netdev - Re: [PATCH net-next] net: add in_softirq() debug checking in napi_consume

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]

Message-ID: <5b04ad33-1611-8d7b-8fec-4269c01ecab3@huawei.com>
Date:   Mon, 2 Nov 2020 11:14:32 +0800
From:   Yunsheng Lin <linyunsheng@...wei.com>
To:     Jakub Kicinski <kuba@...nel.org>
CC:     <davem@...emloft.net>, <linmiaohe@...wei.com>,
        <martin.varghese@...ia.com>, <pabeni@...hat.com>,
        <pshelar@....org>, <fw@...len.de>, <gnault@...hat.com>,
        <steffen.klassert@...unet.com>, <kyk.segfault@...il.com>,
        <viro@...iv.linux.org.uk>, <vladimir.oltean@....com>,
        <edumazet@...gle.com>, <saeed@...nel.org>,
        <netdev@...r.kernel.org>, <linux-kernel@...r.kernel.org>,
        <linuxarm@...wei.com>
Subject: Re: [PATCH net-next] net: add in_softirq() debug checking in
 napi_consume_skb()

On 2020/11/1 6:38, Jakub Kicinski wrote:
> On Thu, 29 Oct 2020 19:34:48 +0800 Yunsheng Lin wrote:
>> The current semantic for napi_consume_skb() is that caller need
>> to provide non-zero budget when calling from NAPI context, and
>> breaking this semantic will cause hard to debug problem, because
>> _kfree_skb_defer() need to run in atomic context in order to push
>> the skb to the particular cpu' napi_alloc_cache atomically.
>>
>> So add a in_softirq() debug checking in napi_consume_skb() to catch
>> this kind of error.
>>
>> Suggested-by: Eric Dumazet <edumazet@...gle.com>
>> Signed-off-by: Yunsheng Lin <linyunsheng@...wei.com>
> 
>> diff --git a/net/core/skbuff.c b/net/core/skbuff.c
>> index 1ba8f01..1834007 100644
>> --- a/net/core/skbuff.c
>> +++ b/net/core/skbuff.c
>> @@ -897,6 +897,10 @@ void napi_consume_skb(struct sk_buff *skb, int budget)
>>  		return;
>>  	}
>>  
>> +	DEBUG_NET_WARN(!in_softirq(),
>> +		       "%s is called with non-zero budget outside softirq context.\n",
>> +		       __func__);
> 
> Can't we use lockdep instead of defining our own knobs?

>From the first look, using the below seems better than defining our
own knobs, because there is similar lockdep_assert_in_irq() checking
already and lockdep_assert_in_*() is NULL-OP when CONFIG_PROVE_LOCKING
is not defined.

> 
> Like this maybe?
> 
> diff --git a/include/linux/lockdep.h b/include/linux/lockdep.h
> index f5594879175a..5253a167d00c 100644
> --- a/include/linux/lockdep.h
> +++ b/include/linux/lockdep.h
> @@ -594,6 +594,14 @@ do {                                                                       \
>                       this_cpu_read(hardirqs_enabled)));                \
>  } while (0)
>  
> +#define lockdep_assert_in_softirq()                                    \
> +do {                                                                   \
> +       WARN_ON_ONCE(__lockdep_enabled                  &&              \
> +                    (softirq_count() == 0              ||              \
> +                     this_cpu_read(hardirq_context)));                 \

Using in_softirq() seems more obvious then using softirq_count()?
And there is below comment above avoiding the using of in_softirq(), maybe
that is why you use softirq_count() directly here?
"softirq_count() == 0" still mean we are not in the SoftIRQ context and
BH is not disabled. right? Perhap lockdep_assert_in_softirq_or_bh_disabled()
is more obvious?

/*
 * Are we doing bottom half or hardware interrupt processing?
 *
 * in_irq()       - We're in (hard) IRQ context
 * in_softirq()   - We have BH disabled, or are processing softirqs
 * in_interrupt() - We're in NMI,IRQ,SoftIRQ context or have BH disabled
 * in_serving_softirq() - We're in softirq context
 * in_nmi()       - We're in NMI context
 * in_task()	  - We're in task context
 *
 * Note: due to the BH disabled confusion: in_softirq(),in_interrupt() really
 *       should not be used in new code.
 */


Also, is there any particular reason we do the "this_cpu_read(hardirq_context)"
checking?

Thanks.

> +} while (0)
> 
> 
> 
>>  	if (!skb_unref(skb))
>>  		return;
>>  
> 
> .
>