Revision 6e474083f3daf3a3546737f5d7d502ad12eb257c authored by Wei Xu on 01 December 2017, 10:10:36 UTC, committed by David S. Miller on 03 December 2017, 02:31:03 UTC
Matthew found a roughly 40% tcp throughput regression with commit c67df11f(vhost_net: try batch dequing from skb array) as discussed in the following thread: https://www.mail-archive.com/netdev@vger.kernel.org/msg187936.html Eventually we figured out that it was a skb leak in handle_rx() when sending packets to the VM. This usually happens when a guest can not drain out vq as fast as vhost fills in, afterwards it sets off the traffic jam and leaks skb(s) which occurs as no headcount to send on the vq from vhost side. This can be avoided by making sure we have got enough headcount before actually consuming a skb from the batched rx array while transmitting, which is simply done by moving checking the zero headcount a bit ahead. Signed-off-by: Wei Xu <wexu@redhat.com> Reported-by: Matthew Rosato <mjrosato@linux.vnet.ibm.com> Acked-by: Michael S. Tsirkin <mst@redhat.com> Signed-off-by: David S. Miller <davem@davemloft.net>
1 parent fa935ca
File | Mode | Size |
---|---|---|
samsung | ||
tegra | ||
Kconfig | -rw-r--r-- | 4.8 KB |
Makefile | -rw-r--r-- | 780 bytes |
atmel-ebi.c | -rw-r--r-- | 14.7 KB |
atmel-sdramc.c | -rw-r--r-- | 2.3 KB |
brcmstb_dpfe.c | -rw-r--r-- | 17.9 KB |
da8xx-ddrctl.c | -rw-r--r-- | 4.1 KB |
emif.c | -rw-r--r-- | 54.7 KB |
emif.h | -rw-r--r-- | 19.1 KB |
fsl-corenet-cf.c | -rw-r--r-- | 6.7 KB |
fsl_ifc.c | -rw-r--r-- | 8.9 KB |
jz4780-nemc.c | -rw-r--r-- | 9.7 KB |
mtk-smi.c | -rw-r--r-- | 11.3 KB |
mvebu-devbus.c | -rw-r--r-- | 9.9 KB |
of_memory.c | -rw-r--r-- | 4.8 KB |
of_memory.h | -rw-r--r-- | 1.1 KB |
omap-gpmc.c | -rw-r--r-- | 67.1 KB |
pl172.c | -rw-r--r-- | 8.1 KB |
tegra20-mc.c | -rw-r--r-- | 5.9 KB |
ti-aemif.c | -rw-r--r-- | 12.0 KB |
![swh spinner](/static/img/swh-spinner.gif)
Computing file changes ...