Re: [PATCH 1/2] delayacct: support re-entrance detection of thrashing accounting

From: Andrew Morton
Date: Sat Aug 13 2022 - 19:23:37 EST


On Sat, 13 Aug 2022 07:41:09 +0000 cgel.zte@xxxxxxxxx wrote:

> From: Yang Yang <yang.yang29@xxxxxxxxxx>
>
> Once upon a time, we only support accounting thrashing of page cache.
> Then Joonsoo introduced workingset detection for anonymous pages and
> we gained the ability to account thrashing of them[1].
>
> For page cache thrashing accounting, there is no suitable place to do
> it in fs level likes swap_readpage(). So we have to do it in
> folio_wait_bit_common().
>
> Then for anonymous pages thrashing accounting, we have to do it in
> both swap_readpage() and folio_wait_bit_common(). This likes PSI,
> so we should let thrashing accounting supports re-entrance detection.
>
> This patch is to prepare complete thrashing accounting, and is based
> on patch "filemap: make the accounting of thrashing more consistent".
>
> -static inline void delayacct_thrashing_start(void)
> +static inline void delayacct_thrashing_start(unsigned long *flags)

I don't think that `flags' is a very descriptive name. It might be
anything.

> --- a/include/linux/sched.h
> +++ b/include/linux/sched.h
> @@ -943,6 +943,10 @@ struct task_struct {
> #ifdef CONFIG_CPU_SUP_INTEL
> unsigned reported_split_lock:1;
> #endif
> +#ifdef CONFIG_TASK_DELAY_ACCT
> + /* delay due to memory thrashing */
> + unsigned in_thrashing:1;
> +#endif

OK, saving space in the task_struct is good.

> unsigned long atomic_flags; /* Flags requiring atomic access. */
>
> diff --git a/kernel/delayacct.c b/kernel/delayacct.c
> index 164ed9ef77a3..a5916196022f 100644
> --- a/kernel/delayacct.c
> +++ b/kernel/delayacct.c
> @@ -214,13 +214,22 @@ void __delayacct_freepages_end(void)
> &current->delays->freepages_count);
> }
>
> -void __delayacct_thrashing_start(void)
> +void __delayacct_thrashing_start(unsigned long *flags)
> {
> + *flags = current->in_thrashing;
> + if (*flags)
> + return;

Can't we rename `flags' to `in_thrashing' throughout?

And may as well give it type `bool'. And convert that bool to/from the
bitfield when moving it in and out of the task_struct with

*in_thrashing = !!current->in_thrashing;

current->in_thrashing = (in_thrashing != 0);