patm -- Device driver for IDT77252 based ATM interfaces (ProSum and IDT)
device patm
device utopia
device atm
options NATM
options LIBMBPOOL
The patm device driver supports ATM cards based on the IDT77252 chip. It
has been tested with ProSum's ProATM-155 cards and with IDT's evaluation
boards. The driver interfaces with the natm(4) framework, netgraph(4)
and HARP. It provides only PVC services. Signalling, ATMARP, ILMI and
other higher layer protocols are implemented using netgraph(4) or HARP.
For configuring the card for IP see natmip(4).
The driver supports UBR, CBR, VBR and ABR traffic. Supported AALs are:
AAL0 (cell payloads), AAL5 and raw AAL. The driver supports opening of
VCI/VPI 0/0 in RX, raw AAL-mode. This VC will receive all incoming cells
(even those with non-zero GFC fields and VPI/VCI values outside the
allowed range) that are not claimed by other open connections. This may
be used for monitoring purposes.
The following sysctls are recognized by the driver additionally to those
handled by utopia(4):
hw.atm.patmN.istats
Returns a list of uint32_t statistic counters with internal driver
statistics.
hw.atm.patmN.eeprom
This is a read-only variable containing the contents of the on-board
EEPROM device.
hw.atm.patmN.lbuf_max
This puts an upper limit on the number of large receive buffers the
driver will allocate. This is a read-only variable that can be set
via a loader tunable.
hw.atm.patmN.tx_maxmaps
This is the upper limit of transmission DMA maps the driver will
allocate. This is read-only but may be set via a loader tunable.
hw.atm.patmN.debug
(only if debugging enabled) These are debugging flags. See
if_patmvar.h() for the possible flags. This may be initialized via
a loader tunable.
hw.atm.patmN.regs
(only if debugging enabled) Returns the current values of the card's
registers.
hw.atm.patmN.tsq
Returns the transmit status queue.
When loaded the driver initializes several variables from loader tunables:
hw.patmN.lbuf_max
This initializes the corresponding sysctl variable and defines an
upper limit on the number of large receive buffers (mbuf clusters).
hw.patmN.tx_maxmaps
This initializes the corresponding sysctl variable and is the maximum
number of DMA maps for transmission that the driver will allocated.
hw.patmN.debug
(only if debugging enabled) Initializes the debugging flags.
The driver supports the media options sdh, noscramb and unassigned (see
utopia(4)) when the card is a 155MBit card. Both PMC-Sierra S/UNI and
IDT77155 PHY chips are supported for these cards. For 25MBit cards the
IDT77105 is supported.
patm1: <NICStAR (77222/77252) ATM adapter> port 0xc000-0xc0ff mem 0xf8000000-0xf83fffff,0xf4000000-0xf4000fff irq 11 at device 8.0 on pci2
patm1: IDT77252 155MBit interface; idt77252 Rev. G; IDT77155 PHY
patm1: 128K x 32 SRAM; 4096 connections
natm(4), natmip(4), utopia(4)
The card fails to generate an interrupt if a cell is received in AAL0
mode that has the MSB of the PTI field cleared. Therefore cells will be
delivered on the next receive interrupt which can happen either when the
receive status queue is full, or a cell with the last bit of the PTI set
is received.
Although the card supports AAL3/4 the driver does not.
The rate tables used by this driver are not the tables recommended by IDT
(they are wrong anyway). The driver's tables are slightly more aggressive
than IDT's. That means, that the actual cell rate can be slightly
higher than the specified. This is in contrast to the IDT tables where
cell rates 5% less than the allowed one have been observed. This can be
changed by changing the program that generates these tables, found in
/usr/src/sys/dev/patm/genrtab, and regenerating them.
The reported media for the 155MBit cards will always be OC3/MM, because
there is no (known to me) way to figure out the actual medium. The
medium should really be coded in the EEPROM by the manufacturer.
The Tx cell counter in the utopia statistics is wrong, because the chip
uses idle cells for spacing and the PHY counts these cells. While there
is a configuration option for the chip to switch of these cells and,
according to the documentation, this should not influence cell spacing,
it does, so the driver let's the chip generate idle cells.
Thanks to Christian Bucari from ProSum for lending two of these cards to
enable the development of this driver. Thanks also for answering my
questions.
Harti Brandt <harti@freebsd.org>.
FreeBSD July 15, 2003 FreeBSD
[ Back ] |