
IBM3206K0424
Preliminary
IBM Processor for Network Resources
pnr25.chapt05.01
August 14, 2000
Cell/Packet Re-assembly (REASM)
Page 327 of 676
enabled in the cut through configuration. This buffer is referred to as the optional header page and uses its 
own page size and receive queue to allow for better storage utilization. When enabled, the second entry in 
the DMA list becomes the optional header page entry, and should be treated accordingly for page recovery. 
In split header mode, the header bytes are DMAed as soon as they are all received. This mode is useful if the 
user environment requires the header bytes to be in a separate buffer from the packet header and DMA list. 
Another optional feature is to DMA the header only. This feature is enabled in the cut through configuration. 
When enabled, only the packet header, DMA list, and header bytes are DMAed. Either a single DMA or two 
DMAs occur based on if the optional header feature is enabled. This feature can be useful when a routing 
decision needs to be made for a packet and the entire packet does not need to be brought into host storage. 
Another possible scenario is the header bytes may determine how the user wants to DMA the packet data to 
the host. 
For smaller packet sizes, it may be more efficient to perform a single DMA and keep the packet header, DMA 
list, and packet data in a single buffer. To do this, single page mode should be enabled in the cut through 
configuration. When enabled, the header page size is used to determine the DMA behavior. If a packet 
completes and the total length of the packet and headers will fit in the header page size, then a single DMA is 
performed. If the data length exceeds the single page size as it is being received, the data is scattered using 
the normal options (might need to catch up). This feature can be useful for optimizing user processing for 
smaller packets. For example, all packets less than 2K might be a candidate for this feature. 
Head Bytes
There are two ways to set the number of head bytes. They are set on a per connection basis in the receive 
LCD using the numHeadbytes and useCrcNumHead fields. When the useCrcNumHead field is set to 
’
0
’
, the 
numHeadbytes field provides a fixed number of bytes that is used as the number of header bytes. When 
useCrcNumHead field is set to 
’
1
’
, then the RXCRC nano code will calculate the number of header bytes 
using the frameType field to index an IP procedure. Currently, RXCRC will only set the number of header 
bytes for recognized TCP/IP headers. Other headers can be recognized. Contact IBM technical support to 
discuss requirements.