1.. _gfp_mask_from_fs_io:
2
3=================================
4GFP masks used from FS/IO context
5=================================
6
7:Date: May, 2018
8:Author: Michal Hocko <mhocko@kernel.org>
9
10Introduction
11============
12
13Code paths in the filesystem and IO stacks must be careful when
14allocating memory to prevent recursion deadlocks caused by direct
15memory reclaim calling back into the FS or IO paths and blocking on
16already held resources (e.g. locks - most commonly those used for the
17transaction context).
18
19The traditional way to avoid this deadlock problem is to clear __GFP_FS
20respectively __GFP_IO (note the latter implies clearing the first as well) in
21the gfp mask when calling an allocator. GFP_NOFS respectively GFP_NOIO can be
22used as shortcut. It turned out though that above approach has led to
23abuses when the restricted gfp mask is used "just in case" without a
24deeper consideration which leads to problems because an excessive use
25of GFP_NOFS/GFP_NOIO can lead to memory over-reclaim or other memory
26reclaim issues.
27
28New API
29========
30
31Since 4.12 we do have a generic scope API for both NOFS and NOIO context
32``memalloc_nofs_save``, ``memalloc_nofs_restore`` respectively ``memalloc_noio_save``,
33``memalloc_noio_restore`` which allow to mark a scope to be a critical
34section from a filesystem or I/O point of view. Any allocation from that
35scope will inherently drop __GFP_FS respectively __GFP_IO from the given
36mask so no memory allocation can recurse back in the FS/IO.
37
38.. kernel-doc:: include/linux/sched/mm.h
39   :functions: memalloc_nofs_save memalloc_nofs_restore
40.. kernel-doc:: include/linux/sched/mm.h
41   :functions: memalloc_noio_save memalloc_noio_restore
42
43FS/IO code then simply calls the appropriate save function before
44any critical section with respect to the reclaim is started - e.g.
45lock shared with the reclaim context or when a transaction context
46nesting would be possible via reclaim. The restore function should be
47called when the critical section ends. All that ideally along with an
48explanation what is the reclaim context for easier maintenance.
49
50Please note that the proper pairing of save/restore functions
51allows nesting so it is safe to call ``memalloc_noio_save`` or
52``memalloc_noio_restore`` respectively from an existing NOIO or NOFS
53scope.
54
55What about __vmalloc(GFP_NOFS)
56==============================
57
58vmalloc doesn't support GFP_NOFS semantic because there are hardcoded
59GFP_KERNEL allocations deep inside the allocator which are quite non-trivial
60to fix up. That means that calling ``vmalloc`` with GFP_NOFS/GFP_NOIO is
61almost always a bug. The good news is that the NOFS/NOIO semantic can be
62achieved by the scope API.
63
64In the ideal world, upper layers should already mark dangerous contexts
65and so no special care is required and vmalloc should be called without
66any problems. Sometimes if the context is not really clear or there are
67layering violations then the recommended way around that is to wrap ``vmalloc``
68by the scope API with a comment explaining the problem.
69