2019-06-19 06:36:53

by Simon Sandström

[permalink] [raw]
Subject: [PATCH] staging: kpc2000: simplify error handling in kp2000_pcie_probe

We can get rid of a few iounmaps in the middle of the function by
re-ordering the error handling labels and adding two new labels.

Signed-off-by: Simon Sandström <[email protected]>
---

This change has not been tested besides by compiling. It might be good
took take an extra look to make sure that I got everything right.

Also, this change was proposed by Dan Carpenter. Should I add anything
in the commit message to show this?

- Simon

drivers/staging/kpc2000/kpc2000/core.c | 22 ++++++++++------------
1 file changed, 10 insertions(+), 12 deletions(-)

diff --git a/drivers/staging/kpc2000/kpc2000/core.c b/drivers/staging/kpc2000/kpc2000/core.c
index 610ea549d240..cb05cca687e1 100644
--- a/drivers/staging/kpc2000/kpc2000/core.c
+++ b/drivers/staging/kpc2000/kpc2000/core.c
@@ -351,12 +351,11 @@ static int kp2000_pcie_probe(struct pci_dev *pdev,

err = pci_request_region(pcard->pdev, REG_BAR, KP_DRIVER_NAME_KP2000);
if (err) {
- iounmap(pcard->regs_bar_base);
dev_err(&pcard->pdev->dev,
"probe: failed to acquire PCI region (%d)\n",
err);
err = -ENODEV;
- goto err_disable_device;
+ goto err_unmap_regs;
}

pcard->regs_base_resource.start = reg_bar_phys_addr;
@@ -374,7 +373,7 @@ static int kp2000_pcie_probe(struct pci_dev *pdev,
dev_err(&pcard->pdev->dev,
"probe: DMA_BAR could not remap memory to virtual space\n");
err = -ENODEV;
- goto err_unmap_regs;
+ goto err_release_regs;
}
dev_dbg(&pcard->pdev->dev,
"probe: DMA_BAR virt hardware address start [%p]\n",
@@ -384,11 +383,10 @@ static int kp2000_pcie_probe(struct pci_dev *pdev,

err = pci_request_region(pcard->pdev, DMA_BAR, "kp2000_pcie");
if (err) {
- iounmap(pcard->dma_bar_base);
dev_err(&pcard->pdev->dev,
"probe: failed to acquire PCI region (%d)\n", err);
err = -ENODEV;
- goto err_unmap_regs;
+ goto err_unmap_dma;
}

pcard->dma_base_resource.start = dma_bar_phys_addr;
@@ -400,7 +398,7 @@ static int kp2000_pcie_probe(struct pci_dev *pdev,
pcard->sysinfo_regs_base = pcard->regs_bar_base;
err = read_system_regs(pcard);
if (err)
- goto err_unmap_dma;
+ goto err_release_dma;

// Disable all "user" interrupts because they're not used yet.
writeq(0xFFFFFFFFFFFFFFFF,
@@ -438,14 +436,14 @@ static int kp2000_pcie_probe(struct pci_dev *pdev,
if (err) {
dev_err(&pcard->pdev->dev,
"CANNOT use DMA mask %0llx\n", DMA_BIT_MASK(64));
- goto err_unmap_dma;
+ goto err_release_dma;
}
dev_dbg(&pcard->pdev->dev,
"Using DMA mask %0llx\n", dma_get_mask(PCARD_TO_DEV(pcard)));

err = pci_enable_msi(pcard->pdev);
if (err < 0)
- goto err_unmap_dma;
+ goto err_release_dma;

rv = request_irq(pcard->pdev->irq, kp2000_irq_handler, IRQF_SHARED,
pcard->name, pcard);
@@ -478,14 +476,14 @@ static int kp2000_pcie_probe(struct pci_dev *pdev,
free_irq(pcard->pdev->irq, pcard);
err_disable_msi:
pci_disable_msi(pcard->pdev);
+err_release_dma:
+ pci_release_region(pdev, DMA_BAR);
err_unmap_dma:
iounmap(pcard->dma_bar_base);
- pci_release_region(pdev, DMA_BAR);
- pcard->dma_bar_base = NULL;
+err_release_regs:
+ pci_release_region(pdev, REG_BAR);
err_unmap_regs:
iounmap(pcard->regs_bar_base);
- pci_release_region(pdev, REG_BAR);
- pcard->regs_bar_base = NULL;
err_disable_device:
pci_disable_device(pcard->pdev);
err_remove_ida:
--
2.20.1


2019-06-19 06:54:59

by Dan Carpenter

[permalink] [raw]
Subject: Re: [PATCH] staging: kpc2000: simplify error handling in kp2000_pcie_probe

On Wed, Jun 19, 2019 at 08:36:07AM +0200, Simon Sandstr?m wrote:
> We can get rid of a few iounmaps in the middle of the function by
> re-ordering the error handling labels and adding two new labels.
>
> Signed-off-by: Simon Sandstr?m <[email protected]>
> ---
>
> This change has not been tested besides by compiling. It might be good
> took take an extra look to make sure that I got everything right.
>

You have the right instincts that when something looks really
complicated that's probably for a reason. That attitude will serve you
well in the future! But in this case it's staging code so the original
code is just strange.

Reviewed-by: Dan Carpenter <[email protected]>

> Also, this change was proposed by Dan Carpenter. Should I add anything
> in the commit message to show this?

There is a Suggested-by: tag for this, but don't resend because I don't
care and I've already reviewed this version so I don't want to review
the patch again.

regards,
dan carpenter