Re: [PATCH mptcp-next v7 07/11] mptcp: pm: initialize and release mptcp_pm_ops

[Matthieu Baerts <matttbe@kernel.org>]

Hi Geliang,

On 03/03/2025 05:22, Geliang Tang wrote:
> From: Geliang Tang <tanggeliang@kylinos.cn>
> 
> Add a struct mptcp_pm_ops pointer "ops" in struct mptcp_pm_data, and two
> functions mptcp_pm_initialize() and mptcp_pm_release(), to set and release
> this pointer. mptcp_pm_initialize() is invoked in mptcp_pm_data_reset(),
> while mptcp_pm_release() is invoked in mptcp_pm_destroy().
> 
> Signed-off-by: Geliang Tang <tanggeliang@kylinos.cn>
> ---
>  net/mptcp/pm.c       | 42 +++++++++++++++++++++++++++++++++++++++---
>  net/mptcp/protocol.h |  3 +++
>  2 files changed, 42 insertions(+), 3 deletions(-)
> 
> diff --git a/net/mptcp/pm.c b/net/mptcp/pm.c
> index 98f81221786f..e8b34f2ecb35 100644
> --- a/net/mptcp/pm.c
> +++ b/net/mptcp/pm.c
> @@ -973,15 +973,15 @@ void mptcp_pm_worker(struct mptcp_sock *msk)
>  void mptcp_pm_destroy(struct mptcp_sock *msk)
>  {
>  	mptcp_pm_free_anno_list(msk);
> -
> -	if (mptcp_pm_is_userspace(msk))
> -		mptcp_userspace_pm_free_local_addr_list(msk);
> +	mptcp_pm_release(msk);
>  }
>  
>  void mptcp_pm_data_reset(struct mptcp_sock *msk)
>  {
> +	const char *path_manager = mptcp_get_path_manager(sock_net((struct sock *)msk));
>  	u8 pm_type = mptcp_get_pm_type(sock_net((struct sock *)msk));
>  	struct mptcp_pm_data *pm = &msk->pm;
> +	int ret;
>  
>  	pm->add_addr_signaled = 0;
>  	pm->add_addr_accepted = 0;
> @@ -991,6 +991,12 @@ void mptcp_pm_data_reset(struct mptcp_sock *msk)
>  	pm->rm_list_rx.nr = 0;
>  	WRITE_ONCE(pm->pm_type, pm_type);
>  
> +	rcu_read_lock();
> +	ret = mptcp_pm_initialize(msk, mptcp_pm_find(path_manager));
> +	rcu_read_unlock();
> +	if (ret)
> +		return;

Mmh, that's annoying if pm->ops has not been set here! I don't think
mptcp_pm_initialize() can fail and if it does, it should stop the
connection.

> +
>  	if (pm_type == MPTCP_PM_TYPE_KERNEL) {>  		bool subflows_allowed = !!mptcp_pm_get_subflows_max(msk);
>  
> @@ -1102,3 +1108,33 @@ void mptcp_pm_get_available(char *buf, size_t maxlen)
>  	}
>  	rcu_read_unlock();
>  }
> +
> +int mptcp_pm_initialize(struct mptcp_sock *msk, struct mptcp_pm_ops *pm)
> +{
> +	if (!pm)

I guess this should never happen, right? Or maybe yes if a BPF PM has
been unloaded?

> +		pm = &mptcp_kernel_pm;
> +
> +	if (!bpf_try_module_get(pm, pm->owner))
> +		return -EBUSY;

Should it not fallback to the kernel PM + print a warning (once?)?

pr_warn_once("%pm %s couldn't be initialized, falling back to 'kernel'",
             pm->name);
pm = &mptcp_kernel_pm;

> +
> +	msk->pm.ops = pm;
> +	if (msk->pm.ops->init)
> +		msk->pm.ops->init(msk);
> +
> +	pr_debug("pm %s initialized\n", pm->name);
> +	return 0;
> +}
> +
> +void mptcp_pm_release(struct mptcp_sock *msk)
> +{
> +	struct mptcp_pm_ops *pm = msk->pm.ops;
> +
> +	if (!pm)
> +		return;
> +
> +	msk->pm.ops = NULL;
> +	if (pm->release)
> +		pm->release(msk);
> +
> +	bpf_module_put(pm, pm->owner);
> +}

Can you not declare mptcp_pm_release() and mptcp_pm_initialize() as
static and move them above mptcp_pm_destroy() and mptcp_pm_data_reset()?
It would make more sense, and no need to export them in protocol.h.

Cheers,
Matt
-- 
Sponsored by the NGI0 Core fund.