[PATCH 0/8 v3] scope GFP_NOFS api

classic Classic list List threaded Threaded
46 messages Options
123
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

[PATCH 0/8 v3] scope GFP_NOFS api

Michal Hocko
Hi,
I have posted the previous version here [1]. Since then I've added some
reviewed bys and fixed some minor issues. I've dropped patch 2 [2] based
on Dave's request [3]. I agree that this can be done later and doing
all at once. I still think that __GFP_NOLOCKDEP should be added by this
series to make the further development easier.

There didn't seem to be any real objections and so I think we should go
and merge this and build further on top. I would like to get rid of all
explicit GFP_NOFS usage in ext4 code. I have something half baked already
and will send it later on. I also hope we can get further with the xfs
as well.

I haven't heard anything from btrfs or other filesystems guys which is a
bit unfortunate but I do not want to wait for them to much longer, they
can join the effort later on.

The patchset is based on next-20170106

Diffstat says
 fs/ext4/acl.c             |  6 +++---
 fs/ext4/extents.c         |  8 ++++----
 fs/ext4/resize.c          |  4 ++--
 fs/ext4/xattr.c           |  4 ++--
 fs/jbd2/journal.c         |  7 +++++++
 fs/jbd2/transaction.c     | 11 +++++++++++
 fs/xfs/kmem.c             | 10 +++++-----
 fs/xfs/kmem.h             |  2 +-
 fs/xfs/libxfs/xfs_btree.c |  2 +-
 fs/xfs/xfs_aops.c         |  6 +++---
 fs/xfs/xfs_buf.c          |  8 ++++----
 fs/xfs/xfs_trans.c        | 12 ++++++------
 include/linux/gfp.h       | 18 +++++++++++++++++-
 include/linux/jbd2.h      |  2 ++
 include/linux/sched.h     | 32 ++++++++++++++++++++++++++------
 kernel/locking/lockdep.c  |  6 +++++-
 lib/radix-tree.c          |  2 ++
 mm/page_alloc.c           |  8 +++++---
 mm/vmscan.c               |  6 +++---
 19 files changed, 109 insertions(+), 45 deletions(-)

Shortlog:
Michal Hocko (8):
      lockdep: allow to disable reclaim lockup detection
      xfs: abstract PF_FSTRANS to PF_MEMALLOC_NOFS
      mm: introduce memalloc_nofs_{save,restore} API
      xfs: use memalloc_nofs_{save,restore} instead of memalloc_noio*
      jbd2: mark the transaction context with the scope GFP_NOFS context
      jbd2: make the whole kjournald2 kthread NOFS safe
      Revert "ext4: avoid deadlocks in the writeback path by using sb_getblk_gfp"
      Revert "ext4: fix wrong gfp type under transaction"

[1] http://lkml.kernel.org/r/20161215140715.12732-1-mhocko@...
[2] http://lkml.kernel.org/r/20161215140715.12732-3-mhocko@...
[3] http://lkml.kernel.org/r/20161219212413.GN4326@dastard



------------------------------------------------------------------------------
Check out the vibrant tech community on one of the world's most
engaging tech sites, SlashDot.org! http://sdm.link/slashdot
_______________________________________________
Linux-NTFS-Dev mailing list
[hidden email]
https://lists.sourceforge.net/lists/listinfo/linux-ntfs-dev
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

[PATCH 1/8] lockdep: allow to disable reclaim lockup detection

Michal Hocko
From: Michal Hocko <[hidden email]>

The current implementation of the reclaim lockup detection can lead to
false positives and those even happen and usually lead to tweak the
code to silence the lockdep by using GFP_NOFS even though the context
can use __GFP_FS just fine. See
http://lkml.kernel.org/r/20160512080321.GA18496@dastard as an example.

=================================
[ INFO: inconsistent lock state ]
4.5.0-rc2+ #4 Tainted: G           O
---------------------------------
inconsistent {RECLAIM_FS-ON-R} -> {IN-RECLAIM_FS-W} usage.
kswapd0/543 [HC0[0]:SC0[0]:HE1:SE1] takes:

(&xfs_nondir_ilock_class){++++-+}, at: [<ffffffffa00781f7>] xfs_ilock+0x177/0x200 [xfs]

{RECLAIM_FS-ON-R} state was registered at:
  [<ffffffff8110f369>] mark_held_locks+0x79/0xa0
  [<ffffffff81113a43>] lockdep_trace_alloc+0xb3/0x100
  [<ffffffff81224623>] kmem_cache_alloc+0x33/0x230
  [<ffffffffa008acc1>] kmem_zone_alloc+0x81/0x120 [xfs]
  [<ffffffffa005456e>] xfs_refcountbt_init_cursor+0x3e/0xa0 [xfs]
  [<ffffffffa0053455>] __xfs_refcount_find_shared+0x75/0x580 [xfs]
  [<ffffffffa00539e4>] xfs_refcount_find_shared+0x84/0xb0 [xfs]
  [<ffffffffa005dcb8>] xfs_getbmap+0x608/0x8c0 [xfs]
  [<ffffffffa007634b>] xfs_vn_fiemap+0xab/0xc0 [xfs]
  [<ffffffff81244208>] do_vfs_ioctl+0x498/0x670
  [<ffffffff81244459>] SyS_ioctl+0x79/0x90
  [<ffffffff81847cd7>] entry_SYSCALL_64_fastpath+0x12/0x6f

       CPU0
       ----
  lock(&xfs_nondir_ilock_class);
  <Interrupt>
    lock(&xfs_nondir_ilock_class);

 *** DEADLOCK ***

3 locks held by kswapd0/543:

stack backtrace:
CPU: 0 PID: 543 Comm: kswapd0 Tainted: G           O    4.5.0-rc2+ #4

Hardware name: innotek GmbH VirtualBox/VirtualBox, BIOS VirtualBox 12/01/2006

 ffffffff82a34f10 ffff88003aa078d0 ffffffff813a14f9 ffff88003d8551c0
 ffff88003aa07920 ffffffff8110ec65 0000000000000000 0000000000000001
 ffff880000000001 000000000000000b 0000000000000008 ffff88003d855aa0
Call Trace:
 [<ffffffff813a14f9>] dump_stack+0x4b/0x72
 [<ffffffff8110ec65>] print_usage_bug+0x215/0x240
 [<ffffffff8110ee85>] mark_lock+0x1f5/0x660
 [<ffffffff8110e100>] ? print_shortest_lock_dependencies+0x1a0/0x1a0
 [<ffffffff811102e0>] __lock_acquire+0xa80/0x1e50
 [<ffffffff8122474e>] ? kmem_cache_alloc+0x15e/0x230
 [<ffffffffa008acc1>] ? kmem_zone_alloc+0x81/0x120 [xfs]
 [<ffffffff811122e8>] lock_acquire+0xd8/0x1e0
 [<ffffffffa00781f7>] ? xfs_ilock+0x177/0x200 [xfs]
 [<ffffffffa0083a70>] ? xfs_reflink_cancel_cow_range+0x150/0x300 [xfs]
 [<ffffffff8110aace>] down_write_nested+0x5e/0xc0
 [<ffffffffa00781f7>] ? xfs_ilock+0x177/0x200 [xfs]
 [<ffffffffa00781f7>] xfs_ilock+0x177/0x200 [xfs]
 [<ffffffffa0083a70>] xfs_reflink_cancel_cow_range+0x150/0x300 [xfs]
 [<ffffffffa0085bdc>] xfs_fs_evict_inode+0xdc/0x1e0 [xfs]
 [<ffffffff8124d7d5>] evict+0xc5/0x190
 [<ffffffff8124d8d9>] dispose_list+0x39/0x60
 [<ffffffff8124eb2b>] prune_icache_sb+0x4b/0x60
 [<ffffffff8123317f>] super_cache_scan+0x14f/0x1a0
 [<ffffffff811e0d19>] shrink_slab.part.63.constprop.79+0x1e9/0x4e0
 [<ffffffff811e50ee>] shrink_zone+0x15e/0x170
 [<ffffffff811e5ef1>] kswapd+0x4f1/0xa80
 [<ffffffff811e5a00>] ? zone_reclaim+0x230/0x230
 [<ffffffff810e6882>] kthread+0xf2/0x110
 [<ffffffff810e6790>] ? kthread_create_on_node+0x220/0x220
 [<ffffffff8184803f>] ret_from_fork+0x3f/0x70
 [<ffffffff810e6790>] ? kthread_create_on_node+0x220/0x220

To quote Dave:
"
Ignoring whether reflink should be doing anything or not, that's a
"xfs_refcountbt_init_cursor() gets called both outside and inside
transactions" lockdep false positive case. The problem here is
lockdep has seen this allocation from within a transaction, hence a
GFP_NOFS allocation, and now it's seeing it in a GFP_KERNEL context.
Also note that we have an active reference to this inode.

So, because the reclaim annotations overload the interrupt level
detections and it's seen the inode ilock been taken in reclaim
("interrupt") context, this triggers a reclaim context warning where
it thinks it is unsafe to do this allocation in GFP_KERNEL context
holding the inode ilock...
"

This sounds like a fundamental problem of the reclaim lock detection.
It is really impossible to annotate such a special usecase IMHO unless
the reclaim lockup detection is reworked completely. Until then it
is much better to provide a way to add "I know what I am doing flag"
and mark problematic places. This would prevent from abusing GFP_NOFS
flag which has a runtime effect even on configurations which have
lockdep disabled.

Introduce __GFP_NOLOCKDEP flag which tells the lockdep gfp tracking to
skip the current allocation request.

While we are at it also make sure that the radix tree doesn't
accidentaly override tags stored in the upper part of the gfp_mask.

Suggested-by: Peter Zijlstra <[hidden email]>
Acked-by: Peter Zijlstra (Intel) <[hidden email]>
Signed-off-by: Michal Hocko <[hidden email]>
---
 include/linux/gfp.h      | 10 +++++++++-
 kernel/locking/lockdep.c |  4 ++++
 lib/radix-tree.c         |  2 ++
 3 files changed, 15 insertions(+), 1 deletion(-)

diff --git a/include/linux/gfp.h b/include/linux/gfp.h
index 4175dca4ac39..1a934383cc20 100644
--- a/include/linux/gfp.h
+++ b/include/linux/gfp.h
@@ -41,6 +41,11 @@ struct vm_area_struct;
 #define ___GFP_OTHER_NODE 0x800000u
 #define ___GFP_WRITE 0x1000000u
 #define ___GFP_KSWAPD_RECLAIM 0x2000000u
+#ifdef CONFIG_LOCKDEP
+#define ___GFP_NOLOCKDEP 0x4000000u
+#else
+#define ___GFP_NOLOCKDEP 0
+#endif
 /* If the above are modified, __GFP_BITS_SHIFT may need updating */
 
 /*
@@ -186,8 +191,11 @@ struct vm_area_struct;
 #define __GFP_NOTRACK_FALSE_POSITIVE (__GFP_NOTRACK)
 #define __GFP_OTHER_NODE ((__force gfp_t)___GFP_OTHER_NODE)
 
+/* Disable lockdep for GFP context tracking */
+#define __GFP_NOLOCKDEP ((__force gfp_t)___GFP_NOLOCKDEP)
+
 /* Room for N __GFP_FOO bits */
-#define __GFP_BITS_SHIFT 26
+#define __GFP_BITS_SHIFT (26 + IS_ENABLED(CONFIG_LOCKDEP))
 #define __GFP_BITS_MASK ((__force gfp_t)((1 << __GFP_BITS_SHIFT) - 1))
 
 /*
diff --git a/kernel/locking/lockdep.c b/kernel/locking/lockdep.c
index 844cd04bb453..59e94ce8a0cf 100644
--- a/kernel/locking/lockdep.c
+++ b/kernel/locking/lockdep.c
@@ -2879,6 +2879,10 @@ static void __lockdep_trace_alloc(gfp_t gfp_mask, unsigned long flags)
  if (DEBUG_LOCKS_WARN_ON(irqs_disabled_flags(flags)))
  return;
 
+ /* Disable lockdep if explicitly requested */
+ if (gfp_mask & __GFP_NOLOCKDEP)
+ return;
+
  mark_held_locks(curr, RECLAIM_FS);
 }
 
diff --git a/lib/radix-tree.c b/lib/radix-tree.c
index 5ccf00277233..c63b719346ae 100644
--- a/lib/radix-tree.c
+++ b/lib/radix-tree.c
@@ -2198,6 +2198,8 @@ static int radix_tree_cpu_dead(unsigned int cpu)
 void __init radix_tree_init(void)
 {
  int ret;
+
+ BUILD_BUG_ON(RADIX_TREE_MAX_TAGS + __GFP_BITS_SHIFT > 32);
  radix_tree_node_cachep = kmem_cache_create("radix_tree_node",
  sizeof(struct radix_tree_node), 0,
  SLAB_PANIC | SLAB_RECLAIM_ACCOUNT,
--
2.11.0


------------------------------------------------------------------------------
Check out the vibrant tech community on one of the world's most
engaging tech sites, SlashDot.org! http://sdm.link/slashdot
_______________________________________________
Linux-NTFS-Dev mailing list
[hidden email]
https://lists.sourceforge.net/lists/listinfo/linux-ntfs-dev
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

[PATCH 2/8] xfs: abstract PF_FSTRANS to PF_MEMALLOC_NOFS

Michal Hocko
In reply to this post by Michal Hocko
From: Michal Hocko <[hidden email]>

xfs has defined PF_FSTRANS to declare a scope GFP_NOFS semantic quite
some time ago. We would like to make this concept more generic and use
it for other filesystems as well. Let's start by giving the flag a
more generic name PF_MEMALLOC_NOFS which is in line with an exiting
PF_MEMALLOC_NOIO already used for the same purpose for GFP_NOIO
contexts. Replace all PF_FSTRANS usage from the xfs code in the first
step before we introduce a full API for it as xfs uses the flag directly
anyway.

This patch doesn't introduce any functional change.

Signed-off-by: Michal Hocko <[hidden email]>
Reviewed-by: Brian Foster <[hidden email]>
---
 fs/xfs/kmem.c             |  4 ++--
 fs/xfs/kmem.h             |  2 +-
 fs/xfs/libxfs/xfs_btree.c |  2 +-
 fs/xfs/xfs_aops.c         |  6 +++---
 fs/xfs/xfs_trans.c        | 12 ++++++------
 include/linux/sched.h     |  2 ++
 6 files changed, 15 insertions(+), 13 deletions(-)

diff --git a/fs/xfs/kmem.c b/fs/xfs/kmem.c
index 339c696bbc01..a76a05dae96b 100644
--- a/fs/xfs/kmem.c
+++ b/fs/xfs/kmem.c
@@ -80,13 +80,13 @@ kmem_zalloc_large(size_t size, xfs_km_flags_t flags)
  * context via PF_MEMALLOC_NOIO to prevent memory reclaim re-entering
  * the filesystem here and potentially deadlocking.
  */
- if ((current->flags & PF_FSTRANS) || (flags & KM_NOFS))
+ if ((current->flags & PF_MEMALLOC_NOFS) || (flags & KM_NOFS))
  noio_flag = memalloc_noio_save();
 
  lflags = kmem_flags_convert(flags);
  ptr = __vmalloc(size, lflags | __GFP_HIGHMEM | __GFP_ZERO, PAGE_KERNEL);
 
- if ((current->flags & PF_FSTRANS) || (flags & KM_NOFS))
+ if ((current->flags & PF_MEMALLOC_NOFS) || (flags & KM_NOFS))
  memalloc_noio_restore(noio_flag);
 
  return ptr;
diff --git a/fs/xfs/kmem.h b/fs/xfs/kmem.h
index 689f746224e7..d973dbfc2bfa 100644
--- a/fs/xfs/kmem.h
+++ b/fs/xfs/kmem.h
@@ -50,7 +50,7 @@ kmem_flags_convert(xfs_km_flags_t flags)
  lflags = GFP_ATOMIC | __GFP_NOWARN;
  } else {
  lflags = GFP_KERNEL | __GFP_NOWARN;
- if ((current->flags & PF_FSTRANS) || (flags & KM_NOFS))
+ if ((current->flags & PF_MEMALLOC_NOFS) || (flags & KM_NOFS))
  lflags &= ~__GFP_FS;
  }
 
diff --git a/fs/xfs/libxfs/xfs_btree.c b/fs/xfs/libxfs/xfs_btree.c
index 21e6a6ab6b9a..a2672ba4dc33 100644
--- a/fs/xfs/libxfs/xfs_btree.c
+++ b/fs/xfs/libxfs/xfs_btree.c
@@ -2866,7 +2866,7 @@ xfs_btree_split_worker(
  struct xfs_btree_split_args *args = container_of(work,
  struct xfs_btree_split_args, work);
  unsigned long pflags;
- unsigned long new_pflags = PF_FSTRANS;
+ unsigned long new_pflags = PF_MEMALLOC_NOFS;
 
  /*
  * we are in a transaction context here, but may also be doing work
diff --git a/fs/xfs/xfs_aops.c b/fs/xfs/xfs_aops.c
index ef382bfb402b..d4094bb55033 100644
--- a/fs/xfs/xfs_aops.c
+++ b/fs/xfs/xfs_aops.c
@@ -189,7 +189,7 @@ xfs_setfilesize_trans_alloc(
  * We hand off the transaction to the completion thread now, so
  * clear the flag here.
  */
- current_restore_flags_nested(&tp->t_pflags, PF_FSTRANS);
+ current_restore_flags_nested(&tp->t_pflags, PF_MEMALLOC_NOFS);
  return 0;
 }
 
@@ -252,7 +252,7 @@ xfs_setfilesize_ioend(
  * thus we need to mark ourselves as being in a transaction manually.
  * Similarly for freeze protection.
  */
- current_set_flags_nested(&tp->t_pflags, PF_FSTRANS);
+ current_set_flags_nested(&tp->t_pflags, PF_MEMALLOC_NOFS);
  __sb_writers_acquired(VFS_I(ip)->i_sb, SB_FREEZE_FS);
 
  /* we abort the update if there was an IO error */
@@ -1015,7 +1015,7 @@ xfs_do_writepage(
  * Given that we do not allow direct reclaim to call us, we should
  * never be called while in a filesystem transaction.
  */
- if (WARN_ON_ONCE(current->flags & PF_FSTRANS))
+ if (WARN_ON_ONCE(current->flags & PF_MEMALLOC_NOFS))
  goto redirty;
 
  /*
diff --git a/fs/xfs/xfs_trans.c b/fs/xfs/xfs_trans.c
index 70f42ea86dfb..f5969c8274fc 100644
--- a/fs/xfs/xfs_trans.c
+++ b/fs/xfs/xfs_trans.c
@@ -134,7 +134,7 @@ xfs_trans_reserve(
  bool rsvd = (tp->t_flags & XFS_TRANS_RESERVE) != 0;
 
  /* Mark this thread as being in a transaction */
- current_set_flags_nested(&tp->t_pflags, PF_FSTRANS);
+ current_set_flags_nested(&tp->t_pflags, PF_MEMALLOC_NOFS);
 
  /*
  * Attempt to reserve the needed disk blocks by decrementing
@@ -144,7 +144,7 @@ xfs_trans_reserve(
  if (blocks > 0) {
  error = xfs_mod_fdblocks(tp->t_mountp, -((int64_t)blocks), rsvd);
  if (error != 0) {
- current_restore_flags_nested(&tp->t_pflags, PF_FSTRANS);
+ current_restore_flags_nested(&tp->t_pflags, PF_MEMALLOC_NOFS);
  return -ENOSPC;
  }
  tp->t_blk_res += blocks;
@@ -221,7 +221,7 @@ xfs_trans_reserve(
  tp->t_blk_res = 0;
  }
 
- current_restore_flags_nested(&tp->t_pflags, PF_FSTRANS);
+ current_restore_flags_nested(&tp->t_pflags, PF_MEMALLOC_NOFS);
 
  return error;
 }
@@ -914,7 +914,7 @@ __xfs_trans_commit(
 
  xfs_log_commit_cil(mp, tp, &commit_lsn, regrant);
 
- current_restore_flags_nested(&tp->t_pflags, PF_FSTRANS);
+ current_restore_flags_nested(&tp->t_pflags, PF_MEMALLOC_NOFS);
  xfs_trans_free(tp);
 
  /*
@@ -944,7 +944,7 @@ __xfs_trans_commit(
  if (commit_lsn == -1 && !error)
  error = -EIO;
  }
- current_restore_flags_nested(&tp->t_pflags, PF_FSTRANS);
+ current_restore_flags_nested(&tp->t_pflags, PF_MEMALLOC_NOFS);
  xfs_trans_free_items(tp, NULLCOMMITLSN, !!error);
  xfs_trans_free(tp);
 
@@ -998,7 +998,7 @@ xfs_trans_cancel(
  xfs_log_done(mp, tp->t_ticket, NULL, false);
 
  /* mark this thread as no longer being in a transaction */
- current_restore_flags_nested(&tp->t_pflags, PF_FSTRANS);
+ current_restore_flags_nested(&tp->t_pflags, PF_MEMALLOC_NOFS);
 
  xfs_trans_free_items(tp, NULLCOMMITLSN, dirty);
  xfs_trans_free(tp);
diff --git a/include/linux/sched.h b/include/linux/sched.h
index 1531c48f56e2..abeb84604d32 100644
--- a/include/linux/sched.h
+++ b/include/linux/sched.h
@@ -2320,6 +2320,8 @@ extern void thread_group_cputime_adjusted(struct task_struct *p, cputime_t *ut,
 #define PF_FREEZER_SKIP 0x40000000 /* Freezer should not count it as freezable */
 #define PF_SUSPEND_TASK 0x80000000      /* this thread called freeze_processes and should not be frozen */
 
+#define PF_MEMALLOC_NOFS PF_FSTRANS /* Transition to a more generic GFP_NOFS scope semantic */
+
 /*
  * Only the _current_ task can read/write to tsk->flags, but other
  * tasks can access tsk->flags in readonly mode for example
--
2.11.0


------------------------------------------------------------------------------
Check out the vibrant tech community on one of the world's most
engaging tech sites, SlashDot.org! http://sdm.link/slashdot
_______________________________________________
Linux-NTFS-Dev mailing list
[hidden email]
https://lists.sourceforge.net/lists/listinfo/linux-ntfs-dev
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

[PATCH 3/8] mm: introduce memalloc_nofs_{save, restore} API

Michal Hocko
In reply to this post by Michal Hocko
From: Michal Hocko <[hidden email]>

GFP_NOFS context is used for the following 5 reasons currently
        - to prevent from deadlocks when the lock held by the allocation
          context would be needed during the memory reclaim
        - to prevent from stack overflows during the reclaim because
          the allocation is performed from a deep context already
        - to prevent lockups when the allocation context depends on
          other reclaimers to make a forward progress indirectly
        - just in case because this would be safe from the fs POV
        - silence lockdep false positives

Unfortunately overuse of this allocation context brings some problems
to the MM. Memory reclaim is much weaker (especially during heavy FS
metadata workloads), OOM killer cannot be invoked because the MM layer
doesn't have enough information about how much memory is freeable by the
FS layer.

In many cases it is far from clear why the weaker context is even used
and so it might be used unnecessarily. We would like to get rid of
those as much as possible. One way to do that is to use the flag in
scopes rather than isolated cases. Such a scope is declared when really
necessary, tracked per task and all the allocation requests from within
the context will simply inherit the GFP_NOFS semantic.

Not only this is easier to understand and maintain because there are
much less problematic contexts than specific allocation requests, this
also helps code paths where FS layer interacts with other layers (e.g.
crypto, security modules, MM etc...) and there is no easy way to convey
the allocation context between the layers.

Introduce memalloc_nofs_{save,restore} API to control the scope
of GFP_NOFS allocation context. This is basically copying
memalloc_noio_{save,restore} API we have for other restricted allocation
context GFP_NOIO. The PF_MEMALLOC_NOFS flag already exists and it is
just an alias for PF_FSTRANS which has been xfs specific until recently.
There are no more PF_FSTRANS users anymore so let's just drop it.

PF_MEMALLOC_NOFS is now checked in the MM layer and drops __GFP_FS
implicitly same as PF_MEMALLOC_NOIO drops __GFP_IO. memalloc_noio_flags
is renamed to current_gfp_context because it now cares about both
PF_MEMALLOC_NOFS and PF_MEMALLOC_NOIO contexts. Xfs code paths preserve
their semantic. kmem_flags_convert() doesn't need to evaluate the flag
anymore.

This patch shouldn't introduce any functional changes.

Let's hope that filesystems will drop direct GFP_NOFS (resp. ~__GFP_FS)
usage as much as possible and only use a properly documented
memalloc_nofs_{save,restore} checkpoints where they are appropriate.

Signed-off-by: Michal Hocko <[hidden email]>
---
 fs/xfs/kmem.h            |  2 +-
 include/linux/gfp.h      |  8 ++++++++
 include/linux/sched.h    | 34 ++++++++++++++++++++++++++--------
 kernel/locking/lockdep.c |  2 +-
 mm/page_alloc.c          |  8 +++++---
 mm/vmscan.c              |  6 +++---
 6 files changed, 44 insertions(+), 16 deletions(-)

diff --git a/fs/xfs/kmem.h b/fs/xfs/kmem.h
index d973dbfc2bfa..ae08cfd9552a 100644
--- a/fs/xfs/kmem.h
+++ b/fs/xfs/kmem.h
@@ -50,7 +50,7 @@ kmem_flags_convert(xfs_km_flags_t flags)
  lflags = GFP_ATOMIC | __GFP_NOWARN;
  } else {
  lflags = GFP_KERNEL | __GFP_NOWARN;
- if ((current->flags & PF_MEMALLOC_NOFS) || (flags & KM_NOFS))
+ if (flags & KM_NOFS)
  lflags &= ~__GFP_FS;
  }
 
diff --git a/include/linux/gfp.h b/include/linux/gfp.h
index 1a934383cc20..bfe53d95c25b 100644
--- a/include/linux/gfp.h
+++ b/include/linux/gfp.h
@@ -217,8 +217,16 @@ struct vm_area_struct;
  *
  * GFP_NOIO will use direct reclaim to discard clean pages or slab pages
  *   that do not require the starting of any physical IO.
+ *   Please try to avoid using this flag directly and instead use
+ *   memalloc_noio_{save,restore} to mark the whole scope which cannot
+ *   perform any IO with a short explanation why. All allocation requests
+ *   will inherit GFP_NOIO implicitly.
  *
  * GFP_NOFS will use direct reclaim but will not use any filesystem interfaces.
+ *   Please try to avoid using this flag directly and instead use
+ *   memalloc_nofs_{save,restore} to mark the whole scope which cannot/shouldn't
+ *   recurse into the FS layer with a short explanation why. All allocation
+ *   requests will inherit GFP_NOFS implicitly.
  *
  * GFP_USER is for userspace allocations that also need to be directly
  *   accessibly by the kernel or hardware. It is typically used by hardware
diff --git a/include/linux/sched.h b/include/linux/sched.h
index abeb84604d32..2032fc642a26 100644
--- a/include/linux/sched.h
+++ b/include/linux/sched.h
@@ -2307,9 +2307,9 @@ extern void thread_group_cputime_adjusted(struct task_struct *p, cputime_t *ut,
 #define PF_USED_ASYNC 0x00004000 /* used async_schedule*(), used by module init */
 #define PF_NOFREEZE 0x00008000 /* this thread should not be frozen */
 #define PF_FROZEN 0x00010000 /* frozen for system suspend */
-#define PF_FSTRANS 0x00020000 /* inside a filesystem transaction */
-#define PF_KSWAPD 0x00040000 /* I am kswapd */
-#define PF_MEMALLOC_NOIO 0x00080000 /* Allocating memory without IO involved */
+#define PF_KSWAPD 0x00020000 /* I am kswapd */
+#define PF_MEMALLOC_NOFS 0x00040000 /* All allocation requests will inherit GFP_NOFS */
+#define PF_MEMALLOC_NOIO 0x00080000 /* All allocation requests will inherit GFP_NOIO */
 #define PF_LESS_THROTTLE 0x00100000 /* Throttle me less: I clean memory */
 #define PF_KTHREAD 0x00200000 /* I am a kernel thread */
 #define PF_RANDOMIZE 0x00400000 /* randomize virtual address space */
@@ -2320,8 +2320,6 @@ extern void thread_group_cputime_adjusted(struct task_struct *p, cputime_t *ut,
 #define PF_FREEZER_SKIP 0x40000000 /* Freezer should not count it as freezable */
 #define PF_SUSPEND_TASK 0x80000000      /* this thread called freeze_processes and should not be frozen */
 
-#define PF_MEMALLOC_NOFS PF_FSTRANS /* Transition to a more generic GFP_NOFS scope semantic */
-
 /*
  * Only the _current_ task can read/write to tsk->flags, but other
  * tasks can access tsk->flags in readonly mode for example
@@ -2347,13 +2345,21 @@ extern void thread_group_cputime_adjusted(struct task_struct *p, cputime_t *ut,
 #define tsk_used_math(p) ((p)->flags & PF_USED_MATH)
 #define used_math() tsk_used_math(current)
 
-/* __GFP_IO isn't allowed if PF_MEMALLOC_NOIO is set in current->flags
- * __GFP_FS is also cleared as it implies __GFP_IO.
+/*
+ * Applies per-task gfp context to the given allocation flags.
+ * PF_MEMALLOC_NOIO implies GFP_NOIO
+ * PF_MEMALLOC_NOFS implies GFP_NOFS
  */
-static inline gfp_t memalloc_noio_flags(gfp_t flags)
+static inline gfp_t current_gfp_context(gfp_t flags)
 {
+ /*
+ * NOIO implies both NOIO and NOFS and it is a weaker context
+ * so always make sure it makes precendence
+ */
  if (unlikely(current->flags & PF_MEMALLOC_NOIO))
  flags &= ~(__GFP_IO | __GFP_FS);
+ else if (unlikely(current->flags & PF_MEMALLOC_NOFS))
+ flags &= ~__GFP_FS;
  return flags;
 }
 
@@ -2369,6 +2375,18 @@ static inline void memalloc_noio_restore(unsigned int flags)
  current->flags = (current->flags & ~PF_MEMALLOC_NOIO) | flags;
 }
 
+static inline unsigned int memalloc_nofs_save(void)
+{
+ unsigned int flags = current->flags & PF_MEMALLOC_NOFS;
+ current->flags |= PF_MEMALLOC_NOFS;
+ return flags;
+}
+
+static inline void memalloc_nofs_restore(unsigned int flags)
+{
+ current->flags = (current->flags & ~PF_MEMALLOC_NOFS) | flags;
+}
+
 /* Per-process atomic flags. */
 #define PFA_NO_NEW_PRIVS 0 /* May not gain new privileges. */
 #define PFA_SPREAD_PAGE  1      /* Spread page cache over cpuset */
diff --git a/kernel/locking/lockdep.c b/kernel/locking/lockdep.c
index 59e94ce8a0cf..cdcd6f249ec8 100644
--- a/kernel/locking/lockdep.c
+++ b/kernel/locking/lockdep.c
@@ -2870,7 +2870,7 @@ static void __lockdep_trace_alloc(gfp_t gfp_mask, unsigned long flags)
  return;
 
  /* We're only interested __GFP_FS allocations for now */
- if (!(gfp_mask & __GFP_FS))
+ if (!(gfp_mask & __GFP_FS) || (curr->flags & PF_MEMALLOC_NOFS))
  return;
 
  /*
diff --git a/mm/page_alloc.c b/mm/page_alloc.c
index 46ad035b955f..5138b46a4295 100644
--- a/mm/page_alloc.c
+++ b/mm/page_alloc.c
@@ -3807,10 +3807,12 @@ __alloc_pages_nodemask(gfp_t gfp_mask, unsigned int order,
  goto out;
 
  /*
- * Runtime PM, block IO and its error handling path can deadlock
- * because I/O on the device might not complete.
+ * Apply scoped allocation constrains. This is mainly about
+ * GFP_NOFS resp. GFP_NOIO which has to be inherited for all
+ * allocation requests from a particular context which has
+ * been marked by memalloc_no{fs,io}_{save,restore}
  */
- alloc_mask = memalloc_noio_flags(gfp_mask);
+ alloc_mask = current_gfp_context(gfp_mask);
  ac.spread_dirty_pages = false;
 
  /*
diff --git a/mm/vmscan.c b/mm/vmscan.c
index 6aa5b01d3e75..4ea6b610f20e 100644
--- a/mm/vmscan.c
+++ b/mm/vmscan.c
@@ -2949,7 +2949,7 @@ unsigned long try_to_free_pages(struct zonelist *zonelist, int order,
  unsigned long nr_reclaimed;
  struct scan_control sc = {
  .nr_to_reclaim = SWAP_CLUSTER_MAX,
- .gfp_mask = (gfp_mask = memalloc_noio_flags(gfp_mask)),
+ .gfp_mask = (gfp_mask = current_gfp_context(gfp_mask)),
  .reclaim_idx = gfp_zone(gfp_mask),
  .order = order,
  .nodemask = nodemask,
@@ -3029,7 +3029,7 @@ unsigned long try_to_free_mem_cgroup_pages(struct mem_cgroup *memcg,
  int nid;
  struct scan_control sc = {
  .nr_to_reclaim = max(nr_pages, SWAP_CLUSTER_MAX),
- .gfp_mask = (gfp_mask & GFP_RECLAIM_MASK) |
+ .gfp_mask = (current_gfp_context(gfp_mask) & GFP_RECLAIM_MASK) |
  (GFP_HIGHUSER_MOVABLE & ~GFP_RECLAIM_MASK),
  .reclaim_idx = MAX_NR_ZONES - 1,
  .target_mem_cgroup = memcg,
@@ -3723,7 +3723,7 @@ static int __node_reclaim(struct pglist_data *pgdat, gfp_t gfp_mask, unsigned in
  int classzone_idx = gfp_zone(gfp_mask);
  struct scan_control sc = {
  .nr_to_reclaim = max(nr_pages, SWAP_CLUSTER_MAX),
- .gfp_mask = (gfp_mask = memalloc_noio_flags(gfp_mask)),
+ .gfp_mask = (gfp_mask = current_gfp_context(gfp_mask)),
  .order = order,
  .priority = NODE_RECLAIM_PRIORITY,
  .may_writepage = !!(node_reclaim_mode & RECLAIM_WRITE),
--
2.11.0


------------------------------------------------------------------------------
Check out the vibrant tech community on one of the world's most
engaging tech sites, SlashDot.org! http://sdm.link/slashdot
_______________________________________________
Linux-NTFS-Dev mailing list
[hidden email]
https://lists.sourceforge.net/lists/listinfo/linux-ntfs-dev
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

[PATCH 4/8] xfs: use memalloc_nofs_{save, restore} instead of memalloc_noio*

Michal Hocko
In reply to this post by Michal Hocko
From: Michal Hocko <[hidden email]>

kmem_zalloc_large and _xfs_buf_map_pages use memalloc_noio_{save,restore}
API to prevent from reclaim recursion into the fs because vmalloc can
invoke unconditional GFP_KERNEL allocations and these functions might be
called from the NOFS contexts. The memalloc_noio_save will enforce
GFP_NOIO context which is even weaker than GFP_NOFS and that seems to be
unnecessary. Let's use memalloc_nofs_{save,restore} instead as it should
provide exactly what we need here - implicit GFP_NOFS context.

Changes since v1
- s@memalloc_noio_restore@memalloc_nofs_restore@ in _xfs_buf_map_pages
  as per Brian Foster

Signed-off-by: Michal Hocko <[hidden email]>
---
 fs/xfs/kmem.c    | 10 +++++-----
 fs/xfs/xfs_buf.c |  8 ++++----
 2 files changed, 9 insertions(+), 9 deletions(-)

diff --git a/fs/xfs/kmem.c b/fs/xfs/kmem.c
index a76a05dae96b..d69ed5e76621 100644
--- a/fs/xfs/kmem.c
+++ b/fs/xfs/kmem.c
@@ -65,7 +65,7 @@ kmem_alloc(size_t size, xfs_km_flags_t flags)
 void *
 kmem_zalloc_large(size_t size, xfs_km_flags_t flags)
 {
- unsigned noio_flag = 0;
+ unsigned nofs_flag = 0;
  void *ptr;
  gfp_t lflags;
 
@@ -80,14 +80,14 @@ kmem_zalloc_large(size_t size, xfs_km_flags_t flags)
  * context via PF_MEMALLOC_NOIO to prevent memory reclaim re-entering
  * the filesystem here and potentially deadlocking.
  */
- if ((current->flags & PF_MEMALLOC_NOFS) || (flags & KM_NOFS))
- noio_flag = memalloc_noio_save();
+ if (flags & KM_NOFS)
+ nofs_flag = memalloc_nofs_save();
 
  lflags = kmem_flags_convert(flags);
  ptr = __vmalloc(size, lflags | __GFP_HIGHMEM | __GFP_ZERO, PAGE_KERNEL);
 
- if ((current->flags & PF_MEMALLOC_NOFS) || (flags & KM_NOFS))
- memalloc_noio_restore(noio_flag);
+ if (flags & KM_NOFS)
+ memalloc_nofs_restore(nofs_flag);
 
  return ptr;
 }
diff --git a/fs/xfs/xfs_buf.c b/fs/xfs/xfs_buf.c
index 7f0a01f7b592..8cb8dd4cdfd8 100644
--- a/fs/xfs/xfs_buf.c
+++ b/fs/xfs/xfs_buf.c
@@ -441,17 +441,17 @@ _xfs_buf_map_pages(
  bp->b_addr = NULL;
  } else {
  int retried = 0;
- unsigned noio_flag;
+ unsigned nofs_flag;
 
  /*
  * vm_map_ram() will allocate auxillary structures (e.g.
  * pagetables) with GFP_KERNEL, yet we are likely to be under
  * GFP_NOFS context here. Hence we need to tell memory reclaim
- * that we are in such a context via PF_MEMALLOC_NOIO to prevent
+ * that we are in such a context via PF_MEMALLOC_NOFS to prevent
  * memory reclaim re-entering the filesystem here and
  * potentially deadlocking.
  */
- noio_flag = memalloc_noio_save();
+ nofs_flag = memalloc_nofs_save();
  do {
  bp->b_addr = vm_map_ram(bp->b_pages, bp->b_page_count,
  -1, PAGE_KERNEL);
@@ -459,7 +459,7 @@ _xfs_buf_map_pages(
  break;
  vm_unmap_aliases();
  } while (retried++ <= 1);
- memalloc_noio_restore(noio_flag);
+ memalloc_nofs_restore(nofs_flag);
 
  if (!bp->b_addr)
  return -ENOMEM;
--
2.11.0


------------------------------------------------------------------------------
Check out the vibrant tech community on one of the world's most
engaging tech sites, SlashDot.org! http://sdm.link/slashdot
_______________________________________________
Linux-NTFS-Dev mailing list
[hidden email]
https://lists.sourceforge.net/lists/listinfo/linux-ntfs-dev
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

[PATCH 5/8] jbd2: mark the transaction context with the scope GFP_NOFS context

Michal Hocko
In reply to this post by Michal Hocko
From: Michal Hocko <[hidden email]>

now that we have memalloc_nofs_{save,restore} api we can mark the whole
transaction context as implicitly GFP_NOFS. All allocations will
automatically inherit GFP_NOFS this way. This means that we do not have
to mark any of those requests with GFP_NOFS and moreover all the
ext4_kv[mz]alloc(GFP_NOFS) are also safe now because even the hardcoded
GFP_KERNEL allocations deep inside the vmalloc will be NOFS now.

Signed-off-by: Michal Hocko <[hidden email]>
Reviewed-by: Jan Kara <[hidden email]>
---
 fs/jbd2/transaction.c | 11 +++++++++++
 include/linux/jbd2.h  |  2 ++
 2 files changed, 13 insertions(+)

diff --git a/fs/jbd2/transaction.c b/fs/jbd2/transaction.c
index e1652665bd93..35a5d3d76182 100644
--- a/fs/jbd2/transaction.c
+++ b/fs/jbd2/transaction.c
@@ -388,6 +388,11 @@ static int start_this_handle(journal_t *journal, handle_t *handle,
 
  rwsem_acquire_read(&journal->j_trans_commit_map, 0, 0, _THIS_IP_);
  jbd2_journal_free_transaction(new_transaction);
+ /*
+ * Make sure that no allocations done while the transaction is
+ * open is going to recurse back to the fs layer.
+ */
+ handle->saved_alloc_context = memalloc_nofs_save();
  return 0;
 }
 
@@ -466,6 +471,7 @@ handle_t *jbd2__journal_start(journal_t *journal, int nblocks, int rsv_blocks,
  trace_jbd2_handle_start(journal->j_fs_dev->bd_dev,
  handle->h_transaction->t_tid, type,
  line_no, nblocks);
+
  return handle;
 }
 EXPORT_SYMBOL(jbd2__journal_start);
@@ -1760,6 +1766,11 @@ int jbd2_journal_stop(handle_t *handle)
  if (handle->h_rsv_handle)
  jbd2_journal_free_reserved(handle->h_rsv_handle);
 free_and_exit:
+ /*
+ * scope of th GFP_NOFS context is over here and so we can
+ * restore the original alloc context.
+ */
+ memalloc_nofs_restore(handle->saved_alloc_context);
  jbd2_free_handle(handle);
  return err;
 }
diff --git a/include/linux/jbd2.h b/include/linux/jbd2.h
index dfaa1f4dcb0c..606b6bce3a5b 100644
--- a/include/linux/jbd2.h
+++ b/include/linux/jbd2.h
@@ -491,6 +491,8 @@ struct jbd2_journal_handle
 
  unsigned long h_start_jiffies;
  unsigned int h_requested_credits;
+
+ unsigned int saved_alloc_context;
 };
 
 
--
2.11.0


------------------------------------------------------------------------------
Check out the vibrant tech community on one of the world's most
engaging tech sites, SlashDot.org! http://sdm.link/slashdot
_______________________________________________
Linux-NTFS-Dev mailing list
[hidden email]
https://lists.sourceforge.net/lists/listinfo/linux-ntfs-dev
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

[PATCH 6/8] jbd2: make the whole kjournald2 kthread NOFS safe

Michal Hocko
In reply to this post by Michal Hocko
From: Michal Hocko <[hidden email]>

kjournald2 is central to the transaction commit processing. As such any
potential allocation from this kernel thread has to be GFP_NOFS. Make
sure to mark the whole kernel thread GFP_NOFS by the memalloc_nofs_save.

Suggested-by: Jan Kara <[hidden email]>
Signed-off-by: Michal Hocko <[hidden email]>
Reviewed-by: Jan Kara <[hidden email]>
---
 fs/jbd2/journal.c | 7 +++++++
 1 file changed, 7 insertions(+)

diff --git a/fs/jbd2/journal.c b/fs/jbd2/journal.c
index a097048ed1a3..3a449150f834 100644
--- a/fs/jbd2/journal.c
+++ b/fs/jbd2/journal.c
@@ -206,6 +206,13 @@ static int kjournald2(void *arg)
  wake_up(&journal->j_wait_done_commit);
 
  /*
+ * Make sure that no allocations from this kernel thread will ever recurse
+ * to the fs layer because we are responsible for the transaction commit
+ * and any fs involvement might get stuck waiting for the trasn. commit.
+ */
+ memalloc_nofs_save();
+
+ /*
  * And now, wait forever for commit wakeup events.
  */
  write_lock(&journal->j_state_lock);
--
2.11.0


------------------------------------------------------------------------------
Check out the vibrant tech community on one of the world's most
engaging tech sites, SlashDot.org! http://sdm.link/slashdot
_______________________________________________
Linux-NTFS-Dev mailing list
[hidden email]
https://lists.sourceforge.net/lists/listinfo/linux-ntfs-dev
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

[PATCH 7/8] Revert "ext4: avoid deadlocks in the writeback path by using sb_getblk_gfp"

Michal Hocko
In reply to this post by Michal Hocko
From: Michal Hocko <[hidden email]>

This reverts commit c45653c341f5c8a0ce19c8f0ad4678640849cb86 because
sb_getblk_gfp is not really needed as
sb_getblk
  __getblk_gfp
    __getblk_slow
      grow_buffers
        grow_dev_page
          gfp_mask = mapping_gfp_constraint(inode->i_mapping, ~__GFP_FS) | gfp

so __GFP_FS is cleared unconditionally and therefore the above commit
didn't have any real effect in fact.

This patch should not introduce any functional change. The main point
of this change is to reduce explicit GFP_NOFS usage inside ext4 code to
make the review of the remaining usage easier.

Signed-off-by: Michal Hocko <[hidden email]>
Reviewed-by: Jan Kara <[hidden email]>
---
 fs/ext4/extents.c | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/fs/ext4/extents.c b/fs/ext4/extents.c
index 3e295d3350a9..9867b9e5ad8f 100644
--- a/fs/ext4/extents.c
+++ b/fs/ext4/extents.c
@@ -518,7 +518,7 @@ __read_extent_tree_block(const char *function, unsigned int line,
  struct buffer_head *bh;
  int err;
 
- bh = sb_getblk_gfp(inode->i_sb, pblk, __GFP_MOVABLE | GFP_NOFS);
+ bh = sb_getblk(inode->i_sb, pblk);
  if (unlikely(!bh))
  return ERR_PTR(-ENOMEM);
 
@@ -1096,7 +1096,7 @@ static int ext4_ext_split(handle_t *handle, struct inode *inode,
  err = -EFSCORRUPTED;
  goto cleanup;
  }
- bh = sb_getblk_gfp(inode->i_sb, newblock, __GFP_MOVABLE | GFP_NOFS);
+ bh = sb_getblk(inode->i_sb, newblock);
  if (unlikely(!bh)) {
  err = -ENOMEM;
  goto cleanup;
@@ -1290,7 +1290,7 @@ static int ext4_ext_grow_indepth(handle_t *handle, struct inode *inode,
  if (newblock == 0)
  return err;
 
- bh = sb_getblk_gfp(inode->i_sb, newblock, __GFP_MOVABLE | GFP_NOFS);
+ bh = sb_getblk(inode->i_sb, newblock);
  if (unlikely(!bh))
  return -ENOMEM;
  lock_buffer(bh);
--
2.11.0


------------------------------------------------------------------------------
Check out the vibrant tech community on one of the world's most
engaging tech sites, SlashDot.org! http://sdm.link/slashdot
_______________________________________________
Linux-NTFS-Dev mailing list
[hidden email]
https://lists.sourceforge.net/lists/listinfo/linux-ntfs-dev
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

[PATCH 8/8] Revert "ext4: fix wrong gfp type under transaction"

Michal Hocko
In reply to this post by Michal Hocko
From: Michal Hocko <[hidden email]>

This reverts commit 216553c4b7f3e3e2beb4981cddca9b2027523928. Now that
the transaction context uses memalloc_nofs_save and all allocations
within the this context inherit GFP_NOFS automatically, there is no
reason to mark specific allocations explicitly.

This patch should not introduce any functional change. The main point
of this change is to reduce explicit GFP_NOFS usage inside ext4 code
to make the review of the remaining usage easier.

Signed-off-by: Michal Hocko <[hidden email]>
Reviewed-by: Jan Kara <[hidden email]>
---
 fs/ext4/acl.c     | 6 +++---
 fs/ext4/extents.c | 2 +-
 fs/ext4/resize.c  | 4 ++--
 fs/ext4/xattr.c   | 4 ++--
 4 files changed, 8 insertions(+), 8 deletions(-)

diff --git a/fs/ext4/acl.c b/fs/ext4/acl.c
index fd389935ecd1..9e98092c2a4b 100644
--- a/fs/ext4/acl.c
+++ b/fs/ext4/acl.c
@@ -32,7 +32,7 @@ ext4_acl_from_disk(const void *value, size_t size)
  return ERR_PTR(-EINVAL);
  if (count == 0)
  return NULL;
- acl = posix_acl_alloc(count, GFP_NOFS);
+ acl = posix_acl_alloc(count, GFP_KERNEL);
  if (!acl)
  return ERR_PTR(-ENOMEM);
  for (n = 0; n < count; n++) {
@@ -94,7 +94,7 @@ ext4_acl_to_disk(const struct posix_acl *acl, size_t *size)
 
  *size = ext4_acl_size(acl->a_count);
  ext_acl = kmalloc(sizeof(ext4_acl_header) + acl->a_count *
- sizeof(ext4_acl_entry), GFP_NOFS);
+ sizeof(ext4_acl_entry), GFP_KERNEL);
  if (!ext_acl)
  return ERR_PTR(-ENOMEM);
  ext_acl->a_version = cpu_to_le32(EXT4_ACL_VERSION);
@@ -159,7 +159,7 @@ ext4_get_acl(struct inode *inode, int type)
  }
  retval = ext4_xattr_get(inode, name_index, "", NULL, 0);
  if (retval > 0) {
- value = kmalloc(retval, GFP_NOFS);
+ value = kmalloc(retval, GFP_KERNEL);
  if (!value)
  return ERR_PTR(-ENOMEM);
  retval = ext4_xattr_get(inode, name_index, "", value, retval);
diff --git a/fs/ext4/extents.c b/fs/ext4/extents.c
index 9867b9e5ad8f..0371e7aa7bea 100644
--- a/fs/ext4/extents.c
+++ b/fs/ext4/extents.c
@@ -2933,7 +2933,7 @@ int ext4_ext_remove_space(struct inode *inode, ext4_lblk_t start,
  le16_to_cpu(path[k].p_hdr->eh_entries)+1;
  } else {
  path = kzalloc(sizeof(struct ext4_ext_path) * (depth + 1),
-       GFP_NOFS);
+       GFP_KERNEL);
  if (path == NULL) {
  ext4_journal_stop(handle);
  return -ENOMEM;
diff --git a/fs/ext4/resize.c b/fs/ext4/resize.c
index cf681004b196..e121f4e048b8 100644
--- a/fs/ext4/resize.c
+++ b/fs/ext4/resize.c
@@ -816,7 +816,7 @@ static int add_new_gdb(handle_t *handle, struct inode *inode,
 
  n_group_desc = ext4_kvmalloc((gdb_num + 1) *
      sizeof(struct buffer_head *),
-     GFP_NOFS);
+     GFP_KERNEL);
  if (!n_group_desc) {
  err = -ENOMEM;
  ext4_warning(sb, "not enough memory for %lu groups",
@@ -943,7 +943,7 @@ static int reserve_backup_gdb(handle_t *handle, struct inode *inode,
  int res, i;
  int err;
 
- primary = kmalloc(reserved_gdb * sizeof(*primary), GFP_NOFS);
+ primary = kmalloc(reserved_gdb * sizeof(*primary), GFP_KERNEL);
  if (!primary)
  return -ENOMEM;
 
diff --git a/fs/ext4/xattr.c b/fs/ext4/xattr.c
index 5a94fa52b74f..172317462238 100644
--- a/fs/ext4/xattr.c
+++ b/fs/ext4/xattr.c
@@ -875,7 +875,7 @@ ext4_xattr_block_set(handle_t *handle, struct inode *inode,
 
  unlock_buffer(bs->bh);
  ea_bdebug(bs->bh, "cloning");
- s->base = kmalloc(bs->bh->b_size, GFP_NOFS);
+ s->base = kmalloc(bs->bh->b_size, GFP_KERNEL);
  error = -ENOMEM;
  if (s->base == NULL)
  goto cleanup;
@@ -887,7 +887,7 @@ ext4_xattr_block_set(handle_t *handle, struct inode *inode,
  }
  } else {
  /* Allocate a buffer where we construct the new block. */
- s->base = kzalloc(sb->s_blocksize, GFP_NOFS);
+ s->base = kzalloc(sb->s_blocksize, GFP_KERNEL);
  /* assert(header == s->base) */
  error = -ENOMEM;
  if (s->base == NULL)
--
2.11.0


------------------------------------------------------------------------------
Check out the vibrant tech community on one of the world's most
engaging tech sites, SlashDot.org! http://sdm.link/slashdot
_______________________________________________
Linux-NTFS-Dev mailing list
[hidden email]
https://lists.sourceforge.net/lists/listinfo/linux-ntfs-dev
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

[DEBUG PATCH 0/2] debug explicit GFP_NO{FS, IO} usage from the scope context

Michal Hocko
In reply to this post by Michal Hocko
These two patches should help to identify explicit GFP_NO{FS,IO} usage
from withing a scope context and reduce such a usage as a result. Such
a usage can be changed to the full GFP_KERNEL because all the calls
from within the NO{FS,IO} scope will drop the __GFP_FS resp. __GFP_IO
automatically and if the function is called outside of the scope then
we do not need to restrict it to NOFS/NOIO as long as all the reclaim
recursion unsafe contexts are marked properly. This means that each such
a reported allocation site has to be checked before converted.

The debugging has to be enabled explicitly by a kernel command line
parameter and then it reports the stack trace of the allocation and
also the function which has started the current scope.

These two patches are _not_ intended to be merged and they are only
aimed at debugging.


------------------------------------------------------------------------------
Check out the vibrant tech community on one of the world's most
engaging tech sites, SlashDot.org! http://sdm.link/slashdot
_______________________________________________
Linux-NTFS-Dev mailing list
[hidden email]
https://lists.sourceforge.net/lists/listinfo/linux-ntfs-dev
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

[DEBUG PATCH 1/2] mm, debug: report when GFP_NO{FS, IO} is used explicitly from memalloc_no{fs, io}_{save, restore} context

Michal Hocko
From: Michal Hocko <[hidden email]>

THIS PATCH IS FOR TESTING ONLY AND NOT MEANT TO HIT LINUS TREE

It is desirable to reduce the direct GFP_NO{FS,IO} usage at minimum and
prefer scope usage defined by memalloc_no{fs,io}_{save,restore} API.

Let's help this process and add a debugging tool to catch when an
explicit allocation request for GFP_NO{FS,IO} is done from the scope
context. The printed stacktrace should help to identify the caller
and evaluate whether it can be changed to use a wider context or whether
it is called from another potentially dangerous context which needs
a scope protection as well.

The checks have to be enabled explicitly by debug_scope_gfp kernel
command line parameter.

Signed-off-by: Michal Hocko <[hidden email]>
---
 include/linux/sched.h | 14 +++++++++++--
 include/linux/slab.h  |  3 +++
 mm/page_alloc.c       | 58 +++++++++++++++++++++++++++++++++++++++++++++++++++
 3 files changed, 73 insertions(+), 2 deletions(-)

diff --git a/include/linux/sched.h b/include/linux/sched.h
index 2032fc642a26..59428926e989 100644
--- a/include/linux/sched.h
+++ b/include/linux/sched.h
@@ -1988,6 +1988,8 @@ struct task_struct {
  /* A live task holds one reference. */
  atomic_t stack_refcount;
 #endif
+ unsigned long nofs_caller;
+ unsigned long noio_caller;
 /* CPU-specific state of this task */
  struct thread_struct thread;
 /*
@@ -2345,6 +2347,8 @@ extern void thread_group_cputime_adjusted(struct task_struct *p, cputime_t *ut,
 #define tsk_used_math(p) ((p)->flags & PF_USED_MATH)
 #define used_math() tsk_used_math(current)
 
+extern void debug_scope_gfp_context(gfp_t gfp_mask);
+
 /*
  * Applies per-task gfp context to the given allocation flags.
  * PF_MEMALLOC_NOIO implies GFP_NOIO
@@ -2363,25 +2367,31 @@ static inline gfp_t current_gfp_context(gfp_t flags)
  return flags;
 }
 
-static inline unsigned int memalloc_noio_save(void)
+static inline unsigned int __memalloc_noio_save(unsigned long caller)
 {
  unsigned int flags = current->flags & PF_MEMALLOC_NOIO;
  current->flags |= PF_MEMALLOC_NOIO;
+ current->noio_caller = caller;
  return flags;
 }
 
+#define memalloc_noio_save() __memalloc_noio_save(_RET_IP_)
+
 static inline void memalloc_noio_restore(unsigned int flags)
 {
  current->flags = (current->flags & ~PF_MEMALLOC_NOIO) | flags;
 }
 
-static inline unsigned int memalloc_nofs_save(void)
+static inline unsigned int __memalloc_nofs_save(unsigned long caller)
 {
  unsigned int flags = current->flags & PF_MEMALLOC_NOFS;
  current->flags |= PF_MEMALLOC_NOFS;
+ current->nofs_caller = caller;
  return flags;
 }
 
+#define memalloc_nofs_save() __memalloc_nofs_save(_RET_IP_)
+
 static inline void memalloc_nofs_restore(unsigned int flags)
 {
  current->flags = (current->flags & ~PF_MEMALLOC_NOFS) | flags;
diff --git a/include/linux/slab.h b/include/linux/slab.h
index 084b12bad198..6559668e29db 100644
--- a/include/linux/slab.h
+++ b/include/linux/slab.h
@@ -477,6 +477,7 @@ static __always_inline void *kmalloc_large(size_t size, gfp_t flags)
  */
 static __always_inline void *kmalloc(size_t size, gfp_t flags)
 {
+ debug_scope_gfp_context(flags);
  if (__builtin_constant_p(size)) {
  if (size > KMALLOC_MAX_CACHE_SIZE)
  return kmalloc_large(size, flags);
@@ -517,6 +518,7 @@ static __always_inline int kmalloc_size(int n)
 
 static __always_inline void *kmalloc_node(size_t size, gfp_t flags, int node)
 {
+ debug_scope_gfp_context(flags);
 #ifndef CONFIG_SLOB
  if (__builtin_constant_p(size) &&
  size <= KMALLOC_MAX_CACHE_SIZE && !(flags & GFP_DMA)) {
@@ -575,6 +577,7 @@ int memcg_update_all_caches(int num_memcgs);
  */
 static inline void *kmalloc_array(size_t n, size_t size, gfp_t flags)
 {
+ debug_scope_gfp_context(flags);
  if (size != 0 && n > SIZE_MAX / size)
  return NULL;
  if (__builtin_constant_p(n) && __builtin_constant_p(size))
diff --git a/mm/page_alloc.c b/mm/page_alloc.c
index 5138b46a4295..87a2bb5262b2 100644
--- a/mm/page_alloc.c
+++ b/mm/page_alloc.c
@@ -3738,6 +3738,63 @@ __alloc_pages_slowpath(gfp_t gfp_mask, unsigned int order,
  return page;
 }
 
+static bool debug_scope_gfp;
+
+static int __init enable_debug_scope_gfp(char *unused)
+{
+ debug_scope_gfp = true;
+ return 0;
+}
+
+/*
+ * spit the stack trace if the given gfp_mask clears flags which are context
+ * wide cleared. Such a caller can remove special flags clearing and rely on
+ * the context wide mask.
+ */
+void debug_scope_gfp_context(gfp_t gfp_mask)
+{
+ gfp_t restrict_mask;
+
+ if (likely(!debug_scope_gfp))
+ return;
+
+ /* both NOFS, NOIO are irrelevant when direct reclaim is disabled */
+ if (!(gfp_mask & __GFP_DIRECT_RECLAIM))
+ return;
+
+ if (current->flags & PF_MEMALLOC_NOIO)
+ restrict_mask = __GFP_IO;
+ else if ((current->flags & PF_MEMALLOC_NOFS) && (gfp_mask & __GFP_IO))
+ restrict_mask = __GFP_FS;
+ else
+ return;
+
+ if ((gfp_mask & restrict_mask) != restrict_mask) {
+ /*
+ * If you see this this warning then the code does:
+ * memalloc_no{fs,io}_save()
+ * ...
+ *    foo()
+ *      alloc_page(GFP_NO{FS,IO})
+ * ...
+ * memalloc_no{fs,io}_restore()
+ *
+ * allocation which is unnecessary because the scope gfp
+ * context will do that for all allocation requests already.
+ * If foo() is called from multiple contexts then make sure other
+ * contexts are safe wrt. GFP_NO{FS,IO} semantic and either add
+ * scope protection into particular paths or change the gfp mask
+ * to GFP_KERNEL.
+ */
+ pr_info("Unnecesarily specific gfp mask:%#x(%pGg) for the %s task wide context from %ps\n", gfp_mask, &gfp_mask,
+ (current->flags & PF_MEMALLOC_NOIO)?"NOIO":"NOFS",
+ (void*)((current->flags & PF_MEMALLOC_NOIO)?current->noio_caller:current->nofs_caller));
+ dump_stack();
+ }
+}
+EXPORT_SYMBOL(debug_scope_gfp_context);
+early_param("debug_scope_gfp", enable_debug_scope_gfp);
+
 /*
  * This is the 'heart' of the zoned buddy allocator.
  */
@@ -3802,6 +3859,7 @@ __alloc_pages_nodemask(gfp_t gfp_mask, unsigned int order,
  }
 
  /* First allocation attempt */
+ debug_scope_gfp_context(gfp_mask);
  page = get_page_from_freelist(alloc_mask, order, alloc_flags, &ac);
  if (likely(page))
  goto out;
--
2.11.0


------------------------------------------------------------------------------
Check out the vibrant tech community on one of the world's most
engaging tech sites, SlashDot.org! http://sdm.link/slashdot
_______________________________________________
Linux-NTFS-Dev mailing list
[hidden email]
https://lists.sourceforge.net/lists/listinfo/linux-ntfs-dev
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

[DEBUG PATCH 2/2] silent warnings which we cannot do anything about

Michal Hocko
In reply to this post by Michal Hocko
From: Michal Hocko <[hidden email]>

THIS PATCH IS FOR TESTING ONLY AND NOT MEANT TO HIT LINUS TREE

There are some code paths used by all the filesystems which we cannot
change to drop the GFP_NOFS, yet they generate a lot of warnings.
Provide {disable,enable}_scope_gfp_check to silence those.
alloc_page_buffers and grow_dev_page are silenced right away.

Signed-off-by: Michal Hocko <[hidden email]>
---
 fs/buffer.c           |  4 ++++
 include/linux/sched.h | 11 +++++++++++
 mm/page_alloc.c       |  3 +++
 3 files changed, 18 insertions(+)

diff --git a/fs/buffer.c b/fs/buffer.c
index 28484b3ebc98..dbe529e7881b 100644
--- a/fs/buffer.c
+++ b/fs/buffer.c
@@ -873,7 +873,9 @@ struct buffer_head *alloc_page_buffers(struct page *page, unsigned long size,
  head = NULL;
  offset = PAGE_SIZE;
  while ((offset -= size) >= 0) {
+ disable_scope_gfp_check();
  bh = alloc_buffer_head(GFP_NOFS);
+ enable_scope_gfp_check();
  if (!bh)
  goto no_grow;
 
@@ -1003,7 +1005,9 @@ grow_dev_page(struct block_device *bdev, sector_t block,
  */
  gfp_mask |= __GFP_NOFAIL;
 
+ disable_scope_gfp_check();
  page = find_or_create_page(inode->i_mapping, index, gfp_mask);
+ enable_scope_gfp_check();
  if (!page)
  return ret;
 
diff --git a/include/linux/sched.h b/include/linux/sched.h
index 59428926e989..f60294732ed5 100644
--- a/include/linux/sched.h
+++ b/include/linux/sched.h
@@ -1988,6 +1988,7 @@ struct task_struct {
  /* A live task holds one reference. */
  atomic_t stack_refcount;
 #endif
+ bool disable_scope_gfp_warn;
  unsigned long nofs_caller;
  unsigned long noio_caller;
 /* CPU-specific state of this task */
@@ -2390,6 +2391,16 @@ static inline unsigned int __memalloc_nofs_save(unsigned long caller)
  return flags;
 }
 
+static inline void disable_scope_gfp_check(void)
+{
+ current->disable_scope_gfp_warn = true;
+}
+
+static inline void enable_scope_gfp_check(void)
+{
+ current->disable_scope_gfp_warn = false;
+}
+
 #define memalloc_nofs_save() __memalloc_nofs_save(_RET_IP_)
 
 static inline void memalloc_nofs_restore(unsigned int flags)
diff --git a/mm/page_alloc.c b/mm/page_alloc.c
index 87a2bb5262b2..5405278bd733 100644
--- a/mm/page_alloc.c
+++ b/mm/page_alloc.c
@@ -3762,6 +3762,9 @@ void debug_scope_gfp_context(gfp_t gfp_mask)
  if (!(gfp_mask & __GFP_DIRECT_RECLAIM))
  return;
 
+ if (current->disable_scope_gfp_warn)
+ return;
+
  if (current->flags & PF_MEMALLOC_NOIO)
  restrict_mask = __GFP_IO;
  else if ((current->flags & PF_MEMALLOC_NOFS) && (gfp_mask & __GFP_IO))
--
2.11.0


------------------------------------------------------------------------------
Check out the vibrant tech community on one of the world's most
engaging tech sites, SlashDot.org! http://sdm.link/slashdot
_______________________________________________
Linux-NTFS-Dev mailing list
[hidden email]
https://lists.sourceforge.net/lists/listinfo/linux-ntfs-dev
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

Re: [PATCH 1/8] lockdep: allow to disable reclaim lockup detection

Vlastimil Babka
In reply to this post by Michal Hocko
On 01/06/2017 03:11 PM, Michal Hocko wrote:

> From: Michal Hocko <[hidden email]>
>
> The current implementation of the reclaim lockup detection can lead to
> false positives and those even happen and usually lead to tweak the
> code to silence the lockdep by using GFP_NOFS even though the context
> can use __GFP_FS just fine. See
> http://lkml.kernel.org/r/20160512080321.GA18496@dastard as an example.
>
> =================================
> [ INFO: inconsistent lock state ]
> 4.5.0-rc2+ #4 Tainted: G           O
> ---------------------------------
> inconsistent {RECLAIM_FS-ON-R} -> {IN-RECLAIM_FS-W} usage.
> kswapd0/543 [HC0[0]:SC0[0]:HE1:SE1] takes:
>
> (&xfs_nondir_ilock_class){++++-+}, at: [<ffffffffa00781f7>] xfs_ilock+0x177/0x200 [xfs]
>
> {RECLAIM_FS-ON-R} state was registered at:
>   [<ffffffff8110f369>] mark_held_locks+0x79/0xa0
>   [<ffffffff81113a43>] lockdep_trace_alloc+0xb3/0x100
>   [<ffffffff81224623>] kmem_cache_alloc+0x33/0x230
>   [<ffffffffa008acc1>] kmem_zone_alloc+0x81/0x120 [xfs]
>   [<ffffffffa005456e>] xfs_refcountbt_init_cursor+0x3e/0xa0 [xfs]
>   [<ffffffffa0053455>] __xfs_refcount_find_shared+0x75/0x580 [xfs]
>   [<ffffffffa00539e4>] xfs_refcount_find_shared+0x84/0xb0 [xfs]
>   [<ffffffffa005dcb8>] xfs_getbmap+0x608/0x8c0 [xfs]
>   [<ffffffffa007634b>] xfs_vn_fiemap+0xab/0xc0 [xfs]
>   [<ffffffff81244208>] do_vfs_ioctl+0x498/0x670
>   [<ffffffff81244459>] SyS_ioctl+0x79/0x90
>   [<ffffffff81847cd7>] entry_SYSCALL_64_fastpath+0x12/0x6f
>
>        CPU0
>        ----
>   lock(&xfs_nondir_ilock_class);
>   <Interrupt>
>     lock(&xfs_nondir_ilock_class);
>
>  *** DEADLOCK ***
>
> 3 locks held by kswapd0/543:
>
> stack backtrace:
> CPU: 0 PID: 543 Comm: kswapd0 Tainted: G           O    4.5.0-rc2+ #4
>
> Hardware name: innotek GmbH VirtualBox/VirtualBox, BIOS VirtualBox 12/01/2006
>
>  ffffffff82a34f10 ffff88003aa078d0 ffffffff813a14f9 ffff88003d8551c0
>  ffff88003aa07920 ffffffff8110ec65 0000000000000000 0000000000000001
>  ffff880000000001 000000000000000b 0000000000000008 ffff88003d855aa0
> Call Trace:
>  [<ffffffff813a14f9>] dump_stack+0x4b/0x72
>  [<ffffffff8110ec65>] print_usage_bug+0x215/0x240
>  [<ffffffff8110ee85>] mark_lock+0x1f5/0x660
>  [<ffffffff8110e100>] ? print_shortest_lock_dependencies+0x1a0/0x1a0
>  [<ffffffff811102e0>] __lock_acquire+0xa80/0x1e50
>  [<ffffffff8122474e>] ? kmem_cache_alloc+0x15e/0x230
>  [<ffffffffa008acc1>] ? kmem_zone_alloc+0x81/0x120 [xfs]
>  [<ffffffff811122e8>] lock_acquire+0xd8/0x1e0
>  [<ffffffffa00781f7>] ? xfs_ilock+0x177/0x200 [xfs]
>  [<ffffffffa0083a70>] ? xfs_reflink_cancel_cow_range+0x150/0x300 [xfs]
>  [<ffffffff8110aace>] down_write_nested+0x5e/0xc0
>  [<ffffffffa00781f7>] ? xfs_ilock+0x177/0x200 [xfs]
>  [<ffffffffa00781f7>] xfs_ilock+0x177/0x200 [xfs]
>  [<ffffffffa0083a70>] xfs_reflink_cancel_cow_range+0x150/0x300 [xfs]
>  [<ffffffffa0085bdc>] xfs_fs_evict_inode+0xdc/0x1e0 [xfs]
>  [<ffffffff8124d7d5>] evict+0xc5/0x190
>  [<ffffffff8124d8d9>] dispose_list+0x39/0x60
>  [<ffffffff8124eb2b>] prune_icache_sb+0x4b/0x60
>  [<ffffffff8123317f>] super_cache_scan+0x14f/0x1a0
>  [<ffffffff811e0d19>] shrink_slab.part.63.constprop.79+0x1e9/0x4e0
>  [<ffffffff811e50ee>] shrink_zone+0x15e/0x170
>  [<ffffffff811e5ef1>] kswapd+0x4f1/0xa80
>  [<ffffffff811e5a00>] ? zone_reclaim+0x230/0x230
>  [<ffffffff810e6882>] kthread+0xf2/0x110
>  [<ffffffff810e6790>] ? kthread_create_on_node+0x220/0x220
>  [<ffffffff8184803f>] ret_from_fork+0x3f/0x70
>  [<ffffffff810e6790>] ? kthread_create_on_node+0x220/0x220
>
> To quote Dave:
> "
> Ignoring whether reflink should be doing anything or not, that's a
> "xfs_refcountbt_init_cursor() gets called both outside and inside
> transactions" lockdep false positive case. The problem here is
> lockdep has seen this allocation from within a transaction, hence a
> GFP_NOFS allocation, and now it's seeing it in a GFP_KERNEL context.
> Also note that we have an active reference to this inode.
>
> So, because the reclaim annotations overload the interrupt level
> detections and it's seen the inode ilock been taken in reclaim
> ("interrupt") context, this triggers a reclaim context warning where
> it thinks it is unsafe to do this allocation in GFP_KERNEL context
> holding the inode ilock...
> "
>
> This sounds like a fundamental problem of the reclaim lock detection.
> It is really impossible to annotate such a special usecase IMHO unless
> the reclaim lockup detection is reworked completely. Until then it
> is much better to provide a way to add "I know what I am doing flag"
> and mark problematic places. This would prevent from abusing GFP_NOFS
> flag which has a runtime effect even on configurations which have
> lockdep disabled.
>
> Introduce __GFP_NOLOCKDEP flag which tells the lockdep gfp tracking to
> skip the current allocation request.
>
> While we are at it also make sure that the radix tree doesn't
> accidentaly override tags stored in the upper part of the gfp_mask.
>
> Suggested-by: Peter Zijlstra <[hidden email]>
> Acked-by: Peter Zijlstra (Intel) <[hidden email]>
> Signed-off-by: Michal Hocko <[hidden email]>

Acked-by: Vlastimil Babka <[hidden email]>


------------------------------------------------------------------------------
Check out the vibrant tech community on one of the world's most
engaging tech sites, SlashDot.org! http://sdm.link/slashdot
_______________________________________________
Linux-NTFS-Dev mailing list
[hidden email]
https://lists.sourceforge.net/lists/listinfo/linux-ntfs-dev
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

Re: [PATCH 2/8] xfs: abstract PF_FSTRANS to PF_MEMALLOC_NOFS

Vlastimil Babka
In reply to this post by Michal Hocko
On 01/06/2017 03:11 PM, Michal Hocko wrote:

> From: Michal Hocko <[hidden email]>
>
> xfs has defined PF_FSTRANS to declare a scope GFP_NOFS semantic quite
> some time ago. We would like to make this concept more generic and use
> it for other filesystems as well. Let's start by giving the flag a
> more generic name PF_MEMALLOC_NOFS which is in line with an exiting
> PF_MEMALLOC_NOIO already used for the same purpose for GFP_NOIO
> contexts. Replace all PF_FSTRANS usage from the xfs code in the first
> step before we introduce a full API for it as xfs uses the flag directly
> anyway.
>
> This patch doesn't introduce any functional change.
>
> Signed-off-by: Michal Hocko <[hidden email]>
> Reviewed-by: Brian Foster <[hidden email]>

Acked-by: Vlastimil Babka <[hidden email]>

A nit:

> --- a/include/linux/sched.h
> +++ b/include/linux/sched.h
> @@ -2320,6 +2320,8 @@ extern void thread_group_cputime_adjusted(struct task_struct *p, cputime_t *ut,
>  #define PF_FREEZER_SKIP 0x40000000 /* Freezer should not count it as freezable */
>  #define PF_SUSPEND_TASK 0x80000000      /* this thread called freeze_processes and should not be frozen */
>  
> +#define PF_MEMALLOC_NOFS PF_FSTRANS /* Transition to a more generic GFP_NOFS scope semantic */

I don't see why this transition is needed, as there are already no users
of PF_FSTRANS after this patch. The next patch doesn't remove any more,
so this is just extra churn IMHO. But not a strong objection.

> +
>  /*
>   * Only the _current_ task can read/write to tsk->flags, but other
>   * tasks can access tsk->flags in readonly mode for example
>


------------------------------------------------------------------------------
Check out the vibrant tech community on one of the world's most
engaging tech sites, SlashDot.org! http://sdm.link/slashdot
_______________________________________________
Linux-NTFS-Dev mailing list
[hidden email]
https://lists.sourceforge.net/lists/listinfo/linux-ntfs-dev
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

Re: [PATCH 3/8] mm: introduce memalloc_nofs_{save, restore} API

Vlastimil Babka
In reply to this post by Michal Hocko
On 01/06/2017 03:11 PM, Michal Hocko wrote:

> From: Michal Hocko <[hidden email]>
>
> GFP_NOFS context is used for the following 5 reasons currently
> - to prevent from deadlocks when the lock held by the allocation
>  context would be needed during the memory reclaim
> - to prevent from stack overflows during the reclaim because
>  the allocation is performed from a deep context already
> - to prevent lockups when the allocation context depends on
>  other reclaimers to make a forward progress indirectly
> - just in case because this would be safe from the fs POV
> - silence lockdep false positives
>
> Unfortunately overuse of this allocation context brings some problems
> to the MM. Memory reclaim is much weaker (especially during heavy FS
> metadata workloads), OOM killer cannot be invoked because the MM layer
> doesn't have enough information about how much memory is freeable by the
> FS layer.
>
> In many cases it is far from clear why the weaker context is even used
> and so it might be used unnecessarily. We would like to get rid of
> those as much as possible. One way to do that is to use the flag in
> scopes rather than isolated cases. Such a scope is declared when really
> necessary, tracked per task and all the allocation requests from within
> the context will simply inherit the GFP_NOFS semantic.
>
> Not only this is easier to understand and maintain because there are
> much less problematic contexts than specific allocation requests, this
> also helps code paths where FS layer interacts with other layers (e.g.
> crypto, security modules, MM etc...) and there is no easy way to convey
> the allocation context between the layers.
>
> Introduce memalloc_nofs_{save,restore} API to control the scope
> of GFP_NOFS allocation context. This is basically copying
> memalloc_noio_{save,restore} API we have for other restricted allocation
> context GFP_NOIO. The PF_MEMALLOC_NOFS flag already exists and it is
> just an alias for PF_FSTRANS which has been xfs specific until recently.
> There are no more PF_FSTRANS users anymore so let's just drop it.
>
> PF_MEMALLOC_NOFS is now checked in the MM layer and drops __GFP_FS
> implicitly same as PF_MEMALLOC_NOIO drops __GFP_IO. memalloc_noio_flags
> is renamed to current_gfp_context because it now cares about both
> PF_MEMALLOC_NOFS and PF_MEMALLOC_NOIO contexts. Xfs code paths preserve
> their semantic. kmem_flags_convert() doesn't need to evaluate the flag
> anymore.
>
> This patch shouldn't introduce any functional changes.
>
> Let's hope that filesystems will drop direct GFP_NOFS (resp. ~__GFP_FS)
> usage as much as possible and only use a properly documented
> memalloc_nofs_{save,restore} checkpoints where they are appropriate.
>
> Signed-off-by: Michal Hocko <[hidden email]>


[...]

> +static inline unsigned int memalloc_nofs_save(void)
> +{
> + unsigned int flags = current->flags & PF_MEMALLOC_NOFS;
> + current->flags |= PF_MEMALLOC_NOFS;

So this is not new, as same goes for memalloc_noio_save, but I've
noticed that e.g. exit_signal() does tsk->flags |= PF_EXITING;
So is it possible that there's a r-m-w hazard here?

> + return flags;
> +}
> +
> +static inline void memalloc_nofs_restore(unsigned int flags)
> +{
> + current->flags = (current->flags & ~PF_MEMALLOC_NOFS) | flags;
> +}
> +
>  /* Per-process atomic flags. */
>  #define PFA_NO_NEW_PRIVS 0 /* May not gain new privileges. */
>  #define PFA_SPREAD_PAGE  1      /* Spread page cache over cpuset */

[...]

> @@ -3029,7 +3029,7 @@ unsigned long try_to_free_mem_cgroup_pages(struct mem_cgroup *memcg,
>   int nid;
>   struct scan_control sc = {
>   .nr_to_reclaim = max(nr_pages, SWAP_CLUSTER_MAX),
> - .gfp_mask = (gfp_mask & GFP_RECLAIM_MASK) |
> + .gfp_mask = (current_gfp_context(gfp_mask) & GFP_RECLAIM_MASK) |

So this function didn't do memalloc_noio_flags() before? Is it a bug
that should be fixed separately or at least mentioned? Because that
looks like a functional change...

Thanks!

>   (GFP_HIGHUSER_MOVABLE & ~GFP_RECLAIM_MASK),
>   .reclaim_idx = MAX_NR_ZONES - 1,
>   .target_mem_cgroup = memcg,
> @@ -3723,7 +3723,7 @@ static int __node_reclaim(struct pglist_data *pgdat, gfp_t gfp_mask, unsigned in
>   int classzone_idx = gfp_zone(gfp_mask);
>   struct scan_control sc = {
>   .nr_to_reclaim = max(nr_pages, SWAP_CLUSTER_MAX),
> - .gfp_mask = (gfp_mask = memalloc_noio_flags(gfp_mask)),
> + .gfp_mask = (gfp_mask = current_gfp_context(gfp_mask)),
>   .order = order,
>   .priority = NODE_RECLAIM_PRIORITY,
>   .may_writepage = !!(node_reclaim_mode & RECLAIM_WRITE),
>


------------------------------------------------------------------------------
Check out the vibrant tech community on one of the world's most
engaging tech sites, SlashDot.org! http://sdm.link/slashdot
_______________________________________________
Linux-NTFS-Dev mailing list
[hidden email]
https://lists.sourceforge.net/lists/listinfo/linux-ntfs-dev
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

Re: [PATCH 3/8] mm: introduce memalloc_nofs_{save, restore} API

Michal Hocko
On Mon 09-01-17 14:04:21, Vlastimil Babka wrote:
[...]
> > +static inline unsigned int memalloc_nofs_save(void)
> > +{
> > + unsigned int flags = current->flags & PF_MEMALLOC_NOFS;
> > + current->flags |= PF_MEMALLOC_NOFS;
>
> So this is not new, as same goes for memalloc_noio_save, but I've
> noticed that e.g. exit_signal() does tsk->flags |= PF_EXITING;
> So is it possible that there's a r-m-w hazard here?

exit_signals operates on current and all task_struct::flags should be
used only on the current.
[...]

> > @@ -3029,7 +3029,7 @@ unsigned long try_to_free_mem_cgroup_pages(struct mem_cgroup *memcg,
> >   int nid;
> >   struct scan_control sc = {
> >   .nr_to_reclaim = max(nr_pages, SWAP_CLUSTER_MAX),
> > - .gfp_mask = (gfp_mask & GFP_RECLAIM_MASK) |
> > + .gfp_mask = (current_gfp_context(gfp_mask) & GFP_RECLAIM_MASK) |
>
> So this function didn't do memalloc_noio_flags() before? Is it a bug
> that should be fixed separately or at least mentioned? Because that
> looks like a functional change...

We didn't need it. Kmem charges are opt-in and current all of them
support GFP_IO. The LRU pages are not charged in NOIO context either.
We need it now because there will be callers to charge GFP_KERNEL while
being inside the NOFS scope.

Now that you have opened this I have noticed that the code is wrong
here because GFP_HIGHUSER_MOVABLE & ~GFP_RECLAIM_MASK would overwrite
the removed GFP_FS. I guess it would be better and less error prone
to move the current_gfp_context part into the direct reclaim entry -
do_try_to_free_pages - and put the comment like this
---
diff --git a/mm/vmscan.c b/mm/vmscan.c
index 4ea6b610f20e..df7975185f11 100644
--- a/mm/vmscan.c
+++ b/mm/vmscan.c
@@ -2756,6 +2756,13 @@ static unsigned long do_try_to_free_pages(struct zonelist *zonelist,
  int initial_priority = sc->priority;
  unsigned long total_scanned = 0;
  unsigned long writeback_threshold;
+
+ /*
+ * Make sure that the gfp context properly handles scope gfp mask.
+ * This might weaken the reclaim context (e.g. make it GFP_NOFS or
+ * GFP_NOIO).
+ */
+ sc->gfp_mask = current_gfp_context(sc->gfp_mask);
 retry:
  delayacct_freepages_start();
 
@@ -2949,7 +2956,7 @@ unsigned long try_to_free_pages(struct zonelist *zonelist, int order,
  unsigned long nr_reclaimed;
  struct scan_control sc = {
  .nr_to_reclaim = SWAP_CLUSTER_MAX,
- .gfp_mask = (gfp_mask = current_gfp_context(gfp_mask)),
+ .gfp_mask = gfp_mask,
  .reclaim_idx = gfp_zone(gfp_mask),
  .order = order,
  .nodemask = nodemask,
@@ -3029,8 +3036,7 @@ unsigned long try_to_free_mem_cgroup_pages(struct mem_cgroup *memcg,
  int nid;
  struct scan_control sc = {
  .nr_to_reclaim = max(nr_pages, SWAP_CLUSTER_MAX),
- .gfp_mask = (current_gfp_context(gfp_mask) & GFP_RECLAIM_MASK) |
- (GFP_HIGHUSER_MOVABLE & ~GFP_RECLAIM_MASK),
+ .gfp_mask = GFP_HIGHUSER_MOVABLE & ~GFP_RECLAIM_MASK,
  .reclaim_idx = MAX_NR_ZONES - 1,
  .target_mem_cgroup = memcg,
  .priority = DEF_PRIORITY,
@@ -3723,7 +3729,7 @@ static int __node_reclaim(struct pglist_data *pgdat, gfp_t gfp_mask, unsigned in
  int classzone_idx = gfp_zone(gfp_mask);
  struct scan_control sc = {
  .nr_to_reclaim = max(nr_pages, SWAP_CLUSTER_MAX),
- .gfp_mask = (gfp_mask = current_gfp_context(gfp_mask)),
+ .gfp_mask = gfp_mask,
  .order = order,
  .priority = NODE_RECLAIM_PRIORITY,
  .may_writepage = !!(node_reclaim_mode & RECLAIM_WRITE),
--
Michal Hocko
SUSE Labs

------------------------------------------------------------------------------
Check out the vibrant tech community on one of the world's most
engaging tech sites, SlashDot.org! http://sdm.link/slashdot
_______________________________________________
Linux-NTFS-Dev mailing list
[hidden email]
https://lists.sourceforge.net/lists/listinfo/linux-ntfs-dev
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

Re: [PATCH 3/8] mm: introduce memalloc_nofs_{save, restore} API

Michal Hocko
On Mon 09-01-17 14:42:10, Michal Hocko wrote:
> On Mon 09-01-17 14:04:21, Vlastimil Babka wrote:
[...]
> Now that you have opened this I have noticed that the code is wrong
> here because GFP_HIGHUSER_MOVABLE & ~GFP_RECLAIM_MASK would overwrite
> the removed GFP_FS.

Blee, it wouldn't because ~GFP_RECLAIM_MASK will not contain neither
GFP_FS nor GFP_IO. So all is good here.

> I guess it would be better and less error prone
> to move the current_gfp_context part into the direct reclaim entry -
> do_try_to_free_pages - and put the comment like this

well, after more thinking about we, should probably keep it where it is.
If for nothing else try_to_free_mem_cgroup_pages has a tracepoint which
prints the gfp mask so we should use the filtered one. So let's just
scratch this follow up fix.

> ---
> diff --git a/mm/vmscan.c b/mm/vmscan.c
> index 4ea6b610f20e..df7975185f11 100644
> --- a/mm/vmscan.c
> +++ b/mm/vmscan.c
> @@ -2756,6 +2756,13 @@ static unsigned long do_try_to_free_pages(struct zonelist *zonelist,
>   int initial_priority = sc->priority;
>   unsigned long total_scanned = 0;
>   unsigned long writeback_threshold;
> +
> + /*
> + * Make sure that the gfp context properly handles scope gfp mask.
> + * This might weaken the reclaim context (e.g. make it GFP_NOFS or
> + * GFP_NOIO).
> + */
> + sc->gfp_mask = current_gfp_context(sc->gfp_mask);
>  retry:
>   delayacct_freepages_start();
>  
> @@ -2949,7 +2956,7 @@ unsigned long try_to_free_pages(struct zonelist *zonelist, int order,
>   unsigned long nr_reclaimed;
>   struct scan_control sc = {
>   .nr_to_reclaim = SWAP_CLUSTER_MAX,
> - .gfp_mask = (gfp_mask = current_gfp_context(gfp_mask)),
> + .gfp_mask = gfp_mask,
>   .reclaim_idx = gfp_zone(gfp_mask),
>   .order = order,
>   .nodemask = nodemask,
> @@ -3029,8 +3036,7 @@ unsigned long try_to_free_mem_cgroup_pages(struct mem_cgroup *memcg,
>   int nid;
>   struct scan_control sc = {
>   .nr_to_reclaim = max(nr_pages, SWAP_CLUSTER_MAX),
> - .gfp_mask = (current_gfp_context(gfp_mask) & GFP_RECLAIM_MASK) |
> - (GFP_HIGHUSER_MOVABLE & ~GFP_RECLAIM_MASK),
> + .gfp_mask = GFP_HIGHUSER_MOVABLE & ~GFP_RECLAIM_MASK,
>   .reclaim_idx = MAX_NR_ZONES - 1,
>   .target_mem_cgroup = memcg,
>   .priority = DEF_PRIORITY,
> @@ -3723,7 +3729,7 @@ static int __node_reclaim(struct pglist_data *pgdat, gfp_t gfp_mask, unsigned in
>   int classzone_idx = gfp_zone(gfp_mask);
>   struct scan_control sc = {
>   .nr_to_reclaim = max(nr_pages, SWAP_CLUSTER_MAX),
> - .gfp_mask = (gfp_mask = current_gfp_context(gfp_mask)),
> + .gfp_mask = gfp_mask,
>   .order = order,
>   .priority = NODE_RECLAIM_PRIORITY,
>   .may_writepage = !!(node_reclaim_mode & RECLAIM_WRITE),
> --
> Michal Hocko
> SUSE Labs

--
Michal Hocko
SUSE Labs

------------------------------------------------------------------------------
Check out the vibrant tech community on one of the world's most
engaging tech sites, SlashDot.org! http://sdm.link/slashdot
_______________________________________________
Linux-NTFS-Dev mailing list
[hidden email]
https://lists.sourceforge.net/lists/listinfo/linux-ntfs-dev
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

Re: [PATCH 3/8] mm: introduce memalloc_nofs_{save, restore} API

Vlastimil Babka
In reply to this post by Michal Hocko
On 01/09/2017 02:42 PM, Michal Hocko wrote:

> On Mon 09-01-17 14:04:21, Vlastimil Babka wrote:
> [...]
>>> +static inline unsigned int memalloc_nofs_save(void)
>>> +{
>>> + unsigned int flags = current->flags & PF_MEMALLOC_NOFS;
>>> + current->flags |= PF_MEMALLOC_NOFS;
>>
>> So this is not new, as same goes for memalloc_noio_save, but I've
>> noticed that e.g. exit_signal() does tsk->flags |= PF_EXITING;
>> So is it possible that there's a r-m-w hazard here?
>
> exit_signals operates on current and all task_struct::flags should be
> used only on the current.
> [...]

Ah, good to know.

>
>>> @@ -3029,7 +3029,7 @@ unsigned long try_to_free_mem_cgroup_pages(struct mem_cgroup *memcg,
>>>   int nid;
>>>   struct scan_control sc = {
>>>   .nr_to_reclaim = max(nr_pages, SWAP_CLUSTER_MAX),
>>> - .gfp_mask = (gfp_mask & GFP_RECLAIM_MASK) |
>>> + .gfp_mask = (current_gfp_context(gfp_mask) & GFP_RECLAIM_MASK) |
>>
>> So this function didn't do memalloc_noio_flags() before? Is it a bug
>> that should be fixed separately or at least mentioned? Because that
>> looks like a functional change...
>
> We didn't need it. Kmem charges are opt-in and current all of them
> support GFP_IO. The LRU pages are not charged in NOIO context either.
> We need it now because there will be callers to charge GFP_KERNEL while
> being inside the NOFS scope.

I see.

> Now that you have opened this I have noticed that the code is wrong
> here because GFP_HIGHUSER_MOVABLE & ~GFP_RECLAIM_MASK would overwrite
> the removed GFP_FS. I guess it would be better and less error prone
> to move the current_gfp_context part into the direct reclaim entry -
> do_try_to_free_pages - and put the comment like this

Agree with your "So let's just scratch this follow up fix in the next
e-mail.

So for the unchanged patch.

Acked-by: Vlastimil Babka <[hidden email]>


------------------------------------------------------------------------------
Check out the vibrant tech community on one of the world's most
engaging tech sites, SlashDot.org! http://sdm.link/slashdot
_______________________________________________
Linux-NTFS-Dev mailing list
[hidden email]
https://lists.sourceforge.net/lists/listinfo/linux-ntfs-dev
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

Re: [PATCH 4/8] xfs: use memalloc_nofs_{save, restore} instead of memalloc_noio*

Vlastimil Babka
In reply to this post by Michal Hocko
On 01/06/2017 03:11 PM, Michal Hocko wrote:

> From: Michal Hocko <[hidden email]>
>
> kmem_zalloc_large and _xfs_buf_map_pages use memalloc_noio_{save,restore}
> API to prevent from reclaim recursion into the fs because vmalloc can
> invoke unconditional GFP_KERNEL allocations and these functions might be
> called from the NOFS contexts. The memalloc_noio_save will enforce
> GFP_NOIO context which is even weaker than GFP_NOFS and that seems to be
> unnecessary. Let's use memalloc_nofs_{save,restore} instead as it should
> provide exactly what we need here - implicit GFP_NOFS context.
>
> Changes since v1
> - s@memalloc_noio_restore@memalloc_nofs_restore@ in _xfs_buf_map_pages
>   as per Brian Foster
>
> Signed-off-by: Michal Hocko <[hidden email]>

Not a xfs expert, but seems correct.

Acked-by: Vlastimil Babka <[hidden email]>

Nit below:

> ---
>  fs/xfs/kmem.c    | 10 +++++-----
>  fs/xfs/xfs_buf.c |  8 ++++----
>  2 files changed, 9 insertions(+), 9 deletions(-)
>
> diff --git a/fs/xfs/kmem.c b/fs/xfs/kmem.c
> index a76a05dae96b..d69ed5e76621 100644
> --- a/fs/xfs/kmem.c
> +++ b/fs/xfs/kmem.c
> @@ -65,7 +65,7 @@ kmem_alloc(size_t size, xfs_km_flags_t flags)
>  void *
>  kmem_zalloc_large(size_t size, xfs_km_flags_t flags)
>  {
> - unsigned noio_flag = 0;
> + unsigned nofs_flag = 0;
>   void *ptr;
>   gfp_t lflags;
>  
> @@ -80,14 +80,14 @@ kmem_zalloc_large(size_t size, xfs_km_flags_t flags)
>   * context via PF_MEMALLOC_NOIO to prevent memory reclaim re-entering
>   * the filesystem here and potentially deadlocking.

The comment above is now largely obsolete, or minimally should be
changed to PF_MEMALLOC_NOFS?

>   */
> - if ((current->flags & PF_MEMALLOC_NOFS) || (flags & KM_NOFS))
> - noio_flag = memalloc_noio_save();
> + if (flags & KM_NOFS)
> + nofs_flag = memalloc_nofs_save();
>  
>   lflags = kmem_flags_convert(flags);
>   ptr = __vmalloc(size, lflags | __GFP_HIGHMEM | __GFP_ZERO, PAGE_KERNEL);
>  
> - if ((current->flags & PF_MEMALLOC_NOFS) || (flags & KM_NOFS))
> - memalloc_noio_restore(noio_flag);
> + if (flags & KM_NOFS)
> + memalloc_nofs_restore(nofs_flag);
>  
>   return ptr;
>  }
> diff --git a/fs/xfs/xfs_buf.c b/fs/xfs/xfs_buf.c
> index 7f0a01f7b592..8cb8dd4cdfd8 100644
> --- a/fs/xfs/xfs_buf.c
> +++ b/fs/xfs/xfs_buf.c
> @@ -441,17 +441,17 @@ _xfs_buf_map_pages(
>   bp->b_addr = NULL;
>   } else {
>   int retried = 0;
> - unsigned noio_flag;
> + unsigned nofs_flag;
>  
>   /*
>   * vm_map_ram() will allocate auxillary structures (e.g.
>   * pagetables) with GFP_KERNEL, yet we are likely to be under
>   * GFP_NOFS context here. Hence we need to tell memory reclaim
> - * that we are in such a context via PF_MEMALLOC_NOIO to prevent
> + * that we are in such a context via PF_MEMALLOC_NOFS to prevent
>   * memory reclaim re-entering the filesystem here and
>   * potentially deadlocking.
>   */
> - noio_flag = memalloc_noio_save();
> + nofs_flag = memalloc_nofs_save();
>   do {
>   bp->b_addr = vm_map_ram(bp->b_pages, bp->b_page_count,
>   -1, PAGE_KERNEL);
> @@ -459,7 +459,7 @@ _xfs_buf_map_pages(
>   break;
>   vm_unmap_aliases();
>   } while (retried++ <= 1);
> - memalloc_noio_restore(noio_flag);
> + memalloc_nofs_restore(nofs_flag);
>  
>   if (!bp->b_addr)
>   return -ENOMEM;
>


------------------------------------------------------------------------------
Check out the vibrant tech community on one of the world's most
engaging tech sites, SlashDot.org! http://sdm.link/slashdot
_______________________________________________
Linux-NTFS-Dev mailing list
[hidden email]
https://lists.sourceforge.net/lists/listinfo/linux-ntfs-dev
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

Re: [PATCH 4/8] xfs: use memalloc_nofs_{save, restore} instead of memalloc_noio*

Michal Hocko
On Mon 09-01-17 15:08:27, Vlastimil Babka wrote:

> On 01/06/2017 03:11 PM, Michal Hocko wrote:
> > From: Michal Hocko <[hidden email]>
> >
> > kmem_zalloc_large and _xfs_buf_map_pages use memalloc_noio_{save,restore}
> > API to prevent from reclaim recursion into the fs because vmalloc can
> > invoke unconditional GFP_KERNEL allocations and these functions might be
> > called from the NOFS contexts. The memalloc_noio_save will enforce
> > GFP_NOIO context which is even weaker than GFP_NOFS and that seems to be
> > unnecessary. Let's use memalloc_nofs_{save,restore} instead as it should
> > provide exactly what we need here - implicit GFP_NOFS context.
> >
> > Changes since v1
> > - s@memalloc_noio_restore@memalloc_nofs_restore@ in _xfs_buf_map_pages
> >   as per Brian Foster
> >
> > Signed-off-by: Michal Hocko <[hidden email]>
>
> Not a xfs expert, but seems correct.
>
> Acked-by: Vlastimil Babka <[hidden email]>

Thanks!

>
> Nit below:
>
> > ---
> >  fs/xfs/kmem.c    | 10 +++++-----
> >  fs/xfs/xfs_buf.c |  8 ++++----
> >  2 files changed, 9 insertions(+), 9 deletions(-)
> >
> > diff --git a/fs/xfs/kmem.c b/fs/xfs/kmem.c
> > index a76a05dae96b..d69ed5e76621 100644
> > --- a/fs/xfs/kmem.c
> > +++ b/fs/xfs/kmem.c
> > @@ -65,7 +65,7 @@ kmem_alloc(size_t size, xfs_km_flags_t flags)
> >  void *
> >  kmem_zalloc_large(size_t size, xfs_km_flags_t flags)
> >  {
> > - unsigned noio_flag = 0;
> > + unsigned nofs_flag = 0;
> >   void *ptr;
> >   gfp_t lflags;
> >  
> > @@ -80,14 +80,14 @@ kmem_zalloc_large(size_t size, xfs_km_flags_t flags)
> >   * context via PF_MEMALLOC_NOIO to prevent memory reclaim re-entering
> >   * the filesystem here and potentially deadlocking.
>
> The comment above is now largely obsolete, or minimally should be
> changed to PF_MEMALLOC_NOFS?
---
diff --git a/fs/xfs/kmem.c b/fs/xfs/kmem.c
index d69ed5e76621..0c9f94f41b6c 100644
--- a/fs/xfs/kmem.c
+++ b/fs/xfs/kmem.c
@@ -77,7 +77,7 @@ kmem_zalloc_large(size_t size, xfs_km_flags_t flags)
  * __vmalloc() will allocate data pages and auxillary structures (e.g.
  * pagetables) with GFP_KERNEL, yet we may be under GFP_NOFS context
  * here. Hence we need to tell memory reclaim that we are in such a
- * context via PF_MEMALLOC_NOIO to prevent memory reclaim re-entering
+ * context via PF_MEMALLOC_NOFS to prevent memory reclaim re-entering
  * the filesystem here and potentially deadlocking.
  */
  if (flags & KM_NOFS)

I will fold it into the original patch.

Thanks!
--
Michal Hocko
SUSE Labs

------------------------------------------------------------------------------
Check out the vibrant tech community on one of the world's most
engaging tech sites, SlashDot.org! http://sdm.link/slashdot
_______________________________________________
Linux-NTFS-Dev mailing list
[hidden email]
https://lists.sourceforge.net/lists/listinfo/linux-ntfs-dev
123
Loading...