diff options
author | Tyrel Datwyler <tyreld@linux.ibm.com> | 2019-11-10 23:21:29 -0600 |
---|---|---|
committer | Michael Ellerman <mpe@ellerman.id.au> | 2019-11-13 16:57:57 +1100 |
commit | 775fa495af04e0bdb3a00085aaa2d915ed51388f (patch) | |
tree | c41f52015997401fa155ba3e30891923cbb6a30b /tools/perf/scripts/python/export-to-postgresql.py | |
parent | 57409d4fb12c185b2c0689e0496878c8f6bb5b58 (diff) |
powerpc/pseries: Fix drc-info mappings of logical cpus to drc-index
There are a couple subtle errors in the mapping between cpu-ids and a
cpus associated drc-index when using the new ibm,drc-info property.
The first is that while drc-info may have been a supported firmware
feature at boot it is possible we have migrated to a CEC with older
firmware that doesn't support the ibm,drc-info property. In that case
the device tree would have been updated after migration to remove the
ibm,drc-info property and replace it with the older style ibm,drc-*
properties for types, indexes, names, and power-domains. PAPR even
goes as far as dictating that if we advertise support for drc-info
that we are capable of supporting either property type at runtime.
The second is that the first value of the ibm,drc-info property is
the int encoded count of drc-info entries. As such "value" returned
by of_prop_next_u32() is pointing at that count, and not the first
element of the first drc-info entry as is expected by the
of_read_drc_info_cell() helper.
Fix the first by ignoring DRC-INFO firmware feature and instead
testing directly for ibm,drc-info, and then falling back to the
old style ibm,drc-indexes in the case it doesn't exit.
Fix the second by incrementing value to the next element prior to
parsing drc-info entries.
Signed-off-by: Tyrel Datwyler <tyreld@linux.ibm.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/1573449697-5448-3-git-send-email-tyreld@linux.ibm.com
Diffstat (limited to 'tools/perf/scripts/python/export-to-postgresql.py')
0 files changed, 0 insertions, 0 deletions