1 #!/bin/sh 1 #!/bin/sh
2 # SPDX-License-Identifier: GPL-2.0-only <<
3 # 2 #
4 # Copyright 2015, Daniel Axtens, IBM Corporati 3 # Copyright 2015, Daniel Axtens, IBM Corporation
5 # 4 #
>> 5 # This program is free software; you can redistribute it and/or modify
>> 6 # it under the terms of the GNU General Public License as published by
>> 7 # the Free Software Foundation; version 2 of the License.
>> 8 #
>> 9 # This program is distributed in the hope that it will be useful,
>> 10 # but WITHOUT ANY WARRANTY; without even the implied warranty of
>> 11 # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
>> 12 # GNU General Public License for more details.
6 13
7 14
8 # do we have ./getscom, ./putscom? 15 # do we have ./getscom, ./putscom?
9 if [ -x ./getscom ] && [ -x ./putscom ]; then 16 if [ -x ./getscom ] && [ -x ./putscom ]; then
10 GETSCOM=./getscom 17 GETSCOM=./getscom
11 PUTSCOM=./putscom 18 PUTSCOM=./putscom
12 elif which getscom > /dev/null; then 19 elif which getscom > /dev/null; then
13 GETSCOM=$(which getscom) 20 GETSCOM=$(which getscom)
14 PUTSCOM=$(which putscom) 21 PUTSCOM=$(which putscom)
15 else 22 else
16 cat <<EOF 23 cat <<EOF
17 Can't find getscom/putscom in . or \$PATH. 24 Can't find getscom/putscom in . or \$PATH.
18 See https://github.com/open-power/skiboot. 25 See https://github.com/open-power/skiboot.
19 The tool is in external/xscom-utils 26 The tool is in external/xscom-utils
20 EOF 27 EOF
21 exit 1 28 exit 1
22 fi 29 fi
23 30
24 # We will get 8 HMI events per injection 31 # We will get 8 HMI events per injection
25 # todo: deal with things being offline 32 # todo: deal with things being offline
26 expected_hmis=8 33 expected_hmis=8
27 COUNT_HMIS() { 34 COUNT_HMIS() {
28 dmesg | grep -c 'Harmless Hypervisor Maint 35 dmesg | grep -c 'Harmless Hypervisor Maintenance interrupt'
29 } 36 }
30 37
31 # massively expand snooze delay, allowing inje 38 # massively expand snooze delay, allowing injection on all cores
32 ppc64_cpu --smt-snooze-delay=1000000000 39 ppc64_cpu --smt-snooze-delay=1000000000
33 40
34 # when we exit, restore it 41 # when we exit, restore it
35 trap "ppc64_cpu --smt-snooze-delay=100" 0 1 42 trap "ppc64_cpu --smt-snooze-delay=100" 0 1
36 43
37 # for each chip+core combination 44 # for each chip+core combination
38 # todo - less fragile parsing 45 # todo - less fragile parsing
39 grep -E -o 'OCC: Chip [0-9a-f]+ Core [0-9a-f]' !! 46 egrep -o 'OCC: Chip [0-9a-f]+ Core [0-9a-f]' < /sys/firmware/opal/msglog |
40 while read chipcore; do 47 while read chipcore; do
41 chip=$(echo "$chipcore"|awk '{print $3 48 chip=$(echo "$chipcore"|awk '{print $3}')
42 core=$(echo "$chipcore"|awk '{print $5 49 core=$(echo "$chipcore"|awk '{print $5}')
43 fir="0x1${core}013100" 50 fir="0x1${core}013100"
44 51
45 # verify that Core FIR is zero as expe 52 # verify that Core FIR is zero as expected
46 if [ "$($GETSCOM -c 0x${chip} $fir)" ! 53 if [ "$($GETSCOM -c 0x${chip} $fir)" != 0 ]; then
47 echo "FIR was not zero before 54 echo "FIR was not zero before injection for chip $chip, core $core. Aborting!"
48 echo "Result of $GETSCOM -c 0x 55 echo "Result of $GETSCOM -c 0x${chip} $fir:"
49 $GETSCOM -c 0x${chip} $fir 56 $GETSCOM -c 0x${chip} $fir
50 echo "If you get a -5 error, t 57 echo "If you get a -5 error, the core may be in idle state. Try stress-ng."
51 echo "Otherwise, try $PUTSCOM 58 echo "Otherwise, try $PUTSCOM -c 0x${chip} $fir 0"
52 exit 1 59 exit 1
53 fi 60 fi
54 61
55 # keep track of the number of HMIs han 62 # keep track of the number of HMIs handled
56 old_hmis=$(COUNT_HMIS) 63 old_hmis=$(COUNT_HMIS)
57 64
58 # do injection, adding a marker to dme 65 # do injection, adding a marker to dmesg for clarity
59 echo "Injecting HMI on core $core, chi 66 echo "Injecting HMI on core $core, chip $chip" | tee /dev/kmsg
60 # inject a RegFile recoverable error 67 # inject a RegFile recoverable error
61 if ! $PUTSCOM -c 0x${chip} $fir 200000 68 if ! $PUTSCOM -c 0x${chip} $fir 2000000000000000 > /dev/null; then
62 echo "Error injecting. Abortin 69 echo "Error injecting. Aborting!"
63 exit 1 70 exit 1
64 fi 71 fi
65 72
66 # now we want to wait for all the HMIs 73 # now we want to wait for all the HMIs to be processed
67 # we expect one per thread on the core 74 # we expect one per thread on the core
68 i=0; 75 i=0;
69 new_hmis=$(COUNT_HMIS) 76 new_hmis=$(COUNT_HMIS)
70 while [ $new_hmis -lt $((old_hmis + ex 77 while [ $new_hmis -lt $((old_hmis + expected_hmis)) ] && [ $i -lt 12 ]; do
71 echo "Seen $((new_hmis - old_hmis) 78 echo "Seen $((new_hmis - old_hmis)) HMI(s) out of $expected_hmis expected, sleeping"
72 sleep 5; 79 sleep 5;
73 i=$((i + 1)) 80 i=$((i + 1))
74 new_hmis=$(COUNT_HMIS) 81 new_hmis=$(COUNT_HMIS)
75 done 82 done
76 if [ $i = 12 ]; then 83 if [ $i = 12 ]; then
77 echo "Haven't seen expected $expec 84 echo "Haven't seen expected $expected_hmis recoveries after 1 min. Aborting."
78 exit 1 85 exit 1
79 fi 86 fi
80 echo "Processed $expected_hmis events; 87 echo "Processed $expected_hmis events; presumed success. Check dmesg."
81 echo "" 88 echo ""
82 done 89 done
Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.