Re: [PATCH v5 2/4] staging: octeon: replace pr_warn with dev_warn in fill path

[Ayush Mukkanwar]

On Mon, May 4, 2026 at 8:26 PM Greg KH <gregkh@xxxxxxxxxxxxxxxxxxx> wrote:
>
> On Mon, Apr 27, 2026 at 08:45:53PM +0530, Ayush Mukkanwar wrote:
> > Replace pr_warn() with dev_warn() in cvm_oct_fill_hw_memory() to
> > include device information in log messages.
> >
> > To make the device pointer accessible from the workqueue callback,
> > introduce struct octeon_ethernet_platform to hold both a struct
> > device pointer and the delayed_work. This replaces the static global
> > cvm_oct_rx_refill_work. The struct is allocated with devm_kzalloc()
> > in probe() and stored via platform_set_drvdata(). The worker
> > retrieves it using container_of().
> >
> > Add a struct device pointer to oct_rx_group and thread it through
> > cvm_oct_rx_initialize() to support the NAPI poll refill path.
>
> Why isn't this the platform device?  And why isn't this data just part
> of the platform device's data anyway?  Why is it a separate "thing"?
>
> thanks,
>
> greg k-h

Hi Greg,

Thank you for the review.

The changes you're pointing out were in the later patches of the v5 series.
My mistake was splitting them across patches rather than doing it
correctly from the start, which left an inconsistent intermediate
state.

I've reworked the series so that struct platform_device is passed
from the first patch, and oct_rx_group is placed inside
octeon_ethernet_platform in the same patch that introduces it.

Will send v6 shortly.

Thanks,
Ayush