| Message ID | 20220408045743.1432968-2-yosryahmed@google.com (mailing list archive) |
|---|---|
| State | New |
| Headers | show |
| Series | memcg: introduce per-memcg proactive reclaim | expand |
On Fri, Apr 08, 2022 at 04:57:40AM +0000, Yosry Ahmed wrote: > +static ssize_t memory_reclaim(struct kernfs_open_file *of, char *buf, > + size_t nbytes, loff_t off) > +{ > + struct mem_cgroup *memcg = mem_cgroup_from_css(of_css(of)); > + unsigned int nr_retries = MAX_RECLAIM_RETRIES; > + unsigned long nr_to_reclaim, nr_reclaimed = 0; > + int err; > + > + buf = strstrip(buf); > + err = page_counter_memparse(buf, "", &nr_to_reclaim); Is there a reason not to support "max"? Empty string seems odd to me here. > + if (err) > + return err; > + > + while (nr_reclaimed < nr_to_reclaim) { > + unsigned long reclaimed; > + > + if (signal_pending(current)) > + break; I think this should be `return -EINTR;` > + > + reclaimed = try_to_free_mem_cgroup_pages(memcg, > + nr_to_reclaim - nr_reclaimed, > + GFP_KERNEL, true); > + > + if (!reclaimed && !nr_retries--) > + break; Here you can just `return -EAGAIN;` > + > + nr_reclaimed += reclaimed; > + } > + > + return nr_reclaimed < nr_to_reclaim ? -EAGAIN : nbytes; Then this can just be `return nbytes;` I'm very much in favor of this new interface. Thanks for working on it!
On Fri 08-04-22 09:43:03, Dan Schatzberg wrote: > On Fri, Apr 08, 2022 at 04:57:40AM +0000, Yosry Ahmed wrote: > > +static ssize_t memory_reclaim(struct kernfs_open_file *of, char *buf, > > + size_t nbytes, loff_t off) > > +{ > > + struct mem_cgroup *memcg = mem_cgroup_from_css(of_css(of)); > > + unsigned int nr_retries = MAX_RECLAIM_RETRIES; > > + unsigned long nr_to_reclaim, nr_reclaimed = 0; > > + int err; > > + > > + buf = strstrip(buf); > > + err = page_counter_memparse(buf, "", &nr_to_reclaim); > > Is there a reason not to support "max"? Empty string seems odd to me > here. I have to say I have missed the special meaning of the empty string here and I agree this would indeed really weird. Does cgroup core even call here? cgroup_file_write seems to drop !nbytes input. Regarding "max" as a possible input. I am not really sure to be honest. I can imagine that it could be legit to simply reclaim all the charges (e.g. before removing the memcg) which should be achieveable by reclaiming the reported consumption. Or what exactly should be the semantic?
On Fri, Apr 08, 2022 at 04:11:05PM +0200, Michal Hocko wrote: > Regarding "max" as a possible input. I am not really sure to be honest. > I can imagine that it could be legit to simply reclaim all the charges > (e.g. before removing the memcg) which should be achieveable by > reclaiming the reported consumption. Or what exactly should be the > semantic? Yeah, it just allows you to avoid reading memory.current to just reclaim everything if you can specify "max" - you're still protected by nretries to eventually bail out. Mostly, though I just feel like supporting "max" makes memory.reclaim semetric with a lot of the cgroup memory control files which tend to support "max".
On Fri, Apr 8, 2022 at 6:43 AM Dan Schatzberg <schatzberg.dan@gmail.com> wrote: > > On Fri, Apr 08, 2022 at 04:57:40AM +0000, Yosry Ahmed wrote: > > +static ssize_t memory_reclaim(struct kernfs_open_file *of, char *buf, > > + size_t nbytes, loff_t off) > > +{ > > + struct mem_cgroup *memcg = mem_cgroup_from_css(of_css(of)); > > + unsigned int nr_retries = MAX_RECLAIM_RETRIES; > > + unsigned long nr_to_reclaim, nr_reclaimed = 0; > > + int err; > > + > > + buf = strstrip(buf); > > + err = page_counter_memparse(buf, "", &nr_to_reclaim); > > Is there a reason not to support "max"? Empty string seems odd to me > here. We can certainly support "max" to reclaim as much as we can with MAX_RECLAIM_RETRIES, if there are no objections from the maintainers. > > > + if (err) > > + return err; > > + > > + while (nr_reclaimed < nr_to_reclaim) { > > + unsigned long reclaimed; > > + > > + if (signal_pending(current)) > > + break; > > I think this should be `return -EINTR;` Yes this makes more sense. I think this was modeled after the if block in memory_high_write(), but maybe it makes sense there to just report success as the new high limit was set anyway. Will change it in the next version. > > > + > > + reclaimed = try_to_free_mem_cgroup_pages(memcg, > > + nr_to_reclaim - nr_reclaimed, > > + GFP_KERNEL, true); > > + > > + if (!reclaimed && !nr_retries--) > > + break; > > Here you can just `return -EAGAIN;` Will do. > > > + > > + nr_reclaimed += reclaimed; > > + } > > + > > + return nr_reclaimed < nr_to_reclaim ? -EAGAIN : nbytes; > > Then this can just be `return nbytes;` Will do. > > I'm very much in favor of this new interface. Thanks for working on > it! Thanks so much for reviewing it!
On Fri, Apr 8, 2022 at 7:55 AM Dan Schatzberg <schatzberg.dan@gmail.com> wrote: > > On Fri, Apr 08, 2022 at 04:11:05PM +0200, Michal Hocko wrote: > > Regarding "max" as a possible input. I am not really sure to be honest. > > I can imagine that it could be legit to simply reclaim all the charges > > (e.g. before removing the memcg) which should be achieveable by > > reclaiming the reported consumption. Or what exactly should be the > > semantic? > > Yeah, it just allows you to avoid reading memory.current to just > reclaim everything if you can specify "max" - you're still protected > by nretries to eventually bail out. Mostly, though I just feel like > supporting "max" makes memory.reclaim semetric with a lot of the > cgroup memory control files which tend to support "max". One possible approach here is to have force_empty behavior when we write "max" to memory.reclaim. From Google's perspective we don't have a preference, but it seems to me like logical behavior. We can do this either by directly calling mem_cgroup_force_empty() or just draining stock and lrus in memory_reclaim(). This actually brings up another interesting point. Do you think we should drain lrus if try_to_free_mem_cgroup_pages() fails to reclaim the request amount? We can do this after the first call or before the last one. It could introduce more evictable pages for try_to_free_mem_cgroup_pages() to free.
On Fri, Apr 08, 2022 at 04:57:40AM +0000, Yosry Ahmed wrote: > From: Shakeel Butt <shakeelb@google.com> > > Introduce a memcg interface to trigger memory reclaim on a memory cgroup. > > Use case: Proactive Reclaim > --------------------------- > > A userspace proactive reclaimer can continuously probe the memcg to > reclaim a small amount of memory. This gives more accurate and > up-to-date workingset estimation as the LRUs are continuously > sorted and can potentially provide more deterministic memory > overcommit behavior. The memory overcommit controller can provide > more proactive response to the changing behavior of the running > applications instead of being reactive. > > A userspace reclaimer's purpose in this case is not a complete replacement > for kswapd or direct reclaim, it is to proactively identify memory savings > opportunities and reclaim some amount of cold pages set by the policy > to free up the memory for more demanding jobs or scheduling new jobs. > > A user space proactive reclaimer is used in Google data centers. > Additionally, Meta's TMO paper recently referenced a very similar > interface used for user space proactive reclaim: > https://dl.acm.org/doi/pdf/10.1145/3503222.3507731 > > Benefits of a user space reclaimer: > ----------------------------------- > > 1) More flexible on who should be charged for the cpu of the memory > reclaim. For proactive reclaim, it makes more sense to be centralized. > > 2) More flexible on dedicating the resources (like cpu). The memory > overcommit controller can balance the cost between the cpu usage and > the memory reclaimed. > > 3) Provides a way to the applications to keep their LRUs sorted, so, > under memory pressure better reclaim candidates are selected. This also > gives more accurate and uptodate notion of working set for an > application. > > Why memory.high is not enough? > ------------------------------ > > - memory.high can be used to trigger reclaim in a memcg and can > potentially be used for proactive reclaim. > However there is a big downside in using memory.high. It can potentially > introduce high reclaim stalls in the target application as the > allocations from the processes or the threads of the application can hit > the temporary memory.high limit. > > - Userspace proactive reclaimers usually use feedback loops to decide > how much memory to proactively reclaim from a workload. The metrics > used for this are usually either refaults or PSI, and these metrics > will become messy if the application gets throttled by hitting the > high limit. > > - memory.high is a stateful interface, if the userspace proactive > reclaimer crashes for any reason while triggering reclaim it can leave > the application in a bad state. > > - If a workload is rapidly expanding, setting memory.high to proactively > reclaim memory can result in actually reclaiming more memory than > intended. > > The benefits of such interface and shortcomings of existing interface > were further discussed in this RFC thread: > https://lore.kernel.org/linux-mm/5df21376-7dd1-bf81-8414-32a73cea45dd@google.com/ > > Interface: > ---------- > > Introducing a very simple memcg interface 'echo 10M > memory.reclaim' to > trigger reclaim in the target memory cgroup. > > The interface is introduced as a nested-keyed file to allow for future > optional arguments to be easily added to configure the behavior of > reclaim. > > Possible Extensions: > -------------------- > > - This interface can be extended with an additional parameter or flags > to allow specifying one or more types of memory to reclaim from (e.g. > file, anon, ..). > > - The interface can also be extended with a node mask to reclaim from > specific nodes. This has use cases for reclaim-based demotion in memory > tiering systens. > > - A similar per-node interface can also be added to support proactive > reclaim and reclaim-based demotion in systems without memcg. > > - Add a timeout parameter to make it easier for user space to call the > interface without worrying about being blocked for an undefined amount > of time. > > For now, let's keep things simple by adding the basic functionality. > > [yosryahmed@google.com: refreshed to current master, updated commit > message based on recent discussions and use cases] > Signed-off-by: Shakeel Butt <shakeelb@google.com> > Signed-off-by: Yosry Ahmed <yosryahmed@google.com> > Acked-by: Johannes Weiner <hannes@cmpxchg.org> > Acked-by: Michal Hocko <mhocko@suse.com> > Acked-by: Wei Xu <weixugc@google.com> Acked-by: Roman Gushchin <roman.gushchin@linux.dev> Thank you for incorporating all the feedback! Nice work!
On Fri 08-04-22 10:55:56, Dan Schatzberg wrote: > On Fri, Apr 08, 2022 at 04:11:05PM +0200, Michal Hocko wrote: > > Regarding "max" as a possible input. I am not really sure to be honest. > > I can imagine that it could be legit to simply reclaim all the charges > > (e.g. before removing the memcg) which should be achieveable by > > reclaiming the reported consumption. Or what exactly should be the > > semantic? > > Yeah, it just allows you to avoid reading memory.current to just > reclaim everything if you can specify "max" The same could be achieved by requesting a really high number (-1Ul) > - you're still protected > by nretries to eventually bail out. The number of retries is an implementation detail and nobody should really rely on that. Bail out on signal can be still used so yeah getting a large input or whatever alias of that should be just fine. > Mostly, though I just feel like > supporting "max" makes memory.reclaim semetric with a lot of the > cgroup memory control files which tend to support "max". max is used for limits now and this doesn't have a semantic of one. But I have to say I do not really feel strongly about this.
On Fri, Apr 8, 2022 at 1:08 PM Yosry Ahmed <yosryahmed@google.com> wrote: > > On Fri, Apr 8, 2022 at 7:55 AM Dan Schatzberg <schatzberg.dan@gmail.com> wrote: > > > > On Fri, Apr 08, 2022 at 04:11:05PM +0200, Michal Hocko wrote: > > > Regarding "max" as a possible input. I am not really sure to be honest. > > > I can imagine that it could be legit to simply reclaim all the charges > > > (e.g. before removing the memcg) which should be achieveable by > > > reclaiming the reported consumption. Or what exactly should be the > > > semantic? > > > > Yeah, it just allows you to avoid reading memory.current to just > > reclaim everything if you can specify "max" - you're still protected > > by nretries to eventually bail out. Mostly, though I just feel like > > supporting "max" makes memory.reclaim semetric with a lot of the > > cgroup memory control files which tend to support "max". > > One possible approach here is to have force_empty behavior when we > write "max" to memory.reclaim. From Google's perspective we don't have > a preference, but it seems to me like logical behavior. We can do this > either by directly calling mem_cgroup_force_empty() or just draining > stock and lrus in memory_reclaim(). > > This actually brings up another interesting point. Do you think we > should drain lrus if try_to_free_mem_cgroup_pages() fails to reclaim > the request amount? We can do this after the first call or before the > last one. It could introduce more evictable pages for > try_to_free_mem_cgroup_pages() to free. Hey Michal, any thoughts on this? I am looking for feedback on this before I send out v4.
On Thu 14-04-22 10:25:29, Yosry Ahmed wrote: > On Fri, Apr 8, 2022 at 1:08 PM Yosry Ahmed <yosryahmed@google.com> wrote: > > > > On Fri, Apr 8, 2022 at 7:55 AM Dan Schatzberg <schatzberg.dan@gmail.com> wrote: > > > > > > On Fri, Apr 08, 2022 at 04:11:05PM +0200, Michal Hocko wrote: > > > > Regarding "max" as a possible input. I am not really sure to be honest. > > > > I can imagine that it could be legit to simply reclaim all the charges > > > > (e.g. before removing the memcg) which should be achieveable by > > > > reclaiming the reported consumption. Or what exactly should be the > > > > semantic? > > > > > > Yeah, it just allows you to avoid reading memory.current to just > > > reclaim everything if you can specify "max" - you're still protected > > > by nretries to eventually bail out. Mostly, though I just feel like > > > supporting "max" makes memory.reclaim semetric with a lot of the > > > cgroup memory control files which tend to support "max". > > > > One possible approach here is to have force_empty behavior when we > > write "max" to memory.reclaim. From Google's perspective we don't have > > a preference, but it seems to me like logical behavior. We can do this > > either by directly calling mem_cgroup_force_empty() or just draining > > stock and lrus in memory_reclaim(). > > > > This actually brings up another interesting point. Do you think we > > should drain lrus if try_to_free_mem_cgroup_pages() fails to reclaim > > the request amount? We can do this after the first call or before the > > last one. It could introduce more evictable pages for > > try_to_free_mem_cgroup_pages() to free. > > Hey Michal, any thoughts on this? I am looking for feedback on this > before I send out v4. As I've already said I do not have strong preferences for the "max" value to be accepted by the interface. Maybe you can add a support for max in a separate patch so the discussion would not delay the rest of the work. For the LRU draining I do not see any problem for that to be added. The overhead of the operation would increase, especially on larger machines, which could be a concern. So the real question is whether not doing so is a big problem. Our force_empty implementation optimistically drains pcp caches but please note that this is not really guranteeing anything as charges can happen at any time.
diff --git a/Documentation/admin-guide/cgroup-v2.rst b/Documentation/admin-guide/cgroup-v2.rst index 69d7a6983f78..19bcd73cad03 100644 --- a/Documentation/admin-guide/cgroup-v2.rst +++ b/Documentation/admin-guide/cgroup-v2.rst @@ -1208,6 +1208,27 @@ PAGE_SIZE multiple when read back. high limit is used and monitored properly, this limit's utility is limited to providing the final safety net. + memory.reclaim + A write-only nested-keyed file which exists for all cgroups. + + This is a simple interface to trigger memory reclaim in the + target cgroup. + + This file accepts a single key, the number of bytes to reclaim. + No nested keys are currently supported. + + Example:: + + echo "1G" > memory.reclaim + + The interface can be later extended with nested keys to + configure the reclaim behavior. For example, specify the + type of memory to reclaim from (anon, file, ..). + + Please note that the kernel can over or under reclaim from + the target cgroup. If less bytes are reclaimed than the + specified amount, -EAGAIN is returned. + memory.oom.group A read-write single value file which exists on non-root cgroups. The default value is "0". diff --git a/mm/memcontrol.c b/mm/memcontrol.c index 725f76723220..2b214b66d333 100644 --- a/mm/memcontrol.c +++ b/mm/memcontrol.c @@ -6355,6 +6355,38 @@ static ssize_t memory_oom_group_write(struct kernfs_open_file *of, return nbytes; } +static ssize_t memory_reclaim(struct kernfs_open_file *of, char *buf, + size_t nbytes, loff_t off) +{ + struct mem_cgroup *memcg = mem_cgroup_from_css(of_css(of)); + unsigned int nr_retries = MAX_RECLAIM_RETRIES; + unsigned long nr_to_reclaim, nr_reclaimed = 0; + int err; + + buf = strstrip(buf); + err = page_counter_memparse(buf, "", &nr_to_reclaim); + if (err) + return err; + + while (nr_reclaimed < nr_to_reclaim) { + unsigned long reclaimed; + + if (signal_pending(current)) + break; + + reclaimed = try_to_free_mem_cgroup_pages(memcg, + nr_to_reclaim - nr_reclaimed, + GFP_KERNEL, true); + + if (!reclaimed && !nr_retries--) + break; + + nr_reclaimed += reclaimed; + } + + return nr_reclaimed < nr_to_reclaim ? -EAGAIN : nbytes; +} + static struct cftype memory_files[] = { { .name = "current", @@ -6413,6 +6445,11 @@ static struct cftype memory_files[] = { .seq_show = memory_oom_group_show, .write = memory_oom_group_write, }, + { + .name = "reclaim", + .flags = CFTYPE_NS_DELEGATABLE, + .write = memory_reclaim, + }, { } /* terminate */ };