* [PATCH] PCI/MSI: add iounmap in msix_capability_init()
@ 2026-01-25 14:44 Haoxiang Li
2026-02-06 22:37 ` Bjorn Helgaas
` (2 more replies)
0 siblings, 3 replies; 4+ messages in thread
From: Haoxiang Li @ 2026-01-25 14:44 UTC (permalink / raw)
To: bhelgaas, tglx, jgross, hans.zhang, himanshu.madhani, roger.pau,
chrisl
Cc: linux-pci, linux-kernel, Haoxiang Li
In msix_capability_init(), if msix_setup_interrupts()
fails, call iounmap() to release the memory allocated
by msix_map_region().
Signed-off-by: Haoxiang Li <lihaoxiang@isrc.iscas.ac.cn>
---
drivers/pci/msi/msi.c | 4 +++-
1 file changed, 3 insertions(+), 1 deletion(-)
diff --git a/drivers/pci/msi/msi.c b/drivers/pci/msi/msi.c
index 34d664139f48..e010ecd9f90d 100644
--- a/drivers/pci/msi/msi.c
+++ b/drivers/pci/msi/msi.c
@@ -737,7 +737,7 @@ static int msix_capability_init(struct pci_dev *dev, struct msix_entry *entries,
ret = msix_setup_interrupts(dev, entries, nvec, affd);
if (ret)
- goto out_disable;
+ goto out_unmap;
/* Disable INTX */
pci_intx_for_msi(dev, 0);
@@ -758,6 +758,8 @@ static int msix_capability_init(struct pci_dev *dev, struct msix_entry *entries,
pcibios_free_irq(dev);
return 0;
+out_unmap:
+ iounmap(dev->msix_base);
out_disable:
dev->msix_enabled = 0;
pci_msix_clear_and_set_ctrl(dev, PCI_MSIX_FLAGS_MASKALL | PCI_MSIX_FLAGS_ENABLE, 0);
--
2.25.1
^ permalink raw reply related [flat|nested] 4+ messages in thread
* Re: [PATCH] PCI/MSI: add iounmap in msix_capability_init()
2026-01-25 14:44 [PATCH] PCI/MSI: add iounmap in msix_capability_init() Haoxiang Li
@ 2026-02-06 22:37 ` Bjorn Helgaas
2026-03-10 14:38 ` Guenter Roeck
2026-03-16 17:22 ` Guenter Roeck
2 siblings, 0 replies; 4+ messages in thread
From: Bjorn Helgaas @ 2026-02-06 22:37 UTC (permalink / raw)
To: Haoxiang Li
Cc: bhelgaas, tglx, jgross, hans.zhang, himanshu.madhani, roger.pau,
chrisl, linux-pci, linux-kernel
On Sun, Jan 25, 2026 at 10:44:52PM +0800, Haoxiang Li wrote:
> In msix_capability_init(), if msix_setup_interrupts()
> fails, call iounmap() to release the memory allocated
> by msix_map_region().
>
> Signed-off-by: Haoxiang Li <lihaoxiang@isrc.iscas.ac.cn>
> ---
> drivers/pci/msi/msi.c | 4 +++-
> 1 file changed, 3 insertions(+), 1 deletion(-)
>
> diff --git a/drivers/pci/msi/msi.c b/drivers/pci/msi/msi.c
> index 34d664139f48..e010ecd9f90d 100644
> --- a/drivers/pci/msi/msi.c
> +++ b/drivers/pci/msi/msi.c
> @@ -737,7 +737,7 @@ static int msix_capability_init(struct pci_dev *dev, struct msix_entry *entries,
>
> ret = msix_setup_interrupts(dev, entries, nvec, affd);
> if (ret)
> - goto out_disable;
> + goto out_unmap;
>
> /* Disable INTX */
> pci_intx_for_msi(dev, 0);
> @@ -758,6 +758,8 @@ static int msix_capability_init(struct pci_dev *dev, struct msix_entry *entries,
> pcibios_free_irq(dev);
> return 0;
>
> +out_unmap:
> + iounmap(dev->msix_base);
Maybe Thomas has already applied this, and if not, he should chime in.
Personally, I would prefer if msix_map_region() were renamed and made
to return the phys_addr instead of returning the ioremapped address.
That way both the ioremap() and the iounmap() on failure would be in
msix_capability_init() so it's easier to review.
> out_disable:
> dev->msix_enabled = 0;
> pci_msix_clear_and_set_ctrl(dev, PCI_MSIX_FLAGS_MASKALL | PCI_MSIX_FLAGS_ENABLE, 0);
> --
> 2.25.1
>
^ permalink raw reply [flat|nested] 4+ messages in thread
* Re: [PATCH] PCI/MSI: add iounmap in msix_capability_init()
2026-01-25 14:44 [PATCH] PCI/MSI: add iounmap in msix_capability_init() Haoxiang Li
2026-02-06 22:37 ` Bjorn Helgaas
@ 2026-03-10 14:38 ` Guenter Roeck
2026-03-16 17:22 ` Guenter Roeck
2 siblings, 0 replies; 4+ messages in thread
From: Guenter Roeck @ 2026-03-10 14:38 UTC (permalink / raw)
To: Haoxiang Li
Cc: bhelgaas, tglx, jgross, hans.zhang, himanshu.madhani, roger.pau,
chrisl, linux-pci, linux-kernel
Hi,
On Sun, Jan 25, 2026 at 10:44:52PM +0800, Haoxiang Li wrote:
> In msix_capability_init(), if msix_setup_interrupts()
> fails, call iounmap() to release the memory allocated
> by msix_map_region().
>
AI review has the following feedback:
In __msix_setup_interrupts(), `dev` is declared with the
`__free(free_msi_irqs)` cleanup attribute. If an error is returned,
pci_free_msi_irqs() is automatically called.
pci_free_msi_irqs() already unmaps dev->msix_base and sets it to NULL:
void pci_free_msi_irqs(struct pci_dev *dev)
{
...
if (dev->msix_base) {
iounmap(dev->msix_base);
dev->msix_base = NULL;
}
}
So if msix_setup_interrupts() fails, dev->msix_base has already been
unmapped and is NULL by the time we reach this new error path.
Is this extra iounmap() necessary?
I checked the code, and agree.
Thanks,
Guenter
> Signed-off-by: Haoxiang Li <lihaoxiang@isrc.iscas.ac.cn>
> ---
> drivers/pci/msi/msi.c | 4 +++-
> 1 file changed, 3 insertions(+), 1 deletion(-)
>
> diff --git a/drivers/pci/msi/msi.c b/drivers/pci/msi/msi.c
> index 34d664139f48..e010ecd9f90d 100644
> --- a/drivers/pci/msi/msi.c
> +++ b/drivers/pci/msi/msi.c
> @@ -737,7 +737,7 @@ static int msix_capability_init(struct pci_dev *dev, struct msix_entry *entries,
>
> ret = msix_setup_interrupts(dev, entries, nvec, affd);
> if (ret)
> - goto out_disable;
> + goto out_unmap;
>
> /* Disable INTX */
> pci_intx_for_msi(dev, 0);
> @@ -758,6 +758,8 @@ static int msix_capability_init(struct pci_dev *dev, struct msix_entry *entries,
> pcibios_free_irq(dev);
> return 0;
>
> +out_unmap:
> + iounmap(dev->msix_base);
> out_disable:
> dev->msix_enabled = 0;
> pci_msix_clear_and_set_ctrl(dev, PCI_MSIX_FLAGS_MASKALL | PCI_MSIX_FLAGS_ENABLE, 0);
> --
> 2.25.1
>
^ permalink raw reply [flat|nested] 4+ messages in thread
* Re: [PATCH] PCI/MSI: add iounmap in msix_capability_init()
2026-01-25 14:44 [PATCH] PCI/MSI: add iounmap in msix_capability_init() Haoxiang Li
2026-02-06 22:37 ` Bjorn Helgaas
2026-03-10 14:38 ` Guenter Roeck
@ 2026-03-16 17:22 ` Guenter Roeck
2 siblings, 0 replies; 4+ messages in thread
From: Guenter Roeck @ 2026-03-16 17:22 UTC (permalink / raw)
To: Haoxiang Li
Cc: bhelgaas, tglx, jgross, hans.zhang, himanshu.madhani, roger.pau,
chrisl, linux-pci, linux-kernel
Hi,
On Sun, Jan 25, 2026 at 10:44:52PM +0800, Haoxiang Li wrote:
> In msix_capability_init(), if msix_setup_interrupts()
> fails, call iounmap() to release the memory allocated
> by msix_map_region().
>
Really ? AI review provided the following feedback. Looking into the code,
I think it has a point.
Thanks,
Guenter
> Signed-off-by: Haoxiang Li <lihaoxiang@isrc.iscas.ac.cn>
> ---
> drivers/pci/msi/msi.c | 4 +++-
> 1 file changed, 3 insertions(+), 1 deletion(-)
>
> diff --git a/drivers/pci/msi/msi.c b/drivers/pci/msi/msi.c
> index 34d664139f48..e010ecd9f90d 100644
> --- a/drivers/pci/msi/msi.c
> +++ b/drivers/pci/msi/msi.c
> @@ -737,7 +737,7 @@ static int msix_capability_init(struct pci_dev *dev, struct msix_entry *entries,
>
> ret = msix_setup_interrupts(dev, entries, nvec, affd);
> if (ret)
> - goto out_disable;
> + goto out_unmap;
Does msix_setup_interrupts() already unmap this region? If it fails,
msix_setup_interrupts() calls pci_free_msi_irqs(), which appears to already
call iounmap(dev->msix_base) and sets dev->msix_base to NULL.
>
> /* Disable INTX */
> pci_intx_for_msi(dev, 0);
> @@ -758,6 +758,8 @@ static int msix_capability_init(struct pci_dev *dev, struct msix_entry *entries,
> pcibios_free_irq(dev);
> return 0;
>
> +out_unmap:
> + iounmap(dev->msix_base);
If dev->msix_base was already set to NULL by pci_free_msi_irqs() during the
msix_setup_interrupts() error path, does this result in iounmap(NULL)?
> out_disable:
> dev->msix_enabled = 0;
> pci_msix_clear_and_set_ctrl(dev, PCI_MSIX_FLAGS_MASKALL | PCI_MSIX_FLAGS_ENABLE, 0);
> --
> 2.25.1
>
^ permalink raw reply [flat|nested] 4+ messages in thread
end of thread, other threads:[~2026-03-16 17:22 UTC | newest]
Thread overview: 4+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2026-01-25 14:44 [PATCH] PCI/MSI: add iounmap in msix_capability_init() Haoxiang Li
2026-02-06 22:37 ` Bjorn Helgaas
2026-03-10 14:38 ` Guenter Roeck
2026-03-16 17:22 ` Guenter Roeck
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox