xref: /freebsd/sys/dev/cxgbe/firmware/t6fw_cfg_fpga.txt (revision 59c8e88e72633afbc47a4ace0d2170d00d51f7dc)
1# Chelsio T6 Factory Default configuration file.
2#
3# Copyright (C) 2014-2015 Chelsio Communications.  All rights reserved.
4#
5#   DO NOT MODIFY THIS FILE UNDER ANY CIRCUMSTANCES.  MODIFICATION OF THIS FILE
6#   WILL RESULT IN A NON-FUNCTIONAL ADAPTER AND MAY RESULT IN PHYSICAL DAMAGE
7#   TO ADAPTERS.
8
9
10# This file provides the default, power-on configuration for 2-port T6-based
11# adapters shipped from the factory.  These defaults are designed to address
12# the needs of the vast majority of Terminator customers.  The basic idea is to
13# have a default configuration which allows a customer to plug a Terminator
14# adapter in and have it work regardless of OS, driver or application except in
15# the most unusual and/or demanding customer applications.
16#
17# Many of the Terminator resources which are described by this configuration
18# are finite.  This requires balancing the configuration/operation needs of
19# device drivers across OSes and a large number of customer application.
20#
21# Some of the more important resources to allocate and their constaints are:
22#  1. Virtual Interfaces: 256.
23#  2. Ingress Queues with Free Lists: 1024.
24#  3. Egress Queues: 128K.
25#  4. MSI-X Vectors: 1088.
26#  5. Multi-Port Support (MPS) TCAM: 336 entries to support MAC destination
27#     address matching on Ingress Packets.
28#
29# Some of the important OS/Driver resource needs are:
30#  6. Some OS Drivers will manage all resources through a single Physical
31#     Function (currently PF4 but it could be any Physical Function).
32#  7. Some OS Drivers will manage different ports and functions (NIC,
33#     storage, etc.) on different Physical Functions.  For example, NIC
34#     functions for ports 0-1 on PF0-1, FCoE on PF4, iSCSI on PF5, etc.
35#
36# Some of the customer application needs which need to be accommodated:
37#  8. Some customers will want to support large CPU count systems with
38#     good scaling.  Thus, we'll need to accommodate a number of
39#     Ingress Queues and MSI-X Vectors to allow up to some number of CPUs
40#     to be involved per port and per application function.  For example,
41#     in the case where all ports and application functions will be
42#     managed via a single Unified PF and we want to accommodate scaling up
43#     to 8 CPUs, we would want:
44#
45#         2 ports *
46#         3 application functions (NIC, FCoE, iSCSI) per port *
47#         16 Ingress Queue/MSI-X Vectors per application function
48#
49#     for a total of 96 Ingress Queues and MSI-X Vectors on the Unified PF.
50#     (Plus a few for Firmware Event Queues, etc.)
51#
52#  9. Some customers will want to use PCI-E SR-IOV Capability to allow Virtual
53#     Machines to directly access T6 functionality via SR-IOV Virtual Functions
54#     and "PCI Device Passthrough" -- this is especially true for the NIC
55#     application functionality.
56#
57
58
59# Global configuration settings.
60#
61[global]
62	rss_glb_config_mode = basicvirtual
63	rss_glb_config_options = tnlmapen,hashtoeplitz,tnlalllkp
64
65	# PL_TIMEOUT register
66	pl_timeout_value = 1000		# the timeout value in units of us
67
68	# The following Scatter Gather Engine (SGE) settings assume a 4KB Host
69	# Page Size and a 64B L1 Cache Line Size. It programs the
70	# EgrStatusPageSize and IngPadBoundary to 64B and the PktShift to 2.
71	# If a Master PF Driver finds itself on a machine with different
72	# parameters, then the Master PF Driver is responsible for initializing
73	# these parameters to appropriate values.
74	#
75	# Notes:
76	#  1. The Free List Buffer Sizes below are raw and the firmware will
77	#     round them up to the Ingress Padding Boundary.
78	#  2. The SGE Timer Values below are expressed below in microseconds.
79	#     The firmware will convert these values to Core Clock Ticks when
80	#     it processes the configuration parameters.
81	#
82	reg[0x1008] = 0x40810/0x21c70	# SGE_CONTROL
83	reg[0x100c] = 0x22222222	# SGE_HOST_PAGE_SIZE
84	reg[0x10a0] = 0x01040810	# SGE_INGRESS_RX_THRESHOLD
85	reg[0x1044] = 4096		# SGE_FL_BUFFER_SIZE0
86	reg[0x1048] = 65536		# SGE_FL_BUFFER_SIZE1
87	reg[0x104c] = 1536		# SGE_FL_BUFFER_SIZE2
88	reg[0x1050] = 9024		# SGE_FL_BUFFER_SIZE3
89	reg[0x1054] = 9216		# SGE_FL_BUFFER_SIZE4
90	reg[0x1058] = 2048		# SGE_FL_BUFFER_SIZE5
91	reg[0x105c] = 128		# SGE_FL_BUFFER_SIZE6
92	reg[0x1060] = 8192		# SGE_FL_BUFFER_SIZE7
93	reg[0x1064] = 16384		# SGE_FL_BUFFER_SIZE8
94	reg[0x10a4] = 0xa000a000/0xf000f000 # SGE_DBFIFO_STATUS
95	reg[0x10a8] = 0x402000/0x402000	# SGE_DOORBELL_CONTROL
96	sge_timer_value = 5, 10, 20, 50, 100, 200 # SGE_TIMER_VALUE* in usecs
97	reg[0x10c4] = 0x20000000/0x20000000 # GK_CONTROL, enable 5th thread
98
99	#DBQ Timer duration = 1 cclk cycle duration * (sge_dbq_timertick+1) * sge_dbq_timer
100	#SGE DBQ tick value. All timers are multiple of this value
101#	sge_dbq_timertick = 5 #in usecs
102#	sge_dbq_timer = 1, 2, 4, 6, 8, 10, 12, 16
103	# enable TP_OUT_CONFIG.IPIDSPLITMODE
104	reg[0x7d04] = 0x00010000/0x00010000
105
106	reg[0x7dc0] = 0x0e2f8849	# TP_SHIFT_CNT
107
108	#Tick granularities in kbps
109	tsch_ticks = 1000, 100, 10, 1
110
111	# TP_VLAN_PRI_MAP to select filter tuples and enable ServerSram
112	# filter control: compact, fcoemask
113	# server sram   : srvrsram
114	# filter tuples : fragmentation, mpshittype, macmatch, ethertype,
115	#		  protocol, tos, vlan, vnic_id, port, fcoe
116	# valid filterModes are described the Terminator 5 Data Book
117	filterMode = fcoemask, srvrsram, fragmentation, mpshittype, protocol, vlan, port, fcoe
118
119	# filter tuples enforced in LE active region (equal to or subset of filterMode)
120	filterMask = protocol, fcoe
121
122	# Percentage of dynamic memory (in either the EDRAM or external MEM)
123	# to use for TP RX payload
124	tp_pmrx = 30
125
126	# TP RX payload page size
127	tp_pmrx_pagesize = 64K
128
129	# TP number of RX channels
130	tp_nrxch = 0		# 0 (auto) = 1
131
132	# Percentage of dynamic memory (in either the EDRAM or external MEM)
133	# to use for TP TX payload
134	tp_pmtx = 50
135
136	# TP TX payload page size
137	tp_pmtx_pagesize = 64K
138
139	# TP number of TX channels
140	tp_ntxch = 0		# 0 (auto) = equal number of ports
141
142	# TP OFLD MTUs
143	tp_mtus = 88, 256, 512, 576, 808, 1024, 1280, 1488, 1500, 2002, 2048, 4096, 4352, 8192, 9000, 9600
144
145	# enable TP_OUT_CONFIG.IPIDSPLITMODE and CRXPKTENC
146	reg[0x7d04] = 0x00010008/0x00010008
147
148	# TP_GLOBAL_CONFIG
149	reg[0x7d08] = 0x00000800/0x00000800 # set IssFromCplEnable
150
151	# TP_PC_CONFIG
152	reg[0x7d48] = 0x00000000/0x00000400 # clear EnableFLMError
153
154	# TP_PARA_REG0
155	reg[0x7d60] = 0x06000000/0x07000000 # set InitCWND to 6
156
157	# LE_DB_CONFIG
158	reg[0x19c04] = 0x00400000/0x00440000 # LE Server SRAM Enable,
159					     # LE IPv4 compression disabled
160	# LE_DB_HASH_CONFIG
161	reg[0x19c28] = 0x00800000/0x01f00000 # LE Hash bucket size 8,
162
163	# ULP_TX_CONFIG
164	reg[0x8dc0] = 0x00000104/0x00000104 # Enable ITT on PI err
165					    # Enable more error msg for ...
166					    # TPT error.
167
168	# ULP_RX_MISC_FEATURE_ENABLE
169	reg[0x1925c] = 0x01003400/0x01003400 # iscsi tag pi bit
170					     # Enable offset decrement after ...
171					     # PI extraction and before DDP
172					     # ulp insert pi source info in DIF
173					     # iscsi_eff_offset_en
174
175	#Enable iscsi completion moderation feature
176	#reg[0x1925c] = 0x000041c0/0x000031c0	# Enable offset decrement after
177						# PI extraction and before DDP.
178						# ulp insert pi source info in
179						# DIF.
180						# Enable iscsi hdr cmd mode.
181						# iscsi force cmd mode.
182						# Enable iscsi cmp mode.
183
184# Some "definitions" to make the rest of this a bit more readable.  We support
185# 4 ports, 3 functions (NIC, FCoE and iSCSI), scaling up to 8 "CPU Queue Sets"
186# per function per port ...
187#
188# NMSIX = 1088			# available MSI-X Vectors
189# NVI = 256			# available Virtual Interfaces
190# NMPSTCAM = 336		# MPS TCAM entries
191#
192# NPORTS = 2			# ports
193# NCPUS = 16			# CPUs we want to support scalably
194# NFUNCS = 3			# functions per port (NIC, FCoE, iSCSI)
195
196# Breakdown of Virtual Interface/Queue/Interrupt resources for the "Unified
197# PF" which many OS Drivers will use to manage most or all functions.
198#
199# Each Ingress Queue can use one MSI-X interrupt but some Ingress Queues can
200# use Forwarded Interrupt Ingress Queues.  For these latter, an Ingress Queue
201# would be created and the Queue ID of a Forwarded Interrupt Ingress Queue
202# will be specified as the "Ingress Queue Asynchronous Destination Index."
203# Thus, the number of MSI-X Vectors assigned to the Unified PF will be less
204# than or equal to the number of Ingress Queues ...
205#
206# NVI_NIC = 4			# NIC access to NPORTS
207# NFLIQ_NIC = 32		# NIC Ingress Queues with Free Lists
208# NETHCTRL_NIC = 32		# NIC Ethernet Control/TX Queues
209# NEQ_NIC = 64			# NIC Egress Queues (FL, ETHCTRL/TX)
210# NMPSTCAM_NIC = 16		# NIC MPS TCAM Entries (NPORTS*4)
211# NMSIX_NIC = 32		# NIC MSI-X Interrupt Vectors (FLIQ)
212#
213# NVI_OFLD = 0			# Offload uses NIC function to access ports
214# NFLIQ_OFLD = 16		# Offload Ingress Queues with Free Lists
215# NETHCTRL_OFLD = 0		# Offload Ethernet Control/TX Queues
216# NEQ_OFLD = 16			# Offload Egress Queues (FL)
217# NMPSTCAM_OFLD = 0		# Offload MPS TCAM Entries (uses NIC's)
218# NMSIX_OFLD = 16		# Offload MSI-X Interrupt Vectors (FLIQ)
219#
220# NVI_RDMA = 0			# RDMA uses NIC function to access ports
221# NFLIQ_RDMA = 4		# RDMA Ingress Queues with Free Lists
222# NETHCTRL_RDMA = 0		# RDMA Ethernet Control/TX Queues
223# NEQ_RDMA = 4			# RDMA Egress Queues (FL)
224# NMPSTCAM_RDMA = 0		# RDMA MPS TCAM Entries (uses NIC's)
225# NMSIX_RDMA = 4		# RDMA MSI-X Interrupt Vectors (FLIQ)
226#
227# NEQ_WD = 128			# Wire Direct TX Queues and FLs
228# NETHCTRL_WD = 64		# Wire Direct TX Queues
229# NFLIQ_WD = 64	`		# Wire Direct Ingress Queues with Free Lists
230#
231# NVI_ISCSI = 4			# ISCSI access to NPORTS
232# NFLIQ_ISCSI = 4		# ISCSI Ingress Queues with Free Lists
233# NETHCTRL_ISCSI = 0		# ISCSI Ethernet Control/TX Queues
234# NEQ_ISCSI = 4			# ISCSI Egress Queues (FL)
235# NMPSTCAM_ISCSI = 4		# ISCSI MPS TCAM Entries (NPORTS)
236# NMSIX_ISCSI = 4		# ISCSI MSI-X Interrupt Vectors (FLIQ)
237#
238# NVI_FCOE = 4			# FCOE access to NPORTS
239# NFLIQ_FCOE = 34		# FCOE Ingress Queues with Free Lists
240# NETHCTRL_FCOE = 32		# FCOE Ethernet Control/TX Queues
241# NEQ_FCOE = 66			# FCOE Egress Queues (FL)
242# NMPSTCAM_FCOE = 32 		# FCOE MPS TCAM Entries (NPORTS)
243# NMSIX_FCOE = 34		# FCOE MSI-X Interrupt Vectors (FLIQ)
244
245# Two extra Ingress Queues per function for Firmware Events and Forwarded
246# Interrupts, and two extra interrupts per function for Firmware Events (or a
247# Forwarded Interrupt Queue) and General Interrupts per function.
248#
249# NFLIQ_EXTRA = 6		# "extra" Ingress Queues 2*NFUNCS (Firmware and
250# 				#   Forwarded Interrupts
251# NMSIX_EXTRA = 6		# extra interrupts 2*NFUNCS (Firmware and
252# 				#   General Interrupts
253
254# Microsoft HyperV resources.  The HyperV Virtual Ingress Queues will have
255# their interrupts forwarded to another set of Forwarded Interrupt Queues.
256#
257# NVI_HYPERV = 16		# VMs we want to support
258# NVIIQ_HYPERV = 2		# Virtual Ingress Queues with Free Lists per VM
259# NFLIQ_HYPERV = 40		# VIQs + NCPUS Forwarded Interrupt Queues
260# NEQ_HYPERV = 32		# VIQs Free Lists
261# NMPSTCAM_HYPERV = 16		# MPS TCAM Entries (NVI_HYPERV)
262# NMSIX_HYPERV = 8		# NCPUS Forwarded Interrupt Queues
263
264# Adding all of the above Unified PF resource needs together: (NIC + OFLD +
265# RDMA + ISCSI + FCOE + EXTRA + HYPERV)
266#
267# NVI_UNIFIED = 28
268# NFLIQ_UNIFIED = 106
269# NETHCTRL_UNIFIED = 32
270# NEQ_UNIFIED = 124
271# NMPSTCAM_UNIFIED = 40
272#
273# The sum of all the MSI-X resources above is 74 MSI-X Vectors but we'll round
274# that up to 128 to make sure the Unified PF doesn't run out of resources.
275#
276# NMSIX_UNIFIED = 128
277#
278# The Storage PFs could need up to NPORTS*NCPUS + NMSIX_EXTRA MSI-X Vectors
279# which is 34 but they're probably safe with 32.
280#
281# NMSIX_STORAGE = 32
282
283# Note: The UnifiedPF is PF4 which doesn't have any Virtual Functions
284# associated with it.  Thus, the MSI-X Vector allocations we give to the
285# UnifiedPF aren't inherited by any Virtual Functions.  As a result we can
286# provision many more Virtual Functions than we can if the UnifiedPF were
287# one of PF0-1.
288#
289
290# All of the below PCI-E parameters are actually stored in various *_init.txt
291# files.  We include them below essentially as comments.
292#
293# For PF0-1 we assign 8 vectors each for NIC Ingress Queues of the associated
294# ports 0-1.
295#
296# For PF4, the Unified PF, we give it an MSI-X Table Size as outlined above.
297#
298# For PF5-6 we assign enough MSI-X Vectors to support FCoE and iSCSI
299# storage applications across all four possible ports.
300#
301# Additionally, since the UnifiedPF isn't one of the per-port Physical
302# Functions, we give the UnifiedPF and the PF0-1 Physical Functions
303# different PCI Device IDs which will allow Unified and Per-Port Drivers
304# to directly select the type of Physical Function to which they wish to be
305# attached.
306#
307# Note that the actual values used for the PCI-E Intelectual Property will be
308# 1 less than those below since that's the way it "counts" things.  For
309# readability, we use the number we actually mean ...
310#
311# PF0_INT = 8			# NCPUS
312# PF1_INT = 8			# NCPUS
313# PF0_3_INT = 32		# PF0_INT + PF1_INT + PF2_INT + PF3_INT
314#
315# PF4_INT = 128			# NMSIX_UNIFIED
316# PF5_INT = 32			# NMSIX_STORAGE
317# PF6_INT = 32			# NMSIX_STORAGE
318# PF7_INT = 0			# Nothing Assigned
319# PF4_7_INT = 192		# PF4_INT + PF5_INT + PF6_INT + PF7_INT
320#
321# PF0_7_INT = 224		# PF0_3_INT + PF4_7_INT
322#
323# With the above we can get 17 VFs/PF0-3 (limited by 336 MPS TCAM entries)
324# but we'll lower that to 16 to make our total 64 and a nice power of 2 ...
325#
326# NVF = 16
327
328
329# For those OSes which manage different ports on different PFs, we need
330# only enough resources to support a single port's NIC application functions
331# on PF0-3.  The below assumes that we're only doing NIC with NCPUS "Queue
332# Sets" for ports 0-3.  The FCoE and iSCSI functions for such OSes will be
333# managed on the "storage PFs" (see below).
334#
335
336# Some OS Drivers manage all application functions for all ports via PF4.
337# Thus we need to provide a large number of resources here.  For Egress
338# Queues we need to account for both TX Queues as well as Free List Queues
339# (because the host is responsible for producing Free List Buffers for the
340# hardware to consume).
341#
342[function "0"]
343	wx_caps = all		# write/execute permissions for all commands
344	r_caps = all		# read permissions for all commands
345	nvi = 28		# NVI_UNIFIED
346	niqflint = 170		# NFLIQ_UNIFIED + NLFIQ_WD
347	nethctrl = 96 		# NETHCTRL_UNIFIED + NETHCTRL_WD
348	neq = 252		# NEQ_UNIFIED + NEQ_WD
349	nexactf = 40		# NMPSTCAM_UNIFIED
350	nrawf = 2
351	cmask = all		# access to all channels
352	pmask = all		# access to all four ports ...
353	nethofld = 1024		# number of user mode ethernet flow contexts
354	ncrypto_lookaside = 32
355	nclip = 32		# number of clip region entries
356	nfilter = 48		# number of filter region entries
357	nserver = 48		# number of server region entries
358	nhash = 2048		# number of hash region entries
359	nhpfilter = 0		# number of high priority filter region entries
360	protocol = nic_vm, ofld, rddp, rdmac, iscsi_initiator_pdu, iscsi_target_pdu, iscsi_t10dif, tlskeys, crypto_lookaside
361	tp_l2t = 3072
362	tp_ddp = 2
363	tp_ddp_iscsi = 2
364	tp_tls_key = 3
365	tp_stag = 2
366	tp_pbl = 5
367	tp_rq = 7
368	tp_srq = 128
369
370# We have FCoE and iSCSI storage functions on PF5 and PF6 each of which may
371# need to have Virtual Interfaces on each of the four ports with up to NCPUS
372# "Queue Sets" each.
373#
374[function "1"]
375	wx_caps = all		# write/execute permissions for all commands
376	r_caps = all		# read permissions for all commands
377	nvi = 4			# NPORTS
378	niqflint = 34		# NPORTS*NCPUS + NMSIX_EXTRA
379	nethctrl = 32		# NPORTS*NCPUS
380	neq = 64		# NPORTS*NCPUS * 2 (FL, ETHCTRL/TX)
381	nexactf = 16		# (NPORTS *(no of snmc grp + 1 hw mac) + 1 anmc grp)) rounded to 16.
382	cmask = all		# access to all channels
383	pmask = all		# access to all four ports ...
384	nserver = 16
385	nhash = 2048
386	tp_l2t = 1020
387	protocol = iscsi_initiator_fofld
388	tp_ddp_iscsi = 2
389	iscsi_ntask = 2048
390	iscsi_nsess = 2048
391	iscsi_nconn_per_session = 1
392	iscsi_ninitiator_instance = 64
393
394
395# The following function, 1023, is not an actual PCIE function but is used to
396# configure and reserve firmware internal resources that come from the global
397# resource pool.
398#
399[function "1023"]
400	wx_caps = all		# write/execute permissions for all commands
401	r_caps = all		# read permissions for all commands
402	nvi = 4			# NVI_UNIFIED
403	cmask = all		# access to all channels
404	pmask = all		# access to all four ports ...
405	nexactf = 8		# NPORTS + DCBX +
406	nfilter = 16		# number of filter region entries
407	#nhpfilter = 0		# number of high priority filter region entries
408
409
410# For Virtual functions, we only allow NIC functionality and we only allow
411# access to one port (1 << PF).  Note that because of limitations in the
412# Scatter Gather Engine (SGE) hardware which checks writes to VF KDOORBELL
413# and GTS registers, the number of Ingress and Egress Queues must be a power
414# of 2.
415#
416[function "0/*"]		# NVF
417	wx_caps = 0x82		# DMAQ | VF
418	r_caps = 0x86		# DMAQ | VF | PORT
419	nvi = 1			# 1 port
420	niqflint = 4		# 2 "Queue Sets" + NXIQ
421	nethctrl = 2		# 2 "Queue Sets"
422	neq = 4			# 2 "Queue Sets" * 2
423	nexactf = 4
424	cmask = all		# access to all channels
425	pmask = 0x1		# access to only one port ...
426
427
428[function "1/*"]		# NVF
429	wx_caps = 0x82		# DMAQ | VF
430	r_caps = 0x86		# DMAQ | VF | PORT
431	nvi = 1			# 1 port
432	niqflint = 4		# 2 "Queue Sets" + NXIQ
433	nethctrl = 2		# 2 "Queue Sets"
434	neq = 4			# 2 "Queue Sets" * 2
435	nexactf = 4
436	cmask = all		# access to all channels
437	pmask = 0x2		# access to only one port ...
438
439
440# MPS features a 196608 bytes ingress buffer that is used for ingress buffering
441# for packets from the wire as well as the loopback path of the L2 switch. The
442# folling params control how the buffer memory is distributed and the L2 flow
443# control settings:
444#
445# bg_mem:	%-age of mem to use for port/buffer group
446# lpbk_mem:	%-age of port/bg mem to use for loopback
447# hwm:		high watermark; bytes available when starting to send pause
448#		frames (in units of 0.1 MTU)
449# lwm:		low watermark; bytes remaining when sending 'unpause' frame
450#		(in inuits of 0.1 MTU)
451# dwm:		minimum delta between high and low watermark (in units of 100
452#		Bytes)
453#
454[port "0"]
455	dcb = ppp, dcbx, b2b	# configure for DCB PPP and enable DCBX offload
456	hwm = 30
457	lwm = 15
458	dwm = 30
459	dcb_app_tlv[0] = 0x8906, ethertype, 3
460	dcb_app_tlv[1] = 0x8914, ethertype, 3
461	dcb_app_tlv[2] = 3260, socketnum, 5
462
463
464[port "1"]
465	dcb = ppp, dcbx, b2b
466	hwm = 30
467	lwm = 15
468	dwm = 30
469	dcb_app_tlv[0] = 0x8906, ethertype, 3
470	dcb_app_tlv[1] = 0x8914, ethertype, 3
471	dcb_app_tlv[2] = 3260, socketnum, 5
472
473
474[fini]
475	version = 0x1425001d
476	checksum = 0x5001af51
477
478# Total resources used by above allocations:
479#   Virtual Interfaces: 104
480#   Ingress Queues/w Free Lists and Interrupts: 526
481#   Egress Queues: 702
482#   MPS TCAM Entries: 336
483#   MSI-X Vectors: 736
484#   Virtual Functions: 64
485#
486#
487