Use androidboot.partition_map as a fallback for matching partition names in the preinit finding.
#8293
Conversation
|
Doesn't
|
|
a script replicates magiskinit SecondStageInit collect_devices parse_device() {
local device_name="$1"
local devpath=$(readlink -f "/sys/dev/block/$device_name")
local uevent_file="/sys/dev/block/$device_name/uevent"
local dm_name_file="/sys/dev/block/$device_name/dm/name"
while IFS='=' read -r key value; do
case "$key" in
MAJOR)
local major="$value"
;;
MINOR)
local minor="$value"
;;
DEVNAME)
local devname="$value"
;;
PARTNAME)
local partname="$value"
;;
esac
done < "$uevent_file"
local dmname=$(cat "$dm_name_file" 2>/dev/null) || dm_name=""
printf "%-18s\t%-18s\t%-18s\t%s\n" "$partname" "$dmname" "$devname" "$devpath"
}
printf "%-18s\t%-18s\t%-18s\t%s\n" "partname" "dmname" "devname" "devpath"
for device in /sys/dev/block/*; do
parse_device "$(basename "$device")"
doneoutput:
kernel cmdline fragment |
|
send mountinfo? |
|
also send /proc/self/mounts? |
|
Do you think maybe the fix should be a magisk |
|
hmm, both of them show |
|
|
|
No, we have no permission to call |
|
Back to the pr state. I tested the change on the production one. turns out it runs the same version of android as the dev one. The patched system boots and the log looks good, but I can't get adb to work to check if magisk is working. |
chsbuffer
force-pushed
the
partition-map-support
branch
3 times, most recently
from
699813b
to
eb9288a
Compare
|
partition_map was added since Android 13; bootconfig was added since Android 12, so I think we only need to parse bootconfig. |
Google Play Games passes The partition_map might be considered dynamic, I assume that the bootconfig is hardcoded in the ROM and can't be changed, so Google Play Games Emulator does the cmdline way. |
|
I will check the kernel cmdline and bootconfig later |
The bootconfig document said so:
|
|
Maybe the bootloader is bugging out, I notice the difference in bootconfig size between before and after patching the |
yujincheng08
force-pushed
the
partition-map-support
branch
2 times, most recently
from
f6ded27
to
9317e55
Compare
yujincheng08
force-pushed
the
partition-map-support
branch
from
9317e55
to
9831c2a
Compare
yujincheng08
force-pushed
the
partition-map-support
branch
2 times, most recently
from
283a1e8
to
15474a2
Compare
|
@yujincheng08 can you review and approve once the change is in an acceptable state? |
yujincheng08
force-pushed
the
partition-map-support
branch
from
15474a2
to
8269dcc
Compare
topjohnwu
force-pushed
the
partition-map-support
branch
from
8851d67
to
28be995
Compare
| else if (key == "PARTNAME") | ||
| strcpy(dev->partname, value.data()); | ||
| strscpy(dev->partname, value.data(), sizeof(dev->devname)); |
There was a problem hiding this comment.
Is it intentional that this is using sizeof(dev->devname) instead of sizeof(dev->partname)? Might be irrelevant, I'm not exactly sure how this code is called, I just happened to glance at it the other day.
There was a problem hiding this comment.
It doesn't really matter, partname and devname are both 32.
@yujincheng08 Can you answer whether this is intentional or not?
I'm also worried that strscpy doesn't add zero terminators, maybe we should use strncat(dst, src, buffer_size - 1) or use strncasecmp instead of strcastcmp when comparing.
Fixes Play Games Developer Emulator
magiskinit: Cannot find preinit metadata, abort!.only tested on the Play Games Developer Emulator, need to do some test on the production one, which runs on older version of Android iirc.