]> Gitweb @ Texas Instruments - Open Source Git Repositories - git.TI.com/gitweb - ti-linux-kernel/ti-linux-kernel-next.git/log
ti-linux-kernel/ti-linux-kernel-next.git
5 years agoMerge branch 'ti-linux-4.14.y' of git.ti.com:ti-linux-kernel/ti-linux-kernel into... ti-rt-linux-4.14.y-next-20180426 ti-rt-linux-4.14.y-next-20180427
LCPD Auto Merger [Wed, 25 Apr 2018 18:31:44 +0000 (13:31 -0500)]
Merge branch 'ti-linux-4.14.y' of git.ti.com:ti-linux-kernel/ti-linux-kernel into ti-rt-linux-4.14.y

TI-Feature: ti_linux_base_rt
TI-Tree: git@git.ti.com:ti-linux-kernel/ti-linux-kernel.git
TI-Branch: ti-linux-4.14.y

* 'ti-linux-4.14.y' of git.ti.com:ti-linux-kernel/ti-linux-kernel:
  memory: omap-gpmc: Avoid redundant NULL check

Signed-off-by: LCPD Auto Merger <lcpd_integration@list.ti.com>
5 years agoMerge branch 'connectivity-ti-linux-4.14.y' of git://git.ti.com/connectivity-integrat...
LCPD Auto Merger [Wed, 25 Apr 2018 18:13:35 +0000 (13:13 -0500)]
Merge branch 'connectivity-ti-linux-4.14.y' of git://git.ti.com/connectivity-integration-tree/connectivity-ti-linux-kernel into ti-linux-4.14.y

TI-Feature: connectivity
TI-Tree: git://git.ti.com/connectivity-integration-tree/connectivity-ti-linux-kernel.git
TI-Branch: connectivity-ti-linux-4.14.y

* 'connectivity-ti-linux-4.14.y' of git://git.ti.com/connectivity-integration-tree/connectivity-ti-linux-kernel:
  memory: omap-gpmc: Avoid redundant NULL check

Signed-off-by: LCPD Auto Merger <lcpd_integration@list.ti.com>
5 years agoMerge branch 'ti-linux-4.14.y' of git.ti.com:ti-linux-kernel/ti-linux-kernel into...
LCPD Auto Merger [Tue, 24 Apr 2018 11:26:01 +0000 (06:26 -0500)]
Merge branch 'ti-linux-4.14.y' of git.ti.com:ti-linux-kernel/ti-linux-kernel into ti-rt-linux-4.14.y

TI-Feature: ti_linux_base_rt
TI-Tree: git@git.ti.com:ti-linux-kernel/ti-linux-kernel.git
TI-Branch: ti-linux-4.14.y

* 'ti-linux-4.14.y' of git.ti.com:ti-linux-kernel/ti-linux-kernel: (164 commits)
  Linux 4.14.36
  writeback: safer lock nesting
  media: staging: lirc_zilog: incorrect reference counting
  Revert "media: lirc_zilog: driver only sends LIRCCODE"
  iwlwifi: add a bunch of new 9000 PCI IDs
  iwlwifi: add shared clock PHY config flag for some devices
  net: dsa: Discard frames from unused ports
  mm/filemap.c: fix NULL pointer in page_cache_tree_insert()
  autofs: mount point create should honour passed in mode
  Don't leak MNT_INTERNAL away from internal mounts
  rpc_pipefs: fix double-dput()
  orangefs_kill_sb(): deal with allocation failures
  hypfs_kill_super(): deal with failed allocations
  jffs2_kill_sb(): deal with failed allocations
  drm/i915: Correctly handle limited range YCbCr data on VLV/CHV
  mmc: sdhci-pci: Only do AMD tuning for HS200
  fanotify: fix logic of events on child
  udf: Fix leak of UTF-16 surrogates into encoded strings
  powerpc/lib: Fix off-by-one in alternate feature patching
  powerpc/xive: Fix trying to "push" an already active pool VP
  ...

Signed-off-by: LCPD Auto Merger <lcpd_integration@list.ti.com>
5 years agoMerge tag 'v4.14.36' of http://git.kernel.org/pub/scm/linux/kernel/git/stable/linux...
LCPD Auto Merger [Tue, 24 Apr 2018 08:17:45 +0000 (03:17 -0500)]
Merge tag 'v4.14.36' of http://git.kernel.org/pub/scm/linux/kernel/git/stable/linux-stable into ti-linux-4.14.y

This is the 4.14.36 stable release

* tag 'v4.14.36' of http://git.kernel.org/pub/scm/linux/kernel/git/stable/linux-stable: (164 commits)
  Linux 4.14.36
  writeback: safer lock nesting
  media: staging: lirc_zilog: incorrect reference counting
  Revert "media: lirc_zilog: driver only sends LIRCCODE"
  iwlwifi: add a bunch of new 9000 PCI IDs
  iwlwifi: add shared clock PHY config flag for some devices
  net: dsa: Discard frames from unused ports
  mm/filemap.c: fix NULL pointer in page_cache_tree_insert()
  autofs: mount point create should honour passed in mode
  Don't leak MNT_INTERNAL away from internal mounts
  rpc_pipefs: fix double-dput()
  orangefs_kill_sb(): deal with allocation failures
  hypfs_kill_super(): deal with failed allocations
  jffs2_kill_sb(): deal with failed allocations
  drm/i915: Correctly handle limited range YCbCr data on VLV/CHV
  mmc: sdhci-pci: Only do AMD tuning for HS200
  fanotify: fix logic of events on child
  udf: Fix leak of UTF-16 surrogates into encoded strings
  powerpc/lib: Fix off-by-one in alternate feature patching
  powerpc/xive: Fix trying to "push" an already active pool VP
  ...

Signed-off-by: LCPD Auto Merger <lcpd_integration@list.ti.com>
5 years agoLinux 4.14.36
Greg Kroah-Hartman [Tue, 24 Apr 2018 07:36:40 +0000 (09:36 +0200)]
Linux 4.14.36

5 years agowriteback: safer lock nesting
Greg Thelen [Fri, 20 Apr 2018 21:55:42 +0000 (14:55 -0700)]
writeback: safer lock nesting

commit 2e898e4c0a3897ccd434adac5abb8330194f527b upstream.

lock_page_memcg()/unlock_page_memcg() use spin_lock_irqsave/restore() if
the page's memcg is undergoing move accounting, which occurs when a
process leaves its memcg for a new one that has
memory.move_charge_at_immigrate set.

unlocked_inode_to_wb_begin,end() use spin_lock_irq/spin_unlock_irq() if
the given inode is switching writeback domains.  Switches occur when
enough writes are issued from a new domain.

This existing pattern is thus suspicious:
    lock_page_memcg(page);
    unlocked_inode_to_wb_begin(inode, &locked);
    ...
    unlocked_inode_to_wb_end(inode, locked);
    unlock_page_memcg(page);

If both inode switch and process memcg migration are both in-flight then
unlocked_inode_to_wb_end() will unconditionally enable interrupts while
still holding the lock_page_memcg() irq spinlock.  This suggests the
possibility of deadlock if an interrupt occurs before unlock_page_memcg().

    truncate
    __cancel_dirty_page
    lock_page_memcg
    unlocked_inode_to_wb_begin
    unlocked_inode_to_wb_end
    <interrupts mistakenly enabled>
                                    <interrupt>
                                    end_page_writeback
                                    test_clear_page_writeback
                                    lock_page_memcg
                                    <deadlock>
    unlock_page_memcg

Due to configuration limitations this deadlock is not currently possible
because we don't mix cgroup writeback (a cgroupv2 feature) and
memory.move_charge_at_immigrate (a cgroupv1 feature).

If the kernel is hacked to always claim inode switching and memcg
moving_account, then this script triggers lockup in less than a minute:

  cd /mnt/cgroup/memory
  mkdir a b
  echo 1 > a/memory.move_charge_at_immigrate
  echo 1 > b/memory.move_charge_at_immigrate
  (
    echo $BASHPID > a/cgroup.procs
    while true; do
      dd if=/dev/zero of=/mnt/big bs=1M count=256
    done
  ) &
  while true; do
    sync
  done &
  sleep 1h &
  SLEEP=$!
  while true; do
    echo $SLEEP > a/cgroup.procs
    echo $SLEEP > b/cgroup.procs
  done

The deadlock does not seem possible, so it's debatable if there's any
reason to modify the kernel.  I suggest we should to prevent future
surprises.  And Wang Long said "this deadlock occurs three times in our
environment", so there's more reason to apply this, even to stable.
Stable 4.4 has minor conflicts applying this patch.  For a clean 4.4 patch
see "[PATCH for-4.4] writeback: safer lock nesting"
https://lkml.org/lkml/2018/4/11/146

Wang Long said "this deadlock occurs three times in our environment"

[gthelen@google.com: v4]
Link: http://lkml.kernel.org/r/20180411084653.254724-1-gthelen@google.com
[akpm@linux-foundation.org: comment tweaks, struct initialization simplification]
Change-Id: Ibb773e8045852978f6207074491d262f1b3fb613
Link: http://lkml.kernel.org/r/20180410005908.167976-1-gthelen@google.com
Fixes: 682aa8e1a6a1 ("writeback: implement unlocked_inode_to_wb transaction and use it for stat updates")
Signed-off-by: Greg Thelen <gthelen@google.com>
Reported-by: Wang Long <wanglong19@meituan.com>
Acked-by: Wang Long <wanglong19@meituan.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Reviewed-by: Andrew Morton <akpm@linux-foundation.org>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Tejun Heo <tj@kernel.org>
Cc: Nicholas Piggin <npiggin@gmail.com>
Cc: <stable@vger.kernel.org> [v4.2+]
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
[natechancellor: Adjust context due to lack of b93b016313b3b]
Signed-off-by: Nathan Chancellor <natechancellor@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agomedia: staging: lirc_zilog: incorrect reference counting
Sean Young [Sun, 15 Apr 2018 09:51:51 +0000 (10:51 +0100)]
media: staging: lirc_zilog: incorrect reference counting

[not upstream as the driver is deleted in 4.16 - gregkh]

Whenever poll is called, the reference count is increased but never
decreased. This means that on rmmod, the lirc_thread is not stopped,
and will trample over freed memory.

Zilog/Hauppauge IR driver unloaded
BUG: unable to handle kernel paging request at ffffffffc17ba640
Oops: 0010 [#1] SMP
CPU: 1 PID: 667 Comm: zilog-rx-i2c-1 Tainted: P         C OE   4.13.16-302.fc27.x86_64 #1
Hardware name: Gigabyte Technology Co., Ltd. GA-MA790FXT-UD5P/GA-MA790FXT-UD5P, BIOS F6 08/06/2009
task: ffff964eb452ca00 task.stack: ffffb254414dc000
RIP: 0010:0xffffffffc17ba640
RSP: 0018:ffffb254414dfe78 EFLAGS: 00010286
RAX: 0000000000000000 RBX: ffff964ec1b35890 RCX: 0000000000000000
RDX: 0000000000000000 RSI: 0000000000000246 RDI: 0000000000000246
RBP: ffffb254414dff00 R08: 000000000000036e R09: ffff964ecfc8dfd0
R10: ffffb254414dfe78 R11: 00000000000f4240 R12: ffff964ec2bf28a0
R13: ffff964ec1b358a8 R14: ffff964ec1b358d0 R15: ffff964ec1b35800
FS:  0000000000000000(0000) GS:ffff964ecfc80000(0000) knlGS:0000000000000000
CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
CR2: ffffffffc17ba640 CR3: 000000023058c000 CR4: 00000000000006e0
Call Trace:
 kthread+0x125/0x140
 ? kthread_park+0x60/0x60
 ? do_syscall_64+0x67/0x140
 ret_from_fork+0x25/0x30
Code:  Bad RIP value.
RIP: 0xffffffffc17ba640 RSP: ffffb254414dfe78
CR2: ffffffffc17ba640

Note that zilog-rx-i2c-1 should have exited by now, but hasn't due to
the missing put in poll().

This code has been replaced completely in kernel v4.16 by a new driver,
see commit acaa34bf06e9 ("media: rc: implement zilog transmitter"), and
commit f95367a7b758 ("media: staging: remove lirc_zilog driver").

Cc: stable@vger.kernel.org # v4.15- (all up to and including v4.15)
Reported-by: Warren Sturm <warren.sturm@gmail.com>
Tested-by: Warren Sturm <warren.sturm@gmail.com>
Signed-off-by: Sean Young <sean@mess.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agoRevert "media: lirc_zilog: driver only sends LIRCCODE"
Sean Young [Sun, 15 Apr 2018 09:51:50 +0000 (10:51 +0100)]
Revert "media: lirc_zilog: driver only sends LIRCCODE"

[not upstream as the driver is deleted in 4.16 - gregkh]

The lirc config documented here
https://www.blushingpenguin.com/mark/blog/?p=24 uses raw_codes for sending
IR. Each key only has one pulse, which in fact is an index into the
haup-ir-blaster.bin file. Changing the driver to LIRCCODE (although more
accurate) breaks this configuration.

This code has been replaced completely in kernel v4.16 by a new driver,
see commit acaa34bf06e9 ("media: rc: implement zilog transmitter"), and
commit f95367a7b758 ("media: staging: remove lirc_zilog driver").

This reverts commit 89d8a2cc51d1f29ea24a0b44dde13253141190a0.

Fixes: 615cd3fe6ccc ("[media] media: lirc_dev: make better use of file->private_data")
Cc: stable@vger.kernel.org # v4.14-v4.15
Reported-by: Warren Sturm <warren.sturm@gmail.com>
Tested-by: Warren Sturm <warren.sturm@gmail.com>
Signed-off-by: Sean Young <sean@mess.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agoiwlwifi: add a bunch of new 9000 PCI IDs
Luca Coelho [Wed, 28 Mar 2018 08:15:09 +0000 (11:15 +0300)]
iwlwifi: add a bunch of new 9000 PCI IDs

commit 9e5053ad9d590e095829a8bb07adbbdbd893f0f9 upstream.

A lot of new PCI IDs were added for the 9000 series.  Add them to the
list of supported PCI IDs.

Cc: stable@vger.kernel.org # 4.13+
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agoiwlwifi: add shared clock PHY config flag for some devices
Luca Coelho [Wed, 21 Feb 2018 20:18:59 +0000 (22:18 +0200)]
iwlwifi: add shared clock PHY config flag for some devices

commit 86a2b2043af79deff5cf000c5a08847faa4f2ee0 upstream.

Some devices use a shared clock which is very sensitive to variations
and cause trouble in some situations.  We need to set a bit in the phy
configuration to indicate that to the FW.  To make this generic, add a
extra_phy_config_flags element to the device configuration and OR it
into the phy_cfg before sending it to the firmware.  And also create a
set of configurations for devices that use shared clocks and need this
extra bit to be set.

Fixes: c62446d2b028 ("iwlwifi: add new 9460 series PCI IDs")
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agonet: dsa: Discard frames from unused ports
Andrew Lunn [Sat, 7 Apr 2018 18:37:40 +0000 (20:37 +0200)]
net: dsa: Discard frames from unused ports

commit fc5f33768cca7144f8d793205b229d46740d183b upstream.

The Marvell switches under some conditions will pass a frame to the
host with the port being the CPU port. Such frames are invalid, and
should be dropped. Not dropping them can result in a crash when
incrementing the receive statistics for an invalid port.

This has been reworked for 4.14, which does not have the central
dsa_master_find_slave() function, so each tag driver needs to check.

Reported-by: Chris Healy <cphealy@gmail.com>
Fixes: 91da11f870f0 ("net: Distributed Switch Architecture protocol support")
Signed-off-by: Andrew Lunn <andrew@lunn.ch>
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>
5 years agomm/filemap.c: fix NULL pointer in page_cache_tree_insert()
Matthew Wilcox [Fri, 20 Apr 2018 21:56:20 +0000 (14:56 -0700)]
mm/filemap.c: fix NULL pointer in page_cache_tree_insert()

commit abc1be13fd113ddef5e2d807a466286b864caed3 upstream.

f2fs specifies the __GFP_ZERO flag for allocating some of its pages.
Unfortunately, the page cache also uses the mapping's GFP flags for
allocating radix tree nodes.  It always masked off the __GFP_HIGHMEM
flag, and masks off __GFP_ZERO in some paths, but not all.  That causes
radix tree nodes to be allocated with a NULL list_head, which causes
backtraces like:

  __list_del_entry+0x30/0xd0
  list_lru_del+0xac/0x1ac
  page_cache_tree_insert+0xd8/0x110

The __GFP_DMA and __GFP_DMA32 flags would also be able to sneak through
if they are ever used.  Fix them all by using GFP_RECLAIM_MASK at the
innermost location, and remove it from earlier in the callchain.

Link: http://lkml.kernel.org/r/20180411060320.14458-2-willy@infradead.org
Fixes: 449dd6984d0e ("mm: keep page cache radix tree nodes in check")
Signed-off-by: Matthew Wilcox <mawilcox@microsoft.com>
Reported-by: Chris Fries <cfries@google.com>
Debugged-by: Minchan Kim <minchan@kernel.org>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Acked-by: Michal Hocko <mhocko@suse.com>
Reviewed-by: Jan Kara <jack@suse.cz>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agoautofs: mount point create should honour passed in mode
Ian Kent [Fri, 20 Apr 2018 21:55:59 +0000 (14:55 -0700)]
autofs: mount point create should honour passed in mode

commit 1e6306652ba18723015d1b4967fe9de55f042499 upstream.

The autofs file system mkdir inode operation blindly sets the created
directory mode to S_IFDIR | 0555, ingoring the passed in mode, which can
cause selinux dac_override denials.

But the function also checks if the caller is the daemon (as no-one else
should be able to do anything here) so there's no point in not honouring
the passed in mode, allowing the daemon to set appropriate mode when
required.

Link: http://lkml.kernel.org/r/152361593601.8051.14014139124905996173.stgit@pluto.themaw.net
Signed-off-by: Ian Kent <raven@themaw.net>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agoDon't leak MNT_INTERNAL away from internal mounts
Al Viro [Fri, 20 Apr 2018 02:03:08 +0000 (22:03 -0400)]
Don't leak MNT_INTERNAL away from internal mounts

commit 16a34adb9392b2fe4195267475ab5b472e55292c upstream.

We want it only for the stuff created by SB_KERNMOUNT mounts, *not* for
their copies.  As it is, creating a deep stack of bindings of /proc/*/ns/*
somewhere in a new namespace and exiting yields a stack overflow.

Cc: stable@kernel.org
Reported-by: Alexander Aring <aring@mojatatu.com>
Bisected-by: Kirill Tkhai <ktkhai@virtuozzo.com>
Tested-by: Kirill Tkhai <ktkhai@virtuozzo.com>
Tested-by: Alexander Aring <aring@mojatatu.com>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agorpc_pipefs: fix double-dput()
Al Viro [Tue, 3 Apr 2018 05:15:46 +0000 (01:15 -0400)]
rpc_pipefs: fix double-dput()

commit 4a3877c4cedd95543f8726b0a98743ed8db0c0fb upstream.

if we ever hit rpc_gssd_dummy_depopulate() dentry passed to
it has refcount equal to 1.  __rpc_rmpipe() drops it and
dput() done after that hits an already freed dentry.

Cc: stable@kernel.org
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agoorangefs_kill_sb(): deal with allocation failures
Al Viro [Tue, 3 Apr 2018 04:13:17 +0000 (00:13 -0400)]
orangefs_kill_sb(): deal with allocation failures

commit 659038428cb43a66e3eff71e2c845c9de3611a98 upstream.

orangefs_fill_sb() might've failed to allocate ORANGEFS_SB(s); don't
oops in that case.

Cc: stable@kernel.org
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agohypfs_kill_super(): deal with failed allocations
Al Viro [Tue, 3 Apr 2018 03:50:31 +0000 (23:50 -0400)]
hypfs_kill_super(): deal with failed allocations

commit a24cd490739586a7d2da3549a1844e1d7c4f4fc4 upstream.

hypfs_fill_super() might fail to allocate sbi; hypfs_kill_super()
should not oops on that.

Cc: stable@vger.kernel.org
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agojffs2_kill_sb(): deal with failed allocations
Al Viro [Tue, 3 Apr 2018 03:56:44 +0000 (23:56 -0400)]
jffs2_kill_sb(): deal with failed allocations

commit c66b23c2840446a82c389e4cb1a12eb2a71fa2e4 upstream.

jffs2_fill_super() might fail to allocate jffs2_sb_info;
jffs2_kill_sb() must survive that.

Cc: stable@kernel.org
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agodrm/i915: Correctly handle limited range YCbCr data on VLV/CHV
Ville Syrjälä [Wed, 14 Feb 2018 19:23:23 +0000 (21:23 +0200)]
drm/i915: Correctly handle limited range YCbCr data on VLV/CHV

commit 5deae9191130db6b617c94fb261804597cf9b508 upstream.

Turns out the VLV/CHV fixed function sprite CSC expects full range
data as input. We've been feeding it limited range data to it all
along. To expand the data out to full range we'll use the color
correction registers (brightness, contrast, and saturation).

On CHV pipe B we were actually doing the right thing already because we
progammed the custom CSC matrix to do expect limited range input. Now
that well pre-expand the data out with the color correction unit, we
need to change the CSC matrix to operate with full range input instead.

This should make the sprite output of the other pipes match the sprite
output of pipe B reasonably well. Looking at the resulting pipe CRCs,
there can be a slight difference in the output, but as I don't know
the formula used by the fixed function CSC of the other pipes, I don't
think it's worth the effort to try to match the output exactly. It
might not even be possible due to difference in internal precision etc.

One slight caveat here is that the color correction registers are single
bufferred, so we should really be updating them during vblank, but we
still don't have a mechanism for that, so just toss in another FIXME.

v2: Rebase
v3: s/bri/brightness/ s/con/contrast/ (Shashank)
v4: Clarify the constants and math (Shashank)

Cc: Harry Wentland <harry.wentland@amd.com>
Cc: Daniel Vetter <daniel@ffwll.ch>
Cc: Daniel Stone <daniel@fooishbar.org>
Cc: Russell King - ARM Linux <linux@armlinux.org.uk>
Cc: Ilia Mirkin <imirkin@alum.mit.edu>
Cc: Hans Verkuil <hverkuil@xs4all.nl>
Cc: Shashank Sharma <shashank.sharma@intel.com>
Cc: Uma Shankar <uma.shankar@intel.com>
Cc: Jyri Sarha <jsarha@ti.com>
Cc: "Tang, Jun" <jun.tang@intel.com>
Reported-by: "Tang, Jun" <jun.tang@intel.com>
Cc: stable@vger.kernel.org
Fixes: 7f1f3851feb0 ("drm/i915: sprite support for ValleyView v4")
Reviewed-by: Shashank Sharma <shashank.sharma@intel.com>
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20180214192327.3250-5-ville.syrjala@linux.intel.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agommc: sdhci-pci: Only do AMD tuning for HS200
Daniel Kurtz [Fri, 6 Apr 2018 22:07:59 +0000 (16:07 -0600)]
mmc: sdhci-pci: Only do AMD tuning for HS200

commit 300ad8992913025b4294d4fc37b6bfff4a8b7ad1 upstream.

Commit c31165d7400b ("mmc: sdhci-pci: Add support for HS200 tuning mode
on AMD, eMMC-4.5.1") added a HS200 tuning method for use with AMD SDHCI
controllers.  As described in the commit subject, this tuning is specific
for HS200.  However, as implemented, this method is used for all host
timings, because platform_execute_tuning, if it exists, is called
unconditionally by sdhci_execute_tuning().  This breaks tuning when using
the AMD controller with, for example, a DDR50 SD card.

Instead, we can implement an amd execute_tuning wrapper callback, and
then conditionally do the HS200 specific tuning for HS200, and otherwise
call back to the standard sdhci_execute_tuning().

Signed-off-by: Daniel Kurtz <djkurtz@chromium.org>
Acked-by: Shyam Sundar S K <Shyam-sundar.S-k@amd.com>
Acked-by: Adrian Hunter <adrian.hunter@intel.com>
Fixes: c31165d7400b ("mmc: sdhci-pci: Add support for HS200 tuning mode on AMD, eMMC-4.5.1")
Cc: stable@vger.kernel.org # v4.11+
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agofanotify: fix logic of events on child
Amir Goldstein [Wed, 4 Apr 2018 20:42:18 +0000 (23:42 +0300)]
fanotify: fix logic of events on child

commit 54a307ba8d3cd00a3902337ffaae28f436eeb1a4 upstream.

When event on child inodes are sent to the parent inode mark and
parent inode mark was not marked with FAN_EVENT_ON_CHILD, the event
will not be delivered to the listener process. However, if the same
process also has a mount mark, the event to the parent inode will be
delivered regadless of the mount mark mask.

This behavior is incorrect in the case where the mount mark mask does
not contain the specific event type. For example, the process adds
a mark on a directory with mask FAN_MODIFY (without FAN_EVENT_ON_CHILD)
and a mount mark with mask FAN_CLOSE_NOWRITE (without FAN_ONDIR).

A modify event on a file inside that directory (and inside that mount)
should not create a FAN_MODIFY event, because neither of the marks
requested to get that event on the file.

Fixes: 1968f5eed54c ("fanotify: use both marks when possible")
Cc: stable <stable@vger.kernel.org>
Signed-off-by: Amir Goldstein <amir73il@gmail.com>
Signed-off-by: Jan Kara <jack@suse.cz>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agoudf: Fix leak of UTF-16 surrogates into encoded strings
Jan Kara [Thu, 12 Apr 2018 15:22:23 +0000 (17:22 +0200)]
udf: Fix leak of UTF-16 surrogates into encoded strings

commit 44f06ba8297c7e9dfd0e49b40cbe119113cca094 upstream.

OSTA UDF specification does not mention whether the CS0 charset in case
of two bytes per character encoding should be treated in UTF-16 or
UCS-2. The sample code in the standard does not treat UTF-16 surrogates
in any special way but on systems such as Windows which work in UTF-16
internally, filenames would be treated as being in UTF-16 effectively.
In Linux it is more difficult to handle characters outside of Base
Multilingual plane (beyond 0xffff) as NLS framework works with 2-byte
characters only. Just make sure we don't leak UTF-16 surrogates into the
resulting string when loading names from the filesystem for now.

CC: stable@vger.kernel.org # >= v4.6
Reported-by: Mingye Wang <arthur200126@gmail.com>
Signed-off-by: Jan Kara <jack@suse.cz>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agopowerpc/lib: Fix off-by-one in alternate feature patching
Michael Ellerman [Mon, 16 Apr 2018 13:25:19 +0000 (23:25 +1000)]
powerpc/lib: Fix off-by-one in alternate feature patching

commit b8858581febb050688e276b956796bc4a78299ed upstream.

When we patch an alternate feature section, we have to adjust any
relative branches that branch out of the alternate section.

But currently we have a bug if we have a branch that points to past
the last instruction of the alternate section, eg:

  FTR_SECTION_ELSE
  1:     b       2f
         or      6,6,6
  2:
  ALT_FTR_SECTION_END(...)
         nop

This will result in a relative branch at 1 with a target that equals
the end of the alternate section.

That branch does not need adjusting when it's moved to the non-else
location. Currently we do adjust it, resulting in a branch that goes
off into the link-time location of the else section, which is junk.

The fix is to not patch branches that have a target == end of the
alternate section.

Fixes: d20fe50a7b3c ("KVM: PPC: Book3S HV: Branch inside feature section")
Fixes: 9b1a735de64c ("powerpc: Add logic to patch alternative feature sections")
Cc: stable@vger.kernel.org # v2.6.27+
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agopowerpc/xive: Fix trying to "push" an already active pool VP
Benjamin Herrenschmidt [Wed, 11 Apr 2018 05:17:59 +0000 (15:17 +1000)]
powerpc/xive: Fix trying to "push" an already active pool VP

commit b32e56e5a87a1f9243db92bc7a5df0ffb4627cfb upstream.

When setting up a CPU, we "push" (activate) a pool VP for it.

However it's an error to do so if it already has an active
pool VP.

This happens when doing soft CPU hotplug on powernv since we
don't tear down the CPU on unplug. The HW flags the error which
gets captured by the diagnostics.

Fix this by making sure to "pull" out any already active pool
first.

Fixes: 243e25112d06 ("powerpc/xive: Native exploitation of the XIVE interrupt controller")
Cc: stable@vger.kernel.org # v4.12+
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agopowerpc/eeh: Fix enabling bridge MMIO windows
Michael Neuling [Wed, 11 Apr 2018 03:37:58 +0000 (13:37 +1000)]
powerpc/eeh: Fix enabling bridge MMIO windows

commit 13a83eac373c49c0a081cbcd137e79210fe78acd upstream.

On boot we save the configuration space of PCIe bridges. We do this so
when we get an EEH event and everything gets reset that we can restore
them.

Unfortunately we save this state before we've enabled the MMIO space
on the bridges. Hence if we have to reset the bridge when we come back
MMIO is not enabled and we end up taking an PE freeze when the driver
starts accessing again.

This patch forces the memory/MMIO and bus mastering on when restoring
bridges on EEH. Ideally we'd do this correctly by saving the
configuration space writes later, but that will have to come later in
a larger EEH rewrite. For now we have this simple fix.

The original bug can be triggered on a boston machine by doing:
  echo 0x8000000000000000 > /sys/kernel/debug/powerpc/PCI0001/err_injct_outbound
On boston, this PHB has a PCIe switch on it.  Without this patch,
you'll see two EEH events, 1 expected and 1 the failure we are fixing
here. The second EEH event causes the anything under the PHB to
disappear (i.e. the i40e eth).

With this patch, only 1 EEH event occurs and devices properly recover.

Fixes: 652defed4875 ("powerpc/eeh: Check PCIe link after reset")
Cc: stable@vger.kernel.org # v3.11+
Reported-by: Pridhiviraj Paidipeddi <ppaidipe@linux.vnet.ibm.com>
Signed-off-by: Michael Neuling <mikey@neuling.org>
Acked-by: Russell Currey <ruscur@russell.cc>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agoMIPS: memset.S: Fix clobber of v1 in last_fixup
Matt Redfearn [Tue, 17 Apr 2018 15:40:00 +0000 (16:40 +0100)]
MIPS: memset.S: Fix clobber of v1 in last_fixup

commit c96eebf07692e53bf4dd5987510d8b550e793598 upstream.

The label .Llast_fixup\@ is jumped to on page fault within the final
byte set loop of memset (on < MIPSR6 architectures). For some reason, in
this fault handler, the v1 register is randomly set to a2 & STORMASK.
This clobbers v1 for the calling function. This can be observed with the
following test code:

static int __init __attribute__((optimize("O0"))) test_clear_user(void)
{
  register int t asm("v1");
  char *test;
  int j, k;

  pr_info("\n\n\nTesting clear_user\n");
  test = vmalloc(PAGE_SIZE);

  for (j = 256; j < 512; j++) {
    t = 0xa5a5a5a5;
    if ((k = clear_user(test + PAGE_SIZE - 256, j)) != j - 256) {
        pr_err("clear_user (%px %d) returned %d\n", test + PAGE_SIZE - 256, j, k);
    }
    if (t != 0xa5a5a5a5) {
       pr_err("v1 was clobbered to 0x%x!\n", t);
    }
  }

  return 0;
}
late_initcall(test_clear_user);

Which demonstrates that v1 is indeed clobbered (MIPS64):

Testing clear_user
v1 was clobbered to 0x1!
v1 was clobbered to 0x2!
v1 was clobbered to 0x3!
v1 was clobbered to 0x4!
v1 was clobbered to 0x5!
v1 was clobbered to 0x6!
v1 was clobbered to 0x7!

Since the number of bytes that could not be set is already contained in
a2, the andi placing a value in v1 is not necessary and actively
harmful in clobbering v1.

Reported-by: James Hogan <jhogan@kernel.org>
Signed-off-by: Matt Redfearn <matt.redfearn@mips.com>
Cc: Ralf Baechle <ralf@linux-mips.org>
Cc: linux-mips@linux-mips.org
Cc: stable@vger.kernel.org
Patchwork: https://patchwork.linux-mips.org/patch/19109/
Signed-off-by: James Hogan <jhogan@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agoMIPS: memset.S: Fix return of __clear_user from Lpartial_fixup
Matt Redfearn [Tue, 17 Apr 2018 14:52:21 +0000 (15:52 +0100)]
MIPS: memset.S: Fix return of __clear_user from Lpartial_fixup

commit daf70d89f80c6e1772233da9e020114b1254e7e0 upstream.

The __clear_user function is defined to return the number of bytes that
could not be cleared. From the underlying memset / bzero implementation
this means setting register a2 to that number on return. Currently if a
page fault is triggered within the memset_partial block, the value
loaded into a2 on return is meaningless.

The label .Lpartial_fixup\@ is jumped to on page fault. In order to work
out how many bytes failed to copy, the exception handler should find how
many bytes left in the partial block (andi a2, STORMASK), add that to
the partial block end address (a2), and subtract the faulting address to
get the remainder. Currently it incorrectly subtracts the partial block
start address (t1), which has additionally been clobbered to generate a
jump target in memset_partial. Fix this by adding the block end address
instead.

This issue was found with the following test code:
      int j, k;
      for (j = 0; j < 512; j++) {
        if ((k = clear_user(NULL, j)) != j) {
           pr_err("clear_user (NULL %d) returned %d\n", j, k);
        }
      }
Which now passes on Creator Ci40 (MIPS32) and Cavium Octeon II (MIPS64).

Suggested-by: James Hogan <jhogan@kernel.org>
Signed-off-by: Matt Redfearn <matt.redfearn@mips.com>
Cc: Ralf Baechle <ralf@linux-mips.org>
Cc: linux-mips@linux-mips.org
Cc: stable@vger.kernel.org
Patchwork: https://patchwork.linux-mips.org/patch/19108/
Signed-off-by: James Hogan <jhogan@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agoMIPS: memset.S: EVA & fault support for small_memset
Matt Redfearn [Thu, 29 Mar 2018 09:28:23 +0000 (10:28 +0100)]
MIPS: memset.S: EVA & fault support for small_memset

commit 8a8158c85e1e774a44fbe81106fa41138580dfd1 upstream.

The MIPS kernel memset / bzero implementation includes a small_memset
branch which is used when the region to be set is smaller than a long (4
bytes on 32bit, 8 bytes on 64bit). The current small_memset
implementation uses a simple store byte loop to write the destination.
There are 2 issues with this implementation:

1. When EVA mode is active, user and kernel address spaces may overlap.
Currently the use of the sb instruction means kernel mode addressing is
always used and an intended write to userspace may actually overwrite
some critical kernel data.

2. If the write triggers a page fault, for example by calling
__clear_user(NULL, 2), instead of gracefully handling the fault, an OOPS
is triggered.

Fix these issues by replacing the sb instruction with the EX() macro,
which will emit EVA compatible instuctions as required. Additionally
implement a fault fixup for small_memset which sets a2 to the number of
bytes that could not be cleared (as defined by __clear_user).

Reported-by: Chuanhua Lei <chuanhua.lei@intel.com>
Signed-off-by: Matt Redfearn <matt.redfearn@mips.com>
Cc: Ralf Baechle <ralf@linux-mips.org>
Cc: linux-mips@linux-mips.org
Cc: stable@vger.kernel.org
Patchwork: https://patchwork.linux-mips.org/patch/18975/
Signed-off-by: James Hogan <jhogan@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agoMIPS: uaccess: Add micromips clobbers to bzero invocation
Matt Redfearn [Tue, 17 Apr 2018 15:40:01 +0000 (16:40 +0100)]
MIPS: uaccess: Add micromips clobbers to bzero invocation

commit b3d7e55c3f886493235bfee08e1e5a4a27cbcce8 upstream.

The micromips implementation of bzero additionally clobbers registers t7
& t8. Specify this in the clobbers list when invoking bzero.

Fixes: 26c5e07d1478 ("MIPS: microMIPS: Optimise 'memset' core library function.")
Reported-by: James Hogan <jhogan@kernel.org>
Signed-off-by: Matt Redfearn <matt.redfearn@mips.com>
Cc: Ralf Baechle <ralf@linux-mips.org>
Cc: linux-mips@linux-mips.org
Cc: <stable@vger.kernel.org> # 3.10+
Patchwork: https://patchwork.linux-mips.org/patch/19110/
Signed-off-by: James Hogan <jhogan@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agoHID: wacom: bluetooth: send exit report for recent Bluetooth devices
Aaron Armstrong Skomra [Wed, 4 Apr 2018 21:24:11 +0000 (14:24 -0700)]
HID: wacom: bluetooth: send exit report for recent Bluetooth devices

commit 619d3a2922ce623ca2eca443cc936810d328317c upstream.

The code path for recent Bluetooth devices omits an exit report which
resets all the values of the device.

Fixes: 4922cd26f0 ("HID: wacom: Support 2nd-gen Intuos Pro's Bluetooth classic interface")
Cc: <stable@vger.kernel.org> # 4.11
Signed-off-by: Aaron Armstrong Skomra <aaron.skomra@wacom.com>
Reviewed-by: Ping Cheng <ping.cheng@wacom.com>
Signed-off-by: Jiri Kosina <jkosina@suse.cz>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agoHID: hidraw: Fix crash on HIDIOCGFEATURE with a destroyed device
Rodrigo Rivas Costa [Thu, 5 Apr 2018 23:09:36 +0000 (01:09 +0200)]
HID: hidraw: Fix crash on HIDIOCGFEATURE with a destroyed device

commit a955358d54695e4ad9f7d6489a7ac4d69a8fc711 upstream.

Doing `ioctl(HIDIOCGFEATURE)` in a tight loop on a hidraw device
and then disconnecting the device, or unloading the driver, can
cause a NULL pointer dereference.

When a hidraw device is destroyed it sets 0 to `dev->exist`.
Most functions check 'dev->exist' before doing its work, but
`hidraw_get_report()` was missing that check.

Cc: stable@vger.kernel.org
Signed-off-by: Rodrigo Rivas Costa <rodrigorivascosta@gmail.com>
Signed-off-by: Jiri Kosina <jkosina@suse.cz>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agoHID: input: fix battery level reporting on BT mice
Dmitry Torokhov [Tue, 3 Apr 2018 17:52:20 +0000 (10:52 -0700)]
HID: input: fix battery level reporting on BT mice

commit 2e210bbb7429cdcf1a1a3ad00c1bf98bd9bf2452 upstream.

The commit 581c4484769e ("HID: input: map digitizer battery usage")
assumed that devices having input (qas opposed to feature) report for
battery strength would report the data on their own, without the need to
be polled by the kernel; unfortunately it is not so. Many wireless mice
do not send unsolicited reports with battery strength data and have to
be polled explicitly. As a complication, stylus devices on digitizers
are not normally connected to the base and thus can not be polled - the
base can only determine battery strength in the stylus when it is in
proximity.

To solve this issue, we add a special flag that tells the kernel
to avoid polling the device (and expect unsolicited reports) and set it
when report field with physical usage of digitizer stylus (HID_DG_STYLUS).
Unless this flag is set, and we have not seen the unsolicited reports,
the kernel will attempt to poll the device when userspace attempts to
read "capacity" and "state" attributes of power_supply object
corresponding to the devices battery.

Fixes: 581c4484769e ("HID: input: map digitizer battery usage")
Bugzilla: https://bugzilla.kernel.org/show_bug.cgi?id=198095
Cc: stable@vger.kernel.org
Reported-and-tested-by: Martin van Es <martin@mrvanes.com>
Signed-off-by: Dmitry Torokhov <dmitry.torokhov@gmail.com>
Signed-off-by: Jiri Kosina <jkosina@suse.cz>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agorandom: add new ioctl RNDRESEEDCRNG
Theodore Ts'o [Wed, 11 Apr 2018 20:32:17 +0000 (16:32 -0400)]
random: add new ioctl RNDRESEEDCRNG

commit d848e5f8e1ebdb227d045db55fe4f825e82965fa upstream.

Add a new ioctl which forces the the crng to be reseeded.

Signed-off-by: Theodore Ts'o <tytso@mit.edu>
Cc: stable@kernel.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agorandom: crng_reseed() should lock the crng instance that it is modifying
Theodore Ts'o [Thu, 12 Apr 2018 04:50:45 +0000 (00:50 -0400)]
random: crng_reseed() should lock the crng instance that it is modifying

commit 0bb29a849a6433b72e249eea7695477b02056e94 upstream.

Reported-by: Jann Horn <jannh@google.com>
Fixes: 1e7f583af67b ("random: make /dev/urandom scalable for silly...")
Cc: stable@kernel.org # 4.8+
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
Reviewed-by: Jann Horn <jannh@google.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agorandom: use a different mixing algorithm for add_device_randomness()
Theodore Ts'o [Wed, 11 Apr 2018 18:58:27 +0000 (14:58 -0400)]
random: use a different mixing algorithm for add_device_randomness()

commit dc12baacb95f205948f64dc936a47d89ee110117 upstream.

add_device_randomness() use of crng_fast_load() was highly
problematic.  Some callers of add_device_randomness() can pass in a
large amount of static information.  This would immediately promote
the crng_init state from 0 to 1, without really doing much to
initialize the primary_crng's internal state with something even
vaguely unpredictable.

Since we don't have the speed constraints of add_interrupt_randomness(),
we can do a better job mixing in the what unpredictability a device
driver or architecture maintainer might see fit to give us, and do it
in a way which does not bump the crng_init_cnt variable.

Also, since add_device_randomness() doesn't bump any entropy
accounting in crng_init state 0, mix the device randomness into the
input_pool entropy pool as well.  This is related to CVE-2018-1108.

Reported-by: Jann Horn <jannh@google.com>
Fixes: ee7998c50c26 ("random: do not ignore early device randomness")
Cc: stable@kernel.org # 4.13+
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agorandom: fix crng_ready() test
Theodore Ts'o [Wed, 11 Apr 2018 17:27:52 +0000 (13:27 -0400)]
random: fix crng_ready() test

commit 43838a23a05fbd13e47d750d3dfd77001536dd33 upstream.

The crng_init variable has three states:

0: The CRNG is not initialized at all
1: The CRNG has a small amount of entropy, hopefully good enough for
   early-boot, non-cryptographical use cases
2: The CRNG is fully initialized and we are sure it is safe for
   cryptographic use cases.

The crng_ready() function should only return true once we are in the
last state.  This addresses CVE-2018-1108.

Reported-by: Jann Horn <jannh@google.com>
Fixes: e192be9d9a30 ("random: replace non-blocking pool...")
Cc: stable@kernel.org # 4.8+
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
Reviewed-by: Jann Horn <jannh@google.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agoALSA: hda/realtek - adjust the location of one mic
Hui Wang [Thu, 19 Apr 2018 05:29:05 +0000 (13:29 +0800)]
ALSA: hda/realtek - adjust the location of one mic

commit a3dafb2200bf3c13905a088e82ae11f1eb275a83 upstream.

There are two front mics on this machine, if we don't adjust the
location for one of them, they will have the same mixer name,
pulseaudio can't handle this situation.

After applying this FIXUP, they will have different mixer name,
then pulseaudio can handle them correctly.

Cc: <stable@vger.kernel.org>
Signed-off-by: Hui Wang <hui.wang@canonical.com>
Signed-off-by: Takashi Iwai <tiwai@suse.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agoALSA: hda/realtek - set PINCFG_HEADSET_MIC to parse_flags
Hui Wang [Thu, 19 Apr 2018 05:29:04 +0000 (13:29 +0800)]
ALSA: hda/realtek - set PINCFG_HEADSET_MIC to parse_flags

commit 3ce0d5aa265bcc0a4b281cb0cabf92491276101b upstream.

Otherwise, the pin will be regarded as microphone, and the jack name
is "Mic Phantom", it is always on in the pulseaudio even nothing is
plugged into the jack. So the UI is confusing to users since the
microphone always shows up in the UI even there is no microphone
plugged.

After adding this flag, the jack name is "Headset Mic Phantom", then
the pulseaudio can handle its detection correctly.

Fixes: f0ba9d699e5c ("ALSA: hda/realtek - Fix Dell headset Mic can't record")
Cc: <stable@vger.kernel.org>
Signed-off-by: Hui Wang <hui.wang@canonical.com>
Signed-off-by: Takashi Iwai <tiwai@suse.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agoALSA: hda - New VIA controller suppor no-snoop path
David Wang [Mon, 16 Apr 2018 09:48:09 +0000 (17:48 +0800)]
ALSA: hda - New VIA controller suppor no-snoop path

commit af52f9982e410edac21ca4b49563053ffc9da1eb upstream.

This patch is used to tell kernel that new VIA HDAC controller also
support no-snoop path.

[ minor coding style fix by tiwai ]

Signed-off-by: David Wang <davidwang@zhaoxin.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Takashi Iwai <tiwai@suse.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agoALSA: rawmidi: Fix missing input substream checks in compat ioctls
Takashi Iwai [Thu, 19 Apr 2018 16:16:15 +0000 (18:16 +0200)]
ALSA: rawmidi: Fix missing input substream checks in compat ioctls

commit 8a56ef4f3ffba9ebf4967b61ef600b0a7ba10f11 upstream.

Some rawmidi compat ioctls lack of the input substream checks
(although they do check only for rfile->output).  This many eventually
lead to an Oops as NULL substream is passed to the rawmidi core
functions.

Fix it by adding the proper checks before each function call.

The bug was spotted by syzkaller.

Reported-by: syzbot+f7a0348affc3b67bc617@syzkaller.appspotmail.com
Cc: <stable@vger.kernel.org>
Signed-off-by: Takashi Iwai <tiwai@suse.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agoALSA: line6: Use correct endpoint type for midi output
Fabián Inostroza [Thu, 12 Apr 2018 03:37:35 +0000 (00:37 -0300)]
ALSA: line6: Use correct endpoint type for midi output

commit 7ecb46e9ee9af18e304eb9e7d6804c59a408e846 upstream.

Sending MIDI messages to a PODxt through the USB connection shows
"usb_submit_urb failed" in dmesg and the message is not received by
the POD.

The error is caused because in the funcion send_midi_async() in midi.c
there is a call to usb_sndbulkpipe() for endpoint 3 OUT, but the PODxt
USB descriptor shows that this endpoint it's an interrupt endpoint.

Patch tested with PODxt only.

[ The bug has been present from the very beginning in the staging
  driver time, but Fixes below points to the commit moving to sound/
  directory so that the fix can be cleanly applied -- tiwai ]

Fixes: 61864d844c29 ("ALSA: move line6 usb driver into sound/usb")
Signed-off-by: Fabián Inostroza <fabianinostroza@udec.cl>
Cc: <stable@vger.kernel.org>
Signed-off-by: Takashi Iwai <tiwai@suse.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agodrm/radeon: Fix PCIe lane width calculation
Paul Parsons [Sat, 2 Apr 2016 11:32:30 +0000 (12:32 +0100)]
drm/radeon: Fix PCIe lane width calculation

commit 85e290d92b4b794d0c758c53007eb4248d385386 upstream.

Two years ago I tried an AMD Radeon E8860 embedded GPU with the drm driver.
The dmesg output included driver warnings about an invalid PCIe lane width.
Tracking the problem back led to si_set_pcie_lane_width_in_smc().
The calculation of the lane widths via ATOM_PPLIB_PCIE_LINK_WIDTH_MASK and
ATOM_PPLIB_PCIE_LINK_WIDTH_SHIFT macros did not increment the resulting
value, per the comment in pptable.h ("lanes - 1"), and per usage elsewhere.
Applying the increment silenced the warnings.
The code has not changed since, so either my analysis was incorrect or the
bug has gone unnoticed. Hence submitting this as an RFC.

Acked-by: Christian König <christian.koenig@amd.com>
Acked-by: Chunming Zhou <david1.zhou@amd.com>
Signed-off-by: Paul Parsons <lost.distance@yahoo.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Cc: stable@vger.kernel.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agodrm/radeon: add PX quirk for Asus K73TK
Nico Sneck [Sat, 7 Apr 2018 15:13:04 +0000 (15:13 +0000)]
drm/radeon: add PX quirk for Asus K73TK

commit b1550359d1eb392ee54f7cf47cffcfe0a602f6a7 upstream.

With this the dGPU turns on correctly.

Signed-off-by: Nico Sneck <nicosneck@hotmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Cc: stable@vger.kernel.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agodrm/rockchip: Clear all interrupts before requesting the IRQ
Marc Zyngier [Tue, 20 Feb 2018 13:01:18 +0000 (13:01 +0000)]
drm/rockchip: Clear all interrupts before requesting the IRQ

commit 5f9e93fed4d45e9a8f84728aff1a8f2ab8922902 upstream.

Calling request_irq() followed by disable_irq() is usually a bad idea,
specially if the interrupt can be pending, and you're not yet in a
position to handle it.

This is exactly what happens on my kevin system when rebooting in a
second kernel using kexec: Some interrupt is left pending from
the previous kernel, and we take it too early, before disable_irq()
could do anything.

Let's clear the pending interrupts as we initialize the HW, and move
the interrupt request after that point. This ensures that we're in
a sane state when the interrupt is requested.

Cc: stable@vger.kernel.org
Signed-off-by: Marc Zyngier <marc.zyngier@arm.com>
[adapted to recent rockchip-drm changes]
Signed-off-by: Heiko Stuebner <heiko@sntech.de>
Link: https://patchwork.freedesktop.org/patch/msgid/20180220130120.5254-2-marc.zyngier@arm.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agodrm/amdgpu/si: implement get/set pcie_lanes asic callback
Alex Deucher [Tue, 3 Apr 2018 17:54:33 +0000 (12:54 -0500)]
drm/amdgpu/si: implement get/set pcie_lanes asic callback

commit 20ca25e86c56f5490bdc80318f4fc06466e4c21b upstream.

Required for dpm setup on some asics. Fixes a NULL dereference
on asics that require it.

Acked-by: Christian König <christian.koenig@amd.com>
Bug: https://bugs.freedesktop.org/show_bug.cgi?id=102553
Tested-by: Abel Garcia Dorta <mercuriete@yahoo.es>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Cc: stable@vger.kernel.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agodrm/amdgpu: Fix PCIe lane width calculation
Alex Deucher [Mon, 2 Apr 2018 17:29:26 +0000 (12:29 -0500)]
drm/amdgpu: Fix PCIe lane width calculation

commit 41212e2fe72b26ded7ed78224d9eab720c2891e2 upstream.

The calculation of the lane widths via ATOM_PPLIB_PCIE_LINK_WIDTH_MASK and
ATOM_PPLIB_PCIE_LINK_WIDTH_SHIFT macros did not increment the resulting
value, per the comment in pptable.h ("lanes - 1"), and per usage elsewhere.
Port of the radeon fix to amdgpu.

Acked-by: Christian König <christian.koenig@amd.com>
Acked-by: Chunming Zhou <david1.zhou@amd.com>
Bug: https://bugs.freedesktop.org/show_bug.cgi?id=102553
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Cc: stable@vger.kernel.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agodrm/amdgpu/sdma: fix mask in emit_pipeline_sync
Alex Deucher [Tue, 27 Mar 2018 20:53:52 +0000 (15:53 -0500)]
drm/amdgpu/sdma: fix mask in emit_pipeline_sync

commit 4a8e06f7aad797e92413a3042d09d3b385fa1fda upstream.

Needs to be a 32 bit mask.

Acked-by: Huang Rui <ray.huang@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Cc: stable@vger.kernel.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agodrm/amdgpu: Fix always_valid bos multiple LRU insertions.
Bas Nieuwenhuizen [Wed, 31 Jan 2018 12:58:55 +0000 (13:58 +0100)]
drm/amdgpu: Fix always_valid bos multiple LRU insertions.

commit a20ee0b1f8b42e2568f3a4408003d22b2dfcc706 upstream.

If these bos are evicted and are in the validated list
things blow up, so do not put them in there. Notably,
that tries to add the bo to the LRU twice, which results
in a BUG_ON in ttm_bo.c.

While for the bo_list an alternative would be to not allow
always valid bos in there, that does not work for the user
fence.

v2: Fixed whitespace issue pointed out by checkpatch.pl

Signed-off-by: Bas Nieuwenhuizen <basni@chromium.org>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Cc: stable@vger.kernel.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agodrm/amdgpu: Add an ATPX quirk for hybrid laptop
Alex Deucher [Thu, 22 Mar 2018 02:05:46 +0000 (21:05 -0500)]
drm/amdgpu: Add an ATPX quirk for hybrid laptop

commit 13b40935cf64f59b93cf1c716a2033488e5a228c upstream.

_PR3 doesn't seem to work properly, use ATPX instead.

Bug: https://bugs.freedesktop.org/show_bug.cgi?id=104064
Reviewed-by: Huang Rui <ray.huang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Cc: stable@vger.kernel.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agoALSA: pcm: Fix endless loop for XRUN recovery in OSS emulation
Takashi Iwai [Sat, 7 Apr 2018 09:48:58 +0000 (11:48 +0200)]
ALSA: pcm: Fix endless loop for XRUN recovery in OSS emulation

commit e15dc99dbb9cf99f6432e8e3c0b3a8f7a3403a86 upstream.

The commit 02a5d6925cd3 ("ALSA: pcm: Avoid potential races between OSS
ioctls and read/write") split the PCM preparation code to a locked
version, and it added a sanity check of runtime->oss.prepare flag
along with the change.  This leaded to an endless loop when the stream
gets XRUN: namely, snd_pcm_oss_write3() and co call
snd_pcm_oss_prepare() without setting runtime->oss.prepare flag and
the loop continues until the PCM state reaches to another one.

As the function is supposed to execute the preparation
unconditionally, drop the invalid state check there.

The bug was triggered by syzkaller.

Fixes: 02a5d6925cd3 ("ALSA: pcm: Avoid potential races between OSS ioctls and read/write")
Reported-by: syzbot+150189c103427d31a053@syzkaller.appspotmail.com
Reported-by: syzbot+7e3f31a52646f939c052@syzkaller.appspotmail.com
Reported-by: syzbot+4f2016cf5185da7759dc@syzkaller.appspotmail.com
Cc: <stable@vger.kernel.org>
Signed-off-by: Takashi Iwai <tiwai@suse.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agoALSA: pcm: Fix mutex unbalance in OSS emulation ioctls
Takashi Iwai [Tue, 27 Mar 2018 12:32:23 +0000 (14:32 +0200)]
ALSA: pcm: Fix mutex unbalance in OSS emulation ioctls

commit f6d297df4dd47ef949540e4a201230d0c5308325 upstream.

The previous fix 40cab6e88cb0 ("ALSA: pcm: Return -EBUSY for OSS
ioctls changing busy streams") introduced some mutex unbalance; the
check of runtime->oss.rw_ref was inserted in a wrong place after the
mutex lock.

This patch fixes the inconsistency by rewriting with the helper
functions to lock/unlock parameters with the stream check.

Fixes: 40cab6e88cb0 ("ALSA: pcm: Return -EBUSY for OSS ioctls changing busy streams")
Reported-by: Dan Carpenter <dan.carpenter@oracle.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Takashi Iwai <tiwai@suse.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agoALSA: pcm: Return -EBUSY for OSS ioctls changing busy streams
Takashi Iwai [Fri, 23 Mar 2018 07:03:26 +0000 (08:03 +0100)]
ALSA: pcm: Return -EBUSY for OSS ioctls changing busy streams

commit 40cab6e88cb0b6c56d3f30b7491a20e803f948f6 upstream.

OSS PCM stream management isn't modal but it allows ioctls issued at
any time for changing the parameters.  In the previous hardening
patch ("ALSA: pcm: Avoid potential races between OSS ioctls and
read/write"), we covered these races and prevent the corruption by
protecting the concurrent accesses via params_lock mutex.  However,
this means that some ioctls that try to change the stream parameter
(e.g. channels or format) would be blocked until the read/write
finishes, and it may take really long.

Basically changing the parameter while reading/writing is an invalid
operation, hence it's even more user-friendly from the API POV if it
returns -EBUSY in such a situation.

This patch adds such checks in the relevant ioctls with the addition
of read/write access refcount.

Cc: <stable@vger.kernel.org>
Signed-off-by: Takashi Iwai <tiwai@suse.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agoALSA: pcm: Avoid potential races between OSS ioctls and read/write
Takashi Iwai [Thu, 22 Mar 2018 17:10:14 +0000 (18:10 +0100)]
ALSA: pcm: Avoid potential races between OSS ioctls and read/write

commit 02a5d6925cd34c3b774bdb8eefb057c40a30e870 upstream.

Although we apply the params_lock mutex to the whole read and write
operations as well as snd_pcm_oss_change_params(), we may still face
some races.

First off, the params_lock is taken inside the read and write loop.
This is intentional for avoiding the too long locking, but it allows
the in-between parameter change, which might lead to invalid
pointers.  We check the readiness of the stream and set up via
snd_pcm_oss_make_ready() at the beginning of read and write, but it's
called only once, by assuming that it remains ready in the rest.

Second, many ioctls that may change the actual parameters
(i.e. setting runtime->oss.params=1) aren't protected, hence they can
be processed in a half-baked state.

This patch is an attempt to plug these holes.  The stream readiness
check is moved inside the read/write inner loop, so that the stream is
always set up in a proper state before further processing.  Also, each
ioctl that may change the parameter is wrapped with the params_lock
for avoiding the races.

The issues were triggered by syzkaller in a few different scenarios,
particularly the one below appearing as GPF in loopback_pos_update.

Reported-by: syzbot+c4227aec125487ec3efa@syzkaller.appspotmail.com
Cc: <stable@vger.kernel.org>
Signed-off-by: Takashi Iwai <tiwai@suse.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agoALSA: pcm: Use ERESTARTSYS instead of EINTR in OSS emulation
Takashi Iwai [Tue, 9 Jan 2018 07:51:02 +0000 (08:51 +0100)]
ALSA: pcm: Use ERESTARTSYS instead of EINTR in OSS emulation

commit c64ed5dd9feba193c76eb460b451225ac2a0d87b upstream.

Fix the last standing EINTR in the whole subsystem.  Use more correct
ERESTARTSYS for pending signals.

Signed-off-by: Takashi Iwai <tiwai@suse.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agovfio/pci: Virtualize Maximum Read Request Size
Alex Williamson [Mon, 2 Oct 2017 18:39:10 +0000 (12:39 -0600)]
vfio/pci: Virtualize Maximum Read Request Size

commit cf0d53ba4947aad6e471491d5b20a567cbe92e56 upstream.

MRRS defines the maximum read request size a device is allowed to
make.  Drivers will often increase this to allow more data transfer
with a single request.  Completions to this request are bound by the
MPS setting for the bus.  Aside from device quirks (none known), it
doesn't seem to make sense to set an MRRS value less than MPS, yet
this is a likely scenario given that user drivers do not have a
system-wide view of the PCI topology.  Virtualize MRRS such that the
user can set MRRS >= MPS, but use MPS as the floor value that we'll
write to hardware.

Signed-off-by: Alex Williamson <alex.williamson@redhat.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agowatchdog: f71808e_wdt: Fix WD_EN register read
Igor Pylypiv [Wed, 7 Mar 2018 07:47:25 +0000 (23:47 -0800)]
watchdog: f71808e_wdt: Fix WD_EN register read

commit 977f6f68331f94bb72ad84ee96b7b87ce737d89d upstream.

F71808FG_FLAG_WD_EN defines bit position, not a bitmask

Signed-off-by: Igor Pylypiv <igor.pylypiv@gmail.com>
Reviewed-by: Guenter Roeck <linux@roeck-us.net>
Signed-off-by: Guenter Roeck <linux@roeck-us.net>
Signed-off-by: Wim Van Sebroeck <wim@iguana.be>
Cc: stable <stable@vger.kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agodt-bindings: clock: mediatek: add binding for fixed-factor clock axisel_d4
Sean Wang [Thu, 1 Mar 2018 03:27:50 +0000 (11:27 +0800)]
dt-bindings: clock: mediatek: add binding for fixed-factor clock axisel_d4

commit 55a5fcafe3a94e8a0777bb993d09107d362258d2 upstream.

Just add binding for a fixed-factor clock axisel_d4, which would be
referenced by PWM devices on MT7623 or MT2701 SoC.

Cc: stable@vger.kernel.org
Fixes: 1de9b21633d6 ("clk: mediatek: Add dt-bindings for MT2701 clocks")
Signed-off-by: Sean Wang <sean.wang@mediatek.com>
Reviewed-by: Rob Herring <robh@kernel.org>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: devicetree@vger.kernel.org
Signed-off-by: Stephen Boyd <sboyd@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agothermal: imx: Fix race condition in imx_thermal_probe()
Mikhail Lappo [Fri, 2 Feb 2018 18:17:46 +0000 (16:17 -0200)]
thermal: imx: Fix race condition in imx_thermal_probe()

commit cf1ba1d73a33944d8c1a75370a35434bf146b8a7 upstream.

When device boots with T > T_trip_1 and requests interrupt,
the race condition takes place. The interrupt comes before
THERMAL_DEVICE_ENABLED is set. This leads to an attempt to
reading sensor value from irq and disabling the sensor, based on
the data->mode field, which expected to be THERMAL_DEVICE_ENABLED,
but still stays as THERMAL_DEVICE_DISABLED. Afher this issue
sensor is never re-enabled, as the driver state is wrong.

Fix this problem by setting the 'data' members prior to
requesting the interrupts.

Fixes: 37713a1e8e4c ("thermal: imx: implement thermal alarm interrupt handling")
Cc: <stable@vger.kernel.org>
Signed-off-by: Mikhail Lappo <mikhail.lappo@esrlabs.com>
Signed-off-by: Fabio Estevam <fabio.estevam@nxp.com>
Reviewed-by: Philipp Zabel <p.zabel@pengutronix.de>
Acked-by: Dong Aisheng <aisheng.dong@nxp.com>
Signed-off-by: Zhang Rui <rui.zhang@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agopwm: rcar: Fix a condition to prevent mismatch value setting to duty
Ryo Kodama [Fri, 9 Mar 2018 11:24:21 +0000 (20:24 +0900)]
pwm: rcar: Fix a condition to prevent mismatch value setting to duty

commit 6225f9c64b40bc8a22503e9cda70f55d7a9dd3c6 upstream.

This patch fixes an issue that is possible to set mismatch value to duty
for R-Car PWM if we input the following commands:

 # cd /sys/class/pwm/<pwmchip>/
 # echo 0 > export
 # cd pwm0
 # echo 30 > period
 # echo 30 > duty_cycle
 # echo 0 > duty_cycle
 # cat duty_cycle
 0
 # echo 1 > enable
 --> Then, the actual duty_cycle is 30, not 0.

So, this patch adds a condition into rcar_pwm_config() to fix this
issue.

Signed-off-by: Ryo Kodama <ryo.kodama.vz@renesas.com>
[shimoda: revise the commit log and add Fixes and Cc tags]
Fixes: ed6c1476bf7f ("pwm: Add support for R-Car PWM Timer")
Cc: Cc: <stable@vger.kernel.org> # v4.4+
Signed-off-by: Yoshihiro Shimoda <yoshihiro.shimoda.uh@renesas.com>
Signed-off-by: Thierry Reding <thierry.reding@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agoclk: bcm2835: De-assert/assert PLL reset signal when appropriate
Boris Brezillon [Thu, 22 Mar 2018 09:11:30 +0000 (10:11 +0100)]
clk: bcm2835: De-assert/assert PLL reset signal when appropriate

commit 753872373b599384ac7df809aa61ea12d1c4d5d1 upstream.

In order to enable a PLL, not only the PLL has to be powered up and
locked, but you also have to de-assert the reset signal. The last part
was missing. Add it so PLLs that were not enabled by the FW/bootloader
can be enabled from Linux.

Fixes: 41691b8862e2 ("clk: bcm2835: Add support for programming the audio domain clocks")
Cc: <stable@vger.kernel.org>
Signed-off-by: Boris Brezillon <boris.brezillon@bootlin.com>
Reviewed-by: Eric Anholt <eric@anholt.net>
Signed-off-by: Stephen Boyd <sboyd@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agoclk: mediatek: fix PWM clock source by adding a fixed-factor clock
Sean Wang [Thu, 1 Mar 2018 03:27:51 +0000 (11:27 +0800)]
clk: mediatek: fix PWM clock source by adding a fixed-factor clock

commit 89cd7aec21af26fd0c117bfc4bfc781724f201de upstream.

The clock for which all PWM devices on MT7623 or MT2701 actually depending
on has to be divided by four from its parent clock axi_sel in the clock
path prior to PWM devices.

Consequently, adding a fixed-factor clock axisel_d4 as one-fourth of
clock axi_sel allows that PWM devices can have the correct resolution
calculation.

Cc: stable@vger.kernel.org
Fixes: e9862118272a ("clk: mediatek: Add MT2701 clock support")
Signed-off-by: Sean Wang <sean.wang@mediatek.com>
Signed-off-by: Stephen Boyd <sboyd@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agoclk: fix false-positive Wmaybe-uninitialized warning
Arnd Bergmann [Fri, 16 Feb 2018 15:27:47 +0000 (16:27 +0100)]
clk: fix false-positive Wmaybe-uninitialized warning

commit ce33f284935e08229046b30635e6aadcbab02b53 upstream.

When we build this driver with on x86-32, gcc produces a false-positive warning:

drivers/clk/renesas/clk-sh73a0.c: In function 'sh73a0_cpg_clocks_init':
drivers/clk/renesas/clk-sh73a0.c:155:10: error: 'parent_name' may be used uninitialized in this function [-Werror=maybe-uninitialized]
   return clk_register_fixed_factor(NULL, name, parent_name, 0,

We can work around that warning by adding a fake initialization, I tried
and failed to come up with any better workaround. This is currently one
of few remaining warnings for a 4.14.y randconfig build, so it would be
good to also have it backported at least to that version. Older versions
have more randconfig warnings, so we might not care.

I had not noticed this earlier, because one patch in my randconfig test
tree removes the '-ffreestanding' option on x86-32, and that avoids
the warning. The -ffreestanding flag was originally global but moved
into arch/i386 by Andi Kleen in commit 6edfba1b33c7 ("[PATCH] x86_64:
Don't define string functions to builtin") as a 'temporary workaround'.

Like many temporary hacks, this turned out to be rather long-lived, from
all I can tell we still need a simple fix to asm/string_32.h before it
can be removed, but I'm not sure about how to best do that.

Cc: stable@vger.kernel.org
Cc: Andi Kleen <ak@linux.intel.com>
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Acked-by: Geert Uytterhoeven <geert+renesas@glider.be>
Signed-off-by: Stephen Boyd <sboyd@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agoclk: mvebu: armada-38x: add support for missing clocks
Richard Genoud [Tue, 13 Mar 2018 15:27:02 +0000 (16:27 +0100)]
clk: mvebu: armada-38x: add support for missing clocks

commit 6a4a4595804548e173f0763a0e7274a3521c59a9 upstream.

Clearfog boards can come with a CPU clocked at 1600MHz (commercial)
or 1333MHz (industrial).

They have also some dip-switches to select a different clock (666, 800,
1066, 1200).

The funny thing is that the recovery button is on the MPP34 fq selector.
So, when booting an industrial board with this button down, the frequency
666MHz is selected (and the kernel didn't boot).

This patch add all the missing clocks.

The only mode I didn't test is 2GHz (uboot found 4294MHz instead :/ ).

Fixes: 0e85aeced4d6 ("clk: mvebu: add clock support for Armada 380/385")
Cc: <stable@vger.kernel.org> # 3.16.x: 9593f4f56cf5: clk: mvebu: armada-38x: add support for 1866MHz variants
Cc: <stable@vger.kernel.org> # 3.16.x
Signed-off-by: Richard Genoud <richard.genoud@gmail.com>
Acked-by: Gregory CLEMENT <gregory.clement@bootlin.com>
Signed-off-by: Stephen Boyd <sboyd@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agoPCI: Mark Broadcom HT1100 and HT2000 Root Port Extended Tags as broken
Sinan Kaya [Tue, 10 Apr 2018 19:44:21 +0000 (14:44 -0500)]
PCI: Mark Broadcom HT1100 and HT2000 Root Port Extended Tags as broken

commit 1b30dfd376e28e7f37eda5e2033f6823cdda222b upstream.

Per PCIe r3.1, sec 2.2.6.2 and 7.8.4, a Requester may not use 8-bit Tags
unless its Extended Tag Field Enable is set, but all Receivers/Completers
must handle 8-bit Tags correctly regardless of their Extended Tag Field
Enable.

Some devices do not handle 8-bit Tags as Completers, so add a quirk for
them.  If we find such a device, we disable Extended Tags for the entire
hierarchy to make peer-to-peer DMA possible.

The Broadcom HT1100/HT2000/HT2100 seems to have issues with handling 8-bit
tags.  Mark it as broken.

This fixes Xorg hangs and unresponsive keyboards with errors like this:

  radeon 0000:06:00.0: GPU lockup (current fence id 0x000000000000000e last fence id 0x0000000000000
  [drm:r600_ring_test [radeon]] *ERROR* radeon: ring 0 test failed (scratch(0x8504)=0xCAFEDEAD)
  [drm:r600_resume [radeon]] *ERROR* r600 startup failed on resume

Fixes: 60db3a4d8cc9 ("PCI: Enable PCIe Extended Tags if supported")
Link: https://bugzilla.kernel.org/show_bug.cgi?id=196197
Signed-off-by: Sinan Kaya <okaya@codeaurora.org>
Signed-off-by: Bjorn Helgaas <helgaas@kernel.org>
CC: stable@vger.kernel.org # v4.11: 62ce94a7a5a5 PCI: Mark Broadcom HT2100 Root Port Extended Tags as broken
CC: stable@vger.kernel.org # v4.11
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agommc: tmio: Fix error handling when issuing CMD23
Masaharu Hayakawa [Tue, 3 Apr 2018 21:57:03 +0000 (23:57 +0200)]
mmc: tmio: Fix error handling when issuing CMD23

commit fc167daff581c01ebce8695e9618231cae3561a1 upstream.

If an error was detected when CMD23 was issued, command sequence should
be terminated with errors and CMD23 should be issued after retuning.

Fixes: 8b22c3c18be5 ("mmc: tmio: add CMD23 support")
Signed-off-by: Masaharu Hayakawa <masaharu.hayakawa.ry@renesas.com>
Signed-off-by: Wolfram Sang <wsa+renesas@sang-engineering.com>
Cc: <stable@vger.kernel.org> # 4.13+
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agommc: jz4740: Fix race condition in IRQ mask update
Alex Smith [Wed, 28 Mar 2018 21:00:43 +0000 (18:00 -0300)]
mmc: jz4740: Fix race condition in IRQ mask update

commit a04f0017c22453613d5f423326b190c61e3b4f98 upstream.

A spinlock is held while updating the internal copy of the IRQ mask,
but not while writing it to the actual IMASK register. After the lock
is released, an IRQ can occur before the IMASK register is written.
If handling this IRQ causes the mask to be changed, when the handler
returns back to the middle of the first mask update, a stale value
will be written to the mask register.

If this causes an IRQ to become unmasked that cannot have its status
cleared by writing a 1 to it in the IREG register, e.g. the SDIO IRQ,
then we can end up stuck with the same IRQ repeatedly being fired but
not handled. Normally the MMC IRQ handler attempts to clear any
unexpected IRQs by writing IREG, but for those that cannot be cleared
in this way then the IRQ will just repeatedly fire.

This was resulting in lockups after a while of using Wi-Fi on the
CI20 (GitHub issue #19).

Resolve by holding the spinlock until after the IMASK register has
been updated.

Cc: stable@vger.kernel.org
Link: https://github.com/MIPS/CI20_linux/issues/19
Fixes: 61bfbdb85687 ("MMC: Add support for the controller on JZ4740 SoCs.")
Tested-by: Mathieu Malaterre <malat@debian.org>
Signed-off-by: Alex Smith <alex.smith@imgtec.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agoiommu/vt-d: Fix a potential memory leak
Lu Baolu [Sat, 24 Feb 2018 05:42:27 +0000 (13:42 +0800)]
iommu/vt-d: Fix a potential memory leak

commit bbe4b3af9d9e3172fb9aa1f8dcdfaedcb381fc64 upstream.

A memory block was allocated in intel_svm_bind_mm() but never freed
in a failure path. This patch fixes this by free it to avoid memory
leakage.

Cc: Ashok Raj <ashok.raj@intel.com>
Cc: Jacob Pan <jacob.jun.pan@linux.intel.com>
Cc: <stable@vger.kernel.org> # v4.4+
Signed-off-by: Lu Baolu <baolu.lu@linux.intel.com>
Fixes: 2f26e0a9c9860 ('iommu/vt-d: Add basic SVM PASID support')
Signed-off-by: Joerg Roedel <jroedel@suse.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agoum: Use POSIX ucontext_t instead of struct ucontext
Krzysztof Mazur [Wed, 15 Nov 2017 10:12:39 +0000 (11:12 +0100)]
um: Use POSIX ucontext_t instead of struct ucontext

commit 4d1a535b8ec5e74b42dfd9dc809142653b2597f6 upstream.

glibc 2.26 removed the 'struct ucontext' to "improve" POSIX compliance
and break programs, including User Mode Linux. Fix User Mode Linux
by using POSIX ucontext_t.

This fixes:

arch/um/os-Linux/signal.c: In function 'hard_handler':
arch/um/os-Linux/signal.c:163:22: error: dereferencing pointer to incomplete type 'struct ucontext'
  mcontext_t *mc = &uc->uc_mcontext;
arch/x86/um/stub_segv.c: In function 'stub_segv_handler':
arch/x86/um/stub_segv.c:16:13: error: dereferencing pointer to incomplete type 'struct ucontext'
          &uc->uc_mcontext);

Cc: stable@vger.kernel.org
Signed-off-by: Krzysztof Mazur <krzysiek@podlesie.net>
Signed-off-by: Richard Weinberger <richard@nod.at>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agoum: Compile with modern headers
Jason A. Donenfeld [Thu, 14 Dec 2017 02:23:37 +0000 (03:23 +0100)]
um: Compile with modern headers

commit 530ba6c7cb3c22435a4d26de47037bb6f86a5329 upstream.

Recent libcs have gotten a bit more strict, so we actually need to
include the right headers and use the right types. This enables UML to
compile again.

Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
Cc: stable@vger.kernel.org
Signed-off-by: Richard Weinberger <richard@nod.at>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agoring-buffer: Check if memory is available before allocation
Steven Rostedt (VMware) [Mon, 2 Apr 2018 14:33:56 +0000 (10:33 -0400)]
ring-buffer: Check if memory is available before allocation

commit 2a872fa4e9c8adc79c830e4009e1cc0c013a9d8a upstream.

The ring buffer is made up of a link list of pages. When making the ring
buffer bigger, it will allocate all the pages it needs before adding to the
ring buffer, and if it fails, it frees them and returns an error. This makes
increasing the ring buffer size an all or nothing action. When this was
first created, the pages were allocated with "NORETRY". This was to not
cause any Out-Of-Memory (OOM) actions from allocating the ring buffer. But
NORETRY was too strict, as the ring buffer would fail to expand even when
there's memory available, but was taken up in the page cache.

Commit 848618857d253 ("tracing/ring_buffer: Try harder to allocate") changed
the allocating from NORETRY to RETRY_MAYFAIL. The RETRY_MAYFAIL would
allocate from the page cache, but if there was no memory available, it would
simple fail the allocation and not trigger an OOM.

This worked fine, but had one problem. As the ring buffer would allocate one
page at a time, it could take up all memory in the system before it failed
to allocate and free that memory. If the allocation is happening and the
ring buffer allocates all memory and then tries to take more than available,
its allocation will not trigger an OOM, but if there's any allocation that
happens someplace else, that could trigger an OOM, even though once the ring
buffer's allocation fails, it would free up all the previous memory it tried
to allocate, and allow other memory allocations to succeed.

Commit d02bd27bd33dd ("mm/page_alloc.c: calculate 'available' memory in a
separate function") separated out si_mem_availble() as a separate function
that could be used to see how much memory is available in the system. Using
this function to make sure that the ring buffer could be allocated before it
tries to allocate pages we can avoid allocating all memory in the system and
making it vulnerable to OOMs if other allocations are taking place.

Link: http://lkml.kernel.org/r/1522320104-6573-1-git-send-email-zhaoyang.huang@spreadtrum.com
CC: stable@vger.kernel.org
Cc: linux-mm@kvack.org
Fixes: 848618857d253 ("tracing/ring_buffer: Try harder to allocate")
Requires: d02bd27bd33dd ("mm/page_alloc.c: calculate 'available' memory in a separate function")
Reported-by: Zhaoyang Huang <huangzhaoyang@gmail.com>
Tested-by: Joel Fernandes <joelaf@google.com>
Signed-off-by: Steven Rostedt (VMware) <rostedt@goodmis.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agonfit: skip region registration for incomplete control regions
Dan Williams [Thu, 22 Mar 2018 04:22:34 +0000 (21:22 -0700)]
nfit: skip region registration for incomplete control regions

commit 0731de476a37c33485af82d64041c9d193208df8 upstream.

Per the ACPI specification the only functional purpose for a DIMM
Control Region to be mapped into the system physical address space, from
an OSPM perspective, is to support block-apertures. However, there are
some BIOSen that publish DIMM Control Region SPA entries for pre-boot
environment consumption.  Undo the kernel policy of generating disabled
'ndblk' regions when this configuration is detected.

Cc: <stable@vger.kernel.org>
Fixes: 1f7df6f88b92 ("libnvdimm, nfit: regions (block-data-window...)")
Reviewed-by: Toshi Kani <toshi.kani@hpe.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agonfit, address-range-scrub: fix scrub in-progress reporting
Dan Williams [Mon, 2 Apr 2018 23:40:04 +0000 (16:40 -0700)]
nfit, address-range-scrub: fix scrub in-progress reporting

commit 78727137fdf49edf9f731bde79d7189067b4047a upstream.

There is a small window whereby ARS scan requests can schedule work that
userspace will miss when polling scrub_show. Hold the init_mutex lock
over calls to report the status to close this potential escape. Also,
make sure that requests to cancel the ARS workqueue are treated as an
idle event.

Cc: <stable@vger.kernel.org>
Cc: Vishal Verma <vishal.l.verma@intel.com>
Fixes: 37b137ff8c83 ("nfit, libnvdimm: allow an ARS scrub...")
Reviewed-by: Dave Jiang <dave.jiang@intel.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agolibnvdimm, namespace: use a safe lookup for dimm device name
Dan Williams [Fri, 6 Apr 2018 23:37:21 +0000 (16:37 -0700)]
libnvdimm, namespace: use a safe lookup for dimm device name

commit 4f8672201b7e7ed4f5f6c3cf6dcd080648580582 upstream.

The following NULL dereference results from incorrectly assuming that
ndd is valid in this print:

  struct nvdimm_drvdata *ndd = to_ndd(&nd_region->mapping[i]);

  /*
   * Give up if we don't find an instance of a uuid at each
   * position (from 0 to nd_region->ndr_mappings - 1), or if we
   * find a dimm with two instances of the same uuid.
   */
  dev_err(&nd_region->dev, "%s missing label for %pUb\n",
                  dev_name(ndd->dev), nd_label->uuid);

 BUG: unable to handle kernel NULL pointer dereference at 0000000000000000
 IP: nd_region_register_namespaces+0xd67/0x13c0 [libnvdimm]
 PGD 0 P4D 0
 Oops: 0000 [#1] SMP PTI
 CPU: 43 PID: 673 Comm: kworker/u609:10 Not tainted 4.16.0-rc4+ #1
 [..]
 RIP: 0010:nd_region_register_namespaces+0xd67/0x13c0 [libnvdimm]
 [..]
 Call Trace:
  ? devres_add+0x2f/0x40
  ? devm_kmalloc+0x52/0x60
  ? nd_region_activate+0x9c/0x320 [libnvdimm]
  nd_region_probe+0x94/0x260 [libnvdimm]
  ? kernfs_add_one+0xe4/0x130
  nvdimm_bus_probe+0x63/0x100 [libnvdimm]

Switch to using the nvdimm device directly.

Fixes: 0e3b0d123c8f ("libnvdimm, namespace: allow multiple pmem...")
Cc: <stable@vger.kernel.org>
Reported-by: Dave Jiang <dave.jiang@intel.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agolibnvdimm, dimm: fix dpa reservation vs uninitialized label area
Dan Williams [Fri, 6 Apr 2018 18:25:38 +0000 (11:25 -0700)]
libnvdimm, dimm: fix dpa reservation vs uninitialized label area

commit c31898c8c711f2bbbcaebe802a55827e288d875a upstream.

At initialization time the 'dimm' driver caches a copy of the memory
device's label area and reserves address space for each of the
namespaces defined.

However, as can be seen below, the reservation occurs even when the
index blocks are invalid:

 nvdimm nmem0: nvdimm_init_config_data: len: 131072 rc: 0
 nvdimm nmem0: config data size: 131072
 nvdimm nmem0: __nd_label_validate: nsindex0 labelsize 1 invalid
 nvdimm nmem0: __nd_label_validate: nsindex1 labelsize 1 invalid
 nvdimm nmem0: : pmem-6025e505: 0x1000000000 @ 0xf50000000 reserve <-- bad

Gate dpa reservation on the presence of valid index blocks.

Cc: <stable@vger.kernel.org>
Fixes: 4a826c83db4e ("libnvdimm: namespace indices: read and validate")
Reported-by: Krzysztof Rusocki <krzysztof.rusocki@intel.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agotpm: self test failure should not cause suspend to fail
Chris Chiu [Tue, 20 Mar 2018 07:36:40 +0000 (15:36 +0800)]
tpm: self test failure should not cause suspend to fail

commit 0803d7befa15cab5717d667a97a66214d2a4c083 upstream.

The Acer Acer Veriton X4110G has a TPM device detected as:
  tpm_tis 00:0b: 1.2 TPM (device-id 0xFE, rev-id 71)

After the first S3 suspend, the following error appears during resume:
  tpm tpm0: A TPM error(38) occurred continue selftest

Any following S3 suspend attempts will now fail with this error:
  tpm tpm0: Error (38) sending savestate before suspend
  PM: Device 00:0b failed to suspend: error 38

Error 38 is TPM_ERR_INVALID_POSTINIT which means the TPM is
not in the correct state. This indicates that the platform BIOS
is not sending the usual TPM_Startup command during S3 resume.
>From this point onwards, all TPM commands will fail.

The same issue was previously reported on Foxconn 6150BK8MC and
Sony Vaio TX3.

The platform behaviour seems broken here, but we should not break
suspend/resume because of this.

When the unexpected TPM state is encountered, set a flag to skip the
affected TPM_SaveState command on later suspends.

Cc: stable@vger.kernel.org
Signed-off-by: Chris Chiu <chiu@endlessm.com>
Signed-off-by: Daniel Drake <drake@endlessm.com>
Link: http://lkml.kernel.org/r/CAB4CAwfSCvj1cudi+MWaB5g2Z67d9DwY1o475YOZD64ma23UiQ@mail.gmail.com
Link: https://lkml.org/lkml/2011/3/28/192
Link: https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=591031
Reviewed-by: Jarkko Sakkinen <jarkko.sakkinen@linux.intel.com>
Signed-off-by: Jarkko Sakkinen <jarkko.sakkinen@linux.intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agocxl: Fix possible deadlock when processing page faults from cxllib
Frederic Barrat [Tue, 3 Apr 2018 13:54:02 +0000 (15:54 +0200)]
cxl: Fix possible deadlock when processing page faults from cxllib

commit ad7b4e8022b9864c075fe71e1328b1d25cad82f6 upstream.

cxllib_handle_fault() is called by an external driver when it needs to
have the host resolve page faults for a buffer. The buffer can cover
several pages and VMAs. The function iterates over all the pages used
by the buffer, based on the page size of the VMA.

To ensure some stability while processing the faults, the thread T1
grabs the mm->mmap_sem semaphore with read access (R1). However, when
processing a page fault for a single page, one of the underlying
functions, copro_handle_mm_fault(), also grabs the same semaphore with
read access (R2). So the thread T1 takes the semaphore twice.

If another thread T2 tries to access the semaphore in write mode W1
(say, because it wants to allocate memory and calls 'brk'), then that
thread T2 will have to wait because there's a reader (R1). If the
thread T1 is processing a new page at that time, it won't get an
automatic grant at R2, because there's now a writer thread
waiting (T2). And we have a deadlock.

The timeline is:
1. thread T1 owns the semaphore with read access R1
2. thread T2 requests write access W1 and waits
3. thread T1 requests read access R2 and waits

The fix is for the thread T1 to release the semaphore R1 once it got
the information it needs from the current VMA. The address space/VMAs
could evolve while T1 iterates over the full buffer, but in the
unlikely case where T1 misses a page, the external driver will raise a
new page fault when retrying the memory access.

Fixes: 3ced8d730063 ("cxl: Export library to support IBM XSL")
Cc: stable@vger.kernel.org # 4.13+
Signed-off-by: Frederic Barrat <fbarrat@linux.vnet.ibm.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agodmaengine: at_xdmac: fix rare residue corruption
Maxime Jayat [Thu, 22 Feb 2018 11:39:55 +0000 (12:39 +0100)]
dmaengine: at_xdmac: fix rare residue corruption

commit c5637476bbf9bb86c7f0413b8f4822a73d8d2d07 upstream.

Despite the efforts made to correctly read the NDA and CUBC registers,
the order in which the registers are read could sometimes lead to an
inconsistent state.

Re-using the timeline from the comments, this following timing of
registers reads could lead to reading NDA with value "@desc2" and
CUBC with value "MAX desc1":

 INITD --------                    ------------
              |____________________|
       _______________________  _______________
 NDA       @desc2             \/   @desc3
       _______________________/\_______________
       __________  ___________  _______________
 CUBC       0    \/ MAX desc1 \/  MAX desc2
       __________/\___________/\_______________
        |  |          |  |
Events:(1)(2)        (3)(4)

(1) check_nda = @desc2
(2) initd = 1
(3) cur_ubc = MAX desc1
(4) cur_nda = @desc2

This is allowed by the condition ((check_nda == cur_nda) && initd),
despite cur_ubc and cur_nda being in the precise state we don't want.

This error leads to incorrect residue computation.

Fix it by inversing the order in which CUBC and INITD are read. This
makes sure that NDA and CUBC are always read together either _before_
INITD goes to 0 or _after_ it is back at 1.
The case where NDA is read before INITD is at 0 and CUBC is read after
INITD is back at 1 will be rejected by check_nda and cur_nda being
different.

Fixes: 53398f488821 ("dmaengine: at_xdmac: fix residue corruption")
Cc: stable@vger.kernel.org
Signed-off-by: Maxime Jayat <maxime.jayat@mobile-devices.fr>
Acked-by: Ludovic Desroches <ludovic.desroches@microchip.com>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agoIB/srp: Fix completion vector assignment algorithm
Bart Van Assche [Mon, 12 Feb 2018 17:50:25 +0000 (09:50 -0800)]
IB/srp: Fix completion vector assignment algorithm

commit 3a148896b24adf8688dc0c59af54531931677a40 upstream.

Ensure that cv_end is equal to ibdev->num_comp_vectors for the
NUMA node with the highest index. This patch improves spreading
of RDMA channels over completion vectors and thereby improves
performance, especially on systems with only a single NUMA node.
This patch drops support for the comp_vector login parameter by
ignoring the value of that parameter since I have not found a
good way to combine support for that parameter and automatic
spreading of RDMA channels over completion vectors.

Fixes: d92c0da71a35 ("IB/srp: Add multichannel support")
Reported-by: Alexander Schmid <alex@modula-shop-systems.de>
Signed-off-by: Bart Van Assche <bart.vanassche@wdc.com>
Cc: Alexander Schmid <alex@modula-shop-systems.de>
Cc: stable@vger.kernel.org
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agoIB/srp: Fix srp_abort()
Bart Van Assche [Fri, 23 Feb 2018 22:09:24 +0000 (14:09 -0800)]
IB/srp: Fix srp_abort()

commit e68088e78d82920632eba112b968e49d588d02a2 upstream.

Before commit e494f6a72839 ("[SCSI] improved eh timeout handler") it
did not really matter whether or not abort handlers like srp_abort()
called .scsi_done() when returning another value than SUCCESS. Since
that commit however this matters. Hence only call .scsi_done() when
returning SUCCESS.

Signed-off-by: Bart Van Assche <bart.vanassche@wdc.com>
Cc: stable@vger.kernel.org
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agoALSA: pcm: Fix UAF at PCM release via PCM timer access
Takashi Iwai [Mon, 2 Apr 2018 20:41:43 +0000 (22:41 +0200)]
ALSA: pcm: Fix UAF at PCM release via PCM timer access

commit a820ccbe21e8ce8e86c39cd1d3bc8c7d1cbb949b upstream.

The PCM runtime object is created and freed dynamically at PCM stream
open / close time.  This is tracked via substream->runtime, and it's
cleared at snd_pcm_detach_substream().

The runtime object assignment is protected by PCM open_mutex, so for
all PCM operations, it's safely handled.  However, each PCM substream
provides also an ALSA timer interface, and user-space can access to
this while closing a PCM substream.  This may eventually lead to a
UAF, as snd_pcm_timer_resolution() tries to access the runtime while
clearing it in other side.

Fortunately, it's the only concurrent access from the PCM timer, and
it merely reads runtime->timer_resolution field.  So, we can avoid the
race by reordering kfree() and wrapping the substream->runtime
clearance with the corresponding timer lock.

Reported-by: syzbot+8e62ff4e07aa2ce87826@syzkaller.appspotmail.com
Cc: <stable@vger.kernel.org>
Signed-off-by: Takashi Iwai <tiwai@suse.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agoRDMA/rxe: Fix an out-of-bounds read
Bart Van Assche [Thu, 1 Mar 2018 22:00:29 +0000 (14:00 -0800)]
RDMA/rxe: Fix an out-of-bounds read

commit a6544a624c3ff92a64e4aca3931fa064607bd3da upstream.

This patch avoids that KASAN reports the following when the SRP initiator
calls srp_post_send():

==================================================================
BUG: KASAN: stack-out-of-bounds in rxe_post_send+0x5c4/0x980 [rdma_rxe]
Read of size 8 at addr ffff880066606e30 by task 02-mq/1074

CPU: 2 PID: 1074 Comm: 02-mq Not tainted 4.16.0-rc3-dbg+ #1
Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS 1.0.0-prebuilt.qemu-project.org 04/01/2014
Call Trace:
dump_stack+0x85/0xc7
print_address_description+0x65/0x270
kasan_report+0x231/0x350
rxe_post_send+0x5c4/0x980 [rdma_rxe]
srp_post_send.isra.16+0x149/0x190 [ib_srp]
srp_queuecommand+0x94d/0x1670 [ib_srp]
scsi_dispatch_cmd+0x1c2/0x550 [scsi_mod]
scsi_queue_rq+0x843/0xa70 [scsi_mod]
blk_mq_dispatch_rq_list+0x143/0xac0
blk_mq_do_dispatch_ctx+0x1c5/0x260
blk_mq_sched_dispatch_requests+0x2bf/0x2f0
__blk_mq_run_hw_queue+0xdb/0x160
__blk_mq_delay_run_hw_queue+0xba/0x100
blk_mq_run_hw_queue+0xf2/0x190
blk_mq_sched_insert_request+0x163/0x2f0
blk_execute_rq+0xb0/0x130
scsi_execute+0x14e/0x260 [scsi_mod]
scsi_probe_and_add_lun+0x366/0x13d0 [scsi_mod]
__scsi_scan_target+0x18a/0x810 [scsi_mod]
scsi_scan_target+0x11e/0x130 [scsi_mod]
srp_create_target+0x1522/0x19e0 [ib_srp]
kernfs_fop_write+0x180/0x210
__vfs_write+0xb1/0x2e0
vfs_write+0xf6/0x250
SyS_write+0x99/0x110
do_syscall_64+0xee/0x2b0
entry_SYSCALL_64_after_hwframe+0x42/0xb7

The buggy address belongs to the page:
page:ffffea0001998180 count:0 mapcount:0 mapping:0000000000000000 index:0x0
flags: 0x4000000000000000()
raw: 4000000000000000 0000000000000000 0000000000000000 00000000ffffffff
raw: dead000000000100 dead000000000200 0000000000000000 0000000000000000
page dumped because: kasan: bad access detected

Memory state around the buggy address:
ffff880066606d00: 00 00 00 00 00 00 00 00 00 00 00 00 00 f1 f1 f1
ffff880066606d80: f1 00 f2 f2 f2 f2 f2 f2 f2 00 00 f2 f2 f2 f2 f2
>ffff880066606e00: f2 00 00 00 00 00 f2 f2 f2 f3 f3 f3 f3 00 00 00
                                    ^
ffff880066606e80: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
ffff880066606f00: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
==================================================================

Fixes: 8700e3e7c485 ("Soft RoCE driver")
Signed-off-by: Bart Van Assche <bart.vanassche@wdc.com>
Cc: Moni Shoua <monis@mellanox.com>
Cc: stable@vger.kernel.org
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agoRDMA/mlx5: Protect from NULL pointer derefence
Leon Romanovsky [Tue, 13 Mar 2018 13:29:24 +0000 (15:29 +0200)]
RDMA/mlx5: Protect from NULL pointer derefence

commit 4289861d88d6c7b5e4c8cc7fe2ad6cdf0cdfc366 upstream.

The mlx5_ib_alloc_implicit_mr() can fail to acquire pages
and the returned mr pointer won't be valid. Ensure that it
is not error prior to access.

Cc: <stable@vger.kernel.org> # 4.10
Fixes: 81713d3788d2 ("IB/mlx5: Add implicit MR support")
Reported-by: Noa Osherovich <noaos@mellanox.com>
Signed-off-by: Leon Romanovsky <leonro@mellanox.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agoRDMA/ucma: Don't allow setting RDMA_OPTION_IB_PATH without an RDMA device
Roland Dreier [Tue, 3 Apr 2018 22:33:01 +0000 (15:33 -0700)]
RDMA/ucma: Don't allow setting RDMA_OPTION_IB_PATH without an RDMA device

commit 8435168d50e66fa5eae01852769d20a36f9e5e83 upstream.

Check to make sure that ctx->cm_id->device is set before we use it.
Otherwise userspace can trigger a NULL dereference by doing
RDMA_USER_CM_CMD_SET_OPTION on an ID that is not bound to a device.

Cc: <stable@vger.kernel.org>
Reported-by: <syzbot+a67bc93e14682d92fc2f@syzkaller.appspotmail.com>
Signed-off-by: Roland Dreier <roland@purestorage.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agodm crypt: limit the number of allocated pages
Mikulas Patocka [Mon, 14 Aug 2017 02:45:08 +0000 (22:45 -0400)]
dm crypt: limit the number of allocated pages

commit 5059353df86e2573ccd9d43fd9d9396dcec47ca2 upstream.

dm-crypt consumes an excessive amount memory when the user attempts to
zero a dm-crypt device with "blkdiscard -z". The command "blkdiscard -z"
calls the BLKZEROOUT ioctl, it goes to the function __blkdev_issue_zeroout,
__blkdev_issue_zeroout sends a large amount of write bios that contain
the zero page as their payload.

For each incoming page, dm-crypt allocates another page that holds the
encrypted data, so when processing "blkdiscard -z", dm-crypt tries to
allocate the amount of memory that is equal to the size of the device.
This can trigger OOM killer or cause system crash.

Fix this by limiting the amount of memory that dm-crypt allocates to 2%
of total system memory. This limit is system-wide and is divided by the
number of active dm-crypt devices and each device receives an equal
share.

Cc: stable@vger.kernel.org
Signed-off-by: Mikulas Patocka <mpatocka@redhat.com>
Signed-off-by: Mike Snitzer <snitzer@redhat.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agoext4: add extra checks to ext4_xattr_block_get()
Theodore Ts'o [Sat, 31 Mar 2018 00:04:11 +0000 (20:04 -0400)]
ext4: add extra checks to ext4_xattr_block_get()

commit 54dd0e0a1b255f115f8647fc6fb93273251b01b9 upstream.

Add explicit checks in ext4_xattr_block_get() just in case the
e_value_offs and e_value_size fields in the the xattr block are
corrupted in memory after the buffer_verified bit is set on the xattr
block.

Signed-off-by: Theodore Ts'o <tytso@mit.edu>
Cc: stable@kernel.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agoext4: add bounds checking to ext4_xattr_find_entry()
Theodore Ts'o [Sat, 31 Mar 2018 00:00:56 +0000 (20:00 -0400)]
ext4: add bounds checking to ext4_xattr_find_entry()

commit 9496005d6ca4cf8f5ee8f828165a8956872dc59d upstream.

Add some paranoia checks to make sure we don't stray beyond the end of
the valid memory region containing ext4 xattr entries while we are
scanning for a match.

Also rename the function to xattr_find_entry() since it is static and
thus only used in fs/ext4/xattr.c

Signed-off-by: Theodore Ts'o <tytso@mit.edu>
Cc: stable@kernel.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agoext4: move call to ext4_error() into ext4_xattr_check_block()
Theodore Ts'o [Fri, 30 Mar 2018 19:42:25 +0000 (15:42 -0400)]
ext4: move call to ext4_error() into ext4_xattr_check_block()

commit de05ca8526796c7e9f7c7282b7f89a818af19818 upstream.

Refactor the call to EXT4_ERROR_INODE() into ext4_xattr_check_block().
This simplifies the code, and fixes a problem where not all callers of
ext4_xattr_check_block() were not resulting in ext4_error() getting
called when the xattr block is corrupted.

Signed-off-by: Theodore Ts'o <tytso@mit.edu>
Cc: stable@vger.kernel.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agoext4: don't allow r/w mounts if metadata blocks overlap the superblock
Theodore Ts'o [Fri, 30 Mar 2018 02:10:35 +0000 (22:10 -0400)]
ext4: don't allow r/w mounts if metadata blocks overlap the superblock

commit 18db4b4e6fc31eda838dd1c1296d67dbcb3dc957 upstream.

If some metadata block, such as an allocation bitmap, overlaps the
superblock, it's very likely that if the file system is mounted
read/write, the results will not be pretty.  So disallow r/w mounts
for file systems corrupted in this particular way.

Signed-off-by: Theodore Ts'o <tytso@mit.edu>
Cc: stable@vger.kernel.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agoext4: always initialize the crc32c checksum driver
Theodore Ts'o [Fri, 30 Mar 2018 02:10:31 +0000 (22:10 -0400)]
ext4: always initialize the crc32c checksum driver

commit a45403b51582a87872927a3e0fc0a389c26867f1 upstream.

The extended attribute code now uses the crc32c checksum for hashing
purposes, so we should just always always initialize it.  We also want
to prevent NULL pointer dereferences if one of the metadata checksum
features is enabled after the file sytsem is originally mounted.

This issue has been assigned CVE-2018-1094.

https://bugzilla.kernel.org/show_bug.cgi?id=199183
https://bugzilla.redhat.com/show_bug.cgi?id=1560788

Signed-off-by: Theodore Ts'o <tytso@mit.edu>
Cc: stable@vger.kernel.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agoext4: fail ext4_iget for root directory if unallocated
Theodore Ts'o [Fri, 30 Mar 2018 01:56:09 +0000 (21:56 -0400)]
ext4: fail ext4_iget for root directory if unallocated

commit 8e4b5eae5decd9dfe5a4ee369c22028f90ab4c44 upstream.

If the root directory has an i_links_count of zero, then when the file
system is mounted, then when ext4_fill_super() notices the problem and
tries to call iput() the root directory in the error return path,
ext4_evict_inode() will try to free the inode on disk, before all of
the file system structures are set up, and this will result in an OOPS
caused by a NULL pointer dereference.

This issue has been assigned CVE-2018-1092.

https://bugzilla.kernel.org/show_bug.cgi?id=199179
https://bugzilla.redhat.com/show_bug.cgi?id=1560777

Reported-by: Wen Xu <wen.xu@gatech.edu>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
Cc: stable@vger.kernel.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agoext4: limit xattr size to INT_MAX
Eric Biggers [Thu, 29 Mar 2018 18:31:42 +0000 (14:31 -0400)]
ext4: limit xattr size to INT_MAX

commit ce3fd194fcc6fbdc00ce095a852f22df97baa401 upstream.

ext4 isn't validating the sizes of xattrs where the value of the xattr
is stored in an external inode.  This is problematic because
->e_value_size is a u32, but ext4_xattr_get() returns an int.  A very
large size is misinterpreted as an error code, which ext4_get_acl()
translates into a bogus ERR_PTR() for which IS_ERR() returns false,
causing a crash.

Fix this by validating that all xattrs are <= INT_MAX bytes.

This issue has been assigned CVE-2018-1095.

https://bugzilla.kernel.org/show_bug.cgi?id=199185
https://bugzilla.redhat.com/show_bug.cgi?id=1560793

Reported-by: Wen Xu <wen.xu@gatech.edu>
Signed-off-by: Eric Biggers <ebiggers@google.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
Cc: stable@vger.kernel.org
Fixes: e50e5129f384 ("ext4: xattr-in-inode support")
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agoext4: protect i_disksize update by i_data_sem in direct write path
Eryu Guan [Thu, 22 Mar 2018 15:41:25 +0000 (11:41 -0400)]
ext4: protect i_disksize update by i_data_sem in direct write path

commit 73fdad00b208b139cf43f3163fbc0f67e4c6047c upstream.

i_disksize update should be protected by i_data_sem, by either taking
the lock explicitly or by using ext4_update_i_disksize() helper. But the
i_disksize updates in ext4_direct_IO_write() are not protected at all,
which may be racing with i_disksize updates in writeback path in
delalloc buffer write path.

This is found by code inspection, and I didn't hit any i_disksize
corruption due to this bug. Thanks to Jan Kara for catching this bug and
suggesting the fix!

Reported-by: Jan Kara <jack@suse.cz>
Suggested-by: Jan Kara <jack@suse.cz>
Signed-off-by: Eryu Guan <guaneryu@gmail.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
Cc: stable@vger.kernel.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agoext4: don't update checksum of new initialized bitmaps
Theodore Ts'o [Mon, 19 Feb 2018 19:16:47 +0000 (14:16 -0500)]
ext4: don't update checksum of new initialized bitmaps

commit 044e6e3d74a3d7103a0c8a9305dfd94d64000660 upstream.

When reading the inode or block allocation bitmap, if the bitmap needs
to be initialized, do not update the checksum in the block group
descriptor.  That's because we're not set up to journal those changes.
Instead, just set the verified bit on the bitmap block, so that it's
not necessary to validate the checksum.

When a block or inode allocation actually happens, at that point the
checksum will be calculated, and update of the bg descriptor block
will be properly journalled.

Signed-off-by: Theodore Ts'o <tytso@mit.edu>
Cc: stable@vger.kernel.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agoext4: pass -ESHUTDOWN code to jbd2 layer
Theodore Ts'o [Mon, 19 Feb 2018 04:45:18 +0000 (23:45 -0500)]
ext4: pass -ESHUTDOWN code to jbd2 layer

commit fb7c02445c497943e7296cd3deee04422b63acb8 upstream.

Previously the jbd2 layer assumed that a file system check would be
required after a journal abort.  In the case of the deliberate file
system shutdown, this should not be necessary.  Allow the jbd2 layer
to distinguish between these two cases by using the ESHUTDOWN errno.

Also add proper locking to __journal_abort_soft().

Signed-off-by: Theodore Ts'o <tytso@mit.edu>
Cc: stable@vger.kernel.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agoext4: eliminate sleep from shutdown ioctl
Theodore Ts'o [Mon, 19 Feb 2018 04:16:28 +0000 (23:16 -0500)]
ext4: eliminate sleep from shutdown ioctl

commit a6d9946bb925293fda9f5ed6d33d8580b001f006 upstream.

The msleep() when processing EXT4_GOING_FLAGS_NOLOGFLUSH was a hack to
avoid some races (that are now fixed), but in fact it introduced its
own race.

Signed-off-by: Theodore Ts'o <tytso@mit.edu>
Cc: stable@vger.kernel.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agoext4: shutdown should not prevent get_write_access
Theodore Ts'o [Mon, 19 Feb 2018 03:07:36 +0000 (22:07 -0500)]
ext4: shutdown should not prevent get_write_access

commit 576d18ed60f5465110087c5e0eb1010de13e374d upstream.

The ext4 forced shutdown flag needs to prevent new handles from being
started, but it needs to allow existing handles to complete.  So the
forced shutdown flag should not force ext4_journal_get_write_access to
fail.

Signed-off-by: Theodore Ts'o <tytso@mit.edu>
Cc: stable@vger.kernel.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agojbd2: if the journal is aborted then don't allow update of the log tail
Theodore Ts'o [Mon, 19 Feb 2018 17:22:53 +0000 (12:22 -0500)]
jbd2: if the journal is aborted then don't allow update of the log tail

commit 85e0c4e89c1b864e763c4e3bb15d0b6d501ad5d9 upstream.

This updates the jbd2 superblock unnecessarily, and on an abort we
shouldn't truncate the log.

Signed-off-by: Theodore Ts'o <tytso@mit.edu>
Cc: stable@vger.kernel.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agoblock: use 32-bit blk_status_t on Alpha
Mikulas Patocka [Wed, 21 Mar 2018 16:42:25 +0000 (12:42 -0400)]
block: use 32-bit blk_status_t on Alpha

commit 6e2fb22103b99c26ae30a46512abe75526d8e4c9 upstream.

Early alpha processors cannot write a single byte or word; they read 8
bytes, modify the value in registers and write back 8 bytes.

The type blk_status_t is defined as one byte, it is often written
asynchronously by I/O completion routines, this asynchronous modification
can corrupt content of nearby bytes if these nearby bytes can be written
simultaneously by another CPU.

- one example of such corruption is the structure dm_io where
  "blk_status_t status" is written by an asynchronous completion routine
  and "atomic_t io_count" is modified synchronously
- another example is the structure dm_buffer where "unsigned hold_count"
  is modified synchronously from process context and "blk_status_t
  write_error" is modified asynchronously from bio completion routine

This patch fixes the bug by changing the type blk_status_t to 32 bits if
we are on Alpha and if we are compiling for a processor that doesn't have
the byte-word-extension.

Signed-off-by: Mikulas Patocka <mpatocka@redhat.com>
Cc: stable@vger.kernel.org # 4.13+
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agoextcon: intel-cht-wc: Set direction and drv flags for V5 boost GPIO
Hans de Goede [Mon, 19 Feb 2018 13:20:46 +0000 (14:20 +0100)]
extcon: intel-cht-wc: Set direction and drv flags for V5 boost GPIO

commit ad49aee401dd1997ec71360df6e51a91ad3cf516 upstream.

Sometimes (firmware bug?) the V5 boost GPIO is not configured as output
by the BIOS, leading to the 5V boost convertor being permanently on,

Explicitly set the direction and drv flags rather then inheriting them
from the firmware to fix this.

Fixes: 585cb239f4de ("extcon: intel-cht-wc: Disable external 5v boost ...")
Cc: stable@vger.kernel.org
Reviewed-by: Andy Shevchenko <andy.shevchenko@gmail.com>
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Chanwoo Choi <cw00.choi@samsung.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agorandom: use a tighter cap in credit_entropy_bits_safe()
Theodore Ts'o [Sat, 25 Feb 2017 22:21:33 +0000 (18:21 -0400)]
random: use a tighter cap in credit_entropy_bits_safe()

commit 9f886f4d1d292442b2f22a0a33321eae821bde40 upstream.

This fixes a harmless UBSAN where root could potentially end up
causing an overflow while bumping the entropy_total field (which is
ignored once the entropy pool has been initialized, and this generally
is completed during the boot sequence).

This is marginal for the stable kernel series, but it's a really
trivial patch, and it fixes UBSAN warning that might cause security
folks to get overly excited for no reason.

Signed-off-by: Theodore Ts'o <tytso@mit.edu>
Reported-by: Chen Feng <puck.chen@hisilicon.com>
Cc: stable@vger.kernel.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>