[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <aLfhyancfP5Na4AN@krava>
Date: Wed, 3 Sep 2025 08:35:53 +0200
From: Jiri Olsa <olsajiri@...il.com>
To: Masami Hiramatsu <mhiramat@...nel.org>
Cc: Oleg Nesterov <oleg@...hat.com>, Peter Zijlstra <peterz@...radead.org>,
Andrii Nakryiko <andrii@...nel.org>, bpf@...r.kernel.org,
linux-kernel@...r.kernel.org, linux-trace-kernel@...r.kernel.org,
x86@...nel.org, Song Liu <songliubraving@...com>,
Yonghong Song <yhs@...com>,
John Fastabend <john.fastabend@...il.com>,
Hao Luo <haoluo@...gle.com>, Steven Rostedt <rostedt@...dmis.org>,
Ingo Molnar <mingo@...nel.org>
Subject: Re: [PATCH perf/core 01/11] uprobes: Add unique flag to uprobe
consumer
On Wed, Sep 03, 2025 at 12:11:33AM +0900, Masami Hiramatsu wrote:
> On Tue, 2 Sep 2025 16:34:54 +0200
> Jiri Olsa <jolsa@...nel.org> wrote:
>
> > Adding unique flag to uprobe consumer to ensure it's the only consumer
> > attached on the uprobe.
> >
> > This is helpful for use cases when consumer wants to change user space
> > registers, which might confuse other consumers. With this change we can
> > ensure there's only one consumer on specific uprobe.
>
> nit: Does this mean one callback (consumer) is exclusively attached?
> If so, "exclusive" will be better wording?
yes, exclusive is better, will change
thanks,
jirka
>
> The logic looks good to me.
>
> Thanks,
>
> >
> > Signed-off-by: Jiri Olsa <jolsa@...nel.org>
> > ---
> > include/linux/uprobes.h | 1 +
> > kernel/events/uprobes.c | 30 ++++++++++++++++++++++++++++--
> > 2 files changed, 29 insertions(+), 2 deletions(-)
> >
> > diff --git a/include/linux/uprobes.h b/include/linux/uprobes.h
> > index 08ef78439d0d..0df849dee720 100644
> > --- a/include/linux/uprobes.h
> > +++ b/include/linux/uprobes.h
> > @@ -60,6 +60,7 @@ struct uprobe_consumer {
> > struct list_head cons_node;
> >
> > __u64 id; /* set when uprobe_consumer is registered */
> > + bool is_unique; /* the only consumer on uprobe */
> > };
> >
> > #ifdef CONFIG_UPROBES
> > diff --git a/kernel/events/uprobes.c b/kernel/events/uprobes.c
> > index 996a81080d56..b9b088f7333a 100644
> > --- a/kernel/events/uprobes.c
> > +++ b/kernel/events/uprobes.c
> > @@ -1024,14 +1024,35 @@ static struct uprobe *alloc_uprobe(struct inode *inode, loff_t offset,
> > return uprobe;
> > }
> >
> > -static void consumer_add(struct uprobe *uprobe, struct uprobe_consumer *uc)
> > +static bool consumer_can_add(struct list_head *head, struct uprobe_consumer *uc)
> > +{
> > + /* Uprobe has no consumer, we can add any. */
> > + if (list_empty(head))
> > + return true;
> > + /* Uprobe has consumer/s, we can't add unique one. */
> > + if (uc->is_unique)
> > + return false;
> > + /*
> > + * Uprobe has consumer/s, we can add nother consumer only if the
> > + * current consumer is not unique.
> > + **/
> > + return !list_first_entry(head, struct uprobe_consumer, cons_node)->is_unique;
> > +}
> > +
> > +static int consumer_add(struct uprobe *uprobe, struct uprobe_consumer *uc)
> > {
> > static atomic64_t id;
> > + int ret = -EBUSY;
> >
> > down_write(&uprobe->consumer_rwsem);
> > + if (!consumer_can_add(&uprobe->consumers, uc))
> > + goto unlock;
> > list_add_rcu(&uc->cons_node, &uprobe->consumers);
> > uc->id = (__u64) atomic64_inc_return(&id);
> > + ret = 0;
> > +unlock:
> > up_write(&uprobe->consumer_rwsem);
> > + return ret;
> > }
> >
> > /*
> > @@ -1420,7 +1441,12 @@ struct uprobe *uprobe_register(struct inode *inode,
> > return uprobe;
> >
> > down_write(&uprobe->register_rwsem);
> > - consumer_add(uprobe, uc);
> > + ret = consumer_add(uprobe, uc);
> > + if (ret) {
> > + put_uprobe(uprobe);
> > + up_write(&uprobe->register_rwsem);
> > + return ERR_PTR(ret);
> > + }
> > ret = register_for_each_vma(uprobe, uc);
> > up_write(&uprobe->register_rwsem);
> >
> > --
> > 2.51.0
> >
>
>
> --
> Masami Hiramatsu (Google) <mhiramat@...nel.org>
Powered by blists - more mailing lists