[net,v3] ixgbe: fix double clean of tx descriptors with xdp
diff mbox series

Message ID 20190822171237.20798-1-i.maximets@samsung.com
State Accepted
Delegated to: Jeff Kirsher
Headers show
Series
  • [net,v3] ixgbe: fix double clean of tx descriptors with xdp
Related show

Commit Message

Ilya Maximets Aug. 22, 2019, 5:12 p.m. UTC
Tx code doesn't clear the descriptors' status after cleaning.
So, if the budget is larger than number of used elems in a ring, some
descriptors will be accounted twice and xsk_umem_complete_tx will move
prod_tail far beyond the prod_head breaking the completion queue ring.

Fix that by limiting the number of descriptors to clean by the number
of used descriptors in the tx ring.

'ixgbe_clean_xdp_tx_irq()' function refactored to look more like
'ixgbe_xsk_clean_tx_ring()' since we're allowed to directly use
'next_to_clean' and 'next_to_use' indexes.

Fixes: 8221c5eba8c1 ("ixgbe: add AF_XDP zero-copy Tx support")
Signed-off-by: Ilya Maximets <i.maximets@samsung.com>
---

Version 3:
  * Reverted some refactoring made for v2.
  * Eliminated 'budget' for tx clean.
  * prefetch returned.

Version 2:
  * 'ixgbe_clean_xdp_tx_irq()' refactored to look more like
    'ixgbe_xsk_clean_tx_ring()'.

 drivers/net/ethernet/intel/ixgbe/ixgbe_xsk.c | 29 ++++++++------------
 1 file changed, 11 insertions(+), 18 deletions(-)

Comments

William Tu Aug. 22, 2019, 5:32 p.m. UTC | #1
On Thu, Aug 22, 2019 at 10:21 AM Alexander Duyck
<alexander.duyck@gmail.com> wrote:
>
> On Thu, Aug 22, 2019 at 10:12 AM Ilya Maximets <i.maximets@samsung.com> wrote:
> >
> > Tx code doesn't clear the descriptors' status after cleaning.
> > So, if the budget is larger than number of used elems in a ring, some
> > descriptors will be accounted twice and xsk_umem_complete_tx will move
> > prod_tail far beyond the prod_head breaking the completion queue ring.
> >
> > Fix that by limiting the number of descriptors to clean by the number
> > of used descriptors in the tx ring.
> >
> > 'ixgbe_clean_xdp_tx_irq()' function refactored to look more like
> > 'ixgbe_xsk_clean_tx_ring()' since we're allowed to directly use
> > 'next_to_clean' and 'next_to_use' indexes.
> >
> > Fixes: 8221c5eba8c1 ("ixgbe: add AF_XDP zero-copy Tx support")
> > Signed-off-by: Ilya Maximets <i.maximets@samsung.com>
> > ---
> >
> > Version 3:
> >   * Reverted some refactoring made for v2.
> >   * Eliminated 'budget' for tx clean.
> >   * prefetch returned.
> >
> > Version 2:
> >   * 'ixgbe_clean_xdp_tx_irq()' refactored to look more like
> >     'ixgbe_xsk_clean_tx_ring()'.
> >
> >  drivers/net/ethernet/intel/ixgbe/ixgbe_xsk.c | 29 ++++++++------------
> >  1 file changed, 11 insertions(+), 18 deletions(-)
>
> Thanks for addressing my concerns.
>
> Acked-by: Alexander Duyck <alexander.h.duyck@linux.intel.com>

Thanks.

Tested-by: William Tu <u9012063@gmail.com>
Björn Töpel Aug. 23, 2019, 6:10 a.m. UTC | #2
On 2019-08-22 19:32, William Tu wrote:
> On Thu, Aug 22, 2019 at 10:21 AM Alexander Duyck
> <alexander.duyck@gmail.com> wrote:
>>
>> On Thu, Aug 22, 2019 at 10:12 AM Ilya Maximets <i.maximets@samsung.com> wrote:
>>>
>>> Tx code doesn't clear the descriptors' status after cleaning.
>>> So, if the budget is larger than number of used elems in a ring, some
>>> descriptors will be accounted twice and xsk_umem_complete_tx will move
>>> prod_tail far beyond the prod_head breaking the completion queue ring.
>>>
>>> Fix that by limiting the number of descriptors to clean by the number
>>> of used descriptors in the tx ring.
>>>
>>> 'ixgbe_clean_xdp_tx_irq()' function refactored to look more like
>>> 'ixgbe_xsk_clean_tx_ring()' since we're allowed to directly use
>>> 'next_to_clean' and 'next_to_use' indexes.
>>>
>>> Fixes: 8221c5eba8c1 ("ixgbe: add AF_XDP zero-copy Tx support")
>>> Signed-off-by: Ilya Maximets <i.maximets@samsung.com>
>>> ---
>>>
>>> Version 3:
>>>    * Reverted some refactoring made for v2.
>>>    * Eliminated 'budget' for tx clean.
>>>    * prefetch returned.
>>>
>>> Version 2:
>>>    * 'ixgbe_clean_xdp_tx_irq()' refactored to look more like
>>>      'ixgbe_xsk_clean_tx_ring()'.
>>>
>>>   drivers/net/ethernet/intel/ixgbe/ixgbe_xsk.c | 29 ++++++++------------
>>>   1 file changed, 11 insertions(+), 18 deletions(-)
>>
>> Thanks for addressing my concerns.
>>
>> Acked-by: Alexander Duyck <alexander.h.duyck@linux.intel.com>
> 
> Thanks.
> 
> Tested-by: William Tu <u9012063@gmail.com>
> 

Will, thanks for testing! For this patch, did you notice any performance
degradation?


Cheers,
Björn
Eelco Chaudron Aug. 23, 2019, 12:10 p.m. UTC | #3
On 22 Aug 2019, at 19:12, Ilya Maximets wrote:

> Tx code doesn't clear the descriptors' status after cleaning.
> So, if the budget is larger than number of used elems in a ring, some
> descriptors will be accounted twice and xsk_umem_complete_tx will move
> prod_tail far beyond the prod_head breaking the completion queue ring.
>
> Fix that by limiting the number of descriptors to clean by the number
> of used descriptors in the tx ring.
>
> 'ixgbe_clean_xdp_tx_irq()' function refactored to look more like
> 'ixgbe_xsk_clean_tx_ring()' since we're allowed to directly use
> 'next_to_clean' and 'next_to_use' indexes.
>
> Fixes: 8221c5eba8c1 ("ixgbe: add AF_XDP zero-copy Tx support")
> Signed-off-by: Ilya Maximets <i.maximets@samsung.com>
> ---
>
> Version 3:
>   * Reverted some refactoring made for v2.
>   * Eliminated 'budget' for tx clean.
>   * prefetch returned.
>
> Version 2:
>   * 'ixgbe_clean_xdp_tx_irq()' refactored to look more like
>     'ixgbe_xsk_clean_tx_ring()'.
>
>  drivers/net/ethernet/intel/ixgbe/ixgbe_xsk.c | 29 
> ++++++++------------
>  1 file changed, 11 insertions(+), 18 deletions(-)
>

Did some test with and without the fix applied. For PVP the results are 
a little different depending on the packet size (note this is a single 
run, no deviation).
For the same physical port in and out it’s faster! Note this was OVS 
AF_XDP using a XENA tester at 10G wire speed.


+--------------------------------------------------------------------------------+
| Physical to Virtual to Physical test, L3 flows[port redirect]          
         |
+-----------------+--------------------------------------------------------------+
|                 | Packet size                                          
         |
+-----------------+--------+--------+--------+--------+--------+--------+--------+
| Number of flows |   64   |  128   |  256   |  512   |  768   |  1024  
|  1514  |
+-----------------+--------+--------+--------+--------+--------+--------+--------+
| [NO FIX]   1000 | 739161 | 700091 | 690034 | 659894 | 618128 | 594223 
| 537504 |
+-----------------+--------+--------+--------+--------+--------+--------+--------+
| [FIX]      1000 | 742317 | 708391 | 689952 | 658034 | 626056 | 587653 
| 530885 |
+-----------------+--------+--------+--------+--------+--------+--------+--------+

+--------------------------------------------------------------------------------------+
| Physical loopback test, L3 flows[port redirect]                        
               |
+-----------------+--------------------------------------------------------------------+
|                 | Packet size                                          
               |
+-----------------+---------+---------+---------+---------+---------+---------+--------+
| Number of flows |   64    |   128   |   256   |   512   |   768   |  
1024   |  1514  |
+-----------------+---------+---------+---------+---------+---------+---------+--------+
| [NO FIX]   1000 | 2573298 | 2227578 | 2514318 | 2298204 | 1081861 | 
1015173 | 788081 |
+-----------------+---------+---------+---------+---------+---------+---------+--------+
| [FIX]      1000 | 3343188 | 3234993 | 3151833 | 2349597 | 1586276 | 
1197304 | 814854 |
+-----------------+---------+---------+---------+---------+---------+---------+--------+


Tested-by: Eelco Chaudron <echaudro@redhat.com>
William Tu Aug. 23, 2019, 1:19 p.m. UTC | #4
On Thu, Aug 22, 2019 at 11:10 PM Björn Töpel <bjorn.topel@intel.com> wrote:
>
> On 2019-08-22 19:32, William Tu wrote:
> > On Thu, Aug 22, 2019 at 10:21 AM Alexander Duyck
> > <alexander.duyck@gmail.com> wrote:
> >>
> >> On Thu, Aug 22, 2019 at 10:12 AM Ilya Maximets <i.maximets@samsung.com> wrote:
> >>>
> >>> Tx code doesn't clear the descriptors' status after cleaning.
> >>> So, if the budget is larger than number of used elems in a ring, some
> >>> descriptors will be accounted twice and xsk_umem_complete_tx will move
> >>> prod_tail far beyond the prod_head breaking the completion queue ring.
> >>>
> >>> Fix that by limiting the number of descriptors to clean by the number
> >>> of used descriptors in the tx ring.
> >>>
> >>> 'ixgbe_clean_xdp_tx_irq()' function refactored to look more like
> >>> 'ixgbe_xsk_clean_tx_ring()' since we're allowed to directly use
> >>> 'next_to_clean' and 'next_to_use' indexes.
> >>>
> >>> Fixes: 8221c5eba8c1 ("ixgbe: add AF_XDP zero-copy Tx support")
> >>> Signed-off-by: Ilya Maximets <i.maximets@samsung.com>
> >>> ---
> >>>
> >>> Version 3:
> >>>    * Reverted some refactoring made for v2.
> >>>    * Eliminated 'budget' for tx clean.
> >>>    * prefetch returned.
> >>>
> >>> Version 2:
> >>>    * 'ixgbe_clean_xdp_tx_irq()' refactored to look more like
> >>>      'ixgbe_xsk_clean_tx_ring()'.
> >>>
> >>>   drivers/net/ethernet/intel/ixgbe/ixgbe_xsk.c | 29 ++++++++------------
> >>>   1 file changed, 11 insertions(+), 18 deletions(-)
> >>
> >> Thanks for addressing my concerns.
> >>
> >> Acked-by: Alexander Duyck <alexander.h.duyck@linux.intel.com>
> >
> > Thanks.
> >
> > Tested-by: William Tu <u9012063@gmail.com>
> >
>
> Will, thanks for testing! For this patch, did you notice any performance
> degradation?
>

No noticeable performance degradation seen this time.
Thanks,
William
Maciej Fijalkowski Aug. 26, 2019, 1:40 p.m. UTC | #5
On Thu, 22 Aug 2019 20:12:37 +0300
Ilya Maximets <i.maximets@samsung.com> wrote:

> Tx code doesn't clear the descriptors' status after cleaning.
> So, if the budget is larger than number of used elems in a ring, some
> descriptors will be accounted twice and xsk_umem_complete_tx will move
> prod_tail far beyond the prod_head breaking the completion queue ring.
> 
> Fix that by limiting the number of descriptors to clean by the number
> of used descriptors in the tx ring.
> 
> 'ixgbe_clean_xdp_tx_irq()' function refactored to look more like
> 'ixgbe_xsk_clean_tx_ring()' since we're allowed to directly use
> 'next_to_clean' and 'next_to_use' indexes.
> 
> Fixes: 8221c5eba8c1 ("ixgbe: add AF_XDP zero-copy Tx support")
> Signed-off-by: Ilya Maximets <i.maximets@samsung.com>
> ---
> 
> Version 3:
>   * Reverted some refactoring made for v2.
>   * Eliminated 'budget' for tx clean.
>   * prefetch returned.
> 
> Version 2:
>   * 'ixgbe_clean_xdp_tx_irq()' refactored to look more like
>     'ixgbe_xsk_clean_tx_ring()'.
> 
>  drivers/net/ethernet/intel/ixgbe/ixgbe_xsk.c | 29 ++++++++------------
>  1 file changed, 11 insertions(+), 18 deletions(-)
> 
> diff --git a/drivers/net/ethernet/intel/ixgbe/ixgbe_xsk.c b/drivers/net/ethernet/intel/ixgbe/ixgbe_xsk.c
> index 6b609553329f..a3b6d8c89127 100644
> --- a/drivers/net/ethernet/intel/ixgbe/ixgbe_xsk.c
> +++ b/drivers/net/ethernet/intel/ixgbe/ixgbe_xsk.c
> @@ -633,19 +633,17 @@ static void ixgbe_clean_xdp_tx_buffer(struct ixgbe_ring *tx_ring,
>  bool ixgbe_clean_xdp_tx_irq(struct ixgbe_q_vector *q_vector,
>  			    struct ixgbe_ring *tx_ring, int napi_budget)

While you're at it, can you please as well remove the 'napi_budget' argument?
It wasn't used at all even before your patch.

I'm jumping late in, but I was really wondering and hesitated with taking
part in discussion since the v1 of this patch - can you elaborate why simply
clearing the DD bit wasn't sufficient?

Maciej

>  {
> +	u16 ntc = tx_ring->next_to_clean, ntu = tx_ring->next_to_use;
>  	unsigned int total_packets = 0, total_bytes = 0;
> -	u32 i = tx_ring->next_to_clean, xsk_frames = 0;
> -	unsigned int budget = q_vector->tx.work_limit;
>  	struct xdp_umem *umem = tx_ring->xsk_umem;
>  	union ixgbe_adv_tx_desc *tx_desc;
>  	struct ixgbe_tx_buffer *tx_bi;
> -	bool xmit_done;
> +	u32 xsk_frames = 0;
>  
> -	tx_bi = &tx_ring->tx_buffer_info[i];
> -	tx_desc = IXGBE_TX_DESC(tx_ring, i);
> -	i -= tx_ring->count;
> +	tx_bi = &tx_ring->tx_buffer_info[ntc];
> +	tx_desc = IXGBE_TX_DESC(tx_ring, ntc);
>  
> -	do {
> +	while (ntc != ntu) {
>  		if (!(tx_desc->wb.status & cpu_to_le32(IXGBE_TXD_STAT_DD)))
>  			break;
>  
> @@ -661,22 +659,18 @@ bool ixgbe_clean_xdp_tx_irq(struct ixgbe_q_vector *q_vector,
>  
>  		tx_bi++;
>  		tx_desc++;
> -		i++;
> -		if (unlikely(!i)) {
> -			i -= tx_ring->count;
> +		ntc++;
> +		if (unlikely(ntc == tx_ring->count)) {
> +			ntc = 0;
>  			tx_bi = tx_ring->tx_buffer_info;
>  			tx_desc = IXGBE_TX_DESC(tx_ring, 0);
>  		}
>  
>  		/* issue prefetch for next Tx descriptor */
>  		prefetch(tx_desc);
> +	}
>  
> -		/* update budget accounting */
> -		budget--;
> -	} while (likely(budget));
> -
> -	i += tx_ring->count;
> -	tx_ring->next_to_clean = i;
> +	tx_ring->next_to_clean = ntc;
>  
>  	u64_stats_update_begin(&tx_ring->syncp);
>  	tx_ring->stats.bytes += total_bytes;
> @@ -688,8 +682,7 @@ bool ixgbe_clean_xdp_tx_irq(struct ixgbe_q_vector *q_vector,
>  	if (xsk_frames)
>  		xsk_umem_complete_tx(umem, xsk_frames);
>  
> -	xmit_done = ixgbe_xmit_zc(tx_ring, q_vector->tx.work_limit);
> -	return budget > 0 && xmit_done;
> +	return ixgbe_xmit_zc(tx_ring, q_vector->tx.work_limit);
>  }
>  
>  int ixgbe_xsk_async_xmit(struct net_device *dev, u32 qid)
Ilya Maximets Aug. 27, 2019, 11 a.m. UTC | #6
On 26.08.2019 16:40, Maciej Fijalkowski wrote:
> On Thu, 22 Aug 2019 20:12:37 +0300
> Ilya Maximets <i.maximets@samsung.com> wrote:
> 
>> Tx code doesn't clear the descriptors' status after cleaning.
>> So, if the budget is larger than number of used elems in a ring, some
>> descriptors will be accounted twice and xsk_umem_complete_tx will move
>> prod_tail far beyond the prod_head breaking the completion queue ring.
>>
>> Fix that by limiting the number of descriptors to clean by the number
>> of used descriptors in the tx ring.
>>
>> 'ixgbe_clean_xdp_tx_irq()' function refactored to look more like
>> 'ixgbe_xsk_clean_tx_ring()' since we're allowed to directly use
>> 'next_to_clean' and 'next_to_use' indexes.
>>
>> Fixes: 8221c5eba8c1 ("ixgbe: add AF_XDP zero-copy Tx support")
>> Signed-off-by: Ilya Maximets <i.maximets@samsung.com>
>> ---
>>
>> Version 3:
>>   * Reverted some refactoring made for v2.
>>   * Eliminated 'budget' for tx clean.
>>   * prefetch returned.
>>
>> Version 2:
>>   * 'ixgbe_clean_xdp_tx_irq()' refactored to look more like
>>     'ixgbe_xsk_clean_tx_ring()'.
>>
>>  drivers/net/ethernet/intel/ixgbe/ixgbe_xsk.c | 29 ++++++++------------
>>  1 file changed, 11 insertions(+), 18 deletions(-)
>>
>> diff --git a/drivers/net/ethernet/intel/ixgbe/ixgbe_xsk.c b/drivers/net/ethernet/intel/ixgbe/ixgbe_xsk.c
>> index 6b609553329f..a3b6d8c89127 100644
>> --- a/drivers/net/ethernet/intel/ixgbe/ixgbe_xsk.c
>> +++ b/drivers/net/ethernet/intel/ixgbe/ixgbe_xsk.c
>> @@ -633,19 +633,17 @@ static void ixgbe_clean_xdp_tx_buffer(struct ixgbe_ring *tx_ring,
>>  bool ixgbe_clean_xdp_tx_irq(struct ixgbe_q_vector *q_vector,
>>  			    struct ixgbe_ring *tx_ring, int napi_budget)
> 
> While you're at it, can you please as well remove the 'napi_budget' argument?
> It wasn't used at all even before your patch.

As you mentioned, this is not related to current patch and this patch doesn't
touch these particular lines of code.  So, I think it's better to make a
separate patch for this if you think it's needed.

> 
> I'm jumping late in, but I was really wondering and hesitated with taking
> part in discussion since the v1 of this patch - can you elaborate why simply
> clearing the DD bit wasn't sufficient?

Clearing the DD bit will end up driver and hardware writing to close memory
locations at the same time leading to cache trashing and poor performance.

Anyway additional write is unnecessary, because we know exactly which descriptors
we need to check.

Best regards, Ilya Maximets.

> 
> Maciej
> 
>>  {
>> +	u16 ntc = tx_ring->next_to_clean, ntu = tx_ring->next_to_use;
>>  	unsigned int total_packets = 0, total_bytes = 0;
>> -	u32 i = tx_ring->next_to_clean, xsk_frames = 0;
>> -	unsigned int budget = q_vector->tx.work_limit;
>>  	struct xdp_umem *umem = tx_ring->xsk_umem;
>>  	union ixgbe_adv_tx_desc *tx_desc;
>>  	struct ixgbe_tx_buffer *tx_bi;
>> -	bool xmit_done;
>> +	u32 xsk_frames = 0;
>>  
>> -	tx_bi = &tx_ring->tx_buffer_info[i];
>> -	tx_desc = IXGBE_TX_DESC(tx_ring, i);
>> -	i -= tx_ring->count;
>> +	tx_bi = &tx_ring->tx_buffer_info[ntc];
>> +	tx_desc = IXGBE_TX_DESC(tx_ring, ntc);
>>  
>> -	do {
>> +	while (ntc != ntu) {
>>  		if (!(tx_desc->wb.status & cpu_to_le32(IXGBE_TXD_STAT_DD)))
>>  			break;
>>  
>> @@ -661,22 +659,18 @@ bool ixgbe_clean_xdp_tx_irq(struct ixgbe_q_vector *q_vector,
>>  
>>  		tx_bi++;
>>  		tx_desc++;
>> -		i++;
>> -		if (unlikely(!i)) {
>> -			i -= tx_ring->count;
>> +		ntc++;
>> +		if (unlikely(ntc == tx_ring->count)) {
>> +			ntc = 0;
>>  			tx_bi = tx_ring->tx_buffer_info;
>>  			tx_desc = IXGBE_TX_DESC(tx_ring, 0);
>>  		}
>>  
>>  		/* issue prefetch for next Tx descriptor */
>>  		prefetch(tx_desc);
>> +	}
>>  
>> -		/* update budget accounting */
>> -		budget--;
>> -	} while (likely(budget));
>> -
>> -	i += tx_ring->count;
>> -	tx_ring->next_to_clean = i;
>> +	tx_ring->next_to_clean = ntc;
>>  
>>  	u64_stats_update_begin(&tx_ring->syncp);
>>  	tx_ring->stats.bytes += total_bytes;
>> @@ -688,8 +682,7 @@ bool ixgbe_clean_xdp_tx_irq(struct ixgbe_q_vector *q_vector,
>>  	if (xsk_frames)
>>  		xsk_umem_complete_tx(umem, xsk_frames);
>>  
>> -	xmit_done = ixgbe_xmit_zc(tx_ring, q_vector->tx.work_limit);
>> -	return budget > 0 && xmit_done;
>> +	return ixgbe_xmit_zc(tx_ring, q_vector->tx.work_limit);
>>  }
>>  
>>  int ixgbe_xsk_async_xmit(struct net_device *dev, u32 qid)
> 
> 
>

Patch
diff mbox series

diff --git a/drivers/net/ethernet/intel/ixgbe/ixgbe_xsk.c b/drivers/net/ethernet/intel/ixgbe/ixgbe_xsk.c
index 6b609553329f..a3b6d8c89127 100644
--- a/drivers/net/ethernet/intel/ixgbe/ixgbe_xsk.c
+++ b/drivers/net/ethernet/intel/ixgbe/ixgbe_xsk.c
@@ -633,19 +633,17 @@  static void ixgbe_clean_xdp_tx_buffer(struct ixgbe_ring *tx_ring,
 bool ixgbe_clean_xdp_tx_irq(struct ixgbe_q_vector *q_vector,
 			    struct ixgbe_ring *tx_ring, int napi_budget)
 {
+	u16 ntc = tx_ring->next_to_clean, ntu = tx_ring->next_to_use;
 	unsigned int total_packets = 0, total_bytes = 0;
-	u32 i = tx_ring->next_to_clean, xsk_frames = 0;
-	unsigned int budget = q_vector->tx.work_limit;
 	struct xdp_umem *umem = tx_ring->xsk_umem;
 	union ixgbe_adv_tx_desc *tx_desc;
 	struct ixgbe_tx_buffer *tx_bi;
-	bool xmit_done;
+	u32 xsk_frames = 0;
 
-	tx_bi = &tx_ring->tx_buffer_info[i];
-	tx_desc = IXGBE_TX_DESC(tx_ring, i);
-	i -= tx_ring->count;
+	tx_bi = &tx_ring->tx_buffer_info[ntc];
+	tx_desc = IXGBE_TX_DESC(tx_ring, ntc);
 
-	do {
+	while (ntc != ntu) {
 		if (!(tx_desc->wb.status & cpu_to_le32(IXGBE_TXD_STAT_DD)))
 			break;
 
@@ -661,22 +659,18 @@  bool ixgbe_clean_xdp_tx_irq(struct ixgbe_q_vector *q_vector,
 
 		tx_bi++;
 		tx_desc++;
-		i++;
-		if (unlikely(!i)) {
-			i -= tx_ring->count;
+		ntc++;
+		if (unlikely(ntc == tx_ring->count)) {
+			ntc = 0;
 			tx_bi = tx_ring->tx_buffer_info;
 			tx_desc = IXGBE_TX_DESC(tx_ring, 0);
 		}
 
 		/* issue prefetch for next Tx descriptor */
 		prefetch(tx_desc);
+	}
 
-		/* update budget accounting */
-		budget--;
-	} while (likely(budget));
-
-	i += tx_ring->count;
-	tx_ring->next_to_clean = i;
+	tx_ring->next_to_clean = ntc;
 
 	u64_stats_update_begin(&tx_ring->syncp);
 	tx_ring->stats.bytes += total_bytes;
@@ -688,8 +682,7 @@  bool ixgbe_clean_xdp_tx_irq(struct ixgbe_q_vector *q_vector,
 	if (xsk_frames)
 		xsk_umem_complete_tx(umem, xsk_frames);
 
-	xmit_done = ixgbe_xmit_zc(tx_ring, q_vector->tx.work_limit);
-	return budget > 0 && xmit_done;
+	return ixgbe_xmit_zc(tx_ring, q_vector->tx.work_limit);
 }
 
 int ixgbe_xsk_async_xmit(struct net_device *dev, u32 qid)