platform/kernel/linux-rpi.git
6 years agonvmet: allow Keep Alive for Discovery controller
Jay Sternberg [Mon, 12 Nov 2018 21:56:35 +0000 (13:56 -0800)]
nvmet: allow Keep Alive for Discovery controller

Per change to specification allowing Discovery controllers to have
explicit persistent connections, remove restriction on Discovery
controllers allowing kato on connect.

Signed-off-by: Jay Sternberg <jay.e.sternberg@intel.com>
Reviewed-by: Hannes Reinecke <hare@suse.com>
Reviewed-by: Sagi Grimberg <sagi@grimberg.me>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
6 years agonvmet: change aen mask functions to use bit numbers
Jay Sternberg [Mon, 12 Nov 2018 21:56:34 +0000 (13:56 -0800)]
nvmet: change aen mask functions to use bit numbers

Functions nvmet_aen_disabled and nvmet_clear_aen were using
values not bit numbers ie 1 << 9 not 9 for bit function clear_bit
and test_and_set_bit.

Signed-off-by: Jay Sternberg <jay.e.sternberg@intel.com>
Reviewed-by: Phil Cayton <phil.cayton@intel.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
6 years agonvmet: provide aen bit functions for multiple controller types
Jay Sternberg [Mon, 12 Nov 2018 21:56:33 +0000 (13:56 -0800)]
nvmet: provide aen bit functions for multiple controller types

Move nvmet_aen_disabled and nvmet_clear_aen in preparation for other types
of controllers to use, initially the discovery controller.

Signed-off-by: Jay Sternberg <jay.e.sternberg@intel.com>
Reviewed-by: Johannes Thumshirn <jthumshirn@suse.de>
Reviewed-by: Sagi Grimberg <sagi@grimberg.me>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
6 years agonvmet: use IOCB_NOWAIT for file-ns buffered I/O
Chaitanya Kulkarni [Wed, 14 Nov 2018 06:12:19 +0000 (01:12 -0500)]
nvmet: use IOCB_NOWAIT for file-ns buffered I/O

This patch optimizes read command behavior when file-ns configured
with buffered I/O. Instead of offloading the buffered I/O read operations
to the worker threads, we first issue the read operation with IOCB_NOWAIT
and try and access the data from the cache. Here we only offload the
request to the worker thread and complete the request in the worker
thread context when IOCB_NOWAIT request fails.

Signed-off-by: Chaitanya Kulkarni <chaitanya.kulkarni@wdc.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
6 years agonvmet: support for traffic based keep-alive
Sagi Grimberg [Fri, 2 Nov 2018 17:28:13 +0000 (10:28 -0700)]
nvmet: support for traffic based keep-alive

A controller that supports traffic based keep-alive can restart the keep
alive timer even when no keep-alive was not received in the kato period
as long as other admin or I/O commands were received.  For each command
set ctrl->cmd_seen to true, and when keep-alive timer expires, if any
commands were seen, resched ka_work instead of escalating to a fatal
error.

Signed-off-by: Sagi Grimberg <sagi@grimberg.me>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
6 years agonvme: support traffic based keep-alive
Sagi Grimberg [Fri, 2 Nov 2018 17:28:15 +0000 (10:28 -0700)]
nvme: support traffic based keep-alive

If the controller supports traffic based keep alive, we restart the keep
alive timer if any admin or io commands was completed during the kato
period.  This prevents a possible starvation of keep alive commands in
the presence of heavy traffic as in such case, we already have a health
indication from the host perspective.

Only set a comp_seen indicator in case the controller supports keep
alive to minimize the overhead for pci controllers.

Signed-off-by: Sagi Grimberg <sagi@grimberg.me>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
6 years agonvme: cache controller attributes
Sagi Grimberg [Fri, 2 Nov 2018 17:28:14 +0000 (10:28 -0700)]
nvme: cache controller attributes

We get the controller attributes in identify, cache them as we'll need
them for traffic based keep alive support.

Reviewed-by: Chaitanya Kulkarni <chaitanya.kulkarni@wdc.com>
Signed-off-by: Sagi Grimberg <sagi@grimberg.me>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
6 years agonvme: introduce ctrl attributes enumeration
Sagi Grimberg [Fri, 2 Nov 2018 17:28:12 +0000 (10:28 -0700)]
nvme: introduce ctrl attributes enumeration

We are growing more controller attributes, so use a proper enumeration
for it.  For now just add the 128-bit hostid which we support.

Reviewed-by: Chaitanya Kulkarni <chaitanya.kulkarni@wdc.com>
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Sagi Grimberg <sagi@grimberg.me>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
6 years agonvme: add a numa_node field to struct nvme_ctrl
Hannes Reinecke [Fri, 16 Nov 2018 08:22:29 +0000 (09:22 +0100)]
nvme: add a numa_node field to struct nvme_ctrl

Instead of directly poking into the struct device add a new numa_node
field to struct nvme_ctrl.  This allows fabrics drivers where ctrl->dev
is a virtual device to support NUMA affinity as well.

Also expose the field as a sysfs attribute, and populate it for the
RDMA and FC transports.

Signed-off-by: Hannes Reinecke <hare@suse.com>
Reviewed-by: Sagi Grimberg <sagi@grimberg.me>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
6 years agonvme: consolidate memset calls in the nvme_setup_cmd path
Chaitanya Kulkarni [Mon, 29 Oct 2018 23:44:18 +0000 (16:44 -0700)]
nvme: consolidate memset calls in the nvme_setup_cmd path

In function nvme_setup_cmd() we call command specific setup function
for flush, rw, and discard. Instead of calling memset in each function
lets call it once in the parent function.

This is purely code cleanup patch and it does not change any existing
functionality.

Signed-off-by: Chaitanya Kulkarni <chaitanya.kulkarni@wdc.com>
Reviewed-by: Sagi Grimberg <sagi@grimberg.me>
Reviewed-by: Johannes Thumshirn <jthumshirn@suse.de>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
6 years agoblk-mq: re-build queue map in case of kdump kernel
Ming Lei [Fri, 7 Dec 2018 03:03:53 +0000 (11:03 +0800)]
blk-mq: re-build queue map in case of kdump kernel

Now almost all .map_queues() implementation based on managed irq
affinity doesn't update queue mapping and it just retrieves the
old built mapping, so if nr_hw_queues is changed, the mapping talbe
includes stale mapping. And only blk_mq_map_queues() may rebuild
the mapping talbe.

One case is that we limit .nr_hw_queues as 1 in case of kdump kernel.
However, drivers often builds queue mapping before allocating tagset
via pci_alloc_irq_vectors_affinity(), but set->nr_hw_queues can be set
as 1 in case of kdump kernel, so wrong queue mapping is used, and
kernel panic[1] is observed during booting.

This patch fixes the kernel panic triggerd on nvme by rebulding the
mapping table via blk_mq_map_queues().

[1] kernel panic log
[    4.438371] nvme nvme0: 16/0/0 default/read/poll queues
[    4.443277] BUG: unable to handle kernel NULL pointer dereference at 0000000000000098
[    4.444681] PGD 0 P4D 0
[    4.445367] Oops: 0000 [#1] SMP NOPTI
[    4.446342] CPU: 3 PID: 201 Comm: kworker/u33:10 Not tainted 4.20.0-rc5-00664-g5eb02f7ee1eb-dirty #459
[    4.447630] Hardware name: QEMU Standard PC (Q35 + ICH9, 2009), BIOS 1.10.2-2.fc27 04/01/2014
[    4.448689] Workqueue: nvme-wq nvme_scan_work [nvme_core]
[    4.449368] RIP: 0010:blk_mq_map_swqueue+0xfb/0x222
[    4.450596] Code: 04 f5 20 28 ef 81 48 89 c6 39 55 30 76 93 89 d0 48 c1 e0 04 48 03 83 f8 05 00 00 48 8b 00 42 8b 3c 28 48 8b 43 58 48 8b 04 f8 <48> 8b b8 98 00 00 00 4c 0f a3 37 72 42 f0 4c 0f ab 37 66 8b b8 f6
[    4.453132] RSP: 0018:ffffc900023b3cd8 EFLAGS: 00010286
[    4.454061] RAX: 0000000000000000 RBX: ffff888174448000 RCX: 0000000000000001
[    4.456480] RDX: 0000000000000001 RSI: ffffe8feffc506c0 RDI: 0000000000000001
[    4.458750] RBP: ffff88810722d008 R08: ffff88817647a880 R09: 0000000000000002
[    4.464580] R10: ffffc900023b3c10 R11: 0000000000000004 R12: ffff888174448538
[    4.467803] R13: 0000000000000004 R14: 0000000000000001 R15: 0000000000000001
[    4.469220] FS:  0000000000000000(0000) GS:ffff88817bac0000(0000) knlGS:0000000000000000
[    4.471554] CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[    4.472464] CR2: 0000000000000098 CR3: 0000000174e4e001 CR4: 0000000000760ee0
[    4.474264] DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
[    4.476007] DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400
[    4.477061] PKRU: 55555554
[    4.477464] Call Trace:
[    4.478731]  blk_mq_init_allocated_queue+0x36a/0x3ad
[    4.479595]  blk_mq_init_queue+0x32/0x4e
[    4.480178]  nvme_validate_ns+0x98/0x623 [nvme_core]
[    4.480963]  ? nvme_submit_sync_cmd+0x1b/0x20 [nvme_core]
[    4.481685]  ? nvme_identify_ctrl.isra.8+0x70/0xa0 [nvme_core]
[    4.482601]  nvme_scan_work+0x23a/0x29b [nvme_core]
[    4.483269]  ? _raw_spin_unlock_irqrestore+0x25/0x38
[    4.483930]  ? try_to_wake_up+0x38d/0x3b3
[    4.484478]  ? process_one_work+0x179/0x2fc
[    4.485118]  process_one_work+0x1d3/0x2fc
[    4.485655]  ? rescuer_thread+0x2ae/0x2ae
[    4.486196]  worker_thread+0x1e9/0x2be
[    4.486841]  kthread+0x115/0x11d
[    4.487294]  ? kthread_park+0x76/0x76
[    4.487784]  ret_from_fork+0x3a/0x50
[    4.488322] Modules linked in: nvme nvme_core qemu_fw_cfg virtio_scsi ip_tables
[    4.489428] Dumping ftrace buffer:
[    4.489939]    (ftrace buffer empty)
[    4.490492] CR2: 0000000000000098
[    4.491052] ---[ end trace 03cd268ad5a86ff7 ]---

Cc: Christoph Hellwig <hch@lst.de>
Cc: linux-nvme@lists.infradead.org
Cc: David Milburn <dmilburn@redhat.com>
Signed-off-by: Ming Lei <ming.lei@redhat.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
6 years agoblkcg: put back rcu lock in blkcg_bio_issue_check()
Dennis Zhou [Thu, 6 Dec 2018 17:49:38 +0000 (12:49 -0500)]
blkcg: put back rcu lock in blkcg_bio_issue_check()

I was a little overzealous in removing the rcu_read_lock() call from
blkcg_bio_issue_check() and it broke blk-throttle. Put it back.

Fixes: e35403a034bf ("blkcg: associate blkg when associating a device")
Signed-off-by: Dennis Zhou <dennis@kernel.org>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
6 years agoblock: convert io-latency to use rq_qos_wait
Josef Bacik [Tue, 4 Dec 2018 17:59:04 +0000 (12:59 -0500)]
block: convert io-latency to use rq_qos_wait

Now that we have this common helper, convert io-latency over to use it
as well.

Signed-off-by: Josef Bacik <josef@toxicpanda.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
6 years agoblock: convert wbt_wait() to use rq_qos_wait()
Josef Bacik [Tue, 4 Dec 2018 17:59:03 +0000 (12:59 -0500)]
block: convert wbt_wait() to use rq_qos_wait()

Now that we have rq_qos_wait() in place, convert wbt_wait() over to
using it with it's specific callbacks.

Signed-off-by: Josef Bacik <josef@toxicpanda.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
6 years agoblock: add rq_qos_wait to rq_qos
Josef Bacik [Tue, 4 Dec 2018 17:59:02 +0000 (12:59 -0500)]
block: add rq_qos_wait to rq_qos

Originally when I split out the common code from blk-wbt into rq_qos I
left the wbt_wait() where it was and simply copied and modified it
slightly to work for io-latency.  However they are both basically the
same thing, and as time has gone on wbt_wait() has ended up much smarter
and kinder than it was when I copied it into io-latency, which means
io-latency has lost out on these improvements.

Since they are the same thing essentially except for a few minor things,
create rq_qos_wait() that replicates what wbt_wait() currently does with
callbacks that can be passed in for the snowflakes to do their own thing
as appropriate.

Signed-off-by: Josef Bacik <josef@toxicpanda.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
6 years agoblkcg: rename blkg_try_get() to blkg_tryget()
Dennis Zhou [Wed, 5 Dec 2018 17:10:39 +0000 (12:10 -0500)]
blkcg: rename blkg_try_get() to blkg_tryget()

blkg reference counting now uses percpu_ref rather than atomic_t. Let's
make this consistent with css_tryget. This renames blkg_try_get to
blkg_tryget and now returns a bool rather than the blkg or %NULL.

Signed-off-by: Dennis Zhou <dennis@kernel.org>
Reviewed-by: Josef Bacik <josef@toxicpanda.com>
Acked-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
6 years agoblkcg: change blkg reference counting to use percpu_ref
Dennis Zhou [Wed, 5 Dec 2018 17:10:38 +0000 (12:10 -0500)]
blkcg: change blkg reference counting to use percpu_ref

Every bio is now associated with a blkg putting blkg_get, blkg_try_get,
and blkg_put on the hot path. Switch over the refcnt in blkg to use
percpu_ref.

Signed-off-by: Dennis Zhou <dennis@kernel.org>
Acked-by: Tejun Heo <tj@kernel.org>
Reviewed-by: Josef Bacik <josef@toxicpanda.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
6 years agoblkcg: remove bio_disassociate_task()
Dennis Zhou [Wed, 5 Dec 2018 17:10:37 +0000 (12:10 -0500)]
blkcg: remove bio_disassociate_task()

Now that a bio only holds a blkg reference, so clean up is simply
putting back that reference. Remove bio_disassociate_task() as it just
calls bio_disassociate_blkg() and call the latter directly.

Signed-off-by: Dennis Zhou <dennis@kernel.org>
Acked-by: Tejun Heo <tj@kernel.org>
Reviewed-by: Josef Bacik <josef@toxicpanda.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
6 years agoblkcg: remove additional reference to the css
Dennis Zhou [Wed, 5 Dec 2018 17:10:36 +0000 (12:10 -0500)]
blkcg: remove additional reference to the css

The previous patch in this series removed carrying around a pointer to
the css in blkg. However, the blkg association logic still relied on
taking a reference on the css to ensure we wouldn't fail in getting a
reference for the blkg.

Here the implicit dependency on the css is removed. The association
continues to rely on the tryget logic walking up the blkg tree. This
streamlines the three ways that association can happen: normal, swap,
and writeback.

Signed-off-by: Dennis Zhou <dennis@kernel.org>
Acked-by: Tejun Heo <tj@kernel.org>
Reviewed-by: Josef Bacik <josef@toxicpanda.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
6 years agoblkcg: remove bio->bi_css and instead use bio->bi_blkg
Dennis Zhou [Wed, 5 Dec 2018 17:10:35 +0000 (12:10 -0500)]
blkcg: remove bio->bi_css and instead use bio->bi_blkg

Prior patches ensured that any bio that interacts with a request_queue
is properly associated with a blkg. This makes bio->bi_css unnecessary
as blkg maintains a reference to blkcg already.

This removes the bio field bi_css and transfers corresponding uses to
access via bi_blkg.

Signed-off-by: Dennis Zhou <dennis@kernel.org>
Reviewed-by: Josef Bacik <josef@toxicpanda.com>
Acked-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
6 years agoblkcg: associate writeback bios with a blkg
Dennis Zhou [Wed, 5 Dec 2018 17:10:34 +0000 (12:10 -0500)]
blkcg: associate writeback bios with a blkg

One of the goals of this series is to remove a separate reference to
the css of the bio. This can and should be accessed via bio_blkcg(). In
this patch, wbc_init_bio() now requires a bio to have a device
associated with it.

Signed-off-by: Dennis Zhou <dennis@kernel.org>
Reviewed-by: Josef Bacik <josef@toxicpanda.com>
Acked-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
6 years agoblkcg: associate a blkg for pages being evicted by swap
Dennis Zhou [Wed, 5 Dec 2018 17:10:33 +0000 (12:10 -0500)]
blkcg: associate a blkg for pages being evicted by swap

A prior patch in this series added blkg association to bios issued by
cgroups. There are two other paths that we want to attribute work back
to the appropriate cgroup: swap and writeback. Here we modify the way
swap tags bios to include the blkg. Writeback will be tackle in the next
patch.

Signed-off-by: Dennis Zhou <dennis@kernel.org>
Reviewed-by: Josef Bacik <josef@toxicpanda.com>
Acked-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
6 years agoblkcg: consolidate bio_issue_init() to be a part of core
Dennis Zhou [Wed, 5 Dec 2018 17:10:32 +0000 (12:10 -0500)]
blkcg: consolidate bio_issue_init() to be a part of core

bio_issue_init among other things initializes the timestamp for an IO.
Rather than have this logic handled by policies, this consolidates it to
be on the init paths (normal, clone, bounce clone).

Signed-off-by: Dennis Zhou <dennis@kernel.org>
Acked-by: Tejun Heo <tj@kernel.org>
Reviewed-by: Liu Bo <bo.liu@linux.alibaba.com>
Reviewed-by: Josef Bacik <josef@toxicpanda.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
6 years agoblkcg: associate blkg when associating a device
Dennis Zhou [Wed, 5 Dec 2018 17:10:31 +0000 (12:10 -0500)]
blkcg: associate blkg when associating a device

Previously, blkg association was handled by controller specific code in
blk-throttle and blk-iolatency. However, because a blkg represents a
relationship between a blkcg and a request_queue, it makes sense to keep
the blkg->q and bio->bi_disk->queue consistent.

This patch moves association into the bio_set_dev macro(). This should
cover the majority of cases where the device is set/changed keeping the
two pointers consistent. Fallback code is added to
blkcg_bio_issue_check() to catch any missing paths.

Signed-off-by: Dennis Zhou <dennis@kernel.org>
Reviewed-by: Josef Bacik <josef@toxicpanda.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
6 years agodm: set the static flush bio device on demand
Dennis Zhou [Wed, 5 Dec 2018 17:10:30 +0000 (12:10 -0500)]
dm: set the static flush bio device on demand

The next patch changes the macro bio_set_dev() to associate a bio with a
blkg based on the device set. However, dm creates a static bio to be
used as the basis for cloning empty flush bios on creation. The
bio_set_dev() call in alloc_dev() will cause problems with the next
patch adding association to bio_set_dev() because the call is before the
bdev is associated with a gendisk (bd_disk is %NULL). To get around
this, set the device on the static bio every time and use that to clone
to the other bios.

Signed-off-by: Dennis Zhou <dennis@kernel.org>
Acked-by: Mike Snitzer <snitzer@redhat.com>
Cc: Alasdair Kergon <agk@redhat.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
6 years agoblkcg: introduce common blkg association logic
Dennis Zhou [Wed, 5 Dec 2018 17:10:29 +0000 (12:10 -0500)]
blkcg: introduce common blkg association logic

There are 3 ways blkg association can happen: association with the
current css, with the page css (swap), or from the wbc css (writeback).

This patch handles how association is done for the first case where we
are associating bsaed on the current css. If there is already a blkg
associated, the css will be reused and association will be redone as the
request_queue may have changed.

Signed-off-by: Dennis Zhou <dennis@kernel.org>
Reviewed-by: Josef Bacik <josef@toxicpanda.com>
Acked-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
6 years agoblkcg: convert blkg_lookup_create() to find closest blkg
Dennis Zhou [Wed, 5 Dec 2018 17:10:28 +0000 (12:10 -0500)]
blkcg: convert blkg_lookup_create() to find closest blkg

There are several scenarios where blkg_lookup_create() can fail such as
the blkcg dying, request_queue is dying, or simply being OOM. Most
handle this by simply falling back to the q->root_blkg and calling it a
day.

This patch implements the notion of closest blkg. During
blkg_lookup_create(), if it fails to create, return the closest blkg
found or the q->root_blkg. blkg_try_get_closest() is introduced and used
during association so a bio is always attached to a blkg.

Signed-off-by: Dennis Zhou <dennis@kernel.org>
Acked-by: Tejun Heo <tj@kernel.org>
Reviewed-by: Josef Bacik <josef@toxicpanda.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
6 years agoblkcg: update blkg_lookup_create() to do locking
Dennis Zhou [Wed, 5 Dec 2018 17:10:27 +0000 (12:10 -0500)]
blkcg: update blkg_lookup_create() to do locking

To know when to create a blkg, the general pattern is to do a
blkg_lookup() and if that fails, lock and do the lookup again, and if
that fails finally create. It doesn't make much sense for everyone who
wants to do creation to write this themselves.

This changes blkg_lookup_create() to do locking and implement this
pattern. The old blkg_lookup_create() is renamed to
__blkg_lookup_create().  If a call site wants to do its own error
handling or already owns the queue lock, they can use
__blkg_lookup_create(). This will be used in upcoming patches.

Signed-off-by: Dennis Zhou <dennis@kernel.org>
Reviewed-by: Josef Bacik <josef@toxicpanda.com>
Acked-by: Tejun Heo <tj@kernel.org>
Reviewed-by: Liu Bo <bo.liu@linux.alibaba.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
6 years agoblkcg: fix ref count issue with bio_blkcg() using task_css
Dennis Zhou [Wed, 5 Dec 2018 17:10:26 +0000 (12:10 -0500)]
blkcg: fix ref count issue with bio_blkcg() using task_css

The bio_blkcg() function turns out to be inconsistent and consequently
dangerous to use. The first part returns a blkcg where a reference is
owned by the bio meaning it does not need to be rcu protected. However,
the third case, the last line, is problematic:

return css_to_blkcg(task_css(current, io_cgrp_id));

This can race against task migration and the cgroup dying. It is also
semantically different as it must be called rcu protected and is
susceptible to failure when trying to get a reference to it.

This patch adds association ahead of calling bio_blkcg() rather than
after. This makes association a required and explicit step along the
code paths for calling bio_blkcg(). In blk-iolatency, association is
moved above the bio_blkcg() call to ensure it will not return %NULL.

BFQ uses the old bio_blkcg() function, but I do not want to address it
in this series due to the complexity. I have created a private version
documenting the inconsistency and noting not to use it.

Signed-off-by: Dennis Zhou <dennis@kernel.org>
Acked-by: Tejun Heo <tj@kernel.org>
Reviewed-by: Josef Bacik <josef@toxicpanda.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
6 years agoblk-mq: remove QUEUE_FLAG_POLL from default MQ flags
Jens Axboe [Wed, 5 Dec 2018 13:50:40 +0000 (06:50 -0700)]
blk-mq: remove QUEUE_FLAG_POLL from default MQ flags

We only support polling if we have poll queues now, but the flag is
being set by default. Remove the default QUEUE_FLAG_POLL setting, we'll
set it in blk_mq_init_allocated_queue() if we have poll queues available
for this device.

Fixes: 6544d229bf43 ("block: enable polling by default if a poll map is initalized")
Reported-by: Kirill Tkhai <ktkhai@virtuozzo.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
6 years agoblock: enable polling by default if a poll map is initalized
Christoph Hellwig [Sun, 2 Dec 2018 16:46:28 +0000 (17:46 +0100)]
block: enable polling by default if a poll map is initalized

If the user did setup polling in the driver we should not require
another know in the block layer to enable it.

Reviewed-by: Sagi Grimberg <sagi@grimberg.me>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
6 years agoblock: only allow polling if a poll queue_map exists
Christoph Hellwig [Sun, 2 Dec 2018 16:46:27 +0000 (17:46 +0100)]
block: only allow polling if a poll queue_map exists

This avoids having to have differnet mq_ops for different setups
with or without poll queues.

Reviewed-by: Sagi Grimberg <sagi@grimberg.me>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
6 years agoblock: remove ->poll_fn
Christoph Hellwig [Sun, 2 Dec 2018 16:46:26 +0000 (17:46 +0100)]
block: remove ->poll_fn

This was intended to support users like nvme multipath, but is just
getting in the way and adding another indirect call.

Reviewed-by: Sagi Grimberg <sagi@grimberg.me>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
6 years agonvme-mpath: remove I/O polling support
Christoph Hellwig [Sun, 2 Dec 2018 16:46:25 +0000 (17:46 +0100)]
nvme-mpath: remove I/O polling support

The ->poll_fn has been stale for a while, as a lot of places check for mq
ops.  But there is no real point in it anyway, as we don't even use
the multipath code for subsystems without multiple ports, which is usually
what we do high performance I/O to.  If it really becomes an issue we
should rework the nvme code to also skip the multipath code for any
private namespace, even if that could mean some trouble when rescanning.

Reviewed-by: Keith Busch <keith.busch@intel.com>
Reviewed-by: Sagi Grimberg <sagi@grimberg.me>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
6 years agonvme-rdma: remove I/O polling support
Christoph Hellwig [Sun, 2 Dec 2018 16:46:24 +0000 (17:46 +0100)]
nvme-rdma: remove I/O polling support

The code was always a bit of a hack that digs far too much into
RDMA core internals.  Lets kick it out and reimplement proper
dedicated poll queues as needed.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
6 years agonvme-pci: remove the CQ lock for interrupt driven queues
Christoph Hellwig [Sun, 2 Dec 2018 16:46:23 +0000 (17:46 +0100)]
nvme-pci: remove the CQ lock for interrupt driven queues

Now that we can't poll regular, interrupt driven I/O queues there
is almost nothing that can race with an interrupt.  The only
possible other contexts polling a CQ are the error handler and
queue shutdown, and both are so far off in the slow path that
we can simply use the big hammer of disabling interrupts.

With that we can stop taking the cq_lock for normal queues.

Reviewed-by: Sagi Grimberg <sagi@grimberg.me>
Reviewed-by: Keith Busch <keith.busch@intel.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
6 years agonvme-pci: don't poll from irq context when deleting queues
Christoph Hellwig [Sun, 2 Dec 2018 16:46:22 +0000 (17:46 +0100)]
nvme-pci: don't poll from irq context when deleting queues

This is the last place outside of nvme_irq that handles CQEs from
interrupt context, and thus is in the way of removing the cq_lock for
normal queues, and avoiding lockdep warnings on the poll queues, for
which we already take it without IRQ disabling.

Reviewed-by: Keith Busch <keith.busch@intel.com>
Reviewed-by: Sagi Grimberg <sagi@grimberg.me>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
6 years agonvme-pci: refactor nvme_disable_io_queues
Christoph Hellwig [Sun, 2 Dec 2018 16:46:21 +0000 (17:46 +0100)]
nvme-pci: refactor nvme_disable_io_queues

Pass the opcode for the delete SQ/CQ command as an argument instead of
the somewhat confusing pass loop.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Keith Busch <keith.busch@intel.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
6 years agonvme-pci: consolidate code for polling non-dedicated queues
Christoph Hellwig [Sun, 2 Dec 2018 16:46:20 +0000 (17:46 +0100)]
nvme-pci: consolidate code for polling non-dedicated queues

We have three places that can poll for I/O completions on a normal
interrupt-enabled queue.  All of them are in slow path code, so
consolidate them to a single helper that uses spin_lock_irqsave and
removes the fast path cqe_pending check.

Reviewed-by: Sagi Grimberg <sagi@grimberg.me>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
6 years agonvme-pci: only allow polling with separate poll queues
Christoph Hellwig [Sun, 2 Dec 2018 16:46:19 +0000 (17:46 +0100)]
nvme-pci: only allow polling with separate poll queues

This will allow us to simplify both the regular NVMe interrupt handler
and the upcoming aio poll code.  In addition to that the separate
queues are generally a good idea for performance reasons.

Reviewed-by: Keith Busch <keith.busch@intel.com>
Reviewed-by: Sagi Grimberg <sagi@grimberg.me>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
6 years agonvme-pci: cleanup SQ allocation a bit
Christoph Hellwig [Sun, 2 Dec 2018 16:46:18 +0000 (17:46 +0100)]
nvme-pci: cleanup SQ allocation a bit

Use a bit flag to mark if the SQ was allocated from the CMB, and clean
up the surrounding code a bit.

Reviewed-by: Sagi Grimberg <sagi@grimberg.me>
Reviewed-by: Keith Busch <keith.busch@intel.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
6 years agonvme-pci: use atomic bitops to mark a queue enabled
Christoph Hellwig [Sun, 2 Dec 2018 16:46:17 +0000 (17:46 +0100)]
nvme-pci: use atomic bitops to mark a queue enabled

This gets rid of all the messing with cq_vector and the ->polled field
by using an atomic bitop to mark the queue enabled or not.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Keith Busch <keith.busch@intel.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
6 years agoblock: move queues types to the block layer
Christoph Hellwig [Sun, 2 Dec 2018 16:46:16 +0000 (17:46 +0100)]
block: move queues types to the block layer

Having another indirect all in the fast path doesn't really help
in our post-spectre world.  Also having too many queue type is just
going to create confusion, so I'd rather manage them centrally.

Note that the queue type naming and ordering changes a bit - the
first index now is the default queue for everything not explicitly
marked, the optional ones are read and poll queues.

Reviewed-by: Sagi Grimberg <sagi@grimberg.me>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
6 years agoaio: clear IOCB_HIPRI
Christoph Hellwig [Thu, 22 Nov 2018 15:44:07 +0000 (16:44 +0100)]
aio: clear IOCB_HIPRI

No one is going to poll for aio (yet), so we must clear the HIPRI
flag, as we would otherwise send it down the poll queues, where no
one will be polling for completions.

Signed-off-by: Christoph Hellwig <hch@lst.de>
IOCB_HIPRI, not RWF_HIPRI.

Reviewed-by: Johannes Thumshirn <jthumshirn@suse.de>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
6 years agoMerge tag 'v4.20-rc5' into for-4.21/block
Jens Axboe [Tue, 4 Dec 2018 16:38:05 +0000 (09:38 -0700)]
Merge tag 'v4.20-rc5' into for-4.21/block

Pull in v4.20-rc5, solving a conflict we'll otherwise get in aio.c and
also getting the merge fix that went into mainline that users are
hitting testing for-4.21/block and/or for-next.

* tag 'v4.20-rc5': (664 commits)
  Linux 4.20-rc5
  PCI: Fix incorrect value returned from pcie_get_speed_cap()
  MAINTAINERS: Update linux-mips mailing list address
  ocfs2: fix potential use after free
  mm/khugepaged: fix the xas_create_range() error path
  mm/khugepaged: collapse_shmem() do not crash on Compound
  mm/khugepaged: collapse_shmem() without freezing new_page
  mm/khugepaged: minor reorderings in collapse_shmem()
  mm/khugepaged: collapse_shmem() remember to clear holes
  mm/khugepaged: fix crashes due to misaccounted holes
  mm/khugepaged: collapse_shmem() stop if punched or truncated
  mm/huge_memory: fix lockdep complaint on 32-bit i_size_read()
  mm/huge_memory: splitting set mapping+index before unfreeze
  mm/huge_memory: rename freeze_page() to unmap_page()
  initramfs: clean old path before creating a hardlink
  kernel/kcov.c: mark funcs in __sanitizer_cov_trace_pc() as notrace
  psi: make disabling/enabling easier for vendor kernels
  proc: fixup map_files test on arm
  debugobjects: avoid recursive calls with kmemleak
  userfaultfd: shmem: UFFDIO_COPY: set the page dirty if VM_WRITE is not set
  ...

6 years agosbitmap: fix sbitmap_for_each_set()
Omar Sandoval [Mon, 3 Dec 2018 22:45:43 +0000 (14:45 -0800)]
sbitmap: fix sbitmap_for_each_set()

We need to ignore bits in the cleared mask when iterating over all set
bits.

Fixes: ea86ea2cdced ("sbitmap: ammortize cost of clearing bits")
Reported-by: Jens Axboe@kernel.dk>
Signed-off-by: Omar Sandoval <osandov@fb.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
6 years agoblk-mq: don't call ktime_get_ns() if we don't need it
Jens Axboe [Wed, 28 Nov 2018 17:50:07 +0000 (10:50 -0700)]
blk-mq: don't call ktime_get_ns() if we don't need it

We only need the request fields and the end_io time if we have
stats enabled, or if we have a scheduler attached as those may
use it for completion time stats.

Reviewed-by: Omar Sandoval <osandov@fb.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
6 years agoLinux 4.20-rc5
Linus Torvalds [Sun, 2 Dec 2018 23:07:55 +0000 (15:07 -0800)]
Linux 4.20-rc5

6 years agoMerge tag 'armsoc-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/arm/arm-soc
Linus Torvalds [Sun, 2 Dec 2018 20:19:44 +0000 (12:19 -0800)]
Merge tag 'armsoc-fixes' of git://git./linux/kernel/git/arm/arm-soc

Pull ARM SoC fixes from Olof Johansson:
 "Volume is a little higher than usual due to a set of gpio fixes for
  Davinci platforms that's been around a while, still seemed appropriate
  to not hold off until next merge window.

  Besides that it's the usual mix of minor fixes, mostly corrections of
  small stuff in device trees.

  Major stability-related one is the removal of a regulator from DT on
  Rock960, since DVFS caused undervoltage. I expect it'll be restored
  once they figure out the underlying issue"

* tag 'armsoc-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/arm/arm-soc: (28 commits)
  MAINTAINERS: Remove unused Qualcomm SoC mailing list
  ARM: davinci: dm644x: set the GPIO base to 0
  ARM: davinci: da830: set the GPIO base to 0
  ARM: davinci: dm355: set the GPIO base to 0
  ARM: davinci: dm646x: set the GPIO base to 0
  ARM: davinci: dm365: set the GPIO base to 0
  ARM: davinci: da850: set the GPIO base to 0
  gpio: davinci: restore a way to manually specify the GPIO base
  ARM: davinci: dm644x: define gpio interrupts as separate resources
  ARM: davinci: dm355: define gpio interrupts as separate resources
  ARM: davinci: dm646x: define gpio interrupts as separate resources
  ARM: davinci: dm365: define gpio interrupts as separate resources
  ARM: davinci: da8xx: define gpio interrupts as separate resources
  ARM: dts: at91: sama5d2: use the divided clock for SMC
  ARM: dts: imx51-zii-rdu1: Remove EEPROM node
  ARM: dts: rockchip: Remove @0 from the veyron memory node
  arm64: dts: rockchip: Fix PCIe reset polarity for rk3399-puma-haikou.
  arm64: dts: qcom: msm8998: Reserve gpio ranges on MTP
  arm64: dts: sdm845-mtp: Reserve reserved gpios
  arm64: dts: ti: k3-am654: Fix wakeup_uart reg address
  ...

6 years agoMerge tag 'for-linus-4.20a-rc5-tag' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Sun, 2 Dec 2018 20:15:55 +0000 (12:15 -0800)]
Merge tag 'for-linus-4.20a-rc5-tag' of git://git./linux/kernel/git/xen/tip

Pull xen fixes from Juergen Gross:

 - A revert of a previous commit as it is no longer necessary and has
   shown to cause problems in some memory hotplug cases.

 - Some small fixes and a minor cleanup.

 - A patch for adding better diagnostic data in a very rare failure
   case.

* tag 'for-linus-4.20a-rc5-tag' of git://git.kernel.org/pub/scm/linux/kernel/git/xen/tip:
  pvcalls-front: fixes incorrect error handling
  Revert "xen/balloon: Mark unallocated host memory as UNUSABLE"
  xen: xlate_mmu: add missing header to fix 'W=1' warning
  xen/x86: add diagnostic printout to xen_mc_flush() in case of error
  x86/xen: cleanup includes in arch/x86/xen/spinlock.c

6 years agoMerge tag 'dmaengine-fix-4.20-rc5' of git://git.infradead.org/users/vkoul/slave-dma
Linus Torvalds [Sun, 2 Dec 2018 20:07:27 +0000 (12:07 -0800)]
Merge tag 'dmaengine-fix-4.20-rc5' of git://git.infradead.org/users/vkoul/slave-dma

Pull dmaengine fixes from Vinod Koul:
 "This contains two fixes to at_hdmac which fixes long standing bus
  reported recently on serial transfers causing memory leak. These fixes
  were done by Richard Genoud"

* tag 'dmaengine-fix-4.20-rc5' of git://git.infradead.org/users/vkoul/slave-dma:
  dmaengine: at_hdmac: fix module unloading
  dmaengine: at_hdmac: fix memory leak in at_dma_xlate()

6 years agoMerge branch 'x86-pti-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Sat, 1 Dec 2018 20:35:48 +0000 (12:35 -0800)]
Merge branch 'x86-pti-for-linus' of git://git./linux/kernel/git/tip/tip

Pull STIBP fallout fixes from Thomas Gleixner:
 "The performance destruction department finally got it's act together
  and came up with a cure for the STIPB regression:

   - Provide a command line option to control the spectre v2 user space
     mitigations. Default is either seccomp or prctl (if seccomp is
     disabled in Kconfig). prctl allows mitigation opt-in, seccomp
     enables the migitation for sandboxed processes.

   - Rework the code to handle the conditional STIBP/IBPB control and
     remove the now unused ptrace_may_access_sched() optimization
     attempt

   - Disable STIBP automatically when SMT is disabled

   - Optimize the switch_to() logic to avoid MSR writes and invocations
     of __switch_to_xtra().

   - Make the asynchronous speculation TIF updates synchronous to
     prevent stale mitigation state.

  As a general cleanup this also makes retpoline directly depend on
  compiler support and removes the 'minimal retpoline' option which just
  pretended to provide some form of security while providing none"

* 'x86-pti-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip: (31 commits)
  x86/speculation: Provide IBPB always command line options
  x86/speculation: Add seccomp Spectre v2 user space protection mode
  x86/speculation: Enable prctl mode for spectre_v2_user
  x86/speculation: Add prctl() control for indirect branch speculation
  x86/speculation: Prepare arch_smt_update() for PRCTL mode
  x86/speculation: Prevent stale SPEC_CTRL msr content
  x86/speculation: Split out TIF update
  ptrace: Remove unused ptrace_may_access_sched() and MODE_IBRS
  x86/speculation: Prepare for conditional IBPB in switch_mm()
  x86/speculation: Avoid __switch_to_xtra() calls
  x86/process: Consolidate and simplify switch_to_xtra() code
  x86/speculation: Prepare for per task indirect branch speculation control
  x86/speculation: Add command line control for indirect branch speculation
  x86/speculation: Unify conditional spectre v2 print functions
  x86/speculataion: Mark command line parser data __initdata
  x86/speculation: Mark string arrays const correctly
  x86/speculation: Reorder the spec_v2 code
  x86/l1tf: Show actual SMT state
  x86/speculation: Rework SMT state change
  sched/smt: Expose sched_smt_present static key
  ...

6 years agoMerge tag 'for-linus-20181201' of git://git.kernel.dk/linux-block
Linus Torvalds [Sat, 1 Dec 2018 19:36:32 +0000 (11:36 -0800)]
Merge tag 'for-linus-20181201' of git://git.kernel.dk/linux-block

Pull block layer fixes from Jens Axboe:

 - Single range elevator discard merge fix, that caused crashes (Ming)

 - Fix for a regression in O_DIRECT, where we could potentially lose the
   error value (Maximilian Heyne)

 - NVMe pull request from Christoph, with little fixes all over the map
   for NVMe.

* tag 'for-linus-20181201' of git://git.kernel.dk/linux-block:
  block: fix single range discard merge
  nvme-rdma: fix double freeing of async event data
  nvme: flush namespace scanning work just before removing namespaces
  nvme: warn when finding multi-port subsystems without multipathing enabled
  fs: fix lost error code in dio_complete
  nvme-pci: fix surprise removal
  nvme-fc: initialize nvme_req(rq)->ctrl after calling __nvme_fc_init_request()
  nvme: Free ctrl device name on init failure

6 years agoMerge tag 'pci-v4.20-fixes-2' of git://git.kernel.org/pub/scm/linux/kernel/git/helgaa...
Linus Torvalds [Sat, 1 Dec 2018 19:32:49 +0000 (11:32 -0800)]
Merge tag 'pci-v4.20-fixes-2' of git://git./linux/kernel/git/helgaas/pci

Pull PCI fixes from Bjorn Helgaas:

 - Fix a link speed checking interface that broke PCIe gen3 cards in
   gen1 slots (Mikulas Patocka)

 - Fix an imx6 link training error (Trent Piepho)

 - Fix a layerscape outbound window accessor calling error (Hou
   Zhiqiang)

 - Fix a DesignWare endpoint MSI-X address calculation error (Gustavo
   Pimentel)

* tag 'pci-v4.20-fixes-2' of git://git.kernel.org/pub/scm/linux/kernel/git/helgaas/pci:
  PCI: Fix incorrect value returned from pcie_get_speed_cap()
  PCI: dwc: Fix MSI-X EP framework address calculation bug
  PCI: layerscape: Fix wrong invocation of outbound window disable accessor
  PCI: imx6: Fix link training status detection in link up check

6 years agoblock: add cmd_flags to print_req_error
Balbir Singh [Tue, 30 Oct 2018 02:40:15 +0000 (02:40 +0000)]
block: add cmd_flags to print_req_error

I ran into a bug where after hibernation due to incompatible
backends, the block driver returned BLK_STS_NOTSUPP, with the
current message it's hard to find out what the command flags
were. Adding req->cmd_flags help make the problem easier to
diagnose.

Reviewed-by: Eduardo Valentin <eduval@amazon.com>
Reviewed-by: Ming Lei <ming.lei@redhat.com>
Signed-off-by: Balbir Singh <sblbir@amzn.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
6 years agoMerge remote-tracking branch 'lorenzo/pci/controller-fixes' into for-linus
Bjorn Helgaas [Sat, 1 Dec 2018 05:42:08 +0000 (23:42 -0600)]
Merge remote-tracking branch 'lorenzo/pci/controller-fixes' into for-linus

  - Fix DesignWare endpoint MSI-X address calculation bug (Gustavo
    Pimentel)

  - Fix Layerscape outbound window disable usage (Hou Zhiqiang)

  - Fix imx6 link up detection (Trent Piepho)

* lorenzo/pci/controller-fixes:
  PCI: dwc: Fix MSI-X EP framework address calculation bug
  PCI: layerscape: Fix wrong invocation of outbound window disable accessor
  PCI: imx6: Fix link training status detection in link up check

6 years agoPCI: Fix incorrect value returned from pcie_get_speed_cap()
Mikulas Patocka [Mon, 26 Nov 2018 16:37:13 +0000 (10:37 -0600)]
PCI: Fix incorrect value returned from pcie_get_speed_cap()

The macros PCI_EXP_LNKCAP_SLS_*GB are values, not bit masks.  We must mask
the register and compare it against them.

This fixes errors like this:

  amdgpu: [powerplay] failed to send message 261 ret is 0

when a PCIe-v3 card is plugged into a PCIe-v1 slot, because the slot is
being incorrectly reported as PCIe-v3 capable.

6cf57be0f78e, which appeared in v4.17, added pcie_get_speed_cap() with the
incorrect test of PCI_EXP_LNKCAP_SLS as a bitmask.  5d9a63304032, which
appeared in v4.19, changed amdgpu to use pcie_get_speed_cap(), so the
amdgpu bug reports below are regressions in v4.19.

Fixes: 6cf57be0f78e ("PCI: Add pcie_get_speed_cap() to find max supported link speed")
Fixes: 5d9a63304032 ("drm/amdgpu: use pcie functions for link width and speed")
Link: https://bugs.freedesktop.org/show_bug.cgi?id=108704
Link: https://bugs.freedesktop.org/show_bug.cgi?id=108778
Signed-off-by: Mikulas Patocka <mpatocka@redhat.com>
[bhelgaas: update comment, remove use of PCI_EXP_LNKCAP_SLS_8_0GB and
PCI_EXP_LNKCAP_SLS_16_0GB since those should be covered by PCI_EXP_LNKCAP2,
remove test of PCI_EXP_LNKCAP for zero, since that register is required]
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Acked-by: Alex Deucher <alexander.deucher@amd.com>
Cc: stable@vger.kernel.org # v4.17+
6 years agoMerge branch 'akpm' (patches from Andrew)
Linus Torvalds [Sat, 1 Dec 2018 02:45:49 +0000 (18:45 -0800)]
Merge branch 'akpm' (patches from Andrew)

Merge misc fixes from Andrew Morton:
 "31 fixes"

* emailed patches from Andrew Morton <akpm@linux-foundation.org>: (31 commits)
  ocfs2: fix potential use after free
  mm/khugepaged: fix the xas_create_range() error path
  mm/khugepaged: collapse_shmem() do not crash on Compound
  mm/khugepaged: collapse_shmem() without freezing new_page
  mm/khugepaged: minor reorderings in collapse_shmem()
  mm/khugepaged: collapse_shmem() remember to clear holes
  mm/khugepaged: fix crashes due to misaccounted holes
  mm/khugepaged: collapse_shmem() stop if punched or truncated
  mm/huge_memory: fix lockdep complaint on 32-bit i_size_read()
  mm/huge_memory: splitting set mapping+index before unfreeze
  mm/huge_memory: rename freeze_page() to unmap_page()
  initramfs: clean old path before creating a hardlink
  kernel/kcov.c: mark funcs in __sanitizer_cov_trace_pc() as notrace
  psi: make disabling/enabling easier for vendor kernels
  proc: fixup map_files test on arm
  debugobjects: avoid recursive calls with kmemleak
  userfaultfd: shmem: UFFDIO_COPY: set the page dirty if VM_WRITE is not set
  userfaultfd: shmem: add i_size checks
  userfaultfd: shmem/hugetlbfs: only allow to register VM_MAYWRITE vmas
  userfaultfd: shmem: allocate anonymous memory for MAP_PRIVATE shmem
  ...

6 years agoMerge tag 'mips_fixes_4.20_4' of git://git.kernel.org/pub/scm/linux/kernel/git/mips...
Linus Torvalds [Sat, 1 Dec 2018 02:41:06 +0000 (18:41 -0800)]
Merge tag 'mips_fixes_4.20_4' of git://git./linux/kernel/git/mips/linux

Pull few more MIPS fixes from Paul Burton:

 - Fix mips_get_syscall_arg() to operate on the task specified when
   detecting o32 tasks running on MIPS64 kernels.

 - Fix some incorrect GPIO pin muxing for the MT7620 SoC.

 - Update the linux-mips mailing list address.

* tag 'mips_fixes_4.20_4' of git://git.kernel.org/pub/scm/linux/kernel/git/mips/linux:
  MAINTAINERS: Update linux-mips mailing list address
  MIPS: ralink: Fix mt7620 nd_sd pinmux
  mips: fix mips_get_syscall_arg o32 check

6 years agoMerge tag 'arm64-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/arm64/linux
Linus Torvalds [Sat, 1 Dec 2018 02:39:07 +0000 (18:39 -0800)]
Merge tag 'arm64-fixes' of git://git./linux/kernel/git/arm64/linux

Pull arm64 fixes from Catalin Marinas:

 - Cortex-A76 erratum workaround

 - ftrace fix to enable syscall events on arm64

 - Fix uninitialised pointer in iort_get_platform_device_domain()

* tag 'arm64-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/arm64/linux:
  ACPI/IORT: Fix iort_get_platform_device_domain() uninitialized pointer value
  arm64: ftrace: Fix to enable syscall events on arm64
  arm64: Add workaround for Cortex-A76 erratum 1286807

6 years agoMerge tag 'gcc-plugins-v4.20-rc5' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Sat, 1 Dec 2018 02:36:30 +0000 (18:36 -0800)]
Merge tag 'gcc-plugins-v4.20-rc5' of git://git./linux/kernel/git/kees/linux

Pull stackleak plugin fix from Kees Cook:
 "Fix crash by not allowing kprobing of stackleak_erase() (Alexander
  Popov)"

* tag 'gcc-plugins-v4.20-rc5' of git://git.kernel.org/pub/scm/linux/kernel/git/kees/linux:
  stackleak: Disable function tracing and kprobes for stackleak_erase()

6 years agoMerge tag 'fscache-fixes-20181130' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Sat, 1 Dec 2018 02:32:33 +0000 (18:32 -0800)]
Merge tag 'fscache-fixes-20181130' of git://git./linux/kernel/git/dhowells/linux-fs

Pull fscache and cachefiles fixes from David Howells:
 "Misc fixes:

   - Fix an assertion failure at fs/cachefiles/xattr.c:138 caused by a
     race between a cache object lookup failing and someone attempting
     to reenable that object, thereby triggering an update of the
     object's attributes.

   - Fix an assertion failure at fs/fscache/operation.c:449 caused by a
     split atomic subtract and atomic read that allows a race to happen.

   - Fix a leak of backing pages when simultaneously reading the same
     page from the same object from two or more threads.

   - Fix a hang due to a race between a cache object being discarded and
     the corresponding cookie being reenabled.

  There are also some minor cleanups:

   - Cast an enum value to a different enum type to prevent clang from
     generating a warning. This shouldn't cause any sort of change in
     the emitted code.

   - Use ktime_get_real_seconds() instead of get_seconds(). This is just
     used to uniquify a filename for an object to be placed in the
     graveyard. Objects placed there are deleted by cachfilesd in
     userspace immediately thereafter.

   - Remove an initialised, but otherwise unused variable. This should
     have been entirely optimised away anyway"

* tag 'fscache-fixes-20181130' of git://git.kernel.org/pub/scm/linux/kernel/git/dhowells/linux-fs:
  fscache, cachefiles: remove redundant variable 'cache'
  cachefiles: avoid deprecated get_seconds()
  cachefiles: Explicitly cast enumerated type in put_object
  fscache: fix race between enablement and dropping of object
  cachefiles: Fix page leak in cachefiles_read_backing_file while vmscan is active
  fscache: Fix race in fscache_op_complete() due to split atomic_sub & read
  cachefiles: Fix an assertion failure when trying to update a failed object

6 years agoMAINTAINERS: Update linux-mips mailing list address
Paul Burton [Fri, 30 Nov 2018 19:57:22 +0000 (11:57 -0800)]
MAINTAINERS: Update linux-mips mailing list address

The linux-mips.org infrastructure has been unreliable recently & nobody
with sufficient access to fix it is around to do so. As a result we're
moving away from it, and part of this is migrating our mailing list to
kernel.org.

Replace all instances of linux-mips@linux-mips.org in MAINTAINERS with
the shiny new linux-mips@vger.kernel.org address.

The new list is now being archived on kernel.org at
https://lore.kernel.org/linux-mips/ which also holds the history of the
old linux-mips.org list.

Signed-off-by: Paul Burton <paul.burton@mips.com>
Cc: linux-mips@vger.kernel.org
Cc: linux-mips@linux-mips.org
6 years agoocfs2: fix potential use after free
Pan Bian [Fri, 30 Nov 2018 22:10:54 +0000 (14:10 -0800)]
ocfs2: fix potential use after free

ocfs2_get_dentry() calls iput(inode) to drop the reference count of
inode, and if the reference count hits 0, inode is freed.  However, in
this function, it then reads inode->i_generation, which may result in a
use after free bug.  Move the put operation later.

Link: http://lkml.kernel.org/r/1543109237-110227-1-git-send-email-bianpan2016@163.com
Fixes: 781f200cb7a("ocfs2: Remove masklog ML_EXPORT.")
Signed-off-by: Pan Bian <bianpan2016@163.com>
Reviewed-by: Andrew Morton <akpm@linux-foundation.org>
Cc: Mark Fasheh <mark@fasheh.com>
Cc: Joel Becker <jlbec@evilplan.org>
Cc: Junxiao Bi <junxiao.bi@oracle.com>
Cc: Joseph Qi <jiangqi903@gmail.com>
Cc: Changwei Ge <ge.changwei@h3c.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
6 years agomm/khugepaged: fix the xas_create_range() error path
Hugh Dickins [Fri, 30 Nov 2018 22:10:50 +0000 (14:10 -0800)]
mm/khugepaged: fix the xas_create_range() error path

collapse_shmem()'s xas_nomem() is very unlikely to fail, but it is
rightly given a failure path, so move the whole xas_create_range() block
up before __SetPageLocked(new_page): so that it does not need to
remember to unlock_page(new_page).

Add the missing mem_cgroup_cancel_charge(), and set (currently unused)
result to SCAN_FAIL rather than SCAN_SUCCEED.

Link: http://lkml.kernel.org/r/alpine.LSU.2.11.1811261531200.2275@eggly.anvils
Fixes: 77da9389b9d5 ("mm: Convert collapse_shmem to XArray")
Signed-off-by: Hugh Dickins <hughd@kernel.org>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Cc: Jerome Glisse <jglisse@redhat.com>
Cc: Konstantin Khlebnikov <khlebnikov@yandex-team.ru>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
6 years agomm/khugepaged: collapse_shmem() do not crash on Compound
Hugh Dickins [Fri, 30 Nov 2018 22:10:47 +0000 (14:10 -0800)]
mm/khugepaged: collapse_shmem() do not crash on Compound

collapse_shmem()'s VM_BUG_ON_PAGE(PageTransCompound) was unsafe: before
it holds page lock of the first page, racing truncation then extension
might conceivably have inserted a hugepage there already.  Fail with the
SCAN_PAGE_COMPOUND result, instead of crashing (CONFIG_DEBUG_VM=y) or
otherwise mishandling the unexpected hugepage - though later we might
code up a more constructive way of handling it, with SCAN_SUCCESS.

Link: http://lkml.kernel.org/r/alpine.LSU.2.11.1811261529310.2275@eggly.anvils
Fixes: f3f0e1d2150b2 ("khugepaged: add support of collapse for tmpfs/shmem pages")
Signed-off-by: Hugh Dickins <hughd@google.com>
Cc: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Cc: Jerome Glisse <jglisse@redhat.com>
Cc: Konstantin Khlebnikov <khlebnikov@yandex-team.ru>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: <stable@vger.kernel.org> [4.8+]
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
6 years agomm/khugepaged: collapse_shmem() without freezing new_page
Hugh Dickins [Fri, 30 Nov 2018 22:10:43 +0000 (14:10 -0800)]
mm/khugepaged: collapse_shmem() without freezing new_page

khugepaged's collapse_shmem() does almost all of its work, to assemble
the huge new_page from 512 scattered old pages, with the new_page's
refcount frozen to 0 (and refcounts of all old pages so far also frozen
to 0).  Including shmem_getpage() to read in any which were out on swap,
memory reclaim if necessary to allocate their intermediate pages, and
copying over all the data from old to new.

Imagine the frozen refcount as a spinlock held, but without any lock
debugging to highlight the abuse: it's not good, and under serious load
heads into lockups - speculative getters of the page are not expecting
to spin while khugepaged is rescheduled.

One can get a little further under load by hacking around elsewhere; but
fortunately, freezing the new_page turns out to have been entirely
unnecessary, with no hacks needed elsewhere.

The huge new_page lock is already held throughout, and guards all its
subpages as they are brought one by one into the page cache tree; and
anything reading the data in that page, without the lock, before it has
been marked PageUptodate, would already be in the wrong.  So simply
eliminate the freezing of the new_page.

Each of the old pages remains frozen with refcount 0 after it has been
replaced by a new_page subpage in the page cache tree, until they are
all unfrozen on success or failure: just as before.  They could be
unfrozen sooner, but cause no problem once no longer visible to
find_get_entry(), filemap_map_pages() and other speculative lookups.

Link: http://lkml.kernel.org/r/alpine.LSU.2.11.1811261527570.2275@eggly.anvils
Fixes: f3f0e1d2150b2 ("khugepaged: add support of collapse for tmpfs/shmem pages")
Signed-off-by: Hugh Dickins <hughd@google.com>
Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Cc: Jerome Glisse <jglisse@redhat.com>
Cc: Konstantin Khlebnikov <khlebnikov@yandex-team.ru>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: <stable@vger.kernel.org> [4.8+]
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
6 years agomm/khugepaged: minor reorderings in collapse_shmem()
Hugh Dickins [Fri, 30 Nov 2018 22:10:39 +0000 (14:10 -0800)]
mm/khugepaged: minor reorderings in collapse_shmem()

Several cleanups in collapse_shmem(): most of which probably do not
really matter, beyond doing things in a more familiar and reassuring
order.  Simplify the failure gotos in the main loop, and on success
update stats while interrupts still disabled from the last iteration.

Link: http://lkml.kernel.org/r/alpine.LSU.2.11.1811261526400.2275@eggly.anvils
Fixes: f3f0e1d2150b2 ("khugepaged: add support of collapse for tmpfs/shmem pages")
Signed-off-by: Hugh Dickins <hughd@google.com>
Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Cc: Jerome Glisse <jglisse@redhat.com>
Cc: Konstantin Khlebnikov <khlebnikov@yandex-team.ru>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: <stable@vger.kernel.org> [4.8+]
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
6 years agomm/khugepaged: collapse_shmem() remember to clear holes
Hugh Dickins [Fri, 30 Nov 2018 22:10:35 +0000 (14:10 -0800)]
mm/khugepaged: collapse_shmem() remember to clear holes

Huge tmpfs testing reminds us that there is no __GFP_ZERO in the gfp
flags khugepaged uses to allocate a huge page - in all common cases it
would just be a waste of effort - so collapse_shmem() must remember to
clear out any holes that it instantiates.

The obvious place to do so, where they are put into the page cache tree,
is not a good choice: because interrupts are disabled there.  Leave it
until further down, once success is assured, where the other pages are
copied (before setting PageUptodate).

Link: http://lkml.kernel.org/r/alpine.LSU.2.11.1811261525080.2275@eggly.anvils
Fixes: f3f0e1d2150b2 ("khugepaged: add support of collapse for tmpfs/shmem pages")
Signed-off-by: Hugh Dickins <hughd@google.com>
Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Cc: Jerome Glisse <jglisse@redhat.com>
Cc: Konstantin Khlebnikov <khlebnikov@yandex-team.ru>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: <stable@vger.kernel.org> [4.8+]
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
6 years agomm/khugepaged: fix crashes due to misaccounted holes
Hugh Dickins [Fri, 30 Nov 2018 22:10:29 +0000 (14:10 -0800)]
mm/khugepaged: fix crashes due to misaccounted holes

Huge tmpfs testing on a shortish file mapped into a pmd-rounded extent
hit shmem_evict_inode()'s WARN_ON(inode->i_blocks) followed by
clear_inode()'s BUG_ON(inode->i_data.nrpages) when the file was later
closed and unlinked.

khugepaged's collapse_shmem() was forgetting to update mapping->nrpages
on the rollback path, after it had added but then needs to undo some
holes.

There is indeed an irritating asymmetry between shmem_charge(), whose
callers want it to increment nrpages after successfully accounting
blocks, and shmem_uncharge(), when __delete_from_page_cache() already
decremented nrpages itself: oh well, just add a comment on that to them
both.

And shmem_recalc_inode() is supposed to be called when the accounting is
expected to be in balance (so it can deduce from imbalance that reclaim
discarded some pages): so change shmem_charge() to update nrpages
earlier (though it's rare for the difference to matter at all).

Link: http://lkml.kernel.org/r/alpine.LSU.2.11.1811261523450.2275@eggly.anvils
Fixes: 800d8c63b2e98 ("shmem: add huge pages support")
Fixes: f3f0e1d2150b2 ("khugepaged: add support of collapse for tmpfs/shmem pages")
Signed-off-by: Hugh Dickins <hughd@google.com>
Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Cc: Jerome Glisse <jglisse@redhat.com>
Cc: Konstantin Khlebnikov <khlebnikov@yandex-team.ru>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: <stable@vger.kernel.org> [4.8+]
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
6 years agomm/khugepaged: collapse_shmem() stop if punched or truncated
Hugh Dickins [Fri, 30 Nov 2018 22:10:25 +0000 (14:10 -0800)]
mm/khugepaged: collapse_shmem() stop if punched or truncated

Huge tmpfs testing showed that although collapse_shmem() recognizes a
concurrently truncated or hole-punched page correctly, its handling of
holes was liable to refill an emptied extent.  Add check to stop that.

Link: http://lkml.kernel.org/r/alpine.LSU.2.11.1811261522040.2275@eggly.anvils
Fixes: f3f0e1d2150b2 ("khugepaged: add support of collapse for tmpfs/shmem pages")
Signed-off-by: Hugh Dickins <hughd@google.com>
Reviewed-by: Matthew Wilcox <willy@infradead.org>
Cc: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Cc: Jerome Glisse <jglisse@redhat.com>
Cc: Konstantin Khlebnikov <khlebnikov@yandex-team.ru>
Cc: <stable@vger.kernel.org> [4.8+]
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
6 years agomm/huge_memory: fix lockdep complaint on 32-bit i_size_read()
Hugh Dickins [Fri, 30 Nov 2018 22:10:21 +0000 (14:10 -0800)]
mm/huge_memory: fix lockdep complaint on 32-bit i_size_read()

Huge tmpfs testing, on 32-bit kernel with lockdep enabled, showed that
__split_huge_page() was using i_size_read() while holding the irq-safe
lru_lock and page tree lock, but the 32-bit i_size_read() uses an
irq-unsafe seqlock which should not be nested inside them.

Instead, read the i_size earlier in split_huge_page_to_list(), and pass
the end offset down to __split_huge_page(): all while holding head page
lock, which is enough to prevent truncation of that extent before the
page tree lock has been taken.

Link: http://lkml.kernel.org/r/alpine.LSU.2.11.1811261520070.2275@eggly.anvils
Fixes: baa355fd33142 ("thp: file pages support for split_huge_page()")
Signed-off-by: Hugh Dickins <hughd@google.com>
Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Cc: Jerome Glisse <jglisse@redhat.com>
Cc: Konstantin Khlebnikov <khlebnikov@yandex-team.ru>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: <stable@vger.kernel.org> [4.8+]
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
6 years agomm/huge_memory: splitting set mapping+index before unfreeze
Hugh Dickins [Fri, 30 Nov 2018 22:10:16 +0000 (14:10 -0800)]
mm/huge_memory: splitting set mapping+index before unfreeze

Huge tmpfs stress testing has occasionally hit shmem_undo_range()'s
VM_BUG_ON_PAGE(page_to_pgoff(page) != index, page).

Move the setting of mapping and index up before the page_ref_unfreeze()
in __split_huge_page_tail() to fix this: so that a page cache lookup
cannot get a reference while the tail's mapping and index are unstable.

In fact, might as well move them up before the smp_wmb(): I don't see an
actual need for that, but if I'm missing something, this way round is
safer than the other, and no less efficient.

You might argue that VM_BUG_ON_PAGE(page_to_pgoff(page) != index, page) is
misplaced, and should be left until after the trylock_page(); but left as
is has not crashed since, and gives more stringent assurance.

Link: http://lkml.kernel.org/r/alpine.LSU.2.11.1811261516380.2275@eggly.anvils
Fixes: e9b61f19858a5 ("thp: reintroduce split_huge_page()")
Requires: 605ca5ede764 ("mm/huge_memory.c: reorder operations in __split_huge_page_tail()")
Signed-off-by: Hugh Dickins <hughd@google.com>
Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Cc: Konstantin Khlebnikov <khlebnikov@yandex-team.ru>
Cc: Jerome Glisse <jglisse@redhat.com>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: <stable@vger.kernel.org> [4.8+]
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
6 years agomm/huge_memory: rename freeze_page() to unmap_page()
Hugh Dickins [Fri, 30 Nov 2018 22:10:13 +0000 (14:10 -0800)]
mm/huge_memory: rename freeze_page() to unmap_page()

The term "freeze" is used in several ways in the kernel, and in mm it
has the particular meaning of forcing page refcount temporarily to 0.
freeze_page() is just too confusing a name for a function that unmaps a
page: rename it unmap_page(), and rename unfreeze_page() remap_page().

Went to change the mention of freeze_page() added later in mm/rmap.c,
but found it to be incorrect: ordinary page reclaim reaches there too;
but the substance of the comment still seems correct, so edit it down.

Link: http://lkml.kernel.org/r/alpine.LSU.2.11.1811261514080.2275@eggly.anvils
Fixes: e9b61f19858a5 ("thp: reintroduce split_huge_page()")
Signed-off-by: Hugh Dickins <hughd@google.com>
Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Cc: Jerome Glisse <jglisse@redhat.com>
Cc: Konstantin Khlebnikov <khlebnikov@yandex-team.ru>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: <stable@vger.kernel.org> [4.8+]
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
6 years agoinitramfs: clean old path before creating a hardlink
Li Zhijian [Fri, 30 Nov 2018 22:10:09 +0000 (14:10 -0800)]
initramfs: clean old path before creating a hardlink

sys_link() can fail due to the new path already existing.  This case
ofen occurs when we use a concated initrd, for example:

1) prepare a basic rootfs, it contains a regular files rc.local
lizhijian@:~/yocto-tiny-i386-2016-04-22$ cat etc/rc.local
 #!/bin/sh
 echo "Running /etc/rc.local..."
yocto-tiny-i386-2016-04-22$ find . | sed 's,^\./,,' | cpio -o -H newc | gzip -n -9 >../rootfs.cgz

2) create a extra initrd which also includes a etc/rc.local
lizhijian@:~/lkp-x86_64/etc$ echo "append initrd" >rc.local
lizhijian@:~/lkp/lkp-x86_64/etc$ cat rc.local
append initrd
lizhijian@:~/lkp/lkp-x86_64/etc$ ln rc.local rc.local.hardlink
append initrd
lizhijian@:~/lkp/lkp-x86_64/etc$ stat rc.local rc.local.hardlink
  File: 'rc.local'
  Size: 14         Blocks: 8          IO Block: 4096   regular file
Device: 801h/2049d Inode: 11296086    Links: 2
Access: (0664/-rw-rw-r--)  Uid: ( 1002/lizhijian)   Gid: ( 1002/lizhijian)
Access: 2018-11-15 16:08:28.654464815 +0800
Modify: 2018-11-15 16:07:57.514903210 +0800
Change: 2018-11-15 16:08:24.180228872 +0800
 Birth: -
  File: 'rc.local.hardlink'
  Size: 14         Blocks: 8          IO Block: 4096   regular file
Device: 801h/2049d Inode: 11296086    Links: 2
Access: (0664/-rw-rw-r--)  Uid: ( 1002/lizhijian)   Gid: ( 1002/lizhijian)
Access: 2018-11-15 16:08:28.654464815 +0800
Modify: 2018-11-15 16:07:57.514903210 +0800
Change: 2018-11-15 16:08:24.180228872 +0800
 Birth: -

lizhijian@:~/lkp/lkp-x86_64$ find . | sed 's,^\./,,' | cpio -o -H newc | gzip -n -9 >../rc-local.cgz
lizhijian@:~/lkp/lkp-x86_64$ gzip -dc ../rc-local.cgz | cpio -t
.
etc
etc/rc.local.hardlink <<< it will be extracted first at this initrd
etc/rc.local

3) concate 2 initrds and boot
lizhijian@:~/lkp$ cat rootfs.cgz rc-local.cgz >concate-initrd.cgz
lizhijian@:~/lkp$ qemu-system-x86_64 -nographic -enable-kvm -cpu host -smp 1 -m 1024 -kernel ~/lkp/linux/arch/x86/boot/bzImage -append "console=ttyS0 earlyprint=ttyS0 ignore_loglevel" -initrd ./concate-initr.cgz -serial stdio -nodefaults

In this case, sys_link(2) will fail and return -EEXIST, so we can only get
the rc.local at rootfs.cgz instead of rc-local.cgz

[akpm@linux-foundation.org: move code to avoid forward declaration]
Link: http://lkml.kernel.org/r/1542352368-13299-1-git-send-email-lizhijian@cn.fujitsu.com
Signed-off-by: Li Zhijian <lizhijian@cn.fujitsu.com>
Cc: Philip Li <philip.li@intel.com>
Cc: Dominik Brodowski <linux@dominikbrodowski.net>
Cc: Li Zhijian <zhijianx.li@intel.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>
6 years agokernel/kcov.c: mark funcs in __sanitizer_cov_trace_pc() as notrace
Anders Roxell [Fri, 30 Nov 2018 22:10:05 +0000 (14:10 -0800)]
kernel/kcov.c: mark funcs in __sanitizer_cov_trace_pc() as notrace

Since __sanitizer_cov_trace_pc() is marked as notrace, function calls in
__sanitizer_cov_trace_pc() shouldn't be traced either.
ftrace_graph_caller() gets called for each function that isn't marked
'notrace', like canonicalize_ip().  This is the call trace from a run:

[  139.644550]  ftrace_graph_caller+0x1c/0x24
[  139.648352]  canonicalize_ip+0x18/0x28
[  139.652313]  __sanitizer_cov_trace_pc+0x14/0x58
[  139.656184]  sched_clock+0x34/0x1e8
[  139.659759]  trace_clock_local+0x40/0x88
[  139.663722]  ftrace_push_return_trace+0x8c/0x1f0
[  139.667767]  prepare_ftrace_return+0xa8/0x100
[  139.671709]  ftrace_graph_caller+0x1c/0x24

Rework so that check_kcov_mode() and canonicalize_ip() that are called
from __sanitizer_cov_trace_pc() are also marked as notrace.

Link: http://lkml.kernel.org/r/20181128081239.18317-1-anders.roxell@linaro.org
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Signen-off-by: Anders Roxell <anders.roxell@linaro.org>
Co-developed-by: Arnd Bergmann <arnd@arndb.de>
Acked-by: Steven Rostedt (VMware) <rostedt@goodmis.org>
Cc: Dmitry Vyukov <dvyukov@google.com>
Cc: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
6 years agopsi: make disabling/enabling easier for vendor kernels
Johannes Weiner [Fri, 30 Nov 2018 22:09:58 +0000 (14:09 -0800)]
psi: make disabling/enabling easier for vendor kernels

Mel Gorman reports a hackbench regression with psi that would prohibit
shipping the suse kernel with it default-enabled, but he'd still like
users to be able to opt in at little to no cost to others.

With the current combination of CONFIG_PSI and the psi_disabled bool set
from the commandline, this is a challenge.  Do the following things to
make it easier:

1. Add a config option CONFIG_PSI_DEFAULT_DISABLED that allows distros
   to enable CONFIG_PSI in their kernel but leave the feature disabled
   unless a user requests it at boot-time.

   To avoid double negatives, rename psi_disabled= to psi=.

2. Make psi_disabled a static branch to eliminate any branch costs
   when the feature is disabled.

In terms of numbers before and after this patch, Mel says:

: The following is a comparision using CONFIG_PSI=n as a baseline against
: your patch and a vanilla kernel
:
:                          4.20.0-rc4             4.20.0-rc4             4.20.0-rc4
:                 kconfigdisable-v1r1                vanilla        psidisable-v1r1
: Amean     1       1.3100 (   0.00%)      1.3923 (  -6.28%)      1.3427 (  -2.49%)
: Amean     3       3.8860 (   0.00%)      4.1230 *  -6.10%*      3.8860 (  -0.00%)
: Amean     5       6.8847 (   0.00%)      8.0390 * -16.77%*      6.7727 (   1.63%)
: Amean     7       9.9310 (   0.00%)     10.8367 *  -9.12%*      9.9910 (  -0.60%)
: Amean     12     16.6577 (   0.00%)     18.2363 *  -9.48%*     17.1083 (  -2.71%)
: Amean     18     26.5133 (   0.00%)     27.8833 *  -5.17%*     25.7663 (   2.82%)
: Amean     24     34.3003 (   0.00%)     34.6830 (  -1.12%)     32.0450 (   6.58%)
: Amean     30     40.0063 (   0.00%)     40.5800 (  -1.43%)     41.5087 (  -3.76%)
: Amean     32     40.1407 (   0.00%)     41.2273 (  -2.71%)     39.9417 (   0.50%)
:
: It's showing that the vanilla kernel takes a hit (as the bisection
: indicated it would) and that disabling PSI by default is reasonably
: close in terms of performance for this particular workload on this
: particular machine so;

Link: http://lkml.kernel.org/r/20181127165329.GA29728@cmpxchg.org
Signed-off-by: Johannes Weiner <hannes@cmpxchg.org>
Tested-by: Mel Gorman <mgorman@techsingularity.net>
Reported-by: Mel Gorman <mgorman@techsingularity.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
6 years agoproc: fixup map_files test on arm
Alexey Dobriyan [Fri, 30 Nov 2018 22:09:53 +0000 (14:09 -0800)]
proc: fixup map_files test on arm

https://bugs.linaro.org/show_bug.cgi?id=3782

Turns out arm doesn't permit mapping address 0, so try minimum virtual
address instead.

Link: http://lkml.kernel.org/r/20181113165446.GA28157@avx2
Signed-off-by: Alexey Dobriyan <adobriyan@gmail.com>
Reported-by: Rafael David Tinoco <rafael.tinoco@linaro.org>
Tested-by: Rafael David Tinoco <rafael.tinoco@linaro.org>
Acked-by: Cyrill Gorcunov <gorcunov@gmail.com>
Cc: Shuah Khan <shuah@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
6 years agodebugobjects: avoid recursive calls with kmemleak
Qian Cai [Fri, 30 Nov 2018 22:09:48 +0000 (14:09 -0800)]
debugobjects: avoid recursive calls with kmemleak

CONFIG_DEBUG_OBJECTS_RCU_HEAD does not play well with kmemleak due to
recursive calls.

fill_pool
  kmemleak_ignore
    make_black_object
      put_object
        __call_rcu (kernel/rcu/tree.c)
          debug_rcu_head_queue
            debug_object_activate
              debug_object_init
                fill_pool
                  kmemleak_ignore
                    make_black_object
                      ...

So add SLAB_NOLEAKTRACE to kmem_cache_create() to not register newly
allocated debug objects at all.

Link: http://lkml.kernel.org/r/20181126165343.2339-1-cai@gmx.us
Signed-off-by: Qian Cai <cai@gmx.us>
Suggested-by: Catalin Marinas <catalin.marinas@arm.com>
Acked-by: Waiman Long <longman@redhat.com>
Acked-by: Catalin Marinas <catalin.marinas@arm.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Yang Shi <yang.shi@linux.alibaba.com>
Cc: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
6 years agouserfaultfd: shmem: UFFDIO_COPY: set the page dirty if VM_WRITE is not set
Andrea Arcangeli [Fri, 30 Nov 2018 22:09:43 +0000 (14:09 -0800)]
userfaultfd: shmem: UFFDIO_COPY: set the page dirty if VM_WRITE is not set

Set the page dirty if VM_WRITE is not set because in such case the pte
won't be marked dirty and the page would be reclaimed without writepage
(i.e.  swapout in the shmem case).

This was found by source review.  Most apps (certainly including QEMU)
only use UFFDIO_COPY on PROT_READ|PROT_WRITE mappings or the app can't
modify the memory in the first place.  This is for correctness and it
could help the non cooperative use case to avoid unexpected data loss.

Link: http://lkml.kernel.org/r/20181126173452.26955-6-aarcange@redhat.com
Reviewed-by: Hugh Dickins <hughd@google.com>
Cc: stable@vger.kernel.org
Fixes: 4c27fe4c4c84 ("userfaultfd: shmem: add shmem_mcopy_atomic_pte for userfaultfd support")
Reported-by: Hugh Dickins <hughd@google.com>
Signed-off-by: Andrea Arcangeli <aarcange@redhat.com>
Cc: "Dr. David Alan Gilbert" <dgilbert@redhat.com>
Cc: Jann Horn <jannh@google.com>
Cc: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Mike Rapoport <rppt@linux.ibm.com>
Cc: Peter Xu <peterx@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
6 years agouserfaultfd: shmem: add i_size checks
Andrea Arcangeli [Fri, 30 Nov 2018 22:09:37 +0000 (14:09 -0800)]
userfaultfd: shmem: add i_size checks

With MAP_SHARED: recheck the i_size after taking the PT lock, to
serialize against truncate with the PT lock.  Delete the page from the
pagecache if the i_size_read check fails.

With MAP_PRIVATE: check the i_size after the PT lock before mapping
anonymous memory or zeropages into the MAP_PRIVATE shmem mapping.

A mostly irrelevant cleanup: like we do the delete_from_page_cache()
pagecache removal after dropping the PT lock, the PT lock is a spinlock
so drop it before the sleepable page lock.

Link: http://lkml.kernel.org/r/20181126173452.26955-5-aarcange@redhat.com
Fixes: 4c27fe4c4c84 ("userfaultfd: shmem: add shmem_mcopy_atomic_pte for userfaultfd support")
Signed-off-by: Andrea Arcangeli <aarcange@redhat.com>
Reviewed-by: Mike Rapoport <rppt@linux.ibm.com>
Reviewed-by: Hugh Dickins <hughd@google.com>
Reported-by: Jann Horn <jannh@google.com>
Cc: <stable@vger.kernel.org>
Cc: "Dr. David Alan Gilbert" <dgilbert@redhat.com>
Cc: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Peter Xu <peterx@redhat.com>
Cc: stable@vger.kernel.org
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
6 years agouserfaultfd: shmem/hugetlbfs: only allow to register VM_MAYWRITE vmas
Andrea Arcangeli [Fri, 30 Nov 2018 22:09:32 +0000 (14:09 -0800)]
userfaultfd: shmem/hugetlbfs: only allow to register VM_MAYWRITE vmas

After the VMA to register the uffd onto is found, check that it has
VM_MAYWRITE set before allowing registration.  This way we inherit all
common code checks before allowing to fill file holes in shmem and
hugetlbfs with UFFDIO_COPY.

The userfaultfd memory model is not applicable for readonly files unless
it's a MAP_PRIVATE.

Link: http://lkml.kernel.org/r/20181126173452.26955-4-aarcange@redhat.com
Fixes: ff62a3421044 ("hugetlb: implement memfd sealing")
Signed-off-by: Andrea Arcangeli <aarcange@redhat.com>
Reviewed-by: Mike Rapoport <rppt@linux.ibm.com>
Reviewed-by: Hugh Dickins <hughd@google.com>
Reported-by: Jann Horn <jannh@google.com>
Fixes: 4c27fe4c4c84 ("userfaultfd: shmem: add shmem_mcopy_atomic_pte for userfaultfd support")
Cc: <stable@vger.kernel.org>
Cc: "Dr. David Alan Gilbert" <dgilbert@redhat.com>
Cc: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Peter Xu <peterx@redhat.com>
Cc: stable@vger.kernel.org
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
6 years agouserfaultfd: shmem: allocate anonymous memory for MAP_PRIVATE shmem
Andrea Arcangeli [Fri, 30 Nov 2018 22:09:28 +0000 (14:09 -0800)]
userfaultfd: shmem: allocate anonymous memory for MAP_PRIVATE shmem

Userfaultfd did not create private memory when UFFDIO_COPY was invoked
on a MAP_PRIVATE shmem mapping.  Instead it wrote to the shmem file,
even when that had not been opened for writing.  Though, fortunately,
that could only happen where there was a hole in the file.

Fix the shmem-backed implementation of UFFDIO_COPY to create private
memory for MAP_PRIVATE mappings.  The hugetlbfs-backed implementation
was already correct.

This change is visible to userland, if userfaultfd has been used in
unintended ways: so it introduces a small risk of incompatibility, but
is necessary in order to respect file permissions.

An app that uses UFFDIO_COPY for anything like postcopy live migration
won't notice the difference, and in fact it'll run faster because there
will be no copy-on-write and memory waste in the tmpfs pagecache
anymore.

Userfaults on MAP_PRIVATE shmem keep triggering only on file holes like
before.

The real zeropage can also be built on a MAP_PRIVATE shmem mapping
through UFFDIO_ZEROPAGE and that's safe because the zeropage pte is
never dirty, in turn even an mprotect upgrading the vma permission from
PROT_READ to PROT_READ|PROT_WRITE won't make the zeropage pte writable.

Link: http://lkml.kernel.org/r/20181126173452.26955-3-aarcange@redhat.com
Fixes: 4c27fe4c4c84 ("userfaultfd: shmem: add shmem_mcopy_atomic_pte for userfaultfd support")
Signed-off-by: Andrea Arcangeli <aarcange@redhat.com>
Reported-by: Mike Rapoport <rppt@linux.ibm.com>
Reviewed-by: Hugh Dickins <hughd@google.com>
Cc: <stable@vger.kernel.org>
Cc: "Dr. David Alan Gilbert" <dgilbert@redhat.com>
Cc: Jann Horn <jannh@google.com>
Cc: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Peter Xu <peterx@redhat.com>
Cc: stable@vger.kernel.org
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
6 years agouserfaultfd: use ENOENT instead of EFAULT if the atomic copy user fails
Andrea Arcangeli [Fri, 30 Nov 2018 22:09:25 +0000 (14:09 -0800)]
userfaultfd: use ENOENT instead of EFAULT if the atomic copy user fails

Patch series "userfaultfd shmem updates".

Jann found two bugs in the userfaultfd shmem MAP_SHARED backend: the
lack of the VM_MAYWRITE check and the lack of i_size checks.

Then looking into the above we also fixed the MAP_PRIVATE case.

Hugh by source review also found a data loss source if UFFDIO_COPY is
used on shmem MAP_SHARED PROT_READ mappings (the production usages
incidentally run with PROT_READ|PROT_WRITE, so the data loss couldn't
happen in those production usages like with QEMU).

The whole patchset is marked for stable.

We verified QEMU postcopy live migration with guest running on shmem
MAP_PRIVATE run as well as before after the fix of shmem MAP_PRIVATE.
Regardless if it's shmem or hugetlbfs or MAP_PRIVATE or MAP_SHARED, QEMU
unconditionally invokes a punch hole if the guest mapping is filebacked
and a MADV_DONTNEED too (needed to get rid of the MAP_PRIVATE COWs and
for the anon backend).

This patch (of 5):

We internally used EFAULT to communicate with the caller, switch to
ENOENT, so EFAULT can be used as a non internal retval.

Link: http://lkml.kernel.org/r/20181126173452.26955-2-aarcange@redhat.com
Fixes: 4c27fe4c4c84 ("userfaultfd: shmem: add shmem_mcopy_atomic_pte for userfaultfd support")
Signed-off-by: Andrea Arcangeli <aarcange@redhat.com>
Reviewed-by: Mike Rapoport <rppt@linux.ibm.com>
Reviewed-by: Hugh Dickins <hughd@google.com>
Cc: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Jann Horn <jannh@google.com>
Cc: Peter Xu <peterx@redhat.com>
Cc: "Dr. David Alan Gilbert" <dgilbert@redhat.com>
Cc: <stable@vger.kernel.org>
Cc: stable@vger.kernel.org
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
6 years agolib/test_kmod.c: fix rmmod double free
Luis Chamberlain [Fri, 30 Nov 2018 22:09:21 +0000 (14:09 -0800)]
lib/test_kmod.c: fix rmmod double free

We free the misc device string twice on rmmod; fix this.  Without this
we cannot remove the module without crashing.

Link: http://lkml.kernel.org/r/20181124050500.5257-1-mcgrof@kernel.org
Signed-off-by: Luis Chamberlain <mcgrof@kernel.org>
Reported-by: Randy Dunlap <rdunlap@infradead.org>
Reviewed-by: Andrew Morton <akpm@linux-foundation.org>
Cc: <stable@vger.kernel.org> [4.12+]
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
6 years agohfsplus: do not free node before using
Pan Bian [Fri, 30 Nov 2018 22:09:18 +0000 (14:09 -0800)]
hfsplus: do not free node before using

hfs_bmap_free() frees node via hfs_bnode_put(node).  However it then
reads node->this when dumping error message on an error path, which may
result in a use-after-free bug.  This patch frees node only when it is
never used.

Link: http://lkml.kernel.org/r/1543053441-66942-1-git-send-email-bianpan2016@163.com
Signed-off-by: Pan Bian <bianpan2016@163.com>
Reviewed-by: Andrew Morton <akpm@linux-foundation.org>
Cc: Ernesto A. Fernandez <ernesto.mnd.fernandez@gmail.com>
Cc: Joe Perches <joe@perches.com>
Cc: Viacheslav Dubeyko <slava@dubeyko.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
6 years agohfs: do not free node before using
Pan Bian [Fri, 30 Nov 2018 22:09:14 +0000 (14:09 -0800)]
hfs: do not free node before using

hfs_bmap_free() frees the node via hfs_bnode_put(node).  However, it
then reads node->this when dumping error message on an error path, which
may result in a use-after-free bug.  This patch frees the node only when
it is never again used.

Link: http://lkml.kernel.org/r/1542963889-128825-1-git-send-email-bianpan2016@163.com
Fixes: a1185ffa2fc ("HFS rewrite")
Signed-off-by: Pan Bian <bianpan2016@163.com>
Reviewed-by: Andrew Morton <akpm@linux-foundation.org>
Cc: Joe Perches <joe@perches.com>
Cc: Ernesto A. Fernandez <ernesto.mnd.fernandez@gmail.com>
Cc: Viacheslav Dubeyko <slava@dubeyko.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
6 years agoproc: update MAINTAINERS with proc.txt
Alexey Dobriyan [Fri, 30 Nov 2018 22:09:10 +0000 (14:09 -0800)]
proc: update MAINTAINERS with proc.txt

Turns out that /proc has official documentation and people even trying
to keep it uptodate.

Link: http://lkml.kernel.org/r/20181116134630.GA8004@avx2
Signed-off-by: Alexey Dobriyan <adobriyan@gmail.com>
Reviewed-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>
6 years agomm/page_alloc.c: fix calculation of pgdat->nr_zones
Wei Yang [Fri, 30 Nov 2018 22:09:07 +0000 (14:09 -0800)]
mm/page_alloc.c: fix calculation of pgdat->nr_zones

init_currently_empty_zone() will adjust pgdat->nr_zones and set it to
'zone_idx(zone) + 1' unconditionally.  This is correct in the normal
case, while not exact in hot-plug situation.

This function is used in two places:

  * free_area_init_core()
  * move_pfn_range_to_zone()

In the first case, we are sure zone index increase monotonically.  While
in the second one, this is under users control.

One way to reproduce this is:
----------------------------

1. create a virtual machine with empty node1

   -m 4G,slots=32,maxmem=32G \
   -smp 4,maxcpus=8          \
   -numa node,nodeid=0,mem=4G,cpus=0-3 \
   -numa node,nodeid=1,mem=0G,cpus=4-7

2. hot-add cpu 3-7

   cpu-add [3-7]

2. hot-add memory to nod1

   object_add memory-backend-ram,id=ram0,size=1G
   device_add pc-dimm,id=dimm0,memdev=ram0,node=1

3. online memory with following order

   echo online_movable > memory47/state
   echo online > memory40/state

After this, node1 will have its nr_zones equals to (ZONE_NORMAL + 1)
instead of (ZONE_MOVABLE + 1).

Michal said:
 "Having an incorrect nr_zones might result in all sorts of problems
  which would be quite hard to debug (e.g. reclaim not considering the
  movable zone). I do not expect many users would suffer from this it
  but still this is trivial and obviously right thing to do so
  backporting to the stable tree shouldn't be harmful (last famous
  words)"

Link: http://lkml.kernel.org/r/20181117022022.9956-1-richard.weiyang@gmail.com
Fixes: f1dd2cd13c4b ("mm, memory_hotplug: do not associate hotadded memory to zones until online")
Signed-off-by: Wei Yang <richard.weiyang@gmail.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Reviewed-by: Oscar Salvador <osalvador@suse.de>
Cc: Anshuman Khandual <anshuman.khandual@arm.com>
Cc: Dave Hansen <dave.hansen@intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
6 years agomm: use swp_offset as key in shmem_replace_page()
Yu Zhao [Fri, 30 Nov 2018 22:09:03 +0000 (14:09 -0800)]
mm: use swp_offset as key in shmem_replace_page()

We changed the key of swap cache tree from swp_entry_t.val to
swp_offset.  We need to do so in shmem_replace_page() as well.

Hugh said:
 "shmem_replace_page() has been wrong since the day I wrote it: good
  enough to work on swap "type" 0, which is all most people ever use
  (especially those few who need shmem_replace_page() at all), but
  broken once there are any non-0 swp_type bits set in the higher order
  bits"

Link: http://lkml.kernel.org/r/20181121215442.138545-1-yuzhao@google.com
Fixes: f6ab1f7f6b2d ("mm, swap: use offset of swap entry as key of swap cache")
Signed-off-by: Yu Zhao <yuzhao@google.com>
Reviewed-by: Matthew Wilcox <willy@infradead.org>
Acked-by: Hugh Dickins <hughd@google.com>
Cc: <stable@vger.kernel.org> [4.9+]
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
6 years agomm: cleancache: fix corruption on missed inode invalidation
Pavel Tikhomirov [Fri, 30 Nov 2018 22:09:00 +0000 (14:09 -0800)]
mm: cleancache: fix corruption on missed inode invalidation

If all pages are deleted from the mapping by memory reclaim and also
moved to the cleancache:

__delete_from_page_cache
  (no shadow case)
  unaccount_page_cache_page
    cleancache_put_page
  page_cache_delete
    mapping->nrpages -= nr
    (nrpages becomes 0)

We don't clean the cleancache for an inode after final file truncation
(removal).

truncate_inode_pages_final
  check (nrpages || nrexceptional) is false
    no truncate_inode_pages
      no cleancache_invalidate_inode(mapping)

These way when reading the new file created with same inode we may get
these trash leftover pages from cleancache and see wrong data instead of
the contents of the new file.

Fix it by always doing truncate_inode_pages which is already ready for
nrpages == 0 && nrexceptional == 0 case and just invalidates inode.

[akpm@linux-foundation.org: add comment, per Jan]
Link: http://lkml.kernel.org/r/20181112095734.17979-1-ptikhomirov@virtuozzo.com
Fixes: commit 91b0abe36a7b ("mm + fs: store shadow entries in page cache")
Signed-off-by: Pavel Tikhomirov <ptikhomirov@virtuozzo.com>
Reviewed-by: Vasily Averin <vvs@virtuozzo.com>
Reviewed-by: Andrey Ryabinin <aryabinin@virtuozzo.com>
Reviewed-by: Jan Kara <jack@suse.cz>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Mel Gorman <mgorman@techsingularity.net>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
6 years agoocfs2: fix deadlock caused by ocfs2_defrag_extent()
Larry Chen [Fri, 30 Nov 2018 22:08:56 +0000 (14:08 -0800)]
ocfs2: fix deadlock caused by ocfs2_defrag_extent()

ocfs2_defrag_extent may fall into deadlock.

ocfs2_ioctl_move_extents
    ocfs2_ioctl_move_extents
      ocfs2_move_extents
        ocfs2_defrag_extent
          ocfs2_lock_allocators_move_extents

            ocfs2_reserve_clusters
              inode_lock GLOBAL_BITMAP_SYSTEM_INODE

  __ocfs2_flush_truncate_log
              inode_lock GLOBAL_BITMAP_SYSTEM_INODE

As backtrace shows above, ocfs2_reserve_clusters() will call inode_lock
against the global bitmap if local allocator has not sufficient cluters.
Once global bitmap could meet the demand, ocfs2_reserve_cluster will
return success with global bitmap locked.

After ocfs2_reserve_cluster(), if truncate log is full,
__ocfs2_flush_truncate_log() will definitely fall into deadlock because
it needs to inode_lock global bitmap, which has already been locked.

To fix this bug, we could remove from
ocfs2_lock_allocators_move_extents() the code which intends to lock
global allocator, and put the removed code after
__ocfs2_flush_truncate_log().

ocfs2_lock_allocators_move_extents() is referred by 2 places, one is
here, the other does not need the data allocator context, which means
this patch does not affect the caller so far.

Link: http://lkml.kernel.org/r/20181101071422.14470-1-lchen@suse.com
Signed-off-by: Larry Chen <lchen@suse.com>
Reviewed-by: Changwei Ge <ge.changwei@h3c.com>
Cc: Mark Fasheh <mark@fasheh.com>
Cc: Joel Becker <jlbec@evilplan.org>
Cc: Junxiao Bi <junxiao.bi@oracle.com>
Cc: Joseph Qi <jiangqi903@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
6 years agomm/gup: finish consolidating error handling
John Hubbard [Fri, 30 Nov 2018 22:08:53 +0000 (14:08 -0800)]
mm/gup: finish consolidating error handling

Commit df06b37ffe5a ("mm/gup: cache dev_pagemap while pinning pages")
attempted to operate on each page that get_user_pages had retrieved.  In
order to do that, it created a common exit point from the routine.
However, one case was missed, which this patch fixes up.

Also, there was still an unnecessary shadow declaration (with a
different type) of the "ret" variable, which this patch removes.

Keith's description of the situation is:

  This also fixes a potentially leaked dev_pagemap reference count if a
  failure occurs when an iteration crosses a vma boundary.  I don't think
  it's normal to have different vma's on a users mapped zone device
  memory, but good to fix anyway.

I actually thought that this code:

    /* first iteration or cross vma bound */
    if (!vma || start >= vma->vm_end) {
        vma = find_extend_vma(mm, start);
        if (!vma && in_gate_area(mm, start)) {
            ret = get_gate_page(mm, start & PAGE_MASK,
                    gup_flags, &vma,
                    pages ? &pages[i] : NULL);
            if (ret)
                goto out;

dealt with the "you're trying to pin the gate page, as part of this
call", rather than the generic case of crossing a vma boundary.  (I
think there's a fine point that I must be overlooking.) But it's still a
valid case, either way.

Link: http://lkml.kernel.org/r/20181121081402.29641-2-jhubbard@nvidia.com
Fixes: df06b37ffe5a4 ("mm/gup: cache dev_pagemap while pinning pages")
Signed-off-by: John Hubbard <jhubbard@nvidia.com>
Reviewed-by: Keith Busch <keith.busch@intel.com>
Cc: Dan Williams <dan.j.williams@intel.com>
Cc: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Cc: Dave Hansen <dave.hansen@intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
6 years agoMAINTAINERS: name change for Luis
Luis Chamberlain [Fri, 30 Nov 2018 22:08:49 +0000 (14:08 -0800)]
MAINTAINERS: name change for Luis

My name has changed, works better than Global Entry I tell ya.

Link: http://lkml.kernel.org/r/20181122003138.7752-1-mcgrof@kernel.org
Signed-off-by: Luis Chamberlain <mcgrof@kernel.org>
Cc: Kees Cook <keescook@chromium.org>
Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
6 years agounifdef: use memcpy instead of strncpy
Linus Torvalds [Fri, 30 Nov 2018 22:45:01 +0000 (14:45 -0800)]
unifdef: use memcpy instead of strncpy

New versions of gcc reasonably warn about the odd pattern of

strncpy(p, q, strlen(q));

which really doesn't make sense: the strncpy() ends up being just a slow
and odd way to write memcpy() in this case.

There was a comment about _why_ the code used strncpy - to avoid the
terminating NUL byte, but memcpy does the same and avoids the warning.

Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
6 years agosbitmap: optimize wakeup check
Jens Axboe [Fri, 30 Nov 2018 00:36:41 +0000 (17:36 -0700)]
sbitmap: optimize wakeup check

Even if we have no waiters on any of the sbitmap_queue wait states, we
still have to loop every entry to check. We do this for every IO, so
the cost adds up.

Shift a bit of the cost to the slow path, when we actually have waiters.
Wrap prepare_to_wait_exclusive() and finish_wait(), so we can maintain
an internal count of how many are currently active. Then we can simply
check this count in sbq_wake_ptr() and not have to loop if we don't
have any sleepers.

Convert the two users of sbitmap with waiting, blk-mq-tag and iSCSI.

Reviewed-by: Omar Sandoval <osandov@fb.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
6 years agosbitmap: ammortize cost of clearing bits
Jens Axboe [Fri, 30 Nov 2018 20:18:06 +0000 (13:18 -0700)]
sbitmap: ammortize cost of clearing bits

sbitmap maintains a set of words that we use to set and clear bits, with
each bit representing a tag for blk-mq. Even though we spread the bits
out and maintain a hint cache, one particular bit allocated will end up
being cleared in the exact same spot.

This introduces batched clearing of bits. Instead of clearing a given
bit, the same bit is set in a cleared/free mask instead. If we fail
allocating a bit from a given word, then we check the free mask, and
batch move those cleared bits at that time. This trades 64 atomic bitops
for 2 cmpxchg().

In a threaded poll test case, half the overhead of getting and clearing
tags is removed with this change. On another poll test case with a
single thread, performance is unchanged.

Reviewed-by: Omar Sandoval <osandov@fb.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
6 years agoMerge tag 'char-misc-4.20-rc5' of git://git.kernel.org/pub/scm/linux/kernel/git/gregk...
Linus Torvalds [Fri, 30 Nov 2018 20:43:17 +0000 (12:43 -0800)]
Merge tag 'char-misc-4.20-rc5' of git://git./linux/kernel/git/gregkh/char-misc

Pull char/misc fixes from Greg KH:
 "Here are a few small char/misc driver fixes for 4.20-rc5 that resolve
  a number of reported issues.

  The "largest" here is the thunderbolt patch, which resolves an issue
  with NVM upgrade, the smallest being some fsi driver fixes. There's
  also a hyperv bugfix, and the usual binder bugfixes.

  All of these have been in linux-next with no reported issues"

* tag 'char-misc-4.20-rc5' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/char-misc:
  misc: mic/scif: fix copy-paste error in scif_create_remote_lookup
  thunderbolt: Prevent root port runtime suspend during NVM upgrade
  Drivers: hv: vmbus: check the creation_status in vmbus_establish_gpadl()
  binder: fix race that allows malicious free of live buffer
  fsi: fsi-scom.c: Remove duplicate header
  fsi: master-ast-cf: select GENERIC_ALLOCATOR

6 years agoMerge tag 'driver-core-4.20-rc5' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Fri, 30 Nov 2018 20:26:06 +0000 (12:26 -0800)]
Merge tag 'driver-core-4.20-rc5' of git://git./linux/kernel/git/gregkh/driver-core

Pull driver core fix from Greg KH:
 "Here is a single driver core fix for 4.20-rc5

  It resolves an issue with the data alignment in 'struct devres' for
  the ARC platform. The full details are in the commit changelog, but
  the short summary is the change is a single line:

-       unsigned long long              data[]; /* guarantee ull alignment */
+       u8 __aligned(ARCH_KMALLOC_MINALIGN) data[];

  This has been in linux-next for a while with no reported issues"

* tag 'driver-core-4.20-rc5' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/driver-core:
  devres: Align data[] to ARCH_KMALLOC_MINALIGN

6 years agoMerge tag 'staging-4.20-rc5' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh...
Linus Torvalds [Fri, 30 Nov 2018 20:23:44 +0000 (12:23 -0800)]
Merge tag 'staging-4.20-rc5' of git://git./linux/kernel/git/gregkh/staging

Pull staging and IIO driver fixes from Greg KH:
 "Here are some small IIO and staging driver fixes for 4.20-rc5.

  Nothing major, the IIO fix ended up touching the HID drivers at the
  same time, but the HID maintainer acked it. The staging fixes are all
  minor patches for reported issues and regressions, full details are in
  the shortlog.

  All of these have been in linux-next for a while with no reported
  issues"

* tag 'staging-4.20-rc5' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/staging:
  iio/hid-sensors: Fix IIO_CHAN_INFO_RAW returning wrong values for signed numbers
  staging: vchiq_arm: fix compat VCHIQ_IOC_AWAIT_COMPLETION
  staging: mt7621-pinctrl: fix uninitialized variable ngroups
  staging: rtl8723bs: Add missing return for cfg80211_rtw_get_station
  staging: most: use format specifier "%s" in snprintf
  staging: rtl8723bs: Fix incorrect sense of ether_addr_equal
  staging: mt7621-dma: fix potentially dereferencing uninitialized 'tx_desc'
  staging: comedi: clarify/unify macros for NI macro-defined terminals
  drivers: staging: cedrus: find ctx before dereferencing it ctx
  staging: rtl8723bs: Fix the return value in case of error in 'rtw_wx_read32()'
  staging: comedi: ni_mio_common: scale ao INSN_CONFIG_GET_CMD_TIMING_CONSTRAINTS
  iio:st_magn: Fix enable device after trigger