Hi Shuai,
On Tue, Feb 11, 2025 at 4:02 PM Shuai Xue <xueshuai@xxxxxxxxxxxxxxxxx> wrote:
>
>
>
> 在 2025/2/8 18:40, Yunhui Cui 写道:
> > During platform_device_register, wrongly using struct device
> > pci_dev as platform_data caused a kmemdup copy of pci_dev. Worse
> > still, accessing the duplicated device leads to list corruption as its
> > mutex content (e.g., list, magic) remains the same as the original.
> >
> > Signed-off-by: Yunhui Cui <cuiyunhui@xxxxxxxxxxxxx>
> > ---
> > drivers/perf/dwc_pcie_pmu.c | 20 +++++++++++++-------
> > 1 file changed, 13 insertions(+), 7 deletions(-)
> >
> > diff --git a/drivers/perf/dwc_pcie_pmu.c b/drivers/perf/dwc_pcie_pmu.c
> > index 19fa2ba8dd67..4f6599e32bba 100644
> > --- a/drivers/perf/dwc_pcie_pmu.c
> > +++ b/drivers/perf/dwc_pcie_pmu.c
> > @@ -565,9 +565,7 @@ static int dwc_pcie_register_dev(struct pci_dev *pdev)
> > u32 sbdf;
> >
> > sbdf = (pci_domain_nr(pdev->bus) << 16) | PCI_DEVID(pdev->bus->number, pdev->devfn);
> > - plat_dev = platform_device_register_data(NULL, "dwc_pcie_pmu", sbdf,
> > - pdev, sizeof(*pdev));
> > -
> > + plat_dev = platform_device_register_simple("platform_dwc_pcie", sbdf, NULL, 0);
> > if (IS_ERR(plat_dev))
> > return PTR_ERR(plat_dev);
> >
> > @@ -616,18 +614,26 @@ static struct notifier_block dwc_pcie_pmu_nb = {
> >
> > static int dwc_pcie_pmu_probe(struct platform_device *plat_dev)
> > {
> > - struct pci_dev *pdev = plat_dev->dev.platform_data;
> > + struct pci_dev *pdev;
> > struct dwc_pcie_pmu *pcie_pmu;
> > char *name;
> > u32 sbdf;
> > u16 vsec;
> > int ret;
> >
> > + sbdf = plat_dev->id;
> > + pdev = pci_get_domain_bus_and_slot(sbdf >> 16, PCI_BUS_NUM(sbdf & 0xffff),
> > + sbdf & 0xff);
> > + if (!pdev) {
> > + pr_err("No pdev found for the sbdf 0x%x\n", sbdf);
> > + return -ENODEV;
> > + }
> > +
> > vsec = dwc_pcie_des_cap(pdev);
> > if (!vsec)
> > return -ENODEV;
>
> pci_dev_put(pdev) should move ahead to aovid return here.
>
> >
> > - sbdf = plat_dev->id;
> > + pci_dev_put(pdev);
> > name = devm_kasprintf(&plat_dev->dev, GFP_KERNEL, "dwc_rootport_%x", sbdf);
> > if (!name)
> > return -ENOMEM;
> > @@ -642,7 +648,7 @@ static int dwc_pcie_pmu_probe(struct platform_device *plat_dev)
> > pcie_pmu->on_cpu = -1;
> > pcie_pmu->pmu = (struct pmu){
> > .name = name,
> > - .parent = &pdev->dev,
> > + .parent = &plat_dev->dev,
> > .module = THIS_MODULE,
> > .attr_groups = dwc_pcie_attr_groups,
> > .capabilities = PERF_PMU_CAP_NO_EXCLUDE,
> > @@ -729,7 +735,7 @@ static int dwc_pcie_pmu_offline_cpu(unsigned int cpu, struct hlist_node *cpuhp_n
> >
> > static struct platform_driver dwc_pcie_pmu_driver = {
> > .probe = dwc_pcie_pmu_probe,
> > - .driver = {.name = "dwc_pcie_pmu",},
> > + .driver = {.name = "platform_dwc_pcie",},
>
> Aha, it is very difficult to come up with a name that satisfies everyone. The
> original name uses the '_pmu' suffix to follow the unwritten convention of
> other PMU drivers.
>
> Personally, I think the original name is more appropriate, but I'll leave the
> decision to @Will.
Since Will hasn't replied, I'll update to the next version to keep the
original name.
>
> Thanks.
> Best Regards.
> Shuai
Thanks,
Yunhui
