Age | Commit message (Collapse) | Author | Files | Lines |
|
OBMC_MACHINE_FEATURES is a variable that probably shouldn't even exist
as all it does is provide a level of indirection to the real
MACHINE_FEATURES (in phosphor-base.inc, OBMC_ is added to the base).
There are a few recipes that have developed checks against
OBMC_MACHINE_FEATURES for enabling recipe-level features when they
should be checking the standard MACHINE_FEATURES variable. Clean up
this recipe-level indirection.
Signed-off-by: Patrick Williams <patrick@stwcx.xyz>
Change-Id: Id437b93aa6c1de9a27726eea9b7e2c30e71ab3bf
|
|
Change-Id: I5ca16c6fa7bc616bfa9bc41c4f86d32fd92b0287
Signed-off-by: Adriana Kobylak <anoo@us.ibm.com>
|
|
Service to collect r/w devtree attribute data from the
devtree and save to known location. This is data
collected prior to concurrent software update and applied
to new device tree in the reboot followed by concurrent
software update.
Version bump to enable service back-end support.
tested:
Jul 30 14:27:35 xxx phal-export-devtree[2624]: Software path
interface add signal received
Jul 30 14:27:35 xxx phal-export-devtree[2624]: Successfully
exported devtree attribute data
root@xx:~# ls /var/lib/phal/
exportdevtree
Signed-off-by: Jayanth Othayoth <ojayanth@in.ibm.com>
Change-Id: Ic2462f596a7ffcc716254577dc693d580111b766
|
|
Service to override r/w devtree attribute data to the
newer version of devtree from older version. This is used
in concurrent software update followed BMC reboot.
This service will only run in the following condition:
- Chassis power is on when the BMC comes out of reset
- version bump to enable service back-end support
Signed-off-by: Jayanth Othayoth <ojayanth@in.ibm.com>
Change-Id: I352c60ba254cfcf50a2339224eab9fa418491cc7
|
|
Enable the new BMC function to look at a CFAM register as a last chance
to see if the host is running.
See the following for more detail:
https://github.com/openbmc/docs/blob/master/designs/bmc-reset-with-host-up.md#note-on-custom-mechanism-for-ibm-systems
openpower-proc-control: srcrev bump f8c14bcfd5..3292c0676e
Andrew Geissler (8):
service-files: move to own sub-directory
primary-proc: move function to common utils
bmc-reset: check if host is running
bmc-reset: ensure error is created if host up
bmc-reset: call new host reset check service
bmc-reset: add support to write a cfam
bmc-reset: add support to clear host running register
bmc-reset: call new clear host running status service
Signed-off-by: Andrew Geissler <geissonator@yahoo.com>
Change-Id: If4f7c3a7a58bbe5c7fda09bdc949ed31f5cb1cbf
|
|
Previous code was piggy backing the memory preserving reboot function
off of the phal machine feature. This had issues in that a
witherspoon-tacoma system also supports mpreboot, but does not support
phal.
Signed-off-by: Andrew Geissler <geissonator@yahoo.com>
Change-Id: Ib044f5252843e84be3304bc0df8ffb8aa947526c
|
|
For phal based systems, move the memory preserving reboot function to
the obmc-host-crash target. This target is currently un-used on phal
based systems.
Make this move so that the obmc-host-diagnostic target can be utilized
as a more generic target. Dumps like hostboot, sbe, and hardware can now
utilize the obmc-host-diagnostic target to ensure the user knows when
any type of diagnostic event is occurring in the system.
Note that the obmc-host-diagnostic target will still be started as a
part of obmc-host-crash since that target is what communicates to
outside users that a diagnostic action is occurring.
To ensure we do not regress the dump functionality on our non-pHAL
systems, ensure the obmc-host-quiesce target remains within the
obmc-host-crash target. Our non pHAL systems go straight to quiesce on a
crash target call.
Tested:
- Verified with all commits in this Topic that the witherspoon system
has the same service layout as before and the rainier system
has the new layout (crash used for mpipl)
(From meta-openpower rev: b3f8d51109a607c6efc158198767395c6441b152)
Signed-off-by: Andrew Geissler <geissonator@yahoo.com>
Change-Id: Ia6d5411f84ec6c83a7edb177a1635c931bbda041
Signed-off-by: Andrew Geissler <geissonator@yahoo.com>
|
|
When the obmc-host-quiesce@.target was initially created the idea was to
leave the system in the state it was in. In most cases the host had
crashed.
As OpenBMC has grown, the use cases for the quiesce target have grown.
For example the Quiesce On Hardware Error design:
https://github.com/openbmc/docs/blob/master/designs/fail-boot-on-hw-error.md
When the QuiesceOnHwError setting is enabled, the host will be put in a
Quiesce state to indicate that an error has occurred. In a lot of cases
though the host actually can (and does) continue to boot. This causes a
lot of confusion for the user (host state shows Quiesce but host is
still booting).
This commit adds the op-stop-instructions.service to the quiesce target
to ensure the host is actually stopped when quiesce is entered.
Tested:
- Disabled AutoReboot and turned on QuiesceOnHwError. Verified when an
error which triggers QuiesceOnHwError was injected, the host state went
to Quiesce and the host was stopped via the stop instructions.
(From meta-openpower rev: ff3e548479bdf4c9db4315f45f1599a4f9adc33f)
Signed-off-by: Andrew Geissler <geissonator@yahoo.com>
Change-Id: If5cbdd03fbb3abcad8942e5ebe4e7b347ecfd599
Signed-off-by: Andrew Geissler <geissonator@yahoo.com>
|
|
Changes:
-Adding support for new service
i.e. proc-pre-poweroff@.service, which will be called
during poweroff.
(From meta-openpower rev: 064663fa543eb32bf55f0657a56a3a6cf47bfe0e)
Signed-off-by: Chirag Sharma <chirshar@in.ibm.com>
Change-Id: I066bb91237f838c87de3da8e10b4d5a4fe7636e8
Signed-off-by: Andrew Geissler <geissonator@yahoo.com>
|
|
The POWER CEC system device tree need reinitialize before doing
normal ipl to mark required targets as functional for ipling.
The service file is added in below patch.
https://gerrit.openbmc-project.xyz/c/openbmc/openpower-proc-control/+/36577
(From meta-openpower rev: 2257014de9964f673b4ae2864deffd25e6022cee)
Signed-off-by: Ramesh Iyyar <rameshi1@in.ibm.com>
Change-Id: I499ee3c6c2b4f3caf33609ca6e70112801d5cbae
Signed-off-by: Andrew Geissler <geissonator@yahoo.com>
|
|
The corresponding change in openpower-proc-control has more details:
https://gerrit.openbmc-project.xyz/c/openbmc/openpower-proc-control/+/31265
The summary is that the cfam-reset is being removed from the power
control application and into its own service. This ensure it is run when
it is needed (i.e. warm reboots) and that future systems which may use
different power control or have different cfam reset requirements can
easily configure the function.
Tested:
- Verified chassison behaved as expected on witherspoon
- Verified normal power on/off works as expected on witherspoon
- Verified warm reboot works as expected on witherspoon
- Verified bmc reset with host up still works as expected
(From meta-openpower rev: fd9969e4221c321d2807c86b16e6fbafac8e3d8b)
Signed-off-by: Andrew Geissler <geissonator@yahoo.com>
Change-Id: If6dc939015dc1647773d20e3909ae2d718512a73
Signed-off-by: Andrew Geissler <geissonator@yahoo.com>
|
|
start_host@.service: Update to start_host service to skip it during
a memory preserve reboot
op-proc-control-systemd-links.bb: update to install services
op-proc-control_git.bb: Added mpreboot services
Bump openpower-proc-control SRCREV to required commit
(From meta-openpower rev: 7e0062d06c12e177f6a767a1c21066dc30ecae44)
Signed-off-by: Dhruvaraj Subhashchandran <dhruvaraj@in.ibm.com>
Change-Id: Ib2e3e34f3faa1ad575def9b26ff3e91d5eebf93f
Signed-off-by: Andrew Geissler <geissonator@yahoo.com>
|
|
There have been some situations where things get so messed up that you
can not even stop the instructions on your processor. When this happens,
the systemd target, obmc-host-stop can not complete. This really should
be a wants relationship vs. a requires so that the target can continue
to stop the host even when this service fails.
(From meta-openpower rev: 8dfb06a8656eeff1c0393607a5bf9cdc6320aae5)
Signed-off-by: Andrew Geissler <geissonator@yahoo.com>
Change-Id: I2704fecab789d6ffa8e274c37d0653bf64f3c8b0
Signed-off-by: Andrew Geissler <geissonator@yahoo.com>
|
|
The cfam_override@.service is required in two situations:
- When chassis power is turned on (to ensure proper workarounds
and overrides are in place)
- When the host is warm rebooted (i.e. no chassis power is turned
off/on)
Install this service in both targets and ensure it only runs once per
boot.
Tested:
- Verified service was only run once during a cold boot
- Verified overrides were correctly applied on a warm reboot
(From meta-openpower rev: 59c6d368452c05a184549c0ee844eba63203250f)
Signed-off-by: Andrew Geissler <geissonator@yahoo.com>
Change-Id: Idb667eb7c0b09d7c00e5893443abbcf58058a465
Signed-off-by: Andrew Geissler <geissonator@yahoo.com>
|
|
Install the new stop instructions and cfam reset services required for a
warm reboot into the appropriate target
Includes following in bump:
Andrew Geissler (4):
warm-reboot: ensure sbe start bit is 0
warm reboot: stop instructions support
warm-reboot: add cfam-reset capabilities
warm-reboot: cfam-reset service
(From meta-openpower rev: 84962a9850f8f7b61698134ecf2d016a9b1e33de)
Signed-off-by: Andrew Geissler <geissonator@yahoo.com>
Change-Id: Ia8fe30f160e0078bbdca2bcef322c9e23ea0e9e0
Signed-off-by: Andrew Geissler <geissonator@yahoo.com>
|