Re: [PATCH net-next v19 1/1] mctp pcc: Implement MCTP over PCC Transport

[Adam Young <admiyo@amperemail.onmicrosoft.com>]

On 4/22/25 10:15, Jeremy Kerr wrote:
> > +       spin_lock_irqsave(&mpnd->lock, flags);
> > +       rc = skb_cow_head(skb, sizeof(struct mctp_pcc_hdr));
> > +       if (rc)
> > +               return rc;
> This will return with mpdn->lock still held.
> And should this return the raw rc value? Or NETDEV_TX_OK?

I based my code off the other drivers in the same directory:

         rc = skb_cow_head(skb, sizeof(struct mctp_i2c_hdr));
         if (rc)
                 return rc;


         rc = skb_cow_head(skb, sizeof(struct mctp_i3c_internal_hdr));
         if (rc)
                 return rc;

However, I just noticed the USB one, that went in last release, does 
this (using a goto)

err_drop:
         dev_dstats_tx_dropped(dev);
         kfree_skb(skb);
         return NETDEV_TX_OK;

Seems strange to returning NETDEV_TX_OK for what is essentially a memory 
allocation failure?  However, I am going to assume that this most recent 
one represents the current wisdom, and will follow suit.