Re: [Xen-devel] [PATCHv2] frontswap: allow multiple backends

[Dan Streetman <ddstreet@xxxxxxxx>]

On Tue, Jun 2, 2015 at 5:06 PM, Andrew Morton <akpm@xxxxxxxxxxxxxxxxxxxx> wrote:
> On Mon,  1 Jun 2015 10:22:24 -0400 Dan Streetman <ddstreet@xxxxxxxx> 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.
>>
>> ...
>>
>> -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);
>
> umm, DECLARE_BITMAP() doesn't initialise the storage.  Either this
> patch wasn't tested very well or you should buy me a lottery ticket!

Doh!  I'll fix and resend.

I did test it, too, but zswap doesn't care if the swap device actually
exists, it just alloc's a tree for whatever it's told.  So likely it
was allocing some extra trees there :)

>
>>       }
>> -     /*
>> -      * We MUST have frontswap_ops set _after_ the frontswap_init's
>> -      * have been called. Otherwise __frontswap_store might fail. Hence
>> -      * the barrier to make sure compiler does not re-order us.
>> +     spin_unlock(&swap_lock);
>> +
>> +     /* the new ops needs to know the currently active swap devices */
>> +     for_each_set_bit(i, a, MAX_SWAPFILES)
>> +             ops->init(i);
>> +
>> +     /* setting frontswap_ops must happen after the ops->init() calls
>> +      * above; cmpxchg implies smp_mb() which will ensure the init is
>> +      * complete at this point
>> +      */
>
> Like this, please:
>
>         /*
>          * Setting ...
>
> and sentences start with capital letters ;)

okay, okay :-)

>
>

_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxx
http://lists.xen.org/xen-devel