netdev - Re: [PATCH 3/3] net: allocate tx queues in register

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:	Mon, 18 Oct 2010 17:17:41 -0700
From:	Tom Herbert <therbert@...gle.com>
To:	John Fastabend <john.r.fastabend@...el.com>
Cc:	Ben Hutchings <bhutchings@...arflare.com>,
	"davem@...emloft.net" <davem@...emloft.net>,
	"netdev@...r.kernel.org" <netdev@...r.kernel.org>,
	"eric.dumazet@...il.com" <eric.dumazet@...il.com>
Subject: Re: [PATCH 3/3] net: allocate tx queues in register_netdevice

On Mon, Oct 18, 2010 at 5:12 PM, John Fastabend
<john.r.fastabend@...el.com> wrote:
> On 10/18/2010 2:41 PM, Ben Hutchings wrote:
>> On Mon, 2010-10-18 at 11:02 -0700, Tom Herbert wrote:
>>> This patch introduces netif_alloc_netdev_queues which is called from
>>> register_device instead of alloc_netdev_mq.  This makes TX queue
>>> allocation symmetric with RX allocation similarly allows drivers to
>>> change dev->num_tx_queues after allocating netdev and before
>>> registering it.
>>
>> Changing num_tx_queues is probably *not* desirable, same as for
>> num_rx_queues.
>>

Okay, so both netif_set_real_num_rx_queues and
netif_set_real_num_tx_queues should return -EINVAL if queue count >
num_[tr]x_queues?

>
> Right, this will break ixgbe and other drivers that may call netif_set_real_num_{rx|tx}_queues() to increase the number of
> queues. Returning an error code seems like a good idea though.
>
> John.
>
>> [...]
>>> --- a/net/core/dev.c
>>> +++ b/net/core/dev.c
>>> @@ -1553,18 +1553,24 @@ static void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev)
>>>   * Routine to help set real_num_tx_queues. To avoid skbs mapped to queues
>>>   * greater then real_num_tx_queues stale skbs on the qdisc must be flushed.
>>>   */
>>> -void netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq)
>>> +int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq)
>>>  {
>>> -    unsigned int real_num = dev->real_num_tx_queues;
>>> +    if (txq < 1)
>>> +            return -EINVAL;
>>>
>>> -    if (unlikely(txq > dev->num_tx_queues))
>>> -            ;
>>> -    else if (txq > real_num)
>>> -            dev->real_num_tx_queues = txq;
>>> -    else if (txq < real_num) {
>>> -            dev->real_num_tx_queues = txq;
>>> -            qdisc_reset_all_tx_gt(dev, txq);
>>> -    }
>>> +    if (dev->reg_state == NETREG_REGISTERED) {
>>> +            ASSERT_RTNL();
>>> +
>>> +            if (txq > dev->num_tx_queues)
>>> +                    return -EINVAL;
>>> +
>>> +            if (txq < dev->real_num_tx_queues)
>>> +                    qdisc_reset_all_tx_gt(dev, txq);
>>> +    } else
>>> +            dev->num_tx_queues = txq;
>> [...]
>>
>> The kernel-doc comment should be updated to reflect the locking
>> requirement and the possibility of failure when called after
>> registration.
>>
>> Ben.
>>
>
>
--
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html