Update the documentation about the interaction between the suspend (S3) call
path and the CPU hotplug infrastructure.
This patch focusses only on the activities of the freezer, cpu hotplug and
the notifications involved. It outlines how regular CPU hotplug differs from
the way it is invoked during suspend and also tries to explain the locking
involved. In addition to that, it discusses the issue of microcode update
during CPU hotplug operations.
Signed-off-by: NSrivatsa S. Bhat <srivatsa.bhat@linux.vnet.ibm.com>
Signed-off-by: NRafael J. Wysocki <rjw@sisk.pl>

As noted by a user in https://bugzilla.redhat.com/show_bug.cgi?id=641789
The Sony VGN-FW21E also needs the nonvs by default workaround added.
Signed-off-by: NDave Jones <davej@redhat.com>
Signed-off-by: NRafael J. Wysocki <rjw@sisk.pl>

* pm-domains:
  ARM: mach-shmobile: sh7372 A4R support (v4)
  ARM: mach-shmobile: sh7372 A3SP support (v4)
  PM / Sleep: Mark devices involved in wakeup signaling during suspend

This change adds support for the sh7372 A4R power domain.

The sh7372 A4R hardware power domain contains the
SH CPU Core and a set of I/O devices including
multimedia accelerators and I2C controllers.

One special case about A4R is the INTCS interrupt
controller that needs to be saved and restored to
keep working as expected. Also the LCDC hardware
blocks are in a different hardware power domain
but have their IRQs routed only through INTCS. So
as long as LCDCs are active we cannot power down
INTCS because that would risk losing interrupts.
Signed-off-by: NMagnus Damm <damm@opensource.se>
Signed-off-by: NRafael J. Wysocki <rjw@sisk.pl>

This change adds support for the sh7372 A3SP power domain.

The sh7372 A3SP hardware power domain contains a
wide range of I/O devices. The list of I/O devices
include SCIF serial ports, DMA Engine hardware,
SD and MMC controller hardware, USB controllers
and I2C master controllers.

This patch adds the A3SP low level code which
powers the hardware power domain on and off. It
also ties in platform devices to the pm domain
support code.

It is worth noting that the serial console is
hooked up to SCIFA0 on most sh7372 boards, and
the SCIFA0 port is included in the A3SP hardware
power domain. For this reason we cannot output
debug messages from the low level power control
code in the case of A3SP.

QoS support is needed in drivers before we can
enable the A3SP power control on the fly.
Signed-off-by: NMagnus Damm <damm@opensource.se>
Signed-off-by: NRafael J. Wysocki <rjw@sisk.pl>

The generic PM domains code in drivers/base/power/domain.c has
to avoid powering off domains that provide power to wakeup devices
during system suspend.  Currently, however, this only works for
wakeup devices directly belonging to the given domain and not for
their children (or the children of their children and so on).
Thus, if there's a wakeup device whose parent belongs to a power
domain handled by the generic PM domains code, the domain will be
powered off during system suspend preventing the device from
signaling wakeup.

To address this problem introduce a device flag, power.wakeup_path,
that will be set during system suspend for all wakeup devices,
their parents, the parents of their parents and so on.  This way,
all wakeup paths in the device hierarchy will be marked and the
generic PM domains code will only need to avoid powering off
domains containing devices whose power.wakeup_path is set.
Signed-off-by: NRafael J. Wysocki <rjw@sisk.pl>

Use threads for LZO compression/decompression on hibernate/thaw.
Improve buffering on hibernate/thaw.
Calculate/verify CRC32 of the image pages on hibernate/thaw.

In my testing, this improved write/read speed by a factor of about two.
Signed-off-by: NBojan Smojver <bojan@rexursive.com>
Signed-off-by: NRafael J. Wysocki <rjw@sisk.pl>

Static and extern variables in kernel/power/hibernate.c need not be
initialized to 0 explicitly, so remove those initializations.

[rjw: Modified subject, added changelog.]
Signed-off-by: NBarry Song <Baohua.Song@csr.com>
Signed-off-by: NRafael J. Wysocki <rjw@sisk.pl>

TASK_KILLABLE is often used to put tasks to sleep for quite some time.
One of the most common uses is to put tasks to sleep while waiting for
replies from a server on a networked filesystem (such as CIFS or NFS).

Unfortunately, fake_signal_wake_up does not currently wake up tasks
that are sleeping in TASK_KILLABLE state. This means that even if the
code were in place to allow them to freeze while in this sleep, it
wouldn't work anyway.

This patch changes this function to wake tasks in this state as well.
This should be harmless -- if the code doing the sleeping doesn't have
handling to deal with freezer events, it should just go back to sleep.
If it does, then this will allow that code to do the right thing.
Signed-off-by: NJeff Layton <jlayton@redhat.com>
Signed-off-by: NRafael J. Wysocki <rjw@sisk.pl>

Patch "PM / Hibernate: Add resumewait param to support MMC-like
devices as resume file" added the resumewait kernel command line
option.  The present patch adds resumedelay so that
resumewait/delay were analogous to rootwait/delay.

[rjw: Modified the subject and changelog slightly.]
Signed-off-by: NBarry Song <baohua.song@csr.com>
Acked-by: NPavel Machek <pavel@ucw.cz>
Signed-off-by: NRafael J. Wysocki <rjw@sisk.pl>

Change the linux-pm list address in MAINTAINERS, as it has been moved
to vger.kernel.org now.
Signed-off-by: NWANG Cong <xiyou.wangcong@gmail.com>
Signed-off-by: NRafael J. Wysocki <rjw@sisk.pl>

Sony Vaio VGN-FW520F does not resume correctly without
acpi_sleep=nonvs, so add it to the ACPI sleep blacklist.

References: https://bugzilla.kernel.org/show_bug.cgi?id=16396#c86Signed-off-by: NBogdan Radulescu <bogdan@nimblex.net>
Signed-off-by: NRafael J. Wysocki <rjw@sisk.pl>

Apparently, Sony Vaio VGN-SR26GN_P does not resume correctly without
acpi_sleep=nonvs, so add it to the ACPI sleep blacklist.

References: https://bugzilla.kernel.org/show_bug.cgi?id=16396#c91Signed-off-by: NRafael J. Wysocki <rjw@sisk.pl>

Some devices like MMC are async detected very slow. For example,
drivers/mmc/host/sdhci.c launches a 200ms delayed work to detect
MMC partitions then add disk.

We have wait_for_device_probe() and scsi_complete_async_scans()
before calling swsusp_check(), but it is not enough to wait for MMC.

This patch adds resumewait kernel param just like rootwait so
that we have enough time to wait until MMC is ready. The difference is
that we wait for resume partition whereas rootwait waits for rootfs
partition (which may be on a different device).

This patch will make hibernation support many embedded products
without SCSI devices, but with devices like MMC.

[rjw: Modified the changelog slightly.]
Signed-off-by: NBarry Song <Baohua.Song@csr.com>
Reviewed-by: NValdis Kletnieks <valdis.kletnieks@vt.edu>
Signed-off-by: NRafael J. Wysocki <rjw@sisk.pl>

Fix a typo in a function name in the kerneldoc comment next to
resume_target_kernel().

[rjw: Changed the subject slightly, added the changelog.]
Signed-off-by: NBarry Song <Baohua.Song@csr.com>
Signed-off-by: NRafael J. Wysocki <rjw@sisk.pl>

There is a problem with the current ordering of hibernate code which
leads to deadlocks in some filesystems' memory shrinkers.  Namely,
some filesystems use freezable kernel threads that are inactive when
the hibernate memory preallocation is carried out.  Those same
filesystems use memory shrinkers that may be triggered by the
hibernate memory preallocation.  If those memory shrinkers wait for
the frozen kernel threads, the hibernate process deadlocks (this
happens with XFS, for one example).

Apparently, it is not technically viable to redesign the filesystems
in question to avoid the situation described above, so the only
possible solution of this issue is to defer the freezing of kernel
threads until the hibernate memory preallocation is done, which is
implemented by this change.

Unfortunately, this requires the memory preallocation to be done
before the "prepare" stage of device freeze, so after this change the
only way drivers can allocate additional memory for their freeze
routines in a clean way is to use PM notifiers.
Reported-by: NChristoph <cr2005@u-club.de>
Signed-off-by: NRafael J. Wysocki <rjw@sisk.pl>

This patch (as1485) documents a change to the kernel's default wakeup
policy.  Devices that forward wakeup requests between buses should be
enabled for wakeup by default.
Signed-off-by: NAlan Stern <stern@rowland.harvard.edu>
Signed-off-by: NRafael J. Wysocki <rjw@sisk.pl>

Introduce the config option CONFIG_VT_CONSOLE_SLEEP in order to cleanup
the #if defined ugliness for the vt suspend support functions. Note that
CONFIG_VT_CONSOLE is already dependant on CONFIG_VT.

The function pm_set_vt_switch is actually dependant on CONFIG_VT and not
CONFIG_PM_SLEEP. This fixes a compile error when CONFIG_PM_SLEEP is
not set:

drivers/tty/vt/vt_ioctl.c:1794: error: redefinition of 'pm_set_vt_switch'
include/linux/suspend.h:17: error: previous definition of 'pm_set_vt_switch' was here

Also, remove the incorrect path from the comment in console.c.

[rjw: Replaced #if defined() with #ifdef in suspend.h.]
Signed-off-by: NH Hartley Sweeten <hsweeten@visionengravers.com>
Acked-by: NArnd Bergmann <arnd@arndb.de>
Signed-off-by: NRafael J. Wysocki <rjw@sisk.pl>

In enter_state() we use "state" as an offset for the pm_states[]
array.  The pm_states[] array only has PM_SUSPEND_MAX elements so
this test is off by one.
Signed-off-by: NDan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: NRafael J. Wysocki <rjw@sisk.pl>
Cc: stable@kernel.org

For s390 there is one additional byte associated with each page,
the storage key. This byte contains the referenced and changed
bits and needs to be included into the hibernation image.
If the storage keys are not restored to their previous state all
original pages would appear to be dirty. This can cause
inconsistencies e.g. with read-only filesystems.
Signed-off-by: NMartin Schwidefsky <schwidefsky@de.ibm.com>
Signed-off-by: NRafael J. Wysocki <rjw@sisk.pl>

Suspend statistics should depend on CONFIG_PM_SLEEP, so make that
happen.
Signed-off-by: NRafael J. Wysocki <rjw@sisk.pl>

Record S3 failure time about each reason and the latest two failed
devices' names in S3 progress.
We can check it through 'suspend_stats' entry in debugfs.

The motivation of the patch:

We are enabling power features on Medfield. Comparing with PC/notebook,
a mobile enters/exits suspend-2-ram (we call it s3 on Medfield) far
more frequently. If it can't enter suspend-2-ram in time, the power
might be used up soon.

We often find sometimes, a device suspend fails. Then, system retries
s3 over and over again. As display is off, testers and developers
don't know what happens.

Some testers and developers complain they don't know if system
tries suspend-2-ram, and what device fails to suspend. They need
such info for a quick check. The patch adds suspend_stats under
debugfs for users to check suspend to RAM statistics quickly.

If not using this patch, we have other methods to get info about
what device fails. One is to turn on  CONFIG_PM_DEBUG, but users
would get too much info and testers need recompile the system.

In addition, dynamic debug is another good tool to dump debug info.
But it still doesn't match our utilization scenario closely.
1) user need write a user space parser to process the syslog output;
2) Our testing scenario is we leave the mobile for at least hours.
   Then, check its status. No serial console available during the
   testing. One is because console would be suspended, and the other
   is serial console connecting with spi or HSU devices would consume
   power. These devices are powered off at suspend-2-ram.
Signed-off-by: NShuoX Liu <shuox.liu@intel.com>
Signed-off-by: NRafael J. Wysocki <rjw@sisk.pl>

* pm-runtime:
  PM / Runtime: Handle .runtime_suspend() failure correctly
  PM / Runtime: Fix kerneldoc comment for rpm_suspend()
  PM / Runtime: Update document about callbacks

If .runtime_suspend() returns -EAGAIN or -EBUSY, the device should
still be in ACTIVE state, so it is not necessary to send an idle
notification to its parent.  If .runtime_suspend() returns other
fatal failure, it doesn't make sense to send idle notification to
its parent.

Skip parent idle notification when failure is returned from
.runtime_suspend() and update comments in rpm_suspend() to reflect
that change.

[rjw: Modified the subject and changelog slightly.]
Signed-off-by: NMing Lei <ming.lei@canonical.com>
Signed-off-by: NRafael J. Wysocki <rjw@sisk.pl>

This patch fix kerneldoc comments for rpm_suspend():

 - 'Cancel a pending idle notification' should be put before, also
   should be changed to 'Cancel a pending idle notification,
   autosuspend or suspend'.

 - idle notification for the device after succeeding suspend has
   been removed, so update the comment accordingly.

[rjw: Modified the subject and changelog slightly.]
Signed-off-by: NMing Lei <ming.lei@canonical.com>
Signed-off-by: NRafael J. Wysocki <rjw@sisk.pl>

Support for device power domains has been introduced in
commit 9659cc06 (PM: Make
system-wide PM and runtime PM treat subsystems consistently),
also power domain callbacks will take precedence over subsystem ones
from commit 4d27e9dc(PM: Make
power domain callbacks take precedence over subsystem ones).

So update part of "Device Runtime PM Callbacks" in
Documentation/power/runtime_pm.txt.
Signed-off-by: NMing Lei <ming.lei@canonical.com>
Signed-off-by: NRafael J. Wysocki <rjw@sisk.pl>

* pm-devfreq:
  PM / devfreq: Add basic governors
  PM / devfreq: Add common sysfs interfaces
  PM: Introduce devfreq: generic DVFS framework with device-specific OPPs
  PM / OPP: Add OPP availability change notifier.

* pm-qos:
  PM / QoS: Update Documentation for the pm_qos and dev_pm_qos frameworks
  PM / QoS: Add function dev_pm_qos_read_value() (v3)
  PM QoS: Add global notification mechanism for device constraints
  PM QoS: Implement per-device PM QoS constraints
  PM QoS: Generalize and export constraints management code
  PM QoS: Reorganize data structs
  PM QoS: Code reorganization
  PM QoS: Minor clean-ups
  PM QoS: Move and rename the implementation files

* pm-domains:
  PM / Domains: Split device PM domain data into base and need_restore
  ARM: mach-shmobile: sh7372 sleep warning fixes
  ARM: mach-shmobile: sh7372 A3SM support
  ARM: mach-shmobile: sh7372 generic suspend/resume support
  PM / Domains: Preliminary support for devices with power.irq_safe set
  PM: Move clock-related definitions and headers to separate file
  PM / Domains: Use power.sybsys_data to reduce overhead
  PM: Reference counting of power.subsys_data
  PM: Introduce struct pm_subsys_data
  ARM / shmobile: Make A3RV be a subdomain of A4LC on SH7372
  PM / Domains: Rename argument of pm_genpd_add_subdomain()
  PM / Domains: Rename GPD_STATE_WAIT_PARENT to GPD_STATE_WAIT_MASTER
  PM / Domains: Allow generic PM domains to have multiple masters
  PM / Domains: Add "wait for parent" status for generic PM domains
  PM / Domains: Make pm_genpd_poweron() always survive parent removal
  PM / Domains: Do not take parent locks to modify subdomain counters
  PM / Domains: Implement subdomain counters as atomic fields

* pm-runtime:
  PM / Tracing: build rpm-traces.c only if CONFIG_PM_RUNTIME is set
  PM / Runtime: Replace dev_dbg() with trace_rpm_*()
  PM / Runtime: Introduce trace points for tracing rpm_* functions
  PM / Runtime: Don't run callbacks under lock for power.irq_safe set
  USB: Add wakeup info to debugging messages
  PM / Runtime: pm_runtime_idle() can be called in atomic context
  PM / Runtime: Add macro to test for runtime PM events
  PM / Runtime: Add might_sleep() to runtime PM functions

* git://github.com/davem330/net:
  net: fix typos in Documentation/networking/scaling.txt
  bridge: leave carrier on for empty bridge
  netfilter: Use proper rwlock init function
  tcp: properly update lost_cnt_hint during shifting
  tcp: properly handle md5sig_pool references
  macvlan/macvtap: Fix unicast between macvtap interfaces in bridge mode

In summary, this DMI quirk uses the _CRS info by default for the ASUS
M2V-MX SE by turning on `pci=use_crs` and is similar to the quirk
added by commit 2491762c ("x86/PCI: use host bridge _CRS info on
ASRock ALiveSATA2-GLAN") whose commit message should be read for further
information.

Since commit 3e3da00c ("x86/pci: AMD one chain system to use pci
read out res") Linux gives the following oops:

    parport0: PC-style at 0x378, irq 7 [PCSPP,TRISTATE]
    HDA Intel 0000:20:01.0: PCI INT A -> GSI 17 (level, low) -> IRQ 17
    HDA Intel 0000:20:01.0: setting latency timer to 64
    BUG: unable to handle kernel paging request at ffffc90011c08000
    IP: [<ffffffffa0578402>] azx_probe+0x3ad/0x86b [snd_hda_intel]
    PGD 13781a067 PUD 13781b067 PMD 1300ba067 PTE 800000fd00000173
    Oops: 0009 [#1] SMP
    last sysfs file: /sys/module/snd_pcm/initstate
    CPU 0
    Modules linked in: snd_hda_intel(+) snd_hda_codec snd_hwdep snd_pcm_oss snd_mixer_oss snd_pcm snd_seq_midi snd_rawmidi snd_seq_midi_event tpm_tis tpm snd_seq tpm_bios psmouse parport_pc snd_timer snd_seq_device parport processor evdev snd i2c_viapro thermal_sys amd64_edac_mod k8temp i2c_core soundcore shpchp pcspkr serio_raw asus_atk0110 pci_hotplug edac_core button snd_page_alloc edac_mce_amd ext3 jbd mbcache sha256_generic cryptd aes_x86_64 aes_generic cbc dm_crypt dm_mod raid1 md_mod usbhid hid sg sd_mod crc_t10dif sr_mod cdrom ata_generic uhci_hcd sata_via pata_via libata ehci_hcd usbcore scsi_mod via_rhine mii nls_base [last unloaded: scsi_wait_scan]
    Pid: 1153, comm: work_for_cpu Not tainted 2.6.37-1-amd64 #1 M2V-MX SE/System Product Name
    RIP: 0010:[<ffffffffa0578402>]  [<ffffffffa0578402>] azx_probe+0x3ad/0x86b [snd_hda_intel]
    RSP: 0018:ffff88013153fe50  EFLAGS: 00010286
    RAX: ffffc90011c08000 RBX: ffff88013029ec00 RCX: 0000000000000006
    RDX: 0000000000000000 RSI: 0000000000000246 RDI: 0000000000000246
    RBP: ffff88013341d000 R08: 0000000000000000 R09: 0000000000000040
    R10: 0000000000000286 R11: 0000000000003731 R12: ffff88013029c400
    R13: 0000000000000000 R14: 0000000000000000 R15: ffff88013341d090
    FS:  0000000000000000(0000) GS:ffff8800bfc00000(0000) knlGS:00000000f7610ab0
    CS:  0010 DS: 0000 ES: 0000 CR0: 000000008005003b
    CR2: ffffc90011c08000 CR3: 0000000132f57000 CR4: 00000000000006f0
    DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
    DR3: 0000000000000000 DR6: 00000000ffff0ff0 DR7: 0000000000000400
    Process work_for_cpu (pid: 1153, threadinfo ffff88013153e000, task ffff8801303c86c0)
    Stack:
     0000000000000005 ffffffff8123ad65 00000000000136c0 ffff88013029c400
     ffff8801303c8998 ffff88013341d000 ffff88013341d090 ffff8801322d9dc8
     ffff88013341d208 0000000000000000 0000000000000000 ffffffff811ad232
    Call Trace:
     [<ffffffff8123ad65>] ? __pm_runtime_set_status+0x162/0x186
     [<ffffffff811ad232>] ? local_pci_probe+0x49/0x92
     [<ffffffff8105afc5>] ? do_work_for_cpu+0x0/0x1b
     [<ffffffff8105afc5>] ? do_work_for_cpu+0x0/0x1b
     [<ffffffff8105afd0>] ? do_work_for_cpu+0xb/0x1b
     [<ffffffff8105fd3f>] ? kthread+0x7a/0x82
     [<ffffffff8100a824>] ? kernel_thread_helper+0x4/0x10
     [<ffffffff8105fcc5>] ? kthread+0x0/0x82
     [<ffffffff8100a820>] ? kernel_thread_helper+0x0/0x10
    Code: f4 01 00 00 ef 31 f6 48 89 df e8 29 dd ff ff 85 c0 0f 88 2b 03 00 00 48 89 ef e8 b4 39 c3 e0 8b 7b 40 e8 fc 9d b1 e0 48 8b 43 38 <66> 8b 10 66 89 14 24 8b 43 14 83 e8 03 83 f8 01 77 32 31 d2 be
    RIP  [<ffffffffa0578402>] azx_probe+0x3ad/0x86b [snd_hda_intel]
     RSP <ffff88013153fe50>
    CR2: ffffc90011c08000
    ---[ end trace 8d1f3ebc136437fd ]---

Trusting the ACPI _CRS information (`pci=use_crs`) fixes this problem.

    $ dmesg | grep -i crs # with the quirk
    PCI: Using host bridge windows from ACPI; if necessary, use "pci=nocrs" and report a bug

The match has to be against the DMI board entries though since the vendor entries are not populated.

    DMI: System manufacturer System Product Name/M2V-MX SE, BIOS 0304    10/30/2007

This quirk should be removed when `pci=use_crs` is enabled for machines
from 2006 or earlier or some other solution is implemented.

Using coreboot [1] with this board the problem does not exist but this
quirk also does not affect it either. To be safe though the check is
tightened to only take effect when the BIOS from American Megatrends is
used.

        15:13 < ruik> but coreboot does not need that
        15:13 < ruik> because i have there only one root bus
        15:13 < ruik> the audio is behind a bridge

        $ sudo dmidecode
        BIOS Information
                Vendor: American Megatrends Inc.
                Version: 0304
                Release Date: 10/30/2007

[1] http://www.coreboot.org/

Bugzilla: https://bugzilla.kernel.org/show_bug.cgi?id=30552

Cc: stable@kernel.org (2.6.34)
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: H. Peter Anvin <hpa@zytor.com>
Cc: x86@kernel.org
Signed-off-by: NPaul Menzel <paulepanter@users.sourceforge.net>
Signed-off-by: NBjorn Helgaas <bhelgaas@google.com>
Acked-by: NJesse Barnes <jbarnes@virtuousgeek.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

The second hunk fixes rps_sock_flow_table but has to re-wrap the paragraph.
Signed-off-by: NBenjamin Poirier <benjamin.poirier@gmail.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

This resolves a regression seen by some users of bridging.
Some users use the bridge like a dummy device.
They expect to be able to put an IPv6 address on the device
with no ports attached. Although there are better ways of doing
this, there is no reason to not allow it.

Note: the bridge still will reflect the state of ports in the
bridge if there are any added.
Signed-off-by: NStephen Hemminger <shemminger@vyatta.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

* 'for-linus' of http://people.redhat.com/agk/git/linux-dm:
  dm crypt: always disable discard_zeroes_data
  dm: raid fix write_mostly arg validation
  dm table: avoid crash if integrity profile changes
  dm: flakey fix corrupt_bio_byte error path

* 'for-linus' of git://neil.brown.name/md:
  md: Avoid waking up a thread after it has been freed.

Replace the open coded initialization with the init function.
Signed-off-by: NThomas Gleixner <tglx@linutronix.de>
Acked-by: NHans Schillstrom <hans.schillstrom@ericsson.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

* 'for-linus' of git://github.com/dtor/input:
  Input: wacom - revert "Cintiq 21UX2 does not have menu strips"

* git://bedivere.hansenpartnership.com/git/scsi-rc-fixes-2.6:
  [SCSI] libsas: fix panic when single phy is disabled on a wide port
  [SCSI] qla2xxx: Fix crash in qla2x00_abort_all_cmds() on unload

This reverts commit 71c86ce5.
The 21UX2 does have touchstrips, but they are in a somewhat-
hidden location.
Signed-off-by: NJason Gerecke <killertofu@gmail.com>
Acked-by: NPing Cheng <pinglinux@gmail.com>
Signed-off-by: NDmitry Torokhov <dtor@mail.ru>