diff mbox

[1/1] IB/rxe: avoid double kfree_skb

Message ID 1524146512-4188-1-git-send-email-yanjun.zhu@oracle.com (mailing list archive)
State Accepted
Headers show

Commit Message

Zhu Yanjun April 19, 2018, 2:01 p.m. UTC
When skb is dropped by iptables rules, the skb is freed at the same time
-EPERM is returned. So in softroce, it is not necessary to free skb again.
Or else, crash will occur.

The steps to reproduce:

     server                       client
    ---------                    ---------
    |1.1.1.1|<----rxe-channel--->|1.1.1.2|
    ---------                    ---------

On server: rping -s -a 1.1.1.1 -v -C 10000 -S 512
On client: rping -c -a 1.1.1.1 -v -C 10000 -S 512

The kernel configs CONFIG_DEBUG_KMEMLEAK and
CONFIG_DEBUG_OBJECTS are enabled on both server and client.

When rping runs, run the following command in server:

iptables -I OUTPUT -p udp  --dport 4791 -j DROP

Without this patch, crash will occur.

CC: Srinivas Eeda <srinivas.eeda@oracle.com>
CC: Junxiao Bi <junxiao.bi@oracle.com>
Signed-off-by: Zhu Yanjun <yanjun.zhu@oracle.com>
Reviewed-by: Yuval Shaia <yuval.shaia@oracle.com>
---
 drivers/infiniband/sw/rxe/rxe_net.c  | 3 +++
 drivers/infiniband/sw/rxe/rxe_req.c  | 5 +++--
 drivers/infiniband/sw/rxe/rxe_resp.c | 9 ++++++---
 3 files changed, 12 insertions(+), 5 deletions(-)

Comments

Doug Ledford April 20, 2018, 2:19 a.m. UTC | #1
On Thu, 2018-04-19 at 10:01 -0400, Zhu Yanjun wrote:
> When skb is dropped by iptables rules, the skb is freed at the same time
> -EPERM is returned. So in softroce, it is not necessary to free skb again.
> Or else, crash will occur.
> 
> The steps to reproduce:
> 
>      server                       client
>     ---------                    ---------
>     |1.1.1.1|<----rxe-channel--->|1.1.1.2|
>     ---------                    ---------
> 
> On server: rping -s -a 1.1.1.1 -v -C 10000 -S 512
> On client: rping -c -a 1.1.1.1 -v -C 10000 -S 512
> 
> The kernel configs CONFIG_DEBUG_KMEMLEAK and
> CONFIG_DEBUG_OBJECTS are enabled on both server and client.
> 
> When rping runs, run the following command in server:
> 
> iptables -I OUTPUT -p udp  --dport 4791 -j DROP
> 
> Without this patch, crash will occur.
> 
> CC: Srinivas Eeda <srinivas.eeda@oracle.com>
> CC: Junxiao Bi <junxiao.bi@oracle.com>
> Signed-off-by: Zhu Yanjun <yanjun.zhu@oracle.com>
> Reviewed-by: Yuval Shaia <yuval.shaia@oracle.com>

I have no reason to doubt your analysis, but if there are a bunch of
error paths for net_xmit and they all return with your skb still being
valid and holding a reference, and then one oddball that returns with
your skb already gone, that just sounds like a mistake waiting to happen
(not to mention a bajillion special cases sprinkled everywhere to deal
with this apparent inconsistency).

Can we get a netdev@ confirmation on this being the right solution?

> ---
>  drivers/infiniband/sw/rxe/rxe_net.c  | 3 +++
>  drivers/infiniband/sw/rxe/rxe_req.c  | 5 +++--
>  drivers/infiniband/sw/rxe/rxe_resp.c | 9 ++++++---
>  3 files changed, 12 insertions(+), 5 deletions(-)
> 
> diff --git a/drivers/infiniband/sw/rxe/rxe_net.c b/drivers/infiniband/sw/rxe/rxe_net.c
> index 9da6e37..2094434 100644
> --- a/drivers/infiniband/sw/rxe/rxe_net.c
> +++ b/drivers/infiniband/sw/rxe/rxe_net.c
> @@ -511,6 +511,9 @@ int rxe_send(struct rxe_pkt_info *pkt, struct sk_buff *skb)
>  
>         if (unlikely(net_xmit_eval(err))) {
>                 pr_debug("error sending packet: %d\n", err);
> +               /* -EPERM means the skb is dropped and freed. */
> +               if (err == -EPERM)
> +                       return -EPERM;
>                 return -EAGAIN;
>         }
>  
> diff --git a/drivers/infiniband/sw/rxe/rxe_req.c b/drivers/infiniband/sw/rxe/rxe_req.c
> index 7bdaf71..9d2efec 100644
> --- a/drivers/infiniband/sw/rxe/rxe_req.c
> +++ b/drivers/infiniband/sw/rxe/rxe_req.c
> @@ -727,8 +727,9 @@ int rxe_requester(void *arg)
>  
>                 rollback_state(wqe, qp, &rollback_wqe, rollback_psn);
>  
> -               if (ret == -EAGAIN) {
> -                       kfree_skb(skb);
> +               if ((ret == -EAGAIN) || (ret == -EPERM)) {
> +                       if (ret == -EAGAIN)
> +                               kfree_skb(skb);
>                         rxe_run_task(&qp->req.task, 1);
>                         goto exit;
>                 }
> diff --git a/drivers/infiniband/sw/rxe/rxe_resp.c b/drivers/infiniband/sw/rxe/rxe_resp.c
> index a65c996..6bdf9b2 100644
> --- a/drivers/infiniband/sw/rxe/rxe_resp.c
> +++ b/drivers/infiniband/sw/rxe/rxe_resp.c
> @@ -742,7 +742,8 @@ static enum resp_states read_reply(struct rxe_qp *qp,
>         err = rxe_xmit_packet(rxe, qp, &ack_pkt, skb);
>         if (err) {
>                 pr_err("Failed sending RDMA reply.\n");
> -               kfree_skb(skb);
> +               if (err != -EPERM)
> +                       kfree_skb(skb);
>                 return RESPST_ERR_RNR;
>         }
>  
> @@ -956,7 +957,8 @@ static int send_ack(struct rxe_qp *qp, struct rxe_pkt_info *pkt,
>         err = rxe_xmit_packet(rxe, qp, &ack_pkt, skb);
>         if (err) {
>                 pr_err_ratelimited("Failed sending ack\n");
> -               kfree_skb(skb);
> +               if (err != -EPERM)
> +                       kfree_skb(skb);
>         }
>  
>  err1:
> @@ -1141,7 +1143,8 @@ static enum resp_states duplicate_request(struct rxe_qp *qp,
>                         if (rc) {
>                                 pr_err("Failed resending result. This flow is not handled - skb ignored\n");
>                                 rxe_drop_ref(qp);
> -                               kfree_skb(skb_copy);
> +                               if (rc != -EPERM)
> +                                       kfree_skb(skb_copy);
>                                 rc = RESPST_CLEANUP;
>                                 goto out;
>                         }
> -- 
> 2.7.4
>
Zhu Yanjun April 20, 2018, 5:46 a.m. UTC | #2
On 2018/4/20 10:19, Doug Ledford wrote:
> On Thu, 2018-04-19 at 10:01 -0400, Zhu Yanjun wrote:
>> When skb is dropped by iptables rules, the skb is freed at the same time
>> -EPERM is returned. So in softroce, it is not necessary to free skb again.
>> Or else, crash will occur.
>>
>> The steps to reproduce:
>>
>>       server                       client
>>      ---------                    ---------
>>      |1.1.1.1|<----rxe-channel--->|1.1.1.2|
>>      ---------                    ---------
>>
>> On server: rping -s -a 1.1.1.1 -v -C 10000 -S 512
>> On client: rping -c -a 1.1.1.1 -v -C 10000 -S 512
>>
>> The kernel configs CONFIG_DEBUG_KMEMLEAK and
>> CONFIG_DEBUG_OBJECTS are enabled on both server and client.
>>
>> When rping runs, run the following command in server:
>>
>> iptables -I OUTPUT -p udp  --dport 4791 -j DROP
>>
>> Without this patch, crash will occur.
>>
>> CC: Srinivas Eeda <srinivas.eeda@oracle.com>
>> CC: Junxiao Bi <junxiao.bi@oracle.com>
>> Signed-off-by: Zhu Yanjun <yanjun.zhu@oracle.com>
>> Reviewed-by: Yuval Shaia <yuval.shaia@oracle.com>
> I have no reason to doubt your analysis, but if there are a bunch of
> error paths for net_xmit and they all return with your skb still being
> valid and holding a reference, and then one oddball that returns with
> your skb already gone, that just sounds like a mistake waiting to happen
> (not to mention a bajillion special cases sprinkled everywhere to deal
> with this apparent inconsistency).
>
> Can we get a netdev@ confirmation on this being the right solution?
Yes. I agree with you.
After iptables rule "iptables -I OUTPUT -p udp  --dport 4791 -j DROP", 
the skb is freed in this function

/* Returns 1 if okfn() needs to be executed by the caller,
  * -EPERM for NF_DROP, 0 otherwise.  Caller must hold rcu_read_lock. */
int nf_hook_slow(struct sk_buff *skb, struct nf_hook_state *state,
                  const struct nf_hook_entries *e, unsigned int s)
{
         unsigned int verdict;
         int ret;

         for (; s < e->num_hook_entries; s++) {
                 verdict = nf_hook_entry_hookfn(&e->hooks[s], skb, state);
                 switch (verdict & NF_VERDICT_MASK) {
                 case NF_ACCEPT:
                         break;
                 case NF_DROP:
kfree_skb(skb);                               <----here, skb is freed
                         ret = NF_DROP_GETERR(verdict);
                         if (ret == 0)
                                 ret = -EPERM;
                         return ret;
                 case NF_QUEUE:
                         ret = nf_queue(skb, state, e, s, verdict);
                         if (ret == 1)
                                 continue;
                         return ret;
                 default:
                         /* Implicit handling for NF_STOLEN, as well as 
any other
                          * non conventional verdicts.
                          */
                         return 0;
                 }
         }

         return 1;
}
EXPORT_SYMBOL(nf_hook_slow);

If I am wrong, please correct me.

And my test environment is still there, any solution can be verified in it.

Zhu Yanjun
>
>> ---
>>   drivers/infiniband/sw/rxe/rxe_net.c  | 3 +++
>>   drivers/infiniband/sw/rxe/rxe_req.c  | 5 +++--
>>   drivers/infiniband/sw/rxe/rxe_resp.c | 9 ++++++---
>>   3 files changed, 12 insertions(+), 5 deletions(-)
>>
>> diff --git a/drivers/infiniband/sw/rxe/rxe_net.c b/drivers/infiniband/sw/rxe/rxe_net.c
>> index 9da6e37..2094434 100644
>> --- a/drivers/infiniband/sw/rxe/rxe_net.c
>> +++ b/drivers/infiniband/sw/rxe/rxe_net.c
>> @@ -511,6 +511,9 @@ int rxe_send(struct rxe_pkt_info *pkt, struct sk_buff *skb)
>>   
>>          if (unlikely(net_xmit_eval(err))) {
>>                  pr_debug("error sending packet: %d\n", err);
>> +               /* -EPERM means the skb is dropped and freed. */
>> +               if (err == -EPERM)
>> +                       return -EPERM;
>>                  return -EAGAIN;
>>          }
>>   
>> diff --git a/drivers/infiniband/sw/rxe/rxe_req.c b/drivers/infiniband/sw/rxe/rxe_req.c
>> index 7bdaf71..9d2efec 100644
>> --- a/drivers/infiniband/sw/rxe/rxe_req.c
>> +++ b/drivers/infiniband/sw/rxe/rxe_req.c
>> @@ -727,8 +727,9 @@ int rxe_requester(void *arg)
>>   
>>                  rollback_state(wqe, qp, &rollback_wqe, rollback_psn);
>>   
>> -               if (ret == -EAGAIN) {
>> -                       kfree_skb(skb);
>> +               if ((ret == -EAGAIN) || (ret == -EPERM)) {
>> +                       if (ret == -EAGAIN)
>> +                               kfree_skb(skb);
>>                          rxe_run_task(&qp->req.task, 1);
>>                          goto exit;
>>                  }
>> diff --git a/drivers/infiniband/sw/rxe/rxe_resp.c b/drivers/infiniband/sw/rxe/rxe_resp.c
>> index a65c996..6bdf9b2 100644
>> --- a/drivers/infiniband/sw/rxe/rxe_resp.c
>> +++ b/drivers/infiniband/sw/rxe/rxe_resp.c
>> @@ -742,7 +742,8 @@ static enum resp_states read_reply(struct rxe_qp *qp,
>>          err = rxe_xmit_packet(rxe, qp, &ack_pkt, skb);
>>          if (err) {
>>                  pr_err("Failed sending RDMA reply.\n");
>> -               kfree_skb(skb);
>> +               if (err != -EPERM)
>> +                       kfree_skb(skb);
>>                  return RESPST_ERR_RNR;
>>          }
>>   
>> @@ -956,7 +957,8 @@ static int send_ack(struct rxe_qp *qp, struct rxe_pkt_info *pkt,
>>          err = rxe_xmit_packet(rxe, qp, &ack_pkt, skb);
>>          if (err) {
>>                  pr_err_ratelimited("Failed sending ack\n");
>> -               kfree_skb(skb);
>> +               if (err != -EPERM)
>> +                       kfree_skb(skb);
>>          }
>>   
>>   err1:
>> @@ -1141,7 +1143,8 @@ static enum resp_states duplicate_request(struct rxe_qp *qp,
>>                          if (rc) {
>>                                  pr_err("Failed resending result. This flow is not handled - skb ignored\n");
>>                                  rxe_drop_ref(qp);
>> -                               kfree_skb(skb_copy);
>> +                               if (rc != -EPERM)
>> +                                       kfree_skb(skb_copy);
>>                                  rc = RESPST_CLEANUP;
>>                                  goto out;
>>                          }
>> -- 
>> 2.7.4
>>

--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Zhu Yanjun April 24, 2018, 8:34 a.m. UTC | #3
Hi, all

rxe_send
     ip_local_out
         __ip_local_out
             nf_hook_slow

In the above call process, nf_hook_slow drops and frees skb, then -EPERM 
is returned when iptables rules(iptables -I OUTPUT -p udp --dport 4791 
-j DROP) is set.

If skb->users is not changed in softroce, kfree_skb should not be called 
in this module.

I will make further investigations about other error handler after 
ip_local_out.
If I am wrong, please correct me.

Any reply is appreciated.

Zhu Yanjun
On 2018/4/20 13:46, Yanjun Zhu wrote:
>
>
> On 2018/4/20 10:19, Doug Ledford wrote:
>> On Thu, 2018-04-19 at 10:01 -0400, Zhu Yanjun wrote:
>>> When skb is dropped by iptables rules, the skb is freed at the same 
>>> time
>>> -EPERM is returned. So in softroce, it is not necessary to free skb 
>>> again.
>>> Or else, crash will occur.
>>>
>>> The steps to reproduce:
>>>
>>>       server                       client
>>>      ---------                    ---------
>>>      |1.1.1.1|<----rxe-channel--->|1.1.1.2|
>>>      ---------                    ---------
>>>
>>> On server: rping -s -a 1.1.1.1 -v -C 10000 -S 512
>>> On client: rping -c -a 1.1.1.1 -v -C 10000 -S 512
>>>
>>> The kernel configs CONFIG_DEBUG_KMEMLEAK and
>>> CONFIG_DEBUG_OBJECTS are enabled on both server and client.
>>>
>>> When rping runs, run the following command in server:
>>>
>>> iptables -I OUTPUT -p udp  --dport 4791 -j DROP
>>>
>>> Without this patch, crash will occur.
>>>
>>> CC: Srinivas Eeda <srinivas.eeda@oracle.com>
>>> CC: Junxiao Bi <junxiao.bi@oracle.com>
>>> Signed-off-by: Zhu Yanjun <yanjun.zhu@oracle.com>
>>> Reviewed-by: Yuval Shaia <yuval.shaia@oracle.com>
>> I have no reason to doubt your analysis, but if there are a bunch of
>> error paths for net_xmit and they all return with your skb still being
>> valid and holding a reference, and then one oddball that returns with
>> your skb already gone, that just sounds like a mistake waiting to happen
>> (not to mention a bajillion special cases sprinkled everywhere to deal
>> with this apparent inconsistency).
>>
>> Can we get a netdev@ confirmation on this being the right solution?
> Yes. I agree with you.
> After iptables rule "iptables -I OUTPUT -p udp  --dport 4791 -j DROP", 
> the skb is freed in this function
>
> /* Returns 1 if okfn() needs to be executed by the caller,
>  * -EPERM for NF_DROP, 0 otherwise.  Caller must hold rcu_read_lock. */
> int nf_hook_slow(struct sk_buff *skb, struct nf_hook_state *state,
>                  const struct nf_hook_entries *e, unsigned int s)
> {
>         unsigned int verdict;
>         int ret;
>
>         for (; s < e->num_hook_entries; s++) {
>                 verdict = nf_hook_entry_hookfn(&e->hooks[s], skb, state);
>                 switch (verdict & NF_VERDICT_MASK) {
>                 case NF_ACCEPT:
>                         break;
>                 case NF_DROP:
> kfree_skb(skb);                               <----here, skb is freed
>                         ret = NF_DROP_GETERR(verdict);
>                         if (ret == 0)
>                                 ret = -EPERM;
>                         return ret;
>                 case NF_QUEUE:
>                         ret = nf_queue(skb, state, e, s, verdict);
>                         if (ret == 1)
>                                 continue;
>                         return ret;
>                 default:
>                         /* Implicit handling for NF_STOLEN, as well as 
> any other
>                          * non conventional verdicts.
>                          */
>                         return 0;
>                 }
>         }
>
>         return 1;
> }
> EXPORT_SYMBOL(nf_hook_slow);
>
> If I am wrong, please correct me.
>
> And my test environment is still there, any solution can be verified 
> in it.
>
> Zhu Yanjun
>>
>>> ---
>>>   drivers/infiniband/sw/rxe/rxe_net.c  | 3 +++
>>>   drivers/infiniband/sw/rxe/rxe_req.c  | 5 +++--
>>>   drivers/infiniband/sw/rxe/rxe_resp.c | 9 ++++++---
>>>   3 files changed, 12 insertions(+), 5 deletions(-)
>>>
>>> diff --git a/drivers/infiniband/sw/rxe/rxe_net.c 
>>> b/drivers/infiniband/sw/rxe/rxe_net.c
>>> index 9da6e37..2094434 100644
>>> --- a/drivers/infiniband/sw/rxe/rxe_net.c
>>> +++ b/drivers/infiniband/sw/rxe/rxe_net.c
>>> @@ -511,6 +511,9 @@ int rxe_send(struct rxe_pkt_info *pkt, struct 
>>> sk_buff *skb)
>>>            if (unlikely(net_xmit_eval(err))) {
>>>                  pr_debug("error sending packet: %d\n", err);
>>> +               /* -EPERM means the skb is dropped and freed. */
>>> +               if (err == -EPERM)
>>> +                       return -EPERM;
>>>                  return -EAGAIN;
>>>          }
>>>   diff --git a/drivers/infiniband/sw/rxe/rxe_req.c 
>>> b/drivers/infiniband/sw/rxe/rxe_req.c
>>> index 7bdaf71..9d2efec 100644
>>> --- a/drivers/infiniband/sw/rxe/rxe_req.c
>>> +++ b/drivers/infiniband/sw/rxe/rxe_req.c
>>> @@ -727,8 +727,9 @@ int rxe_requester(void *arg)
>>>                    rollback_state(wqe, qp, &rollback_wqe, 
>>> rollback_psn);
>>>   -               if (ret == -EAGAIN) {
>>> -                       kfree_skb(skb);
>>> +               if ((ret == -EAGAIN) || (ret == -EPERM)) {
>>> +                       if (ret == -EAGAIN)
>>> +                               kfree_skb(skb);
>>>                          rxe_run_task(&qp->req.task, 1);
>>>                          goto exit;
>>>                  }
>>> diff --git a/drivers/infiniband/sw/rxe/rxe_resp.c 
>>> b/drivers/infiniband/sw/rxe/rxe_resp.c
>>> index a65c996..6bdf9b2 100644
>>> --- a/drivers/infiniband/sw/rxe/rxe_resp.c
>>> +++ b/drivers/infiniband/sw/rxe/rxe_resp.c
>>> @@ -742,7 +742,8 @@ static enum resp_states read_reply(struct rxe_qp 
>>> *qp,
>>>          err = rxe_xmit_packet(rxe, qp, &ack_pkt, skb);
>>>          if (err) {
>>>                  pr_err("Failed sending RDMA reply.\n");
>>> -               kfree_skb(skb);
>>> +               if (err != -EPERM)
>>> +                       kfree_skb(skb);
>>>                  return RESPST_ERR_RNR;
>>>          }
>>>   @@ -956,7 +957,8 @@ static int send_ack(struct rxe_qp *qp, struct 
>>> rxe_pkt_info *pkt,
>>>          err = rxe_xmit_packet(rxe, qp, &ack_pkt, skb);
>>>          if (err) {
>>>                  pr_err_ratelimited("Failed sending ack\n");
>>> -               kfree_skb(skb);
>>> +               if (err != -EPERM)
>>> +                       kfree_skb(skb);
>>>          }
>>>     err1:
>>> @@ -1141,7 +1143,8 @@ static enum resp_states 
>>> duplicate_request(struct rxe_qp *qp,
>>>                          if (rc) {
>>>                                  pr_err("Failed resending result. 
>>> This flow is not handled - skb ignored\n");
>>>                                  rxe_drop_ref(qp);
>>> -                               kfree_skb(skb_copy);
>>> +                               if (rc != -EPERM)
>>> +                                       kfree_skb(skb_copy);
>>>                                  rc = RESPST_CLEANUP;
>>>                                  goto out;
>>>                          }
>>> -- 
>>> 2.7.4
>>>
>
> -- 
> To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html

--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Zhu Yanjun April 25, 2018, 6:56 a.m. UTC | #4
Hi, all

rxe_send [rdma_rxe]
     ip_local_out
         __ip_local_out
         ip_output
             ip_finish_output
                 ip_finish_output2
                     dev_queue_xmit
                         __dev_queue_xmit
                             dev_hard_start_xmit
                                 e1000_xmit_frame [e1000]

When skb is sent, it will pass the above functions. I checked all the 
above functions. If error occurs in the above functions after 
ip_local_out, kfree_skb will be called.
So when ip_local_out returns an error, skb should be freed. It is not 
necessary to call kfree_skb in soft roce module again.

If I am wrong, please correct me.

Zhu Yanjun
On 2018/4/24 16:34, Yanjun Zhu wrote:
> Hi, all
>
> rxe_send
>     ip_local_out
>         __ip_local_out
>             nf_hook_slow
>
> In the above call process, nf_hook_slow drops and frees skb, then 
> -EPERM is returned when iptables rules(iptables -I OUTPUT -p udp 
> --dport 4791 -j DROP) is set.
>
> If skb->users is not changed in softroce, kfree_skb should not be 
> called in this module.
>
> I will make further investigations about other error handler after 
> ip_local_out.
> If I am wrong, please correct me.
>
> Any reply is appreciated.
>
> Zhu Yanjun
> On 2018/4/20 13:46, Yanjun Zhu wrote:
>>
>>
>> On 2018/4/20 10:19, Doug Ledford wrote:
>>> On Thu, 2018-04-19 at 10:01 -0400, Zhu Yanjun wrote:
>>>> When skb is dropped by iptables rules, the skb is freed at the same 
>>>> time
>>>> -EPERM is returned. So in softroce, it is not necessary to free skb 
>>>> again.
>>>> Or else, crash will occur.
>>>>
>>>> The steps to reproduce:
>>>>
>>>>       server                       client
>>>>      ---------                    ---------
>>>>      |1.1.1.1|<----rxe-channel--->|1.1.1.2|
>>>>      ---------                    ---------
>>>>
>>>> On server: rping -s -a 1.1.1.1 -v -C 10000 -S 512
>>>> On client: rping -c -a 1.1.1.1 -v -C 10000 -S 512
>>>>
>>>> The kernel configs CONFIG_DEBUG_KMEMLEAK and
>>>> CONFIG_DEBUG_OBJECTS are enabled on both server and client.
>>>>
>>>> When rping runs, run the following command in server:
>>>>
>>>> iptables -I OUTPUT -p udp  --dport 4791 -j DROP
>>>>
>>>> Without this patch, crash will occur.
>>>>
>>>> CC: Srinivas Eeda <srinivas.eeda@oracle.com>
>>>> CC: Junxiao Bi <junxiao.bi@oracle.com>
>>>> Signed-off-by: Zhu Yanjun <yanjun.zhu@oracle.com>
>>>> Reviewed-by: Yuval Shaia <yuval.shaia@oracle.com>
>>> I have no reason to doubt your analysis, but if there are a bunch of
>>> error paths for net_xmit and they all return with your skb still being
>>> valid and holding a reference, and then one oddball that returns with
>>> your skb already gone, that just sounds like a mistake waiting to 
>>> happen
>>> (not to mention a bajillion special cases sprinkled everywhere to deal
>>> with this apparent inconsistency).
>>>
>>> Can we get a netdev@ confirmation on this being the right solution?
>> Yes. I agree with you.
>> After iptables rule "iptables -I OUTPUT -p udp  --dport 4791 -j 
>> DROP", the skb is freed in this function
>>
>> /* Returns 1 if okfn() needs to be executed by the caller,
>>  * -EPERM for NF_DROP, 0 otherwise.  Caller must hold rcu_read_lock. */
>> int nf_hook_slow(struct sk_buff *skb, struct nf_hook_state *state,
>>                  const struct nf_hook_entries *e, unsigned int s)
>> {
>>         unsigned int verdict;
>>         int ret;
>>
>>         for (; s < e->num_hook_entries; s++) {
>>                 verdict = nf_hook_entry_hookfn(&e->hooks[s], skb, 
>> state);
>>                 switch (verdict & NF_VERDICT_MASK) {
>>                 case NF_ACCEPT:
>>                         break;
>>                 case NF_DROP:
>> kfree_skb(skb);                               <----here, skb is freed
>>                         ret = NF_DROP_GETERR(verdict);
>>                         if (ret == 0)
>>                                 ret = -EPERM;
>>                         return ret;
>>                 case NF_QUEUE:
>>                         ret = nf_queue(skb, state, e, s, verdict);
>>                         if (ret == 1)
>>                                 continue;
>>                         return ret;
>>                 default:
>>                         /* Implicit handling for NF_STOLEN, as well 
>> as any other
>>                          * non conventional verdicts.
>>                          */
>>                         return 0;
>>                 }
>>         }
>>
>>         return 1;
>> }
>> EXPORT_SYMBOL(nf_hook_slow);
>>
>> If I am wrong, please correct me.
>>
>> And my test environment is still there, any solution can be verified 
>> in it.
>>
>> Zhu Yanjun
>>>
>>>> ---
>>>>   drivers/infiniband/sw/rxe/rxe_net.c  | 3 +++
>>>>   drivers/infiniband/sw/rxe/rxe_req.c  | 5 +++--
>>>>   drivers/infiniband/sw/rxe/rxe_resp.c | 9 ++++++---
>>>>   3 files changed, 12 insertions(+), 5 deletions(-)
>>>>
>>>> diff --git a/drivers/infiniband/sw/rxe/rxe_net.c 
>>>> b/drivers/infiniband/sw/rxe/rxe_net.c
>>>> index 9da6e37..2094434 100644
>>>> --- a/drivers/infiniband/sw/rxe/rxe_net.c
>>>> +++ b/drivers/infiniband/sw/rxe/rxe_net.c
>>>> @@ -511,6 +511,9 @@ int rxe_send(struct rxe_pkt_info *pkt, struct 
>>>> sk_buff *skb)
>>>>            if (unlikely(net_xmit_eval(err))) {
>>>>                  pr_debug("error sending packet: %d\n", err);
>>>> +               /* -EPERM means the skb is dropped and freed. */
>>>> +               if (err == -EPERM)
>>>> +                       return -EPERM;
>>>>                  return -EAGAIN;
>>>>          }
>>>>   diff --git a/drivers/infiniband/sw/rxe/rxe_req.c 
>>>> b/drivers/infiniband/sw/rxe/rxe_req.c
>>>> index 7bdaf71..9d2efec 100644
>>>> --- a/drivers/infiniband/sw/rxe/rxe_req.c
>>>> +++ b/drivers/infiniband/sw/rxe/rxe_req.c
>>>> @@ -727,8 +727,9 @@ int rxe_requester(void *arg)
>>>>                    rollback_state(wqe, qp, &rollback_wqe, 
>>>> rollback_psn);
>>>>   -               if (ret == -EAGAIN) {
>>>> -                       kfree_skb(skb);
>>>> +               if ((ret == -EAGAIN) || (ret == -EPERM)) {
>>>> +                       if (ret == -EAGAIN)
>>>> +                               kfree_skb(skb);
>>>>                          rxe_run_task(&qp->req.task, 1);
>>>>                          goto exit;
>>>>                  }
>>>> diff --git a/drivers/infiniband/sw/rxe/rxe_resp.c 
>>>> b/drivers/infiniband/sw/rxe/rxe_resp.c
>>>> index a65c996..6bdf9b2 100644
>>>> --- a/drivers/infiniband/sw/rxe/rxe_resp.c
>>>> +++ b/drivers/infiniband/sw/rxe/rxe_resp.c
>>>> @@ -742,7 +742,8 @@ static enum resp_states read_reply(struct 
>>>> rxe_qp *qp,
>>>>          err = rxe_xmit_packet(rxe, qp, &ack_pkt, skb);
>>>>          if (err) {
>>>>                  pr_err("Failed sending RDMA reply.\n");
>>>> -               kfree_skb(skb);
>>>> +               if (err != -EPERM)
>>>> +                       kfree_skb(skb);
>>>>                  return RESPST_ERR_RNR;
>>>>          }
>>>>   @@ -956,7 +957,8 @@ static int send_ack(struct rxe_qp *qp, struct 
>>>> rxe_pkt_info *pkt,
>>>>          err = rxe_xmit_packet(rxe, qp, &ack_pkt, skb);
>>>>          if (err) {
>>>>                  pr_err_ratelimited("Failed sending ack\n");
>>>> -               kfree_skb(skb);
>>>> +               if (err != -EPERM)
>>>> +                       kfree_skb(skb);
>>>>          }
>>>>     err1:
>>>> @@ -1141,7 +1143,8 @@ static enum resp_states 
>>>> duplicate_request(struct rxe_qp *qp,
>>>>                          if (rc) {
>>>>                                  pr_err("Failed resending result. 
>>>> This flow is not handled - skb ignored\n");
>>>>                                  rxe_drop_ref(qp);
>>>> -                               kfree_skb(skb_copy);
>>>> +                               if (rc != -EPERM)
>>>> +                                       kfree_skb(skb_copy);
>>>>                                  rc = RESPST_CLEANUP;
>>>>                                  goto out;
>>>>                          }
>>>> -- 
>>>> 2.7.4
>>>>
>>
>> -- 
>> To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
>> the body of a message to majordomo@vger.kernel.org
>> More majordomo info at http://vger.kernel.org/majordomo-info.html
>

--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Doug Ledford April 27, 2018, 4:24 p.m. UTC | #5
On Wed, 2018-04-25 at 14:56 +0800, Yanjun Zhu wrote:
> Hi, all
> 
> rxe_send [rdma_rxe]
>      ip_local_out
>          __ip_local_out
>          ip_output
>              ip_finish_output
>                  ip_finish_output2
>                      dev_queue_xmit
>                          __dev_queue_xmit
>                              dev_hard_start_xmit
>                                  e1000_xmit_frame [e1000]
> 
> When skb is sent, it will pass the above functions. I checked all the 
> above functions. If error occurs in the above functions after 
> ip_local_out, kfree_skb will be called.
> So when ip_local_out returns an error, skb should be freed. It is not 
> necessary to call kfree_skb in soft roce module again.
> 
> If I am wrong, please correct me.

No one from netdev has spoken up, and I don't see where you are wrong,
so I've applied this to for-rc.  Thanks.

> Zhu Yanjun
> On 2018/4/24 16:34, Yanjun Zhu wrote:
> > Hi, all
> > 
> > rxe_send
> >     ip_local_out
> >         __ip_local_out
> >             nf_hook_slow
> > 
> > In the above call process, nf_hook_slow drops and frees skb, then 
> > -EPERM is returned when iptables rules(iptables -I OUTPUT -p udp 
> > --dport 4791 -j DROP) is set.
> > 
> > If skb->users is not changed in softroce, kfree_skb should not be 
> > called in this module.
> > 
> > I will make further investigations about other error handler after 
> > ip_local_out.
> > If I am wrong, please correct me.
> > 
> > Any reply is appreciated.
> > 
> > Zhu Yanjun
> > On 2018/4/20 13:46, Yanjun Zhu wrote:
> > > 
> > > 
> > > On 2018/4/20 10:19, Doug Ledford wrote:
> > > > On Thu, 2018-04-19 at 10:01 -0400, Zhu Yanjun wrote:
> > > > > When skb is dropped by iptables rules, the skb is freed at the same 
> > > > > time
> > > > > -EPERM is returned. So in softroce, it is not necessary to free skb 
> > > > > again.
> > > > > Or else, crash will occur.
> > > > > 
> > > > > The steps to reproduce:
> > > > > 
> > > > >       server                       client
> > > > >      ---------                    ---------
> > > > >      |1.1.1.1|<----rxe-channel--->|1.1.1.2|
> > > > >      ---------                    ---------
> > > > > 
> > > > > On server: rping -s -a 1.1.1.1 -v -C 10000 -S 512
> > > > > On client: rping -c -a 1.1.1.1 -v -C 10000 -S 512
> > > > > 
> > > > > The kernel configs CONFIG_DEBUG_KMEMLEAK and
> > > > > CONFIG_DEBUG_OBJECTS are enabled on both server and client.
> > > > > 
> > > > > When rping runs, run the following command in server:
> > > > > 
> > > > > iptables -I OUTPUT -p udp  --dport 4791 -j DROP
> > > > > 
> > > > > Without this patch, crash will occur.
> > > > > 
> > > > > CC: Srinivas Eeda <srinivas.eeda@oracle.com>
> > > > > CC: Junxiao Bi <junxiao.bi@oracle.com>
> > > > > Signed-off-by: Zhu Yanjun <yanjun.zhu@oracle.com>
> > > > > Reviewed-by: Yuval Shaia <yuval.shaia@oracle.com>
> > > > 
> > > > I have no reason to doubt your analysis, but if there are a bunch of
> > > > error paths for net_xmit and they all return with your skb still being
> > > > valid and holding a reference, and then one oddball that returns with
> > > > your skb already gone, that just sounds like a mistake waiting to 
> > > > happen
> > > > (not to mention a bajillion special cases sprinkled everywhere to deal
> > > > with this apparent inconsistency).
> > > > 
> > > > Can we get a netdev@ confirmation on this being the right solution?
> > > 
> > > Yes. I agree with you.
> > > After iptables rule "iptables -I OUTPUT -p udp  --dport 4791 -j 
> > > DROP", the skb is freed in this function
> > > 
> > > /* Returns 1 if okfn() needs to be executed by the caller,
> > >  * -EPERM for NF_DROP, 0 otherwise.  Caller must hold rcu_read_lock. */
> > > int nf_hook_slow(struct sk_buff *skb, struct nf_hook_state *state,
> > >                  const struct nf_hook_entries *e, unsigned int s)
> > > {
> > >         unsigned int verdict;
> > >         int ret;
> > > 
> > >         for (; s < e->num_hook_entries; s++) {
> > >                 verdict = nf_hook_entry_hookfn(&e->hooks[s], skb, 
> > > state);
> > >                 switch (verdict & NF_VERDICT_MASK) {
> > >                 case NF_ACCEPT:
> > >                         break;
> > >                 case NF_DROP:
> > > kfree_skb(skb);                               <----here, skb is freed
> > >                         ret = NF_DROP_GETERR(verdict);
> > >                         if (ret == 0)
> > >                                 ret = -EPERM;
> > >                         return ret;
> > >                 case NF_QUEUE:
> > >                         ret = nf_queue(skb, state, e, s, verdict);
> > >                         if (ret == 1)
> > >                                 continue;
> > >                         return ret;
> > >                 default:
> > >                         /* Implicit handling for NF_STOLEN, as well 
> > > as any other
> > >                          * non conventional verdicts.
> > >                          */
> > >                         return 0;
> > >                 }
> > >         }
> > > 
> > >         return 1;
> > > }
> > > EXPORT_SYMBOL(nf_hook_slow);
> > > 
> > > If I am wrong, please correct me.
> > > 
> > > And my test environment is still there, any solution can be verified 
> > > in it.
> > > 
> > > Zhu Yanjun
> > > > 
> > > > > ---
> > > > >   drivers/infiniband/sw/rxe/rxe_net.c  | 3 +++
> > > > >   drivers/infiniband/sw/rxe/rxe_req.c  | 5 +++--
> > > > >   drivers/infiniband/sw/rxe/rxe_resp.c | 9 ++++++---
> > > > >   3 files changed, 12 insertions(+), 5 deletions(-)
> > > > > 
> > > > > diff --git a/drivers/infiniband/sw/rxe/rxe_net.c 
> > > > > b/drivers/infiniband/sw/rxe/rxe_net.c
> > > > > index 9da6e37..2094434 100644
> > > > > --- a/drivers/infiniband/sw/rxe/rxe_net.c
> > > > > +++ b/drivers/infiniband/sw/rxe/rxe_net.c
> > > > > @@ -511,6 +511,9 @@ int rxe_send(struct rxe_pkt_info *pkt, struct 
> > > > > sk_buff *skb)
> > > > >            if (unlikely(net_xmit_eval(err))) {
> > > > >                  pr_debug("error sending packet: %d\n", err);
> > > > > +               /* -EPERM means the skb is dropped and freed. */
> > > > > +               if (err == -EPERM)
> > > > > +                       return -EPERM;
> > > > >                  return -EAGAIN;
> > > > >          }
> > > > >   diff --git a/drivers/infiniband/sw/rxe/rxe_req.c 
> > > > > b/drivers/infiniband/sw/rxe/rxe_req.c
> > > > > index 7bdaf71..9d2efec 100644
> > > > > --- a/drivers/infiniband/sw/rxe/rxe_req.c
> > > > > +++ b/drivers/infiniband/sw/rxe/rxe_req.c
> > > > > @@ -727,8 +727,9 @@ int rxe_requester(void *arg)
> > > > >                    rollback_state(wqe, qp, &rollback_wqe, 
> > > > > rollback_psn);
> > > > >   -               if (ret == -EAGAIN) {
> > > > > -                       kfree_skb(skb);
> > > > > +               if ((ret == -EAGAIN) || (ret == -EPERM)) {
> > > > > +                       if (ret == -EAGAIN)
> > > > > +                               kfree_skb(skb);
> > > > >                          rxe_run_task(&qp->req.task, 1);
> > > > >                          goto exit;
> > > > >                  }
> > > > > diff --git a/drivers/infiniband/sw/rxe/rxe_resp.c 
> > > > > b/drivers/infiniband/sw/rxe/rxe_resp.c
> > > > > index a65c996..6bdf9b2 100644
> > > > > --- a/drivers/infiniband/sw/rxe/rxe_resp.c
> > > > > +++ b/drivers/infiniband/sw/rxe/rxe_resp.c
> > > > > @@ -742,7 +742,8 @@ static enum resp_states read_reply(struct 
> > > > > rxe_qp *qp,
> > > > >          err = rxe_xmit_packet(rxe, qp, &ack_pkt, skb);
> > > > >          if (err) {
> > > > >                  pr_err("Failed sending RDMA reply.\n");
> > > > > -               kfree_skb(skb);
> > > > > +               if (err != -EPERM)
> > > > > +                       kfree_skb(skb);
> > > > >                  return RESPST_ERR_RNR;
> > > > >          }
> > > > >   @@ -956,7 +957,8 @@ static int send_ack(struct rxe_qp *qp, struct 
> > > > > rxe_pkt_info *pkt,
> > > > >          err = rxe_xmit_packet(rxe, qp, &ack_pkt, skb);
> > > > >          if (err) {
> > > > >                  pr_err_ratelimited("Failed sending ack\n");
> > > > > -               kfree_skb(skb);
> > > > > +               if (err != -EPERM)
> > > > > +                       kfree_skb(skb);
> > > > >          }
> > > > >     err1:
> > > > > @@ -1141,7 +1143,8 @@ static enum resp_states 
> > > > > duplicate_request(struct rxe_qp *qp,
> > > > >                          if (rc) {
> > > > >                                  pr_err("Failed resending result. 
> > > > > This flow is not handled - skb ignored\n");
> > > > >                                  rxe_drop_ref(qp);
> > > > > -                               kfree_skb(skb_copy);
> > > > > +                               if (rc != -EPERM)
> > > > > +                                       kfree_skb(skb_copy);
> > > > >                                  rc = RESPST_CLEANUP;
> > > > >                                  goto out;
> > > > >                          }
> > > > > -- 
> > > > > 2.7.4
> > > > > 
> > > 
> > > -- 
> > > To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
> > > the body of a message to majordomo@vger.kernel.org
> > > More majordomo info at http://vger.kernel.org/majordomo-info.html
> 
>
diff mbox

Patch

diff --git a/drivers/infiniband/sw/rxe/rxe_net.c b/drivers/infiniband/sw/rxe/rxe_net.c
index 9da6e37..2094434 100644
--- a/drivers/infiniband/sw/rxe/rxe_net.c
+++ b/drivers/infiniband/sw/rxe/rxe_net.c
@@ -511,6 +511,9 @@  int rxe_send(struct rxe_pkt_info *pkt, struct sk_buff *skb)
 
 	if (unlikely(net_xmit_eval(err))) {
 		pr_debug("error sending packet: %d\n", err);
+		/* -EPERM means the skb is dropped and freed. */
+		if (err == -EPERM)
+			return -EPERM;
 		return -EAGAIN;
 	}
 
diff --git a/drivers/infiniband/sw/rxe/rxe_req.c b/drivers/infiniband/sw/rxe/rxe_req.c
index 7bdaf71..9d2efec 100644
--- a/drivers/infiniband/sw/rxe/rxe_req.c
+++ b/drivers/infiniband/sw/rxe/rxe_req.c
@@ -727,8 +727,9 @@  int rxe_requester(void *arg)
 
 		rollback_state(wqe, qp, &rollback_wqe, rollback_psn);
 
-		if (ret == -EAGAIN) {
-			kfree_skb(skb);
+		if ((ret == -EAGAIN) || (ret == -EPERM)) {
+			if (ret == -EAGAIN)
+				kfree_skb(skb);
 			rxe_run_task(&qp->req.task, 1);
 			goto exit;
 		}
diff --git a/drivers/infiniband/sw/rxe/rxe_resp.c b/drivers/infiniband/sw/rxe/rxe_resp.c
index a65c996..6bdf9b2 100644
--- a/drivers/infiniband/sw/rxe/rxe_resp.c
+++ b/drivers/infiniband/sw/rxe/rxe_resp.c
@@ -742,7 +742,8 @@  static enum resp_states read_reply(struct rxe_qp *qp,
 	err = rxe_xmit_packet(rxe, qp, &ack_pkt, skb);
 	if (err) {
 		pr_err("Failed sending RDMA reply.\n");
-		kfree_skb(skb);
+		if (err != -EPERM)
+			kfree_skb(skb);
 		return RESPST_ERR_RNR;
 	}
 
@@ -956,7 +957,8 @@  static int send_ack(struct rxe_qp *qp, struct rxe_pkt_info *pkt,
 	err = rxe_xmit_packet(rxe, qp, &ack_pkt, skb);
 	if (err) {
 		pr_err_ratelimited("Failed sending ack\n");
-		kfree_skb(skb);
+		if (err != -EPERM)
+			kfree_skb(skb);
 	}
 
 err1:
@@ -1141,7 +1143,8 @@  static enum resp_states duplicate_request(struct rxe_qp *qp,
 			if (rc) {
 				pr_err("Failed resending result. This flow is not handled - skb ignored\n");
 				rxe_drop_ref(qp);
-				kfree_skb(skb_copy);
+				if (rc != -EPERM)
+					kfree_skb(skb_copy);
 				rc = RESPST_CLEANUP;
 				goto out;
 			}