[PATCH i-g-t 0/3] Drop device scan cache logic

["Zbigniew Kempczyński" <zbigniew.kempczynski@intel.com>]

In this series I haven't added limiting fetching attributes - after
reading udevadm-info.c I'm not sure it makes sense to limit it. Udevadm
in general does same things current igt_device_scan does, especially
it also skips some attributes. Additionally it skips attributes which
are binaries but I've some resistence to do this similar as strlen()
on blob is imo risky and it may result in segfault. I may send attrs
limiting in separate patch after this series will be reviewed (and
hopefully merged).

Series addresses main concern - stale device cache. Now rescans are
unconditional what prevents from opening old device after operations
on module.

Test added which checks device scanning in multithreading code likely
is not needed for merge. Adding to BAT should reveal early potential
problems with the series.

Cc: Lucas De Marchi <lucas.demarchi@intel.com>
Cc: Kamil Konieczny <kamil.konieczny@linux.intel.com>

Zbigniew Kempczyński (3):
  lib/igt_device_scan: drop device scan cache logic
  lib/igt_device_scan: change device list variable visibility
  tests/xe_intel_bb: check multithreading rescans won't segfault

 benchmarks/gem_wsim.c                    |  2 +-
 lib/drmtest.c                            |  2 +-
 lib/igt_device_scan.c                    | 20 +++-------
 lib/igt_device_scan.h                    |  2 +-
 lib/igt_multigpu.c                       |  2 +-
 tests/core_hotunplug.c                   |  2 +-
 tests/device_reset.c                     |  2 +-
 tests/intel-ci/xe-fast-feedback.testlist |  1 +
 tests/intel/i915_suspend.c               |  4 +-
 tests/intel/xe_intel_bb.c                | 47 ++++++++++++++++++++++++
 tools/intel_gpu_top.c                    |  2 +-
 tools/intel_pm_rpm.c                     |  2 +-
 tools/lsgpu.c                            |  2 +-
 13 files changed, 64 insertions(+), 26 deletions(-)

-- 
2.34.1