openwrt/staging/blogic.git
5 years agomm, memcg: proportional memory.{low,min} reclaim
Chris Down [Mon, 7 Oct 2019 00:58:32 +0000 (17:58 -0700)]
mm, memcg: proportional memory.{low,min} reclaim

cgroup v2 introduces two memory protection thresholds: memory.low
(best-effort) and memory.min (hard protection).  While they generally do
what they say on the tin, there is a limitation in their implementation
that makes them difficult to use effectively: that cliff behaviour often
manifests when they become eligible for reclaim.  This patch implements
more intuitive and usable behaviour, where we gradually mount more
reclaim pressure as cgroups further and further exceed their protection
thresholds.

This cliff edge behaviour happens because we only choose whether or not
to reclaim based on whether the memcg is within its protection limits
(see the use of mem_cgroup_protected in shrink_node), but we don't vary
our reclaim behaviour based on this information.  Imagine the following
timeline, with the numbers the lruvec size in this zone:

1. memory.low=1000000, memory.current=999999. 0 pages may be scanned.
2. memory.low=1000000, memory.current=1000000. 0 pages may be scanned.
3. memory.low=1000000, memory.current=10000011000001* pages may be
   scanned. (?!)

* Of course, we won't usually scan all available pages in the zone even
  without this patch because of scan control priority, over-reclaim
  protection, etc.  However, as shown by the tests at the end, these
  techniques don't sufficiently throttle such an extreme change in input,
  so cliff-like behaviour isn't really averted by their existence alone.

Here's an example of how this plays out in practice.  At Facebook, we are
trying to protect various workloads from "system" software, like
configuration management tools, metric collectors, etc (see this[0] case
study).  In order to find a suitable memory.low value, we start by
determining the expected memory range within which the workload will be
comfortable operating.  This isn't an exact science -- memory usage deemed
"comfortable" will vary over time due to user behaviour, differences in
composition of work, etc, etc.  As such we need to ballpark memory.low,
but doing this is currently problematic:

1. If we end up setting it too low for the workload, it won't have
   *any* effect (see discussion above).  The group will receive the full
   weight of reclaim and won't have any priority while competing with the
   less important system software, as if we had no memory.low configured
   at all.

2. Because of this behaviour, we end up erring on the side of setting
   it too high, such that the comfort range is reliably covered.  However,
   protected memory is completely unavailable to the rest of the system,
   so we might cause undue memory and IO pressure there when we *know* we
   have some elasticity in the workload.

3. Even if we get the value totally right, smack in the middle of the
   comfort zone, we get extreme jumps between no pressure and full
   pressure that cause unpredictable pressure spikes in the workload due
   to the current binary reclaim behaviour.

With this patch, we can set it to our ballpark estimation without too much
worry.  Any undesirable behaviour, such as too much or too little reclaim
pressure on the workload or system will be proportional to how far our
estimation is off.  This means we can set memory.low much more
conservatively and thus waste less resources *without* the risk of the
workload falling off a cliff if we overshoot.

As a more abstract technical description, this unintuitive behaviour
results in having to give high-priority workloads a large protection
buffer on top of their expected usage to function reliably, as otherwise
we have abrupt periods of dramatically increased memory pressure which
hamper performance.  Having to set these thresholds so high wastes
resources and generally works against the principle of work conservation.
In addition, having proportional memory reclaim behaviour has other
benefits.  Most notably, before this patch it's basically mandatory to set
memory.low to a higher than desirable value because otherwise as soon as
you exceed memory.low, all protection is lost, and all pages are eligible
to scan again.  By contrast, having a gradual ramp in reclaim pressure
means that you now still get some protection when thresholds are exceeded,
which means that one can now be more comfortable setting memory.low to
lower values without worrying that all protection will be lost.  This is
important because workingset size is really hard to know exactly,
especially with variable workloads, so at least getting *some* protection
if your workingset size grows larger than you expect increases user
confidence in setting memory.low without a huge buffer on top being
needed.

Thanks a lot to Johannes Weiner and Tejun Heo for their advice and
assistance in thinking about how to make this work better.

In testing these changes, I intended to verify that:

1. Changes in page scanning become gradual and proportional instead of
   binary.

   To test this, I experimented stepping further and further down
   memory.low protection on a workload that floats around 19G workingset
   when under memory.low protection, watching page scan rates for the
   workload cgroup:

   +------------+-----------------+--------------------+--------------+
   | memory.low | test (pgscan/s) | control (pgscan/s) | % of control |
   +------------+-----------------+--------------------+--------------+
   |        21G |               0 |                  0 | N/A          |
   |        17G |             867 |               3799 | 23%          |
   |        12G |            1203 |               3543 | 34%          |
   |         8G |            2534 |               3979 | 64%          |
   |         4G |            3980 |               4147 | 96%          |
   |          0 |            3799 |               3980 | 95%          |
   +------------+-----------------+--------------------+--------------+

   As you can see, the test kernel (with a kernel containing this
   patch) ramps up page scanning significantly more gradually than the
   control kernel (without this patch).

2. More gradual ramp up in reclaim aggression doesn't result in
   premature OOMs.

   To test this, I wrote a script that slowly increments the number of
   pages held by stress(1)'s --vm-keep mode until a production system
   entered severe overall memory contention.  This script runs in a highly
   protected slice taking up the majority of available system memory.
   Watching vmstat revealed that page scanning continued essentially
   nominally between test and control, without causing forward reclaim
   progress to become arrested.

[0]: https://facebookmicrosites.github.io/cgroup2/docs/overview.html#case-study-the-fbtax2-project

[akpm@linux-foundation.org: reflow block comments to fit in 80 cols]
[chris@chrisdown.name: handle cgroup_disable=memory when getting memcg protection]
Link: http://lkml.kernel.org/r/20190201045711.GA18302@chrisdown.name
Link: http://lkml.kernel.org/r/20190124014455.GA6396@chrisdown.name
Signed-off-by: Chris Down <chris@chrisdown.name>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Reviewed-by: Roman Gushchin <guro@fb.com>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Tejun Heo <tj@kernel.org>
Cc: Dennis Zhou <dennis@kernel.org>
Cc: Tetsuo Handa <penguin-kernel@i-love.sakura.ne.jp>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
5 years agomm/vmpressure.c: fix a signedness bug in vmpressure_register_event()
Dan Carpenter [Mon, 7 Oct 2019 00:58:28 +0000 (17:58 -0700)]
mm/vmpressure.c: fix a signedness bug in vmpressure_register_event()

The "mode" and "level" variables are enums and in this context GCC will
treat them as unsigned ints so the error handling is never triggered.

I also removed the bogus initializer because it isn't required any more
and it's sort of confusing.

[akpm@linux-foundation.org: reduce implicit and explicit typecasting]
[akpm@linux-foundation.org: fix return value, add comment, per Matthew]
Link: http://lkml.kernel.org/r/20190925110449.GO3264@mwanda
Fixes: 3cadfa2b9497 ("mm/vmpressure.c: convert to use match_string() helper")
Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Reviewed-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Acked-by: David Rientjes <rientjes@google.com>
Reviewed-by: Matthew Wilcox <willy@infradead.org>
Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Enrico Weigelt <info@metux.net>
Cc: Kate Stewart <kstewart@linuxfoundation.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
5 years agomm/page_alloc.c: fix a crash in free_pages_prepare()
Qian Cai [Mon, 7 Oct 2019 00:58:25 +0000 (17:58 -0700)]
mm/page_alloc.c: fix a crash in free_pages_prepare()

On architectures like s390, arch_free_page() could mark the page unused
(set_page_unused()) and any access later would trigger a kernel panic.
Fix it by moving arch_free_page() after all possible accessing calls.

 Hardware name: IBM 2964 N96 400 (z/VM 6.4.0)
 Krnl PSW : 0404e00180000000 0000000026c2b96e (__free_pages_ok+0x34e/0x5d8)
            R:0 T:1 IO:0 EX:0 Key:0 M:1 W:0 P:0 AS:3 CC:2 PM:0 RI:0 EA:3
 Krnl GPRS: 0000000088d43af7 0000000000484000 000000000000007c 000000000000000f
            000003d080012100 000003d080013fc0 0000000000000000 0000000000100000
            00000000275cca48 0000000000000100 0000000000000008 000003d080010000
            00000000000001d0 000003d000000000 0000000026c2b78a 000000002717fdb0
 Krnl Code: 0000000026c2b95cec1100b30659 risbgn %r1,%r1,0,179,6
            0000000026c2b962e32014000036 pfd 2,1024(%r1)
           #0000000026c2b968d7ff10001000 xc 0(256,%r1),0(%r1)
           >0000000026c2b96e41101100  la %r1,256(%r1)
            0000000026c2b972a737fff8  brctg %r3,26c2b962
            0000000026c2b976d7ff10001000 xc 0(256,%r1),0(%r1)
            0000000026c2b97ce31003400004 lg %r1,832
            0000000026c2b982ebff1430016a asi 5168(%r1),-1
 Call Trace:
 __free_pages_ok+0x16a/0x5d8)
 memblock_free_all+0x206/0x290
 mem_init+0x58/0x120
 start_kernel+0x2b0/0x570
 startup_continue+0x6a/0xc0
 INFO: lockdep is turned off.
 Last Breaking-Event-Address:
 __free_pages_ok+0x372/0x5d8
 Kernel panic - not syncing: Fatal exception: panic_on_oops
 00: HCPGIR450W CP entered; disabled wait PSW 00020001 80000000 00000000 26A2379C

In the past, only kernel_poison_pages() would trigger this but it needs
"page_poison=on" kernel cmdline, and I suspect nobody tested that on
s390.  Recently, kernel_init_free_pages() (commit 6471384af2a6 ("mm:
security: introduce init_on_alloc=1 and init_on_free=1 boot options"))
was added and could trigger this as well.

[akpm@linux-foundation.org: add comment]
Link: http://lkml.kernel.org/r/1569613623-16820-1-git-send-email-cai@lca.pw
Fixes: 8823b1dbc05f ("mm/page_poison.c: enable PAGE_POISONING as a separate option")
Fixes: 6471384af2a6 ("mm: security: introduce init_on_alloc=1 and init_on_free=1 boot options")
Signed-off-by: Qian Cai <cai@lca.pw>
Reviewed-by: Heiko Carstens <heiko.carstens@de.ibm.com>
Acked-by: Christian Borntraeger <borntraeger@de.ibm.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Cc: "Kirill A. Shutemov" <kirill@shutemov.name>
Cc: Vasily Gorbik <gor@linux.ibm.com>
Cc: Alexander Duyck <alexander.duyck@gmail.com>
Cc: <stable@vger.kernel.org> [5.3+]
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
5 years agomm/z3fold.c: claim page in the beginning of free
Vitaly Wool [Mon, 7 Oct 2019 00:58:22 +0000 (17:58 -0700)]
mm/z3fold.c: claim page in the beginning of free

There's a really hard to reproduce race in z3fold between z3fold_free()
and z3fold_reclaim_page().  z3fold_reclaim_page() can claim the page
after z3fold_free() has checked if the page was claimed and
z3fold_free() will then schedule this page for compaction which may in
turn lead to random page faults (since that page would have been
reclaimed by then).

Fix that by claiming page in the beginning of z3fold_free() and not
forgetting to clear the claim in the end.

[vitalywool@gmail.com: v2]
Link: http://lkml.kernel.org/r/20190928113456.152742cf@bigdell
Link: http://lkml.kernel.org/r/20190926104844.4f0c6efa1366b8f5741eaba9@gmail.com
Signed-off-by: Vitaly Wool <vitalywool@gmail.com>
Reported-by: Markus Linnala <markus.linnala@gmail.com>
Cc: Dan Streetman <ddstreet@ieee.org>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Henry Burns <henrywolfeburns@gmail.com>
Cc: Shakeel Butt <shakeelb@google.com>
Cc: Markus Linnala <markus.linnala@gmail.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
5 years agokernel/sysctl.c: do not override max_threads provided by userspace
Michal Hocko [Mon, 7 Oct 2019 00:58:19 +0000 (17:58 -0700)]
kernel/sysctl.c: do not override max_threads provided by userspace

Partially revert 16db3d3f1170 ("kernel/sysctl.c: threads-max observe
limits") because the patch is causing a regression to any workload which
needs to override the auto-tuning of the limit provided by kernel.

set_max_threads is implementing a boot time guesstimate to provide a
sensible limit of the concurrently running threads so that runaways will
not deplete all the memory.  This is a good thing in general but there
are workloads which might need to increase this limit for an application
to run (reportedly WebSpher MQ is affected) and that is simply not
possible after the mentioned change.  It is also very dubious to
override an admin decision by an estimation that doesn't have any direct
relation to correctness of the kernel operation.

Fix this by dropping set_max_threads from sysctl_max_threads so any
value is accepted as long as it fits into MAX_THREADS which is important
to check because allowing more threads could break internal robust futex
restriction.  While at it, do not use MIN_THREADS as the lower boundary
because it is also only a heuristic for automatic estimation and admin
might have a good reason to stop new threads to be created even when
below this limit.

This became more severe when we switched x86 from 4k to 8k kernel
stacks.  Starting since 6538b8ea886e ("x86_64: expand kernel stack to
16K") (3.16) we use THREAD_SIZE_ORDER = 2 and that halved the auto-tuned
value.

In the particular case

  3.12
  kernel.threads-max = 515561

  4.4
  kernel.threads-max = 200000

Neither of the two values is really insane on 32GB machine.

I am not sure we want/need to tune the max_thread value further.  If
anything the tuning should be removed altogether if proven not useful in
general.  But we definitely need a way to override this auto-tuning.

Link: http://lkml.kernel.org/r/20190922065801.GB18814@dhcp22.suse.cz
Fixes: 16db3d3f1170 ("kernel/sysctl.c: threads-max observe limits")
Signed-off-by: Michal Hocko <mhocko@suse.com>
Reviewed-by: "Eric W. Biederman" <ebiederm@xmission.com>
Cc: Heinrich Schuchardt <xypron.glpk@gmx.de>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
5 years agomemcg: only record foreign writebacks with dirty pages when memcg is not disabled
Baoquan He [Mon, 7 Oct 2019 00:58:15 +0000 (17:58 -0700)]
memcg: only record foreign writebacks with dirty pages when memcg is not disabled

In kdump kernel, memcg usually is disabled with 'cgroup_disable=memory'
for saving memory.  Now kdump kernel will always panic when dump vmcore
to local disk:

  BUG: kernel NULL pointer dereference, address: 0000000000000ab8
  Oops: 0000 [#1] SMP NOPTI
  CPU: 0 PID: 598 Comm: makedumpfile Not tainted 5.3.0+ #26
  Hardware name: HPE ProLiant DL385 Gen10/ProLiant DL385 Gen10, BIOS A40 10/02/2018
  RIP: 0010:mem_cgroup_track_foreign_dirty_slowpath+0x38/0x140
  Call Trace:
   __set_page_dirty+0x52/0xc0
   iomap_set_page_dirty+0x50/0x90
   iomap_write_end+0x6e/0x270
   iomap_write_actor+0xce/0x170
   iomap_apply+0xba/0x11e
   iomap_file_buffered_write+0x62/0x90
   xfs_file_buffered_aio_write+0xca/0x320 [xfs]
   new_sync_write+0x12d/0x1d0
   vfs_write+0xa5/0x1a0
   ksys_write+0x59/0xd0
   do_syscall_64+0x59/0x1e0
   entry_SYSCALL_64_after_hwframe+0x44/0xa9

And this will corrupt the 1st kernel too with 'cgroup_disable=memory'.

Via the trace and with debugging, it is pointing to commit 97b27821b485
("writeback, memcg: Implement foreign dirty flushing") which introduced
this regression.  Disabling memcg causes the null pointer dereference at
uninitialized data in function mem_cgroup_track_foreign_dirty_slowpath().

Fix it by returning directly if memcg is disabled, but not trying to
record the foreign writebacks with dirty pages.

Link: http://lkml.kernel.org/r/20190924141928.GD31919@MiWiFi-R3L-srv
Fixes: 97b27821b485 ("writeback, memcg: Implement foreign dirty flushing")
Signed-off-by: Baoquan He <bhe@redhat.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Jan Kara <jack@suse.cz>
Cc: Tejun Heo <tj@kernel.org>
Cc: Jens Axboe <axboe@kernel.dk>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
5 years agomm: fix -Wmissing-prototypes warnings
Yi Wang [Mon, 7 Oct 2019 00:58:12 +0000 (17:58 -0700)]
mm: fix -Wmissing-prototypes warnings

We get two warnings when build kernel W=1:

  mm/shuffle.c:36:12: warning: no previous prototype for `shuffle_show' [-Wmissing-prototypes]
  mm/sparse.c:220:6: warning: no previous prototype for `subsection_mask_set' [-Wmissing-prototypes]

Make the functions static to fix this.

Link: http://lkml.kernel.org/r/1566978161-7293-1-git-send-email-wang.yi59@zte.com.cn
Signed-off-by: Yi Wang <wang.yi59@zte.com.cn>
Reviewed-by: David Hildenbrand <david@redhat.com>
Reviewed-by: Kees Cook <keescook@chromium.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
5 years agowriteback: fix use-after-free in finish_writeback_work()
Tejun Heo [Mon, 7 Oct 2019 00:58:09 +0000 (17:58 -0700)]
writeback: fix use-after-free in finish_writeback_work()

finish_writeback_work() reads @done->waitq after decrementing
@done->cnt.  However, once @done->cnt reaches zero, @done may be freed
(from stack) at any moment and @done->waitq can contain something
unrelated by the time finish_writeback_work() tries to read it.  This
led to the following crash.

  "BUG: kernel NULL pointer dereference, address: 0000000000000002"
  #PF: supervisor write access in kernel mode
  #PF: error_code(0x0002) - not-present page
  PGD 0 P4D 0
  Oops: 0002 [#1] SMP DEBUG_PAGEALLOC
  CPU: 40 PID: 555153 Comm: kworker/u98:50 Kdump: loaded Not tainted
  ...
  Workqueue: writeback wb_workfn (flush-btrfs-1)
  RIP: 0010:_raw_spin_lock_irqsave+0x10/0x30
  Code: 48 89 d8 5b c3 e8 50 db 6b ff eb f4 0f 1f 40 00 66 2e 0f 1f 84 00 00 00 00 00 0f 1f 44 00 00 53 9c 5b fa 31 c0 ba 01 00 00 00 <f0> 0f b1 17 75 05 48 89 d8 5b c3 89 c6 e8 fe ca 6b ff eb f2 66 90
  RSP: 0018:ffffc90049b27d98 EFLAGS: 00010046
  RAX: 0000000000000000 RBX: 0000000000000246 RCX: 0000000000000000
  RDX: 0000000000000001 RSI: 0000000000000003 RDI: 0000000000000002
  RBP: 0000000000000000 R08: 0000000000000000 R09: 0000000000000001
  R10: ffff889fff407600 R11: ffff88ba9395d740 R12: 000000000000e300
  R13: 0000000000000003 R14: 0000000000000000 R15: 0000000000000000
  FS:  0000000000000000(0000) GS:ffff88bfdfa00000(0000) knlGS:0000000000000000
  CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
  CR2: 0000000000000002 CR3: 0000000002409005 CR4: 00000000001606e0
  Call Trace:
   __wake_up_common_lock+0x63/0xc0
   wb_workfn+0xd2/0x3e0
   process_one_work+0x1f5/0x3f0
   worker_thread+0x2d/0x3d0
   kthread+0x111/0x130
   ret_from_fork+0x1f/0x30

Fix it by reading and caching @done->waitq before decrementing
@done->cnt.

Link: http://lkml.kernel.org/r/20190924010631.GH2233839@devbig004.ftw2.facebook.com
Fixes: 5b9cce4c7eb069 ("writeback: Generalize and expose wb_completion")
Signed-off-by: Tejun Heo <tj@kernel.org>
Debugged-by: Chris Mason <clm@fb.com>
Reviewed-by: Jens Axboe <axboe@kernel.dk>
Cc: Jan Kara <jack@suse.cz>
Cc: <stable@vger.kernel.org> [5.2+]
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
5 years agomm/memremap: drop unused SECTION_SIZE and SECTION_MASK
Anshuman Khandual [Mon, 7 Oct 2019 00:58:05 +0000 (17:58 -0700)]
mm/memremap: drop unused SECTION_SIZE and SECTION_MASK

SECTION_SIZE and SECTION_MASK macros are not getting used anymore.  But
they do conflict with existing definitions on arm64 platform causing
following warning during build.  Lets drop these unused macros.

  mm/memremap.c:16: warning: "SECTION_MASK" redefined
   #define SECTION_MASK ~((1UL << PA_SECTION_SHIFT) - 1)
  arch/arm64/include/asm/pgtable-hwdef.h:79: note: this is the location of the previous definition
   #define SECTION_MASK  (~(SECTION_SIZE-1))

  mm/memremap.c:17: warning: "SECTION_SIZE" redefined
   #define SECTION_SIZE (1UL << PA_SECTION_SHIFT)
  arch/arm64/include/asm/pgtable-hwdef.h:78: note: this is the location of the previous definition
   #define SECTION_SIZE  (_AC(1, UL) << SECTION_SHIFT)

Link: http://lkml.kernel.org/r/1569312010-31313-1-git-send-email-anshuman.khandual@arm.com
Signed-off-by: Anshuman Khandual <anshuman.khandual@arm.com>
Reported-by: kbuild test robot <lkp@intel.com>
Reviewed-by: David Hildenbrand <david@redhat.com>
Cc: Dan Williams <dan.j.williams@intel.com>
Cc: Jason Gunthorpe <jgg@ziepe.ca>
Cc: Logan Gunthorpe <logang@deltatee.com>
Cc: Ira Weiny <ira.weiny@intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
5 years agopanic: ensure preemption is disabled during panic()
Will Deacon [Mon, 7 Oct 2019 00:58:00 +0000 (17:58 -0700)]
panic: ensure preemption is disabled during panic()

Calling 'panic()' on a kernel with CONFIG_PREEMPT=y can leave the
calling CPU in an infinite loop, but with interrupts and preemption
enabled.  From this state, userspace can continue to be scheduled,
despite the system being "dead" as far as the kernel is concerned.

This is easily reproducible on arm64 when booting with "nosmp" on the
command line; a couple of shell scripts print out a periodic "Ping"
message whilst another triggers a crash by writing to
/proc/sysrq-trigger:

  | sysrq: Trigger a crash
  | Kernel panic - not syncing: sysrq triggered crash
  | CPU: 0 PID: 1 Comm: init Not tainted 5.2.15 #1
  | Hardware name: linux,dummy-virt (DT)
  | Call trace:
  |  dump_backtrace+0x0/0x148
  |  show_stack+0x14/0x20
  |  dump_stack+0xa0/0xc4
  |  panic+0x140/0x32c
  |  sysrq_handle_reboot+0x0/0x20
  |  __handle_sysrq+0x124/0x190
  |  write_sysrq_trigger+0x64/0x88
  |  proc_reg_write+0x60/0xa8
  |  __vfs_write+0x18/0x40
  |  vfs_write+0xa4/0x1b8
  |  ksys_write+0x64/0xf0
  |  __arm64_sys_write+0x14/0x20
  |  el0_svc_common.constprop.0+0xb0/0x168
  |  el0_svc_handler+0x28/0x78
  |  el0_svc+0x8/0xc
  | Kernel Offset: disabled
  | CPU features: 0x0002,24002004
  | Memory Limit: none
  | ---[ end Kernel panic - not syncing: sysrq triggered crash ]---
  |  Ping 2!
  |  Ping 1!
  |  Ping 1!
  |  Ping 2!

The issue can also be triggered on x86 kernels if CONFIG_SMP=n,
otherwise local interrupts are disabled in 'smp_send_stop()'.

Disable preemption in 'panic()' before re-enabling interrupts.

Link: http://lkml.kernel.org/r/20191002123538.22609-1-will@kernel.org
Link: https://lore.kernel.org/r/BX1W47JXPMR8.58IYW53H6M5N@dragonstone
Signed-off-by: Will Deacon <will@kernel.org>
Reported-by: Xogium <contact@xogium.me>
Reviewed-by: Kees Cook <keescook@chromium.org>
Cc: Russell King <linux@armlinux.org.uk>
Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Petr Mladek <pmladek@suse.com>
Cc: Feng Tang <feng.tang@intel.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
5 years agofs: ocfs2: fix a possible null-pointer dereference in ocfs2_info_scan_inode_alloc()
Jia-Ju Bai [Mon, 7 Oct 2019 00:57:57 +0000 (17:57 -0700)]
fs: ocfs2: fix a possible null-pointer dereference in ocfs2_info_scan_inode_alloc()

In ocfs2_info_scan_inode_alloc(), there is an if statement on line 283
to check whether inode_alloc is NULL:

    if (inode_alloc)

When inode_alloc is NULL, it is used on line 287:

    ocfs2_inode_lock(inode_alloc, &bh, 0);
        ocfs2_inode_lock_full_nested(inode, ...)
            struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);

Thus, a possible null-pointer dereference may occur.

To fix this bug, inode_alloc is checked on line 286.

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

Link: http://lkml.kernel.org/r/20190726033717.32359-1-baijiaju1990@gmail.com
Signed-off-by: Jia-Ju Bai <baijiaju1990@gmail.com>
Reviewed-by: Joseph Qi <joseph.qi@linux.alibaba.com>
Cc: Mark Fasheh <mark@fasheh.com>
Cc: Joel Becker <jlbec@evilplan.org>
Cc: Junxiao Bi <junxiao.bi@oracle.com>
Cc: Changwei Ge <gechangwei@live.cn>
Cc: Gang He <ghe@suse.com>
Cc: Jun Piao <piaojun@huawei.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
5 years agofs: ocfs2: fix a possible null-pointer dereference in ocfs2_write_end_nolock()
Jia-Ju Bai [Mon, 7 Oct 2019 00:57:54 +0000 (17:57 -0700)]
fs: ocfs2: fix a possible null-pointer dereference in ocfs2_write_end_nolock()

In ocfs2_write_end_nolock(), there are an if statement on lines 1976,
2047 and 2058, to check whether handle is NULL:

    if (handle)

When handle is NULL, it is used on line 2045:

ocfs2_update_inode_fsync_trans(handle, inode, 1);
        oi->i_sync_tid = handle->h_transaction->t_tid;

Thus, a possible null-pointer dereference may occur.

To fix this bug, handle is checked before calling
ocfs2_update_inode_fsync_trans().

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

Link: http://lkml.kernel.org/r/20190726033705.32307-1-baijiaju1990@gmail.com
Signed-off-by: Jia-Ju Bai <baijiaju1990@gmail.com>
Reviewed-by: Joseph Qi <joseph.qi@linux.alibaba.com>
Cc: Mark Fasheh <mark@fasheh.com>
Cc: Joel Becker <jlbec@evilplan.org>
Cc: Junxiao Bi <junxiao.bi@oracle.com>
Cc: Changwei Ge <gechangwei@live.cn>
Cc: Gang He <ghe@suse.com>
Cc: Jun Piao <piaojun@huawei.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
5 years agofs: ocfs2: fix possible null-pointer dereferences in ocfs2_xa_prepare_entry()
Jia-Ju Bai [Mon, 7 Oct 2019 00:57:50 +0000 (17:57 -0700)]
fs: ocfs2: fix possible null-pointer dereferences in ocfs2_xa_prepare_entry()

In ocfs2_xa_prepare_entry(), there is an if statement on line 2136 to
check whether loc->xl_entry is NULL:

    if (loc->xl_entry)

When loc->xl_entry is NULL, it is used on line 2158:

    ocfs2_xa_add_entry(loc, name_hash);
        loc->xl_entry->xe_name_hash = cpu_to_le32(name_hash);
        loc->xl_entry->xe_name_offset = cpu_to_le16(loc->xl_size);

and line 2164:

    ocfs2_xa_add_namevalue(loc, xi);
        loc->xl_entry->xe_value_size = cpu_to_le64(xi->xi_value_len);
        loc->xl_entry->xe_name_len = xi->xi_name_len;

Thus, possible null-pointer dereferences may occur.

To fix these bugs, if loc-xl_entry is NULL, ocfs2_xa_prepare_entry()
abnormally returns with -EINVAL.

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

[akpm@linux-foundation.org: remove now-unused ocfs2_xa_add_entry()]
Link: http://lkml.kernel.org/r/20190726101447.9153-1-baijiaju1990@gmail.com
Signed-off-by: Jia-Ju Bai <baijiaju1990@gmail.com>
Reviewed-by: Joseph Qi <joseph.qi@linux.alibaba.com>
Cc: Mark Fasheh <mark@fasheh.com>
Cc: Joel Becker <jlbec@evilplan.org>
Cc: Junxiao Bi <junxiao.bi@oracle.com>
Cc: Changwei Ge <gechangwei@live.cn>
Cc: Gang He <ghe@suse.com>
Cc: Jun Piao <piaojun@huawei.com>
Cc: Stephen Rothwell <sfr@canb.auug.org.au>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
5 years agoocfs2: clear zero in unaligned direct IO
Jia Guo [Mon, 7 Oct 2019 00:57:47 +0000 (17:57 -0700)]
ocfs2: clear zero in unaligned direct IO

Unused portion of a part-written fs-block-sized block is not set to zero
in unaligned append direct write.This can lead to serious data
inconsistencies.

Ocfs2 manage disk with cluster size(for example, 1M), part-written in
one cluster will change the cluster state from UN-WRITTEN to WRITTEN,
VFS(function dio_zero_block) doesn't do the cleaning because bh's state
is not set to NEW in function ocfs2_dio_wr_get_block when we write a
WRITTEN cluster.  For example, the cluster size is 1M, file size is 8k
and we direct write from 14k to 15k, then 12k~14k and 15k~16k will
contain dirty data.

We have to deal with two cases:
 1.The starting position of direct write is outside the file.
 2.The starting position of direct write is located in the file.

We need set bh's state to NEW in the first case.  In the second case, we
need mapped twice because bh's state of area out file should be set to
NEW while area in file not.

[akpm@linux-foundation.org: coding style fixes]
Link: http://lkml.kernel.org/r/5292e287-8f1a-fd4a-1a14-661e555e0bed@huawei.com
Signed-off-by: Jia Guo <guojia12@huawei.com>
Reviewed-by: Yiwen Jiang <jiangyiwen@huawei.com>
Cc: Mark Fasheh <mark@fasheh.com>
Cc: Joel Becker <jlbec@evilplan.org>
Cc: Junxiao Bi <junxiao.bi@oracle.com>
Cc: Joseph Qi <joseph.qi@huawei.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
5 years agoMerge tag 'mips_fixes_5.4_1' of git://git.kernel.org/pub/scm/linux/kernel/git/mips...
Linus Torvalds [Fri, 4 Oct 2019 20:31:56 +0000 (13:31 -0700)]
Merge tag 'mips_fixes_5.4_1' of git://git./linux/kernel/git/mips/linux

Pull MIPS fixes from Paul Burton:

 - Build fixes for Cavium Octeon & PMC-Sierra MSP systems, as well as
   all pre-MIPSr6 configurations built with binutils < 2.25.

 - Boot fixes for 64-bit Loongson systems & SGI IP28 systems.

 - Wire up the new clone3 syscall.

 - Clean ups for a few build-time warnings.

* tag 'mips_fixes_5.4_1' of git://git.kernel.org/pub/scm/linux/kernel/git/mips/linux:
  MIPS: fw/arc: Remove unused addr variable
  MIPS: pmcs-msp71xx: Remove unused addr variable
  MIPS: pmcs-msp71xx: Add missing MAX_PROM_MEM definition
  mips: Loongson: Fix the link time qualifier of 'serial_exit()'
  MIPS: init: Prevent adding memory before PHYS_OFFSET
  MIPS: init: Fix reservation of memory between PHYS_OFFSET and mem start
  MIPS: VDSO: Fix build for binutils < 2.25
  MIPS: VDSO: Remove unused gettimeofday.c
  MIPS: Wire up clone3 syscall
  MIPS: octeon: Include required header; fix octeon ethernet build
  MIPS: cpu-bugs64: Mark inline functions as __always_inline
  MIPS: dts: ar9331: fix interrupt-controller size
  MIPS: Loongson64: Fix boot failure after dropping boot_mem_map

5 years agoMerge tag 'riscv/for-v5.4-rc2' of git://git.kernel.org/pub/scm/linux/kernel/git/riscv...
Linus Torvalds [Fri, 4 Oct 2019 20:02:47 +0000 (13:02 -0700)]
Merge tag 'riscv/for-v5.4-rc2' of git://git./linux/kernel/git/riscv/linux

Pull RISC-V fixes from Paul Walmsley:

 - Ensure that exclusive-load reservations are terminated after system
   call or exception handling. This primarily affects QEMU, which does
   not expire load reservations.

 - Fix an issue primarily affecting RV32 platforms that can cause the DT
   header to be corrupted, causing boot failures.

* tag 'riscv/for-v5.4-rc2' of git://git.kernel.org/pub/scm/linux/kernel/git/riscv/linux:
  riscv: Fix memblock reservation for device tree blob
  RISC-V: Clear load reservations while restoring hart contexts

5 years agoMerge tag 'devicetree-fixes-for-5.4' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Fri, 4 Oct 2019 19:57:45 +0000 (12:57 -0700)]
Merge tag 'devicetree-fixes-for-5.4' of git://git./linux/kernel/git/robh/linux

Pull DeviceTree fixes from Rob Herring:
 "Fix several 'dt_binding_check' build failures"

* tag 'devicetree-fixes-for-5.4' of git://git.kernel.org/pub/scm/linux/kernel/git/robh/linux:
  dt-bindings: phy: lantiq: Fix Property Name
  dt-bindings: iio: ad7192: Fix DTC warning in the example
  dt-bindings: iio: ad7192: Fix Regulator Properties
  dt-bindings: media: rc: Fix redundant string
  dt-bindings: dsp: Fix fsl,dsp example

5 years agoMIPS: fw/arc: Remove unused addr variable
Paul Burton [Fri, 4 Oct 2019 17:41:02 +0000 (17:41 +0000)]
MIPS: fw/arc: Remove unused addr variable

The addr variable in prom_free_prom_memory() has been unused since
commit 0df1007677d5 ("MIPS: fw: Record prom memory"), leading to a
compiler warning:

  arch/mips/fw/arc/memory.c:163:16:
    warning: unused variable 'addr' [-Wunused-variable]

Fix this by removing the unused variable.

Signed-off-by: Paul Burton <paul.burton@mips.com>
Fixes: 0df1007677d5 ("MIPS: fw: Record prom memory")
Cc: Jiaxun Yang <jiaxun.yang@flygoat.com>
Cc: linux-mips@vger.kernel.org
5 years agoMerge tag 'for-linus' of git://git.kernel.org/pub/scm/virt/kvm/kvm
Linus Torvalds [Fri, 4 Oct 2019 18:17:51 +0000 (11:17 -0700)]
Merge tag 'for-linus' of git://git./virt/kvm/kvm

Pull KVM fixes from Paolo Bonzini:
 "ARM and x86 bugfixes of all kinds.

  The most visible one is that migrating a nested hypervisor has always
  been busted on Broadwell and newer processors, and that has finally
  been fixed"

* tag 'for-linus' of git://git.kernel.org/pub/scm/virt/kvm/kvm: (22 commits)
  KVM: x86: omit "impossible" pmu MSRs from MSR list
  KVM: nVMX: Fix consistency check on injected exception error code
  KVM: x86: omit absent pmu MSRs from MSR list
  selftests: kvm: Fix libkvm build error
  kvm: vmx: Limit guest PMCs to those supported on the host
  kvm: x86, powerpc: do not allow clearing largepages debugfs entry
  KVM: selftests: x86: clarify what is reported on KVM_GET_MSRS failure
  KVM: VMX: Set VMENTER_L1D_FLUSH_NOT_REQUIRED if !X86_BUG_L1TF
  selftests: kvm: add test for dirty logging inside nested guests
  KVM: x86: fix nested guest live migration with PML
  KVM: x86: assign two bits to track SPTE kinds
  KVM: x86: Expose XSAVEERPTR to the guest
  kvm: x86: Enumerate support for CLZERO instruction
  kvm: x86: Use AMD CPUID semantics for AMD vCPUs
  kvm: x86: Improve emulation of CPUID leaves 0BH and 1FH
  KVM: X86: Fix userspace set invalid CR4
  kvm: x86: Fix a spurious -E2BIG in __do_cpuid_func
  KVM: LAPIC: Loosen filter for adaptive tuning of lapic_timer_advance_ns
  KVM: arm/arm64: vgic: Use the appropriate TRACE_INCLUDE_PATH
  arm64: KVM: Kill hyp_alternate_select()
  ...

5 years agoMerge tag 'for-linus-5.4-rc2-tag' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Fri, 4 Oct 2019 18:13:09 +0000 (11:13 -0700)]
Merge tag 'for-linus-5.4-rc2-tag' of git://git./linux/kernel/git/xen/tip

Pull xen fixes and cleanups from Juergen Gross:

 - a fix in the Xen balloon driver avoiding hitting a BUG_ON() in some
   cases, plus a follow-on cleanup series for that driver

 - a patch for introducing non-blocking EFI callbacks in Xen's EFI
   driver, plu a cleanup patch for Xen EFI handling merging the x86 and
   ARM arch specific initialization into the Xen EFI driver

 - a fix of the Xen xenbus driver avoiding a self-deadlock when cleaning
   up after a user process has died

 - a fix for Xen on ARM after removal of ZONE_DMA

 - a cleanup patch for avoiding build warnings for Xen on ARM

* tag 'for-linus-5.4-rc2-tag' of git://git.kernel.org/pub/scm/linux/kernel/git/xen/tip:
  xen/xenbus: fix self-deadlock after killing user process
  xen/efi: have a common runtime setup function
  arm: xen: mm: use __GPF_DMA32 for arm64
  xen/balloon: Clear PG_offline in balloon_retrieve()
  xen/balloon: Mark pages PG_offline in balloon_append()
  xen/balloon: Drop __balloon_append()
  xen/balloon: Set pages PageOffline() in balloon_add_region()
  ARM: xen: unexport HYPERVISOR_platform_op function
  xen/efi: Set nonblocking callbacks

5 years agoMerge tag 'copy-struct-from-user-v5.4-rc2' of git://git.kernel.org/pub/scm/linux...
Linus Torvalds [Fri, 4 Oct 2019 17:36:31 +0000 (10:36 -0700)]
Merge tag 'copy-struct-from-user-v5.4-rc2' of git://git./linux/kernel/git/brauner/linux

Pull copy_struct_from_user() helper from Christian Brauner:
 "This contains the copy_struct_from_user() helper which got split out
  from the openat2() patchset. It is a generic interface designed to
  copy a struct from userspace.

  The helper will be especially useful for structs versioned by size of
  which we have quite a few. This allows for backwards compatibility,
  i.e. an extended struct can be passed to an older kernel, or a legacy
  struct can be passed to a newer kernel. For the first case (extended
  struct, older kernel) the new fields in an extended struct can be set
  to zero and the struct safely passed to an older kernel.

  The most obvious benefit is that this helper lets us get rid of
  duplicate code present in at least sched_setattr(), perf_event_open(),
  and clone3(). More importantly it will also help to ensure that users
  implementing versioning-by-size end up with the same core semantics.

  This point is especially crucial since we have at least one case where
  versioning-by-size is used but with slighly different semantics:
  sched_setattr(), perf_event_open(), and clone3() all do do similar
  checks to copy_struct_from_user() while rt_sigprocmask(2) always
  rejects differently-sized struct arguments.

  With this pull request we also switch over sched_setattr(),
  perf_event_open(), and clone3() to use the new helper"

* tag 'copy-struct-from-user-v5.4-rc2' of git://git.kernel.org/pub/scm/linux/kernel/git/brauner/linux:
  usercopy: Add parentheses around assignment in test_copy_struct_from_user
  perf_event_open: switch to copy_struct_from_user()
  sched_setattr: switch to copy_struct_from_user()
  clone3: switch to copy_struct_from_user()
  lib: introduce copy_struct_from_user() helper

5 years agoMerge tag 'for-linus-20191003' of git://git.kernel.org/pub/scm/linux/kernel/git/braun...
Linus Torvalds [Fri, 4 Oct 2019 17:18:56 +0000 (10:18 -0700)]
Merge tag 'for-linus-20191003' of git://git./linux/kernel/git/brauner/linux

Pull clone3/pidfd fixes from Christian Brauner:
 "This contains a couple of fixes:

   - Fix pidfd selftest compilation (Shuah Kahn)

     Due to a false linking instruction in the Makefile compilation for
     the pidfd selftests would fail on some systems.

   - Fix compilation for glibc on RISC-V systems (Seth Forshee)

     In some scenarios linux/uapi/linux/sched.h is included where
     __ASSEMBLY__ is defined causing a build failure because struct
     clone_args was not guarded by an #ifndef __ASSEMBLY__.

   - Add missing clone3() and struct clone_args kernel-doc (Christian Brauner)

     clone3() and struct clone_args were missing kernel-docs. (The goal
     is to use kernel-doc for any function or type where it's worth it.)
     For struct clone_args this also contains a comment about the fact
     that it's versioned by size"

* tag 'for-linus-20191003' of git://git.kernel.org/pub/scm/linux/kernel/git/brauner/linux:
  sched: add kernel-doc for struct clone_args
  fork: add kernel-doc for clone3
  selftests: pidfd: Fix undefined reference to pthread_create()
  sched: Add __ASSEMBLY__ guards around struct clone_args

5 years agoMerge tag 'drm-fixes-2019-10-04' of git://anongit.freedesktop.org/drm/drm
Linus Torvalds [Fri, 4 Oct 2019 17:12:37 +0000 (10:12 -0700)]
Merge tag 'drm-fixes-2019-10-04' of git://anongit.freedesktop.org/drm/drm

Pull drm fixes from Dave Airlie:
 "Been offline for 3 days, got back and had some fixes queued up.

  Nothing too major, the i915 dp-mst fix is important, and amdgpu has a
  bulk move speedup fix and some regressions, but nothing too insane for
  an rc2 pull. The intel fixes are also 2 weeks worth, they missed the
  boat last week.

  core:
   - writeback fixes

  i915:
   - Fix DP-MST crtc_mask
   - Fix dsc dpp calculations
   - Fix g4x sprite scaling stride check with GTT remapping
   - Fix concurrence on cases where requests where getting retired at
     same time as resubmitted to HW
   - Fix gen9 display resolutions by setting the right max plane width
   - Fix GPU hang on preemption
   - Mark contents as dirty on a write fault. This was breaking cursor
     sprite with dumb buffers.

  komeda:
   - memory leak fix

  tilcdc:
   - include fix

  amdgpu:
   - Enable bulk moves
   - Power metrics fixes for Navi
   - Fix S4 regression
   - Add query for tcc disabled mask
   - Fix several leaks in error paths
   - randconfig fixes
   - clang fixes"

* tag 'drm-fixes-2019-10-04' of git://anongit.freedesktop.org/drm/drm: (21 commits)
  Revert "drm/i915: Fix DP-MST crtc_mask"
  drm/omap: fix max fclk divider for omap36xx
  drm/i915: Fix g4x sprite scaling stride check with GTT remapping
  drm/i915/dp: Fix dsc bpp calculations, v5.
  drm/amd/display: fix dcn21 Makefile for clang
  drm/amd/display: hide an unused variable
  drm/amdgpu: display_mode_vba_21: remove uint typedef
  drm/amdgpu: hide another #warning
  drm/amdgpu: make pmu support optional, again
  drm/amd/display: memory leak
  drm/amdgpu: fix multiple memory leaks in acp_hw_init
  drm/amdgpu: return tcc_disabled_mask to userspace
  drm/amdgpu: don't increment vram lost if we are in hibernation
  Revert "drm/amdgpu: disable stutter mode for renoir"
  drm/amd/powerplay: add sensor lock support for smu
  drm/amd/powerplay: change metrics update period from 1ms to 100ms
  drm/amdgpu: revert "disable bulk moves for now"
  drm/tilcdc: include linux/pinctrl/consumer.h again
  drm/komeda: prevent memory leak in komeda_wb_connector_add
  drm: Clear the fence pointer when writeback job signaled
  ...

5 years agoMerge tag 'for-linus-2019-10-03' of git://git.kernel.dk/linux-block
Linus Torvalds [Fri, 4 Oct 2019 16:56:51 +0000 (09:56 -0700)]
Merge tag 'for-linus-2019-10-03' of git://git.kernel.dk/linux-block

Pull block fixes from Jens Axboe:

 - Mandate timespec64 for the io_uring timeout ABI (Arnd)

 - Set of NVMe changes via Sagi:
     - controller removal race fix from Balbir
     - quirk additions from Gabriel and Jian-Hong
     - nvme-pci power state save fix from Mario
     - Add 64bit user commands (for 64bit registers) from Marta
     - nvme-rdma/nvme-tcp fixes from Max, Mark and Me
     - Minor cleanups and nits from James, Dan and John

 - Two s390 dasd fixes (Jan, Stefan)

 - Have loop change block size in DIO mode (Martijn)

 - paride pg header ifdef guard (Masahiro)

 - Two blk-mq queue scheduler tweaks, fixing an ordering issue on zoned
   devices and suboptimal performance on others (Ming)

* tag 'for-linus-2019-10-03' of git://git.kernel.dk/linux-block: (22 commits)
  block: sed-opal: fix sparse warning: convert __be64 data
  block: sed-opal: fix sparse warning: obsolete array init.
  block: pg: add header include guard
  Revert "s390/dasd: Add discard support for ESE volumes"
  s390/dasd: Fix error handling during online processing
  io_uring: use __kernel_timespec in timeout ABI
  loop: change queue block size to match when using DIO
  blk-mq: apply normal plugging for HDD
  blk-mq: honor IO scheduler for multiqueue devices
  nvme-rdma: fix possible use-after-free in connect timeout
  nvme: Move ctrl sqsize to generic space
  nvme: Add ctrl attributes for queue_count and sqsize
  nvme: allow 64-bit results in passthru commands
  nvme: Add quirk for Kingston NVME SSD running FW E8FK11.T
  nvmet-tcp: remove superflous check on request sgl
  Added QUIRKs for ADATA XPG SX8200 Pro 512GB
  nvme-rdma: Fix max_hw_sectors calculation
  nvme: fix an error code in nvme_init_subsystem()
  nvme-pci: Save PCI state before putting drive into deepest state
  nvme-tcp: fix wrong stop condition in io_work
  ...

5 years agoKVM: x86: omit "impossible" pmu MSRs from MSR list
Paolo Bonzini [Tue, 1 Oct 2019 13:33:07 +0000 (15:33 +0200)]
KVM: x86: omit "impossible" pmu MSRs from MSR list

INTEL_PMC_MAX_GENERIC is currently 32, which exceeds the 18
contiguous MSR indices reserved by Intel for event selectors.
Since some machines actually have MSRs past the reserved range,
filtering them against x86_pmu.num_counters_gp may have false
positives.  Cut the list to 18 entries to avoid this.

Reported-by: Vitaly Kuznetsov <vkuznets@redhat.com>
Suggested-by: Vitaly Kuznetsov <vkuznets@redhat.com>
Cc: Jim Mattson <jamttson@google.com>
Fixes: e2ada66ec418 ("kvm: x86: Add Intel PMU MSRs to msrs_to_save[]", 2019-08-21)
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
5 years agoMerge tag 'drm-intel-fixes-2019-10-03-1' of git://anongit.freedesktop.org/drm/drm...
Dave Airlie [Fri, 4 Oct 2019 06:31:03 +0000 (16:31 +1000)]
Merge tag 'drm-intel-fixes-2019-10-03-1' of git://anongit.freedesktop.org/drm/drm-intel into drm-fixes

- Fix DP-MST crtc_mask
- Fix dsc dpp calculations
- Fix g4x sprite scaling stride check with GTT remapping

Signed-off-by: Dave Airlie <airlied@redhat.com>
From: Rodrigo Vivi <rodrigo.vivi@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20191003193051.GA26421@intel.com
5 years agoMerge tag 'drm-misc-fixes-2019-10-03' of git://anongit.freedesktop.org/drm/drm-misc...
Dave Airlie [Fri, 4 Oct 2019 06:30:35 +0000 (16:30 +1000)]
Merge tag 'drm-misc-fixes-2019-10-03' of git://anongit.freedesktop.org/drm/drm-misc into drm-fixes

 - One include fix for tilcdc
 - A clock fix for OMAP
 - A memory leak fix for Komeda
 - Some fixes for resources cleanups with writeback

Signed-off-by: Dave Airlie <airlied@redhat.com>
From: Maxime Ripard <mripard@kernel.org>
Link: https://patchwork.freedesktop.org/patch/msgid/20191003081031.oykms5fg4tijvdri@gilmour
5 years agoMerge tag 'drm-fixes-5.4-2019-10-02' of git://people.freedesktop.org/~agd5f/linux...
Dave Airlie [Fri, 4 Oct 2019 06:27:51 +0000 (16:27 +1000)]
Merge tag 'drm-fixes-5.4-2019-10-02' of git://people.freedesktop.org/~agd5f/linux into drm-fixes

drm-fixes-5.4-2019-10-02:

amdgpu:
- Enable bulk moves
- Power metrics fixes for Navi
- Fix S4 regression
- Add query for tcc disabled mask
- Fix several leaks in error paths
- randconfig fixes
- clang fixes

Signed-off-by: Dave Airlie <airlied@redhat.com>
From: Alex Deucher <alexdeucher@gmail.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20191002204909.3519-1-alexander.deucher@amd.com
5 years agoMIPS: pmcs-msp71xx: Remove unused addr variable
Paul Burton [Thu, 3 Oct 2019 22:46:36 +0000 (22:46 +0000)]
MIPS: pmcs-msp71xx: Remove unused addr variable

The addr variable in prom_free_prom_memory() has been unused since
commit b3c948e2c00f ("MIPS: msp: Record prom memory"), causing a warning
& build failure due to -Werror. Remove the unused variable.

Signed-off-by: Paul Burton <paul.burton@mips.com>
Fixes: b3c948e2c00f ("MIPS: msp: Record prom memory")
Cc: Jiaxun Yang <jiaxun.yang@flygoat.com>
Cc: linux-mips@vger.kernel.org
5 years agoMIPS: pmcs-msp71xx: Add missing MAX_PROM_MEM definition
Paul Burton [Thu, 3 Oct 2019 22:46:23 +0000 (22:46 +0000)]
MIPS: pmcs-msp71xx: Add missing MAX_PROM_MEM definition

Commit b3c948e2c00f ("MIPS: msp: Record prom memory") introduced use of
a MAX_PROM_MEM value but didn't define it. A bounds check in
prom_meminit() suggests its value was supposed to be 5, so define it as
such & adjust the bounds check to use the macro rather than a magic
number.

Signed-off-by: Paul Burton <paul.burton@mips.com>
Fixes: b3c948e2c00f ("MIPS: msp: Record prom memory")
Cc: Jiaxun Yang <jiaxun.yang@flygoat.com>
Cc: linux-mips@vger.kernel.org
5 years agovfs: Fix EOVERFLOW testing in put_compat_statfs64
Eric Sandeen [Wed, 2 Oct 2019 21:17:54 +0000 (16:17 -0500)]
vfs: Fix EOVERFLOW testing in put_compat_statfs64

Today, put_compat_statfs64() disallows nearly any field value over
2^32 if f_bsize is only 32 bits, but that makes no sense.
compat_statfs64 is there for the explicit purpose of providing 64-bit
fields for f_files, f_ffree, etc.  And f_bsize is always only 32 bits.

As a result, 32-bit userspace gets -EOVERFLOW for i.e.  large file
counts even with -D_FILE_OFFSET_BITS=64 set.

In reality, only f_bsize and f_frsize can legitimately overflow
(fields like f_type and f_namelen should never be large), so test
only those fields.

This bug was discussed at length some time ago, and this is the proposal
Al suggested at https://lkml.org/lkml/2018/8/6/640.  It seemed to get
dropped amid the discussion of other related changes, but this
part seems obviously correct on its own, so I've picked it up and
sent it, for expediency.

Fixes: 64d2ab32efe3 ("vfs: fix put_compat_statfs64() does not handle errors")
Signed-off-by: Eric Sandeen <sandeen@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
5 years agoblock: sed-opal: fix sparse warning: convert __be64 data
Randy Dunlap [Thu, 3 Oct 2019 02:23:15 +0000 (19:23 -0700)]
block: sed-opal: fix sparse warning: convert __be64 data

sparse warns about incorrect type when using __be64 data.
It is not being converted to CPU-endian but it should be.

Fixes these sparse warnings:

../block/sed-opal.c:375:20: warning: incorrect type in assignment (different base types)
../block/sed-opal.c:375:20:    expected unsigned long long [usertype] align
../block/sed-opal.c:375:20:    got restricted __be64 const [usertype] alignment_granularity
../block/sed-opal.c:376:25: warning: incorrect type in assignment (different base types)
../block/sed-opal.c:376:25:    expected unsigned long long [usertype] lowest_lba
../block/sed-opal.c:376:25:    got restricted __be64 const [usertype] lowest_aligned_lba

Fixes: 455a7b238cd6 ("block: Add Sed-opal library")
Cc: Scott Bauer <scott.bauer@intel.com>
Cc: Rafael Antognolli <rafael.antognolli@intel.com>
Cc: linux-block@vger.kernel.org
Reviewed-by: Jon Derrick <jonathan.derrick@intel.com>
Signed-off-by: Randy Dunlap <rdunlap@infradead.org>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
5 years agoblock: sed-opal: fix sparse warning: obsolete array init.
Randy Dunlap [Thu, 3 Oct 2019 02:23:05 +0000 (19:23 -0700)]
block: sed-opal: fix sparse warning: obsolete array init.

Fix sparse warning: (missing '=')
../block/sed-opal.c:133:17: warning: obsolete array initializer, use C99 syntax

Fixes: ff91064ea37c ("block: sed-opal: check size of shadow mbr")
Cc: linux-block@vger.kernel.org
Cc: Jonas Rabenstein <jonas.rabenstein@studium.uni-erlangen.de>
Cc: David Kozub <zub@linux.fjfi.cvut.cz>
Reviewed-by: Scott Bauer <sbauer@plzdonthack.me>
Reviewed-by: Revanth Rajashekar <revanth.rajashekar@intel.com>
Signed-off-by: Randy Dunlap <rdunlap@infradead.org>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
5 years agoRevert "drm/i915: Fix DP-MST crtc_mask"
Ville Syrjälä [Tue, 3 Sep 2019 15:40:18 +0000 (18:40 +0300)]
Revert "drm/i915: Fix DP-MST crtc_mask"

This reverts commit 4eaceea3a00f8e936a7f48dcd0c975a57f88930f.

Several userspace clients (modesetting ddx and mutter+wayland at least)
handle encoder.possible_crtcs incorrectly. What they essentially do is
the following:

possible_crtcs = ~0;
for_each_possible_encoder(connector)
possible_crtcs &= encoder->possible_crtcs;

Ie. they calculate the intersection of the possible_crtcs
for the connector when they really should be calculating the
union instead.

In our case each MST encoder now has just one unique bit set,
and so the intersection is always zero. The end result is that
MST connectors can't be lit up because no crtc can be found to
drive them.

I've submitted a fix for the modesetting ddx [1], and complained
on #wayland about mutter, so hopefully the situation will improve
in the future. In the meantime we have regression, and so must go
back to the old way of misconfiguring possible_crtcs in the kernel.

[1] https://gitlab.freedesktop.org/xorg/xserver/merge_requests/277

Cc: Jonas Ådahl <jadahl@gmail.com>
Cc: Stanislav Lisovskiy <stanislav.lisovskiy@intel.com>
Cc: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Cc: Dhinakaran Pandiyan <dhinakaran.pandiyan@intel.com>
Cc: Lucas De Marchi <lucas.demarchi@intel.com>
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=111507
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20190903154018.26357-1-ville.syrjala@linux.intel.com
Reviewed-by: José Roberto de Souza <jose.souza@intel.com>
(cherry picked from commit e838bfa8e170415fa3cc8e83ecb171e809c0c422)
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
5 years agosched: add kernel-doc for struct clone_args
Christian Brauner [Fri, 27 Sep 2019 15:29:05 +0000 (17:29 +0200)]
sched: add kernel-doc for struct clone_args

Add kernel-doc for struct clone_args for the clone3() syscall.

Link: https://lore.kernel.org/r/20191001114701.24661-3-christian.brauner@ubuntu.com
Signed-off-by: Christian Brauner <christian.brauner@ubuntu.com>
5 years agofork: add kernel-doc for clone3
Christian Brauner [Fri, 27 Sep 2019 15:28:42 +0000 (17:28 +0200)]
fork: add kernel-doc for clone3

Add kernel-doc for the clone3() syscall.

Link: https://lore.kernel.org/r/20191001114701.24661-2-christian.brauner@ubuntu.com
Signed-off-by: Christian Brauner <christian.brauner@ubuntu.com>
5 years agousercopy: Add parentheses around assignment in test_copy_struct_from_user
Nathan Chancellor [Thu, 3 Oct 2019 17:11:21 +0000 (10:11 -0700)]
usercopy: Add parentheses around assignment in test_copy_struct_from_user

Clang warns:

lib/test_user_copy.c:96:10: warning: using the result of an assignment
as a condition without parentheses [-Wparentheses]
        if (ret |= test(umem_src == NULL, "kmalloc failed"))
            ~~~~^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
lib/test_user_copy.c:96:10: note: place parentheses around the
assignment to silence this warning
        if (ret |= test(umem_src == NULL, "kmalloc failed"))
                ^
            (                                              )
lib/test_user_copy.c:96:10: note: use '!=' to turn this compound
assignment into an inequality comparison
        if (ret |= test(umem_src == NULL, "kmalloc failed"))
                ^~
                !=

Add the parentheses as it suggests because this is intentional.

Fixes: f5a1a536fa14 ("lib: introduce copy_struct_from_user() helper")
Link: https://github.com/ClangBuiltLinux/linux/issues/731
Signed-off-by: Nathan Chancellor <natechancellor@gmail.com>
Acked-by: Aleksa Sarai <cyphar@cyphar.com>
Acked-by: Christian Brauner <christian.brauner@ubuntu.com>
Link: https://lore.kernel.org/r/20191003171121.2723619-1-natechancellor@gmail.com
Signed-off-by: Christian Brauner <christian.brauner@ubuntu.com>
5 years agoMerge tag 'kgdb-5.4-rc2' of git://git.kernel.org/pub/scm/linux/kernel/git/danielt...
Linus Torvalds [Thu, 3 Oct 2019 18:17:57 +0000 (11:17 -0700)]
Merge tag 'kgdb-5.4-rc2' of git://git./linux/kernel/git/danielt/linux

Pull kgdb update from Daniel Thompson:
 "This is just a single patch adding a new reviewer for kgdb. New
  reviewers will be a big help so I decided to consider this to be a
  fix!

  I'm looking forward to working more closely with Doug"

* tag 'kgdb-5.4-rc2' of git://git.kernel.org/pub/scm/linux/kernel/git/danielt/linux:
  MAINTAINERS: kgdb: Add myself as a reviewer for kgdb/kdb

5 years agoMAINTAINERS: kgdb: Add myself as a reviewer for kgdb/kdb
Douglas Anderson [Fri, 20 Sep 2019 17:44:47 +0000 (10:44 -0700)]
MAINTAINERS: kgdb: Add myself as a reviewer for kgdb/kdb

I'm interested in kdb / kgdb and have sent various fixes over the
years.  I'd like to get CCed on patches so I can be aware of them and
also help review.

Signed-off-by: Douglas Anderson <dianders@chromium.org>
Acked-by: Daniel Thompson <daniel.thompson@linaro.org>
Acked-by: Will Deacon <will@kernel.org>
Signed-off-by: Daniel Thompson <daniel.thompson@linaro.org>
5 years agoKVM: nVMX: Fix consistency check on injected exception error code
Sean Christopherson [Tue, 1 Oct 2019 16:21:23 +0000 (09:21 -0700)]
KVM: nVMX: Fix consistency check on injected exception error code

Current versions of Intel's SDM incorrectly state that "bits 31:15 of
the VM-Entry exception error-code field" must be zero.  In reality, bits
31:16 must be zero, i.e. error codes are 16-bit values.

The bogus error code check manifests as an unexpected VM-Entry failure
due to an invalid code field (error number 7) in L1, e.g. when injecting
a #GP with error_code=0x9f00.

Nadav previously reported the bug[*], both to KVM and Intel, and fixed
the associated kvm-unit-test.

[*] https://patchwork.kernel.org/patch/11124749/

Reported-by: Nadav Amit <namit@vmware.com>
Cc: stable@vger.kernel.org
Signed-off-by: Sean Christopherson <sean.j.christopherson@intel.com>
Reviewed-by: Jim Mattson <jmattson@google.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
5 years agoMerge tag 'kvmarm-fixes-5.4-1' of git://git.kernel.org/pub/scm/linux/kernel/git/kvmar...
Paolo Bonzini [Thu, 3 Oct 2019 10:08:50 +0000 (12:08 +0200)]
Merge tag 'kvmarm-fixes-5.4-1' of git://git./linux/kernel/git/kvmarm/kvmarm into HEAD

KVM/arm fixes for 5.4, take #1

- Remove the now obsolete hyp_alternate_select construct
- Fix the TRACE_INCLUDE_PATH macro in the vgic code

5 years agoKVM: x86: omit absent pmu MSRs from MSR list
Paolo Bonzini [Tue, 1 Oct 2019 13:18:26 +0000 (15:18 +0200)]
KVM: x86: omit absent pmu MSRs from MSR list

INTEL_PMC_MAX_GENERIC is currently 32, which exceeds the 18 contiguous
MSR indices reserved by Intel for event selectors.  Since some machines
actually have MSRs past the reserved range, these may survive the
filtering of msrs_to_save array and would be rejected by KVM_GET/SET_MSR.
To avoid this, cut the list to whatever CPUID reports for the host's
architectural PMU.

Reported-by: Vitaly Kuznetsov <vkuznets@redhat.com>
Suggested-by: Vitaly Kuznetsov <vkuznets@redhat.com>
Cc: Jim Mattson <jmattson@google.com>
Fixes: e2ada66ec418 ("kvm: x86: Add Intel PMU MSRs to msrs_to_save[]", 2019-08-21)
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
5 years agoselftests: kvm: Fix libkvm build error
Shuah Khan [Wed, 2 Oct 2019 23:14:30 +0000 (17:14 -0600)]
selftests: kvm: Fix libkvm build error

Fix the following build error from "make TARGETS=kvm kselftest":

libkvm.a(assert.o): relocation R_X86_64_32 against `.rodata.str1.1' can not be used when making a PIE object; recompile with -fPIC

This error is seen when build is done from the main Makefile using
kselftest target. In this case KBUILD_CPPFLAGS and CC_OPTION_CFLAGS
are defined.

When build is invoked using:

"make -C tools/testing/selftests/kvm" KBUILD_CPPFLAGS and CC_OPTION_CFLAGS
aren't defined.

There is no need to pass in KBUILD_CPPFLAGS and CC_OPTION_CFLAGS for the
check to determine if --no-pie is necessary, which is the case when these
two aren't defined when "make -C tools/testing/selftests/kvm" runs.

Fix it by simplifying the no-pie-option logic. With this change, both
build variations work.

"make TARGETS=kvm kselftest"
"make -C tools/testing/selftests/kvm"

Signed-off-by: Shuah Khan <skhan@linuxfoundation.org>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
5 years agoMerge drm-misc-next-fixes-2019-10-02 into drm-misc-fixes
Maxime Ripard [Thu, 3 Oct 2019 08:00:13 +0000 (10:00 +0200)]
Merge drm-misc-next-fixes-2019-10-02 into drm-misc-fixes

One tilcdc fix was left out in drm-misc-next-fixes and didn't make it
during the merge window. Let's bring it into drm-misc-fixes.

Signed-off-by: Maxime Ripard <mripard@kernel.org>
5 years agoMerge drm/drm-fixes into drm-misc-fixes
Maxime Ripard [Thu, 3 Oct 2019 07:59:29 +0000 (09:59 +0200)]
Merge drm/drm-fixes into drm-misc-fixes

We haven't backmerged for a while, let's start the -rc period by pulling
rc1.

Signed-off-by: Maxime Ripard <mripard@kernel.org>
5 years agodrm/omap: fix max fclk divider for omap36xx
Tomi Valkeinen [Wed, 2 Oct 2019 12:25:42 +0000 (15:25 +0300)]
drm/omap: fix max fclk divider for omap36xx

The OMAP36xx and AM/DM37x TRMs say that the maximum divider for DSS fclk
(in CM_CLKSEL_DSS) is 32. Experimentation shows that this is not
correct, and using divider of 32 breaks DSS with a flood or underflows
and sync losts. Dividers up to 31 seem to work fine.

There is another patch to the DT files to limit the divider correctly,
but as the DSS driver also needs to know the maximum divider to be able
to iteratively find good rates, we also need to do the fix in the DSS
driver.

Signed-off-by: Tomi Valkeinen <tomi.valkeinen@ti.com>
Cc: Adam Ford <aford173@gmail.com>
Cc: stable@vger.kernel.org
Link: https://patchwork.freedesktop.org/patch/msgid/20191002122542.8449-1-tomi.valkeinen@ti.com
Tested-by: Adam Ford <aford173@gmail.com>
Reviewed-by: Jyri Sarha <jsarha@ti.com>
5 years agodrm/i915: Fix g4x sprite scaling stride check with GTT remapping
Ville Syrjälä [Mon, 30 Sep 2019 18:30:45 +0000 (21:30 +0300)]
drm/i915: Fix g4x sprite scaling stride check with GTT remapping

I forgot to update the g4x sprite scaling stride check when GTT
remapping was introduced. The stride of the original framebuffer
is irrelevant when remapping is used and instead we want to check
the stride of the remapped view.

Also drop the duplicate width_bytes check. We already check that
a few lines earlier.

Fixes: df79cf441910 ("drm/i915: Store the final plane stride in plane_state")
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20190930183045.662-1-ville.syrjala@linux.intel.com
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
(cherry picked from commit 006e570128f413759b9df64b51bae79903679c9b)
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
5 years agodrm/i915/dp: Fix dsc bpp calculations, v5.
Maarten Lankhorst [Wed, 25 Sep 2019 08:21:09 +0000 (10:21 +0200)]
drm/i915/dp: Fix dsc bpp calculations, v5.

There was a integer wraparound when mode_clock became too high,
and we didn't correct for the FEC overhead factor when dividing,
with the calculations breaking at HBR3.

As a result our calculated bpp was way too high, and the link width
limitation never came into effect.

Print out the resulting bpp calcululations as a sanity check, just
in case we ever have to debug it later on again.

We also used the wrong factor for FEC. While bspec mentions 2.4%,
all the calculations use 1/0.972261, and the same ratio should be
applied to data M/N as well, so use it there when FEC is enabled.

This fixes the FIFO underrun we are seeing with FEC enabled.

Changes since v2:
- Handle fec_enable in intel_link_compute_m_n, so only data M/N is adjusted. (Ville)
- Fix initial hardware readout for FEC. (Ville)
Changes since v3:
- Remove bogus fec_to_mode_clock. (Ville)
Changes since v4:
- Use the correct register for icl. (Ville)
- Split hw readout to a separate patch.

Signed-off-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Fixes: d9218c8f6cf4 ("drm/i915/dp: Add helpers for Compressed BPP and Slice Count for DSC")
Cc: <stable@vger.kernel.org> # v5.0+
Cc: Manasi Navare <manasi.d.navare@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20190925082110.17439-1-maarten.lankhorst@linux.intel.com
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
(cherry picked from commit ed06efb801bd291e935238d3fba46fa03d098f0e)
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
5 years agoblock: pg: add header include guard
Masahiro Yamada [Sat, 20 Jul 2019 12:05:25 +0000 (21:05 +0900)]
block: pg: add header include guard

Add a header include guard just in case.

Signed-off-by: Masahiro Yamada <yamada.masahiro@socionext.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
5 years agotimer-of: don't use conditional expression with mixed 'void' types
Linus Torvalds [Wed, 2 Oct 2019 23:16:07 +0000 (16:16 -0700)]
timer-of: don't use conditional expression with mixed 'void' types

Randy Dunlap reports on the sparse list that sparse warns about this
expression:

        of_irq->percpu ? free_percpu_irq(of_irq->irq, clkevt) :
                free_irq(of_irq->irq, clkevt);

and honestly, sparse is correct to warn.  The return type of
free_percpu_irq() is 'void', while free_irq() returns a 'const void *'
that is the devname argument passed in to the request_irq().

You can't mix a void type with a non-void types in a conditional
expression according to the C standard.  It so happens that gcc seems to
accept it - and the resulting type of the expression is void - but
there's really no reason for the kernel to have this kind of
non-standard expression with no real upside.

The natural way to write that expression is with an if-statement:

        if (of_irq->percpu)
                free_percpu_irq(of_irq->irq, clkevt);
        else
                free_irq(of_irq->irq, clkevt);

which is more legible anyway.

I'm not sure why that timer-of code seems to have this odd pattern.  It
does the same at allocation time, but at least there the types match,
and it makes sense as an expression.

Reported-by: Randy Dunlap <rdunlap@infradead.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
5 years agoMerge branch 'timers-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Wed, 2 Oct 2019 22:54:19 +0000 (15:54 -0700)]
Merge branch 'timers-urgent-for-linus' of git://git./linux/kernel/git/tip/tip

Pull timer fix from Ingo Molnar:
 "Fix a broadcast-timer handling race that can result in spuriously and
  indefinitely delayed hrtimers and even RCU stalls if the system is
  otherwise quiet"

* 'timers-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  tick: broadcast-hrtimer: Fix a race in bc_set_next

5 years agoMerge branch 'sched-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Wed, 2 Oct 2019 22:52:56 +0000 (15:52 -0700)]
Merge branch 'sched-urgent-for-linus' of git://git./linux/kernel/git/tip/tip

Pull membarrier fix from Ingo Molnar:
 "Fix broken locking within membarrier_private_expedited()"

* 'sched-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  membarrier: Fix RCU locking bug caused by faulty merge

5 years agomips: Loongson: Fix the link time qualifier of 'serial_exit()'
Christophe JAILLET [Tue, 10 Sep 2019 03:59:07 +0000 (05:59 +0200)]
mips: Loongson: Fix the link time qualifier of 'serial_exit()'

'exit' functions should be marked as __exit, not __init.

Fixes: 85cc028817ef ("mips: make loongsoon serial driver explicitly modular")
Signed-off-by: Christophe JAILLET <christophe.jaillet@wanadoo.fr>
Signed-off-by: Paul Burton <paul.burton@mips.com>
Cc: chenhc@lemote.com
Cc: ralf@linux-mips.org
Cc: jhogan@kernel.org
Cc: linux-mips@vger.kernel.org
Cc: linux-kernel@vger.kernel.org
Cc: kernel-janitors@vger.kernel.org
5 years agoMIPS: init: Prevent adding memory before PHYS_OFFSET
Thomas Bogendoerfer [Tue, 24 Sep 2019 15:20:51 +0000 (17:20 +0200)]
MIPS: init: Prevent adding memory before PHYS_OFFSET

On some SGI machines (IP28 and IP30) a small region of memory is mirrored
to pyhsical address 0 for exception vectors while rest of the memory
is reachable at a higher physical address. ARC PROM marks this
region as reserved, but with commit a94e4f24ec83 ("MIPS: init: Drop
boot_mem_map") this chunk is used, when searching for start of ram,
which breaks at least IP28 and IP30 machines. To fix this
add_region_memory() checks for start address < PHYS_OFFSET and ignores
these chunks.

Fixes: a94e4f24ec83 ("MIPS: init: Drop boot_mem_map")
Signed-off-by: Thomas Bogendoerfer <tbogendoerfer@suse.de>
Signed-off-by: Paul Burton <paul.burton@mips.com>
Cc: Ralf Baechle <ralf@linux-mips.org>
Cc: James Hogan <jhogan@kernel.org>
Cc: linux-mips@vger.kernel.org
Cc: linux-kernel@vger.kernel.org
5 years agoMIPS: init: Fix reservation of memory between PHYS_OFFSET and mem start
Thomas Bogendoerfer [Tue, 24 Sep 2019 15:19:56 +0000 (17:19 +0200)]
MIPS: init: Fix reservation of memory between PHYS_OFFSET and mem start

Fix calculation of the size for reserving memory between PHYS_OFFSET
and real memory start.

Fixes: a94e4f24ec83 ("MIPS: init: Drop boot_mem_map")
Signed-off-by: Thomas Bogendoerfer <tbogendoerfer@suse.de>
Signed-off-by: Paul Burton <paul.burton@mips.com>
Cc: Ralf Baechle <ralf@linux-mips.org>
Cc: James Hogan <jhogan@kernel.org>
Cc: linux-mips@vger.kernel.org
Cc: linux-kernel@vger.kernel.org
5 years agoMIPS: VDSO: Fix build for binutils < 2.25
Paul Burton [Wed, 2 Oct 2019 17:46:36 +0000 (17:46 +0000)]
MIPS: VDSO: Fix build for binutils < 2.25

Versions of binutils prior to 2.25 are unable to link our VDSO due to an
unsupported R_MIPS_PC32 relocation generated by the ".word _start - ."
line of the inline asm in get_vdso_base(). As such, the intent is that
when building with binutils older than 2.25 we don't build code for
gettimeofday() & friends in the VDSO that rely upon get_vdso_base().

Commit 24640f233b46 ("mips: Add support for generic vDSO") converted us
to using generic VDSO infrastructure, and as part of that the
gettimeofday() functionality moved to a new vgettimeofday.c file. The
check for binutils < 2.25 wasn't updated to handle this new filename,
and so it continues trying to remove the old unused filename from the
build. The end result is that we try to include the gettimeofday() code
in builds that will fail to link.

Fix this by updating the binutils < 2.25 case to remove vgettimeofday.c
from obj-vdso-y, rather than gettimeofday.c.

Signed-off-by: Paul Burton <paul.burton@mips.com>
Fixes: 24640f233b46 ("mips: Add support for generic vDSO")
Cc: Vincenzo Frascino <vincenzo.frascino@arm.com>
Cc: linux-mips@vger.kernel.org
5 years agoMIPS: VDSO: Remove unused gettimeofday.c
Paul Burton [Wed, 2 Oct 2019 17:46:17 +0000 (17:46 +0000)]
MIPS: VDSO: Remove unused gettimeofday.c

arch/mips/vdso/gettimeofday.c has been unused since commit 24640f233b46
("mips: Add support for generic vDSO"). Remove the dead code.

Signed-off-by: Paul Burton <paul.burton@mips.com>
Fixes: 24640f233b46 ("mips: Add support for generic vDSO")
Cc: Vincenzo Frascino <vincenzo.frascino@arm.com>
Cc: linux-mips@vger.kernel.org
5 years agoMIPS: Wire up clone3 syscall
Paul Burton [Wed, 2 Oct 2019 18:59:49 +0000 (18:59 +0000)]
MIPS: Wire up clone3 syscall

Wire up the new clone3 syscall for MIPS, using save_static_function() to
generate a wrapper that saves registers $s0-$s7 prior to invoking the
generic sys_clone3 function just like we do for plain old clone.

Tested atop 64r6el_defconfig using o32, n32 & n64 builds of the simple
test program from:

  https://lore.kernel.org/lkml/20190716130631.tohj4ub54md25dys@brauner.io/

Signed-off-by: Paul Burton <paul.burton@mips.com>
Cc: Christian Brauner <christian@brauner.io>
Acked-by: Christian Brauner <christian.brauner@ubuntu.com>
Cc: linux-mips@vger.kernel.org
5 years agoMerge tag 'erofs-for-5.4-rc2-fixes' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Wed, 2 Oct 2019 20:53:48 +0000 (13:53 -0700)]
Merge tag 'erofs-for-5.4-rc2-fixes' of git://git./linux/kernel/git/xiang/erofs

Pull erofs fixes from Gao Xiang:
 "Three patches to address regressions due to recent cleanups, mainly
  found by stress test on latest mainline kernel (no more regression out
  compared with older kernels for more than a week)

  One additional patch updates sub-entries in MAINTAINERS.

  Summary:

   - Fix error handling in erofs_read_superblock

   - Fix locking in erofs_get_meta_page

   - Fix inplace behavior due to decompression frontend cleanup

   - Update sub-entries in MAINTAINERS in order to better blame"

* tag 'erofs-for-5.4-rc2-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/xiang/erofs:
  erofs: fix mis-inplace determination related with noio chain
  erofs: fix erofs_get_meta_page locking due to a cleanup
  MAINTAINERS: erofs: complete sub-entries for erofs
  erofs: fix return value check in erofs_read_superblock()

5 years agochar/random: Add a newline at the end of the file
Borislav Petkov [Tue, 1 Oct 2019 17:50:23 +0000 (19:50 +0200)]
char/random: Add a newline at the end of the file

On Tue, Oct 01, 2019 at 10:14:40AM -0700, Linus Torvalds wrote:
> The previous state of the file didn't have that 0xa at the end, so you get that
>
>
>   -EXPORT_SYMBOL_GPL(add_bootloader_randomness);
>   \ No newline at end of file
>   +EXPORT_SYMBOL_GPL(add_bootloader_randomness);
>
> which is "the '-' line doesn't have a newline, the '+' line does" marker.

Aaha, that makes total sense, thanks for explaining. Oh well, let's fix
it then so that people don't scratch heads like me.

Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
5 years agoxen/xenbus: fix self-deadlock after killing user process
Juergen Gross [Tue, 1 Oct 2019 15:03:55 +0000 (17:03 +0200)]
xen/xenbus: fix self-deadlock after killing user process

In case a user process using xenbus has open transactions and is killed
e.g. via ctrl-C the following cleanup of the allocated resources might
result in a deadlock due to trying to end a transaction in the xenbus
worker thread:

[ 2551.474706] INFO: task xenbus:37 blocked for more than 120 seconds.
[ 2551.492215]       Tainted: P           OE     5.0.0-29-generic #5
[ 2551.510263] "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message.
[ 2551.528585] xenbus          D    0    37      2 0x80000080
[ 2551.528590] Call Trace:
[ 2551.528603]  __schedule+0x2c0/0x870
[ 2551.528606]  ? _cond_resched+0x19/0x40
[ 2551.528632]  schedule+0x2c/0x70
[ 2551.528637]  xs_talkv+0x1ec/0x2b0
[ 2551.528642]  ? wait_woken+0x80/0x80
[ 2551.528645]  xs_single+0x53/0x80
[ 2551.528648]  xenbus_transaction_end+0x3b/0x70
[ 2551.528651]  xenbus_file_free+0x5a/0x160
[ 2551.528654]  xenbus_dev_queue_reply+0xc4/0x220
[ 2551.528657]  xenbus_thread+0x7de/0x880
[ 2551.528660]  ? wait_woken+0x80/0x80
[ 2551.528665]  kthread+0x121/0x140
[ 2551.528667]  ? xb_read+0x1d0/0x1d0
[ 2551.528670]  ? kthread_park+0x90/0x90
[ 2551.528673]  ret_from_fork+0x35/0x40

Fix this by doing the cleanup via a workqueue instead.

Reported-by: James Dingwall <james@dingwall.me.uk>
Fixes: fd8aa9095a95c ("xen: optimize xenbus driver for multiple concurrent xenstore accesses")
Cc: <stable@vger.kernel.org> # 4.11
Signed-off-by: Juergen Gross <jgross@suse.com>
Reviewed-by: Boris Ostrovsky <boris.ostrovsky@oracle.com>
Signed-off-by: Boris Ostrovsky <boris.ostrovsky@oracle.com>
5 years agodt-bindings: phy: lantiq: Fix Property Name
Maxime Ripard [Wed, 2 Oct 2019 14:46:40 +0000 (16:46 +0200)]
dt-bindings: phy: lantiq: Fix Property Name

The binding has a typo where resets-names should read reset-names, which in
turn leads to a warning when the example is validated, since reset-names is
being used, and the binding prevent the usage of any property that isn't
described.

Fixes: 088e88be5a38 ("dt-bindings: phy: add binding for the Lantiq VRX200 and ARX300 PCIe PHYs")
Signed-off-by: Maxime Ripard <mripard@kernel.org>
Signed-off-by: Rob Herring <robh@kernel.org>
5 years agodt-bindings: iio: ad7192: Fix DTC warning in the example
Maxime Ripard [Wed, 2 Oct 2019 14:45:42 +0000 (16:45 +0200)]
dt-bindings: iio: ad7192: Fix DTC warning in the example

The example contains an SPI bus and device, but doesn't have the
appropriate size and address cells size.

This creates a DTC warning when the example is compiled since the default
ones will not match what the device uses. Let's add them to remove that
warning.

Fixes: f7356e47032c ("dt-bindings: iio: adc: ad7192: Add binding documentation for AD7192")
Signed-off-by: Maxime Ripard <mripard@kernel.org>
Signed-off-by: Rob Herring <robh@kernel.org>
5 years agodt-bindings: iio: ad7192: Fix Regulator Properties
Maxime Ripard [Wed, 2 Oct 2019 14:45:41 +0000 (16:45 +0200)]
dt-bindings: iio: ad7192: Fix Regulator Properties

The AD7192 binding describes two regulator properties, avdd-supply and
dvdd-supply, but describes it as a constant string that must be avdd and
dvdd. This is wrong since a *-supply property is actually a phandle, and
results in warnings when the example is validated (or any device tree using
that device, for that matter).

Let's remove that requirement.

Fixes: f7356e47032c ("dt-bindings: iio: adc: ad7192: Add binding documentation for AD7192")
Signed-off-by: Maxime Ripard <mripard@kernel.org>
Signed-off-by: Rob Herring <robh@kernel.org>
5 years agodt-bindings: media: rc: Fix redundant string
Maxime Ripard [Wed, 2 Oct 2019 14:44:00 +0000 (16:44 +0200)]
dt-bindings: media: rc: Fix redundant string

The linux,rc-map-name property is described using an enum, yet a value has
been put in that enum twice, resulting in a warning. Let's fix that.

Fixes: 7c31b9d67342 ("media: dt-bindings: media: Add YAML schemas for the generic RC bindings")
Signed-off-by: Maxime Ripard <mripard@kernel.org>
Signed-off-by: Rob Herring <robh@kernel.org>
5 years agodt-bindings: dsp: Fix fsl,dsp example
Maxime Ripard [Wed, 2 Oct 2019 14:42:04 +0000 (16:42 +0200)]
dt-bindings: dsp: Fix fsl,dsp example

The fsl,dsp binding requires a memory-region, yet its example doesn't have
one which results in a warning. Let's add a memory-region phandle to the
example.

Fixes: 7db2f2dfc701 ("dt-bindings: dsp: fsl: Add DSP core binding support")
Signed-off-by: Maxime Ripard <mripard@kernel.org>
Reviewed-by: Daniel Baluta <daniel.baluta@nxp.com>
Signed-off-by: Rob Herring <robh@kernel.org>
5 years agodrm/amd/display: fix dcn21 Makefile for clang
Arnd Bergmann [Wed, 2 Oct 2019 12:01:25 +0000 (14:01 +0200)]
drm/amd/display: fix dcn21 Makefile for clang

Just like all the other variants, this one passes invalid
compile-time options with clang after the new code got
merged:

clang: error: unknown argument: '-mpreferred-stack-boundary=4'
scripts/Makefile.build:265: recipe for target 'drivers/gpu/drm/amd/amdgpu/../display/dc/dcn21/dcn21_resource.o' failed

Use the same variant that we have for dcn20 to fix compilation.

Fixes: eced51f9babb ("drm/amd/display: Add hubp block for Renoir (v2)")
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
5 years agodrm/amd/display: hide an unused variable
Arnd Bergmann [Wed, 18 Sep 2019 19:53:58 +0000 (21:53 +0200)]
drm/amd/display: hide an unused variable

Without CONFIG_DEBUG_FS, we get a warning for an unused
variable:

drivers/gpu/drm/amd/amdgpu/../display/amdgpu_dm/amdgpu_dm.c:6020:33: error: unused variable 'source' [-Werror,-Wunused-variable]

Hide the variable in an #ifdef like its only users.

Fixes: 14b2584636c6 ("drm/amd/display: add functionality to grab DPRX CRC entries.")
Reviewed-by: Harry Wentland <harry.wentland@amd.com>
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
5 years agodrm/amdgpu: display_mode_vba_21: remove uint typedef
Arnd Bergmann [Wed, 2 Oct 2019 12:01:24 +0000 (14:01 +0200)]
drm/amdgpu: display_mode_vba_21: remove uint typedef

The type definition for 'uint' clashes with the generic kernel
headers:

drivers/gpu/drm/amd/amdgpu/../display/dc/dml/dcn21/display_mode_vba_21.c:43:22: error: redefinition of typedef 'uint' is a C11 feature [-Werror,-Wtypedef-redefinition]
include/linux/types.h:92:23: note: previous definition is here

Just remove this type and use plain 'unsigned int' consistently,
as it is already use almost everywhere in this file.

Fixes: b04641a3f4c5 ("drm/amd/display: Add Renoir DML")
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
5 years agodrm/amdgpu: hide another #warning
Arnd Bergmann [Wed, 2 Oct 2019 12:01:23 +0000 (14:01 +0200)]
drm/amdgpu: hide another #warning

An earlier patch of mine disabled some #warning statements
that get in the way of build testing, but then another
instance was added around the same time.

Remove that as well.

Fixes: b5203d16aef4 ("drm/amd/amdgpu: hide #warning for missing DC config")
Fixes: e1c14c43395c ("drm/amdgpu: Enable DC on Renoir")
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
5 years agodrm/amdgpu: make pmu support optional, again
Arnd Bergmann [Wed, 2 Oct 2019 12:01:22 +0000 (14:01 +0200)]
drm/amdgpu: make pmu support optional, again

When CONFIG_PERF_EVENTS is disabled, we cannot compile the pmu
portion of the amdgpu driver:

drivers/gpu/drm/amd/amdgpu/amdgpu_pmu.c:48:38: error: no member named 'hw' in 'struct perf_event'
        struct hw_perf_event *hwc = &event->hw;
                                     ~~~~~  ^
drivers/gpu/drm/amd/amdgpu/amdgpu_pmu.c:51:13: error: no member named 'attr' in 'struct perf_event'
        if (event->attr.type != event->pmu->type)
            ~~~~~  ^
...

The same bug was already fixed by commit d155bef0636e ("amdgpu: make pmu
support optional") but broken again by what looks like an incorrectly
rebased patch.

Fixes: 64f55e629237 ("drm/amdgpu: Add RAS EEPROM table.")
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
5 years agodrm/amd/display: memory leak
Navid Emamdoost [Tue, 17 Sep 2019 03:20:44 +0000 (22:20 -0500)]
drm/amd/display: memory leak

In dcn*_clock_source_create when dcn20_clk_src_construct fails allocated
clk_src needs release.

Signed-off-by: Navid Emamdoost <navid.emamdoost@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
5 years agodrm/amdgpu: fix multiple memory leaks in acp_hw_init
Navid Emamdoost [Wed, 2 Oct 2019 03:46:07 +0000 (22:46 -0500)]
drm/amdgpu: fix multiple memory leaks in acp_hw_init

In acp_hw_init there are some allocations that needs to be released in
case of failure:

1- adev->acp.acp_genpd should be released if any allocation attemp for
adev->acp.acp_cell, adev->acp.acp_res or i2s_pdata fails.
2- all of those allocations should be released if
mfd_add_hotplug_devices or pm_genpd_add_device fail.
3- Release is needed in case of time out values expire.

Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Navid Emamdoost <navid.emamdoost@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
5 years agodrm/amdgpu: return tcc_disabled_mask to userspace
Marek Olšák [Tue, 24 Sep 2019 21:53:25 +0000 (17:53 -0400)]
drm/amdgpu: return tcc_disabled_mask to userspace

UMDs need this for correct programming of harvested chips.

Signed-off-by: Marek Olšák <marek.olsak@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
5 years agodrm/amdgpu: don't increment vram lost if we are in hibernation
Alex Deucher [Tue, 1 Oct 2019 21:45:27 +0000 (16:45 -0500)]
drm/amdgpu: don't increment vram lost if we are in hibernation

We reset the GPU as part of our hibernation sequence so we need
to make sure we don't mark vram as lost in that case.

Bug: https://bugs.freedesktop.org/show_bug.cgi?id=111879
Acked-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Huang Rui <ray.huang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
5 years agoRevert "drm/amdgpu: disable stutter mode for renoir"
Aaron Liu [Mon, 16 Sep 2019 02:05:09 +0000 (10:05 +0800)]
Revert "drm/amdgpu: disable stutter mode for renoir"

This reverts commit 5813f97a5969bf1e7e723397a74e00b5de7278d6.

Since SBIOS WCD9925N, NMI printing disappeared. Hence enable stutter
mode.

Signed-off-by: Aaron Liu <aaron.liu@amd.com>
Reviewed-by: Huang Rui <ray.huang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
5 years agodrm/amd/powerplay: add sensor lock support for smu
Kevin Wang [Thu, 26 Sep 2019 08:22:13 +0000 (16:22 +0800)]
drm/amd/powerplay: add sensor lock support for smu

when multithreading access sysfs of amdgpu_pm_info at the sametime.
the swsmu driver cause smu firmware hang.

eg:
single thread access:
Message A + Param A ==> right
Message B + Param B ==> right
Message C + Param C ==> right
multithreading access:
Message A + Param B ==> error
Message B + Param A ==> error
Message C + Param C ==> right

the patch will add sensor lock(mutex) to avoid this error.

Signed-off-by: Kevin Wang <kevin1.wang@amd.com>
Reviewed-by: Kenneth Feng <kenneth.feng@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Cc: stable@vger.kernel.org # 5.3.x
5 years agodrm/amd/powerplay: change metrics update period from 1ms to 100ms
Kevin Wang [Thu, 26 Sep 2019 08:16:41 +0000 (16:16 +0800)]
drm/amd/powerplay: change metrics update period from 1ms to 100ms

v2:
change period from 10ms to 100ms (typo error)

too high frequence to update mertrics table will cause smu firmware
error,so change mertrics table update period from 1ms to 100ms
(navi10, 12, 14)

Signed-off-by: Kevin Wang <kevin1.wang@amd.com>
Reviewed-by: Kenneth Feng <kenneth.feng@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Cc: stable@vger.kernel.org # 5.3.x
5 years agodrm/amdgpu: revert "disable bulk moves for now"
Christian König [Thu, 12 Sep 2019 10:13:50 +0000 (12:13 +0200)]
drm/amdgpu: revert "disable bulk moves for now"

This reverts commit a213c2c7e235cfc0e0a161a558f7fdf2fb3a624a.

The changes to fix this should have landed in 5.1.

Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Zhou, David(ChunMing) <David1.Zhou@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
5 years agoMIPS: octeon: Include required header; fix octeon ethernet build
Paul Burton [Tue, 1 Oct 2019 21:56:37 +0000 (21:56 +0000)]
MIPS: octeon: Include required header; fix octeon ethernet build

Commit 171a9bae68c7 ("staging/octeon: Allow test build on !MIPS") moved
the inclusion of a bunch of headers by various files in the Octeon
ethernet driver into a common header, but in doing so it changed the
order in which those headers are included.

Prior to the referenced commit drivers/staging/octeon/ethernet.c
included asm/octeon/cvmx-pip.h before asm/octeon/cvmx-ipd.h, which makes
use of the CVMX_PIP_SFT_RST definition pulled in by the former. After
commit 171a9bae68c7 ("staging/octeon: Allow test build on !MIPS") we
pull in asm/octeon/cvmx-ipd.h first & builds fail with:

  In file included from drivers/staging/octeon/octeon-ethernet.h:27,
                   from drivers/staging/octeon/ethernet.c:22:
  arch/mips/include/asm/octeon/cvmx-ipd.h: In function 'cvmx_ipd_free_ptr':
  arch/mips/include/asm/octeon/cvmx-ipd.h:330:27: error: storage size of
    'pip_sft_rst' isn't known
      union cvmx_pip_sft_rst pip_sft_rst;
                             ^~~~~~~~~~~
  arch/mips/include/asm/octeon/cvmx-ipd.h:331:36: error: 'CVMX_PIP_SFT_RST'
    undeclared (first use in this function); did you mean 'CVMX_CIU_SOFT_RST'?
      pip_sft_rst.u64 = cvmx_read_csr(CVMX_PIP_SFT_RST);
                                      ^~~~~~~~~~~~~~~~
                                      CVMX_CIU_SOFT_RST
  arch/mips/include/asm/octeon/cvmx-ipd.h:331:36: note: each undeclared
    identifier is reported only once for each function it appears in
  arch/mips/include/asm/octeon/cvmx-ipd.h:330:27: warning: unused variable
    'pip_sft_rst' [-Wunused-variable]
      union cvmx_pip_sft_rst pip_sft_rst;
                             ^~~~~~~~~~~
  make[4]: *** [scripts/Makefile.build:266: drivers/staging/octeon/ethernet.o]
    Error 1
  make[3]: *** [scripts/Makefile.build:509: drivers/staging/octeon] Error 2

Fix this by having asm/octeon/cvmx-ipd.h include the
asm/octeon/cvmx-pip-defs.h header that it is reliant upon, rather than
requiring its users to pull in that header before it.

Signed-off-by: Paul Burton <paul.burton@mips.com>
Fixes: 171a9bae68c7 ("staging/octeon: Allow test build on !MIPS")
Cc: David S. Miller <davem@davemloft.net>
Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Cc: Matthew Wilcox (Oracle) <willy@infradead.org>
Cc: linux-mips@vger.kernel.org
Cc: David S . Miller <davem@davemloft.net>
Cc: Matthew Wilcox <willy@infradead.org>
5 years agoxen/efi: have a common runtime setup function
Juergen Gross [Tue, 1 Oct 2019 08:25:34 +0000 (10:25 +0200)]
xen/efi: have a common runtime setup function

Today the EFI runtime functions are setup in architecture specific
code (x86 and arm), with the functions themselves living in drivers/xen
as they are not architecture dependent.

As the setup is exactly the same for arm and x86 move the setup to
drivers/xen, too. This at once removes the need to make the single
functions global visible.

Signed-off-by: Juergen Gross <jgross@suse.com>
Reviewed-by: Jan Beulich <jbeulich@suse.com>
[boris: "Dropped EXPORT_SYMBOL_GPL(xen_efi_runtime_setup)"]
Signed-off-by: Boris Ostrovsky <boris.ostrovsky@oracle.com>
5 years agoarm: xen: mm: use __GPF_DMA32 for arm64
Peng Fan [Mon, 16 Sep 2019 09:51:33 +0000 (09:51 +0000)]
arm: xen: mm: use __GPF_DMA32 for arm64

arm64 shares some code under arch/arm/xen, including mm.c.
However ZONE_DMA is removed by commit
ad67f5a6545("arm64: replace ZONE_DMA with ZONE_DMA32").

So add a check if CONFIG_ZONE_DMA32 is enabled use __GFP_DMA32.

Signed-off-by: Peng Fan <peng.fan@nxp.com>
Acked-by: Stefano Stabellini <sstabellini@kernel.org>
Signed-off-by: Stefano Stabellini <sstabellini@kernel.org>
5 years agoriscv: Fix memblock reservation for device tree blob
Albert Ou [Fri, 27 Sep 2019 23:14:18 +0000 (16:14 -0700)]
riscv: Fix memblock reservation for device tree blob

This fixes an error with how the FDT blob is reserved in memblock.
An incorrect physical address calculation exposed the FDT header to
unintended corruption, which typically manifested with of_fdt_raw_init()
faulting during late boot after fdt_totalsize() returned a wrong value.
Systems with smaller physical memory sizes more frequently trigger this
issue, as the kernel is more likely to allocate from the DMA32 zone
where bbl places the DTB after the kernel image.

Commit 671f9a3e2e24 ("RISC-V: Setup initial page tables in two stages")
changed the mapping of the DTB to reside in the fixmap area.
Consequently, early_init_fdt_reserve_self() cannot be used anymore in
setup_bootmem() since it relies on __pa() to derive a physical address,
which does not work with dtb_early_va that is no longer a valid kernel
logical address.

The reserved[0x1] region shows the effect of the pointer underflow
resulting from the __pa(initial_boot_params) offset subtraction:

[    0.000000] MEMBLOCK configuration:
[    0.000000]  memory size = 0x000000001fe00000 reserved size = 0x0000000000a2e514
[    0.000000]  memory.cnt  = 0x1
[    0.000000]  memory[0x0]     [0x0000000080200000-0x000000009fffffff], 0x000000001fe00000 bytes flags: 0x0
[    0.000000]  reserved.cnt  = 0x2
[    0.000000]  reserved[0x0]   [0x0000000080200000-0x0000000080c2dfeb], 0x0000000000a2dfec bytes flags: 0x0
[    0.000000]  reserved[0x1]   [0xfffffff080100000-0xfffffff080100527], 0x0000000000000528 bytes flags: 0x0

With the fix applied:

[    0.000000] MEMBLOCK configuration:
[    0.000000]  memory size = 0x000000001fe00000 reserved size = 0x0000000000a2e514
[    0.000000]  memory.cnt  = 0x1
[    0.000000]  memory[0x0]     [0x0000000080200000-0x000000009fffffff], 0x000000001fe00000 bytes flags: 0x0
[    0.000000]  reserved.cnt  = 0x2
[    0.000000]  reserved[0x0]   [0x0000000080200000-0x0000000080c2dfeb], 0x0000000000a2dfec bytes flags: 0x0
[    0.000000]  reserved[0x1]   [0x0000000080e00000-0x0000000080e00527], 0x0000000000000528 bytes flags: 0x0

Fixes: 671f9a3e2e24 ("RISC-V: Setup initial page tables in two stages")
Signed-off-by: Albert Ou <aou@eecs.berkeley.edu>
Tested-by: Bin Meng <bmeng.cn@gmail.com>
Reviewed-by: Anup Patel <anup@brainfault.org>
Signed-off-by: Paul Walmsley <paul.walmsley@sifive.com>
5 years agoRISC-V: Clear load reservations while restoring hart contexts
Palmer Dabbelt [Wed, 25 Sep 2019 00:15:56 +0000 (17:15 -0700)]
RISC-V: Clear load reservations while restoring hart contexts

This is almost entirely a comment.  The bug is unlikely to manifest on
existing hardware because there is a timeout on load reservations, but
manifests on QEMU because there is no timeout.

Signed-off-by: Palmer Dabbelt <palmer@sifive.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Paul Walmsley <paul.walmsley@sifive.com>
5 years agoMIPS: cpu-bugs64: Mark inline functions as __always_inline
Jiaxun Yang [Fri, 27 Sep 2019 05:33:39 +0000 (13:33 +0800)]
MIPS: cpu-bugs64: Mark inline functions as __always_inline

Commit ac7c3e4ff401 ("compiler: enable CONFIG_OPTIMIZE_INLINING
forcibly") allows compiler to uninline functions marked as 'inline'.
Leading to section mismatch in this case.

Since we're using const variables to pass assembly flags, 'inline's
can't be dropped. So we simply mark them as __always_inline.

Signed-off-by: Jiaxun Yang <jiaxun.yang@flygoat.com>
Cc: linux-mips@vger.kernel.org
[paul.burton@mips.com:
  - Annotate these functions with __init, even if it only serves to
    inform human readers when the code can be used.
  - Drop the __always_inline from check_daddi() & check_daddiu() which
    don't use arguments as immediates in inline asm.
  - Rewrap the commit message.]
Signed-off-by: Paul Burton <paul.burton@mips.com>
5 years agoxen/balloon: Clear PG_offline in balloon_retrieve()
David Hildenbrand [Tue, 1 Oct 2019 09:01:52 +0000 (11:01 +0200)]
xen/balloon: Clear PG_offline in balloon_retrieve()

Let's move the clearing to balloon_retrieve(). In
bp_state increase_reservation(), we now clear the flag a little earlier
than before, however, this should not matter for XEN.

Suggested-by: Boris Ostrovsky <boris.ostrovsky@oracle.com>
Cc: Boris Ostrovsky <boris.ostrovsky@oracle.com>
Cc: Juergen Gross <jgross@suse.com>
Cc: Stefano Stabellini <sstabellini@kernel.org>
Signed-off-by: David Hildenbrand <david@redhat.com>
Reviewed-by: Boris Ostrovsky <boris.ostrovsky@oracle.com>
Signed-off-by: Boris Ostrovsky <boris.ostrovsky@oracle.com>
5 years agoxen/balloon: Mark pages PG_offline in balloon_append()
David Hildenbrand [Tue, 1 Oct 2019 09:01:51 +0000 (11:01 +0200)]
xen/balloon: Mark pages PG_offline in balloon_append()

Let's move the __SetPageOffline() call which all callers perform into
balloon_append().

In bp_state decrease_reservation(), pages are now marked PG_offline a
little later than before, however, this should not matter for XEN.

Suggested-by: Boris Ostrovsky <boris.ostrovsky@oracle.com>
Cc: Boris Ostrovsky <boris.ostrovsky@oracle.com>
Cc: Juergen Gross <jgross@suse.com>
Cc: Stefano Stabellini <sstabellini@kernel.org>
Signed-off-by: David Hildenbrand <david@redhat.com>
Reviewed-by: Boris Ostrovsky <boris.ostrovsky@oracle.com>
Signed-off-by: Boris Ostrovsky <boris.ostrovsky@oracle.com>
5 years agoxen/balloon: Drop __balloon_append()
David Hildenbrand [Tue, 1 Oct 2019 09:01:50 +0000 (11:01 +0200)]
xen/balloon: Drop __balloon_append()

Let's simply use balloon_append() directly.

Cc: Boris Ostrovsky <boris.ostrovsky@oracle.com>
Cc: Juergen Gross <jgross@suse.com>
Cc: Stefano Stabellini <sstabellini@kernel.org>
Signed-off-by: David Hildenbrand <david@redhat.com>
Reviewed-by: Boris Ostrovsky <boris.ostrovsky@oracle.com>
Signed-off-by: Boris Ostrovsky <boris.ostrovsky@oracle.com>
5 years agoxen/balloon: Set pages PageOffline() in balloon_add_region()
David Hildenbrand [Fri, 27 Sep 2019 15:46:28 +0000 (17:46 +0200)]
xen/balloon: Set pages PageOffline() in balloon_add_region()

We are missing a __SetPageOffline(), which is why we can get
!PageOffline() pages onto the balloon list, where
alloc_xenballooned_pages() will complain:

page:ffffea0003e7ffc0 refcount:1 mapcount:0 mapping:0000000000000000 index:0x0
flags: 0xffffe00001000(reserved)
raw: 000ffffe00001000 dead000000000100 dead000000000200 0000000000000000
raw: 0000000000000000 0000000000000000 00000001ffffffff 0000000000000000
page dumped because: VM_BUG_ON_PAGE(!PageOffline(page))
------------[ cut here ]------------
kernel BUG at include/linux/page-flags.h:744!
invalid opcode: 0000 [#1] SMP NOPTI

Reported-by: Marek Marczykowski-Górecki <marmarek@invisiblethingslab.com>
Tested-by: Marek Marczykowski-Górecki <marmarek@invisiblethingslab.com>
Fixes: 77c4adf6a6df ("xen/balloon: mark inflated pages PG_offline")
Cc: stable@vger.kernel.org # v5.1+
Cc: Boris Ostrovsky <boris.ostrovsky@oracle.com>
Cc: Juergen Gross <jgross@suse.com>
Cc: Stefano Stabellini <sstabellini@kernel.org>
Signed-off-by: David Hildenbrand <david@redhat.com>
Reviewed-by: Boris Ostrovsky <boris.ostrovsky@oracle.com>
Signed-off-by: Boris Ostrovsky <boris.ostrovsky@oracle.com>
5 years agoARM: xen: unexport HYPERVISOR_platform_op function
Stefano Stabellini [Tue, 1 Oct 2019 17:38:03 +0000 (10:38 -0700)]
ARM: xen: unexport HYPERVISOR_platform_op function

HYPERVISOR_platform_op() is an inline function and should not
be exported. Since commit 15bfc2348d54 ("modpost: check for
static EXPORT_SYMBOL* functions"), this causes a warning:

WARNING: "HYPERVISOR_platform_op" [vmlinux] is a static EXPORT_SYMBOL_GPL

Instead, export the underlying function called by the static inline:
HYPERVISOR_platform_op_raw.

Fixes: 15bfc2348d54 ("modpost: check for static EXPORT_SYMBOL* functions")
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Stefano Stabellini <stefano.stabellini@xilinx.com>
Signed-off-by: Stefano Stabellini <sstabellini@kernel.org>
5 years agomembarrier: Fix RCU locking bug caused by faulty merge
Peter Zijlstra [Tue, 1 Oct 2019 08:50:33 +0000 (10:50 +0200)]
membarrier: Fix RCU locking bug caused by faulty merge

The following commit:

  227a4aadc75b ("sched/membarrier: Fix p->mm->membarrier_state racy load")

got fat fingered by me when merging it with other patches. It meant to move
the RCU section out of the for loop but ended up doing it partially, leaving
a superfluous rcu_read_lock() inside, causing havok.

Reported-by: Ingo Molnar <mingo@kernel.org>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Chris Metcalf <cmetcalf@ezchip.com>
Cc: Christoph Lameter <cl@linux.com>
Cc: Eric W. Biederman <ebiederm@xmission.com>
Cc: Kirill Tkhai <tkhai@yandex.ru>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Oleg Nesterov <oleg@redhat.com>
Cc: Paul E. McKenney <paulmck@linux.ibm.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Russell King - ARM Linux admin <linux@armlinux.org.uk>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: linux-tip-commits@vger.kernel.org
Fixes: 227a4aadc75b ("sched/membarrier: Fix p->mm->membarrier_state racy load")
Link: https://lkml.kernel.org/r/20191001085033.GP4519@hirez.programming.kicks-ass.net
Signed-off-by: Ingo Molnar <mingo@kernel.org>
5 years agoRevert "s390/dasd: Add discard support for ESE volumes"
Stefan Haberland [Tue, 1 Oct 2019 15:34:39 +0000 (17:34 +0200)]
Revert "s390/dasd: Add discard support for ESE volumes"

This reverts commit 7e64db1597fe114b83fe17d0ba96c6aa5fca419a.

The thin provisioning feature introduces an IOCTL and the discard support
to allow userspace tools and filesystems to release unused and previously
allocated space respectively.

During some internal performance improvements and further tests, the
release of allocated space revealed some issues that may lead to data
corruption in some configurations when filesystems are mounted with
discard support enabled.

While we're working on a fix and trying to clarify the situation,
this commit reverts the discard support for ESE volumes to prevent
potential data corruption.

Cc: <stable@vger.kernel.org> # 5.3
Signed-off-by: Stefan Haberland <sth@linux.ibm.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
5 years agos390/dasd: Fix error handling during online processing
Jan Höppner [Tue, 1 Oct 2019 15:34:38 +0000 (17:34 +0200)]
s390/dasd: Fix error handling during online processing

It is possible that the CCW commands for reading volume and extent pool
information are not supported, either by the storage server (for
dedicated DASDs) or by z/VM (for virtual devices, such as MDISKs).

As a command reject will occur in such a case, the current error
handling leads to a failing online processing and thus the DASD can't be
used at all.

Since the data being read is not essential for an fully operational
DASD, the error handling can be removed. Information about the failing
command is sent to the s390dbf debug feature.

Fixes: c729696bcf8b ("s390/dasd: Recognise data for ESE volumes")
Cc: <stable@vger.kernel.org> # 5.3
Reported-by: Frank Heimes <frank.heimes@canonical.com>
Signed-off-by: Jan Höppner <hoeppner@linux.ibm.com>
Signed-off-by: Stefan Haberland <sth@linux.ibm.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
5 years agoio_uring: use __kernel_timespec in timeout ABI
Arnd Bergmann [Tue, 1 Oct 2019 15:53:29 +0000 (09:53 -0600)]
io_uring: use __kernel_timespec in timeout ABI

All system calls use struct __kernel_timespec instead of the old struct
timespec, but this one was just added with the old-style ABI. Change it
now to enforce the use of __kernel_timespec, avoiding ABI confusion and
the need for compat handlers on 32-bit architectures.

Any user space caller will have to use __kernel_timespec now, but this
is unambiguous and works for any C library regardless of the time_t
definition. A nicer way to specify the timeout would have been a less
ambiguous 64-bit nanosecond value, but I suppose it's too late now to
change that as this would impact both 32-bit and 64-bit users.

Fixes: 5262f567987d ("io_uring: IORING_OP_TIMEOUT support")
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
5 years agoloop: change queue block size to match when using DIO
Martijn Coenen [Wed, 4 Sep 2019 19:49:01 +0000 (21:49 +0200)]
loop: change queue block size to match when using DIO

The loop driver assumes that if the passed in fd is opened with
O_DIRECT, the caller wants to use direct I/O on the loop device.
However, if the underlying block device has a different block size than
the loop block queue, direct I/O can't be enabled. Instead of requiring
userspace to manually change the blocksize and re-enable direct I/O,
just change the queue block sizes to match, as well as the io_min size.

Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Martijn Coenen <maco@android.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
5 years agoperf_event_open: switch to copy_struct_from_user()
Aleksa Sarai [Tue, 1 Oct 2019 01:10:55 +0000 (11:10 +1000)]
perf_event_open: switch to copy_struct_from_user()

Switch perf_event_open() syscall from it's own copying
struct perf_event_attr from userspace to the new dedicated
copy_struct_from_user() helper.

The change is very straightforward, and helps unify the syscall
interface for struct-from-userspace syscalls.

Signed-off-by: Aleksa Sarai <cyphar@cyphar.com>
Reviewed-by: Kees Cook <keescook@chromium.org>
Reviewed-by: Christian Brauner <christian.brauner@ubuntu.com>
[christian.brauner@ubuntu.com: improve commit message]
Link: https://lore.kernel.org/r/20191001011055.19283-5-cyphar@cyphar.com
Signed-off-by: Christian Brauner <christian.brauner@ubuntu.com>
5 years agosched_setattr: switch to copy_struct_from_user()
Aleksa Sarai [Tue, 1 Oct 2019 01:10:54 +0000 (11:10 +1000)]
sched_setattr: switch to copy_struct_from_user()

Switch sched_setattr() syscall from it's own copying struct sched_attr
from userspace to the new dedicated copy_struct_from_user() helper.

The change is very straightforward, and helps unify the syscall
interface for struct-from-userspace syscalls. Ideally we could also
unify sched_getattr(2)-style syscalls as well, but unfortunately the
correct semantics for such syscalls are much less clear (see [1] for
more detail). In future we could come up with a more sane idea for how
the syscall interface should look.

[1]: commit 1251201c0d34 ("sched/core: Fix uclamp ABI bug, clean up and
     robustify sched_read_attr() ABI logic and code")

Signed-off-by: Aleksa Sarai <cyphar@cyphar.com>
Reviewed-by: Kees Cook <keescook@chromium.org>
Reviewed-by: Christian Brauner <christian.brauner@ubuntu.com>
[christian.brauner@ubuntu.com: improve commit message]
Link: https://lore.kernel.org/r/20191001011055.19283-4-cyphar@cyphar.com
Signed-off-by: Christian Brauner <christian.brauner@ubuntu.com>
5 years agoclone3: switch to copy_struct_from_user()
Aleksa Sarai [Tue, 1 Oct 2019 01:10:53 +0000 (11:10 +1000)]
clone3: switch to copy_struct_from_user()

Switch clone3() syscall from it's own copying struct clone_args from
userspace to the new dedicated copy_struct_from_user() helper.

The change is very straightforward, and helps unify the syscall
interface for struct-from-userspace syscalls. Additionally, explicitly
define CLONE_ARGS_SIZE_VER0 to match the other users of the
struct-extension pattern.

Signed-off-by: Aleksa Sarai <cyphar@cyphar.com>
Reviewed-by: Kees Cook <keescook@chromium.org>
Reviewed-by: Christian Brauner <christian.brauner@ubuntu.com>
[christian.brauner@ubuntu.com: improve commit message]
Link: https://lore.kernel.org/r/20191001011055.19283-3-cyphar@cyphar.com
Signed-off-by: Christian Brauner <christian.brauner@ubuntu.com>
5 years agolib: introduce copy_struct_from_user() helper
Aleksa Sarai [Tue, 1 Oct 2019 01:10:52 +0000 (11:10 +1000)]
lib: introduce copy_struct_from_user() helper

A common pattern for syscall extensions is increasing the size of a
struct passed from userspace, such that the zero-value of the new fields
result in the old kernel behaviour (allowing for a mix of userspace and
kernel vintages to operate on one another in most cases).

While this interface exists for communication in both directions, only
one interface is straightforward to have reasonable semantics for
(userspace passing a struct to the kernel). For kernel returns to
userspace, what the correct semantics are (whether there should be an
error if userspace is unaware of a new extension) is very
syscall-dependent and thus probably cannot be unified between syscalls
(a good example of this problem is [1]).

Previously there was no common lib/ function that implemented
the necessary extension-checking semantics (and different syscalls
implemented them slightly differently or incompletely[2]). Future
patches replace common uses of this pattern to make use of
copy_struct_from_user().

Some in-kernel selftests that insure that the handling of alignment and
various byte patterns are all handled identically to memchr_inv() usage.

[1]: commit 1251201c0d34 ("sched/core: Fix uclamp ABI bug, clean up and
     robustify sched_read_attr() ABI logic and code")

[2]: For instance {sched_setattr,perf_event_open,clone3}(2) all do do
     similar checks to copy_struct_from_user() while rt_sigprocmask(2)
     always rejects differently-sized struct arguments.

Suggested-by: Rasmus Villemoes <linux@rasmusvillemoes.dk>
Signed-off-by: Aleksa Sarai <cyphar@cyphar.com>
Reviewed-by: Kees Cook <keescook@chromium.org>
Reviewed-by: Christian Brauner <christian.brauner@ubuntu.com>
Link: https://lore.kernel.org/r/20191001011055.19283-2-cyphar@cyphar.com
Signed-off-by: Christian Brauner <christian.brauner@ubuntu.com>
5 years agokvm: vmx: Limit guest PMCs to those supported on the host
Jim Mattson [Mon, 30 Sep 2019 23:38:54 +0000 (16:38 -0700)]
kvm: vmx: Limit guest PMCs to those supported on the host

KVM can only virtualize as many PMCs as the host supports.

Limit the number of generic counters and fixed counters to the number
of corresponding counters supported on the host, rather than to
INTEL_PMC_MAX_GENERIC and INTEL_PMC_MAX_FIXED, respectively.

Note that INTEL_PMC_MAX_GENERIC is currently 32, which exceeds the 18
contiguous MSR indices reserved by Intel for event selectors. Since
the existing code relies on a contiguous range of MSR indices for
event selectors, it can't possibly work for more than 18 general
purpose counters.

Fixes: f5132b01386b5a ("KVM: Expose a version 2 architectural PMU to a guests")
Signed-off-by: Jim Mattson <jmattson@google.com>
Reviewed-by: Marc Orr <marcorr@google.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>