[v5] TUN/TAP & vhost_net: netdev queue flow control to avoid ptr_ring tail drop

[PATCH net-next v5 0/8] TUN/TAP & vhost_net: netdev queue flow control to avoid ptr_ring tail drop

Posted by Simon Schippers 1 week ago

On 23.09.25 16:55, Michael S. Tsirkin wrote:
> On Tue, Sep 23, 2025 at 12:15:45AM +0200, Simon Schippers wrote:
>> This patch series deals with TUN, TAP and vhost_net which drop incoming 
>> SKBs whenever their internal ptr_ring buffer is full. Instead, with this 
>> patch series, the associated netdev queue is stopped before this happens. 
>> This allows the connected qdisc to function correctly as reported by [1] 
>> and improves application-layer performance, see our paper [2]. Meanwhile 
>> the theoretical performance differs only slightly:
>>
>> +------------------------+----------+----------+
>> | pktgen benchmarks      | Stock    | Patched  |
>> | i5 6300HQ, 20M packets |          |          |
>> +------------------------+----------+----------+
>> | TAP                    | 2.10Mpps | 1.99Mpps |
>> +------------------------+----------+----------+
>> | TAP+vhost_net          | 6.05Mpps | 6.14Mpps |
>> +------------------------+----------+----------+
>> | Note: Patched had no TX drops at all,        |
>> | while stock suffered numerous drops.         |
>> +----------------------------------------------+
>>
>> This patch series includes TUN, TAP, and vhost_net because they share 
>> logic. Adjusting only one of them would break the others. Therefore, the 
>> patch series is structured as follows:
>> 1+2: New ptr_ring helpers for 3 & 4
>> 3: TUN & TAP: Stop netdev queue upon reaching a full ptr_ring
> 
> 
> so what happens if you only apply patches 1-3?
> 

The netdev queue of vhost_net would be stopped by tun_net_xmit but will
never be woken again.

>> 4: TUN & TAP: Wake netdev queue after consuming an entry
>> 5+6+7: TUN & TAP: ptr_ring wrappers and other helpers to be called by 
>> vhost_net
>> 8: vhost_net: Call the wrappers & helpers
>>
>> Possible future work:
>> - Introduction of Byte Queue Limits as suggested by Stephen Hemminger
>> - Adaption of the netdev queue flow control for ipvtap & macvtap
>>
>> [1] Link: 
>> https://unix.stackexchange.com/questions/762935/traffic-shaping-ineffective-on-tun-device
>> [2] Link: 
>> https://cni.etit.tu-dortmund.de/storages/cni-etit/r/Research/Publications/2025/Gebauer_2025_VTCFall/Gebauer_VTCFall2025_AuthorsVersion.pdf
>>
>> Links to previous versions:
>> V4: 
>> https://lore.kernel.org/netdev/20250902080957.47265-1-simon.schippers@tu-dortmund.de/T/#u
>> V3: 
>> https://lore.kernel.org/netdev/20250825211832.84901-1-simon.schippers@tu-dortmund.de/T/#u
>> V2: 
>> https://lore.kernel.org/netdev/20250811220430.14063-1-simon.schippers@tu-dortmund.de/T/#u
>> V1: 
>> https://lore.kernel.org/netdev/20250808153721.261334-1-simon.schippers@tu-dortmund.de/T/#u
>>
>> Changelog:
>> V4 -> V5:
>> - Stop the netdev queue prior to producing the final fitting ptr_ring entry
>> -> Ensures the consumer has the latest netdev queue state, making it safe 
>> to wake the queue
>> -> Resolves an issue in vhost_net where the netdev queue could remain 
>> stopped despite being empty
>> -> For TUN/TAP, the netdev queue no longer needs to be woken in the 
>> blocking loop
>> -> Introduces new helpers __ptr_ring_full_next and 
>> __ptr_ring_will_invalidate for this purpose
>>
>> - vhost_net now uses wrappers of TUN/TAP for ptr_ring consumption rather 
>> than maintaining its own rx_ring pointer
>>
>> V3 -> V4:
>> - Target net-next instead of net
>> - Changed to patch series instead of single patch
>> - Changed to new title from old title
>> "TUN/TAP: Improving throughput and latency by avoiding SKB drops"
>> - Wake netdev queue with new helpers wake_netdev_queue when there is any 
>> spare capacity in the ptr_ring instead of waiting for it to be empty
>> - Use tun_file instead of tun_struct in tun_ring_recv as a more consistent 
>> logic
>> - Use smp_wmb() and smp_rmb() barrier pair, which avoids any packet drops 
>> that happened rarely before
>> - Use safer logic for vhost_net using RCU read locks to access TUN/TAP data
>>
>> V2 -> V3: Added support for TAP and TAP+vhost_net.
>>
>> V1 -> V2: Removed NETDEV_TX_BUSY return case in tun_net_xmit and removed 
>> unnecessary netif_tx_wake_queue in tun_ring_recv.
>>
>> Thanks,
>> Simon :)
>>
>> Simon Schippers (8):
>>   __ptr_ring_full_next: Returns if ring will be full after next
>>     insertion
>>   Move the decision of invalidation out of __ptr_ring_discard_one
>>   TUN, TAP & vhost_net: Stop netdev queue before reaching a full
>>     ptr_ring
>>   TUN & TAP: Wake netdev queue after consuming an entry
>>   TUN & TAP: Provide ptr_ring_consume_batched wrappers for vhost_net
>>   TUN & TAP: Provide ptr_ring_unconsume wrappers for vhost_net
>>   TUN & TAP: Methods to determine whether file is TUN/TAP for vhost_net
>>   vhost_net: Replace rx_ring with calls of TUN/TAP wrappers
>>
>>  drivers/net/tap.c        | 115 +++++++++++++++++++++++++++++++--
>>  drivers/net/tun.c        | 136 +++++++++++++++++++++++++++++++++++----
>>  drivers/vhost/net.c      |  90 +++++++++++++++++---------
>>  include/linux/if_tap.h   |  15 +++++
>>  include/linux/if_tun.h   |  18 ++++++
>>  include/linux/ptr_ring.h |  54 +++++++++++++---
>>  6 files changed, 367 insertions(+), 61 deletions(-)
>>
>> -- 
>> 2.43.0
>

Re: [PATCH net-next v5 0/8] TUN/TAP & vhost_net: netdev queue flow control to avoid ptr_ring tail drop

Posted by Michael S. Tsirkin 1 week ago

On Wed, Sep 24, 2025 at 07:59:46AM +0200, Simon Schippers wrote:
> On 23.09.25 16:55, Michael S. Tsirkin wrote:
> > On Tue, Sep 23, 2025 at 12:15:45AM +0200, Simon Schippers wrote:
> >> This patch series deals with TUN, TAP and vhost_net which drop incoming 
> >> SKBs whenever their internal ptr_ring buffer is full. Instead, with this 
> >> patch series, the associated netdev queue is stopped before this happens. 
> >> This allows the connected qdisc to function correctly as reported by [1] 
> >> and improves application-layer performance, see our paper [2]. Meanwhile 
> >> the theoretical performance differs only slightly:
> >>
> >> +------------------------+----------+----------+
> >> | pktgen benchmarks      | Stock    | Patched  |
> >> | i5 6300HQ, 20M packets |          |          |
> >> +------------------------+----------+----------+
> >> | TAP                    | 2.10Mpps | 1.99Mpps |
> >> +------------------------+----------+----------+
> >> | TAP+vhost_net          | 6.05Mpps | 6.14Mpps |
> >> +------------------------+----------+----------+
> >> | Note: Patched had no TX drops at all,        |
> >> | while stock suffered numerous drops.         |
> >> +----------------------------------------------+
> >>
> >> This patch series includes TUN, TAP, and vhost_net because they share 
> >> logic. Adjusting only one of them would break the others. Therefore, the 
> >> patch series is structured as follows:
> >> 1+2: New ptr_ring helpers for 3 & 4
> >> 3: TUN & TAP: Stop netdev queue upon reaching a full ptr_ring
> > 
> > 
> > so what happens if you only apply patches 1-3?
> > 
> 
> The netdev queue of vhost_net would be stopped by tun_net_xmit but will
> never be woken again.

So this breaks bisect. Don't split patches like this please.


> >> 4: TUN & TAP: Wake netdev queue after consuming an entry
> >> 5+6+7: TUN & TAP: ptr_ring wrappers and other helpers to be called by 
> >> vhost_net
> >> 8: vhost_net: Call the wrappers & helpers
> >>
> >> Possible future work:
> >> - Introduction of Byte Queue Limits as suggested by Stephen Hemminger
> >> - Adaption of the netdev queue flow control for ipvtap & macvtap
> >>
> >> [1] Link: 
> >> https://unix.stackexchange.com/questions/762935/traffic-shaping-ineffective-on-tun-device
> >> [2] Link: 
> >> https://cni.etit.tu-dortmund.de/storages/cni-etit/r/Research/Publications/2025/Gebauer_2025_VTCFall/Gebauer_VTCFall2025_AuthorsVersion.pdf
> >>
> >> Links to previous versions:
> >> V4: 
> >> https://lore.kernel.org/netdev/20250902080957.47265-1-simon.schippers@tu-dortmund.de/T/#u
> >> V3: 
> >> https://lore.kernel.org/netdev/20250825211832.84901-1-simon.schippers@tu-dortmund.de/T/#u
> >> V2: 
> >> https://lore.kernel.org/netdev/20250811220430.14063-1-simon.schippers@tu-dortmund.de/T/#u
> >> V1: 
> >> https://lore.kernel.org/netdev/20250808153721.261334-1-simon.schippers@tu-dortmund.de/T/#u
> >>
> >> Changelog:
> >> V4 -> V5:
> >> - Stop the netdev queue prior to producing the final fitting ptr_ring entry
> >> -> Ensures the consumer has the latest netdev queue state, making it safe 
> >> to wake the queue
> >> -> Resolves an issue in vhost_net where the netdev queue could remain 
> >> stopped despite being empty
> >> -> For TUN/TAP, the netdev queue no longer needs to be woken in the 
> >> blocking loop
> >> -> Introduces new helpers __ptr_ring_full_next and 
> >> __ptr_ring_will_invalidate for this purpose
> >>
> >> - vhost_net now uses wrappers of TUN/TAP for ptr_ring consumption rather 
> >> than maintaining its own rx_ring pointer
> >>
> >> V3 -> V4:
> >> - Target net-next instead of net
> >> - Changed to patch series instead of single patch
> >> - Changed to new title from old title
> >> "TUN/TAP: Improving throughput and latency by avoiding SKB drops"
> >> - Wake netdev queue with new helpers wake_netdev_queue when there is any 
> >> spare capacity in the ptr_ring instead of waiting for it to be empty
> >> - Use tun_file instead of tun_struct in tun_ring_recv as a more consistent 
> >> logic
> >> - Use smp_wmb() and smp_rmb() barrier pair, which avoids any packet drops 
> >> that happened rarely before
> >> - Use safer logic for vhost_net using RCU read locks to access TUN/TAP data
> >>
> >> V2 -> V3: Added support for TAP and TAP+vhost_net.
> >>
> >> V1 -> V2: Removed NETDEV_TX_BUSY return case in tun_net_xmit and removed 
> >> unnecessary netif_tx_wake_queue in tun_ring_recv.
> >>
> >> Thanks,
> >> Simon :)
> >>
> >> Simon Schippers (8):
> >>   __ptr_ring_full_next: Returns if ring will be full after next
> >>     insertion
> >>   Move the decision of invalidation out of __ptr_ring_discard_one
> >>   TUN, TAP & vhost_net: Stop netdev queue before reaching a full
> >>     ptr_ring
> >>   TUN & TAP: Wake netdev queue after consuming an entry
> >>   TUN & TAP: Provide ptr_ring_consume_batched wrappers for vhost_net
> >>   TUN & TAP: Provide ptr_ring_unconsume wrappers for vhost_net
> >>   TUN & TAP: Methods to determine whether file is TUN/TAP for vhost_net
> >>   vhost_net: Replace rx_ring with calls of TUN/TAP wrappers
> >>
> >>  drivers/net/tap.c        | 115 +++++++++++++++++++++++++++++++--
> >>  drivers/net/tun.c        | 136 +++++++++++++++++++++++++++++++++++----
> >>  drivers/vhost/net.c      |  90 +++++++++++++++++---------
> >>  include/linux/if_tap.h   |  15 +++++
> >>  include/linux/if_tun.h   |  18 ++++++
> >>  include/linux/ptr_ring.h |  54 +++++++++++++---
> >>  6 files changed, 367 insertions(+), 61 deletions(-)
> >>
> >> -- 
> >> 2.43.0
> >