From: Honnappa Nagarahalli <Honnappa.Nagarahalli@arm.com>
To: "Abdullah Ömer Yamaç" <aomeryamac@gmail.com>
Cc: "Medvedkin, Vladimir" <vladimir.medvedkin@intel.com>,
"dev@dpdk.org" <dev@dpdk.org>,
Yipeng Wang <yipeng1.wang@intel.com>,
Sameh Gobriel <sameh.gobriel@intel.com>,
Bruce Richardson <bruce.richardson@intel.com>,
"thomas@monjalon.net" <thomas@monjalon.net>, nd <nd@arm.com>
Subject: Re: [PATCH] lib/hash,lib/rcu: feature hidden key count in hash
Date: Wed, 28 Feb 2024 14:51:10 +0000 [thread overview]
Message-ID: <ACC49002-958F-49F3-8D5A-155CD5CEF86E@arm.com> (raw)
In-Reply-To: <CA+sj1i9SoLL1XfyqW9dC53JGkcLeCOrtEbpBmzyCrC7_GeokpA@mail.gmail.com>
> On Feb 28, 2024, at 5:44 AM, Abdullah Ömer Yamaç <aomeryamac@gmail.com> wrote:
>
> While I was implementing the new API, I realized one issue, and it would be good to discuss it here. First of all rte_rcu_qsbr_dq_reclaim function checks the state of the qsbr values. It means that all threads should report the quiescent states. It conflicts with my aim.
>
> Let's think about below scenario:
> Eight threads use a hash table and periodically report their quiescent states. One additional thread (main thread) periodically reports the hash size. I implemented the reclaim function in that thread. I mean, the main thread calls reclaim before the rte_hash_count.
>
> Here is the exceptional case that I couldn't retrieve the correct hash size:
> Assume that 6 of 8 threads reported quiescent states and 2 of them are still working on some process and haven't reported quiescent states yet. The main thread calls reclaim functions every time, but elements in dq will not be freed because 2 of the worker threads haven't reported their states (especially if they are waiting for some packets). So, my first proposed method is more suitable for this case. Any idea?
If 2 out of 8 threads have not reported their quiescent state then the elements that have not been acknowledged by those 2 threads cannot be reclaimed and cannot be allocated for further use. Using this you can calculate the most accurate number of entries in the hash table available for allocation.
>
> On Thu, Feb 22, 2024 at 7:44 PM Honnappa Nagarahalli <Honnappa.Nagarahalli@arm.com> wrote:
>
>
> > On Feb 22, 2024, at 6:39 AM, Abdullah Ömer Yamaç <aomeryamac@gmail.com> wrote:
> >
> > As a final decision, I will add a new hash API that forces the reclaim. Is it ok for everyone?
> Ack from my side
>
> >
> > On Thu, Feb 22, 2024 at 5:37 AM Honnappa Nagarahalli <Honnappa.Nagarahalli@arm.com> wrote:
> >
> >
> > > On Feb 21, 2024, at 3:51 PM, Abdullah Ömer Yamaç <aomeryamac@gmail.com> wrote:
> > >
> > >
> > >
> > > On Wed, Feb 21, 2024 at 6:24 AM Honnappa Nagarahalli <Honnappa.Nagarahalli@arm.com> wrote:
> > >
> > >
> > > > On Feb 20, 2024, at 12:58 PM, Abdullah Ömer Yamaç <aomeryamac@gmail.com> wrote:
> > > >
> > > > I appreciate that you gave me suggestions and comments. I will make changes according to all your recommendations, but before that, I want to make everyone's minds clear. Then, I will apply modifications.
> > > >
> > > > On Tue, Feb 20, 2024 at 2:35 AM Honnappa Nagarahalli <Honnappa.Nagarahalli@arm.com> wrote:
> > > >
> > > >
> > > > > On Feb 19, 2024, at 3:28 PM, Abdullah Ömer Yamaç <aomeryamac@gmail.com> wrote:
> > > > >
> > > > > Hello,
> > > > >
> > > > > Let me explain a use case;
> > > > >
> > > > > I have a hash table whose key value is IP addresses, and data (let's say the username of the IP) is related to the IP address. The key point is matching these data with flows. Flows are dynamic, and this hash table is dynamic, as well; both can change anytime. For example, when a flow starts, we look up the hash table with the corresponding IP and retrieve the username. We need to hold this username until the flow terminates, although we removed this IP key from the hash table (multithread). That's why we have RCU and defer queue is necessary for high performance. In my application, I need to know the number of IP-username entries. These numbers can be calculated by rte_hash_count - defer queue size.
> > > > The entries in the defer queue are not reclaimed (there is a probability that all of them can be reclaimed) and hence they are not available for allocation. So, rte_hash_count - defer queue size might not give you the correct number you are expecting.
> > > >
> > > > Currently, there is no API in hash library that forces a reclaim. Does it makes sense to have an API that just does the reclaim (and returns the number of entries pending in the defer queue)? A call to rte_hash_count should provide the exact count you are looking for.
> > > > You are right; no API in the hash library forces a reclaim. In my application, I periodically call rte_count to retrieve hash size, and this data is shown in my GUI. So that means I need to call regularly reclaim. I am trying to figure out which is better, calling reclaim or retrieving the defer queue size. Any comment about this?
> > > Retrieving the defer queue size will be cheaper. However, calling the reclaim API will ensure the entries are freed hence providing an accurate number. Calling the reclaim API on an empty defer queue does not consume many cycles. If needed we could add a check for empty defer queue in the reclaim API and return early.
> > >
> > > I am also wondering if a reclaim API in hash library is needed. Why not call rte_rcu_qsbr_dq_reclaim API from the application?
> > > The reason is simple. struct rte_hash *h is an internal structure and we cannot access the h->dq. So it is not possible to call reclaim.
> > Ack. This will be just a wrapper around the rte_rcu_qsbr_dq_reclaim.
> >
> > >
> > >
> > > > >
> > > > > I think if you need a non-blocking and multithreaded hash table, an RCU-enabled hash table is necessary. Also, this API is necessary if you need to get the actual matchable size.
> > > > >
> > > > >
> > > > >
> > > > >
> > > > >
> > > > > On Mon, Feb 19, 2024 at 8:36 PM Medvedkin, Vladimir <vladimir.medvedkin@intel.com> wrote:
> > > > > Hi Abdullah,
> > > > >
> > > > > Could you please tell more about use cases where this API may be useful?
> > > > >
> > > > > >a new API to get the hidden key count in the hash table if the rcu qsbr is enabled
> > > > >
> > > > > Here in commit message and down below in doxygen comments, I think this
> > > > > statement should be more specific because rcu can be created with
> > > > > RTE_HASH_QSBR_MODE_SYNC mode i.e. without defer queue.
> > > > >
> > > > > Also, new API must be reflected in release notes
> > > > >
> > > > > On 07/02/2024 15:33, Abdullah Ömer Yamaç wrote:
> > > > > > This patch introduce a new API to get the hidden key count in the hash
> > > > > > table if the rcu qsbr is enabled. When using rte_hash_count with rcu
> > > > > > qsbr enabled, it will return the number of elements that are not in the
> > > > > > free queue. Unless rte_rcu_qsbr_dq_reclaim is called, the number of
> > > > > > elements in the defer queue will not be counted and freed. Therefore I
> > > > > > added a new API to get the number of hidden (defer queue) elements
> > > > > > in the hash table. Then the user can calculate the total number of
> > > > > > elements that are available in the hash table.
> > > > > >
> > > > > > Signed-off-by: Abdullah Ömer Yamaç <aomeryamac@gmail.com>
> > > > > >
> > > > > > ---
> > > > > > Cc: Honnappa Nagarahalli <honnappa.nagarahalli@arm.com>
> > > > > > Cc: Yipeng Wang <yipeng1.wang@intel.com>
> > > > > > Cc: Sameh Gobriel <sameh.gobriel@intel.com>
> > > > > > Cc: Bruce Richardson <bruce.richardson@intel.com>
> > > > > > Cc: Vladimir Medvedkin <vladimir.medvedkin@intel.com>
> > > > > > ---
> > > > > > lib/hash/rte_cuckoo_hash.c | 9 +++++++++
> > > > > > lib/hash/rte_hash.h | 13 +++++++++++++
> > > > > > lib/hash/version.map | 1 +
> > > > > > lib/rcu/rte_rcu_qsbr.c | 8 ++++++++
> > > > > > lib/rcu/rte_rcu_qsbr.h | 11 +++++++++++
> > > > > > lib/rcu/version.map | 1 +
> > > > > > 6 files changed, 43 insertions(+)
> > > > > >
> > > > > > diff --git a/lib/hash/rte_cuckoo_hash.c b/lib/hash/rte_cuckoo_hash.c
> > > > > > index 70456754c4..3553f3efc7 100644
> > > > > > --- a/lib/hash/rte_cuckoo_hash.c
> > > > > > +++ b/lib/hash/rte_cuckoo_hash.c
> > > > > > @@ -555,6 +555,15 @@ rte_hash_max_key_id(const struct rte_hash *h)
> > > > > > return h->entries;
> > > > > > }
> > > > > >
> > > > > > +int32_t
> > > > > > +rte_hash_dq_count(const struct rte_hash *h)
> > > > > > +{
> > > > > > + if (h->dq == NULL)
> > > > > input arguments must be checked since this is a public API, the same is
> > > > > true for rte_rcu_qsbr_dq_count()
> > > > > > + return -EINVAL;
> > > > > why not just return 0?
> > > > > > +
> > > > > > + return rte_rcu_qsbr_dq_count(h->dq);
> > > > > > +}
> > > > > > +
> > > > > > int32_t
> > > > > > rte_hash_count(const struct rte_hash *h)
> > > > > > {
> > > > > > diff --git a/lib/hash/rte_hash.h b/lib/hash/rte_hash.h
> > > > > > index 7ecc021111..8ea97e297d 100644
> > > > > > --- a/lib/hash/rte_hash.h
> > > > > > +++ b/lib/hash/rte_hash.h
> > > > > > @@ -193,6 +193,19 @@ rte_hash_free(struct rte_hash *h);
> > > > > > void
> > > > > > rte_hash_reset(struct rte_hash *h);
> > > > > >
> > > > > > +
> > > > > > +/**
> > > > > > + * Return the number of records in the defer queue of the hash table
> > > > > > + * if RCU is enabled.
> > > > > > + * @param h
> > > > > > + * Hash table to query from
> > > > > > + * @return
> > > > > > + * - -EINVAL if parameters are invalid
> > > > > > + * - A value indicating how many records were inserted in the table.
> > > > > did you mean how many records are kept in defer queue?
> > > > > > + */
> > > > > > +int32_t
> > > > > > +rte_hash_dq_count(const struct rte_hash *h);
> > > > > > +
> > > > > > /**
> > > > > > * Return the number of keys in the hash table
> > > > > > * @param h
> > > > > > diff --git a/lib/hash/version.map b/lib/hash/version.map
> > > > > > index 6b2afebf6b..7f7b158cf1 100644
> > > > > > --- a/lib/hash/version.map
> > > > > > +++ b/lib/hash/version.map
> > > > > > @@ -9,6 +9,7 @@ DPDK_24 {
> > > > > > rte_hash_add_key_with_hash;
> > > > > > rte_hash_add_key_with_hash_data;
> > > > > > rte_hash_count;
> > > > > > + rte_hash_dq_count;
> > > > > new API must introduced as an experimental API. The same is true for
> > > > > rte_rcu_qsbr_dq_count()
> > > > > > rte_hash_crc32_alg;
> > > > > > rte_hash_crc_set_alg;
> > > > > > rte_hash_create;
> > > > > > diff --git a/lib/rcu/rte_rcu_qsbr.c b/lib/rcu/rte_rcu_qsbr.c
> > > > > > index bd0b83be0c..89f8da4c4c 100644
> > > > > > --- a/lib/rcu/rte_rcu_qsbr.c
> > > > > > +++ b/lib/rcu/rte_rcu_qsbr.c
> > > > > > @@ -450,6 +450,14 @@ rte_rcu_qsbr_dq_reclaim(struct rte_rcu_qsbr_dq *dq, unsigned int n,
> > > > > > return 0;
> > > > > > }
> > > > > >
> > > > > > +/**
> > > > > > + * Return the number of entries in a defer queue.
> > > > > > + */
> > > > > > +unsigned int rte_rcu_qsbr_dq_count(struct rte_rcu_qsbr_dq *dq)
> > > > > > +{
> > > > Please validate dq here.
> > > >
> > > > > > + return rte_ring_count(dq->r);
> > > > > > +}
> > > > > > +
> > > > > > /* Delete a defer queue. */
> > > > > > int
> > > > > > rte_rcu_qsbr_dq_delete(struct rte_rcu_qsbr_dq *dq)
> > > > > > diff --git a/lib/rcu/rte_rcu_qsbr.h b/lib/rcu/rte_rcu_qsbr.h
> > > > > > index 23c9f89805..ed5a590edd 100644
> > > > > > --- a/lib/rcu/rte_rcu_qsbr.h
> > > > > > +++ b/lib/rcu/rte_rcu_qsbr.h
> > > > > > @@ -794,6 +794,17 @@ int
> > > > > > rte_rcu_qsbr_dq_reclaim(struct rte_rcu_qsbr_dq *dq, unsigned int n,
> > > > > > unsigned int *freed, unsigned int *pending, unsigned int *available);
> > > > > >
> > > > > > +/**
> > > > > > + * Return the number of entries in a defer queue.
> > > > > > + *
> > > > > > + * @param dq
> > > > > > + * Defer queue.
> > > > > > + * @return
> > > > > > + * The number of entries in the defer queue.
> > > > > > + */
> > > > > > +unsigned int
> > > > > > +rte_rcu_qsbr_dq_count(struct rte_rcu_qsbr_dq *dq);
> > > > Agree on the need for this API in RCU
> > > >
> > > > > > +
> > > > > > /**
> > > > > > * Delete a defer queue.
> > > > > > *
> > > > > > diff --git a/lib/rcu/version.map b/lib/rcu/version.map
> > > > > > index 982ffd59d9..f410ab41e7 100644
> > > > > > --- a/lib/rcu/version.map
> > > > > > +++ b/lib/rcu/version.map
> > > > > > @@ -5,6 +5,7 @@ DPDK_24 {
> > > > > > rte_rcu_qsbr_dq_create;
> > > > > > rte_rcu_qsbr_dq_delete;
> > > > > > rte_rcu_qsbr_dq_enqueue;
> > > > > > + rte_rcu_qsbr_dq_count;
> > > > > > rte_rcu_qsbr_dq_reclaim;
> > > > > > rte_rcu_qsbr_dump;
> > > > > > rte_rcu_qsbr_get_memsize;
> > > > >
> > > > > --
> > > > > Regards,
> > > > > Vladimir
> > > > >
> > > >
> > >
> >
>
next prev parent reply other threads:[~2024-02-28 14:51 UTC|newest]
Thread overview: 35+ messages / expand[flat|nested] mbox.gz Atom feed top
2024-02-07 15:33 Abdullah Ömer Yamaç
2024-02-16 12:41 ` Thomas Monjalon
2024-02-19 17:36 ` Medvedkin, Vladimir
2024-02-19 21:28 ` Abdullah Ömer Yamaç
2024-02-19 23:35 ` Honnappa Nagarahalli
2024-02-20 18:58 ` Abdullah Ömer Yamaç
2024-02-21 3:23 ` Honnappa Nagarahalli
2024-02-21 21:51 ` Abdullah Ömer Yamaç
2024-02-22 2:37 ` Honnappa Nagarahalli
2024-02-22 12:39 ` Abdullah Ömer Yamaç
2024-02-22 16:44 ` Honnappa Nagarahalli
2024-02-28 11:44 ` Abdullah Ömer Yamaç
2024-02-28 14:51 ` Honnappa Nagarahalli [this message]
2024-03-02 20:22 ` Abdullah Ömer Yamaç
2024-03-02 21:08 ` [PATCH v2] lib/hash: feature reclaim defer queue Abdullah Ömer Yamaç
2024-03-02 21:27 ` Abdullah Ömer Yamaç
2024-03-03 19:14 ` Honnappa Nagarahalli
2024-03-04 8:27 ` Abdullah Ömer Yamaç
2024-03-04 21:58 ` Honnappa Nagarahalli
2024-03-06 8:55 ` [PATCH v3] lib/hash: add defer queue reclaim API Abdullah Ömer Yamaç
2024-03-06 8:59 ` Abdullah Ömer Yamaç
2024-03-06 10:13 ` Abdullah Ömer Yamaç
2024-03-14 7:04 ` Abdullah Ömer Yamaç
2024-04-04 10:11 ` Abdullah Ömer Yamaç
2024-04-05 2:11 ` Honnappa Nagarahalli
2024-04-15 11:26 ` [PATCH v4] " Abdullah Ömer Yamaç
2024-04-23 13:49 ` Abdullah Ömer Yamaç
2024-04-23 21:24 ` Stephen Hemminger
2024-04-25 14:03 ` Abdullah Ömer Yamaç
2024-04-27 19:54 ` [PATCH v5] " Abdullah Ömer Yamaç
2024-05-10 3:05 ` Abdullah Ömer Yamaç
2024-05-13 5:24 ` Honnappa Nagarahalli
2024-05-13 15:34 ` Abdullah Ömer Yamaç
2024-05-15 10:49 ` Abdullah Ömer Yamaç
2024-05-15 10:54 ` [PATCH v6] " Abdullah Ömer Yamaç
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=ACC49002-958F-49F3-8D5A-155CD5CEF86E@arm.com \
--to=honnappa.nagarahalli@arm.com \
--cc=aomeryamac@gmail.com \
--cc=bruce.richardson@intel.com \
--cc=dev@dpdk.org \
--cc=nd@arm.com \
--cc=sameh.gobriel@intel.com \
--cc=thomas@monjalon.net \
--cc=vladimir.medvedkin@intel.com \
--cc=yipeng1.wang@intel.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).