2021-10-03 09:15:10

by Max Gurtovoy

[permalink] [raw]
Subject: [PATCH v2 1/2] driver core: use NUMA_NO_NODE during device_initialize

Don't use (-1) constant for setting initial device node. Instead, use
the generic NUMA_NO_NODE definition to indicate that "no node id
specified".

Reviewed-by: Stefan Hajnoczi <[email protected]>
Signed-off-by: Max Gurtovoy <[email protected]>
---

changes from v1:
- added patch 2/2 (Krzysztof)
- added Reviewed-by signature (Stefan)

---
drivers/base/core.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/base/core.c b/drivers/base/core.c
index e65dd803a453..2b4b46f6c676 100644
--- a/drivers/base/core.c
+++ b/drivers/base/core.c
@@ -2838,7 +2838,7 @@ void device_initialize(struct device *dev)
spin_lock_init(&dev->devres_lock);
INIT_LIST_HEAD(&dev->devres_head);
device_pm_init(dev);
- set_dev_node(dev, -1);
+ set_dev_node(dev, NUMA_NO_NODE);
#ifdef CONFIG_GENERIC_MSI_IRQ
raw_spin_lock_init(&dev->msi_lock);
INIT_LIST_HEAD(&dev->msi_list);
--
2.18.1


2021-10-03 09:16:19

by Max Gurtovoy

[permalink] [raw]
Subject: [PATCH v2 2/2] PCI/sysfs: use NUMA_NO_NODE macro

Use the proper macro instead of hard-coded (-1) value.

Suggested-by: Krzysztof Wilczyński <[email protected]>
Signed-off-by: Max Gurtovoy <[email protected]>
---
drivers/pci/pci-sysfs.c | 4 ++--
1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/drivers/pci/pci-sysfs.c b/drivers/pci/pci-sysfs.c
index 7fb5cd17cc98..b21065222c87 100644
--- a/drivers/pci/pci-sysfs.c
+++ b/drivers/pci/pci-sysfs.c
@@ -81,8 +81,8 @@ static ssize_t pci_dev_show_local_cpu(struct device *dev, bool list,
const struct cpumask *mask;

#ifdef CONFIG_NUMA
- mask = (dev_to_node(dev) == -1) ? cpu_online_mask :
- cpumask_of_node(dev_to_node(dev));
+ mask = (dev_to_node(dev) == NUMA_NO_NODE) ? cpu_online_mask :
+ cpumask_of_node(dev_to_node(dev));
#else
mask = cpumask_of_pcibus(to_pci_dev(dev)->bus);
#endif
--
2.18.1

2021-10-04 01:52:49

by Krzysztof Wilczyński

[permalink] [raw]
Subject: Re: [PATCH v2 2/2] PCI/sysfs: use NUMA_NO_NODE macro

[+cc Bjorn as he has strong code formatting preference in the PCI tree]

Hi Max,

> Use the proper macro instead of hard-coded (-1) value.
>
> Suggested-by: Krzysztof Wilczyński <[email protected]>
> Signed-off-by: Max Gurtovoy <[email protected]>

Thank you for taking care of this! Much appreciated!

> ---
> drivers/pci/pci-sysfs.c | 4 ++--
> 1 file changed, 2 insertions(+), 2 deletions(-)
>
> diff --git a/drivers/pci/pci-sysfs.c b/drivers/pci/pci-sysfs.c
> index 7fb5cd17cc98..b21065222c87 100644
> --- a/drivers/pci/pci-sysfs.c
> +++ b/drivers/pci/pci-sysfs.c
> @@ -81,8 +81,8 @@ static ssize_t pci_dev_show_local_cpu(struct device *dev, bool list,
> const struct cpumask *mask;
>
> #ifdef CONFIG_NUMA
> - mask = (dev_to_node(dev) == -1) ? cpu_online_mask :
> - cpumask_of_node(dev_to_node(dev));
> + mask = (dev_to_node(dev) == NUMA_NO_NODE) ? cpu_online_mask :
> + cpumask_of_node(dev_to_node(dev));

Oh this somewhat awkward indentation we have with this ternary now,
and so I wonder if either:

mask = (dev_to_node(dev) == NUMA_NO_NODE) ?
cpu_online_mask : cpumask_of_node(dev_to_node(dev));

Or, perhaps (yes, a few more lines):

if (dev_to_node(dev) == NUMA_NO_NODE)
mask = cpu_online_mask;
else
mask = cpumask_of_node(node);

Would be easier on the eyes, so to speak. What do you think (not a problem
if you don't want to change anything, thoguh)?

Thank you!

Reviewed-by: Krzysztof Wilczyński <[email protected]>

Krzysztof

2021-10-04 08:14:54

by Stefan Hajnoczi

[permalink] [raw]
Subject: Re: [PATCH v2 2/2] PCI/sysfs: use NUMA_NO_NODE macro

On Sun, Oct 03, 2021 at 12:13:44PM +0300, Max Gurtovoy wrote:
> Use the proper macro instead of hard-coded (-1) value.
>
> Suggested-by: Krzysztof Wilczyński <[email protected]>
> Signed-off-by: Max Gurtovoy <[email protected]>
> ---
> drivers/pci/pci-sysfs.c | 4 ++--
> 1 file changed, 2 insertions(+), 2 deletions(-)

Reviewed-by: Stefan Hajnoczi <[email protected]>


Attachments:
(No filename) (383.00 B)
signature.asc (499.00 B)
Download all attachments

2021-10-04 19:17:07

by Krzysztof Wilczyński

[permalink] [raw]
Subject: Re: [PATCH v2 2/2] PCI/sysfs: use NUMA_NO_NODE macro

Hi Max,

[...]
> > #ifdef CONFIG_NUMA
> > - mask = (dev_to_node(dev) == -1) ? cpu_online_mask :
> > - cpumask_of_node(dev_to_node(dev));
> > + mask = (dev_to_node(dev) == NUMA_NO_NODE) ? cpu_online_mask :
> > + cpumask_of_node(dev_to_node(dev));
>
> Oh this somewhat awkward indentation we have with this ternary now,
> and so I wonder if either:
>
> mask = (dev_to_node(dev) == NUMA_NO_NODE) ?
> cpu_online_mask : cpumask_of_node(dev_to_node(dev));
>
> Or, perhaps (yes, a few more lines):
>
> if (dev_to_node(dev) == NUMA_NO_NODE)
> mask = cpu_online_mask;
> else
> mask = cpumask_of_node(node);

Doh! I should be cpumask_of_node(dev_to_node(dev)) in the above, of
course. Apologies! Albeit, v3 you sent looks great! Thank you!

Krzysztof