
IBM32NPCXX1EPABBE66
Preliminary
IBM Processor for Network Resources
pnr261_1intro.fm.06
August 14, 2000
Functional Description
Page 11 of 706
Transmit opportunities are repeatedly provided to SEGBF by CSKED at the desired rate until all the data in
the buffer has been passed to LINKC via the cell buffer array. When SEGBF detects that no more data exists
for a buffer, an indication is passed to CSKED. If more buffers are queued, the LCD is updated and the seg-
mentation process continues until all buffers on the LCD queue are serviced. A bit in the buffer header gener-
ates a transmit complete event when all of the data in the buffer has been transmitted.
In POS mode, CSKED will only provide one transmit request per buffer to SEGBF. SEGBF will segment the
entire buffer before accepting another request from CSKED. Up to four physical drops can be configured in
LINKC. A field in the LCD is used to assign a connection to a physical drop.
1.9.2 Receive Path
As cells arrive, they pass from LINKC to REASM. REASM uses a portion of the ATM packet header to look up
the LCD address for this cell. The LCD address is then passed to RXLCD. RXLCD reads the receive portion
of the LCD, and then REASM processes the cell based on the LCD information. For example, the LCD spec-
ifies what AAL to use and maintains the current reassembly state. Using the current reassembly state, the cell
data is written to Packet Memory. While the data is written to Packet Memory, other functions such as CRC
and TCP/IP checksum verification are performed in parallel. If a packet is complete, all trailer verification is
performed. If the packet is good, an event is placed on a receive queue in the RXQUE entity. For error sce-
narios, see
3.15 Receive Queues (RXQUE)
on page 351. At this point, software can dequeue the packet
event from RXQUE using the dequeue operation. It can then examine headers, DMA the data into user
space, and perform TCP checksums. When these actions are complete, the buffer is returned to the PNR by
performing a POOLS free buffer operation.
Figure 4: Transmit Scheduling Capabilities
VCI desc.
Frame
Frame
Frame
Head
Pointer
Tail
Pointer
Frame
Frame
Frame
Frame
VCI desc.
VCI desc.
Pointer
FIFO
Byte
Count
Threshold
Extra Pin
External
Control
Fixed
Rate
VCI
Start variable
length frames
as fixed intervals
VPI/VCI
VPI
3 Level
Priority
Scheduling
Wheels
Cell based Transmission
Frame Based Transmission
ADD
Block