platform/adaptation/renesas_rcar/renesas_kernel.git
9 years agonfsd: correctly define v4.2 support attributes
Christoph Hellwig [Sat, 8 Nov 2014 12:11:03 +0000 (13:11 +0100)]
nfsd: correctly define v4.2 support attributes

commit 6d0ba0432a5e10bc714ba9c5adc460e726e5fbb4 upstream.

Even when security labels are disabled we support at least the same
attributes as v4.1.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: J. Bruce Fields <bfields@redhat.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agort2x00: do not align payload on modern H/W
Stanislaw Gruszka [Tue, 11 Nov 2014 13:28:47 +0000 (14:28 +0100)]
rt2x00: do not align payload on modern H/W

commit cfd9167af14eb4ec21517a32911d460083ee3d59 upstream.

RT2800 and newer hardware require padding between header and payload if
header length is not multiple of 4.

For historical reasons we also align payload to to 4 bytes boundary, but
such alignment is not needed on modern H/W.

Patch fixes skb_under_panic problems reported from time to time:

https://bugzilla.kernel.org/show_bug.cgi?id=84911
https://bugzilla.kernel.org/show_bug.cgi?id=72471
http://marc.info/?l=linux-wireless&m=139108549530402&w=2
https://bugs.launchpad.net/ubuntu/+source/linux/+bug/1087591

Panic happened because we eat 4 bytes of skb headroom on each
(re)transmission when sending frame without the payload and the header
length not being multiple of 4 (i.e. QoS header has 26 bytes). On such
case because paylad_aling=2 is bigger than header_align=0 we increase
header_align by 4 bytes. To prevent that we could change the check to:

if (payload_length && payload_align > header_align)
header_align += 4;

but not aligning payload at all is more effective and alignment is not
really needed by H/W (that has been tested on OpenWrt project for few
years now).

Reported-and-tested-by: Antti S. Lankila <alankila@bel.fi>
Debugged-by: Antti S. Lankila <alankila@bel.fi>
Reported-by: Henrik Asp <solenskiner@gmail.com>
Originally-From: Helmut Schaa <helmut.schaa@googlemail.com>
Signed-off-by: Stanislaw Gruszka <sgruszka@redhat.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agocan: dev: avoid calling kfree_skb() from interrupt context
Thomas Körper [Fri, 31 Oct 2014 06:33:54 +0000 (07:33 +0100)]
can: dev: avoid calling kfree_skb() from interrupt context

commit 5247a589c24022ab34e780039cc8000c48f2035e upstream.

ikfree_skb() is Called in can_free_echo_skb(), which might be called from (TX
Error) interrupt, which triggers the folloing warning:

[ 1153.360705] ------------[ cut here ]------------
[ 1153.360715] WARNING: CPU: 0 PID: 31 at net/core/skbuff.c:563 skb_release_head_state+0xb9/0xd0()
[ 1153.360772] Call Trace:
[ 1153.360778]  [<c167906f>] dump_stack+0x41/0x52
[ 1153.360782]  [<c105bb7e>] warn_slowpath_common+0x7e/0xa0
[ 1153.360784]  [<c158b909>] ? skb_release_head_state+0xb9/0xd0
[ 1153.360786]  [<c158b909>] ? skb_release_head_state+0xb9/0xd0
[ 1153.360788]  [<c105bc42>] warn_slowpath_null+0x22/0x30
[ 1153.360791]  [<c158b909>] skb_release_head_state+0xb9/0xd0
[ 1153.360793]  [<c158be90>] skb_release_all+0x10/0x30
[ 1153.360795]  [<c158bf06>] kfree_skb+0x36/0x80
[ 1153.360799]  [<f8486938>] ? can_free_echo_skb+0x28/0x40 [can_dev]
[ 1153.360802]  [<f8486938>] can_free_echo_skb+0x28/0x40 [can_dev]
[ 1153.360805]  [<f849a12c>] esd_pci402_interrupt+0x34c/0x57a [esd402]
[ 1153.360809]  [<c10a75b5>] handle_irq_event_percpu+0x35/0x180
[ 1153.360811]  [<c10a7623>] ? handle_irq_event_percpu+0xa3/0x180
[ 1153.360813]  [<c10a7731>] handle_irq_event+0x31/0x50
[ 1153.360816]  [<c10a9c7f>] handle_fasteoi_irq+0x6f/0x120
[ 1153.360818]  [<c10a9c10>] ? handle_edge_irq+0x110/0x110
[ 1153.360822]  [<c1011b61>] handle_irq+0x71/0x90
[ 1153.360823]  <IRQ>  [<c168152c>] do_IRQ+0x3c/0xd0
[ 1153.360829]  [<c1680b6c>] common_interrupt+0x2c/0x34
[ 1153.360834]  [<c107d277>] ? finish_task_switch+0x47/0xf0
[ 1153.360836]  [<c167c27b>] __schedule+0x35b/0x7e0
[ 1153.360839]  [<c10a5334>] ? console_unlock+0x2c4/0x4d0
[ 1153.360842]  [<c13df500>] ? n_tty_receive_buf_common+0x890/0x890
[ 1153.360845]  [<c10707b6>] ? process_one_work+0x196/0x370
[ 1153.360847]  [<c167c723>] schedule+0x23/0x60
[ 1153.360849]  [<c1070de1>] worker_thread+0x161/0x460
[ 1153.360852]  [<c1090fcf>] ? __wake_up_locked+0x1f/0x30
[ 1153.360854]  [<c1070c80>] ? rescuer_thread+0x2f0/0x2f0
[ 1153.360856]  [<c1074f01>] kthread+0xa1/0xc0
[ 1153.360859]  [<c1680401>] ret_from_kernel_thread+0x21/0x30
[ 1153.360861]  [<c1074e60>] ? kthread_create_on_node+0x110/0x110
[ 1153.360863] ---[ end trace 5ff83639cbb74b35 ]---

This patch replaces the kfree_skb() by dev_kfree_skb_any().

Signed-off-by: Thomas Körper <thomas.koerper@esd.eu>
Signed-off-by: Marc Kleine-Budde <mkl@pengutronix.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agoscsi: add Intel Multi-Flex to scsi scan blacklist
Christian Sünkenberg [Tue, 18 Nov 2014 19:23:32 +0000 (20:23 +0100)]
scsi: add Intel Multi-Flex to scsi scan blacklist

commit 1899045510ff109980d9cc34e330fd8ca3631871 upstream.

Intel Multi-Flex LUNs choke on REPORT SUPPORTED OPERATION CODES
resulting in sd_mod hanging for several minutes on startup.
The issue was introduced with WRITE SAME discovery heuristics.

Fixes: 5db44863b6eb ("[SCSI] sd: Implement support for WRITE SAME")
Signed-off-by: Christian Sünkenberg <christian.suenkenberg@hfg-karlsruhe.de>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agovhost-scsi: Take configfs group dependency during VHOST_SCSI_SET_ENDPOINT
Nicholas Bellinger [Wed, 8 Oct 2014 06:19:20 +0000 (06:19 +0000)]
vhost-scsi: Take configfs group dependency during VHOST_SCSI_SET_ENDPOINT

commit ab8edab132829b26dd13db6caca3c242cce35dc1 upstream.

This patch addresses a bug where individual vhost-scsi configfs endpoint
groups can be removed from below while active exports to QEMU userspace
still exist, resulting in an OOPs.

It adds a configfs_depend_item() in vhost_scsi_set_endpoint() to obtain
an explicit dependency on se_tpg->tpg_group in order to prevent individual
vhost-scsi WWPN endpoints from being released via normal configfs methods
while an QEMU ioctl reference still exists.

Also, add matching configfs_undepend_item() in vhost_scsi_clear_endpoint()
to release the dependency, once QEMU's reference to the individual group
at /sys/kernel/config/target/vhost/$WWPN/$TPGT is released.

(Fix up vhost_scsi_clear_endpoint() error path - DanC)

Cc: Michael S. Tsirkin <mst@redhat.com>
Cc: Paolo Bonzini <pbonzini@redhat.com>
Cc: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Nicholas Bellinger <nab@linux-iscsi.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agospi: dw: Fix dynamic speed change.
Thor Thayer [Thu, 6 Nov 2014 19:54:27 +0000 (13:54 -0600)]
spi: dw: Fix dynamic speed change.

commit 0a8727e69778683495058852f783eeda141a754e upstream.

An IOCTL call that calls spi_setup() and then dw_spi_setup() will
overwrite the persisted last transfer speed. On each transfer, the
SPI speed is compared to the last transfer speed to determine if the
clock divider registers need to be updated (did the speed change?).
This bug was observed with the spidev driver using spi-config to
update the max transfer speed.

This fix: Don't overwrite the persisted last transaction clock speed
when updating the SPI parameters in dw_spi_setup(). On the next
transaction, the new speed won't match the persisted last speed
and the hardware registers will be updated.
On initialization, the persisted last transaction clock
speed will be 0 but will be updated after the first SPI
transaction.

Move zeroed clock divider check into clock change test because
chip->clk_div is zero on startup and would cause a divide-by-zero
error. The calculation was wrong as well (can't support odd #).

Reported-by: Vlastimil Setka <setka@vsis.cz>
Signed-off-by: Vlastimil Setka <setka@vsis.cz>
Signed-off-by: Thor Thayer <tthayer@opensource.altera.com>
Signed-off-by: Mark Brown <broonie@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agoiser-target: Disable TX completion interrupt coalescing
Nicholas Bellinger [Sun, 5 Oct 2014 09:13:03 +0000 (02:13 -0700)]
iser-target: Disable TX completion interrupt coalescing

commit 0d0f660d882c1c02748ced13966a2413aa5d6cc2 upstream.

This patch explicitly disables TX completion interrupt coalescing logic
in isert_put_response() and isert_put_datain() that was originally added
as an efficiency optimization in commit 95b60f07.

It has been reported that this change can trigger ABORT_TASK timeouts
under certain small block workloads, where disabling coalescing was
required for stability.  According to Sagi, this doesn't impact
overall performance, so go ahead and disable it for now.

Reported-by: Moussa Ba <moussaba@micron.com>
Reported-by: Sagi Grimberg <sagig@dev.mellanox.co.il>
Signed-off-by: Nicholas Bellinger <nab@linux-iscsi.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agoTarget/iser: Fix a wrong dereference in case discovery session is over iser
Sagi Grimberg [Tue, 10 Jun 2014 10:41:41 +0000 (13:41 +0300)]
Target/iser: Fix a wrong dereference in case discovery session is over iser

commit e0546fc1ba66c90cb38a5764357366267d3e58e4 upstream.

In case the discovery session is carried over iser, we can't
access the assumed network portal since the default portal is
used. In this case we don't really need to allocate the fastreg
pool, just prepare to the text pdu that will follow.

Signed-off-by: Sagi Grimberg <sagig@mellanox.com>
Reported-by: Alex Tabachnik <alext@mellanox.com>
Signed-off-by: Nicholas Bellinger <nab@linux-iscsi.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agoiser-target: Handle DEVICE_REMOVAL event on network portal listener correctly
Sagi Grimberg [Tue, 28 Oct 2014 20:45:03 +0000 (13:45 -0700)]
iser-target: Handle DEVICE_REMOVAL event on network portal listener correctly

commit 3b726ae2de02a406cc91903f80132daee37b6f1b upstream.

In this case the cm_id->context is the isert_np, and the cm_id->qp
is NULL, so use that to distinct the cases.

Since we don't expect any other events on this cm_id we can
just return -1 for explicit termination of the cm_id by the
cma layer.

Signed-off-by: Sagi Grimberg <sagig@mellanox.com>
Signed-off-by: Nicholas Bellinger <nab@linux-iscsi.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agotarget: Don't call TFO->write_pending if data_length == 0
Roland Dreier [Tue, 14 Oct 2014 21:16:24 +0000 (14:16 -0700)]
target: Don't call TFO->write_pending if data_length == 0

commit 885e7b0e181c14e4d0ddd26c688bad2b84c1ada9 upstream.

If an initiator sends a zero-length command (e.g. TEST UNIT READY) but
sets the transfer direction in the transport layer to indicate a
data-out phase, we still shouldn't try to transfer data.  At best it's
a NOP, and depending on the transport, we might crash on an
uninitialized sg list.

Reported-by: Craig Watson <craig.watson@vanguard-rugged.com>
Signed-off-by: Roland Dreier <roland@purestorage.com>
Signed-off-by: Nicholas Bellinger <nab@linux-iscsi.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agosrp-target: Retry when QP creation fails with ENOMEM
Bart Van Assche [Sun, 19 Oct 2014 15:05:33 +0000 (18:05 +0300)]
srp-target: Retry when QP creation fails with ENOMEM

commit ab477c1ff5e0a744c072404bf7db51bfe1f05b6e upstream.

It is not guaranteed to that srp_sq_size is supported
by the HCA. So if we failed to create the QP with ENOMEM,
try with a smaller srp_sq_size. Keep it up until we hit
MIN_SRPT_SQ_SIZE, then fail the connection.

Reported-by: Mark Lehrer <lehrer@gmail.com>
Signed-off-by: Bart Van Assche <bvanassche@acm.org>
Signed-off-by: Sagi Grimberg <sagig@mellanox.com>
Signed-off-by: Nicholas Bellinger <nab@linux-iscsi.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agoInput: xpad - use proper endpoint type
Greg Kroah-Hartman [Tue, 25 Nov 2014 08:38:17 +0000 (00:38 -0800)]
Input: xpad - use proper endpoint type

commit a1f9a4072655843fc03186acbad65990cc05dd2d upstream.

The xpad wireless endpoint is not a bulk endpoint on my devices, but
rather an interrupt one, so the USB core complains when it is submitted.
I'm guessing that the author really did mean that this should be an
interrupt urb, but as there are a zillion different xpad devices out
there, let's cover out bases and handle both bulk and interrupt
endpoints just as easily.

Signed-off-by: "Pierre-Loup A. Griffais" <pgriffais@valvesoftware.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Signed-off-by: Dmitry Torokhov <dmitry.torokhov@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agoInput: synaptics - adjust min/max on Thinkpad E540
Ben Sagal [Mon, 17 Nov 2014 01:23:40 +0000 (17:23 -0800)]
Input: synaptics - adjust min/max on Thinkpad E540

commit bce4f9e764c36bc35dd5c9cf9e057c09f422397d upstream.

The LEN2006 Synaptics touchpad (as found in Thinkpad E540) returns wrong
min max values.

touchpad-edge-detector output:
>  Touchpad SynPS/2 Synaptics TouchPad on /dev/input/event6
>  Move one finger around the touchpad to detect the actual edges
>  Kernel says:    x [1472..5674], y [1408..4684]
>  Touchpad sends: x [1264..5675], y [1171..4688]

Bugzilla: https://bugzilla.kernel.org/show_bug.cgi?id=88211
Signed-off-by: Binyamin Sagal <bensagal@gmail.com>
Signed-off-by: Dmitry Torokhov <dmitry.torokhov@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agoARM: 8226/1: cacheflush: get rid of restarting block
Vladimir Murzin [Thu, 27 Nov 2014 10:39:04 +0000 (11:39 +0100)]
ARM: 8226/1: cacheflush: get rid of restarting block

commit 3f4aa45ceea5789a4aade536acc27f2e0d3da5e1 upstream.

We cannot restart cacheflush safely if a process provides user-defined
signal handler and signal is pending. In this case -EINTR is returned
and it is expected that process re-invokes syscall. However, there are
a few problems with that:
 * looks like nobody bothers checking return value from cacheflush
 * but if it did, we don't provide the restart address for that, so the
   process has to use the same range again
 * ...and again, what might lead to looping forever

So, remove cacheflush restarting code and terminate cache flushing
as early as fatal signal is pending.

Reported-by: Chanho Min <chanho.min@lge.com>
Signed-off-by: Vladimir Murzin <vladimir.murzin@arm.com>
Acked-by: Will Deacon <will.deacon@arm.com>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agoARM: 8222/1: mvebu: enable strex backoff delay
Thomas Petazzoni [Tue, 25 Nov 2014 17:43:15 +0000 (18:43 +0100)]
ARM: 8222/1: mvebu: enable strex backoff delay

commit 995ab5189d1d7264e79e665dfa032a19b3ac646e upstream.

Under extremely rare conditions, in an MPCore node consisting of at
least 3 CPUs, two CPUs trying to perform a STREX to data on the same
shared cache line can enter a livelock situation.

This patch enables the HW mechanism that overcomes the bug. This fixes
the incorrect setup of the STREX backoff delay bit due to a wrong
description in the specification.

Note that enabling the STREX backoff delay mechanism is done by
leaving the bit *cleared*, while the bit was currently being set by
the proc-v7.S code.

[Thomas: adapt to latest mainline, slightly reword the commit log, add
stable markers.]

Fixes: de4901933f6d ("arm: mm: Add support for PJ4B cpu and init routines")

Signed-off-by: Nadav Haklai <nadavh@marvell.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Acked-by: Gregory CLEMENT <gregory.clement@free-electrons.com>
Acked-by: Jason Cooper <jason@lakedaemon.net>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agoARM: 8216/1: xscale: correct auxiliary register in suspend/resume
Dmitry Eremin-Solenikov [Fri, 21 Nov 2014 14:29:00 +0000 (15:29 +0100)]
ARM: 8216/1: xscale: correct auxiliary register in suspend/resume

commit ef59a20ba375aeb97b3150a118318884743452a8 upstream.

According to the manuals I have, XScale auxiliary register should be
reached with opc_2 = 1 instead of crn = 1. cpu_xscale_proc_init
correctly uses c1, c0, 1 arguments, but cpu_xscale_do_suspend and
cpu_xscale_do_resume use c1, c1, 0. Correct suspend/resume functions to
also use c1, c0, 1.

The issue was primarily noticed thanks to qemu reporing "unsupported
instruction" on the pxa suspend path. Confirmed in PXA210/250 and PXA255
XScale Core manuals and in PXA270 and PXA320 Developers Guides.

Harware tested by me on tosa (pxa255). Robert confirmed on pxa270 board.

Tested-by: Robert Jarzmik <robert.jarzmik@free.fr>
Signed-off-by: Dmitry Eremin-Solenikov <dbaryshkov@gmail.com>
Acked-by: Robert Jarzmik <robert.jarzmik@free.fr>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agoof/irq: Drop obsolete 'interrupts' vs 'interrupts-extended' text
Bjorn Helgaas [Sat, 1 Nov 2014 23:35:31 +0000 (17:35 -0600)]
of/irq: Drop obsolete 'interrupts' vs 'interrupts-extended' text

commit 66865de4314caca30598244b86817e774c188afa upstream.

a9ecdc0fdc54 ("of/irq: Fix lookup to use 'interrupts-extended' property
first") updated the description to say that:

  - Both 'interrupts' and 'interrupts-extended' may be present
  - Software should prefer 'interrupts-extended'
  - Software that doesn't comprehend 'interrupts-extended' may use
    'interrupts'

But there is still a paragraph at the end that prohibits having both and
says 'interrupts' should be preferred.

Remove the contradictory text.

Fixes: a9ecdc0fdc54 ("of/irq: Fix lookup to use 'interrupts-extended' property first")
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Acked-by: Brian Norris <computersforpeace@gmail.com>
Acked-by: Mark Rutland <mark.rutland@arm.com>
Signed-off-by: Rob Herring <robh@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agoaio: fix uncorrent dirty pages accouting when truncating AIO ring buffer
Gu Zheng [Thu, 6 Nov 2014 09:46:21 +0000 (17:46 +0800)]
aio: fix uncorrent dirty pages accouting when truncating AIO ring buffer

commit 835f252c6debd204fcd607c79975089b1ecd3472 upstream.

https://bugzilla.kernel.org/show_bug.cgi?id=86831

Markus reported that when shutting down mysqld (with AIO support,
on a ext3 formatted Harddrive) leads to a negative number of dirty pages
(underrun to the counter). The negative number results in a drastic reduction
of the write performance because the page cache is not used, because the kernel
thinks it is still 2 ^ 32 dirty pages open.

Add a warn trace in __dec_zone_state will catch this easily:

static inline void __dec_zone_state(struct zone *zone, enum
zone_stat_item item)
{
     atomic_long_dec(&zone->vm_stat[item]);
+    WARN_ON_ONCE(item == NR_FILE_DIRTY &&
atomic_long_read(&zone->vm_stat[item]) < 0);
     atomic_long_dec(&vm_stat[item]);
}

[   21.341632] ------------[ cut here ]------------
[   21.346294] WARNING: CPU: 0 PID: 309 at include/linux/vmstat.h:242
cancel_dirty_page+0x164/0x224()
[   21.355296] Modules linked in: wutbox_cp sata_mv
[   21.359968] CPU: 0 PID: 309 Comm: kworker/0:1 Not tainted 3.14.21-WuT #80
[   21.366793] Workqueue: events free_ioctx
[   21.370760] [<c0016a64>] (unwind_backtrace) from [<c0012f88>]
(show_stack+0x20/0x24)
[   21.378562] [<c0012f88>] (show_stack) from [<c03f8ccc>]
(dump_stack+0x24/0x28)
[   21.385840] [<c03f8ccc>] (dump_stack) from [<c0023ae4>]
(warn_slowpath_common+0x84/0x9c)
[   21.393976] [<c0023ae4>] (warn_slowpath_common) from [<c0023bb8>]
(warn_slowpath_null+0x2c/0x34)
[   21.402800] [<c0023bb8>] (warn_slowpath_null) from [<c00c0688>]
(cancel_dirty_page+0x164/0x224)
[   21.411524] [<c00c0688>] (cancel_dirty_page) from [<c00c080c>]
(truncate_inode_page+0x8c/0x158)
[   21.420272] [<c00c080c>] (truncate_inode_page) from [<c00c0a94>]
(truncate_inode_pages_range+0x11c/0x53c)
[   21.429890] [<c00c0a94>] (truncate_inode_pages_range) from
[<c00c0f6c>] (truncate_pagecache+0x88/0xac)
[   21.439252] [<c00c0f6c>] (truncate_pagecache) from [<c00c0fec>]
(truncate_setsize+0x5c/0x74)
[   21.447731] [<c00c0fec>] (truncate_setsize) from [<c013b3a8>]
(put_aio_ring_file.isra.14+0x34/0x90)
[   21.456826] [<c013b3a8>] (put_aio_ring_file.isra.14) from
[<c013b424>] (aio_free_ring+0x20/0xcc)
[   21.465660] [<c013b424>] (aio_free_ring) from [<c013b4f4>]
(free_ioctx+0x24/0x44)
[   21.473190] [<c013b4f4>] (free_ioctx) from [<c003d8d8>]
(process_one_work+0x134/0x47c)
[   21.481132] [<c003d8d8>] (process_one_work) from [<c003e988>]
(worker_thread+0x130/0x414)
[   21.489350] [<c003e988>] (worker_thread) from [<c00448ac>]
(kthread+0xd4/0xec)
[   21.496621] [<c00448ac>] (kthread) from [<c000ec18>]
(ret_from_fork+0x14/0x20)
[   21.503884] ---[ end trace 79c4bf42c038c9a1 ]---

The cause is that we set the aio ring file pages as *DIRTY* via SetPageDirty
(bypasses the VFS dirty pages increment) when init, and aio fs uses
*default_backing_dev_info* as the backing dev, which does not disable
the dirty pages accounting capability.
So truncating aio ring file will contribute to accounting dirty pages (VFS
dirty pages decrement), then error occurs.

The original goal is keeping these pages in memory (can not be reclaimed
or swapped) in life-time via marking it dirty. But thinking more, we have
already pinned pages via elevating the page's refcount, which can already
achieve the goal, so the SetPageDirty seems unnecessary.

In order to fix the issue, using the __set_page_dirty_no_writeback instead
of the nop .set_page_dirty, and dropped the SetPageDirty (don't manually
set the dirty flags, don't disable set_page_dirty(), rely on default behaviour).

With the above change, the dirty pages accounting can work well. But as we
known, aio fs is an anonymous one, which should never cause any real write-back,
we can ignore the dirty pages (write back) accounting by disabling the dirty
pages (write back) accounting capability. So we introduce an aio private
backing dev info (disabled the ACCT_DIRTY/WRITEBACK/ACCT_WB capabilities) to
replace the default one.

Reported-by: Markus Königshaus <m.koenigshaus@wut.de>
Signed-off-by: Gu Zheng <guz.fnst@cn.fujitsu.com>
Acked-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Benjamin LaHaise <bcrl@kvack.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agoALSA: usb-audio: Add ctrl message delay quirk for Marantz/Denon devices
Jurgen Kramer [Sat, 15 Nov 2014 13:01:21 +0000 (14:01 +0100)]
ALSA: usb-audio: Add ctrl message delay quirk for Marantz/Denon devices

commit 6e84a8d7ac3ba246ef44e313e92bc16a1da1b04a upstream.

This patch adds a USB control message delay quirk for a few specific Marantz/Denon
devices. Without the delay the DACs will not work properly and produces the
following type of messages:

Nov 15 10:09:21 orwell kernel: [   91.342880] usb 3-13: clock source 41 is not valid, cannot use
Nov 15 10:09:21 orwell kernel: [   91.343775] usb 3-13: clock source 41 is not valid, cannot use

There are likely other Marantz/Denon devices using the same USB module which exhibit the
same problems. But as this cannot be verified I limited the patch to the devices
I could test.

The following two devices are covered by this path:
- Marantz SA-14S1
- Marantz HD-DAC1

Signed-off-by: Jurgen Kramer <gtmkramer@xs4all.nl>
Signed-off-by: Takashi Iwai <tiwai@suse.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agocan: esd_usb2: fix memory leak on disconnect
Alexey Khoroshilov [Fri, 10 Oct 2014 20:31:07 +0000 (00:31 +0400)]
can: esd_usb2: fix memory leak on disconnect

commit efbd50d2f62fc1f69a3dcd153e63ba28cc8eb27f upstream.

It seems struct esd_usb2 dev is not deallocated on disconnect. The patch adds
the missing deallocation.

Found by Linux Driver Verification project (linuxtesting.org).

Signed-off-by: Alexey Khoroshilov <khoroshilov@ispras.ru>
Acked-by: Matthias Fuchs <matthias.fuchs@esd.eu>
Signed-off-by: Marc Kleine-Budde <mkl@pengutronix.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agousb: xhci: rework root port wake bits if controller isn't allowed to wakeup
Lu Baolu [Tue, 18 Nov 2014 09:27:14 +0000 (11:27 +0200)]
usb: xhci: rework root port wake bits if controller isn't allowed to wakeup

commit a1377e5397ab321e21b793ec8cd2b6f12bd3c718 upstream.

When system is being suspended, if host device is not allowed to do wakeup,
xhci_suspend() needs to clear all root port wake on bits. Otherwise, some
platforms may generate spurious wakeup, even if PCI PME# is disabled.

The initial commit ff8cbf250b44 ("xhci: clear root port wake on bits"),
which also got into stable, turned out to not work correctly and had to
be reverted, and is now rewritten.

Signed-off-by: Lu Baolu <baolu.lu@linux.intel.com>
Suggested-by: Alan Stern <stern@rowland.harvard.edu>
Acked-by: Alan Stern <stern@rowland.harvard.edu>
[Mathias Nyman: reword commit message]
Signed-off-by: Mathias Nyman <mathias.nyman@linux.intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agoUSB: xhci: don't start a halted endpoint before its new dequeue is set
Mathias Nyman [Tue, 18 Nov 2014 09:27:11 +0000 (11:27 +0200)]
USB: xhci: don't start a halted endpoint before its new dequeue is set

commit c3492dbfa1050debf23a5b5cd2bc7514c5b37896 upstream.

A halted endpoint ring must first be reset, then move the ring
dequeue pointer past the problematic TRB. If we start the ring too
early after reset, but before moving the dequeue pointer we
will end up executing the same problematic TRB again.

As we always issue a set transfer dequeue command after a reset
endpoint command we can skip starting endpoint rings at reset endpoint
command completion.

Without this fix we end up trying to handle the same faulty TD for
contol endpoints. causing timeout, and failing testusb ctrl_out write
tests.

Fixes: e9df17e (USB: xhci: Correct assumptions about number of rings per endpoint.)
Tested-by: Felipe Balbi <balbi@ti.com>
Signed-off-by: Mathias Nyman <mathias.nyman@linux.intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agousb-quirks: Add reset-resume quirk for MS Wireless Laser Mouse 6000
Hans de Goede [Mon, 24 Nov 2014 10:22:38 +0000 (11:22 +0100)]
usb-quirks: Add reset-resume quirk for MS Wireless Laser Mouse 6000

commit 263e80b43559a6103e178a9176938ce171b23872 upstream.

This wireless mouse receiver needs a reset-resume quirk to properly come
out of reset.

BugLink: https://bugzilla.redhat.com/show_bug.cgi?id=1165206
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agousb: serial: ftdi_sio: add PIDs for Matrix Orbital products
Troy Clark [Mon, 17 Nov 2014 22:33:17 +0000 (14:33 -0800)]
usb: serial: ftdi_sio: add PIDs for Matrix Orbital products

commit 204ec6e07ea7aff863df0f7c53301f9cbbfbb9d3 upstream.

Add PIDs for new Matrix Orbital GTT series products.

Signed-off-by: Troy Clark <tclark@matrixorbital.ca>
[johan: shorten commit message ]
Signed-off-by: Johan Hovold <johan@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agoUSB: serial: cp210x: add IDs for CEL MeshConnect USB Stick
Preston Fick [Sat, 8 Nov 2014 05:26:11 +0000 (23:26 -0600)]
USB: serial: cp210x: add IDs for CEL MeshConnect USB Stick

commit ffcfe30ebd8dd703d0fc4324ffe56ea21f5479f4 upstream.

Signed-off-by: Preston Fick <pffick@gmail.com>
Signed-off-by: Johan Hovold <johan@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agoUSB: keyspan: fix tty line-status reporting
Johan Hovold [Tue, 18 Nov 2014 10:25:19 +0000 (11:25 +0100)]
USB: keyspan: fix tty line-status reporting

commit 5d1678a33c731b56e245e888fdae5e88efce0997 upstream.

Fix handling of TTY error flags, which are not bitmasks and must
specifically not be ORed together as this prevents the line discipline
from recognising them.

Also insert null characters when reporting overrun errors as these are
not associated with the received character.

Fixes: 1da177e4c3f4 ("Linux-2.6.12-rc2")
Signed-off-by: Johan Hovold <johan@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agoUSB: keyspan: fix overrun-error reporting
Johan Hovold [Tue, 18 Nov 2014 10:25:20 +0000 (11:25 +0100)]
USB: keyspan: fix overrun-error reporting

commit 855515a6d3731242d85850a206f2ec084c917338 upstream.

Fix reporting of overrun errors, which are not associated with a
character. Instead insert a null character and report only once.

Fixes: 1da177e4c3f4 ("Linux-2.6.12-rc2")
Signed-off-by: Johan Hovold <johan@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agoUSB: ssu100: fix overrun-error reporting
Johan Hovold [Tue, 18 Nov 2014 10:25:21 +0000 (11:25 +0100)]
USB: ssu100: fix overrun-error reporting

commit 75bcbf29c284dd0154c3e895a0bd1ef0e796160e upstream.

Fix reporting of overrun errors, which should only be reported once
using the inserted null character.

Fixes: 6b8f1ca5581b ("USB: ssu100: set tty_flags in ssu100_process_packet")
Signed-off-by: Johan Hovold <johan@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agostaging: r8188eu: Add new device ID for DLink GO-USB-N150
Larry Finger [Thu, 27 Nov 2014 16:10:21 +0000 (10:10 -0600)]
staging: r8188eu: Add new device ID for DLink GO-USB-N150

commit 6d4556fc0309608f760f1d329df56d77fdd0c31a upstream.

The DLink GO-USB-N150 with revision B1 uses this driver.

Signed-off-by: Larry Finger <Larry.Finger@lwfinger.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agoiio: Fix IIO_EVENT_CODE_EXTRACT_DIR bit mask
Cristina Ciocan [Tue, 11 Nov 2014 14:07:42 +0000 (16:07 +0200)]
iio: Fix IIO_EVENT_CODE_EXTRACT_DIR bit mask

commit ccf54555da9a5e91e454b909ca6a5303c7d6b910 upstream.

The direction field is set on 7 bits, thus we need to AND it with 0111 111 mask
in order to retrieve it, that is 0x7F, not 0xCF as it is now.

Fixes: ade7ef7ba (staging:iio: Differential channel handling)
Signed-off-by: Cristina Ciocan <cristina.ciocan@intel.com>
Signed-off-by: Jonathan Cameron <jic23@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agopowerpc/pseries: Fix endiannes issue in RTAS call from xmon
Laurent Dufour [Mon, 24 Nov 2014 14:07:53 +0000 (15:07 +0100)]
powerpc/pseries: Fix endiannes issue in RTAS call from xmon

commit 3b8a3c01096925a824ed3272601082289d9c23a5 upstream.

On pseries system (LPAR) xmon failed to enter when running in LE mode,
system is hunging. Inititating xmon will lead to such an output on the
console:

SysRq : Entering xmon
cpu 0x15: Vector: 0  at [c0000003f39ffb10]
    pc: c00000000007ed7c: sysrq_handle_xmon+0x5c/0x70
    lr: c00000000007ed7c: sysrq_handle_xmon+0x5c/0x70
    sp: c0000003f39ffc70
   msr: 8000000000009033
  current = 0xc0000003fafa7180
  paca    = 0xc000000007d75e80  softe: 0  irq_happened: 0x01
    pid   = 14617, comm = bash
Bad kernel stack pointer fafb4b0 at eca7cc4
cpu 0x15: Vector: 300 (Data Access) at [c000000007f07d40]
    pc: 000000000eca7cc4
    lr: 000000000eca7c44
    sp: fafb4b0
   msr: 8000000000001000
   dar: 10000000
 dsisr: 42000000
  current = 0xc0000003fafa7180
  paca    = 0xc000000007d75e80  softe: 0  irq_happened: 0x01
    pid   = 14617, comm = bash
cpu 0x15: Exception 300 (Data Access) in xmon, returning to main loop
xmon: WARNING: bad recursive fault on cpu 0x15

The root cause is that xmon is calling RTAS to turn off the surveillance
when entering xmon, and RTAS is requiring big endian parameters.

This patch is byte swapping the RTAS arguments when running in LE mode.

Signed-off-by: Laurent Dufour <ldufour@linux.vnet.ibm.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agopowerpc/pseries: Honor the generic "no_64bit_msi" flag
Benjamin Herrenschmidt [Tue, 7 Oct 2014 05:12:55 +0000 (16:12 +1100)]
powerpc/pseries: Honor the generic "no_64bit_msi" flag

commit 415072a041bf50dbd6d56934ffc0cbbe14c97be8 upstream.

Instead of the arch specific quirk which we are deprecating

Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agoof/base: Fix PowerPC address parsing hack
Benjamin Herrenschmidt [Fri, 14 Nov 2014 06:55:03 +0000 (17:55 +1100)]
of/base: Fix PowerPC address parsing hack

commit 746c9e9f92dde2789908e51a354ba90a1962a2eb upstream.

We have a historical hack that treats missing ranges properties as the
equivalent of an empty one. This is needed for ancient PowerMac "bad"
device-trees, and shouldn't be enabled for any other PowerPC platform,
otherwise we get some nasty layout of devices in sysfs or even
duplication when a set of otherwise identically named devices is
created multiple times under a different parent node with no ranges
property.

This fix is needed for the PowerNV i2c busses to be exposed properly
and will fix a number of other embedded cases.

Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Acked-by: Grant Likely <grant.likely@linaro.org>
Signed-off-by: Rob Herring <robh@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agoath9k: Fix RTC_DERIVED_CLK usage
Miaoqing Pan [Thu, 6 Nov 2014 05:22:23 +0000 (10:52 +0530)]
ath9k: Fix RTC_DERIVED_CLK usage

commit 4e6ce4dc7ce71d0886908d55129d5d6482a27ff9 upstream.

Based on the reference clock, which could be 25MHz or 40MHz,
AR_RTC_DERIVED_CLK is programmed differently for AR9340 and AR9550.
But, when a chip reset is done, processing the initvals
sets the register back to the default value.

Fix this by moving the code in ath9k_hw_init_pll() to
ar9003_hw_override_ini(). Also, do this override for AR9531.

Signed-off-by: Miaoqing Pan <miaoqing@qca.qualcomm.com>
Signed-off-by: Sujith Manoharan <c_manoha@qca.qualcomm.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agoASoC: dpcm: Fix race between FE/BE updates and trigger
Takashi Iwai [Tue, 4 Nov 2014 15:52:28 +0000 (16:52 +0100)]
ASoC: dpcm: Fix race between FE/BE updates and trigger

commit ea9d0d771fcd32cd56070819749477d511ec9117 upstream.

DPCM can update the FE/BE connection states totally asynchronously
from the FE's PCM state.  Most of FE/BE state changes are protected by
mutex, so that they won't race, but there are still some actions that
are uncovered.  For example, suppose to switch a BE while a FE's
stream is running.  This would call soc_dpcm_runtime_update(), which
sets FE's runtime_update flag, then sets up and starts BEs, and clears
FE's runtime_update flag again.

When a device emits XRUN during this operation, the PCM core triggers
snd_pcm_stop(XRUN).  Since the trigger action is an atomic ops, this
isn't blocked by the mutex, thus it kicks off DPCM's trigger action.
It eventually updates and clears FE's runtime_update flag while
soc_dpcm_runtime_update() is running concurrently, and it results in
confusion.

Usually, for avoiding such a race, we take a lock.  There is a PCM
stream lock for that purpose.  However, as already mentioned, the
trigger action is atomic, and we can't take the lock for the whole
soc_dpcm_runtime_update() or other operations that include the lengthy
jobs like hw_params or prepare.

This patch provides an alternative solution.  This adds a way to defer
the conflicting trigger callback to be executed at the end of FE/BE
state changes.  For doing it, two things are introduced:

- Each runtime_update state change of FEs is protected via PCM stream
  lock.
- The FE's trigger callback checks the runtime_update flag.  If it's
  not set, the trigger action is executed there.  If set, mark the
  pending trigger action and returns immediately.
- At the exit of runtime_update state change, it checks whether the
  pending trigger is present.  If yes, it executes the trigger action
  at this point.

Reported-and-tested-by: Qiao Zhou <zhouqiao@marvell.com>
Signed-off-by: Takashi Iwai <tiwai@suse.de>
Acked-by: Liam Girdwood <liam.r.girdwood@linux.intel.com>
Signed-off-by: Mark Brown <broonie@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agoASoC: wm_adsp: Avoid attempt to free buffers that might still be in use
Charles Keepax [Mon, 17 Nov 2014 10:48:21 +0000 (10:48 +0000)]
ASoC: wm_adsp: Avoid attempt to free buffers that might still be in use

commit 9da7a5a9fdeeb76b2243f6b473363a7e6147ab6f upstream.

We should not free any buffers associated with writing out coefficients
to the DSP until all the async writes have completed. This patch updates
the out of memory path when allocating a new buffer to include a call to
regmap_async_complete.

Reported-by: JS Park <aitdark.park@samsung.com>
Signed-off-by: Charles Keepax <ckeepax@opensource.wolfsonmicro.com>
Signed-off-by: Mark Brown <broonie@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agoASoC: sgtl5000: Fix SMALL_POP bit definition
Fabio Estevam [Fri, 14 Nov 2014 04:14:47 +0000 (02:14 -0200)]
ASoC: sgtl5000: Fix SMALL_POP bit definition

commit c251ea7bd7a04f1f2575467e0de76e803cf59149 upstream.

On a mx28evk with a sgtl5000 codec we notice a loud 'click' sound  to happen
5 seconds after the end of a playback.

The SMALL_POP bit should fix this, but its definition is incorrect:
according to the sgtl5000 manual it is bit 0 of CHIP_REF_CTRL register, not
bit 1.

Fix the definition accordingly and enable the bit as intended per the code
comment.

After applying this change, no loud 'click' sound is heard after playback

Signed-off-by: Fabio Estevam <fabio.estevam@freescale.com>
Signed-off-by: Mark Brown <broonie@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agoASoC: fsi: remove unsupported PAUSE flag
Kuninori Morimoto [Wed, 29 Oct 2014 04:01:53 +0000 (21:01 -0700)]
ASoC: fsi: remove unsupported PAUSE flag

commit c1b9b9b1ad2df6144ca3fbe6989f7bd9ea5c5562 upstream.

FSI doesn't support PAUSE.
Remove SNDRV_PCM_INFO_PAUSE flags from snd_pcm_hardware info

Signed-off-by: Kuninori Morimoto <kuninori.morimoto.gx@renesas.com>
Signed-off-by: Mark Brown <broonie@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agoASoC: rsnd: remove unsupported PAUSE flag
Kuninori Morimoto [Wed, 29 Oct 2014 04:02:03 +0000 (21:02 -0700)]
ASoC: rsnd: remove unsupported PAUSE flag

commit 706c66213e5e623e23f521b1acbd8171af7a3549 upstream.

R-Car sound doesn't support PAUSE.
Remove SNDRV_PCM_INFO_PAUSE flags from snd_pcm_hardware info

Signed-off-by: Kuninori Morimoto <kuninori.morimoto.gx@renesas.com>
Signed-off-by: Mark Brown <broonie@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agoib_isert: Add max_send_sge=2 minimum for control PDU responses
Or Gerlitz [Wed, 22 Oct 2014 21:55:49 +0000 (14:55 -0700)]
ib_isert: Add max_send_sge=2 minimum for control PDU responses

commit f57915cfa5b2b14c1cffa2e83c034f55e3f0e70d upstream.

This patch adds a max_send_sge=2 minimum in isert_conn_setup_qp()
to ensure outgoing control PDU responses with tx_desc->num_sge=2
are able to function correctly.

This addresses a bug with RDMA hardware using dev_attr.max_sge=3,
that in the original code with the ConnectX-2 work-around would
result in isert_conn->max_sge=1 being negotiated.

Originally reported by Chris with ocrdma driver.

Reported-by: Chris Moore <Chris.Moore@emulex.com>
Tested-by: Chris Moore <Chris.Moore@emulex.com>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: Nicholas Bellinger <nab@linux-iscsi.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agoIB/isert: Adjust CQ size to HW limits
Chris Moore [Tue, 4 Nov 2014 16:28:29 +0000 (16:28 +0000)]
IB/isert: Adjust CQ size to HW limits

commit b1a5ad006b34ded9dc7ec64988deba1b3ecad367 upstream.

isert has an issue of trying to create a CQ with more CQEs than are
supported by the hardware, that currently results in failures during
isert_device creation during first session login.

This is the isert version of the patch that Minh Tran submitted for
iser, and is simple a workaround required to function with existing
ocrdma hardware.

Signed-off-by: Chris Moore <chris.moore@emulex.com>
Reviewied-by: Sagi Grimberg <sagig@mellanox.com>
Signed-off-by: Nicholas Bellinger <nab@linux-iscsi.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agoclockevent: sun4i: Fix race condition in the probe code
Maxime Ripard [Tue, 18 Nov 2014 22:59:33 +0000 (23:59 +0100)]
clockevent: sun4i: Fix race condition in the probe code

commit 6bab4a8a1888729f17f4923cc5867e4674f66333 upstream.

The interrupts were activated and the handler registered before the clockevent
was registered in the probe function.

The interrupt handler, however, was making the assumption that the clockevent
device was registered.

That could cause a null pointer dereference if the timer interrupt was firing
during this narrow window.

Fix that by moving the clockevent registration before the interrupt is enabled.

Reported-by: Roman Byshko <rbyshko@gmail.com>
Signed-off-by: Maxime Ripard <maxime.ripard@free-electrons.com>
Signed-off-by: Daniel Lezcano <daniel.lezcano@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agoPCI/MSI: Add device flag indicating that 64-bit MSIs don't work
Benjamin Herrenschmidt [Fri, 3 Oct 2014 05:13:24 +0000 (15:13 +1000)]
PCI/MSI: Add device flag indicating that 64-bit MSIs don't work

commit f144d1496b47e7450f41b767d0d91c724c2198bc upstream.

This can be set by quirks/drivers to be used by the architecture code
that assigns the MSI addresses.

We additionally add verification in the core MSI code that the values
assigned by the architecture do satisfy the limitation in order to fail
gracefully if they don't (ie. the arch hasn't been updated to deal with
that quirk yet).

Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Acked-by: Bjorn Helgaas <bhelgaas@google.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agoPCI: Support 64-bit bridge windows if we have 64-bit dma_addr_t
Yinghai Lu [Wed, 19 Nov 2014 21:30:32 +0000 (14:30 -0700)]
PCI: Support 64-bit bridge windows if we have 64-bit dma_addr_t

commit 7fc986d8a9727e5d40da3c2c1c343da6142e82a9 upstream.

Aaron reported that a 32-bit x86 kernel with Physical Address Extension
(PAE) support complains about bridge prefetchable memory windows above 4GB:

  pci_bus 0000:00: root bus resource [mem 0x380000000000-0x383fffffffff]
  ...
  pci 0000:03:00.0: reg 0x10: [mem 0x383fffc00000-0x383fffdfffff 64bit pref]
  pci 0000:03:00.0: reg 0x20: [mem 0x383fffe04000-0x383fffe07fff 64bit pref]
  pci 0000:03:00.1: reg 0x10: [mem 0x383fffa00000-0x383fffbfffff 64bit pref]
  pci 0000:03:00.1: reg 0x20: [mem 0x383fffe00000-0x383fffe03fff 64bit pref]
  pci 0000:00:02.2: PCI bridge to [bus 03-04]
  pci 0000:00:02.2:   bridge window [io  0x1000-0x1fff]
  pci 0000:00:02.2:   bridge window [mem 0x91900000-0x91cfffff]
  pci 0000:00:02.2: can't handle 64-bit address space for bridge

In this kernel, unsigned long is 32 bits and dma_addr_t is 64 bits.
Previously we used "unsigned long" to hold the bridge window address.  But
this is a bus address, so we should use dma_addr_t instead.

Use dma_addr_t to hold the bridge window base and limit.

The question of whether the CPU can actually *address* the window is
separate and depends on what the physical address space of the CPU is and
whether the host bridge does any address translation.

[bhelgaas: fix "shift count > width of type", changelog, stable tag]
Fixes: d56dbf5bab8c ("PCI: Allocate 64-bit BARs above 4G when possible")
Link: https://bugzilla.kernel.org/show_bug.cgi?id=88131
Reported-by: Aaron Ma <mapengyu@gmail.com>
Tested-by: Aaron Ma <mapengyu@gmail.com>
Signed-off-by: Yinghai Lu <yinghai@kernel.org>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agoipx: fix locking regression in ipx_sendmsg and ipx_recvmsg
Jiri Bohac [Wed, 19 Nov 2014 22:05:49 +0000 (23:05 +0100)]
ipx: fix locking regression in ipx_sendmsg and ipx_recvmsg

[ Upstream commit 01462405f0c093b2f8dfddafcadcda6c9e4c5cdf ]

This fixes an old regression introduced by commit
b0d0d915 (ipx: remove the BKL).

When a recvmsg syscall blocks waiting for new data, no data can be sent on the
same socket with sendmsg because ipx_recvmsg() sleeps with the socket locked.

This breaks mars-nwe (NetWare emulator):
- the ncpserv process reads the request using recvmsg
- ncpserv forks and spawns nwconn
- ncpserv calls a (blocking) recvmsg and waits for new requests
- nwconn deadlocks in sendmsg on the same socket

Commit b0d0d915 has simply replaced BKL locking with
lock_sock/release_sock. Unlike now, BKL got unlocked while
sleeping, so a blocking recvmsg did not block a concurrent
sendmsg.

Only keep the socket locked while actually working with the socket data and
release it prior to calling skb_recv_datagram().

Signed-off-by: Jiri Bohac <jbohac@suse.cz>
Reviewed-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agopptp: fix stack info leak in pptp_getname()
Mathias Krause [Wed, 19 Nov 2014 17:05:26 +0000 (18:05 +0100)]
pptp: fix stack info leak in pptp_getname()

[ Upstream commit a5f6fc28d6e6cc379c6839f21820e62262419584 ]

pptp_getname() only partially initializes the stack variable sa,
particularly only fills the pptp part of the sa_addr union. The code
thereby discloses 16 bytes of kernel stack memory via getsockname().

Fix this by memset(0)'ing the union before.

Cc: Dmitry Kozlov <xeb@mail.ru>
Signed-off-by: Mathias Krause <minipli@googlemail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agobonding: fix curr_active_slave/carrier with loadbalance arp monitoring
Nikolay Aleksandrov [Tue, 18 Nov 2014 14:14:44 +0000 (15:14 +0100)]
bonding: fix curr_active_slave/carrier with loadbalance arp monitoring

[ Upstream commit b8e4500f42fe4464a33a887579147050bed8fcef ]

Since commit 6fde8f037e60 ("bonding: fix locking in
bond_loadbalance_arp_mon()") we can have a stale bond carrier state and
stale curr_active_slave when using arp monitoring in loadbalance modes. The
reason is that in bond_loadbalance_arp_mon() we can't have
do_failover == true but slave_state_changed == false, whenever do_failover
is true then slave_state_changed is also true. Then the following piece
from bond_loadbalance_arp_mon():
                if (slave_state_changed) {
                        bond_slave_state_change(bond);
                        if (BOND_MODE(bond) == BOND_MODE_XOR)
                                bond_update_slave_arr(bond, NULL);
                } else if (do_failover) {
                        block_netpoll_tx();
                        bond_select_active_slave(bond);
                        unblock_netpoll_tx();
                }

will execute only the first branch, always and regardless of do_failover.
Since these two events aren't related in such way, we need to decouple and
consider them separately.

For example this issue could lead to the following result:
Bonding Mode: load balancing (round-robin)
*MII Status: down*
MII Polling Interval (ms): 0
Up Delay (ms): 0
Down Delay (ms): 0
ARP Polling Interval (ms): 100
ARP IP target/s (n.n.n.n form): 192.168.9.2

Slave Interface: ens12
*MII Status: up*
Speed: 10000 Mbps
Duplex: full
Link Failure Count: 2
Permanent HW addr: 00:0f:53:01:42:2c
Slave queue ID: 0

Slave Interface: eth1
*MII Status: up*
Speed: Unknown
Duplex: Unknown
Link Failure Count: 70
Permanent HW addr: 52:54:00:2f:0f:8e
Slave queue ID: 0

Since some interfaces are up, then the status of the bond should also be
up, but it will never change unless something invokes bond_set_carrier()
(i.e. enslave, bond_select_active_slave etc). Now, if I force the
calling of bond_select_active_slave via for example changing
primary_reselect (it can change in any mode), then the MII status goes to
"up" because it calls bond_select_active_slave() which should've been done
from bond_loadbalance_arp_mon() itself.

CC: Veaceslav Falico <vfalico@gmail.com>
CC: Jay Vosburgh <j.vosburgh@gmail.com>
CC: Andy Gospodarek <andy@greyhouse.net>
CC: Ding Tianhong <dingtianhong@huawei.com>
Fixes: 6fde8f037e60 ("bonding: fix locking in bond_loadbalance_arp_mon()")
Signed-off-by: Nikolay Aleksandrov <nikolay@redhat.com>
Acked-by: Veaceslav Falico <vfalico@gmail.com>
Acked-by: Andy Gospodarek <gospo@cumulusnetworks.com>
Acked-by: Ding Tianhong <dingtianhong@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agoqmi_wwan: Add support for HP lt4112 LTE/HSPA+ Gobi 4G Modem
Martin Hauke [Sun, 16 Nov 2014 18:55:25 +0000 (19:55 +0100)]
qmi_wwan: Add support for HP lt4112 LTE/HSPA+ Gobi 4G Modem

[ Upstream commit bb2bdeb83fb125c95e47fc7eca2a3e8f868e2a74 ]

Added the USB VID/PID for the HP lt4112 LTE/HSPA+ Gobi 4G Modem (Huawei me906e)

Signed-off-by: Martin Hauke <mardnh@gmx.de>
Acked-by: Bjørn Mork <bjorn@mork.no>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agoieee802154: fix error handling in ieee802154fake_probe()
Alexey Khoroshilov [Fri, 14 Nov 2014 23:11:59 +0000 (02:11 +0300)]
ieee802154: fix error handling in ieee802154fake_probe()

[ Upstream commit 8c2dd54485ccee7fc4086611e188478584758c8d ]

In case of any failure ieee802154fake_probe() just calls unregister_netdev().
But it does not look safe to unregister netdevice before it was registered.

The patch implements straightforward resource deallocation in case of
failure in ieee802154fake_probe().

Found by Linux Driver Verification project (linuxtesting.org).

Signed-off-by: Alexey Khoroshilov <khoroshilov@ispras.ru>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agoipv4: Fix incorrect error code when adding an unreachable route
Panu Matilainen [Fri, 14 Nov 2014 11:14:32 +0000 (13:14 +0200)]
ipv4: Fix incorrect error code when adding an unreachable route

[ Upstream commit 49dd18ba4615eaa72f15c9087dea1c2ab4744cf5 ]

Trying to add an unreachable route incorrectly returns -ESRCH if
if custom FIB rules are present:

[root@localhost ~]# ip route add 74.125.31.199 dev eth0 via 1.2.3.4
RTNETLINK answers: Network is unreachable
[root@localhost ~]# ip rule add to 55.66.77.88 table 200
[root@localhost ~]# ip route add 74.125.31.199 dev eth0 via 1.2.3.4
RTNETLINK answers: No such process
[root@localhost ~]#

Commit 83886b6b636173b206f475929e58fac75c6f2446 ("[NET]: Change "not found"
return value for rule lookup") changed fib_rules_lookup()
to use -ESRCH as a "not found" code internally, but for user space it
should be translated into -ENETUNREACH. Handle the translation centrally in
ipv4-specific fib_lookup(), leaving the DECnet case alone.

On a related note, commit b7a71b51ee37d919e4098cd961d59a883fd272d8
("ipv4: removed redundant conditional") removed a similar translation from
ip_route_input_slow() prematurely AIUI.

Fixes: b7a71b51ee37 ("ipv4: removed redundant conditional")
Signed-off-by: Panu Matilainen <pmatilai@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agoinetdevice: fixed signed integer overflow
Vincent BENAYOUN [Thu, 13 Nov 2014 12:47:26 +0000 (13:47 +0100)]
inetdevice: fixed signed integer overflow

[ Upstream commit 84bc88688e3f6ef843aa8803dbcd90168bb89faf ]

There could be a signed overflow in the following code.

The expression, (32-logmask) is comprised between 0 and 31 included.
It may be equal to 31.
In such a case the left shift will produce a signed integer overflow.
According to the C99 Standard, this is an undefined behavior.
A simple fix is to replace the signed int 1 with the unsigned int 1U.

Signed-off-by: Vincent BENAYOUN <vincent.benayoun@trust-in-soft.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agosparc64: Fix constraints on swab helpers.
David S. Miller [Sun, 16 Nov 2014 21:19:32 +0000 (13:19 -0800)]
sparc64: Fix constraints on swab helpers.

[ Upstream commit 5a2b59d3993e8ca4f7788a48a23e5cb303f26954 ]

We are reading the memory location, so we have to have a memory
constraint in there purely for the sake of showing the data flow
to the compiler.

Reported-by: Martin K. Petersen <martin.petersen@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agouprobes, x86: Fix _TIF_UPROBE vs _TIF_NOTIFY_RESUME
Andy Lutomirski [Fri, 21 Nov 2014 21:26:07 +0000 (13:26 -0800)]
uprobes, x86: Fix _TIF_UPROBE vs _TIF_NOTIFY_RESUME

commit 82975bc6a6df743b9a01810fb32cb65d0ec5d60b upstream.

x86 call do_notify_resume on paranoid returns if TIF_UPROBE is set but
not on non-paranoid returns.  I suspect that this is a mistake and that
the code only works because int3 is paranoid.

Setting _TIF_NOTIFY_RESUME in the uprobe code was probably a workaround
for the x86 bug.  With that bug fixed, we can remove _TIF_NOTIFY_RESUME
from the uprobes code.

Reported-by: Oleg Nesterov <oleg@redhat.com>
Acked-by: Srikar Dronamraju <srikar@linux.vnet.ibm.com>
Acked-by: Borislav Petkov <bp@suse.de>
Signed-off-by: Andy Lutomirski <luto@amacapital.net>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agox86, kaslr: Handle Gold linker for finding bss/brk
Kees Cook [Tue, 18 Nov 2014 00:16:04 +0000 (16:16 -0800)]
x86, kaslr: Handle Gold linker for finding bss/brk

commit 70b61e362187b5fccac206506d402f3424e3e749 upstream.

When building with the Gold linker, the .bss and .brk areas of vmlinux
are shown as consecutive instead of having the same file offset. Allow
for either state, as long as things add up correctly.

Fixes: e6023367d779 ("x86, kaslr: Prevent .bss from overlaping initrd")
Reported-by: Markus Trippelsdorf <markus@trippelsdorf.de>
Signed-off-by: Kees Cook <keescook@chromium.org>
Cc: Junjie Mao <eternal.n08@gmail.com>
Link: http://lkml.kernel.org/r/20141118001604.GA25045@www.outflux.net
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agox86, mm: Set NX across entire PMD at boot
Kees Cook [Fri, 14 Nov 2014 19:47:37 +0000 (11:47 -0800)]
x86, mm: Set NX across entire PMD at boot

commit 45e2a9d4701d8c624d4a4bcdd1084eae31e92f58 upstream.

When setting up permissions on kernel memory at boot, the end of the
PMD that was split from bss remained executable. It should be NX like
the rest. This performs a PMD alignment instead of a PAGE alignment to
get the correct span of memory.

Before:
---[ High Kernel Mapping ]---
...
0xffffffff8202d000-0xffffffff82200000  1868K     RW       GLB NX pte
0xffffffff82200000-0xffffffff82c00000    10M     RW   PSE GLB NX pmd
0xffffffff82c00000-0xffffffff82df5000  2004K     RW       GLB NX pte
0xffffffff82df5000-0xffffffff82e00000    44K     RW       GLB x  pte
0xffffffff82e00000-0xffffffffc0000000   978M                     pmd

After:
---[ High Kernel Mapping ]---
...
0xffffffff8202d000-0xffffffff82200000  1868K     RW       GLB NX pte
0xffffffff82200000-0xffffffff82e00000    12M     RW   PSE GLB NX pmd
0xffffffff82e00000-0xffffffffc0000000   978M                     pmd

[ tglx: Changed it to roundup(_brk_end, PMD_SIZE) and added a comment.
        We really should unmap the reminder along with the holes
        caused by init,initdata etc. but thats a different issue ]

Signed-off-by: Kees Cook <keescook@chromium.org>
Cc: Andy Lutomirski <luto@amacapital.net>
Cc: Toshi Kani <toshi.kani@hp.com>
Cc: Yasuaki Ishimatsu <isimatu.yasuaki@jp.fujitsu.com>
Cc: David Vrabel <david.vrabel@citrix.com>
Cc: Wang Nan <wangnan0@huawei.com>
Cc: Yinghai Lu <yinghai@kernel.org>
Link: http://lkml.kernel.org/r/20141114194737.GA3091@www.outflux.net
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agox86: Require exact match for 'noxsave' command line option
Dave Hansen [Tue, 11 Nov 2014 22:01:33 +0000 (14:01 -0800)]
x86: Require exact match for 'noxsave' command line option

commit 2cd3949f702692cf4c5d05b463f19cd706a92dd3 upstream.

We have some very similarly named command-line options:

arch/x86/kernel/cpu/common.c:__setup("noxsave", x86_xsave_setup);
arch/x86/kernel/cpu/common.c:__setup("noxsaveopt", x86_xsaveopt_setup);
arch/x86/kernel/cpu/common.c:__setup("noxsaves", x86_xsaves_setup);

__setup() is designed to match options that take arguments, like
"foo=bar" where you would have:

__setup("foo", x86_foo_func...);

The problem is that "noxsave" actually _matches_ "noxsaves" in
the same way that "foo" matches "foo=bar".  If you boot an old
kernel that does not know about "noxsaves" with "noxsaves" on the
command line, it will interpret the argument as "noxsave", which
is not what you want at all.

This makes the "noxsave" handler only return success when it finds
an *exact* match.

[ tglx: We really need to make __setup() more robust. ]

Signed-off-by: Dave Hansen <dave.hansen@linux.intel.com>
Cc: Dave Hansen <dave@sr71.net>
Cc: Fenghua Yu <fenghua.yu@intel.com>
Cc: x86@kernel.org
Link: http://lkml.kernel.org/r/20141111220133.FE053984@viggo.jf.intel.com
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agox86_64, traps: Rework bad_iret
Andy Lutomirski [Sun, 23 Nov 2014 02:00:33 +0000 (18:00 -0800)]
x86_64, traps: Rework bad_iret

commit b645af2d5905c4e32399005b867987919cbfc3ae upstream.

It's possible for iretq to userspace to fail.  This can happen because
of a bad CS, SS, or RIP.

Historically, we've handled it by fixing up an exception from iretq to
land at bad_iret, which pretends that the failed iret frame was really
the hardware part of #GP(0) from userspace.  To make this work, there's
an extra fixup to fudge the gs base into a usable state.

This is suboptimal because it loses the original exception.  It's also
buggy because there's no guarantee that we were on the kernel stack to
begin with.  For example, if the failing iret happened on return from an
NMI, then we'll end up executing general_protection on the NMI stack.
This is bad for several reasons, the most immediate of which is that
general_protection, as a non-paranoid idtentry, will try to deliver
signals and/or schedule from the wrong stack.

This patch throws out bad_iret entirely.  As a replacement, it augments
the existing swapgs fudge into a full-blown iret fixup, mostly written
in C.  It's should be clearer and more correct.

Signed-off-by: Andy Lutomirski <luto@amacapital.net>
Reviewed-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agox86_64, traps: Stop using IST for #SS
Andy Lutomirski [Sun, 23 Nov 2014 02:00:32 +0000 (18:00 -0800)]
x86_64, traps: Stop using IST for #SS

commit 6f442be2fb22be02cafa606f1769fa1e6f894441 upstream.

On a 32-bit kernel, this has no effect, since there are no IST stacks.

On a 64-bit kernel, #SS can only happen in user code, on a failed iret
to user space, a canonical violation on access via RSP or RBP, or a
genuine stack segment violation in 32-bit kernel code.  The first two
cases don't need IST, and the latter two cases are unlikely fatal bugs,
and promoting them to double faults would be fine.

This fixes a bug in which the espfix64 code mishandles a stack segment
violation.

This saves 4k of memory per CPU and a tiny bit of code.

Signed-off-by: Andy Lutomirski <luto@amacapital.net>
Reviewed-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agox86_64, traps: Fix the espfix64 #DF fixup and rewrite it in C
Andy Lutomirski [Sun, 23 Nov 2014 02:00:31 +0000 (18:00 -0800)]
x86_64, traps: Fix the espfix64 #DF fixup and rewrite it in C

commit af726f21ed8af2cdaa4e93098dc211521218ae65 upstream.

There's nothing special enough about the espfix64 double fault fixup to
justify writing it in assembly.  Move it to C.

This also fixes a bug: if the double fault came from an IST stack, the
old asm code would return to a partially uninitialized stack frame.

Fixes: 3891a04aafd668686239349ea58f3314ea2af86b
Signed-off-by: Andy Lutomirski <luto@amacapital.net>
Reviewed-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agoMIPS: Loongson: Make platform serial setup always built-in.
Aaro Koskinen [Wed, 19 Nov 2014 23:05:38 +0000 (01:05 +0200)]
MIPS: Loongson: Make platform serial setup always built-in.

commit 26927f76499849e095714452b8a4e09350f6a3b9 upstream.

If SERIAL_8250 is compiled as a module, the platform specific setup
for Loongson will be a module too, and it will not work very well.
At least on Loongson 3 it will trigger a build failure,
since loongson_sysconf is not exported to modules.

Fix by making the platform specific serial code always built-in.

Signed-off-by: Aaro Koskinen <aaro.koskinen@iki.fi>
Reported-by: Ralf Baechle <ralf@linux-mips.org>
Cc: linux-mips@linux-mips.org
Cc: Huacai Chen <chenhc@lemote.com>
Cc: Markos Chandras <Markos.Chandras@imgtec.com>
Patchwork: https://patchwork.linux-mips.org/patch/8533/
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agoMIPS: oprofile: Fix backtrace on 64-bit kernel
Aaro Koskinen [Fri, 17 Oct 2014 15:10:24 +0000 (18:10 +0300)]
MIPS: oprofile: Fix backtrace on 64-bit kernel

commit bbaf113a481b6ce32444c125807ad3618643ce57 upstream.

Fix incorrect cast that always results in wrong address for the new
frame on 64-bit kernels.

Signed-off-by: Aaro Koskinen <aaro.koskinen@nsn.com>
Cc: linux-mips@linux-mips.org
Patchwork: https://patchwork.linux-mips.org/patch/8110/
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agoLinux 3.14.25
Greg Kroah-Hartman [Fri, 21 Nov 2014 17:23:44 +0000 (09:23 -0800)]
Linux 3.14.25

9 years agomm/page_alloc: prevent MIGRATE_RESERVE pages from being misplaced
Vlastimil Babka [Wed, 4 Jun 2014 23:07:22 +0000 (16:07 -0700)]
mm/page_alloc: prevent MIGRATE_RESERVE pages from being misplaced

commit 5bcc9f86ef09a933255ee66bd899d4601785dad5 upstream.

For the MIGRATE_RESERVE pages, it is useful when they do not get
misplaced on free_list of other migratetype, otherwise they might get
allocated prematurely and e.g.  fragment the MIGRATE_RESEVE pageblocks.
While this cannot be avoided completely when allocating new
MIGRATE_RESERVE pageblocks in min_free_kbytes sysctl handler, we should
prevent the misplacement where possible.

Currently, it is possible for the misplacement to happen when a
MIGRATE_RESERVE page is allocated on pcplist through rmqueue_bulk() as a
fallback for other desired migratetype, and then later freed back
through free_pcppages_bulk() without being actually used.  This happens
because free_pcppages_bulk() uses get_freepage_migratetype() to choose
the free_list, and rmqueue_bulk() calls set_freepage_migratetype() with
the *desired* migratetype and not the page's original MIGRATE_RESERVE
migratetype.

This patch fixes the problem by moving the call to
set_freepage_migratetype() from rmqueue_bulk() down to
__rmqueue_smallest() and __rmqueue_fallback() where the actual page's
migratetype (e.g.  from which free_list the page is taken from) is used.
Note that this migratetype might be different from the pageblock's
migratetype due to freepage stealing decisions.  This is OK, as page
stealing never uses MIGRATE_RESERVE as a fallback, and also takes care
to leave all MIGRATE_CMA pages on the correct freelist.

Therefore, as an additional benefit, the call to
get_pageblock_migratetype() from rmqueue_bulk() when CMA is enabled, can
be removed completely.  This relies on the fact that MIGRATE_CMA
pageblocks are created only during system init, and the above.  The
related is_migrate_isolate() check is also unnecessary, as memory
isolation has other ways to move pages between freelists, and drain pcp
lists containing pages that should be isolated.  The buffered_rmqueue()
can also benefit from calling get_freepage_migratetype() instead of
get_pageblock_migratetype().

Signed-off-by: Vlastimil Babka <vbabka@suse.cz>
Reported-by: Yong-Taek Lee <ytk.lee@samsung.com>
Reported-by: Bartlomiej Zolnierkiewicz <b.zolnierkie@samsung.com>
Suggested-by: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Acked-by: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Suggested-by: Mel Gorman <mgorman@suse.de>
Acked-by: Minchan Kim <minchan@kernel.org>
Cc: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Cc: Marek Szyprowski <m.szyprowski@samsung.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Rik van Riel <riel@redhat.com>
Cc: Michal Nazarewicz <mina86@mina86.com>
Cc: "Wang, Yalin" <Yalin.Wang@sonymobile.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Mel Gorman <mgorman@suse.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agomm: vmscan: use proportional scanning during direct reclaim and full scan at DEF_PRIORITY
Mel Gorman [Wed, 4 Jun 2014 23:10:49 +0000 (16:10 -0700)]
mm: vmscan: use proportional scanning during direct reclaim and full scan at DEF_PRIORITY

commit 1a501907bbea8e6ebb0b16cf6db9e9cbf1d2c813 upstream.

Commit "mm: vmscan: obey proportional scanning requirements for kswapd"
ensured that file/anon lists were scanned proportionally for reclaim from
kswapd but ignored it for direct reclaim.  The intent was to minimse
direct reclaim latency but Yuanhan Liu pointer out that it substitutes one
long stall for many small stalls and distorts aging for normal workloads
like streaming readers/writers.  Hugh Dickins pointed out that a
side-effect of the same commit was that when one LRU list dropped to zero
that the entirety of the other list was shrunk leading to excessive
reclaim in memcgs.  This patch scans the file/anon lists proportionally
for direct reclaim to similarly age page whether reclaimed by kswapd or
direct reclaim but takes care to abort reclaim if one LRU drops to zero
after reclaiming the requested number of pages.

Based on ext4 and using the Intel VM scalability test

                                              3.15.0-rc5            3.15.0-rc5
                                                shrinker            proportion
Unit  lru-file-readonce    elapsed      5.3500 (  0.00%)      5.4200 ( -1.31%)
Unit  lru-file-readonce time_range      0.2700 (  0.00%)      0.1400 ( 48.15%)
Unit  lru-file-readonce time_stddv      0.1148 (  0.00%)      0.0536 ( 53.33%)
Unit lru-file-readtwice    elapsed      8.1700 (  0.00%)      8.1700 (  0.00%)
Unit lru-file-readtwice time_range      0.4300 (  0.00%)      0.2300 ( 46.51%)
Unit lru-file-readtwice time_stddv      0.1650 (  0.00%)      0.0971 ( 41.16%)

The test cases are running multiple dd instances reading sparse files. The results are within
the noise for the small test machine. The impact of the patch is more noticable from the vmstats

                            3.15.0-rc5  3.15.0-rc5
                              shrinker  proportion
Minor Faults                     35154       36784
Major Faults                       611        1305
Swap Ins                           394        1651
Swap Outs                         4394        5891
Allocation stalls               118616       44781
Direct pages scanned           4935171     4602313
Kswapd pages scanned          15921292    16258483
Kswapd pages reclaimed        15913301    16248305
Direct pages reclaimed         4933368     4601133
Kswapd efficiency                  99%         99%
Kswapd velocity             670088.047  682555.961
Direct efficiency                  99%         99%
Direct velocity             207709.217  193212.133
Percentage direct scans            23%         22%
Page writes by reclaim        4858.000    6232.000
Page writes file                   464         341
Page writes anon                  4394        5891

Note that there are fewer allocation stalls even though the amount
of direct reclaim scanning is very approximately the same.

Signed-off-by: Mel Gorman <mgorman@suse.de>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Hugh Dickins <hughd@google.com>
Cc: Tim Chen <tim.c.chen@linux.intel.com>
Cc: Dave Chinner <david@fromorbit.com>
Tested-by: Yuanhan Liu <yuanhan.liu@linux.intel.com>
Cc: Bob Liu <bob.liu@oracle.com>
Cc: Jan Kara <jack@suse.cz>
Cc: Rik van Riel <riel@redhat.com>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Mel Gorman <mgorman@suse.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agofs/superblock: avoid locking counting inodes and dentries before reclaiming them
Tim Chen [Wed, 4 Jun 2014 23:10:47 +0000 (16:10 -0700)]
fs/superblock: avoid locking counting inodes and dentries before reclaiming them

commit d23da150a37c9fe3cc83dbaf71b3e37fd434ed52 upstream.

We remove the call to grab_super_passive in call to super_cache_count.
This becomes a scalability bottleneck as multiple threads are trying to do
memory reclamation, e.g.  when we are doing large amount of file read and
page cache is under pressure.  The cached objects quickly got reclaimed
down to 0 and we are aborting the cache_scan() reclaim.  But counting
creates a log jam acquiring the sb_lock.

We are holding the shrinker_rwsem which ensures the safety of call to
list_lru_count_node() and s_op->nr_cached_objects.  The shrinker is
unregistered now before ->kill_sb() so the operation is safe when we are
doing unmount.

The impact will depend heavily on the machine and the workload but for a
small machine using postmark tuned to use 4xRAM size the results were

                                  3.15.0-rc5            3.15.0-rc5
                                     vanilla         shrinker-v1r1
Ops/sec Transactions         21.00 (  0.00%)       24.00 ( 14.29%)
Ops/sec FilesCreate          39.00 (  0.00%)       44.00 ( 12.82%)
Ops/sec CreateTransact       10.00 (  0.00%)       12.00 ( 20.00%)
Ops/sec FilesDeleted       6202.00 (  0.00%)     6202.00 (  0.00%)
Ops/sec DeleteTransact       11.00 (  0.00%)       12.00 (  9.09%)
Ops/sec DataRead/MB          25.97 (  0.00%)       29.10 ( 12.05%)
Ops/sec DataWrite/MB         49.99 (  0.00%)       56.02 ( 12.06%)

ffsb running in a configuration that is meant to simulate a mail server showed

                                 3.15.0-rc5             3.15.0-rc5
                                    vanilla          shrinker-v1r1
Ops/sec readall           9402.63 (  0.00%)      9567.97 (  1.76%)
Ops/sec create            4695.45 (  0.00%)      4735.00 (  0.84%)
Ops/sec delete             173.72 (  0.00%)       179.83 (  3.52%)
Ops/sec Transactions     14271.80 (  0.00%)     14482.81 (  1.48%)
Ops/sec Read                37.00 (  0.00%)        37.60 (  1.62%)
Ops/sec Write               18.20 (  0.00%)        18.30 (  0.55%)

Signed-off-by: Tim Chen <tim.c.chen@linux.intel.com>
Signed-off-by: Mel Gorman <mgorman@suse.de>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Hugh Dickins <hughd@google.com>
Cc: Dave Chinner <david@fromorbit.com>
Tested-by: Yuanhan Liu <yuanhan.liu@linux.intel.com>
Cc: Bob Liu <bob.liu@oracle.com>
Cc: Jan Kara <jack@suse.cz>
Acked-by: Rik van Riel <riel@redhat.com>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Mel Gorman <mgorman@suse.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agofs/superblock: unregister sb shrinker before ->kill_sb()
Dave Chinner [Wed, 4 Jun 2014 23:10:46 +0000 (16:10 -0700)]
fs/superblock: unregister sb shrinker before ->kill_sb()

commit 28f2cd4f6da24a1aa06c226618ed5ad69e13df64 upstream.

This series is aimed at regressions noticed during reclaim activity.  The
first two patches are shrinker patches that were posted ages ago but never
merged for reasons that are unclear to me.  I'm posting them again to see
if there was a reason they were dropped or if they just got lost.  Dave?
Time?  The last patch adjusts proportional reclaim.  Yuanhan Liu, can you
retest the vm scalability test cases on a larger machine?  Hugh, does this
work for you on the memcg test cases?

Based on ext4, I get the following results but unfortunately my larger
test machines are all unavailable so this is based on a relatively small
machine.

postmark
                                  3.15.0-rc5            3.15.0-rc5
                                     vanilla       proportion-v1r4
Ops/sec Transactions         21.00 (  0.00%)       25.00 ( 19.05%)
Ops/sec FilesCreate          39.00 (  0.00%)       45.00 ( 15.38%)
Ops/sec CreateTransact       10.00 (  0.00%)       12.00 ( 20.00%)
Ops/sec FilesDeleted       6202.00 (  0.00%)     6202.00 (  0.00%)
Ops/sec DeleteTransact       11.00 (  0.00%)       12.00 (  9.09%)
Ops/sec DataRead/MB          25.97 (  0.00%)       30.02 ( 15.59%)
Ops/sec DataWrite/MB         49.99 (  0.00%)       57.78 ( 15.58%)

ffsb (mail server simulator)
                                 3.15.0-rc5             3.15.0-rc5
                                    vanilla        proportion-v1r4
Ops/sec readall           9402.63 (  0.00%)      9805.74 (  4.29%)
Ops/sec create            4695.45 (  0.00%)      4781.39 (  1.83%)
Ops/sec delete             173.72 (  0.00%)       177.23 (  2.02%)
Ops/sec Transactions     14271.80 (  0.00%)     14764.37 (  3.45%)
Ops/sec Read                37.00 (  0.00%)        38.50 (  4.05%)
Ops/sec Write               18.20 (  0.00%)        18.50 (  1.65%)

dd of a large file
                                3.15.0-rc5            3.15.0-rc5
                                   vanilla       proportion-v1r4
WallTime DownloadTar       75.00 (  0.00%)       61.00 ( 18.67%)
WallTime DD               423.00 (  0.00%)      401.00 (  5.20%)
WallTime Delete             2.00 (  0.00%)        5.00 (-150.00%)

stutter (times mmap latency during large amounts of IO)

                            3.15.0-rc5            3.15.0-rc5
                               vanilla       proportion-v1r4
Unit >5ms Delays  80252.0000 (  0.00%)  81523.0000 ( -1.58%)
Unit Mmap min         8.2118 (  0.00%)      8.3206 ( -1.33%)
Unit Mmap mean       17.4614 (  0.00%)     17.2868 (  1.00%)
Unit Mmap stddev     24.9059 (  0.00%)     34.6771 (-39.23%)
Unit Mmap max      2811.6433 (  0.00%)   2645.1398 (  5.92%)
Unit Mmap 90%        20.5098 (  0.00%)     18.3105 ( 10.72%)
Unit Mmap 93%        22.9180 (  0.00%)     20.1751 ( 11.97%)
Unit Mmap 95%        25.2114 (  0.00%)     22.4988 ( 10.76%)
Unit Mmap 99%        46.1430 (  0.00%)     43.5952 (  5.52%)
Unit Ideal  Tput     85.2623 (  0.00%)     78.8906 (  7.47%)
Unit Tput min        44.0666 (  0.00%)     43.9609 (  0.24%)
Unit Tput mean       45.5646 (  0.00%)     45.2009 (  0.80%)
Unit Tput stddev      0.9318 (  0.00%)      1.1084 (-18.95%)
Unit Tput max        46.7375 (  0.00%)     46.7539 ( -0.04%)

This patch (of 3):

We will like to unregister the sb shrinker before ->kill_sb().  This will
allow cached objects to be counted without call to grab_super_passive() to
update ref count on sb.  We want to avoid locking during memory
reclamation especially when we are skipping the memory reclaim when we are
out of cached objects.

This is safe because grab_super_passive does a try-lock on the
sb->s_umount now, and so if we are in the unmount process, it won't ever
block.  That means what used to be a deadlock and races we were avoiding
by using grab_super_passive() is now:

        shrinker                        umount

        down_read(shrinker_rwsem)
                                        down_write(sb->s_umount)
                                        shrinker_unregister
                                          down_write(shrinker_rwsem)
                                            <blocks>
        grab_super_passive(sb)
          down_read_trylock(sb->s_umount)
            <fails>
        <shrinker aborts>
        ....
        <shrinkers finish running>
        up_read(shrinker_rwsem)
                                          <unblocks>
                                          <removes shrinker>
                                          up_write(shrinker_rwsem)
                                        ->kill_sb()
                                        ....

So it is safe to deregister the shrinker before ->kill_sb().

Signed-off-by: Tim Chen <tim.c.chen@linux.intel.com>
Signed-off-by: Mel Gorman <mgorman@suse.de>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Hugh Dickins <hughd@google.com>
Cc: Dave Chinner <david@fromorbit.com>
Tested-by: Yuanhan Liu <yuanhan.liu@linux.intel.com>
Cc: Bob Liu <bob.liu@oracle.com>
Cc: Jan Kara <jack@suse.cz>
Acked-by: Rik van Riel <riel@redhat.com>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Mel Gorman <mgorman@suse.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agomm: fix direct reclaim writeback regression
Hugh Dickins [Sat, 26 Jul 2014 19:58:23 +0000 (12:58 -0700)]
mm: fix direct reclaim writeback regression

commit 8bdd638091605dc66d92c57c4b80eb87fffc15f7 upstream.

Shortly before 3.16-rc1, Dave Jones reported:

  WARNING: CPU: 3 PID: 19721 at fs/xfs/xfs_aops.c:971
           xfs_vm_writepage+0x5ce/0x630 [xfs]()
  CPU: 3 PID: 19721 Comm: trinity-c61 Not tainted 3.15.0+ #3
  Call Trace:
    xfs_vm_writepage+0x5ce/0x630 [xfs]
    shrink_page_list+0x8f9/0xb90
    shrink_inactive_list+0x253/0x510
    shrink_lruvec+0x563/0x6c0
    shrink_zone+0x3b/0x100
    shrink_zones+0x1f1/0x3c0
    try_to_free_pages+0x164/0x380
    __alloc_pages_nodemask+0x822/0xc90
    alloc_pages_vma+0xaf/0x1c0
    handle_mm_fault+0xa31/0xc50
  etc.

 970   if (WARN_ON_ONCE((current->flags & (PF_MEMALLOC|PF_KSWAPD)) ==
 971                   PF_MEMALLOC))

I did not respond at the time, because a glance at the PageDirty block
in shrink_page_list() quickly shows that this is impossible: we don't do
writeback on file pages (other than tmpfs) from direct reclaim nowadays.
Dave was hallucinating, but it would have been disrespectful to say so.

However, my own /var/log/messages now shows similar complaints

  WARNING: CPU: 1 PID: 28814 at fs/ext4/inode.c:1881 ext4_writepage+0xa7/0x38b()
  WARNING: CPU: 0 PID: 27347 at fs/ext4/inode.c:1764 ext4_writepage+0xa7/0x38b()

from stressing some mmotm trees during July.

Could a dirty xfs or ext4 file page somehow get marked PageSwapBacked,
so fail shrink_page_list()'s page_is_file_cache() test, and so proceed
to mapping->a_ops->writepage()?

Yes, 3.16-rc1's commit 68711a746345 ("mm, migration: add destination
page freeing callback") has provided such a way to compaction: if
migrating a SwapBacked page fails, its newpage may be put back on the
list for later use with PageSwapBacked still set, and nothing will clear
it.

Whether that can do anything worse than issue WARN_ON_ONCEs, and get
some statistics wrong, is unclear: easier to fix than to think through
the consequences.

Fixing it here, before the put_new_page(), addresses the bug directly,
but is probably the worst place to fix it.  Page migration is doing too
many parts of the job on too many levels: fixing it in
move_to_new_page() to complement its SetPageSwapBacked would be
preferable, except why is it (and newpage->mapping and newpage->index)
done there, rather than down in migrate_page_move_mapping(), once we are
sure of success? Not a cleanup to get into right now, especially not
with memcg cleanups coming in 3.17.

Reported-by: Dave Jones <davej@redhat.com>
Signed-off-by: Hugh Dickins <hughd@google.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Mel Gorman <mgorman@suse.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agox86/mm: In the PTE swapout page reclaim case clear the accessed bit instead of flushi...
Shaohua Li [Tue, 8 Apr 2014 07:58:09 +0000 (15:58 +0800)]
x86/mm: In the PTE swapout page reclaim case clear the accessed bit instead of flushing the TLB

commit b13b1d2d8692b437203de7a404c6b809d2cc4d99 upstream.

We use the accessed bit to age a page at page reclaim time,
and currently we also flush the TLB when doing so.

But in some workloads TLB flush overhead is very heavy. In my
simple multithreaded app with a lot of swap to several pcie
SSDs, removing the tlb flush gives about 20% ~ 30% swapout
speedup.

Fortunately just removing the TLB flush is a valid optimization:
on x86 CPUs, clearing the accessed bit without a TLB flush
doesn't cause data corruption.

It could cause incorrect page aging and the (mistaken) reclaim of
hot pages, but the chance of that should be relatively low.

So as a performance optimization don't flush the TLB when
clearing the accessed bit, it will eventually be flushed by
a context switch or a VM operation anyway. [ In the rare
event of it not getting flushed for a long time the delay
shouldn't really matter because there's no real memory
pressure for swapout to react to. ]

Suggested-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Shaohua Li <shli@fusionio.com>
Acked-by: Rik van Riel <riel@redhat.com>
Acked-by: Mel Gorman <mgorman@suse.de>
Acked-by: Hugh Dickins <hughd@google.com>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Cc: linux-mm@kvack.org
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Link: http://lkml.kernel.org/r/20140408075809.GA1764@kernel.org
[ Rewrote the changelog and the code comments. ]
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Signed-off-by: Mel Gorman <mgorman@suse.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agomm, compaction: properly signal and act upon lock and need_sched() contention
Vlastimil Babka [Wed, 4 Jun 2014 23:10:41 +0000 (16:10 -0700)]
mm, compaction: properly signal and act upon lock and need_sched() contention

commit be9765722e6b7ece8263cbab857490332339bd6f upstream.

Compaction uses compact_checklock_irqsave() function to periodically check
for lock contention and need_resched() to either abort async compaction,
or to free the lock, schedule and retake the lock.  When aborting,
cc->contended is set to signal the contended state to the caller.  Two
problems have been identified in this mechanism.

First, compaction also calls directly cond_resched() in both scanners when
no lock is yet taken.  This call either does not abort async compaction,
or set cc->contended appropriately.  This patch introduces a new
compact_should_abort() function to achieve both.  In isolate_freepages(),
the check frequency is reduced to once by SWAP_CLUSTER_MAX pageblocks to
match what the migration scanner does in the preliminary page checks.  In
case a pageblock is found suitable for calling isolate_freepages_block(),
the checks within there are done on higher frequency.

Second, isolate_freepages() does not check if isolate_freepages_block()
aborted due to contention, and advances to the next pageblock.  This
violates the principle of aborting on contention, and might result in
pageblocks not being scanned completely, since the scanning cursor is
advanced.  This problem has been noticed in the code by Joonsoo Kim when
reviewing related patches.  This patch makes isolate_freepages_block()
check the cc->contended flag and abort.

In case isolate_freepages() has already isolated some pages before
aborting due to contention, page migration will proceed, which is OK since
we do not want to waste the work that has been done, and page migration
has own checks for contention.  However, we do not want another isolation
attempt by either of the scanners, so cc->contended flag check is added
also to compaction_alloc() and compact_finished() to make sure compaction
is aborted right after the migration.

The outcome of the patch should be reduced lock contention by async
compaction and lower latencies for higher-order allocations where direct
compaction is involved.

[akpm@linux-foundation.org: fix typo in comment]
Reported-by: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Signed-off-by: Vlastimil Babka <vbabka@suse.cz>
Reviewed-by: Naoya Horiguchi <n-horiguchi@ah.jp.nec.com>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Mel Gorman <mgorman@suse.de>
Cc: Bartlomiej Zolnierkiewicz <b.zolnierkie@samsung.com>
Cc: Michal Nazarewicz <mina86@mina86.com>
Cc: Christoph Lameter <cl@linux.com>
Cc: Rik van Riel <riel@redhat.com>
Acked-by: Michal Nazarewicz <mina86@mina86.com>
Tested-by: Shawn Guo <shawn.guo@linaro.org>
Tested-by: Kevin Hilman <khilman@linaro.org>
Tested-by: Stephen Warren <swarren@nvidia.com>
Tested-by: Fabio Estevam <fabio.estevam@freescale.com>
Cc: David Rientjes <rientjes@google.com>
Cc: Stephen Rothwell <sfr@canb.auug.org.au>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Mel Gorman <mgorman@suse.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agomm/compaction: avoid rescanning pageblocks in isolate_freepages
Vlastimil Babka [Wed, 4 Jun 2014 23:08:34 +0000 (16:08 -0700)]
mm/compaction: avoid rescanning pageblocks in isolate_freepages

commit e9ade569910a82614ff5f2c2cea2b65a8d785da4 upstream.

The compaction free scanner in isolate_freepages() currently remembers PFN
of the highest pageblock where it successfully isolates, to be used as the
starting pageblock for the next invocation.  The rationale behind this is
that page migration might return free pages to the allocator when
migration fails and we don't want to skip them if the compaction
continues.

Since migration now returns free pages back to compaction code where they
can be reused, this is no longer a concern.  This patch changes
isolate_freepages() so that the PFN for restarting is updated with each
pageblock where isolation is attempted.  Using stress-highalloc from
mmtests, this resulted in 10% reduction of the pages scanned by the free
scanner.

Note that the somewhat similar functionality that records highest
successful pageblock in zone->compact_cached_free_pfn, remains unchanged.
This cache is used when the whole compaction is restarted, not for
multiple invocations of the free scanner during single compaction.

Signed-off-by: Vlastimil Babka <vbabka@suse.cz>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Mel Gorman <mgorman@suse.de>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: Bartlomiej Zolnierkiewicz <b.zolnierkie@samsung.com>
Acked-by: Michal Nazarewicz <mina86@mina86.com>
Reviewed-by: Naoya Horiguchi <n-horiguchi@ah.jp.nec.com>
Cc: Christoph Lameter <cl@linux.com>
Cc: Rik van Riel <riel@redhat.com>
Acked-by: David Rientjes <rientjes@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Mel Gorman <mgorman@suse.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agomm/compaction: do not count migratepages when unnecessary
Vlastimil Babka [Wed, 4 Jun 2014 23:08:32 +0000 (16:08 -0700)]
mm/compaction: do not count migratepages when unnecessary

commit f8c9301fa5a2a8b873c67f2a3d8230d5c13f61b7 upstream.

During compaction, update_nr_listpages() has been used to count remaining
non-migrated and free pages after a call to migrage_pages().  The
freepages counting has become unneccessary, and it turns out that
migratepages counting is also unnecessary in most cases.

The only situation when it's needed to count cc->migratepages is when
migrate_pages() returns with a negative error code.  Otherwise, the
non-negative return value is the number of pages that were not migrated,
which is exactly the count of remaining pages in the cc->migratepages
list.

Furthermore, any non-zero count is only interesting for the tracepoint of
mm_compaction_migratepages events, because after that all remaining
unmigrated pages are put back and their count is set to 0.

This patch therefore removes update_nr_listpages() completely, and changes
the tracepoint definition so that the manual counting is done only when
the tracepoint is enabled, and only when migrate_pages() returns a
negative error code.

Furthermore, migrate_pages() and the tracepoints won't be called when
there's nothing to migrate.  This potentially avoids some wasted cycles
and reduces the volume of uninteresting mm_compaction_migratepages events
where "nr_migrated=0 nr_failed=0".  In the stress-highalloc mmtest, this
was about 75% of the events.  The mm_compaction_isolate_migratepages event
is better for determining that nothing was isolated for migration, and
this one was just duplicating the info.

Signed-off-by: Vlastimil Babka <vbabka@suse.cz>
Reviewed-by: Naoya Horiguchi <n-horiguchi@ah.jp.nec.com>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Mel Gorman <mgorman@suse.de>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: Bartlomiej Zolnierkiewicz <b.zolnierkie@samsung.com>
Acked-by: Michal Nazarewicz <mina86@mina86.com>
Cc: Christoph Lameter <cl@linux.com>
Cc: Rik van Riel <riel@redhat.com>
Acked-by: David Rientjes <rientjes@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Mel Gorman <mgorman@suse.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agomm, compaction: terminate async compaction when rescheduling
David Rientjes [Wed, 4 Jun 2014 23:08:31 +0000 (16:08 -0700)]
mm, compaction: terminate async compaction when rescheduling

commit aeef4b83806f49a0c454b7d4578671b71045bee2 upstream.

Async compaction terminates prematurely when need_resched(), see
compact_checklock_irqsave().  This can never trigger, however, if the
cond_resched() in isolate_migratepages_range() always takes care of the
scheduling.

If the cond_resched() actually triggers, then terminate this pageblock
scan for async compaction as well.

Signed-off-by: David Rientjes <rientjes@google.com>
Acked-by: Mel Gorman <mgorman@suse.de>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: Mel Gorman <mgorman@suse.de>
Cc: Vlastimil Babka <vbabka@suse.cz>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Mel Gorman <mgorman@suse.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agomm, compaction: embed migration mode in compact_control
David Rientjes [Wed, 4 Jun 2014 23:08:28 +0000 (16:08 -0700)]
mm, compaction: embed migration mode in compact_control

commit e0b9daeb453e602a95ea43853dc12d385558ce1f upstream.

We're going to want to manipulate the migration mode for compaction in the
page allocator, and currently compact_control's sync field is only a bool.

Currently, we only do MIGRATE_ASYNC or MIGRATE_SYNC_LIGHT compaction
depending on the value of this bool.  Convert the bool to enum
migrate_mode and pass the migration mode in directly.  Later, we'll want
to avoid MIGRATE_SYNC_LIGHT for thp allocations in the pagefault patch to
avoid unnecessary latency.

This also alters compaction triggered from sysfs, either for the entire
system or for a node, to force MIGRATE_SYNC.

[akpm@linux-foundation.org: fix build]
[iamjoonsoo.kim@lge.com: use MIGRATE_SYNC in alloc_contig_range()]
Signed-off-by: David Rientjes <rientjes@google.com>
Suggested-by: Mel Gorman <mgorman@suse.de>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: Greg Thelen <gthelen@google.com>
Cc: Naoya Horiguchi <n-horiguchi@ah.jp.nec.com>
Signed-off-by: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Mel Gorman <mgorman@suse.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agomm, compaction: add per-zone migration pfn cache for async compaction
David Rientjes [Wed, 4 Jun 2014 23:08:27 +0000 (16:08 -0700)]
mm, compaction: add per-zone migration pfn cache for async compaction

commit 35979ef3393110ff3c12c6b94552208d3bdf1a36 upstream.

Each zone has a cached migration scanner pfn for memory compaction so that
subsequent calls to memory compaction can start where the previous call
left off.

Currently, the compaction migration scanner only updates the per-zone
cached pfn when pageblocks were not skipped for async compaction.  This
creates a dependency on calling sync compaction to avoid having subsequent
calls to async compaction from scanning an enormous amount of non-MOVABLE
pageblocks each time it is called.  On large machines, this could be
potentially very expensive.

This patch adds a per-zone cached migration scanner pfn only for async
compaction.  It is updated everytime a pageblock has been scanned in its
entirety and when no pages from it were successfully isolated.  The cached
migration scanner pfn for sync compaction is updated only when called for
sync compaction.

Signed-off-by: David Rientjes <rientjes@google.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Reviewed-by: Naoya Horiguchi <n-horiguchi@ah.jp.nec.com>
Cc: Greg Thelen <gthelen@google.com>
Cc: Mel Gorman <mgorman@suse.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Mel Gorman <mgorman@suse.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agomm, compaction: return failed migration target pages back to freelist
David Rientjes [Wed, 4 Jun 2014 23:08:26 +0000 (16:08 -0700)]
mm, compaction: return failed migration target pages back to freelist

commit d53aea3d46d64e95da9952887969f7533b9ab25e upstream.

Greg reported that he found isolated free pages were returned back to the
VM rather than the compaction freelist.  This will cause holes behind the
free scanner and cause it to reallocate additional memory if necessary
later.

He detected the problem at runtime seeing that ext4 metadata pages (esp
the ones read by "sbi->s_group_desc[i] = sb_bread(sb, block)") were
constantly visited by compaction calls of migrate_pages().  These pages
had a non-zero b_count which caused fallback_migrate_page() ->
try_to_release_page() -> try_to_free_buffers() to fail.

Memory compaction works by having a "freeing scanner" scan from one end of
a zone which isolates pages as migration targets while another "migrating
scanner" scans from the other end of the same zone which isolates pages
for migration.

When page migration fails for an isolated page, the target page is
returned to the system rather than the freelist built by the freeing
scanner.  This may require the freeing scanner to continue scanning memory
after suitable migration targets have already been returned to the system
needlessly.

This patch returns destination pages to the freeing scanner freelist when
page migration fails.  This prevents unnecessary work done by the freeing
scanner but also encourages memory to be as compacted as possible at the
end of the zone.

Signed-off-by: David Rientjes <rientjes@google.com>
Reported-by: Greg Thelen <gthelen@google.com>
Acked-by: Mel Gorman <mgorman@suse.de>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Reviewed-by: Naoya Horiguchi <n-horiguchi@ah.jp.nec.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Mel Gorman <mgorman@suse.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agomm, migration: add destination page freeing callback
David Rientjes [Wed, 4 Jun 2014 23:08:25 +0000 (16:08 -0700)]
mm, migration: add destination page freeing callback

commit 68711a746345c44ae00c64d8dbac6a9ce13ac54a upstream.

Memory migration uses a callback defined by the caller to determine how to
allocate destination pages.  When migration fails for a source page,
however, it frees the destination page back to the system.

This patch adds a memory migration callback defined by the caller to
determine how to free destination pages.  If a caller, such as memory
compaction, builds its own freelist for migration targets, this can reuse
already freed memory instead of scanning additional memory.

If the caller provides a function to handle freeing of destination pages,
it is called when page migration fails.  If the caller passes NULL then
freeing back to the system will be handled as usual.  This patch
introduces no functional change.

Signed-off-by: David Rientjes <rientjes@google.com>
Reviewed-by: Naoya Horiguchi <n-horiguchi@ah.jp.nec.com>
Acked-by: Mel Gorman <mgorman@suse.de>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: Greg Thelen <gthelen@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Mel Gorman <mgorman@suse.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agomm/compaction: cleanup isolate_freepages()
Vlastimil Babka [Wed, 4 Jun 2014 23:07:26 +0000 (16:07 -0700)]
mm/compaction: cleanup isolate_freepages()

commit c96b9e508f3d06ddb601dcc9792d62c044ab359e upstream.

isolate_freepages() is currently somewhat hard to follow thanks to many
looks like it is related to the 'low_pfn' variable, but in fact it is not.

This patch renames the 'high_pfn' variable to a hopefully less confusing name,
and slightly changes its handling without a functional change. A comment made
obsolete by recent changes is also updated.

[akpm@linux-foundation.org: comment fixes, per Minchan]
[iamjoonsoo.kim@lge.com: cleanups]
Signed-off-by: Vlastimil Babka <vbabka@suse.cz>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Mel Gorman <mgorman@suse.de>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: Bartlomiej Zolnierkiewicz <b.zolnierkie@samsung.com>
Cc: Michal Nazarewicz <mina86@mina86.com>
Cc: Naoya Horiguchi <n-horiguchi@ah.jp.nec.com>
Cc: Christoph Lameter <cl@linux.com>
Cc: Rik van Riel <riel@redhat.com>
Cc: Dongjun Shin <d.j.shin@samsung.com>
Cc: Sunghwan Yun <sunghwan.yun@samsung.com>
Signed-off-by: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Mel Gorman <mgorman@suse.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agomm/compaction: clean up unused code lines
Heesub Shin [Wed, 4 Jun 2014 23:07:24 +0000 (16:07 -0700)]
mm/compaction: clean up unused code lines

commit 13fb44e4b0414d7e718433a49e6430d5b76bd46e upstream.

Remove code lines currently not in use or never called.

Signed-off-by: Heesub Shin <heesub.shin@samsung.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: Dongjun Shin <d.j.shin@samsung.com>
Cc: Sunghwan Yun <sunghwan.yun@samsung.com>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Mel Gorman <mgorman@suse.de>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: Bartlomiej Zolnierkiewicz <b.zolnierkie@samsung.com>
Cc: Michal Nazarewicz <mina86@mina86.com>
Cc: Naoya Horiguchi <n-horiguchi@ah.jp.nec.com>
Cc: Christoph Lameter <cl@linux.com>
Cc: Rik van Riel <riel@redhat.com>
Cc: Dongjun Shin <d.j.shin@samsung.com>
Cc: Sunghwan Yun <sunghwan.yun@samsung.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Mel Gorman <mgorman@suse.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agomm/readahead.c: inline ra_submit
Fabian Frederick [Mon, 7 Apr 2014 22:37:55 +0000 (15:37 -0700)]
mm/readahead.c: inline ra_submit

commit 29f175d125f0f3a9503af8a5596f93d714cceb08 upstream.

Commit f9acc8c7b35a ("readahead: sanify file_ra_state names") left
ra_submit with a single function call.

Move ra_submit to internal.h and inline it to save some stack.  Thanks
to Andrew Morton for commenting different versions.

Signed-off-by: Fabian Frederick <fabf@skynet.be>
Suggested-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Mel Gorman <mgorman@suse.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agocallers of iov_copy_from_user_atomic() don't need pagecache_disable()
Al Viro [Mon, 3 Feb 2014 03:10:25 +0000 (22:10 -0500)]
callers of iov_copy_from_user_atomic() don't need pagecache_disable()

commit 9e8c2af96e0d2d5fe298dd796fb6bc16e888a48d upstream.

... it does that itself (via kmap_atomic())

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Mel Gorman <mgorman@suse.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agomm: remove read_cache_page_async()
Sasha Levin [Thu, 3 Apr 2014 21:48:18 +0000 (14:48 -0700)]
mm: remove read_cache_page_async()

commit 67f9fd91f93c582b7de2ab9325b6e179db77e4d5 upstream.

This patch removes read_cache_page_async() which wasn't really needed
anywhere and simplifies the code around it a bit.

read_cache_page_async() is useful when we want to read a page into the
cache without waiting for it to complete.  This happens when the
appropriate callback 'filler' doesn't complete its read operation and
releases the page lock immediately, and instead queues a different
completion routine to do that.  This never actually happened anywhere in
the code.

read_cache_page_async() had 3 different callers:

- read_cache_page() which is the sync version, it would just wait for
  the requested read to complete using wait_on_page_read().

- JFFS2 would call it from jffs2_gc_fetch_page(), but the filler
  function it supplied doesn't do any async reads, and would complete
  before the filler function returns - making it actually a sync read.

- CRAMFS would call it using the read_mapping_page_async() wrapper, with
  a similar story to JFFS2 - the filler function doesn't do anything that
  reminds async reads and would always complete before the filler function
  returns.

To sum it up, the code in mm/filemap.c never took advantage of having
read_cache_page_async().  While there are filler callbacks that do async
reads (such as the block one), we always called it with the
read_cache_page().

This patch adds a mandatory wait for read to complete when adding a new
page to the cache, and removes read_cache_page_async() and its wrappers.

Signed-off-by: Sasha Levin <sasha.levin@oracle.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Mel Gorman <mgorman@suse.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agomm: madvise: fix MADV_WILLNEED on shmem swapouts
Johannes Weiner [Thu, 22 May 2014 18:54:17 +0000 (11:54 -0700)]
mm: madvise: fix MADV_WILLNEED on shmem swapouts

commit 55231e5c898c5c03c14194001e349f40f59bd300 upstream.

MADV_WILLNEED currently does not read swapped out shmem pages back in.

Commit 0cd6144aadd2 ("mm + fs: prepare for non-page entries in page
cache radix trees") made find_get_page() filter exceptional radix tree
entries but failed to convert all find_get_page() callers that WANT
exceptional entries over to find_get_entry().  One of them is shmem swap
readahead in madvise, which now skips over any swap-out records.

Convert it to find_get_entry().

Fixes: 0cd6144aadd2 ("mm + fs: prepare for non-page entries in page cache radix trees")
Signed-off-by: Johannes Weiner <hannes@cmpxchg.org>
Reported-by: Hugh Dickins <hughd@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Mel Gorman <mgorman@suse.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agomm + fs: prepare for non-page entries in page cache radix trees
Johannes Weiner [Thu, 3 Apr 2014 21:47:46 +0000 (14:47 -0700)]
mm + fs: prepare for non-page entries in page cache radix trees

commit 0cd6144aadd2afd19d1aca880153530c52957604 upstream.

shmem mappings already contain exceptional entries where swap slot
information is remembered.

To be able to store eviction information for regular page cache, prepare
every site dealing with the radix trees directly to handle entries other
than pages.

The common lookup functions will filter out non-page entries and return
NULL for page cache holes, just as before.  But provide a raw version of
the API which returns non-page entries as well, and switch shmem over to
use it.

Signed-off-by: Johannes Weiner <hannes@cmpxchg.org>
Reviewed-by: Rik van Riel <riel@redhat.com>
Reviewed-by: Minchan Kim <minchan@kernel.org>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Bob Liu <bob.liu@oracle.com>
Cc: Christoph Hellwig <hch@infradead.org>
Cc: Dave Chinner <david@fromorbit.com>
Cc: Greg Thelen <gthelen@google.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Jan Kara <jack@suse.cz>
Cc: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Cc: Luigi Semenzato <semenzato@google.com>
Cc: Mel Gorman <mgorman@suse.de>
Cc: Metin Doslu <metin@citusdata.com>
Cc: Michel Lespinasse <walken@google.com>
Cc: Ozgun Erdogan <ozgun@citusdata.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Roman Gushchin <klamm@yandex-team.ru>
Cc: Ryan Mallon <rmallon@gmail.com>
Cc: Tejun Heo <tj@kernel.org>
Cc: Vlastimil Babka <vbabka@suse.cz>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Mel Gorman <mgorman@suse.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agomm: filemap: move radix tree hole searching here
Johannes Weiner [Thu, 3 Apr 2014 21:47:44 +0000 (14:47 -0700)]
mm: filemap: move radix tree hole searching here

commit e7b563bb2a6f4d974208da46200784b9c5b5a47e upstream.

The radix tree hole searching code is only used for page cache, for
example the readahead code trying to get a a picture of the area
surrounding a fault.

It sufficed to rely on the radix tree definition of holes, which is
"empty tree slot".  But this is about to change, though, as shadow page
descriptors will be stored in the page cache after the actual pages get
evicted from memory.

Move the functions over to mm/filemap.c and make them native page cache
operations, where they can later be adapted to handle the new definition
of "page cache hole".

Signed-off-by: Johannes Weiner <hannes@cmpxchg.org>
Reviewed-by: Rik van Riel <riel@redhat.com>
Reviewed-by: Minchan Kim <minchan@kernel.org>
Acked-by: Mel Gorman <mgorman@suse.de>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Bob Liu <bob.liu@oracle.com>
Cc: Christoph Hellwig <hch@infradead.org>
Cc: Dave Chinner <david@fromorbit.com>
Cc: Greg Thelen <gthelen@google.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Jan Kara <jack@suse.cz>
Cc: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Cc: Luigi Semenzato <semenzato@google.com>
Cc: Metin Doslu <metin@citusdata.com>
Cc: Michel Lespinasse <walken@google.com>
Cc: Ozgun Erdogan <ozgun@citusdata.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Roman Gushchin <klamm@yandex-team.ru>
Cc: Ryan Mallon <rmallon@gmail.com>
Cc: Tejun Heo <tj@kernel.org>
Cc: Vlastimil Babka <vbabka@suse.cz>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Mel Gorman <mgorman@suse.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agomm: shmem: save one radix tree lookup when truncating swapped pages
Johannes Weiner [Thu, 3 Apr 2014 21:47:41 +0000 (14:47 -0700)]
mm: shmem: save one radix tree lookup when truncating swapped pages

commit 6dbaf22ce1f1dfba33313198eb5bd989ae76dd87 upstream.

Page cache radix tree slots are usually stabilized by the page lock, but
shmem's swap cookies have no such thing.  Because the overall truncation
loop is lockless, the swap entry is currently confirmed by a tree lookup
and then deleted by another tree lookup under the same tree lock region.

Use radix_tree_delete_item() instead, which does the verification and
deletion with only one lookup.  This also allows removing the
delete-only special case from shmem_radix_tree_replace().

Signed-off-by: Johannes Weiner <hannes@cmpxchg.org>
Reviewed-by: Minchan Kim <minchan@kernel.org>
Reviewed-by: Rik van Riel <riel@redhat.com>
Acked-by: Mel Gorman <mgorman@suse.de>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Bob Liu <bob.liu@oracle.com>
Cc: Christoph Hellwig <hch@infradead.org>
Cc: Dave Chinner <david@fromorbit.com>
Cc: Greg Thelen <gthelen@google.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Jan Kara <jack@suse.cz>
Cc: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Cc: Luigi Semenzato <semenzato@google.com>
Cc: Metin Doslu <metin@citusdata.com>
Cc: Michel Lespinasse <walken@google.com>
Cc: Ozgun Erdogan <ozgun@citusdata.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Roman Gushchin <klamm@yandex-team.ru>
Cc: Ryan Mallon <rmallon@gmail.com>
Cc: Tejun Heo <tj@kernel.org>
Cc: Vlastimil Babka <vbabka@suse.cz>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Mel Gorman <mgorman@suse.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agolib: radix-tree: add radix_tree_delete_item()
Johannes Weiner [Thu, 3 Apr 2014 21:47:39 +0000 (14:47 -0700)]
lib: radix-tree: add radix_tree_delete_item()

commit 53c59f262d747ea82e7414774c59a489501186a0 upstream.

Provide a function that does not just delete an entry at a given index,
but also allows passing in an expected item.  Delete only if that item
is still located at the specified index.

This is handy when lockless tree traversals want to delete entries as
well because they don't have to do an second, locked lookup to verify
the slot has not changed under them before deleting the entry.

Signed-off-by: Johannes Weiner <hannes@cmpxchg.org>
Reviewed-by: Minchan Kim <minchan@kernel.org>
Reviewed-by: Rik van Riel <riel@redhat.com>
Acked-by: Mel Gorman <mgorman@suse.de>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Bob Liu <bob.liu@oracle.com>
Cc: Christoph Hellwig <hch@infradead.org>
Cc: Dave Chinner <david@fromorbit.com>
Cc: Greg Thelen <gthelen@google.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Jan Kara <jack@suse.cz>
Cc: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Cc: Luigi Semenzato <semenzato@google.com>
Cc: Metin Doslu <metin@citusdata.com>
Cc: Michel Lespinasse <walken@google.com>
Cc: Ozgun Erdogan <ozgun@citusdata.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Roman Gushchin <klamm@yandex-team.ru>
Cc: Ryan Mallon <rmallon@gmail.com>
Cc: Tejun Heo <tj@kernel.org>
Cc: Vlastimil Babka <vbabka@suse.cz>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Mel Gorman <mgorman@suse.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agoregmap: fix kernel hang on regmap_bulk_write with zero val_count.
Quentin Casasnovas [Wed, 12 Nov 2014 10:19:23 +0000 (11:19 +0100)]
regmap: fix kernel hang on regmap_bulk_write with zero val_count.

Fixes commit 2f06fa04cf35da5c24481da3ac84a2900d0b99c3 which was an
incorrect backported version of commit
d6b41cb06044a7d895db82bdd54f6e4219970510 upstream.

If val_count is zero we return -EINVAL with map->lock_arg locked, which
will deadlock the kernel next time we try to acquire this lock.

This was introduced by f5942dd ("regmap: fix possible ZERO_SIZE_PTR pointer
dereferencing error.") which improperly back-ported d6b41cb0.

This issue was found during review of Ubuntu Trusty 3.13.0-40.68 kernel to
prepare Ksplice rebootless updates.

Fixes: f5942dd ("regmap: fix possible ZERO_SIZE_PTR pointer dereferencing error.")
Signed-off-by: Quentin Casasnovas <quentin.casasnovas@oracle.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agoiwlwifi: configure the LTR
Emmanuel Grumbach [Tue, 23 Sep 2014 20:02:41 +0000 (23:02 +0300)]
iwlwifi: configure the LTR

commit 9180ac50716a097a407c6d7e7e4589754a922260 upstream.

The LTR is the handshake between the device and the root
complex about the latency allowed when the bus exits power
save. This configuration was missing and this led to high
latency in the link power up. The end user could experience
high latency in the network because of this.

Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agonet: sctp: fix skb_over_panic when receiving malformed ASCONF chunks
Daniel Borkmann [Thu, 9 Oct 2014 20:55:31 +0000 (22:55 +0200)]
net: sctp: fix skb_over_panic when receiving malformed ASCONF chunks

commit 9de7922bc709eee2f609cd01d98aaedc4cf5ea74 upstream.

Commit 6f4c618ddb0 ("SCTP : Add paramters validity check for
ASCONF chunk") added basic verification of ASCONF chunks, however,
it is still possible to remotely crash a server by sending a
special crafted ASCONF chunk, even up to pre 2.6.12 kernels:

skb_over_panic: text:ffffffffa01ea1c3 len:31056 put:30768
 head:ffff88011bd81800 data:ffff88011bd81800 tail:0x7950
 end:0x440 dev:<NULL>
 ------------[ cut here ]------------
kernel BUG at net/core/skbuff.c:129!
[...]
Call Trace:
 <IRQ>
 [<ffffffff8144fb1c>] skb_put+0x5c/0x70
 [<ffffffffa01ea1c3>] sctp_addto_chunk+0x63/0xd0 [sctp]
 [<ffffffffa01eadaf>] sctp_process_asconf+0x1af/0x540 [sctp]
 [<ffffffff8152d025>] ? _read_unlock_bh+0x15/0x20
 [<ffffffffa01e0038>] sctp_sf_do_asconf+0x168/0x240 [sctp]
 [<ffffffffa01e3751>] sctp_do_sm+0x71/0x1210 [sctp]
 [<ffffffff8147645d>] ? fib_rules_lookup+0xad/0xf0
 [<ffffffffa01e6b22>] ? sctp_cmp_addr_exact+0x32/0x40 [sctp]
 [<ffffffffa01e8393>] sctp_assoc_bh_rcv+0xd3/0x180 [sctp]
 [<ffffffffa01ee986>] sctp_inq_push+0x56/0x80 [sctp]
 [<ffffffffa01fcc42>] sctp_rcv+0x982/0xa10 [sctp]
 [<ffffffffa01d5123>] ? ipt_local_in_hook+0x23/0x28 [iptable_filter]
 [<ffffffff8148bdc9>] ? nf_iterate+0x69/0xb0
 [<ffffffff81496d10>] ? ip_local_deliver_finish+0x0/0x2d0
 [<ffffffff8148bf86>] ? nf_hook_slow+0x76/0x120
 [<ffffffff81496d10>] ? ip_local_deliver_finish+0x0/0x2d0
 [<ffffffff81496ded>] ip_local_deliver_finish+0xdd/0x2d0
 [<ffffffff81497078>] ip_local_deliver+0x98/0xa0
 [<ffffffff8149653d>] ip_rcv_finish+0x12d/0x440
 [<ffffffff81496ac5>] ip_rcv+0x275/0x350
 [<ffffffff8145c88b>] __netif_receive_skb+0x4ab/0x750
 [<ffffffff81460588>] netif_receive_skb+0x58/0x60

This can be triggered e.g., through a simple scripted nmap
connection scan injecting the chunk after the handshake, for
example, ...

  -------------- INIT[ASCONF; ASCONF_ACK] ------------->
  <----------- INIT-ACK[ASCONF; ASCONF_ACK] ------------
  -------------------- COOKIE-ECHO -------------------->
  <-------------------- COOKIE-ACK ---------------------
  ------------------ ASCONF; UNKNOWN ------------------>

... where ASCONF chunk of length 280 contains 2 parameters ...

  1) Add IP address parameter (param length: 16)
  2) Add/del IP address parameter (param length: 255)

... followed by an UNKNOWN chunk of e.g. 4 bytes. Here, the
Address Parameter in the ASCONF chunk is even missing, too.
This is just an example and similarly-crafted ASCONF chunks
could be used just as well.

The ASCONF chunk passes through sctp_verify_asconf() as all
parameters passed sanity checks, and after walking, we ended
up successfully at the chunk end boundary, and thus may invoke
sctp_process_asconf(). Parameter walking is done with
WORD_ROUND() to take padding into account.

In sctp_process_asconf()'s TLV processing, we may fail in
sctp_process_asconf_param() e.g., due to removal of the IP
address that is also the source address of the packet containing
the ASCONF chunk, and thus we need to add all TLVs after the
failure to our ASCONF response to remote via helper function
sctp_add_asconf_response(), which basically invokes a
sctp_addto_chunk() adding the error parameters to the given
skb.

When walking to the next parameter this time, we proceed
with ...

  length = ntohs(asconf_param->param_hdr.length);
  asconf_param = (void *)asconf_param + length;

... instead of the WORD_ROUND()'ed length, thus resulting here
in an off-by-one that leads to reading the follow-up garbage
parameter length of 12336, and thus throwing an skb_over_panic
for the reply when trying to sctp_addto_chunk() next time,
which implicitly calls the skb_put() with that length.

Fix it by using sctp_walk_params() [ which is also used in
INIT parameter processing ] macro in the verification *and*
in ASCONF processing: it will make sure we don't spill over,
that we walk parameters WORD_ROUND()'ed. Moreover, we're being
more defensive and guard against unknown parameter types and
missized addresses.

Joint work with Vlad Yasevich.

Fixes: b896b82be4ae ("[SCTP] ADDIP: Support for processing incoming ASCONF_ACK chunks.")
Signed-off-by: Daniel Borkmann <dborkman@redhat.com>
Signed-off-by: Vlad Yasevich <vyasevich@gmail.com>
Acked-by: Neil Horman <nhorman@tuxdriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Cc: Josh Boyer <jwboyer@fedoraproject.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agonet: sctp: fix panic on duplicate ASCONF chunks
Daniel Borkmann [Thu, 9 Oct 2014 20:55:32 +0000 (22:55 +0200)]
net: sctp: fix panic on duplicate ASCONF chunks

commit b69040d8e39f20d5215a03502a8e8b4c6ab78395 upstream.

When receiving a e.g. semi-good formed connection scan in the
form of ...

  -------------- INIT[ASCONF; ASCONF_ACK] ------------->
  <----------- INIT-ACK[ASCONF; ASCONF_ACK] ------------
  -------------------- COOKIE-ECHO -------------------->
  <-------------------- COOKIE-ACK ---------------------
  ---------------- ASCONF_a; ASCONF_b ----------------->

... where ASCONF_a equals ASCONF_b chunk (at least both serials
need to be equal), we panic an SCTP server!

The problem is that good-formed ASCONF chunks that we reply with
ASCONF_ACK chunks are cached per serial. Thus, when we receive a
same ASCONF chunk twice (e.g. through a lost ASCONF_ACK), we do
not need to process them again on the server side (that was the
idea, also proposed in the RFC). Instead, we know it was cached
and we just resend the cached chunk instead. So far, so good.

Where things get nasty is in SCTP's side effect interpreter, that
is, sctp_cmd_interpreter():

While incoming ASCONF_a (chunk = event_arg) is being marked
!end_of_packet and !singleton, and we have an association context,
we do not flush the outqueue the first time after processing the
ASCONF_ACK singleton chunk via SCTP_CMD_REPLY. Instead, we keep it
queued up, although we set local_cork to 1. Commit 2e3216cd54b1
changed the precedence, so that as long as we get bundled, incoming
chunks we try possible bundling on outgoing queue as well. Before
this commit, we would just flush the output queue.

Now, while ASCONF_a's ASCONF_ACK sits in the corked outq, we
continue to process the same ASCONF_b chunk from the packet. As
we have cached the previous ASCONF_ACK, we find it, grab it and
do another SCTP_CMD_REPLY command on it. So, effectively, we rip
the chunk->list pointers and requeue the same ASCONF_ACK chunk
another time. Since we process ASCONF_b, it's correctly marked
with end_of_packet and we enforce an uncork, and thus flush, thus
crashing the kernel.

Fix it by testing if the ASCONF_ACK is currently pending and if
that is the case, do not requeue it. When flushing the output
queue we may relink the chunk for preparing an outgoing packet,
but eventually unlink it when it's copied into the skb right
before transmission.

Joint work with Vlad Yasevich.

Fixes: 2e3216cd54b1 ("sctp: Follow security requirement of responding with 1 packet")
Signed-off-by: Daniel Borkmann <dborkman@redhat.com>
Signed-off-by: Vlad Yasevich <vyasevich@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Cc: Josh Boyer <jwboyer@fedoraproject.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agonet: sctp: fix remote memory pressure from excessive queueing
Daniel Borkmann [Thu, 9 Oct 2014 20:55:33 +0000 (22:55 +0200)]
net: sctp: fix remote memory pressure from excessive queueing

commit 26b87c7881006311828bb0ab271a551a62dcceb4 upstream.

This scenario is not limited to ASCONF, just taken as one
example triggering the issue. When receiving ASCONF probes
in the form of ...

  -------------- INIT[ASCONF; ASCONF_ACK] ------------->
  <----------- INIT-ACK[ASCONF; ASCONF_ACK] ------------
  -------------------- COOKIE-ECHO -------------------->
  <-------------------- COOKIE-ACK ---------------------
  ---- ASCONF_a; [ASCONF_b; ...; ASCONF_n;] JUNK ------>
  [...]
  ---- ASCONF_m; [ASCONF_o; ...; ASCONF_z;] JUNK ------>

... where ASCONF_a, ASCONF_b, ..., ASCONF_z are good-formed
ASCONFs and have increasing serial numbers, we process such
ASCONF chunk(s) marked with !end_of_packet and !singleton,
since we have not yet reached the SCTP packet end. SCTP does
only do verification on a chunk by chunk basis, as an SCTP
packet is nothing more than just a container of a stream of
chunks which it eats up one by one.

We could run into the case that we receive a packet with a
malformed tail, above marked as trailing JUNK. All previous
chunks are here goodformed, so the stack will eat up all
previous chunks up to this point. In case JUNK does not fit
into a chunk header and there are no more other chunks in
the input queue, or in case JUNK contains a garbage chunk
header, but the encoded chunk length would exceed the skb
tail, or we came here from an entirely different scenario
and the chunk has pdiscard=1 mark (without having had a flush
point), it will happen, that we will excessively queue up
the association's output queue (a correct final chunk may
then turn it into a response flood when flushing the
queue ;)): I ran a simple script with incremental ASCONF
serial numbers and could see the server side consuming
excessive amount of RAM [before/after: up to 2GB and more].

The issue at heart is that the chunk train basically ends
with !end_of_packet and !singleton markers and since commit
2e3216cd54b1 ("sctp: Follow security requirement of responding
with 1 packet") therefore preventing an output queue flush
point in sctp_do_sm() -> sctp_cmd_interpreter() on the input
chunk (chunk = event_arg) even though local_cork is set,
but its precedence has changed since then. In the normal
case, the last chunk with end_of_packet=1 would trigger the
queue flush to accommodate possible outgoing bundling.

In the input queue, sctp_inq_pop() seems to do the right thing
in terms of discarding invalid chunks. So, above JUNK will
not enter the state machine and instead be released and exit
the sctp_assoc_bh_rcv() chunk processing loop. It's simply
the flush point being missing at loop exit. Adding a try-flush
approach on the output queue might not work as the underlying
infrastructure might be long gone at this point due to the
side-effect interpreter run.

One possibility, albeit a bit of a kludge, would be to defer
invalid chunk freeing into the state machine in order to
possibly trigger packet discards and thus indirectly a queue
flush on error. It would surely be better to discard chunks
as in the current, perhaps better controlled environment, but
going back and forth, it's simply architecturally not possible.
I tried various trailing JUNK attack cases and it seems to
look good now.

Joint work with Vlad Yasevich.

Fixes: 2e3216cd54b1 ("sctp: Follow security requirement of responding with 1 packet")
Signed-off-by: Daniel Borkmann <dborkman@redhat.com>
Signed-off-by: Vlad Yasevich <vyasevich@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Cc: Josh Boyer <jwboyer@fedoraproject.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agoKVM: x86: Don't report guest userspace emulation error to userspace
Nadav Amit [Tue, 16 Sep 2014 23:50:50 +0000 (02:50 +0300)]
KVM: x86: Don't report guest userspace emulation error to userspace

commit a2b9e6c1a35afcc0973acb72e591c714e78885ff upstream.

Commit fc3a9157d314 ("KVM: X86: Don't report L2 emulation failures to
user-space") disabled the reporting of L2 (nested guest) emulation failures to
userspace due to race-condition between a vmexit and the instruction emulator.
The same rational applies also to userspace applications that are permitted by
the guest OS to access MMIO area or perform PIO.

This patch extends the current behavior - of injecting a #UD instead of
reporting it to userspace - also for guest userspace code.

Signed-off-by: Nadav Amit <namit@cs.technion.ac.il>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agoperf/x86/intel: Use proper dTLB-load-misses event on IvyBridge
Vince Weaver [Mon, 14 Jul 2014 19:33:25 +0000 (15:33 -0400)]
perf/x86/intel: Use proper dTLB-load-misses event on IvyBridge

commit 1996388e9f4e3444db8273bc08d25164d2967c21 upstream.

This was discussed back in February:

https://lkml.org/lkml/2014/2/18/956

But I never saw a patch come out of it.

On IvyBridge we share the SandyBridge cache event tables, but the
dTLB-load-miss event is not compatible.  Patch it up after
the fact to the proper DTLB_LOAD_MISSES.DEMAND_LD_MISS_CAUSES_A_WALK

Signed-off-by: Vince Weaver <vincent.weaver@maine.edu>
Signed-off-by: Peter Zijlstra <peterz@infradead.org>
Cc: Arnaldo Carvalho de Melo <acme@kernel.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Link: http://lkml.kernel.org/r/alpine.DEB.2.11.1407141528200.17214@vincent-weaver-1.umelst.maine.edu
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Cc: Hou Pengyang <houpengyang@huawei.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agoperf: Handle compat ioctl
Pawel Moll [Fri, 13 Jun 2014 15:03:32 +0000 (16:03 +0100)]
perf: Handle compat ioctl

commit b3f207855f57b9c8f43a547a801340bb5cbc59e5 upstream.

When running a 32-bit userspace on a 64-bit kernel (eg. i386
application on x86_64 kernel or 32-bit arm userspace on arm64
kernel) some of the perf ioctls must be treated with special
care, as they have a pointer size encoded in the command.

For example, PERF_EVENT_IOC_ID in 32-bit world will be encoded
as 0x80042407, but 64-bit kernel will expect 0x80082407. In
result the ioctl will fail returning -ENOTTY.

This patch solves the problem by adding code fixing up the
size as compat_ioctl file operation.

Reported-by: Drew Richardson <drew.richardson@arm.com>
Signed-off-by: Pawel Moll <pawel.moll@arm.com>
Signed-off-by: Peter Zijlstra <peterz@infradead.org>
Cc: Arnaldo Carvalho de Melo <acme@kernel.org>
Cc: Jiri Olsa <jolsa@redhat.com>
Link: http://lkml.kernel.org/r/1402671812-9078-1-git-send-email-pawel.moll@arm.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Signed-off-by: David Ahern <daahern@cisco.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agodell-wmi: Fix access out of memory
Pali Rohár [Mon, 29 Sep 2014 13:10:51 +0000 (15:10 +0200)]
dell-wmi: Fix access out of memory

commit a666b6ffbc9b6705a3ced704f52c3fe9ea8bf959 upstream.

Without this patch, dell-wmi is trying to access elements of dynamically
allocated array without checking the array size. This can lead to memory
corruption or a kernel panic. This patch adds the missing checks for
array size.

Signed-off-by: Pali Rohár <pali.rohar@gmail.com>
Signed-off-by: Darren Hart <dvhart@linux.intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agorcu: Use rcu_gp_kthread_wake() to wake up grace period kthreads
Pranith Kumar [Tue, 12 Aug 2014 17:07:47 +0000 (13:07 -0400)]
rcu: Use rcu_gp_kthread_wake() to wake up grace period kthreads

commit 2aa792e6faf1a00f5accf1f69e87e11a390ba2cd upstream.

The rcu_gp_kthread_wake() function checks for three conditions before
waking up grace period kthreads:

*  Is the thread we are trying to wake up the current thread?
*  Are the gp_flags zero? (all threads wait on non-zero gp_flags condition)
*  Is there no thread created for this flavour, hence nothing to wake up?

If any one of these condition is true, we do not call wake_up().
It was found that there are quite a few avoidable wake ups both during
idle time and under stress induced by rcutorture.

Idle:

Total:66000, unnecessary:66000, case1:61827, case2:66000, case3:0
Total:68000, unnecessary:68000, case1:63696, case2:68000, case3:0

rcutorture:

Total:254000, unnecessary:254000, case1:199913, case2:254000, case3:0
Total:256000, unnecessary:256000, case1:201784, case2:256000, case3:0

Here case{1-3} are the cases listed above. We can avoid these wake
ups by using rcu_gp_kthread_wake() to conditionally wake up the grace
period kthreads.

There is a comment about an implied barrier supplied by the wake_up()
logic.  This barrier is necessary for the awakened thread to see the
updated ->gp_flags.  This flag is always being updated with the root node
lock held. Also, the awakened thread tries to acquire the root node lock
before reading ->gp_flags because of which there is proper ordering.

Hence this commit tries to avoid calling wake_up() whenever we can by
using rcu_gp_kthread_wake() function.

Signed-off-by: Pranith Kumar <bobby.prani@gmail.com>
CC: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Cc: Kamal Mostafa <kamal@canonical.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agorcu: Make callers awaken grace-period kthread
Paul E. McKenney [Tue, 11 Mar 2014 20:02:16 +0000 (13:02 -0700)]
rcu: Make callers awaken grace-period kthread

commit 48a7639ce80cf279834d0d44865e49ecd714f37d upstream.

The rcu_start_gp_advanced() function currently uses irq_work_queue()
to defer wakeups of the RCU grace-period kthread.  This deferring
is necessary to avoid RCU-scheduler deadlocks involving the rcu_node
structure's lock, meaning that RCU cannot call any of the scheduler's
wake-up functions while holding one of these locks.

Unfortunately, the second and subsequent calls to irq_work_queue() are
ignored, and the first call will be ignored (aside from queuing the work
item) if the scheduler-clock tick is turned off.  This is OK for many
uses, especially those where irq_work_queue() is called from an interrupt
or softirq handler, because in those cases the scheduler-clock-tick state
will be re-evaluated, which will turn the scheduler-clock tick back on.
On the next tick, any deferred work will then be processed.

However, this strategy does not always work for RCU, which can be invoked
at process level from idle CPUs.  In this case, the tick might never
be turned back on, indefinitely defering a grace-period start request.
Note that the RCU CPU stall detector cannot see this condition, because
there is no RCU grace period in progress.  Therefore, we can (and do!)
see long tens-of-seconds stalls in grace-period handling.  In theory,
we could see a full grace-period hang, but rcutorture testing to date
has seen only the tens-of-seconds stalls.  Event tracing demonstrates
that irq_work_queue() is being called repeatedly to no effect during
these stalls: The "newreq" event appears repeatedly from a task that is
not one of the grace-period kthreads.

In theory, irq_work_queue() might be fixed to avoid this sort of issue,
but RCU's requirements are unusual and it is quite straightforward to pass
wake-up responsibility up through RCU's call chain, so that the wakeup
happens when the offending locks are released.

This commit therefore makes this change.  The rcu_start_gp_advanced(),
rcu_start_future_gp(), rcu_accelerate_cbs(), rcu_advance_cbs(),
__note_gp_changes(), and rcu_start_gp() functions now return a boolean
which indicates when a wake-up is needed.  A new rcu_gp_kthread_wake()
does the wakeup when it is necessary and safe to do so: No self-wakes,
no wake-ups if the ->gp_flags field indicates there is no need (as in
someone else did the wake-up before we got around to it), and no wake-ups
before the grace-period kthread has been created.

Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Steven Rostedt <rostedt@goodmis.org>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Reviewed-by: Josh Triplett <josh@joshtriplett.org>
[ Pranith: backport to 3.13-stable: just rcu_gp_kthread_wake(),
  prereq for 2aa792e "rcu: Use rcu_gp_kthread_wake() to wake up grace
  period kthreads" ]
Signed-off-by: Pranith Kumar <bobby.prani@gmail.com>
Signed-off-by: Kamal Mostafa <kamal@canonical.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agoGFS2: Fix address space from page function
Steven Whitehouse [Mon, 31 Mar 2014 16:48:27 +0000 (17:48 +0100)]
GFS2: Fix address space from page function

commit 1b2ad41214c9bf6e8befa000f0522629194bf540 upstream.

Now that rgrps use the address space which is part of the super
block, we need to update gfs2_mapping2sbd() to take account of
that. The only way to do that easily is to use a different set
of address_space_operations for rgrps.

Reported-by: Abhi Das <adas@redhat.com>
Tested-by: Abhi Das <adas@redhat.com>
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agoARM: probes: fix instruction fetch order with <asm/opcodes.h>
Ben Dooks [Fri, 8 Nov 2013 18:29:25 +0000 (18:29 +0000)]
ARM: probes: fix instruction fetch order with <asm/opcodes.h>

commit 888be25402021a425da3e85e2d5a954d7509286e upstream.

If we are running BE8, the data and instruction endianness do not
match, so use <asm/opcodes.h> to correctly translate memory accesses
into ARM instructions.

Acked-by: Jon Medhurst <tixy@linaro.org>
Signed-off-by: Ben Dooks <ben.dooks@codethink.co.uk>
[taras.kondratiuk@linaro.org: fixed Thumb instruction fetch order]
Signed-off-by: Taras Kondratiuk <taras.kondratiuk@linaro.org>
[wangnan: backport to 3.10 and 3.14:
 - adjust context
 - backport all changes on arch/arm/kernel/probes.c to
   arch/arm/kernel/kprobes-common.c since we don't have
   commit c18377c303787ded44b7decd7dee694db0f205e9.
 - After the above adjustments, becomes same to Taras Kondratiuk's
   original patch:
     http://lists.linaro.org/pipermail/linaro-kernel/2014-January/010346.html
]
Signed-off-by: Wang Nan <wangnan0@huawei.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agonetfilter: xt_bpf: add mising opaque struct sk_filter definition
Pablo Neira [Tue, 29 Jul 2014 16:12:15 +0000 (18:12 +0200)]
netfilter: xt_bpf: add mising opaque struct sk_filter definition

commit e10038a8ec06ac819b7552bb67aaa6d2d6f850c1 upstream.

This structure is not exposed to userspace, so fix this by defining
struct sk_filter; so we skip the casting in kernelspace. This is safe
since userspace has no way to lurk with that internal pointer.

Fixes: e6f30c7 ("netfilter: x_tables: add xt_bpf match")
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
Acked-by: Willem de Bruijn <willemb@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>