Align units to a segment unit when adjusting free segment number.
Test: run the smart idle maint service.
Change-Id: I4fd74ac92adc4ae1a0ded4a7df75a690d829eb20
Signed-off-by: Daeho Jeong <daehojeong@google.com>
when the idlemaint run is working, if two or more idlemaint abort concurrency are working.
all idlemaint abort will hold the wakelock, until idlemaint run finished,
but it just nofity one idlemaint abort thread to stop and release the wakelock.
Remaining idlemaint abort threads will hold the wakelock until next idlemaint run to notify them.
Bug: 292383791
Test: see the issue Description
Change-Id: I5312be2e0e92c41483cab26e627a4ae562e5aed8
Since Android platform codespace doesn't support exception handling, we
use strtoll() instead of stoll for direct error handling.
Bug: 274369737
Test: check smart idle maintenace service log
Change-Id: I57c709b1e329228790e0a883edb64dc023135a24
We introduce a new parameter of target dirty segment ratio,
which can be used to set a target dirty / (dirty + free) segments
ratio. For example, if we set this as 80%, GC sleep time will be
calculated to achieve this ratio in a GC period.
Test: check smart idle maint log of StorageManagerService
Signed-off-by: Daeho Jeong <daehojeong@google.com>
Change-Id: I73f2bcf4bdb810164c174bd0d2518b15d577d5d5
Sometimes, waiting for the HAL makes infinite calls to HAL and ending
up with power consuming issues. While tracking the root cause, we will
temporally turn off HAL for storage device GC.
Bug: 235470321
Test: run "sm idle-maint run"
Ignore-AOSP-First: This is a temporal fix for Android TM devices.
Signed-off-by: Daeho Jeong <daehojeong@google.com>
Change-Id: Ieb371b7fdebfe938206a45547bb24dfbf2c2e7be
When it is stuck in runDevGC() for an unexpected reason, now it cannot
prevent the other callers from entering into it again. Fix it.
Bug: 232297944
Test: run "sm idle-maint run" twice & check whether they are overlapped
Signed-off-by: Daeho Jeong <daehojeong@google.com>
Change-Id: I785c8aeebd8fcf58c34d9be9968d99634d0b420a
Merged-In: I785c8aeebd8fcf58c34d9be9968d99634d0b420a
When it is stuck in runDevGC() for an unexpected reason, now it cannot
prevent the other callers from entering into it again. Fix it.
Bug: 232297944
Test: run "sm idle-maint run" twice & check whether they are overlapped
Signed-off-by: Daeho Jeong <daehojeong@google.com>
Change-Id: I785c8aeebd8fcf58c34d9be9968d99634d0b420a
overprovision space and reserved area should be subtracted when we
calculate free segments count in GC for data blocks.
Test: check Vold setGCUrgentPace log
Bug: 202283480
Bug: 181079477
Signed-off-by: Daeho Jeong <daehojeong@google.com>
Change-Id: I7b749588ff794ff0429e17a787d83bcc19af0ec1
When there is not enough dirty segments to reclaim, we don't have to set
gc sleep time.
Test: Run smart idle maint with zero dirty segments
Bug: 202283480
Bug: 181079477
Signed-off-by: Daeho Jeong <daehojeong@google.com>
Change-Id: If6a9965ecb9b1a91bf5d0dd763d5fa0af63348e0
To make gc period changable, added a new gcPeriod parameter to
setGCUrgentPace.
Test: adb shell device_config put storage_native_boot smart_idle_maint_period 10
Bug: 202283480
Bug: 181079477
Signed-off-by: Daeho Jeong <daehojeong@google.com>
Change-Id: I4e44a80ea5b51d9a7cde69d25e129dd0673b271f
We need to reduce F2FS GC intensity for smart idle maintenance service
not to interfere with other I/O activities.
Test: check smart idle maintenance log every hour
Bug: 202283480
Bug: 181079477
Signed-off-by: Daeho Jeong <daehojeong@google.com>
Change-Id: Ideb1641a1539a23828bc8b00d9739c8128a984c0
Added interfaces required by smart idle maintenance service in
StorageManagerService, whose goal is to determine when to trigger
filesystem defragmentation while keeping the best user experience
as long as possible, and avoiding hurting UFS lifetime.
Test: check smart idle maintenance log every hour
Bug: 202283480
Bug: 181079477
Signed-off-by: Daeho Jeong <daehojeong@google.com>
Change-Id: I012cfb9b01e5d21ec71700c3c52ac9c096cd1a90
Acquiring a wakelock can fail if the suspend service is unavailable.
Explicitly check that wakelock was acquired before performing
operations that require the device to stay on.
Bug: b/179229598
Test: Boot test on Pixel 4 device
Change-Id: If30087223e44098801a31d1bfd239ac22e891abe
Prefix all HIDL HAL types with the letter H.
Also make the invocation more generic so that we can use
the AIDL implementation in the future.
Test: TH
Bug: 177470478
Change-Id: Iee0af9727c48c532315203c138a4996a08fcaa34
When the idle maintenance task runs on Cuttlefish, there are several
warnings printed to logcat after the expected messages:
D vold : Starting trim of /data
I vold : Trimmed 0 bytes on /data in 36ms
D vold : Starting trim of /cache
I vold : Trimmed 58662912 bytes on /cache in 0ms
D vold : Starting trim of /metadata
I vold : Trimmed 7725056 bytes on /metadata in 12ms
D vold : Starting trim of none
W vold : Failed to open none: No such file or directory
D vold : Starting trim of /sdcard
W vold : Failed to open /sdcard: Not a directory
D vold : Starting trim of /mnt/vendor/shared
W vold : Trim failed on /mnt/vendor/shared: Inappropriate ioctl for device
This is because vold gathers the filesystems to trim from the fstab, but
it fails to exclude some entries that aren't appropriate to trim:
/dev/block/zram0 none swap defaults zramsize=75%
/tmp /sdcard none defaults,bind recoveryonly
shared /mnt/vendor/shared virtiofs nosuid,nodev,noatime nofail
These should be excluded because they are swap space, a bind mount, and
a virtual filesystem respectively.
Fix addFromFstab() to exclude the above cases.
Afterwards, the messages on Cuttlefish are:
D vold : Starting trim of /data
I vold : Trimmed 0 bytes on /data in 39ms
D vold : Starting trim of /cache
I vold : Trimmed 58662912 bytes on /cache in 0ms
D vold : Starting trim of /metadata
I vold : Trimmed 9822208 bytes on /metadata in 12ms
Change-Id: Idc575106fe6f81c737f684429d58dba4bd5478ad
vold already reads the default fstab into memory when starting up, and
the default fstab isn't allowed to change later. So in IdleMaint.cpp,
just use 'fstab_default' instead of reading it again.
This also has the advantage that fstab entries for "logical partitions"
now get a properly updated blk_device, which is needed in order to start
using blk_device to exclude virtual filesystems in addFromFstab().
Change-Id: Id6457a2b7972d01dde4bca0c5f2da86374d930af
F2FS changes proper configurations along with gc_urgent, so idle-maint doesn't
need to set this redundantly.
Change-Id: I4a71a5d877a3bb9636e2b65132ec806edc56a8fe
Signed-off-by: Jaegeuk Kim <jaegeuk@google.com>
This fixes F2FS GC failure in idle-maint.
Bug: 111953875
Change-Id: I1d10802121d5641cf9ba780fee249affd2cf6ffe
Merged-In: I1d10802121d5641cf9ba780fee249affd2cf6ffe
Test: Tested by hand
Signed-off-by: Jaegeuk Kim <jaegeuk@google.com>
This fixes F2FS GC failure in idle-maint.
07-28 18:25:54.838 603 11187 D vold : idle maintenance started
07-28 18:25:54.846 603 11187 D vold : Start GC on /sys/fs/f2fs/sda21
07-28 18:25:54.847 603 11187 W vold : Set discard gralunarity failed on/sys/fs/f2fs/sda21: No such file or directory
07-28 18:25:54.847 603 11187 W vold : Start GC failed on /sys/fs/f2fs/sda21: No such file or directory
Bug: 111953875
Change-Id: I1d10802121d5641cf9ba780fee249affd2cf6ffe
Signed-off-by: Jaegeuk Kim <jaegeuk@google.com>
This activates device GCs in idle time.
F2FS GC = 7 mins
Trim = 1 mins
Dev GC = 2 mins
Bug: 63264275
Bug: 68721792
Change-Id: I843a742ef192ebe00d77c47a216d4200bc9ccb9d
Signed-off-by: Jaegeuk Kim <jaegeuk@google.com>
runIdleMaint is equivalent with:
1. echo 1 > /sys/fs/f2fs/sdX/gc_urgent
2. wait until /sys/fs/f2fs/sdX/dirty_segments
<= threshold or timeout
3. echo 0 > /sys/fs/f2fs/sdX/gc_urgent
4. fstrim
abortIdleMaint forces the wait loop above to exit and
skips fstrim. However, if fstrim is already running,
abortIdleMaint will just leave it run to completion.
Test: adb shell sm idle-maint [run|abort]
Bug: 67776637
Change-Id: I4adff8d9b6bbd63bce41368cea55dc9e9b117eb6
We've been allocating task objects without freeing them, oops. We
don't really need full classes for these tasks, so move them to
blocking methods, and invoke them from a detached thread.
Remove FIDTRIM support, which isn't meaningful on UFS-based flash
devices. Modern devices require FBE/FDE which gives us better
protection against trimmed data lingering around.
Rename "Trim" to more generic "IdleMaint", since it'll soon extend
to include custom F2FS optimization logic.
Check for shady ".." when validating paths.
Test: cts-tradefed run commandAndExit cts-dev -m CtsOsTestCases -t android.os.storage.cts.StorageManagerTest
Test: cts-tradefed run commandAndExit cts-dev --abi armeabi-v7a -m CtsAppSecurityHostTestCases -t android.appsecurity.cts.AdoptableHostTest
Bug: 67041047
Change-Id: I4fb194c5d5ef13f413c02acedfbaaf79c567582b
