Re: [PATCH 1/5] proc: use down_read_killable for /proc/pid/maps

[Michal Hocko]

On Wed 15-05-19 11:41:12, Konstantin Khlebnikov wrote:
> Do not stuck forever if something wrong.
> This function also used for /proc/pid/smaps.

I do agree that the killable variant is better but I do not understand
the changelog. What would keep the lock blocked for ever? I do not think
we have writer lock held for unbound amount of time anywhere, do we?

> Signed-off-by: Konstantin Khlebnikov <khlebnikov@xxxxxxxxxxxxxx>

Other than that
Acked-by: Michal Hocko <mhocko@xxxxxxxx>

> ---
>  fs/proc/task_mmu.c   |    6 +++++-
>  fs/proc/task_nommu.c |    6 +++++-
>  2 files changed, 10 insertions(+), 2 deletions(-)
> 
> diff --git a/fs/proc/task_mmu.c b/fs/proc/task_mmu.c
> index 01d4eb0e6bd1..2bf210229daf 100644
> --- a/fs/proc/task_mmu.c
> +++ b/fs/proc/task_mmu.c
> @@ -166,7 +166,11 @@ static void *m_start(struct seq_file *m, loff_t *ppos)
>  	if (!mm || !mmget_not_zero(mm))
>  		return NULL;
>  
> -	down_read(&mm->mmap_sem);
> +	if (down_read_killable(&mm->mmap_sem)) {
> +		mmput(mm);
> +		return ERR_PTR(-EINTR);
> +	}
> +
>  	hold_task_mempolicy(priv);
>  	priv->tail_vma = get_gate_vma(mm);
>  
> diff --git a/fs/proc/task_nommu.c b/fs/proc/task_nommu.c
> index 36bf0f2e102e..7907e6419e57 100644
> --- a/fs/proc/task_nommu.c
> +++ b/fs/proc/task_nommu.c
> @@ -211,7 +211,11 @@ static void *m_start(struct seq_file *m, loff_t *pos)
>  	if (!mm || !mmget_not_zero(mm))
>  		return NULL;
>  
> -	down_read(&mm->mmap_sem);
> +	if (down_read_killable(&mm->mmap_sem)) {
> +		mmput(mm);
> +		return ERR_PTR(-EINTR);
> +	}
> +
>  	/* start from the Nth VMA */
>  	for (p = rb_first(&mm->mm_rb); p; p = rb_next(p))
>  		if (n-- == 0)

-- 
Michal Hocko
SUSE Labs