cxgb4: Fix kernel panic while accessing sge_info

[ Upstream commit 479a0d1376f6d97c60871442911f1394d4446a25 ]

The sge_info debugfs collects offload queue info even when offload
capability is disabled and leads to panic.

[  144.139871] CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[  144.139874] CR2: 0000000000000000 CR3: 000000082d456005 CR4: 00000000001606e0
[  144.139876] Call Trace:
[  144.139887]  sge_queue_start+0x12/0x30 [cxgb4]
[  144.139897]  seq_read+0x1d4/0x3d0
[  144.139906]  full_proxy_read+0x50/0x70
[  144.139913]  vfs_read+0x89/0x140
[  144.139916]  ksys_read+0x55/0xd0
[  144.139924]  do_syscall_64+0x5b/0x1d0
[  144.139933]  entry_SYSCALL_64_after_hwframe+0x44/0xa9
[  144.139936] RIP: 0033:0x7f4b01493990

Fix this crash by skipping the offload queue access in sge_qinfo when
offload capability is disabled

Signed-off-by: Herat Ramani <herat@chelsio.com>
Signed-off-by: Vishal Kulkarni <vishal@chelsio.com>
Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

bnx2x: Fix logic to get total no. of PFs per engine

[ Upstream commit ee699f89bdbaa19c399804504241b5c531b48888 ]

Driver doesn't calculate total number of PFs configured on a
given engine correctly which messed up resources in the PFs
loaded on that engine, leading driver to exceed configuration
of resources (like vlan filters etc.) beyond the limit per
engine, which ended up with asserts from the firmware.

Signed-off-by: Manish Chopra <manishc@marvell.com>
Signed-off-by: Ariel Elior <aelior@marvell.com>
Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

bnx2x: Do not handle requests from VFs after parity

[ Upstream commit 7113f796bbbced2470cd6d7379d50d7a7a78bf34 ]

Parity error from the hardware will cause PF to lose the state
of their VFs due to PF's internal reload and hardware reset following
the parity error. Restrict any configuration request from the VFs after
the parity as it could cause unexpected hardware behavior, only way
for VFs to recover would be to trigger FLR on VFs and reload them.

Signed-off-by: Manish Chopra <manishc@marvell.com>
Signed-off-by: Ariel Elior <aelior@marvell.com>
Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

habanalabs: remove variable 'val' set but not used

[ Upstream commit 68a1fdf2451f38b4ada0607eb6e1303f8a02e0b7 ]

Fixes gcc '-Wunused-but-set-variable' warning:

drivers/misc/habanalabs/goya/goya.c: In function goya_pldm_init_cpu:
drivers/misc/habanalabs/goya/goya.c:2195:6: warning: variable val set but not used [-Wunused-but-set-variable]
drivers/misc/habanalabs/goya/goya.c: In function goya_hw_init:
drivers/misc/habanalabs/goya/goya.c:2505:6: warning: variable val set but not used [-Wunused-but-set-variable]

Fixes: 9494a8dd8d22 ("habanalabs: add h/w queues module")
Signed-off-by: Chen Wandun <chenwandun@huawei.com>
Reviewed-by: Oded Gabbay <oded.gabbay@gmail.com>
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

habanalabs: rate limit error msg on waiting for CS

[ Upstream commit 018e0e3594f7dcd029d258e368c485e742fa9cdb ]

In case a user submits a CS, and the submission fails, and the user doesn't
check the return value and instead use the error return value as a valid
sequence number of a CS and ask to wait on it, the driver will print an
error and return an error code for that wait.

The real problem happens if now the user ignores the error of the wait, and
try to wait again and again. This can lead to a flood of error messages
from the driver and even soft lockup event.

Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
Reviewed-by: Tomer Tayar <ttayar@habana.ai>
Signed-off-by: Sasha Levin <sashal@kernel.org>

bpf: Clear skb->tstamp in bpf_redirect when necessary

[ Upstream commit 5133498f4ad1123a5ffd4c08df6431dab882cc32 ]

Redirecting a packet from ingress to egress by using bpf_redirect
breaks if the egress interface has an fq qdisc installed. This is the same
problem as fixed in 'commit 8203e2d844d3 ("net: clear skb->tstamp in forwarding paths")

Clear skb->tstamp when redirecting into the egress path.

Fixes: 80b14dee2bea ("net: Add a new socket option for a future transmit time.")
Fixes: fb420d5d91c1 ("tcp/fq: move back to CLOCK_MONOTONIC")
Signed-off-by: Lorenz Bauer <lmb@cloudflare.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Link: https://lore.kernel.org/bpf/20191213180817.2510-1-lmb@cloudflare.com
Signed-off-by: Sasha Levin <sashal@kernel.org>

ocxl: Fix potential memory leak on context creation

[ Upstream commit 913e73c77d48aeeb50c16450a653dca9c71ae2e2 ]

If we couldn't fully init a context, we were leaking memory.

Fixes: b9721d275cc2 ("ocxl: Allow external drivers to use OpenCAPI contexts")
Signed-off-by: Frederic Barrat <fbarrat@linux.ibm.com>
Acked-by: Andrew Donnellan <ajd@linux.ibm.com>
Reviewed-by: Greg Kurz <groug@kaod.org>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20191209105513.8566-1-fbarrat@linux.ibm.com
Signed-off-by: Sasha Levin <sashal@kernel.org>

Btrfs: fix hole extent items with a zero size after range cloning

[ Upstream commit 147271e35ba267506dde6550f58ccf8d287db3ef ]

Normally when cloning a file range if we find an implicit hole at the end
of the range we assume it is because the NO_HOLES feature is enabled.
However that is not always the case. One well known case [1] is when we
have a power failure after mixing buffered and direct IO writes against
the same file.

In such cases we need to punch a hole in the destination file, and if
the NO_HOLES feature is not enabled, we need to insert explicit file
extent items to represent the hole. After commit 690a5dbfc51315
("Btrfs: fix ENOSPC errors, leading to transaction aborts, when cloning
extents"), we started to insert file extent items representing the hole
with an item size of 0, which is invalid and should be 53 bytes (the size
of a btrfs_file_extent_item structure), resulting in all sorts of
corruptions and invalid memory accesses. This is detected by the tree
checker when we attempt to write a leaf to disk.

The problem can be sporadically triggered by test case generic/561 from
fstests. That test case does not exercise power failure and creates a new
filesystem when it starts, so it does not use a filesystem created by any
previous test that tests power failure. However the test does both
buffered and direct IO writes (through fsstress) and it's precisely that
which is creating the implicit holes in files. That happens even before
the commit mentioned earlier. I need to investigate why we get those
implicit holes to check if there is a real problem or not. For now this
change fixes the regression of introducing file extent items with an item
size of 0 bytes.

Fix the issue by calling btrfs_punch_hole_range() without passing a
btrfs_clone_extent_info structure, which ensures file extent items are
inserted to represent the hole with a correct item size. We were passing
a btrfs_clone_extent_info with a value of 0 for its 'item_size' field,
which was causing the insertion of file extent items with an item size
of 0.

[1] https://www.spinics.net/lists/linux-btrfs/msg75350.html

Reported-by: David Sterba <dsterba@suse.com>
Fixes: 690a5dbfc51315 ("Btrfs: fix ENOSPC errors, leading to transaction aborts, when cloning extents")
Signed-off-by: Filipe Manana <fdmanana@suse.com>
Reviewed-by: Josef Bacik <josef@toxicpanda.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

btrfs: handle error in btrfs_cache_block_group

[ Upstream commit db8fe64f9ce61d1d89d3c3c34d111a43afb9f053 ]

We have a BUG_ON(ret < 0) in find_free_extent from
btrfs_cache_block_group.  If we fail to allocate our ctl we'll just
panic, which is not good.  Instead just go on to another block group.
If we fail to find a block group we don't want to return ENOSPC, because
really we got a ENOMEM and that's the root of the problem.  Save our
return from btrfs_cache_block_group(), and then if we still fail to make
our allocation return that ret so we get the right error back.

Tested with inject-error.py from bcc.

Reviewed-by: Johannes Thumshirn <jthumshirn@suse.de>
Signed-off-by: Josef Bacik <josef@toxicpanda.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

powerpc/spinlocks: Include correct header for static key

commit 6da3eced8c5f3b03340b0c395bacd552c4d52411 upstream.

Recently, the spinlock implementation grew a static key optimization,
but the jump_label.h header include was left out, leading to build
errors:

  linux/arch/powerpc/include/asm/spinlock.h:44:7: error: implicit declaration of function ‘static_branch_unlikely’
   44 |  if (!static_branch_unlikely(&shared_processor))

This commit adds the missing header.

mpe: The build break is only seen with CONFIG_JUMP_LABEL=n.

Fixes: 656c21d6af5d ("powerpc/shared: Use static key to detect shared processor")
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
Reviewed-by: Srikar Dronamraju <srikar@linux.ibm.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20191223133147.129983-1-Jason@zx2c4.com
Cc: Srikar Dronamraju <srikar@linux.vnet.ibm.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

powerpc/vcpu: Assume dedicated processors as non-preempt

commit 14c73bd344da60abaf7da3ea2e7733ddda35bbac upstream.

With commit 247f2f6f3c70 ("sched/core: Don't schedule threads on
pre-empted vCPUs"), the scheduler avoids preempted vCPUs to schedule
tasks on wakeup. This leads to wrong choice of CPU, which in-turn
leads to larger wakeup latencies. Eventually, it leads to performance
regression in latency sensitive benchmarks like soltp, schbench etc.

On Powerpc, vcpu_is_preempted() only looks at yield_count. If the
yield_count is odd, the vCPU is assumed to be preempted. However
yield_count is increased whenever the LPAR enters CEDE state (idle).
So any CPU that has entered CEDE state is assumed to be preempted.

Even if vCPU of dedicated LPAR is preempted/donated, it should have
right of first-use since they are supposed to own the vCPU.

On a Power9 System with 32 cores:
  # lscpu
  Architecture:        ppc64le
  Byte Order:          Little Endian
  CPU(s):              128
  On-line CPU(s) list: 0-127
  Thread(s) per core:  8
  Core(s) per socket:  1
  Socket(s):           16
  NUMA node(s):        2
  Model:               2.2 (pvr 004e 0202)
  Model name:          POWER9 (architected), altivec supported
  Hypervisor vendor:   pHyp
  Virtualization type: para
  L1d cache:           32K
  L1i cache:           32K
  L2 cache:            512K
  L3 cache:            10240K
  NUMA node0 CPU(s):   0-63
  NUMA node1 CPU(s):   64-127

  # perf stat -a -r 5 ./schbench
  v5.4                               v5.4 + patch
  Latency percentiles (usec)         Latency percentiles (usec)
        50.0000th: 45                      50.0th: 45
        75.0000th: 62                      75.0th: 63
        90.0000th: 71                      90.0th: 74
        95.0000th: 77                      95.0th: 78
        *99.0000th: 91                     *99.0th: 82
        99.5000th: 707                     99.5th: 83
        99.9000th: 6920                    99.9th: 86
        min=0, max=10048                   min=0, max=96
  Latency percentiles (usec)         Latency percentiles (usec)
        50.0000th: 45                      50.0th: 46
        75.0000th: 61                      75.0th: 64
        90.0000th: 72                      90.0th: 75
        95.0000th: 79                      95.0th: 79
        *99.0000th: 691                    *99.0th: 83
        99.5000th: 3972                    99.5th: 85
        99.9000th: 8368                    99.9th: 91
        min=0, max=16606                   min=0, max=117
  Latency percentiles (usec)         Latency percentiles (usec)
        50.0000th: 45                      50.0th: 46
        75.0000th: 61                      75.0th: 64
        90.0000th: 71                      90.0th: 75
        95.0000th: 77                      95.0th: 79
        *99.0000th: 106                    *99.0th: 83
        99.5000th: 2364                    99.5th: 84
        99.9000th: 7480                    99.9th: 90
        min=0, max=10001                   min=0, max=95
  Latency percentiles (usec)         Latency percentiles (usec)
        50.0000th: 45                      50.0th: 47
        75.0000th: 62                      75.0th: 65
        90.0000th: 72                      90.0th: 75
        95.0000th: 78                      95.0th: 79
        *99.0000th: 93                     *99.0th: 84
        99.5000th: 108                     99.5th: 85
        99.9000th: 6792                    99.9th: 90
        min=0, max=17681                   min=0, max=117
  Latency percentiles (usec)         Latency percentiles (usec)
        50.0000th: 46                      50.0th: 45
        75.0000th: 62                      75.0th: 64
        90.0000th: 73                      90.0th: 75
        95.0000th: 79                      95.0th: 79
        *99.0000th: 113                    *99.0th: 82
        99.5000th: 2724                    99.5th: 83
        99.9000th: 6184                    99.9th: 93
        min=0, max=9887                    min=0, max=111

   Performance counter stats for 'system wide' (5 runs):

  context-switches    43,373  ( +-  0.40% )   44,597 ( +-  0.55% )
  cpu-migrations       1,211  ( +-  5.04% )      220 ( +-  6.23% )
  page-faults         15,983  ( +-  5.21% )   15,360 ( +-  3.38% )

Waiman Long suggested using static_keys.

Fixes: 247f2f6f3c70 ("sched/core: Don't schedule threads on pre-empted vCPUs")
Cc: stable@vger.kernel.org # v4.18+
Reported-by: Parth Shah <parth@linux.ibm.com>
Reported-by: Ihor Pasichnyk <Ihor.Pasichnyk@ibm.com>
Tested-by: Juri Lelli <juri.lelli@redhat.com>
Acked-by: Waiman Long <longman@redhat.com>
Reviewed-by: Gautham R. Shenoy <ego@linux.vnet.ibm.com>
Signed-off-by: Srikar Dronamraju <srikar@linux.vnet.ibm.com>
Acked-by: Phil Auld <pauld@redhat.com>
Reviewed-by: Vaidyanathan Srinivasan <svaidy@linux.ibm.com>
Tested-by: Parth Shah <parth@linux.ibm.com>
[mpe: Move the key and setting of the key to pseries/setup.c]
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20191213035036.6913-1-mpe@ellerman.id.au
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

Btrfs: fix cloning range with a hole when using the NO_HOLES feature

[ Upstream commit fcb970581dd900675c4371c2b688a57924a8368c ]

When using the NO_HOLES feature if we clone a range that contains a hole
and a temporary ENOSPC happens while dropping extents from the target
inode's range, we can end up failing and aborting the transaction with
-EEXIST or with a corrupt file extent item, that has a length greater
than it should and overlaps with other extents. For example when cloning
the following range from inode A to inode B:

  Inode A:

    extent A1                                          extent A2
  [ ----------- ]  [ hole, implicit, 4MB length ]  [ ------------- ]
  0            1MB                                 5MB            6MB

  Range to clone: [1MB, 6MB)

  Inode B:

    extent B1       extent B2        extent B3         extent B4
  [ ---------- ]  [ --------- ]    [ ---------- ]    [ ---------- ]
  0           1MB 1MB        2MB   2MB        5MB    5MB         6MB

  Target range: [1MB, 6MB) (same as source, to make it easier to explain)

The following can happen:

1) btrfs_punch_hole_range() gets -ENOSPC from __btrfs_drop_extents();

2) At that point, 'cur_offset' is set to 1MB and __btrfs_drop_extents()
   set 'drop_end' to 2MB, meaning it was able to drop only extent B2;

3) We then compute 'clone_len' as 'drop_end' - 'cur_offset' = 2MB - 1MB =
   1MB;

4) We then attempt to insert a file extent item at inode B with a file
   offset of 5MB, which is the value of clone_info->file_offset. This
   fails with error -EEXIST because there's already an extent at that
   offset (extent B4);

5) We abort the current transaction with -EEXIST and return that error
   to user space as well.

Another example, for extent corruption:

  Inode A:

    extent A1                                           extent A2
  [ ----------- ]   [ hole, implicit, 10MB length ]  [ ------------- ]
  0            1MB                                  11MB            12MB

  Inode B:

    extent B1         extent B2
  [ ----------- ]   [ --------- ]    [ ----------------------------- ]
  0            1MB 1MB         5MB  5MB                             12MB

  Target range: [1MB, 12MB) (same as source, to make it easier to explain)

1) btrfs_punch_hole_range() gets -ENOSPC from __btrfs_drop_extents();

2) At that point, 'cur_offset' is set to 1MB and __btrfs_drop_extents()
   set 'drop_end' to 5MB, meaning it was able to drop only extent B2;

3) We then compute 'clone_len' as 'drop_end' - 'cur_offset' = 5MB - 1MB =
   4MB;

4) We then insert a file extent item at inode B with a file offset of 11MB
   which is the value of clone_info->file_offset, and a length of 4MB (the
   value of 'clone_len'). So we get 2 extents items with ranges that
   overlap and an extent length of 4MB, larger then the extent A2 from
   inode A (1MB length);

5) After that we end the transaction, balance the btree dirty pages and
   then start another or join the previous transaction. It might happen
   that the transaction which inserted the incorrect extent was committed
   by another task so we end up with extent corruption if a power failure
   happens.

So fix this by making sure we attempt to insert the extent to clone at
the destination inode only if we are past dropping the sub-range that
corresponds to a hole.

Fixes: 690a5dbfc51315 ("Btrfs: fix ENOSPC errors, leading to transaction aborts, when cloning extents")
Signed-off-by: Filipe Manana <fdmanana@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

btrfs: Fix error messages in qgroup_rescan_init

[ Upstream commit 37d02592f11bb76e4ab1dcaa5b8a2a0715403207 ]

The branch of qgroup_rescan_init which is executed from the mount
path prints wrong errors messages. The textual print out in case
BTRFS_QGROUP_STATUS_FLAG_RESCAN/BTRFS_QGROUP_STATUS_FLAG_ON are not
set are transposed. Fix it by exchanging their place.

Signed-off-by: Nikolay Borisov <nborisov@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

powerpc: Ensure that swiotlb buffer is allocated from low memory

[ Upstream commit 8fabc623238e68b3ac63c0dd1657bf86c1fa33af ]

Some powerpc platforms (e.g. 85xx) limit DMA-able memory way below 4G.
If a system has more physical memory than this limit, the swiotlb
buffer is not addressable because it is allocated from memblock using
top-down mode.

Force memblock to bottom-up mode before calling swiotlb_init() to
ensure that the swiotlb buffer is DMA-able.

Reported-by: Christian Zigotzky <chzigotzky@xenosoft.de>
Signed-off-by: Mike Rapoport <rppt@linux.ibm.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20191204123524.22919-1-rppt@kernel.org
Signed-off-by: Sasha Levin <sashal@kernel.org>

pinctrl: pinmux: fix a possible null pointer in pinmux_can_be_used_for_gpio

[ Upstream commit 6ba2fd391ac58c1a26874f10c3054a1ea4aca2d0 ]

This commit adds a check on ops pointer to avoid a kernel panic when
ops->strict is used. Indeed, on some pinctrl driver (at least for
pinctrl-stmfx) the pinmux ops is not implemented. Let's assume than gpio
can be used in this case.

Fixes: 472a61e777fe ("pinctrl/gpio: Take MUX usage into account")
Signed-off-by: Alexandre Torgue <alexandre.torgue@st.com>
Link: https://lore.kernel.org/r/20191204144106.10876-1-alexandre.torgue@st.com
Signed-off-by: Linus Walleij <linus.walleij@linaro.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

cfg80211: fix double-free after changing network namespace

[ Upstream commit 56cb31e185adb61f930743a9b70e700a43625386 ]

If wdev->wext.keys was initialized it didn't get reset to NULL on
unregister (and it doesn't get set in cfg80211_init_wdev either), but
wdev is reused if unregister was triggered through
cfg80211_switch_netns.

The next unregister (for whatever reason) will try to free
wdev->wext.keys again.

Signed-off-by: Stefan Bühler <source@stbuehler.de>
Link: https://lore.kernel.org/r/20191126100543.782023-1-stefan.buehler@tik.uni-stuttgart.de
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

mac80211: fix TID field in monitor mode transmit

[ Upstream commit 753ffad3d6243303994227854d951ff5c70fa9e0 ]

Fix overwriting of the qos_ctrl.tid field for encrypted frames injected on
a monitor interface. While qos_ctrl.tid is not encrypted, it's used as an
input into the encryption algorithm so it's protected, and thus cannot be
modified after encryption. For injected frames, the encryption may already
have been done in userspace, so we cannot change any fields.

Before passing the frame to the driver, the qos_ctrl.tid field is updated
from skb->priority. Prior to dbd50a851c50 skb->priority was updated in
ieee80211_select_queue_80211(), but this function is no longer always
called.

Update skb->priority in ieee80211_monitor_start_xmit() so that the value
is stored, and when later code 'modifies' the TID it really sets it to
the same value as before, preserving the encryption.

Fixes: dbd50a851c50 ("mac80211: only allocate one queue when using iTXQs")
Signed-off-by: Fredrik Olofsson <fredrik.olofsson@anyfinetworks.com>
Link: https://lore.kernel.org/r/20191119133451.14711-1-fredrik.olofsson@anyfinetworks.com
[rewrite commit message based on our discussion]
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

clk: walk orphan list on clock provider registration

[ Upstream commit 66d9506440bb05289eb4867059e7b8c6ed209717 ]

So far, we walked the orphan list every time a new clock was registered
in CCF. This was fine since the clocks were only referenced by name.

Now that the clock can be referenced through DT, it is not enough:
* Controller A register first a reference clocks from controller B
  through DT.
* Controller B register all its clocks then register the provider.

Each time controller B registers a new clock, the orphan list is walked
but it can't match since the provider is registered yet. When the
provider is finally registered, the orphan list is not walked unless
another clock is registered afterward.

This can lead to situation where some clocks remain orphaned even if
the parent is available.

Walking the orphan list on provider registration solves the problem.

Reported-by: Jian Hu <jian.hu@amlogic.com>
Fixes: fc0c209c147f ("clk: Allow parents to be specified without string names")
Signed-off-by: Jerome Brunet <jbrunet@baylibre.com>
Link: https://lkml.kernel.org/r/20191203080805.104628-1-jbrunet@baylibre.com
Signed-off-by: Stephen Boyd <sboyd@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

bus: ti-sysc: Fix missing reset delay handling

[ Upstream commit e709ed70d122e94cb426b1e1f905829eae19a009 ]

We have dts property for "ti,sysc-delay-us", and we're using it, but the
wait after OCP softreset only happens if devices are probed in legacy mode.

Let's add a delay after writing the OCP softreset when specified.

Fixes: e0db94fe87da ("bus: ti-sysc: Make OCP reset work for sysstatus and sysconfig reset bits")
Cc: Keerthy <j-keerthy@ti.com>
Signed-off-by: Tony Lindgren <tony@atomide.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

pinctrl: aspeed-g6: Fix LPC/eSPI mux configuration

[ Upstream commit eb45f2110b036e4e35d3f3aaee1c2ccf49d92425 ]

Early revisions of the AST2600 datasheet are conflicted about the state
of the LPC/eSPI strapping bit (SCU510[6]). Conversations with ASPEED
determined that the reference pinmux configuration tables were in error
and the SCU documentation contained the correct configuration. Update
the driver to reflect the state described in the SCU documentation.

Fixes: 2eda1cdec49f ("pinctrl: aspeed: Add AST2600 pinmux support")
Signed-off-by: Andrew Jeffery <andrew@aj.id.au>
Link: https://lore.kernel.org/r/20191202050110.15340-1-andrew@aj.id.au
Signed-off-by: Linus Walleij <linus.walleij@linaro.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

ARM: imx_v6_v7_defconfig: Explicitly restore CONFIG_DEBUG_FS

[ Upstream commit 46db63abb79524209c15c683feccfba116746757 ]

This is currently off and that's not desirable: default imx config is
meant to be generally useful for development and debugging.

Running git bisect between v5.4 and v5.5-rc1 finds this started from
commit 0e4a459f56c3 ("tracing: Remove unnecessary DEBUG_FS dependency")

Explicit CONFIG_DEBUG_FS=y was earlier removed by
commit c29d541f590c ("ARM: imx_v6_v7_defconfig: Remove unneeded options")

A very similar fix was required before:
commit 7e9eb6268809 ("ARM: imx_v6_v7_defconfig: Explicitly restore CONFIG_DEBUG_FS")

Signed-off-by: Leonard Crestez <leonard.crestez@nxp.com>
Signed-off-by: Shawn Guo <shawnguo@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

arm64: dts: ls1028a: fix reboot node

[ Upstream commit 3f0fb37b22b460e3dec62bee284932881574acb9 ]

The reboot register isn't located inside the DCFG controller, but in its
own RST controller. Fix it.

Fixes: 8897f3255c9c ("arm64: dts: Add support for NXP LS1028A SoC")
Signed-off-by: Michael Walle <michael@walle.cc>
Acked-by: Li Yang <leoyang.li@nxp.com>
Signed-off-by: Shawn Guo <shawnguo@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

samples: bpf: fix syscall_tp due to unused syscall

[ Upstream commit fe3300897cbfd76c6cb825776e5ac0ca50a91ca4 ]

Currently, open() is called from the user program and it calls the syscall
'sys_openat', not the 'sys_open'. This leads to an error of the program
of user side, due to the fact that the counter maps are zero since no
function such 'sys_open' is called.

This commit adds the kernel bpf program which are attached to the
tracepoint 'sys_enter_openat' and 'sys_enter_openat'.

Fixes: 1da236b6be963 ("bpf: add a test case for syscalls/sys_{enter|exit}_* tracepoints")
Signed-off-by: Daniel T. Lee <danieltimlee@gmail.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

samples: bpf: Replace symbol compare of trace_event

[ Upstream commit bba1b2a890253528c45aa66cf856f289a215bfbc ]

Previously, when this sample is added, commit 1c47910ef8013
("samples/bpf: add perf_event+bpf example"), a symbol 'sys_read' and
'sys_write' has been used without no prefixes. But currently there are
no exact symbols with these under kallsyms and this leads to failure.

This commit changes exact compare to substring compare to keep compatible
with exact symbol or prefixed symbol.

Fixes: 1c47910ef8013 ("samples/bpf: add perf_event+bpf example")
Signed-off-by: Daniel T. Lee <danieltimlee@gmail.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Link: https://lore.kernel.org/bpf/20191205080114.19766-2-danieltimlee@gmail.com
Signed-off-by: Sasha Levin <sashal@kernel.org>

kselftest: Support old perl versions

[ Upstream commit 4eac734486fd431e0756cc5e929f140911a36a53 ]

On an old perl such as v5.10.1, `kselftest/prefix.pl` gives below error
message:

    Can't locate object method "autoflush" via package "IO::Handle" at kselftest/prefix.pl line 10.

This commit fixes the error by explicitly specifying the use of the
`IO::Handle` package.

Signed-off-by: SeongJae Park <sjpark@amazon.de>
Acked-by: Kees Cook <keescook@chromium.org>
Signed-off-by: Shuah Khan <skhan@linuxfoundation.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

kselftest/runner: Print new line in print of timeout log

[ Upstream commit d187801d1a46519d2a322f879f7c8f85c685372e ]

If a timeout failure occurs, kselftest kills the test process and prints
the timeout log.  If the test process has killed while printing a log
that ends with new line, the timeout log can be printed in middle of the
test process output so that it can be seems like a comment, as below:

    # test_process_log not ok 3 selftests: timers: nsleep-lat # TIMEOUT

This commit avoids such problem by printing one more line before the
TIMEOUT failure log.

Signed-off-by: SeongJae Park <sjpark@amazon.de>
Acked-by: Kees Cook <keescook@chromium.org>
Signed-off-by: Shuah Khan <skhan@linuxfoundation.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

ARM: dts: am437x-gp/epos-evm: fix panel compatible

[ Upstream commit c6b16761c6908d3dc167a0a566578b4b0b972905 ]

The LCD panel on AM4 GP EVMs and ePOS boards seems to be
osd070t1718-19ts. The current dts files say osd057T0559-34ts. Possibly
the panel has changed since the early EVMs, or there has been a mistake
with the panel type.

Update the DT files accordingly.

Acked-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Tomi Valkeinen <tomi.valkeinen@ti.com>
Signed-off-by: Tony Lindgren <tony@atomide.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

spi: spi-ti-qspi: Fix a bug when accessing non default CS

[ Upstream commit c52c91bb9aa6bd8c38dbf9776158e33038aedd43 ]

When switching ChipSelect from default CS0 to any other CS, driver fails
to update the bits in system control module register that control which
CS is mapped for MMIO access. This causes reads to fail when driver
tries to access QSPI flash on CS1/2/3.

Fix this by updating appropriate bits whenever active CS changes.

Reported-by: Andreas Dannenberg <dannenberg@ti.com>
Signed-off-by: Vignesh Raghavendra <vigneshr@ti.com>
Link: https://lore.kernel.org/r/20191211155216.30212-1-vigneshr@ti.com
Signed-off-by: Mark Brown <broonie@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

perf header: Fix false warning when there are no duplicate cache entries

[ Upstream commit 28707826877f84bce0977845ea529cbdd08e4e8d ]

Before this patch, perf expected that there might be NPROC*4 unique
cache entries at max, however, it also expected that some of them would
be shared and/or of the same size, thus the final number of entries
would be reduced to be lower than NPROC*4. In case the number of entries
hadn't been reduced (was NPROC*4), the warning was printed.

However, some systems might have unusual cache topology, such as the
following two-processor KVM guest:

cpu  level  shared_cpu_list  size
  0     1         0           32K
  0     1         0           64K
  0     2         0           512K
  0     3         0           8192K
  1     1         1           32K
  1     1         1           64K
  1     2         1           512K
  1     3         1           8192K

This KVM guest has 8 (NPROC*4) unique cache entries, which used to make
perf printing the message, although there actually aren't "way too many
cpu caches".

v2: Removing unused argument.

v3: Unifying the way we obtain number of cpus.

v4: Removed '& UINT_MAX' construct which is redundant.

Signed-off-by: Michael Petlan <mpetlan@redhat.com>
Acked-by: Jiri Olsa <jolsa@redhat.com>
LPU-Reference: 20191208162056.20772-1-mpetlan@redhat.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

perf metricgroup: Fix printing event names of metric group with multiple events

[ Upstream commit eb573e746b9d4f0921dcb2449be3df41dae3caea ]

Commit f01642e4912b ("perf metricgroup: Support multiple events for
metricgroup") introduced support for multiple events in a metric group.
But with the current upstream, metric events names are not printed
properly

In power9 platform:

command:# ./perf stat --metric-only -M translation -C 0 -I 1000 sleep 2
     1.000208486
     2.000368863
     2.001400558

Similarly in skylake platform:

command:./perf stat --metric-only -M Power -I 1000
     1.000579994
     2.002189493

With current upstream version, issue is with event name comparison logic
in find_evsel_group(). Current logic is to compare events belonging to a
metric group to the events in perf_evlist.  Since the break statement is
missing in the loop used for comparison between metric group and
perf_evlist events, the loop continues to execute even after getting a
pattern match, and end up in discarding the matches.

Incase of single metric event belongs to metric group, its working fine,
because in case of single event once it compare all events it reaches to
end of perf_evlist.

Example for single metric event in power9 platform:

command:# ./perf stat --metric-only  -M branches_per_inst -I 1000 sleep 1
     1.000094653                  0.2
     1.001337059                  0.0

This patch fixes the issue by making sure once we found all events
belongs to that metric event matched in find_evsel_group(), we
successfully break from that loop by adding corresponding condition.

With this patch:
In power9 platform:

command:# ./perf stat --metric-only -M translation -C 0 -I 1000 sleep 2
result:#
            time  derat_4k_miss_rate_percent  derat_4k_miss_ratio derat_miss_ratio derat_64k_miss_rate_percent  derat_64k_miss_ratio dslb_miss_rate_percent islb_miss_rate_percent
     1.000135672                         0.0                  0.3              1.0                         0.0                   0.2                    0.0                    0.0
     2.000380617                         0.0                  0.0              0.0                         0.0                   0.0                    0.0                    0.0

command:# ./perf stat --metric-only -M Power -I 1000

Similarly in skylake platform:
result:#
            time    Turbo_Utilization    C3_Core_Residency  C6_Core_Residency  C7_Core_Residency    C2_Pkg_Residency  C3_Pkg_Residency     C6_Pkg_Residency   C7_Pkg_Residency
     1.000563580                  0.3                  0.0                2.6               44.2                21.9               0.0                  0.0               0.0
     2.002235027                  0.4                  0.0                2.7               43.0                20.7               0.0                  0.0               0.0

Committer testing:

  Before:

  [root@seventh ~]# perf stat --metric-only -M Power -I 1000
  #           time
       1.000383223
       2.001168182
       3.001968545
       4.002741200
       5.003442022
  ^C     5.777687244

  [root@seventh ~]#

  After the patch:

  [root@seventh ~]# perf stat --metric-only -M Power -I 1000
  #           time    Turbo_Utilization    C3_Core_Residency    C6_Core_Residency    C7_Core_Residency     C2_Pkg_Residency     C3_Pkg_Residency     C6_Pkg_Residency     C7_Pkg_Residency
       1.000406577                  0.4                  0.1                  1.4                 97.0                  0.0                  0.0                  0.0                  0.0
       2.001481572                  0.3                  0.0                  0.6                 97.9                  0.0                  0.0                  0.0                  0.0
       3.002332585                  0.2                  0.0                  1.0                 97.5                  0.0                  0.0                  0.0                  0.0
       4.003196624                  0.2                  0.0                  0.3                 98.6                  0.0                  0.0                  0.0                  0.0
       5.004063851                  0.3                  0.0                  0.7                 97.7                  0.0                  0.0                  0.0                  0.0
  ^C     5.471260276                  0.2                  0.0                  0.5                 49.3                  0.0                  0.0                  0.0                  0.0

  [root@seventh ~]#
  [root@seventh ~]# dmesg | grep -i skylake
  [    0.187807] Performance Events: PEBS fmt3+, Skylake events, 32-deep LBR, full-width counters, Intel PMU driver.
  [root@seventh ~]#

Fixes: f01642e4912b ("perf metricgroup: Support multiple events for metricgroup")
Signed-off-by: Kajol Jain <kjain@linux.ibm.com>
Reviewed-by: Ravi Bangoria <ravi.bangoria@linux.ibm.com>
Tested-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: Anju T Sudhakar <anju@linux.vnet.ibm.com>
Cc: Jin Yao <yao.jin@linux.intel.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Kan Liang <kan.liang@linux.intel.com>
Cc: Madhavan Srinivasan <maddy@linux.vnet.ibm.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Link: http://lore.kernel.org/lkml/20191120084059.24458-1-kjain@linux.ibm.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

bpftool: Don't crash on missing jited insns or ksyms

[ Upstream commit 5b79bcdf03628a3a9ee04d9cd5fabcf61a8e20be ]

When the kptr_restrict sysctl is set, the kernel can fail to return
jited_ksyms or jited_prog_insns, but still have positive values in
nr_jited_ksyms and jited_prog_len. This causes bpftool to crash when
trying to dump the program because it only checks the len fields not
the actual pointers to the instructions and ksyms.

Fix this by adding the missing checks.

Fixes: 71bb428fe2c1 ("tools: bpf: add bpftool")
Fixes: f84192ee00b7 ("tools: bpftool: resolve calls without using imm field")
Signed-off-by: Toke Høiland-Jørgensen <toke@redhat.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Martin KaFai Lau <kafai@fb.com>
Link: https://lore.kernel.org/bpf/20191210181412.151226-1-toke@redhat.com
Signed-off-by: Sasha Levin <sashal@kernel.org>

bpf, mips: Limit to 33 tail calls

[ Upstream commit e49e6f6db04e915dccb494ae10fa14888fea6f89 ]

All BPF JIT compilers except RISC-V's and MIPS' enforce a 33-tail calls
limit at runtime.  In addition, a test was recently added, in tailcalls2,
to check this limit.

This patch updates the tail call limit in MIPS' JIT compiler to allow
33 tail calls.

Fixes: b6bd53f9c4e8 ("MIPS: Add missing file for eBPF JIT.")
Reported-by: Mahshid Khezri <khezri.mahshid@gmail.com>
Signed-off-by: Paul Chaignon <paul.chaignon@orange.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Martin KaFai Lau <kafai@fb.com>
Link: https://lore.kernel.org/bpf/b8eb2caac1c25453c539248e56ca22f74b5316af.1575916815.git.paul.chaignon@gmail.com
Signed-off-by: Sasha Levin <sashal@kernel.org>

bpf, riscv: Limit to 33 tail calls

[ Upstream commit 96bc4432f5ade1045521f3b247f516b1478166bd ]

All BPF JIT compilers except RISC-V's and MIPS' enforce a 33-tail calls
limit at runtime.  In addition, a test was recently added, in tailcalls2,
to check this limit.

This patch updates the tail call limit in RISC-V's JIT compiler to allow
33 tail calls.  I tested it using the above selftest on an emulated
RISCV64.

Fixes: 2353ecc6f91f ("bpf, riscv: add BPF JIT for RV64G")
Reported-by: Mahshid Khezri <khezri.mahshid@gmail.com>
Signed-off-by: Paul Chaignon <paul.chaignon@orange.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Björn Töpel <bjorn.topel@gmail.com>
Acked-by: Martin KaFai Lau <kafai@fb.com>
Link: https://lore.kernel.org/bpf/966fe384383bf23a0ee1efe8d7291c78a3fb832b.1575916815.git.paul.chaignon@gmail.com
Signed-off-by: Sasha Levin <sashal@kernel.org>

arm64: dts: ls1028a: fix typo in TMU calibration data

[ Upstream commit 961f8209c8d5ef5d33da42e6656d7c8179899da0 ]

The temperature sensor may jump backwards because there is a wrong
calibration value. Both values have to be monotonically increasing.
Fix it.

This was tested on a custom board.

Fixes: 571cebfe8e2b ("arm64: dts: ls1028a: Add Thermal Monitor Unit node")
Signed-off-by: Michael Walle <michael@walle.cc>
Acked-by: Tang Yuantian <andy.tang@nxp.com>
Signed-off-by: Shawn Guo <shawnguo@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

ARM: dts: bcm283x: Fix critical trip point

[ Upstream commit 30e647a764d446723a7e0fb08d209e0104f16173 ]

During definition of the CPU thermal zone of BCM283x SoC family there
was a misunderstanding of the meaning "criticial trip point" and the
thermal throttling range of the VideoCore firmware. The latter one takes
effect when the core temperature is at least 85 degree celsius or higher

So the current critical trip point doesn't make sense, because the
thermal shutdown appears before the firmware has a chance to throttle
the ARM core(s).

Fix these unwanted shutdowns by increasing the critical trip point
to a value which shouldn't be reached with working thermal throttling.

Fixes: 0fe4d2181cc4 ("ARM: dts: bcm283x: Add CPU thermal zone with 1 trip point")
Signed-off-by: Stefan Wahren <wahrenst@gmx.net>
Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

ARM: omap2plus_defconfig: Add back DEBUG_FS

[ Upstream commit e00b59d30506dc9ef91caf2f3c584209cc9f61e4 ]

Commit 0e4a459f56c3 ("tracing: Remove unnecessary DEBUG_FS dependency")
removed select for DEBUG_FS but we still need it at least for enabling
deeper idle states for the SoCs.

Signed-off-by: Tony Lindgren <tony@atomide.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

ARM: dts: am335x-sancloud-bbe: fix phy mode

[ Upstream commit c842b8c4ff9859f750447f3ca08f64b2ed23cebc ]

The phy mode should be rgmii-id.  For some reason, it used to work with
rgmii-txid but doesn't any more.

Signed-off-by: Mans Rullgard <mans@mansr.com>
Signed-off-by: Tony Lindgren <tony@atomide.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

ASoC: SOF: Intel: split cht and byt debug window sizes

[ Upstream commit f84337c3fb8ff4d533ccbed0d2db4e8587d0ff58 ]

Turns out SSP 3-5 are only available on cht, to avoid dumping on
undefined registers let's split the definition.

Signed-off-by: Curtis Malainey <cujomalainey@chromium.org>
Signed-off-by: Pierre-Louis Bossart <pierre-louis.bossart@linux.intel.com>
Link: https://lore.kernel.org/r/20191210004854.16845-7-pierre-louis.bossart@linux.intel.com
Signed-off-by: Mark Brown <broonie@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

ASoC: SOF: loader: snd_sof_fw_parse_ext_data log warning on unknown header

[ Upstream commit 8edc95667646a75f0fc97e08ecb180581fdff300 ]

Added warning log when found some unknown FW boot ext header,
to improve debuggability.

Signed-off-by: Karol Trzcinski <karolx.trzcinski@linux.intel.com>
Signed-off-by: Pierre-Louis Bossart <pierre-louis.bossart@linux.intel.com>
Link: https://lore.kernel.org/r/20191210004854.16845-3-pierre-louis.bossart@linux.intel.com
Signed-off-by: Mark Brown <broonie@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

ASoC: topology: Check return value for soc_tplg_pcm_create()

[ Upstream commit a3039aef52d9ffeb67e9211899cd3e8a2953a01f ]

The return value of soc_tplg_pcm_create() is currently not checked
in soc_tplg_pcm_elems_load(). If an error is to occur there, the
topology ignores it and continues loading.

Fix that by checking the status and rejecting the topology on error.

Reviewed-by: Ranjani Sridharan <ranjani.sridharan@linux.intel.com>
Signed-off-by: Dragos Tarcatu <dragos_tarcatu@mentor.com>
Signed-off-by: Pierre-Louis Bossart <pierre-louis.bossart@linux.intel.com>
Link: https://lore.kernel.org/r/20191210003939.15752-3-pierre-louis.bossart@linux.intel.com
Signed-off-by: Mark Brown <broonie@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

ASoC: topology: Check return value for snd_soc_add_dai_link()

[ Upstream commit 76d2703649321c296df7ec0dafd50add96215de4 ]

snd_soc_add_dai_link() might fail. This situation occurs for
instance in a very specific use case where a PCM device and a
Back End DAI link are given identical names in the topology.
When this happens, soc_new_pcm_runtime() fails and then
snd_soc_add_dai_link() returns -ENOMEM when called from
soc_tplg_fe_link_create(). Because of that, the link will not
get added into the card list, so any attempt to remove it later
ends up in a panic.

Fix that by checking the return status and free the memory in case
of an error.

Reviewed-by: Ranjani Sridharan <ranjani.sridharan@linux.intel.com>
Signed-off-by: Dragos Tarcatu <dragos_tarcatu@mentor.com>
Signed-off-by: Pierre-Louis Bossart <pierre-louis.bossart@linux.intel.com>
Link: https://lore.kernel.org/r/20191210003939.15752-2-pierre-louis.bossart@linux.intel.com
Signed-off-by: Mark Brown <broonie@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

reset: Do not register resource data for missing resets

[ Upstream commit db23808615e29d9a04f96806cac56f78b0fee0ef ]

When an optional reset is not present, __devm_reset_control_get() and
devm_reset_control_array_get() still register resource data to release
the non-existing reset on cleanup, which is futile.

Fix this by skipping NULL reset control pointers.

Signed-off-by: Geert Uytterhoeven <geert+renesas@glider.be>
Signed-off-by: Philipp Zabel <p.zabel@pengutronix.de>
Signed-off-by: Sasha Levin <sashal@kernel.org>

spi: spi-cavium-thunderx: Add missing pci_release_regions()

[ Upstream commit a841e2853e1afecc2ee692b8cc5bff606bc84e4c ]

The driver forgets to call pci_release_regions() in probe failure
and remove.
Add the missed calls to fix it.

Signed-off-by: Chuhong Yuan <hslester96@gmail.com>
Link: https://lore.kernel.org/r/20191206075500.18525-1-hslester96@gmail.com
Signed-off-by: Mark Brown <broonie@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

ARM: dts: Cygnus: Fix MDIO node address/size cells

[ Upstream commit fac2c2da3596d77c343988bb0d41a8c533b2e73c ]

The MDIO node on Cygnus had an reversed #address-cells and
 #size-cells properties, correct those.

Fixes: 40c26d3af60a ("ARM: dts: Cygnus: Add the ethernet switch and ethernet PHY")
Reported-by: Simon Horman <simon.horman@netronome.com>
Reviewed-by: Ray Jui <ray.jui@broadcom.com>
Reviewed-by: Simon Horman <simon.horman@netronome.com>
Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

ARM: exynos_defconfig: Restore debugfs support

[ Upstream commit a2315d3aea5976acd919d3d3fcf82f752562c25b ]

Commit 9f532d26c75c ("ARM: exynos_defconfig: Trim and reorganize with
savedefconfig") removed explicit enable line for CONFIG_DEBUG_FS, because
that feature has been selected by other enabled options: CONFIG_TRACING,
which in turn had been selected by CONFIG_PERF_EVENTS and
CONFIG_PROVE_LOCKING.

In meantime, commit 0e4a459f56c3 ("tracing: Remove unnecessary DEBUG_FS
dependency") removed the dependency between CONFIG_DEBUG_FS and
CONFIG_TRACING, so CONFIG_DEBUG_FS is no longer enabled in default builds.

Enable it again explicitly, as debugfs support is essential for various
automated testing tools.

Signed-off-by: Marek Szyprowski <m.szyprowski@samsung.com>
Signed-off-by: Krzysztof Kozlowski <krzk@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

selftests: safesetid: Fix Makefile to set correct test program

[ Upstream commit 8ef1ec0ca32c6f8a87f5b4c24b1db26da67c5609 ]

Fix Makefile to set safesetid-test.sh to TEST_PROGS instead
of non existing run_tests.sh.

Without this fix, I got following error.
  ----
  TAP version 13
  1..1
  # selftests: safesetid: run_tests.sh
  # Warning: file run_tests.sh is missing!
  not ok 1 selftests: safesetid: run_tests.sh
  ----

Fixes: c67e8ec03f3f ("LSM: SafeSetID: add selftest")
Signed-off-by: Masami Hiramatsu <mhiramat@kernel.org>
Signed-off-by: Shuah Khan <skhan@linuxfoundation.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

selftests: safesetid: Check the return value of setuid/setgid

[ Upstream commit 295c4e21cf27ac9af542140e3e797df9e0cf7b5f ]

Check the return value of setuid() and setgid().
This fixes the following warnings and improves test result.

safesetid-test.c: In function ‘main’:
safesetid-test.c:294:2: warning: ignoring return value of ‘setuid’, declared with attribute warn_unused_result [-Wunused-result]
  setuid(NO_POLICY_USER);
  ^~~~~~~~~~~~~~~~~~~~~~
safesetid-test.c:295:2: warning: ignoring return value of ‘setgid’, declared with attribute warn_unused_result [-Wunused-result]
  setgid(NO_POLICY_USER);
  ^~~~~~~~~~~~~~~~~~~~~~
safesetid-test.c:309:2: warning: ignoring return value of ‘setuid’, declared with attribute warn_unused_result [-Wunused-result]
  setuid(RESTRICTED_PARENT);
  ^~~~~~~~~~~~~~~~~~~~~~~~~
safesetid-test.c:310:2: warning: ignoring return value of ‘setgid’, declared with attribute warn_unused_result [-Wunused-result]
  setgid(RESTRICTED_PARENT);
  ^~~~~~~~~~~~~~~~~~~~~~~~~
safesetid-test.c: In function ‘test_setuid’:
safesetid-test.c:216:3: warning: ignoring return value of ‘setuid’, declared with attribute warn_unused_result [-Wunused-result]
   setuid(child_uid);
   ^~~~~~~~~~~~~~~~~

Fixes: c67e8ec03f3f ("LSM: SafeSetID: add selftest")
Signed-off-by: Masami Hiramatsu <mhiramat@kernel.org>
Signed-off-by: Shuah Khan <skhan@linuxfoundation.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

selftests: safesetid: Move link library to LDLIBS

[ Upstream commit be12252212fa3dfed6e75112865095c484c0ce87 ]

Move -lcap to LDLIBS from CFLAGS because it is a library
to be linked.

Without this, safesetid failed to build with link error
as below.

----
/usr/bin/ld: /tmp/ccL8rZHT.o: in function `drop_caps':
safesetid-test.c:(.text+0xe7): undefined reference to `cap_get_proc'
/usr/bin/ld: safesetid-test.c:(.text+0x107): undefined reference to `cap_set_flag'
/usr/bin/ld: safesetid-test.c:(.text+0x10f): undefined reference to `cap_set_proc'
/usr/bin/ld: safesetid-test.c:(.text+0x117): undefined reference to `cap_free'
/usr/bin/ld: safesetid-test.c:(.text+0x136): undefined reference to `cap_clear'
collect2: error: ld returned 1 exit status
----

Fixes: c67e8ec03f3f ("LSM: SafeSetID: add selftest")
Signed-off-by: Masami Hiramatsu <mhiramat@kernel.org>
Signed-off-by: Shuah Khan <skhan@linuxfoundation.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

selftests/ftrace: Fix multiple kprobe testcase

[ Upstream commit 5cc6c8d4a99d0ee4d5466498e258e593df1d3eb6 ]

Fix multiple kprobe event testcase to work it correctly.
There are 2 bugfixes.
 - Since `wc -l FILE` returns not only line number but also
   FILE filename, following "if" statement always failed.
   Fix this bug by replacing it with 'cat FILE | wc -l'
 - Since "while do-done loop" block with pipeline becomes a
   subshell, $N local variable is not update outside of
   the loop.
   Fix this bug by using actual target number (256) instead
   of $N.

Signed-off-by: Masami Hiramatsu <mhiramat@kernel.org>
Reviewed-by: Steven Rostedt (VMware) <rostedt@goodmis.org>
Signed-off-by: Shuah Khan <skhan@linuxfoundation.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

selftests/ftrace: Do not to use absolute debugfs path

[ Upstream commit ba1b9c5048e43716921abe3a1db19cebebf4a5f5 ]

Use relative path to trigger file instead of absolute debugfs path,
because if the user uses tracefs instead of debugfs, it can be
mounted at /sys/kernel/tracing.
Anyway, since the ftracetest is designed to be run at the tracing
directory, user doesn't need to use absolute path.

Signed-off-by: Masami Hiramatsu <mhiramat@kernel.org>
Reviewed-by: Steven Rostedt (VMware) <rostedt@goodmis.org>
Signed-off-by: Shuah Khan <skhan@linuxfoundation.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

selftests/ftrace: Fix ftrace test cases to check unsupported

[ Upstream commit 25deae098e748d8d36bc35129a66734b8f6925c9 ]

Since dynamic function tracer can be disabled, set_ftrace_filter
can be disappeared. Test cases which depends on it, must check
whether the set_ftrace_filter exists or not before testing
and if not, return as unsupported.

Also, if the function tracer itself is disabled, we can not
set "function" to current_tracer. Test cases must check it
before testing, and return as unsupported.

Signed-off-by: Masami Hiramatsu <mhiramat@kernel.org>
Reviewed-by: Steven Rostedt (VMware) <rostedt@goodmis.org>
Signed-off-by: Shuah Khan <skhan@linuxfoundation.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

selftests/ftrace: Fix to check the existence of set_ftrace_filter

[ Upstream commit fd1baf6ca2ea3550ea47f2bb0bdcf34ec764a779 ]

If we run ftracetest on the kernel with CONFIG_DYNAMIC_FTRACE=n,
there is no set_ftrace_filter and all test cases are failed, because
reset_ftrace_filter() returns an error.
Let's check whether set_ftrace_filter exists in reset_ftrace_filter()
and clean up only set_ftrace_notrace in initialize_ftrace().

Signed-off-by: Masami Hiramatsu <mhiramat@kernel.org>
Reviewed-by: Steven Rostedt (VMware) <rostedt@goodmis.org>
Signed-off-by: Shuah Khan <skhan@linuxfoundation.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

ARM: dts: BCM5301X: Fix MDIO node address/size cells

[ Upstream commit 093c3f94e922d83a734fc4da08cc5814990f32c6 ]

The MDIO node on BCM5301X had an reversed #address-cells and
 #size-cells properties, correct those, silencing checker warnings:

.../linux/arch/arm/boot/dts/bcm4708-asus-rt-ac56u.dt.yaml: mdio@18003000: #address-cells:0:0: 1 was expected

Reported-by: Simon Horman <simon.horman@netronome.com>
Fixes: 23f1eca6d59b ("ARM: dts: BCM5301X: Specify MDIO bus in the DT")
Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

netfilter: nf_tables_offload: return EOPNOTSUPP if rule specifies no actions

[ Upstream commit 81ec61074bcf68acfcb2820cda3ff9d9984419c7 ]

If the rule only specifies the matching side, return EOPNOTSUPP.
Otherwise, the front-end relies on the drivers to reject this rule.

Fixes: c9626a2cbdb2 ("netfilter: nf_tables: add hardware offload support")
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

netfilter: nf_tables: skip module reference count bump on object updates

[ Upstream commit fd57d0cbe187e93f63777d36e9f49293311d417f ]

Use __nft_obj_type_get() instead, otherwise there is a module reference
counter leak.

Fixes: d62d0ba97b58 ("netfilter: nf_tables: Introduce stateful object update operation")
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

netfilter: nf_tables: validate NFT_DATA_VALUE after nft_data_init()

[ Upstream commit 0d2c96af797ba149e559c5875c0151384ab6dd14 ]

Userspace might bogusly sent NFT_DATA_VERDICT in several netlink
attributes that assume NFT_DATA_VALUE. Moreover, make sure that error
path invokes nft_data_release() to decrement the reference count on the
chain object.

Fixes: 96518518cc41 ("netfilter: add nftables")
Fixes: 0f3cd9b36977 ("netfilter: nf_tables: add range expression")
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

netfilter: nf_tables: validate NFT_SET_ELEM_INTERVAL_END

[ Upstream commit bffc124b6fe37d0ae9b428d104efb426403bb5c9 ]

Only NFTA_SET_ELEM_KEY and NFTA_SET_ELEM_FLAGS make sense for elements
whose NFT_SET_ELEM_INTERVAL_END flag is set on.

Fixes: 96518518cc41 ("netfilter: add nftables")
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

netfilter: nft_set_rbtree: bogus lookup/get on consecutive elements in named sets

[ Upstream commit db3b665dd77b34e34df00e17d7b299c98fcfb2c5 ]

The existing rbtree implementation might store consecutive elements
where the closing element and the opening element might overlap, eg.

[ a, a+1) [ a+1, a+2)

This patch removes the optimization for non-anonymous sets in the exact
matching case, where it is assumed to stop searching in case that the
closing element is found. Instead, invalidate candidate interval and
keep looking further in the tree.

The lookup/get operation might return false, while there is an element
in the rbtree. Moreover, the get operation returns true as if a+2 would
be in the tree. This happens with named sets after several set updates.

The existing lookup optimization (that only works for the anonymous
sets) might not reach the opening [ a+1,... element if the closing
...,a+1) is found in first place when walking over the rbtree. Hence,
walking the full tree in that case is needed.

This patch fixes the lookup and get operations.

Fixes: e701001e7cbe ("netfilter: nft_rbtree: allow adjacent intervals with dynamic updates")
Fixes: ba0e4d9917b4 ("netfilter: nf_tables: get set elements via netlink")
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

netfilter: uapi: Avoid undefined left-shift in xt_sctp.h

[ Upstream commit 164166558aacea01b99c8c8ffb710d930405ba69 ]

With 'bytes(__u32)' being 32, a left-shift of 31 may happen which is
undefined for the signed 32-bit value 1. Avoid this by declaring 1 as
unsigned.

Signed-off-by: Phil Sutter <phil@nwl.cc>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

ARM: vexpress: Set-up shared OPP table instead of individual for each CPU

[ Upstream commit 2a76352ad2cc6b78e58f737714879cc860903802 ]

Currently we add individual copy of same OPP table for each CPU within
the cluster. This is redundant and doesn't reflect the reality.

We can't use core cpumask to set policy->cpus in ve_spc_cpufreq_init()
anymore as it gets called via cpuhp_cpufreq_online()->cpufreq_online()
->cpufreq_driver->init() and the cpumask gets updated upon CPU hotplug
operations. It also may cause issues when the vexpress_spc_cpufreq
driver is built as a module.

Since ve_spc_clk_init is built-in device initcall, we should be able to
use the same topology_core_cpumask to set the opp sharing cpumask via
dev_pm_opp_set_sharing_cpus and use the same later in the driver via
dev_pm_opp_get_sharing_cpus.

Cc: Liviu Dudau <liviu.dudau@arm.com>
Cc: Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>
Acked-by: Viresh Kumar <viresh.kumar@linaro.org>
Tested-by: Dietmar Eggemann <dietmar.eggemann@arm.com>
Signed-off-by: Sudeep Holla <sudeep.holla@arm.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

ARM: dts: imx6ul: imx6ul-14x14-evk.dtsi: Fix SPI NOR probing

[ Upstream commit 0aeb1f2b74f3402e9cdb7c0b8e2c369c9767301e ]

Without this "jedec,spi-nor" compatible property, probing of the SPI NOR
does not work on the NXP i.MX6ULL EVK. Fix this by adding this
compatible property to the DT.

Fixes: 7d77b8505aa9 ("ARM: dts: imx6ull: fix the imx6ull-14x14-evk configuration")
Signed-off-by: Stefan Roese <sr@denx.de>
Reviewed-by: Fabio Estevam <festevam@gmail.com>
Reviewed-by: Frieder Schrempf <frieder.schrempf@kontron.de>
Signed-off-by: Shawn Guo <shawnguo@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

efi/earlycon: Remap entire framebuffer after page initialization

[ Upstream commit b418d660bb9798d2249ac6a46c844389ef50b6a5 ]

When commit:

  69c1f396f25b ("efi/x86: Convert x86 EFI earlyprintk into generic earlycon implementation")

moved the x86 specific EFI earlyprintk implementation to a shared location,
it also tweaked the behaviour. In particular, it dropped a trick with full
framebuffer remapping after page initialization, leading to two regressions:

  1) very slow scrolling after page initialization,
  2) kernel hang when the 'keep_bootcon' command line argument is passed.

Putting the tweak back fixes #2 and mitigates #1, i.e., it limits the slow
behavior to the early boot stages, presumably due to eliminating heavy
map()/unmap() operations per each pixel line on the screen.

 [ ardb: ensure efifb is unmapped again unless keep_bootcon is in effect. ]
 [ mingo: speling fixes. ]

Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Signed-off-by: Ard Biesheuvel <ardb@kernel.org>
Cc: Arvind Sankar <nivedita@alum.mit.edu>
Cc: Bhupesh Sharma <bhsharma@redhat.com>
Cc: Masayoshi Mizuma <m.mizuma@jp.fujitsu.com>
Cc: linux-efi@vger.kernel.org
Fixes: 69c1f396f25b ("efi/x86: Convert x86 EFI earlyprintk into generic earlycon implementation")
Link: https://lkml.kernel.org/r/20191206165542.31469-7-ardb@kernel.org
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

efi/gop: Fix memory leak in __gop_query32/64()

[ Upstream commit ff397be685e410a59c34b21ce0c55d4daa466bb7 ]

efi_graphics_output_protocol::query_mode() returns info in
callee-allocated memory which must be freed by the caller, which
we aren't doing.

We don't actually need to call query_mode() in order to obtain the
info for the current graphics mode, which is already there in
gop->mode->info, so just access it directly in the setup_gop32/64()
functions.

Also nothing uses the size of the info structure, so don't update the
passed-in size (which is the size of the gop_handle table in bytes)
unnecessarily.

Signed-off-by: Arvind Sankar <nivedita@alum.mit.edu>
Signed-off-by: Ard Biesheuvel <ardb@kernel.org>
Cc: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Cc: Bhupesh Sharma <bhsharma@redhat.com>
Cc: Masayoshi Mizuma <m.mizuma@jp.fujitsu.com>
Cc: linux-efi@vger.kernel.org
Link: https://lkml.kernel.org/r/20191206165542.31469-5-ardb@kernel.org
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

efi/gop: Return EFI_SUCCESS if a usable GOP was found

[ Upstream commit dbd89c303b4420f6cdb689fd398349fc83b059dd ]

If we've found a usable instance of the Graphics Output Protocol
(GOP) with a framebuffer, it is possible that one of the later EFI
calls fails while checking if any support console output. In this
case status may be an EFI error code even though we found a usable
GOP.

Fix this by explicitly return EFI_SUCCESS if a usable GOP has been
located.

Signed-off-by: Arvind Sankar <nivedita@alum.mit.edu>
Signed-off-by: Ard Biesheuvel <ardb@kernel.org>
Cc: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Cc: Bhupesh Sharma <bhsharma@redhat.com>
Cc: Masayoshi Mizuma <m.mizuma@jp.fujitsu.com>
Cc: linux-efi@vger.kernel.org
Link: https://lkml.kernel.org/r/20191206165542.31469-4-ardb@kernel.org
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

efi/gop: Return EFI_NOT_FOUND if there are no usable GOPs

[ Upstream commit 6fc3cec30dfeee7d3c5db8154016aff9d65503c5 ]

If we don't find a usable instance of the Graphics Output Protocol
(GOP) because none of them have a framebuffer (i.e. they were all
PIXEL_BLT_ONLY), but all the EFI calls succeeded, we will return
EFI_SUCCESS even though we didn't find a usable GOP.

Fix this by explicitly returning EFI_NOT_FOUND if no usable GOPs are
found, allowing the caller to probe for UGA instead.

Signed-off-by: Arvind Sankar <nivedita@alum.mit.edu>
Signed-off-by: Ard Biesheuvel <ardb@kernel.org>
Cc: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Cc: Bhupesh Sharma <bhsharma@redhat.com>
Cc: Masayoshi Mizuma <m.mizuma@jp.fujitsu.com>
Cc: linux-efi@vger.kernel.org
Link: https://lkml.kernel.org/r/20191206165542.31469-3-ardb@kernel.org
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

selftests: netfilter: use randomized netns names

[ Upstream commit 5a2e6af81807d4616f9839ad0ae7d1313b45c64d ]

Using ns0, ns1, etc. isn't a good idea, they might exist already.
Use a random suffix.

Also, older nft versions don't support "-" as alias for stdin, so
use /dev/stdin instead.

Signed-off-by: Florian Westphal <fw@strlen.de>
Acked-by: Phil Sutter <phil@nwl.cc>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

ASoC: Intel: bytcr_rt5640: Update quirk for Teclast X89

[ Upstream commit 7eccc05c7101f34cc36afe9405d15de6d4099fb4 ]

When the Teclast X89 quirk was added we did not have jack-detection
support yet.

Note the over-current detection limit is set to 2mA instead of the usual
1.5mA because this tablet tends to give false-positive button-presses
when it is set to 1.5mA.

Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Acked-by: Pierre-Louis Bossart <pierre-louis.bossart@linux.intel.com>
Link: https://lore.kernel.org/r/20191203221442.2657-1-hdegoede@redhat.com
Signed-off-by: Mark Brown <broonie@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

x86/efi: Update e820 with reserved EFI boot services data to fix kexec breakage

[ Upstream commit af164898482817a1d487964b68f3c21bae7a1beb ]

Michael Weiser reported that he got this error during a kexec rebooting:

  esrt: Unsupported ESRT version 2904149718861218184.

The ESRT memory stays in EFI boot services data, and it was reserved
in kernel via efi_mem_reserve().  The initial purpose of the reservation
is to reuse the EFI boot services data across kexec reboot. For example
the BGRT image data and some ESRT memory like Michael reported.

But although the memory is reserved it is not updated in the X86 E820 table,
and kexec_file_load() iterates system RAM in the IO resource list to find places
for kernel, initramfs and other stuff. In Michael's case the kexec loaded
initramfs overwrote the ESRT memory and then the failure happened.

Since kexec_file_load() depends on the E820 table being updated, just fix this
by updating the reserved EFI boot services memory as reserved type in E820.

Originally any memory descriptors with EFI_MEMORY_RUNTIME attribute are
bypassed in the reservation code path because they are assumed as reserved.

But the reservation is still needed for multiple kexec reboots,
and it is the only possible case we come here thus just drop the code
chunk, then everything works without side effects.

On my machine the ESRT memory sits in an EFI runtime data range, it does
not trigger the problem, but I successfully tested with BGRT instead.
both kexec_load() and kexec_file_load() work and kdump works as well.

[ mingo: Edited the changelog. ]

Reported-by: Michael Weiser <michael@weiser.dinsnail.net>
Tested-by: Michael Weiser <michael@weiser.dinsnail.net>
Signed-off-by: Dave Young <dyoung@redhat.com>
Cc: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Eric W. Biederman <ebiederm@xmission.com>
Cc: H. Peter Anvin <hpa@zytor.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: kexec@lists.infradead.org
Cc: linux-efi@vger.kernel.org
Link: https://lkml.kernel.org/r/20191204075233.GA10520@dhcp-128-65.nay.redhat.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

regulator: core: fix regulator_register() error paths to properly release rdev

[ Upstream commit a3cde9534ebdafe18a9bbab208df724c57e6c8e8 ]

There are several issues with the error handling code of
the regulator_register() function:
        ret = device_register(&rdev->dev);
        if (ret != 0) {
                put_device(&rdev->dev); --> rdev released
                goto unset_supplies;
        }
...
unset_supplies:
...
        unset_regulator_supplies(rdev); --> use-after-free
...
clean:
        if (dangling_of_gpiod)
                gpiod_put(config->ena_gpiod);
        kfree(rdev);                     --> double free

We add a variable to record the failure of device_register() and
move put_device() down a bit to avoid the above issues.

Fixes: c438b9d01736 ("regulator: core: Move registration of regulator device")
Signed-off-by: Wen Yang <wenyang@linux.alibaba.com>
Cc: Liam Girdwood <lgirdwood@gmail.com>
Cc: Mark Brown <broonie@kernel.org>
Cc: linux-kernel@vger.kernel.org
Link: https://lore.kernel.org/r/20191201030250.38074-1-wenyang@linux.alibaba.com
Signed-off-by: Mark Brown <broonie@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

libtraceevent: Copy pkg-config file to output folder when using O=

[ Upstream commit 15b3904f8e884e0d34d5f09906cf6526d0b889a2 ]

When we use 'O=' with make to build libtraceevent in a separate folder
it still copies 'libtraceevent.pc' to its source folder. Modify the
Makefile so that it uses the output folder to copy the pkg-config file
and install from there.

Signed-off-by: Sudipm Mukherjee <sudipm.mukherjee@gmail.com>
Reviewed-by: Steven Rostedt (VMware) <rostedt@goodmis.org>
Cc: linux-trace-devel@vger.kernel.org
Link: http://lore.kernel.org/lkml/20191115113610.21493-2-sudipm.mukherjee@gmail.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

libtraceevent: Fix lib installation with O=

[ Upstream commit 587db8ebdac2c5eb3a8851e16b26f2e2711ab797 ]

When we use 'O=' with make to build libtraceevent in a separate folder
it fails to install libtraceevent.a and libtraceevent.so.1.1.0 with the
error:

  INSTALL  /home/sudip/linux/obj-trace/libtraceevent.a
  INSTALL  /home/sudip/linux/obj-trace/libtraceevent.so.1.1.0

  cp: cannot stat 'libtraceevent.a': No such file or directory
  Makefile:225: recipe for target 'install_lib' failed
  make: *** [install_lib] Error 1

I used the command:

  make O=../../../obj-trace DESTDIR=~/test prefix==/usr  install

It turns out libtraceevent Makefile, even though it builds in a separate
folder, searches for libtraceevent.a and libtraceevent.so.1.1.0 in its
source folder.

So, add the 'OUTPUT' prefix to the source path so that 'make' looks for
the files in the correct place.

Signed-off-by: Sudipm Mukherjee <sudipm.mukherjee@gmail.com>
Reviewed-by: Steven Rostedt (VMware) <rostedt@goodmis.org>
Cc: linux-trace-devel@vger.kernel.org
Link: http://lore.kernel.org/lkml/20191115113610.21493-1-sudipm.mukherjee@gmail.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

mwifiex: Fix heap overflow in mmwifiex_process_tdls_action_frame()

[ Upstream commit 1e58252e334dc3f3756f424a157d1b7484464c40 ]

mwifiex_process_tdls_action_frame() without checking
the incoming tdls infomation element's vality before use it,
this may cause multi heap buffer overflows.

Fix them by putting vality check before use it.

IE is TLV struct, but ht_cap and  ht_oper aren’t TLV struct.
the origin marvell driver code is wrong:

memcpy(&sta_ptr->tdls_cap.ht_oper, pos,....
memcpy((u8 *)&sta_ptr->tdls_cap.ht_capb, pos,...

Fix the bug by changing pos(the address of IE) to
pos+2 ( the address of IE value ).

Signed-off-by: qize wang <wangqize888888888@gmail.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

netfilter: nf_tables_offload: Check for the NETDEV_UNREGISTER event

[ Upstream commit d1f4c966475c6dd2545c6625022cb24e878bee11 ]

Check for the NETDEV_UNREGISTER event from the nft_offload_netdev_event
function, which is the event that actually triggers the clean up.

Fixes: 06d392cbe3db ("netfilter: nf_tables_offload: remove rules when the device unregisters")
Signed-off-by: wenxu <wenxu@ucloud.cn>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

x86/intel: Disable HPET on Intel Ice Lake platforms

[ Upstream commit e0748539e3d594dd26f0d27a270f14720b22a406 ]

Like CFL and CFL-H, ICL SoC has skewed HPET timer once it hits PC10.
So let's disable HPET on ICL.

Signed-off-by: Kai-Heng Feng <kai.heng.feng@canonical.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: bp@alien8.de
Cc: feng.tang@intel.com
Cc: harry.pan@intel.com
Cc: hpa@zytor.com
Link: https://lkml.kernel.org/r/20191129062303.18982-2-kai.heng.feng@canonical.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

netfilter: ctnetlink: netns exit must wait for callbacks

[ Upstream commit 18a110b022a5c02e7dc9f6109d0bd93e58ac6ebb ]

Curtis Taylor and Jon Maxwell reported and debugged a crash on 3.10
based kernel.

Crash occurs in ctnetlink_conntrack_events because net->nfnl socket is
NULL.  The nfnl socket was set to NULL by netns destruction running on
another cpu.

The exiting network namespace calls the relevant destructors in the
following order:

1. ctnetlink_net_exit_batch

This nulls out the event callback pointer in struct netns.

2. nfnetlink_net_exit_batch

This nulls net->nfnl socket and frees it.

3. nf_conntrack_cleanup_net_list

This removes all remaining conntrack entries.

This is order is correct. The only explanation for the crash so ar is:

cpu1: conntrack is dying, eviction occurs:
 -> nf_ct_delete()
   -> nf_conntrack_event_report \
     -> nf_conntrack_eventmask_report
       -> notify->fcn() (== ctnetlink_conntrack_events).

cpu1: a. fetches rcu protected pointer to obtain ctnetlink event callback.
      b. gets interrupted.
 cpu2: runs netns exit handlers:
     a runs ctnetlink destructor, event cb pointer set to NULL.
     b runs nfnetlink destructor, nfnl socket is closed and set to NULL.
cpu1: c. resumes and trips over NULL net->nfnl.

Problem appears to be that ctnetlink_net_exit_batch only prevents future
callers of nf_conntrack_eventmask_report() from obtaining the callback.
It doesn't wait of other cpus that might have already obtained the
callbacks address.

I don't see anything in upstream kernels that would prevent similar
crash: We need to wait for all cpus to have exited the event callback.

Fixes: 9592a5c01e79dbc59eb56fa ("netfilter: ctnetlink: netns support")
Signed-off-by: Florian Westphal <fw@strlen.de>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

locking/spinlock/debug: Fix various data races

[ Upstream commit 1a365e822372ba24c9da0822bc583894f6f3d821 ]

This fixes various data races in spinlock_debug. By testing with KCSAN,
it is observable that the console gets spammed with data races reports,
suggesting these are extremely frequent.

Example data race report:

  read to 0xffff8ab24f403c48 of 4 bytes by task 221 on cpu 2:
   debug_spin_lock_before kernel/locking/spinlock_debug.c:85 [inline]
   do_raw_spin_lock+0x9b/0x210 kernel/locking/spinlock_debug.c:112
   __raw_spin_lock include/linux/spinlock_api_smp.h:143 [inline]
   _raw_spin_lock+0x39/0x40 kernel/locking/spinlock.c:151
   spin_lock include/linux/spinlock.h:338 [inline]
   get_partial_node.isra.0.part.0+0x32/0x2f0 mm/slub.c:1873
   get_partial_node mm/slub.c:1870 [inline]
  <snip>

  write to 0xffff8ab24f403c48 of 4 bytes by task 167 on cpu 3:
   debug_spin_unlock kernel/locking/spinlock_debug.c:103 [inline]
   do_raw_spin_unlock+0xc9/0x1a0 kernel/locking/spinlock_debug.c:138
   __raw_spin_unlock_irqrestore include/linux/spinlock_api_smp.h:159 [inline]
   _raw_spin_unlock_irqrestore+0x2d/0x50 kernel/locking/spinlock.c:191
   spin_unlock_irqrestore include/linux/spinlock.h:393 [inline]
   free_debug_processing+0x1b3/0x210 mm/slub.c:1214
   __slab_free+0x292/0x400 mm/slub.c:2864
  <snip>

As a side-effect, with KCSAN, this eventually locks up the console, most
likely due to deadlock, e.g. .. -> printk lock -> spinlock_debug ->
KCSAN detects data race -> kcsan_print_report() -> printk lock ->
deadlock.

This fix will 1) avoid the data races, and 2) allow using lock debugging
together with KCSAN.

Reported-by: Qian Cai <cai@lca.pw>
Signed-off-by: Marco Elver <elver@google.com>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Paul E. McKenney <paulmck@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Will Deacon <will.deacon@arm.com>
Link: https://lkml.kernel.org/r/20191120155715.28089-1-elver@google.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

spi: fsl: Handle the single hardwired chipselect case

[ Upstream commit 7251953d784baf7e5416afabe030a0e81de1a938 ]

The Freescale MPC8xxx had a special quirk for handling a
single hardwired chipselect, the case when we're using neither
GPIO nor native chip select: when inspecting the device tree
and finding zero "cs-gpios" on the device node the code would
assume we have a single hardwired chipselect that leaves the
device always selected.

This quirk is not handled by the new core code, so we need
to check the "cs-gpios" explicitly in the driver and set
pdata->max_chipselect = 1 which will later fall through to
the SPI master ->num_chipselect.

Make sure not to assign the chip select handler in this
case: there is no handling needed since the chip is always
selected, and this is what the old code did as well.

Cc: Christophe Leroy <christophe.leroy@c-s.fr>
Reported-by: Christophe Leroy <christophe.leroy@c-s.fr>
Fixes: 0f0581b24bd0 ("spi: fsl: Convert to use CS GPIO descriptors")
Signed-off-by: Linus Walleij <linus.walleij@linaro.org>
Tested-by: Christophe Leroy <christophe.leroy@c-s.fr> (No tested the
Link: https://lore.kernel.org/r/20191128083718.39177-3-linus.walleij@linaro.org
Signed-off-by: Mark Brown <broonie@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

gpio: Handle counting of Freescale chipselects

[ Upstream commit 71b8f600b034c7f5780f6fb311dabfe331c64feb ]

We have a special quirk to handle the Freescale
nonstandard SPI chipselect GPIOs in the gpiolib-of.c
file, but it currently only handles the case where
the GPIOs are actually requested (gpiod_*get()).

We also need to handle that the SPI core attempts
to count the GPIOs before use, and that needs a
similar quirk in the OF part of the library.

Cc: Christophe Leroy <christophe.leroy@c-s.fr>
Reported-by: Christophe Leroy <christophe.leroy@c-s.fr>
Fixes: 0f0581b24bd0 ("spi: fsl: Convert to use CS GPIO descriptors")
Signed-off-by: Linus Walleij <linus.walleij@linaro.org>
Tested-by: Christophe Leroy <christophe.leroy@c-s.fr>
Link: https://lore.kernel.org/r/20191128083718.39177-2-linus.walleij@linaro.org
Signed-off-by: Mark Brown <broonie@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

spi: fsl: Fix GPIO descriptor support

[ Upstream commit f106904968e2a075e64653b9b79dda9f0f070ab5 ]

This makes the driver actually support looking up GPIO
descriptor. A coding mistake in the initial descriptor
support patch was that it was failing to turn on the very
feature it was implementing. Mea culpa.

Cc: Christophe Leroy <christophe.leroy@c-s.fr>
Reported-by: Christophe Leroy <christophe.leroy@c-s.fr>
Fixes: 0f0581b24bd0 ("spi: fsl: Convert to use CS GPIO descriptors")
Signed-off-by: Linus Walleij <linus.walleij@linaro.org>
Tested-by: Christophe Leroy <christophe.leroy@c-s.fr>
Link: https://lore.kernel.org/r/20191128083718.39177-1-linus.walleij@linaro.org
Signed-off-by: Mark Brown <broonie@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

ASoC: max98090: fix possible race conditions

[ Upstream commit 45dfbf56975994822cce00b7475732a49f8aefed ]

max98090_interrupt() and max98090_pll_work() run in 2 different threads.
There are 2 possible races:

Note: M98090_REG_DEVICE_STATUS = 0x01.
Note: ULK == 0, PLL is locked; ULK == 1, PLL is unlocked.

max98090_interrupt      max98090_pll_work
----------------------------------------------
schedule max98090_pll_work
                        restart max98090 codec
receive ULK INT
                        assert ULK == 0
schedule max98090_pll_work (1).

In the case (1), the PLL is locked but max98090_interrupt unnecessarily
schedules another max98090_pll_work.

max98090_interrupt      max98090_pll_work      max98090 codec
----------------------------------------------------------------------
                                               ULK = 1
receive ULK INT
read 0x01
                                               ULK = 0 (clear on read)
schedule max98090_pll_work
                        restart max98090 codec
                                               ULK = 1
receive ULK INT
read 0x01
                                               ULK = 0 (clear on read)
                        read 0x01
                        assert ULK == 0 (2).

In the case (2), both max98090_interrupt and max98090_pll_work read
the same clear-on-read register.  max98090_pll_work would falsely
thought PLL is locked.
Note: the case (2) race is introduced by the previous commit ("ASoC:
max98090: exit workaround earlier if PLL is locked") to check the status
and exit the loop earlier in max98090_pll_work.

There are 2 possible solution options:
A. turn off ULK interrupt before scheduling max98090_pll_work; and turn
on again before exiting max98090_pll_work.
B. remove the second thread of execution.

Option A cannot fix the case (2) race because it still has 2 threads
access the same clear-on-read register simultaneously.  Although we
could suppose the register is volatile and read the status via I2C could
be much slower than the hardware raises the bits.

Option B introduces a maximum 10~12 msec penalty delay in the interrupt
handler.  However, it could only punish the jack detection by extra
10~12 msec.

Adopts option B which is the better solution overall.

Signed-off-by: Tzung-Bi Shih <tzungbi@google.com>
Link: https://lore.kernel.org/r/20191122073114.219945-4-tzungbi@google.com
Reviewed-by: Pierre-Louis Bossart <pierre-louis.bossart@linux.intel.com>
Signed-off-by: Mark Brown <broonie@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

regulator: fix use after free issue

[ Upstream commit 4affd79a125ac91e6a53be843ea3960a8fc00cbb ]

This is caused by dereferencing 'rdev' after put_device() in
the _regulator_get()/_regulator_put() functions.
This patch just moves the put_device() down a bit to avoid the
issue.

Signed-off-by: Wen Yang <wenyang@linux.alibaba.com>
Cc: Liam Girdwood <lgirdwood@gmail.com>
Cc: Mark Brown <broonie@kernel.org>
Cc: linux-kernel@vger.kernel.org
Link: https://lore.kernel.org/r/20191124145835.25999-1-wenyang@linux.alibaba.com
Signed-off-by: Mark Brown <broonie@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

spi: pxa2xx: Add support for Intel Jasper Lake

[ Upstream commit 9c7315c9fca5de203538163cf42699bb10328902 ]

LPSS SPI on Intel Jasper Lake is compatible with Intel Ice Lake which
follows Intel Cannon Lake. Add PCI IDs of Jasper Lake.

Signed-off-by: Jarkko Nikula <jarkko.nikula@linux.intel.com>
Link: https://lore.kernel.org/r/20191125125159.15404-1-jarkko.nikula@linux.intel.com
Signed-off-by: Mark Brown <broonie@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

ASoC: rt5682: fix i2c arbitration lost issue

[ Upstream commit bc094709de0192a756c6946a7c89c543243ae609 ]

This patch modified the HW initial setting to fix i2c arbitration lost issue.

Signed-off-by: Shuming Fan <shumingf@realtek.com>
Link: https://lore.kernel.org/r/20191125091940.11953-1-shumingf@realtek.com
Signed-off-by: Mark Brown <broonie@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

bpf: Fix passing modified ctx to ld/abs/ind instruction

commit 6d4f151acf9a4f6fab09b615f246c717ddedcf0c upstream.

Anatoly has been fuzzing with kBdysch harness and reported a KASAN
slab oob in one of the outcomes:

  [...]
  [   77.359642] BUG: KASAN: slab-out-of-bounds in bpf_skb_load_helper_8_no_cache+0x71/0x130
  [   77.360463] Read of size 4 at addr ffff8880679bac68 by task bpf/406
  [   77.361119]
  [   77.361289] CPU: 2 PID: 406 Comm: bpf Not tainted 5.5.0-rc2-xfstests-00157-g2187f215eba #1
  [   77.362134] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS 1.12.0-1 04/01/2014
  [   77.362984] Call Trace:
  [   77.363249]  dump_stack+0x97/0xe0
  [   77.363603]  print_address_description.constprop.0+0x1d/0x220
  [   77.364251]  ? bpf_skb_load_helper_8_no_cache+0x71/0x130
  [   77.365030]  ? bpf_skb_load_helper_8_no_cache+0x71/0x130
  [   77.365860]  __kasan_report.cold+0x37/0x7b
  [   77.366365]  ? bpf_skb_load_helper_8_no_cache+0x71/0x130
  [   77.366940]  kasan_report+0xe/0x20
  [   77.367295]  bpf_skb_load_helper_8_no_cache+0x71/0x130
  [   77.367821]  ? bpf_skb_load_helper_8+0xf0/0xf0
  [   77.368278]  ? mark_lock+0xa3/0x9b0
  [   77.368641]  ? kvm_sched_clock_read+0x14/0x30
  [   77.369096]  ? sched_clock+0x5/0x10
  [   77.369460]  ? sched_clock_cpu+0x18/0x110
  [   77.369876]  ? bpf_skb_load_helper_8+0xf0/0xf0
  [   77.370330]  ___bpf_prog_run+0x16c0/0x28f0
  [   77.370755]  __bpf_prog_run32+0x83/0xc0
  [   77.371153]  ? __bpf_prog_run64+0xc0/0xc0
  [   77.371568]  ? match_held_lock+0x1b/0x230
  [   77.371984]  ? rcu_read_lock_held+0xa1/0xb0
  [   77.372416]  ? rcu_is_watching+0x34/0x50
  [   77.372826]  sk_filter_trim_cap+0x17c/0x4d0
  [   77.373259]  ? sock_kzfree_s+0x40/0x40
  [   77.373648]  ? __get_filter+0x150/0x150
  [   77.374059]  ? skb_copy_datagram_from_iter+0x80/0x280
  [   77.374581]  ? do_raw_spin_unlock+0xa5/0x140
  [   77.375025]  unix_dgram_sendmsg+0x33a/0xa70
  [   77.375459]  ? do_raw_spin_lock+0x1d0/0x1d0
  [   77.375893]  ? unix_peer_get+0xa0/0xa0
  [   77.376287]  ? __fget_light+0xa4/0xf0
  [   77.376670]  __sys_sendto+0x265/0x280
  [   77.377056]  ? __ia32_sys_getpeername+0x50/0x50
  [   77.377523]  ? lock_downgrade+0x350/0x350
  [   77.377940]  ? __sys_setsockopt+0x2a6/0x2c0
  [   77.378374]  ? sock_read_iter+0x240/0x240
  [   77.378789]  ? __sys_socketpair+0x22a/0x300
  [   77.379221]  ? __ia32_sys_socket+0x50/0x50
  [   77.379649]  ? mark_held_locks+0x1d/0x90
  [   77.380059]  ? trace_hardirqs_on_thunk+0x1a/0x1c
  [   77.380536]  __x64_sys_sendto+0x74/0x90
  [   77.380938]  do_syscall_64+0x68/0x2a0
  [   77.381324]  entry_SYSCALL_64_after_hwframe+0x49/0xbe
  [   77.381878] RIP: 0033:0x44c070
  [...]

After further debugging, turns out while in case of other helper functions
we disallow passing modified ctx, the special case of ld/abs/ind instruction
which has similar semantics (except r6 being the ctx argument) is missing
such check. Modified ctx is impossible here as bpf_skb_load_helper_8_no_cache()
and others are expecting skb fields in original position, hence, add
check_ctx_reg() to reject any modified ctx. Issue was first introduced back
in f1174f77b50c ("bpf/verifier: rework value tracking").

Fixes: f1174f77b50c ("bpf/verifier: rework value tracking")
Reported-by: Anatoly Trosinenko <anatoly.trosinenko@gmail.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Link: https://lore.kernel.org/bpf/20200106215157.3553-1-daniel@iogearbox.net
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

USB: dummy-hcd: use usb_urb_dir_in instead of usb_pipein

commit 6dabeb891c001c592645df2f477fed9f5d959987 upstream.

Commit fea3409112a9 ("USB: add direction bit to urb->transfer_flags") has
added a usb_urb_dir_in() helper function that can be used to determine
the direction of the URB. With that patch USB_DIR_IN control requests with
wLength == 0 are considered out requests by real USB HCDs. This patch
changes dummy-hcd to use the usb_urb_dir_in() helper to match that
behavior.

Signed-off-by: Andrey Konovalov <andreyknvl@google.com>
Link: https://lore.kernel.org/r/4ae9e68ebca02f08a93ac61fe065057c9a01f0a8.1571667489.git.andreyknvl@google.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

Linux 5.4.10

powerpc/pmem: Fix kernel crash due to wrong range value usage in flush_dcache_range

commit 6f4679b956741d2da6ad3ebb738cbe1264ac8781 upstream.

This patch fix the below kernel crash.

 BUG: Unable to handle kernel data access on read at 0xc000000380000000
 Faulting instruction address: 0xc00000000008b6f0
cpu 0x5: Vector: 300 (Data Access) at [c0000000d8587790]
    pc: c00000000008b6f0: arch_remove_memory+0x150/0x210
    lr: c00000000008b720: arch_remove_memory+0x180/0x210
    sp: c0000000d8587a20
   msr: 800000000280b033
   dar: c000000380000000
 dsisr: 40000000
  current = 0xc0000000d8558600
  paca    = 0xc00000000fff8f00   irqmask: 0x03   irq_happened: 0x01
    pid   = 1220, comm = ndctl
enter ? for help
 memunmap_pages+0x33c/0x410
 devm_action_release+0x30/0x50
 release_nodes+0x30c/0x3a0
 device_release_driver_internal+0x178/0x240
 unbind_store+0x74/0x190
 drv_attr_store+0x44/0x60
 sysfs_kf_write+0x74/0xa0
 kernfs_fop_write+0x1b0/0x260
 __vfs_write+0x3c/0x70
 vfs_write+0xe4/0x200
 ksys_write+0x7c/0x140
 system_call+0x5c/0x68

Fixes: 076265907cf9 ("powerpc: Chunk calls to flush_dcache_range in arch_*_memory")
Reported-by: Sachin Sant <sachinp@linux.vnet.ibm.com>
Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.ibm.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20191204052909.59145-1-aneesh.kumar@linux.ibm.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

Linux 5.4.9

mm/hugetlb: defer freeing of huge pages if in non-task context

[ Upstream commit c77c0a8ac4c522638a8242fcb9de9496e3cdbb2d ]

The following lockdep splat was observed when a certain hugetlbfs test
was run:

  ================================
  WARNING: inconsistent lock state
  4.18.0-159.el8.x86_64+debug #1 Tainted: G        W --------- -  -
  --------------------------------
  inconsistent {SOFTIRQ-ON-W} -> {IN-SOFTIRQ-W} usage.
  swapper/30/0 [HC0[0]:SC1[1]:HE1:SE0] takes:
  ffffffff9acdc038 (hugetlb_lock){+.?.}, at: free_huge_page+0x36f/0xaa0
  {SOFTIRQ-ON-W} state was registered at:
    lock_acquire+0x14f/0x3b0
    _raw_spin_lock+0x30/0x70
    __nr_hugepages_store_common+0x11b/0xb30
    hugetlb_sysctl_handler_common+0x209/0x2d0
    proc_sys_call_handler+0x37f/0x450
    vfs_write+0x157/0x460
    ksys_write+0xb8/0x170
    do_syscall_64+0xa5/0x4d0
    entry_SYSCALL_64_after_hwframe+0x6a/0xdf
  irq event stamp: 691296
  hardirqs last  enabled at (691296): [<ffffffff99bb034b>] _raw_spin_unlock_irqrestore+0x4b/0x60
  hardirqs last disabled at (691295): [<ffffffff99bb0ad2>] _raw_spin_lock_irqsave+0x22/0x81
  softirqs last  enabled at (691284): [<ffffffff97ff0c63>] irq_enter+0xc3/0xe0
  softirqs last disabled at (691285): [<ffffffff97ff0ebe>] irq_exit+0x23e/0x2b0

  other info that might help us debug this:
   Possible unsafe locking scenario:

         CPU0
         ----
    lock(hugetlb_lock);
    <Interrupt>
      lock(hugetlb_lock);

   *** DEADLOCK ***
      :
  Call Trace:
   <IRQ>
   __lock_acquire+0x146b/0x48c0
   lock_acquire+0x14f/0x3b0
   _raw_spin_lock+0x30/0x70
   free_huge_page+0x36f/0xaa0
   bio_check_pages_dirty+0x2fc/0x5c0
   clone_endio+0x17f/0x670 [dm_mod]
   blk_update_request+0x276/0xe50
   scsi_end_request+0x7b/0x6a0
   scsi_io_completion+0x1c6/0x1570
   blk_done_softirq+0x22e/0x350
   __do_softirq+0x23d/0xad8
   irq_exit+0x23e/0x2b0
   do_IRQ+0x11a/0x200
   common_interrupt+0xf/0xf
   </IRQ>

Both the hugetbl_lock and the subpool lock can be acquired in
free_huge_page().  One way to solve the problem is to make both locks
irq-safe.  However, Mike Kravetz had learned that the hugetlb_lock is
held for a linear scan of ALL hugetlb pages during a cgroup reparentling
operation.  So it is just too long to have irq disabled unless we can
break hugetbl_lock down into finer-grained locks with shorter lock hold
times.

Another alternative is to defer the freeing to a workqueue job.  This
patch implements the deferred freeing by adding a free_hpage_workfn()
work function to do the actual freeing.  The free_huge_page() call in a
non-task context saves the page to be freed in the hpage_freelist linked
list in a lockless manner using the llist APIs.

The generic workqueue is used to process the work, but a dedicated
workqueue can be used instead if it is desirable to have the huge page
freed ASAP.

Thanks to Kirill Tkhai <ktkhai@virtuozzo.com> for suggesting the use of
llist APIs which simplfy the code.

Link: http://lkml.kernel.org/r/20191217170331.30893-1-longman@redhat.com
Signed-off-by: Waiman Long <longman@redhat.com>
Reviewed-by: Mike Kravetz <mike.kravetz@oracle.com>
Acked-by: Davidlohr Bueso <dbueso@suse.de>
Acked-by: Michal Hocko <mhocko@suse.com>
Reviewed-by: Kirill Tkhai <ktkhai@virtuozzo.com>
Cc: Aneesh Kumar K.V <aneesh.kumar@linux.ibm.com>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Andi Kleen <ak@linux.intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

hsr: fix a race condition in node list insertion and deletion

[ Upstream commit 92a35678ec075100ce666a2fb6969151affb0e5d ]

hsr nodes are protected by RCU and there is no write side lock.
But node insertions and deletions could be being operated concurrently.
So write side locking is needed.

Test commands:
    ip netns add nst
    ip link add veth0 type veth peer name veth1
    ip link add veth2 type veth peer name veth3
    ip link set veth1 netns nst
    ip link set veth3 netns nst
    ip link set veth0 up
    ip link set veth2 up
    ip link add hsr0 type hsr slave1 veth0 slave2 veth2
    ip a a 192.168.100.1/24 dev hsr0
    ip link set hsr0 up
    ip netns exec nst ip link set veth1 up
    ip netns exec nst ip link set veth3 up
    ip netns exec nst ip link add hsr1 type hsr slave1 veth1 slave2 veth3
    ip netns exec nst ip a a 192.168.100.2/24 dev hsr1
    ip netns exec nst ip link set hsr1 up

    for i in {0..9}
    do
        for j in {0..9}
do
    for k in {0..9}
    do
        for l in {0..9}
do
        arping 192.168.100.2 -I hsr0 -s 00:01:3$i:4$j:5$k:6$l -c1 &
done
    done
done
    done

Splat looks like:
[  236.066091][ T3286] list_add corruption. next->prev should be prev (ffff8880a5940300), but was ffff8880a5940d0.
[  236.069617][ T3286] ------------[ cut here ]------------
[  236.070545][ T3286] kernel BUG at lib/list_debug.c:25!
[  236.071391][ T3286] invalid opcode: 0000 [#1] SMP DEBUG_PAGEALLOC KASAN PTI
[  236.072343][ T3286] CPU: 0 PID: 3286 Comm: arping Tainted: G        W         5.5.0-rc1+ #209
[  236.073463][ T3286] Hardware name: innotek GmbH VirtualBox/VirtualBox, BIOS VirtualBox 12/01/2006
[  236.074695][ T3286] RIP: 0010:__list_add_valid+0x74/0xd0
[  236.075499][ T3286] Code: 48 39 da 75 27 48 39 f5 74 36 48 39 dd 74 31 48 83 c4 08 b8 01 00 00 00 5b 5d c3 48 b
[  236.078277][ T3286] RSP: 0018:ffff8880aaa97648 EFLAGS: 00010286
[  236.086991][ T3286] RAX: 0000000000000075 RBX: ffff8880d4624c20 RCX: 0000000000000000
[  236.088000][ T3286] RDX: 0000000000000075 RSI: 0000000000000008 RDI: ffffed1015552ebf
[  236.098897][ T3286] RBP: ffff88809b53d200 R08: ffffed101b3c04f9 R09: ffffed101b3c04f9
[  236.099960][ T3286] R10: 00000000308769a1 R11: ffffed101b3c04f8 R12: ffff8880d4624c28
[  236.100974][ T3286] R13: ffff8880d4624c20 R14: 0000000040310100 R15: ffff8880ce17ee02
[  236.138967][ T3286] FS:  00007f23479fa680(0000) GS:ffff8880d9c00000(0000) knlGS:0000000000000000
[  236.144852][ T3286] CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[  236.145720][ T3286] CR2: 00007f4a14bab210 CR3: 00000000a61c6001 CR4: 00000000000606f0
[  236.146776][ T3286] Call Trace:
[  236.147222][ T3286]  hsr_add_node+0x314/0x490 [hsr]
[  236.153633][ T3286]  hsr_forward_skb+0x2b6/0x1bc0 [hsr]
[  236.154362][ T3286]  ? rcu_read_lock_sched_held+0x90/0xc0
[  236.155091][ T3286]  ? rcu_read_lock_bh_held+0xa0/0xa0
[  236.156607][ T3286]  hsr_dev_xmit+0x70/0xd0 [hsr]
[  236.157254][ T3286]  dev_hard_start_xmit+0x160/0x740
[  236.157941][ T3286]  __dev_queue_xmit+0x1961/0x2e10
[  236.158565][ T3286]  ? netdev_core_pick_tx+0x2e0/0x2e0
[ ... ]

Reported-by: syzbot+3924327f9ad5f4d2b343@syzkaller.appspotmail.com
Fixes: f421436a591d ("net/hsr: Add support for the High-availability Seamless Redundancy protocol (HSRv0)")
Signed-off-by: Taehee Yoo <ap420073@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Sasha Levin <sashal@kernel.org>

hsr: fix error handling routine in hsr_dev_finalize()

[ Upstream commit 1d19e2d53e8ed9e4c98fc95e0067492cda7288b0 ]

hsr_dev_finalize() is called to create new hsr interface.
There are some wrong error handling codes.

1. wrong checking return value of debugfs_create_{dir/file}.
These function doesn't return NULL. If error occurs in there,
it returns error pointer.
So, it should check error pointer instead of NULL.

2. It doesn't unregister interface if it fails to setup hsr interface.
If it fails to initialize hsr interface after register_netdevice(),
it should call unregister_netdevice().

3. Ignore failure of creation of debugfs
If creating of debugfs dir and file is failed, creating hsr interface
will be failed. But debugfs doesn't affect actual logic of hsr module.
So, ignoring this is more correct and this behavior is more general.

Fixes: c5a759117210 ("net/hsr: Use list_head (and rcu) instead of array for slave devices.")
Signed-off-by: Taehee Yoo <ap420073@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Sasha Levin <sashal@kernel.org>

hsr: avoid debugfs warning message when module is remove

[ Upstream commit 84bb59d773853bc2dda2ac1ef8474c40eb33a3c6 ]

When hsr module is being removed, debugfs_remove() is called to remove
both debugfs directory and file.

When module is being removed, module state is changed to
MODULE_STATE_GOING then exit() is called.
At this moment, module couldn't be held so try_module_get()
will be failed.

debugfs's open() callback tries to hold the module if .owner is existing.
If it fails, warning message is printed.

CPU0 CPU1
delete_module()
    try_stop_module()
    hsr_exit() open() <-- WARNING
        debugfs_remove()

In order to avoid the warning message, this patch makes hsr module does
not set .owner. Unsetting .owner is safe because these are protected by
inode_lock().

Test commands:
    #SHELL1
    ip link add dummy0 type dummy
    ip link add dummy1 type dummy
    while :
    do
        ip link add hsr0 type hsr slave1 dummy0 slave2 dummy1
modprobe -rv hsr
    done

    #SHELL2
    while :
    do
        cat /sys/kernel/debug/hsr0/node_table
    done

Splat looks like:
[  101.223783][ T1271] ------------[ cut here ]------------
[  101.230309][ T1271] debugfs file owner did not clean up at exit: node_table
[  101.230380][ T1271] WARNING: CPU: 3 PID: 1271 at fs/debugfs/file.c:309 full_proxy_open+0x10f/0x650
[  101.233153][ T1271] Modules linked in: hsr(-) dummy veth openvswitch nsh nf_conncount nf_nat nf_conntrack nf_d]
[  101.237112][ T1271] CPU: 3 PID: 1271 Comm: cat Tainted: G        W         5.5.0-rc1+ #204
[  101.238270][ T1271] Hardware name: innotek GmbH VirtualBox/VirtualBox, BIOS VirtualBox 12/01/2006
[  101.240379][ T1271] RIP: 0010:full_proxy_open+0x10f/0x650
[  101.241166][ T1271] Code: 48 c1 ea 03 80 3c 02 00 0f 85 c1 04 00 00 49 8b 3c 24 e8 04 86 7e ff 84 c0 75 2d 4c 8
[  101.251985][ T1271] RSP: 0018:ffff8880ca22fa38 EFLAGS: 00010286
[  101.273355][ T1271] RAX: dffffc0000000008 RBX: ffff8880cc6e6200 RCX: 0000000000000000
[  101.274466][ T1271] RDX: 0000000000000000 RSI: 0000000000000006 RDI: ffff8880c4dd5c14
[  101.275581][ T1271] RBP: 0000000000000000 R08: fffffbfff2922f5d R09: 0000000000000000
[  101.276733][ T1271] R10: 0000000000000001 R11: 0000000000000000 R12: ffffffffc0551bc0
[  101.277853][ T1271] R13: ffff8880c4059a48 R14: ffff8880be50a5e0 R15: ffffffff941adaa0
[  101.278956][ T1271] FS:  00007f8871cda540(0000) GS:ffff8880da800000(0000) knlGS:0000000000000000
[  101.280216][ T1271] CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[  101.282832][ T1271] CR2: 00007f88717cfd10 CR3: 00000000b9440005 CR4: 00000000000606e0
[  101.283974][ T1271] Call Trace:
[  101.285328][ T1271]  do_dentry_open+0x63c/0xf50
[  101.286077][ T1271]  ? open_proxy_open+0x270/0x270
[  101.288271][ T1271]  ? __x64_sys_fchdir+0x180/0x180
[  101.288987][ T1271]  ? inode_permission+0x65/0x390
[  101.289682][ T1271]  path_openat+0x701/0x2810
[  101.290294][ T1271]  ? path_lookupat+0x880/0x880
[  101.290957][ T1271]  ? check_chain_key+0x236/0x5d0
[  101.291676][ T1271]  ? __lock_acquire+0xdfe/0x3de0
[  101.292358][ T1271]  ? sched_clock+0x5/0x10
[  101.292962][ T1271]  ? sched_clock_cpu+0x18/0x170
[  101.293644][ T1271]  ? find_held_lock+0x39/0x1d0
[  101.305616][ T1271]  do_filp_open+0x17a/0x270
[  101.306061][ T1271]  ? may_open_dev+0xc0/0xc0
[ ... ]

Fixes: fc4ecaeebd26 ("net: hsr: add debugfs support for display node list")
Signed-off-by: Taehee Yoo <ap420073@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Sasha Levin <sashal@kernel.org>

net: annotate lockless accesses to sk->sk_pacing_shift

[ Upstream commit 7c68fa2bddda6d942bd387c9ba5b4300737fd991 ]

sk->sk_pacing_shift can be read and written without lock
synchronization. This patch adds annotations to
document this fact and avoid future syzbot complains.

This might also avoid unexpected false sharing
in sk_pacing_shift_update(), as the compiler
could remove the conditional check and always
write over sk->sk_pacing_shift :

if (sk->sk_pacing_shift != val)
sk->sk_pacing_shift = val;

Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Sasha Levin <sashal@kernel.org>

perf/x86/intel/bts: Fix the use of page_private()

[ Upstream commit ff61541cc6c1962957758ba433c574b76f588d23 ]

Commit

  8062382c8dbe2 ("perf/x86/intel/bts: Add BTS PMU driver")

brought in a warning with the BTS buffer initialization
that is easily tripped with (assuming KPTI is disabled):

instantly throwing:

> ------------[ cut here ]------------
> WARNING: CPU: 2 PID: 326 at arch/x86/events/intel/bts.c:86 bts_buffer_setup_aux+0x117/0x3d0
> Modules linked in:
> CPU: 2 PID: 326 Comm: perf Not tainted 5.4.0-rc8-00291-gceb9e77324fa #904
> RIP: 0010:bts_buffer_setup_aux+0x117/0x3d0
> Call Trace:
>  rb_alloc_aux+0x339/0x550
>  perf_mmap+0x607/0xc70
>  mmap_region+0x76b/0xbd0
...

It appears to assume (for lost raisins) that PagePrivate() is set,
while later it actually tests for PagePrivate() before using
page_private().

Make it consistent and always check PagePrivate() before using
page_private().

Fixes: 8062382c8dbe2 ("perf/x86/intel/bts: Add BTS PMU driver")
Signed-off-by: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Vince Weaver <vincent.weaver@maine.edu>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Link: https://lkml.kernel.org/r/20191205142853.28894-2-alexander.shishkin@linux.intel.com
Signed-off-by: Sasha Levin <sashal@kernel.org>

efi: Don't attempt to map RCI2 config table if it doesn't exist

[ Upstream commit a470552ee8965da0fe6fd4df0aa39c4cda652c7c ]

Commit:

  1c5fecb61255aa12 ("efi: Export Runtime Configuration Interface table to sysfs")

... added support for a Dell specific UEFI configuration table, but
failed to take into account that mapping the table should not be
attempted unless the table actually exists. If it doesn't exist,
the code usually fails silently unless pr_debug() prints are
enabled. However, on 32-bit PAE x86, the splat below is produced due
to the attempt to map the placeholder value EFI_INVALID_TABLE_ADDR
which we use for non-existing UEFI configuration tables, and which
equals ULONG_MAX.

   memremap attempted on mixed range 0x00000000ffffffff size: 0x1e
   WARNING: CPU: 1 PID: 1 at kernel/iomem.c:81 memremap+0x1a3/0x1c0
   Modules linked in:
   CPU: 1 PID: 1 Comm: swapper/0 Not tainted 5.4.2-smp-mine #1
   Hardware name: Hewlett-Packard HP Z400 Workstation/0B4Ch, BIOS 786G3 v03.61 03/05/2018
   EIP: memremap+0x1a3/0x1c0
  ...
   Call Trace:
    ? map_properties+0x473/0x473
    ? efi_rci2_sysfs_init+0x2c/0x154
    ? map_properties+0x473/0x473
    ? do_one_initcall+0x49/0x1d4
    ? parse_args+0x1e8/0x2a0
    ? do_early_param+0x7a/0x7a
    ? kernel_init_freeable+0x139/0x1c2
    ? rest_init+0x8e/0x8e
    ? kernel_init+0xd/0xf2
    ? ret_from_fork+0x2e/0x38

Fix this by checking whether the table exists before attempting to map it.

Reported-by: Richard Narron <comet.berkeley@gmail.com>
Tested-by: Richard Narron <comet.berkeley@gmail.com>
Signed-off-by: Ard Biesheuvel <ardb@kernel.org>
Cc: linux-efi@vger.kernel.org
Fixes: 1c5fecb61255aa12 ("efi: Export Runtime Configuration Interface table to sysfs")
Link: https://lkml.kernel.org/r/20191210090945.11501-2-ardb@kernel.org
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

lib/ubsan: don't serialize UBSAN report

[ Upstream commit ce5c31db3645b649a31044a4d8b6057f6c723702 ]

At the moment, UBSAN report will be serialized using a spin_lock().  On
RT-systems, spinlocks are turned to rt_spin_lock and may sleep.  This
will result to the following splat if the undefined behavior is in a
context that can sleep:

  BUG: sleeping function called from invalid context at /src/linux/kernel/locking/rtmutex.c:968
  in_atomic(): 1, irqs_disabled(): 128, pid: 3447, name: make
  1 lock held by make/3447:
   #0: 000000009a966332 (&mm->mmap_sem){++++}, at: do_page_fault+0x140/0x4f8
  irq event stamp: 6284
  hardirqs last  enabled at (6283): [<ffff000011326520>] _raw_spin_unlock_irqrestore+0x90/0xa0
  hardirqs last disabled at (6284): [<ffff0000113262b0>] _raw_spin_lock_irqsave+0x30/0x78
  softirqs last  enabled at (2430): [<ffff000010088ef8>] fpsimd_restore_current_state+0x60/0xe8
  softirqs last disabled at (2427): [<ffff000010088ec0>] fpsimd_restore_current_state+0x28/0xe8
  Preemption disabled at:
  [<ffff000011324a4c>] rt_mutex_futex_unlock+0x4c/0xb0
  CPU: 3 PID: 3447 Comm: make Tainted: G        W         5.2.14-rt7-01890-ge6e057589653 #911
  Call trace:
    dump_backtrace+0x0/0x148
    show_stack+0x14/0x20
    dump_stack+0xbc/0x104
    ___might_sleep+0x154/0x210
    rt_spin_lock+0x68/0xa0
    ubsan_prologue+0x30/0x68
    handle_overflow+0x64/0xe0
    __ubsan_handle_add_overflow+0x10/0x18
    __lock_acquire+0x1c28/0x2a28
    lock_acquire+0xf0/0x370
    _raw_spin_lock_irqsave+0x58/0x78
    rt_mutex_futex_unlock+0x4c/0xb0
    rt_spin_unlock+0x28/0x70
    get_page_from_freelist+0x428/0x2b60
    __alloc_pages_nodemask+0x174/0x1708
    alloc_pages_vma+0x1ac/0x238
    __handle_mm_fault+0x4ac/0x10b0
    handle_mm_fault+0x1d8/0x3b0
    do_page_fault+0x1c8/0x4f8
    do_translation_fault+0xb8/0xe0
    do_mem_abort+0x3c/0x98
    el0_da+0x20/0x24

The spin_lock() will protect against multiple CPUs to output a report
together, I guess to prevent them from being interleaved.  However, they
can still interleave with other messages (and even splat from
__might_sleep).

So the lock usefulness seems pretty limited.  Rather than trying to
accomodate RT-system by switching to a raw_spin_lock(), the lock is now
completely dropped.

Link: http://lkml.kernel.org/r/20190920100835.14999-1-julien.grall@arm.com
Signed-off-by: Julien Grall <julien.grall@arm.com>
Reported-by: Andre Przywara <andre.przywara@arm.com>
Acked-by: Andrey Ryabinin <aryabinin@virtuozzo.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Cc: Steven Rostedt <rostedt@goodmis.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

xen/blkback: Avoid unmapping unmapped grant pages

[ Upstream commit f9bd84a8a845d82f9b5a081a7ae68c98a11d2e84 ]

For each I/O request, blkback first maps the foreign pages for the
request to its local pages.  If an allocation of a local page for the
mapping fails, it should unmap every mapping already made for the
request.

However, blkback's handling mechanism for the allocation failure does
not mark the remaining foreign pages as unmapped.  Therefore, the unmap
function merely tries to unmap every valid grant page for the request,
including the pages not mapped due to the allocation failure.  On a
system that fails the allocation frequently, this problem leads to
following kernel crash.

  [  372.012538] BUG: unable to handle kernel NULL pointer dereference at 0000000000000001
  [  372.012546] IP: [<ffffffff814071ac>] gnttab_unmap_refs.part.7+0x1c/0x40
  [  372.012557] PGD 16f3e9067 PUD 16426e067 PMD 0
  [  372.012562] Oops: 0002 [#1] SMP
  [  372.012566] Modules linked in: act_police sch_ingress cls_u32
  ...
  [  372.012746] Call Trace:
  [  372.012752]  [<ffffffff81407204>] gnttab_unmap_refs+0x34/0x40
  [  372.012759]  [<ffffffffa0335ae3>] xen_blkbk_unmap+0x83/0x150 [xen_blkback]
  ...
  [  372.012802]  [<ffffffffa0336c50>] dispatch_rw_block_io+0x970/0x980 [xen_blkback]
  ...
  Decompressing Linux... Parsing ELF... done.
  Booting the kernel.
  [    0.000000] Initializing cgroup subsys cpuset

This commit fixes this problem by marking the grant pages of the given
request that didn't mapped due to the allocation failure as invalid.

Fixes: c6cc142dac52 ("xen-blkback: use balloon pages for all mappings")

Reviewed-by: David Woodhouse <dwmw@amazon.de>
Reviewed-by: Maximilian Heyne <mheyne@amazon.de>
Reviewed-by: Paul Durrant <pdurrant@amazon.co.uk>
Reviewed-by: Roger Pau Monné <roger.pau@citrix.com>
Signed-off-by: SeongJae Park <sjpark@amazon.de>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Signed-off-by: Sasha Levin <sashal@kernel.org>

mm/sparse.c: mark populate_section_memmap as __meminit

[ Upstream commit 030eab4f9ffb469344c10a46bc02c5149db0a2a9 ]

Building the kernel on s390 with -Og produces the following warning:

  WARNING: vmlinux.o(.text+0x28dabe): Section mismatch in reference from the function populate_section_memmap() to the function .meminit.text:__populate_section_memmap()
  The function populate_section_memmap() references
  the function __meminit __populate_section_memmap().
  This is often because populate_section_memmap lacks a __meminit
  annotation or the annotation of __populate_section_memmap is wrong.

While -Og is not supported, in theory this might still happen with
another compiler or on another architecture.  So fix this by using the
correct section annotations.

[iii@linux.ibm.com: v2]
Link: http://lkml.kernel.org/r/20191030151639.41486-1-iii@linux.ibm.com
Link: http://lkml.kernel.org/r/20191028165549.14478-1-iii@linux.ibm.com
Signed-off-by: Ilya Leoshkevich <iii@linux.ibm.com>
Acked-by: David Hildenbrand <david@redhat.com>
Cc: Heiko Carstens <heiko.carstens@de.ibm.com>
Cc: Vasily Gorbik <gor@linux.ibm.com>
Cc: Oscar Salvador <OSalvador@suse.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

s390/smp: fix physical to logical CPU map for SMT

[ Upstream commit 72a81ad9d6d62dcb79f7e8ad66ffd1c768b72026 ]

If an SMT capable system is not IPL'ed from the first CPU the setup of
the physical to logical CPU mapping is broken: the IPL core gets CPU
number 0, but then the next core gets CPU number 1. Correct would be
that all SMT threads of CPU 0 get the subsequent logical CPU numbers.

This is important since a lot of code (like e.g. the CPU topology
code) assumes that CPU maps are setup like this. If the mapping is
broken the system will not IPL due to broken topology masks:

[    1.716341] BUG: arch topology broken
[    1.716342]      the SMT domain not a subset of the MC domain
[    1.716343] BUG: arch topology broken
[    1.716344]      the MC domain not a subset of the BOOK domain

This scenario can usually not happen since LPARs are always IPL'ed
from CPU 0 and also re-IPL is intiated from CPU 0. However older
kernels did initiate re-IPL on an arbitrary CPU. If therefore a re-IPL
from an old kernel into a new kernel is initiated this may lead to
crash.

Fix this by setting up the physical to logical CPU mapping correctly.

Signed-off-by: Heiko Carstens <heiko.carstens@de.ibm.com>
Signed-off-by: Vasily Gorbik <gor@linux.ibm.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

Btrfs: only associate the locked page with one async_chunk struct

[ Upstream commit 1d53c9e6723022b12e4a5ed4b141f67c834b7f6f ]

The btrfs writepages function collects a large range of pages flagged
for delayed allocation, and then sends them down through the COW code
for processing.  When compression is on, we allocate one async_chunk
structure for every 512K, and then run those pages through the
compression code for IO submission.

writepages starts all of this off with a single page, locked by the
original call to extent_write_cache_pages(), and it's important to keep
track of this page because it has already been through
clear_page_dirty_for_io().

The btrfs async_chunk struct has a pointer to the locked_page, and when
we're redirtying the page because compression had to fallback to
uncompressed IO, we use page->index to decide if a given async_chunk
struct really owns that page.

But, this is racey.  If a given delalloc range is broken up into two
async_chunks (chunkA and chunkB), we can end up with something like
this:

 compress_file_range(chunkA)
 submit_compress_extents(chunkA)
 submit compressed bios(chunkA)
 put_page(locked_page)

 compress_file_range(chunkB)
 ...

Or:

 async_cow_submit
  submit_compressed_extents <--- falls back to buffered writeout
   cow_file_range
    extent_clear_unlock_delalloc
     __process_pages_contig
       put_page(locked_pages)

    async_cow_submit

The end result is that chunkA is completed and cleaned up before chunkB
even starts processing.  This means we can free locked_page() and reuse
it elsewhere.  If we get really lucky, it'll have the same page->index
in its new home as it did before.

While we're processing chunkB, we might decide we need to fall back to
uncompressed IO, and so compress_file_range() will call
__set_page_dirty_nobufers() on chunkB->locked_page.

Without cgroups in use, this creates as a phantom dirty page, which
isn't great but isn't the end of the world. What can happen, it can go
through the fixup worker and the whole COW machinery again:

in submit_compressed_extents():
  while (async extents) {
  ...
    cow_file_range
    if (!page_started ...)
      extent_write_locked_range
    else if (...)
      unlock_page
    continue;

This hasn't been observed in practice but is still possible.

With cgroups in use, we might crash in the accounting code because
page->mapping->i_wb isn't set.

  BUG: unable to handle kernel NULL pointer dereference at 00000000000000d0
  IP: percpu_counter_add_batch+0x11/0x70
  PGD 66534e067 P4D 66534e067 PUD 66534f067 PMD 0
  Oops: 0000 [#1] SMP DEBUG_PAGEALLOC
  CPU: 16 PID: 2172 Comm: rm Not tainted
  RIP: 0010:percpu_counter_add_batch+0x11/0x70
  RSP: 0018:ffffc9000a97bbe0 EFLAGS: 00010286
  RAX: 0000000000000005 RBX: 0000000000000090 RCX: 0000000000026115
  RDX: 0000000000000030 RSI: ffffffffffffffff RDI: 0000000000000090
  RBP: 0000000000000000 R08: fffffffffffffff5 R09: 0000000000000000
  R10: 00000000000260c0 R11: ffff881037fc26c0 R12: ffffffffffffffff
  R13: ffff880fe4111548 R14: ffffc9000a97bc90 R15: 0000000000000001
  FS:  00007f5503ced480(0000) GS:ffff880ff7200000(0000) knlGS:0000000000000000
  CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
  CR2: 00000000000000d0 CR3: 00000001e0459005 CR4: 0000000000360ee0
  DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
  DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400
  Call Trace:
   account_page_cleaned+0x15b/0x1f0
   __cancel_dirty_page+0x146/0x200
   truncate_cleanup_page+0x92/0xb0
   truncate_inode_pages_range+0x202/0x7d0
   btrfs_evict_inode+0x92/0x5a0
   evict+0xc1/0x190
   do_unlinkat+0x176/0x280
   do_syscall_64+0x63/0x1a0
   entry_SYSCALL_64_after_hwframe+0x42/0xb7

The fix here is to make asyc_chunk->locked_page NULL everywhere but the
one async_chunk struct that's allowed to do things to the locked page.

Link: https://lore.kernel.org/linux-btrfs/c2419d01-5c84-3fb4-189e-4db519d08796@suse.com/
Fixes: 771ed689d2cd ("Btrfs: Optimize compressed writeback and reads")
Reviewed-by: Josef Bacik <josef@toxicpanda.com>
Signed-off-by: Chris Mason <clm@fb.com>
[ update changelog from mail thread discussion ]
Signed-off-by: David Sterba <dsterba@suse.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>