使用Valgrind运行DPDK示例程序dpdk-testpmd时启动卡住求助

问题现象 #

使用Valgrind运行DPDK示例程序（如dpdk-testpmd）时，程序在启动阶段卡住，相关日志如下：

[root@node1 app]# valgrind --leak-check=yes  ./dpdk-testpmd -a 0000:af:00.3 
==322313== Memcheck, a memory error detector
==322313== Copyright (C) 2002-2022, and GNU GPL'd, by Julian Seward et al.
==322313== Using Valgrind-3.21.0.GIT and LibVEX; rerun with -h for copyright info
==322313== Command: ./dpdk-testpmd -a 0000:af:00.3
==322313== 
EAL: Detected CPU lcores: 96
EAL: Detected NUMA nodes: 2
EAL: Detected static linkage of DPDK
EAL: Multi-process socket /var/run/dpdk/rte/mp_socket
EAL: Selected IOVA mode 'VA'
EAL: No available 2048 kB hugepages reported
EAL: VFIO support initialized
==322313== Warning: set address range perms: large range [0x140000000, 0x980000000) (noaccess)
==322313== Warning: set address range perms: large range [0x940000000, 0x980000000) (noaccess)
==322313== Warning: set address range perms: large range [0x100b967000, 0x184b967000) (noaccess)
==322313== Warning: set address range perms: large range [0x100b967000, 0x184b967000) (noaccess)
==322313== Warning: set address range perms: large range [0x100b967000, 0x184b967000) (noaccess)
==322313== Warning: set address range perms: large range [0x100b967000, 0x184b967000) (noaccess)
==322313== Warning: set address range perms: large range [0x100b967000, 0x184b967000) (noaccess)
==322313== Warning: set address range perms: large range [0x100b967000, 0x184b967000) (noaccess)
==322313== Warning: set address range perms: large range [0x100b967000, 0x184b967000) (noaccess)
==322313== Warning: set address range perms: large range [0x100b967000, 0x184b967000) (noaccess)
==322313== Warning: set address range perms: large range [0x100b967000, 0x184b967000) (noaccess)
==322313== Warning: set address range perms: large range [0x100b967000, 0x184b967000) (noaccess)
==322313== Warning: set address range perms: large range [0x100b967000, 0x184b967000) (noaccess)
EAL: WARNING! Base virtual address hint (0xa80001000 != 0x1040000000) not respected!
EAL:    This may cause issues with mapping into secondary processes
==322313== Warning: set address range perms: large range [0x100b967000, 0x1040000000) (noaccess)

而直接运行程序时完全正常：

[root@node1 app]# ./dpdk-testpmd -a 0000:af:00.3 
EAL: Detected CPU lcores: 96
EAL: Detected NUMA nodes: 2
EAL: Detected static linkage of DPDK
EAL: Multi-process socket /var/run/dpdk/rte/mp_socket
EAL: Selected IOVA mode 'VA'
EAL: No available 2048 kB hugepages reported
EAL: VFIO support initialized
EAL: Using IOMMU type 1 (Type 1)
EAL: Probe PCI driver: net_ice (8086:1593) device: 0000:af:00.3 (socket 1)
ice_load_pkg_type(): Active package is: 1.3.28.0, ICE OS Default Package (double VLAN mode)
TELEMETRY: No legacy callbacks, legacy socket not created
testpmd: create a new mbuf pool <mb_pool_0>: n=203456, size=2176, socket=0
testpmd: preferred mempool ops selected: ring_mp_mc
testpmd: create a new mbuf pool <mb_pool_1>: n=203456, size=2176, socket=1
testpmd: preferred mempool ops selected: ring_mp_mc

Warning! port-topology=paired and odd forward ports number, the last port will pair with itself.

Configuring Port 0 (socket 1)
ice_set_rx_function(): Using AVX2 Vector Rx (port 0).
Port 0: B4:96:91:A7:66:57
Checking link statuses...
Done
No commandline core given, start packet forwarding
io packet forwarding - ports=1 - cores=1 - streams=1 - NUMA support enabled, MP allocation mode: native
Logical Core 1 (socket 0) forwards packets on 1 streams:
  RX P=0/Q=0 (socket 1) -> TX P=0/Q=0 (socket 1) peer=02:00:00:00:00:00

  io packet forwarding packets/burst=32
  nb forwarding cores=1 - nb forwarding ports=1
  port 0: RX queue number: 1 Tx queue number: 1
    Rx offloads=0x0 Tx offloads=0x10000
    RX queue: 0
      RX desc=1024 - RX free threshold=32
      RX threshold registers: pthresh=0 hthresh=0  wthresh=0
      RX Offloads=0x0
    TX queue: 0
      TX desc=1024 - TX free threshold=32
      TX threshold registers: pthresh=32 hthresh=0  wthresh=0
      TX offloads=0x10000 - TX RS bit threshold=32
Press enter to exit

需求为调试链接DPDK库的应用中常规malloc调用分配的内存区域的泄漏或损坏问题，而非DPDK自身rte_*系列内存接口的问题。

原因分析 #

Valgrind默认会拦截并跟踪所有内存操作，但DPDK依赖大页内存、自定义内存管理框架，这些内存区域的地址空间范围大、权限设置特殊，会被Valgrind反复拦截处理，导致性能急剧下降甚至卡住。同时DPDK的内存布局逻辑与Valgrind的地址空间管理存在冲突，触发大量地址权限警告，最终卡在内存初始化阶段。

解决方案 #

1. 排除DPDK内存区域的Valgrind检测 #

从Valgrind日志中提取所有警告的内存范围，用--exclude-range参数跳过这些区域的检测，让Valgrind只关注常规malloc的内存：

valgrind --leak-check=yes \
  --exclude-range=0x140000000-0x980000000 \
  --exclude-range=0x100b967000-0x184b967000 \
  --exclude-range=0x100b967000-0x1040000000 \
  ./dpdk-testpmd -a 0000:af:00.3

如果范围较多，可将所有范围写入文件（每行一个范围，格式为0x起始地址-0x结束地址），用--exclude-range-file参数加载：

# 先创建exclude_ranges.txt文件，写入所有需要排除的内存范围
valgrind --leak-check=yes --exclude-range-file=exclude_ranges.txt ./dpdk-testpmd -a 0000:af:00.3

2. 降低Valgrind的拦截开销 #

添加以下参数减少Valgrind对DPDK内存操作的拦截和优化，提升运行效率：

• --vex-iropt-level=0：降低VEX中间代码的优化级别，减少对DPDK特殊内存操作的处理
• --trace-children=no：禁止跟踪DPDK创建的子进程（如果有）
  示例命令：

valgrind --leak-check=yes --vex-iropt-level=0 --trace-children=no \
  --exclude-range=0x140000000-0x980000000 \
  ./dpdk-testpmd -a 0000:af:00.3

3. 优化DPDK启动参数 #

限制DPDK的内存使用量，减少Valgrind需要处理的内存范围：

• --socket-mem=1024,1024：指定每个NUMA节点使用1GB内存（根据实际需求调整）
• --no-telemetry：关闭DPDK遥测功能，减少不必要的内存操作
• --disable-hugepages：如果应用允许，禁用大页（部分DPDK功能依赖大页，需谨慎使用）
  示例命令：

valgrind --leak-check=yes --exclude-range=0x140000000-0x980000000 \
  ./dpdk-testpmd -a 0000:af:00.3 --socket-mem=1024,1024 --no-telemetry

4. 编译DPDK时启用Valgrind兼容选项 #

如果可以重新编译DPDK，添加CONFIG_RTE_VALGRIND=y配置，让DPDK在Valgrind环境下自动调整内存操作逻辑，避免冲突：

# 进入DPDK源码目录
meson setup build -Dvalgrind=true
ninja -C build

使用编译后的DPDK库链接应用后，再用Valgrind运行即可。

验证 #

按照上述方法运行后，DPDK程序可正常启动，Valgrind仅检测常规malloc/free分配的内存区域，不会再卡在启动阶段，满足调试需求。

内容的提问来源于stack exchange，提问作者ssrini78