Stable 1.8

test: update persist count for ppc (drain after ulog processing)

test: fix fail on return pmempool_feature

test: Fix obj_persist_count/TEST2 for ppc64le

Signed-off-by: Lucas A. M. Magalhaes <lamm@linux.ibm.com>

Ref: #4962

merge stable-1.8 into stable-1.9

pmem: Fix ppc64le build

Typo broke ppc64le build.
Fix commit 28050429

Signed-off-by: Lucas A. M. Magalhaes <lamm@linux.ibm.com>

Disable SDS if is not supported

Merge Stable 1.7 in to Stable 1.8

obj: fix pmreorder and two existing failure atomicity bugs

Without this option, the consistency checking function
can mask issues.

This prevented pmreorder tests from failling in certain
scenarios.

This missing drain might have lead to failure atomicity
bug if the store that invalidated the ulog was reordered
by the CPU with any of the log entries and a crash
happened.

This is a regression introduced in 5e137e0b. That patch accidentally
modified the order in which huge chunk headers are updated, leading
to possible issue with failure atomicity where the chunk headers could
be inconsistent if a crash happened in between two header updates.

common: fix cirrus CI builds

Signed-off-by: Lucas A. M. Magalhaes <lamm@linux.ibm.com>

Signed-off-by: Lucas A. M. Magalhaes <lamm@linux.ibm.com>

Signed-off-by: Lucas A. M. Magalhaes <lamm@linux.ibm.com>

pmempool SDS checks where not a automaticaly skipping if SDS is not
supported.

Signed-off-by: Lucas A. M. Magalhaes <lamm@linux.ibm.com>

Signed-off-by: Lucas A. M. Magalhaes <lamm@linux.ibm.com>

Signed-off-by: Lucas A. M. Magalhaes <lamm@linux.ibm.com>

Signed-off-by: Lucas A. M. Magalhaes <lamm@linux.ibm.com>

Signed-off-by: Lucas A. M. Magalhaes <lamm@linux.ibm.com>

Some devices have no support for SDS and will not be able to create a
pool unless build without SDS_ENABLE or forcing disable SDS during
creation.

This commit checks if the device has SDS support on ndctl before
durint poolset creation and disables it if don't.

Signed-off-by: Lucas A. M. Magalhaes <lamm@linux.ibm.com>

common: fix LIBFABRIC flags

If BUILD_RPMEM is set, LIBFABRIC_* are not set up by common.inc and
 cause rpmemd to fail building

Signed-off-by: Nicolas Morey-Chaisemartin <nmoreychaisemartin@suse.com>

test: add support for ndctl list v63

test: disable bytecode on module import (py)

common: fix GIT_VERSION string