Re: [Linux-ATM-General] [PATCH] atm: he: Replace GFP_ATOMIC with GFP_KERNEL in he_open

[Chas W. <3c...@gm...>]

On Fri, Jan 26, 2018 at 3:23 AM, Jia-Ju Bai <bai...@gm...> wrote:

> After checking all possible call chains to he_open() here,
> my tool finds that he_open() is never called in atomic context.
> And this function is assigned to a function pointer "dev->ops->open",
> which is only called by __vcc_connect() (net/atm/common.c)
> through dev->ops->send(), and __vcc_connect() is only called by
> vcc_connect(), which calls mutex_lock(),
> so it indicates that he_open() can call functions which may sleep.
> Thus GFP_ATOMIC is not necessary, and it can be replaced with GFP_KERNEL.
>
> This is found by a static analysis tool named DCNS written by myself.
>
> Signed-off-by: Jia-Ju Bai <bai...@gm...>
>

Acked-by: Chas Williams <3c...@gm...>

At one point, there was talk of doing open/close in the bottom half of the
kernel.  Since this only happens
in user space as of now, this doesn't need to be ATOMIC.


> ---
>  drivers/atm/he.c |    2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/drivers/atm/he.c b/drivers/atm/he.c
> index e58538c..fea5bf0 100644
> --- a/drivers/atm/he.c
> +++ b/drivers/atm/he.c
> @@ -2135,7 +2135,7 @@ static int he_start(struct atm_dev *dev)
>
>         cid = he_mkcid(he_dev, vpi, vci);
>
> -       he_vcc = kmalloc(sizeof(struct he_vcc), GFP_ATOMIC);
> +       he_vcc = kmalloc(sizeof(struct he_vcc), GFP_KERNEL);
>         if (he_vcc == NULL) {
>                 hprintk("unable to allocate he_vcc during open\n");
>                 return -ENOMEM;
> --
> 1.7.9.5
>
>