diff options
Diffstat (limited to 'block')
-rw-r--r-- | block/Kconfig | 13 | ||||
-rw-r--r-- | block/Kconfig.iosched | 1 | ||||
-rw-r--r-- | block/Makefile | 1 | ||||
-rw-r--r-- | block/blk-cgroup.c | 177 | ||||
-rw-r--r-- | block/blk-cgroup.h | 58 |
5 files changed, 250 insertions, 0 deletions
diff --git a/block/Kconfig b/block/Kconfig index 9be0b56eaee1..6ba1a8e3388b 100644 --- a/block/Kconfig +++ b/block/Kconfig | |||
@@ -77,6 +77,19 @@ config BLK_DEV_INTEGRITY | |||
77 | T10/SCSI Data Integrity Field or the T13/ATA External Path | 77 | T10/SCSI Data Integrity Field or the T13/ATA External Path |
78 | Protection. If in doubt, say N. | 78 | Protection. If in doubt, say N. |
79 | 79 | ||
80 | config BLK_CGROUP | ||
81 | bool | ||
82 | depends on CGROUPS | ||
83 | default n | ||
84 | ---help--- | ||
85 | Generic block IO controller cgroup interface. This is the common | ||
86 | cgroup interface which should be used by various IO controlling | ||
87 | policies. | ||
88 | |||
89 | Currently, CFQ IO scheduler uses it to recognize task groups and | ||
90 | control disk bandwidth allocation (proportional time slice allocation) | ||
91 | to such task groups. | ||
92 | |||
80 | endif # BLOCK | 93 | endif # BLOCK |
81 | 94 | ||
82 | config BLOCK_COMPAT | 95 | config BLOCK_COMPAT |
diff --git a/block/Kconfig.iosched b/block/Kconfig.iosched index 8bd105115a69..be0280deec29 100644 --- a/block/Kconfig.iosched +++ b/block/Kconfig.iosched | |||
@@ -23,6 +23,7 @@ config IOSCHED_DEADLINE | |||
23 | 23 | ||
24 | config IOSCHED_CFQ | 24 | config IOSCHED_CFQ |
25 | tristate "CFQ I/O scheduler" | 25 | tristate "CFQ I/O scheduler" |
26 | select BLK_CGROUP | ||
26 | default y | 27 | default y |
27 | ---help--- | 28 | ---help--- |
28 | The CFQ I/O scheduler tries to distribute bandwidth equally | 29 | The CFQ I/O scheduler tries to distribute bandwidth equally |
diff --git a/block/Makefile b/block/Makefile index 7914108952f2..cb2d515ebd6e 100644 --- a/block/Makefile +++ b/block/Makefile | |||
@@ -8,6 +8,7 @@ obj-$(CONFIG_BLOCK) := elevator.o blk-core.o blk-tag.o blk-sysfs.o \ | |||
8 | blk-iopoll.o ioctl.o genhd.o scsi_ioctl.o | 8 | blk-iopoll.o ioctl.o genhd.o scsi_ioctl.o |
9 | 9 | ||
10 | obj-$(CONFIG_BLK_DEV_BSG) += bsg.o | 10 | obj-$(CONFIG_BLK_DEV_BSG) += bsg.o |
11 | obj-$(CONFIG_BLK_CGROUP) += blk-cgroup.o | ||
11 | obj-$(CONFIG_IOSCHED_NOOP) += noop-iosched.o | 12 | obj-$(CONFIG_IOSCHED_NOOP) += noop-iosched.o |
12 | obj-$(CONFIG_IOSCHED_DEADLINE) += deadline-iosched.o | 13 | obj-$(CONFIG_IOSCHED_DEADLINE) += deadline-iosched.o |
13 | obj-$(CONFIG_IOSCHED_CFQ) += cfq-iosched.o | 14 | obj-$(CONFIG_IOSCHED_CFQ) += cfq-iosched.o |
diff --git a/block/blk-cgroup.c b/block/blk-cgroup.c new file mode 100644 index 000000000000..4f6afd76ec59 --- /dev/null +++ b/block/blk-cgroup.c | |||
@@ -0,0 +1,177 @@ | |||
1 | /* | ||
2 | * Common Block IO controller cgroup interface | ||
3 | * | ||
4 | * Based on ideas and code from CFQ, CFS and BFQ: | ||
5 | * Copyright (C) 2003 Jens Axboe <axboe@kernel.dk> | ||
6 | * | ||
7 | * Copyright (C) 2008 Fabio Checconi <fabio@gandalf.sssup.it> | ||
8 | * Paolo Valente <paolo.valente@unimore.it> | ||
9 | * | ||
10 | * Copyright (C) 2009 Vivek Goyal <vgoyal@redhat.com> | ||
11 | * Nauman Rafique <nauman@google.com> | ||
12 | */ | ||
13 | #include <linux/ioprio.h> | ||
14 | #include "blk-cgroup.h" | ||
15 | |||
16 | struct blkio_cgroup blkio_root_cgroup = { .weight = 2*BLKIO_WEIGHT_DEFAULT }; | ||
17 | |||
18 | struct blkio_cgroup *cgroup_to_blkio_cgroup(struct cgroup *cgroup) | ||
19 | { | ||
20 | return container_of(cgroup_subsys_state(cgroup, blkio_subsys_id), | ||
21 | struct blkio_cgroup, css); | ||
22 | } | ||
23 | |||
24 | void blkiocg_add_blkio_group(struct blkio_cgroup *blkcg, | ||
25 | struct blkio_group *blkg, void *key) | ||
26 | { | ||
27 | unsigned long flags; | ||
28 | |||
29 | spin_lock_irqsave(&blkcg->lock, flags); | ||
30 | rcu_assign_pointer(blkg->key, key); | ||
31 | hlist_add_head_rcu(&blkg->blkcg_node, &blkcg->blkg_list); | ||
32 | spin_unlock_irqrestore(&blkcg->lock, flags); | ||
33 | } | ||
34 | |||
35 | int blkiocg_del_blkio_group(struct blkio_group *blkg) | ||
36 | { | ||
37 | /* Implemented later */ | ||
38 | return 0; | ||
39 | } | ||
40 | |||
41 | /* called under rcu_read_lock(). */ | ||
42 | struct blkio_group *blkiocg_lookup_group(struct blkio_cgroup *blkcg, void *key) | ||
43 | { | ||
44 | struct blkio_group *blkg; | ||
45 | struct hlist_node *n; | ||
46 | void *__key; | ||
47 | |||
48 | hlist_for_each_entry_rcu(blkg, n, &blkcg->blkg_list, blkcg_node) { | ||
49 | __key = blkg->key; | ||
50 | if (__key == key) | ||
51 | return blkg; | ||
52 | } | ||
53 | |||
54 | return NULL; | ||
55 | } | ||
56 | |||
57 | #define SHOW_FUNCTION(__VAR) \ | ||
58 | static u64 blkiocg_##__VAR##_read(struct cgroup *cgroup, \ | ||
59 | struct cftype *cftype) \ | ||
60 | { \ | ||
61 | struct blkio_cgroup *blkcg; \ | ||
62 | \ | ||
63 | blkcg = cgroup_to_blkio_cgroup(cgroup); \ | ||
64 | return (u64)blkcg->__VAR; \ | ||
65 | } | ||
66 | |||
67 | SHOW_FUNCTION(weight); | ||
68 | #undef SHOW_FUNCTION | ||
69 | |||
70 | static int | ||
71 | blkiocg_weight_write(struct cgroup *cgroup, struct cftype *cftype, u64 val) | ||
72 | { | ||
73 | struct blkio_cgroup *blkcg; | ||
74 | |||
75 | if (val < BLKIO_WEIGHT_MIN || val > BLKIO_WEIGHT_MAX) | ||
76 | return -EINVAL; | ||
77 | |||
78 | blkcg = cgroup_to_blkio_cgroup(cgroup); | ||
79 | blkcg->weight = (unsigned int)val; | ||
80 | return 0; | ||
81 | } | ||
82 | |||
83 | struct cftype blkio_files[] = { | ||
84 | { | ||
85 | .name = "weight", | ||
86 | .read_u64 = blkiocg_weight_read, | ||
87 | .write_u64 = blkiocg_weight_write, | ||
88 | }, | ||
89 | }; | ||
90 | |||
91 | static int blkiocg_populate(struct cgroup_subsys *subsys, struct cgroup *cgroup) | ||
92 | { | ||
93 | return cgroup_add_files(cgroup, subsys, blkio_files, | ||
94 | ARRAY_SIZE(blkio_files)); | ||
95 | } | ||
96 | |||
97 | static void blkiocg_destroy(struct cgroup_subsys *subsys, struct cgroup *cgroup) | ||
98 | { | ||
99 | struct blkio_cgroup *blkcg = cgroup_to_blkio_cgroup(cgroup); | ||
100 | |||
101 | free_css_id(&blkio_subsys, &blkcg->css); | ||
102 | kfree(blkcg); | ||
103 | } | ||
104 | |||
105 | static struct cgroup_subsys_state * | ||
106 | blkiocg_create(struct cgroup_subsys *subsys, struct cgroup *cgroup) | ||
107 | { | ||
108 | struct blkio_cgroup *blkcg, *parent_blkcg; | ||
109 | |||
110 | if (!cgroup->parent) { | ||
111 | blkcg = &blkio_root_cgroup; | ||
112 | goto done; | ||
113 | } | ||
114 | |||
115 | /* Currently we do not support hierarchy deeper than two level (0,1) */ | ||
116 | parent_blkcg = cgroup_to_blkio_cgroup(cgroup->parent); | ||
117 | if (css_depth(&parent_blkcg->css) > 0) | ||
118 | return ERR_PTR(-EINVAL); | ||
119 | |||
120 | blkcg = kzalloc(sizeof(*blkcg), GFP_KERNEL); | ||
121 | if (!blkcg) | ||
122 | return ERR_PTR(-ENOMEM); | ||
123 | |||
124 | blkcg->weight = BLKIO_WEIGHT_DEFAULT; | ||
125 | done: | ||
126 | spin_lock_init(&blkcg->lock); | ||
127 | INIT_HLIST_HEAD(&blkcg->blkg_list); | ||
128 | |||
129 | return &blkcg->css; | ||
130 | } | ||
131 | |||
132 | /* | ||
133 | * We cannot support shared io contexts, as we have no mean to support | ||
134 | * two tasks with the same ioc in two different groups without major rework | ||
135 | * of the main cic data structures. For now we allow a task to change | ||
136 | * its cgroup only if it's the only owner of its ioc. | ||
137 | */ | ||
138 | static int blkiocg_can_attach(struct cgroup_subsys *subsys, | ||
139 | struct cgroup *cgroup, struct task_struct *tsk, | ||
140 | bool threadgroup) | ||
141 | { | ||
142 | struct io_context *ioc; | ||
143 | int ret = 0; | ||
144 | |||
145 | /* task_lock() is needed to avoid races with exit_io_context() */ | ||
146 | task_lock(tsk); | ||
147 | ioc = tsk->io_context; | ||
148 | if (ioc && atomic_read(&ioc->nr_tasks) > 1) | ||
149 | ret = -EINVAL; | ||
150 | task_unlock(tsk); | ||
151 | |||
152 | return ret; | ||
153 | } | ||
154 | |||
155 | static void blkiocg_attach(struct cgroup_subsys *subsys, struct cgroup *cgroup, | ||
156 | struct cgroup *prev, struct task_struct *tsk, | ||
157 | bool threadgroup) | ||
158 | { | ||
159 | struct io_context *ioc; | ||
160 | |||
161 | task_lock(tsk); | ||
162 | ioc = tsk->io_context; | ||
163 | if (ioc) | ||
164 | ioc->cgroup_changed = 1; | ||
165 | task_unlock(tsk); | ||
166 | } | ||
167 | |||
168 | struct cgroup_subsys blkio_subsys = { | ||
169 | .name = "blkio", | ||
170 | .create = blkiocg_create, | ||
171 | .can_attach = blkiocg_can_attach, | ||
172 | .attach = blkiocg_attach, | ||
173 | .destroy = blkiocg_destroy, | ||
174 | .populate = blkiocg_populate, | ||
175 | .subsys_id = blkio_subsys_id, | ||
176 | .use_id = 1, | ||
177 | }; | ||
diff --git a/block/blk-cgroup.h b/block/blk-cgroup.h new file mode 100644 index 000000000000..ba5703f69b42 --- /dev/null +++ b/block/blk-cgroup.h | |||
@@ -0,0 +1,58 @@ | |||
1 | #ifndef _BLK_CGROUP_H | ||
2 | #define _BLK_CGROUP_H | ||
3 | /* | ||
4 | * Common Block IO controller cgroup interface | ||
5 | * | ||
6 | * Based on ideas and code from CFQ, CFS and BFQ: | ||
7 | * Copyright (C) 2003 Jens Axboe <axboe@kernel.dk> | ||
8 | * | ||
9 | * Copyright (C) 2008 Fabio Checconi <fabio@gandalf.sssup.it> | ||
10 | * Paolo Valente <paolo.valente@unimore.it> | ||
11 | * | ||
12 | * Copyright (C) 2009 Vivek Goyal <vgoyal@redhat.com> | ||
13 | * Nauman Rafique <nauman@google.com> | ||
14 | */ | ||
15 | |||
16 | #include <linux/cgroup.h> | ||
17 | |||
18 | struct blkio_cgroup { | ||
19 | struct cgroup_subsys_state css; | ||
20 | unsigned int weight; | ||
21 | spinlock_t lock; | ||
22 | struct hlist_head blkg_list; | ||
23 | }; | ||
24 | |||
25 | struct blkio_group { | ||
26 | /* An rcu protected unique identifier for the group */ | ||
27 | void *key; | ||
28 | struct hlist_node blkcg_node; | ||
29 | }; | ||
30 | |||
31 | #define BLKIO_WEIGHT_MIN 100 | ||
32 | #define BLKIO_WEIGHT_MAX 1000 | ||
33 | #define BLKIO_WEIGHT_DEFAULT 500 | ||
34 | |||
35 | #ifdef CONFIG_BLK_CGROUP | ||
36 | extern struct blkio_cgroup blkio_root_cgroup; | ||
37 | extern struct blkio_cgroup *cgroup_to_blkio_cgroup(struct cgroup *cgroup); | ||
38 | extern void blkiocg_add_blkio_group(struct blkio_cgroup *blkcg, | ||
39 | struct blkio_group *blkg, void *key); | ||
40 | extern int blkiocg_del_blkio_group(struct blkio_group *blkg); | ||
41 | extern struct blkio_group *blkiocg_lookup_group(struct blkio_cgroup *blkcg, | ||
42 | void *key); | ||
43 | #else | ||
44 | static inline struct blkio_cgroup * | ||
45 | cgroup_to_blkio_cgroup(struct cgroup *cgroup) { return NULL; } | ||
46 | |||
47 | static inline void blkiocg_add_blkio_group(struct blkio_cgroup *blkcg, | ||
48 | struct blkio_group *blkg, void *key) | ||
49 | { | ||
50 | } | ||
51 | |||
52 | static inline int | ||
53 | blkiocg_del_blkio_group(struct blkio_group *blkg) { return 0; } | ||
54 | |||
55 | static inline struct blkio_group * | ||
56 | blkiocg_lookup_group(struct blkio_cgroup *blkcg, void *key) { return NULL; } | ||
57 | #endif | ||
58 | #endif /* _BLK_CGROUP_H */ | ||