2018-07-14 06:27:47

by Davidlohr Bueso

Subject: [PATCH] lib/rhashtable: consider param->min_size when setting initial table size

rhashtable_init() currently does not take into account the user-passed
min_size parameter unless param->nelem_hint is set as well. As such,
the default size (number of buckets) will always be HASH_DEFAULT_SIZE
even if the smallest allowed size is larger than that. Remediate this
by unconditionally calling into rounded_hashtable_size() and handling
things accordingly.

Signed-off-by: Davidlohr Bueso <[email protected]>
---
lib/rhashtable.c | 17 +++++++++++------
1 file changed, 11 insertions(+), 6 deletions(-)

diff --git a/lib/rhashtable.c b/lib/rhashtable.c
index 411c4041ce83..89c3cfc8334d 100644
--- a/lib/rhashtable.c
+++ b/lib/rhashtable.c
@@ -962,8 +962,16 @@ EXPORT_SYMBOL_GPL(rhashtable_walk_stop);

static size_t rounded_hashtable_size(const struct rhashtable_params *params)
{
- return max(roundup_pow_of_two(params->nelem_hint * 4 / 3),
- (unsigned long)params->min_size);
+ size_t retsize;
+
+ if (params->nelem_hint)
+ retsize = max(roundup_pow_of_two(params->nelem_hint * 4 / 3),
+ (unsigned long)params->min_size);
+ else
+ retsize = max(HASH_DEFAULT_SIZE,
+ (unsigned long)params->min_size);
+
+ return retsize;
}

static u32 rhashtable_jhash2(const void *key, u32 length, u32 seed)
@@ -1020,8 +1028,6 @@ int rhashtable_init(struct rhashtable *ht,
struct bucket_table *tbl;
size_t size;

- size = HASH_DEFAULT_SIZE;
-
if ((!params->key_len && !params->obj_hashfn) ||
(params->obj_hashfn && !params->obj_cmpfn))
return -EINVAL;
@@ -1048,8 +1054,7 @@ int rhashtable_init(struct rhashtable *ht,

ht->p.min_size = max_t(u16, ht->p.min_size, HASH_MIN_SIZE);

- if (params->nelem_hint)
- size = rounded_hashtable_size(&ht->p);
+ size = rounded_hashtable_size(&ht->p);

if (params->locks_mul)
ht->p.locks_mul = roundup_pow_of_two(params->locks_mul);
--
2.16.4

2018-07-16 00:01:44

by Herbert Xu

Subject: Re: [PATCH] lib/rhashtable: consider param->min_size when setting initial table size

On Fri, Jul 13, 2018 at 11:25:16PM -0700, Davidlohr Bueso wrote:
> rhashtable_init() currently does not take into account the user-passed
> min_size parameter unless param->nelem_hint is set as well. As such,
> the default size (number of buckets) will always be HASH_DEFAULT_SIZE
> even if the smallest allowed size is larger than that. Remediate this
> by unconditionally calling into rounded_hashtable_size() and handling
> things accordingly.
>
> Signed-off-by: Davidlohr Bueso <[email protected]>

Acked-by: Herbert Xu <[email protected]>

Normally these patches go through netdev so could you please resend
it with my ack to [email protected]?

Thanks,
--
Email: Herbert Xu <[email protected]>
Home Page: http://gondor.apana.org.au/~herbert/
PGP Key: http://gondor.apana.org.au/~herbert/pubkey.txt

2018-07-16 20:27:12

by Davidlohr Bueso

Subject: [PATCH resend -next] lib/rhashtable: consider param->min_size when setting initial table size

rhashtable_init() currently does not take into account the user-passed
min_size parameter unless param->nelem_hint is set as well. As such,
the default size (number of buckets) will always be HASH_DEFAULT_SIZE
even if the smallest allowed size is larger than that. Remediate this
by unconditionally calling into rounded_hashtable_size() and handling
things accordingly.

Signed-off-by: Davidlohr Bueso <[email protected]>
Acked-by: Herbert Xu <[email protected]>
---
Resending per Herbert's request with Cc to netdev.

lib/rhashtable.c | 17 +++++++++++------
1 file changed, 11 insertions(+), 6 deletions(-)

diff --git a/lib/rhashtable.c b/lib/rhashtable.c
index 62d3a976614e..310e29b51507 100644
--- a/lib/rhashtable.c
+++ b/lib/rhashtable.c
@@ -953,8 +953,16 @@ EXPORT_SYMBOL_GPL(rhashtable_walk_stop);

static size_t rounded_hashtable_size(const struct rhashtable_params *params)
{
- return max(roundup_pow_of_two(params->nelem_hint * 4 / 3),
- (unsigned long)params->min_size);
+ size_t retsize;
+
+ if (params->nelem_hint)
+ retsize = max(roundup_pow_of_two(params->nelem_hint * 4 / 3),
+ (unsigned long)params->min_size);
+ else
+ retsize = max(HASH_DEFAULT_SIZE,
+ (unsigned long)params->min_size);
+
+ return retsize;
}

static u32 rhashtable_jhash2(const void *key, u32 length, u32 seed)
@@ -1010,8 +1018,6 @@ int rhashtable_init(struct rhashtable *ht,
struct bucket_table *tbl;
size_t size;

- size = HASH_DEFAULT_SIZE;
-
if ((!params->key_len && !params->obj_hashfn) ||
(params->obj_hashfn && !params->obj_cmpfn))
return -EINVAL;
@@ -1035,8 +1041,7 @@ int rhashtable_init(struct rhashtable *ht,

ht->p.min_size = max_t(u16, ht->p.min_size, HASH_MIN_SIZE);

- if (params->nelem_hint)
- size = rounded_hashtable_size(&ht->p);
+ size = rounded_hashtable_size(&ht->p);

if (params->locks_mul)
ht->p.locks_mul = roundup_pow_of_two(params->locks_mul);
--
2.16.4

2018-07-17 22:31:53

by Davidlohr Bueso

Subject: Re: [PATCH] lib/rhashtable: consider param->min_size when setting initial table size

On Mon, 16 Jul 2018, Herbert Xu wrote:

>On Fri, Jul 13, 2018 at 11:25:16PM -0700, Davidlohr Bueso wrote:
>> rhashtable_init() currently does not take into account the user-passed
>> min_size parameter unless param->nelem_hint is set as well. As such,
>> the default size (number of buckets) will always be HASH_DEFAULT_SIZE
>> even if the smallest allowed size is larger than that. Remediate this
>> by unconditionally calling into rounded_hashtable_size() and handling
>> things accordingly.
>>
>> Signed-off-by: Davidlohr Bueso <[email protected]>
>
>Acked-by: Herbert Xu <[email protected]>
>
>Normally these patches go through netdev so could you please resend
>it with my ack to [email protected]?

So I've done the resend, but at least would think that routing the
patch through Andrew would work best as he picked up the rhashtable
changes regarding ipc and this touches the same call.

Thanks,
Davidlohr

2018-07-18 01:59:14

by Andrew Morton

Subject: Re: [PATCH] lib/rhashtable: consider param->min_size when setting initial table size

On Tue, 17 Jul 2018 15:30:57 -0700 Davidlohr Bueso <[email protected]> wrote:

> On Mon, 16 Jul 2018, Herbert Xu wrote:
>
> >On Fri, Jul 13, 2018 at 11:25:16PM -0700, Davidlohr Bueso wrote:
> >> rhashtable_init() currently does not take into account the user-passed
> >> min_size parameter unless param->nelem_hint is set as well. As such,
> >> the default size (number of buckets) will always be HASH_DEFAULT_SIZE
> >> even if the smallest allowed size is larger than that. Remediate this
> >> by unconditionally calling into rounded_hashtable_size() and handling
> >> things accordingly.
> >>
> >> Signed-off-by: Davidlohr Bueso <[email protected]>
> >
> >Acked-by: Herbert Xu <[email protected]>
> >
> >Normally these patches go through netdev so could you please resend
> >it with my ack to [email protected]?
>
> So I've done the resend, but at least would think that routing the
> patch through Andrew would work best as he picked up the rhashtable
> changes regarding ipc and this touches the same call.
>

Either approach works.

Reviewed-by: Andrew Morton <[email protected]>

2018-07-18 20:29:06

by David Miller

Subject: Re: [PATCH resend -next] lib/rhashtable: consider param->min_size when setting initial table size

From: Davidlohr Bueso <[email protected]>
Date: Mon, 16 Jul 2018 13:26:13 -0700

> rhashtable_init() currently does not take into account the user-passed
> min_size parameter unless param->nelem_hint is set as well. As such,
> the default size (number of buckets) will always be HASH_DEFAULT_SIZE
> even if the smallest allowed size is larger than that. Remediate this
> by unconditionally calling into rounded_hashtable_size() and handling
> things accordingly.
>
> Signed-off-by: Davidlohr Bueso <[email protected]>
> Acked-by: Herbert Xu <[email protected]>
> ---
> Resending per Herbert's request with Cc to netdev.

Applied and queued up for -stable, thanks.