From: Julia Lawall <[email protected]>
The out_msi_disable label should be before cleanup_nomem to additionally
benefit from the call to iounmap. Subsequent gotos are adjusted to go to
out_msi_disable instead of cleanup_nomem, which now follows it. This is
safe because pci_disable_msi does nothing if pci_enable_msi was not called.
A simplified version of the semantic match that finds this problem is as
follows: (http://coccinelle.lip6.fr/)
// <smpl>
@r@
expression e1,e2;
statement S;
@@
e1 = pci_ioremap_bar(...);
... when != e1 = e2
when != iounmap(e1)
when any
(
if (<+...e1...+>) S
|
if(...) { ... return 0; }
|
if (...) { ... when != iounmap(e1)
when != if (...) { ... iounmap(e1) ... }
* return ...;
} else S
)
// </smpl>
Signed-off-by: Julia Lawall <[email protected]>
---
drivers/scsi/ipr.c | 12 ++++++------
1 file changed, 6 insertions(+), 6 deletions(-)
diff --git a/drivers/scsi/ipr.c b/drivers/scsi/ipr.c
index 888086c..8d63630 100644
--- a/drivers/scsi/ipr.c
+++ b/drivers/scsi/ipr.c
@@ -8778,14 +8778,14 @@ static int __devinit ipr_probe_ioa(struct pci_dev *pdev,
if (rc != PCIBIOS_SUCCESSFUL) {
dev_err(&pdev->dev, "Failed to save PCI config space\n");
rc = -EIO;
- goto cleanup_nomem;
+ goto out_msi_disable;
}
if ((rc = ipr_save_pcix_cmd_reg(ioa_cfg)))
- goto cleanup_nomem;
+ goto out_msi_disable;
if ((rc = ipr_set_pcix_cmd_reg(ioa_cfg)))
- goto cleanup_nomem;
+ goto out_msi_disable;
if (ioa_cfg->sis64)
ioa_cfg->cfg_table_size = (sizeof(struct ipr_config_table_hdr64)
@@ -8800,7 +8800,7 @@ static int __devinit ipr_probe_ioa(struct pci_dev *pdev,
if (rc < 0) {
dev_err(&pdev->dev,
"Couldn't allocate enough memory for device driver!\n");
- goto cleanup_nomem;
+ goto out_msi_disable;
}
/*
@@ -8845,10 +8845,10 @@ out:
cleanup_nolog:
ipr_free_mem(ioa_cfg);
-cleanup_nomem:
- iounmap(ipr_regs);
out_msi_disable:
pci_disable_msi(pdev);
+cleanup_nomem:
+ iounmap(ipr_regs);
out_release_regions:
pci_release_regions(pdev);
out_scsi_host_put:
Acked-by: Brian King <[email protected]>
--
Brian King
Linux on Power Virtualization
IBM Linux Technology Center
On 05/31/2011 07:16 AM, Julia Lawall wrote:
> From: Julia Lawall <[email protected]>
>
> The out_msi_disable label should be before cleanup_nomem to additionally
> benefit from the call to iounmap.
Yes, this is a problem. I propose the following patch instead.
---
In the case where ipr_test_msi returns an error that is not -EOPNOTSUPP, the
execution jumps to the out_msi_disable label. This misses the call to iounmap
for ipr_regs which was initialized earlier.
The fix is to do the call to pci_disable_msi when the error case is detected
and then goto the cleanup_nomem label if the error is not -EOPNOTSUPP.
Signed-off-by: Wayne Boyer <[email protected]>
---
drivers/scsi/ipr.c | 10 ++++------
1 file changed, 4 insertions(+), 6 deletions(-)
Index: b/drivers/scsi/ipr.c
===================================================================
--- a/drivers/scsi/ipr.c 2011-06-09 08:14:44.927740117 -0700
+++ b/drivers/scsi/ipr.c 2011-06-09 08:50:10.105630466 -0700
@@ -8763,11 +8763,11 @@ static int __devinit ipr_probe_ioa(struc
/* Enable MSI style interrupts if they are supported. */
if (ioa_cfg->ipr_chip->intr_type == IPR_USE_MSI && !pci_enable_msi(pdev)) {
rc = ipr_test_msi(ioa_cfg, pdev);
- if (rc == -EOPNOTSUPP)
+ if (rc) {
pci_disable_msi(pdev);
- else if (rc)
- goto out_msi_disable;
- else
+ if (rc != -EOPNOTSUPP)
+ goto cleanup_nomem;
+ } else
dev_info(&pdev->dev, "MSI enabled with IRQ: %d\n", pdev->irq);
} else if (ipr_debug)
dev_info(&pdev->dev, "Cannot enable MSI.\n");
@@ -8847,8 +8847,6 @@ cleanup_nolog:
ipr_free_mem(ioa_cfg);
cleanup_nomem:
iounmap(ipr_regs);
-out_msi_disable:
- pci_disable_msi(pdev);
out_release_regions:
pci_release_regions(pdev);
out_scsi_host_put:
--
Wayne Boyer
IBM - Beaverton, Oregon
LTC S/W Development
(503) 578-5236, T/L 775-5236
On 06/09/2011 10:51 AM, Wayne Boyer wrote:
> On 05/31/2011 07:16 AM, Julia Lawall wrote:
>> From: Julia Lawall <[email protected]>
>>
>> The out_msi_disable label should be before cleanup_nomem to additionally
>> benefit from the call to iounmap.
>
> Yes, this is a problem. I propose the following patch instead.
By removing the out_msi_disable label, if you fail initialization later
on and goto cleanup_nomem, you will end up leaving MSI enabled when you exit
with this patch.
-Brian
>
> ---
>
> In the case where ipr_test_msi returns an error that is not -EOPNOTSUPP, the
> execution jumps to the out_msi_disable label. This misses the call to iounmap
> for ipr_regs which was initialized earlier.
>
> The fix is to do the call to pci_disable_msi when the error case is detected
> and then goto the cleanup_nomem label if the error is not -EOPNOTSUPP.
>
> Signed-off-by: Wayne Boyer <[email protected]>
> ---
>
> drivers/scsi/ipr.c | 10 ++++------
> 1 file changed, 4 insertions(+), 6 deletions(-)
>
> Index: b/drivers/scsi/ipr.c
> ===================================================================
> --- a/drivers/scsi/ipr.c 2011-06-09 08:14:44.927740117 -0700
> +++ b/drivers/scsi/ipr.c 2011-06-09 08:50:10.105630466 -0700
> @@ -8763,11 +8763,11 @@ static int __devinit ipr_probe_ioa(struc
> /* Enable MSI style interrupts if they are supported. */
> if (ioa_cfg->ipr_chip->intr_type == IPR_USE_MSI && !pci_enable_msi(pdev)) {
> rc = ipr_test_msi(ioa_cfg, pdev);
> - if (rc == -EOPNOTSUPP)
> + if (rc) {
> pci_disable_msi(pdev);
> - else if (rc)
> - goto out_msi_disable;
> - else
> + if (rc != -EOPNOTSUPP)
> + goto cleanup_nomem;
> + } else
> dev_info(&pdev->dev, "MSI enabled with IRQ: %d\n", pdev->irq);
> } else if (ipr_debug)
> dev_info(&pdev->dev, "Cannot enable MSI.\n");
> @@ -8847,8 +8847,6 @@ cleanup_nolog:
> ipr_free_mem(ioa_cfg);
> cleanup_nomem:
> iounmap(ipr_regs);
> -out_msi_disable:
> - pci_disable_msi(pdev);
> out_release_regions:
> pci_release_regions(pdev);
> out_scsi_host_put:
>
>
--
Brian King
Linux on Power Virtualization
IBM Linux Technology Center
On Tue, 14 Jun 2011, Brian King wrote:
> On 06/09/2011 10:51 AM, Wayne Boyer wrote:
> > On 05/31/2011 07:16 AM, Julia Lawall wrote:
> >> From: Julia Lawall <[email protected]>
> >>
> >> The out_msi_disable label should be before cleanup_nomem to additionally
> >> benefit from the call to iounmap.
> >
> > Yes, this is a problem. I propose the following patch instead.
>
> By removing the out_msi_disable label, if you fail initialization later
> on and goto cleanup_nomem, you will end up leaving MSI enabled when you exit
> with this patch.
I agree.
julia
> -Brian
>
>
>
> >
> > ---
> >
> > In the case where ipr_test_msi returns an error that is not -EOPNOTSUPP, the
> > execution jumps to the out_msi_disable label. This misses the call to iounmap
> > for ipr_regs which was initialized earlier.
> >
> > The fix is to do the call to pci_disable_msi when the error case is detected
> > and then goto the cleanup_nomem label if the error is not -EOPNOTSUPP.
> >
> > Signed-off-by: Wayne Boyer <[email protected]>
> > ---
> >
> > drivers/scsi/ipr.c | 10 ++++------
> > 1 file changed, 4 insertions(+), 6 deletions(-)
> >
> > Index: b/drivers/scsi/ipr.c
> > ===================================================================
> > --- a/drivers/scsi/ipr.c 2011-06-09 08:14:44.927740117 -0700
> > +++ b/drivers/scsi/ipr.c 2011-06-09 08:50:10.105630466 -0700
> > @@ -8763,11 +8763,11 @@ static int __devinit ipr_probe_ioa(struc
> > /* Enable MSI style interrupts if they are supported. */
> > if (ioa_cfg->ipr_chip->intr_type == IPR_USE_MSI && !pci_enable_msi(pdev)) {
> > rc = ipr_test_msi(ioa_cfg, pdev);
> > - if (rc == -EOPNOTSUPP)
> > + if (rc) {
> > pci_disable_msi(pdev);
> > - else if (rc)
> > - goto out_msi_disable;
> > - else
> > + if (rc != -EOPNOTSUPP)
> > + goto cleanup_nomem;
> > + } else
> > dev_info(&pdev->dev, "MSI enabled with IRQ: %d\n", pdev->irq);
> > } else if (ipr_debug)
> > dev_info(&pdev->dev, "Cannot enable MSI.\n");
> > @@ -8847,8 +8847,6 @@ cleanup_nolog:
> > ipr_free_mem(ioa_cfg);
> > cleanup_nomem:
> > iounmap(ipr_regs);
> > -out_msi_disable:
> > - pci_disable_msi(pdev);
> > out_release_regions:
> > pci_release_regions(pdev);
> > out_scsi_host_put:
> >
> >
>
>
> --
> Brian King
> Linux on Power Virtualization
> IBM Linux Technology Center
>
>
> --
> To unsubscribe from this list: send the line "unsubscribe kernel-janitors" in
> the body of a message to [email protected]
> More majordomo info at http://vger.kernel.org/majordomo-info.html
>
On 06/15/2011 01:50 AM, Julia Lawall wrote:
> On Tue, 14 Jun 2011, Brian King wrote:
>
>> On 06/09/2011 10:51 AM, Wayne Boyer wrote:
>>> On 05/31/2011 07:16 AM, Julia Lawall wrote:
>>>> From: Julia Lawall <[email protected]>
>>>>
>>>> The out_msi_disable label should be before cleanup_nomem to additionally
>>>> benefit from the call to iounmap.
>>>
>>> Yes, this is a problem. I propose the following patch instead.
>>
>> By removing the out_msi_disable label, if you fail initialization later
>> on and goto cleanup_nomem, you will end up leaving MSI enabled when you exit
>> with this patch.
>
> I agree.
>
> julia
>
I also agree. Please disregard my patch.
--
Wayne Boyer
IBM - Beaverton, Oregon
LTC S/W Development
(503) 578-5236, T/L 775-5236