood of the compiler generating
a reload of either f_op or the function ptr is basically nil, just define
a hung up fasync() file operation that returns an error.
Signed-off-by: NPeter Hurley <peter@hurleysoftware.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

Only the N_TTY line discipline implements the signal-driven i/o
notification enabled/disabled by fcntl(F_SETFL, O_ASYNC). The ldisc
fasync() notification is sent to the ldisc when the enable state has
changed (the tty core is notified via the fasync() VFS file operation).

The N_TTY line discipline used the enable state to change the wakeup
condition (minimum_to_wake = 1) for notifying the signal handler i/o is
available. However, just the presence of data is sufficient and necessary
to signal i/o is available, so changing minimum_to_wake is unnecessary
(and creates a race condition with read() and poll() which may be
concurrently updating minimum_to_wake).

Furthermore, since the kill_fasync() VFS helper performs no action if
the fasync list is empty, calling unconditionally is preferred; if
signal driven i/o just has been disabled, no signal will be sent by
kill_fasync() anyway so notification of the change via the ldisc
fasync() method is superfluous.
Signed-off-by: NPeter Hurley <peter@hurleysoftware.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

A read() in non-canonical mode when VMIN > 0 and VTIME == 0 does not
complete until at least VMIN chars have been read (or the user buffer is
full). In this infrequent read mode, n_tty_read() attempts to reduce
wakeups by computing the amount of data still necessary to complete the
read (minimum_to_wake) and only waking the read()/poll() when that much
unread data has been processed. This is the only read mode for which
new data does not necessarily generate a wakeup.

However, this optimization is broken and commonly leads to hung reads
even though the necessary amount of data has been received. Since the
optimization is of marginal value anyway, just remove the whole
thing. This also remedies a race between a concurrent poll() and
read() in this mode, where the poll() can reset the minimum_to_wake
of the read() (and vice versa).
Signed-off-by: NPeter Hurley <peter@hurleysoftware.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

Warn if tty_audit_buf use is attempted after tty_audit_exit() has
already freed it.
Signed-off-by: NPeter Hurley <peter@hurleysoftware.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

The data read from another tty may be relevant to the action of
the TIOCSTI ioctl; log the audit buffer immediately.
Signed-off-by: NPeter Hurley <peter@hurleysoftware.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

Audit is unlikely to be enabled; check first to exit asap.
Signed-off-by: NPeter Hurley <peter@hurleysoftware.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

The first-use tty audit buffer allocation is a potential race
amongst multiple attempts at 'first-use'; only one 'winner' is
acceptable.

The successful buffer assignment occurs if tty_audit_buf == NULL
(which will also be the return from cmpxchg()); otherwise, another
racer 'won' and this buffer allocation is freed.
Signed-off-by: NPeter Hurley <peter@hurleysoftware.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

When tty_audit_exit() is called from do_exit(), the process is
single-threaded. Since the tty_audit_buf is only shared by threads
of a process, no other thread can be concurrently accessing the
tty_audit_buf during or after tty_audit_exit().

Thus, no other thread can be holding an extra tty_audit_buf reference
which would prevent tty_audit_exit() from freeing the tty_audit_buf.
As that is the only purpose of the ref counting, remove the reference
counting and free the tty_audit_buf directly.
Signed-off-by: NPeter Hurley <peter@hurleysoftware.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

The tty audit buffer is allocated at first use and not freed until
the process exits. If tty audit is turned off after the audit buffer
has been allocated, no effort is made to release the buffer.
So re-checking if tty audit has just been turned off when tty audit
was just on is false optimization; the likelihood of triggering this
condition is exceedingly small.
Signed-off-by: NPeter Hurley <peter@hurleysoftware.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

The audit_tty and audit_tty_log_passwd fields are actually bool
values, so merge into single memory location to access atomically.

NB: audit log operations may still occur after tty audit is disabled
which is consistent with the existing functionality
Signed-off-by: NPeter Hurley <peter@hurleysoftware.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

Use dev_t instead of separate major/minor fields to track tty
audit buffer association.
Signed-off-by: NPeter Hurley <peter@hurleysoftware.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

tty_audit_push() and tty_audit_push_current() perform identical
tasks; eliminate the tty_audit_push() implementation and the
tty_audit_push_current() name.
Signed-off-by: NPeter Hurley <peter@hurleysoftware.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

In canonical read mode, each line read and logged is pushed separately
with tty_audit_push(). For all single-threaded processes and multi-threaded
processes reading from only one tty, this patch has no effect; the last line
read will still be the entry pushed to the audit log because the tty
association cannot have changed between tty_audit_add_data() and
tty_audit_push().

For multi-threaded processes reading from different ttys concurrently,
the audit log will have mixed log entries anyway. Consider two ttys
audited concurrently:

CPU0                           CPU1
----------                     ------------
tty_audit_add_data(ttyA)
                               tty_audit_add_data(ttyB)
tty_audit_push()
                               tty_audit_add_data(ttyB)
                               tty_audit_push()

This patch will now cause the ttyB output to be split into separate
audit log entries.

However, this possibility is equally likely without this patch:

CPU0                           CPU1
----------                     ------------
                               tty_audit_add_data(ttyB)
tty_audit_add_data(ttyA)
tty_audit_push()
                               tty_audit_add_data(ttyB)
                               tty_audit_push()

Mixed canonical and non-canonical reads have similar races.
Signed-off-by: NPeter Hurley <peter@hurleysoftware.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

lock_task_sighand() is for situations where the struct task_struct*
may disappear while trying to deref the sighand; this never applies
to 'current'.
Signed-off-by: NPeter Hurley <peter@hurleysoftware.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

The tty audit buffer used to audit/record tty input is allocated on
the process's first call to tty_audit_add_data(), and not freed until
the process exits. On each call to tty_audit_add_data(), the current
tty is compared (by major:minor) with the last tty associated with
the audit buffer, and if the tty has changed the existing data is
logged to the audit log. The audit buffer is then re-associated with
the new tty.

Currently, the audit buffer is immediately associated with the tty;
however, the association must be re-checked when the buffer is locked
prior to copying the tty input. This extra step is always necessary,
since a concurrent read of a different tty by another thread of the
process may have used the buffer in between allocation and buffer
lock.

Rather than associate the audit buffer with the tty at allocation,
leave the buffer initially un-associated (null dev_t); simply let the
re-association check also perform the initial association.
Signed-off-by: NPeter Hurley <peter@hurleysoftware.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

The tty termios bits cannot change while n_tty_read() is in the
i/o loop; the termios_rwsem ensures mutual exclusion with termios
changes in n_tty_set_termios(). Check L_ICANON() directly and
eliminate icanon parameter.

NB: tty_audit_add_data() => tty_audit_buf_get() => tty_audit_buf_alloc()
is a single path; ie., tty_audit_buf_get() and tty_audit_buf_alloc()
have no other callers.
Signed-off-by: NPeter Hurley <peter@hurleysoftware.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

tty audit never logs pty master reads, but packet mode only works for
pty masters, so tty_audit_add_data() was never logging packet mode
anyway.

Don't audit packet mode data. As those are the lone call sites, remove
tty_put_user().
Signed-off-by: NPeter Hurley <peter@hurleysoftware.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

Reads from pty masters are not logged; early-out before taking
locks.
Signed-off-by: NPeter Hurley <peter@hurleysoftware.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

Access to tty->tty_files list is always per-tty, never for all ttys
simultaneously. Replace global tty_files_lock spinlock with per-tty
->files_lock. Initialize when the ->tty_files list is inited, in
alloc_tty_struct().
Signed-off-by: NPeter Hurley <peter@hurleysoftware.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

The TTY_DEBUG macro is not used; remove.
Signed-off-by: NPeter Hurley <peter@hurleysoftware.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

Move is_ignored() to drivers/tty/tty_io.c and re-declare in file
scope.
Signed-off-by: NPeter Hurley <peter@hurleysoftware.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

tty_read_raw_data() and tty_signal() no longer exist; remove
declarations.
Signed-off-by: NPeter Hurley <peter@hurleysoftware.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

Reduce global tty symbols; move and rename tty_ldisc_begin() as
n_tty_init() and redefine the N_TTY ldisc ops as file scope.
Signed-off-by: NPeter Hurley <peter@hurleysoftware.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

tty_mutex is a core, system-wide lock; there is no reason for any
code outside the tty core to have direct access.
Signed-off-by: NPeter Hurley <peter@hurleysoftware.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

tty_ldisc_setup() is race-free and can reference tty->ldisc without
snapshots.
Signed-off-by: NPeter Hurley <peter@hurleysoftware.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

The line discipline id is stored in the tty's termios; document the
implicit initial value of N_TTY.
Signed-off-by: NPeter Hurley <peter@hurleysoftware.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

Currently, when the tty is hungup, the ldisc is re-instanced; ie., the
current instance is destroyed and a new instance is created. The purpose
of this design was to guarantee a valid, open ldisc for the lifetime of
the tty.

However, now that tty buffers are owned by and have lifetime equivalent
to the tty_port (since v3.10), any data received immediately after the
ldisc is re-instanced may cause continued driver i/o operations
concurrently with the driver's hangup() operation. For drivers that
shutdown h/w on hangup, this is unexpected and usually bad. For example,
the serial core may free the xmit buffer page concurrently with an
in-progress write() operation (triggered by echo).

With the existing stable and robust ldisc reference handling, the
cleaned-up tty_reopen(), the straggling unsafe ldisc use cleaned up, and
the preparation to properly handle a NULL tty->ldisc, the ldisc instance
can be destroyed and only re-instanced when the tty is re-opened.

If the tty was opened as /dev/console or /dev/tty0, the original behavior
of re-instancing the ldisc is retained (the 'reinit' parameter to
tty_ldisc_hangup() is true). This is required since those file descriptors
are never hungup.

This patch has neglible impact on userspace; the tty file_operations ptr
is changed to point to the hungup file operations _before_ the ldisc
instance is destroyed, so only racing file operations might now retrieve
a NULL ldisc reference (which is simply handled as if the hungup file
operation had been called instead -- see "tty: Prepare for destroying
line discipline on hangup").

This resolves a long-standing FIXME and several crash reports.
Signed-off-by: NPeter Hurley <peter@hurleysoftware.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

At tty hangup, the line discipline instance is reinitialized by
closing the current ldisc instance and opening a new instance.
This operation is complicated by error recovery: if the attempt
to reinit the current line discipline fails, the line discipline
is reset to N_TTY (which should not but can fail).

Re-purpose tty_ldisc_reinit() to return a valid, open line discipline
instance, or otherwise, an error.
Signed-off-by: NPeter Hurley <peter@hurleysoftware.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

tty->ldisc is a ptr to struct tty_ldisc, but unfortunately 'ldisc' is
also used as a parameter or local name to refer to the line discipline
index value (ie, N_TTY, N_GSM, etc.); instead prefer the name used
by the line discipline registration/ref counting functions.
Signed-off-by: NPeter Hurley <peter@hurleysoftware.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

In preparation for destroying the line discipline instance on hangup,
move tty_ldisc_kill() to eliminate needless forward declarations.
No functional change.
Signed-off-by: NPeter Hurley <peter@hurleysoftware.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

In preparation of destroying line discipline on hangup, fix
ldisc core operations to properly handle when the tty's ldisc is
NULL.
Signed-off-by: NPeter Hurley <peter@hurleysoftware.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

tty file_operations (read/write/ioctl) wait for the ldisc reference
indefinitely (until ldisc lifetime events, such as hangup or TIOCSETD,
finish). Since hangup now destroys the ldisc and does not instance
another copy, file_operations must now be prepared to receive a NULL
ldisc reference from tty_ldisc_ref_wait():

CPU 0                                   CPU 1
-----                                   -----
(*f_op->read)() => tty_read()
                                        __tty_hangup()
                                        ...
                                        f_op = &hung_up_tty_fops;
                                        ...
                                        tty_ldisc_hangup()
                                           tty_ldisc_lock()
                                           tty_ldisc_kill()
                                              tty->ldisc = NULL
                                           tty_ldisc_unlock()
ld = tty_ldisc_ref_wait()
/* ld == NULL */

Instead, the action taken now is to return the same value as if the
tty had been hungup a moment earlier:

CPU 0                                   CPU 1
-----                                   -----
                                        __tty_hangup()
                                        ...
                                        f_op = &hung_up_tty_fops;
(*f_op->read)() => hung_up_tty_read()
return 0;
                                        ...
                                        tty_ldisc_hangup()
                                           tty_ldisc_lock()
                                           tty_ldisc_kill()
                                              tty->ldisc = NULL
                                           tty_ldisc_unlock()
Signed-off-by: NPeter Hurley <peter@hurleysoftware.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

tty_ldisc_kill() sets tty->ldisc to NULL; _not_ to N_TTY with a valid
but unopened ldisc. Fix function header documentation.
Signed-off-by: NPeter Hurley <peter@hurleysoftware.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

tty_ldisc_get() returns ERR_PTR() values if unsuccessful, not NULL;
fix function header documentation.
Signed-off-by: NPeter Hurley <peter@hurleysoftware.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

After the ldisc is released, but before the tty is destroyed, the termios
is saved (in tty_free_termios()); this termios is restored if a new
tty is created on next open(). However, the line discipline is always
reset, which is not obvious in the current method. Instead, reset
as part of the restore.

Restore the original line discipline, which may not have been N_TTY.
Signed-off-by: NPeter Hurley <peter@hurleysoftware.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

The chars_in_buffer() line discipline method serves no functional
purpose, other than as a (dubious) debugging aid for mostly bit-rotting
drivers. Despite being documented as an optional method, every caller
is unconditionally executed (although conditionally compiled).
Furthermore, direct tty->ldisc access without an ldisc ref is unsafe.
Lastly, N_TTY's chars_in_buffer() has warned of removal since 3.12.
Signed-off-by: NPeter Hurley <peter@hurleysoftware.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>