| Message ID | 20221206171340.139790-4-hannes@cmpxchg.org (mailing list archive) |
|---|---|
| State | New |
| Headers | show |
| Series | mm: push down lock_page_memcg() | expand |
On Tue, Dec 6, 2022 at 9:14 AM Johannes Weiner <hannes@cmpxchg.org> wrote: > > Charge moving mode in cgroup1 allows memory to follow tasks as they > migrate between cgroups. This is, and always has been, a questionable > thing to do - for several reasons. > > First, it's expensive. Pages need to be identified, locked and > isolated from various MM operations, and reassigned, one by one. > > Second, it's unreliable. Once pages are charged to a cgroup, there > isn't always a clear owner task anymore. Cache isn't moved at all, for > example. Mapped memory is moved - but if trylocking or isolating a > page fails, it's arbitrarily left behind. Frequent moving between > domains may leave a task's memory scattered all over the place. > > Third, it isn't really needed. Launcher tasks can kick off workload > tasks directly in their target cgroup. Using dedicated per-workload > groups allows fine-grained policy adjustments - no need to move tasks > and their physical pages between control domains. The feature was > never forward-ported to cgroup2, and it hasn't been missed. > > Despite it being a niche usecase, the maintenance overhead of > supporting it is enormous. Because pages are moved while they are live > and subject to various MM operations, the synchronization rules are > complicated. There are lock_page_memcg() in MM and FS code, which > non-cgroup people don't understand. In some cases we've been able to > shift code and cgroup API calls around such that we can rely on native > locking as much as possible. But that's fragile, and sometimes we need > to hold MM locks for longer than we otherwise would (pte lock e.g.). > > Mark the feature deprecated. Hopefully we can remove it soon. > > Signed-off-by: Johannes Weiner <hannes@cmpxchg.org> Acked-by: Shakeel Butt <shakeelb@google.com> I would request this patch to be backported to stable kernels as well for early warnings to users which update to newer kernels very late.
On Tue, 6 Dec 2022, Johannes Weiner wrote: > Charge moving mode in cgroup1 allows memory to follow tasks as they > migrate between cgroups. This is, and always has been, a questionable > thing to do - for several reasons. > > First, it's expensive. Pages need to be identified, locked and > isolated from various MM operations, and reassigned, one by one. > > Second, it's unreliable. Once pages are charged to a cgroup, there > isn't always a clear owner task anymore. Cache isn't moved at all, for > example. Mapped memory is moved - but if trylocking or isolating a > page fails, it's arbitrarily left behind. Frequent moving between > domains may leave a task's memory scattered all over the place. > > Third, it isn't really needed. Launcher tasks can kick off workload > tasks directly in their target cgroup. Using dedicated per-workload > groups allows fine-grained policy adjustments - no need to move tasks > and their physical pages between control domains. The feature was > never forward-ported to cgroup2, and it hasn't been missed. > > Despite it being a niche usecase, the maintenance overhead of > supporting it is enormous. Because pages are moved while they are live > and subject to various MM operations, the synchronization rules are > complicated. There are lock_page_memcg() in MM and FS code, which > non-cgroup people don't understand. In some cases we've been able to > shift code and cgroup API calls around such that we can rely on native > locking as much as possible. But that's fragile, and sometimes we need > to hold MM locks for longer than we otherwise would (pte lock e.g.). > > Mark the feature deprecated. Hopefully we can remove it soon. > > Signed-off-by: Johannes Weiner <hannes@cmpxchg.org> Acked-by: Hugh Dickins <hughd@google.com> but I wonder if it would be helpful to mention move_charge_at_immigrate in the deprecation message: maybe the first line should be "Cgroup memory moving (move_charge_at_immigrate) is deprecated.\n" > --- > Documentation/admin-guide/cgroup-v1/memory.rst | 11 ++++++++++- > mm/memcontrol.c | 4 ++++ > 2 files changed, 14 insertions(+), 1 deletion(-) > > diff --git a/Documentation/admin-guide/cgroup-v1/memory.rst b/Documentation/admin-guide/cgroup-v1/memory.rst > index 60370f2c67b9..87d7877b98ec 100644 > --- a/Documentation/admin-guide/cgroup-v1/memory.rst > +++ b/Documentation/admin-guide/cgroup-v1/memory.rst > @@ -86,6 +86,8 @@ Brief summary of control files. > memory.swappiness set/show swappiness parameter of vmscan > (See sysctl's vm.swappiness) > memory.move_charge_at_immigrate set/show controls of moving charges > + This knob is deprecated and shouldn't be > + used. > memory.oom_control set/show oom controls. > memory.numa_stat show the number of memory usage per numa > node > @@ -717,9 +719,16 @@ Soft limits can be setup by using the following commands (in this example we > It is recommended to set the soft limit always below the hard limit, > otherwise the hard limit will take precedence. > > -8. Move charges at task migration > +8. Move charges at task migration (DEPRECATED!) > ================================= > > +THIS IS DEPRECATED! > + > +It's expensive and unreliable! It's better practice to launch workload > +tasks directly from inside their target cgroup. Use dedicated workload > +cgroups to allow fine-grained policy adjustments without having to > +move physical pages between control domains. > + > Users can move charges associated with a task along with task migration, that > is, uncharge task's pages from the old cgroup and charge them to the new cgroup. > This feature is not supported in !CONFIG_MMU environments because of lack of > diff --git a/mm/memcontrol.c b/mm/memcontrol.c > index b696354c1b21..e650a38d9a90 100644 > --- a/mm/memcontrol.c > +++ b/mm/memcontrol.c > @@ -3919,6 +3919,10 @@ static int mem_cgroup_move_charge_write(struct cgroup_subsys_state *css, > { > struct mem_cgroup *memcg = mem_cgroup_from_css(css); > > + pr_warn_once("Cgroup memory moving is deprecated. " > + "Please report your usecase to linux-mm@kvack.org if you " > + "depend on this functionality.\n"); > + > if (val & ~MOVE_MASK) > return -EINVAL; > > -- > 2.38.1 > >
On Tue, Dec 06, 2022 at 05:58:14PM -0800, Hugh Dickins wrote: > On Tue, 6 Dec 2022, Johannes Weiner wrote: > > > Charge moving mode in cgroup1 allows memory to follow tasks as they > > migrate between cgroups. This is, and always has been, a questionable > > thing to do - for several reasons. > > > > First, it's expensive. Pages need to be identified, locked and > > isolated from various MM operations, and reassigned, one by one. > > > > Second, it's unreliable. Once pages are charged to a cgroup, there > > isn't always a clear owner task anymore. Cache isn't moved at all, for > > example. Mapped memory is moved - but if trylocking or isolating a > > page fails, it's arbitrarily left behind. Frequent moving between > > domains may leave a task's memory scattered all over the place. > > > > Third, it isn't really needed. Launcher tasks can kick off workload > > tasks directly in their target cgroup. Using dedicated per-workload > > groups allows fine-grained policy adjustments - no need to move tasks > > and their physical pages between control domains. The feature was > > never forward-ported to cgroup2, and it hasn't been missed. > > > > Despite it being a niche usecase, the maintenance overhead of > > supporting it is enormous. Because pages are moved while they are live > > and subject to various MM operations, the synchronization rules are > > complicated. There are lock_page_memcg() in MM and FS code, which > > non-cgroup people don't understand. In some cases we've been able to > > shift code and cgroup API calls around such that we can rely on native > > locking as much as possible. But that's fragile, and sometimes we need > > to hold MM locks for longer than we otherwise would (pte lock e.g.). > > > > Mark the feature deprecated. Hopefully we can remove it soon. > > > > Signed-off-by: Johannes Weiner <hannes@cmpxchg.org> > > Acked-by: Hugh Dickins <hughd@google.com> Thanks > but I wonder if it would be helpful to mention move_charge_at_immigrate > in the deprecation message: maybe the first line should be > "Cgroup memory moving (move_charge_at_immigrate) is deprecated.\n" Fair enough! Here is the updated patch. --- From 0e791e6ab8ba2f75dd4205684c06bcc7308d9867 Mon Sep 17 00:00:00 2001 From: Johannes Weiner <hannes@cmpxchg.org> Date: Mon, 5 Dec 2022 19:57:06 +0100 Subject: [PATCH] mm: memcontrol: deprecate charge moving Charge moving mode in cgroup1 allows memory to follow tasks as they migrate between cgroups. This is, and always has been, a questionable thing to do - for several reasons. First, it's expensive. Pages need to be identified, locked and isolated from various MM operations, and reassigned, one by one. Second, it's unreliable. Once pages are charged to a cgroup, there isn't always a clear owner task anymore. Cache isn't moved at all, for example. Mapped memory is moved - but if trylocking or isolating a page fails, it's arbitrarily left behind. Frequent moving between domains may leave a task's memory scattered all over the place. Third, it isn't really needed. Launcher tasks can kick off workload tasks directly in their target cgroup. Using dedicated per-workload groups allows fine-grained policy adjustments - no need to move tasks and their physical pages between control domains. The feature was never forward-ported to cgroup2, and it hasn't been missed. Despite it being a niche usecase, the maintenance overhead of supporting it is enormous. Because pages are moved while they are live and subject to various MM operations, the synchronization rules are complicated. There are lock_page_memcg() in MM and FS code, which non-cgroup people don't understand. In some cases we've been able to shift code and cgroup API calls around such that we can rely on native locking as much as possible. But that's fragile, and sometimes we need to hold MM locks for longer than we otherwise would (pte lock e.g.). Mark the feature deprecated. Hopefully we can remove it soon. Signed-off-by: Johannes Weiner <hannes@cmpxchg.org> Acked-by: Shakeel Butt <shakeelb@google.com> Acked-by: Hugh Dickins <hughd@google.com> Cc: stable@vger.kernel.org --- Documentation/admin-guide/cgroup-v1/memory.rst | 11 ++++++++++- mm/memcontrol.c | 4 ++++ 2 files changed, 14 insertions(+), 1 deletion(-) diff --git a/Documentation/admin-guide/cgroup-v1/memory.rst b/Documentation/admin-guide/cgroup-v1/memory.rst index 60370f2c67b9..87d7877b98ec 100644 --- a/Documentation/admin-guide/cgroup-v1/memory.rst +++ b/Documentation/admin-guide/cgroup-v1/memory.rst @@ -86,6 +86,8 @@ Brief summary of control files. memory.swappiness set/show swappiness parameter of vmscan (See sysctl's vm.swappiness) memory.move_charge_at_immigrate set/show controls of moving charges + This knob is deprecated and shouldn't be + used. memory.oom_control set/show oom controls. memory.numa_stat show the number of memory usage per numa node @@ -717,9 +719,16 @@ Soft limits can be setup by using the following commands (in this example we It is recommended to set the soft limit always below the hard limit, otherwise the hard limit will take precedence. -8. Move charges at task migration +8. Move charges at task migration (DEPRECATED!) ================================= +THIS IS DEPRECATED! + +It's expensive and unreliable! It's better practice to launch workload +tasks directly from inside their target cgroup. Use dedicated workload +cgroups to allow fine-grained policy adjustments without having to +move physical pages between control domains. + Users can move charges associated with a task along with task migration, that is, uncharge task's pages from the old cgroup and charge them to the new cgroup. This feature is not supported in !CONFIG_MMU environments because of lack of diff --git a/mm/memcontrol.c b/mm/memcontrol.c index b696354c1b21..9c9a42153b76 100644 --- a/mm/memcontrol.c +++ b/mm/memcontrol.c @@ -3919,6 +3919,10 @@ static int mem_cgroup_move_charge_write(struct cgroup_subsys_state *css, { struct mem_cgroup *memcg = mem_cgroup_from_css(css); + pr_warn_once("Cgroup memory moving (move_charge_at_immigrate) is deprecated. " + "Please report your usecase to linux-mm@kvack.org if you " + "depend on this functionality.\n"); + if (val & ~MOVE_MASK) return -EINVAL;
On Tue, 6 Dec 2022 16:03:54 -0800 Shakeel Butt <shakeelb@google.com> wrote: > On Tue, Dec 6, 2022 at 9:14 AM Johannes Weiner <hannes@cmpxchg.org> wrote: > > > > Charge moving mode in cgroup1 allows memory to follow tasks as they > > migrate between cgroups. This is, and always has been, a questionable > > thing to do - for several reasons. > > > > First, it's expensive. Pages need to be identified, locked and > > isolated from various MM operations, and reassigned, one by one. > > > > Second, it's unreliable. Once pages are charged to a cgroup, there > > isn't always a clear owner task anymore. Cache isn't moved at all, for > > example. Mapped memory is moved - but if trylocking or isolating a > > page fails, it's arbitrarily left behind. Frequent moving between > > domains may leave a task's memory scattered all over the place. > > > > Third, it isn't really needed. Launcher tasks can kick off workload > > tasks directly in their target cgroup. Using dedicated per-workload > > groups allows fine-grained policy adjustments - no need to move tasks > > and their physical pages between control domains. The feature was > > never forward-ported to cgroup2, and it hasn't been missed. > > > > Despite it being a niche usecase, the maintenance overhead of > > supporting it is enormous. Because pages are moved while they are live > > and subject to various MM operations, the synchronization rules are > > complicated. There are lock_page_memcg() in MM and FS code, which > > non-cgroup people don't understand. In some cases we've been able to > > shift code and cgroup API calls around such that we can rely on native > > locking as much as possible. But that's fragile, and sometimes we need > > to hold MM locks for longer than we otherwise would (pte lock e.g.). > > > > Mark the feature deprecated. Hopefully we can remove it soon. > > > > Signed-off-by: Johannes Weiner <hannes@cmpxchg.org> > > Acked-by: Shakeel Butt <shakeelb@google.com> > > I would request this patch to be backported to stable kernels as well > for early warnings to users which update to newer kernels very late. Sounds reasonable, but the changelog should have a few words in it explaining why we're requesting the backport. I guess I can type those in. We're at -rc8 and I'm not planning on merging these up until after 6.2-rc1 is out. Please feel free to argue with me on that score.
On Wed, Dec 7, 2022 at 1:51 PM Andrew Morton <akpm@linux-foundation.org> wrote: > > On Tue, 6 Dec 2022 16:03:54 -0800 Shakeel Butt <shakeelb@google.com> wrote: > > > On Tue, Dec 6, 2022 at 9:14 AM Johannes Weiner <hannes@cmpxchg.org> wrote: > > > > > > Charge moving mode in cgroup1 allows memory to follow tasks as they > > > migrate between cgroups. This is, and always has been, a questionable > > > thing to do - for several reasons. > > > > > > First, it's expensive. Pages need to be identified, locked and > > > isolated from various MM operations, and reassigned, one by one. > > > > > > Second, it's unreliable. Once pages are charged to a cgroup, there > > > isn't always a clear owner task anymore. Cache isn't moved at all, for > > > example. Mapped memory is moved - but if trylocking or isolating a > > > page fails, it's arbitrarily left behind. Frequent moving between > > > domains may leave a task's memory scattered all over the place. > > > > > > Third, it isn't really needed. Launcher tasks can kick off workload > > > tasks directly in their target cgroup. Using dedicated per-workload > > > groups allows fine-grained policy adjustments - no need to move tasks > > > and their physical pages between control domains. The feature was > > > never forward-ported to cgroup2, and it hasn't been missed. > > > > > > Despite it being a niche usecase, the maintenance overhead of > > > supporting it is enormous. Because pages are moved while they are live > > > and subject to various MM operations, the synchronization rules are > > > complicated. There are lock_page_memcg() in MM and FS code, which > > > non-cgroup people don't understand. In some cases we've been able to > > > shift code and cgroup API calls around such that we can rely on native > > > locking as much as possible. But that's fragile, and sometimes we need > > > to hold MM locks for longer than we otherwise would (pte lock e.g.). > > > > > > Mark the feature deprecated. Hopefully we can remove it soon. > > > > > > Signed-off-by: Johannes Weiner <hannes@cmpxchg.org> > > > > Acked-by: Shakeel Butt <shakeelb@google.com> > > > > I would request this patch to be backported to stable kernels as well > > for early warnings to users which update to newer kernels very late. > > Sounds reasonable, but the changelog should have a few words in it > explaining why we're requesting the backport. I guess I can type those > in. Thanks a lot. > > We're at -rc8 and I'm not planning on merging these up until after > 6.2-rc1 is out. Please feel free to argue with me on that score. No, I totally agree with you. There is no such urgency in merging these and a couple of weeks delay is totally fine.
diff --git a/Documentation/admin-guide/cgroup-v1/memory.rst b/Documentation/admin-guide/cgroup-v1/memory.rst index 60370f2c67b9..87d7877b98ec 100644 --- a/Documentation/admin-guide/cgroup-v1/memory.rst +++ b/Documentation/admin-guide/cgroup-v1/memory.rst @@ -86,6 +86,8 @@ Brief summary of control files. memory.swappiness set/show swappiness parameter of vmscan (See sysctl's vm.swappiness) memory.move_charge_at_immigrate set/show controls of moving charges + This knob is deprecated and shouldn't be + used. memory.oom_control set/show oom controls. memory.numa_stat show the number of memory usage per numa node @@ -717,9 +719,16 @@ Soft limits can be setup by using the following commands (in this example we It is recommended to set the soft limit always below the hard limit, otherwise the hard limit will take precedence. -8. Move charges at task migration +8. Move charges at task migration (DEPRECATED!) ================================= +THIS IS DEPRECATED! + +It's expensive and unreliable! It's better practice to launch workload +tasks directly from inside their target cgroup. Use dedicated workload +cgroups to allow fine-grained policy adjustments without having to +move physical pages between control domains. + Users can move charges associated with a task along with task migration, that is, uncharge task's pages from the old cgroup and charge them to the new cgroup. This feature is not supported in !CONFIG_MMU environments because of lack of diff --git a/mm/memcontrol.c b/mm/memcontrol.c index b696354c1b21..e650a38d9a90 100644 --- a/mm/memcontrol.c +++ b/mm/memcontrol.c @@ -3919,6 +3919,10 @@ static int mem_cgroup_move_charge_write(struct cgroup_subsys_state *css, { struct mem_cgroup *memcg = mem_cgroup_from_css(css); + pr_warn_once("Cgroup memory moving is deprecated. " + "Please report your usecase to linux-mm@kvack.org if you " + "depend on this functionality.\n"); + if (val & ~MOVE_MASK) return -EINVAL;
Charge moving mode in cgroup1 allows memory to follow tasks as they migrate between cgroups. This is, and always has been, a questionable thing to do - for several reasons. First, it's expensive. Pages need to be identified, locked and isolated from various MM operations, and reassigned, one by one. Second, it's unreliable. Once pages are charged to a cgroup, there isn't always a clear owner task anymore. Cache isn't moved at all, for example. Mapped memory is moved - but if trylocking or isolating a page fails, it's arbitrarily left behind. Frequent moving between domains may leave a task's memory scattered all over the place. Third, it isn't really needed. Launcher tasks can kick off workload tasks directly in their target cgroup. Using dedicated per-workload groups allows fine-grained policy adjustments - no need to move tasks and their physical pages between control domains. The feature was never forward-ported to cgroup2, and it hasn't been missed. Despite it being a niche usecase, the maintenance overhead of supporting it is enormous. Because pages are moved while they are live and subject to various MM operations, the synchronization rules are complicated. There are lock_page_memcg() in MM and FS code, which non-cgroup people don't understand. In some cases we've been able to shift code and cgroup API calls around such that we can rely on native locking as much as possible. But that's fragile, and sometimes we need to hold MM locks for longer than we otherwise would (pte lock e.g.). Mark the feature deprecated. Hopefully we can remove it soon. Signed-off-by: Johannes Weiner <hannes@cmpxchg.org> --- Documentation/admin-guide/cgroup-v1/memory.rst | 11 ++++++++++- mm/memcontrol.c | 4 ++++ 2 files changed, 14 insertions(+), 1 deletion(-)