[PATCH v3 04/25] dma/idxd: clear device at scan

[David Marchand]

On Tue, 26 May 2026 at 11:42, Bruce Richardson
<bruce.richardson at intel.com> wrote:
>
> On Tue, May 26, 2026 at 10:41:44AM +0200, David Marchand wrote:
> > Currently, the device object is only manipulated by the dma/idxd bus
> > callbacks and EAL is not looking too much into this object.
> >
> > However, in the next refactoring, EAL will expect a clean object, like
> > when checking that the device has been already probed
> > (iow dev->driver != NULL).
> >
> > Request a 0'd object when allocating.
> >
> > Reported-by: Bruce Richardson <bruce.richardson at intel.com>
> > Signed-off-by: David Marchand <david.marchand at redhat.com>
> > ---
> >  drivers/dma/idxd/idxd_bus.c | 2 +-
> >  1 file changed, 1 insertion(+), 1 deletion(-)
> >
>
> Out of interest, did you (or Claude :-) ) run a scan over the other bus
> drivers to see if any others are similarly not properly zeroed on init?

I did check, the "legacy" way.

All other are calling calloc(), or rte_zmalloc_*, or manually clearing.
I hesitated at changing the malloc+memset pattern to calloc, but I
left it for later.
The series is big enough.


-- 
David Marchand