Commit d846687d7f84e45f23ecf3846dbb43312a1206dd
Committed by
Linus Torvalds
Linus Torvalds
1 parent
74a1166dfe
cgroups: use flex_array in attach_proc
Convert cgroup_attach_proc to use flex_array. The cgroup_attach_proc implementation requires a pre-allocated array to store task pointers to atomically move a thread-group, but asking for a monolithic array with kmalloc() may be unreliable for very large groups. Using flex_array provides the same functionality with less risk of failure. This is a post-patch for cgroup-procs-write.patch. Signed-off-by: Ben Blum <bblum@andrew.cmu.edu> Cc: "Eric W. Biederman" <ebiederm@xmission.com> Cc: Li Zefan <lizf@cn.fujitsu.com> Cc: Matt Helsley <matthltc@us.ibm.com> Reviewed-by: Paul Menage <menage@google.com> Cc: Oleg Nesterov <oleg@redhat.com> Cc: David Rientjes <rientjes@google.com> Cc: Miao Xie <miaox@cn.fujitsu.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Showing 1 changed file with 24 additions and 9 deletions Side-by-side Diff
kernel/cgroup.c
| ... | ... | @@ -57,6 +57,7 @@ |
| 57 | 57 | #include <linux/vmalloc.h> /* TODO: replace with more sophisticated array */ |
| 58 | 58 | #include <linux/eventfd.h> |
| 59 | 59 | #include <linux/poll.h> |
| 60 | +#include <linux/flex_array.h> /* used in cgroup_attach_proc */ | |
| 60 | 61 | |
| 61 | 62 | #include <asm/atomic.h> |
| 62 | 63 | |
| ... | ... | @@ -1995,7 +1996,7 @@ |
| 1995 | 1996 | struct cgroupfs_root *root = cgrp->root; |
| 1996 | 1997 | /* threadgroup list cursor and array */ |
| 1997 | 1998 | struct task_struct *tsk; |
| 1998 | - struct task_struct **group; | |
| 1999 | + struct flex_array *group; | |
| 1999 | 2000 | /* |
| 2000 | 2001 | * we need to make sure we have css_sets for all the tasks we're |
| 2001 | 2002 | * going to move -before- we actually start moving them, so that in |
| 2002 | 2003 | |
| ... | ... | @@ -2012,9 +2013,15 @@ |
| 2012 | 2013 | * and if threads exit, this will just be an over-estimate. |
| 2013 | 2014 | */ |
| 2014 | 2015 | group_size = get_nr_threads(leader); |
| 2015 | - group = kmalloc(group_size * sizeof(*group), GFP_KERNEL); | |
| 2016 | + /* flex_array supports very large thread-groups better than kmalloc. */ | |
| 2017 | + group = flex_array_alloc(sizeof(struct task_struct *), group_size, | |
| 2018 | + GFP_KERNEL); | |
| 2016 | 2019 | if (!group) |
| 2017 | 2020 | return -ENOMEM; |
| 2021 | + /* pre-allocate to guarantee space while iterating in rcu read-side. */ | |
| 2022 | + retval = flex_array_prealloc(group, 0, group_size - 1, GFP_KERNEL); | |
| 2023 | + if (retval) | |
| 2024 | + goto out_free_group_list; | |
| 2018 | 2025 | |
| 2019 | 2026 | /* prevent changes to the threadgroup list while we take a snapshot. */ |
| 2020 | 2027 | rcu_read_lock(); |
| ... | ... | @@ -2037,7 +2044,12 @@ |
| 2037 | 2044 | /* as per above, nr_threads may decrease, but not increase. */ |
| 2038 | 2045 | BUG_ON(i >= group_size); |
| 2039 | 2046 | get_task_struct(tsk); |
| 2040 | - group[i] = tsk; | |
| 2047 | + /* | |
| 2048 | + * saying GFP_ATOMIC has no effect here because we did prealloc | |
| 2049 | + * earlier, but it's good form to communicate our expectations. | |
| 2050 | + */ | |
| 2051 | + retval = flex_array_put_ptr(group, i, tsk, GFP_ATOMIC); | |
| 2052 | + BUG_ON(retval != 0); | |
| 2041 | 2053 | i++; |
| 2042 | 2054 | } while_each_thread(leader, tsk); |
| 2043 | 2055 | /* remember the number of threads in the array for later. */ |
| ... | ... | @@ -2059,7 +2071,8 @@ |
| 2059 | 2071 | if (ss->can_attach_task) { |
| 2060 | 2072 | /* run on each task in the threadgroup. */ |
| 2061 | 2073 | for (i = 0; i < group_size; i++) { |
| 2062 | - retval = ss->can_attach_task(cgrp, group[i]); | |
| 2074 | + tsk = flex_array_get_ptr(group, i); | |
| 2075 | + retval = ss->can_attach_task(cgrp, tsk); | |
| 2063 | 2076 | if (retval) { |
| 2064 | 2077 | failed_ss = ss; |
| 2065 | 2078 | cancel_failed_ss = true; |
| ... | ... | @@ -2075,7 +2088,7 @@ |
| 2075 | 2088 | */ |
| 2076 | 2089 | INIT_LIST_HEAD(&newcg_list); |
| 2077 | 2090 | for (i = 0; i < group_size; i++) { |
| 2078 | - tsk = group[i]; | |
| 2091 | + tsk = flex_array_get_ptr(group, i); | |
| 2079 | 2092 | /* nothing to do if this task is already in the cgroup */ |
| 2080 | 2093 | oldcgrp = task_cgroup_from_root(tsk, root); |
| 2081 | 2094 | if (cgrp == oldcgrp) |
| ... | ... | @@ -2114,7 +2127,7 @@ |
| 2114 | 2127 | ss->pre_attach(cgrp); |
| 2115 | 2128 | } |
| 2116 | 2129 | for (i = 0; i < group_size; i++) { |
| 2117 | - tsk = group[i]; | |
| 2130 | + tsk = flex_array_get_ptr(group, i); | |
| 2118 | 2131 | /* leave current thread as it is if it's already there */ |
| 2119 | 2132 | oldcgrp = task_cgroup_from_root(tsk, root); |
| 2120 | 2133 | if (cgrp == oldcgrp) |
| 2121 | 2134 | |
| ... | ... | @@ -2167,10 +2180,12 @@ |
| 2167 | 2180 | } |
| 2168 | 2181 | } |
| 2169 | 2182 | /* clean up the array of referenced threads in the group. */ |
| 2170 | - for (i = 0; i < group_size; i++) | |
| 2171 | - put_task_struct(group[i]); | |
| 2183 | + for (i = 0; i < group_size; i++) { | |
| 2184 | + tsk = flex_array_get_ptr(group, i); | |
| 2185 | + put_task_struct(tsk); | |
| 2186 | + } | |
| 2172 | 2187 | out_free_group_list: |
| 2173 | - kfree(group); | |
| 2188 | + flex_array_free(group); | |
| 2174 | 2189 | return retval; |
| 2175 | 2190 | } |
| 2176 | 2191 |