Age | Commit message (Collapse) | Author | Files | Lines |
|
fanotify wants to drop fsnotify_mark_srcu lock when waiting for response
from userspace so that the whole notification subsystem is not blocked
during that time. This patch provides a framework for safely getting
mark reference for a mark found in the object list which pins the mark
in that list. We can then drop fsnotify_mark_srcu, wait for userspace
response and then safely continue iteration of the object list once we
reaquire fsnotify_mark_srcu.
Reviewed-by: Miklos Szeredi <mszeredi@redhat.com>
Reviewed-by: Amir Goldstein <amir73il@gmail.com>
Signed-off-by: Jan Kara <jack@suse.cz>
|
|
Currently we queue all marks for destruction on group shutdown and then
destroy them from fsnotify_destroy_group() instead from a worker thread
which is the usual path. However worker can already be processing some
list of marks to destroy so this does not make 100% all marks are really
destroyed by the time group is shut down. This isn't a big problem as
each mark holds group reference and thus group stays partially alive
until all marks are really freed but there's no point in complicating
our lives - just wait for the delayed work to be finished instead.
Reviewed-by: Miklos Szeredi <mszeredi@redhat.com>
Reviewed-by: Amir Goldstein <amir73il@gmail.com>
Signed-off-by: Jan Kara <jack@suse.cz>
|
|
Instead of removing mark from object list from fsnotify_detach_mark(),
remove the mark when last reference to the mark is dropped. This will
allow fanotify to wait for userspace response to event without having to
hold onto fsnotify_mark_srcu.
To avoid pinning inodes by elevated refcount (and thus e.g. delaying
file deletion) while someone holds mark reference, we detach connector
from the object also from fsnotify_destroy_marks() and not only after
removing last mark from the list as it was now.
Reviewed-by: Miklos Szeredi <mszeredi@redhat.com>
Reviewed-by: Amir Goldstein <amir73il@gmail.com>
Signed-off-by: Jan Kara <jack@suse.cz>
|
|
Currently we queue mark into a list of marks for destruction in
__fsnotify_free_mark() and keep the last mark reference dangling. After the
worker waits for SRCU period, it drops the last reference to the mark
which frees it. This scheme has the disadvantage that if we hold
reference to a mark and drop and reacquire SRCU lock, the mark can get
freed immediately which is slightly inconvenient and we will need to
avoid this in the future.
Move to a scheme where queueing of mark into a list of marks for
destruction happens when the last reference to the mark is dropped. Also
drop reference to the mark held by group list already when mark is
removed from that list instead of dropping it only from the destruction
worker.
Reviewed-by: Miklos Szeredi <mszeredi@redhat.com>
Reviewed-by: Amir Goldstein <amir73il@gmail.com>
Signed-off-by: Jan Kara <jack@suse.cz>
|
|
Dropping mark reference can result in mark being freed. Although it
should not happen in inotify_remove_from_idr() since caller should hold
another reference, just don't risk lock up just after WARN_ON
unnecessarily. Also fold do_inotify_remove_from_idr() into the single
callsite as that function really is just two lines of real code.
Reviewed-by: Miklos Szeredi <mszeredi@redhat.com>
Reviewed-by: Amir Goldstein <amir73il@gmail.com>
Signed-off-by: Jan Kara <jack@suse.cz>
|
|
Currently we free fsnotify_mark_connector structure only when inode /
vfsmount is getting freed. This can however impose noticeable memory
overhead when marks get attached to inodes only temporarily. So free the
connector structure once the last mark is detached from the object.
Since notification infrastructure can be working with the connector
under the protection of fsnotify_mark_srcu, we have to be careful and
free the fsnotify_mark_connector only after SRCU period passes.
Reviewed-by: Miklos Szeredi <mszeredi@redhat.com>
Reviewed-by: Amir Goldstein <amir73il@gmail.com>
Signed-off-by: Jan Kara <jack@suse.cz>
|
|
So far list of marks attached to an object (inode / vfsmount) was
protected by i_lock or mnt_root->d_lock. This dictates that the list
must be empty before the object can be destroyed although the list is
now anchored in the fsnotify_mark_connector structure. Protect the list
by a spinlock in the fsnotify_mark_connector structure to decouple
lifetime of a list of marks from a lifetime of the object. This also
simplifies the code quite a bit since we don't have to differentiate
between inode and vfsmount lists in quite a few places anymore.
Reviewed-by: Miklos Szeredi <mszeredi@redhat.com>
Reviewed-by: Amir Goldstein <amir73il@gmail.com>
Signed-off-by: Jan Kara <jack@suse.cz>
|
|
After removing all the indirection it is clear that
hlist_del_init_rcu(&mark->obj_list);
in fsnotify_destroy_marks() is not needed as the mark gets removed from
the list shortly afterwards in fsnotify_destroy_mark() ->
fsnotify_detach_mark() -> fsnotify_detach_from_object(). Also there is
no problem with mark being visible on object list while we call
fsnotify_destroy_mark() as parallel destruction of marks from several
places is properly handled (as mentioned in the comment in
fsnotify_destroy_marks(). So just remove the list removal and also the
stale comment.
Reviewed-by: Miklos Szeredi <mszeredi@redhat.com>
Reviewed-by: Amir Goldstein <amir73il@gmail.com>
Signed-off-by: Jan Kara <jack@suse.cz>
|
|
We lock object list lock in fsnotify_detach_from_object() twice - once
to detach mark and second time to recalculate mask. That is unnecessary
and later it will become problematic as we will free the connector as
soon as there is no mark in it. So move recalculation of fsnotify mask
into the same critical section that is detaching mark.
This also removes recalculation of child dentry flags from
fsnotify_detach_from_object(). That is however fine. Those marks will
get recalculated once some event happens on a child.
Reviewed-by: Miklos Szeredi <mszeredi@redhat.com>
Signed-off-by: Jan Kara <jack@suse.cz>
|
|
fsnotify_detach_mark() calls fsnotify_destroy_inode_mark() or
fsnotify_destroy_vfsmount_mark() to remove mark from object list. These
two functions are however very similar and differ only in the lock they
use to protect the object list of marks. Simplify the code by removing
the indirection and removing mark from the object list in a common
function.
Reviewed-by: Miklos Szeredi <mszeredi@redhat.com>
Reviewed-by: Amir Goldstein <amir73il@gmail.com>
Signed-off-by: Jan Kara <jack@suse.cz>
|
|
Instead of passing spinlock into fsnotify_destroy_marks() determine it
directly in that function from the connector type. This will reduce code
churn when changing lock protecting list of marks.
Reviewed-by: Miklos Szeredi <mszeredi@redhat.com>
Reviewed-by: Amir Goldstein <amir73il@gmail.com>
Signed-off-by: Jan Kara <jack@suse.cz>
|
|
Move locking of a mark list into fsnotify_find_mark(). This reduces code
churn in the following patch changing lock protecting the list.
Reviewed-by: Miklos Szeredi <mszeredi@redhat.com>
Reviewed-by: Amir Goldstein <amir73il@gmail.com>
Signed-off-by: Jan Kara <jack@suse.cz>
|
|
Move locking of locks protecting a list of marks into
fsnotify_recalc_mask(). This reduces code churn in the following patch
which changes the lock protecting the list of marks.
Reviewed-by: Miklos Szeredi <mszeredi@redhat.com>
Reviewed-by: Amir Goldstein <amir73il@gmail.com>
Signed-off-by: Jan Kara <jack@suse.cz>
|
|
Move fsnotify_destroy_marks() to be later in the fs/notify/mark.c. It
will need some functions that are declared after its current
declaration. No functional change.
Reviewed-by: Miklos Szeredi <mszeredi@redhat.com>
Reviewed-by: Amir Goldstein <amir73il@gmail.com>
Signed-off-by: Jan Kara <jack@suse.cz>
|
|
Adding notification mark to object list has been currently done through
fsnotify_add_{inode|vfsmount}_mark() helpers from
fsnotify_add_mark_locked() which call fsnotify_add_mark_list(). Remove
this unnecessary indirection to simplify the code.
Pushing all the locking to fsnotify_add_mark_list() also allows us to
allocate the connector structure with GFP_KERNEL mode.
Reviewed-by: Miklos Szeredi <mszeredi@redhat.com>
Reviewed-by: Amir Goldstein <amir73il@gmail.com>
Signed-off-by: Jan Kara <jack@suse.cz>
|
|
Currently inode reference is held by fsnotify marks. Change the rules so
that inode reference is held by fsnotify_mark_connector structure
whenever the list is non-empty. This simplifies the code and is more
logical.
Reviewed-by: Miklos Szeredi <mszeredi@redhat.com>
Reviewed-by: Amir Goldstein <amir73il@gmail.com>
Signed-off-by: Jan Kara <jack@suse.cz>
|
|
Move pointer to inode / vfsmount from mark itself to the
fsnotify_mark_connector structure. This is another step on the path
towards decoupling inode / vfsmount lifetime from notification mark
lifetime.
Reviewed-by: Miklos Szeredi <mszeredi@redhat.com>
Reviewed-by: Amir Goldstein <amir73il@gmail.com>
Signed-off-by: Jan Kara <jack@suse.cz>
|
|
Currently notification marks are attached to object (inode or vfsmnt) by
a hlist_head in the object. The list is also protected by a spinlock in
the object. So while there is any mark attached to the list of marks,
the object must be pinned in memory (and thus e.g. last iput() deleting
inode cannot happen). Also for list iteration in fsnotify() to work, we
must hold fsnotify_mark_srcu lock so that mark itself and
mark->obj_list.next cannot get freed. Thus we are required to wait for
response to fanotify events from userspace process with
fsnotify_mark_srcu lock held. That causes issues when userspace process
is buggy and does not reply to some event - basically the whole
notification subsystem gets eventually stuck.
So to be able to drop fsnotify_mark_srcu lock while waiting for
response, we have to pin the mark in memory and make sure it stays in
the object list (as removing the mark waiting for response could lead to
lost notification events for groups later in the list). However we don't
want inode reclaim to block on such mark as that would lead to system
just locking up elsewhere.
This commit is the first in the series that paves way towards solving
these conflicting lifetime needs. Instead of anchoring the list of marks
directly in the object, we anchor it in a dedicated structure
(fsnotify_mark_connector) and just point to that structure from the
object. The following commits will also add spinlock protecting the list
and object pointer to the structure.
Reviewed-by: Miklos Szeredi <mszeredi@redhat.com>
Reviewed-by: Amir Goldstein <amir73il@gmail.com>
Signed-off-by: Jan Kara <jack@suse.cz>
|
|
Add a comment that lifetime of a notification mark is protected by SRCU
and remove a comment about clearing of marks attached to the inode. It
is stale and more uptodate version is at fsnotify_destroy_marks() which
is the function handling this case.
Reviewed-by: Miklos Szeredi <mszeredi@redhat.com>
Reviewed-by: Amir Goldstein <amir73il@gmail.com>
Signed-off-by: Jan Kara <jack@suse.cz>
|
|
Move recalculation of inode / vfsmount notification mask under
group->mark_mutex of the mark which was modified. These are the only
places where mask recalculation happens without mark being protected
from detaching from inode / vfsmount which will cause issues with the
following patches.
Reviewed-by: Miklos Szeredi <mszeredi@redhat.com>
Reviewed-by: Amir Goldstein <amir73il@gmail.com>
Signed-off-by: Jan Kara <jack@suse.cz>
|
|
Printing inode pointers in warnings has dubious value and with future
changes we won't be able to easily get them without either locking or
chances we oops along the way. So just remove inode pointers from the
warning messages.
Reviewed-by: Miklos Szeredi <mszeredi@redhat.com>
Reviewed-by: Amir Goldstein <amir73il@gmail.com>
Signed-off-by: Jan Kara <jack@suse.cz>
|
|
show_fdinfo() iterates group's list of marks. All marks found there are
guaranteed to be alive and they stay so until we release
group->mark_mutex. So remove uncecessary tests whether mark is alive.
Reviewed-by: Miklos Szeredi <mszeredi@redhat.com>
Reviewed-by: Amir Goldstein <amir73il@gmail.com>
Signed-off-by: Jan Kara <jack@suse.cz>
|
|
Merge 5-level page table prep from Kirill Shutemov:
"Here's relatively low-risk part of 5-level paging patchset. Merging it
now will make x86 5-level paging enabling in v4.12 easier.
The first patch is actually x86-specific: detect 5-level paging
support. It boils down to single define.
The rest of patchset converts Linux MMU abstraction from 4- to 5-level
paging.
Enabling of new abstraction in most cases requires adding single line
of code in arch-specific code. The rest is taken care by asm-generic/.
Changes to mm/ code are mostly mechanical: add support for new page
table level -- p4d_t -- where we deal with pud_t now.
v2:
- fix build on microblaze (Michal);
- comment for __ARCH_HAS_5LEVEL_HACK in kasan_populate_zero_shadow();
- acks from Michal"
* emailed patches from Kirill A Shutemov <kirill.shutemov@linux.intel.com>:
mm: introduce __p4d_alloc()
mm: convert generic code to 5-level paging
asm-generic: introduce <asm-generic/pgtable-nop4d.h>
arch, mm: convert all architectures to use 5level-fixup.h
asm-generic: introduce __ARCH_USE_5LEVEL_HACK
asm-generic: introduce 5level-fixup.h
x86/cpufeature: Add 5-level paging detection
|
|
Merge fixes from Andrew Morton:
"26 fixes"
* emailed patches from Andrew Morton <akpm@linux-foundation.org>: (26 commits)
userfaultfd: remove wrong comment from userfaultfd_ctx_get()
fat: fix using uninitialized fields of fat_inode/fsinfo_inode
sh: cayman: IDE support fix
kasan: fix races in quarantine_remove_cache()
kasan: resched in quarantine_remove_cache()
mm: do not call mem_cgroup_free() from within mem_cgroup_alloc()
thp: fix another corner case of munlock() vs. THPs
rmap: fix NULL-pointer dereference on THP munlocking
mm/memblock.c: fix memblock_next_valid_pfn()
userfaultfd: selftest: vm: allow to build in vm/ directory
userfaultfd: non-cooperative: userfaultfd_remove revalidate vma in MADV_DONTNEED
userfaultfd: non-cooperative: fix fork fctx->new memleak
mm/cgroup: avoid panic when init with low memory
drivers/md/bcache/util.h: remove duplicate inclusion of blkdev.h
mm/vmstats: add thp_split_pud event for clarity
include/linux/fs.h: fix unsigned enum warning with gcc-4.2
userfaultfd: non-cooperative: release all ctx in dup_userfaultfd_complete
userfaultfd: non-cooperative: robustness check
userfaultfd: non-cooperative: rollback userfaultfd_exit
x86, mm: unify exit paths in gup_pte_range()
...
|
|
Pull xfs fixes from Darrick Wong:
"Here are some bug fixes for -rc2 to clean up the copy on write
handling and to remove a cause of hangs.
- Fix various iomap bugs
- Fix overly aggressive CoW preallocation garbage collection
- Fixes to CoW endio error handling
- Fix some incorrect geometry calculations
- Remove a potential system hang in bulkstat
- Try to allocate blocks more aggressively to reduce ENOSPC errors"
* tag 'xfs-4.11-fixes-1' of git://git.kernel.org/pub/scm/fs/xfs/xfs-linux:
xfs: try any AG when allocating the first btree block when reflinking
xfs: use iomap new flag for newly allocated delalloc blocks
xfs: remove kmem_zalloc_greedy
xfs: Use xfs_icluster_size_fsb() to calculate inode alignment mask
xfs: fix and streamline error handling in xfs_end_io
xfs: only reclaim unwritten COW extents periodically
iomap: invalidate page caches should be after iomap_dio_complete() in direct write
|
|
It's a void function, so there is no return value;
Link: http://lkml.kernel.org/r/20170309150817.7510-1-david@redhat.com
Signed-off-by: David Hildenbrand <david@redhat.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
|
|
Recently fallocate patch was merged and it uses
MSDOS_I(inode)->mmu_private at fat_evict_inode(). However,
fat_inode/fsinfo_inode that was introduced in past didn't initialize
MSDOS_I(inode) properly.
With those combinations, it became the cause of accessing random entry
in FAT area.
Link: http://lkml.kernel.org/r/87pohrj4i8.fsf@mail.parknet.co.jp
Signed-off-by: OGAWA Hirofumi <hirofumi@mail.parknet.co.jp>
Reported-by: Moreno Bartalucci <moreno.bartalucci@tecnorama.it>
Tested-by: Moreno Bartalucci <moreno.bartalucci@tecnorama.it>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
|
|
userfaultfd_remove() has to be execute before zapping the pagetables or
UFFDIO_COPY could keep filling pages after zap_page_range returned,
which would result in non zero data after a MADV_DONTNEED.
However userfaultfd_remove() may have to release the mmap_sem. This was
handled correctly in MADV_REMOVE, but MADV_DONTNEED accessed a
potentially stale vma (the very vma passed to zap_page_range(vma, ...)).
The fix consists in revalidating the vma in case userfaultfd_remove()
had to release the mmap_sem.
This also optimizes away an unnecessary down_read/up_read in the
MADV_REMOVE case if UFFD_EVENT_FORK had to be delivered.
It all remains zero runtime cost in case CONFIG_USERFAULTFD=n as
userfaultfd_remove() will be defined as "true" at build time.
Link: http://lkml.kernel.org/r/20170302173738.18994-3-aarcange@redhat.com
Signed-off-by: Andrea Arcangeli <aarcange@redhat.com>
Acked-by: Mike Rapoport <rppt@linux.vnet.ibm.com>
Cc: "Dr. David Alan Gilbert" <dgilbert@redhat.com>
Cc: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Pavel Emelyanov <xemul@parallels.com>
Cc: Hillf Danton <hillf.zj@alibaba-inc.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
|
|
We have a memleak in the ->new ctx if the uffd of the parent is closed
before the fork event is read, nothing frees the new context.
Link: http://lkml.kernel.org/r/20170302173738.18994-2-aarcange@redhat.com
Signed-off-by: Mike Rapoport <rppt@linux.vnet.ibm.com>
Signed-off-by: Andrea Arcangeli <aarcange@redhat.com>
Reported-by: Andrea Arcangeli <aarcange@redhat.com>
Cc: "Dr. David Alan Gilbert" <dgilbert@redhat.com>
Cc: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Pavel Emelyanov <xemul@parallels.com>
Cc: Hillf Danton <hillf.zj@alibaba-inc.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
|
|
Don't stop running dup_fctx() even if userfaultfd_event_wait_completion
fails as it has to run userfaultfd_ctx_put on all ctx to pair against
the userfaultfd_ctx_get that was run on all fctx->orig in
dup_userfaultfd.
Link: http://lkml.kernel.org/r/20170224181957.19736-4-aarcange@redhat.com
Signed-off-by: Andrea Arcangeli <aarcange@redhat.com>
Acked-by: Mike Rapoport <rppt@linux.vnet.ibm.com>
Cc: "Dr. David Alan Gilbert" <dgilbert@redhat.com>
Cc: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Pavel Emelyanov <xemul@parallels.com>
Cc: Hillf Danton <hillf.zj@alibaba-inc.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
|
|
Similar to the handle_userfault() case, also make sure to never attempt
to send any event past the PF_EXITING point of no return.
This is purely a robustness check.
Link: http://lkml.kernel.org/r/20170224181957.19736-3-aarcange@redhat.com
Signed-off-by: Andrea Arcangeli <aarcange@redhat.com>
Acked-by: Mike Rapoport <rppt@linux.vnet.ibm.com>
Cc: "Dr. David Alan Gilbert" <dgilbert@redhat.com>
Cc: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Pavel Emelyanov <xemul@parallels.com>
Cc: Hillf Danton <hillf.zj@alibaba-inc.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
|
|
Patch series "userfaultfd non-cooperative further update for 4.11 merge
window".
Unfortunately I noticed one relevant bug in userfaultfd_exit while doing
more testing. I've been doing testing before and this was also tested
by kbuild bot and exercised by the selftest, but this bug never
reproduced before.
I dropped userfaultfd_exit as result. I dropped it because of
implementation difficulty in receiving signals in __mmput and because I
think -ENOSPC as result from the background UFFDIO_COPY should be enough
already.
Before I decided to remove userfaultfd_exit, I noticed userfaultfd_exit
wasn't exercised by the selftest and when I tried to exercise it, after
moving it to a more correct place in __mmput where it would make more
sense and where the vma list is stable, it resulted in the
event_wait_completion in D state. So then I added the second patch to
be sure even if we call userfaultfd_event_wait_completion too late
during task exit(), we won't risk to generate tasks in D state. The
same check exists in handle_userfault() for the same reason, except it
makes a difference there, while here is just a robustness check and it's
run under WARN_ON_ONCE.
While looking at the userfaultfd_event_wait_completion() function I
looked back at its callers too while at it and I think it's not ok to
stop executing dup_fctx on the fcs list because we relay on
userfaultfd_event_wait_completion to execute
userfaultfd_ctx_put(fctx->orig) which is paired against
userfaultfd_ctx_get(fctx->orig) in dup_userfault just before
list_add(fcs). This change only takes care of fctx->orig but this area
also needs further review looking for similar problems in fctx->new.
The only patch that is urgent is the first because it's an use after
free during a SMP race condition that affects all processes if
CONFIG_USERFAULTFD=y. Very hard to reproduce though and probably
impossible without SLUB poisoning enabled.
This patch (of 3):
I once reproduced this oops with the userfaultfd selftest, it's not
easily reproducible and it requires SLUB poisoning to reproduce.
general protection fault: 0000 [#1] SMP
Modules linked in:
CPU: 2 PID: 18421 Comm: userfaultfd Tainted: G ------------ T 3.10.0+ #15
Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS rel-1.10.1-0-g8891697-prebuilt.qemu-project.org 04/01/2014
task: ffff8801f83b9440 ti: ffff8801f833c000 task.ti: ffff8801f833c000
RIP: 0010:[<ffffffff81451299>] [<ffffffff81451299>] userfaultfd_exit+0x29/0xa0
RSP: 0018:ffff8801f833fe80 EFLAGS: 00010202
RAX: ffff8801f833ffd8 RBX: 6b6b6b6b6b6b6b6b RCX: ffff8801f83b9440
RDX: 0000000000000000 RSI: 0000000000000000 RDI: ffff8800baf18600
RBP: ffff8801f833fee8 R08: 0000000000000000 R09: 0000000000000001
R10: 0000000000000000 R11: ffffffff8127ceb3 R12: 0000000000000000
R13: ffff8800baf186b0 R14: ffff8801f83b99f8 R15: 00007faed746c700
FS: 0000000000000000(0000) GS:ffff88023fc80000(0000) knlGS:0000000000000000
CS: 0010 DS: 0000 ES: 0000 CR0: 000000008005003b
CR2: 00007faf0966f028 CR3: 0000000001bc6000 CR4: 00000000000006e0
DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
DR3: 0000000000000000 DR6: 00000000ffff0ff0 DR7: 0000000000000400
Call Trace:
do_exit+0x297/0xd10
SyS_exit+0x17/0x20
tracesys+0xdd/0xe2
Code: 00 00 66 66 66 66 90 55 48 89 e5 41 54 53 48 83 ec 58 48 8b 1f 48 85 db 75 11 eb 73 66 0f 1f 44 00 00 48 8b 5b 10 48 85 db 74 64 <4c> 8b a3 b8 00 00 00 4d 85 e4 74 eb 41 f6 84 24 2c 01 00 00 80
RIP [<ffffffff81451299>] userfaultfd_exit+0x29/0xa0
RSP <ffff8801f833fe80>
---[ end trace 9fecd6dcb442846a ]---
In the debugger I located the "mm" pointer in the stack and walking
mm->mmap->vm_next through the end shows the vma->vm_next list is fully
consistent and it is null terminated list as expected. So this has to
be an SMP race condition where userfaultfd_exit was running while the
vma list was being modified by another CPU.
When userfaultfd_exit() run one of the ->vm_next pointers pointed to
SLAB_POISON (RBX is the vma pointer and is 0x6b6b..).
The reason is that it's not running in __mmput but while there are still
other threads running and it's not holding the mmap_sem (it can't as it
has to wait the even to be received by the manager). So this is an use
after free that was happening for all processes.
One more implementation problem aside from the race condition:
userfaultfd_exit has really to check a flag in mm->flags before walking
the vma or it's going to slowdown the exit() path for regular tasks.
One more implementation problem: at that point signals can't be
delivered so it would also create a task in D state if the manager
doesn't read the event.
The major design issue: it overall looks superfluous as the manager can
check for -ENOSPC in the background transfer:
if (mmget_not_zero(ctx->mm)) {
[..]
} else {
return -ENOSPC;
}
It's safer to roll it back and re-introduce it later if at all.
[rppt@linux.vnet.ibm.com: documentation fixup after removal of UFFD_EVENT_EXIT]
Link: http://lkml.kernel.org/r/1488345437-4364-1-git-send-email-rppt@linux.vnet.ibm.com
Link: http://lkml.kernel.org/r/20170224181957.19736-2-aarcange@redhat.com
Signed-off-by: Andrea Arcangeli <aarcange@redhat.com>
Signed-off-by: Mike Rapoport <rppt@linux.vnet.ibm.com>
Acked-by: Mike Rapoport <rppt@linux.vnet.ibm.com>
Cc: "Dr. David Alan Gilbert" <dgilbert@redhat.com>
Cc: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Pavel Emelyanov <xemul@parallels.com>
Cc: Hillf Danton <hillf.zj@alibaba-inc.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
|
|
__do_fault assumes vmf->page has been initialized and is valid if
VM_FAULT_NOPAGE is not returned by vma->vm_ops->fault(vma, vmf).
handle_userfault() in turn should return VM_FAULT_NOPAGE if it doesn't
return VM_FAULT_SIGBUS or VM_FAULT_RETRY (the other two possibilities).
This VM_FAULT_NOPAGE case is only invoked when signal are pending and it
didn't matter for anonymous memory before. It only started to matter
since shmem was introduced. hugetlbfs also takes a different path and
doesn't exercise __do_fault.
Link: http://lkml.kernel.org/r/20170228154201.GH5816@redhat.com
Signed-off-by: Andrea Arcangeli <aarcange@redhat.com>
Reported-by: Dmitry Vyukov <dvyukov@google.com>
Cc: "Kirill A. Shutemov" <kirill@shutemov.name>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
|
|
Convert all non-architecture-specific code to 5-level paging.
It's mostly mechanical adding handling one more page table level in
places where we deal with pud_t.
Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
|
|
This removes the extra include header file that was added in commit
e58bc927835a "Pull overlayfs updates from Miklos Szeredi" now that it
is no longer needed.
There are probably other such includes that got added during the
scheduler header splitup series, but this is the one that annoyed me
personally and I know about.
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
|
|
When a reflink operation causes the bmap code to allocate a btree block
we're currently doing single-AG allocations due to having ->firstblock
set and then try any higher AG due a little reflink quirk we've put in
when adding the reflink code. But given that we do not have a minleft
reservation of any kind in this AG we can still not have any space in
the same or higher AG even if the file system has enough free space.
To fix this use a XFS_ALLOCTYPE_FIRST_AG allocation in this fall back
path instead.
[And yes, we need to redo this properly instead of piling hacks over
hacks. I'm working on that, but it's not going to be a small series.
In the meantime this fixes the customer reported issue]
Also add a warning for failing allocations to make it easier to debug.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
|
|
Commit fa7f138 ("xfs: clear delalloc and cache on buffered write
failure") fixed one regression in the iomap error handling code and
exposed another. The fundamental problem is that if a buffered write
is a rewrite of preexisting delalloc blocks and the write fails, the
failure handling code can punch out preexisting blocks with valid
file data.
This was reproduced directly by sub-block writes in the LTP
kernel/syscalls/write/write03 test. A first 100 byte write allocates
a single block in a file. A subsequent 100 byte write fails and
punches out the block, including the data successfully written by
the previous write.
To address this problem, update the ->iomap_begin() handler to
distinguish newly allocated delalloc blocks from preexisting
delalloc blocks via the IOMAP_F_NEW flag. Use this flag in the
->iomap_end() handler to decide when a failed or short write should
punch out delalloc blocks.
This introduces the subtle requirement that ->iomap_begin() should
never combine newly allocated delalloc blocks with existing blocks
in the resulting iomap descriptor. This can occur when a new
delalloc reservation merges with a neighboring extent that is part
of the current write, for example. Therefore, drop the
post-allocation extent lookup from xfs_bmapi_reserve_delalloc() and
just return the record inserted into the fork. This ensures only new
blocks are returned and thus that preexisting delalloc blocks are
always handled as "found" blocks and not punched out on a failed
rewrite.
Reported-by: Xiong Zhou <xzhou@redhat.com>
Signed-off-by: Brian Foster <bfoster@redhat.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
|
|
The sole remaining caller of kmem_zalloc_greedy is bulkstat, which uses
it to grab 1-4 pages for staging of inobt records. The infinite loop in
the greedy allocation function is causing hangs[1] in generic/269, so
just get rid of the greedy allocator in favor of kmem_zalloc_large.
This makes bulkstat somewhat more likely to ENOMEM if there's really no
pages to spare, but eliminates a source of hangs.
[1] http://lkml.kernel.org/r/20170301044634.rgidgdqqiiwsmfpj%40XZHOUW.usersys.redhat.com
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
---
v2: remove single-page fallback
|
|
When block size is larger than inode cluster size, the call to
XFS_B_TO_FSBT(mp, mp->m_inode_cluster_size) returns 0. Also, mkfs.xfs
would have set xfs_sb->sb_inoalignmt to 0. Hence in
xfs_set_inoalignment(), xfs_mount->m_inoalign_mask gets initialized to
-1 instead of 0. However, xfs_mount->m_sinoalign would get correctly
intialized to 0 because for every positive value of xfs_mount->m_dalign,
the condition "!(mp->m_dalign & mp->m_inoalign_mask)" would evaluate to
false.
Also, xfs_imap() worked fine even with xfs_mount->m_inoalign_mask having
-1 as the value because blks_per_cluster variable would have the value 1
and hence we would never have a need to use xfs_mount->m_inoalign_mask
to compute the inode chunk's agbno and offset within the chunk.
Signed-off-by: Chandan Rajendra <chandan@linux.vnet.ibm.com>
Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
|
|
There are two different cases of buffered I/O errors:
- first we can have an already shutdown fs. In that case we should skip
any on-disk operations and just clean up the appen transaction if
present and destroy the ioend
- a real I/O error. In that case we should cleanup any lingering COW
blocks. This gets skipped in the current code and is fixed by this
patch.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
|
|
We only want to reclaim preallocations from our periodic work item.
Currently this is archived by looking for a dirty inode, but that check
is rather fragile. Instead add a flag to xfs_reflink_cancel_cow_* so
that the caller can ask for just cancelling unwritten extents in the COW
fork.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com>
[darrick: fix typos in commit message]
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
|
|
git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip
Pull timer fixes from Ingo Molnar:
"This includes a fix for lockups caused by incorrect nsecs related
cleanup, and a capabilities check fix for timerfd"
* 'timers-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
jiffies: Revert bogus conversion of NSEC_PER_SEC to TICK_NSEC
timerfd: Only check CAP_WAKE_ALARM when it is needed
|
|
write
After XFS switching to iomap based DIO (commit acdda3aae146 ("xfs:
use iomap_dio_rw")), I started to notice dio29/dio30 tests failures
from LTP run on ppc64 hosts, and they can be reproduced on x86_64
hosts with 512B/1k block size XFS too.
dio29 diotest3 -b 65536 -n 100 -i 1000 -o 1024000
dio30 diotest6 -b 65536 -n 100 -i 1000 -o 1024000
The failure message is like:
bufcmp: offset 0: Expected: 0x62, got 0x0
diotest03 1 TPASS : Read with Direct IO, Write without
diotest03 2 TFAIL : diotest3.c:142: comparsion failed; child=98 offset=1425408
diotest03 3 TFAIL : diotest3.c:194: Write Direct-child 98 failed
Direct write wrote 0x62 but buffer read got zero. This is because,
when doing direct write to a hole or preallocated file, we
invalidate the page caches before converting the extent from
unwritten state to normal state, which is done by
iomap_dio_complete(), thus leave a window for other buffer reader to
cache the unwritten state extent.
Consider this case, with sub-page blocksize XFS, two processes are
direct writing to different blocksize-aligned regions (say 512B) of
the same preallocated file, and reading the region back via buffered
I/O to compare contents.
process A, region [0,512] process B, region [512,1024]
xfs_file_write_iter
xfs_file_aio_dio_write
iomap_dio_rw
iomap_apply
invalidate_inode_pages2_range
xfs_file_write_iter
xfs_file_aio_dio_write
iomap_dio_rw
iomap_apply
invalidate_inode_pages2_range
iomap_dio_complete
xfs_file_read_iter
xfs_file_buffered_aio_read
generic_file_read_iter
do_generic_file_read
<readahead fills pagecache with 0>
iomap_dio_complete
xfs_file_read_iter
<read gets 0 from pagecache>
Process A first invalidates page caches, at this point the
underlying extent is still in unwritten state (iomap_dio_complete
not called yet), and process B finishs direct write and populates
page caches via readahead, which caches zeros in page for region A,
then process A reads zeros from page cache, instead of the actual
data.
Fix it by invalidating page caches after converting unwritten extent
to make sure we read content from disk after extent state changed,
as what we did before switching to iomap based dio.
Also introduce a new 'start' variable to save the original write
offset (iomap_dio_complete() updates iocb->ki_pos), and a 'err'
variable for invalidating caches result, cause we can't reuse 'ret'
anymore.
Signed-off-by: Eryu Guan <eguan@redhat.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
|
|
git://git.kernel.org/pub/scm/linux/kernel/git/viro/vfs
Pull misc final vfs updates from Al Viro:
"A few unrelated patches that got beating in -next.
Everything else will have to go into the next window ;-/"
* 'work.misc' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/vfs:
hfs: fix hfs_readdir()
selftest for default_file_splice_read() infoleak
9p: constify ->d_name handling
|
|
Pull SMB3 fixes from Steve French:
"Some small bug fixes as well as SMB2.1/SMB3 enablement for DFS (global
namespace) which previously was only enabled for CIFS"
* 'for-next' of git://git.samba.org/sfrench/cifs-2.6:
smb2: Enforce sec= mount option
CIFS: Fix sparse warnings
CIFS: implement get_dfs_refer for SMB2+
CIFS: use DFS pathnames in SMB2+ Create requests
CIFS: set signing flag in SMB2+ TreeConnect if needed
CIFS: let ses->ipc_tid hold smb2 TreeIds
CIFS: add use_ipc flag to SMB2_ioctl()
CIFS: add build_path_from_dentry_optional_prefix()
CIFS: move DFS response parsing out of SMB1 code
CIFS: Fix possible use after free in demultiplex thread
|
|
git://git.kernel.org/pub/scm/linux/kernel/git/mszeredi/fuse
Pull fuse update from Miklos Szeredi:
"A bugfix and cleanups"
* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mszeredi/fuse:
fuse: release: private_data cannot be NULL
fuse: cleanup fuse_file refcounting
fuse: add missing FR_FORCE
|
|
git://git.kernel.org/pub/scm/linux/kernel/git/mszeredi/vfs
Pull overlayfs updates from Miklos Szeredi:
"Because copy up can take a long time, serialized copy ups could be a
big performance bottleneck. This update allows concurrent copy up of
regular files eliminating this potential problem.
There are also minor fixes"
* 'overlayfs-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mszeredi/vfs:
ovl: drop CAP_SYS_RESOURCE from saved mounter's credentials
ovl: properly implement sync_filesystem()
ovl: concurrent copy up of regular files
ovl: introduce copy up waitqueue
ovl: copy up regular file using O_TMPFILE
ovl: rearrange code in ovl_copy_up_locked()
ovl: check if upperdir fs supports O_TMPFILE
|
|
git://git.kernel.org/pub/scm/linux/kernel/git/viro/vfs
Pull vfs 'statx()' update from Al Viro.
This adds the new extended stat() interface that internally subsumes our
previous stat interfaces, and allows user mode to specify in more detail
what kind of information it wants.
It also allows for some explicit synchronization information to be
passed to the filesystem, which can be relevant for network filesystems:
is the cached value ok, or do you need open/close consistency, or what?
From David Howells.
Andreas Dilger points out that the first version of the extended statx
interface was posted June 29, 2010:
https://www.spinics.net/lists/linux-fsdevel/msg33831.html
* 'rebased-statx' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/vfs:
statx: Add a system call to make enhanced file info available
|
|
Pull block layer fixes from Jens Axboe:
"A collection of fixes for this merge window, either fixes for existing
issues, or parts that were waiting for acks to come in. This pull
request contains:
- Allocation of nvme queues on the right node from Shaohua.
This was ready long before the merge window, but waiting on an ack
from Bjorn on the PCI bit. Now that we have that, the three patches
can go in.
- Two fixes for blk-mq-sched with nvmeof, which uses hctx specific
request allocations. This caused an oops. One part from Sagi, one
part from Omar.
- A loop partition scan deadlock fix from Omar, fixing a regression
in this merge window.
- A three-patch series from Keith, closing up a hole on clearing out
requests on shutdown/resume.
- A stable fix for nbd from Josef, fixing a leak of sockets.
- Two fixes for a regression in this window from Jan, fixing a
problem with one of his earlier patches dealing with queue vs bdi
life times.
- A fix for a regression with virtio-blk, causing an IO stall if
scheduling is used. From me.
- A fix for an io context lock ordering problem. From me"
* 'for-linus' of git://git.kernel.dk/linux-block:
block: Move bdi_unregister() to del_gendisk()
blk-mq: ensure that bd->last is always set correctly
block: don't call ioc_exit_icq() with the queue lock held for blk-mq
block: Initialize bd_bdi on inode initialization
loop: fix LO_FLAGS_PARTSCAN hang
nvme: Complete all stuck requests
blk-mq: Provide freeze queue timeout
blk-mq: Export blk_mq_freeze_queue_wait
nbd: stop leaking sockets
blk-mq: move update of tags->rqs to __blk_mq_alloc_request()
blk-mq: kill blk_mq_set_alloc_data()
blk-mq: make blk_mq_alloc_request_hctx() allocate a scheduler request
blk-mq-sched: Allocate sched reserved tags as specified in the original queue tagset
nvme: allocate nvme_queue in correct node
PCI: add an API to get node from vector
blk-mq: allocate blk_mq_tags and requests in correct node
|
|
git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip
Pull sched.h split-up from Ingo Molnar:
"The point of these changes is to significantly reduce the
<linux/sched.h> header footprint, to speed up the kernel build and to
have a cleaner header structure.
After these changes the new <linux/sched.h>'s typical preprocessed
size goes down from a previous ~0.68 MB (~22K lines) to ~0.45 MB (~15K
lines), which is around 40% faster to build on typical configs.
Not much changed from the last version (-v2) posted three weeks ago: I
eliminated quirks, backmerged fixes plus I rebased it to an upstream
SHA1 from yesterday that includes most changes queued up in -next plus
all sched.h changes that were pending from Andrew.
I've re-tested the series both on x86 and on cross-arch defconfigs,
and did a bisectability test at a number of random points.
I tried to test as many build configurations as possible, but some
build breakage is probably still left - but it should be mostly
limited to architectures that have no cross-compiler binaries
available on kernel.org, and non-default configurations"
* 'WIP.sched-core-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip: (146 commits)
sched/headers: Clean up <linux/sched.h>
sched/headers: Remove #ifdefs from <linux/sched.h>
sched/headers: Remove the <linux/topology.h> include from <linux/sched.h>
sched/headers, hrtimer: Remove the <linux/wait.h> include from <linux/hrtimer.h>
sched/headers, x86/apic: Remove the <linux/pm.h> header inclusion from <asm/apic.h>
sched/headers, timers: Remove the <linux/sysctl.h> include from <linux/timer.h>
sched/headers: Remove <linux/magic.h> from <linux/sched/task_stack.h>
sched/headers: Remove <linux/sched.h> from <linux/sched/init.h>
sched/core: Remove unused prefetch_stack()
sched/headers: Remove <linux/rculist.h> from <linux/sched.h>
sched/headers: Remove the 'init_pid_ns' prototype from <linux/sched.h>
sched/headers: Remove <linux/signal.h> from <linux/sched.h>
sched/headers: Remove <linux/rwsem.h> from <linux/sched.h>
sched/headers: Remove the runqueue_is_locked() prototype
sched/headers: Remove <linux/sched.h> from <linux/sched/hotplug.h>
sched/headers: Remove <linux/sched.h> from <linux/sched/debug.h>
sched/headers: Remove <linux/sched.h> from <linux/sched/nohz.h>
sched/headers: Remove <linux/sched.h> from <linux/sched/stat.h>
sched/headers: Remove the <linux/gfp.h> include from <linux/sched.h>
sched/headers: Remove <linux/rtmutex.h> from <linux/sched.h>
...
|