user namespaces: document CFS behavior
PATCH 1/1 user namespaces: document CFS behavior
Documented the currently bogus state of support for CFS user groups with user namespaces. In particular, all users in a user namespace should be children of the user which created the user namespace. This is yet to be implemented.
--- Documentation/scheduler/sched-design-CFS.txt | 21 +++++++++++++++++++++ kernel/user.c | 8 +++++++- 2 files changed, 28 insertions(+), 1 deletions(-)
Documentation/scheduler/sched-design-CFS.txt +21 -0
@@ -273,3 +273,24 @@ task groups and modify their CPU share using the "cgroups" pseudo filesystem. # #Launch gmplayer (or your favourite movie player) # echo <movie_player_pid> > multimedia/tasks + +8. Implementation note: user namespaces + +User namespaces are intended to be hierarchical. But they are currently +only partially implemented. Each of those has ramifications for CFS. + +First, since user namespaces are hierarchical, the /sys/kernel/uids +presentation is inadequate. Eventually we will likely want to use sysfs +tagging to provide private views of /sys/kernel/uids within each user +namespace. + +Second, the hierarchical nature is intended to support completely +unprivileged use of user namespaces. So if using user groups, then +we want the users in a user namespace to be children of the user +who created it. + +That is currently unimplemented. So instead, every user in a new +user namespace will receive 1024 shares just like any user in the +initial user namespace. Note that at the moment creation of a new +user namespace requires each of CAP_SYS_ADMIN, CAP_SETUID, and +CAP_SETGID.
kernel/user.c +7 -1
@@ -239,7 +239,13 @@ static struct kobj_type uids_ktype = { .release = uids_release, }; -/* create /sys/kernel/uids/<uid>/cpu_share file for this user */ +/* +* Create /sys/kernel/uids/<uid>/cpu_share file for this user +* We do not create this file for users in a user namespace (until +* sysfs tagging is implemented). +* +* See Documentation/scheduler/sched-design-CFS.txt for ramifications. +*/ static int uids_user_create (struct user_struct *up) {struct kobject *kobj = &up->kobj;
Signature · 1 line
1.5.4.3
James Morris
AppliedQuoted 9 lines · Serge E. Hallyn
Documented the currently bogus state of support for CFS user groups with user namespaces. In particular, all users in a user namespace should be children of the user which created the user namespace. This is yet to be implemented.
Signed-off-by Serge E. Hallyn <[email protected]>Acked-by Dhaval Giani <[email protected]>Signed-off-by Serge E. Hallyn <[email protected]>
Applied.
Signature · 2 lines
James Morris <[email protected]>
Quoted 7 lines · James Morris
Quoted 5 lines · Serge E. Hallyn
Documented the currently bogus state of support for CFS user groups with user namespaces. In particular, all users in a user namespace should be children of the user which created the user namespace. This is yet to be implemented.
Quoted 6 lines
Quoted 4 lines
Signed-off-by Serge E. Hallyn <[email protected]>Acked-by Dhaval Giani <[email protected]>Signed-off-by Serge E. Hallyn <[email protected]>Applied.
Thanks, James.
-serge
Quoted 7 lines · James Morris
Quoted 5 lines · Serge E. Hallyn
Documented the currently bogus state of support for CFS user groups with user namespaces. In particular, all users in a user namespace should be children of the user which created the user namespace. This is yet to be implemented.
Quoted 6 lines
Quoted 4 lines
Signed-off-by Serge E. Hallyn <[email protected]>Acked-by Dhaval Giani <[email protected]>Signed-off-by Serge E. Hallyn <[email protected]>Applied.
hm, i'd rather carry these in the scheduler tree as we've got some changes in this area already and because these are arguably scheduler changes:
Documentation/scheduler/sched-design-CFS.txt | 21 +++++++++++++++++++++ kernel/user.c | 8 +++++++-
the incompleteness of CONFIG_USER_SCHED is not addressed yet.
Ok?
Ingo
Quoted 10 lines · Ingo Molnar
hm, i'd rather carry these in the scheduler tree as we've got some changes in this area already and because these are arguably scheduler changes:
Documentation/scheduler/sched-design-CFS.txt | 21 +++++++++++++++++++++ kernel/user.c | 8 +++++++-the incompleteness of
CONFIG_USER_SCHEDis not addressed yet.Ok?
They're already published in my tree. Do you want me to revert them?
Signature · 2 lines
James Morris <[email protected]>
Quoted 6 lines · James Morris
Quoted 4 lines · Ingo Molnar
hm, i'd rather carry these in the scheduler tree as we've got some changes in this area already and because these are arguably scheduler changes:
Quoted 8 lines
Quoted 6 lines
Documentation/scheduler/sched-design-CFS.txt | 21 +++++++++++++++++++++ kernel/user.c | 8 +++++++-the incompleteness of
CONFIG_USER_SCHEDis not addressed yet.Ok?
They're already published in my tree. Do you want me to revert them?
I've also got it published :-/ If it's at the tail of your commits, can you reset it?
Ingo
Quoted 4 lines · Ingo Molnar
Quoted 1 line
They're already published in my tree. Do you want me to revert them?
I've also got it published :-/ If it's at the tail of your commits, can you reset it?
There have been commits since, and people are using the tree in any case.
- James
Signature · 2 lines
James Morris <[email protected]>
Quoted 8 lines · James Morris
Quoted 4 lines · Ingo Molnar
Quoted 1 line
They're already published in my tree. Do you want me to revert them?
I've also got it published :-/ If it's at the tail of your commits, can you reset it?
There have been commits since, and people are using the tree in any case.
Yikes. Maybe it's best (henceforth) to keep user namespaces patches in their own tree toward the end of linux-next? There certainly are security implications with all of them, but other things (filesystem, sched) will be impacted as well...
Sorry.
-serge
Quoted 8 lines · Serge E. Hallyn
Quoted 7 lines · James Morris
Quoted 4 lines · Ingo Molnar
Quoted 1 line
They're already published in my tree. Do you want me to revert them?
I've also got it published :-/ If it's at the tail of your commits, can you reset it?
Quoted 6 lines
Quoted 1 line
There have been commits since, and people are using the tree in any case.
Yikes. Maybe it's best (henceforth) to keep user namespaces patches in their own tree toward the end of linux-next? There certainly are security implications with all of them, but other things (filesystem, sched) will be impacted as well...
i'm fine with having this in James's tree - i didnt realize that you had another commit that created dependencies.
(if such situations ever cause real damage then we can always create a separate branch to track it cleanly. That's not needed now.)
Ingo