[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <dda53b5a-3298-500f-45c5-f5d123559e8e@bytedance.com>
Date: Tue, 15 Aug 2023 10:28:21 +0800
From: Chuyi Zhou <zhouchuyi@...edance.com>
To: Jonathan Corbet <corbet@....net>, hannes@...xchg.org,
mhocko@...nel.org, roman.gushchin@...ux.dev, ast@...nel.org,
daniel@...earbox.net, andrii@...nel.org, muchun.song@...ux.dev
Cc: bpf@...r.kernel.org, linux-kernel@...r.kernel.org,
wuyun.abel@...edance.com, robin.lu@...edance.com
Subject: Re: [RFC PATCH v2 2/5] mm: Add policy_name to identify OOM policies
Hello,
在 2023/8/15 04:51, Jonathan Corbet 写道:
> Chuyi Zhou <zhouchuyi@...edance.com> writes:
>
>> This patch adds a new metadata policy_name in oom_control and report it
>> in dump_header(), so we can know what has been the selection policy. In
>> BPF program, we can call kfunc set_oom_policy_name to set the current
>> user-defined policy name. The in-kernel policy_name is "default".
>>
>> Signed-off-by: Chuyi Zhou <zhouchuyi@...edance.com>
>> ---
>> include/linux/oom.h | 7 +++++++
>> mm/oom_kill.c | 42 +++++++++++++++++++++++++++++++++++++++---
>> 2 files changed, 46 insertions(+), 3 deletions(-)
>
> So I have a possibly dumb question here...
>
>> diff --git a/include/linux/oom.h b/include/linux/oom.h
>> index 7d0c9c48a0c5..69d0f2ec6ea6 100644
>> --- a/include/linux/oom.h
>> +++ b/include/linux/oom.h
>> @@ -22,6 +22,10 @@ enum oom_constraint {
>> CONSTRAINT_MEMCG,
>> };
>>
>> +enum {
>> + POLICY_NAME_LEN = 16,
>> +};
>
> We've defined our name length, fine...
>
>> /*
>> * Details of the page allocation that triggered the oom killer that are used to
>> * determine what should be killed.
>> @@ -52,6 +56,9 @@ struct oom_control {
>>
>> /* Used to print the constraint info. */
>> enum oom_constraint constraint;
>> +
>> + /* Used to report the policy info. */
>> + char policy_name[POLICY_NAME_LEN];
>> };
>
> ...that is the length of the array, appended to the structure...
>
>> extern struct mutex oom_lock;
>> diff --git a/mm/oom_kill.c b/mm/oom_kill.c
>> index 255c9ef1d808..3239dcdba4d7 100644
>> --- a/mm/oom_kill.c
>> +++ b/mm/oom_kill.c
>> @@ -443,6 +443,35 @@ static int dump_task(struct task_struct *p, void *arg)
>> return 0;
>> }
>>
>> +__bpf_kfunc void set_oom_policy_name(struct oom_control *oc, const char *src, size_t sz)
>> +{
>> + memset(oc->policy_name, 0, sizeof(oc->policy_name));
>> +
>> + if (sz > POLICY_NAME_LEN)
>> + sz = POLICY_NAME_LEN;
>> +
>> + memcpy(oc->policy_name, src, sz);
>> +}
>
> This truncates the name, possibly leaving it without a terminating NUL
> character, right?
>
Yes, indeed. We should guarantee that the policy_name is terminated with
a NULL character to avoid some undefined behavior.
Thanks for your helpful review.
------
Chuyi Zhou
Powered by blists - more mailing lists