Re: [PATCH v3 1/9] driver core: Don't let a device probe until it's ready

[Doug Anderson]

Hi,

On Fri, Apr 3, 2026 at 8:44 AM Greg Kroah-Hartman
<gregkh@xxxxxxxxxxxxxxxxxxx> wrote:
>
> > If you wish, I can turn this into something like:
> >
> > #define DEV_FLAG_READY_TO_PROBE         0
> > #define DEV_FLAG_CAN_MATCH              1
> > #define DEV_FLAG_DMA_IOMMU              2
> > ...
> >
> > ...but that seemed worse (to me) than the enum. Please shout if I
> > misunderstood or you disagree.
>
> If you make it a numbered enum, that's fine (I hate unnumbered ones),

Sounds like a plan.


> and a comment somewhere saying we will "max out" at 63, or have a
> DEV_FLAG_MAX_BIT entry in there.

Sure. To be compatible across all architectures, it should probably
max at 31, right? Atomic bitops works with an "unsigned long" which
might be only 32-bits. Oh, actually I should just add "DEV_FLAG_COUNT"
at the end of the enum and declare flags with DECLARE_BITMAP(). Then
there is no max.

In total, I've now got this:

enum struct_device_flags {
        DEV_FLAG_READY_TO_PROBE = 0,
        DEV_FLAG_CAN_MATCH = 1,
        DEV_FLAG_DMA_IOMMU = 2,
        DEV_FLAG_DMA_SKIP_SYNC = 3,
        DEV_FLAG_DMA_OPS_BYPASS = 4,
        DEV_FLAG_STATE_SYNCED = 5,
        DEV_FLAG_DMA_COHERENT = 6,
        DEV_FLAG_OF_NODE_REUSED = 7,
        DEV_FLAG_OFFLINE_DISABLED = 8,
        DEV_FLAG_OFFLINE = 9,

        DEV_FLAG_COUNT
};

...and "flags" is now:

DECLARE_BITMAP(flags, DEV_FLAG_COUNT);

Yell if you want anything changed. I'll restart my allmodconfig
compile tests now.

-Doug