platform/kernel/linux-rpi.git
5 years agoMerge branch 'akpm'
Greg Kroah-Hartman [Fri, 5 Oct 2018 23:33:03 +0000 (16:33 -0700)]
Merge branch 'akpm'

* akpm:
  mm: madvise(MADV_DODUMP): allow hugetlbfs pages
  ocfs2: fix locking for res->tracking and dlm->tracking_list
  mm/vmscan.c: fix int overflow in callers of do_shrink_slab()
  mm/vmstat.c: skip NR_TLB_REMOTE_FLUSH* properly
  mm/vmstat.c: fix outdated vmstat_text
  proc: restrict kernel stack dumps to root
  mm/hugetlb: add mmap() encodings for 32MB and 512MB page sizes
  mm/migrate.c: split only transparent huge pages when allocation fails
  ipc/shm.c: use ERR_CAST() for shm_lock() error return
  mm/gup_benchmark: fix unsigned comparison to zero in __gup_benchmark_ioctl
  mm, thp: fix mlocking THP page with migration enabled
  ocfs2: fix crash in ocfs2_duplicate_clusters_by_page()
  hugetlb: take PMD sharing into account when flushing tlb/caches
  mm: migration: fix migration of huge PMD shared pages

5 years agomm: madvise(MADV_DODUMP): allow hugetlbfs pages
Daniel Black [Fri, 5 Oct 2018 22:52:19 +0000 (15:52 -0700)]
mm: madvise(MADV_DODUMP): allow hugetlbfs pages

Reproducer, assuming 2M of hugetlbfs available:

Hugetlbfs mounted, size=2M and option user=testuser

  # mount | grep ^hugetlbfs
  hugetlbfs on /dev/hugepages type hugetlbfs (rw,pagesize=2M,user=dan)
  # sysctl vm.nr_hugepages=1
  vm.nr_hugepages = 1
  # grep Huge /proc/meminfo
  AnonHugePages:         0 kB
  ShmemHugePages:        0 kB
  HugePages_Total:       1
  HugePages_Free:        1
  HugePages_Rsvd:        0
  HugePages_Surp:        0
  Hugepagesize:       2048 kB
  Hugetlb:            2048 kB

Code:

  #include <sys/mman.h>
  #include <stddef.h>
  #define SIZE 2*1024*1024
  int main()
  {
    void *ptr;
    ptr = mmap(NULL, SIZE, PROT_READ | PROT_WRITE, MAP_PRIVATE | MAP_HUGETLB | MAP_ANONYMOUS, -1, 0);
    madvise(ptr, SIZE, MADV_DONTDUMP);
    madvise(ptr, SIZE, MADV_DODUMP);
  }

Compile and strace:

  mmap(NULL, 2097152, PROT_READ|PROT_WRITE, MAP_PRIVATE|MAP_ANONYMOUS|MAP_HUGETLB, -1, 0) = 0x7ff7c9200000
  madvise(0x7ff7c9200000, 2097152, MADV_DONTDUMP) = 0
  madvise(0x7ff7c9200000, 2097152, MADV_DODUMP) = -1 EINVAL (Invalid argument)

hugetlbfs pages have VM_DONTEXPAND in the VmFlags driver pages based on
author testing with analysis from Florian Weimer[1].

The inclusion of VM_DONTEXPAND into the VM_SPECIAL defination was a
consequence of the large useage of VM_DONTEXPAND in device drivers.

A consequence of [2] is that VM_DONTEXPAND marked pages are unable to be
marked DODUMP.

A user could quite legitimately madvise(MADV_DONTDUMP) their hugetlbfs
memory for a while and later request that madvise(MADV_DODUMP) on the same
memory.  We correct this omission by allowing madvice(MADV_DODUMP) on
hugetlbfs pages.

[1] https://stackoverflow.com/questions/52548260/madvisedodump-on-the-same-ptr-size-as-a-successful-madvisedontdump-fails-wit
[2] commit 0103bd16fb90 ("mm: prepare VM_DONTDUMP for using in drivers")

Link: http://lkml.kernel.org/r/20180930054629.29150-1-daniel@linux.ibm.com
Link: https://lists.launchpad.net/maria-discuss/msg05245.html
Fixes: 0103bd16fb90 ("mm: prepare VM_DONTDUMP for using in drivers")
Reported-by: Kenneth Penza <kpenza@gmail.com>
Signed-off-by: Daniel Black <daniel@linux.ibm.com>
Reviewed-by: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Konstantin Khlebnikov <khlebnikov@openvz.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agoocfs2: fix locking for res->tracking and dlm->tracking_list
Ashish Samant [Fri, 5 Oct 2018 22:52:15 +0000 (15:52 -0700)]
ocfs2: fix locking for res->tracking and dlm->tracking_list

In dlm_init_lockres() we access and modify res->tracking and
dlm->tracking_list without holding dlm->track_lock.  This can cause list
corruptions and can end up in kernel panic.

Fix this by locking res->tracking and dlm->tracking_list with
dlm->track_lock instead of dlm->spinlock.

Link: http://lkml.kernel.org/r/1529951192-4686-1-git-send-email-ashish.samant@oracle.com
Signed-off-by: Ashish Samant <ashish.samant@oracle.com>
Reviewed-by: Changwei Ge <ge.changwei@h3c.com>
Acked-by: Joseph Qi <jiangqi903@gmail.com>
Acked-by: Jun Piao <piaojun@huawei.com>
Cc: Mark Fasheh <mark@fasheh.com>
Cc: Joel Becker <jlbec@evilplan.org>
Cc: Junxiao Bi <junxiao.bi@oracle.com>
Cc: Changwei Ge <ge.changwei@h3c.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agomm/vmscan.c: fix int overflow in callers of do_shrink_slab()
Kirill Tkhai [Fri, 5 Oct 2018 22:52:10 +0000 (15:52 -0700)]
mm/vmscan.c: fix int overflow in callers of do_shrink_slab()

do_shrink_slab() returns unsigned long value, and the placing into int
variable cuts high bytes off.  Then we compare ret and 0xfffffffe (since
SHRINK_EMPTY is converted to ret type).

Thus a large number of objects returned by do_shrink_slab() may be
interpreted as SHRINK_EMPTY, if low bytes of their value are equal to
0xfffffffe.  Fix that by declaration ret as unsigned long in these
functions.

Link: http://lkml.kernel.org/r/153813407177.17544.14888305435570723973.stgit@localhost.localdomain
Signed-off-by: Kirill Tkhai <ktkhai@virtuozzo.com>
Reported-by: Cyrill Gorcunov <gorcunov@openvz.org>
Acked-by: Cyrill Gorcunov <gorcunov@openvz.org>
Reviewed-by: Josef Bacik <josef@toxicpanda.com>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Andrey Ryabinin <aryabinin@virtuozzo.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp>
Cc: Shakeel Butt <shakeelb@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agomm/vmstat.c: skip NR_TLB_REMOTE_FLUSH* properly
Jann Horn [Fri, 5 Oct 2018 22:52:07 +0000 (15:52 -0700)]
mm/vmstat.c: skip NR_TLB_REMOTE_FLUSH* properly

5dd0b16cdaff ("mm/vmstat: Make NR_TLB_REMOTE_FLUSH_RECEIVED available even
on UP") made the availability of the NR_TLB_REMOTE_FLUSH* counters inside
the kernel unconditional to reduce #ifdef soup, but (either to avoid
showing dummy zero counters to userspace, or because that code was missed)
didn't update the vmstat_array, meaning that all following counters would
be shown with incorrect values.

This only affects kernel builds with
CONFIG_VM_EVENT_COUNTERS=y && CONFIG_DEBUG_TLBFLUSH=y && CONFIG_SMP=n.

Link: http://lkml.kernel.org/r/20181001143138.95119-2-jannh@google.com
Fixes: 5dd0b16cdaff ("mm/vmstat: Make NR_TLB_REMOTE_FLUSH_RECEIVED available even on UP")
Signed-off-by: Jann Horn <jannh@google.com>
Reviewed-by: Kees Cook <keescook@chromium.org>
Reviewed-by: Andrew Morton <akpm@linux-foundation.org>
Acked-by: Michal Hocko <mhocko@suse.com>
Acked-by: Roman Gushchin <guro@fb.com>
Cc: Davidlohr Bueso <dave@stgolabs.net>
Cc: Oleg Nesterov <oleg@redhat.com>
Cc: Christoph Lameter <clameter@sgi.com>
Cc: Kemi Wang <kemi.wang@intel.com>
Cc: Andy Lutomirski <luto@kernel.org>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agomm/vmstat.c: fix outdated vmstat_text
Jann Horn [Fri, 5 Oct 2018 22:52:03 +0000 (15:52 -0700)]
mm/vmstat.c: fix outdated vmstat_text

7a9cdebdcc17 ("mm: get rid of vmacache_flush_all() entirely") removed the
VMACACHE_FULL_FLUSHES statistics, but didn't remove the corresponding
entry in vmstat_text.  This causes an out-of-bounds access in
vmstat_show().

Luckily this only affects kernels with CONFIG_DEBUG_VM_VMACACHE=y, which
is probably very rare.

Link: http://lkml.kernel.org/r/20181001143138.95119-1-jannh@google.com
Fixes: 7a9cdebdcc17 ("mm: get rid of vmacache_flush_all() entirely")
Signed-off-by: Jann Horn <jannh@google.com>
Reviewed-by: Kees Cook <keescook@chromium.org>
Reviewed-by: Andrew Morton <akpm@linux-foundation.org>
Acked-by: Michal Hocko <mhocko@suse.com>
Acked-by: Roman Gushchin <guro@fb.com>
Cc: Davidlohr Bueso <dave@stgolabs.net>
Cc: Oleg Nesterov <oleg@redhat.com>
Cc: Christoph Lameter <clameter@sgi.com>
Cc: Kemi Wang <kemi.wang@intel.com>
Cc: Andy Lutomirski <luto@kernel.org>
Cc: Ingo Molnar <mingo@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agoproc: restrict kernel stack dumps to root
Jann Horn [Fri, 5 Oct 2018 22:51:58 +0000 (15:51 -0700)]
proc: restrict kernel stack dumps to root

Currently, you can use /proc/self/task/*/stack to cause a stack walk on
a task you control while it is running on another CPU.  That means that
the stack can change under the stack walker.  The stack walker does
have guards against going completely off the rails and into random
kernel memory, but it can interpret random data from your kernel stack
as instruction pointers and stack pointers.  This can cause exposure of
kernel stack contents to userspace.

Restrict the ability to inspect kernel stacks of arbitrary tasks to root
in order to prevent a local attacker from exploiting racy stack unwinding
to leak kernel task stack contents.  See the added comment for a longer
rationale.

There don't seem to be any users of this userspace API that can't
gracefully bail out if reading from the file fails.  Therefore, I believe
that this change is unlikely to break things.  In the case that this patch
does end up needing a revert, the next-best solution might be to fake a
single-entry stack based on wchan.

Link: http://lkml.kernel.org/r/20180927153316.200286-1-jannh@google.com
Fixes: 2ec220e27f50 ("proc: add /proc/*/stack")
Signed-off-by: Jann Horn <jannh@google.com>
Acked-by: Kees Cook <keescook@chromium.org>
Cc: Alexey Dobriyan <adobriyan@gmail.com>
Cc: Ken Chen <kenchen@google.com>
Cc: Will Deacon <will.deacon@arm.com>
Cc: Laura Abbott <labbott@redhat.com>
Cc: Andy Lutomirski <luto@amacapital.net>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: Josh Poimboeuf <jpoimboe@redhat.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: "H . Peter Anvin" <hpa@zytor.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agomm/hugetlb: add mmap() encodings for 32MB and 512MB page sizes
Anshuman Khandual [Fri, 5 Oct 2018 22:51:54 +0000 (15:51 -0700)]
mm/hugetlb: add mmap() encodings for 32MB and 512MB page sizes

ARM64 architecture also supports 32MB and 512MB HugeTLB page sizes.  This
just adds mmap() system call argument encoding for them.

Link: http://lkml.kernel.org/r/1537841300-6979-1-git-send-email-anshuman.khandual@arm.com
Signed-off-by: Anshuman Khandual <anshuman.khandual@arm.com>
Acked-by: Punit Agrawal <punit.agrawal@arm.com>
Acked-by: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Will Deacon <will.deacon@arm.com>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agomm/migrate.c: split only transparent huge pages when allocation fails
Anshuman Khandual [Fri, 5 Oct 2018 22:51:51 +0000 (15:51 -0700)]
mm/migrate.c: split only transparent huge pages when allocation fails

split_huge_page_to_list() fails on HugeTLB pages.  I was experimenting
with moving 32MB contig HugeTLB pages on arm64 (with a debug patch
applied) and hit the following stack trace when the kernel crashed.

[ 3732.462797] Call trace:
[ 3732.462835]  split_huge_page_to_list+0x3b0/0x858
[ 3732.462913]  migrate_pages+0x728/0xc20
[ 3732.462999]  soft_offline_page+0x448/0x8b0
[ 3732.463097]  __arm64_sys_madvise+0x724/0x850
[ 3732.463197]  el0_svc_handler+0x74/0x110
[ 3732.463297]  el0_svc+0x8/0xc
[ 3732.463347] Code: d1000400 f90b0e60 f2fbd5a2 a94982a1 (f9000420)

When unmap_and_move[_huge_page]() fails due to lack of memory, the
splitting should happen only for transparent huge pages not for HugeTLB
pages.  PageTransHuge() returns true for both THP and HugeTLB pages.
Hence the conditonal check should test PagesHuge() flag to make sure that
given pages is not a HugeTLB one.

Link: http://lkml.kernel.org/r/1537798495-4996-1-git-send-email-anshuman.khandual@arm.com
Fixes: 94723aafb9 ("mm: unclutter THP migration")
Signed-off-by: Anshuman Khandual <anshuman.khandual@arm.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Acked-by: Naoya Horiguchi <n-horiguchi@ah.jp.nec.com>
Cc: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Cc: Zi Yan <zi.yan@cs.rutgers.edu>
Cc: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agoipc/shm.c: use ERR_CAST() for shm_lock() error return
Kees Cook [Fri, 5 Oct 2018 22:51:48 +0000 (15:51 -0700)]
ipc/shm.c: use ERR_CAST() for shm_lock() error return

This uses ERR_CAST() instead of an open-coded cast, as it is casting
across structure pointers, which upsets __randomize_layout:

ipc/shm.c: In function `shm_lock':
ipc/shm.c:209:9: note: randstruct: casting between randomized structure pointer types (ssa): `struct shmid_kernel' and `struct kern_ipc_perm'

  return (void *)ipcp;
         ^~~~~~~~~~~~

Link: http://lkml.kernel.org/r/20180919180722.GA15073@beast
Fixes: 82061c57ce93 ("ipc: drop ipc_lock()")
Signed-off-by: Kees Cook <keescook@chromium.org>
Cc: Davidlohr Bueso <dbueso@suse.de>
Cc: Manfred Spraul <manfred@colorfullife.com>
Cc: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agomm/gup_benchmark: fix unsigned comparison to zero in __gup_benchmark_ioctl
YueHaibing [Fri, 5 Oct 2018 22:51:44 +0000 (15:51 -0700)]
mm/gup_benchmark: fix unsigned comparison to zero in __gup_benchmark_ioctl

get_user_pages_fast() will return negative value if no pages were pinned,
then be converted to a unsigned, which is compared to zero, giving the
wrong result.

Link: http://lkml.kernel.org/r/20180921095015.26088-1-yuehaibing@huawei.com
Fixes: 09e35a4a1ca8 ("mm/gup_benchmark: handle gup failures")
Signed-off-by: YueHaibing <yuehaibing@huawei.com>
Reviewed-by: Andrew Morton <akpm@linux-foundation.org>
Cc: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agomm, thp: fix mlocking THP page with migration enabled
Kirill A. Shutemov [Fri, 5 Oct 2018 22:51:41 +0000 (15:51 -0700)]
mm, thp: fix mlocking THP page with migration enabled

A transparent huge page is represented by a single entry on an LRU list.
Therefore, we can only make unevictable an entire compound page, not
individual subpages.

If a user tries to mlock() part of a huge page, we want the rest of the
page to be reclaimable.

We handle this by keeping PTE-mapped huge pages on normal LRU lists: the
PMD on border of VM_LOCKED VMA will be split into PTE table.

Introduction of THP migration breaks[1] the rules around mlocking THP
pages.  If we had a single PMD mapping of the page in mlocked VMA, the
page will get mlocked, regardless of PTE mappings of the page.

For tmpfs/shmem it's easy to fix by checking PageDoubleMap() in
remove_migration_pmd().

Anon THP pages can only be shared between processes via fork().  Mlocked
page can only be shared if parent mlocked it before forking, otherwise CoW
will be triggered on mlock().

For Anon-THP, we can fix the issue by munlocking the page on removing PTE
migration entry for the page.  PTEs for the page will always come after
mlocked PMD: rmap walks VMAs from oldest to newest.

Test-case:

#include <unistd.h>
#include <sys/mman.h>
#include <sys/wait.h>
#include <linux/mempolicy.h>
#include <numaif.h>

int main(void)
{
        unsigned long nodemask = 4;
        void *addr;

addr = mmap((void *)0x20000000UL, 2UL << 20, PROT_READ | PROT_WRITE,
MAP_PRIVATE | MAP_ANONYMOUS | MAP_LOCKED, -1, 0);

        if (fork()) {
wait(NULL);
return 0;
        }

        mlock(addr, 4UL << 10);
        mbind(addr, 2UL << 20, MPOL_PREFERRED | MPOL_F_RELATIVE_NODES,
                &nodemask, 4, MPOL_MF_MOVE);

        return 0;
}

[1] https://lkml.kernel.org/r/CAOMGZ=G52R-30rZvhGxEbkTw7rLLwBGadVYeo--iizcD3upL3A@mail.gmail.com

Link: http://lkml.kernel.org/r/20180917133816.43995-1-kirill.shutemov@linux.intel.com
Fixes: 616b8371539a ("mm: thp: enable thp migration in generic path")
Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Reported-by: Vegard Nossum <vegard.nossum@oracle.com>
Reviewed-by: Zi Yan <zi.yan@cs.rutgers.edu>
Cc: Naoya Horiguchi <n-horiguchi@ah.jp.nec.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: <stable@vger.kernel.org> [4.14+]
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agoocfs2: fix crash in ocfs2_duplicate_clusters_by_page()
Larry Chen [Fri, 5 Oct 2018 22:51:37 +0000 (15:51 -0700)]
ocfs2: fix crash in ocfs2_duplicate_clusters_by_page()

ocfs2_duplicate_clusters_by_page() may crash if one of the extent's pages
is dirty.  When a page has not been written back, it is still in dirty
state.  If ocfs2_duplicate_clusters_by_page() is called against the dirty
page, the crash happens.

To fix this bug, we can just unlock the page and wait until the page until
its not dirty.

The following is the backtrace:

kernel BUG at /root/code/ocfs2/refcounttree.c:2961!
[exception RIP: ocfs2_duplicate_clusters_by_page+822]
__ocfs2_move_extent+0x80/0x450 [ocfs2]
? __ocfs2_claim_clusters+0x130/0x250 [ocfs2]
ocfs2_defrag_extent+0x5b8/0x5e0 [ocfs2]
__ocfs2_move_extents_range+0x2a4/0x470 [ocfs2]
ocfs2_move_extents+0x180/0x3b0 [ocfs2]
? ocfs2_wait_for_recovery+0x13/0x70 [ocfs2]
ocfs2_ioctl_move_extents+0x133/0x2d0 [ocfs2]
ocfs2_ioctl+0x253/0x640 [ocfs2]
do_vfs_ioctl+0x90/0x5f0
SyS_ioctl+0x74/0x80
do_syscall_64+0x74/0x140
entry_SYSCALL_64_after_hwframe+0x3d/0xa2

Once we find the page is dirty, we do not wait until it's clean, rather we
use write_one_page() to write it back

Link: http://lkml.kernel.org/r/20180829074740.9438-1-lchen@suse.com
[lchen@suse.com: update comments]
Link: http://lkml.kernel.org/r/20180830075041.14879-1-lchen@suse.com
[akpm@linux-foundation.org: coding-style fixes]
Signed-off-by: Larry Chen <lchen@suse.com>
Acked-by: Changwei Ge <ge.changwei@h3c.com>
Cc: Mark Fasheh <mark@fasheh.com>
Cc: Joel Becker <jlbec@evilplan.org>
Cc: Junxiao Bi <junxiao.bi@oracle.com>
Cc: Joseph Qi <jiangqi903@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agohugetlb: take PMD sharing into account when flushing tlb/caches
Mike Kravetz [Fri, 5 Oct 2018 22:51:33 +0000 (15:51 -0700)]
hugetlb: take PMD sharing into account when flushing tlb/caches

When fixing an issue with PMD sharing and migration, it was discovered via
code inspection that other callers of huge_pmd_unshare potentially have an
issue with cache and tlb flushing.

Use the routine adjust_range_if_pmd_sharing_possible() to calculate worst
case ranges for mmu notifiers.  Ensure that this range is flushed if
huge_pmd_unshare succeeds and unmaps a PUD_SUZE area.

Link: http://lkml.kernel.org/r/20180823205917.16297-3-mike.kravetz@oracle.com
Signed-off-by: Mike Kravetz <mike.kravetz@oracle.com>
Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Reviewed-by: Naoya Horiguchi <n-horiguchi@ah.jp.nec.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Davidlohr Bueso <dave@stgolabs.net>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Jerome Glisse <jglisse@redhat.com>
Cc: Mike Kravetz <mike.kravetz@oracle.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agomm: migration: fix migration of huge PMD shared pages
Mike Kravetz [Fri, 5 Oct 2018 22:51:29 +0000 (15:51 -0700)]
mm: migration: fix migration of huge PMD shared pages

The page migration code employs try_to_unmap() to try and unmap the source
page.  This is accomplished by using rmap_walk to find all vmas where the
page is mapped.  This search stops when page mapcount is zero.  For shared
PMD huge pages, the page map count is always 1 no matter the number of
mappings.  Shared mappings are tracked via the reference count of the PMD
page.  Therefore, try_to_unmap stops prematurely and does not completely
unmap all mappings of the source page.

This problem can result is data corruption as writes to the original
source page can happen after contents of the page are copied to the target
page.  Hence, data is lost.

This problem was originally seen as DB corruption of shared global areas
after a huge page was soft offlined due to ECC memory errors.  DB
developers noticed they could reproduce the issue by (hotplug) offlining
memory used to back huge pages.  A simple testcase can reproduce the
problem by creating a shared PMD mapping (note that this must be at least
PUD_SIZE in size and PUD_SIZE aligned (1GB on x86)), and using
migrate_pages() to migrate process pages between nodes while continually
writing to the huge pages being migrated.

To fix, have the try_to_unmap_one routine check for huge PMD sharing by
calling huge_pmd_unshare for hugetlbfs huge pages.  If it is a shared
mapping it will be 'unshared' which removes the page table entry and drops
the reference on the PMD page.  After this, flush caches and TLB.

mmu notifiers are called before locking page tables, but we can not be
sure of PMD sharing until page tables are locked.  Therefore, check for
the possibility of PMD sharing before locking so that notifiers can
prepare for the worst possible case.

Link: http://lkml.kernel.org/r/20180823205917.16297-2-mike.kravetz@oracle.com
[mike.kravetz@oracle.com: make _range_in_vma() a static inline]
Link: http://lkml.kernel.org/r/6063f215-a5c8-2f0c-465a-2c515ddc952d@oracle.com
Fixes: 39dde65c9940 ("shared page table for hugetlb page")
Signed-off-by: Mike Kravetz <mike.kravetz@oracle.com>
Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Reviewed-by: Naoya Horiguchi <n-horiguchi@ah.jp.nec.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Davidlohr Bueso <dave@stgolabs.net>
Cc: Jerome Glisse <jglisse@redhat.com>
Cc: Mike Kravetz <mike.kravetz@oracle.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agoMerge tag 'pci-v4.19-fixes-3' of ssh://gitolite.kernel.org/pub/scm/linux/kernel/git...
Greg Kroah-Hartman [Fri, 5 Oct 2018 23:11:16 +0000 (16:11 -0700)]
Merge tag 'pci-v4.19-fixes-3' of ssh://gitolite./linux/kernel/git/helgaas/pci

Bjorn writes:
  "PCI fixes for v4.19:

   - Reprogram bridge prefetch registers to fix NVIDIA and Radeon issues
     after suspend/resume (Daniel Drake)

   - Fix mvebu I/O mapping creation sequence (Thomas Petazzoni)

   - Fix minor MAINTAINERS file match issue (Bjorn Helgaas)"

* tag 'pci-v4.19-fixes-3' of ssh://gitolite.kernel.org/pub/scm/linux/kernel/git/helgaas/pci:
  PCI: mvebu: Fix PCI I/O mapping creation sequence
  MAINTAINERS: Remove obsolete drivers/pci pattern from ACPI section
  PCI: Reprogram bridge prefetch registers on resume

5 years agoMerge tag 'for-4.19/dm-fixes-2' of git://git.kernel.org/pub/scm/linux/kernel/git...
Greg Kroah-Hartman [Fri, 5 Oct 2018 23:09:56 +0000 (16:09 -0700)]
Merge tag 'for-4.19/dm-fixes-2' of git://git./linux/kernel/git/device-mapper/linux-dm

Mike writes:
  "device mapper fixes

   - Fix a DM thinp __udivdi3 undefined on 32-bit bug introduced during
     4.19 merge window.

   - Fix leak and dangling pointer in DM multipath's scsi_dh related code.

   - A couple stable@ fixes for DM cache's resize support.

   - A DM raid fix to remove "const" from decipher_sync_action()'s return
     type."

* tag 'for-4.19/dm-fixes-2' of git://git.kernel.org/pub/scm/linux/kernel/git/device-mapper/linux-dm:
  dm cache: fix resize crash if user doesn't reload cache table
  dm cache metadata: ignore hints array being too small during resize
  dm raid: remove bogus const from decipher_sync_action() return type
  dm mpath: fix attached_handler_name leak and dangling hw_handler_name pointer
  dm thin metadata: fix __udivdi3 undefined on 32-bit

5 years agoMerge tag 'gpio-v4.19-3' of git://git.kernel.org/pub/scm/linux/kernel/git/linusw...
Greg Kroah-Hartman [Fri, 5 Oct 2018 23:09:11 +0000 (16:09 -0700)]
Merge tag 'gpio-v4.19-3' of git://git./linux/kernel/git/linusw/linux-gpio

Linus writes:
  "A single GPIO fix:
   Free the last used descriptor, an off by one error.
   This is tagged for stable as well."

* tag 'gpio-v4.19-3' of git://git.kernel.org/pub/scm/linux/kernel/git/linusw/linux-gpio:
  gpiolib: Free the last requested descriptor

5 years agoMerge tag 'pm-4.19-rc7' of git://git.kernel.org/pub/scm/linux/kernel/git/rafael/linux-pm
Greg Kroah-Hartman [Fri, 5 Oct 2018 23:08:12 +0000 (16:08 -0700)]
Merge tag 'pm-4.19-rc7' of git://git./linux/kernel/git/rafael/linux-pm

Rafael writes:
  "Power management fix for 4.19-rc7

   Fix a bug that may cause runtime PM to misbehave for some devices
   after a failing or aborted system suspend which is nasty enough for
   an -rc7 time frame fix."

* tag 'pm-4.19-rc7' of git://git.kernel.org/pub/scm/linux/kernel/git/rafael/linux-pm:
  PM / core: Clear the direct_complete flag on errors

5 years agoMerge branch 'perf-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel...
Greg Kroah-Hartman [Fri, 5 Oct 2018 23:07:13 +0000 (16:07 -0700)]
Merge branch 'perf-urgent-for-linus' of git://git./linux/kernel/git/tip/tip

Ingo writes:
  "perf fixes:
    - fix a CPU#0 hot unplug bug and a PCI enumeration bug in the x86 Intel uncore PMU driver
    - fix a CPU event enumeration bug in the x86 AMD PMU driver
    - fix a perf ring-buffer corruption bug when using tracepoints
    - fix a PMU unregister locking bug"

* 'perf-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  perf/x86/amd/uncore: Set ThreadMask and SliceMask for L3 Cache perf events
  perf/x86/intel/uncore: Fix PCI BDF address of M3UPI on SKX
  perf/ring_buffer: Prevent concurent ring buffer access
  perf/x86/intel/uncore: Use boot_cpu_data.phys_proc_id instead of hardcorded physical package ID 0
  perf/core: Fix perf_pmu_unregister() locking

5 years agoMerge branch 'x86-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel...
Greg Kroah-Hartman [Fri, 5 Oct 2018 22:40:57 +0000 (15:40 -0700)]
Merge branch 'x86-urgent-for-linus' of git://git./linux/kernel/git/tip/tip

Ingo writes:
  "x86 fixes:

   Misc fixes:

    - fix various vDSO bugs: asm constraints and retpolines
    - add vDSO test units to make sure they never re-appear
    - fix UV platform TSC initialization bug
    - fix build warning on Clang"

* 'x86-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  x86/vdso: Fix vDSO syscall fallback asm constraint regression
  x86/cpu/amd: Remove unnecessary parentheses
  x86/vdso: Only enable vDSO retpolines when enabled and supported
  x86/tsc: Fix UV TSC initialization
  x86/platform/uv: Provide is_early_uv_system()
  selftests/x86: Add clock_gettime() tests to test_vdso
  x86/vdso: Fix asm constraints on vDSO syscall fallbacks

5 years agoMerge branch 'sched-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel...
Greg Kroah-Hartman [Fri, 5 Oct 2018 22:39:38 +0000 (15:39 -0700)]
Merge branch 'sched-urgent-for-linus' of git://git./linux/kernel/git/tip/tip

Ingo writes:
  "scheduler fixes:

   These fixes address a rather involved performance regression between
   v4.17->v4.19 in the sched/numa auto-balancing code. Since distros
   really need this fix we accelerated it to sched/urgent for a faster
   upstream merge.

   NUMA scheduling and balancing performance is now largely back to
   v4.17 levels, without reintroducing the NUMA placement bugs that
   v4.18 and v4.19 fixed.

   Many thanks to Srikar Dronamraju, Mel Gorman and Jirka Hladky, for
   reporting, testing, re-testing and solving this rather complex set of
   bugs."

* 'sched-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  sched/numa: Migrate pages to local nodes quicker early in the lifetime of a task
  mm, sched/numa: Remove rate-limiting of automatic NUMA balancing migration
  sched/numa: Avoid task migration for small NUMA improvement
  mm/migrate: Use spin_trylock() while resetting rate limit
  sched/numa: Limit the conditions where scan period is reset
  sched/numa: Reset scan rate whenever task moves across nodes
  sched/numa: Pass destination CPU as a parameter to migrate_task_rq
  sched/numa: Stop multiple tasks from moving to the CPU at the same time

5 years agoMerge branch 'locking-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel...
Greg Kroah-Hartman [Fri, 5 Oct 2018 22:38:32 +0000 (15:38 -0700)]
Merge branch 'locking-urgent-for-linus' of git://git./linux/kernel/git/tip/tip

Ingo writes:
  "locking fixes:

   A fix in the ww_mutex self-test that produces a scary splat, plus an
   updates to the maintained-filed patters in MAINTAINER."

* 'locking-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  locking/ww_mutex: Fix runtime warning in the WW mutex selftest
  MAINTAINERS: Remove dead path from LOCKING PRIMITIVES entry

5 years agoMerge tag 'sound-4.19-rc7' of git://git.kernel.org/pub/scm/linux/kernel/git/tiwai...
Greg Kroah-Hartman [Fri, 5 Oct 2018 22:37:22 +0000 (15:37 -0700)]
Merge tag 'sound-4.19-rc7' of git://git./linux/kernel/git/tiwai/sound

Takashi writes:
  "sound fixes for 4.19-rc7

   Just two small fixes for HD-audio: one is for a typo in completion
   timeout, and another a fixup for Dell machines as usual"

* tag 'sound-4.19-rc7' of git://git.kernel.org/pub/scm/linux/kernel/git/tiwai/sound:
  ALSA: hda/realtek - Cannot adjust speaker's volume on Dell XPS 27 7760
  ALSA: hda: Fix the audio-component completion timeout

5 years agoMerge tag 'iommu-fixes-v4.19-rc6' of git://git.kernel.org/pub/scm/linux/kernel/git...
Greg Kroah-Hartman [Fri, 5 Oct 2018 15:30:40 +0000 (08:30 -0700)]
Merge tag 'iommu-fixes-v4.19-rc6' of git://git./linux/kernel/git/joro/iommu

Joerg writes:
  "IOMMU Fix for Linux v4.19-rc6

   One important fix:
- Fix a memory leak with AMD IOMMU when SME is active and a VM
  has assigned devices. In that case the complete guest memory
  will be leaked without this fix."

* tag 'iommu-fixes-v4.19-rc6' of git://git.kernel.org/pub/scm/linux/kernel/git/joro/iommu:
  iommu/amd: Clear memory encryption mask from physical address

5 years agoMerge tag 'for-linus' of git://git.kernel.org/pub/scm/virt/kvm/kvm
Greg Kroah-Hartman [Fri, 5 Oct 2018 15:29:44 +0000 (08:29 -0700)]
Merge tag 'for-linus' of git://git./virt/kvm/kvm

Paolo writes:
  "KVM changes for 4.19-rc7

   x86 and PPC bugfixes, mostly introduced in 4.19-rc1."

* tag 'for-linus' of git://git.kernel.org/pub/scm/virt/kvm/kvm:
  kvm: nVMX: fix entry with pending interrupt if APICv is enabled
  KVM: VMX: hide flexpriority from guest when disabled at the module level
  KVM: VMX: check for existence of secondary exec controls before accessing
  KVM: PPC: Book3S HV: Avoid crash from THP collapse during radix page fault
  KVM: x86: fix L1TF's MMIO GFN calculation
  tools/kvm_stat: cut down decimal places in update interval dialog
  KVM: nVMX: Fix emulation of VM_ENTRY_LOAD_BNDCFGS
  KVM: x86: Do not use kvm_x86_ops->mpx_supported() directly
  KVM: nVMX: Do not expose MPX VMX controls when guest MPX disabled
  KVM: x86: never trap MSR_KERNEL_GS_BASE

5 years agoMerge branch 'linus' of git://git.kernel.org/pub/scm/linux/kernel/git/herbert/crypto-2.6
Greg Kroah-Hartman [Fri, 5 Oct 2018 15:28:35 +0000 (08:28 -0700)]
Merge branch 'linus' of git://git./linux/kernel/git/herbert/crypto-2.6

Herbert writes:
  "Crypto Fixes for 4.19

   This push fixes the following issues:
   - Out-of-bound stack access in qat.
   - Illegal schedule in mxs-dcp.
   - Memory corruption in chelsio.
   - Incorrect pointer computation in caam."

* 'linus' of git://git.kernel.org/pub/scm/linux/kernel/git/herbert/crypto-2.6:
  crypto: qat - Fix KASAN stack-out-of-bounds bug in adf_probe()
  crypto: mxs-dcp - Fix wait logic on chan threads
  crypto: chelsio - Fix memory corruption in DMA Mapped buffers.
  crypto: caam/jr - fix ablkcipher_edesc pointer arithmetic

5 years agoMerge tag '4.19-rc6-smb3-fixes' of git://git.samba.org/sfrench/cifs-2.6
Greg Kroah-Hartman [Fri, 5 Oct 2018 15:27:47 +0000 (08:27 -0700)]
Merge tag '4.19-rc6-smb3-fixes' of git://git.samba.org/sfrench/cifs-2.6

Steve writes:
  "SMB3 fixes

   four small SMB3 fixes: one for stable, the others to address a more
   recent regression"

* tag '4.19-rc6-smb3-fixes' of git://git.samba.org/sfrench/cifs-2.6:
  smb3: fix lease break problem introduced by compounding
  cifs: only wake the thread for the very last PDU in a compound
  cifs: add a warning if we try to to dequeue a deleted mid
  smb2: fix missing files in root share directory listing

5 years agoiommu/amd: Clear memory encryption mask from physical address
Singh, Brijesh [Thu, 4 Oct 2018 21:40:23 +0000 (21:40 +0000)]
iommu/amd: Clear memory encryption mask from physical address

Boris Ostrovsky reported a memory leak with device passthrough when SME
is active.

The VFIO driver uses iommu_iova_to_phys() to get the physical address for
an iova. This physical address is later passed into vfio_unmap_unpin() to
unpin the memory. The vfio_unmap_unpin() uses pfn_valid() before unpinning
the memory. The pfn_valid() check was failing because encryption mask was
part of the physical address returned. This resulted in the memory not
being unpinned and therefore leaked after the guest terminates.

The memory encryption mask must be cleared from the physical address in
iommu_iova_to_phys().

Fixes: 2543a786aa25 ("iommu/amd: Allow the AMD IOMMU to work with memory encryption")
Reported-by: Boris Ostrovsky <boris.ostrovsky@oracle.com>
Cc: Tom Lendacky <thomas.lendacky@amd.com>
Cc: Joerg Roedel <joro@8bytes.org>
Cc: <iommu@lists.linux-foundation.org>
Cc: Borislav Petkov <bp@suse.de>
Cc: Paolo Bonzini <pbonzini@redhat.com>
Cc: Radim Krčmář <rkrcmar@redhat.com>
Cc: kvm@vger.kernel.org
Cc: Boris Ostrovsky <boris.ostrovsky@oracle.com>
Cc: <stable@vger.kernel.org> # 4.14+
Signed-off-by: Brijesh Singh <brijesh.singh@amd.com>
Signed-off-by: Joerg Roedel <jroedel@suse.de>
5 years agoMerge tag 'kvm-ppc-fixes-4.19-3' of git://git.kernel.org/pub/scm/linux/kernel/git...
Paolo Bonzini [Fri, 5 Oct 2018 07:39:53 +0000 (09:39 +0200)]
Merge tag 'kvm-ppc-fixes-4.19-3' of git://git./linux/kernel/git/paulus/powerpc into kvm-master

Third set of PPC KVM fixes for 4.19

One patch here, fixing a potential host crash introduced (or at least
exacerbated) by a previous fix for corruption relating to radix guest
page faults and THP operations.

5 years agoMerge tag 'drm-fixes-2018-10-05' of git://anongit.freedesktop.org/drm/drm
Greg Kroah-Hartman [Fri, 5 Oct 2018 00:23:58 +0000 (17:23 -0700)]
Merge tag 'drm-fixes-2018-10-05' of git://anongit.freedesktop.org/drm/drm

Dave writes:
  "amdgpu and two core fixes

   Two fixes for amdgpu:
   one corrects a use of process->mm
   one fix for display code race condition that can result in a crash

   Two core fixes:
   One for a use-after-free in the leasing code
   One for a cma/fbdev crash."

* tag 'drm-fixes-2018-10-05' of git://anongit.freedesktop.org/drm/drm:
  drm/amdkfd: Fix incorrect use of process->mm
  drm/amd/display: Signal hw_done() after waiting for flip_done()
  drm/cma-helper: Fix crash in fbdev error path
  drm: fix use-after-free read in drm_mode_create_lease_ioctl()

5 years agoMerge branch 'drm-fixes-4.19' of git://people.freedesktop.org/~agd5f/linux into drm...
Dave Airlie [Thu, 4 Oct 2018 22:39:31 +0000 (08:39 +1000)]
Merge branch 'drm-fixes-4.19' of git://people.freedesktop.org/~agd5f/linux into drm-fixes

- Fix an ordering issue in DC with respect to atomic flips that could result
  in a crash
- Fix incorrect use of process->mm in KFD

Signed-off-by: Dave Airlie <airlied@redhat.com>
From: Alex Deucher <alexdeucher@gmail.com>
Link: https://patchwork.freedesktop.org/patch/msgid/1538668374-22334-1-git-send-email-alexander.deucher@amd.com
5 years agoMerge tag 'drm-misc-fixes-2018-10-04' of git://anongit.freedesktop.org/drm/drm-misc...
Dave Airlie [Thu, 4 Oct 2018 22:39:01 +0000 (08:39 +1000)]
Merge tag 'drm-misc-fixes-2018-10-04' of git://anongit.freedesktop.org/drm/drm-misc into drm-fixes

drm-misc-fixes for v4.19-rc7:
- Fix use-after-free in drm_mode_create_lease_ioctl()
- Fix crash in fbdev error path.

Signed-off-by: Dave Airlie <airlied@redhat.com>
From: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/13b2c3ac-9a96-710e-ceb9-890af164f10e@linux.intel.com
5 years agoMerge tag 'ovl-fixes-4.19-rc7' of git://git.kernel.org/pub/scm/linux/kernel/git/mszer...
Greg Kroah-Hartman [Thu, 4 Oct 2018 20:24:38 +0000 (13:24 -0700)]
Merge tag 'ovl-fixes-4.19-rc7' of git://git./linux/kernel/git/mszeredi/vfs

Miklos writes:
  "overlayfs fixes for 4.19-rc7

   This update fixes a couple of regressions in the stacked file update
   added in this cycle, as well as some older bugs uncovered by
   syzkaller.

   There's also one trivial naming change that touches other parts of
   the fs subsystem."

* tag 'ovl-fixes-4.19-rc7' of git://git.kernel.org/pub/scm/linux/kernel/git/mszeredi/vfs:
  ovl: fix format of setxattr debug
  ovl: fix access beyond unterminated strings
  ovl: make symbol 'ovl_aops' static
  vfs: swap names of {do,vfs}_clone_file_range()
  ovl: fix freeze protection bypass in ovl_clone_file_range()
  ovl: fix freeze protection bypass in ovl_write_iter()
  ovl: fix memory leak on unlink of indexed file

5 years agodm cache: fix resize crash if user doesn't reload cache table
Mike Snitzer [Wed, 26 Sep 2018 00:56:02 +0000 (20:56 -0400)]
dm cache: fix resize crash if user doesn't reload cache table

A reload of the cache's DM table is needed during resize because
otherwise a crash will occur when attempting to access smq policy
entries associated with the portion of the cache that was recently
extended.

The reason is cache-size based data structures in the policy will not be
resized, the only way to safely extend the cache is to allow for a
proper cache policy initialization that occurs when the cache table is
loaded.  For example the smq policy's space_init(), init_allocator(),
calc_hotspot_params() must be sized based on the extended cache size.

The fix for this is to disallow cache resizes of this pattern:
1) suspend "cache" target's device
2) resize the fast device used for the cache
3) resume "cache" target's device

Instead, the last step must be a full reload of the cache's DM table.

Fixes: 66a636356 ("dm cache: add stochastic-multi-queue (smq) policy")
Cc: stable@vger.kernel.org
Signed-off-by: Mike Snitzer <snitzer@redhat.com>
5 years agodm cache metadata: ignore hints array being too small during resize
Joe Thornber [Mon, 24 Sep 2018 20:19:30 +0000 (16:19 -0400)]
dm cache metadata: ignore hints array being too small during resize

Commit fd2fa9541 ("dm cache metadata: save in-core policy_hint_size to
on-disk superblock") enabled previously written policy hints to be
used after a cache is reactivated.  But in doing so the cache
metadata's hint array was left exposed to out of bounds access because
on resize the metadata's on-disk hint array wasn't ever extended.

Fix this by ignoring that there are no on-disk hints associated with the
newly added cache blocks.  An expanded on-disk hint array is later
rewritten upon the next clean shutdown of the cache.

Fixes: fd2fa9541 ("dm cache metadata: save in-core policy_hint_size to on-disk superblock")
Cc: stable@vger.kernel.org
Signed-off-by: Joe Thornber <ejt@redhat.com>
Signed-off-by: Mike Snitzer <snitzer@redhat.com>
5 years agoPM / core: Clear the direct_complete flag on errors
Rafael J. Wysocki [Thu, 4 Oct 2018 09:08:12 +0000 (11:08 +0200)]
PM / core: Clear the direct_complete flag on errors

If __device_suspend() runs asynchronously (in which case the device
passed to it is in dpm_suspended_list at that point) and it returns
early on an error or pending wakeup, and the power.direct_complete
flag has been set for the device already, the subsequent
device_resume() will be confused by that and it will call
pm_runtime_enable() incorrectly, as runtime PM has not been
disabled for the device by __device_suspend().

To avoid that, clear power.direct_complete if __device_suspend()
is not going to disable runtime PM for the device before returning.

Fixes: aae4518b3124 (PM / sleep: Mechanism to avoid resuming runtime-suspended devices unnecessarily)
Reported-by: Al Cooper <alcooperx@gmail.com>
Tested-by: Al Cooper <alcooperx@gmail.com>
Reviewed-by: Ulf Hansson <ulf.hansson@linaro.org>
Cc: 3.16+ <stable@vger.kernel.org> # 3.16+
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
5 years agoMerge branch 'fixes' of git://git.armlinux.org.uk/~rmk/linux-arm
Greg Kroah-Hartman [Thu, 4 Oct 2018 16:48:10 +0000 (09:48 -0700)]
Merge branch 'fixes' of git://git.armlinux.org.uk/~rmk/linux-arm

Russell writes:
  "A couple of small ARM fixes from Stefan and Thomas:
   - Adding the io_pgetevents syscall
   - Fixing a bounds check in pci_ioremap_io()"

* 'fixes' of git://git.armlinux.org.uk/~rmk/linux-arm:
  ARM: 8799/1: mm: fix pci_ioremap_io() offset check
  ARM: 8787/1: wire up io_pgetevents syscall

5 years agoMerge tag 'drm-fixes-2018-10-04' of git://anongit.freedesktop.org/drm/drm
Greg Kroah-Hartman [Thu, 4 Oct 2018 16:18:44 +0000 (09:18 -0700)]
Merge tag 'drm-fixes-2018-10-04' of git://anongit.freedesktop.org/drm/drm

Dave writes:
  "drm exynos, tda9950 and intel fixes

   3 i915 fixes:
     compressed error handling zlib fix
     compiler warning cleanup
     and a minor code cleanup

   2 tda9950:
     Two fixes for the HDMI CEC

   1 exynos:
     A fix required for IOMMU interaction."

* tag 'drm-fixes-2018-10-04' of git://anongit.freedesktop.org/drm/drm:
  drm/i915: Handle incomplete Z_FINISH for compressed error states
  drm/i915: Avoid compiler warning for maybe unused gu_misc_iir
  drm/i915: Do not redefine the has_csr parameter.
  drm/exynos: Use selected dma_dev default iommu domain instead of a fake one
  drm/i2c: tda9950: set MAX_RETRIES for errors only
  drm/i2c: tda9950: fix timeout counter check

5 years agoMerge tag 'xfs-fixes-for-4.19-rc6' of git://git.kernel.org/pub/scm/fs/xfs/xfs-linux
Greg Kroah-Hartman [Thu, 4 Oct 2018 16:17:38 +0000 (09:17 -0700)]
Merge tag 'xfs-fixes-for-4.19-rc6' of git://git./fs/xfs/xfs-linux

Dave writes:
  "XFS fixes for 4.19-rc6

   Accumlated regression and bug fixes for 4.19-rc6, including:

   o make iomap correctly mark dirty pages for sub-page block sizes
   o fix regression in handling extent-to-btree format conversion errors
   o fix torn log wrap detection for new logs
   o various corrupt inode detection fixes
   o various delalloc state fixes
   o cleanup all the missed transaction cancel cases missed from changes merged
     in 4.19-rc1
   o fix lockdep false positive on transaction allocation
   o fix locking and reference counting on buffer log items"

* tag 'xfs-fixes-for-4.19-rc6' of git://git.kernel.org/pub/scm/fs/xfs/xfs-linux:
  xfs: fix error handling in xfs_bmap_extents_to_btree
  iomap: set page dirty after partial delalloc on mkwrite
  xfs: remove invalid log recovery first/last cycle check
  xfs: validate inode di_forkoff
  xfs: skip delalloc COW blocks in xfs_reflink_end_cow
  xfs: don't treat unknown di_flags2 as corruption in scrub
  xfs: remove duplicated include from alloc.c
  xfs: don't bring in extents in xfs_bmap_punch_delalloc_range
  xfs: fix transaction leak in xfs_reflink_allocate_cow()
  xfs: avoid lockdep false positives in xfs_trans_alloc
  xfs: refactor xfs_buf_log_item reference count handling
  xfs: clean up xfs_trans_brelse()
  xfs: don't unlock invalidated buf on aborted tx commit
  xfs: remove last of unnecessary xfs_defer_cancel() callers
  xfs: don't crash the vfs on a garbage inline symlink

5 years agoMerge tag 'riscv-for-linus-4.19-rc7' of git://git.kernel.org/pub/scm/linux/kernel...
Greg Kroah-Hartman [Thu, 4 Oct 2018 16:16:11 +0000 (09:16 -0700)]
Merge tag 'riscv-for-linus-4.19-rc7' of git://git./linux/kernel/git/palmer/riscv-linux

Palmer writes:
  "A Single RISC-V Fix for 4.19-rc7

   This tag contains a single patch that managed to get lost in the
   shuffle, which explains why it's so late.  This single line has been
   floating around in various patch sets for months, and fixes our DMA32
   region."

* tag 'riscv-for-linus-4.19-rc7' of git://git.kernel.org/pub/scm/linux/kernel/git/palmer/riscv-linux:
  RISCV: Fix end PFN for low memory

5 years agodrm/amdkfd: Fix incorrect use of process->mm
Felix Kuehling [Tue, 2 Oct 2018 22:41:12 +0000 (18:41 -0400)]
drm/amdkfd: Fix incorrect use of process->mm

This mm_struct pointer should never be dereferenced. If running in
a user thread, just use current->mm. If running in a kernel worker
use get_task_mm to get a safe reference to the mm_struct.

Reviewed-by: Oded Gabbay <oded.gabbay@gmail.com>
Acked-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Felix Kuehling <Felix.Kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
5 years agodrm/amd/display: Signal hw_done() after waiting for flip_done()
Shirish S [Mon, 24 Sep 2018 13:31:47 +0000 (19:01 +0530)]
drm/amd/display: Signal hw_done() after waiting for flip_done()

In amdgpu_dm_commit_tail(), wait until flip_done() is signaled before
we signal hw_done().

[Why]

This is to temporarily address a paging error that occurs when a
nonblocking commit contends with another commit, particularly in a
mirrored display configuration where at least 2 CRTCs are updated.
The error occurs in drm_atomic_helper_wait_for_flip_done(), when we
attempt to access the contents of new_crtc_state->commit.

Here's the sequence for a mirrored 2 display setup (irrelevant steps
left out for clarity):

**THREAD 1**                        | **THREAD 2**
                                    |
Initialize atomic state for flip    |
                                    |
Queue worker                        |
                                   ...

                                    | Do work for flip
                                    |
                                    | Signal hw_done() on CRTC 1
                                    | Signal hw_done() on CRTC 2
                                    |
                                    | Wait for flip_done() on CRTC 1

                                <---- **PREEMPTED BY THREAD 1**

Initialize atomic state for cursor  |
update (1)                          |
                                    |
Do cursor update work on both CRTCs |
                                    |
Clear atomic state (2)              |
**DONE**                            |
                                   ...
                                    |
                                    | Wait for flip_done() on CRTC 2
                                    | *ERROR*
                                    |

The issue starts with (1). When the atomic state is initialized, the
current CRTC states are duplicated to be the new_crtc_states, and
referenced to be the old_crtc_states. (The new_crtc_states are to be
filled with update data.)

Some things to note:

* Due to the mirrored configuration, the cursor updates on both CRTCs.

* At this point, the pflip IRQ has already been handled, and flip_done
  signaled on all CRTCs. The cursor commit can therefore continue.

* The old_crtc_states used by the cursor update are the **same states**
  as the new_crtc_states used by the flip worker.

At (2), the old_crtc_state is freed (*), and the cursor commit
completes. We then context switch back to the flip worker, where we
attempt to access the new_crtc_state->commit object. This is
problematic, as this state has already been freed.

(*) Technically, 'state->crtcs[i].state' is freed, which was made to
    reference old_crtc_state in drm_atomic_helper_swap_state()

[How]

By moving hw_done() after wait_for_flip_done(), we're guaranteed that
the new_crtc_state (from the flip worker's perspective) still exists.
This is because any other commit will be blocked, waiting for the
hw_done() signal.

Note that both the i915 and imx drivers have this sequence flipped
already, masking this problem.

Signed-off-by: Shirish S <shirish.s@amd.com>
Signed-off-by: Leo Li <sunpeng.li@amd.com>
Reviewed-by: Harry Wentland <harry.wentland@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
5 years agokvm: nVMX: fix entry with pending interrupt if APICv is enabled
Paolo Bonzini [Wed, 3 Oct 2018 11:44:26 +0000 (13:44 +0200)]
kvm: nVMX: fix entry with pending interrupt if APICv is enabled

Commit b5861e5cf2fcf83031ea3e26b0a69d887adf7d21 introduced a check on
the interrupt-window and NMI-window CPU execution controls in order to
inject an external interrupt vmexit before the first guest instruction
executes.  However, when APIC virtualization is enabled the host does not
need a vmexit in order to inject an interrupt at the next interrupt window;
instead, it just places the interrupt vector in RVI and the processor will
inject it as soon as possible.  Therefore, on machines with APICv it is
not enough to check the CPU execution controls: the same scenario can also
happen if RVI>vPPR.

Fixes: b5861e5cf2fcf83031ea3e26b0a69d887adf7d21
Reviewed-by: Nikita Leshchenko <nikita.leshchenko@oracle.com>
Cc: Sean Christopherson <sean.j.christopherson@intel.com>
Cc: Liran Alon <liran.alon@oracle.com>
Cc: Radim Krčmář <rkrcmar@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
5 years agoovl: fix format of setxattr debug
Miklos Szeredi [Thu, 4 Oct 2018 12:49:10 +0000 (14:49 +0200)]
ovl: fix format of setxattr debug

Format has a typo: it was meant to be "%.*s", not "%*s".  But at some point
callers grew nonprintable values as well, so use "%*pE" instead with a
maximized length.

Reported-by: Amir Goldstein <amir73il@gmail.com>
Signed-off-by: Miklos Szeredi <mszeredi@redhat.com>
Fixes: 3a1e819b4e80 ("ovl: store file handle of lower inode on copy up")
Cc: <stable@vger.kernel.org> # v4.12
5 years agoovl: fix access beyond unterminated strings
Amir Goldstein [Fri, 28 Sep 2018 18:00:48 +0000 (21:00 +0300)]
ovl: fix access beyond unterminated strings

KASAN detected slab-out-of-bounds access in printk from overlayfs,
because string format used %*s instead of %.*s.

> BUG: KASAN: slab-out-of-bounds in string+0x298/0x2d0 lib/vsprintf.c:604
> Read of size 1 at addr ffff8801c36c66ba by task syz-executor2/27811
>
> CPU: 0 PID: 27811 Comm: syz-executor2 Not tainted 4.19.0-rc5+ #36
...
>  printk+0xa7/0xcf kernel/printk/printk.c:1996
>  ovl_lookup_index.cold.15+0xe8/0x1f8 fs/overlayfs/namei.c:689

Reported-by: syzbot+376cea2b0ef340db3dd4@syzkaller.appspotmail.com
Signed-off-by: Amir Goldstein <amir73il@gmail.com>
Signed-off-by: Miklos Szeredi <mszeredi@redhat.com>
Fixes: 359f392ca53e ("ovl: lookup index entry for copy up origin")
Cc: <stable@vger.kernel.org> # v4.13
5 years agoKVM: VMX: hide flexpriority from guest when disabled at the module level
Paolo Bonzini [Wed, 3 Oct 2018 08:34:00 +0000 (10:34 +0200)]
KVM: VMX: hide flexpriority from guest when disabled at the module level

As of commit 8d860bbeedef ("kvm: vmx: Basic APIC virtualization controls
have three settings"), KVM will disable VIRTUALIZE_APIC_ACCESSES when
a nested guest writes APIC_BASE MSR and kvm-intel.flexpriority=0,
whereas previously KVM would allow a nested guest to enable
VIRTUALIZE_APIC_ACCESSES so long as it's supported in hardware.  That is,
KVM now advertises VIRTUALIZE_APIC_ACCESSES to a guest but doesn't
(always) allow setting it when kvm-intel.flexpriority=0, and may even
initially allow the control and then clear it when the nested guest
writes APIC_BASE MSR, which is decidedly odd even if it doesn't cause
functional issues.

Hide the control completely when the module parameter is cleared.

reported-by: Sean Christopherson <sean.j.christopherson@intel.com>
Fixes: 8d860bbeedef ("kvm: vmx: Basic APIC virtualization controls have three settings")
Cc: Jim Mattson <jmattson@google.com>
Cc: stable@vger.kernel.org
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
5 years agoKVM: VMX: check for existence of secondary exec controls before accessing
Sean Christopherson [Mon, 1 Oct 2018 21:25:34 +0000 (14:25 -0700)]
KVM: VMX: check for existence of secondary exec controls before accessing

Return early from vmx_set_virtual_apic_mode() if the processor doesn't
support VIRTUALIZE_APIC_ACCESSES or VIRTUALIZE_X2APIC_MODE, both of
which reside in SECONDARY_VM_EXEC_CONTROL.  This eliminates warnings
due to VMWRITEs to SECONDARY_VM_EXEC_CONTROL (VMCS field 401e) failing
on processors without secondary exec controls.

Remove the similar check for TPR shadowing as it is incorporated in the
flexpriority_enabled check and the APIC-related code in
vmx_update_msr_bitmap() is further gated by VIRTUALIZE_X2APIC_MODE.

Reported-by: Gerhard Wiesinger <redhat@wiesinger.com>
Fixes: 8d860bbeedef ("kvm: vmx: Basic APIC virtualization controls have three settings")
Cc: Jim Mattson <jmattson@google.com>
Cc: stable@vger.kernel.org
Signed-off-by: Sean Christopherson <sean.j.christopherson@intel.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
5 years agox86/vdso: Fix vDSO syscall fallback asm constraint regression
Andy Lutomirski [Wed, 3 Oct 2018 23:23:49 +0000 (16:23 -0700)]
x86/vdso: Fix vDSO syscall fallback asm constraint regression

When I added the missing memory outputs, I failed to update the
index of the first argument (ebx) on 32-bit builds, which broke the
fallbacks.  Somehow I must have screwed up my testing or gotten
lucky.

Add another test to cover gettimeofday() as well.

Signed-off-by: Andy Lutomirski <luto@kernel.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: stable@vger.kernel.org
Fixes: 715bd9d12f84 ("x86/vdso: Fix asm constraints on vDSO syscall fallbacks")
Link: http://lkml.kernel.org/r/21bd45ab04b6d838278fa5bebfa9163eceffa13c.1538608971.git.luto@kernel.org
Signed-off-by: Ingo Molnar <mingo@kernel.org>
5 years agoALSA: hda/realtek - Cannot adjust speaker's volume on Dell XPS 27 7760
Kai-Heng Feng [Thu, 4 Oct 2018 03:39:42 +0000 (11:39 +0800)]
ALSA: hda/realtek - Cannot adjust speaker's volume on Dell XPS 27 7760

The issue is the same as commit dd9aa335c880 ("ALSA: hda/realtek - Can't
adjust speaker's volume on a Dell AIO"), the output requires to connect
to a node with Amp-out capability.

Applying the same fixup ALC298_FIXUP_SPK_VOLUME can fix the issue.

BugLink: https://bugs.launchpad.net/bugs/1775068
Signed-off-by: Kai-Heng Feng <kai.heng.feng@canonical.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Takashi Iwai <tiwai@suse.de>
5 years agoKVM: PPC: Book3S HV: Avoid crash from THP collapse during radix page fault
Paul Mackerras [Thu, 4 Oct 2018 04:51:11 +0000 (14:51 +1000)]
KVM: PPC: Book3S HV: Avoid crash from THP collapse during radix page fault

Commit 71d29f43b633 ("KVM: PPC: Book3S HV: Don't use compound_order to
determine host mapping size", 2018-09-11) added a call to
__find_linux_pte() and a dereference of the returned PTE pointer to the
radix page fault path in the common case where the page is normal
system memory.  Previously, __find_linux_pte() was only called for
mappings to physical addresses which don't have a page struct (e.g.
memory-mapped I/O) or where the page struct is marked as reserved
memory.

This exposes us to the possibility that the returned PTE pointer
could be NULL, for example in the case of a concurrent THP collapse
operation.  Dereferencing the returned NULL pointer causes a host
crash.

To fix this, we check for NULL, and if it is NULL, we retry the
operation by returning to the guest, with the expectation that it
will generate the same page fault again (unless of course it has
been fixed up by another CPU in the meantime).

Fixes: 71d29f43b633 ("KVM: PPC: Book3S HV: Don't use compound_order to determine host mapping size")
Signed-off-by: Paul Mackerras <paulus@ozlabs.org>
5 years agoMerge branch 'drm-tda9950-fixes' of git://git.armlinux.org.uk/~rmk/linux-arm into...
Dave Airlie [Thu, 4 Oct 2018 00:28:27 +0000 (10:28 +1000)]
Merge branch 'drm-tda9950-fixes' of git://git.armlinux.org.uk/~rmk/linux-arm into drm-fixes

two tda9950 fixes.

Signed-off-by: Dave Airlie <airlied@redhat.com>
From: Russell King <rmk@armlinux.org.uk>
Link: https://patchwork.freedesktop.org/patch/msgid/20181001162948.GA9508@rmk-PC.armlinux.org.uk
5 years agoMerge tag 'drm-intel-fixes-2018-10-03' of git://anongit.freedesktop.org/drm/drm-intel...
Dave Airlie [Thu, 4 Oct 2018 00:04:38 +0000 (10:04 +1000)]
Merge tag 'drm-intel-fixes-2018-10-03' of git://anongit.freedesktop.org/drm/drm-intel into drm-fixes

There's one fix for our zlib incomlete Z_FINISH on our error state handling,
plus a compilation warning fix and a tiny code clean up.

Signed-off-by: Dave Airlie <airlied@redhat.com>
From: Rodrigo Vivi <rodrigo.vivi@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20181003202840.GA23560@intel.com
5 years agoMerge gitolite.kernel.org:/pub/scm/linux/kernel/git/davem/net
Greg Kroah-Hartman [Wed, 3 Oct 2018 23:09:11 +0000 (16:09 -0700)]
Merge gitolite./pub/scm/linux/kernel/git/davem/net

David writes:
  "Networking fixes:
   1) Prefix length validation in xfrm layer, from Steffen Klassert.

   2) TX status reporting fix in mac80211, from Andrei Otcheretianski.

   3) Fix hangs due to TX_DROP in mac80211, from Bob Copeland.

   4) Fix DMA error regression in b43, from Larry Finger.

   5) Add input validation to xenvif_set_hash_mapping(), from Jan Beulich.

   6) SMMU unmapping fix in hns driver, from Yunsheng Lin.

   7) Bluetooh crash in unpairing on SMP, from Matias Karhumaa.

   8) WoL handling fixes in the phy layer, from Heiner Kallweit.

   9) Fix deadlock in bonding, from Mahesh Bandewar.

   10) Fill ttl inherit infor in vxlan driver, from Hangbin Liu.

   11) Fix TX timeouts during netpoll, from Michael Chan.

   12) RXRPC layer fixes from David Howells.

   13) Another batch of ndo_poll_controller() removals to deal with
       excessive resource consumption during load.  From Eric Dumazet.

   14) Fix a specific TIPC failure secnario, from LUU Duc Canh.

   15) Really disable clocks in r8169 during suspend so that low
       power states can actually be reached.

   16) Fix SYN backlog lockdep issue in tcp and dccp, from Eric Dumazet.

   17) Fix RCU locking in netpoll SKB send, which shows up in bonding,
       from Dave Jones.

   18) Fix TX stalls in r8169, from Heiner Kallweit.

   19) Fix locksup in nfp due to control message storms, from Jakub
       Kicinski.

   20) Various rmnet bug fixes from Subash Abhinov Kasiviswanathan and
       Sean Tranchetti.

   21) Fix use after free in ip_cmsg_recv_dstaddr(), from Eric Dumazet."

* gitolite.kernel.org:/pub/scm/linux/kernel/git/davem/net: (122 commits)
  ixgbe: check return value of napi_complete_done()
  sctp: fix fall-through annotation
  r8169: always autoneg on resume
  ipv4: fix use-after-free in ip_cmsg_recv_dstaddr()
  net: qualcomm: rmnet: Fix incorrect allocation flag in receive path
  net: qualcomm: rmnet: Fix incorrect allocation flag in transmit
  net: qualcomm: rmnet: Skip processing loopback packets
  net: systemport: Fix wake-up interrupt race during resume
  rtnl: limit IFLA_NUM_TX_QUEUES and IFLA_NUM_RX_QUEUES to 4096
  bonding: fix warning message
  inet: make sure to grab rcu_read_lock before using ireq->ireq_opt
  nfp: avoid soft lockups under control message storm
  declance: Fix continuation with the adapter identification message
  net: fec: fix rare tx timeout
  r8169: fix network stalls due to missing bit TXCFG_AUTO_FIFO
  tun: napi flags belong to tfile
  tun: initialize napi_mutex unconditionally
  tun: remove unused parameters
  bond: take rcu lock in netpoll_send_skb_on_dev
  rtnetlink: Fail dump if target netnsid is invalid
  ...

5 years agoixgbe: check return value of napi_complete_done()
Song Liu [Wed, 3 Oct 2018 18:30:35 +0000 (11:30 -0700)]
ixgbe: check return value of napi_complete_done()

The NIC driver should only enable interrupts when napi_complete_done()
returns true. This patch adds the check for ixgbe.

Cc: stable@vger.kernel.org # 4.10+
Suggested-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: Song Liu <songliubraving@fb.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoMerge tag 'linux-kselftest-4.19-rc7' of git://git.kernel.org/pub/scm/linux/kernel...
Greg Kroah-Hartman [Wed, 3 Oct 2018 18:06:49 +0000 (11:06 -0700)]
Merge tag 'linux-kselftest-4.19-rc7' of git://git./linux/kernel/git/shuah/linux-kselftest

Shuah writes:
  "kselftest fixes for 4.19-rc7

   This fixes update for 4.19-rc7 consists one fix to rseq test to
   prevent it from seg-faulting when compiled with -fpie."

* tag 'linux-kselftest-4.19-rc7' of git://git.kernel.org/pub/scm/linux/kernel/git/shuah/linux-kselftest:
  rseq/selftests: fix parametrized test with -fpie

5 years agosctp: fix fall-through annotation
Gustavo A. R. Silva [Wed, 3 Oct 2018 10:45:56 +0000 (12:45 +0200)]
sctp: fix fall-through annotation

Replace "fallthru" with a proper "fall through" annotation.

This fix is part of the ongoing efforts to enabling
-Wimplicit-fallthrough

Signed-off-by: Gustavo A. R. Silva <gustavo@embeddedor.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agodrm/i915: Handle incomplete Z_FINISH for compressed error states
Chris Wilson [Wed, 3 Oct 2018 08:24:22 +0000 (09:24 +0100)]
drm/i915: Handle incomplete Z_FINISH for compressed error states

The final call to zlib_deflate(Z_FINISH) may require more output
space to be allocated and so needs to re-invoked. Failure to do so in
the current code leads to incomplete zlib streams (albeit intact due to
the use of Z_SYNC_FLUSH) resulting in the occasional short object
capture.

v2: Check against overrunning our pre-allocated page array
v3: Drop Z_SYNC_FLUSH entirely

Testcase: igt/i915-error-capture.js
Fixes: 0a97015d45ee ("drm/i915: Compress GPU objects in error state")
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Cc: <stable@vger.kernel.org> # v4.10+
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20181003082422.23214-1-chris@chris-wilson.co.uk
(cherry picked from commit 83bc0f5b432f60394466deef16fc753e27371d0b)
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
5 years agoMerge tag 'media/v4.19-3' of git://git.kernel.org/pub/scm/linux/kernel/git/mchehab...
Greg Kroah-Hartman [Wed, 3 Oct 2018 11:23:46 +0000 (04:23 -0700)]
Merge tag 'media/v4.19-3' of git://git./linux/kernel/git/mchehab/linux-media

Mauro writes:
  "media fixes for v4.19-rc6"

* tag 'media/v4.19-3' of git://git.kernel.org/pub/scm/linux/kernel/git/mchehab/linux-media:
  media: v4l: event: Prevent freeing event subscriptions while accessed

5 years agoMerge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jikos/hid
Greg Kroah-Hartman [Wed, 3 Oct 2018 11:22:30 +0000 (04:22 -0700)]
Merge branch 'for-linus' of git://git./linux/kernel/git/jikos/hid

Jiri writes:
  "HID fixes:
   - hantick touchpad fix from Anisse Astier
   - device ID addition for Ice Lake mobile from Srinivas Pandruvada
   - touchscreen resume fix for certain i2c-hid driven devices from Hans
     de Goede"

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jikos/hid:
  HID: intel-ish-hid: Enable Ice Lake mobile
  HID: i2c-hid: Remove RESEND_REPORT_DESCR quirk and its handling
  HID: i2c-hid: disable runtime PM operations on hantick touchpad

5 years agoMerge branch 'fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/vfs
Greg Kroah-Hartman [Wed, 3 Oct 2018 11:21:23 +0000 (04:21 -0700)]
Merge branch 'fixes' of git://git./linux/kernel/git/viro/vfs

Al writes:
  "xattrs regression fix from Andreas; sat in -next for quite a while."

* 'fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/vfs:
  sysfs: Do not return POSIX ACL xattrs via listxattr

5 years agomedia: v4l: event: Prevent freeing event subscriptions while accessed
Sakari Ailus [Tue, 11 Sep 2018 09:32:37 +0000 (05:32 -0400)]
media: v4l: event: Prevent freeing event subscriptions while accessed

The event subscriptions are added to the subscribed event list while
holding a spinlock, but that lock is subsequently released while still
accessing the subscription object. This makes it possible to unsubscribe
the event --- and freeing the subscription object's memory --- while
the subscription object is simultaneously accessed.

Prevent this by adding a mutex to serialise the event subscription and
unsubscription. This also gives a guarantee to the callback ops that the
add op has returned before the del op is called.

This change also results in making the elems field less special:
subscriptions are only added to the event list once they are fully
initialised.

Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Reviewed-by: Hans Verkuil <hans.verkuil@cisco.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Cc: stable@vger.kernel.org # for 4.14 and up
Fixes: c3b5b0241f62 ("V4L/DVB: V4L: Events: Add backend")
Signed-off-by: Mauro Carvalho Chehab <mchehab+samsung@kernel.org>
5 years agolocking/ww_mutex: Fix runtime warning in the WW mutex selftest
Guenter Roeck [Tue, 2 Oct 2018 21:48:49 +0000 (14:48 -0700)]
locking/ww_mutex: Fix runtime warning in the WW mutex selftest

If CONFIG_WW_MUTEX_SELFTEST=y is enabled, booting an image
in an arm64 virtual machine results in the following
traceback if 8 CPUs are enabled:

  DEBUG_LOCKS_WARN_ON(__owner_task(owner) != current)
  WARNING: CPU: 2 PID: 537 at kernel/locking/mutex.c:1033 __mutex_unlock_slowpath+0x1a8/0x2e0
  ...
  Call trace:
   __mutex_unlock_slowpath()
   ww_mutex_unlock()
   test_cycle_work()
   process_one_work()
   worker_thread()
   kthread()
   ret_from_fork()

If requesting b_mutex fails with -EDEADLK, the error variable
is reassigned to the return value from calling ww_mutex_lock
on a_mutex again. If this call fails, a_mutex is not locked.
It is, however, unconditionally unlocked subsequently, causing
the reported warning. Fix the problem by using two error variables.

With this change, the selftest still fails as follows:

  cyclic deadlock not resolved, ret[7/8] = -35

However, the traceback is gone.

Signed-off-by: Guenter Roeck <linux@roeck-us.net>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Will Deacon <will.deacon@arm.com>
Fixes: d1b42b800e5d0 ("locking/ww_mutex: Add kselftests for resolving ww_mutex cyclic deadlocks")
Link: http://lkml.kernel.org/r/1538516929-9734-1-git-send-email-linux@roeck-us.net
Signed-off-by: Ingo Molnar <mingo@kernel.org>
5 years agoMerge tag 'exynos-drm-fixes-for-v4.19-rc7' of git://git.kernel.org/pub/scm/linux...
Dave Airlie [Wed, 3 Oct 2018 06:31:11 +0000 (16:31 +1000)]
Merge tag 'exynos-drm-fixes-for-v4.19-rc7' of git://git./linux/kernel/git/daeinki/drm-exynos into drm-fixes

Use default iommu domain instead of fake one
- This patch makes it to reuse default IOMMU domain instead of
  allocating a fake IOMMU domain, and allows some design changes
  for enhancement of IOMMU framework[1] without breaking Exynos DRM.

[1] https://www.spinics.net/lists/arm-kernel/msg676098.html

Signed-off-by: Dave Airlie <airlied@redhat.com>
From: Inki Dae <inki.dae@samsung.com>
Link: https://patchwork.freedesktop.org/patch/msgid/1538360696-23579-1-git-send-email-inki.dae@samsung.com
5 years agox86/cpu/amd: Remove unnecessary parentheses
Nathan Chancellor [Tue, 2 Oct 2018 22:45:11 +0000 (15:45 -0700)]
x86/cpu/amd: Remove unnecessary parentheses

Clang warns when multiple pairs of parentheses are used for a single
conditional statement.

arch/x86/kernel/cpu/amd.c:925:14: warning: equality comparison with
extraneous parentheses [-Wparentheses-equality]
        if ((c->x86 == 6)) {
             ~~~~~~~^~~~
arch/x86/kernel/cpu/amd.c:925:14: note: remove extraneous parentheses
around the comparison to silence this warning
        if ((c->x86 == 6)) {
            ~       ^   ~
arch/x86/kernel/cpu/amd.c:925:14: note: use '=' to turn this equality
comparison into an assignment
        if ((c->x86 == 6)) {
                    ^~
                    =
1 warning generated.

Signed-off-by: Nathan Chancellor <natechancellor@gmail.com>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Link: http://lkml.kernel.org/r/20181002224511.14929-1-natechancellor@gmail.com
Link: https://github.com/ClangBuiltLinux/linux/issues/187
Signed-off-by: Ingo Molnar <mingo@kernel.org>
5 years agox86/vdso: Only enable vDSO retpolines when enabled and supported
Andy Lutomirski [Wed, 3 Oct 2018 04:26:50 +0000 (21:26 -0700)]
x86/vdso: Only enable vDSO retpolines when enabled and supported

When I fixed the vDSO build to use inline retpolines, I messed up
the Makefile logic and made it unconditional.  It should have
depended on CONFIG_RETPOLINE and on the availability of compiler
support.  This broke the build on some older compilers.

Reported-by: nikola.ciprich@linuxbox.cz
Signed-off-by: Andy Lutomirski <luto@kernel.org>
Cc: Borislav Petkov <bp@alien8.de>
Cc: David Woodhouse <dwmw2@infradead.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Matt Rickard <matt@softrans.com.au>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: jason.vas.dias@gmail.com
Cc: stable@vger.kernel.org
Fixes: 2e549b2ee0e3 ("x86/vdso: Fix vDSO build if a retpoline is emitted")
Link: http://lkml.kernel.org/r/08a1f29f2c238dd1f493945e702a521f8a5aa3ae.1538540801.git.luto@kernel.org
Signed-off-by: Ingo Molnar <mingo@kernel.org>
5 years agor8169: always autoneg on resume
Alex Xu (Hello71) [Sun, 30 Sep 2018 15:06:39 +0000 (11:06 -0400)]
r8169: always autoneg on resume

This affects at least versions 25 and 33, so assume all cards are broken
and just renegotiate by default.

Fixes: 10bc6a6042c9 ("r8169: fix autoneg issue on resume with RTL8168E")
Signed-off-by: Alex Xu (Hello71) <alex_y_xu@yahoo.ca>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoipv4: fix use-after-free in ip_cmsg_recv_dstaddr()
Eric Dumazet [Sun, 30 Sep 2018 18:33:39 +0000 (11:33 -0700)]
ipv4: fix use-after-free in ip_cmsg_recv_dstaddr()

Caching ip_hdr(skb) before a call to pskb_may_pull() is buggy,
do not do it.

Fixes: 2efd4fca703a ("ip: in cmsg IP(V6)_ORIGDSTADDR call pskb_may_pull")
Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Willem de Bruijn <willemb@google.com>
Reported-by: syzbot <syzkaller@googlegroups.com>
Acked-by: Willem de Bruijn <willemb@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoMerge tag 'mlx5-fixes-2018-10-01' of git://git.kernel.org/pub/scm/linux/kernel/git...
David S. Miller [Wed, 3 Oct 2018 05:20:24 +0000 (22:20 -0700)]
Merge tag 'mlx5-fixes-2018-10-01' of git://git./linux/kernel/git/saeed/linux

Saeed Mahameed says:

====================
Mellanox, mlx5 fixes 2018-10-01

This pull request includes some fixes to mlx5 driver,
Please pull and let me know if there's any problem.

For -stable v4.11:
"6e0a4a23c59a ('net/mlx5: E-Switch, Fix out of bound access when setting vport rate')"

For -stable v4.18:
"98d6627c372a ('net/mlx5e: Set vlan masks for all offloaded TC rules')"
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoMerge branch 'rmnet-fixes'
David S. Miller [Wed, 3 Oct 2018 05:16:00 +0000 (22:16 -0700)]
Merge branch 'rmnet-fixes'

Subash Abhinov Kasiviswanathan says:

====================
net: qualcomm: rmnet: Updates 2018-10-02

This series is a set of small fixes for rmnet driver

Patch 1 is a fix for a scenario reported by syzkaller
Patch 2 & 3 are fixes for incorrect allocation flags
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: qualcomm: rmnet: Fix incorrect allocation flag in receive path
Subash Abhinov Kasiviswanathan [Wed, 3 Oct 2018 00:52:03 +0000 (18:52 -0600)]
net: qualcomm: rmnet: Fix incorrect allocation flag in receive path

The incoming skb needs to be reallocated in case the headroom
is not sufficient to adjust the ethernet header. This allocation
needs to be atomic otherwise it results in this splat

 [<600601bb>] ___might_sleep+0x185/0x1a3
 [<603f6314>] ? _raw_spin_unlock_irqrestore+0x0/0x27
 [<60069bb0>] ? __wake_up_common_lock+0x95/0xd1
 [<600602b0>] __might_sleep+0xd7/0xe2
 [<60065598>] ? enqueue_task_fair+0x112/0x209
 [<600eea13>] __kmalloc_track_caller+0x5d/0x124
 [<600ee9b6>] ? __kmalloc_track_caller+0x0/0x124
 [<602696d5>] __kmalloc_reserve.isra.34+0x30/0x7e
 [<603f629b>] ? _raw_spin_lock_irqsave+0x0/0x3d
 [<6026b744>] pskb_expand_head+0xbf/0x310
 [<6025ca6a>] rmnet_rx_handler+0x7e/0x16b
 [<6025c9ec>] ? rmnet_rx_handler+0x0/0x16b
 [<6027ad0c>] __netif_receive_skb_core+0x301/0x96f
 [<60033c17>] ? set_signals+0x0/0x40
 [<6027bbcb>] __netif_receive_skb+0x24/0x8e

Fixes: 74692caf1b0b ("net: qualcomm: rmnet: Process packets over ethernet")
Signed-off-by: Sean Tranchetti <stranche@codeaurora.org>
Signed-off-by: Subash Abhinov Kasiviswanathan <subashab@codeaurora.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: qualcomm: rmnet: Fix incorrect allocation flag in transmit
Subash Abhinov Kasiviswanathan [Wed, 3 Oct 2018 00:52:02 +0000 (18:52 -0600)]
net: qualcomm: rmnet: Fix incorrect allocation flag in transmit

The incoming skb needs to be reallocated in case the headroom
is not sufficient to add the MAP header. This allocation needs to
be atomic otherwise it results in the following splat

[32805.801456] BUG: sleeping function called from invalid context
[32805.841141] Internal error: Oops - BUG: 0 [#1] PREEMPT SMP
[32805.904773] task: ffffffd7c5f62280 task.stack: ffffff80464a8000
[32805.910851] pc : ___might_sleep+0x180/0x188
[32805.915143] lr : ___might_sleep+0x180/0x188
[32806.131520] Call trace:
[32806.134041]  ___might_sleep+0x180/0x188
[32806.137980]  __might_sleep+0x50/0x84
[32806.141653]  __kmalloc_track_caller+0x80/0x3bc
[32806.146215]  __kmalloc_reserve+0x3c/0x88
[32806.150241]  pskb_expand_head+0x74/0x288
[32806.154269]  rmnet_egress_handler+0xb0/0x1d8
[32806.162239]  rmnet_vnd_start_xmit+0xc8/0x13c
[32806.166627]  dev_hard_start_xmit+0x148/0x280
[32806.181181]  sch_direct_xmit+0xa4/0x198
[32806.185125]  __qdisc_run+0x1f8/0x310
[32806.188803]  net_tx_action+0x23c/0x26c
[32806.192655]  __do_softirq+0x220/0x408
[32806.196420]  do_softirq+0x4c/0x70

Fixes: ceed73a2cf4a ("drivers: net: ethernet: qualcomm: rmnet: Initial implementation")
Signed-off-by: Subash Abhinov Kasiviswanathan <subashab@codeaurora.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: qualcomm: rmnet: Skip processing loopback packets
Sean Tranchetti [Wed, 3 Oct 2018 00:52:01 +0000 (18:52 -0600)]
net: qualcomm: rmnet: Skip processing loopback packets

RMNET RX handler was processing invalid packets that were
originally sent on the real device and were looped back via
dev_loopback_xmit(). This was detected using syzkaller.

Fixes: ceed73a2cf4a ("drivers: net: ethernet: qualcomm: rmnet: Initial implementation")
Signed-off-by: Sean Tranchetti <stranche@codeaurora.org>
Signed-off-by: Subash Abhinov Kasiviswanathan <subashab@codeaurora.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: systemport: Fix wake-up interrupt race during resume
Florian Fainelli [Tue, 2 Oct 2018 23:52:03 +0000 (16:52 -0700)]
net: systemport: Fix wake-up interrupt race during resume

The AON_PM_L2 is normally used to trigger and identify the source of a
wake-up event. Since the RX_SYS clock is no longer turned off, we also
have an interrupt being sent to the SYSTEMPORT INTRL_2_0 controller, and
that interrupt remains active up until the magic packet detector is
disabled which happens much later during the driver resumption.

The race happens if we have a CPU that is entering the SYSTEMPORT
INTRL2_0 handler during resume, and another CPU has managed to clear the
wake-up interrupt during bcm_sysport_resume_from_wol(). In that case, we
have the first CPU stuck in the interrupt handler with an interrupt
cause that has been cleared under its feet, and so we keep returning
IRQ_NONE and we never make any progress.

This was not a problem before because we would always turn off the
RX_SYS clock during WoL, so the SYSTEMPORT INTRL2_0 would also be turned
off as well, thus not latching the interrupt.

The fix is to make sure we do not enable either the MPD or
BRCM_TAG_MATCH interrupts since those are redundant with what the
AON_PM_L2 interrupt controller already processes and they would cause
such a race to occur.

Fixes: bb9051a2b230 ("net: systemport: Add support for WAKE_FILTER")
Fixes: 83e82f4c706b ("net: systemport: add Wake-on-LAN support")
Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agosmb3: fix lease break problem introduced by compounding
Steve French [Tue, 2 Oct 2018 23:54:09 +0000 (18:54 -0500)]
smb3: fix lease break problem introduced by compounding

Fixes problem (discovered by Aurelien) introduced by recent commit:
commit b24df3e30cbf48255db866720fb71f14bf9d2f39
("cifs: update receive_encrypted_standard to handle compounded responses")

which broke the ability to respond to some lease breaks
(lease breaks being ignored is a problem since can block
server response for duration of the lease break timeout).

Signed-off-by: Steve French <stfrench@microsoft.com>
Reviewed-by: Ronnie Sahlberg <lsahlber@redhat.com>
Reviewed-by: Pavel Shilovsky <pshilov@microsoft.com>
5 years agocifs: only wake the thread for the very last PDU in a compound
Ronnie Sahlberg [Thu, 30 Aug 2018 00:13:00 +0000 (10:13 +1000)]
cifs: only wake the thread for the very last PDU in a compound

For compounded PDUs we whould only wake the waiting thread for the
very last PDU of the compound.
We do this so that we are guaranteed that the demultiplex_thread will
not process or access any of those MIDs any more once the send/recv
thread starts processing.

Else there is a race where at the end of the send/recv processing we
will try to delete all the mids of the compound. If the multiplex
thread still has other mids to process at this point for this compound
this can lead to an oops.

Needed to fix recent commit:
commit 730928c8f4be88e9d6a027a16b1e8fa9c59fc077
("cifs: update smb2_queryfs() to use compounding")

Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com>
Signed-off-by: Steve French <stfrench@microsoft.com>
Reviewed-by: Pavel Shilovsky <pshilov@microsoft.com>
5 years agoMerge tag 'wireless-drivers-for-davem-2018-10-01' of git://git.kernel.org/pub/scm...
David S. Miller [Tue, 2 Oct 2018 23:16:59 +0000 (16:16 -0700)]
Merge tag 'wireless-drivers-for-davem-2018-10-01' of git://git./linux/kernel/git/kvalo/wireless-drivers

Kalle Valo says:

====================
wireless-drivers fixes for 4.19

First, and also hopefully the last, set of fixes for 4.19. All small
but still important fixes

mt76x0

* fix a bug when a virtual interface is removed multiple times

b43

* fix DMA error related regression with proprietary firmware

iwlwifi

* fix an oops which was a regression in v4.19-rc1
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agocifs: add a warning if we try to to dequeue a deleted mid
Ronnie Sahlberg [Thu, 30 Aug 2018 00:12:59 +0000 (10:12 +1000)]
cifs: add a warning if we try to to dequeue a deleted mid

cifs_delete_mid() is called once we are finished handling a mid and we
expect no more work done on this mid.

Needed to fix recent commit:
commit 730928c8f4be88e9d6a027a16b1e8fa9c59fc077
("cifs: update smb2_queryfs() to use compounding")

Add a warning if someone tries to dequeue a mid that has already been
flagged to be deleted.
Also change list_del() to list_del_init() so that if we have similar bugs
resurface in the future we will not oops.

Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com>
Signed-off-by: Steve French <stfrench@microsoft.com>
Reviewed-by: Pavel Shilovsky <pshilov@microsoft.com>
5 years agosmb2: fix missing files in root share directory listing
Aurelien Aptel [Thu, 17 May 2018 14:35:07 +0000 (16:35 +0200)]
smb2: fix missing files in root share directory listing

When mounting a Windows share that is the root of a drive (eg. C$)
the server does not return . and .. directory entries. This results in
the smb2 code path erroneously skipping the 2 first entries.

Pseudo-code of the readdir() code path:

cifs_readdir(struct file, struct dir_context)
    initiate_cifs_search            <-- if no reponse cached yet
        server->ops->query_dir_first

    dir_emit_dots
        dir_emit                    <-- adds "." and ".." if we're at pos=0

    find_cifs_entry
        initiate_cifs_search        <-- if pos < start of current response
                                         (restart search)
        server->ops->query_dir_next <-- if pos > end of current response
                                         (fetch next search res)

    for(...)                        <-- loops over cur response entries
                                          starting at pos
        cifs_filldir                <-- skip . and .., emit entry
            cifs_fill_dirent
            dir_emit
pos++

A) dir_emit_dots() always adds . & ..
   and sets the current dir pos to 2 (0 and 1 are done).

Therefore we always want the index_to_find to be 2 regardless of if
the response has . and ..

B) smb1 code initializes index_of_last_entry with a +2 offset

  in cifssmb.c CIFSFindFirst():
psrch_inf->index_of_last_entry = 2 /* skip . and .. */ +
psrch_inf->entries_in_buffer;

Later in find_cifs_entry() we want to find the next dir entry at pos=2
as a result of (A)

first_entry_in_buffer = cfile->srch_inf.index_of_last_entry -
cfile->srch_inf.entries_in_buffer;

This var is the dir pos that the first entry in the buffer will
have therefore it must be 2 in the first call.

If we don't offset index_of_last_entry by 2 (like in (B)),
first_entry_in_buffer=0 but we were instructed to get pos=2 so this
code in find_cifs_entry() skips the 2 first which is ok for non-root
shares, as it skips . and .. from the response but is not ok for root
shares where the 2 first are actual files

pos_in_buf = index_to_find - first_entry_in_buffer;
                // pos_in_buf=2
// we skip 2 first response entries :(
for (i = 0; (i < (pos_in_buf)) && (cur_ent != NULL); i++) {
/* go entry by entry figuring out which is first */
cur_ent = nxt_dir_entry(cur_ent, end_of_smb,
cfile->srch_inf.info_level);
}

C) cifs_filldir() skips . and .. so we can safely ignore them for now.

Sample program:

int main(int argc, char **argv)
{
const char *path = argc >= 2 ? argv[1] : ".";
DIR *dh;
struct dirent *de;

printf("listing path <%s>\n", path);
dh = opendir(path);
if (!dh) {
printf("opendir error %d\n", errno);
return 1;
}

while (1) {
de = readdir(dh);
if (!de) {
if (errno) {
printf("readdir error %d\n", errno);
return 1;
}
printf("end of listing\n");
break;
}
printf("off=%lu <%s>\n", de->d_off, de->d_name);
}

return 0;
}

Before the fix with SMB1 on root shares:

<.>            off=1
<..>           off=2
<$Recycle.Bin> off=3
<bootmgr>      off=4

and on non-root shares:

<.>    off=1
<..>   off=4  <-- after adding .., the offsets jumps to +2 because
<2536> off=5       we skipped . and .. from response buffer (C)
<411>  off=6       but still incremented pos
<file> off=7
<fsx>  off=8

Therefore the fix for smb2 is to mimic smb1 behaviour and offset the
index_of_last_entry by 2.

Test results comparing smb1 and smb2 before/after the fix on root
share, non-root shares and on large directories (ie. multi-response
dir listing):

PRE FIX
=======
pre-1-root VS pre-2-root:
        ERR pre-2-root is missing [bootmgr, $Recycle.Bin]
pre-1-nonroot VS pre-2-nonroot:
        OK~ same files, same order, different offsets
pre-1-nonroot-large VS pre-2-nonroot-large:
        OK~ same files, same order, different offsets

POST FIX
========
post-1-root VS post-2-root:
        OK same files, same order, same offsets
post-1-nonroot VS post-2-nonroot:
        OK same files, same order, same offsets
post-1-nonroot-large VS post-2-nonroot-large:
        OK same files, same order, same offsets

REGRESSION?
===========
pre-1-root VS post-1-root:
        OK same files, same order, same offsets
pre-1-nonroot VS post-1-nonroot:
        OK same files, same order, same offsets

BugLink: https://bugzilla.samba.org/show_bug.cgi?id=13107
Signed-off-by: Aurelien Aptel <aaptel@suse.com>
Signed-off-by: Paulo Alcantara <palcantara@suse.deR>
Reviewed-by: Ronnie Sahlberg <lsahlber@redhat.com>
Signed-off-by: Steve French <stfrench@microsoft.com>
CC: Stable <stable@vger.kernel.org>
5 years agortnl: limit IFLA_NUM_TX_QUEUES and IFLA_NUM_RX_QUEUES to 4096
Eric Dumazet [Tue, 2 Oct 2018 22:47:35 +0000 (15:47 -0700)]
rtnl: limit IFLA_NUM_TX_QUEUES and IFLA_NUM_RX_QUEUES to 4096

We have an impressive number of syzkaller bugs that are linked
to the fact that syzbot was able to create a networking device
with millions of TX (or RX) queues.

Let's limit the number of RX/TX queues to 4096, this really should
cover all known cases.

A separate patch will add various cond_resched() in the loops
handling sysfs entries at device creation and dismantle.

Tested:

lpaa6:~# ip link add gre-4097 numtxqueues 4097 numrxqueues 4097 type ip6gretap
RTNETLINK answers: Invalid argument

lpaa6:~# time ip link add gre-4096 numtxqueues 4096 numrxqueues 4096 type ip6gretap

real 0m0.180s
user 0m0.000s
sys 0m0.107s

Fixes: 76ff5cc91935 ("rtnl: allow to specify number of rx and tx queues on device creation")
Signed-off-by: Eric Dumazet <edumazet@google.com>
Reported-by: syzbot <syzkaller@googlegroups.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agobonding: fix warning message
Mahesh Bandewar [Tue, 2 Oct 2018 19:14:34 +0000 (12:14 -0700)]
bonding: fix warning message

RX queue config for bonding master could be different from its slave
device(s). With the commit 6a9e461f6fe4 ("bonding: pass link-local
packets to bonding master also."), the packet is reinjected into stack
with skb->dev as bonding master. This potentially triggers the
message:

   "bondX received packet on queue Y, but number of RX queues is Z"

whenever the queue that packet is received on is higher than the
numrxqueues on bonding master (Y > Z).

Fixes: 6a9e461f6fe4 ("bonding: pass link-local packets to bonding master also.")
Reported-by: John Sperbeck <jsperbeck@google.com>
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: Mahesh Bandewar <maheshb@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoinet: make sure to grab rcu_read_lock before using ireq->ireq_opt
Eric Dumazet [Tue, 2 Oct 2018 19:35:05 +0000 (12:35 -0700)]
inet: make sure to grab rcu_read_lock before using ireq->ireq_opt

Timer handlers do not imply rcu_read_lock(), so my recent fix
triggered a LOCKDEP warning when SYNACK is retransmit.

Lets add rcu_read_lock()/rcu_read_unlock() pairs around ireq->ireq_opt
usages instead of guessing what is done by callers, since it is
not worth the pain.

Get rid of ireq_opt_deref() helper since it hides the logic
without real benefit, since it is now a standard rcu_dereference().

Fixes: 1ad98e9d1bdf ("tcp/dccp: fix lockdep issue when SYN is backlogged")
Signed-off-by: Eric Dumazet <edumazet@google.com>
Reported-by: Willem de Bruijn <willemb@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoRISCV: Fix end PFN for low memory
Atish Patra [Tue, 11 Sep 2018 18:30:18 +0000 (11:30 -0700)]
RISCV: Fix end PFN for low memory

Use memblock_end_of_DRAM which provides correct last low memory
PFN. Without that, DMA32 region becomes empty resulting in zero
pages being allocated for DMA32.

This patch is based on earlier patch from palmer which never
merged into 4.19. I just edited the commit text to make more
sense.

Signed-off-by: Atish Patra <atish.patra@wdc.com>
Signed-off-by: Palmer Dabbelt <palmer@sifive.com>
5 years agox86/tsc: Fix UV TSC initialization
Mike Travis [Tue, 2 Oct 2018 18:01:46 +0000 (13:01 -0500)]
x86/tsc: Fix UV TSC initialization

The recent rework of the TSC calibration code introduced a regression on UV
systems as it added a call to tsc_early_init() which initializes the TSC
ADJUST values before acpi_boot_table_init().  In the case of UV systems,
that is a necessary step that calls uv_system_init().  This informs
tsc_sanitize_first_cpu() that the kernel runs on a platform with async TSC
resets as documented in commit 341102c3ef29 ("x86/tsc: Add option that TSC
on Socket 0 being non-zero is valid")

Fix it by skipping the early tsc initialization on UV systems and let TSC
init tests take place later in tsc_init().

Fixes: cf7a63ef4e02 ("x86/tsc: Calibrate tsc only once")
Suggested-by: Hedi Berriche <hedi.berriche@hpe.com>
Signed-off-by: Mike Travis <mike.travis@hpe.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Reviewed-by: Russ Anderson <rja@hpe.com>
Reviewed-by: Dimitri Sivanich <sivanich@hpe.com>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Cc: Russ Anderson <russ.anderson@hpe.com>
Cc: Dimitri Sivanich <dimitri.sivanich@hpe.com>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Kate Stewart <kstewart@linuxfoundation.org>
Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Cc: Philippe Ombredanne <pombredanne@nexb.com>
Cc: Pavel Tatashin <pasha.tatashin@oracle.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Len Brown <len.brown@intel.com>
Cc: Dou Liyang <douly.fnst@cn.fujitsu.com>
Cc: Xiaoming Gao <gxm.linux.kernel@gmail.com>
Cc: Rajvi Jingar <rajvi.jingar@intel.com>
Link: https://lkml.kernel.org/r/20181002180144.923579706@stormcage.americas.sgi.com
5 years agox86/platform/uv: Provide is_early_uv_system()
Mike Travis [Tue, 2 Oct 2018 18:01:45 +0000 (13:01 -0500)]
x86/platform/uv: Provide is_early_uv_system()

Introduce is_early_uv_system() which uses efi.uv_systab to decide early
in the boot process whether the kernel runs on a UV system.

This is needed to skip other early setup/init code that might break
the UV platform if done too early such as before necessary ACPI tables
parsing takes place.

Suggested-by: Hedi Berriche <hedi.berriche@hpe.com>
Signed-off-by: Mike Travis <mike.travis@hpe.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Reviewed-by: Russ Anderson <rja@hpe.com>
Reviewed-by: Dimitri Sivanich <sivanich@hpe.com>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Cc: Russ Anderson <russ.anderson@hpe.com>
Cc: Dimitri Sivanich <dimitri.sivanich@hpe.com>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Kate Stewart <kstewart@linuxfoundation.org>
Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Cc: Philippe Ombredanne <pombredanne@nexb.com>
Cc: Pavel Tatashin <pasha.tatashin@oracle.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Len Brown <len.brown@intel.com>
Cc: Dou Liyang <douly.fnst@cn.fujitsu.com>
Cc: Xiaoming Gao <gxm.linux.kernel@gmail.com>
Cc: Rajvi Jingar <rajvi.jingar@intel.com>
Link: https://lkml.kernel.org/r/20181002180144.801700401@stormcage.americas.sgi.com
5 years agonfp: avoid soft lockups under control message storm
Jakub Kicinski [Tue, 2 Oct 2018 17:10:14 +0000 (10:10 -0700)]
nfp: avoid soft lockups under control message storm

When FW floods the driver with control messages try to exit the cmsg
processing loop every now and then to avoid soft lockups.  Cmsg
processing is generally very lightweight so 512 seems like a reasonable
budget, which should not be exceeded under normal conditions.

Fixes: 77ece8d5f196 ("nfp: add control vNIC datapath")
Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Reviewed-by: Simon Horman <simon.horman@netronome.com>
Tested-by: Pieter Jansen van Vuuren <pieter.jansenvanvuuren@netronome.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agodeclance: Fix continuation with the adapter identification message
Maciej W. Rozycki [Tue, 2 Oct 2018 13:23:45 +0000 (14:23 +0100)]
declance: Fix continuation with the adapter identification message

Fix a commit 4bcc595ccd80 ("printk: reinstate KERN_CONT for printing
continuation lines") regression with the `declance' driver, which caused
the adapter identification message to be split between two lines, e.g.:

declance.c: v0.011 by Linux MIPS DECstation task force
tc6: PMAD-AA
, addr = 08:00:2b:1b:2a:6a, irq = 14
tc6: registered as eth0.

Address that properly, by printing identification with a single call,
making the messages now look like:

declance.c: v0.011 by Linux MIPS DECstation task force
tc6: PMAD-AA, addr = 08:00:2b:1b:2a:6a, irq = 14
tc6: registered as eth0.

Signed-off-by: Maciej W. Rozycki <macro@linux-mips.org>
Fixes: 4bcc595ccd80 ("printk: reinstate KERN_CONT for printing continuation lines")
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: fec: fix rare tx timeout
Rickard x Andersson [Tue, 2 Oct 2018 12:49:32 +0000 (14:49 +0200)]
net: fec: fix rare tx timeout

During certain heavy network loads TX could time out
with TX ring dump.
TX is sometimes never restarted after reaching
"tx_stop_threshold" because function "fec_enet_tx_queue"
only tests the first queue.

In addition the TX timeout callback function failed to
recover because it also operated only on the first queue.

Signed-off-by: Rickard x Andersson <rickaran@axis.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoMerge tag 'fbdev-v4.19-rc7' of https://github.com/bzolnier/linux
Greg Kroah-Hartman [Tue, 2 Oct 2018 12:19:43 +0000 (05:19 -0700)]
Merge tag 'fbdev-v4.19-rc7' of https://github.com/bzolnier/linux

Bartlomiej writes:
  "fbdev fixes for v4.19-rc7:

   - fix OMAPFB_MEMORY_READ ioctl to not leak kernel memory in omapfb driver
     (Tomi Valkeinen)

   - add missing prepare/unprepare clock operations in pxa168fb driver
     (Lubomir Rintel)

   - add nobgrt option in efifb driver to disable ACPI BGRT logo restore
     (Hans de Goede)

   - fix spelling mistake in fall-through annotation in stifb driver
     (Gustavo A. R. Silva)

   - fix URL for uvesafb repository in the documentation (Adam Jackson)"

* tag 'fbdev-v4.19-rc7' of https://github.com/bzolnier/linux:
  video/fbdev/stifb: Fix spelling mistake in fall-through annotation
  uvesafb: Fix URLs in the documentation
  efifb: BGRT: Add nobgrt option
  fbdev/omapfb: fix omapfb_memory_read infoleak
  pxa168fb: prepare the clock

5 years agoMerge tag 'mmc-v4.19-rc4' of git://git.kernel.org/pub/scm/linux/kernel/git/ulfh/mmc
Greg Kroah-Hartman [Tue, 2 Oct 2018 12:19:04 +0000 (05:19 -0700)]
Merge tag 'mmc-v4.19-rc4' of git://git./linux/kernel/git/ulfh/mmc

Ulf writes:
  "MMC core:
    - Fixup conversion of debounce time to/from ms/us

   MMC host:
    - sdhi: Fixup whitelisting for Gen3 types"

* tag 'mmc-v4.19-rc4' of git://git.kernel.org/pub/scm/linux/kernel/git/ulfh/mmc:
  mmc: slot-gpio: Fix debounce time to use miliseconds again
  mmc: core: Fix debounce time to use microseconds
  mmc: sdhi: sys_dmac: check for all Gen3 types when whitelisting

5 years agodrm/cma-helper: Fix crash in fbdev error path
Noralf Trønnes [Mon, 1 Oct 2018 19:45:36 +0000 (21:45 +0200)]
drm/cma-helper: Fix crash in fbdev error path

Sergey Suloev reported a crash happening in drm_client_dev_hotplug()
when fbdev had failed to register.

[    9.124598] vc4_hdmi 3f902000.hdmi: ASoC: Failed to create component debugfs directory
[    9.147667] vc4_hdmi 3f902000.hdmi: vc4-hdmi-hifi <-> 3f902000.hdmi mapping ok
[    9.155184] vc4_hdmi 3f902000.hdmi: ASoC: no DMI vendor name!
[    9.166544] vc4-drm soc:gpu: bound 3f902000.hdmi (ops vc4_hdmi_ops [vc4])
[    9.173840] vc4-drm soc:gpu: bound 3f806000.vec (ops vc4_vec_ops [vc4])
[    9.181029] vc4-drm soc:gpu: bound 3f004000.txp (ops vc4_txp_ops [vc4])
[    9.188519] vc4-drm soc:gpu: bound 3f400000.hvs (ops vc4_hvs_ops [vc4])
[    9.195690] vc4-drm soc:gpu: bound 3f206000.pixelvalve (ops vc4_crtc_ops [vc4])
[    9.203523] vc4-drm soc:gpu: bound 3f207000.pixelvalve (ops vc4_crtc_ops [vc4])
[    9.215032] vc4-drm soc:gpu: bound 3f807000.pixelvalve (ops vc4_crtc_ops [vc4])
[    9.274785] vc4-drm soc:gpu: bound 3fc00000.v3d (ops vc4_v3d_ops [vc4])
[    9.290246] [drm] Initialized vc4 0.0.0 20140616 for soc:gpu on minor 0
[    9.297464] [drm] Supports vblank timestamp caching Rev 2 (21.10.2013).
[    9.304600] [drm] Driver supports precise vblank timestamp query.
[    9.382856] vc4-drm soc:gpu: [drm:drm_fb_helper_fbdev_setup [drm_kms_helper]] *ERROR* Failed to set fbdev configuration
[   10.404937] Unable to handle kernel paging request at virtual address 00330a656369768a
[   10.441620] [00330a656369768a] address between user and kernel address ranges
[   10.449087] Internal error: Oops: 96000004 [#1] PREEMPT SMP
[   10.454762] Modules linked in: brcmfmac vc4 drm_kms_helper cfg80211 drm rfkill smsc95xx brcmutil usbnet drm_panel_orientation_quirks raspberrypi_hwmon bcm2835_dma crc32_ce pwm_bcm2835 bcm2835_rng virt_dma rng_core i2c_bcm2835 ip_tables x_tables ipv6
[   10.477296] CPU: 2 PID: 45 Comm: kworker/2:1 Not tainted 4.19.0-rc5 #3
[   10.483934] Hardware name: Raspberry Pi 3 Model B Rev 1.2 (DT)
[   10.489966] Workqueue: events output_poll_execute [drm_kms_helper]
[   10.596515] Process kworker/2:1 (pid: 45, stack limit = 0x000000007e8924dc)
[   10.603590] Call trace:
[   10.606259]  drm_client_dev_hotplug+0x5c/0xb0 [drm]
[   10.611303]  drm_kms_helper_hotplug_event+0x30/0x40 [drm_kms_helper]
[   10.617849]  output_poll_execute+0xc4/0x1e0 [drm_kms_helper]
[   10.623616]  process_one_work+0x1c8/0x318
[   10.627695]  worker_thread+0x48/0x428
[   10.631420]  kthread+0xf8/0x128
[   10.634615]  ret_from_fork+0x10/0x18
[   10.638255] Code: 54000220 f9401261 aa1303e0 b4000141 (f9400c21)
[   10.644456] ---[ end trace c75b4a4b0e141908 ]---

The reason for this is that drm_fbdev_cma_init() removes the drm_client
when fbdev registration fails, but it doesn't remove the client from the
drm_device client list. So the client list now has a pointer that points
into the unknown and we have a 'use after free' situation.

Split drm_client_new() into drm_client_init() and drm_client_add() to fix
removal in the error path.

Fixes: 894a677f4b3e ("drm/cma-helper: Use the generic fbdev emulation")
Reported-by: Sergey Suloev <ssuloev@orpaltech.com>
Cc: Stefan Wahren <stefan.wahren@i2se.com>
Cc: Eric Anholt <eric@anholt.net>
Cc: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Noralf Trønnes <noralf@tronnes.org>
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Link: https://patchwork.freedesktop.org/patch/msgid/20181001194536.57756-1-noralf@tronnes.org
5 years agosched/numa: Migrate pages to local nodes quicker early in the lifetime of a task
Mel Gorman [Mon, 1 Oct 2018 10:05:25 +0000 (11:05 +0100)]
sched/numa: Migrate pages to local nodes quicker early in the lifetime of a task

Automatic NUMA Balancing uses a multi-stage pass to decide whether a page
should migrate to a local node. This filter avoids excessive ping-ponging
if a page is shared or used by threads that migrate cross-node frequently.

Threads inherit both page tables and the preferred node ID from the
parent. This means that threads can trigger hinting faults earlier than
a new task which delays scanning for a number of seconds. As it can be
load balanced very early in its lifetime there can be an unnecessary delay
before it starts migrating thread-local data. This patch migrates private
pages faster early in the lifetime of a thread using the sequence counter
as an identifier of new tasks.

With this patch applied, STREAM performance is the same as 4.17 even though
processes are not spread cross-node prematurely. Other workloads showed
a mix of minor gains and losses. This is somewhat expected most workloads
are not very sensitive to the starting conditions of a process.

                         4.19.0-rc5             4.19.0-rc5                 4.17.0
                         numab-v1r1       fastmigrate-v1r1                vanilla
MB/sec copy     43298.52 (   0.00%)    47335.46 (   9.32%)    47219.24 (   9.06%)
MB/sec scale    30115.06 (   0.00%)    32568.12 (   8.15%)    32527.56 (   8.01%)
MB/sec add      32825.12 (   0.00%)    36078.94 (   9.91%)    35928.02 (   9.45%)
MB/sec triad    32549.52 (   0.00%)    35935.94 (  10.40%)    35969.88 (  10.51%)

Signed-off-by: Mel Gorman <mgorman@techsingularity.net>
Reviewed-by: Rik van Riel <riel@surriel.com>
Acked-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Jirka Hladky <jhladky@redhat.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Linux-MM <linux-mm@kvack.org>
Cc: Srikar Dronamraju <srikar@linux.vnet.ibm.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Link: http://lkml.kernel.org/r/20181001100525.29789-3-mgorman@techsingularity.net
Signed-off-by: Ingo Molnar <mingo@kernel.org>
5 years agomm, sched/numa: Remove rate-limiting of automatic NUMA balancing migration
Mel Gorman [Mon, 1 Oct 2018 10:05:24 +0000 (11:05 +0100)]
mm, sched/numa: Remove rate-limiting of automatic NUMA balancing migration

Rate limiting of page migrations due to automatic NUMA balancing was
introduced to mitigate the worst-case scenario of migrating at high
frequency due to false sharing or slowly ping-ponging between nodes.
Since then, a lot of effort was spent on correctly identifying these
pages and avoiding unnecessary migrations and the safety net may no longer
be required.

Jirka Hladky reported a regression in 4.17 due to a scheduler patch that
avoids spreading STREAM tasks wide prematurely. However, once the task
was properly placed, it delayed migrating the memory due to rate limiting.
Increasing the limit fixed the problem for him.

Currently, the limit is hard-coded and does not account for the real
capabilities of the hardware. Even if an estimate was attempted, it would
not properly account for the number of memory controllers and it could
not account for the amount of bandwidth used for normal accesses. Rather
than fudging, this patch simply eliminates the rate limiting.

However, Jirka reports that a STREAM configuration using multiple
processes achieved similar performance to 4.16. In local tests, this patch
improved performance of STREAM relative to the baseline but it is somewhat
machine-dependent. Most workloads show little or not performance difference
implying that there is not a heavily reliance on the throttling mechanism
and it is safe to remove.

STREAM on 2-socket machine
                         4.19.0-rc5             4.19.0-rc5
                         numab-v1r1       noratelimit-v1r1
MB/sec copy     43298.52 (   0.00%)    44673.38 (   3.18%)
MB/sec scale    30115.06 (   0.00%)    31293.06 (   3.91%)
MB/sec add      32825.12 (   0.00%)    34883.62 (   6.27%)
MB/sec triad    32549.52 (   0.00%)    34906.60 (   7.24%

Signed-off-by: Mel Gorman <mgorman@techsingularity.net>
Reviewed-by: Rik van Riel <riel@surriel.com>
Acked-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Jirka Hladky <jhladky@redhat.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Linux-MM <linux-mm@kvack.org>
Cc: Srikar Dronamraju <srikar@linux.vnet.ibm.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Link: http://lkml.kernel.org/r/20181001100525.29789-2-mgorman@techsingularity.net
Signed-off-by: Ingo Molnar <mingo@kernel.org>
5 years agoMAINTAINERS: Remove dead path from LOCKING PRIMITIVES entry
Will Deacon [Mon, 1 Oct 2018 14:28:56 +0000 (15:28 +0100)]
MAINTAINERS: Remove dead path from LOCKING PRIMITIVES entry

Since 890658b7ab48 ("locking/mutex: Kill arch specific code"), there
are no mutex header files under arch/, so we can remove the redundant
entry from MAINTAINERS.

Reported-by: Joe Perches <joe@perches.com>
Signed-off-by: Will Deacon <will.deacon@arm.com>
Cc: Jason Low <jason.low2@hpe.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Link: http://lkml.kernel.org/r/20181001142856.GC9716@arm.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
5 years agodrm: fix use-after-free read in drm_mode_create_lease_ioctl()
Jann Horn [Mon, 1 Oct 2018 15:31:17 +0000 (17:31 +0200)]
drm: fix use-after-free read in drm_mode_create_lease_ioctl()

fd_install() moves the reference given to it into the file descriptor table
of the current process. If the current process is multithreaded, then
immediately after fd_install(), another thread can close() the file
descriptor and cause the file's resources to be cleaned up.

Since the reference to "lessee" is held by the file, we must not access
"lessee" after the fd_install() call.

As far as I can tell, to reach this codepath, the caller must have an open
file descriptor to a DRI device in master mode. I'm not sure what the
requirements for that are.

Signed-off-by: Jann Horn <jannh@google.com>
Fixes: 62884cd386b8 ("drm: Add four ioctls for managing drm mode object leases [v7]")
Cc: stable@vger.kernel.org
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Link: https://patchwork.freedesktop.org/patch/msgid/20181001153117.216923-1-jannh@google.com
5 years agosched/numa: Avoid task migration for small NUMA improvement
Srikar Dronamraju [Fri, 21 Sep 2018 17:49:01 +0000 (23:19 +0530)]
sched/numa: Avoid task migration for small NUMA improvement

If NUMA improvement from the task migration is going to be very
minimal, then avoid task migration.

Specjbb2005 results (8 warehouses)
Higher bops are better

2 Socket - 2  Node Haswell - X86
JVMS  Prev    Current  %Change
4     198512  205910   3.72673
1     313559  318491   1.57291

2 Socket - 4 Node Power8 - PowerNV
JVMS  Prev     Current  %Change
8     74761.9  74935.9  0.232739
1     214874   226796   5.54837

2 Socket - 2  Node Power9 - PowerNV
JVMS  Prev    Current  %Change
4     180536  189780   5.12031
1     210281  205695   -2.18089

4 Socket - 4  Node Power7 - PowerVM
JVMS  Prev     Current  %Change
8     56511.4  60370    6.828
1     104899   108100   3.05151

1/7 cases is regressing, if we look at events migrate_pages seem
to vary the most especially in the regressing case. Also some
amount of variance is expected between different runs of
Specjbb2005.

Some events stats before and after applying the patch.

perf stats 8th warehouse Multi JVM 2 Socket - 2  Node Haswell - X86
Event                     Before          After
cs                        13,818,546      13,801,554
migrations                1,149,960       1,151,541
faults                    385,583         433,246
cache-misses              55,259,546,768  55,168,691,835
sched:sched_move_numa     2,257           2,551
sched:sched_stick_numa    9               24
sched:sched_swap_numa     512             904
migrate:mm_migrate_pages  2,225           1,571

vmstat 8th warehouse Multi JVM 2 Socket - 2  Node Haswell - X86
Event                   Before  After
numa_hint_faults        72692   113682
numa_hint_faults_local  62270   102163
numa_hit                238762  240181
numa_huge_pte_updates   48      36
numa_interleave         75      64
numa_local              238676  240103
numa_other              86      78
numa_pages_migrated     2225    1564
numa_pte_updates        98557   134080

perf stats 8th warehouse Single JVM 2 Socket - 2  Node Haswell - X86
Event                     Before          After
cs                        3,173,490       3,079,150
migrations                36,966          31,455
faults                    108,776         99,081
cache-misses              12,200,075,320  11,588,126,740
sched:sched_move_numa     1,264           1
sched:sched_stick_numa    0               0
sched:sched_swap_numa     0               0
migrate:mm_migrate_pages  899             36

vmstat 8th warehouse Single JVM 2 Socket - 2  Node Haswell - X86
Event                   Before  After
numa_hint_faults        21109   430
numa_hint_faults_local  17120   77
numa_hit                72934   71277
numa_huge_pte_updates   42      0
numa_interleave         33      22
numa_local              72866   71218
numa_other              68      59
numa_pages_migrated     915     23
numa_pte_updates        42326   0

perf stats 8th warehouse Multi JVM 2 Socket - 2  Node Power9 - PowerNV
Event                     Before       After
cs                        8,312,022    8,707,565
migrations                231,705      171,342
faults                    310,242      310,820
cache-misses              402,324,573  136,115,400
sched:sched_move_numa     193          215
sched:sched_stick_numa    0            6
sched:sched_swap_numa     3            24
migrate:mm_migrate_pages  93           162

vmstat 8th warehouse Multi JVM 2 Socket - 2  Node Power9 - PowerNV
Event                   Before  After
numa_hint_faults        11838   8985
numa_hint_faults_local  11216   8154
numa_hit                90689   93819
numa_huge_pte_updates   0       0
numa_interleave         1579    882
numa_local              89634   93496
numa_other              1055    323
numa_pages_migrated     92      169
numa_pte_updates        12109   9217

perf stats 8th warehouse Single JVM 2 Socket - 2  Node Power9 - PowerNV
Event                     Before      After
cs                        2,170,481   2,152,072
migrations                10,126      10,704
faults                    160,962     164,376
cache-misses              10,834,845  3,818,437
sched:sched_move_numa     10          16
sched:sched_stick_numa    0           0
sched:sched_swap_numa     0           7
migrate:mm_migrate_pages  2           199

vmstat 8th warehouse Single JVM 2 Socket - 2  Node Power9 - PowerNV
Event                   Before  After
numa_hint_faults        403     2248
numa_hint_faults_local  358     1666
numa_hit                25898   25704
numa_huge_pte_updates   0       0
numa_interleave         207     200
numa_local              25860   25679
numa_other              38      25
numa_pages_migrated     2       197
numa_pte_updates        400     2234

perf stats 8th warehouse Multi JVM 4 Socket - 4  Node Power7 - PowerVM
Event                     Before           After
cs                        110,339,633      93,330,595
migrations                4,139,812        4,122,061
faults                    863,622          865,979
cache-misses              231,838,045,660  225,395,083,479
sched:sched_move_numa     2,196            2,372
sched:sched_stick_numa    33               24
sched:sched_swap_numa     544              769
migrate:mm_migrate_pages  2,469            1,677

vmstat 8th warehouse Multi JVM 4 Socket - 4  Node Power7 - PowerVM
Event                   Before  After
numa_hint_faults        85748   91638
numa_hint_faults_local  66831   78096
numa_hit                242213  242225
numa_huge_pte_updates   0       0
numa_interleave         0       2
numa_local              242211  242219
numa_other              2       6
numa_pages_migrated     2376    1515
numa_pte_updates        86233   92274

perf stats 8th warehouse Single JVM 4 Socket - 4  Node Power7 - PowerVM
Event                     Before          After
cs                        59,331,057      51,487,271
migrations                552,019         537,170
faults                    266,586         256,921
cache-misses              73,796,312,990  70,073,831,187
sched:sched_move_numa     981             576
sched:sched_stick_numa    54              24
sched:sched_swap_numa     286             327
migrate:mm_migrate_pages  713             726

vmstat 8th warehouse Single JVM 4 Socket - 4  Node Power7 - PowerVM
Event                   Before  After
numa_hint_faults        14807   12000
numa_hint_faults_local  5738    5024
numa_hit                36230   36470
numa_huge_pte_updates   0       0
numa_interleave         0       0
numa_local              36228   36465
numa_other              2       5
numa_pages_migrated     703     726
numa_pte_updates        14742   11930

Signed-off-by: Srikar Dronamraju <srikar@linux.vnet.ibm.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Jirka Hladky <jhladky@redhat.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Mel Gorman <mgorman@techsingularity.net>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Rik van Riel <riel@surriel.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Link: http://lkml.kernel.org/r/1537552141-27815-7-git-send-email-srikar@linux.vnet.ibm.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
5 years agomm/migrate: Use spin_trylock() while resetting rate limit
Srikar Dronamraju [Fri, 21 Sep 2018 17:49:00 +0000 (23:19 +0530)]
mm/migrate: Use spin_trylock() while resetting rate limit

Since this spinlock will only serialize the migrate rate limiting,
convert the spin_lock() to a spin_trylock(). If another thread is updating, this
task can move on.

Specjbb2005 results (8 warehouses)
Higher bops are better

2 Socket - 2  Node Haswell - X86
JVMS  Prev    Current  %Change
4     205332  198512   -3.32145
1     319785  313559   -1.94693

2 Socket - 4 Node Power8 - PowerNV
JVMS  Prev    Current  %Change
8     74912   74761.9  -0.200368
1     206585  214874   4.01239

2 Socket - 2  Node Power9 - PowerNV
JVMS  Prev    Current  %Change
4     189162  180536   -4.56011
1     213760  210281   -1.62753

4 Socket - 4  Node Power7 - PowerVM
JVMS  Prev     Current  %Change
8     58736.8  56511.4  -3.78877
1     105419   104899   -0.49327

Avoiding stretching of window intervals may be the reason for the
regression. Also code now uses READ_ONCE/WRITE_ONCE. That may
also be hurting performance to some extent.

Some events stats before and after applying the patch.

perf stats 8th warehouse Multi JVM 2 Socket - 2  Node Haswell - X86
Event                     Before          After
cs                        14,285,708      13,818,546
migrations                1,180,621       1,149,960
faults                    339,114         385,583
cache-misses              55,205,631,894  55,259,546,768
sched:sched_move_numa     843             2,257
sched:sched_stick_numa    6               9
sched:sched_swap_numa     219             512
migrate:mm_migrate_pages  365             2,225

vmstat 8th warehouse Multi JVM 2 Socket - 2  Node Haswell - X86
Event                   Before  After
numa_hint_faults        26907   72692
numa_hint_faults_local  24279   62270
numa_hit                239771  238762
numa_huge_pte_updates   0       48
numa_interleave         68      75
numa_local              239688  238676
numa_other              83      86
numa_pages_migrated     363     2225
numa_pte_updates        27415   98557

perf stats 8th warehouse Single JVM 2 Socket - 2  Node Haswell - X86
Event                     Before          After
cs                        3,202,779       3,173,490
migrations                37,186          36,966
faults                    106,076         108,776
cache-misses              12,024,873,744  12,200,075,320
sched:sched_move_numa     931             1,264
sched:sched_stick_numa    0               0
sched:sched_swap_numa     1               0
migrate:mm_migrate_pages  637             899

vmstat 8th warehouse Single JVM 2 Socket - 2  Node Haswell - X86
Event                   Before  After
numa_hint_faults        17409   21109
numa_hint_faults_local  14367   17120
numa_hit                73953   72934
numa_huge_pte_updates   20      42
numa_interleave         25      33
numa_local              73892   72866
numa_other              61      68
numa_pages_migrated     668     915
numa_pte_updates        27276   42326

perf stats 8th warehouse Multi JVM 2 Socket - 2  Node Power9 - PowerNV
Event                     Before       After
cs                        8,474,013    8,312,022
migrations                254,934      231,705
faults                    320,506      310,242
cache-misses              110,580,458  402,324,573
sched:sched_move_numa     725          193
sched:sched_stick_numa    0            0
sched:sched_swap_numa     7            3
migrate:mm_migrate_pages  145          93

vmstat 8th warehouse Multi JVM 2 Socket - 2  Node Power9 - PowerNV
Event                   Before  After
numa_hint_faults        22797   11838
numa_hint_faults_local  21539   11216
numa_hit                89308   90689
numa_huge_pte_updates   0       0
numa_interleave         865     1579
numa_local              88955   89634
numa_other              353     1055
numa_pages_migrated     149     92
numa_pte_updates        22930   12109

perf stats 8th warehouse Single JVM 2 Socket - 2  Node Power9 - PowerNV
Event                     Before     After
cs                        2,195,628  2,170,481
migrations                11,179     10,126
faults                    149,656    160,962
cache-misses              8,117,515  10,834,845
sched:sched_move_numa     49         10
sched:sched_stick_numa    0          0
sched:sched_swap_numa     0          0
migrate:mm_migrate_pages  5          2

vmstat 8th warehouse Single JVM 2 Socket - 2  Node Power9 - PowerNV
Event                   Before  After
numa_hint_faults        3577    403
numa_hint_faults_local  3476    358
numa_hit                26142   25898
numa_huge_pte_updates   0       0
numa_interleave         358     207
numa_local              26042   25860
numa_other              100     38
numa_pages_migrated     5       2
numa_pte_updates        3587    400

perf stats 8th warehouse Multi JVM 4 Socket - 4  Node Power7 - PowerVM
Event                     Before           After
cs                        100,602,296      110,339,633
migrations                4,135,630        4,139,812
faults                    789,256          863,622
cache-misses              226,160,621,058  231,838,045,660
sched:sched_move_numa     1,366            2,196
sched:sched_stick_numa    16               33
sched:sched_swap_numa     374              544
migrate:mm_migrate_pages  1,350            2,469

vmstat 8th warehouse Multi JVM 4 Socket - 4  Node Power7 - PowerVM
Event                   Before  After
numa_hint_faults        47857   85748
numa_hint_faults_local  39768   66831
numa_hit                240165  242213
numa_huge_pte_updates   0       0
numa_interleave         0       0
numa_local              240165  242211
numa_other              0       2
numa_pages_migrated     1224    2376
numa_pte_updates        48354   86233

perf stats 8th warehouse Single JVM 4 Socket - 4  Node Power7 - PowerVM
Event                     Before          After
cs                        58,515,496      59,331,057
migrations                564,845         552,019
faults                    245,807         266,586
cache-misses              73,603,757,976  73,796,312,990
sched:sched_move_numa     996             981
sched:sched_stick_numa    10              54
sched:sched_swap_numa     193             286
migrate:mm_migrate_pages  646             713

vmstat 8th warehouse Single JVM 4 Socket - 4  Node Power7 - PowerVM
Event                   Before  After
numa_hint_faults        13422   14807
numa_hint_faults_local  5619    5738
numa_hit                36118   36230
numa_huge_pte_updates   0       0
numa_interleave         0       0
numa_local              36116   36228
numa_other              2       2
numa_pages_migrated     616     703
numa_pte_updates        13374   14742

Suggested-by: Peter Zijlstra <peterz@infradead.org>
Signed-off-by: Srikar Dronamraju <srikar@linux.vnet.ibm.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Jirka Hladky <jhladky@redhat.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Mel Gorman <mgorman@techsingularity.net>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Rik van Riel <riel@surriel.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Link: http://lkml.kernel.org/r/1537552141-27815-6-git-send-email-srikar@linux.vnet.ibm.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
5 years agosched/numa: Limit the conditions where scan period is reset
Mel Gorman [Fri, 21 Sep 2018 17:48:59 +0000 (23:18 +0530)]
sched/numa: Limit the conditions where scan period is reset

migrate_task_rq_fair() resets the scan rate for NUMA balancing on every
cross-node migration. In the event of excessive load balancing due to
saturation, this may result in the scan rate being pegged at maximum and
further overloading the machine.

This patch only resets the scan if NUMA balancing is active, a preferred
node has been selected and the task is being migrated from the preferred
node as these are the most harmful. For example, a migration to the preferred
node does not justify a faster scan rate. Similarly, a migration between two
nodes that are not preferred is probably bouncing due to over-saturation of
the machine.  In that case, scanning faster and trapping more NUMA faults
will further overload the machine.

Specjbb2005 results (8 warehouses)
Higher bops are better

2 Socket - 2  Node Haswell - X86
JVMS  Prev    Current  %Change
4     203370  205332   0.964744
1     328431  319785   -2.63252

2 Socket - 4 Node Power8 - PowerNV
JVMS  Prev    Current  %Change
1     206070  206585   0.249915

2 Socket - 2  Node Power9 - PowerNV
JVMS  Prev    Current  %Change
4     188386  189162   0.41192
1     201566  213760   6.04963

4 Socket - 4  Node Power7 - PowerVM
JVMS  Prev     Current  %Change
8     59157.4  58736.8  -0.710985
1     105495   105419   -0.0720413

Some events stats before and after applying the patch.

perf stats 8th warehouse Multi JVM 2 Socket - 2  Node Haswell - X86
Event                     Before          After
cs                        13,825,492      14,285,708
migrations                1,152,509       1,180,621
faults                    371,948         339,114
cache-misses              55,654,206,041  55,205,631,894
sched:sched_move_numa     1,856           843
sched:sched_stick_numa    4               6
sched:sched_swap_numa     428             219
migrate:mm_migrate_pages  898             365

vmstat 8th warehouse Multi JVM 2 Socket - 2  Node Haswell - X86
Event                   Before  After
numa_hint_faults        57146   26907
numa_hint_faults_local  51612   24279
numa_hit                238164  239771
numa_huge_pte_updates   16      0
numa_interleave         63      68
numa_local              238085  239688
numa_other              79      83
numa_pages_migrated     883     363
numa_pte_updates        67540   27415

perf stats 8th warehouse Single JVM 2 Socket - 2  Node Haswell - X86
Event                     Before          After
cs                        3,288,525       3,202,779
migrations                38,652          37,186
faults                    111,678         106,076
cache-misses              12,111,197,376  12,024,873,744
sched:sched_move_numa     900             931
sched:sched_stick_numa    0               0
sched:sched_swap_numa     5               1
migrate:mm_migrate_pages  714             637

vmstat 8th warehouse Single JVM 2 Socket - 2  Node Haswell - X86
Event                   Before  After
numa_hint_faults        18572   17409
numa_hint_faults_local  14850   14367
numa_hit                73197   73953
numa_huge_pte_updates   11      20
numa_interleave         25      25
numa_local              73138   73892
numa_other              59      61
numa_pages_migrated     712     668
numa_pte_updates        24021   27276

perf stats 8th warehouse Multi JVM 2 Socket - 2  Node Power9 - PowerNV
Event                     Before       After
cs                        8,451,543    8,474,013
migrations                202,804      254,934
faults                    310,024      320,506
cache-misses              253,522,507  110,580,458
sched:sched_move_numa     213          725
sched:sched_stick_numa    0            0
sched:sched_swap_numa     2            7
migrate:mm_migrate_pages  88           145

vmstat 8th warehouse Multi JVM 2 Socket - 2  Node Power9 - PowerNV
Event                   Before  After
numa_hint_faults        11830   22797
numa_hint_faults_local  11301   21539
numa_hit                90038   89308
numa_huge_pte_updates   0       0
numa_interleave         855     865
numa_local              89796   88955
numa_other              242     353
numa_pages_migrated     88      149
numa_pte_updates        12039   22930

perf stats 8th warehouse Single JVM 2 Socket - 2  Node Power9 - PowerNV
Event                     Before     After
cs                        2,049,153  2,195,628
migrations                11,405     11,179
faults                    162,309    149,656
cache-misses              7,203,343  8,117,515
sched:sched_move_numa     22         49
sched:sched_stick_numa    0          0
sched:sched_swap_numa     0          0
migrate:mm_migrate_pages  1          5

vmstat 8th warehouse Single JVM 2 Socket - 2  Node Power9 - PowerNV
Event                   Before  After
numa_hint_faults        1693    3577
numa_hint_faults_local  1669    3476
numa_hit                25177   26142
numa_huge_pte_updates   0       0
numa_interleave         194     358
numa_local              24993   26042
numa_other              184     100
numa_pages_migrated     1       5
numa_pte_updates        1577    3587

perf stats 8th warehouse Multi JVM 4 Socket - 4  Node Power7 - PowerVM
Event                     Before           After
cs                        94,515,937       100,602,296
migrations                4,203,554        4,135,630
faults                    832,697          789,256
cache-misses              226,248,698,331  226,160,621,058
sched:sched_move_numa     1,730            1,366
sched:sched_stick_numa    14               16
sched:sched_swap_numa     432              374
migrate:mm_migrate_pages  1,398            1,350

vmstat 8th warehouse Multi JVM 4 Socket - 4  Node Power7 - PowerVM
Event                   Before  After
numa_hint_faults        80079   47857
numa_hint_faults_local  68620   39768
numa_hit                241187  240165
numa_huge_pte_updates   0       0
numa_interleave         0       0
numa_local              241186  240165
numa_other              1       0
numa_pages_migrated     1347    1224
numa_pte_updates        80729   48354

perf stats 8th warehouse Single JVM 4 Socket - 4  Node Power7 - PowerVM
Event                     Before          After
cs                        63,704,961      58,515,496
migrations                573,404         564,845
faults                    230,878         245,807
cache-misses              76,568,222,781  73,603,757,976
sched:sched_move_numa     509             996
sched:sched_stick_numa    31              10
sched:sched_swap_numa     182             193
migrate:mm_migrate_pages  541             646

vmstat 8th warehouse Single JVM 4 Socket - 4  Node Power7 - PowerVM
Event                   Before  After
numa_hint_faults        8501    13422
numa_hint_faults_local  2960    5619
numa_hit                35526   36118
numa_huge_pte_updates   0       0
numa_interleave         0       0
numa_local              35526   36116
numa_other              0       2
numa_pages_migrated     539     616
numa_pte_updates        8433    13374

Signed-off-by: Mel Gorman <mgorman@techsingularity.net>
Signed-off-by: Srikar Dronamraju <srikar@linux.vnet.ibm.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Jirka Hladky <jhladky@redhat.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Rik van Riel <riel@surriel.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Link: http://lkml.kernel.org/r/1537552141-27815-5-git-send-email-srikar@linux.vnet.ibm.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
5 years agosched/numa: Reset scan rate whenever task moves across nodes
Srikar Dronamraju [Fri, 21 Sep 2018 17:48:58 +0000 (23:18 +0530)]
sched/numa: Reset scan rate whenever task moves across nodes

Currently task scan rate is reset when NUMA balancer migrates the task
to a different node. If NUMA balancer initiates a swap, reset is only
applicable to the task that initiates the swap. Similarly no scan rate
reset is done if the task is migrated across nodes by traditional load
balancer.

Instead move the scan reset to the migrate_task_rq. This ensures the
task moved out of its preferred node, either gets back to its preferred
node quickly or finds a new preferred node. Doing so, would be fair to
all tasks migrating across nodes.

Specjbb2005 results (8 warehouses)
Higher bops are better

2 Socket - 2  Node Haswell - X86
JVMS  Prev    Current  %Change
4     200668  203370   1.3465
1     321791  328431   2.06345

2 Socket - 4 Node Power8 - PowerNV
JVMS  Prev    Current  %Change
1     204848  206070   0.59654

2 Socket - 2  Node Power9 - PowerNV
JVMS  Prev    Current  %Change
4     188098  188386   0.153112
1     200351  201566   0.606436

4 Socket - 4  Node Power7 - PowerVM
JVMS  Prev     Current  %Change
8     58145.9  59157.4  1.73959
1     103798   105495   1.63491

Some events stats before and after applying the patch.

perf stats 8th warehouse Multi JVM 2 Socket - 2  Node Haswell - X86
Event                     Before          After
cs                        13,912,183      13,825,492
migrations                1,155,931       1,152,509
faults                    367,139         371,948
cache-misses              54,240,196,814  55,654,206,041
sched:sched_move_numa     1,571           1,856
sched:sched_stick_numa    9               4
sched:sched_swap_numa     463             428
migrate:mm_migrate_pages  703             898

vmstat 8th warehouse Multi JVM 2 Socket - 2  Node Haswell - X86
Event                   Before  After
numa_hint_faults        50155   57146
numa_hint_faults_local  45264   51612
numa_hit                239652  238164
numa_huge_pte_updates   36      16
numa_interleave         68      63
numa_local              239576  238085
numa_other              76      79
numa_pages_migrated     680     883
numa_pte_updates        71146   67540

perf stats 8th warehouse Single JVM 2 Socket - 2  Node Haswell - X86
Event                     Before          After
cs                        3,156,720       3,288,525
migrations                30,354          38,652
faults                    97,261          111,678
cache-misses              12,400,026,826  12,111,197,376
sched:sched_move_numa     4               900
sched:sched_stick_numa    0               0
sched:sched_swap_numa     1               5
migrate:mm_migrate_pages  20              714

vmstat 8th warehouse Single JVM 2 Socket - 2  Node Haswell - X86
Event                   Before  After
numa_hint_faults        272     18572
numa_hint_faults_local  186     14850
numa_hit                71362   73197
numa_huge_pte_updates   0       11
numa_interleave         23      25
numa_local              71299   73138
numa_other              63      59
numa_pages_migrated     2       712
numa_pte_updates        0       24021

perf stats 8th warehouse Multi JVM 2 Socket - 2  Node Power9 - PowerNV
Event                     Before       After
cs                        8,606,824    8,451,543
migrations                155,352      202,804
faults                    301,409      310,024
cache-misses              157,759,224  253,522,507
sched:sched_move_numa     168          213
sched:sched_stick_numa    0            0
sched:sched_swap_numa     3            2
migrate:mm_migrate_pages  125          88

vmstat 8th warehouse Multi JVM 2 Socket - 2  Node Power9 - PowerNV
Event                   Before  After
numa_hint_faults        4650    11830
numa_hint_faults_local  3946    11301
numa_hit                90489   90038
numa_huge_pte_updates   0       0
numa_interleave         892     855
numa_local              90034   89796
numa_other              455     242
numa_pages_migrated     124     88
numa_pte_updates        4818    12039

perf stats 8th warehouse Single JVM 2 Socket - 2  Node Power9 - PowerNV
Event                     Before     After
cs                        2,113,167  2,049,153
migrations                10,533     11,405
faults                    142,727    162,309
cache-misses              5,594,192  7,203,343
sched:sched_move_numa     10         22
sched:sched_stick_numa    0          0
sched:sched_swap_numa     0          0
migrate:mm_migrate_pages  6          1

vmstat 8th warehouse Single JVM 2 Socket - 2  Node Power9 - PowerNV
Event                   Before  After
numa_hint_faults        744     1693
numa_hint_faults_local  584     1669
numa_hit                25551   25177
numa_huge_pte_updates   0       0
numa_interleave         263     194
numa_local              25302   24993
numa_other              249     184
numa_pages_migrated     6       1
numa_pte_updates        744     1577

perf stats 8th warehouse Multi JVM 4 Socket - 4  Node Power7 - PowerVM
Event                     Before           After
cs                        101,227,352      94,515,937
migrations                4,151,829        4,203,554
faults                    745,233          832,697
cache-misses              224,669,561,766  226,248,698,331
sched:sched_move_numa     617              1,730
sched:sched_stick_numa    2                14
sched:sched_swap_numa     187              432
migrate:mm_migrate_pages  316              1,398

vmstat 8th warehouse Multi JVM 4 Socket - 4  Node Power7 - PowerVM
Event                   Before  After
numa_hint_faults        24195   80079
numa_hint_faults_local  21639   68620
numa_hit                238331  241187
numa_huge_pte_updates   0       0
numa_interleave         0       0
numa_local              238331  241186
numa_other              0       1
numa_pages_migrated     204     1347
numa_pte_updates        24561   80729

perf stats 8th warehouse Single JVM 4 Socket - 4  Node Power7 - PowerVM
Event                     Before          After
cs                        62,738,978      63,704,961
migrations                562,702         573,404
faults                    228,465         230,878
cache-misses              75,778,067,952  76,568,222,781
sched:sched_move_numa     648             509
sched:sched_stick_numa    13              31
sched:sched_swap_numa     137             182
migrate:mm_migrate_pages  733             541

vmstat 8th warehouse Single JVM 4 Socket - 4  Node Power7 - PowerVM
Event                   Before  After
numa_hint_faults        10281   8501
numa_hint_faults_local  3242    2960
numa_hit                36338   35526
numa_huge_pte_updates   0       0
numa_interleave         0       0
numa_local              36338   35526
numa_other              0       0
numa_pages_migrated     706     539
numa_pte_updates        10176   8433

Signed-off-by: Srikar Dronamraju <srikar@linux.vnet.ibm.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Jirka Hladky <jhladky@redhat.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Mel Gorman <mgorman@techsingularity.net>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Rik van Riel <riel@surriel.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Link: http://lkml.kernel.org/r/1537552141-27815-4-git-send-email-srikar@linux.vnet.ibm.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
5 years agosched/numa: Pass destination CPU as a parameter to migrate_task_rq
Srikar Dronamraju [Fri, 21 Sep 2018 17:48:57 +0000 (23:18 +0530)]
sched/numa: Pass destination CPU as a parameter to migrate_task_rq

This additional parameter (new_cpu) is used later for identifying if
task migration is across nodes.

No functional change.

Specjbb2005 results (8 warehouses)
Higher bops are better

2 Socket - 2  Node Haswell - X86
JVMS  Prev    Current  %Change
4     203353  200668   -1.32036
1     328205  321791   -1.95427

2 Socket - 4 Node Power8 - PowerNV
JVMS  Prev    Current  %Change
1     214384  204848   -4.44809

2 Socket - 2  Node Power9 - PowerNV
JVMS  Prev    Current  %Change
4     188553  188098   -0.241311
1     196273  200351   2.07772

4 Socket - 4  Node Power7 - PowerVM
JVMS  Prev     Current  %Change
8     57581.2  58145.9  0.980702
1     103468   103798   0.318939

Brings out the variance between different specjbb2005 runs.

Some events stats before and after applying the patch.

perf stats 8th warehouse Multi JVM 2 Socket - 2  Node Haswell - X86
Event                     Before          After
cs                        13,941,377      13,912,183
migrations                1,157,323       1,155,931
faults                    382,175         367,139
cache-misses              54,993,823,500  54,240,196,814
sched:sched_move_numa     2,005           1,571
sched:sched_stick_numa    14              9
sched:sched_swap_numa     529             463
migrate:mm_migrate_pages  1,573           703

vmstat 8th warehouse Multi JVM 2 Socket - 2  Node Haswell - X86
Event                   Before  After
numa_hint_faults        67099   50155
numa_hint_faults_local  58456   45264
numa_hit                240416  239652
numa_huge_pte_updates   18      36
numa_interleave         65      68
numa_local              240339  239576
numa_other              77      76
numa_pages_migrated     1574    680
numa_pte_updates        77182   71146

perf stats 8th warehouse Single JVM 2 Socket - 2  Node Haswell - X86
Event                     Before          After
cs                        3,176,453       3,156,720
migrations                30,238          30,354
faults                    87,869          97,261
cache-misses              12,544,479,391  12,400,026,826
sched:sched_move_numa     23              4
sched:sched_stick_numa    0               0
sched:sched_swap_numa     6               1
migrate:mm_migrate_pages  10              20

vmstat 8th warehouse Single JVM 2 Socket - 2  Node Haswell - X86
Event                   Before  After
numa_hint_faults        236     272
numa_hint_faults_local  201     186
numa_hit                72293   71362
numa_huge_pte_updates   0       0
numa_interleave         26      23
numa_local              72233   71299
numa_other              60      63
numa_pages_migrated     8       2
numa_pte_updates        0       0

perf stats 8th warehouse Multi JVM 2 Socket - 2  Node Power9 - PowerNV
Event                     Before       After
cs                        8,478,820    8,606,824
migrations                171,323      155,352
faults                    307,499      301,409
cache-misses              240,353,599  157,759,224
sched:sched_move_numa     214          168
sched:sched_stick_numa    0            0
sched:sched_swap_numa     4            3
migrate:mm_migrate_pages  89           125

vmstat 8th warehouse Multi JVM 2 Socket - 2  Node Power9 - PowerNV
Event                   Before  After
numa_hint_faults        5301    4650
numa_hint_faults_local  4745    3946
numa_hit                92943   90489
numa_huge_pte_updates   0       0
numa_interleave         899     892
numa_local              92345   90034
numa_other              598     455
numa_pages_migrated     88      124
numa_pte_updates        5505    4818

perf stats 8th warehouse Single JVM 2 Socket - 2  Node Power9 - PowerNV
Event                     Before      After
cs                        2,066,172   2,113,167
migrations                11,076      10,533
faults                    149,544     142,727
cache-misses              10,398,067  5,594,192
sched:sched_move_numa     43          10
sched:sched_stick_numa    0           0
sched:sched_swap_numa     0           0
migrate:mm_migrate_pages  6           6

vmstat 8th warehouse Single JVM 2 Socket - 2  Node Power9 - PowerNV
Event                   Before  After
numa_hint_faults        3552    744
numa_hint_faults_local  3347    584
numa_hit                25611   25551
numa_huge_pte_updates   0       0
numa_interleave         213     263
numa_local              25583   25302
numa_other              28      249
numa_pages_migrated     6       6
numa_pte_updates        3535    744

perf stats 8th warehouse Multi JVM 4 Socket - 4  Node Power7 - PowerVM
Event                     Before           After
cs                        99,358,136       101,227,352
migrations                4,041,607        4,151,829
faults                    749,653          745,233
cache-misses              225,562,543,251  224,669,561,766
sched:sched_move_numa     771              617
sched:sched_stick_numa    14               2
sched:sched_swap_numa     204              187
migrate:mm_migrate_pages  1,180            316

vmstat 8th warehouse Multi JVM 4 Socket - 4  Node Power7 - PowerVM
Event                   Before  After
numa_hint_faults        27409   24195
numa_hint_faults_local  20677   21639
numa_hit                239988  238331
numa_huge_pte_updates   0       0
numa_interleave         0       0
numa_local              239983  238331
numa_other              5       0
numa_pages_migrated     1016    204
numa_pte_updates        27916   24561

perf stats 8th warehouse Single JVM 4 Socket - 4  Node Power7 - PowerVM
Event                     Before          After
cs                        60,899,307      62,738,978
migrations                544,668         562,702
faults                    270,834         228,465
cache-misses              74,543,455,635  75,778,067,952
sched:sched_move_numa     735             648
sched:sched_stick_numa    25              13
sched:sched_swap_numa     174             137
migrate:mm_migrate_pages  816             733

vmstat 8th warehouse Single JVM 4 Socket - 4  Node Power7 - PowerVM
Event                   Before  After
numa_hint_faults        11059   10281
numa_hint_faults_local  4733    3242
numa_hit                41384   36338
numa_huge_pte_updates   0       0
numa_interleave         0       0
numa_local              41383   36338
numa_other              1       0
numa_pages_migrated     815     706
numa_pte_updates        11323   10176

Signed-off-by: Srikar Dronamraju <srikar@linux.vnet.ibm.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Jirka Hladky <jhladky@redhat.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Mel Gorman <mgorman@techsingularity.net>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Rik van Riel <riel@surriel.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Link: http://lkml.kernel.org/r/1537552141-27815-3-git-send-email-srikar@linux.vnet.ibm.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>