From: "Eelco Chaudron" <echaudro@redhat.com>
To: "Sebastian Andrzej Siewior" <bigeasy@linutronix.de>
Cc: netdev@vger.kernel.org, davem@davemloft.net, dev@openvswitch.org,
kuba@kernel.org, pabeni@redhat.com, pshelar@ovn.org,
jlelli@redhat.com
Subject: Re: [PATCH net-next] net: openvswitch: fix to make sure flow_lookup() is not preempted
Date: Wed, 14 Oct 2020 12:44:23 +0200 [thread overview]
Message-ID: <3D834ADB-09E7-4E28-B62F-CB6281987E41@redhat.com> (raw)
In-Reply-To: <20201013125307.ugz4nvjvyxrfhi6n@linutronix.de>
On 13 Oct 2020, at 14:53, Sebastian Andrzej Siewior wrote:
> On 2020-10-13 14:44:19 [+0200], Eelco Chaudron wrote:
>> The flow_lookup() function uses per CPU variables, which must not be
>> preempted. However, this is fine in the general napi use case where
>> the local BH is disabled. But, it's also called in the netlink
>> context, which is preemptible. The below patch makes sure that even
>> in the netlink path, preemption is disabled.
>>
>> Fixes: eac87c413bf9 ("net: openvswitch: reorder masks array based on
>> usage")
>> Reported-by: Juri Lelli <jlelli@redhat.com>
>> Signed-off-by: Eelco Chaudron <echaudro@redhat.com>
>> ---
>> net/openvswitch/flow_table.c | 10 +++++++++-
>> 1 file changed, 9 insertions(+), 1 deletion(-)
>>
>> diff --git a/net/openvswitch/flow_table.c
>> b/net/openvswitch/flow_table.c
>> index 87c286ad660e..16289386632b 100644
>> --- a/net/openvswitch/flow_table.c
>> +++ b/net/openvswitch/flow_table.c
>> @@ -850,9 +850,17 @@ struct sw_flow *ovs_flow_tbl_lookup(struct
>> flow_table *tbl,
>> struct mask_array *ma = rcu_dereference_ovsl(tbl->mask_array);
>> u32 __always_unused n_mask_hit;
>> u32 __always_unused n_cache_hit;
>> + struct sw_flow *flow;
>> u32 index = 0;
>>
>> - return flow_lookup(tbl, ti, ma, key, &n_mask_hit, &n_cache_hit,
>> &index);
>> + /* This function gets called trough the netlink interface and
>> therefore
>> + * is preemptible. However, flow_lookup() function needs to be
>> called
>> + * with preemption disabled due to CPU specific variables.
>> + */
>
> Once again. u64_stats_update_begin(). What protects you against
> concurrent access.
Thanks Sebastian for repeating this, as I thought I went over the
seqcount code and thought it should be fine for my use case. However
based on this comment I went over it again, and found the logic part I
was constantly missing :)
My idea is to send a v2 patch and in addition to the preempt_disable()
also make the seqcount part per CPU. I noticed other parts of the
networking stack doing it the same way. So the patch would look
something like:
@@ -731,7 +732,7 @@ static struct sw_flow *flow_lookup(struct flow_table
*tbl,
u32 *n_cache_hit,
u32 *index)
{
- u64 *usage_counters = this_cpu_ptr(ma->masks_usage_cntr);
+ struct mask_array_stats *stats =
this_cpu_ptr(ma->masks_usage_stats);
struct sw_flow *flow;
struct sw_flow_mask *mask;
int i;
@@ -741,9 +742,9 @@ static struct sw_flow *flow_lookup(struct flow_table
*tbl,
if (mask) {
flow = masked_flow_lookup(ti, key, mask,
n_mask_hit);
if (flow) {
- u64_stats_update_begin(&ma->syncp);
- usage_counters[*index]++;
- u64_stats_update_end(&ma->syncp);
+ u64_stats_update_begin(&stats->syncp);
+ stats->usage_cntr[*index]++;
+ u64_stats_update_end(&stats->syncp);
(*n_cache_hit)++;
return flow;
}
Let me know your thoughts.
Thanks,
Eelco
next prev parent reply other threads:[~2020-10-14 10:44 UTC|newest]
Thread overview: 6+ messages / expand[flat|nested] mbox.gz Atom feed top
2020-10-13 12:44 [PATCH net-next] net: openvswitch: fix to make sure flow_lookup() is not preempted Eelco Chaudron
2020-10-13 12:53 ` Sebastian Andrzej Siewior
2020-10-14 10:44 ` Eelco Chaudron [this message]
2020-10-15 7:55 ` Sebastian Andrzej Siewior
2020-10-15 8:11 ` Eelco Chaudron
2020-10-15 8:15 ` Sebastian Andrzej Siewior
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=3D834ADB-09E7-4E28-B62F-CB6281987E41@redhat.com \
--to=echaudro@redhat.com \
--cc=bigeasy@linutronix.de \
--cc=davem@davemloft.net \
--cc=dev@openvswitch.org \
--cc=jlelli@redhat.com \
--cc=kuba@kernel.org \
--cc=netdev@vger.kernel.org \
--cc=pabeni@redhat.com \
--cc=pshelar@ovn.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).