As L0 Syman is admin and it should have flexible to provide an API for end user to give GPU dump/crashlog and collect debug information/ logs. In a scenario like in multi GPU connected platform, if one of GPU goes hung and user would like to diagnosis for hung or get gpu dump logs will help to analysis purpose.
This might implement at spec 2.0 level and Sysman has to completely de-link from L0 core and this API should work without zeInit() as zeInit has limitation of require device handle and which may not be possible in case of GPU hung. Sysman use PCI BDF info and access to sysfs directly ang get the detailed log in the new API (which could be global API)
As L0 Syman is admin and it should have flexible to provide an API for end user to give GPU dump/crashlog and collect debug information/ logs. In a scenario like in multi GPU connected platform, if one of GPU goes hung and user would like to diagnosis for hung or get gpu dump logs will help to analysis purpose.
This might implement at spec 2.0 level and Sysman has to completely de-link from L0 core and this API should work without zeInit() as zeInit has limitation of require device handle and which may not be possible in case of GPU hung. Sysman use PCI BDF info and access to sysfs directly ang get the detailed log in the new API (which could be global API)