Hi Yang, On Thu, Aug 29, 2024 at 5:35 AM Yang Ruibin <11162571@xxxxxxxx> wrote: > Increase the reference count by calling pci_get_slot(), and remember to > decrement the reference count by calling pci_dev_put(). > > Signed-off-by: Yang Ruibin <11162571@xxxxxxxx> Thanks for your patch, which is now commit 8a0ec8c2d736961f ("spi: Insert the missing pci_dev_put()before return") in spi/for-next. > --- a/drivers/spi/spi-pxa2xx-pci.c > +++ b/drivers/spi/spi-pxa2xx-pci.c > @@ -146,8 +146,10 @@ static int lpss_spi_setup(struct pci_dev *dev, struct pxa2xx_spi_controller *c) > c->num_chipselect = 1; > > ret = pxa2xx_spi_pci_clk_register(dev, ssp, 50000000); > - if (ret) > + if (ret) { > + pci_dev_put(dma_dev); dma_dev is still uninitialized at this point. > return ret; > + } > > dma_dev = pci_get_slot(dev->bus, PCI_DEVFN(PCI_SLOT(dev->devfn), 0)); dma_dev is initialized only here... > ret = devm_add_action_or_reset(&dev->dev, lpss_dma_put_device, dma_dev); ... and freed automatically by lpss_dma_put_device() in case of any later failures since commit 609d7ffdc42199a0 ("spi: pxa2xx-pci: Balance reference count for PCI DMA device") in v5.18. > @@ -222,8 +224,10 @@ static int mrfld_spi_setup(struct pci_dev *dev, struct pxa2xx_spi_controller *c) > } > > ret = pxa2xx_spi_pci_clk_register(dev, ssp, 25000000); > - if (ret) > + if (ret) { > + pci_dev_put(dma_dev); > return ret; > + } > > dma_dev = pci_get_slot(dev->bus, PCI_DEVFN(21, 0)); > ret = devm_add_action_or_reset(&dev->dev, lpss_dma_put_device, dma_dev); Likewise. Hence this patch is not needed, and introduced two bugs. Gr{oetje,eeting}s, Geert -- Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@xxxxxxxxxxxxxx In personal conversations with technical people, I call myself a hacker. But when I'm talking to journalists I just say "programmer" or something like that. -- Linus Torvalds