Re: [PATCH v3 04/10] mm: frontswap: split out __frontswap_unuse_pages

[Konrad Rzeszutek Wilk]

On Mon, Jun 11, 2012 at 10:27 AM, Dan Magenheimer
<dan.magenheimer@xxxxxxxxxx> wrote:
>> From: Sasha Levin [mailto:levinsasha928@xxxxxxxxx]
>> Subject: Re: [PATCH v3 04/10] mm: frontswap: split out __frontswap_unuse_pages
>>
>> > > + assert_spin_locked(&swap_lock);
>> >
>> > Normally, we should use this assertion when we can't find swap_lock is hold or not easily
>> > by complicated call depth or unexpected use-case like general function.
>> > But I expect this function's caller is very limited, not complicated.
>> > Just comment write down isn't enough?
>>
>> Is there a reason not to do it though? Debugging a case where this
>> function is called without a swaplock and causes corruption won't be
>> easy.
>
> I'm not sure of the correct kernel style but I like the fact
> that assert_spin_locked both documents the lock requirement and tests
> it at runtime.

The kernel style is to do "
3) Separate your changes.

Separate _logical changes_ into a single patch file.
"

So it is fine, but it should be in its own patch.
>
> I don't know the correct kernel syntax but is it possible
> to make this code be functional when the kernel "debug"
> option is on, but a no-op when "debug" is disabled?
> IMHO, that would be the ideal solution.
>
>> > > + for (type = swap_list.head; type >= 0; type = si->next) {
>> > > +         si = swap_info[type];
>> > > +         si_frontswap_pages = atomic_read(&si->frontswap_pages);
>> > > +         if (total_pages_to_unuse < si_frontswap_pages) {
>> > > +                 pages = pages_to_unuse = total_pages_to_unuse;
>> > > +         } else {
>> > > +                 pages = si_frontswap_pages;
>> > > +                 pages_to_unuse = 0; /* unuse all */
>> > > +         }
>> > > +         /* ensure there is enough RAM to fetch pages from frontswap */
>> > > +         if (security_vm_enough_memory_mm(current->mm, pages)) {
>> > > +                 ret = -ENOMEM;
>> >
>> >
>> > Nipick:
>> > I am not sure detailed error returning would be good.
>> > Caller doesn't matter it now but it can consider it in future.
>> > Hmm,
>>
>> Is there a reason to avoid returning a meaningful error when it's pretty
>> easy?
>
> I'm certainly not an expert on kernel style (as this whole series
> of patches demonstrates :-) but I think setting a meaningful
> error code is useful documentation and plans for future users
> that might use the error code.

Aye.
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/