Difference between revisions of "Talos II/Firmware"

From RCS Wiki
Jump to navigation Jump to search
Line 18: Line 18:
  
 
Known Issues
 
Known Issues
* While significantly lessened in this release, under certain rare circumstances during IPL the OCC can malfunction and lock the FSI bus.  This leads to loss of communication and fan controls, with all fans stuck on high, and can in some situations cause the second CPU to be guarded out.  This issue is exacerbated by the fast reboot functionality, which is disabled by default in the current PNOR.  The underlying fault is being tracked upstream in [https://github.com/openbmc/openbmc/issues/1699 Issue 1699].  Should the second CPU guard out, issue "pflash -P GUARD -c" from the BMC shell to remove the spurious guard entries.
+
* When fast reboot is enabled, under certain rare circumstances during IPL the OCC can malfunction and lock the FSI bus.  This leads to loss of communication and fan controls, with all fans stuck on high, and can in some situations cause the second CPU to be guarded out.  As a result, the fast reboot functionality is disabled by default in the current PNOR.
  
 
== System Package v1.05 ==
 
== System Package v1.05 ==

Revision as of 13:22, 20 July 2018

Talos II Official Firmware Builds

For upgrade instructions, please visit the Firmware Upgrade Quick Start page. Alternatively, for those wishing to build from source, please visit the Compiling Firmware page.

All firmware builds are cryptographically signed with the current Raptor Computing Systems firmware signing key at time of release [1]. The intermediate firmware signer keys, in turn, are signed by our master umbrella key [2].

System Package v1.06

Released: 06-19-2018

Change Log

  • Upgrade BMC kernel
  • Add Lite support to BMC
  • Upgrade PNOR stack to latest upstream versions
  • Update FPGA logic to support the Lite hardware configuration

Known Issues

  • When fast reboot is enabled, under certain rare circumstances during IPL the OCC can malfunction and lock the FSI bus. This leads to loss of communication and fan controls, with all fans stuck on high, and can in some situations cause the second CPU to be guarded out. As a result, the fast reboot functionality is disabled by default in the current PNOR.

System Package v1.05

Released: 05-28-2018

Change Log

  • Upgrade BMC kernel to Linux 4.13
  • Add FSI bus driver error recovery
  • Upgrade PNOR stack to latest upstream versions
  • Modify FPGA logic to conform to ATX specifications

Known Issues

  • While significantly lessened in this release, under certain rare circumstances during IPL the OCC can malfunction and lock the FSI bus. This leads to loss of communication and fan controls, with all fans stuck on high, and can in some situations cause the second CPU to be guarded out. This issue is exacerbated by the fast reboot functionality, which is disabled by default in the current PNOR. The underlying fault is being tracked upstream in Issue 1699. Should the second CPU guard out, issue "pflash -P GUARD -c" from the BMC shell to remove the spurious guard entries.

System Package v1.04

Released: 05-04-2018

Change Log

  • Fix regression accidentally introduced in System Package v1.03 where the fan controls do not engage on single CPU systems
  • Tweak chassis fan settings to minimize audible hunting

Known Issues

  • Under certain rare circumstances during IPL, the OCC can malfunction and lock the FSI bus. This leads to loss of communication and fan controls, with all fans stuck on high, and can in some situations cause the second CPU to be guarded out. This issue is exacerbated by the fast reboot functionality, which is disabled by default in the current PNOR. The underlying fault is being tracked upstream in Issue 1699. Should the second CPU guard out, issue "pflash -P GUARD -c" from the BMC shell to remove the spurious guard entries.

System Package v1.03

Released: 04-30-2018

Change Log

  • Upgrade PNOR stack to latest upstream versions
  • Disable fast reboot by default (work around Issue 1699 causing fans stuck on full speed)
  • Use PID control loop for fans instead of original (limited functionality) IBM fan control loop

Known Issues

  • Under certain rare circumstances during IPL, the OCC can malfunction and lock the FSI bus. This leads to loss of communication and fan controls, with all fans stuck on high, and can in some situations cause the second CPU to be guarded out. This issue is exacerbated by the fast reboot functionality, which is disabled by default in the current PNOR. The underlying fault is being tracked upstream in Issue 1699. Should the second CPU guard out, issue "pflash -P GUARD -c" from the BMC shell to remove the spurious guard entries.

System Package v1.02

Released: 04-20-2018

Change Log

  • Fix certain DIMMs with unusual SPD frequency values
  • Raise CPU core temperature setpoints
  • Load less aggressive fan curves for CPU temperature control

Known Issues

  • When fast reboot is enabled, the fan controls may stop working after a reboot. A normal reboot (host shutdown, power off, power on, IPL) restores the fan control to normal operation. The FSI bus lockups remain a significant upstream bug in the standard OpenPOWER firmware, and Raptor Computing Systems is waiting for a fix from IBM for the FSI lockup that will also fix the fan controls once and for all.

System Package v1.01

Released: 04-15-2018

Change Log

  • Fix spurious guard of second CPU package.
  • Fix fan control disengaging during IPL
  • Add on-board VGA disable jumper (J10109)
  • Revup host PNOR packages to latest upstream versions
  • Enable WoF on 18 and 22 core packages
  • Enable 2666MHz DDR4 memory DIMMs
  • Add on-board VGA disable jumper support

System Package v1.00

Released: 03-26-2018

Change Log

  • Initial release