On 2024/9/18 15:28, David Marchand wrote:
> On Wed, Sep 18, 2024 at 8:10 AM WanRenyong <wa...@yunsilicon.com> wrote:
>> +static const struct rte_pci_id xsc_ethdev_pci_id_map[] = {
>> + { RTE_PCI_DEVICE(XSC_PCI_VENDOR_ID, XSC_PCI_DEV_ID_MS) },
> You need to null terminate this array with something like:
> { .vendor_id = 0, /* sentinel */ },
>
> Otherwise the bus pci code may read a next symbol or data present in
> the .data section.
>
> ASan caught this issue when running the unit tests:
>
> ==70261==ERROR: AddressSanitizer: global-buffer-overflow on address
> 0x7f8e46bf45f0 at pc 0x7f8e56be523b bp 0x7ffe2ef88ca0 sp
> 0x7ffe2ef88c98
> READ of size 2 at 0x7f8e46bf45f0 thread T0
> #0 0x7f8e56be523a in rte_pci_match
> /home/runner/work/dpdk/dpdk/build/../drivers/bus/pci/pci_common.c:178:47
> #1 0x7f8e56be523a in rte_pci_probe_one_driver
> /home/runner/work/dpdk/dpdk/build/../drivers/bus/pci/pci_common.c:223:7
> #2 0x7f8e56be523a in pci_probe_all_drivers
> /home/runner/work/dpdk/dpdk/build/../drivers/bus/pci/pci_common.c:391:8
> #3 0x7f8e56be3297 in pci_probe
> /home/runner/work/dpdk/dpdk/build/../drivers/bus/pci/pci_common.c:418:9
> #4 0x7f8e56fe9ea8 in rte_bus_probe
> /home/runner/work/dpdk/dpdk/build/../lib/eal/common/eal_common_bus.c:78:9
> #5 0x7f8e570580d1 in rte_eal_init
> /home/runner/work/dpdk/dpdk/build/../lib/eal/linux/eal.c:1288:6
> #6 0x5573a597d65d in main
> /home/runner/work/dpdk/dpdk/build/../app/test/test.c:145:9
> #7 0x7f8e55829d8f in __libc_start_call_main
> csu/../sysdeps/nptl/libc_start_call_main.h:58:16
> #8 0x7f8e55829e3f in __libc_start_main csu/../csu/libc-start.c:392:3
> #9 0x5573a58bf114 in _start
> (/home/runner/work/dpdk/dpdk/build/app/dpdk-test+0x1e9114) (BuildId:
> 8d4741d712c15395a67005124e1f908d96acf7ff)
>
>
> 172 int
> 173 rte_pci_match(const struct rte_pci_driver *pci_drv,
> 174 const struct rte_pci_device *pci_dev)
> 175 {
> 176 const struct rte_pci_id *id_table;
> 177
> 178 for (id_table = pci_drv->id_table; id_table->vendor_id !=
> 0;
> 179 id_table++) {
>
>
Thanks for review, will fix it in the next version.
--
Thanks,
WanRenyong
