[Intel-wired-lan] [PATCH net v3] ixgbe: fix double clean of tx descriptors with xdp

Eelco Chaudron echaudro at redhat.com
Fri Aug 23 12:10:45 UTC 2019



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 at 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 at redhat.com>


More information about the Intel-wired-lan mailing list