linux-toradex.git/mm, branch v4.12

slub: make sysfs file removal asynchronous

2017-06-23T23:15:55+00:00

Commit bf5eb3de3847 ("slub: separate out sysfs_slab_release() from
sysfs_slab_remove()") made slub sysfs file removals synchronous to
kmem_cache shutdown.

Unfortunately, this created a possible ABBA deadlock between slab_mutex
and sysfs draining mechanism triggering the following lockdep warning.

  ======================================================
  [ INFO: possible circular locking dependency detected ]
  4.10.0-test+ #48 Not tainted
  -------------------------------------------------------
  rmmod/1211 is trying to acquire lock:
   (s_active#120){++++.+}, at: [] kernfs_remove+0x23/0x40

  but task is already holding lock:
   (slab_mutex){+.+.+.}, at: [] kmem_cache_destroy+0x41/0x2d0

  which lock already depends on the new lock.

  the existing dependency chain (in reverse order) is:

  -> #1 (slab_mutex){+.+.+.}:
	 lock_acquire+0xf6/0x1f0
	 __mutex_lock+0x75/0x950
	 mutex_lock_nested+0x1b/0x20
	 slab_attr_store+0x75/0xd0
	 sysfs_kf_write+0x45/0x60
	 kernfs_fop_write+0x13c/0x1c0
	 __vfs_write+0x28/0x120
	 vfs_write+0xc8/0x1e0
	 SyS_write+0x49/0xa0
	 entry_SYSCALL_64_fastpath+0x1f/0xc2

  -> #0 (s_active#120){++++.+}:
	 __lock_acquire+0x10ed/0x1260
	 lock_acquire+0xf6/0x1f0
	 __kernfs_remove+0x254/0x320
	 kernfs_remove+0x23/0x40
	 sysfs_remove_dir+0x51/0x80
	 kobject_del+0x18/0x50
	 __kmem_cache_shutdown+0x3e6/0x460
	 kmem_cache_destroy+0x1fb/0x2d0
	 kvm_exit+0x2d/0x80 [kvm]
	 vmx_exit+0x19/0xa1b [kvm_intel]
	 SyS_delete_module+0x198/0x1f0
	 entry_SYSCALL_64_fastpath+0x1f/0xc2

  other info that might help us debug this:

   Possible unsafe locking scenario:

	 CPU0                    CPU1
	 ----                    ----
    lock(slab_mutex);
				 lock(s_active#120);
				 lock(slab_mutex);
    lock(s_active#120);

   *** DEADLOCK ***

  2 locks held by rmmod/1211:
   #0:  (cpu_hotplug.dep_map){++++++}, at: [] get_online_cpus+0x37/0x80
   #1:  (slab_mutex){+.+.+.}, at: [] kmem_cache_destroy+0x41/0x2d0

  stack backtrace:
  CPU: 3 PID: 1211 Comm: rmmod Not tainted 4.10.0-test+ #48
  Hardware name: Hewlett-Packard HP Compaq Pro 6300 SFF/339A, BIOS K01 v02.05 05/07/2012
  Call Trace:
   print_circular_bug+0x1be/0x210
   __lock_acquire+0x10ed/0x1260
   lock_acquire+0xf6/0x1f0
   __kernfs_remove+0x254/0x320
   kernfs_remove+0x23/0x40
   sysfs_remove_dir+0x51/0x80
   kobject_del+0x18/0x50
   __kmem_cache_shutdown+0x3e6/0x460
   kmem_cache_destroy+0x1fb/0x2d0
   kvm_exit+0x2d/0x80 [kvm]
   vmx_exit+0x19/0xa1b [kvm_intel]
   SyS_delete_module+0x198/0x1f0
   ? SyS_delete_module+0x5/0x1f0
   entry_SYSCALL_64_fastpath+0x1f/0xc2

It'd be the cleanest to deal with the issue by removing sysfs files
without holding slab_mutex before the rest of shutdown; however, given
the current code structure, it is pretty difficult to do so.

This patch punts sysfs file removal to a work item.  Before commit
bf5eb3de3847, the removal was punted to a RCU delayed work item which is
executed after release.  Now, we're punting to a different work item on
shutdown which still maintains the goal removing the sysfs files earlier
when destroying kmem_caches.

Link: http://lkml.kernel.org/r/20170620204512.GI21326@htj.duckdns.org
Fixes: bf5eb3de3847 ("slub: separate out sysfs_slab_release() from sysfs_slab_remove()")
Signed-off-by: Tejun Heo 
Reported-by: Steven Rostedt (VMware) 
Tested-by: Steven Rostedt (VMware) 
Cc: Christoph Lameter 
Cc: Pekka Enberg 
Cc: David Rientjes 
Cc: Joonsoo Kim 
Signed-off-by: Andrew Morton 
Signed-off-by: Linus Torvalds

mm/vmalloc.c: huge-vmap: fail gracefully on unexpected huge vmap mappings

2017-06-23T23:15:55+00:00

Existing code that uses vmalloc_to_page() may assume that any address
for which is_vmalloc_addr() returns true may be passed into
vmalloc_to_page() to retrieve the associated struct page.

This is not un unreasonable assumption to make, but on architectures
that have CONFIG_HAVE_ARCH_HUGE_VMAP=y, it no longer holds, and we need
to ensure that vmalloc_to_page() does not go off into the weeds trying
to dereference huge PUDs or PMDs as table entries.

Given that vmalloc() and vmap() themselves never create huge mappings or
deal with compound pages at all, there is no correct answer in this
case, so return NULL instead, and issue a warning.

When reading /proc/kcore on arm64, you will hit an oops as soon as you
hit the huge mappings used for the various segments that make up the
mapping of vmlinux.  With this patch applied, you will no longer hit the
oops, but the kcore contents willl be incorrect (these regions will be
zeroed out)

We are fixing this for kcore specifically, so it avoids vread() for
those regions.  At least one other problematic user exists, i.e.,
/dev/kmem, but that is currently broken on arm64 for other reasons.

Link: http://lkml.kernel.org/r/20170609082226.26152-1-ard.biesheuvel@linaro.org
Signed-off-by: Ard Biesheuvel 
Acked-by: Mark Rutland 
Reviewed-by: Laura Abbott 
Cc: Michal Hocko 
Cc: zhong jiang 
Cc: Dave Hansen 
Signed-off-by: Andrew Morton 
Signed-off-by: Linus Torvalds

mm, thp: remove cond_resched from __collapse_huge_page_copy

2017-06-23T23:15:55+00:00

This is a partial revert of commit 338a16ba1549 ("mm, thp: copying user
pages must schedule on collapse") which added a cond_resched() to
__collapse_huge_page_copy().

On x86 with CONFIG_HIGHPTE, __collapse_huge_page_copy is called in
atomic context and thus scheduling is not possible.  This is only a
possible config on arm and i386.

Although need_resched has been shown to be set for over 100 jiffies
while doing the iteration in __collapse_huge_page_copy, this is better
than doing

	if (in_atomic())
		cond_resched()

to cover only non-CONFIG_HIGHPTE configs.

Link: http://lkml.kernel.org/r/alpine.DEB.2.10.1706191341550.97821@chino.kir.corp.google.com
Signed-off-by: David Rientjes 
Reported-by: Larry Finger 
Tested-by: Larry Finger 
Acked-by: Michal Hocko 
Cc: Vlastimil Babka 
Signed-off-by: Andrew Morton 
Signed-off-by: Linus Torvalds

Allow stack to grow up to address space limit

2017-06-21T18:07:18+00:00

Fix expand_upwards() on architectures with an upward-growing stack (parisc,
metag and partly IA-64) to allow the stack to reliably grow exactly up to
the address space limit given by TASK_SIZE.

Signed-off-by: Helge Deller 
Acked-by: Hugh Dickins 
Signed-off-by: Linus Torvalds

mm: fix new crash in unmapped_area_topdown()

2017-06-21T17:56:11+00:00

Trinity gets kernel BUG at mm/mmap.c:1963! in about 3 minutes of
mmap testing.  That's the VM_BUG_ON(gap_end < gap_start) at the
end of unmapped_area_topdown().  Linus points out how MAP_FIXED
(which does not have to respect our stack guard gap intentions)
could result in gap_end below gap_start there.  Fix that, and
the similar case in its alternative, unmapped_area().

Cc: stable@vger.kernel.org
Fixes: 1be7107fbe18 ("mm: larger stack guard gap, between vmas")
Reported-by: Dave Jones 
Debugged-by: Linus Torvalds 
Signed-off-by: Hugh Dickins 
Acked-by: Michal Hocko 
Signed-off-by: Linus Torvalds

mm: larger stack guard gap, between vmas

2017-06-19T13:50:20+00:00

Stack guard page is a useful feature to reduce a risk of stack smashing
into a different mapping. We have been using a single page gap which
is sufficient to prevent having stack adjacent to a different mapping.
But this seems to be insufficient in the light of the stack usage in
userspace. E.g. glibc uses as large as 64kB alloca() in many commonly
used functions. Others use constructs liks gid_t buffer[NGROUPS_MAX]
which is 256kB or stack strings with MAX_ARG_STRLEN.

This will become especially dangerous for suid binaries and the default
no limit for the stack size limit because those applications can be
tricked to consume a large portion of the stack and a single glibc call
could jump over the guard page. These attacks are not theoretical,
unfortunatelly.

Make those attacks less probable by increasing the stack guard gap
to 1MB (on systems with 4k pages; but make it depend on the page size
because systems with larger base pages might cap stack allocations in
the PAGE_SIZE units) which should cover larger alloca() and VLA stack
allocations. It is obviously not a full fix because the problem is
somehow inherent, but it should reduce attack space a lot.

One could argue that the gap size should be configurable from userspace,
but that can be done later when somebody finds that the new 1MB is wrong
for some special case applications.  For now, add a kernel command line
option (stack_guard_gap) to specify the stack gap size (in page units).

Implementation wise, first delete all the old code for stack guard page:
because although we could get away with accounting one extra page in a
stack vma, accounting a larger gap can break userspace - case in point,
a program run with "ulimit -S -v 20000" failed when the 1MB gap was
counted for RLIMIT_AS; similar problems could come with RLIMIT_MLOCK
and strict non-overcommit mode.

Instead of keeping gap inside the stack vma, maintain the stack guard
gap as a gap between vmas: using vm_start_gap() in place of vm_start
(or vm_end_gap() in place of vm_end if VM_GROWSUP) in just those few
places which need to respect the gap - mainly arch_get_unmapped_area(),
and and the vma tree's subtree_gap support for that.

Original-patch-by: Oleg Nesterov 
Original-patch-by: Michal Hocko 
Signed-off-by: Hugh Dickins 
Acked-by: Michal Hocko 
Tested-by: Helge Deller  # parisc
Signed-off-by: Linus Torvalds

mm: correct the comment when reclaimed pages exceed the scanned pages

2017-06-16T21:37:05+00:00

Commit e1587a494540 ("mm: vmpressure: fix sending wrong events on
underflow") declared that reclaimed pages exceed the scanned pages due
to the thp reclaim.

That is incorrect because THP will be spilt to normal page and loop
again, which will result in the scanned pages increment.

[akpm@linux-foundation.org: tweak comment text]
Link: http://lkml.kernel.org/r/1496824266-25235-1-git-send-email-zhongjiang@huawei.com
Signed-off-by: zhongjiang 
Acked-by: Minchan Kim 
Signed-off-by: Andrew Morton 
Signed-off-by: Linus Torvalds

mm: numa: avoid waiting on freed migrated pages

2017-06-16T21:37:05+00:00

In do_huge_pmd_numa_page(), we attempt to handle a migrating thp pmd by
waiting until the pmd is unlocked before we return and retry.  However,
we can race with migrate_misplaced_transhuge_page():

    // do_huge_pmd_numa_page                // migrate_misplaced_transhuge_page()
    // Holds 0 refs on page                 // Holds 2 refs on page

    vmf->ptl = pmd_lock(vma->vm_mm, vmf->pmd);
    /* ... */
    if (pmd_trans_migrating(*vmf->pmd)) {
            page = pmd_page(*vmf->pmd);
            spin_unlock(vmf->ptl);
                                            ptl = pmd_lock(mm, pmd);
                                            if (page_count(page) != 2)) {
                                                    /* roll back */
                                            }
                                            /* ... */
                                            mlock_migrate_page(new_page, page);
                                            /* ... */
                                            spin_unlock(ptl);
                                            put_page(page);
                                            put_page(page); // page freed here
            wait_on_page_locked(page);
            goto out;
    }

This can result in the freed page having its waiters flag set
unexpectedly, which trips the PAGE_FLAGS_CHECK_AT_PREP checks in the
page alloc/free functions.  This has been observed on arm64 KVM guests.

We can avoid this by having do_huge_pmd_numa_page() take a reference on
the page before dropping the pmd lock, mirroring what we do in
__migration_entry_wait().

When we hit the race, migrate_misplaced_transhuge_page() will see the
reference and abort the migration, as it may do today in other cases.

Fixes: b8916634b77bffb2 ("mm: Prevent parallel splits during THP migration")
Link: http://lkml.kernel.org/r/1497349722-6731-2-git-send-email-will.deacon@arm.com
Signed-off-by: Mark Rutland 
Signed-off-by: Will Deacon 
Acked-by: Steve Capper 
Acked-by: Kirill A. Shutemov 
Acked-by: Vlastimil Babka 
Cc: Mel Gorman 
Cc: 
Signed-off-by: Andrew Morton 
Signed-off-by: Linus Torvalds

swap: cond_resched in swap_cgroup_prepare()

2017-06-16T21:37:05+00:00

I saw need_resched() warnings when swapping on large swapfile (TBs)
because continuously allocating many pages in swap_cgroup_prepare() took
too long.

We already cond_resched when freeing page in swap_cgroup_swapoff().  Do
the same for the page allocation.

Link: http://lkml.kernel.org/r/20170604200109.17606-1-yuzhao@google.com
Signed-off-by: Yu Zhao 
Acked-by: Michal Hocko 
Acked-by: Vladimir Davydov 
Cc: 
Signed-off-by: Andrew Morton 
Signed-off-by: Linus Torvalds

mm/memory-failure.c: use compound_head() flags for huge pages

2017-06-16T21:37:05+00:00

memory_failure() chooses a recovery action function based on the page
flags.  For huge pages it uses the tail page flags which don't have
anything interesting set, resulting in:

> Memory failure: 0x9be3b4: Unknown page state
> Memory failure: 0x9be3b4: recovery action for unknown page: Failed

Instead, save a copy of the head page's flags if this is a huge page,
this means if there are no relevant flags for this tail page, we use the
head pages flags instead.  This results in the me_huge_page() recovery
action being called:

> Memory failure: 0x9b7969: recovery action for huge page: Delayed

For hugepages that have not yet been allocated, this allows the hugepage
to be dequeued.

Fixes: 524fca1e7356 ("HWPOISON: fix misjudgement of page_action() for errors on mlocked pages")
Link: http://lkml.kernel.org/r/20170524130204.21845-1-james.morse@arm.com
Signed-off-by: James Morse 
Tested-by: Punit Agrawal 
Acked-by: Punit Agrawal 
Acked-by: Naoya Horiguchi 
Cc: 
Signed-off-by: Andrew Morton 
Signed-off-by: Linus Torvalds