1=====
2zswap
3=====
4
5Overview
6========
7
8Zswap is a lightweight compressed cache for swap pages. It takes pages that are
9in the process of being swapped out and attempts to compress them into a
10dynamically allocated RAM-based memory pool.  zswap basically trades CPU cycles
11for potentially reduced swap I/O.  This trade-off can also result in a
12significant performance improvement if reads from the compressed cache are
13faster than reads from a swap device.
14
15Some potential benefits:
16
17* Desktop/laptop users with limited RAM capacities can mitigate the
18  performance impact of swapping.
19* Overcommitted guests that share a common I/O resource can
20  dramatically reduce their swap I/O pressure, avoiding heavy handed I/O
21  throttling by the hypervisor. This allows more work to get done with less
22  impact to the guest workload and guests sharing the I/O subsystem
23* Users with SSDs as swap devices can extend the life of the device by
24  drastically reducing life-shortening writes.
25
26Zswap evicts pages from compressed cache on an LRU basis to the backing swap
27device when the compressed pool reaches its size limit.  This requirement had
28been identified in prior community discussions.
29
30Whether Zswap is enabled at the boot time depends on whether
31the ``CONFIG_ZSWAP_DEFAULT_ON`` Kconfig option is enabled or not.
32This setting can then be overridden by providing the kernel command line
33``zswap.enabled=`` option, for example ``zswap.enabled=0``.
34Zswap can also be enabled and disabled at runtime using the sysfs interface.
35An example command to enable zswap at runtime, assuming sysfs is mounted
36at ``/sys``, is::
37
38	echo 1 > /sys/module/zswap/parameters/enabled
39
40When zswap is disabled at runtime it will stop storing pages that are
41being swapped out.  However, it will _not_ immediately write out or fault
42back into memory all of the pages stored in the compressed pool.  The
43pages stored in zswap will remain in the compressed pool until they are
44either invalidated or faulted back into memory.  In order to force all
45pages out of the compressed pool, a swapoff on the swap device(s) will
46fault back into memory all swapped out pages, including those in the
47compressed pool.
48
49Design
50======
51
52Zswap receives pages for compression from the swap subsystem and is able to
53evict pages from its own compressed pool on an LRU basis and write them back to
54the backing swap device in the case that the compressed pool is full.
55
56Zswap makes use of zpool for the managing the compressed memory pool.  Each
57allocation in zpool is not directly accessible by address.  Rather, a handle is
58returned by the allocation routine and that handle must be mapped before being
59accessed.  The compressed memory pool grows on demand and shrinks as compressed
60pages are freed.  The pool is not preallocated.  By default, a zpool
61of type selected in ``CONFIG_ZSWAP_ZPOOL_DEFAULT`` Kconfig option is created,
62but it can be overridden at boot time by setting the ``zpool`` attribute,
63e.g. ``zswap.zpool=zsmalloc``. It can also be changed at runtime using the sysfs
64``zpool`` attribute, e.g.::
65
66	echo zsmalloc > /sys/module/zswap/parameters/zpool
67
68The zsmalloc type zpool has a complex compressed page storage method, and it
69can achieve great storage densities.
70
71When a swap page is passed from swapout to zswap, zswap maintains a mapping
72of the swap entry, a combination of the swap type and swap offset, to the zpool
73handle that references that compressed swap page.  This mapping is achieved
74with a red-black tree per swap type.  The swap offset is the search key for the
75tree nodes.
76
77During a page fault on a PTE that is a swap entry, the swapin code calls the
78zswap load function to decompress the page into the page allocated by the page
79fault handler.
80
81Once there are no PTEs referencing a swap page stored in zswap (i.e. the count
82in the swap_map goes to 0) the swap code calls the zswap invalidate function
83to free the compressed entry.
84
85Zswap seeks to be simple in its policies.  Sysfs attributes allow for one user
86controlled policy:
87
88* max_pool_percent - The maximum percentage of memory that the compressed
89  pool can occupy.
90
91The default compressor is selected in ``CONFIG_ZSWAP_COMPRESSOR_DEFAULT``
92Kconfig option, but it can be overridden at boot time by setting the
93``compressor`` attribute, e.g. ``zswap.compressor=lzo``.
94It can also be changed at runtime using the sysfs "compressor"
95attribute, e.g.::
96
97	echo lzo > /sys/module/zswap/parameters/compressor
98
99When the zpool and/or compressor parameter is changed at runtime, any existing
100compressed pages are not modified; they are left in their own zpool.  When a
101request is made for a page in an old zpool, it is uncompressed using its
102original compressor.  Once all pages are removed from an old zpool, the zpool
103and its compressor are freed.
104
105Some of the pages in zswap are same-value filled pages (i.e. contents of the
106page have same value or repetitive pattern). These pages include zero-filled
107pages and they are handled differently. During store operation, a page is
108checked if it is a same-value filled page before compressing it. If true, the
109compressed length of the page is set to zero and the pattern or same-filled
110value is stored.
111
112To prevent zswap from shrinking pool when zswap is full and there's a high
113pressure on swap (this will result in flipping pages in and out zswap pool
114without any real benefit but with a performance drop for the system), a
115special parameter has been introduced to implement a sort of hysteresis to
116refuse taking pages into zswap pool until it has sufficient space if the limit
117has been hit. To set the threshold at which zswap would start accepting pages
118again after it became full, use the sysfs ``accept_threshold_percent``
119attribute, e. g.::
120
121	echo 80 > /sys/module/zswap/parameters/accept_threshold_percent
122
123Setting this parameter to 100 will disable the hysteresis.
124
125Some users cannot tolerate the swapping that comes with zswap store failures
126and zswap writebacks. Swapping can be disabled entirely (without disabling
127zswap itself) on a cgroup-basis as follows::
128
129	echo 0 > /sys/fs/cgroup/<cgroup-name>/memory.zswap.writeback
130
131Note that if the store failures are recurring (for e.g if the pages are
132incompressible), users can observe reclaim inefficiency after disabling
133writeback (because the same pages might be rejected again and again).
134
135When there is a sizable amount of cold memory residing in the zswap pool, it
136can be advantageous to proactively write these cold pages to swap and reclaim
137the memory for other use cases. By default, the zswap shrinker is disabled.
138User can enable it as follows::
139
140  echo Y > /sys/module/zswap/parameters/shrinker_enabled
141
142This can be enabled at the boot time if ``CONFIG_ZSWAP_SHRINKER_DEFAULT_ON`` is
143selected.
144
145A debugfs interface is provided for various statistic about pool size, number
146of pages stored, same-value filled pages and various counters for the reasons
147pages are rejected.
148