Btrfs: keep pages dirty when using btrfs_writepage_fixup_worker

[ Upstream commit 25f3c5021985e885292980d04a1423fd83c967bb ]

For COW, btrfs expects pages dirty pages to have been through a few setup
steps.  This includes reserving space for the new block allocations and marking
the range in the state tree for delayed allocation.

A few places outside btrfs will dirty pages directly, especially when unmapping
mmap'd pages.  In order for these to properly go through COW, we run them
through a fixup worker to wait for stable pages, and do the delalloc prep.

87826df0ec36 added a window where the dirty pages were cleaned, but pending
more action from the fixup worker.  We clear_page_dirty_for_io() before
we call into writepage, so the page is no longer dirty.  The commit
changed it so now we leave the page clean between unlocking it here and
the fixup worker starting at some point in the future.

During this window, page migration can jump in and relocate the page.  Once our
fixup work actually starts, it finds page->mapping is NULL and we end up
freeing the page without ever writing it.

This leads to crc errors and other exciting problems, since it screws up the
whole statemachine for waiting for ordered extents.  The fix here is to keep
the page dirty while we're waiting for the fixup worker to get to work.
This is accomplished by returning -EAGAIN from btrfs_writepage_cow_fixup
if we queued the page up for fixup, which will cause the writepage
function to redirty the page.

Because we now expect the page to be dirty once it gets to the fixup
worker we must adjust the error cases to call clear_page_dirty_for_io()
on the page.  That is the bulk of the patch, but it is not the fix, the
fix is the -EAGAIN from btrfs_writepage_cow_fixup.  We cannot separate
these two changes out because the error conditions change with the new
expectations.

Signed-off-by: Chris Mason <clm@fb.com>
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>

KVM: s390: ENOTSUPP -> EOPNOTSUPP fixups

[ Upstream commit c611990844c28c61ca4b35ff69d3a2ae95ccd486 ]

There is no ENOTSUPP for userspace.

Reported-by: Julian Wiedmann <jwi@linux.ibm.com>
Fixes: 519783935451 ("KVM: s390: introduce ais mode modify function")
Fixes: 2c1a48f2e5ed ("KVM: S390: add new group for flic")
Reviewed-by: Cornelia Huck <cohuck@redhat.com>
Reviewed-by: Thomas Huth <thuth@redhat.com>
Signed-off-by: Christian Borntraeger <borntraeger@de.ibm.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

nbd: add a flush_workqueue in nbd_start_device

[ Upstream commit 5c0dd228b5fc30a3b732c7ae2657e0161ec7ed80 ]

When kzalloc fail, may cause trying to destroy the
workqueue from inside the workqueue.

If num_connections is m (2 < m), and NO.1 ~ NO.n
(1 < n < m) kzalloc are successful. The NO.(n + 1)
failed. Then, nbd_start_device will return ENOMEM
to nbd_start_device_ioctl, and nbd_start_device_ioctl
will return immediately without running flush_workqueue.
However, we still have n recv threads. If nbd_release
run first, recv threads may have to drop the last
config_refs and try to destroy the workqueue from
inside the workqueue.

To fix it, add a flush_workqueue in nbd_start_device.

Fixes: e9e006f5fcf2 ("nbd: fix max number of supported devs")
Signed-off-by: Sun Ke <sunke32@huawei.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Signed-off-by: Sasha Levin <sashal@kernel.org>

tracing: Simplify assignment parsing for hist triggers

[ Upstream commit b527b638fd63ba791dc90a0a6e9a3035b10df52b ]

In the process of adding better error messages for sorting, I realized
that strsep was being used incorrectly and some of the error paths I
was expecting to be hit weren't and just fell through to the common
invalid key error case.

It also became obvious that for keyword assignments, it wasn't
necessary to save the full assignment and reparse it later, and having
a common empty-assignment check would also make more sense in terms of
error processing.

Change the code to fix these problems and simplify it for new error
message changes in a subsequent patch.

Link: http://lkml.kernel.org/r/1c3ef0b6655deaf345f6faee2584a0298ac2d743.1561743018.git.zanussi@kernel.org
Fixes: e62347d24534 ("tracing: Add hist trigger support for user-defined sorting ('sort=' param)")
Fixes: 7ef224d1d0e3 ("tracing: Add 'hist' event trigger command")
Fixes: a4072fe85ba3 ("tracing: Add a clock attribute for hist triggers")
Reported-by: Masami Hiramatsu <mhiramat@kernel.org>
Reviewed-by: Masami Hiramatsu <mhiramat@kernel.org>
Signed-off-by: Tom Zanussi <zanussi@kernel.org>
Signed-off-by: Steven Rostedt (VMware) <rostedt@goodmis.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

drm/amd/display: Retrain dongles when SINK_COUNT becomes non-zero

[ Upstream commit 3eb6d7aca53d81ce888624f09cd44dc0302161e8 ]

[WHY]
Two years ago the patch referenced by the Fixes tag stopped running
dp_verify_link_cap_with_retries during DP detection when the reason
for the detection was a short-pulse interrupt. This effectively meant
that we were no longer doing the verify_link_cap training on active
dongles when their SINK_COUNT changed from 0 to 1.

A year ago this was partly remedied with:
commit 80adaebd2d41 ("drm/amd/display: Don't skip link training for empty dongle")

This made sure that we trained the dongle on initial hotplug (without
connected downstream devices).

This is all fine and dandy if it weren't for the fact that there are
some dongles on the market that don't like link training when SINK_COUNT
is 0 These dongles will in fact indicate a SINK_COUNT of 0 immediately
after hotplug, even when a downstream device is connected, and then
trigger a shortpulse interrupt indicating a SINK_COUNT change to 1.

In order to play nicely we will need our policy to not link train an
active DP dongle when SINK_COUNT is 0 but ensure we train it when the
SINK_COUNT changes to 1.

[HOW]
Call dp_verify_link_cap_with_retries on detection even when the detection
is triggered from a short pulse interrupt.

With this change we can also revert this commit which we'll do in a separate
follow-up change:
commit 80adaebd2d41 ("drm/amd/display: Don't skip link training for empty dongle")

Fixes: 0301ccbaf67d ("drm/amd/display: DP Compliance 400.1.1 failure")
Suggested-by: Louis Li <Ching-shih.Li@amd.com>
Tested-by: Louis Li <Ching-shih.Li@amd.com>
Cc: Wenjing Liu <Wenjing.Liu@amd.com>
Cc: Hersen Wu <hersenxs.wu@amd.com>
Cc: Eric Yang <Eric.Yang2@amd.com>
Reviewed-by: Wenjing Liu <Wenjing.Liu@amd.com>
Signed-off-by: Harry Wentland <harry.wentland@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

rtc: i2c/spi: Avoid inclusion of REGMAP support when not needed

[ Upstream commit 34719de919af07682861cb0fa2bcf64da33ecf44 ]

Merely enabling I2C and RTC selects REGMAP_I2C and REGMAP_SPI, even when
no driver needs it.  While the former can be moduler, the latter cannot,
and thus becomes built-in.

Fix this by moving the select statements for REGMAP_I2C and REGMAP_SPI
from the RTC_I2C_AND_SPI helper to the individual drivers that depend on
it.

Note that the comment for RTC_I2C_AND_SPI refers to SND_SOC_I2C_AND_SPI
for more information, but the latter does not select REGMAP_{I2C,SPI}
itself, and defers that to the individual drivers, too.

Fixes: 080481f54ef62121 ("rtc: merge ds3232 and ds3234")
Signed-off-by: Geert Uytterhoeven <geert@linux-m68k.org>
Reported-by: kbuild test robot <lkp@intel.com>
Reported-by: kbuild test robot <lkp@intel.com>
Link: https://lore.kernel.org/r/20200112171349.22268-1-geert@linux-m68k.org
Signed-off-by: Alexandre Belloni <alexandre.belloni@bootlin.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

selftests: settings: tests can be in subsubdirs

[ Upstream commit ac87813d4372f4c005264acbe3b7f00c1dee37c4 ]

Commit 852c8cbf34d3 ("selftests/kselftest/runner.sh: Add 45 second
timeout per test") adds support for a new per-test-directory "settings"
file. But this only works for tests not in a sub-subdirectories, e.g.

 - tools/testing/selftests/rtc (rtc) is OK,
 - tools/testing/selftests/net/mptcp (net/mptcp) is not.

We have to increase the timeout for net/mptcp tests which are not
upstreamed yet but this fix is valid for other tests if they need to add
a "settings" file, see the full list with:

  tools/testing/selftests/*/*/**/Makefile

Note that this patch changes the text header message printed at the end
of the execution but this text is modified only for the tests that are
in sub-subdirectories, e.g.

  ok 1 selftests: net/mptcp: mptcp_connect.sh

Before we had:

  ok 1 selftests: mptcp: mptcp_connect.sh

But showing the full target name is probably better, just in case a
subsubdir has the same name as another one in another subdirectory.

Fixes: 852c8cbf34d3 (selftests/kselftest/runner.sh: Add 45 second timeout per test)
Signed-off-by: Matthieu Baerts <matthieu.baerts@tessares.net>
Reviewed-by: Kees Cook <keescook@chromium.org>
Signed-off-by: Shuah Khan <skhan@linuxfoundation.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

brcmfmac: sdio: Fix OOB interrupt initialization on brcm43362

[ Upstream commit 8c8e60fb86a90a30721bbd797f58f96b3980dcc1 ]

Commit 262f2b53f679 ("brcmfmac: call brcmf_attach() just before calling
brcmf_bus_started()") changed the initialization order of the brcmfmac
SDIO driver. Unfortunately since brcmf_sdiod_intr_register() is now
called before the sdiodev->bus_if initialization, it reads the wrong
chip ID and fails to initialize the GPIO on brcm43362. Thus the chip
cannot send interrupts and fails to probe:

[   12.517023] brcmfmac: brcmf_sdio_bus_rxctl: resumed on timeout
[   12.531214] ieee80211 phy0: brcmf_bus_started: failed: -110
[   12.536976] ieee80211 phy0: brcmf_attach: dongle is not responding: err=-110
[   12.566467] brcmfmac: brcmf_sdio_firmware_callback: brcmf_attach failed

Initialize the bus interface earlier to ensure that
brcmf_sdiod_intr_register() properly sets up the OOB interrupt.

BugLink: https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=908438
Fixes: 262f2b53f679 ("brcmfmac: call brcmf_attach() just before calling brcmf_bus_started()")
Signed-off-by: Jean-Philippe Brucker <jean-philippe@linaro.org>
Reviewed-by: Arend van Spriel <arend.vanspriel@broadcom.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

rtw88: fix rate mask for 1SS chip

[ Upstream commit 35a68fa5f96a80797e11b6952a47c5a84939a7bf ]

The rate mask is used to tell firmware the supported rate depends on
negotiation. We loop 2 times for all VHT/HT 2SS rate mask first, and then
only keep the part according to chip's NSS.

This commit fixes the logic error of '&' operations for VHT/HT rate, and
we should run this logic before adding legacy rate.

To access HT MCS map, index 0/1 represent MCS 0-7/8-15 respectively. Use
NL80211_BAND_xxx is incorrect, so fix it as well.

Signed-off-by: Ping-Ke Shih <pkshih@realtek.com>
Signed-off-by: Yan-Hsuan Chuang <yhchuang@realtek.com>
Reviewed-by: Chris Chiu <chiu@endlessm.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

ath10k: Correct the DMA direction for management tx buffers

[ Upstream commit 6ba8b3b6bd772f575f7736c8fd893c6981fcce16 ]

The management packets, send to firmware via WMI, are
mapped using the direction DMA_TO_DEVICE. Currently in
case of wmi cleanup, these buffers are being unmapped
using an incorrect DMA direction. This can cause unwanted
behavior when the host driver is handling a restart
of the wlan firmware.

We might see a trace like below

[<ffffff8008098b18>] __dma_inv_area+0x28/0x58
[<ffffff8001176734>] ath10k_wmi_mgmt_tx_clean_up_pending+0x60/0xb0 [ath10k_core]
[<ffffff80088c7c50>] idr_for_each+0x78/0xe4
[<ffffff80011766a4>] ath10k_wmi_detach+0x4c/0x7c [ath10k_core]
[<ffffff8001163d7c>] ath10k_core_stop+0x58/0x68 [ath10k_core]
[<ffffff800114fb74>] ath10k_halt+0xec/0x13c [ath10k_core]
[<ffffff8001165110>] ath10k_core_restart+0x11c/0x1a8 [ath10k_core]
[<ffffff80080c36bc>] process_one_work+0x16c/0x31c

Fix the incorrect DMA direction during the wmi
management tx buffer cleanup.

Tested HW: WCN3990
Tested FW: WLAN.HL.3.1-00784-QCAHLSWMTPLZ-1

Fixes: dc405152bb6 ("ath10k: handle mgmt tx completion event")
Signed-off-by: Rakesh Pillai <pillair@codeaurora.org>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

ext4, jbd2: ensure panic when aborting with zero errno

[ Upstream commit 51f57b01e4a3c7d7bdceffd84de35144e8c538e7 ]

JBD2_REC_ERR flag used to indicate the errno has been updated when jbd2
aborted, and then __ext4_abort() and ext4_handle_error() can invoke
panic if ERRORS_PANIC is specified. But if the journal has been aborted
with zero errno, jbd2_journal_abort() didn't set this flag so we can
no longer panic. Fix this by always record the proper errno in the
journal superblock.

Fixes: 4327ba52afd03 ("ext4, jbd2: ensure entering into panic after recording an error in superblock")
Signed-off-by: zhangyi (F) <yi.zhang@huawei.com>
Reviewed-by: Jan Kara <jack@suse.cz>
Link: https://lore.kernel.org/r/20191204124614.45424-3-yi.zhang@huawei.com
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
Signed-off-by: Sasha Levin <sashal@kernel.org>

ARM: 8952/1: Disable kmemleak on XIP kernels

[ Upstream commit bc420c6ceefbb86cbbc8c00061bd779c17fa6997 ]

Kmemleak relies on specific symbols to register the read only data
during init (e.g. __start_ro_after_init).
Trying to build an XIP kernel on arm results in the linking error
reported below because when this option is selected read only data
after init are not allowed since .data is read only (.rodata).

  arm-linux-gnueabihf-ld: mm/kmemleak.o: in function `kmemleak_init':
  kmemleak.c:(.init.text+0x148): undefined reference to `__end_ro_after_init'
  arm-linux-gnueabihf-ld: kmemleak.c:(.init.text+0x14c):
     undefined reference to `__end_ro_after_init'
  arm-linux-gnueabihf-ld: kmemleak.c:(.init.text+0x150):
     undefined reference to `__start_ro_after_init'
  arm-linux-gnueabihf-ld: kmemleak.c:(.init.text+0x156):
     undefined reference to `__start_ro_after_init'
  arm-linux-gnueabihf-ld: kmemleak.c:(.init.text+0x162):
     undefined reference to `__start_ro_after_init'
  arm-linux-gnueabihf-ld: kmemleak.c:(.init.text+0x16a):
     undefined reference to `__start_ro_after_init'
  linux/Makefile:1078: recipe for target 'vmlinux' failed

Fix the issue enabling kmemleak only on non XIP kernels.

Signed-off-by: Vincenzo Frascino <vincenzo.frascino@arm.com>
Signed-off-by: Russell King <rmk+kernel@armlinux.org.uk>
Signed-off-by: Sasha Levin <sashal@kernel.org>

tracing: Fix very unlikely race of registering two stat tracers

[ Upstream commit dfb6cd1e654315168e36d947471bd2a0ccd834ae ]

Looking through old emails in my INBOX, I came across a patch from Luis
Henriques that attempted to fix a race of two stat tracers registering the
same stat trace (extremely unlikely, as this is done in the kernel, and
probably doesn't even exist). The submitted patch wasn't quite right as it
needed to deal with clean up a bit better (if two stat tracers were the
same, it would have the same files).

But to make the code cleaner, all we needed to do is to keep the
all_stat_sessions_mutex held for most of the registering function.

Link: http://lkml.kernel.org/r/1410299375-20068-1-git-send-email-luis.henriques@canonical.com
Fixes: 002bb86d8d42f ("tracing/ftrace: separate events tracing and stats tracing engine")
Reported-by: Luis Henriques <luis.henriques@canonical.com>
Signed-off-by: Steven Rostedt (VMware) <rostedt@goodmis.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

tracing: Fix tracing_stat return values in error handling paths

[ Upstream commit afccc00f75bbbee4e4ae833a96c2d29a7259c693 ]

tracing_stat_init() was always returning '0', even on the error paths.  It
now returns -ENODEV if tracing_init_dentry() fails or -ENOMEM if it fails
to created the 'trace_stat' debugfs directory.

Link: http://lkml.kernel.org/r/1410299381-20108-1-git-send-email-luis.henriques@canonical.com
Fixes: ed6f1c996bfe4 ("tracing: Check return value of tracing_init_dentry()")
Signed-off-by: Luis Henriques <luis.henriques@canonical.com>
[ Pulled from the archeological digging of my INBOX ]
Signed-off-by: Steven Rostedt (VMware) <rostedt@goodmis.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

powerpc/iov: Move VF pdev fixup into pcibios_fixup_iov()

[ Upstream commit 965c94f309be58fbcc6c8d3e4f123376c5970d79 ]

An ioda_pe for each VF is allocated in pnv_pci_sriov_enable() before
the pci_dev for the VF is created. We need to set the pe->pdev pointer
at some point after the pci_dev is created. Currently we do that in:

pcibios_bus_add_device()
pnv_pci_dma_dev_setup() (via phb->ops.dma_dev_setup)
/* fixup is done here */
pnv_pci_ioda_dma_dev_setup() (via pnv_phb->dma_dev_setup)

The fixup needs to be done before setting up DMA for for the VF's PE,
but there's no real reason to delay it until this point. Move the
fixup into pnv_pci_ioda_fixup_iov() so the ordering is:

pcibios_add_device()
pnv_pci_ioda_fixup_iov() (via ppc_md.pcibios_fixup_sriov)

pcibios_bus_add_device()
...

This isn't strictly required, but it's a slightly more logical place
to do the fixup and it simplifies pnv_pci_dma_dev_setup().

Signed-off-by: Oliver O'Halloran <oohall@gmail.com>
Reviewed-by: Alexey Kardashevskiy <aik@ozlabs.ru>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20200110070207.439-4-oohall@gmail.com
Signed-off-by: Sasha Levin <sashal@kernel.org>

s390/pci: Fix possible deadlock in recover_store()

[ Upstream commit 576c75e36c689bec6a940e807bae27291ab0c0de ]

With zpci_disable() working, lockdep detected a potential deadlock
(lockdep output at the end).

The deadlock is between recovering a PCI function via the

/sys/bus/pci/devices/<dev>/recover

attribute vs powering it off via

/sys/bus/pci/slots/<slot>/power.

The fix is analogous to the changes in commit 0ee223b2e1f6 ("scsi: core:
Avoid that SCSI device removal through sysfs triggers a deadlock")
that fixed a potential deadlock on removing a SCSI device via sysfs.

[  204.830107] ======================================================
[  204.830109] WARNING: possible circular locking dependency detected
[  204.830111] 5.5.0-rc2-06072-gbc03ecc9a672 #6 Tainted: G        W
[  204.830112] ------------------------------------------------------
[  204.830113] bash/1034 is trying to acquire lock:
[  204.830115] 0000000192a1a610 (kn->count#200){++++}, at: kernfs_remove_by_name_ns+0x5c/0xa8
[  204.830122]
               but task is already holding lock:
[  204.830123] 00000000c16134a8 (pci_rescan_remove_lock){+.+.}, at: pci_stop_and_remove_bus_device_locked+0x26/0x48
[  204.830128]
               which lock already depends on the new lock.

[  204.830129]
               the existing dependency chain (in reverse order) is:
[  204.830130]
               -> #1 (pci_rescan_remove_lock){+.+.}:
[  204.830134]        validate_chain+0x93a/0xd08
[  204.830136]        __lock_acquire+0x4ae/0x9d0
[  204.830137]        lock_acquire+0x114/0x280
[  204.830140]        __mutex_lock+0xa2/0x960
[  204.830142]        mutex_lock_nested+0x32/0x40
[  204.830145]        recover_store+0x4c/0xa8
[  204.830147]        kernfs_fop_write+0xe6/0x218
[  204.830151]        vfs_write+0xb0/0x1b8
[  204.830152]        ksys_write+0x6c/0xf8
[  204.830154]        system_call+0xd8/0x2d8
[  204.830155]
               -> #0 (kn->count#200){++++}:
[  204.830187]        check_noncircular+0x1e6/0x240
[  204.830189]        check_prev_add+0xfc/0xdb0
[  204.830190]        validate_chain+0x93a/0xd08
[  204.830192]        __lock_acquire+0x4ae/0x9d0
[  204.830193]        lock_acquire+0x114/0x280
[  204.830194]        __kernfs_remove.part.0+0x2e4/0x360
[  204.830196]        kernfs_remove_by_name_ns+0x5c/0xa8
[  204.830198]        remove_files.isra.0+0x4c/0x98
[  204.830199]        sysfs_remove_group+0x66/0xc8
[  204.830201]        sysfs_remove_groups+0x46/0x68
[  204.830204]        device_remove_attrs+0x52/0x90
[  204.830207]        device_del+0x182/0x418
[  204.830208]        pci_remove_bus_device+0x8a/0x130
[  204.830210]        pci_stop_and_remove_bus_device_locked+0x3a/0x48
[  204.830212]        disable_slot+0x68/0x100
[  204.830213]        power_write_file+0x7c/0x130
[  204.830215]        kernfs_fop_write+0xe6/0x218
[  204.830217]        vfs_write+0xb0/0x1b8
[  204.830218]        ksys_write+0x6c/0xf8
[  204.830220]        system_call+0xd8/0x2d8
[  204.830221]
               other info that might help us debug this:

[  204.830223]  Possible unsafe locking scenario:

[  204.830224]        CPU0                    CPU1
[  204.830225]        ----                    ----
[  204.830226]   lock(pci_rescan_remove_lock);
[  204.830227]                                lock(kn->count#200);
[  204.830229]                                lock(pci_rescan_remove_lock);
[  204.830231]   lock(kn->count#200);
[  204.830233]
                *** DEADLOCK ***

[  204.830234] 4 locks held by bash/1034:
[  204.830235]  #0: 00000001b6fbc498 (sb_writers#4){.+.+}, at: vfs_write+0x158/0x1b8
[  204.830239]  #1: 000000018c9f5090 (&of->mutex){+.+.}, at: kernfs_fop_write+0xaa/0x218
[  204.830242]  #2: 00000001f7da0810 (kn->count#235){.+.+}, at: kernfs_fop_write+0xb6/0x218
[  204.830245]  #3: 00000000c16134a8 (pci_rescan_remove_lock){+.+.}, at: pci_stop_and_remove_bus_device_locked+0x26/0x48
[  204.830248]
               stack backtrace:
[  204.830250] CPU: 2 PID: 1034 Comm: bash Tainted: G        W         5.5.0-rc2-06072-gbc03ecc9a672 #6
[  204.830252] Hardware name: IBM 8561 T01 703 (LPAR)
[  204.830253] Call Trace:
[  204.830257]  [<00000000c05e10c0>] show_stack+0x88/0xf0
[  204.830260]  [<00000000c112dca4>] dump_stack+0xa4/0xe0
[  204.830261]  [<00000000c0694c06>] check_noncircular+0x1e6/0x240
[  204.830263]  [<00000000c0695bec>] check_prev_add+0xfc/0xdb0
[  204.830264]  [<00000000c06971da>] validate_chain+0x93a/0xd08
[  204.830266]  [<00000000c06994c6>] __lock_acquire+0x4ae/0x9d0
[  204.830267]  [<00000000c069867c>] lock_acquire+0x114/0x280
[  204.830269]  [<00000000c09ca15c>] __kernfs_remove.part.0+0x2e4/0x360
[  204.830270]  [<00000000c09cb5c4>] kernfs_remove_by_name_ns+0x5c/0xa8
[  204.830272]  [<00000000c09cee14>] remove_files.isra.0+0x4c/0x98
[  204.830274]  [<00000000c09cf2ae>] sysfs_remove_group+0x66/0xc8
[  204.830276]  [<00000000c09cf356>] sysfs_remove_groups+0x46/0x68
[  204.830278]  [<00000000c0e3dfe2>] device_remove_attrs+0x52/0x90
[  204.830280]  [<00000000c0e40382>] device_del+0x182/0x418
[  204.830281]  [<00000000c0dcfd7a>] pci_remove_bus_device+0x8a/0x130
[  204.830283]  [<00000000c0dcfe92>] pci_stop_and_remove_bus_device_locked+0x3a/0x48
[  204.830285]  [<00000000c0de7190>] disable_slot+0x68/0x100
[  204.830286]  [<00000000c0de6514>] power_write_file+0x7c/0x130
[  204.830288]  [<00000000c09cc846>] kernfs_fop_write+0xe6/0x218
[  204.830290]  [<00000000c08f3480>] vfs_write+0xb0/0x1b8
[  204.830291]  [<00000000c08f378c>] ksys_write+0x6c/0xf8
[  204.830293]  [<00000000c1154374>] system_call+0xd8/0x2d8
[  204.830294] INFO: lockdep is turned off.

Signed-off-by: Niklas Schnelle <schnelle@linux.ibm.com>
Reviewed-by: Peter Oberparleiter <oberpar@linux.ibm.com>
Signed-off-by: Vasily Gorbik <gor@linux.ibm.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

wan/hdlc_x25: fix skb handling

[ Upstream commit 953c4a08dfc9ffe763a8340ac10f459d6c6cc4eb ]

o call skb_reset_network_header() before hdlc->xmit()
 o change skb proto to HDLC (0x0019) before hdlc->xmit()
 o call dev_queue_xmit_nit() before hdlc->xmit()

This changes make it possible to trace (tcpdump) outgoing layer2
(ETH_P_HDLC) packets

Additionally call skb_reset_network_header() after each skb_push() /
skb_pull().

Signed-off-by: Martin Schiller <ms@dev.tdt.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Sasha Levin <sashal@kernel.org>

dmaengine: fsl-qdma: fix duplicated argument to &&

[ Upstream commit 4b048178854da11656596d36a107577d66fd1e08 ]

There is duplicated argument to && in function fsl_qdma_free_chan_resources,
which looks like a typo, pointer fsl_queue->desc_pool also needs NULL check,
fix it.
Detected with coccinelle.

Fixes: b092529e0aa0 ("dmaengine: fsl-qdma: Add qDMA controller driver for Layerscape SoCs")
Signed-off-by: Chen Zhou <chenzhou10@huawei.com>
Reviewed-by: Peng Ma <peng.ma@nxp.com>
Tested-by: Peng Ma <peng.ma@nxp.com>
Link: https://lore.kernel.org/r/20200120125843.34398-1-chenzhou10@huawei.com
Signed-off-by: Vinod Koul <vkoul@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

udf: Allow writing to 'Rewritable' partitions

[ Upstream commit 15fb05fd286ac57a0802d71624daeb5c1c2d5b07 ]

UDF 2.60 standard states in section 2.2.14.2:

    A partition with Access Type 3 (rewritable) shall define a Freed
    Space Bitmap or a Freed Space Table, see 2.3.3. All other partitions
    shall not define a Freed Space Bitmap or a Freed Space Table.

    Rewritable partitions are used on media that require some form of
    preprocessing before re-writing data (for example legacy MO). Such
    partitions shall use Access Type 3.

    Overwritable partitions are used on media that do not require
    preprocessing before overwriting data (for example: CD-RW, DVD-RW,
    DVD+RW, DVD-RAM, BD-RE, HD DVD-Rewritable). Such partitions shall
    use Access Type 4.

however older versions of the standard didn't have this wording and
there are tools out there that create UDF filesystems with rewritable
partitions but that don't contain a Freed Space Bitmap or a Freed Space
Table on media that does not require pre-processing before overwriting a
block. So instead of forcing media with rewritable partition read-only,
base this decision on presence of a Freed Space Bitmap or a Freed Space
Table.

Reported-by: Pali Rohár <pali.rohar@gmail.com>
Reviewed-by: Pali Rohár <pali.rohar@gmail.com>
Fixes: b085fbe2ef7f ("udf: Fix crash during mount")
Link: https://lore.kernel.org/linux-fsdevel/20200112144735.hj2emsoy4uwsouxz@pali
Signed-off-by: Jan Kara <jack@suse.cz>
Signed-off-by: Sasha Levin <sashal@kernel.org>

pwm: omap-dmtimer: Simplify error handling

[ Upstream commit c4cf7aa57eb83b108d2d9c6c37c143388fee2a4d ]

Instead of doing error handling in the middle of ->probe(), move error
handling and freeing the reference to timer to the end.

This fixes a resource leak as dm_timer wasn't freed when allocating
*omap failed.

Implementation note: The put: label was never reached without a goto and
ret being unequal to 0, so the removed return statement is fine.

Fixes: 6604c6556db9 ("pwm: Add PWM driver for OMAP using dual-mode timers")
Signed-off-by: Uwe Kleine-König <u.kleine-koenig@pengutronix.de>
Signed-off-by: Thierry Reding <thierry.reding@gmail.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

x86/sysfb: Fix check for bad VRAM size

[ Upstream commit dacc9092336be20b01642afe1a51720b31f60369 ]

When checking whether the reported lfb_size makes sense, the height
* stride result is page-aligned before seeing whether it exceeds the
reported size.

This doesn't work if height * stride is not an exact number of pages.
For example, as reported in the kernel bugzilla below, an 800x600x32 EFI
framebuffer gets skipped because of this.

Move the PAGE_ALIGN to after the check vs size.

Reported-by: Christopher Head <chead@chead.ca>
Tested-by: Christopher Head <chead@chead.ca>
Signed-off-by: Arvind Sankar <nivedita@alum.mit.edu>
Signed-off-by: Borislav Petkov <bp@suse.de>
Link: https://bugzilla.kernel.org/show_bug.cgi?id=206051
Link: https://lkml.kernel.org/r/20200107230410.2291947-1-nivedita@alum.mit.edu
Signed-off-by: Sasha Levin <sashal@kernel.org>

clk: ti: dra7: fix parent for gmac_clkctrl

[ Upstream commit 69e300283796dae7e8c2e6acdabcd31336c0c93e ]

The parent clk for gmac clk ctrl has to be gmac_main_clk (125MHz) instead
of dpll_gmac_ck (1GHz). This is caused incorrect CPSW MDIO operation.
Hence, fix it.

Fixes: dffa9051d546 ('clk: ti: dra7: add new clkctrl data')
Signed-off-by: Grygorii Strashko <grygorii.strashko@ti.com>
Signed-off-by: Tero Kristo <t-kristo@ti.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

ext4: fix deadlock allocating bio_post_read_ctx from mempool

[ Upstream commit 68e45330e341dad2d3a0a3f8ef2ec46a2a0a3bbc ]

Without any form of coordination, any case where multiple allocations
from the same mempool are needed at a time to make forward progress can
deadlock under memory pressure.

This is the case for struct bio_post_read_ctx, as one can be allocated
to decrypt a Merkle tree page during fsverity_verify_bio(), which itself
is running from a post-read callback for a data bio which has its own
struct bio_post_read_ctx.

Fix this by freeing the first bio_post_read_ctx before calling
fsverity_verify_bio().  This works because verity (if enabled) is always
the last post-read step.

This deadlock can be reproduced by trying to read from an encrypted
verity file after reducing NUM_PREALLOC_POST_READ_CTXS to 1 and patching
mempool_alloc() to pretend that pool->alloc() always fails.

Note that since NUM_PREALLOC_POST_READ_CTXS is actually 128, to actually
hit this bug in practice would require reading from lots of encrypted
verity files at the same time.  But it's theoretically possible, as N
available objects isn't enough to guarantee forward progress when > N/2
threads each need 2 objects at a time.

Fixes: 22cfe4b48ccb ("ext4: add fs-verity read support")
Signed-off-by: Eric Biggers <ebiggers@google.com>
Link: https://lore.kernel.org/r/20191231181222.47684-1-ebiggers@kernel.org
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
Signed-off-by: Sasha Levin <sashal@kernel.org>

jbd2: clear JBD2_ABORT flag before journal_reset to update log tail info when load journal

[ Upstream commit a09decff5c32060639a685581c380f51b14e1fc2 ]

If the journal is dirty when the filesystem is mounted, jbd2 will replay
the journal but the journal superblock will not be updated by
journal_reset() because JBD2_ABORT flag is still set (it was set in
journal_init_common()). This is problematic because when a new transaction
is then committed, it will be recorded in block 1 (journal->j_tail was set
to 1 in journal_reset()). If unclean shutdown happens again before the
journal superblock is updated, the new recorded transaction will not be
replayed during the next mount (because of stale sb->s_start and
sb->s_sequence values) which can lead to filesystem corruption.

Fixes: 85e0c4e89c1b ("jbd2: if the journal is aborted then don't allow update of the log tail")
Signed-off-by: Kai Li <li.kai4@h3c.com>
Link: https://lore.kernel.org/r/20200111022542.5008-1-li.kai4@h3c.com
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
Signed-off-by: Sasha Levin <sashal@kernel.org>

kselftest: Minimise dependency of get_size on C library interfaces

[ Upstream commit 6b64a650f0b2ae3940698f401732988699eecf7a ]

It was observed[1] on arm64 that __builtin_strlen led to an infinite
loop in the get_size selftest.  This is because __builtin_strlen (and
other builtins) may sometimes result in a call to the C library
function.  The C library implementation of strlen uses an IFUNC
resolver to load the most efficient strlen implementation for the
underlying machine and hence has a PLT indirection even for static
binaries.  Because this binary avoids the C library startup routines,
the PLT initialization never happens and hence the program gets stuck
in an infinite loop.

On x86_64 the __builtin_strlen just happens to expand inline and avoid
the call but that is not always guaranteed.

Further, while testing on x86_64 (Fedora 31), it was observed that the
test also failed with a segfault inside write() because the generated
code for the write function in glibc seems to access TLS before the
syscall (probably due to the cancellation point check) and fails
because TLS is not initialised.

To mitigate these problems, this patch reduces the interface with the
C library to just the syscall function.  The syscall function still
sets errno on failure, which is undesirable but for now it only
affects cases where syscalls fail.

[1] https://bugs.linaro.org/show_bug.cgi?id=5479

Signed-off-by: Siddhesh Poyarekar <siddhesh@gotplt.org>
Reported-by: Masami Hiramatsu <masami.hiramatsu@linaro.org>
Tested-by: Masami Hiramatsu <masami.hiramatsu@linaro.org>
Reviewed-by: Tim Bird <tim.bird@sony.com>
Signed-off-by: Shuah Khan <skhan@linuxfoundation.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

drm/amd/display: Clear state after exiting fixed active VRR state

[ Upstream commit 6f8f76444baf405bacb0591d97549a71a9aaa1ac ]

[why]
Upon exiting a fixed active VRR state, the state isn't cleared. This
leads to the variable VRR range to be calculated incorrectly.

[how]
Set fixed active state to false when updating vrr params

Signed-off-by: Amanda Liu <amanda.liu@amd.com>
Reviewed-by: Anthony Koo <Anthony.Koo@amd.com>
Acked-by: Harry Wentland <harry.wentland@amd.com>
Acked-by: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

clocksource/drivers/bcm2835_timer: Fix memory leak of timer

[ Upstream commit 2052d032c06761330bca4944bb7858b00960e868 ]

Currently when setup_irq fails the error exit path will leak the
recently allocated timer structure.  Originally the code would
throw a panic but a later commit changed the behaviour to return
via the err_iounmap path and hence we now have a memory leak. Fix
this by adding a err_timer_free error path that kfree's timer.

Addresses-Coverity: ("Resource Leak")
Fixes: 524a7f08983d ("clocksource/drivers/bcm2835_timer: Convert init function to return error")
Signed-off-by: Colin Ian King <colin.king@canonical.com>
Signed-off-by: Daniel Lezcano <daniel.lezcano@linaro.org>
Link: https://lore.kernel.org/r/20191219213246.34437-1-colin.king@canonical.com
Signed-off-by: Sasha Levin <sashal@kernel.org>

usb: dwc2: Fix IN FIFO allocation

[ Upstream commit 644139f8b64d818f6345351455f14471510879a5 ]

On chips with fewer FIFOs than endpoints (for example RK3288 which has 9
endpoints, but only 6 which are cabable of input), the DPTXFSIZN
registers above the FIFO count may return invalid values.

With logging added on startup, I see:

dwc2 ff580000.usb: dwc2_hsotg_init_fifo: ep=1 sz=256
dwc2 ff580000.usb: dwc2_hsotg_init_fifo: ep=2 sz=128
dwc2 ff580000.usb: dwc2_hsotg_init_fifo: ep=3 sz=128
dwc2 ff580000.usb: dwc2_hsotg_init_fifo: ep=4 sz=64
dwc2 ff580000.usb: dwc2_hsotg_init_fifo: ep=5 sz=64
dwc2 ff580000.usb: dwc2_hsotg_init_fifo: ep=6 sz=32
dwc2 ff580000.usb: dwc2_hsotg_init_fifo: ep=7 sz=0
dwc2 ff580000.usb: dwc2_hsotg_init_fifo: ep=8 sz=0
dwc2 ff580000.usb: dwc2_hsotg_init_fifo: ep=9 sz=0
dwc2 ff580000.usb: dwc2_hsotg_init_fifo: ep=10 sz=0
dwc2 ff580000.usb: dwc2_hsotg_init_fifo: ep=11 sz=0
dwc2 ff580000.usb: dwc2_hsotg_init_fifo: ep=12 sz=0
dwc2 ff580000.usb: dwc2_hsotg_init_fifo: ep=13 sz=0
dwc2 ff580000.usb: dwc2_hsotg_init_fifo: ep=14 sz=0
dwc2 ff580000.usb: dwc2_hsotg_init_fifo: ep=15 sz=0

but:

# cat /sys/kernel/debug/ff580000.usb/fifo
Non-periodic FIFOs:
RXFIFO: Size 275
NPTXFIFO: Size 16, Start 0x00000113

Periodic TXFIFOs:
DPTXFIFO 1: Size 256, Start 0x00000123
DPTXFIFO 2: Size 128, Start 0x00000223
DPTXFIFO 3: Size 128, Start 0x000002a3
DPTXFIFO 4: Size 64, Start 0x00000323
DPTXFIFO 5: Size 64, Start 0x00000363
DPTXFIFO 6: Size 32, Start 0x000003a3
DPTXFIFO 7: Size 0, Start 0x000003e3
DPTXFIFO 8: Size 0, Start 0x000003a3
DPTXFIFO 9: Size 256, Start 0x00000123

so it seems that FIFO 9 is mirroring FIFO 1.

Fix the allocation by using the FIFO count instead of the endpoint count
when selecting a FIFO for an endpoint.

Acked-by: Minas Harutyunyan <hminas@synopsys.com>
Signed-off-by: John Keeping <john@metanate.com>
Signed-off-by: Felipe Balbi <balbi@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

usb: gadget: udc: fix possible sleep-in-atomic-context bugs in gr_probe()

[ Upstream commit 9c1ed62ae0690dfe5d5e31d8f70e70a95cb48e52 ]

The driver may sleep while holding a spinlock.
The function call path (from bottom to top) in Linux 4.19 is:

drivers/usb/gadget/udc/core.c, 1175:
kzalloc(GFP_KERNEL) in usb_add_gadget_udc_release
drivers/usb/gadget/udc/core.c, 1272:
usb_add_gadget_udc_release in usb_add_gadget_udc
drivers/usb/gadget/udc/gr_udc.c, 2186:
usb_add_gadget_udc in gr_probe
drivers/usb/gadget/udc/gr_udc.c, 2183:
spin_lock in gr_probe

drivers/usb/gadget/udc/core.c, 1195:
mutex_lock in usb_add_gadget_udc_release
drivers/usb/gadget/udc/core.c, 1272:
usb_add_gadget_udc_release in usb_add_gadget_udc
drivers/usb/gadget/udc/gr_udc.c, 2186:
usb_add_gadget_udc in gr_probe
drivers/usb/gadget/udc/gr_udc.c, 2183:
spin_lock in gr_probe

drivers/usb/gadget/udc/gr_udc.c, 212:
debugfs_create_file in gr_probe
drivers/usb/gadget/udc/gr_udc.c, 2197:
gr_dfs_create in gr_probe
drivers/usb/gadget/udc/gr_udc.c, 2183:
    spin_lock in gr_probe

drivers/usb/gadget/udc/gr_udc.c, 2114:
devm_request_threaded_irq in gr_request_irq
drivers/usb/gadget/udc/gr_udc.c, 2202:
gr_request_irq in gr_probe
drivers/usb/gadget/udc/gr_udc.c, 2183:
    spin_lock in gr_probe

kzalloc(GFP_KERNEL), mutex_lock(), debugfs_create_file() and
devm_request_threaded_irq() can sleep at runtime.

To fix these possible bugs, usb_add_gadget_udc(), gr_dfs_create() and
gr_request_irq() are called without handling the spinlock.

These bugs are found by a static analysis tool STCheck written by myself.

Signed-off-by: Jia-Ju Bai <baijiaju1990@gmail.com>
Signed-off-by: Felipe Balbi <balbi@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

drm/nouveau/nouveau: fix incorrect sizeof on args.src an args.dst

[ Upstream commit f42e4b337b327b1336c978c4b5174990a25f68a0 ]

The sizeof is currently on args.src and args.dst and should be on
*args.src and *args.dst. Fortunately these sizes just so happen
to be the same size so it worked, however, this should be fixed
and it also cleans up static analysis warnings

Addresses-Coverity: ("sizeof not portable")
Fixes: f268307ec7c7 ("nouveau: simplify nouveau_dmem_migrate_vma")
Signed-off-by: Colin Ian King <colin.king@canonical.com>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

spi: fsl-lpspi: fix only one cs-gpio working

[ Upstream commit bc3a8b295e5bca9d1ec2622a6ba38289f9fd3d8a ]

Why it does not work at the moment:
- num_chipselect sets the number of cs-gpios that are in the DT.
  This comes from drivers/spi/spi.c
- num_chipselect gets set with devm_spi_register_controller, that is
  called in drivers/spi/spi.c
- devm_spi_register_controller got called after num_chipselect has
  been used.

How this commit fixes the issue:
- devm_spi_register_controller gets called before num_chipselect is
  being used.

Fixes: c7a402599504 ("spi: lpspi: use the core way to implement cs-gpio function")
Signed-off-by: Philippe Schenker <philippe.schenker@toradex.com>
Link: https://lore.kernel.org/r/20191204141312.1411251-1-philippe.schenker@toradex.com
Signed-off-by: Mark Brown <broonie@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

drm/amdgpu/sriov: workaround on rev_id for Navi12 under sriov

[ Upstream commit df5e984c8bd414561c320d6cbbb66d53abf4c7e2 ]

guest vm gets 0xffffffff when reading RCC_DEV0_EPF0_STRAP0,
as a consequence, the rev_id and external_rev_id are wrong.

workaround it by hardcoding the rev_id to 0, which is the default value.

v2. add comment in the code

Signed-off-by: Tiecheng Zhou <Tiecheng.Zhou@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

uio: fix a sleep-in-atomic-context bug in uio_dmem_genirq_irqcontrol()

[ Upstream commit b74351287d4bd90636c3f48bc188c2f53824c2d4 ]

The driver may sleep while holding a spinlock.
The function call path (from bottom to top) in Linux 4.19 is:

kernel/irq/manage.c, 523:
synchronize_irq in disable_irq
drivers/uio/uio_dmem_genirq.c, 140:
disable_irq in uio_dmem_genirq_irqcontrol
drivers/uio/uio_dmem_genirq.c, 134:
_raw_spin_lock_irqsave in uio_dmem_genirq_irqcontrol

synchronize_irq() can sleep at runtime.

To fix this bug, disable_irq() is called without holding the spinlock.

This bug is found by a static analysis tool STCheck written by myself.

Signed-off-by: Jia-Ju Bai <baijiaju1990@gmail.com>
Link: https://lore.kernel.org/r/20191218094405.6009-1-baijiaju1990@gmail.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

raid6/test: fix a compilation error

[ Upstream commit 6b8651aac1dca6140dd7fb4c9fec2736ed3f6223 ]

The compilation error is redeclaration showed as following:

        In file included from ../../../include/linux/limits.h:6,
                         from /usr/include/x86_64-linux-gnu/bits/local_lim.h:38,
                         from /usr/include/x86_64-linux-gnu/bits/posix1_lim.h:161,
                         from /usr/include/limits.h:183,
                         from /usr/lib/gcc/x86_64-linux-gnu/8/include-fixed/limits.h:194,
                         from /usr/lib/gcc/x86_64-linux-gnu/8/include-fixed/syslimits.h:7,
                         from /usr/lib/gcc/x86_64-linux-gnu/8/include-fixed/limits.h:34,
                         from ../../../include/linux/raid/pq.h:30,
                         from algos.c:14:
        ../../../include/linux/types.h:114:15: error: conflicting types for ‘int64_t’
         typedef s64   int64_t;
                       ^~~~~~~
        In file included from /usr/include/stdint.h:34,
                         from /usr/lib/gcc/x86_64-linux-gnu/8/include/stdint.h:9,
                         from /usr/include/inttypes.h:27,
                         from ../../../include/linux/raid/pq.h:29,
                         from algos.c:14:
        /usr/include/x86_64-linux-gnu/bits/stdint-intn.h:27:19: note: previous \
        declaration of ‘int64_t’ was here
         typedef __int64_t int64_t;

Fixes: 54d50897d544 ("linux/kernel.h: split *_MAX and *_MIN macros into <linux/limits.h>")
Signed-off-by: Zhengyuan Liu <liuzhengyuan@kylinos.cn>
Signed-off-by: Song Liu <songliubraving@fb.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

net: ethernet: ixp4xx: Standard module init

[ Upstream commit c83db9ef5640548631707e8b4a7bcddc115fdbae ]

The IXP4xx driver was initializing the MDIO bus before even
probing, in the callbacks supposed to be used for setting up
the module itself, and with the side effect of trying to
register the MDIO bus as soon as this module was loaded or
compiled into the kernel whether the device was discovered
or not.

This does not work with multiplatform environments.

To get rid of this: set up the MDIO bus from the probe()
callback and remove it in the remove() callback. Rename
the probe() and remove() calls to reflect the most common
conventions.

Since there is a bit of checking for the ethernet feature
to be present in the MDIO registering function, making the
whole module not even be registered if we can't find an
MDIO bus, we need something similar: register the MDIO
bus when the corresponding ethernet is probed, and
return -EPROBE_DEFER on the other interfaces until this
happens. If no MDIO bus is present on any of the
registered interfaces we will eventually bail out.

None of the platforms I've seen has e.g. MDIO on EthB
and only uses EthC, there is always a Ethernet hardware
on the NPE (B, C) that has the MDIO bus, we just might
have to wait for it.

Signed-off-by: Linus Walleij <linus.walleij@linaro.org>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

sparc: Add .exit.data section.

[ Upstream commit 548f0b9a5f4cffa0cecf62eb12aa8db682e4eee6 ]

This fixes build errors of all sorts.

Also, emit .exit.text unconditionally.

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

MIPS: Loongson: Fix potential NULL dereference in loongson3_platform_init()

[ Upstream commit 72d052e28d1d2363f9107be63ef3a3afdea6143c ]

If kzalloc fails, it should return -ENOMEM, otherwise may trigger a NULL
pointer dereference.

Fixes: 3adeb2566b9b ("MIPS: Loongson: Improve LEFI firmware interface")
Signed-off-by: Tiezhu Yang <yangtiezhu@loongson.cn>
Signed-off-by: Paul Burton <paulburton@kernel.org>
Cc: Ralf Baechle <ralf@linux-mips.org>
Cc: Huacai Chen <chenhc@lemote.com>
Cc: Jiaxun Yang <jiaxun.yang@flygoat.com>
Cc: linux-mips@vger.kernel.org
Cc: linux-kernel@vger.kernel.org
Signed-off-by: Sasha Levin <sashal@kernel.org>

efi/x86: Map the entire EFI vendor string before copying it

[ Upstream commit ffc2760bcf2dba0dbef74013ed73eea8310cc52c ]

Fix a couple of issues with the way we map and copy the vendor string:
- we map only 2 bytes, which usually works since you get at least a
  page, but if the vendor string happens to cross a page boundary,
  a crash will result
- only call early_memunmap() if early_memremap() succeeded, or we will
  call it with a NULL address which it doesn't like,
- while at it, switch to early_memremap_ro(), and array indexing rather
  than pointer dereferencing to read the CHAR16 characters.

Signed-off-by: Ard Biesheuvel <ardb@kernel.org>
Cc: Andy Lutomirski <luto@kernel.org>
Cc: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Cc: Arvind Sankar <nivedita@alum.mit.edu>
Cc: Matthew Garrett <mjg59@google.com>
Cc: linux-efi@vger.kernel.org
Fixes: 5b83683f32b1 ("x86: EFI runtime service support")
Link: https://lkml.kernel.org/r/20200103113953.9571-5-ardb@kernel.org
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

pinctrl: baytrail: Do not clear IRQ flags on direct-irq enabled pins

[ Upstream commit a23680594da7a9e2696dbcf4f023e9273e2fa40b ]

Suspending Goodix touchscreens requires changing the interrupt pin to
output before sending them a power-down command. Followed by wiggling
the interrupt pin to wake the device up, after which it is put back
in input mode.

On Bay Trail devices with a Goodix touchscreen direct-irq mode is used
in combination with listing the pin as a normal GpioIo resource.

This works fine, until the goodix driver gets rmmod-ed and then insmod-ed
again. In this case byt_gpio_disable_free() calls
byt_gpio_clear_triggering() which clears the IRQ flags and after that the
(direct) IRQ no longer triggers.

This commit fixes this by adding a check for the BYT_DIRECT_IRQ_EN flag
to byt_gpio_clear_triggering().

Note that byt_gpio_clear_triggering() only gets called from
byt_gpio_disable_free() for direct-irq enabled pins, as these are excluded
from the irq_valid mask by byt_init_irq_valid_mask().

Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Acked-by: Mika Westerberg <mika.westerberg@linux.intel.com>
Reviewed-by: Linus Walleij <linus.walleij@linaro.org>
Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

IB/core: Let IB core distribute cache update events

[ Upstream commit 6b57cea9221b0247ad5111b348522625e489a8e4 ]

Currently when the low level driver notifies Pkey, GID, and port change
events they are notified to the registered handlers in the order they are
registered.

IB core and other ULPs such as IPoIB are interested in GID, LID, Pkey
change events.

Since all GID queries done by ULPs are serviced by IB core, and the IB
core deferes cache updates to a work queue, it is possible for other
clients to see stale cache data when they handle their own events.

For example, the below call tree shows how ipoib will call
rdma_query_gid() concurrently with the update to the cache sitting in the
WQ.

mlx5_ib_handle_event()
  ib_dispatch_event()
    ib_cache_event()
       queue_work() -> slow cache update

    [..]
    ipoib_event()
     queue_work()
       [..]
       work handler
         ipoib_ib_dev_flush_light()
           __ipoib_ib_dev_flush()
              ipoib_dev_addr_changed_valid()
                rdma_query_gid() <- Returns old GID, cache not updated.

Move all the event dispatch to a work queue so that the cache update is
always done before any clients are notified.

Fixes: f35faa4ba956 ("IB/core: Simplify ib_query_gid to always refer to cache")
Link: https://lore.kernel.org/r/20191212113024.336702-3-leon@kernel.org
Signed-off-by: Parav Pandit <parav@mellanox.com>
Signed-off-by: Leon Romanovsky <leonro@mellanox.com>
Reviewed-by: Jason Gunthorpe <jgg@mellanox.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

kernel/module: Fix memleak in module_add_modinfo_attrs()

[ Upstream commit f6d061d617124abbd55396a3bc37b9bf7d33233c ]

In module_add_modinfo_attrs() if sysfs_create_file() fails
on the first iteration of the loop (so i = 0), we forget to
free the modinfo_attrs.

Fixes: bc6f2a757d52 ("kernel/module: Fix mem leak in module_add_modinfo_attrs")
Reviewed-by: Miroslav Benes <mbenes@suse.cz>
Signed-off-by: YueHaibing <yuehaibing@huawei.com>
Signed-off-by: Jessica Yu <jeyu@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

media: sti: bdisp: fix a possible sleep-in-atomic-context bug in bdisp_device_run()

[ Upstream commit bb6d42061a05d71dd73f620582d9e09c8fbf7f5b ]

The driver may sleep while holding a spinlock.
The function call path (from bottom to top) in Linux 4.19 is:

drivers/media/platform/sti/bdisp/bdisp-hw.c, 385:
    msleep in bdisp_hw_reset
drivers/media/platform/sti/bdisp/bdisp-v4l2.c, 341:
    bdisp_hw_reset in bdisp_device_run
drivers/media/platform/sti/bdisp/bdisp-v4l2.c, 317:
    _raw_spin_lock_irqsave in bdisp_device_run

To fix this bug, msleep() is replaced with udelay().

This bug is found by a static analysis tool STCheck written by myself.

Signed-off-by: Jia-Ju Bai <baijiaju1990@gmail.com>
Reviewed-by: Fabien Dessenne <fabien.dessenne@st.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

char/random: silence a lockdep splat with printk()

[ Upstream commit 1b710b1b10eff9d46666064ea25f079f70bc67a8 ]

Sergey didn't like the locking order,

uart_port->lock  ->  tty_port->lock

uart_write (uart_port->lock)
  __uart_start
    pl011_start_tx
      pl011_tx_chars
        uart_write_wakeup
          tty_port_tty_wakeup
            tty_port_default
              tty_port_tty_get (tty_port->lock)

but those code is so old, and I have no clue how to de-couple it after
checking other locks in the splat. There is an onging effort to make all
printk() as deferred, so until that happens, workaround it for now as a
short-term fix.

LTP: starting iogen01 (export LTPROOT; rwtest -N iogen01 -i 120s -s
read,write -Da -Dv -n 2 500b:$TMPDIR/doio.f1.$$
1000b:$TMPDIR/doio.f2.$$)
WARNING: possible circular locking dependency detected
------------------------------------------------------
doio/49441 is trying to acquire lock:
ffff008b7cff7290 (&(&zone->lock)->rlock){..-.}, at: rmqueue+0x138/0x2050

but task is already holding lock:
60ff000822352818 (&pool->lock/1){-.-.}, at: start_flush_work+0xd8/0x3f0

  which lock already depends on the new lock.

  the existing dependency chain (in reverse order) is:

  -> #4 (&pool->lock/1){-.-.}:
       lock_acquire+0x320/0x360
       _raw_spin_lock+0x64/0x80
       __queue_work+0x4b4/0xa10
       queue_work_on+0xac/0x11c
       tty_schedule_flip+0x84/0xbc
       tty_flip_buffer_push+0x1c/0x28
       pty_write+0x98/0xd0
       n_tty_write+0x450/0x60c
       tty_write+0x338/0x474
       __vfs_write+0x88/0x214
       vfs_write+0x12c/0x1a4
       redirected_tty_write+0x90/0xdc
       do_loop_readv_writev+0x140/0x180
       do_iter_write+0xe0/0x10c
       vfs_writev+0x134/0x1cc
       do_writev+0xbc/0x130
       __arm64_sys_writev+0x58/0x8c
       el0_svc_handler+0x170/0x240
       el0_sync_handler+0x150/0x250
       el0_sync+0x164/0x180

  -> #3 (&(&port->lock)->rlock){-.-.}:
       lock_acquire+0x320/0x360
       _raw_spin_lock_irqsave+0x7c/0x9c
       tty_port_tty_get+0x24/0x60
       tty_port_default_wakeup+0x1c/0x3c
       tty_port_tty_wakeup+0x34/0x40
       uart_write_wakeup+0x28/0x44
       pl011_tx_chars+0x1b8/0x270
       pl011_start_tx+0x24/0x70
       __uart_start+0x5c/0x68
       uart_write+0x164/0x1c8
       do_output_char+0x33c/0x348
       n_tty_write+0x4bc/0x60c
       tty_write+0x338/0x474
       redirected_tty_write+0xc0/0xdc
       do_loop_readv_writev+0x140/0x180
       do_iter_write+0xe0/0x10c
       vfs_writev+0x134/0x1cc
       do_writev+0xbc/0x130
       __arm64_sys_writev+0x58/0x8c
       el0_svc_handler+0x170/0x240
       el0_sync_handler+0x150/0x250
       el0_sync+0x164/0x180

  -> #2 (&port_lock_key){-.-.}:
       lock_acquire+0x320/0x360
       _raw_spin_lock+0x64/0x80
       pl011_console_write+0xec/0x2cc
       console_unlock+0x794/0x96c
       vprintk_emit+0x260/0x31c
       vprintk_default+0x54/0x7c
       vprintk_func+0x218/0x254
       printk+0x7c/0xa4
       register_console+0x734/0x7b0
       uart_add_one_port+0x734/0x834
       pl011_register_port+0x6c/0xac
       sbsa_uart_probe+0x234/0x2ec
       platform_drv_probe+0xd4/0x124
       really_probe+0x250/0x71c
       driver_probe_device+0xb4/0x200
       __device_attach_driver+0xd8/0x188
       bus_for_each_drv+0xbc/0x110
       __device_attach+0x120/0x220
       device_initial_probe+0x20/0x2c
       bus_probe_device+0x54/0x100
       device_add+0xae8/0xc2c
       platform_device_add+0x278/0x3b8
       platform_device_register_full+0x238/0x2ac
       acpi_create_platform_device+0x2dc/0x3a8
       acpi_bus_attach+0x390/0x3cc
       acpi_bus_attach+0x108/0x3cc
       acpi_bus_attach+0x108/0x3cc
       acpi_bus_attach+0x108/0x3cc
       acpi_bus_scan+0x7c/0xb0
       acpi_scan_init+0xe4/0x304
       acpi_init+0x100/0x114
       do_one_initcall+0x348/0x6a0
       do_initcall_level+0x190/0x1fc
       do_basic_setup+0x34/0x4c
       kernel_init_freeable+0x19c/0x260
       kernel_init+0x18/0x338
       ret_from_fork+0x10/0x18

  -> #1 (console_owner){-...}:
       lock_acquire+0x320/0x360
       console_lock_spinning_enable+0x6c/0x7c
       console_unlock+0x4f8/0x96c
       vprintk_emit+0x260/0x31c
       vprintk_default+0x54/0x7c
       vprintk_func+0x218/0x254
       printk+0x7c/0xa4
       get_random_u64+0x1c4/0x1dc
       shuffle_pick_tail+0x40/0xac
       __free_one_page+0x424/0x710
       free_one_page+0x70/0x120
       __free_pages_ok+0x61c/0xa94
       __free_pages_core+0x1bc/0x294
       memblock_free_pages+0x38/0x48
       __free_pages_memory+0xcc/0xfc
       __free_memory_core+0x70/0x78
       free_low_memory_core_early+0x148/0x18c
       memblock_free_all+0x18/0x54
       mem_init+0xb4/0x17c
       mm_init+0x14/0x38
       start_kernel+0x19c/0x530

  -> #0 (&(&zone->lock)->rlock){..-.}:
       validate_chain+0xf6c/0x2e2c
       __lock_acquire+0x868/0xc2c
       lock_acquire+0x320/0x360
       _raw_spin_lock+0x64/0x80
       rmqueue+0x138/0x2050
       get_page_from_freelist+0x474/0x688
       __alloc_pages_nodemask+0x3b4/0x18dc
       alloc_pages_current+0xd0/0xe0
       alloc_slab_page+0x2b4/0x5e0
       new_slab+0xc8/0x6bc
       ___slab_alloc+0x3b8/0x640
       kmem_cache_alloc+0x4b4/0x588
       __debug_object_init+0x778/0x8b4
       debug_object_init_on_stack+0x40/0x50
       start_flush_work+0x16c/0x3f0
       __flush_work+0xb8/0x124
       flush_work+0x20/0x30
       xlog_cil_force_lsn+0x88/0x204 [xfs]
       xfs_log_force_lsn+0x128/0x1b8 [xfs]
       xfs_file_fsync+0x3c4/0x488 [xfs]
       vfs_fsync_range+0xb0/0xd0
       generic_write_sync+0x80/0xa0 [xfs]
       xfs_file_buffered_aio_write+0x66c/0x6e4 [xfs]
       xfs_file_write_iter+0x1a0/0x218 [xfs]
       __vfs_write+0x1cc/0x214
       vfs_write+0x12c/0x1a4
       ksys_write+0xb0/0x120
       __arm64_sys_write+0x54/0x88
       el0_svc_handler+0x170/0x240
       el0_sync_handler+0x150/0x250
       el0_sync+0x164/0x180

       other info that might help us debug this:

 Chain exists of:
   &(&zone->lock)->rlock --> &(&port->lock)->rlock --> &pool->lock/1

 Possible unsafe locking scenario:

       CPU0                    CPU1
       ----                    ----
  lock(&pool->lock/1);
                               lock(&(&port->lock)->rlock);
                               lock(&pool->lock/1);
  lock(&(&zone->lock)->rlock);

                *** DEADLOCK ***

4 locks held by doio/49441:
 #0: a0ff00886fc27408 (sb_writers#8){.+.+}, at: vfs_write+0x118/0x1a4
 #1: 8fff00080810dfe0 (&xfs_nondir_ilock_class){++++}, at:
xfs_ilock+0x2a8/0x300 [xfs]
 #2: ffff9000129f2390 (rcu_read_lock){....}, at:
rcu_lock_acquire+0x8/0x38
 #3: 60ff000822352818 (&pool->lock/1){-.-.}, at:
start_flush_work+0xd8/0x3f0

               stack backtrace:
CPU: 48 PID: 49441 Comm: doio Tainted: G        W
Hardware name: HPE Apollo 70             /C01_APACHE_MB         , BIOS
L50_5.13_1.11 06/18/2019
Call trace:
 dump_backtrace+0x0/0x248
 show_stack+0x20/0x2c
 dump_stack+0xe8/0x150
 print_circular_bug+0x368/0x380
 check_noncircular+0x28c/0x294
 validate_chain+0xf6c/0x2e2c
 __lock_acquire+0x868/0xc2c
 lock_acquire+0x320/0x360
 _raw_spin_lock+0x64/0x80
 rmqueue+0x138/0x2050
 get_page_from_freelist+0x474/0x688
 __alloc_pages_nodemask+0x3b4/0x18dc
 alloc_pages_current+0xd0/0xe0
 alloc_slab_page+0x2b4/0x5e0
 new_slab+0xc8/0x6bc
 ___slab_alloc+0x3b8/0x640
 kmem_cache_alloc+0x4b4/0x588
 __debug_object_init+0x778/0x8b4
 debug_object_init_on_stack+0x40/0x50
 start_flush_work+0x16c/0x3f0
 __flush_work+0xb8/0x124
 flush_work+0x20/0x30
 xlog_cil_force_lsn+0x88/0x204 [xfs]
 xfs_log_force_lsn+0x128/0x1b8 [xfs]
 xfs_file_fsync+0x3c4/0x488 [xfs]
 vfs_fsync_range+0xb0/0xd0
 generic_write_sync+0x80/0xa0 [xfs]
 xfs_file_buffered_aio_write+0x66c/0x6e4 [xfs]
 xfs_file_write_iter+0x1a0/0x218 [xfs]
 __vfs_write+0x1cc/0x214
 vfs_write+0x12c/0x1a4
 ksys_write+0xb0/0x120
 __arm64_sys_write+0x54/0x88
 el0_svc_handler+0x170/0x240
 el0_sync_handler+0x150/0x250
 el0_sync+0x164/0x180

Reviewed-by: Sergey Senozhatsky <sergey.senozhatsky.work@gmail.com>
Signed-off-by: Qian Cai <cai@lca.pw>
Link: https://lore.kernel.org/r/1573679785-21068-1-git-send-email-cai@lca.pw
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
Signed-off-by: Sasha Levin <sashal@kernel.org>

x86/fpu: Deactivate FPU state after failure during state load

[ Upstream commit bbc55341b9c67645d1a5471506370caf7dd4a203 ]

In __fpu__restore_sig(), fpu_fpregs_owner_ctx needs to be reset if the
FPU state was not fully restored. Otherwise the following may happen (on
the same CPU):

  Task A                     Task B               fpu_fpregs_owner_ctx
  *active*                                        A.fpu
  __fpu__restore_sig()
                             ctx switch           load B.fpu
                             *active*             B.fpu
  fpregs_lock()
  copy_user_to_fpregs_zeroing()
    copy_kernel_to_xregs() *modify*
    copy_user_to_xregs() *fails*
  fpregs_unlock()
                            ctx switch            skip loading B.fpu,
                            *active*              B.fpu

In the success case, fpu_fpregs_owner_ctx is set to the current task.

In the failure case, the FPU state might have been modified by loading
the init state.

In this case, fpu_fpregs_owner_ctx needs to be reset in order to ensure
that the FPU state of the following task is loaded from saved state (and
not skipped because it was the previous state).

Reset fpu_fpregs_owner_ctx after a failure during restore occurred, to
ensure that the FPU state for the next task is always loaded.

The problem was debugged-by Yu-cheng Yu <yu-cheng.yu@intel.com>.

 [ bp: Massage commit message. ]

Fixes: 5f409e20b7945 ("x86/fpu: Defer FPU state load until return to userspace")
Reported-by: Yu-cheng Yu <yu-cheng.yu@intel.com>
Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Signed-off-by: Borislav Petkov <bp@suse.de>
Cc: Andy Lutomirski <luto@kernel.org>
Cc: Dave Hansen <dave.hansen@linux.intel.com>
Cc: Fenghua Yu <fenghua.yu@intel.com>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Jann Horn <jannh@google.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: "Ravi V. Shankar" <ravi.v.shankar@intel.com>
Cc: Rik van Riel <riel@surriel.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Tony Luck <tony.luck@intel.com>
Cc: x86-ml <x86@kernel.org>
Link: https://lkml.kernel.org/r/20191220195906.plk6kpmsrikvbcfn@linutronix.de
Signed-off-by: Sasha Levin <sashal@kernel.org>

iommu/vt-d: Fix off-by-one in PASID allocation

[ Upstream commit 39d630e332144028f56abba83d94291978e72df1 ]

PASID allocator uses IDR which is exclusive for the end of the
allocation range. There is no need to decrement pasid_max.

Fixes: af39507305fb ("iommu/vt-d: Apply global PASID in SVA")
Reported-by: Eric Auger <eric.auger@redhat.com>
Signed-off-by: Jacob Pan <jacob.jun.pan@linux.intel.com>
Reviewed-by: Eric Auger <eric.auger@redhat.com>
Signed-off-by: Lu Baolu <baolu.lu@linux.intel.com>
Signed-off-by: Joerg Roedel <jroedel@suse.de>
Signed-off-by: Sasha Levin <sashal@kernel.org>

gpio: gpio-grgpio: fix possible sleep-in-atomic-context bugs in grgpio_irq_map/unmap()

[ Upstream commit e36eaf94be8f7bc4e686246eed3cf92d845e2ef8 ]

The driver may sleep while holding a spinlock.
The function call path (from bottom to top) in Linux 4.19 is:

drivers/gpio/gpio-grgpio.c, 261:
request_irq in grgpio_irq_map
drivers/gpio/gpio-grgpio.c, 255:
_raw_spin_lock_irqsave in grgpio_irq_map

drivers/gpio/gpio-grgpio.c, 318:
free_irq in grgpio_irq_unmap
drivers/gpio/gpio-grgpio.c, 299:
_raw_spin_lock_irqsave in grgpio_irq_unmap

request_irq() and free_irq() can sleep at runtime.

To fix these bugs, request_irq() and free_irq() are called without
holding the spinlock.

These bugs are found by a static analysis tool STCheck written by myself.

Signed-off-by: Jia-Ju Bai <baijiaju1990@gmail.com>
Link: https://lore.kernel.org/r/20191218132605.10594-1-baijiaju1990@gmail.com
Signed-off-by: Linus Walleij <linus.walleij@linaro.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

clk: meson: meson8b: make the CCF use the glitch-free mali mux

[ Upstream commit 8daeaea99caabe24a0929fac17977ebfb882fa86 ]

The "mali_0" or "mali_1" clock trees should not be updated while the
clock is running. Enforce this by setting CLK_SET_RATE_GATE on the
"mali_0" and "mali_1" gates. This makes the CCF switch to the "mali_1"
tree when "mali_0" is currently active and vice versa, which is exactly
what the vendor driver does when updating the frequency of the mali
clock.

This fixes a potential hang when changing the GPU frequency at runtime.

Fixes: 74e1f2521f16ff ("clk: meson: meson8b: add the GPU clock tree")
Signed-off-by: Martin Blumenstingl <martin.blumenstingl@googlemail.com>
Signed-off-by: Jerome Brunet <jbrunet@baylibre.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

powerpc/powernv/iov: Ensure the pdn for VFs always contains a valid PE number

[ Upstream commit 3b5b9997b331e77ce967eba2c4bc80dc3134a7fe ]

On pseries there is a bug with adding hotplugged devices to an IOMMU
group. For a number of dumb reasons fixing that bug first requires
re-working how VFs are configured on PowerNV. For background, on
PowerNV we use the pcibios_sriov_enable() hook to do two things:

  1. Create a pci_dn structure for each of the VFs, and
  2. Configure the PHB's internal BARs so the MMIO range for each VF
     maps to a unique PE.

Roughly speaking a PE is the hardware counterpart to a Linux IOMMU
group since all the devices in a PE share the same IOMMU table. A PE
also defines the set of devices that should be isolated in response to
a PCI error (i.e. bad DMA, UR/CA, AER events, etc). When isolated all
MMIO and DMA traffic to and from devicein the PE is blocked by the
root complex until the PE is recovered by the OS.

The requirement to block MMIO causes a giant headache because the P8
PHB generally uses a fixed mapping between MMIO addresses and PEs. As
a result we need to delay configuring the IOMMU groups for device
until after MMIO resources are assigned. For physical devices (i.e.
non-VFs) the PE assignment is done in pcibios_setup_bridge() which is
called immediately after the MMIO resources for downstream
devices (and the bridge's windows) are assigned. For VFs the setup is
more complicated because:

  a) pcibios_setup_bridge() is not called again when VFs are activated, and
  b) The pci_dev for VFs are created by generic code which runs after
     pcibios_sriov_enable() is called.

The work around for this is a two step process:

  1. A fixup in pcibios_add_device() is used to initialised the cached
     pe_number in pci_dn, then
  2. A bus notifier then adds the device to the IOMMU group for the PE
     specified in pci_dn->pe_number.

A side effect fixing the pseries bug mentioned in the first paragraph
is moving the fixup out of pcibios_add_device() and into
pcibios_bus_add_device(), which is called much later. This results in
step 2. failing because pci_dn->pe_number won't be initialised when
the bus notifier is run.

We can fix this by removing the need for the fixup. The PE for a VF is
known before the VF is even scanned so we can initialise
pci_dn->pe_number pcibios_sriov_enable() instead. Unfortunately,
moving the initialisation causes two problems:

  1. We trip the WARN_ON() in the current fixup code, and
  2. The EEH core clears pdn->pe_number when recovering a VF and
     relies on the fixup to correctly re-set it.

The only justification for either of these is a comment in
eeh_rmv_device() suggesting that pdn->pe_number *must* be set to
IODA_INVALID_PE in order for the VF to be scanned. However, this
comment appears to have no basis in reality. Both bugs can be fixed by
just deleting the code.

Tested-by: Alexey Kardashevskiy <aik@ozlabs.ru>
Reviewed-by: Alexey Kardashevskiy <aik@ozlabs.ru>
Signed-off-by: Oliver O'Halloran <oohall@gmail.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20191028085424.12006-1-oohall@gmail.com
Signed-off-by: Sasha Levin <sashal@kernel.org>

clk: at91: sam9x60: fix programmable clock prescaler

[ Upstream commit 66d9f5214c9ba1c151478f99520b6817302d50dc ]

The prescaler works as parent rate divided by (PRES + 1) (is_pres_direct == 1)
It does not work in the way of parent rate shifted to the right by (PRES + 1),
which means division by 2^(PRES + 1) (is_pres_direct == 0)
Thus is_pres_direct must be enabled for this SoC, to make the right computation.
This field was added in
commit 45b06682113b ("clk: at91: fix programmable clock for sama5d2")
SAM9X60 has the same field as SAMA5D2 in the PCK

Fixes: 01e2113de9a5 ("clk: at91: add sam9x60 pmc driver")
Signed-off-by: Eugen Hristev <eugen.hristev@microchip.com>
Link: https://lkml.kernel.org/r/1575977088-16781-1-git-send-email-eugen.hristev@microchip.com
Signed-off-by: Stephen Boyd <sboyd@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

media: sun4i-csi: Fix [HV]sync polarity handling

[ Upstream commit 1948dcf0f928b8bcdca57ca3fba8545ba380fc29 ]

The Allwinner camera sensor interface has a different definition of
[HV]sync. While the timing diagram uses the names HSYNC and VSYNC,
the note following the diagram and register names use HREF and VREF.
Combined they imply the hardware uses either [HV]REF or inverted
[HV]SYNC. There are also registers to set horizontal skip lengths
in pixels and vertical skip lengths in lines, also known as back
porches.

Fix the polarity handling by using the opposite polarity flag for
the checks. Also rename `[hv]sync_pol` to `[hv]ref_pol` to better
match the hardware register description.

Fixes: 577bbf23b758 ("media: sunxi: Add A10 CSI driver")
Signed-off-by: Chen-Yu Tsai <wens@csie.org>
Acked-by: Maxime Ripard <mripard@kernel.org>
Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

media: sun4i-csi: Fix data sampling polarity handling

[ Upstream commit cf9e6d5dbdd56ef2aa72f28c806711c4293c8848 ]

The CLK_POL field specifies whether data is sampled on the falling or
rising edge of PCLK, not whether the data lines are active high or low.
Evidence of this can be found in the timing diagram labeled "horizontal
size setting and pixel clock timing".

Fix the setting by checking the correct flag, V4L2_MBUS_PCLK_SAMPLE_RISING.
While at it, reorder the three polarity flag checks so HSYNC and VSYNC
are grouped together.

Fixes: 577bbf23b758 ("media: sunxi: Add A10 CSI driver")
Signed-off-by: Chen-Yu Tsai <wens@csie.org>
Acked-by: Maxime Ripard <mripard@kernel.org>
Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

media: sun4i-csi: Deal with DRAM offset

[ Upstream commit 249b286171fa9c358e8d5c825b48c4ebea97c498 ]

On Allwinner SoCs, some high memory bandwidth devices do DMA directly
over the memory bus (called MBUS), instead of the system bus. These
devices include the CSI camera sensor interface, video (codec) engine,
display subsystem, etc.. The memory bus has a different addressing
scheme without the DRAM starting offset.

Deal with this using the "interconnects" property from the device tree,
or if that is not available, set dev->dma_pfn_offset to PHYS_PFN_OFFSET.

Fixes: 577bbf23b758 ("media: sunxi: Add A10 CSI driver")
Signed-off-by: Chen-Yu Tsai <wens@csie.org>
Acked-by: Maxime Ripard <mripard@kernel.org>
Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

media: i2c: mt9v032: fix enum mbus codes and frame sizes

[ Upstream commit 1451d5ae351d938a0ab1677498c893f17b9ee21d ]

This driver supports both the mt9v032 (color) and the mt9v022 (mono)
sensors. Depending on which sensor is used, the format from the sensor is
different. The format.code inside the dev struct holds this information.
The enum mbus and enum frame sizes need to take into account both type of
sensors, not just the color one. To solve this, use the format.code in
these functions instead of the hardcoded bayer color format (which is only
used for mt9v032).

[Sakari Ailus: rewrapped commit message]

Suggested-by: Wenyou Yang <wenyou.yang@microchip.com>
Signed-off-by: Eugen Hristev <eugen.hristev@microchip.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

media: ov5640: Fix check for PLL1 exceeding max allowed rate

[ Upstream commit 2e3df204f9af42a47823ee955c08950373417420 ]

The variable _rate is by ov5640_compute_sys_clk() which returns
zero if the PLL exceeds 1GHz.  Unfortunately, the check to see
if the max PLL1 output is checking 'rate' and not '_rate' and
'rate' does not ever appear to be 0.

This patch changes the check against the returned value of
'_rate' to determine if the PLL1 output exceeds 1GHz.

Fixes: aa2882481cad ("media: ov5640: Adjust the clock based on the expected rate")
Signed-off-by: Adam Ford <aford173@gmail.com>
Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

pxa168fb: Fix the function used to release some memory in an error handling path

[ Upstream commit 3c911fe799d1c338d94b78e7182ad452c37af897 ]

In the probe function, some resources are allocated using 'dma_alloc_wc()',
they should be released with 'dma_free_wc()', not 'dma_free_coherent()'.

We already use 'dma_free_wc()' in the remove function, but not in the
error handling path of the probe function.

Also, remove a useless 'PAGE_ALIGN()'. 'info->fix.smem_len' is already
PAGE_ALIGNed.

Fixes: 638772c7553f ("fb: add support of LCD display controller on pxa168/910 (base layer)")
Signed-off-by: Christophe JAILLET <christophe.jaillet@wanadoo.fr>
Reviewed-by: Lubomir Rintel <lkundrak@v3.sk>
CC: YueHaibing <yuehaibing@huawei.com>
Signed-off-by: Bartlomiej Zolnierkiewicz <b.zolnierkie@samsung.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20190831100024.3248-1-christophe.jaillet@wanadoo.fr
Signed-off-by: Sasha Levin <sashal@kernel.org>

drm/msm/adreno: fix zap vs no-zap handling

[ Upstream commit 15ab987c423df561e0949d77fb5043921ae59956 ]

We can have two cases, when it comes to "zap" fw.  Either the fw
requires zap fw to take the GPU out of secure mode at boot, or it does
not and we can write RBBM_SECVID_TRUST_CNTL directly.  Previously we
decided based on whether zap fw load succeeded, but this is not a great
plan because:

1) we could have zap fw in the filesystem on a device where it is not
   required
2) we could have the inverse case

Instead, shift to deciding based on whether we have a 'zap-shader' node
in dt.  In practice, there is only one device (currently) with upstream
dt that does not use zap (cheza), and it already has a /delete-node/ for
the zap-shader node.

Fixes: abccb9fe3267 ("drm/msm/a6xx: Add zap shader load")
Signed-off-by: Rob Clark <robdclark@chromium.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

drm/mipi_dbi: Fix off-by-one bugs in mipi_dbi_blank()

[ Upstream commit 2ce18249af5a28031b3f909cfafccc88ea966c9d ]

When configuring the frame memory window, the last column and row
numbers are written to the column resp. page address registers.  These
numbers are thus one less than the actual window width resp. height.

While this is handled correctly in mipi_dbi_fb_dirty() since commit
03ceb1c8dfd1e293 ("drm/tinydrm: Fix setting of the column/page end
addresses."), it is not in mipi_dbi_blank().  The latter still forgets
to subtract one when calculating the most significant bytes of the
column and row numbers, thus programming wrong values when the display
width or height is a multiple of 256.

Fixes: 02dd95fe31693626 ("drm/tinydrm: Add MIPI DBI support")
Signed-off-by: Geert Uytterhoeven <geert+renesas@glider.be>
Signed-off-by: Noralf Trønnes <noralf@tronnes.org>
Link: https://patchwork.freedesktop.org/patch/msgid/20191230130604.31006-1-geert+renesas@glider.be
Signed-off-by: Sasha Levin <sashal@kernel.org>

printk: fix exclusive_console replaying

[ Upstream commit def97da136515cb289a14729292c193e0a93bc64 ]

Commit f92b070f2dc8 ("printk: Do not miss new messages when replaying
the log") introduced a new variable @exclusive_console_stop_seq to
store when an exclusive console should stop printing. It should be
set to the @console_seq value at registration. However, @console_seq
is previously set to @syslog_seq so that the exclusive console knows
where to begin. This results in the exclusive console immediately
reactivating all the other consoles and thus repeating the messages
for those consoles.

Set @console_seq after @exclusive_console_stop_seq has stored the
current @console_seq value.

Fixes: f92b070f2dc8 ("printk: Do not miss new messages when replaying the log")
Link: http://lkml.kernel.org/r/20191219115322.31160-1-john.ogness@linutronix.de
Cc: Steven Rostedt <rostedt@goodmis.org>
Cc: linux-kernel@vger.kernel.org
Signed-off-by: John Ogness <john.ogness@linutronix.de>
Acked-by: Sergey Senozhatsky <sergey.senozhatsky@gmail.com>
Signed-off-by: Petr Mladek <pmladek@suse.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

pinctrl: sh-pfc: sh7264: Fix CAN function GPIOs

[ Upstream commit 55b1cb1f03ad5eea39897d0c74035e02deddcff2 ]

pinmux_func_gpios[] contains a hole due to the missing function GPIO
definition for the "CTX0&CTX1" signal, which is the logical "AND" of the
two CAN outputs.

Fix this by:
  - Renaming CRX0_CRX1_MARK to CTX0_CTX1_MARK, as PJ2MD[2:0]=010
    configures the combined "CTX0&CTX1" output signal,
  - Renaming CRX0X1_MARK to CRX0_CRX1_MARK, as PJ3MD[1:0]=10 configures
    the shared "CRX0/CRX1" input signal, which is fed to both CAN
    inputs,
  - Adding the missing function GPIO definition for "CTX0&CTX1" to
    pinmux_func_gpios[],
  - Moving all CAN enums next to each other.

See SH7262 Group, SH7264 Group User's Manual: Hardware, Rev. 4.00:
  [1] Figure 1.2 (3) (Pin Assignment for the SH7264 Group (1-Mbyte
      Version),
  [2] Figure 1.2 (4) Pin Assignment for the SH7264 Group (640-Kbyte
      Version,
  [3] Table 1.4 List of Pins,
  [4] Figure 20.29 Connection Example when Using This Module as 1-Channel
      Module (64 Mailboxes x 1 Channel),
  [5] Table 32.10 Multiplexed Pins (Port J),
  [6] Section 32.2.30 (3) Port J Control Register 0 (PJCR0).

Note that the last 2 disagree about PJ2MD[2:0], which is probably the
root cause of this bug.  But considering [4], "CTx0&CTx1" in [5] must
be correct, and "CRx0&CRx1" in [6] must be wrong.

Signed-off-by: Geert Uytterhoeven <geert+renesas@glider.be>
Link: https://lore.kernel.org/r/20191218194812.12741-4-geert+renesas@glider.be
Signed-off-by: Sasha Levin <sashal@kernel.org>

gianfar: Fix TX timestamping with a stacked DSA driver

[ Upstream commit c26a2c2ddc0115eb088873f5c309cf46b982f522 ]

The driver wrongly assumes that it is the only entity that can set the
SKBTX_IN_PROGRESS bit of the current skb. Therefore, in the
gfar_clean_tx_ring function, where the TX timestamp is collected if
necessary, the aforementioned bit is used to discriminate whether or not
the TX timestamp should be delivered to the socket's error queue.

But a stacked driver such as a DSA switch can also set the
SKBTX_IN_PROGRESS bit, which is actually exactly what it should do in
order to denote that the hardware timestamping process is undergoing.

Therefore, gianfar would misinterpret the "in progress" bit as being its
own, and deliver a second skb clone in the socket's error queue,
completely throwing off a PTP process which is not expecting to receive
it, _even though_ TX timestamping is not enabled for gianfar.

There have been discussions [0] as to whether non-MAC drivers need or
not to set SKBTX_IN_PROGRESS at all (whose purpose is to avoid sending 2
timestamps, a sw and a hw one, to applications which only expect one).
But as of this patch, there are at least 2 PTP drivers that would break
in conjunction with gianfar: the sja1105 DSA switch and the felix
switch, by way of its ocelot core driver.

So regardless of that conclusion, fix the gianfar driver to not do stuff
based on flags set by others and not intended for it.

[0]: https://www.spinics.net/lists/netdev/msg619699.html

Fixes: f0ee7acfcdd4 ("gianfar: Add hardware TX timestamping support")
Signed-off-by: Vladimir Oltean <olteanv@gmail.com>
Acked-by: Richard Cochran <richardcochran@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Sasha Levin <sashal@kernel.org>

ALSA: ctl: allow TLV read operation for callback type of element in locked case

[ Upstream commit d61fe22c2ae42d9fd76c34ef4224064cca4b04b0 ]

A design of ALSA control core allows applications to execute three
operations for TLV feature; read, write and command. Furthermore, it
allows driver developers to process the operations by two ways; allocated
array or callback function. In the former, read operation is just allowed,
thus developers uses the latter when device driver supports variety of
models or the target model is expected to dynamically change information
stored in TLV container.

The core also allows applications to lock any element so that the other
applications can't perform write operation to the element for element
value and TLV information. When the element is locked, write and command
operation for TLV information are prohibited as well as element value.
Any read operation should be allowed in the case.

At present, when an element has callback function for TLV information,
TLV read operation returns EPERM if the element is locked. On the
other hand, the read operation is success when an element has allocated
array for TLV information. In both cases, read operation is success for
element value expectedly.

This commit fixes the bug. This change can be backported to v4.14
kernel or later.

Signed-off-by: Takashi Sakamoto <o-takashi@sakamocchi.jp>
Reviewed-by: Jaroslav Kysela <perex@perex.cz>
Link: https://lore.kernel.org/r/20191223093347.15279-1-o-takashi@sakamocchi.jp
Signed-off-by: Takashi Iwai <tiwai@suse.de>
Signed-off-by: Sasha Levin <sashal@kernel.org>

ext4: fix ext4_dax_read/write inode locking sequence for IOCB_NOWAIT

[ Upstream commit f629afe3369e9885fd6e9cc7a4f514b6a65cf9e9 ]

Apparently our current rwsem code doesn't like doing the trylock, then
lock for real scheme.  So change our dax read/write methods to just do the
trylock for the RWF_NOWAIT case.
This seems to fix AIM7 regression in some scalable filesystems upto ~25%
in some cases. Claimed in commit 942491c9e6d6 ("xfs: fix AIM7 regression")

Reviewed-by: Jan Kara <jack@suse.cz>
Reviewed-by: Matthew Bobrowski <mbobrowski@mbobrowski.org>
Tested-by: Joseph Qi <joseph.qi@linux.alibaba.com>
Signed-off-by: Ritesh Harjani <riteshh@linux.ibm.com>
Link: https://lore.kernel.org/r/20191212055557.11151-2-riteshh@linux.ibm.com
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
Signed-off-by: Sasha Levin <sashal@kernel.org>

leds: pca963x: Fix open-drain initialization

[ Upstream commit 697529091ac7a0a90ca349b914bb30641c13c753 ]

Before commit bb29b9cccd95 ("leds: pca963x: Add bindings to invert
polarity") Mode register 2 was initialized directly with either 0x01
or 0x05 for open-drain or totem pole (push-pull) configuration.

Afterwards, MODE2 initialization started using bitwise operations on
top of the default MODE2 register value (0x05). Using bitwise OR for
setting OUTDRV with 0x01 and 0x05 does not produce correct results.
When open-drain is used, instead of setting OUTDRV to 0, the driver
keeps it as 1:

Open-drain: 0x05 | 0x01 -> 0x05 (0b101 - incorrect)
Totem pole: 0x05 | 0x05 -> 0x05 (0b101 - correct but still wrong)

Now OUTDRV setting uses correct bitwise operations for initialization:

Open-drain: 0x05 & ~0x04 -> 0x01 (0b001 - correct)
Totem pole: 0x05 | 0x04 -> 0x05 (0b101 - correct)

Additional MODE2 register definitions are introduced now as well.

Fixes: bb29b9cccd95 ("leds: pca963x: Add bindings to invert polarity")
Signed-off-by: Zahari Petkov <zahari@balena.io>
Signed-off-by: Pavel Machek <pavel@ucw.cz>
Signed-off-by: Sasha Levin <sashal@kernel.org>

drm/amd/display: Map ODM memory correctly when doing ODM combine

[ Upstream commit ec5b356c58941bb8930858155d9ce14ceb3d30a0 ]

[why]
Up to 4 ODM memory pieces are required per ODM combine and cannot
overlap, i.e. each ODM "session" has to use its own memory pieces.
The ODM-memory mapping is currently broken for generic case.

The maximum number of memory pieces is ASIC-dependent, but it's always
big enough to satisfy maximum number of ODM combines. Memory pieces
are mapped as a bit-map, i.e. one memory piece corresponds to one bit.
The OPTC doing ODM needs to select memory pieces by setting the
corresponding bits, making sure there's no overlap with other OPTC
instances that might be doing ODM.

The current mapping works only for OPTC instance indexes smaller than
3. For instance indexes 3 and up it practically maps no ODM memory,
causing black, gray or white screen in display configs that include
ODM on OPTC instance 3 or up.

[how]
Statically map two unique ODM memory pieces for each OPTC instance
and piece them together when programming ODM combine mode.

Signed-off-by: Nikola Cornij <nikola.cornij@amd.com>
Reviewed-by: Jun Lei <Jun.Lei@amd.com>
Acked-by: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

PCI: Fix pci_add_dma_alias() bitmask size

[ Upstream commit f8bf2aeb651b3460a4b36fd7ba1ba1d31777d35c ]

The number of possible devfns is 256, but pci_add_dma_alias() allocated a
bitmap of size 255.  Fix this off-by-one error.

This fixes commits 338c3149a221 ("PCI: Add support for multiple DMA
aliases") and c6635792737b ("PCI: Allocate dma_alias_mask with
bitmap_zalloc()"), but I doubt it was possible to see a problem because
it takes 4 64-bit longs (or 8 32-bit longs) to hold 255 bits, and
bitmap_zalloc() doesn't save the 255-bit size anywhere.

[bhelgaas: commit log, move #define to drivers/pci/pci.h, include loop
limit fix from Qian Cai:
https://lore.kernel.org/r/20191218170004.5297-1-cai@lca.pw]
Signed-off-by: James Sewart <jamessewart@arista.com>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Logan Gunthorpe <logang@deltatee.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

brcmfmac: Fix use after free in brcmf_sdio_readframes()

[ Upstream commit 216b44000ada87a63891a8214c347e05a4aea8fe ]

The brcmu_pkt_buf_free_skb() function frees "pkt" so it leads to a
static checker warning:

    drivers/net/wireless/broadcom/brcm80211/brcmfmac/sdio.c:1974 brcmf_sdio_readframes()
    error: dereferencing freed memory 'pkt'

It looks like there was supposed to be a continue after we free "pkt".

Fixes: 4754fceeb9a6 ("brcmfmac: streamline SDIO read frame routine")
Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Acked-by: Franky Lin <franky.lin@broadcom.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

brcmfmac: Fix memory leak in brcmf_p2p_create_p2pdev()

[ Upstream commit 5cc509aa83c6acd2c5cd94f99065c39d2bd0a490 ]

In the implementation of brcmf_p2p_create_p2pdev() the allocated memory
for p2p_vif is leaked when the mac address is the same as primary
interface. To fix this, go to error path to release p2p_vif via
brcmf_free_vif().

Fixes: cb746e47837a ("brcmfmac: check p2pdev mac address uniqueness")
Signed-off-by: Navid Emamdoost <navid.emamdoost@gmail.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

cpu/hotplug, stop_machine: Fix stop_machine vs hotplug order

[ Upstream commit 45178ac0cea853fe0e405bf11e101bdebea57b15 ]

Paul reported a very sporadic, rcutorture induced, workqueue failure.
When the planets align, the workqueue rescuer's self-migrate fails and
then triggers a WARN for running a work on the wrong CPU.

Tejun then figured that set_cpus_allowed_ptr()'s stop_one_cpu() call
could be ignored! When stopper->enabled is false, stop_machine will
insta complete the work, without actually doing the work. Worse, it
will not WARN about this (we really should fix this).

It turns out there is a small window where a freshly online'ed CPU is
marked 'online' but doesn't yet have the stopper task running:

BP AP

bringup_cpu()
  __cpu_up(cpu, idle)  --> start_secondary()
...
cpu_startup_entry()
  bringup_wait_for_ap()
    wait_for_ap_thread() <--   cpuhp_online_idle()
  while (1)
    do_idle()

... available to run kthreads ...

    stop_machine_unpark()
      stopper->enable = true;

Close this by moving the stop_machine_unpark() into
cpuhp_online_idle(), such that the stopper thread is ready before we
start the idle loop and schedule.

Reported-by: "Paul E. McKenney" <paulmck@kernel.org>
Debugged-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Tested-by: "Paul E. McKenney" <paulmck@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

clk: meson: pll: Fix by 0 division in __pll_params_to_rate()

[ Upstream commit d8488a41800d9f5c80bc0d17b9cc2c91b4841464 ]

Some meson pll registers can be initialized with 0 as N value, introducing
the following division by 0 when computing rate :

  UBSAN: Undefined behaviour in drivers/clk/meson/clk-pll.c:75:9
  division by zero
  CPU: 0 PID: 1 Comm: swapper/0 Not tainted 5.4.0-rc3-608075-g86c9af8630e1-dirty #400
  Call trace:
   dump_backtrace+0x0/0x1c0
   show_stack+0x14/0x20
   dump_stack+0xc4/0x100
   ubsan_epilogue+0x14/0x68
   __ubsan_handle_divrem_overflow+0x98/0xb8
   __pll_params_to_rate+0xdc/0x140
   meson_clk_pll_recalc_rate+0x278/0x3a0
   __clk_register+0x7c8/0xbb0
   devm_clk_hw_register+0x54/0xc0
   meson_eeclkc_probe+0xf4/0x1a0
   platform_drv_probe+0x54/0xd8
   really_probe+0x16c/0x438
   driver_probe_device+0xb0/0xf0
   device_driver_attach+0x94/0xa0
   __driver_attach+0x70/0x108
   bus_for_each_dev+0xd8/0x128
   driver_attach+0x30/0x40
   bus_add_driver+0x1b0/0x2d8
   driver_register+0xbc/0x1d0
   __platform_driver_register+0x78/0x88
   axg_driver_init+0x18/0x20
   do_one_initcall+0xc8/0x24c
   kernel_init_freeable+0x2b0/0x344
   kernel_init+0x10/0x128
   ret_from_fork+0x10/0x18

This checks if N is null before doing the division.

Fixes: 7a29a869434e ("clk: meson: Add support for Meson clock controller")
Reviewed-by: Martin Blumenstingl <martin.blumenstingl@googlemail.com>
Signed-off-by: Remi Pommarel <repk@triplefau.lt>
[jbrunet@baylibre.com: update the comment in above the fix]
Signed-off-by: Jerome Brunet <jbrunet@baylibre.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

media: meson: add missing allocation failure check on new_buf

[ Upstream commit 11e0e167d071a28288a7a0a211d48c571d19b56f ]

Currently if the allocation of new_buf fails then a null pointer
dereference occurs when assiging new_buf->vb. Avoid this by returning
early on a memory allocation failure as there is not much more can
be done at this point.

Addresses-Coverity: ("Dereference null return")

Fixes: 3e7f51bd9607 ("media: meson: add v4l2 m2m video decoder driver")
Signed-off-by: Colin Ian King <colin.king@canonical.com>
Reviewed-by: Kevin Hilman <khilman@baylibre.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

f2fs: call f2fs_balance_fs outside of locked page

[ Upstream commit bdf03299248916640a835a05d32841bb3d31912d ]

Otherwise, we can hit deadlock by waiting for the locked page in
move_data_block in GC.

 Thread A                     Thread B
 - do_page_mkwrite
  - f2fs_vm_page_mkwrite
   - lock_page
                              - f2fs_balance_fs
                                  - mutex_lock(gc_mutex)
                               - f2fs_gc
                                - do_garbage_collect
                                 - ra_data_block
                                  - grab_cache_page
   - f2fs_balance_fs
    - mutex_lock(gc_mutex)

Fixes: 39a8695824510 ("f2fs: refactor ->page_mkwrite() flow")
Reviewed-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

f2fs: preallocate DIO blocks when forcing buffered_io

[ Upstream commit 47501f87c61ad2aa234add63e1ae231521dbc3f5 ]

The previous preallocation and DIO decision like below.

                         allow_outplace_dio              !allow_outplace_dio
f2fs_force_buffered_io   (*) No_Prealloc / Buffered_IO   Prealloc / Buffered_IO
!f2fs_force_buffered_io  No_Prealloc / DIO               Prealloc / DIO

But, Javier reported Case (*) where zoned device bypassed preallocation but
fell back to buffered writes in f2fs_direct_IO(), resulting in stale data
being read.

In order to fix the issue, actually we need to preallocate blocks whenever
we fall back to buffered IO like this. No change is made in the other cases.

                         allow_outplace_dio              !allow_outplace_dio
f2fs_force_buffered_io   (*) Prealloc / Buffered_IO      Prealloc / Buffered_IO
!f2fs_force_buffered_io  No_Prealloc / DIO               Prealloc / DIO

Reported-and-tested-by: Javier Gonzalez <javier@javigon.com>
Signed-off-by: Damien Le Moal <damien.lemoal@wdc.com>
Tested-by: Shin'ichiro Kawasaki <shinichiro.kawasaki@wdc.com>
Reviewed-by: Chao Yu <yuchao0@huawei.com>
Reviewed-by: Javier González <javier@javigon.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

rcu: Fix data-race due to atomic_t copy-by-value

[ Upstream commit 6cf539a87a61a4fbc43f625267dbcbcf283872ed ]

This fixes a data-race where `atomic_t dynticks` is copied by value. The
copy is performed non-atomically, resulting in a data-race if `dynticks`
is updated concurrently.

This data-race was found with KCSAN:
==================================================================
BUG: KCSAN: data-race in dyntick_save_progress_counter / rcu_irq_enter

write to 0xffff989dbdbe98e0 of 4 bytes by task 10 on cpu 3:
 atomic_add_return include/asm-generic/atomic-instrumented.h:78 [inline]
 rcu_dynticks_snap kernel/rcu/tree.c:310 [inline]
 dyntick_save_progress_counter+0x43/0x1b0 kernel/rcu/tree.c:984
 force_qs_rnp+0x183/0x200 kernel/rcu/tree.c:2286
 rcu_gp_fqs kernel/rcu/tree.c:1601 [inline]
 rcu_gp_fqs_loop+0x71/0x880 kernel/rcu/tree.c:1653
 rcu_gp_kthread+0x22c/0x3b0 kernel/rcu/tree.c:1799
 kthread+0x1b5/0x200 kernel/kthread.c:255
 <snip>

read to 0xffff989dbdbe98e0 of 4 bytes by task 154 on cpu 7:
 rcu_nmi_enter_common kernel/rcu/tree.c:828 [inline]
 rcu_irq_enter+0xda/0x240 kernel/rcu/tree.c:870
 irq_enter+0x5/0x50 kernel/softirq.c:347
 <snip>

Reported by Kernel Concurrency Sanitizer on:
CPU: 7 PID: 154 Comm: kworker/7:1H Not tainted 5.3.0+ #5
Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS 1.12.0-1 04/01/2014
Workqueue: kblockd blk_mq_run_work_fn
==================================================================

Signed-off-by: Marco Elver <elver@google.com>
Cc: Paul E. McKenney <paulmck@kernel.org>
Cc: Josh Triplett <josh@joshtriplett.org>
Cc: Steven Rostedt <rostedt@goodmis.org>
Cc: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Cc: Joel Fernandes <joel@joelfernandes.org>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Dmitry Vyukov <dvyukov@google.com>
Cc: rcu@vger.kernel.org
Cc: linux-kernel@vger.kernel.org
Reviewed-by: Joel Fernandes (Google) <joel@joelfernandes.org>
Signed-off-by: Paul E. McKenney <paulmck@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

rcu: Fix missed wakeup of exp_wq waiters

[ Upstream commit fd6bc19d7676a060a171d1cf3dcbf6fd797eb05f ]

Tasks waiting within exp_funnel_lock() for an expedited grace period to
elapse can be starved due to the following sequence of events:

1. Tasks A and B both attempt to start an expedited grace
period at about the same time. This grace period will have
completed when the lower four bits of the rcu_state structure's
->expedited_sequence field are 0b'0100', for example, when the
initial value of this counter is zero. Task A wins, and thus
does the actual work of starting the grace period, including
acquiring the rcu_state structure's .exp_mutex and sets the
counter to 0b'0001'.

2. Because task B lost the race to start the grace period, it
waits on ->expedited_sequence to reach 0b'0100' inside of
exp_funnel_lock(). This task therefore blocks on the rcu_node
structure's ->exp_wq[1] field, keeping in mind that the
end-of-grace-period value of ->expedited_sequence (0b'0100')
is shifted down two bits before indexing the ->exp_wq[] field.

3. Task C attempts to start another expedited grace period,
but blocks on ->exp_mutex, which is still held by Task A.

4. The aforementioned expedited grace period completes, so that
->expedited_sequence now has the value 0b'0100'.  A kworker task
therefore acquires the rcu_state structure's ->exp_wake_mutex
and starts awakening any tasks waiting for this grace period.

5. One of the first tasks awakened happens to be Task A.  Task A
therefore releases the rcu_state structure's ->exp_mutex,
which allows Task C to start the next expedited grace period,
which causes the lower four bits of the rcu_state structure's
->expedited_sequence field to become 0b'0101'.

6. Task C's expedited grace period completes, so that the lower four
bits of the rcu_state structure's ->expedited_sequence field now
become 0b'1000'.

7. The kworker task from step 4 above continues its wakeups.
Unfortunately, the wake_up_all() refetches the rcu_state
structure's .expedited_sequence field:

wake_up_all(&rnp->exp_wq[rcu_seq_ctr(rcu_state.expedited_sequence) & 0x3]);

This results in the wakeup being applied to the rcu_node
structure's ->exp_wq[2] field, which is unfortunate given that
Task B is instead waiting on ->exp_wq[1].

On a busy system, no harm is done (or at least no permanent harm is done).
Some later expedited grace period will redo the wakeup.  But on a quiet
system, such as many embedded systems, it might be a good long time before
there was another expedited grace period.  On such embedded systems,
this situation could therefore result in a system hang.

This issue manifested as DPM device timeout during suspend (which
usually qualifies as a quiet time) due to a SCSI device being stuck in
_synchronize_rcu_expedited(), with the following stack trace:

schedule()
synchronize_rcu_expedited()
synchronize_rcu()
scsi_device_quiesce()
scsi_bus_suspend()
dpm_run_callback()
__device_suspend()

This commit therefore prevents such delays, timeouts, and hangs by
making rcu_exp_wait_wake() use its "s" argument consistently instead of
refetching from rcu_state.expedited_sequence.

Fixes: 3b5f668e715b ("rcu: Overlap wakeups with next expedited grace period")
Signed-off-by: Neeraj Upadhyay <neeraju@codeaurora.org>
Signed-off-by: Paul E. McKenney <paulmck@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

rcu/nocb: Fix dump_tree hierarchy print always active

[ Upstream commit 610dea36d3083a977e4f156206cbe1eaa2a532f0 ]

Commit 18cd8c93e69e ("rcu/nocb: Print gp/cb kthread hierarchy if
dump_tree") added print statements to rcu_organize_nocb_kthreads for
debugging, but incorrectly guarded them, causing the function to always
spew out its message.

This patch fixes it by guarding both pr_alert statements with dump_tree,
while also changing the second pr_alert to a pr_cont, to print the
hierarchy in a single line (assuming that's how it was supposed to
work).

Fixes: 18cd8c93e69e ("rcu/nocb: Print gp/cb kthread hierarchy if dump_tree")
Signed-off-by: Stefan Reiter <stefan@pimaker.at>
[ paulmck: Make single-nocbs-CPU GP kthreads look less erroneous. ]
Signed-off-by: Paul E. McKenney <paulmck@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

drm/qxl: Complete exception handling in qxl_device_init()

[ Upstream commit dbe3ad61dcebc49fe3efca70a0f752a95b4600f2 ]

A coccicheck run provided information like the following.

drivers/gpu/drm/qxl/qxl_kms.c:295:1-7: ERROR: missing iounmap;
ioremap on line 178 and execution via conditional on line 185

Generated by: scripts/coccinelle/free/iounmap.cocci

A jump target was specified in an if branch. The corresponding function
call did not release the desired system resource then.
Thus use the label “rom_unmap” instead to fix the exception handling
for this function implementation.

Fixes: 5043348a4969ae1661c008efe929abd0d76e3792 ("drm: qxl: Fix error handling at qxl_device_init")
Signed-off-by: Markus Elfring <elfring@users.sourceforge.net>
Link: http://patchwork.freedesktop.org/patch/msgid/5e5ef9c4-4d85-3c93-cf28-42cfcb5b0649@web.de
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

wil6210: fix break that is never reached because of zero'ing of a retry counter

[ Upstream commit 5b1413f00b5beb9f5fed94e43ea0c497d5db9633 ]

There is a check on the retry counter invalid_buf_id_retry that is always
false because invalid_buf_id_retry is initialized to zero on each iteration
of a while-loop.  Fix this by initializing the retry counter before the
while-loop starts.

Addresses-Coverity: ("Logically dead code")
Fixes: b4a967b7d0f5 ("wil6210: reset buff id in status message after completion")
Signed-off-by: Colin Ian King <colin.king@canonical.com>
Reviewed-by: Maya Erez <merez@codeaurora.org>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

ath10k: Fix qmi init error handling

[ Upstream commit f8a595a87e93a33a10879f4b856be818d2f53c84 ]

When ath10k_qmi_init() fails, the error handling does not free the irq
resources, which causes an issue if we EPROBE_DEFER as we'll attempt to
(re-)register irqs which are already registered.

Fix this by doing a power off since we just powered on the hardware, and
freeing the irqs as error handling.

Fixes: ba94c753ccb4 ("ath10k: add QMI message handshake for wcn3990 client")
Signed-off-by: Jeffrey Hugo <jeffrey.l.hugo@gmail.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

drm/gma500: Fixup fbdev stolen size usage evaluation

[ Upstream commit fd1a5e521c3c083bb43ea731aae0f8b95f12b9bd ]

psbfb_probe performs an evaluation of the required size from the stolen
GTT memory, but gets it wrong in two distinct ways:
- The resulting size must be page-size-aligned;
- The size to allocate is derived from the surface dimensions, not the fb
  dimensions.

When two connectors are connected with different modes, the smallest will
be stored in the fb dimensions, but the size that needs to be allocated must
match the largest (surface) dimensions. This is what is used in the actual
allocation code.

Fix this by correcting the evaluation to conform to the two points above.
It allows correctly switching to 16bpp when one connector is e.g. 1920x1080
and the other is 1024x768.

Signed-off-by: Paul Kocialkowski <paul.kocialkowski@bootlin.com>
Signed-off-by: Patrik Jakobsson <patrik.r.jakobsson@gmail.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20191107153048.843881-1-paul.kocialkowski@bootlin.com
Signed-off-by: Sasha Levin <sashal@kernel.org>

net/sched: flower: add missing validation of TCA_FLOWER_FLAGS

[ Upstream commit e2debf0852c4d66ba1a8bde12869b196094c70a7 ]

unlike other classifiers that can be offloaded (i.e. users can set flags
like 'skip_hw' and 'skip_sw'), 'cls_flower' doesn't validate the size of
netlink attribute 'TCA_FLOWER_FLAGS' provided by user: add a proper entry
to fl_policy.

Fixes: 5b33f48842fa ("net/flower: Introduce hardware offload support")
Signed-off-by: Davide Caratti <dcaratti@redhat.com>
Acked-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

net/sched: matchall: add missing validation of TCA_MATCHALL_FLAGS

[ Upstream commit 1afa3cc90f8fb745c777884d79eaa1001d6927a6 ]

unlike other classifiers that can be offloaded (i.e. users can set flags
like 'skip_hw' and 'skip_sw'), 'cls_matchall' doesn't validate the size
of netlink attribute 'TCA_MATCHALL_FLAGS' provided by user: add a proper
entry to mall_policy.

Fixes: b87f7936a932 ("net/sched: Add match-all classifier hw offloading.")
Signed-off-by: Davide Caratti <dcaratti@redhat.com>
Acked-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

net: dsa: tag_qca: Make sure there is headroom for tag

[ Upstream commit 04fb91243a853dbde216d829c79d9632e52aa8d9 ]

Passing tag size to skb_cow_head will make sure
there is enough headroom for the tag data.
This change does not introduce any overhead in case there
is already available headroom for tag.

Signed-off-by: Per Forlin <perfn@axis.com>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

net/smc: fix leak of kernel memory to user space

[ Upstream commit 457fed775c97ac2c0cd1672aaf2ff2c8a6235e87 ]

As nlmsg_put() does not clear the memory that is reserved,
it this the caller responsability to make sure all of this
memory will be written, in order to not reveal prior content.

While we are at it, we can provide the socket cookie even
if clsock is not set.

syzbot reported :

BUG: KMSAN: uninit-value in __arch_swab32 arch/x86/include/uapi/asm/swab.h:10 [inline]
BUG: KMSAN: uninit-value in __fswab32 include/uapi/linux/swab.h:59 [inline]
BUG: KMSAN: uninit-value in __swab32p include/uapi/linux/swab.h:179 [inline]
BUG: KMSAN: uninit-value in __be32_to_cpup include/uapi/linux/byteorder/little_endian.h:82 [inline]
BUG: KMSAN: uninit-value in get_unaligned_be32 include/linux/unaligned/access_ok.h:30 [inline]
BUG: KMSAN: uninit-value in ____bpf_skb_load_helper_32 net/core/filter.c:240 [inline]
BUG: KMSAN: uninit-value in ____bpf_skb_load_helper_32_no_cache net/core/filter.c:255 [inline]
BUG: KMSAN: uninit-value in bpf_skb_load_helper_32_no_cache+0x14a/0x390 net/core/filter.c:252
CPU: 1 PID: 5262 Comm: syz-executor.5 Not tainted 5.5.0-rc5-syzkaller #0
Hardware name: Google Google Compute Engine/Google Compute Engine, BIOS Google 01/01/2011
Call Trace:
 __dump_stack lib/dump_stack.c:77 [inline]
 dump_stack+0x1c9/0x220 lib/dump_stack.c:118
 kmsan_report+0xf7/0x1e0 mm/kmsan/kmsan_report.c:118
 __msan_warning+0x58/0xa0 mm/kmsan/kmsan_instr.c:215
 __arch_swab32 arch/x86/include/uapi/asm/swab.h:10 [inline]
 __fswab32 include/uapi/linux/swab.h:59 [inline]
 __swab32p include/uapi/linux/swab.h:179 [inline]
 __be32_to_cpup include/uapi/linux/byteorder/little_endian.h:82 [inline]
 get_unaligned_be32 include/linux/unaligned/access_ok.h:30 [inline]
 ____bpf_skb_load_helper_32 net/core/filter.c:240 [inline]
 ____bpf_skb_load_helper_32_no_cache net/core/filter.c:255 [inline]
 bpf_skb_load_helper_32_no_cache+0x14a/0x390 net/core/filter.c:252

Uninit was created at:
 kmsan_save_stack_with_flags mm/kmsan/kmsan.c:144 [inline]
 kmsan_internal_poison_shadow+0x66/0xd0 mm/kmsan/kmsan.c:127
 kmsan_kmalloc_large+0x73/0xc0 mm/kmsan/kmsan_hooks.c:128
 kmalloc_large_node_hook mm/slub.c:1406 [inline]
 kmalloc_large_node+0x282/0x2c0 mm/slub.c:3841
 __kmalloc_node_track_caller+0x44b/0x1200 mm/slub.c:4368
 __kmalloc_reserve net/core/skbuff.c:141 [inline]
 __alloc_skb+0x2fd/0xac0 net/core/skbuff.c:209
 alloc_skb include/linux/skbuff.h:1049 [inline]
 netlink_dump+0x44b/0x1ab0 net/netlink/af_netlink.c:2224
 __netlink_dump_start+0xbb2/0xcf0 net/netlink/af_netlink.c:2352
 netlink_dump_start include/linux/netlink.h:233 [inline]
 smc_diag_handler_dump+0x2ba/0x300 net/smc/smc_diag.c:242
 sock_diag_rcv_msg+0x211/0x610 net/core/sock_diag.c:256
 netlink_rcv_skb+0x451/0x650 net/netlink/af_netlink.c:2477
 sock_diag_rcv+0x63/0x80 net/core/sock_diag.c:275
 netlink_unicast_kernel net/netlink/af_netlink.c:1302 [inline]
 netlink_unicast+0xf9e/0x1100 net/netlink/af_netlink.c:1328
 netlink_sendmsg+0x1248/0x14d0 net/netlink/af_netlink.c:1917
 sock_sendmsg_nosec net/socket.c:639 [inline]
 sock_sendmsg net/socket.c:659 [inline]
 kernel_sendmsg+0x433/0x440 net/socket.c:679
 sock_no_sendpage+0x235/0x300 net/core/sock.c:2740
 kernel_sendpage net/socket.c:3776 [inline]
 sock_sendpage+0x1e1/0x2c0 net/socket.c:937
 pipe_to_sendpage+0x38c/0x4c0 fs/splice.c:458
 splice_from_pipe_feed fs/splice.c:512 [inline]
 __splice_from_pipe+0x539/0xed0 fs/splice.c:636
 splice_from_pipe fs/splice.c:671 [inline]
 generic_splice_sendpage+0x1d5/0x2d0 fs/splice.c:844
 do_splice_from fs/splice.c:863 [inline]
 do_splice fs/splice.c:1170 [inline]
 __do_sys_splice fs/splice.c:1447 [inline]
 __se_sys_splice+0x2380/0x3350 fs/splice.c:1427
 __x64_sys_splice+0x6e/0x90 fs/splice.c:1427
 do_syscall_64+0xb8/0x160 arch/x86/entry/common.c:296
 entry_SYSCALL_64_after_hwframe+0x44/0xa9

Fixes: f16a7dd5cf27 ("smc: netlink interface for SMC sockets")
Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Ursula Braun <ubraun@linux.vnet.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

enic: prevent waking up stopped tx queues over watchdog reset

[ Upstream commit 0f90522591fd09dd201065c53ebefdfe3c6b55cb ]

Recent months, our customer reported several kernel crashes all
preceding with following message:
NETDEV WATCHDOG: eth2 (enic): transmit queue 0 timed out
Error message of one of those crashes:
BUG: unable to handle kernel paging request at ffffffffa007e090

After analyzing severl vmcores, I found that most of crashes are
caused by memory corruption. And all the corrupted memory areas
are overwritten by data of network packets. Moreover, I also found
that the tx queues were enabled over watchdog reset.

After going through the source code, I found that in enic_stop(),
the tx queues stopped by netif_tx_disable() could be woken up over
a small time window between netif_tx_disable() and the
napi_disable() by the following code path:
napi_poll->
  enic_poll_msix_wq->
     vnic_cq_service->
        enic_wq_service->
           netif_wake_subqueue(enic->netdev, q_number)->
              test_and_clear_bit(__QUEUE_STATE_DRV_XOFF, &txq->state)
In turn, upper netowrk stack could queue skb to ENIC NIC though
enic_hard_start_xmit(). And this might introduce some race condition.

Our customer comfirmed that this kind of kernel crash doesn't occur over
90 days since they applied this patch.

Signed-off-by: Firo Yang <firo.yang@suse.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

core: Don't skip generic XDP program execution for cloned SKBs

[ Upstream commit ad1e03b2b3d4430baaa109b77bc308dc73050de3 ]

The current generic XDP handler skips execution of XDP programs entirely if
an SKB is marked as cloned. This leads to some surprising behaviour, as
packets can end up being cloned in various ways, which will make an XDP
program not see all the traffic on an interface.

This was discovered by a simple test case where an XDP program that always
returns XDP_DROP is installed on a veth device. When combining this with
the Scapy packet sniffer (which uses an AF_PACKET) socket on the sending
side, SKBs reliably end up in the cloned state, causing them to be passed
through to the receiving interface instead of being dropped. A minimal
reproducer script for this is included below.

This patch fixed the issue by simply triggering the existing linearisation
code for cloned SKBs instead of skipping the XDP program execution. This
behaviour is in line with the behaviour of the native XDP implementation
for the veth driver, which will reallocate and copy the SKB data if the SKB
is marked as shared.

Reproducer Python script (requires BCC and Scapy):

from scapy.all import TCP, IP, Ether, sendp, sniff, AsyncSniffer, Raw, UDP
from bcc import BPF
import time, sys, subprocess, shlex

SKB_MODE = (1 << 1)
DRV_MODE = (1 << 2)
PYTHON=sys.executable

def client():
    time.sleep(2)
    # Sniffing on the sender causes skb_cloned() to be set
    s = AsyncSniffer()
    s.start()

    for p in range(10):
        sendp(Ether(dst="aa:aa:aa:aa:aa:aa", src="cc:cc:cc:cc:cc:cc")/IP()/UDP()/Raw("Test"),
              verbose=False)
        time.sleep(0.1)

    s.stop()
    return 0

def server(mode):
    prog = BPF(text="int dummy_drop(struct xdp_md *ctx) {return XDP_DROP;}")
    func = prog.load_func("dummy_drop", BPF.XDP)
    prog.attach_xdp("a_to_b", func, mode)

    time.sleep(1)

    s = sniff(iface="a_to_b", count=10, timeout=15)
    if len(s):
        print(f"Got {len(s)} packets - should have gotten 0")
        return 1
    else:
        print("Got no packets - as expected")
        return 0

if len(sys.argv) < 2:
    print(f"Usage: {sys.argv[0]} <skb|drv>")
    sys.exit(1)

if sys.argv[1] == "client":
    sys.exit(client())
elif sys.argv[1] == "server":
    mode = SKB_MODE if sys.argv[2] == 'skb' else DRV_MODE
    sys.exit(server(mode))
else:
    try:
        mode = sys.argv[1]
        if mode not in ('skb', 'drv'):
            print(f"Usage: {sys.argv[0]} <skb|drv>")
            sys.exit(1)
        print(f"Running in {mode} mode")

        for cmd in [
                'ip netns add netns_a',
                'ip netns add netns_b',
                'ip -n netns_a link add a_to_b type veth peer name b_to_a netns netns_b',
                # Disable ipv6 to make sure there's no address autoconf traffic
                'ip netns exec netns_a sysctl -qw net.ipv6.conf.a_to_b.disable_ipv6=1',
                'ip netns exec netns_b sysctl -qw net.ipv6.conf.b_to_a.disable_ipv6=1',
                'ip -n netns_a link set dev a_to_b address aa:aa:aa:aa:aa:aa',
                'ip -n netns_b link set dev b_to_a address cc:cc:cc:cc:cc:cc',
                'ip -n netns_a link set dev a_to_b up',
                'ip -n netns_b link set dev b_to_a up']:
            subprocess.check_call(shlex.split(cmd))

        server = subprocess.Popen(shlex.split(f"ip netns exec netns_a {PYTHON} {sys.argv[0]} server {mode}"))
        client = subprocess.Popen(shlex.split(f"ip netns exec netns_b {PYTHON} {sys.argv[0]} client"))

        client.wait()
        server.wait()
        sys.exit(server.returncode)

    finally:
        subprocess.run(shlex.split("ip netns delete netns_a"))
        subprocess.run(shlex.split("ip netns delete netns_b"))

Fixes: d445516966dc ("net: xdp: support xdp generic on virtual devices")
Reported-by: Stepan Horacek <shoracek@redhat.com>
Suggested-by: Paolo Abeni <pabeni@redhat.com>
Signed-off-by: Toke Høiland-Jørgensen <toke@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

Linux 5.4.21

mmc: core: Rework wp-gpio handling

[ Upstream commit 9073d10b098973519044f5fcdc25586810b435da ]

Use MMC_CAP2_RO_ACTIVE_HIGH flag as indicator if GPIO line is to be
inverted compared to DT/platform-specified polarity. The flag is not used
after init in GPIO mode anyway. No functional changes intended.

Signed-off-by: Michał Mirosław <mirq-linux@rere.qmqm.pl>
Link: https://lore.kernel.org/r/a60f563f11bbff821da2fa2949ca82922b144860.1576031637.git.mirq-linux@rere.qmqm.pl
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

gpio: add gpiod_toggle_active_low()

[ Upstream commit d3a5bcb4a17f1ad072484bb92c42519ff3aba6e1 ]

Add possibility to toggle active-low flag of a gpio descriptor. This is
useful for compatibility code, where defaults are inverted vs DT gpio
flags or the active-low flag is taken from elsewhere.

Acked-by: Linus Walleij <linus.walleij@linaro.org>
Signed-off-by: Michał Mirosław <mirq-linux@rere.qmqm.pl>
Link: https://lore.kernel.org/r/7ce0338e01ad17fa5a227176813941b41a7c35c1.1576031637.git.mirq-linux@rere.qmqm.pl
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

KVM: x86/mmu: Fix struct guest_walker arrays for 5-level paging

[ Upstream commit f6ab0107a4942dbf9a5cf0cca3f37e184870a360 ]

Define PT_MAX_FULL_LEVELS as PT64_ROOT_MAX_LEVEL, i.e. 5, to fix shadow
paging for 5-level guest page tables.  PT_MAX_FULL_LEVELS is used to
size the arrays that track guest pages table information, i.e. using a
"max levels" of 4 causes KVM to access garbage beyond the end of an
array when querying state for level 5 entries.  E.g. FNAME(gpte_changed)
will read garbage and most likely return %true for a level 5 entry,
soft-hanging the guest because FNAME(fetch) will restart the guest
instead of creating SPTEs because it thinks the guest PTE has changed.

Note, KVM doesn't yet support 5-level nested EPT, so PT_MAX_FULL_LEVELS
gets to stay "4" for the PTTYPE_EPT case.

Fixes: 855feb673640 ("KVM: MMU: Add 5 level EPT & Shadow page table support.")
Cc: stable@vger.kernel.org
Signed-off-by: Sean Christopherson <sean.j.christopherson@intel.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

ext4: choose hardlimit when softlimit is larger than hardlimit in ext4_statfs_project()

[ Upstream commit 57c32ea42f8e802bda47010418e25043e0c9337f ]

Setting softlimit larger than hardlimit seems meaningless
for disk quota but currently it is allowed. In this case,
there may be a bit of comfusion for users when they run
df comamnd to directory which has project quota.

For example, we set 20M softlimit and 10M hardlimit of
block usage limit for project quota of test_dir(project id 123).

[root@hades mnt_ext4]# repquota -P -a
*** Report for project quotas on device /dev/loop0
Block grace time: 7days; Inode grace time: 7days
                        Block limits                File limits
Project         used    soft    hard  grace    used  soft  hard  grace
----------------------------------------------------------------------
 0        --      13       0       0              2     0     0
 123      --   10237   20480   10240              5   200   100

The result of df command as below:

[root@hades mnt_ext4]# df -h test_dir
Filesystem      Size  Used Avail Use% Mounted on
/dev/loop0       20M   10M   10M  50% /home/cgxu/test/mnt_ext4

Even though it looks like there is another 10M free space to use,
if we write new data to diretory test_dir(inherit project id),
the write will fail with errno(-EDQUOT).

After this patch, the df result looks like below.

[root@hades mnt_ext4]# df -h test_dir
Filesystem      Size  Used Avail Use% Mounted on
/dev/loop0       10M   10M  3.0K 100% /home/cgxu/test/mnt_ext4

Signed-off-by: Chengguang Xu <cgxu519@mykernel.net>
Reviewed-by: Jan Kara <jack@suse.cz>
Link: https://lore.kernel.org/r/20191016022501.760-1-cgxu519@mykernel.net
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
Signed-off-by: Sasha Levin <sashal@kernel.org>

jbd2: do not clear the BH_Mapped flag when forgetting a metadata buffer

[ Upstream commit c96dceeabf765d0b1b1f29c3bf50a5c01315b820 ]

Commit 904cdbd41d74 ("jbd2: clear dirty flag when revoking a buffer from
an older transaction") set the BH_Freed flag when forgetting a metadata
buffer which belongs to the committing transaction, it indicate the
committing process clear dirty bits when it is done with the buffer. But
it also clear the BH_Mapped flag at the same time, which may trigger
below NULL pointer oops when block_size < PAGE_SIZE.

rmdir 1             kjournald2                 mkdir 2
                    jbd2_journal_commit_transaction
    commit transaction N
jbd2_journal_forget
set_buffer_freed(bh1)
                    jbd2_journal_commit_transaction
                     commit transaction N+1
                     ...
                     clear_buffer_mapped(bh1)
                                               ext4_getblk(bh2 ummapped)
                                               ...
                                               grow_dev_page
                                                init_page_buffers
                                                 bh1->b_private=NULL
                                                 bh2->b_private=NULL
                     jbd2_journal_put_journal_head(jh1)
                      __journal_remove_journal_head(hb1)
       jh1 is NULL and trigger oops

*) Dir entry block bh1 and bh2 belongs to one page, and the bh2 has
   already been unmapped.

For the metadata buffer we forgetting, we should always keep the mapped
flag and clear the dirty flags is enough, so this patch pick out the
these buffers and keep their BH_Mapped flag.

Link: https://lore.kernel.org/r/20200213063821.30455-3-yi.zhang@huawei.com
Fixes: 904cdbd41d74 ("jbd2: clear dirty flag when revoking a buffer from an older transaction")
Reviewed-by: Jan Kara <jack@suse.cz>
Signed-off-by: zhangyi (F) <yi.zhang@huawei.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
Cc: stable@kernel.org
Signed-off-by: Sasha Levin <sashal@kernel.org>

jbd2: move the clearing of b_modified flag to the journal_unmap_buffer()

[ Upstream commit 6a66a7ded12baa6ebbb2e3e82f8cb91382814839 ]

There is no need to delay the clearing of b_modified flag to the
transaction committing time when unmapping the journalled buffer, so
just move it to the journal_unmap_buffer().

Link: https://lore.kernel.org/r/20200213063821.30455-2-yi.zhang@huawei.com
Reviewed-by: Jan Kara <jack@suse.cz>
Signed-off-by: zhangyi (F) <yi.zhang@huawei.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
Cc: stable@kernel.org
Signed-off-by: Sasha Levin <sashal@kernel.org>

Revert "drm/sun4i: drv: Allow framebuffer modifiers in mode config"

commit cf913e9683273f2640501094fa63a67e29f437b3 upstream.

This reverts commit 9db9c0cf5895e4ddde2814360cae7bea9282edd2.

Setting mode_config.allow_fb_modifiers manually is completely
unnecessary. It is set automatically by drm_universal_plane_init() based
on the fact if modifier list is provided or not. Even more, it breaks
DE2 and DE3 as they don't support any modifiers beside linear. Modifiers
aware applications can be confused by provided empty modifier list - at
least linear modifier should be included, but it's not for DE2 and DE3.

Fixes: 9db9c0cf5895 ("drm/sun4i: drv: Allow framebuffer modifiers in mode config")
Signed-off-by: Jernej Skrabec <jernej.skrabec@siol.net>
Reviewed-by: Paul Kocialkowski <paul.kocialkowski@bootlin.com>
Signed-off-by: Maxime Ripard <maxime@cerno.tech>
Link: https://patchwork.freedesktop.org/patch/msgid/20200126065937.9564-1-jernej.skrabec@siol.net
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

NFSv4.1 make cachethis=no for writes

commit cd1b659d8ce7697ee9799b64f887528315b9097b upstream.

Turning caching off for writes on the server should improve performance.

Fixes: fba83f34119a ("NFS: Pass "privileged" value to nfs4_init_sequence()")
Signed-off-by: Olga Kornievskaia <kolga@netapp.com>
Reviewed-by: Trond Myklebust <trond.myklebust@hammerspace.com>
Signed-off-by: Anna Schumaker <Anna.Schumaker@Netapp.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

perf stat: Don't report a null stalled cycles per insn metric

commit 80cc7bb6c104d733bff60ddda09f19139c61507c upstream.

For data collected on machines with front end stalled cycles supported,
such as found on modern AMD CPU families, commit 146540fb545b ("perf
stat: Always separate stalled cycles per insn") introduces a new line in
CSV output with a leading comma that upsets some automated scripts.
Scripts have to use "-e ex_ret_instr" to work around this issue, after
upgrading to a version of perf with that commit.

We could add "if (have_frontend_stalled && !config->csv_sep)" to the not
(total && avg) else clause, to emphasize that CSV users are usually
scripts, and are written to do only what is needed, i.e., they wouldn't
typically invoke "perf stat" without specifying an explicit event list.

But - let alone CSV output - why should users now tolerate a constant
0-reporting extra line in regular terminal output?:

BEFORE:

$ sudo perf stat --all-cpus -einstructions,cycles -- sleep 1

 Performance counter stats for 'system wide':

       181,110,981      instructions              #    0.58  insn per cycle
                                                  #    0.00  stalled cycles per insn
       309,876,469      cycles

       1.002202582 seconds time elapsed

The user would not like to see the now permanent:

  "0.00  stalled cycles per insn"

line fixture, as it gives no useful information.

So this patch removes the printing of the zeroed stalled cycles line
altogether, almost reverting the very original commit fb4605ba47e7
("perf stat: Check for frontend stalled for metrics"), which seems like
it was written to normalize --metric-only column output of common Intel
machines at the time: modern Intel machines have ceased to support the
genericised frontend stalled metrics AFAICT.

AFTER:

$ sudo perf stat --all-cpus -einstructions,cycles -- sleep 1

 Performance counter stats for 'system wide':

       244,071,432      instructions              #    0.69  insn per cycle
       355,353,490      cycles

       1.001862516 seconds time elapsed

Output behaviour when stalled cycles is indeed measured is not affected
(BEFORE == AFTER):

$ sudo perf stat --all-cpus -einstructions,cycles,stalled-cycles-frontend -- sleep 1

 Performance counter stats for 'system wide':

       247,227,799      instructions              #    0.63  insn per cycle
                                                  #    0.26  stalled cycles per insn
       394,745,636      cycles
        63,194,485      stalled-cycles-frontend   #   16.01% frontend cycles idle

       1.002079770 seconds time elapsed

Fixes: 146540fb545b ("perf stat: Always separate stalled cycles per insn")
Signed-off-by: Kim Phillips <kim.phillips@amd.com>
Acked-by: Andi Kleen <ak@linux.intel.com>
Acked-by: Jiri Olsa <jolsa@redhat.com>
Acked-by: Song Liu <songliubraving@fb.com>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Cong Wang <xiyou.wangcong@gmail.com>
Cc: Davidlohr Bueso <dave@stgolabs.net>
Cc: Jin Yao <yao.jin@linux.intel.com>
Cc: Kan Liang <kan.liang@linux.intel.com>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Link: http://lore.kernel.org/lkml/20200207230613.26709-1-kim.phillips@amd.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

KVM: x86: Mask off reserved bit from #DB exception payload

commit 307f1cfa269657c63cfe2c932386fcc24684d9dd upstream.

KVM defines the #DB payload as compatible with the 'pending debug
exceptions' field under VMX, not DR6. Mask off bit 12 when applying the
payload to DR6, as it is reserved on DR6 but not the 'pending debug
exceptions' field.

Fixes: f10c729ff965 ("kvm: vmx: Defer setting of DR6 until #DB delivery")
Signed-off-by: Oliver Upton <oupton@google.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

arm64: dts: fast models: Fix FVP PCI interrupt-map property

commit 3543d7ddd55fe12c37e8a9db846216c51846015b upstream.

The interrupt map for the FVP's PCI node is missing the
parent-unit-address cells for each of the INTx entries, leading to the
kernel code failing to parse the entries correctly.

Add the missing zero cells, which are pretty useless as far as the GIC
is concerned, but that the spec requires. This allows INTx to be usable
on the model, and VFIO to work correctly.

Fixes: fa083b99eb28 ("arm64: dts: fast models: Add DTS fo Base RevC FVP")
Signed-off-by: Marc Zyngier <maz@kernel.org>
Signed-off-by: Sudeep Holla <sudeep.holla@arm.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

cifs: fix mount option display for sec=krb5i

commit 3f6166aaf19902f2f3124b5426405e292e8974dd upstream.

Fix display for sec=krb5i which was wrongly interleaved by cruid,
resulting in string "sec=krb5,cruid=<...>i" instead of
"sec=krb5i,cruid=<...>".

Fixes: 96281b9e46eb ("smb3: for kerberos mounts display the credential uid used")
Signed-off-by: Petr Pavlu <petr.pavlu@suse.com>
Signed-off-by: Steve French <stfrench@microsoft.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

mac80211: fix quiet mode activation in action frames

commit 2bf973ff9b9aeceb8acda629ae65341820d4b35b upstream.

Previously I intended to ignore quiet mode in probe response, however
I ended up ignoring it instead for action frames. As a matter of fact,
this path isn't invoked for probe responses to start with. Just revert
this patch.

Signed-off-by: Sara Sharon <sara.sharon@intel.com>
Fixes: 7976b1e9e3bf ("mac80211: ignore quiet mode in probe")
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>
Link: https://lore.kernel.org/r/20200131111300.891737-15-luca@coelho.fi
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

hwmon: (pmbus/ltc2978) Fix PMBus polling of MFR_COMMON definitions.

commit cf2b012c90e74e85d8aea7d67e48868069cfee0c upstream.

Change 21537dc driver PMBus polling of MFR_COMMON from bits 5/4 to
bits 6/5. This fixs a LTC297X family bug where polling always returns
not busy even when the part is busy. This fixes a LTC388X and
LTM467X bug where polling used PEND and NOT_IN_TRANS, and BUSY was
not polled, which can lead to NACKing of commands. LTC388X and
LTM467X modules now poll BUSY and PEND, increasing reliability by
eliminating NACKing of commands.

Signed-off-by: Mike Jones <michael-a1.jones@analog.com>
Link: https://lore.kernel.org/r/1580234400-2829-2-git-send-email-michael-a1.jones@analog.com
Fixes: e04d1ce9bbb49 ("hwmon: (ltc2978) Add polling for chips requiring it")
Signed-off-by: Guenter Roeck <linux@roeck-us.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>