요약·해설과 원문, 전문 번역을 서로 분리했습니다. API 이름, symbol, source path는 원문 표기를 사용합니다.
1. 요약·해설
원문의 핵심 논리와 kernel programming 관점의 보충 설명입니다. 아래의 전문 번역과는 별도로 작성했습니다.
Topology, protection과 BIOS status
sysfs-driver-mlxreg-io:147-249Static topology ID, PCIe reset retention, VPD write protection, voltage-regulator update와 BIOS image 상태를 다룹니다.
Chassis line-card와 power management
sysfs-driver-mlxreg-io:250-367최대 8개 line card의 reset release, power, ASIC reset mask와 hardware·software power-budget ownership을 제어합니다.
Line-card programmable device
sysfs-driver-mlxreg-io:368-469Line-card CPLD·FPGA version, VPD protection, reset cause, JTAG·SPI burning, QSFP power와 board configuration을 제공합니다.
PHY, ASIC와 External Root of Trust
sysfs-driver-mlxreg-io:471-614PHY·ASIC reset과 health, host communication readiness, EROT AP reset monitoring, recovery·write-protect·SPI selection을 다룹니다.
Power, clock과 field update
sysfs-driver-mlxreg-io:615-754ASIC power-good, clock board status, programming enable, CPLD5, JTAG update method와 BMC write-protection grant handshake를 제공합니다.
DPU board control
sysfs-driver-mlxreg-io:756-815DPU boot progress와 hardware ID, last-reset cause, PCIe·PHY·TPM·USB PHY reset 및 UFM upgrade status를 설명합니다.
2. 영어 원문 전체
번역 기준이 된 Linux v6.18.37 원문입니다. 줄 번호는 이 버전의 파일 좌표입니다.
원문 전체 펼치기
1
What: /sys/devices/platform/mlxplat/mlxreg-io/hwmon/hwmon*/asic_health
2
Date: June 2018
3
KernelVersion: 4.19
4
Contact: Vadim Pasternak <[email protected]>
5
Description: This file shows ASIC health status. The possible values are:
6
0 - health failed, 2 - health OK, 3 - ASIC in booting state.
8
The files are read only.
10
What: /sys/devices/platform/mlxplat/mlxreg-io/hwmon/hwmon*/cpld1_version
11
What: /sys/devices/platform/mlxplat/mlxreg-io/hwmon/hwmon*/cpld2_version
12
Date: June 2018
13
KernelVersion: 4.19
14
Contact: Vadim Pasternak <[email protected]>
15
Description: These files show with which CPLD versions have been burned
16
on carrier and switch boards.
18
The files are read only.
20
What: /sys/devices/platform/mlxplat/mlxreg-io/hwmon/hwmon*/fan_dir
21
Date: December 2018
22
KernelVersion: 5.0
23
Contact: Vadim Pasternak <[email protected]>
24
Description: This file shows the system fans direction:
25
forward direction - relevant bit is set 0;
26
reversed direction - relevant bit is set 1.
28
The files are read only.
30
What: /sys/devices/platform/mlxplat/mlxreg-io/hwmon/hwmon*/cpld3_version
31
Date: November 2018
32
KernelVersion: 5.0
33
Contact: Vadim Pasternak <[email protected]>
34
Description: These files show with which CPLD versions have been burned
35
on LED or Gearbox board.
37
The files are read only.
39
What: /sys/devices/platform/mlxplat/mlxreg-io/hwmon/hwmon*/jtag_enable
40
Date: November 2018
41
KernelVersion: 5.0
42
Contact: Vadim Pasternak <[email protected]>
43
Description: These files enable and disable the access to the JTAG domain.
44
By default access to the JTAG domain is disabled.
46
The file is read/write.
48
What: /sys/devices/platform/mlxplat/mlxreg-io/hwmon/hwmon*/select_iio
49
Date: June 2018
50
KernelVersion: 4.19
51
Contact: Vadim Pasternak <[email protected]>
52
Description: This file allows iio devices selection.
54
Attribute select_iio can be written with 0 or with 1. It
55
selects which one of iio devices can be accessed.
57
The file is read/write.
59
What: /sys/devices/platform/mlxplat/mlxreg-io/hwmon/hwmon*/psu1_on
60
/sys/devices/platform/mlxplat/mlxreg-io/hwmon/hwmon*/psu2_on
61
/sys/devices/platform/mlxplat/mlxreg-io/hwmon/hwmon*/pwr_cycle
62
/sys/devices/platform/mlxplat/mlxreg-io/hwmon/hwmon*/pwr_down
63
Date: June 2018
64
KernelVersion: 4.19
65
Contact: Vadim Pasternak <[email protected]>
66
Description: These files allow asserting system power cycling, switching
67
power supply units on and off and system's main power domain
68
shutdown.
69
Expected behavior:
70
When pwr_cycle is written 1: auxiliary power domain will go
71
down and after short period (about 1 second) up.
72
When psu1_on or psu2_on is written 1, related unit will be
73
disconnected from the power source, when written 0 - connected.
74
If both are written 1 - power supplies main power domain will
75
go down.
76
When pwr_down is written 1, system's main power domain will go
77
down.
79
The files are write only.
81
What: /sys/devices/platform/mlxplat/mlxreg-io/hwmon/hwmon*/reset_aux_pwr_or_ref
82
What: /sys/devices/platform/mlxplat/mlxreg-io/hwmon/hwmon*/reset_asic_thermal
83
What: /sys/devices/platform/mlxplat/mlxreg-io/hwmon/hwmon*/reset_hotswap_or_halt
84
What: /sys/devices/platform/mlxplat/mlxreg-io/hwmon/hwmon*/reset_hotswap_or_wd
85
What: /sys/devices/platform/mlxplat/mlxreg-io/hwmon/hwmon*/reset_fw_reset
86
What: /sys/devices/platform/mlxplat/mlxreg-io/hwmon/hwmon*/reset_long_pb
87
What: /sys/devices/platform/mlxplat/mlxreg-io/hwmon/hwmon*/reset_main_pwr_fail
88
What: /sys/devices/platform/mlxplat/mlxreg-io/hwmon/hwmon*/reset_short_pb
89
What: /sys/devices/platform/mlxplat/mlxreg-io/hwmon/hwmon*/reset_sw_reset
90
Date: June 2018
91
KernelVersion: 4.19
92
Contact: Vadim Pasternak <[email protected]>
93
Description: These files show the system reset cause, as following: power
94
auxiliary outage or power refresh, ASIC thermal shutdown, halt,
95
hotswap, watchdog, firmware reset, long press power button,
96
short press power button, software reset. Value 1 in file means
97
this is reset cause, 0 - otherwise. Only one of the above
98
causes could be 1 at the same time, representing only last
99
reset cause.
101
The files are read only.
103
What: /sys/devices/platform/mlxplat/mlxreg-io/hwmon/hwmon*/reset_comex_pwr_fail
104
What: /sys/devices/platform/mlxplat/mlxreg-io/hwmon/hwmon*/reset_from_comex
105
What: /sys/devices/platform/mlxplat/mlxreg-io/hwmon/hwmon*/reset_system
106
What: /sys/devices/platform/mlxplat/mlxreg-io/hwmon/hwmon*/reset_voltmon_upgrade_fail
107
Date: November 2018
108
KernelVersion: 5.0
109
Contact: Vadim Pasternak <[email protected]>
110
Description: These files show the system reset cause, as following: ComEx
111
power fail, reset from ComEx, system platform reset, reset
112
due to voltage monitor devices upgrade failure,
113
Value 1 in file means this is reset cause, 0 - otherwise.
114
Only one bit could be 1 at the same time, representing only
115
the last reset cause.
117
The files are read only.
119
What: /sys/devices/platform/mlxplat/mlxreg-io/hwmon/hwmon*/cpld4_version
120
Date: November 2018
121
KernelVersion: 5.0
122
Contact: Vadim Pasternak <[email protected]>
123
Description: These files show with which CPLD versions have been burned
124
on LED board.
126
The files are read only.
128
What: /sys/devices/platform/mlxplat/mlxreg-io/hwmon/hwmon*/reset_comex_thermal
129
What: /sys/devices/platform/mlxplat/mlxreg-io/hwmon/hwmon*/reset_comex_wd
130
What: /sys/devices/platform/mlxplat/mlxreg-io/hwmon/hwmon*/reset_from_asic
131
What: /sys/devices/platform/mlxplat/mlxreg-io/hwmon/hwmon*/reset_reload_bios
132
What: /sys/devices/platform/mlxplat/mlxreg-io/hwmon/hwmon*/reset_sff_wd
133
What: /sys/devices/platform/mlxplat/mlxreg-io/hwmon/hwmon*/reset_swb_wd
134
Date: June 2019
135
KernelVersion: 5.3
136
Contact: Vadim Pasternak <[email protected]>
137
Description: These files show the system reset cause, as following:
138
COMEX thermal shutdown; wathchdog power off or reset was derived
139
by one of the next components: COMEX, switch board or by Small Form
140
Factor mezzanine, reset requested from ASIC, reset caused by BIOS
141
reload. Value 1 in file means this is reset cause, 0 - otherwise.
142
Only one of the above causes could be 1 at the same time, representing
143
only last reset cause.
145
The files are read only.
147
What: /sys/devices/platform/mlxplat/mlxreg-io/hwmon/hwmon*/config1
148
What: /sys/devices/platform/mlxplat/mlxreg-io/hwmon/hwmon*/config2
149
Date: January 2020
150
KernelVersion: 5.6
151
Contact: Vadim Pasternak <[email protected]>
152
Description: These files show system static topology identification
153
like system's static I2C topology, number and type of FPGA
154
devices within the system and so on.
156
The files are read only.
158
What: /sys/devices/platform/mlxplat/mlxreg-io/hwmon/hwmon*/reset_ac_pwr_fail
159
What: /sys/devices/platform/mlxplat/mlxreg-io/hwmon/hwmon*/reset_platform
160
What: /sys/devices/platform/mlxplat/mlxreg-io/hwmon/hwmon*/reset_soc
161
What: /sys/devices/platform/mlxplat/mlxreg-io/hwmon/hwmon*/reset_sw_pwr_off
162
Date: January 2020
163
KernelVersion: 5.6
164
Contact: Vadim Pasternak <[email protected]>
165
Description: These files show the system reset causes, as following: reset
166
due to AC power failure, reset invoked from software by
167
assertion reset signal through CPLD. reset caused by signal
168
asserted by SOC through ACPI register, reset invoked from
169
software by assertion power off signal through CPLD.
171
The files are read only.
173
What: /sys/devices/platform/mlxplat/mlxreg-io/hwmon/hwmon*/pcie_asic_reset_dis
174
Date: January 2020
175
KernelVersion: 5.6
176
Contact: Vadim Pasternak <[email protected]>
177
Description: This file allows to retain ASIC up during PCIe root complex
178
reset, when attribute is set 1.
180
The file is read/write.
182
What: /sys/devices/platform/mlxplat/mlxreg-io/hwmon/hwmon*/vpd_wp
183
Date: January 2020
184
KernelVersion: 5.6
185
Contact: Vadim Pasternak <[email protected]>
186
Description: This file allows to overwrite system VPD hardware write
187
protection when attribute is set 1.
189
The file is read/write.
191
What: /sys/devices/platform/mlxplat/mlxreg-io/hwmon/hwmon*/voltreg_update_status
192
Date: January 2020
193
KernelVersion: 5.6
194
Contact: Vadim Pasternak <[email protected]>
195
Description: This file exposes the configuration update status of burnable
196
voltage regulator devices. The status values are as following:
197
0 - OK; 1 - CRC failure; 2 = I2C failure; 3 - in progress.
199
The file is read only.
201
What: /sys/devices/platform/mlxplat/mlxreg-io/hwmon/hwmon*/ufm_version
202
Date: January 2020
203
KernelVersion: 5.6
204
Contact: Vadim Pasternak <[email protected]>
205
Description: This file exposes the firmware version of burnable voltage
206
regulator devices.
208
The file is read only.
210
What: /sys/devices/platform/mlxplat/mlxreg-io/hwmon/hwmon*/cpld1_pn
211
What: /sys/devices/platform/mlxplat/mlxreg-io/hwmon/hwmon*/cpld2_pn
212
What: /sys/devices/platform/mlxplat/mlxreg-io/hwmon/hwmon*/cpld3_pn
213
What: /sys/devices/platform/mlxplat/mlxreg-io/hwmon/hwmon*/cpld4_pn
214
What: /sys/devices/platform/mlxplat/mlxreg-io/hwmon/hwmon*/cpld1_version_min
215
What: /sys/devices/platform/mlxplat/mlxreg-io/hwmon/hwmon*/cpld2_version_min
216
What: /sys/devices/platform/mlxplat/mlxreg-io/hwmon/hwmon*/cpld3_version_min
217
What: /sys/devices/platform/mlxplat/mlxreg-io/hwmon/hwmon*/cpld4_version_min
218
Date: July 2020
219
KernelVersion: 5.9
220
Contact: Vadim Pasternak <[email protected]>
221
Description: These files show with which CPLD part numbers and minor
222
versions have been burned CPLD devices equipped on a
223
system.
225
The files are read only.
227
What: /sys/devices/platform/mlxplat/mlxreg-io/hwmon/hwmon*/bios_active_image
228
What: /sys/devices/platform/mlxplat/mlxreg-io/hwmon/hwmon*/bios_auth_fail
229
What: /sys/devices/platform/mlxplat/mlxreg-io/hwmon/hwmon*/bios_upgrade_fail
230
Date: October 2021
231
KernelVersion: 5.16
232
Contact: Vadim Pasternak <[email protected]>
233
Description: The files represent BIOS statuses:
235
bios_active_image: location of current active BIOS image:
236
0: Top, 1: Bottom.
237
The reported value should correspond to value expected by OS
238
in case of BIOS safe mode is 0. This bit is related to Intel
239
top-swap feature of DualBios on the same flash.
241
bios_auth_fail: BIOS upgrade is failed because provided BIOS
242
image is not signed correctly.
244
bios_upgrade_fail: BIOS upgrade is failed by some other
245
reason not because authentication. For example due to
246
physical SPI flash problem.
248
The files are read only.
250
What: /sys/devices/platform/mlxplat/mlxreg-io/hwmon/hwmon*/lc1_enable
251
What: /sys/devices/platform/mlxplat/mlxreg-io/hwmon/hwmon*/lc2_enable
252
What: /sys/devices/platform/mlxplat/mlxreg-io/hwmon/hwmon*/lc3_enable
253
What: /sys/devices/platform/mlxplat/mlxreg-io/hwmon/hwmon*/lc4_enable
254
What: /sys/devices/platform/mlxplat/mlxreg-io/hwmon/hwmon*/lc5_enable
255
What: /sys/devices/platform/mlxplat/mlxreg-io/hwmon/hwmon*/lc6_enable
256
What: /sys/devices/platform/mlxplat/mlxreg-io/hwmon/hwmon*/lc7_enable
257
What: /sys/devices/platform/mlxplat/mlxreg-io/hwmon/hwmon*/lc8_enable
258
Date: October 2021
259
KernelVersion: 5.16
260
Contact: Vadim Pasternak <[email protected]>
261
Description: These files allow line cards enable state control.
262
Expected behavior:
263
When lc{n}_enable is written 1, related line card is released
264
from the reset state, when 0 - is hold in reset state.
266
The files are read/write.
268
What: /sys/devices/platform/mlxplat/mlxreg-io/hwmon/hwmon*/lc1_pwr
269
What: /sys/devices/platform/mlxplat/mlxreg-io/hwmon/hwmon*/lc2_pwr
270
What: /sys/devices/platform/mlxplat/mlxreg-io/hwmon/hwmon*/lc3_pwr
271
What: /sys/devices/platform/mlxplat/mlxreg-io/hwmon/hwmon*/lc4_pwr
272
What: /sys/devices/platform/mlxplat/mlxreg-io/hwmon/hwmon*/lc5_pwr
273
What: /sys/devices/platform/mlxplat/mlxreg-io/hwmon/hwmon*/lc6_pwr
274
What: /sys/devices/platform/mlxplat/mlxreg-io/hwmon/hwmon*/lc7_pwr
275
What: /sys/devices/platform/mlxplat/mlxreg-io/hwmon/hwmon*/lc8_pwr
276
Date: October 2021
277
KernelVersion: 5.16
278
Contact: Vadim Pasternak <[email protected]>
279
Description: These files switching line cards power on and off.
280
Expected behavior:
281
When lc{n}_pwr is written 1, related line card is powered
282
on, when written 0 - powered off.
284
The files are read/write.
286
What: /sys/devices/platform/mlxplat/mlxreg-io/hwmon/hwmon*/lc1_rst_mask
287
What: /sys/devices/platform/mlxplat/mlxreg-io/hwmon/hwmon*/lc2_rst_mask
288
What: /sys/devices/platform/mlxplat/mlxreg-io/hwmon/hwmon*/lc3_rst_mask
289
What: /sys/devices/platform/mlxplat/mlxreg-io/hwmon/hwmon*/lc4_rst_mask
290
What: /sys/devices/platform/mlxplat/mlxreg-io/hwmon/hwmon*/lc5_rst_mask
291
What: /sys/devices/platform/mlxplat/mlxreg-io/hwmon/hwmon*/lc6_rst_mask
292
What: /sys/devices/platform/mlxplat/mlxreg-io/hwmon/hwmon*/lc7_rst_mask
293
What: /sys/devices/platform/mlxplat/mlxreg-io/hwmon/hwmon*/lc8_rst_mask
294
Date: October 2021
295
KernelVersion: 5.16
296
Contact: Vadim Pasternak <[email protected]>
297
Description: These files clear line card reset bit enforced by ASIC, when it
298
sets it due to some abnormal ASIC behavior.
299
Expected behavior:
300
When lc{n}_rst_mask is written 1, related line card reset bit
301
is cleared, when written 0 - no effect.
303
The files are write only.
305
What: /sys/devices/platform/mlxplat/mlxreg-io/hwmon/hwmon*/os_started
306
Date: October 2021
307
KernelVersion: 5.16
308
Contact: Vadim Pasternak <[email protected]>
309
Description: This file, when written 1, indicates to programmable devices
310
that OS is taking control over it.
312
The file is read/write.
314
What: /sys/devices/platform/mlxplat/mlxreg-io/hwmon/hwmon*/pm_mgmt_en
315
Date: October 2021
316
KernelVersion: 5.16
317
Contact: Vadim Pasternak <[email protected]>
318
Description: This file assigns power management control ownership.
319
When power management control is provided by hardware, hardware
320
will automatically power off one or more line previously
321
powered line cards in case system power budget is getting
322
insufficient. It could be in case when some of power units lost
323
power good state.
324
When pm_mgmt_en is written 1, power management control by
325
software is enabled, 0 - power management control by hardware.
326
Note that for any setting of pm_mgmt_en attribute hardware will
327
not allow to power on any new line card in case system power
328
budget is insufficient.
329
Same in case software will try to power on several line cards
330
at once - hardware will power line cards while system has
331
enough power budget.
332
Default is 0.
334
The file is read/write.
336
What: /sys/devices/platform/mlxplat/mlxreg-io/hwmon/hwmon*/psu3_on
337
What: /sys/devices/platform/mlxplat/mlxreg-io/hwmon/hwmon*/psu4_on
338
Date: October 2021
339
KernelVersion: 5.16
340
Contact: Vadim Pasternak <[email protected]>
341
Description: These files switching power supply units on and off.
342
Expected behavior:
343
When psu3_on or psu4_on is written 1, related unit will be
344
disconnected from the power source, when written 0 - connected.
346
The files are write only.
348
What: /sys/devices/platform/mlxplat/mlxreg-io/hwmon/hwmon*/shutdown_unlock
349
Date: October 2021
350
KernelVersion: 5.16
351
Contact: Vadim Pasternak <[email protected]>
352
Description: This file allows to unlock ASIC after thermal shutdown event.
353
When system thermal shutdown is enforced by ASIC, ASIC is
354
getting locked and after system boot it will not be available.
355
Software can decide to unlock it by setting this attribute to
356
1 and then perform system power cycle by setting pwr_cycle
357
attribute to 1 (power cycle of main power domain).
358
Before setting shutdown_unlock to 1 it is recommended to
359
validate that system reboot cause is reset_asic_thermal or
360
reset_thermal_spc_or_pciesw.
361
In case shutdown_unlock is not set 1, the only way to release
362
ASIC from locking - is full system power cycle through the
363
external power distribution unit.
364
Default is 1.
366
The file is read/write.
368
What: /sys/devices/platform/mlxplat/i2c_mlxcpld.*/i2c-*/i2c-*/i2c-*/*-0032/mlxreg-io.*/hwmon/hwmon*/cpld1_pn
369
What: /sys/devices/platform/mlxplat/i2c_mlxcpld.*/i2c-*/i2c-*/i2c-*/*-0032/mlxreg-io.*/hwmon/hwmon*/cpld1_version
370
What: /sys/devices/platform/mlxplat/i2c_mlxcpld.*/i2c-*/i2c-*/i2c-*/*-0032/mlxreg-io.*/hwmon/hwmon*/cpld1_version_min
371
Date: October 2021
372
KernelVersion: 5.16
373
Contact: Vadim Pasternak <[email protected]>
374
Description: These files show with which CPLD major and minor versions
375
and part number has been burned CPLD device on line card.
377
The files are read only.
379
What: /sys/devices/platform/mlxplat/i2c_mlxcpld.*/i2c-*/i2c-*/i2c-*/*-0032/mlxreg-io.*/hwmon/hwmon*/fpga1_pn
380
What: /sys/devices/platform/mlxplat/i2c_mlxcpld.*/i2c-*/i2c-*/i2c-*/*-0032/mlxreg-io.*/hwmon/hwmon*/fpga1_version
381
What: /sys/devices/platform/mlxplat/i2c_mlxcpld.*/i2c-*/i2c-*/i2c-*/*-0032/mlxreg-io.*/hwmon/hwmon*/fpga1_version_min
382
Date: October 2021
383
KernelVersion: 5.16
384
Contact: Vadim Pasternak <[email protected]>
385
Description: These files show with which FPGA major and minor versions
386
and part number has been burned FPGA device on line card.
388
The files are read only.
390
What: /sys/devices/platform/mlxplat/i2c_mlxcpld.*/i2c-*/i2c-*/i2c-*/*-0032/mlxreg-io.*/hwmon/hwmon*/vpd_wp
391
Date: October 2021
392
KernelVersion: 5.16
393
Contact: Vadim Pasternak <[email protected]>
394
Description: This file allow to overwrite line card VPD hardware write
395
protection mode. When attribute is set 1 - write protection is
396
disabled, when 0 - enabled.
397
Default is 0.
398
If the system is in locked-down mode writing this file will not
399
be allowed.
400
The purpose if this file is to allow line card VPD burning
401
during production flow.
403
The file is read/write.
405
What: /sys/devices/platform/mlxplat/i2c_mlxcpld.*/i2c-*/i2c-*/i2c-*/*-0032/mlxreg-io.*/hwmon/hwmon*/reset_aux_pwr_or_ref
406
What: /sys/devices/platform/mlxplat/i2c_mlxcpld.*/i2c-*/i2c-*/i2c-*/*-0032/mlxreg-io.*/hwmon/hwmon*/reset_dc_dc_pwr_fail
407
What: /sys/devices/platform/mlxplat/i2c_mlxcpld.*/i2c-*/i2c-*/i2c-*/*-0032/mlxreg-io.*/hwmon/hwmon*/reset_fpga_not_done
408
What: /sys/devices/platform/mlxplat/i2c_mlxcpld.*/i2c-*/i2c-*/i2c-*/*-0032/mlxreg-io.*/hwmon/hwmon*/reset_from_chassis
409
What: /sys/devices/platform/mlxplat/i2c_mlxcpld.*/i2c-*/i2c-*/i2c-*/*-0032/mlxreg-io.*/hwmon/hwmon*/reset_line_card
410
What: /sys/devices/platform/mlxplat/i2c_mlxcpld.*/i2c-*/i2c-*/i2c-*/*-0032/mlxreg-io.*/hwmon/hwmon*/reset_pwr_off_from_chassis
411
Date: October 2021
412
KernelVersion: 5.16
413
Contact: Vadim Pasternak <[email protected]>
414
Description: These files show the line reset cause, as following: power
415
auxiliary outage or power refresh, DC-to-DC power failure, FPGA reset
416
failed, line card reset failed, power off from chassis.
417
Value 1 in file means this is reset cause, 0 - otherwise. Only one of
418
the above causes could be 1 at the same time, representing only last
419
reset cause.
421
The files are read only.
423
What: /sys/devices/platform/mlxplat/i2c_mlxcpld.*/i2c-*/i2c-*/i2c-*/*-0032/mlxreg-io.*/hwmon/hwmon*/cpld_upgrade_en
424
What: /sys/devices/platform/mlxplat/i2c_mlxcpld.*/i2c-*/i2c-*/i2c-*/*-0032/mlxreg-io.*/hwmon/hwmon*/fpga_upgrade_en
425
Date: October 2021
426
KernelVersion: 5.16
427
Contact: Vadim Pasternak <[email protected]>
428
Description: These files allow CPLD and FPGA burning. Value 1 in file means burning
429
is enabled, 0 - otherwise.
430
If the system is in locked-down mode writing these files will
431
not be allowed.
432
The purpose of these files to allow line card CPLD and FPGA
433
upgrade through the JTAG daisy-chain.
435
The files are read/write.
437
What: /sys/devices/platform/mlxplat/i2c_mlxcpld.*/i2c-*/i2c-*/i2c-*/*-0032/mlxreg-io.*/hwmon/hwmon*/qsfp_pwr_en
438
What: /sys/devices/platform/mlxplat/i2c_mlxcpld.*/i2c-*/i2c-*/i2c-*/*-0032/mlxreg-io.*/hwmon/hwmon*/pwr_en
439
Date: October 2021
440
KernelVersion: 5.16
441
Contact: Vadim Pasternak <[email protected]>
442
Description: These files allow to power on/off all QSFP ports and whole line card.
443
The attributes are set 1 for power on, 0 - for power off.
445
The files are read/write.
447
What: /sys/devices/platform/mlxplat/i2c_mlxcpld.*/i2c-*/i2c-*/i2c-*/*-0032/mlxreg-io.*/hwmon/hwmon*/agb_spi_burn_en
448
What: /sys/devices/platform/mlxplat/i2c_mlxcpld.*/i2c-*/i2c-*/i2c-*/*-0032/mlxreg-io.*/hwmon/hwmon*/fpga_spi_burn_en
449
Date: October 2021
450
KernelVersion: 5.16
451
Contact: Vadim Pasternak <[email protected]>
452
Description: These files allow gearboxes and FPGA SPI flash burning.
453
The attributes are set 1 to enable burning, 0 - to disable.
454
If the system is in locked-down mode writing these files will
455
not be allowed.
456
The purpose of these files to allow line card Gearboxes and FPGA
457
burning during production flow.
459
The file is read/write.
461
What: /sys/devices/platform/mlxplat/i2c_mlxcpld.*/i2c-*/i2c-*/i2c-*/*-0032/mlxreg-io.*/hwmon/hwmon*/max_power
462
What: /sys/devices/platform/mlxplat/i2c_mlxcpld.*/i2c-*/i2c-*/i2c-*/*-0032/mlxreg-io.*/hwmon/hwmon*/config
463
Date: October 2021
464
KernelVersion: 5.16
465
Contact: Vadim Pasternak <[email protected]>
466
Description: These files provide the maximum powered required for line card
467
feeding and line card configuration Id.
469
The files are read only.
471
What: /sys/devices/platform/mlxplat/mlxreg-io/hwmon/hwmon*/phy_reset
472
Date: May 2022
473
KernelVersion: 5.19
474
Contact: Vadim Pasternak <[email protected]>
475
Description: This file allows to reset PHY 88E1548 when attribute is set 0
476
due to some abnormal PHY behavior.
477
Expected behavior:
478
When phy_reset is written 1, all PHY 88E1548 are released
479
from the reset state, when 0 - are hold in reset state.
481
The files are read/write.
483
What: /sys/devices/platform/mlxplat/mlxreg-io/hwmon/hwmon*/mac_reset
484
Date: May 2022
485
KernelVersion: 5.19
486
Contact: Vadim Pasternak <[email protected]>
487
Description: This file allows to reset ASIC MT52132 when attribute is set 0
488
due to some abnormal ASIC behavior.
489
Expected behavior:
490
When mac_reset is written 1, the ASIC MT52132 is released
491
from the reset state, when 0 - is hold in reset state.
493
The files are read/write.
495
What: /sys/devices/platform/mlxplat/mlxreg-io/hwmon/hwmon*/qsfp_pwr_good
496
Date: May 2022
497
KernelVersion: 5.19
498
Contact: Vadim Pasternak <[email protected]>
499
Description: This file shows QSFP ports power status. The value is set to 0
500
when one of any QSFP ports is plugged. The value is set to 1 when
501
there are no any QSFP ports are plugged.
502
The possible values are:
503
0 - Power good, 1 - Not power good.
505
The files are read only.
507
What: /sys/devices/platform/mlxplat/mlxreg-io/hwmon/hwmon*/asic2_health
508
Date: July 2022
509
KernelVersion: 5.20
510
Contact: Vadim Pasternak <[email protected]>
511
Description: This file shows 2-nd ASIC health status. The possible values are:
512
0 - health failed, 2 - health OK, 3 - ASIC in booting state.
514
The file is read only.
516
What: /sys/devices/platform/mlxplat/mlxreg-io/hwmon/hwmon*/asic_reset
517
What: /sys/devices/platform/mlxplat/mlxreg-io/hwmon/hwmon*/asic2_reset
518
Date: July 2022
519
KernelVersion: 5.20
520
Contact: Vadim Pasternak <[email protected]>
521
Description: These files allow to each of ASICs by writing 1.
523
The files are write only.
525
What: /sys/devices/platform/mlxplat/mlxreg-io/hwmon/hwmon*/comm_chnl_ready
526
Date: July 2022
527
KernelVersion: 5.20
528
Contact: Vadim Pasternak <[email protected]>
529
Description: This file is used to indicate remote end (for example BMC) that system
530
host CPU is ready for sending telemetry data to remote end.
531
For indication the file should be written 1.
533
The file is write only.
535
What: /sys/devices/platform/mlxplat/mlxreg-io/hwmon/hwmon*/config3
536
Date: January 2020
537
KernelVersion: 5.6
538
Contact: Vadim Pasternak <[email protected]>
539
Description: The file indicates COME module hardware configuration.
540
The value is pushed by hardware through GPIO pins.
541
The purpose is to expose some minor BOM changes for the same system SKU.
543
The file is read only.
545
What: /sys/devices/platform/mlxplat/mlxreg-io/hwmon/hwmon*/reset_pwr_converter_fail
546
Date: February 2023
547
KernelVersion: 6.3
548
Contact: Vadim Pasternak <[email protected]>
549
Description: This file shows the system reset cause due to power converter
550
devices failure.
551
Value 1 in file means this is reset cause, 0 - otherwise.
553
The file is read only.
555
What: /sys/devices/platform/mlxplat/mlxreg-io/hwmon/hwmon*/erot1_ap_reset
556
What: /sys/devices/platform/mlxplat/mlxreg-io/hwmon/hwmon*/erot2_ap_reset
557
Date: February 2023
558
KernelVersion: 6.3
559
Contact: Vadim Pasternak <[email protected]>
560
Description: These files aim to monitor the status of the External Root of Trust (EROT)
561
processor's RESET output to the Application Processor (AP).
562
By reading this file, could be determined if the EROT has invalidated or
563
revoked AP Firmware, at which point it will hold the AP in RESET until a
564
valid firmware is loaded. This protects the AP from running an
565
unauthorized firmware. In the normal flow, the AP reset should be released
566
after the EROT validates the integrity of the FW, and it should be done so
567
as quickly as possible so that the AP boots before the CPU starts to
568
communicate to each ASIC.
570
The files are read only.
572
What: /sys/devices/platform/mlxplat/mlxreg-io/hwmon/hwmon*/erot1_recovery
573
What: /sys/devices/platform/mlxplat/mlxreg-io/hwmon/hwmon*/erot2_recovery
574
What: /sys/devices/platform/mlxplat/mlxreg-io/hwmon/hwmon*/erot1_reset
575
What: /sys/devices/platform/mlxplat/mlxreg-io/hwmon/hwmon*/erot2_reset
576
Date: February 2023
577
KernelVersion: 6.3
578
Contact: Vadim Pasternak <[email protected]>
579
Description: These files aim to perform External Root of Trust (EROT) recovery
580
sequence after EROT device failure.
581
These EROT devices protect ASICs from unauthorized access and in normal
582
flow their reset should be released with system power – earliest power
583
up stage, so that EROTs can begin boot and authentication process before
584
CPU starts to communicate to ASICs.
585
Issuing a reset to the EROT while asserting the recovery signal will cause
586
the EROT Application Processor to enter recovery mode so that the EROT FW
587
can be updated/recovered.
588
For reset/recovery the related file should be toggled by 1/0.
590
The files are read/write.
592
What: /sys/devices/platform/mlxplat/mlxreg-io/hwmon/hwmon*/erot1_wp
593
What: /sys/devices/platform/mlxplat/mlxreg-io/hwmon/hwmon*/erot2_wp
594
Date: February 2023
595
KernelVersion: 6.3
596
Contact: Vadim Pasternak <[email protected]>
597
Description: These files allow access to External Root of Trust (EROT) for reset
598
and recovery sequence after EROT device failure.
599
Default is 0 (programming disabled).
600
If the system is in locked-down mode writing this file will not be allowed.
602
The files are read/write.
604
What: /sys/devices/platform/mlxplat/mlxreg-io/hwmon/hwmon*/spi_chnl_select
605
Date: February 2023
606
KernelVersion: 6.3
607
Contact: Vadim Pasternak <[email protected]>
608
Description: This file allows SPI chip selection for External Root of Trust (EROT)
609
device Out-of-Band recovery.
610
File can be written with 0 or with 1. It selects which EROT can be accessed
611
through SPI device.
613
The file is read/write.
615
What: /sys/devices/platform/mlxplat/mlxreg-io/hwmon/hwmon*/asic_pg_fail
616
Date: February 2023
617
KernelVersion: 6.3
618
Contact: Vadim Pasternak [email protected]
619
Description: This file shows ASIC Power Good status.
620
Value 1 in file means ASIC Power Good failed, 0 - otherwise.
622
The file is read only.
624
What: /sys/devices/platform/mlxplat/mlxreg-io/hwmon/hwmon*/clk_brd1_boot_fail
625
What: /sys/devices/platform/mlxplat/mlxreg-io/hwmon/hwmon*/clk_brd2_boot_fail
626
What: /sys/devices/platform/mlxplat/mlxreg-io/hwmon/hwmon*/clk_brd_fail
627
Date: February 2023
628
KernelVersion: 6.3
629
Contact: Vadim Pasternak [email protected]
630
Description: These files are related to clock boards status in system.
631
- clk_brd1_boot_fail: warning about 1-st clock board failed to boot from CI.
632
- clk_brd2_boot_fail: warning about 2-nd clock board failed to boot from CI.
633
- clk_brd_fail: error about common clock board boot failure.
635
The files are read only.
637
What: /sys/devices/platform/mlxplat/mlxreg-io/hwmon/hwmon*/clk_brd_prog_en
638
Date: February 2023
639
KernelVersion: 6.3
640
Contact: Vadim Pasternak <[email protected]>
641
Description: This file enables programming of clock boards.
642
Default is 0 (programming disabled).
643
If the system is in locked-down mode writing this file will not be allowed.
645
The file is read/write.
647
What: /sys/devices/platform/mlxplat/mlxreg-io/hwmon/hwmon*/pwr_converter_prog_en
648
Date: February 2023
649
KernelVersion: 6.3
650
Contact: Vadim Pasternak <[email protected]>
651
Description: This file enables programming of power converters.
652
Default is 0 (programming disabled).
653
If the system is in locked-down mode writing this file will not be allowed.
655
The file is read/write.
657
What: /sys/devices/platform/mlxplat/mlxreg-io/hwmon/hwmon*/reset_ac_ok_fail
658
Date: February 2023
659
KernelVersion: 6.3
660
Contact: Vadim Pasternak <[email protected]>
661
Description: This file shows the system reset cause due to AC power failure.
662
Value 1 in file means this is reset cause, 0 - otherwise.
664
The file is read only.
666
What: /sys/devices/platform/mlxplat/mlxreg-io/hwmon/hwmon*/cpld5_pn
667
What: /sys/devices/platform/mlxplat/mlxreg-io/hwmon/hwmon*/cpld5_version
668
What: /sys/devices/platform/mlxplat/mlxreg-io/hwmon/hwmon*/cpld5_version_min
669
Date: August 2023
670
KernelVersion: 6.6
671
Contact: Vadim Pasternak <[email protected]>
672
Description: These files show with which CPLD part numbers, version and minor
673
versions have been burned the 5-th CPLD device equipped on a
674
system.
676
The files are read only.
678
What: /sys/devices/platform/mlxplat/mlxreg-io/hwmon/hwmon*/jtag_cap
679
Date: August 2023
680
KernelVersion: 6.6
681
Contact: Vadim Pasternak <[email protected]>
682
Description: This file indicates the available method of CPLD/FPGA devices
683
field update through the JTAG chain:
685
b00 - field update through LPC bus register memory space.
686
b01 - Reserved.
687
b10 - Reserved.
688
b11 - field update through CPU GPIOs bit-banging.
690
The file is read only.
692
What: /sys/devices/platform/mlxplat/mlxreg-io/hwmon/hwmon*/lid_open
693
Date: August 2023
694
KernelVersion: 6.6
695
Contact: Vadim Pasternak <[email protected]>
696
Description: 1 - indicates that system lid is opened, otherwise 0.
698
The file is read only.
700
What: /sys/devices/platform/mlxplat/mlxreg-io/hwmon/hwmon*/reset_long_pwr_pb
701
Date: August 2023
702
KernelVersion: 6.6
703
Contact: Vadim Pasternak <[email protected]>
704
Description: This file if set 1 indicates that system has been reset by
705
long press of power button.
707
The file is read only.
709
What: /sys/devices/platform/mlxplat/mlxreg-io/hwmon/hwmon*/reset_swb_dc_dc_pwr_fail
710
Date: August 2023
711
KernelVersion: 6.6
712
Contact: Vadim Pasternak <[email protected]>
713
Description: This file shows 1 in case the system reset happened due to the
714
failure of any DC-DC power converter devices equipped on the
715
switch board.
717
The file is read only.
719
What: /sys/devices/platform/mlxplat/mlxreg-io/hwmon/hwmon*/global_wp_request
720
Date: May 2025
721
KernelVersion: 6.16
722
Contact: Vadim Pasternak <[email protected]>
723
Description: This file when written 1 activates request to allow access to
724
the write protected flashes. Such request can be performed only
725
for system equipped with BMC (Board Management Controller),
726
which can grant access to protected flashes. In case BMC allows
727
access - it will respond with "global_wp_response". BMC decides
728
regarding time window of granted access. After granted window is
729
expired, BMC will change value back to 0.
730
Default value is 0.
732
The file is read/write.
734
What: /sys/devices/platform/mlxplat/mlxreg-io/hwmon/hwmon*/global_wp_response
735
Date: May 2025
736
KernelVersion: 6.16
737
Contact: Vadim Pasternak <[email protected]>
738
Description: This file, when set 1, indicates that access to protected
739
flashes have been granted to host CPU by BMC.
740
Default value is 0.
742
The file is read only.
744
What: /sys/devices/platform/mlxplat/mlxreg-io/hwmon/hwmon*/shutdown_unlock
745
Date: May 2025
746
KernelVersion: 6.16
747
Contact: Vadim Pasternak <[email protected]>
748
Description: When ASICs are getting overheated, system protection
749
hardware mechanism enforces system reboot. After system
750
reboot ASICs come up in locked state. To unlock ASICs,
751
this file should be written 1
752
Default value is 0.
754
The file is read/write.
756
What: /sys/devices/platform/mlxplat/i2c_mlxcpld.*/i2c-*/i2c-*/*-00**/mlxreg-io.*/hwmon/hwmon*/boot_progress
757
Date: May 2025
758
KernelVersion: 6.16
759
Contact: Vadim Pasternak <[email protected]>
760
Description: These files show the Data Process Unit board boot progress
761
state. Valid states are:
762
- 4 : OS starting.
763
- 5 : OS running.
764
- 6 : Low-Power Standby.
766
The file is read only.
768
What: /sys/devices/platform/mlxplat/i2c_mlxcpld.*/i2c-*/i2c-*/*-00**/mlxreg-io.*/hwmon/hwmon*/dpu_id
769
Date: May 2025
770
KernelVersion: 6.16
771
Contact: Vadim Pasternak <[email protected]>
772
Description: This file shows hardware Id of Data Process Unit board.
774
The file is read only.
776
What: /sys/devices/platform/mlxplat/i2c_mlxcpld.*/i2c-*/i2c-*/*-00**/mlxreg-io.*/hwmon/hwmon*/reset_aux_pwr_or_reload
777
What: /sys/devices/platform/mlxplat/i2c_mlxcpld.*/i2c-*/i2c-*/*-00**/mlxreg-io.*/hwmon/hwmon*/reset_dpu_thermal
778
What: /sys/devices/platform/mlxplat/i2c_mlxcpld.*/i2c-*/i2c-*/*-00**/mlxreg-io.*/hwmon/hwmon*/reset_from_main_board
779
Date: May 2025
780
KernelVersion: 6.16
781
Contact: Vadim Pasternak <[email protected]>
782
Description: These files expose the cause of the most recent reset of the Data
783
Processing Unit (DPU) board. The possible causes are:
784
- Power auxiliary outage or power reload.
785
- Thermal shutdown.
786
- Reset request from the main board.
787
Value 1 in file means this is reset cause, 0 - otherwise. Only one of
788
the above causes could be 1 at the same time, representing only last
789
reset cause.
791
The files are read only.
793
What: /sys/devices/platform/mlxplat/i2c_mlxcpld.*/i2c-*/i2c-*/*-00**/mlxreg-io.*/hwmon/hwmon*/perst_rst
794
What: /sys/devices/platform/mlxplat/i2c_mlxcpld.*/i2c-*/i2c-*/*-00**/mlxreg-io.*/hwmon/hwmon*/phy_rst
795
What: /sys/devices/platform/mlxplat/i2c_mlxcpld.*/i2c-*/i2c-*/*-00**/mlxreg-io.*/hwmon/hwmon*/tpm_rst
796
What: /sys/devices/platform/mlxplat/i2c_mlxcpld.*/i2c-*/i2c-*/*-00**/mlxreg-io.*/hwmon/hwmon*/usbphy_rst
797
Date: May 2025
798
KernelVersion: 6.16
799
Contact: Vadim Pasternak <[email protected]>
800
Description: These files allow to reset hardware components of Data Process
801
Unit board. Respectively PCI, Ethernet PHY, TPM and USB PHY
802
resets.
803
Default values for all the attributes is 1. Writing 0 will
804
cause reset of the related component.
806
The files are read/write.
808
What: /sys/devices/platform/mlxplat/i2c_mlxcpld.*/i2c-*/i2c-*/*-00**/mlxreg-io.*/hwmon/hwmon*/ufm_upgrade
809
Date: May 2025
810
KernelVersion: 6.16
811
Contact: Vadim Pasternak <[email protected]>
812
Description: These files show status of Unified Fabric Manager upgrade.
813
state. 0 - means upgrade is done, 1 - otherwise.
815
The file is read only.
3. 한국어 전문 번역
영어 원문의 문단 순서와 의미를 유지한 전체 번역입니다. 코드, 함수명, symbol과 URL은 원문 표기를 유지합니다.
경로와 값 해석
1-815별도 표기가 없는 모든 항목의 담당자는 Vadim Pasternak <[email protected]>다. Chassis-level 공통 prefix는 /sys/devices/platform/mlxplat/mlxreg-io/hwmon/hwmon*/이다. 각 표는 이 prefix 뒤의 attribute 이름을 표시한다.
Reset-cause 파일은 대부분 1이면 해당 원인이 가장 최근 reset 원인이고 0이면 아니라고 해석한다. 같은 group에서는 한 원인만 1이 될 수 있다. Control 파일은 이름과 달리 active-low 의미를 갖는 경우가 있으므로 각 항목의 0·1 정의를 따라야 한다.
ASIC, CPLD, fan, JTAG와 IIO
1-57| Attribute | 도입 | 접근 | 설명 |
|---|---|---|---|
| asic_health | 2018-06 · 4.19 | read-only | ASIC health status다. 0은 health failed, 2는 health OK, 3은 ASIC booting state다. |
| cpld1_version, cpld2_version | 2018-06 · 4.19 | read-only | Carrier board와 switch board에 burn된 CPLD version을 표시한다. |
| fan_dir | 2018-12 · 5.0 | read-only | System fan 방향이다. 관련 bit가 0이면 forward, 1이면 reversed direction이다. |
| cpld3_version | 2018-11 · 5.0 | read-only | LED board 또는 Gearbox board에 burn된 CPLD version을 표시한다. |
| jtag_enable | 2018-11 · 5.0 | read/write | JTAG domain 접근을 enable 또는 disable한다. 기본적으로 JTAG 접근은 disabled다. |
| select_iio | 2018-06 · 4.19 | read/write | 0 또는 1을 써서 접근할 IIO device 하나를 선택한다. |
PSU와 system power action
59-80psu1_on, psu2_on, pwr_cycle, pwr_down은 2018년 6월 kernel 4.19에 도입된 write-only 파일이다.
| Attribute | 기록값 | 동작 |
|---|---|---|
| pwr_cycle | 1 | Auxiliary power domain을 내린 뒤 약 1초 후 다시 올린다. |
| psu1_on, psu2_on | 1 / 0 | 1이면 해당 PSU를 power source에서 분리하고 0이면 연결한다. 둘 다 1이면 PSU main power domain이 내려간다. |
| pwr_down | 1 | System main power domain을 내린다. |
초기 platform reset-cause group
81-145| Attribute group | 도입 | 가능한 reset 원인 |
|---|---|---|
| reset_aux_pwr_or_ref, reset_asic_thermal, reset_hotswap_or_halt, reset_hotswap_or_wd, reset_fw_reset, reset_long_pb, reset_main_pwr_fail, reset_short_pb, reset_sw_reset | 2018-06 · 4.19 | Auxiliary power outage 또는 refresh, ASIC thermal shutdown, halt, hotswap, watchdog, firmware reset, power button long·short press, main-power failure와 software reset을 나타내는 read-only bit다. |
| reset_comex_pwr_fail, reset_from_comex, reset_system, reset_voltmon_upgrade_fail | 2018-11 · 5.0 | ComEx power failure, ComEx가 요청한 reset, system platform reset, voltage-monitor device upgrade failure를 나타내는 read-only bit다. |
| cpld4_version | 2018-11 · 5.0 | LED board에 burn된 CPLD version을 표시하는 read-only 값이다. |
| reset_comex_thermal, reset_comex_wd, reset_from_asic, reset_reload_bios, reset_sff_wd, reset_swb_wd | 2019-06 · 5.3 | COMEX thermal shutdown, COMEX·switch board·SFF mezzanine watchdog power-off 또는 reset, ASIC 요청 reset, BIOS reload reset을 나타내는 read-only bit다. |
Power, thermal, watchdog, firmware 또는 software event 발생→CPLD가 last-reset cause bit 기록→Boot 후 reset_* 파일을 읽음→값 1인 attribute가 최근 reset 원인→같은 cause group의 나머지 값은 0
각 platform generation은 원인별 bit를 별도 파일로 내보내지만 가장 최근 reset 원인 하나만 유효하다.
Topology ID, reset retention과 firmware status
147-225| Attribute | 도입 | 접근 | 설명 |
|---|---|---|---|
| config1, config2 | 2020-01 · 5.6 | read-only | Static I2C topology와 system 안 FPGA device의 수·type 등 static topology identification을 표시한다. |
| reset_ac_pwr_fail, reset_platform, reset_soc, reset_sw_pwr_off | 2020-01 · 5.6 | read-only | AC power failure, software가 CPLD reset signal을 assert한 platform reset, SoC가 ACPI register를 통해 assert한 signal, software가 CPLD power-off signal을 assert한 reset 원인을 표시한다. |
| pcie_asic_reset_dis | 2020-01 · 5.6 | read/write | 1로 설정하면 PCIe root-complex reset 동안 ASIC을 up 상태로 유지한다. |
| vpd_wp | 2020-01 · 5.6 | read/write | 1로 설정하면 system VPD의 hardware write protection을 override한다. |
| voltreg_update_status | 2020-01 · 5.6 | read-only | Burn 가능한 voltage regulator의 configuration update 상태다. 0: OK, 1: CRC failure, 2: I2C failure, 3: in progress. |
| ufm_version | 2020-01 · 5.6 | read-only | Burn 가능한 voltage-regulator device의 firmware version을 표시한다. |
| cpld1_pn, cpld2_pn, cpld3_pn, cpld4_pn, cpld1_version_min, cpld2_version_min, cpld3_version_min, cpld4_version_min | 2020-07 · 5.9 | read-only | System에 장착된 CPLD device에 burn된 part number와 minor version을 표시한다. |
Dual-BIOS image와 upgrade failure
227-249bios_active_image, bios_auth_fail, bios_upgrade_fail은 2021년 10월 kernel 5.16에 도입된 read-only 파일이다.
| Attribute | 값과 의미 |
|---|---|
| bios_active_image | 현재 active BIOS image 위치다. 0은 Top, 1은 Bottom이다. BIOS safe mode에서는 OS가 기대하는 값과 맞는 보고값이 0이어야 한다. 같은 flash의 DualBIOS를 위한 Intel top-swap 기능과 관련된 bit다. |
| bios_auth_fail | 제공된 BIOS image의 서명이 올바르지 않아 BIOS upgrade가 실패했음을 나타낸다. |
| bios_upgrade_fail | Authentication 이외의 이유로 BIOS upgrade가 실패했음을 나타낸다. 예를 들면 physical SPI flash 문제다. |
Line-card enable, power와 reset-mask
250-304다음 세 group은 2021년 10월 kernel 5.16에 도입됐으며 n은 1부터 8까지다.
| Attribute | 접근 | 0·1 동작 |
|---|---|---|
| lc1_enable, lc2_enable, lc3_enable, lc4_enable, lc5_enable, lc6_enable, lc7_enable, lc8_enable | read/write | lc{n}_enable=1이면 해당 line card를 reset state에서 release하고 0이면 reset 상태로 유지한다. |
| lc1_pwr, lc2_pwr, lc3_pwr, lc4_pwr, lc5_pwr, lc6_pwr, lc7_pwr, lc8_pwr | read/write | lc{n}_pwr=1이면 해당 line card power를 켜고 0이면 끈다. |
| lc1_rst_mask, lc2_rst_mask, lc3_rst_mask, lc4_rst_mask, lc5_rst_mask, lc6_rst_mask, lc7_rst_mask, lc8_rst_mask | write-only | ASIC의 비정상 동작 때문에 enforce된 해당 line-card reset bit를 1로 써서 clear한다. 0은 아무 효과가 없다. |
OS handoff와 power-budget ownership
305-346| Attribute | 도입 | 접근 | 설명 |
|---|---|---|---|
| os_started | 2021-10 · 5.16 | read/write | 1을 쓰면 OS가 programmable device 제어권을 인수했음을 device에 알린다. |
| pm_mgmt_en | 2021-10 · 5.16 | read/write | Power-management control owner를 정한다. 1은 software control, 0은 hardware control이며 기본값은 0이다. |
| psu3_on, psu4_on | 2021-10 · 5.16 | write-only | 1이면 해당 PSU를 power source에서 분리하고 0이면 연결한다. |
pm_mgmt_en=0에서 hardware는 system power budget이 부족해지면 이미 켜진 line card 하나 이상을 자동으로 끌 수 있다. 예를 들어 PSU가 power-good 상태를 잃었을 때다. pm_mgmt_en 값과 관계없이 budget이 부족하면 hardware는 새 line card power-on을 허용하지 않는다. Software가 여러 card를 동시에 켜려 해도 budget이 허용하는 범위까지만 hardware가 켠다.
Thermal shutdown 뒤 ASIC unlock: 2021 정의
348-367| 항목 | 내용 |
|---|---|
| Attribute | shutdown_unlock |
| Date / Kernel | 2021-10 · 5.16 |
| Access | read/write |
| Default | 1 |
ASIC이 system thermal shutdown을 enforce하면 ASIC이 lock되고 다음 boot에서도 사용할 수 없다. Software는 shutdown_unlock=1로 unlock을 요청한 뒤 pwr_cycle=1로 main power domain을 power-cycle할 수 있다.
1을 쓰기 전에 reboot 원인이 reset_asic_thermal 또는 reset_thermal_spc_or_pciesw인지 확인하는 것이 권장된다. shutdown_unlock을 1로 설정하지 않으면 external power-distribution unit을 통한 full-system power cycle만 ASIC lock을 풀 수 있다.
Line-card CPLD·FPGA와 VPD
368-404Line-card attribute의 공통 prefix는 /sys/devices/platform/mlxplat/i2c_mlxcpld.*/i2c-*/i2c-*/i2c-*/*-0032/mlxreg-io.*/hwmon/hwmon*/이다. 이 절의 항목은 2021년 10월 kernel 5.16에 도입됐다.
| Attribute | 접근 | 설명 |
|---|---|---|
| cpld1_pn, cpld1_version, cpld1_version_min | read-only | Line card CPLD에 burn된 part number와 major·minor version을 표시한다. |
| fpga1_pn, fpga1_version, fpga1_version_min | read-only | Line card FPGA에 burn된 part number와 major·minor version을 표시한다. |
| vpd_wp | read/write | Line-card VPD hardware write-protection mode를 override한다. 1은 write protection disabled, 0은 enabled이며 기본값은 0이다. System lockdown mode에서는 쓸 수 없다. Production flow에서 line-card VPD를 burn하기 위한 interface다. |
Line-card reset cause와 programmable-device upgrade
405-459| Attribute group | 접근 | 설명 |
|---|---|---|
| reset_aux_pwr_or_ref, reset_dc_dc_pwr_fail, reset_fpga_not_done, reset_from_chassis, reset_line_card, reset_pwr_off_from_chassis | read-only | Line-card의 auxiliary-power outage 또는 refresh, DC-to-DC power failure, FPGA reset failure, chassis 요청 reset, line-card reset failure, chassis power-off 원인을 표시한다. 가장 최근 원인 하나만 1이다. |
| cpld_upgrade_en, fpga_upgrade_en | read/write | 1이면 CPLD 또는 FPGA burning을 enable하고 0이면 disable한다. Lockdown mode에서는 쓸 수 없다. JTAG daisy-chain을 통한 line-card CPLD·FPGA upgrade에 사용한다. |
| agb_spi_burn_en, fpga_spi_burn_en | read/write | 1이면 Gearbox와 FPGA SPI-flash burning을 enable하고 0이면 disable한다. Lockdown mode에서는 쓸 수 없다. Production flow에서 line-card Gearbox·FPGA를 burn하는 용도다. |
Line-card power와 board configuration
437-469| Attribute | 접근 | 설명 |
|---|---|---|
| qsfp_pwr_en | read/write | 모든 QSFP port power를 제어한다. 1은 power on, 0은 power off다. |
| pwr_en | read/write | Line card 전체 power를 제어한다. 1은 power on, 0은 power off다. |
| max_power | read-only | Line card 공급에 필요한 최대 power를 제공한다. |
| config | read-only | Line-card configuration ID를 제공한다. |
qsfp_pwr_en·pwr_en과 max_power·config는 모두 2021년 10월 kernel 5.16에 도입됐다. agb_spi_burn_en·fpga_spi_burn_en은 앞 절과 source line 447-459가 겹치지만 원문 구조를 보존하기 위해 upgrade group에서 설명했다.
PHY, MAC, QSFP와 ASIC
471-524| Attribute | 도입 | 접근 | 값과 동작 |
|---|---|---|---|
| phy_reset | 2022-05 · 5.19 | read/write | 1이면 모든 PHY 88E1548을 reset에서 release하고 0이면 reset 상태로 유지한다. |
| mac_reset | 2022-05 · 5.19 | read/write | 1이면 ASIC MT52132를 reset에서 release하고 0이면 reset 상태로 유지한다. |
| qsfp_pwr_good | 2022-05 · 5.19 | read-only | QSFP port power status다. 원문은 QSFP가 하나라도 꽂히면 0, 아무 QSFP도 없으면 1이라고 설명하며 값 정의는 0: Power good, 1: Not power good다. |
| asic2_health | 2022-07 · 5.20 | read-only | 두 번째 ASIC health다. 0: failed, 2: OK, 3: booting. |
| asic_reset, asic2_reset | 2022-07 · 5.20 | write-only | 각 파일에 1을 써서 대응 ASIC을 reset한다. |
Host communication, COME config와 converter reset
525-554| Attribute | 도입 | 접근 | 설명 |
|---|---|---|---|
| comm_chnl_ready | 2022-07 · 5.20 | write-only | 1을 써서 system host CPU가 remote endpoint, 예를 들어 BMC에 telemetry data를 보낼 준비가 됐음을 알린다. |
| config3 | 2020-01 · 5.6 | read-only | COME module hardware configuration을 나타낸다. Hardware가 GPIO pin을 통해 값을 전달하며 같은 system SKU의 minor BOM 변경을 노출한다. |
| reset_pwr_converter_fail | 2023-02 · 6.3 | read-only | Power-converter device failure 때문에 reset됐다면 1, 아니면 0이다. |
External Root of Trust monitor와 recovery
555-614| Attribute | 접근 | 설명 |
|---|---|---|
| erot1_ap_reset, erot2_ap_reset | read-only | EROT processor가 Application Processor(AP)로 내보내는 RESET 상태를 표시한다. EROT가 AP firmware를 invalid 또는 revoked로 판정하면 valid firmware가 load될 때까지 AP를 reset 상태로 유지해 unauthorized firmware 실행을 막는다. 정상 흐름에서는 firmware integrity 검증 직후 AP reset을 가능한 빨리 release한다. |
| erot1_recovery, erot2_recovery, erot1_reset, erot2_reset | read/write | EROT failure 뒤 recovery sequence를 수행한다. EROT는 system power-up의 가장 이른 단계에서 reset이 release되어 CPU가 ASIC과 통신하기 전 boot·authentication을 시작해야 한다. Recovery signal을 assert한 채 EROT를 reset하면 AP가 recovery mode로 들어가 firmware를 update 또는 복구할 수 있다. 대응 파일을 1/0으로 toggle한다. |
| erot1_wp, erot2_wp | read/write | EROT reset·recovery 접근을 허용한다. 기본값은 0, 즉 programming disabled다. Lockdown mode에서는 쓸 수 없다. |
| spi_chnl_select | read/write | EROT out-of-band recovery에 사용할 SPI chip을 0 또는 1로 선택한다. |
이 절의 모든 ABI는 2023년 2월 kernel 6.3에 도입됐다.
System power-up과 EROT boot→EROT가 AP firmware integrity 검증→Invalid·revoked이면 AP RESET 유지→Valid이면 AP RESET release→AP boot 후 CPU가 ASIC과 통신→EROT failure 시 recovery assert와 reset toggle로 recovery mode 진입
EROT 검증과 AP reset release가 먼저 완료되어야 CPU가 ASIC 통신을 안전하게 시작할 수 있다.
ASIC power-good, clock board와 converter programming
615-664| Attribute | 접근 | 설명 |
|---|---|---|
| asic_pg_fail | read-only | ASIC Power Good failure 상태다. 1이면 Power Good failed, 0이면 아니다. |
| clk_brd1_boot_fail | read-only | 첫 번째 clock board가 CI에서 boot하지 못했다는 warning이다. |
| clk_brd2_boot_fail | read-only | 두 번째 clock board가 CI에서 boot하지 못했다는 warning이다. |
| clk_brd_fail | read-only | 공통 clock-board boot failure error다. |
| clk_brd_prog_en | read/write | Clock-board programming을 enable한다. 기본값은 0이며 lockdown mode에서는 쓸 수 없다. |
| pwr_converter_prog_en | read/write | Power-converter programming을 enable한다. 기본값은 0이며 lockdown mode에서는 쓸 수 없다. |
| reset_ac_ok_fail | read-only | AC power failure 때문에 system reset이 발생했다면 1, 아니면 0이다. |
이 절의 모든 ABI는 2023년 2월 kernel 6.3에 도입됐다. asic_pg_fail과 clock-board status의 원문 Contact 표기는 angle bracket 없이 적혀 있지만 같은 Vadim Pasternak 주소다.
CPLD5, JTAG update method와 chassis 상태
666-717| Attribute | 도입 | 접근 | 설명 |
|---|---|---|---|
| cpld5_pn, cpld5_version, cpld5_version_min | 2023-08 · 6.6 | read-only | System의 다섯 번째 CPLD에 burn된 part number, major version과 minor version을 표시한다. |
| jtag_cap | 2023-08 · 6.6 | read-only | JTAG chain을 통한 CPLD·FPGA field-update 방법이다. b00은 LPC-bus register memory space, b01·b10은 reserved, b11은 CPU GPIO bit-banging이다. |
| lid_open | 2023-08 · 6.6 | read-only | 1이면 system lid가 열렸고, 그 밖에는 0이다. |
| reset_long_pwr_pb | 2023-08 · 6.6 | read-only | Power button을 길게 눌러 system이 reset됐다면 1이다. |
| reset_swb_dc_dc_pwr_fail | 2023-08 · 6.6 | read-only | Switch board에 장착된 DC-DC power-converter device 하나의 failure 때문에 reset됐다면 1이다. |
BMC를 통한 global write-protection grant
719-742global_wp_request와 global_wp_response는 2025년 5월 kernel 6.16에 도입됐고 기본값은 모두 0이다.
Host CPUmlxreg-ioBMC
01
global_wp_request=1접근 요청 전달정책 확인
02
응답 대기global_wp_response허용 시 1 기록
03
Protected flash 접근승인 window 유지시간 창 관리
04
Window 만료response가 0으로 복귀접근 회수
Host는 request만 만들고 실제 접근 허용 여부와 시간 창은 BMC가 결정한다.
| Attribute | 접근 | 설명 |
|---|---|---|
| global_wp_request | read/write | 1을 쓰면 write-protected flash 접근 요청을 활성화한다. BMC가 있는 system에서만 가능하며 BMC가 허용하면 global_wp_response로 응답한다. 허용 시간 창은 BMC가 정하고 만료 후 값을 0으로 되돌린다. |
| global_wp_response | read-only | 1이면 BMC가 host CPU에 protected flash 접근을 허용했음을 뜻한다. |
Overheat 뒤 ASIC unlock: 2025 정의
744-754| 항목 | 내용 |
|---|---|
| Attribute | shutdown_unlock |
| Date / Kernel | 2025-05 · 6.16 |
| Access | read/write |
| Default | 0 |
ASIC이 과열되면 system protection hardware가 reboot를 enforce하고, reboot 뒤 ASIC은 locked 상태로 올라온다. ASIC을 unlock하려면 이 파일에 1을 쓴다.
같은 경로의 2021년 정의는 기본값 1과 power-cycle 절차를 설명하지만 이 2025년 정의는 기본값 0이라고 명시한다. 원문에 존재하는 platform·generation별 차이를 합치지 않고 둘 다 보존했다.
DPU boot, reset cause와 component reset
756-806DPU attribute 공통 prefix는 /sys/devices/platform/mlxplat/i2c_mlxcpld.*/i2c-*/i2c-*/*-00**/mlxreg-io.*/hwmon/hwmon*/이다. 이 절의 모든 항목은 2025년 5월 kernel 6.16에 도입됐다.
| Attribute | 접근 | 설명 |
|---|---|---|
| boot_progress | read-only | DPU board boot 진행 상태다. 4: OS starting, 5: OS running, 6: Low-Power Standby. |
| dpu_id | read-only | DPU board의 hardware ID를 표시한다. |
| reset_aux_pwr_or_reload, reset_dpu_thermal, reset_from_main_board | read-only | 가장 최근 DPU reset 원인을 나타낸다. 각각 auxiliary-power outage 또는 reload, thermal shutdown, main-board reset request다. 원인 하나만 1이고 나머지는 0이다. |
| perst_rst, phy_rst, tpm_rst, usbphy_rst | read/write | 각각 DPU의 PCIe, Ethernet PHY, TPM, USB PHY를 reset한다. 기본값은 모두 1이며 0을 쓰면 해당 component가 reset된다. |
DPU Unified Fabric Manager upgrade
808-815| 항목 | 내용 |
|---|---|
| Attribute | ufm_upgrade |
| Date / Kernel | 2025-05 · 6.16 |
| Access | read-only |
| Description | Unified Fabric Manager upgrade 상태다. 0이면 upgrade 완료, 1이면 아직 완료되지 않았다. |
Platform health, version과 reset cause
sysfs-driver-mlxreg-io:1-145ASIC health, CPLD version, fan direction, JTAG·IIO 선택, power command와 여러 generation의 last-reset cause bit를 제공합니다.