DPDK patches and discussions
 help / color / mirror / Atom feed
From: Jia He <hejianet@gmail.com>
To: jerin.jacob@caviumnetworks.com, dev@dpdk.org, olivier.matz@6wind.com
Cc: konstantin.ananyev@intel.com, bruce.richardson@intel.com,
	jianbo.liu@arm.com, hemant.agrawal@nxp.com,
	Jia He <hejianet@gmail.com>, Jia He <jia.he@hxt-semitech.com>,
	jie2.liu@hxt-semitech.com, bing.zhao@hxt-semitech.com,
	stable@dpdk.org
Subject: [dpdk-dev] [PATCH v6] ring: guarantee load/load order in enqueue and dequeue
Date: Fri, 10 Nov 2017 03:30:42 +0000	[thread overview]
Message-ID: <1510284642-7442-2-git-send-email-hejianet@gmail.com> (raw)
In-Reply-To: <1510284642-7442-1-git-send-email-hejianet@gmail.com>

We watched a rte panic of mbuf_autotest in our qualcomm arm64 server
(Amberwing).

Root cause:
In __rte_ring_move_cons_head()
...
        do {
                /* Restore n as it may change every loop */
                n = max;

                *old_head = r->cons.head;                //1st load
                const uint32_t prod_tail = r->prod.tail; //2nd load

In weak memory order architectures(powerpc,arm), the 2nd load might be
reodered before the 1st load, that makes *entries is bigger than we wanted.
This nasty reording messed enque/deque up.

cpu1(producer)          cpu2(consumer)          cpu3(consumer)
                        load r->prod.tail
in enqueue:
load r->cons.tail
load r->prod.head

store r->prod.tail

                                                load r->cons.head
                                                load r->prod.tail
                                                ...
                                                store r->cons.{head,tail}
                        load r->cons.head

Then, r->cons.head will be bigger than prod_tail, then make *entries very
big and the consumer will go forward incorrectly.

After this patch, the old cons.head will be recaculated after failure of
rte_atomic32_cmpset

There is no such issue on X86, because X86 is strong memory order model.
But rte_smp_rmb() doesn't have impact on runtime performance on X86, so
keep the same code without architectures specific concerns.

Signed-off-by: Jia He <jia.he@hxt-semitech.com>
Signed-off-by: jie2.liu@hxt-semitech.com
Signed-off-by: bing.zhao@hxt-semitech.com
Acked-by: Jerin Jacob <jerin.jacob@caviumnetworks.com>
Acked-by: Jianbo Liu <jianbo.liu@arm.com>
Cc: stable@dpdk.org

---
 lib/librte_ring/rte_ring.h | 12 ++++++++++++
 1 file changed, 12 insertions(+)

diff --git a/lib/librte_ring/rte_ring.h b/lib/librte_ring/rte_ring.h
index 5e9b3b7..e924438 100644
--- a/lib/librte_ring/rte_ring.h
+++ b/lib/librte_ring/rte_ring.h
@@ -409,6 +409,12 @@ __rte_ring_move_prod_head(struct rte_ring *r, int is_sp,
 		n = max;
 
 		*old_head = r->prod.head;
+
+		/* add rmb barrier to avoid load/load reorder in weak
+		 * memory model. It is noop on x86
+		 */
+		rte_smp_rmb();
+
 		const uint32_t cons_tail = r->cons.tail;
 		/*
 		 *  The subtraction is done between two unsigned 32bits value
@@ -517,6 +523,12 @@ __rte_ring_move_cons_head(struct rte_ring *r, int is_sc,
 		n = max;
 
 		*old_head = r->cons.head;
+
+		/* add rmb barrier to avoid load/load reorder in weak
+		 * memory model. It is noop on x86
+		 */
+		rte_smp_rmb();
+
 		const uint32_t prod_tail = r->prod.tail;
 		/* The subtraction is done between two unsigned 32bits value
 		 * (the result is always modulo 32 bits even if we have
-- 
2.7.4

  reply	other threads:[~2017-11-10  3:31 UTC|newest]

Thread overview: 39+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
     [not found] <1510118764-29697-1-git-send-email-hejianet@gmail.com>
2017-11-08  9:54 ` [dpdk-dev] [PATCH v4 0/4] fix race condition in enqueue/dequeue because of cpu reorder Jia He
2017-11-08  9:54   ` [dpdk-dev] [PATCH v4 1/4] eal/arm64: remove the braces {} for dmb() and dsb() Jia He
2017-11-08  9:54   ` [dpdk-dev] [PATCH v4 2/4] ring: guarantee load/load order in enqueue and dequeue Jia He
2017-11-08  9:54   ` [dpdk-dev] [PATCH v4 3/4] ring: introduce new header file to include common functions Jia He
2017-11-08  9:54   ` [dpdk-dev] [PATCH v4 4/4] ring: introduce new header file to support C11 memory model Jia He
2017-11-08 12:15   ` [dpdk-dev] [PATCH v4 0/4] fix race condition in enqueue/dequeue because of cpu reorder Bruce Richardson
2017-11-08 15:11     ` Jia He
2017-11-08 16:29       ` Jerin Jacob
2017-11-08 18:36       ` Ananyev, Konstantin
     [not found]       ` <2459a535-920e-9ac5-2f46-1d1dd00e275b@gmail.com>
2017-11-24  9:24         ` Bruce Richardson
2017-11-10  1:51 ` [dpdk-dev] [PATCH v5 0/1] " Jia He
2017-11-10  1:51   ` [dpdk-dev] [PATCH v5 1/1] ring: guarantee load/load order in enqueue and dequeue Jia He
2017-11-10  2:46     ` Jerin Jacob
2017-11-10  3:12       ` Jianbo Liu
2017-11-10  9:59     ` Ananyev, Konstantin
2017-11-10  3:30   ` [dpdk-dev] [PATCH v6] " Jia He
2017-11-10  3:30     ` Jia He [this message]
2017-11-12 17:51       ` [dpdk-dev] [dpdk-stable] [PATCH v6] ring: " Thomas Monjalon
2017-11-10  5:23   ` [dpdk-dev] [PATCH v5 0/3] support c11 memory model barrier in librte_ring Jia He
2017-11-10  5:23     ` [dpdk-dev] [PATCH v5 1/3] eal/arm64: remove the braces {} for dmb() and dsb() Jia He
2017-11-10  5:23     ` [dpdk-dev] [PATCH v5 2/3] ring: introduce new header file to include common functions Jia He
2017-11-10  5:23     ` [dpdk-dev] [PATCH v6 3/3] ring: introduce new header file to support C11 memory model Jia He
2017-11-27  2:00     ` [dpdk-dev] [PATCH V6 0/3] support c11 memory model barrier in librte_ring Jia He
2017-11-27  2:00       ` [dpdk-dev] [PATCH V6 1/3] eal/arm64: remove the braces {} for dmb() and dsb() Jia He
2017-12-03 11:11         ` Jerin Jacob
2017-11-27  2:00       ` [dpdk-dev] [PATCH V6 2/3] ring: introduce new header file to include common functions Jia He
2017-12-03 12:13         ` Jerin Jacob
2017-11-27  2:00       ` [dpdk-dev] [PATCH V6 3/3] ring: introduce new header file to support C11 memory model Jia He
2017-12-03 12:14         ` Jerin Jacob
2017-12-04  1:50       ` [dpdk-dev] [PATCH V7 0/3] support c11 memory model barrier in librte_ring Jia He
2017-12-04  1:50         ` [dpdk-dev] [PATCH V7 1/3] eal/arm64: remove the braces {} for dmb() and dsb() Jia He
2017-12-04  1:50         ` [dpdk-dev] [PATCH V7 2/3] ring: introduce new header file to include common functions Jia He
2018-01-12 17:09           ` Thomas Monjalon
2018-01-16  2:06             ` Jia He
2018-01-16 15:19           ` Olivier Matz
2017-12-04  1:50         ` [dpdk-dev] [PATCH V7 3/3] ring: introduce new header file to support C11 memory model Jia He
2017-12-04  8:05           ` Jianbo Liu
2018-01-12 17:18           ` Thomas Monjalon
2018-01-16 15:18           ` Olivier Matz

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=1510284642-7442-2-git-send-email-hejianet@gmail.com \
    --to=hejianet@gmail.com \
    --cc=bing.zhao@hxt-semitech.com \
    --cc=bruce.richardson@intel.com \
    --cc=dev@dpdk.org \
    --cc=hemant.agrawal@nxp.com \
    --cc=jerin.jacob@caviumnetworks.com \
    --cc=jia.he@hxt-semitech.com \
    --cc=jianbo.liu@arm.com \
    --cc=jie2.liu@hxt-semitech.com \
    --cc=konstantin.ananyev@intel.com \
    --cc=olivier.matz@6wind.com \
    --cc=stable@dpdk.org \
    /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).