In message <[EMAIL PROTECTED]>,David Miller writes: >From: Christoph Hellwig <[EMAIL PROTECTED]> >Date: Thu, 13 Jul 2006 22:36:16 +0200 >> - drivers/atm/{idt77252.c, nicstar.c}, drivers/usb/atm/usbatm.c: >> These use private skb queues and do odd things. I can't see >> any point for using dev_alloc_skb with it's additional headroom >> reservation here.
as i recall, for simple rfc1483 encapsulation, the LLC/SNAP header of 8 bytes needs to be replaced by an ether header of 14 bytes. its likely this is the reason dev_alloc_skb() is used. most of the other drivers use atm_alloc_charge() which does a skb_reserve(), but these drivers dont dma directly into a skb. >Any of these cases that DMA into the skb allocated will need >the headroom, as explained by the PowerPC folks. this only works for "ordinary" ip traffic (which i admit is the most common case). but since the nicstar/idt77252 devices have a peak speed of 155Mb/s i doubt they would suffer much. considering they still use virt_to_bus/bus_to_virt i would be a bit surprised if they worked (atleast on ppc64). they can be simply converted to alloc_skb() and manually skb_reserve() a little overhead. usbatm should probably be converted to use atm_alloc_charge() since the vcc is known at this point unlike the nicstar/idt77252 case. - To unsubscribe from this list: send the line "unsubscribe netdev" in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html