Releases: Mellanox/hw-mgmt
V.7.0010.2200-rc1
Features
Add support for SDK OFFLINE event for handling flow with in service firmware upgrade
Bug fixes
fix fan_dir attribute on fixed fan systems.
V.7.0010.2100-3
================================================================================
- V.7.0010.2100
- Sun, 10 Jan 2021
-
New Feature
o For MSN4700 A1 (Leopard re-spin): add support 5 mp2975 instead of 7 xdpe12284
o For MQM9700 (Gorilla)
add support for mp2975 and mp2888, (Align driver mp2975 with upstream)
*Kernel 4.19 - 0032-hwmon-pmbus-Add-support-for-MPS-Multi-phase-mp2975-c.patch
*Kernel 4.19 - 0037-hwmon-pmbus-Add-support-for-MPS-Multi-phase-mp2888-c.patch
*Kernel 4.9 - 0065-hwmon-pmbus-Add-support-for-MPS-Multi-phase-mp2975-c.patch
*Kernel 4.9 - 0068-hwmon-pmbus-Add-support-for-MPS-Multi-phase-mp2888-c.patchadd support for 7th FAN and 7th LED and tachometers
*Kernel 4.19 - 0035-platform-x86-mlx-platform-Extend-FAN-and-LED-config-.patch
*Kernel 4.9 - 0066-platform-x86-mlx-platform-Extend-FAN-and-LED-config-.patcho For any non-Broadwell CPU: Initialize temperature limits critical, max, min and hysteresis of all discovered SODIMMs with temp sensor.
o Export appropriate GPIO pins (through sysfs) that are used for JTAG bit-bang CPLD burning. For platforms that support JTAG-GPIO FU. (Broadwell & Coffeelake)
o Add common fan direction attribute for SPC1/2/3 using sysfs, and replace existing thermal algorithm temperature difference mechanism for detecting fan direction with new fan_dir attribute
o For MSN4600 A1 (Liger re-spin like existing Leopard re-spin) add support 5 mp2975 instead of 7 xdpe12284. -
Bug fixes
o Fix dmesg dump
o Minimizing hw-mgmt dump when used in mlnx OS to not include overlapping data with sysdump
o Change during ASIC restart without active resume, the thermal control default from suspend to non-suspend.
o Add capability to build hw-management directly in vdi(centos)
o Fix issue mlxsw_minimal init failure - can't read asic_bus from config/asic_bus file not created yet.o Remove Patch: In systems w/o SMBIOS, predict system based on CPU (stays CL only)
*Kernel 4.19 - 0005-platform-x86-mlx-platform-DMI-match-for-marginal-sys.patch from ONL
o Add support for NVMe : Backport capability from kernel 5.6 to kernel 4.19 which allows NVME temperature reading using sysfs
*Kernel 4.19 - 0033-hwmon-Add-convience-macro-to-define-simple-static-se.patch
*Kernel 4.19 - 0034-backport-nvme-Add-hardware-monitoring-support.patch
o Modify ASIC shut down temperature threshold to 140c (SX, SPC 1/2/3)
*Kernel 4.19 - 0036-mlxsw-core-Increase-critical-threshold-for-ASIC-ther.patch
o Validate critical temperature threshold read from port cage.
*Kernel 4.19 - 0038-mlxsw-core-Add-validation-of-transceiver-temperature.patcho Modify ASIC shut down temperature threshold to 140c (SX, SPC 1/2/3)
*Kernel 4.9 - 0067-mlxsw-core-Increase-critical-threshold-for-ASIC-ther.patch
o Validate critical temperature threshold read from port cage.
*Kernel 4.9 - 0069-mlxsw-core-Add-validation-of-transceiver-temperature.patch -
Known issues and limitations:
o Kernel patch 4.9 #60 is available upstream from kernel 4.9.207 and
Kernel patch 4.19 #28 is available upstream from kernel 4.19.89.
- No need to apply these patches when working with these kernel versions
or above
o ethtool for QSFP-DD is working only in raw mode.
o SN4700 PSU (Murata) sensors PSU2 and PSU3 might be not available after insertion/removal.
o PSUs inventory read via PMBus require the following packages:
- i2c-tools_4.1-1_amd64.deb
- libi2c0_4.1-1_amd64.deb
o I2C Asic driver take up to 5 second to complete initialization. When
sending ADD even need to make sure to wait at list 5 second before
reset of ASIC.
o Systems SN2010, SN2100, SN2410, SN2700 and SN2740 (and their
"-B" variants) require the following flag in kernel cmdline:
"acpi_enforce_resources=lax acpi=noirq".
o Few bug fixes introduced in upstream kernel 4.19, whoever use older
v4.19 kernel then v4.19.58, should cherry pick the following commits:
- Fix wrong order in probing routine initialization:
d2d8f64012543898a0158b1fc5c07af3d41c89d8 (available in v4.19.49)
- Fix parent device in i2c-mux-reg device registration
c241f3fbfa1af86f572a92f2e4d708358e163806 (available in v4.19.58)
o Kernel patch 4.9 #37 is available upstream from kernel 4.9.197 and
Kernel patch 4.19 #9 is available upstream from kernel 4.19.79.
- No need to apply these patches when working with these kernel versions
or above
o This version requires FW version 29.2000.1886 or higher for spectrum-2
and 13.2000.1886 or higher for spectrum-1.
================================================================================
V.7.0010.2000
================================================================================
- V.7.0010.2000
- Wed, 22 Sep 2020
-
New Feature
-
Bug fixes
HW Mgmt core:
o In rare cases when thermal algorithem is Re-activiated after FAN/PSU insertion, FAN remians at high rpm
o When stop hw-managment code received error in the log instead of exit code '0'.
o In SPC1 i2c sometimes collide with chip reset coming from SDKo Remove raw eeprom data link, when working with PSU which don't have eeprom for "msn274x", "msn24xx" and "msn27xx" systems
*Kernel 4.19 - 0030-platform-x86-mlx-platform-Remove-PSU-EEPROM-configur.patch
*Kernel 4.19 - 0031-platform-x86-mlx-platform-Remove-PSU-EEPROM-configur.patch*Kernel 4.9 - 0063-platform-x86-mlx-platform-Remove-PSU-EEPROM-configur.patch *Kernel 4.9 - 0064-platform-x86-mlx-platform-Remove-PSU-EEPROM-configur.patch
o Fix memory leak on mlxsw_core_bus_device module removal
*Kernel 4.19 - 0029-mlxsw-core-Fix-memory-leak-on-module-removal.patch
*Kernel 4.9 - 0062-mlxsw-core-Fix-memory-leak-on-module-removal.patch -
Known issues and limitations:
o Kernel patch 4.9 #60 is available upstream from kernel 4.9.207 and
Kernel patch 4.19 #28 is available upstream from kernel 4.19.89.
- No need to apply these patches when working with these kernel versions
or above
o ethtool for QSFP-DD is working only in raw mode.
o SN4700 PSU (Murata) sensors PSU2 and PSU3 might be not available after insertion/removal.
o PSUs inventory read via PMBus require the following packages:
- i2c-tools_4.1-1_amd64.deb
- libi2c0_4.1-1_amd64.deb
o I2C Asic driver take up to 5 second to complete initialization. When
sending ADD even need to make sure to wait at list 5 second before
reset of ASIC.
o Systems SN2010, SN2100, SN2410, SN2700 and SN2740 (and their
"-B" variants) require the following flag in kernel cmdline:
"acpi_enforce_resources=lax acpi=noirq".
o Few bug fixes introduced in upstream kernel 4.19, whoever use older
v4.19 kernel then v4.19.58, should cherry pick the following commits:
- Fix wrong order in probing routine initialization:
d2d8f64012543898a0158b1fc5c07af3d41c89d8 (available in v4.19.49)
- Fix parent device in i2c-mux-reg device registration
c241f3fbfa1af86f572a92f2e4d708358e163806 (available in v4.19.58)
o Kernel patch 4.9 #37 is available upstream from kernel 4.9.197 and
Kernel patch 4.19 #9 is available upstream from kernel 4.19.79.
- No need to apply these patches when working with these kernel versions
or above
o This version requires FW version 29.2000.1886 or higher for spectrum-2
and 13.2000.1886 or higher for spectrum-1.
V.7.0010.1900
================================================================================
- V.7.0010.1900
- Wed, 16 Sep 2020
-
New Feature
o Add support for MSN4600 -
Bug fixes
HW Mgmt core:
o Fix thermal-controll dynamic minimum tables for the systems MSN2010, MSN2700, MSN2740, MSN3700, MSN3420, MSN3510.
MSN4700, MSN3800, MSN4600Co Fix thermal-controll stop service command. Previously in some cases it can take up to 60 sec.
-
Known issues and limitations:
o Kernel patch 4.9 #60 is available upstream from kernel 4.9.207 and
Kernel patch 4.19 #28 is available upstream from kernel 4.19.89.
- No need to apply these patches when working with these kernel versions
or above
o ethtool for QSFP-DD is working only in raw mode.
o SN4700 PSU (Murata) sensors PSU2 and PSU3 might be not available after insertion/removal.
o PSUs inventory read via PMBus require the following packages:
- i2c-tools_4.1-1_amd64.deb
- libi2c0_4.1-1_amd64.deb
o I2C Asic driver take up to 5 second to complete initialization. When
sending ADD even need to make sure to wait at list 5 second before
reset of ASIC.
o Systems SN2010, SN2100, SN2410, SN2700 and SN2740 (and their
"-B" variants) require the following flag in kernel cmdline:
"acpi_enforce_resources=lax acpi=noirq".
o Few bug fixes introduced in upstream kernel 4.19, whoever use older
v4.19 kernel then v4.19.58, should cherry pick the following commits:
- Fix wrong order in probing routine initialization:
d2d8f64012543898a0158b1fc5c07af3d41c89d8 (available in v4.19.49)
- Fix parent device in i2c-mux-reg device registration
c241f3fbfa1af86f572a92f2e4d708358e163806 (available in v4.19.58)
o Kernel patch 4.9 #37 is available upstream from kernel 4.9.197 and
Kernel patch 4.19 #9 is available upstream from kernel 4.19.79.
- No need to apply these patches when working with these kernel versions
or above
o This version requires FW version 29.2000.1886 or higher for spectrum-2
and 13.2000.1886 or higher for spectrum-1.
V.7.0010.1300
================================================================================
- V.7.0010.1300
- Sun, 23 Aug 2020
-
New Features
o Add dynamic minimum tables for MSN3700X, MSN3800, MSN3420, MSN4600,
MSN4700 systems
o Split hw-management to one-shot init hw-management service
and thermal control services.
o Add man pages for hw-management.service, hw-management-tc.service
and hw-management.sh -
Bug fixes
HW Mgmt core:
o Fix servics hang issue when doing multiple restart to thermal service in a loop
*Kernel 4.9- 0060-thermal-Fix-deadlock-in-thermal-thermal_zone_device_.patch
*Kernel 4.19 - 0028-thermal-Fix-deadlock-in-thermal-thermal_zone_device_.patch
o Move PSU EEPROM configuration from kerenl to user space for SPC2 / SPC3 system
*Kernel 4.9 - 0059-platform-x86-mlx-platform-Remove-PSU-EEPROM-configur.patch
*Kernel 4.19 - 0027-platform-x86-mlx-platform-Remove-PSU-EEPROM-configur.patch -
Known issues and limitations:
o Kernel patch 4.9 #60 is available upstream from kernel 4.9.207 and
Kernel patch 4.19 #28 is available upstream from kernel 4.19.89.
- No need to apply these patches when working with these kernel versions
or above
o ethtool for QSFP-DD is working only in raw mode.
o SN4700 PSU (Murata) sensors PSU2 and PSU3 might be not available after insertion/removal.
o PSUs inventory read via PMBus require the following packages:
- i2c-tools_4.1-1_amd64.deb
- libi2c0_4.1-1_amd64.deb
o I2C Asic driver take up to 5 second to complete initialization. When
sending ADD even need to make sure to wait at list 5 second before
reset of ASIC.
o Systems SN2010, SN2100, SN2410, SN2700 and SN2740 (and their
"-B" variants) require the following flag in kernel cmdline:
"acpi_enforce_resources=lax acpi=noirq".
o Few bug fixes introduced in upstream kernel 4.19, whoever use older
v4.19 kernel then v4.19.58, should cherry pick the following commits:
- Fix wrong order in probing routine initialization:
d2d8f64012543898a0158b1fc5c07af3d41c89d8 (available in v4.19.49)
- Fix parent device in i2c-mux-reg device registration
c241f3fbfa1af86f572a92f2e4d708358e163806 (available in v4.19.58)
o Kernel patch 4.9 #37 is available upstream from kernel 4.9.197 and
Kernel patch 4.19 #9 is available upstream from kernel 4.19.79.
- No need to apply these patches when working with these kernel versions
or above
o This version requires FW version 29.2000.1886 or higher for spectrum-2
and 13.2000.1886 or higher for spectrum-1.
V.7.0010.1000
- V.7.0010.1000
- Sun, 31 May 2020
-
New Features
o Add support for QSFP-DD cabel info
*Kernel 4.9 - 0054-Add-support-for-new-transceivers-types-QSFP-DD-and-Q.patch
*Kernel 4.19 - 0022-Add-support-for-new-transceivers-types-QSFP-DD-and-Q.patcho Thermal zone detection for thermal algorithm had move from user space to
kernel driver.
*Kernel 4.9 - 0055-mlxsw-core-Add-the-hottest-thermal-zone-detection.patch
- 0056-mlxsw-core-thermal-Separate-temperature-trend-read-c.patch
*Kernel 4.19 - 0023-mlxsw-core-Add-the-hottest-thermal-zone-detection.patch
- 0024-mlxsw-core-thermal-Separate-temperature-trend-read-c.patcho Support of watchdog long timeout period in the kernel
New watchdog type is discovered by CPLD watchdog capability.
The timeout period can be set to more the 32/255 sec and up to 24 hours.
*Kernel 4.9 - 0057-watchdog-mlx-wdt-support-new-watchdog-type-with-long.patch
*Kernel 4.19 - 0025-watchdog-mlx-wdt-support-new-watchdog-type-with-long.patcho Add support for reporting PSU FAN speed capabilities psu vpd data file.
o Setting initial FAN to 100% speed until thermal control is started. -
Known issues and limitations:
o ethtool for QSFP-DD is working only in raw mode.
o SN4700 PSU (Murata) sensors PSU2 and PSU3 might be not available after insertion/removal.
o PSUs inventory read via PMBus require the following packages:
- i2c-tools_4.1-1_amd64.deb
- libi2c0_4.1-1_amd64.deb
o I2C Asic driver take up to 5 second to complete initialization.
When sending ADD event, you need to make sure
to wait at list 5 seconds before reset of ASIC.
o Systems SN2010, SN2100, SN2410, SN2700 and SN2740 (and their
"-B" variants) require the following flag in kernel cmdline:
"acpi_enforce_resources=lax acpi=noirq".
o Few bug fixes introduced in upstream kernel 4.19, whoever use older
v4.19 kernel then v4.19.58, should cherry pick the following commits:
- Fix wrong order in probing routine initialization:
d2d8f64012543898a0158b1fc5c07af3d41c89d8 (available in v4.19.49)
- Fix parent device in i2c-mux-reg device registration
c241f3fbfa1af86f572a92f2e4d708358e163806 (available in v4.19.58)
o Kernel patch 4.9 #37 is available upstream from kernel 4.9.197 and
Kernel patch 4.19 #9 is available upstream from kernel 4.19.79.
- No need to apply these patches when working with these kernel versions
or above
o This version requires FW version 29.2000.1886 or higher for spectrum-2
and 13.2000.1886 or higher for spectrum-1.
V.7.0010.0001
RC Candidate for May 2020 release
V.7.0000.3034
=========================================================================
- V.7.0000.3034
- Sun, 26 Apr 2020
-
Bug fixes
HW Mgmt core:
o Fix in SN2010 wrong number of CPLD display.
o Remove error in dmsg in power On/Off flow of PSU in MSN2700/MSN2410 -
Known issues and limitations:
o SN4700 PSU (Murata) sensors might be not available after insertion/removal.
o PSUs inventory read via PMBus require the following packages:
- i2c-tools_4.1-1_amd64.deb
- libi2c0_4.1-1_amd64.deb
o I2C Asic driver takes up to 5 seconds to complete initialization. When
sending ADD even need to make sure to wait at list 5 seconds before
reset of ASIC.
o Systems SN2010, SN2100, SN2410, SN2700 and SN2740 (and their
"-B" variants) require the following flag in kernel cmdline:
"acpi_enforce_resources=lax acpi=noirq".
o Few bug fixes introduced in upstream kernel 4.19, whoever use older
v4.19 kernel then v4.19.58, should cherry pick the following commits:
- Fix wrong order in probing routine initialization:
d2d8f64012543898a0158b1fc5c07af3d41c89d8 (available in v4.19.49)
- Fix parent device in i2c-mux-reg device registration
c241f3fbfa1af86f572a92f2e4d708358e163806 (available in v4.19.58)
o Kernel patch 4.9 #37 is available upstream from kernel 4.9.197 and
Kernel patch 4.19 #9 is available upstream from kernel 4.19.79.
- No need to apply these patches when working with these kernel versions
or above
o This version requires FW version 29.2000.1886 or higher for spectrum-2
and 13.2000.1886 or higher for spectrum-1.
=========================================================================
V.7.0000.3032
- V.7.0000.3032
- Mon, 19 Apr 2020
-
Bug fixes
HW Mgmt core:
o Fix flow of attaching PSU EEPROM driver.
o Fix start systemd hw-management service with disabled thermal control. -
Known issues and limitations:
o SN4700 PSU (Murata) sensors might be not available after insertion/removal.
o PSUs inventory read via PMBus require the following packages:
- i2c-tools_4.1-1_amd64.deb
- libi2c0_4.1-1_amd64.deb
o I2C Asic driver take up to 5 second to complete initialization. When
sending ADD even need to make sure to wait at list 5 second before
reset of ASIC.
o Systems SN2010, SN2100, SN2410, SN2700 and SN2740 (and their
"-B" variants) require the following flag in kernel cmdline:
"acpi_enforce_resources=lax acpi=noirq".
o Few bug fixes introduced in upstream kernel 4.19, whoever use older
v4.19 kernel then v4.19.58, should cherry pick the following commits:
- Fix wrong order in probing routine initialization:
d2d8f64012543898a0158b1fc5c07af3d41c89d8 (available in v4.19.49)
- Fix parent device in i2c-mux-reg device registration
c241f3fbfa1af86f572a92f2e4d708358e163806 (available in v4.19.58)
o Kernel patch 4.9 #37 is available upstream from kernel 4.9.197 and
Kernel patch 4.19 #9 is available upstream from kernel 4.19.79.
- No need to apply these patches when working with these kernel versions
or above
o This version requires FW version 29.2000.1886 or higher for spectrum-2
and 13.2000.1886 or higher for spectrum-1.
V.7.0000.3030
======================================================================
- V.7.0000.3030
- Mon, 13 Apr 2020
- New Features
o Add support for MSN4600C.
o Add CPLD minor and part version.
o Add module status indicate - sfp_[index]_status
o Chassis Mgmt user manual - rev 1.7 - Bug fixes
HW Mgmt core:
o Fix - Prevent double notification when udev event sent with environmental data.
- see backport patch-kernel 4.9 #53
- see backport patch-kernel 4.19 #21
o Fix - number of fans for MSN3700C
Thermal:
o Fix - Remove the option of the thermal event, which might cause the failure of hw-management.
service start in case the thermal event received before the service start.
o Fix - Thermal vector update for MSN2700.
HW Mgmt doc:
o Fix - README.ME file - update the kernel config list. - Known issues and limitations:
o SN4700 PSU (Murata) sensors might be not available after insertion/removal.
o PSUs inventory read via PMBus require the following packages:
- i2c-tools_4.1-1_amd64.deb
- libi2c0_4.1-1_amd64.deb
o I2C Asic driver takes up to 5 seconds to complete initialization. When
sending ADD even needs to make sure to wait at list 5 seconds before reset of ASIC.
o Systems SN2010, SN2100, SN2410, SN2700 and SN2740 (and their
"-B" variants) require the following flag in kernel cmdline:
"acpi_enforce_resources=lax acpi=noirq".
o Few bug fixes introduced in upstream kernel 4.19, whoever use older
v4.19 kernel then v4.19.58, should cherry pick the following commits:
- Fix wrong order in probing routine initialization:
d2d8f64012543898a0158b1fc5c07af3d41c89d8 (available in v4.19.49)
- Fix parent device in i2c-mux-reg device registration
c241f3fbfa1af86f572a92f2e4d708358e163806 (available in v4.19.58)
o Kernel patch 4.9 #37 is available upstream from kernel 4.9.197 and.
Kernel patch 4.19 #9 is available upstream from kernel 4.19.79.
- No need to apply these patches when working with these kernel versions
or above
o This version requires FW version 29.2000.1886 or higher for spectrum-2 and 13.2000.1886 or higher for spectrum-1.
======================================================================