blackbird-hostboot/src/usr/intr, branch 07-25-2019

Automatically include config.h

2019-12-06T16:28:47+00:00

Rather than having to remember to include config.h anywhere we reference a CONFIG variable (and usually forgetting), this adds it to the default compiler flags so that it gets included in every source file we build. Change-Id: I53622ab4d46c55d942e98cae6ec03049fd5b3d08 Reviewed-on: http://rchgit01.rchland.ibm.com/gerrit1/87475 Tested-by: Jenkins Server Tested-by: Jenkins OP Build CI Tested-by: Jenkins OP HW Tested-by: FSP CI Jenkins Reviewed-by: Zachary Clark Reviewed-by: Roland Veloz Reviewed-by: Christian R Geddes Reviewed-by: Nicholas E Bofferding

Modify INTRP Initialization Order to route LSI interrupts over PSHIHB

2019-11-08T16:58:07+00:00

- Currently the code sets up routing over PSIHIB and then a second function call overwrites that setting which forces things back on the local wire Change-Id: I7d3b44d48c66f994e8a8758eb75568038492fdaf CQ: SW464174 Reviewed-on: http://rchgit01.rchland.ibm.com/gerrit1/86669 Tested-by: Jenkins Server Tested-by: Jenkins OP Build CI Tested-by: Jenkins OP HW Tested-by: FSP CI Jenkins Reviewed-by: Ilya Smirnov Reviewed-by: Michael Baiocchi Reviewed-by: Daniel M Crowell

INTRP HW Workaround for Clockgating Issue

2019-03-27T16:13:15+00:00

- The issue occurs because the slave procs route their interrupts to the master proc, which forces the LSI Interrupt event over the PSIHB. However, the master is still using the local wire for its LSI Interrupt events. There is a situation where undesirable hw behavior can occur because of this (dropped/missed interrupts) - This workaround will set the ESB Notification Register on the master proc, which will force the lsi interrupts over the PSIHB and not the local wire. Change-Id: If595f62d268b6ac504f714b7e50e452b9f926015 CQ: SW441117 Reviewed-on: http://rchgit01.rchland.ibm.com/gerrit1/74719 Tested-by: Jenkins Server Reviewed-by: Christian R. Geddes Tested-by: Jenkins OP Build CI Tested-by: Jenkins OP HW Tested-by: FSP CI Jenkins Reviewed-by: Daniel M. Crowell

Clear INT_CQ related firs after reseting INTRRP logic in HB

2018-11-01T15:09:43+00:00

With multi-node systems during the MPIPL steps up until the re-init of the intrrp the nodes will be running completely independently. Because of this we have seen errors where node C is still sending hypervisor traffic to node A while node A has already torn down its interrupt logic. These errors are irrelevant as we reset the interrupt logic anyways. CQ: SW448121 Change-Id: I910ab80da662b6ad4fb4fbc680dbf2957d2736da Reviewed-on: http://rchgit01.rchland.ibm.com/gerrit1/68182 Reviewed-by: Daniel M. Crowell Reviewed-by: Dean Sanner Tested-by: Jenkins Server Tested-by: Jenkins OP Build CI Tested-by: FSP CI Jenkins Reviewed-by: Zane C. Shelley Tested-by: Jenkins OP HW Reviewed-by: William G. Hoffa

Remove printTimaInfo function until we figure out how to avoid issues

2018-10-11T13:30:34+00:00

We are getting data storage exceptions when we try to read out of the Thread Interrupt Management Area. We thought this was tested and was working but if we go down this path we get a data storage exception. Remove this code until we get it working. Change-Id: Ic8b0d72d05eeac3037643fb017e8327ddb083711 CQ: SW447225 Reviewed-on: http://rchgit01.rchland.ibm.com/gerrit1/67260 Tested-by: Jenkins Server Tested-by: Jenkins OP Build CI Tested-by: Jenkins OP HW Tested-by: FSP CI Jenkins Reviewed-by: Matt Derksen Reviewed-by: Ilya Smirnov Reviewed-by: Daniel M. Crowell

intr: Use the correct trace handle in intrrp TRACDCOMP calls

2018-09-25T17:34:13+00:00

Change-Id: If352854176ada680856bba0ea6cb5df92e8ed8f4 Signed-off-by: Andrew Jeffery Reviewed-on: http://rchgit01.rchland.ibm.com/gerrit1/66569 Tested-by: Jenkins Server Tested-by: Jenkins OP Build CI Reviewed-by: Daniel M. Crowell

Support HB running in SMF

2018-09-24T17:17:16+00:00

Support SMF for P9N/P9C. Lots of minor tweaks to make this work, but the biggest is to run userspace in problem state This is needed because for SMF Hostboot will need to run in S=1, HV=0,PR=1 (and kernel in S=1, HV=1, PR=0) This commit makes P9 HB userpsace run in HV=0 PR=1 and kernel in HV=1, PR=0. Change-Id: Ia4771df5e8858c6b7ae54b0746e62b283afb4bc4 RTC: 197243 Reviewed-on: http://rchgit01.rchland.ibm.com/gerrit1/50530 Tested-by: Jenkins Server Tested-by: Jenkins OP Build CI Tested-by: Jenkins OP HW Tested-by: FSP CI Jenkins Reviewed-by: Dean Sanner Reviewed-by: Nicholas E. Bofferding Reviewed-by: Christian R. Geddes Reviewed-by: Daniel M. Crowell

Don't Process Interrupts During Shutdown

2018-09-19T16:53:31+00:00

An issue has come up on OpenPOWER where we would checkstop during istep 10.2 or 16.1 when completing the interrupt processing due to the code attempting to access a memory area that got locked during the shutdown sequence. This change no-ops the complete interrupt routine if a shutdown has been requested to remedy this issue. Change-Id: Ie444149884c6a42a0010593fb94b91298b4bc798 Reviewed-on: http://rchgit01.rchland.ibm.com/gerrit1/66049 Tested-by: Jenkins Server Tested-by: Jenkins OP Build CI Tested-by: Jenkins OP HW Tested-by: FSP CI Jenkins Reviewed-by: Nicholas E. Bofferding Reviewed-by: Christian R. Geddes Reviewed-by: Michael Baiocchi Reviewed-by: Daniel M. Crowell

Additional tracing for shutdown events

2018-09-12T15:31:05+00:00

Hit an issue where a shutdown was called but all that is traced is a msgQ pointer. These traces should help associate the msgQ pointer to an actual task being shutdown. Change-Id: Ic22fdde1a146eec90f26d557e8af1ddec3c85c4e CQ:SW443508 Reviewed-on: http://rchgit01.rchland.ibm.com/gerrit1/65464 Reviewed-by: Christian R. Geddes Tested-by: Jenkins Server Tested-by: Jenkins OP Build CI Tested-by: Jenkins OP HW Tested-by: FSP CI Jenkins Reviewed-by: Corey V. Swenson Reviewed-by: Daniel M. Crowell

Only unmask source on proc targ passed to unmask function in intrrp

2018-08-01T21:13:13+00:00

There was a bug in the code where if an interrupt was resolved for a source the code to clean up the pending interrupt in the intrrp will unmask the source. This was actually unmasking the source on all processors. This causes odd behavior if there was an outstanding int on that source on one of the other processors. Essentially the intrrp would think that a new interrupt came in but we will already be handling it. The logic breaks down here and we end up getting in a locked state. Fixing this caused issues with Hostboot management of the maskList which kept track of which sources were masked. Instead of managing this list of masked sources hostboot will instead keep track of sources which are not registered to any msg Qs. When new processors are added to the interrupt resource provider after fabric is up, all LSI sources who are not registered to a msg Q will be masked for the new processor. CQ: SW419101 Change-Id: I86f5bc2a748383e18b1853d9bf9480f265c214fd Reviewed-on: http://rchgit01.rchland.ibm.com/gerrit1/63158 Tested-by: Jenkins Server Tested-by: Jenkins OP Build CI Tested-by: Jenkins OP HW Tested-by: FSP CI Jenkins Reviewed-by: Dean Sanner Reviewed-by: Daniel M. Crowell