[PATCH] mm/hmm: remove redundant check non_swap_entry()

[PATCH] mm/hmm: remove redundant check non_swap_entry()

[Ralph Campbell]

In zap_pte_range(), the check for non_swap_entry() and
is_device_private_entry() is redundant since the latter is a subset of the
former. Remove the redundant check to simplify the code and for clarity.

Signed-off-by: Ralph Campbell <rcampbell@nvidia.com>
---

This is based on the current linux tree and is intended for Andrew's mm
tree. There is no rush so it could go into 5.9 but I think it is safe
enough to go into an rc after the patch is reviewed.

 mm/memory.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/mm/memory.c b/mm/memory.c
index dc7f3543b1fd..bdbb4f97e7d0 100644
--- a/mm/memory.c
+++ b/mm/memory.c
@@ -1098,7 +1098,7 @@ static unsigned long zap_pte_range(struct mmu_gather *tlb,
 		}
 
 		entry = pte_to_swp_entry(ptent);
-		if (non_swap_entry(entry) && is_device_private_entry(entry)) {
+		if (is_device_private_entry(entry)) {
 			struct page *page = device_private_entry_to_page(entry);
 
 			if (unlikely(details && details->check_mapping)) {
-- 
2.20.1

Re: [PATCH] mm/hmm: remove redundant check non_swap_entry()

[Jason Gunthorpe]

On Fri, Jun 12, 2020 at 12:26:18PM -0700, Ralph Campbell wrote:
> In zap_pte_range(), the check for non_swap_entry() and
> is_device_private_entry() is redundant since the latter is a subset of the
> former. Remove the redundant check to simplify the code and for clarity.
> 
> Signed-off-by: Ralph Campbell <rcampbell@nvidia.com>
> 
> This is based on the current linux tree and is intended for Andrew's mm
> tree. There is no rush so it could go into 5.9 but I think it is safe
> enough to go into an rc after the patch is reviewed.

Probably shouldn't mark it as mm/hmm if Andrew is going to pick it

Reviewed-by: Jason Gunthorpe <jgg@mellanox.com>

Jason

Re: [PATCH] mm/hmm: remove redundant check non_swap_entry()

[Matthew Wilcox]

On Fri, Jun 12, 2020 at 12:26:18PM -0700, Ralph Campbell wrote:
> In zap_pte_range(), the check for non_swap_entry() and
> is_device_private_entry() is redundant since the latter is a subset of the
> former. Remove the redundant check to simplify the code and for clarity.

That is highly configuration dependent.

#else /* CONFIG_DEVICE_PRIVATE */
...
static inline bool is_device_private_entry(swp_entry_t entry)
{
        return false;
}

...
#if defined(CONFIG_MEMORY_FAILURE) || defined(CONFIG_MIGRATION) || \
    defined(CONFIG_DEVICE_PRIVATE)
static inline int non_swap_entry(swp_entry_t entry)
{
        return swp_type(entry) >= MAX_SWAPFILES;
}
#else
static inline int non_swap_entry(swp_entry_t entry)
{
        return 0;
}
#endif

Re: [PATCH] mm/hmm: remove redundant check non_swap_entry()

[Jason Gunthorpe]

On Fri, Jun 12, 2020 at 12:35:24PM -0700, Matthew Wilcox wrote:
> On Fri, Jun 12, 2020 at 12:26:18PM -0700, Ralph Campbell wrote:
> > In zap_pte_range(), the check for non_swap_entry() and
> > is_device_private_entry() is redundant since the latter is a subset of the
> > former. Remove the redundant check to simplify the code and for clarity.
> 
> That is highly configuration dependent.
> 
> #else /* CONFIG_DEVICE_PRIVATE */
> ...
> static inline bool is_device_private_entry(swp_entry_t entry)
> {
>         return false;
> }

The commit message might be a bit confusing, as it is not a subset, I
would say that device_private_entry alone is sufficient to tell if the
entry is private or not.

For the !CONFIG_DEVICE_PRIVATE case having it wired to false is
right.

Jason

Re: [PATCH] mm/hmm: remove redundant check non_swap_entry()

[Ralph Campbell]

On 6/12/20 12:33 PM, Jason Gunthorpe wrote:
> On Fri, Jun 12, 2020 at 12:26:18PM -0700, Ralph Campbell wrote:
>> In zap_pte_range(), the check for non_swap_entry() and
>> is_device_private_entry() is redundant since the latter is a subset of the
>> former. Remove the redundant check to simplify the code and for clarity.
>>
>> Signed-off-by: Ralph Campbell <rcampbell@nvidia.com>
>>
>> This is based on the current linux tree and is intended for Andrew's mm
>> tree. There is no rush so it could go into 5.9 but I think it is safe
>> enough to go into an rc after the patch is reviewed.
> 
> Probably shouldn't mark it as mm/hmm if Andrew is going to pick it

OK, I'll mark it as just mm: if I need to repost.

> Reviewed-by: Jason Gunthorpe <jgg@mellanox.com>
> 
> Jason
> 

Thanks!

Re: [PATCH] mm/hmm: remove redundant check non_swap_entry()

[Ralph Campbell]

On 6/12/20 12:42 PM, Jason Gunthorpe wrote:
> On Fri, Jun 12, 2020 at 12:35:24PM -0700, Matthew Wilcox wrote:
>> On Fri, Jun 12, 2020 at 12:26:18PM -0700, Ralph Campbell wrote:
>>> In zap_pte_range(), the check for non_swap_entry() and
>>> is_device_private_entry() is redundant since the latter is a subset of the
>>> former. Remove the redundant check to simplify the code and for clarity.
>>
>> That is highly configuration dependent.
>>
>> #else /* CONFIG_DEVICE_PRIVATE */
>> ...
>> static inline bool is_device_private_entry(swp_entry_t entry)
>> {
>>          return false;
>> }
> 
> The commit message might be a bit confusing, as it is not a subset, I
> would say that device_private_entry alone is sufficient to tell if the
> entry is private or not.
> 
> For the !CONFIG_DEVICE_PRIVATE case having it wired to false is
> right.
> 
> Jason
> 

How about the following message instead?

In zap_pte_range(), the check for non_swap_entry() and
is_device_private_entry() is unnecessary since the latter is sufficient
to determine if the page is a device private page. Remove the test for
non_swap_entry() to simplify the code and for clarity.

Re: [PATCH] mm/hmm: remove redundant check non_swap_entry()

[Jason Gunthorpe]

On Fri, Jun 12, 2020 at 12:53:23PM -0700, Ralph Campbell wrote:
> 
> On 6/12/20 12:42 PM, Jason Gunthorpe wrote:
> > On Fri, Jun 12, 2020 at 12:35:24PM -0700, Matthew Wilcox wrote:
> > > On Fri, Jun 12, 2020 at 12:26:18PM -0700, Ralph Campbell wrote:
> > > > In zap_pte_range(), the check for non_swap_entry() and
> > > > is_device_private_entry() is redundant since the latter is a subset of the
> > > > former. Remove the redundant check to simplify the code and for clarity.
> > > 
> > > That is highly configuration dependent.
> > > 
> > > #else /* CONFIG_DEVICE_PRIVATE */
> > > ...
> > > static inline bool is_device_private_entry(swp_entry_t entry)
> > > {
> > >          return false;
> > > }
> > 
> > The commit message might be a bit confusing, as it is not a subset, I
> > would say that device_private_entry alone is sufficient to tell if the
> > entry is private or not.
> > 
> > For the !CONFIG_DEVICE_PRIVATE case having it wired to false is
> > right.
> > 
> > Jason
> > 
> 
> How about the following message instead?
> 
> In zap_pte_range(), the check for non_swap_entry() and
> is_device_private_entry() is unnecessary since the latter is sufficient
> to determine if the page is a device private page. Remove the test for
> non_swap_entry() to simplify the code and for clarity.

Yes, that is clearer to me

Jason