Re: [PATCH] frontswap: allow multiple backends

[Dan Streetman <ddstreet@ieee.org>]

On Fri, May 29, 2015 at 6:07 PM, Andrew Morton
<akpm@linux-foundation.org> wrote:
> On Thu, 28 May 2015 16:28:37 -0400 Dan Streetman <ddstreet@ieee.org> wrote:
>
>> Change frontswap single pointer to a singly linked list of frontswap
>> implementations.  Update Xen tmem implementation as register no longer
>> returns anything.
>>
>> Frontswap only keeps track of a single implementation; any implementation
>> that registers second (or later) will replace the previously registered
>> implementation, and gets a pointer to the previous implementation that
>> the new implementation is expected to pass all frontswap functions to
>> if it can't handle the function itself.  However that method doesn't
>> really make much sense, as passing that work on to every implementation
>> adds unnecessary work to implementations; instead, frontswap should
>> simply keep a list of all registered implementations and try each
>> implementation for any function.  Most importantly, neither of the
>> two currently existing frontswap implementations in the kernel actually
>> do anything with any previous frontswap implementation that they
>> replace when registering.
>>
>> This allows frontswap to successfully manage multiple implementations
>> by keeping a list of them all.
>
> Looks OK to me.  The "you can never deregister" thing makes life
> simpler.
>
> But we need to have a fight over style issues.  Just because you *can*
> do something doesn't mean you should.  Don't make you poor readers sit
> there going crosseyed at elaborate `for' statements.  Try to keep the
> code as simple and straightforward as possible.
>
>> ...
>>
>>  /*
>> - * Register operations for frontswap, returning previous thus allowing
>> - * detection of multiple backends and possible nesting.
>> + * Register operations for frontswap
>>   */
>> -struct frontswap_ops *frontswap_register_ops(struct frontswap_ops *ops)
>> +void frontswap_register_ops(struct frontswap_ops *ops)
>>  {
>> -     struct frontswap_ops *old = frontswap_ops;
>> -     int i;
>> -
>> -     for (i = 0; i < MAX_SWAPFILES; i++) {
>> -             if (test_and_clear_bit(i, need_init)) {
>> -                     struct swap_info_struct *sis = swap_info[i];
>> -                     /* __frontswap_init _should_ have set it! */
>> -                     if (!sis->frontswap_map)
>> -                             return ERR_PTR(-EINVAL);
>> -                     ops->init(i);
>> +     DECLARE_BITMAP(a, MAX_SWAPFILES);
>> +     DECLARE_BITMAP(b, MAX_SWAPFILES);
>> +     struct swap_info_struct *si;
>> +     unsigned int i;
>> +
>> +     spin_lock(&swap_lock);
>> +     plist_for_each_entry(si, &swap_active_head, list) {
>> +             if (!WARN_ON(!si->frontswap_map))
>> +                     set_bit(si->type, a);
>> +     }
>> +     spin_unlock(&swap_lock);
>> +
>> +     for (i = find_first_bit(a, MAX_SWAPFILES);
>> +          i < MAX_SWAPFILES;
>> +          i = find_next_bit(a, MAX_SWAPFILES, i + 1))
>> +             ops->init(i);
>
>         i = find_first_bit(a, MAX_SWAPFILES);
>         while (i < MAX_SWAPFILES) {
>                 ops->init(i);
>                 i = find_next_bit(a, MAX_SWAPFILES, i + 1);
>         }

i was looking for a for_each_set_bit(), but didn't find it, until now
when i looked in bitops.h (why is that in bitops instead of bitmap?).

Anyway, this can be replaced with a simple for_each_set_bit() which
should be much clearer.

>
>> +     do {
>> +             ops->next = frontswap_ops;
>> +     } while (cmpxchg(&frontswap_ops, ops->next, ops) != ops->next);
>> +
>> +     spin_lock(&swap_lock);
>> +     plist_for_each_entry(si, &swap_active_head, list) {
>> +             if (si->frontswap_map)
>> +                     set_bit(si->type, b);
>> +     }
>> +     spin_unlock(&swap_lock);
>> +
>> +     if (!bitmap_equal(a, b, MAX_SWAPFILES)) {
>> +             for (i = 0; i < MAX_SWAPFILES; i++) {
>> +                     if (!test_bit(i, a) && test_bit(i, b))
>> +                             ops->init(i);
>> +                     else if (test_bit(i, a) && !test_bit(i, b))
>> +                             ops->invalidate_area(i);
>>               }
>> ...
>>
>> @@ -215,24 +216,25 @@ static inline void __frontswap_clear(struct swap_info_struct *sis,
>>   */
>>  int __frontswap_store(struct page *page)
>>  {
>> -     int ret = -1, dup = 0;
>> +     int ret, dup;
>>       swp_entry_t entry = { .val = page_private(page), };
>>       int type = swp_type(entry);
>>       struct swap_info_struct *sis = swap_info[type];
>>       pgoff_t offset = swp_offset(entry);
>> +     struct frontswap_ops *ops;
>>
>>       /*
>>        * Return if no backend registed.
>>        * Don't need to inc frontswap_failed_stores here.
>>        */
>>       if (!frontswap_ops)
>> -             return ret;
>> +             return -1;
>>
>>       BUG_ON(!PageLocked(page));
>>       BUG_ON(sis == NULL);
>> -     if (__frontswap_test(sis, offset))
>> -             dup = 1;
>> -     ret = frontswap_ops->store(type, offset, page);
>> +     dup = __frontswap_test(sis, offset);
>> +     for (ops = frontswap_ops, ret = -1; ops && ret; ops = ops->next)
>> +             ret = ops->store(type, offset, page);
>
>         ret = -1;
>         for (ops = frontswap_ops; ops; ops = ops->next) {
>                 ret = ops->store(type, offset, page);
>                 if (!ret)
>                         break;
>         }
>
> One advantage of doing it this way is that it leaves room for comments.
> And this code would benefit from a comment above the "if (!ret)".
> What's going on here?  What could cause ->store to return zero and is
> this an error?  We should explain this somewhere;  `struct
> frontswap_ops' is cheerily undocumented, so where?

you're right of course, i think i was sleep-deprived when i wrote that
for loop.  I forgot the rule about what's most important: clarity not
brevity (or, clarity not cleverity...).  ;-)

I'll add some comments, both to the frontswap_ops struct as well as in
functions like here for clarification.

>
>
> Is the `ret = -1' really needed?  Can this function ever be called if
> there aren't any registered frontswap_ops?

in this case, the ret = -1 initialization was only needed because of
the ops && ret check (which would fail if ret started at 0).  Putting
the ret check into the for loop both makes it clearer and removes the
need to initialize ret.

The (!frontswap_ops) above the for loop will ensure the for loop is
always called with at least 1 ops.

>
>
> Also, __frontswap_store() disturbs me:
>
> : int __frontswap_store(struct page *page)
> : {
> :       int ret, dup;
> :       swp_entry_t entry = { .val = page_private(page), };
> :       int type = swp_type(entry);
> :       struct swap_info_struct *sis = swap_info[type];
> :       pgoff_t offset = swp_offset(entry);
> :       struct frontswap_ops *ops;
> :
> :       /*
> :        * Return if no backend registed.
> :        * Don't need to inc frontswap_failed_stores here.
> :        */
> :       if (!frontswap_ops)
> :               return -1;
> :
> :       BUG_ON(!PageLocked(page));
> :       BUG_ON(sis == NULL);
> :       dup = __frontswap_test(sis, offset);
> :       ret = -1;
> :       for (ops = frontswap_ops; ops; ops = ops->next) {
> :               ret = ops->store(type, offset, page);
> :               if (!ret)
> :                       break;
> :       }
>
> Here we've just iterated through all the registered operations.
>
> :       if (ret == 0) {
> :               set_bit(offset, sis->frontswap_map);
> :               inc_frontswap_succ_stores();
> :               if (!dup)
> :                       atomic_inc(&sis->frontswap_pages);
> :       } else {
> :               /*
> :                 failed dup always results in automatic invalidate of
> :                 the (older) page from frontswap
> :                */
> :               inc_frontswap_failed_stores();
> :               if (dup) {
> :                       __frontswap_clear(sis, offset);
> :                       frontswap_ops->invalidate_page(type, offset);
>
> But here we call ->invalidate_page on just one of teh registered
> operations.  Seems wrong.

oops, yep it is.  I missed updating this to invalidate all the
registered ops.  Will fix.

>
> Maybe some careful code commentary would clear this up.
>
> :               }
> :       }
> :       if (frontswap_writethrough_enabled)
> :               /* report failure so swap also writes to swap device */
> :               ret = -1;
> :       return ret;
> : }
>
> Please review:

yep, plus the additional comments and changes i mentioned above.  I'll
try re-sending the patch with updates.

Thanks!


>
> --- a/mm/frontswap.c~frontswap-allow-multiple-backends-fix
> +++ a/mm/frontswap.c
> @@ -97,7 +97,7 @@ static inline void inc_frontswap_invalid
>   *
>   * Obviously the opposite (unloading the backend) must be done after all
>   * the frontswap_[store|load|invalidate_area|invalidate_page] start
> - * ignorning or failing the requests.  However, there is currently no way
> + * ignoring or failing the requests.  However, there is currently no way
>   * to unload a backend once it is registered.
>   */
>
> @@ -118,10 +118,11 @@ void frontswap_register_ops(struct front
>         }
>         spin_unlock(&swap_lock);
>
> -       for (i = find_first_bit(a, MAX_SWAPFILES);
> -            i < MAX_SWAPFILES;
> -            i = find_next_bit(a, MAX_SWAPFILES, i + 1))
> +       i = find_first_bit(a, MAX_SWAPFILES);
> +       while (i < MAX_SWAPFILES) {
>                 ops->init(i);
> +               i = find_next_bit(a, MAX_SWAPFILES, i + 1);
> +       }
>
>         do {
>                 ops->next = frontswap_ops;
> @@ -233,8 +234,12 @@ int __frontswap_store(struct page *page)
>         BUG_ON(!PageLocked(page));
>         BUG_ON(sis == NULL);
>         dup = __frontswap_test(sis, offset);
> -       for (ops = frontswap_ops, ret = -1; ops && ret; ops = ops->next)
> +       ret = -1;
> +       for (ops = frontswap_ops; ops; ops = ops->next) {
>                 ret = ops->store(type, offset, page);
> +               if (!ret)
> +                       break;
> +       }
>         if (ret == 0) {
>                 set_bit(offset, sis->frontswap_map);
>                 inc_frontswap_succ_stores();
> @@ -279,8 +284,12 @@ int __frontswap_load(struct page *page)
>         BUG_ON(sis == NULL);
>         if (!__frontswap_test(sis, offset))
>                 return -1;
> -       for (ops = frontswap_ops, ret = -1; ops && ret; ops = ops->next)
> +       ret = -1;
> +       for (ops = frontswap_ops; ops; ops = ops->next) {
>                 ret = ops->load(type, offset, page);
> +               if (!ret)
> +                       break;
> +       }
>         if (ret == 0) {
>                 inc_frontswap_loads();
>                 if (frontswap_tmem_exclusive_gets_enabled) {
> _
>

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>