Re: [PATCH] net: airoha: Fix an error handling path in airoha_probe()

[Christophe JAILLET]

Le 18/04/2025 à 14:28, Lorenzo Bianconi a écrit :
> > If an error occurs after a successful airoha_hw_init() call,
> > airoha_ppe_deinit() needs to be called as already done in the remove
> > function.
> >
> > Fixes: 00a7678310fe ("net: airoha: Introduce flowtable offload support")
> > Signed-off-by: Christophe JAILLET <christophe.jaillet@xxxxxxxxxx>
> > ---
> > Compile tested-only
> > ---
> >   drivers/net/ethernet/airoha/airoha_eth.c | 2 ++
> >   1 file changed, 2 insertions(+)
> >
> > diff --git a/drivers/net/ethernet/airoha/airoha_eth.c b/drivers/net/ethernet/airoha/airoha_eth.c
> > index 69e523dd4186..252b32ceb064 100644
> > --- a/drivers/net/ethernet/airoha/airoha_eth.c
> > +++ b/drivers/net/ethernet/airoha/airoha_eth.c
> > @@ -2631,6 +2631,8 @@ static int airoha_probe(struct platform_device *pdev)
> >   		}
> >   	}
> >   	free_netdev(eth->napi_dev);
> > +
> > +	airoha_ppe_deinit(eth);
> >   	platform_set_drvdata(pdev, NULL);
> >   
> >   	return err;
> > --
> > 2.49.0
>
> Hi Christophe,
>
> I agree we are missing a airoha_ppe_deinit() call in the probe error path,
> but we should move it above after stopping the NAPI since if airoha_hw_init()
> fails we will undo the work done by airoha_ppe_init(). Something like:

Agreed.

I'll send a v2 with as a small series, because of another leak I found 
while looking at it.

And while at it, I'll propose a few clean-ups.

CJ


> diff --git a/drivers/net/ethernet/airoha/airoha_eth.c b/drivers/net/ethernet/airoha/airoha_eth.c
> index 16c7896f931f..37d9678798d1 100644
> --- a/drivers/net/ethernet/airoha/airoha_eth.c
> +++ b/drivers/net/ethernet/airoha/airoha_eth.c
> @@ -2959,6 +2959,7 @@ static int airoha_probe(struct platform_device *pdev)
>   error_napi_stop:
>   	for (i = 0; i < ARRAY_SIZE(eth->qdma); i++)
>   		airoha_qdma_stop_napi(&eth->qdma[i]);
> +	airoha_ppe_init(eth);
>   error_hw_cleanup:
>   	for (i = 0; i < ARRAY_SIZE(eth->qdma); i++)
>   		airoha_hw_cleanup(&eth->qdma[i]);
>
>
> Agree?
>
> Regards,
> Lorenzo