summaryrefslogtreecommitdiffstats
path: root/src/kernel/doorbell.C
Commit message (Collapse)AuthorAgeFilesLines
* Add sync to doorbell function to avoid weak consistency bugDan Crowell2019-01-111-1/+9
| | | | | | | | | | | | | | | | | | | | | | | | | Due to the way POWER works, there is no guarantee that a memory location written by 1 thread is actually visible to another thread unless you explicitly assert a sync of some kind. In the code that runs to wake up new cores, there is a path where memory is written by 1 thread (usually on the master processor) but is immediately consumed by a new thread that just woke up. It is possible for the new thread to consume the memory and not see the contents that were written by the other thread. The fix is to add a sync command before we send the doorbell msgsnd operation. This will ensure the memory contents are visible to the thread as soon as it wakes up. Change-Id: I8a1483dd7bbda5af064ba6d004dc9e0a3a61ce78 CQ: SW453195 Reviewed-on: http://rchgit01.rchland.ibm.com/gerrit1/70311 Reviewed-by: Dean Sanner <dsanner@us.ibm.com> Reviewed-by: Nicholas E. Bofferding <bofferdn@us.ibm.com> Tested-by: Jenkins Server <pfd-jenkins+hostboot@us.ibm.com> Tested-by: Jenkins OP Build CI <op-jenkins+hostboot@us.ibm.com> Tested-by: Jenkins OP HW <op-hw-jenkins+hostboot@us.ibm.com> Reviewed-by: Daniel M. Crowell <dcrowell@us.ibm.com>
* Restore Timebase on Master Core Threads 1-3 after Sleep/WinkleBill Hoffa2018-07-241-0/+14
| | | | | | | | | | | | | | Change-Id: I329dd64345f2474cb0dad628ccc2244d85be86c2 CQ: SW429364 Reviewed-on: http://rchgit01.rchland.ibm.com/gerrit1/63147 Reviewed-by: ILYA SMIRNOV <ismirno@us.ibm.com> Tested-by: Jenkins Server <pfd-jenkins+hostboot@us.ibm.com> Tested-by: Jenkins OP Build CI <op-jenkins+hostboot@us.ibm.com> Reviewed-by: Prachi Gupta <pragupta@us.ibm.com> Reviewed-by: Christian R. Geddes <crgeddes@us.ibm.com> Tested-by: Jenkins OP HW <op-hw-jenkins+hostboot@us.ibm.com> Tested-by: FSP CI Jenkins <fsp-CI-jenkins+hostboot@us.ibm.com> Reviewed-by: Daniel M. Crowell <dcrowell@us.ibm.com>
* Updates to make multinode IPC workDean Sanner2018-02-131-7/+1
| | | | | | | | | | | | | | | | | | -Fixed node/group id calculations -Reduced dbell printk to prevent printk overflow -Fixed architectual hole in how internode IPC works Workitems won't work, instead just always check for IPC on any doorbells to master thread -Changed PIR tracing to print out in hex Change-Id: I25eb7f87fd812a90f98a7724b1ac1100f764fe7b Reviewed-on: http://ralgit01.raleigh.ibm.com/gerrit1/53187 Tested-by: Jenkins Server <pfd-jenkins+hostboot@us.ibm.com> Reviewed-by: Prachi Gupta <pragupta@us.ibm.com> Reviewed-by: William G. Hoffa <wghoffa@us.ibm.com> Tested-by: Jenkins OP Build CI <op-jenkins+hostboot@us.ibm.com> Tested-by: FSP CI Jenkins <fsp-CI-jenkins+hostboot@us.ibm.com> Reviewed-by: Dean Sanner <dsanner@us.ibm.com>
* Multi-Drawer (IPC) Interrupt/Messaging SupportBill Hoffa2018-01-251-9/+6
| | | | | | | | | | | | | | | | | | | | | | | | | | - Use doorbells instead of IPIs (no IPI support using LSI interupts in the XIVE intr architecture) - New message type from kernel to userspace so the kernel can notify the HB userspace Interrupt Resource Provider (INTRP) that an IPC message was sent to the particular HB instance (in P8 this happened automatically as that was part of the IPI architecture). - Re-enable testcase that validates that an IPC message can be successfully sent. Change-Id: Ic846f8dca45217205ed61d8381a573e995cb16f2 RTC: 150861 Reviewed-on: http://ralgit01.raleigh.ibm.com/gerrit1/52004 Tested-by: Jenkins Server <pfd-jenkins+hostboot@us.ibm.com> Tested-by: Jenkins OP Build CI <op-jenkins+hostboot@us.ibm.com> Tested-by: Jenkins OP HW <op-hw-jenkins+hostboot@us.ibm.com> Reviewed-by: Christian R. Geddes <crgeddes@us.ibm.com> Tested-by: FSP CI Jenkins <fsp-CI-jenkins+hostboot@us.ibm.com> Reviewed-by: Prachi Gupta <pragupta@us.ibm.com> Reviewed-by: Daniel M. Crowell <dcrowell@us.ibm.com>
* Doorbell Interrupt Base Support for Core/Thread WakeupBill Hoffa2016-05-061-2/+46
| | | | | | | | | | | | | | | | | | | | | | | | | This change includes: - Implementation of a generic KernelWorkItem Class - Kernel functionality for doorbell send to specific PIRs - Kernel changes to send core/thread Wakeup doorbells using doorbell_send() + placing KernelWorkItems on a cpu stack obj to be executed during doorbell wakeup - Kernel Interrupt Message handler changes to send wakeup msgs - Interrupt Resource Provider (INTRRP) Changes to handle wakeup msgs and monitor for timeouts - Changes to the IPL flow to invoke proper Core/Thread Wakeup - A basic outline (commented out) for how IPC messages can be implemented in the future Change-Id: I547fb8719bac657def561565ae11ab18cde72096 CMVC-Prereq: 992722 RTC:137564 Reviewed-on: http://ralgit01.raleigh.ibm.com/gerrit1/22815 Tested-by: Jenkins Server Tested-by: FSP CI Jenkins Reviewed-by: Andrew J. Geissler <andrewg@us.ibm.com> Reviewed-by: Prachi Gupta <pragupta@us.ibm.com> Reviewed-by: Daniel M. Crowell <dcrowell@us.ibm.com>
* Add doorbell support.Patrick Williams2016-04-041-0/+51
Change-Id: I4c5ef96c4793f6da26d54d1d61a51f6395e6b34b Backport: master-p8 Reviewed-on: http://ralgit01.raleigh.ibm.com/gerrit1/930 Tested-by: Jenkins Server Reviewed-by: Daniel M. Crowell <dcrowell@us.ibm.com> Reviewed-by: William G. Hoffa <wghoffa@us.ibm.com> Tested-by: FSP CI Jenkins Reviewed-by: Matthew A. Ploetz <maploetz@us.ibm.com>
OpenPOWER on IntegriCloud