blob: 00dc32c0ed75c9963201ad43d9190fc7f25eb6ce [file] [log] [blame]
b.liue9582032025-04-17 19:18:16 +08001#!/bin/sh
2# SPDX-License-Identifier: GPL-2.0-only
3
4pe_ok() {
5 local dev="$1"
6 local path="/sys/bus/pci/devices/$dev/eeh_pe_state"
7
8 # if a driver doesn't support the error handling callbacks then the
9 # device is recovered by removing and re-probing it. This causes the
10 # sysfs directory to disappear so read the PE state once and squash
11 # any potential error messages
12 local eeh_state="$(cat $path 2>/dev/null)"
13 if [ -z "$eeh_state" ]; then
14 return 1;
15 fi
16
17 local fw_state="$(echo $eeh_state | cut -d' ' -f1)"
18 local sw_state="$(echo $eeh_state | cut -d' ' -f2)"
19
20 # If EEH_PE_ISOLATED or EEH_PE_RECOVERING are set then the PE is in an
21 # error state or being recovered. Either way, not ok.
22 if [ "$((sw_state & 0x3))" -ne 0 ] ; then
23 return 1
24 fi
25
26 # A functioning PE should have the EEH_STATE_MMIO_ACTIVE and
27 # EEH_STATE_DMA_ACTIVE flags set. For some goddamn stupid reason
28 # the platform backends set these when the PE is in reset. The
29 # RECOVERING check above should stop any false positives though.
30 if [ "$((fw_state & 0x18))" -ne "$((0x18))" ] ; then
31 return 1
32 fi
33
34 return 0;
35}
36
37eeh_supported() {
38 test -e /proc/powerpc/eeh && \
39 grep -q 'EEH Subsystem is enabled' /proc/powerpc/eeh
40}
41
42eeh_one_dev() {
43 local dev="$1"
44
45 # Using this function from the command line is sometimes useful for
46 # testing so check that the argument is a well-formed sysfs device
47 # name.
48 if ! test -e /sys/bus/pci/devices/$dev/ ; then
49 echo "Error: '$dev' must be a sysfs device name (DDDD:BB:DD.F)"
50 return 1;
51 fi
52
53 # Break it
54 echo $dev >/sys/kernel/debug/powerpc/eeh_dev_break
55
56 # Force an EEH device check. If the kernel has already
57 # noticed the EEH (due to a driver poll or whatever), this
58 # is a no-op.
59 echo $dev >/sys/kernel/debug/powerpc/eeh_dev_check
60
61 # Default to a 60s timeout when waiting for a device to recover. This
62 # is an arbitrary default which can be overridden by setting the
63 # EEH_MAX_WAIT environmental variable when required.
64
65 # The current record holder for longest recovery time is:
66 # "Adaptec Series 8 12G SAS/PCIe 3" at 39 seconds
67 max_wait=${EEH_MAX_WAIT:=60}
68
69 for i in `seq 0 ${max_wait}` ; do
70 if pe_ok $dev ; then
71 break;
72 fi
73 echo "$dev, waited $i/${max_wait}"
74 sleep 1
75 done
76
77 if ! pe_ok $dev ; then
78 echo "$dev, Failed to recover!"
79 return 1;
80 fi
81
82 echo "$dev, Recovered after $i seconds"
83 return 0;
84}
85