[dpdk-dev] [Bug 570] [dpdk-20.11] vhost_pmd_xstats: Failed to Launch virtio_user and vhost side core dumped

[bugzilla at dpdk.org]

https://bugs.dpdk.org/show_bug.cgi?id=570

            Bug ID: 570
           Summary: [dpdk-20.11] vhost_pmd_xstats: Failed to Launch
                    virtio_user and vhost side core dumped
           Product: DPDK
           Version: 20.11
          Hardware: All
                OS: All
            Status: UNCONFIRMED
          Severity: normal
          Priority: Normal
         Component: vhost/virtio
          Assignee: dev at dpdk.org
          Reporter: yux.jiang at intel.com
  Target Milestone: ---

Created attachment 133
  --> https://bugs.dpdk.org/attachment.cgi?id=133&action=edit
Failed to Launch virtio_user and vhost side core dumped

Environment

DPDK version: Use make showversion or for a non-released version: git remote -v
&& git show-ref --heads 
[60af9bd63dd46ebd709371e88b4096fd2bc048cf] version: 20.11-rc3
Other software versions: name/version for QEMU, OVS, etc. Repeat as required.
OS: Ubuntu20.04.1 LTS (Focal Fossa)/5.4.0-38-generic
Compiler: gcc version 9.3.0
Hardware platform: Intel(R) Xeon(R) Gold 6139 CPU @ 2.30GHz
NIC hardware: Ethernet Controller XL710 for 40GbE QSFP+ 1583
NIC firmware: if applicable

Test Setup
Steps to reproduce
List the steps to reproduce the issue.
TG --> NIC --> Vhost --> Virtio--> Vhost --> NIC --> TG
Test Case 1: xstats test with packed ring mergeable path
========================================================
Step0: Build DPDK
    rm -rf x86_64-native-linuxapp-gcc
    CC=gcc meson --werror -Denable_kmods=True --default-library=static
x86_64-native-linuxapp-gcc
    ninja -C x86_64-native-linuxapp-gcc -j 110  

Step1. Bind one port to vfio-pci, then launch vhost by below command::    
    rm -rf vhost-net*
    ./x86_64-native-linuxapp-gcc/app/dpdk-testpmd -n 4 -l 2-4  \
    --file-prefix=vhost --vdev 'net_vhost0,iface=vhost-net,queues=2,client=0'
-- -i --nb-cores=2 --rxq=2 --txq=2

Step2. Launch virtio-user by below command::    
    ./x86_64-native-linuxapp-gcc/app/dpdk-testpmd -n 4 -l 5-7 \
    --no-pci --file-prefix=virtio \
   
--vdev=net_virtio_user0,mac=00:01:02:03:04:05,path=./vhost-net,queues=2,packed_vq=1,mrg_rxbuf=1,in_order=0
\
    -- -i --tx-offloads=0x0 --enable-hw-vlan-strip --rss-ip --nb-cores=2
--rxq=2 --txq=2


# Put output in a noformat block like this.
root at dpdk:~/dpdk# ./x86_64-native-linuxapp-gcc/app/dpdk-testpmd -n 4 -l 5-7
--no-pci --file-prefix=virtio
--vdev=net_virtio_user0,mac=00:01:02:03:04:05,path=./vhost-net,queues=2,packed_vq=1,mrg_rxbuf=1,in_order=0
-- -i --tx-offloads=0x0 --enable-hw-vlan-strip --rss-ip --nb-cores=2 --rxq=2
--txq=2
EAL: Detected 72 lcore(s)
EAL: Detected 2 NUMA nodes
EAL: Multi-process socket /var/run/dpdk/virtio/mp_socket
EAL: Selected IOVA mode 'VA'
EAL: Probing VFIO support...
EAL: VFIO support initialized
vhost_user_read(): Failed to recv msg hdr: -1 instead of 12.
vhost_user_sock(): Received msg failed: Connection reset by peer
virtio_user_send_status_update(): VHOST_USER_SET_STATUS failed (-1): Connection
reset by peer
root at dpdk:~/dpdk#

And vhost side log:
VHOST_CONFIG: read message VHOST_USER_GET_STATUS
VHOST_CONFIG: read message VHOST_USER_SET_STATUS
VHOST_CONFIG: New device status(0x00000003):
        -RESET: 0
        -ACKNOWLEDGE: 1
        -DRIVER: 1
        -FEATURES_OK: 0
        -DRIVER_OK: 0
        -DEVICE_NEED_RESET: 0
        -FAILED: 0
VHOST_CONFIG: read message VHOST_USER_GET_STATUS
VHOST_CONFIG: read message VHOST_USER_SET_VRING_CALL
VHOST_CONFIG: vring call idx:0 file:38
VHOST_CONFIG: read message VHOST_USER_SET_VRING_CALL
Segmentation fault (core dumped)
root at dpdk:~/dpdk#


Expected Result
Explain what is the expected result in text or as an example output:
# Put output in a noformat block like this.
Launch virtio-user successfully and vhost side works well.


Regression
Is this issue a regression: (Y)
Version the regression was introduced: Specify git id if known.
root at dpdk:~/dpdk# git bisect bad
8acd7c21335305426fbbc0a3aea2978734d0b2a8 is the first bad commit
commit 8acd7c21335305426fbbc0a3aea2978734d0b2a8
Author: Maxime Coquelin <maxime.coquelin at redhat.com>
Date:   Mon Oct 19 19:34:09 2020 +0200
    vhost: fix virtqueues metadata allocation
    The Vhost-user backend implementation assumes there will be
    no holes in the device's array of virtqueues metadata
    pointers.
    It can happen though, and would cause segmentation faults,
    memory leaks or undefined behaviour.
    This patch keep the assumption that there is no holes in this
    array, and allocate all uninitialized virtqueues metadata up
    to requested index.
    Fixes: 160cbc815b41 ("vhost: remove a hack on queue allocation")
    Cc: stable at dpdk.org
    Suggested-by: Adrian Moreno <amorenoz at redhat.com>
    Signed-off-by: Maxime Coquelin <maxime.coquelin at redhat.com>
    Reviewed-by: Chenbo Xia <chenbo.xia at intel.com>
 lib/librte_vhost/vhost.c | 33 ++++++++++++++++++++-------------
 1 file changed, 20 insertions(+), 13 deletions(-)

-- 
You are receiving this mail because:
You are the assignee for the bug.