linux-kernel - Re: [PATCH 12/14] net: sched: retry action check-insert on concurrent modification

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives
Hash Suite: Windows password security audit tool. GUI, reports in PDF.
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Date:   Wed, 16 May 2018 15:21:35 +0200
From:   Jiri Pirko <jiri@...nulli.us>
To:     Vlad Buslov <vladbu@...lanox.com>
Cc:     netdev@...r.kernel.org, davem@...emloft.net, jhs@...atatu.com,
        xiyou.wangcong@...il.com, pablo@...filter.org,
        kadlec@...ckhole.kfki.hu, fw@...len.de, ast@...nel.org,
        daniel@...earbox.net, edumazet@...gle.com, keescook@...omium.org,
        linux-kernel@...r.kernel.org, netfilter-devel@...r.kernel.org,
        coreteam@...filter.org, kliteyn@...lanox.com
Subject: Re: [PATCH 12/14] net: sched: retry action check-insert on
 concurrent modification

Wed, May 16, 2018 at 02:43:58PM CEST, vladbu@...lanox.com wrote:
>
>On Wed 16 May 2018 at 12:26, Jiri Pirko <jiri@...nulli.us> wrote:
>> Wed, May 16, 2018 at 01:55:06PM CEST, vladbu@...lanox.com wrote:
>>>
>>>On Wed 16 May 2018 at 09:59, Jiri Pirko <jiri@...nulli.us> wrote:
>>>> Mon, May 14, 2018 at 04:27:13PM CEST, vladbu@...lanox.com wrote:
>>>>>Retry check-insert sequence in action init functions if action with same
>>>>>index was inserted concurrently.
>>>>>
>>>>>Signed-off-by: Vlad Buslov <vladbu@...lanox.com>
>>>>>---
>>>>> net/sched/act_bpf.c        | 8 +++++++-
>>>>> net/sched/act_connmark.c   | 8 +++++++-
>>>>> net/sched/act_csum.c       | 8 +++++++-
>>>>> net/sched/act_gact.c       | 8 +++++++-
>>>>> net/sched/act_ife.c        | 8 +++++++-
>>>>> net/sched/act_ipt.c        | 8 +++++++-
>>>>> net/sched/act_mirred.c     | 8 +++++++-
>>>>> net/sched/act_nat.c        | 8 +++++++-
>>>>> net/sched/act_pedit.c      | 8 +++++++-
>>>>> net/sched/act_police.c     | 9 ++++++++-
>>>>> net/sched/act_sample.c     | 8 +++++++-
>>>>> net/sched/act_simple.c     | 9 ++++++++-
>>>>> net/sched/act_skbedit.c    | 8 +++++++-
>>>>> net/sched/act_skbmod.c     | 8 +++++++-
>>>>> net/sched/act_tunnel_key.c | 9 ++++++++-
>>>>> net/sched/act_vlan.c       | 9 ++++++++-
>>>>> 16 files changed, 116 insertions(+), 16 deletions(-)
>>>>>
>>>>>diff --git a/net/sched/act_bpf.c b/net/sched/act_bpf.c
>>>>>index 5554bf7..7e20fdc 100644
>>>>>--- a/net/sched/act_bpf.c
>>>>>+++ b/net/sched/act_bpf.c
>>>>>@@ -299,10 +299,16 @@ static int tcf_bpf_init(struct net *net, struct nlattr *nla,
>>>>> 
>>>>> 	parm = nla_data(tb[TCA_ACT_BPF_PARMS]);
>>>>> 
>>>>>+replay:
>>>>> 	if (!tcf_idr_check(tn, parm->index, act, bind)) {
>>>>> 		ret = tcf_idr_create(tn, parm->index, est, act,
>>>>> 				     &act_bpf_ops, bind, true);
>>>>>-		if (ret < 0)
>>>>>+		/* Action with specified index was created concurrently.
>>>>>+		 * Check again.
>>>>>+		 */
>>>>>+		if (parm->index && ret == -ENOSPC)
>>>>>+			goto replay;
>>>>>+		else if (ret)
>>>>
>>>> Hmm, looks like you are doing the same/very similar thing in every act
>>>> code. I think it would make sense to introduce a helper function for
>>>> this purpose.
>>>
>>>This code uses goto so it can't be easily refactored into standalone
>>>function. Could you specify which part of this code you suggest to
>>>extract?
>>
>> Hmm, looking at the code, I think that what would help is to have a
>> helper that would atomically check if index exists and if not, it would
>> allocate one. Something like:
>>
>>
>> int tcf_idr_check_alloc(struct tc_action_net *tn, u32 *index,
>> 			struct tc_action **a, int bind)
>> {
>> 	struct tcf_idrinfo *idrinfo = tn->idrinfo;
>> 	struct tc_action *p;
>> 	int err;
>>
>> 	spin_lock(&idrinfo->lock);
>> 	if (*index) {
>> 		p = idr_find(&idrinfo->action_idr, *index);
>> 		if (p) {
>> 			if (bind)
>> 	   			p->tcfa_bindcnt++;
>> 			p->tcfa_refcnt++;
>> 			*a = p;
>> 			err = 0;
>> 		} else {
>> 			*a = NULL;
>> 			err = idr_alloc_u32(idr, NULL, index,
>> 					    *index, GFP_ATOMIC);
>> 		}
>> 	} else {
>> 		*index = 1;
>> 		*a = NULL;
>> 		err = idr_alloc_u32(idr, NULL, index, UINT_MAX, GFP_ATOMIC);
>> 	}
>> 	spin_unlock(&idrinfo->lock);
>> 	return err;
>> }
>>
>> The act code would just check if "a" is NULL and if so, it would call
>> tcf_idr_create() with allocated index as arg.
>
>What about multiple actions that have arbitrary code between initial
>check and idr allocation that is currently inside tcf_idr_create()?

Why it would be a problem to have them after the allocation?

There is one issue though with my draft. tcf_idr_insert() function
which actually assigns a "p" pointer to the idr index is called later on.
Until that happens, the idr_find() would return NULL even if the index
is actually allocated. We cannot assign "p" in tcf_idr_check_alloc()
because it is allocated only later on in tcf_idr_create(). But that is
resolvable by the following trick:

int tcf_idr_check_alloc(struct tc_action_net *tn, u32 *index,
			struct tc_action **a, int bind)
{
	struct tcf_idrinfo *idrinfo = tn->idrinfo;
	struct tc_action *p;
	int err;

again:
	spin_lock(&idrinfo->lock);
	if (*index) {
 		p = idr_find(&idrinfo->action_idr, *index);
		if (IS_ERR(p)) {
			/* This means that another process allocated
			 * index but did not assign the pointer yet.
			 */
			spin_unlock(&idrinfo->lock);
			goto again;
		}
 		if (p) {
 			if (bind)
 	   			p->tcfa_bindcnt++;
 			p->tcfa_refcnt++;
 			*a = p;
 			err = 0;
 		} else {
 			*a = NULL;
 			err = idr_alloc_u32(idr, NULL, index,
 					    *index, GFP_ATOMIC);
			idr_replace(&idrinfo->action_idr,
				    ERR_PTR(-EBUSY), *index);
 		}
 	} else {
 		*index = 1;
		*a = NULL;
 		err = idr_alloc_u32(idr, NULL, index, UINT_MAX, GFP_ATOMIC);
		idr_replace(&idrinfo->action_idr, ERR_PTR(-EBUSY), *index);
 	}
 	spin_unlock(&idrinfo->lock);
 	return err;
}





>
>>
>>
>>>
>>>>
>>>> [...]
>>>
>