Re: [External] Re: [PATCH V3 2/3] mm: add max swappiness arg to lru_gen for anonymous memory only

[Zhongkun He]

On Thu, Apr 10, 2025 at 10:10 AM Andrew Morton
<akpm@xxxxxxxxxxxxxxxxxxxx> wrote:
>
> On Wed,  9 Apr 2025 15:06:19 +0800 Zhongkun He <hezhongkun.hzk@xxxxxxxxxxxxx> wrote:
>
> > The MGLRU
>
> paging yuzhao?

I have cc yuzhao and look forward to the relay.

>
> > already supports reclaiming only from anonymous memory
> > via the /sys/kernel/debug/lru_gen interface. Now, memory.reclaim
> > also supports the swappiness=max parameter to enable reclaiming
> > solely from anonymous memory. To unify the semantics of proactive
> > reclaiming from anonymous folios, the max parameter is introduced.
> >
> > Additionally, the use of SWAPPINESS_ANON_ONLY in place of
> > 'MAX_SWAPPINESS + 1' improves code clarity and makes the intention
> > more explicit.
> >
> > --- a/mm/vmscan.c
> > +++ b/mm/vmscan.c
> > @@ -2697,8 +2697,11 @@ static bool should_clear_pmd_young(void)
> >               READ_ONCE((lruvec)->lrugen.min_seq[LRU_GEN_FILE]),      \
> >       }
> >
> > +#define max_evictable_type(swappiness)                                               \
> > +     ((swappiness) != SWAPPINESS_ANON_ONLY)
> > +
> >  #define evictable_min_seq(min_seq, swappiness)                               \
> > -     min((min_seq)[!(swappiness)], (min_seq)[(swappiness) <= MAX_SWAPPINESS])
> > +     min((min_seq)[!(swappiness)], (min_seq)[max_evictable_type(swappiness)])
>
> Why oh why did we implement these in cpp?

Just want to make the code more clear.  Maybe we should do more like this

/* The range of swappiness is [0,1-200,201], 0 means file type only;
 * 1-200 anon and file type; 201 anon type only
 */
#define max_type(swappiness) ((swappiness) != SWAPPINESS_ANON_ONLY)
#define min_type(swappiness) !(swappiness)

#define evictable_min_seq(min_seq, swappiness)              \
    min((min_seq)[min_type(swappiness)], (min_seq)[max_type(swappiness)])

 #define for_each_evictable_type(type, swappiness)                      \
-       for ((type) = !(swappiness); (type) <= ((swappiness) <=
MAX_SWAPPINESS); (type)++)
+       for ((type) = min_type(swappiness) ; (type) <=
max_type(swappiness); (type)++)

>
> >
> > @@ -3857,7 +3860,7 @@ static bool inc_min_seq(struct lruvec *lruvec, int type, int swappiness)
> >       int hist = lru_hist_from_seq(lrugen->min_seq[type]);
> >       int new_gen, old_gen = lru_gen_from_seq(lrugen->min_seq[type]);
> >
> > -     if (type ? swappiness > MAX_SWAPPINESS : !swappiness)
> > +     if (type ? (swappiness == SWAPPINESS_ANON_ONLY) : !swappiness)
>
> This expression makes my brain bleed.
>
>         if (type) {
>                 if (swappiness == SWAPPINESS_ANON_ONLY) {
>                         /*
>                          * Nice comment explaining why we're doing this
>                          */
>                         goto done;;
>                 }
>         } else {
>                 if (!swappiness) {
>                         /*
>                          * Nice comment explaining why we're doing this
>                          */
>                         goto done;
>                 }
>         }
>
> or
>
>         if (type && (swappiness == SWAPPINESS_ANON_ONLY)) {
>                 /*
>                  * Nice comment explaining why we're doing this
>                  */
>                 goto done;
>         }
>
>         if (!type && !swappiness) {
>                 /*
>                  * Nice comment explaining why we're doing this
>                  */
>                 goto done;
>         }
>
> It's much more verbose, but it has the huge advantage that it creates
> locations where we can add comments which tell readers what's going on.
> Which is pretty important, no?
>

Yes, I agree. Will do, thanks.

> >               goto done;
> >
> >       /* prevent cold/hot inversion if the type is evictable */
> > @@ -5523,7 +5526,7 @@ static int run_cmd(char cmd, int memcg_id, int nid, unsigned long seq,
> >
> >       if (swappiness < MIN_SWAPPINESS)
> >               swappiness = get_swappiness(lruvec, sc);
> > -     else if (swappiness > MAX_SWAPPINESS + 1)
> > +     else if (swappiness > SWAPPINESS_ANON_ONLY)
> >               goto done;
> >
> >       switch (cmd) {
> > @@ -5580,7 +5583,7 @@ static ssize_t lru_gen_seq_write(struct file *file, const char __user *src,
> >       while ((cur = strsep(&next, ",;\n"))) {
> >               int n;
> >               int end;
> > -             char cmd;
> > +             char cmd, swap_string[5];
> >               unsigned int memcg_id;
> >               unsigned int nid;
> >               unsigned long seq;
> > @@ -5591,13 +5594,22 @@ static ssize_t lru_gen_seq_write(struct file *file, const char __user *src,
> >               if (!*cur)
> >                       continue;
> >
> > -             n = sscanf(cur, "%c %u %u %lu %n %u %n %lu %n", &cmd, &memcg_id, &nid,
> > -                        &seq, &end, &swappiness, &end, &opt, &end);
> > +             n = sscanf(cur, "%c %u %u %lu %n %4s %n %lu %n", &cmd, &memcg_id, &nid,
> > +                        &seq, &end, swap_string, &end, &opt, &end);
>
> Permits userspace to easily overrun swap_string[].  OK, it's root-only,
> but still, why permit this?
>

IICC, the arg in sscanf is %4s meaning the length of the string will
not be allowed to overrun
the swap_string, thanks.

> >               if (n < 4 || cur[end]) {
> >                       err = -EINVAL;
> >                       break;
> >               }
> >
> > +             /* set by userspace for anonymous memory only */
> > +             if (!strncmp("max", swap_string, sizeof("max"))) {
>
> Can sscanf() give us a non null-terminated string?
>

No, the sscanf will add '\0' to the end, so the maximum number of
input characters is 4,
and the length of swap_string is 5, with one character reserved for
the null terminator '\0'
for sscanf.

Thanks for your time.

> > +                     swappiness = SWAPPINESS_ANON_ONLY;
> > +             } else {
> > +                     err = kstrtouint(swap_string, 0, &swappiness);
> > +                     if (err)
> > +                             break;
> > +             }
> > +
> >               err = run_cmd(cmd, memcg_id, nid, seq, &sc, swappiness, opt);
> >               if (err)
> >                       break;
>