Re: [PATCH net] r8169: Fix sleeping function called during get_stats64

[Corinna Vinschen <vinschen@redhat.com>]

[-- Attachment #1: Type: text/plain, Size: 1361 bytes --]

On Sep  9 22:23, Francois Romieu wrote:
> Corinna Vinschen <vinschen@redhat.com> :
> [...]
> > diff --git a/drivers/net/ethernet/realtek/r8169.c b/drivers/net/ethernet/realtek/r8169.c
> > index 24dcbe6..630811a 100644
> > --- a/drivers/net/ethernet/realtek/r8169.c
> > +++ b/drivers/net/ethernet/realtek/r8169.c
> [...]
> > +	if (!paddr)
> > +		return false;
> 
> I guess this is the secret recipe.

Actually, no.  I started out stress testing this combined with a printk
to show if paddr can be NULL, but this never occurs.  ndo_get_stats64 is
apparently only called after registering the device, and this occurs
after calling dma_alloc_coherent in rtl_init_one.

Rather than removing the entire test, I acidentally only removed the
printk.  Sorry about that.

> [...]
> > @@ -8447,9 +8411,14 @@ static int rtl_init_one(struct pci_dev *pdev, const struct pci_device_id *ent)
> >  
> >  	tp->rtl_fw = RTL_FIRMWARE_UNKNOWN;
> >  
> > +	tp->CntArray = dma_alloc_coherent (&pdev->dev, sizeof(*tp->CntArray),
> > +					   &tp->CntPhysAddr, GFP_KERNEL);
> > +	if (!tp->CntArray)
> > +		goto err_out_cnt_4;
> > +
> 
> rc is still zero here so rtl_init_one will return success.

Thanks for catching!  I'm sending a fixed patch in a minute, removing
the paddr test and setting rc to -ENOMEM if dma_alloc_coherent fails.


Corinna

[-- Attachment #2: Type: application/pgp-signature, Size: 819 bytes --]