On 01/12/2018 03:21 AM, Andrew Morton wrote: > On Thu, 11 Jan 2018 14:59:23 +0300 Andrey Ryabinin <aryabinin@xxxxxxxxxxxxx> wrote: > >> On 01/11/2018 01:31 AM, Andrew Morton wrote: >>> On Wed, 10 Jan 2018 15:43:17 +0300 Andrey Ryabinin <aryabinin@xxxxxxxxxxxxx> wrote: >>> >>>> mem_cgroup_resize_[memsw]_limit() tries to free only 32 (SWAP_CLUSTER_MAX) >>>> pages on each iteration. This makes practically impossible to decrease >>>> limit of memory cgroup. Tasks could easily allocate back 32 pages, >>>> so we can't reduce memory usage, and once retry_count reaches zero we return >>>> -EBUSY. >>>> >>>> Easy to reproduce the problem by running the following commands: >>>> >>>> mkdir /sys/fs/cgroup/memory/test >>>> echo $$ >> /sys/fs/cgroup/memory/test/tasks >>>> cat big_file > /dev/null & >>>> sleep 1 && echo $((100*1024*1024)) > /sys/fs/cgroup/memory/test/memory.limit_in_bytes >>>> -bash: echo: write error: Device or resource busy >>>> >>>> Instead of relying on retry_count, keep retrying the reclaim until >>>> the desired limit is reached or fail if the reclaim doesn't make >>>> any progress or a signal is pending. >>>> >>> >>> Is there any situation under which that mem_cgroup_resize_limit() can >>> get stuck semi-indefinitely in a livelockish state? It isn't very >>> obvious that we're protected from this, so perhaps it would help to >>> have a comment which describes how loop termination is assured? >>> >> >> We are not protected from this. If tasks in cgroup *indefinitely* generate reclaimable memory at high rate >> and user asks to set unreachable limit, like 'echo 4096 > memory.limit_in_bytes', than >> try_to_free_mem_cgroup_pages() will return non-zero indefinitely. >> >> Is that a big deal? At least loop can be interrupted by a signal, and we don't hold any locks here. > > It may be better to detect this condition, give up and return an error? > That's basically what how v1 worked, "if (curusage >= oldusage)" used to be the way to detect this potential livelock. So we can just go back to it? -- To unsubscribe from this list: send the line "unsubscribe cgroups" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
