| 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
| ||
|
Date: Fri, 22 Nov 2013 15:56:39 +0400
From: Denis Kirjanov <kirjanov@...il.com>
To: ZHAO Gang <gamerh2o@...il.com>
Cc: Mark Einon <mark.einon@...il.com>,
Greg Kroah-Hartman <gregkh@...uxfoundation.org>,
LKML <linux-kernel@...r.kernel.org>,
STAGING SUBSYSTEM <devel@...verdev.osuosl.org>
Subject: Re: [PATCH v3 2/5] staging: et131x: drop packet when error occurs in et131x_tx
If you have no free TX descriptors that means that something went
wrong and it's a BUG. You have to tell the stack to stop sending
packets using netif_stop_queue() and reenable transmissions once tx
descriptors will be available. There are a lot of live examples in the
source tree.
On 11/22/13, ZHAO Gang <gamerh2o@...il.com> wrote:
> On Fri, Nov 22, 2013 at 5:17 PM, Denis Kirjanov <kirjanov@...il.com> wrote:
>> On 11/22/13, Mark Einon <mark.einon@...il.com> wrote:
>>> On Wed, Nov 20, 2013 at 03:55:27PM +0800, ZHAO Gang wrote:
>>>> As TODO file suggested, drop packet instead of return NETDEV_TX_BUSY
>>>> when tx failed.
>>>>
>>>> et131x_tx calls function et131x_send_packets, I put the work of
>>>> et131x_send_packets directly into et131x_tx, and made some changes to
>>>> let the code more readable.
>>>>
>>>> Signed-off-by: ZHAO Gang <gamerh2o@...il.com>
>>>
>>> Acked-by: Mark Einon <mark.einon@...il.com>
>>>
>>>> ---
>>>> drivers/staging/et131x/et131x.c | 84
>>>> +++++++++++------------------------------
>>>> 1 file changed, 22 insertions(+), 62 deletions(-)
>>>>
>>>> diff --git a/drivers/staging/et131x/et131x.c
>>>> b/drivers/staging/et131x/et131x.c
>>>> index 836a4db..cda037a 100644
>>>> --- a/drivers/staging/et131x/et131x.c
>>>> +++ b/drivers/staging/et131x/et131x.c
>>>> @@ -3123,55 +3123,6 @@ static int send_packet(struct sk_buff *skb,
>>>> struct
>>>> et131x_adapter *adapter)
>>>> return 0;
>>>> }
>>>>
>>>> -/* et131x_send_packets - This function is called by the OS to send
>>>> packets
>>>> - * @skb: the packet(s) to send
>>>> - * @netdev:device on which to TX the above packet(s)
>>>> - *
>>>> - * Return 0 in almost all cases; non-zero value in extreme hard
>>>> failure
>>>> only
>>>> - */
>>>> -static int et131x_send_packets(struct sk_buff *skb, struct net_device
>>>> *netdev)
>>>> -{
>>>> - int status = 0;
>>>> - struct et131x_adapter *adapter = netdev_priv(netdev);
>>>> -
>>>> - /* Send these packets
>>>> - *
>>>> - * NOTE: The Linux Tx entry point is only given one packet at a
>>>> time
>>>> - * to Tx, so the PacketCount and it's array used makes no sense
>>>> here
>>>> - */
>>>> -
>>>> - /* TCB is not available */
>>>> - if (adapter->tx_ring.used >= NUM_TCB) {
>>>> - /* NOTE: If there's an error on send, no need to queue the
>>>> - * packet under Linux; if we just send an error up to the
>>>> - * netif layer, it will resend the skb to us.
>>>> - */
>>>> - status = -ENOMEM;
>>>> - } else {
>>>> - /* We need to see if the link is up; if it's not, make the
>>>> - * netif layer think we're good and drop the packet
>>>> - */
>>>> - if ((adapter->flags & FMP_ADAPTER_FAIL_SEND_MASK) ||
>>>> - !netif_carrier_ok(netdev)) {
>>>> - dev_kfree_skb_any(skb);
>>>> - skb = NULL;
>>>> -
>>>> - adapter->net_stats.tx_dropped++;
>>>> - } else {
>>>> - status = send_packet(skb, adapter);
>>>> - if (status != 0 && status != -ENOMEM) {
>>>> - /* On any other error, make netif think
>>>> we're
>>>> - * OK and drop the packet
>>>> - */
>>>> - dev_kfree_skb_any(skb);
>>>> - skb = NULL;
>>>> - adapter->net_stats.tx_dropped++;
>>>> - }
>>>> - }
>>>> - }
>>>> - return status;
>>>> -}
>>>> -
>>>> /* free_send_packet - Recycle a struct tcb
>>>> * @adapter: pointer to our adapter
>>>> * @tcb: pointer to struct tcb
>>>> @@ -4537,12 +4488,9 @@ static void et131x_multicast(struct net_device
>>>> *netdev)
>>>> /* et131x_tx - The handler to tx a packet on the device
>>>> * @skb: data to be Tx'd
>>>> * @netdev: device on which data is to be Tx'd
>>>> - *
>>>> - * Returns 0 on success, errno on failure (as defined in errno.h)
>>>> */
>>>> -static int et131x_tx(struct sk_buff *skb, struct net_device *netdev)
>>>> +static netdev_tx_t et131x_tx(struct sk_buff *skb, struct net_device
>>>> *netdev)
>>>> {
>>>> - int status = 0;
>>>> struct et131x_adapter *adapter = netdev_priv(netdev);
>>>>
>>>> /* stop the queue if it's getting full */
>>>> @@ -4553,17 +4501,29 @@ static int et131x_tx(struct sk_buff *skb,
>>>> struct
>>>> net_device *netdev)
>>>> /* Save the timestamp for the TX timeout watchdog */
>>>> netdev->trans_start = jiffies;
>>>>
>>>> - /* Call the device-specific data Tx routine */
>>>> - status = et131x_send_packets(skb, netdev);
>>>> + /* TCB is not available */
>>>> + if (adapter->tx_ring.used >= NUM_TCB)
>>>> + goto drop;
>>
>> That's wrong. You have to properly manage tx queue and not just drop
>> packets
>
> Could you kindly tell me what should be done to tx queue?
>
>>
>>>> - /* Check status and manage the netif queue if necessary */
>>>> - if (status != 0) {
>>>> - if (status == -ENOMEM)
>>>> - status = NETDEV_TX_BUSY;
>>>> - else
>>>> - status = NETDEV_TX_OK;
>>>> + /* We need to see if the link is up; if it's not, make the
>>>> + * netif layer think we're good and drop the packet
>>>> + */
>>>> + if ((adapter->flags & FMP_ADAPTER_FAIL_SEND_MASK) ||
>>>> + !netif_carrier_ok(netdev)) {
>>>> + goto drop;
>>>> + } else {
>>>> + if (!send_packet(skb, adapter))
>>>> + goto out;
>>>> + /* On error (send_packet returns != 0), make netif
>>>> + * think we're OK and drop the packet
>>>> + */
>>>> }
>>>> - return status;
>>>> +
>>>> +drop:
>>>> + dev_kfree_skb_any(skb);
>>>> + adapter->net_stats.tx_dropped++;
>>>> +out:
>>>> + return NETDEV_TX_OK;
>>>> }
>>>>
>>>> /* et131x_tx_timeout - Timeout handler
>>>> --
>>>> 1.8.3.1
>>>>
>>> _______________________________________________
>>> devel mailing list
>>> devel@...uxdriverproject.org
>>> http://driverdev.linuxdriverproject.org/mailman/listinfo/driverdev-devel
>>>
>>
>>
>> --
>> Regards,
>> Denis
>
--
Regards,
Denis
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
Powered by blists - more mailing lists