inotify, memcg: account inotify instances to kmemcg
Subsystems (4)
PATCH v2 1/1 inotify, memcg: account inotify instances to kmemcg
Currently the fs sysctl inotify/max_user_instances is used to limit the number of inotify instances on the system. For systems running multiple workloads, the per-user namespace sysctl max_inotify_instances can be used to further partition inotify instances. However there is no easy way to set a sensible system level max limit on inotify instances and further partition it between the workloads. It is much easier to charge the underlying resource (i.e. memory) behind the inotify instances to the memcg of the workload and let their memory limits limit the number of inotify instances they can create.
With inotify instances charged to memcg, the admin can simply set max_user_instances to INT_MAX and let the memcg limits of the jobs limit their inotify instances.
---
Changes since v1:
- introduce fsnotify_alloc_user_group() and convert fanotify in addition to inotify to use that function. [suggested by Amir]
fs/notify/fanotify/fanotify_user.c | 2 +- fs/notify/group.c | 25 ++++++++++++++++++++----- fs/notify/inotify/inotify_user.c | 4 ++-- include/linux/fsnotify_backend.h | 1 + 4 files changed, 24 insertions(+), 8 deletions(-)
fs/notify/fanotify/fanotify_user.c +1 -1
@@ -976,7 +976,7 @@ SYSCALL_DEFINE2(fanotify_init, unsigned int, flags, unsigned int, event_f_flags) f_flags |=O_NONBLOCK ;/* fsnotify_alloc_group takes a ref. Dropped in fanotify_release */ - group =fsnotify_alloc_group (&fanotify_fsnotify_ops); + group =fsnotify_alloc_user_group (&fanotify_fsnotify_ops);if (IS_ERR (group)) {free_uid (user);return PTR_ERR (group);
fs/notify/group.c +20 -5
@@ -111,14 +111,12 @@ void fsnotify_put_group(struct fsnotify_group *group) }EXPORT_SYMBOL_GPL (fsnotify_put_group); -/* -* Create a new fsnotify_group and hold a reference for the group returned. -*/ -struct fsnotify_group *fsnotify_alloc_group (const struct fsnotify_ops *ops) +static struct fsnotify_group *__fsnotify_alloc_group ( +const struct fsnotify_ops *ops,gfp_t gfp) {struct fsnotify_group *group; - group =kzalloc (sizeof (struct fsnotify_group ),GFP_KERNEL ); + group =kzalloc (sizeof (struct fsnotify_group ), gfp);if (!group)return ERR_PTR (-ENOMEM );
@@ -139,8 +137,25 @@ struct fsnotify_group *fsnotify_alloc_group(const struct fsnotify_ops *ops)return group; } + +/* +* Create a new fsnotify_group and hold a reference for the group returned. +*/ +struct fsnotify_group *fsnotify_alloc_group (const struct fsnotify_ops *ops) +{ +return __fsnotify_alloc_group (ops,GFP_KERNEL ); +}EXPORT_SYMBOL_GPL (fsnotify_alloc_group); +/* +* Create a new fsnotify_group and hold a reference for the group returned. +*/ +struct fsnotify_group *fsnotify_alloc_user_group (const struct fsnotify_ops *ops) +{ +return __fsnotify_alloc_group (ops,GFP_KERNEL_ACCOUNT ); +} +EXPORT_SYMBOL_GPL (fsnotify_alloc_user_group); +int fsnotify_fasync (int fd,struct file *file,int on) {struct fsnotify_group *group = file->private_data;
fs/notify/inotify/inotify_user.c +2 -2
@@ -632,11 +632,11 @@ static struct fsnotify_group *inotify_new_group(unsigned int max_events)struct fsnotify_group *group;struct inotify_event_info *oevent; - group =fsnotify_alloc_group (&inotify_fsnotify_ops); + group =fsnotify_alloc_user_group (&inotify_fsnotify_ops);if (IS_ERR (group))return group; - oevent =kmalloc (sizeof (struct inotify_event_info ),GFP_KERNEL ); + oevent =kmalloc (sizeof (struct inotify_event_info ),GFP_KERNEL_ACCOUNT );if (unlikely (!oevent)) {fsnotify_destroy_group (group);return ERR_PTR (-ENOMEM );
include/linux/fsnotify_backend.h +1 -0
@@ -470,6 +470,7 @@ static inline void fsnotify_update_flags(struct dentry *dentry)/* create a new group */ extern struct fsnotify_group *fsnotify_alloc_group (const struct fsnotify_ops *ops); +extern struct fsnotify_group *fsnotify_alloc_user_group (const struct fsnotify_ops *ops);/* get reference to a group */ extern void fsnotify_get_group (struct fsnotify_group *group);/* drop reference on a group from fsnotify_alloc_group */
Signature · 1 line
2.29.2.684.gfbc64c5ab5-goog
Quoted 11 lines · Shakeel Butt
Currently the fs sysctl inotify/max_user_instances is used to limit the number of inotify instances on the system. For systems running multiple workloads, the per-user namespace sysctl
max_inotify_instancescan be used to further partition inotify instances. However there is no easy way to set a sensible system level max limit on inotify instances and further partition it between the workloads. It is much easier to charge the underlying resource (i.e. memory) behind the inotify instances to the memcg of the workload and let their memory limits limit the number of inotify instances they can create.
Quoted 5 lines
With inotify instances charged to memcg, the admin can simply set
max_user_instancestoINT_MAXand let the memcg limits of the jobs limit their inotify instances.Signed-off-by Shakeel Butt <[email protected]>
Quoted 105 lines
---
Changes since v1:
- introducefsnotify_alloc_user_group()and convert fanotify in addition to inotify to use that function. [suggested by Amir]fs/notify/fanotify/fanotify_user.c | 2 +- fs/notify/group.c | 25 ++++++++++++++++++++----- fs/notify/inotify/inotify_user.c | 4 ++-- include/linux/fsnotify_backend.h | 1 + 4 files changed, 24 insertions(+), 8 deletions(-)fs/notify/fanotify/fanotify_user.c +1
-1@@ -976,7 +976,7 @@ SYSCALL_DEFINE2(fanotify_init, unsigned int, flags, unsigned int, event_f_flags) f_flags |=O_NONBLOCK ;/* fsnotify_alloc_group takes a ref. Dropped in fanotify_release */ - group =fsnotify_alloc_group (&fanotify_fsnotify_ops); + group =fsnotify_alloc_user_group (&fanotify_fsnotify_ops);if (IS_ERR (group)) {free_uid (user);return PTR_ERR (group);fs/notify/group.c +20
-5@@ -111,14 +111,12 @@ void fsnotify_put_group(struct fsnotify_group *group) }EXPORT_SYMBOL_GPL (fsnotify_put_group); -/* -* Create a new fsnotify_group and hold a reference for the group returned. -*/ -struct fsnotify_group *fsnotify_alloc_group (const struct fsnotify_ops *ops) +static struct fsnotify_group *__fsnotify_alloc_group ( +const struct fsnotify_ops *ops,gfp_t gfp) {struct fsnotify_group *group; - group =kzalloc (sizeof (struct fsnotify_group ),GFP_KERNEL ); + group =kzalloc (sizeof (struct fsnotify_group ), gfp);if (!group)return ERR_PTR (-ENOMEM );@@ -139,8 +137,25 @@ struct fsnotify_group *fsnotify_alloc_group(const struct fsnotify_ops *ops)return group; } + +/* +* Create a new fsnotify_group and hold a reference for the group returned. +*/ +struct fsnotify_group *fsnotify_alloc_group (const struct fsnotify_ops *ops) +{ +return __fsnotify_alloc_group (ops,GFP_KERNEL ); +}EXPORT_SYMBOL_GPL (fsnotify_alloc_group); +/* +* Create a new fsnotify_group and hold a reference for the group returned. +*/ +struct fsnotify_group *fsnotify_alloc_user_group (const struct fsnotify_ops *ops) +{ +return __fsnotify_alloc_group (ops,GFP_KERNEL_ACCOUNT ); +} +EXPORT_SYMBOL_GPL (fsnotify_alloc_user_group); +int fsnotify_fasync (int fd,struct file *file,int on) {struct fsnotify_group *group = file->private_data;fs/notify/inotify/inotify_user.c +2
-2@@ -632,11 +632,11 @@ static struct fsnotify_group *inotify_new_group(unsigned int max_events)struct fsnotify_group *group;struct inotify_event_info *oevent; - group =fsnotify_alloc_group (&inotify_fsnotify_ops); + group =fsnotify_alloc_user_group (&inotify_fsnotify_ops);if (IS_ERR (group))return group; - oevent =kmalloc (sizeof (struct inotify_event_info ),GFP_KERNEL ); + oevent =kmalloc (sizeof (struct inotify_event_info ),GFP_KERNEL_ACCOUNT );if (unlikely (!oevent)) {fsnotify_destroy_group (group);return ERR_PTR (-ENOMEM );include/linux/fsnotify_backend.h +1
-1@@ -470,6 +470,7 @@ static inline void fsnotify_update_flags(struct dentry *dentry)/* create a new group */ extern struct fsnotify_group *fsnotify_alloc_group (const struct fsnotify_ops *ops); +extern struct fsnotify_group *fsnotify_alloc_user_group (const struct fsnotify_ops *ops);/* get reference to a group */ extern void fsnotify_get_group (struct fsnotify_group *group);/* drop reference on a group from fsnotify_alloc_group */ --2.29.2.684.gfbc64c5ab5-goog
Quoted 13 lines · Amir Goldstein
Quoted 11 lines · Shakeel Butt
Currently the fs sysctl inotify/max_user_instances is used to limit the number of inotify instances on the system. For systems running multiple workloads, the per-user namespace sysctl
max_inotify_instancescan be used to further partition inotify instances. However there is no easy way to set a sensible system level max limit on inotify instances and further partition it between the workloads. It is much easier to charge the underlying resource (i.e. memory) behind the inotify instances to the memcg of the workload and let their memory limits limit the number of inotify instances they can create.
Quoted 6 lines
Quoted 5 lines
With inotify instances charged to memcg, the admin can simply set
max_user_instancestoINT_MAXand let the memcg limits of the jobs limit their inotify instances.Signed-off-by Shakeel Butt <[email protected]>Reviewed-by Amir Goldstein <[email protected]>
Thanks a lot.
Jan Kara
maintainerQuoted 12 lines · Amir Goldstein
Quoted 11 lines · Shakeel Butt
Currently the fs sysctl inotify/max_user_instances is used to limit the number of inotify instances on the system. For systems running multiple workloads, the per-user namespace sysctl
max_inotify_instancescan be used to further partition inotify instances. However there is no easy way to set a sensible system level max limit on inotify instances and further partition it between the workloads. It is much easier to charge the underlying resource (i.e. memory) behind the inotify instances to the memcg of the workload and let their memory limits limit the number of inotify instances they can create.
Quoted 6 lines
Quoted 5 lines
With inotify instances charged to memcg, the admin can simply set
max_user_instancestoINT_MAXand let the memcg limits of the jobs limit their inotify instances.Signed-off-by Shakeel Butt <[email protected]>Reviewed-by Amir Goldstein <[email protected]>
The patch looks good to me. Thanks for review Amir. I've picked the patch to my tree although it's currently in flux as we're currently in the middle of the merge window. So I won't push out the branch with the patch yet since I plan on rebasing it on top of -rc1 anyway.
Honza
Quoted 104 lines
Quoted 104 lines
Changes since v1:
- introducefsnotify_alloc_user_group()and convert fanotify in addition to inotify to use that function. [suggested by Amir]fs/notify/fanotify/fanotify_user.c | 2 +- fs/notify/group.c | 25 ++++++++++++++++++++----- fs/notify/inotify/inotify_user.c | 4 ++-- include/linux/fsnotify_backend.h | 1 + 4 files changed, 24 insertions(+), 8 deletions(-)fs/notify/fanotify/fanotify_user.c +1
-1@@ -976,7 +976,7 @@ SYSCALL_DEFINE2(fanotify_init, unsigned int, flags, unsigned int, event_f_flags) f_flags |=O_NONBLOCK ;/* fsnotify_alloc_group takes a ref. Dropped in fanotify_release */ - group =fsnotify_alloc_group (&fanotify_fsnotify_ops); + group =fsnotify_alloc_user_group (&fanotify_fsnotify_ops);if (IS_ERR (group)) {free_uid (user);return PTR_ERR (group);fs/notify/group.c +20
-5@@ -111,14 +111,12 @@ void fsnotify_put_group(struct fsnotify_group *group) }EXPORT_SYMBOL_GPL (fsnotify_put_group); -/* -* Create a new fsnotify_group and hold a reference for the group returned. -*/ -struct fsnotify_group *fsnotify_alloc_group (const struct fsnotify_ops *ops) +static struct fsnotify_group *__fsnotify_alloc_group ( +const struct fsnotify_ops *ops,gfp_t gfp) {struct fsnotify_group *group; - group =kzalloc (sizeof (struct fsnotify_group ),GFP_KERNEL ); + group =kzalloc (sizeof (struct fsnotify_group ), gfp);if (!group)return ERR_PTR (-ENOMEM );@@ -139,8 +137,25 @@ struct fsnotify_group *fsnotify_alloc_group(const struct fsnotify_ops *ops)return group; } + +/* +* Create a new fsnotify_group and hold a reference for the group returned. +*/ +struct fsnotify_group *fsnotify_alloc_group (const struct fsnotify_ops *ops) +{ +return __fsnotify_alloc_group (ops,GFP_KERNEL ); +}EXPORT_SYMBOL_GPL (fsnotify_alloc_group); +/* +* Create a new fsnotify_group and hold a reference for the group returned. +*/ +struct fsnotify_group *fsnotify_alloc_user_group (const struct fsnotify_ops *ops) +{ +return __fsnotify_alloc_group (ops,GFP_KERNEL_ACCOUNT ); +} +EXPORT_SYMBOL_GPL (fsnotify_alloc_user_group); +int fsnotify_fasync (int fd,struct file *file,int on) {struct fsnotify_group *group = file->private_data;fs/notify/inotify/inotify_user.c +2
-2@@ -632,11 +632,11 @@ static struct fsnotify_group *inotify_new_group(unsigned int max_events)struct fsnotify_group *group;struct inotify_event_info *oevent; - group =fsnotify_alloc_group (&inotify_fsnotify_ops); + group =fsnotify_alloc_user_group (&inotify_fsnotify_ops);if (IS_ERR (group))return group; - oevent =kmalloc (sizeof (struct inotify_event_info ),GFP_KERNEL ); + oevent =kmalloc (sizeof (struct inotify_event_info ),GFP_KERNEL_ACCOUNT );if (unlikely (!oevent)) {fsnotify_destroy_group (group);return ERR_PTR (-ENOMEM );include/linux/fsnotify_backend.h +1
-1@@ -470,6 +470,7 @@ static inline void fsnotify_update_flags(struct dentry *dentry)/* create a new group */ extern struct fsnotify_group *fsnotify_alloc_group (const struct fsnotify_ops *ops); +extern struct fsnotify_group *fsnotify_alloc_user_group (const struct fsnotify_ops *ops);/* get reference to a group */ extern void fsnotify_get_group (struct fsnotify_group *group);/* drop reference on a group from fsnotify_alloc_group */ --2.29.2.684.gfbc64c5ab5-goog
Signature · 2 lines
Jan Kara <[email protected]> SUSE Labs, CR