Database Info

Unable to mount loopback

I created a zeroed file with dd and attempted to mount it with:
mount -o loop /sdcard/test.img /data/test
and
losetup /dev/block/loop0 /sdcard/test.img
However, no loopback devices are displayed when I invoke mount. Can anyone clue me in? Feel like i'm missing something basic. By the way, I am running these commands from Saurik's busybox with the losetup /dev/block/loop fix.
p.s. I'd ask on the G1 hackers mailing list, but it's gone a bit bonkers in the past few days.

I'm not sure, but doesn't the image file you're trying to mount have to contain an actual file system?

Yea, I subsequently created a filesystem for it with mke2fs (which is mysteriously missing from busybox for android?). Still no go. Same when I tried to mount other partition-in-a-files like the module.cramfs or recovery.img
If I try to losetup the loop device then mount that, it returns error:
mount: mounting /dev/block/loop0 on /data/test failed: No such device
If I try to mount directly with -o loop, it returns:
mount: cannot setup loop device: No such file or directory
And yet Saurik seems to have been able to loop mount fine. Perhaps it is because I am testing on the qemu instead of hardware?

That might be it. You might try busybox's mount also.
But yeah, It definitely does work on the phone at least.

I've been trying to figure out how to load debian on the emulator, but I can't get the ext2 image mounted. I had to recompile the kernel first so support loadable modules, but even after that, it would not work (can't remember the specific error). Has anyone gotten an ext2 file system mounted on the emulator successfully?
The reason I'm doing this is so I can do native compiles, without actually having to do it on the phone. It has advantages in that I can crank up the clock speed and the RAM.

So I was able to mount loopback just fine on the actual hardware, but i'm having trouble getting it to run from the init.rc file. This is what i'm typing at the shell (it works):
(after insmod ext2.ko)
mount -o loop,noatime,nodiratime,nosuid,nodev -t ext2 /sdcard/loopy/app.img /data/app
It works fine and i'm able to install and run apks from the sdcard (/dev/block/loop2 on /data/app type ext2). Of course this is no good unless it can be mounted before the launcher starts (or it won't see any of our apks). The syntax for init mounting is different from sh, so I tried this (this is in /data/init.rc btw):
on userdata
#mount loopback file from sdcard
insmod /system/modules/2.6.25-01843-gfea26b0/kernel/fs/ext2/ext2.ko
mount ext2 /sdcard/loopy/app.img /data/app noatime nodiratime nosuid nodev loop
I know the script is being run since the ext2.ko module is getting installed. But the mount command isn't being executed. Can anyone shine a light on this? I'm not very familiar with the init daemon.

jashsu said:
So I was able to mount loopback just fine on the actual hardware, but i'm having trouble getting it to run from the init.rc file. This is what i'm typing at the shell (it works):
(after insmod ext2.ko)
mount -o loop,noatime,nodiratime,nosuid,nodev -t ext2 /sdcard/loopy/app.img /data/app
It works fine and i'm able to install and run apks from the sdcard (/dev/block/loop2 on /data/app type ext2). Of course this is no good unless it can be mounted before the launcher starts (or it won't see any of our apks). The syntax for init mounting is different from sh, so I tried this (this is in /data/init.rc btw):
on userdata
#mount loopback file from sdcard
insmod /system/modules/2.6.25-01843-gfea26b0/kernel/fs/ext2/ext2.ko
mount ext2 /sdcard/loopy/app.img /data/app noatime nodiratime nosuid nodev loop
I know the script is being run since the ext2.ko module is getting installed. But the mount command isn't being executed. Can anyone shine a light on this? I'm not very familiar with the init daemon.
Click to expand...
Click to collapse
Try
mount ext2 [email protected]/sdcard/loop/app.img /data/app noatime nodiratime nosuid nodev

JesusFreke said:
Try
mount ext2 [email protected]/sdcard/loop/app.img /data/app noatime nodiratime nosuid nodev
Click to expand...
Click to collapse
Tried that one earlier too. No love.

jashsu said:
Tried that one earlier too. No love.
Click to expand...
Click to collapse
I betcha the sdcard hasn't been mounted yet.
Is there anything in dmesg that looks relevant? Or maybe logcat?

Try temporarily moving the image file to /data, and see if you can get it to mount. That should verify that it has something to do with it being on the sdcard

JesusFreke said:
I betcha the sdcard hasn't been mounted yet.
Is there anything in dmesg that looks relevant? Or maybe logcat?
Click to expand...
Click to collapse
Agh! You're right. The mount daemon runs at line 192, in the services section. Long after the /data/init.rc is parsed.
service mountd /system/bin/mountd
socket mountd stream 0660 root mount
Any way around this short of manually mounting the sdcard in the /data/init.rc?
Edit: dmesg bingo:
<3>[ 1.920368] init: unable to mount [email protected]/sdcard/loopy/app.img to /data/app
as ext2
<6>[ 1.921498] yaffs: dev is 32505860 name is "mtdblock4"
<6>[ 1.922016] yaffs: passed flags ""
<4>[ 1.922322] yaffs: Attempting MTD mount on 31.4, "mtdblock4"
<4>[ 1.927204] yaffs: restored from checkpoint
<4>[ 1.927845] yaffs_read_super: isCheckpointed 1
<3>[ 1.960002] init: unable to mount /system/xbin.cramfs to /system/xbin as c
ramfs
<3>[ 1.965221] init: service 'console' requires console
<4>[ 2.161895] mmc_sd_init_card: Delay 250 ms
<4>[ 2.522535] mmc1: host does not support reading read-only switch. assuming
write-enable.
<6>[ 2.523359] mmc1: new SDHC card at address cd4e
On a side note, it looks like the mount action for xbin.cramfs in the main init.rc is failing because it's redundant with an earlier mount in /system/init.rc. I'm guessing this is a vestige of development on cramfs loopback mounting (mentioned here by Saurik).
6) at Jesus Freke's suggestion, setup /system/xbin (and later /system/modules) to mount from a compressed image
Click to expand...
Click to collapse

Yeah, I had left in an extra mount for xbin.cramfs in the base init.rc. I believe that's fixed in v1.3. I'll check again and make sure

So I gather the reason for wanting to mount a loopback device is to be able to create an ext2 fs as a file on the sdcard fat fs instead of creating 2 separate parittions on the sdcard , one for fat and one for ext2?
If this can be done, this would be a much better solution for getting apps to run on the sdcard than the previous hack which requires creating an ext2 partition on the sdcard.

Yeah that was what I was looking into. However as noted above, the I couldn't mount the image on the sdcard because the point at which /data/init.rc gets parsed occurs before the sdcard service starts up. The partition method doesn't have this problem because it creates a node for paritition #2 in the init.rc

jashsu said:
Yeah that was what I was looking into. However as noted above, the I couldn't mount the image on the sdcard because the point at which /data/init.rc gets parsed occurs before the sdcard service starts up. The partition method doesn't have this problem because it creates a node for paritition #2 in the init.rc
Click to expand...
Click to collapse
so u can't mount the sdcard in /data/init.rc?

dwang said:
so u can't mount the sdcard in /data/init.rc?
Click to expand...
Click to collapse
Not sure, I didn't try mounting partition #1. You could give it a shot by using mknod in /data/init.rc like the parted sdcard method. The sdcard service might not like that the card has already been mounted though.
There are two major reasons I abandoned this concept. The first is I didn't want to figure out how to get permission to mount /sdcard (since / is ro). The second is that making the sdcard available as a usb mass storage device makes it exclusively unaccessible from Android. I suppose that might not be so bad if you are only offloading /data/app, as long as you don't try to execute any new apps while your mass storage connection is going. It's non-ideal.

jashsu said:
Not sure, I didn't try mounting partition #1. You could give it a shot by using mknod in /data/init.rc like the parted sdcard method. The sdcard service might not like that the card has already been mounted though.
There are two major reasons I abandoned this concept. The first is I didn't want to figure out how to get permission to mount /sdcard (since / is ro). The second is that making the sdcard available as a usb mass storage device makes it exclusively unaccessible from Android. I suppose that might not be so bad if you are only offloading /data/app, as long as you don't try to execute any new apps while your mass storage connection is going. It's non-ideal.
Click to expand...
Click to collapse
I don't think the 2nd reason is that big of a deal. If google is planning on implementing the file based method for apps on the sdcard, they could just block access to apps when the sdcard is mounted to a pc. Its a lot better than the alternative of having to create a 2nd ext2 partition on the sdcard. Of course google should have really solved this issue before the g1 came out.
Anyways, I don't see /data/init.rc on my jf rc 1.2 g1. I only see /init.rc and /system/init.rc. Is /data/init.rc something that you created just for the purpose of mounting the file.

dwang said:
I don't think the 2nd reason is that big of a deal. If google is planning on implementing the file based method for apps on the sdcard, they could just block access to apps when the sdcard is mounted to a pc. Its a lot better than the alternative of having to create a 2nd ext2 partition on the sdcard. Of course google should have really solved this issue before the g1 came out.
Anyways, I don't see /data/init.rc on my jf rc 1.2 g1. I only see /init.rc and /system/init.rc. Is /data/init.rc something that you created just for the purpose of mounting the file.
Click to expand...
Click to collapse
Yeah it's referenced in /init.rc Easier to use since unlike /system/init.rc, it's not in a ro folder. You could use /system/init.rc too. Either way really.

jashsu said:
Yeah it's referenced in /init.rc Easier to use since unlike /system/init.rc, it's not in a ro folder. You could use /system/init.rc too. Either way really.
Click to expand...
Click to collapse
Ok.. I'll probably just stick with partitioning the sdcard into fat and ext2. Its working well for me and there really isn't a whole lot of advantages having the ext2 as a file on the fat partition.
I have 41 apps installed on the sdcard now (including some huge games like bonzai blast and brain genius deluxe) and I still have 59 megs of free internal storage. I think everybody that has a rooted g1 should put apps to the sdcard.

[INFO] Clarification: reformatting ext partition

I'm posting this because I am seeing lots of confusion in this area.
Most of the guides found on this forum as well as answers given in the q/a section pertaining to this give the following advice for formatting/wiping a ext parition:
Code:
rm -rf /system/sd/*
Although this may fix a lot of problems it does not really reformat the partion.
"rm -rf" just means "remove(delete) --recursively(subfolders as well) --force(don't ask if ok)". This only deletes the files by removing the inode entries from the filesystem table.
To truly reformat and ensure there are no remnants on the partition you need to make a new filesystem on that partition. This will build a new inode table completely instead of just removing entries in the table.
Enter recovery console (filesystem should NOT be mounted when formatted)
Code:
mke2fs /dev/block/mmcblk0p2 #for ext2 partitions
mke2fs -j /dev/block/mmcblk0p2 #for ext3 partitions
"mke2f2" means "make extended2 filesystem"
"-j" means "add a journal"(ext2+journal=ext3)
Hope this helps

Printing this out right now this is a keeper
Thankyou

Thanks buddy...printing now...

Ok, even though I am just begining at Linux I do understand the difference here between the 2 methods (its like the quick format option under windows). My question is, does it matter? Can you explain why the quick method may be detramental or unsafe.

Thanks for the advice dumfuq, but what if you have a ext4 partition?

Total Noob:
I've just been removing the partition and creating a new one.. is that just as effective (obviously the longer route)?
Also, how would you go about reformatting a linux-swap partition? Again, I've just been removing and recreating every time I flash.. I realize this isn't the fastest way.

Thanks a lot. This is very helpful.

Baldyman1966 said:
Ok, even though I am just begining at Linux I do understand the difference here between the 2 methods (its like the quick format option under windows). My question is, does it matter? Can you explain why the quick method may be detramental or unsafe.
Click to expand...
Click to collapse
I don't think just deleting the files is detrimental or unsafe...its just not formatting. For the most part it doesn't matter but if you want a fresh filesystem make a new one.
Thanks for the advice dumfuq, but what if you have a ext4 partition?
Click to expand...
Click to collapse
It should be
Code:
mke2fs -T ext4 /dev/block/mmcblk0p2
But I haven't tried it
I've just been removing the partition and creating a new one.. is that just as effective (obviously the longer route)? Also, how would you go about reformatting a linux-swap partition? Again, I've just been removing and recreating every time I flash.. I realize this isn't the fastest way
Click to expand...
Click to collapse
Yup...just as effective.
For swap
Code:
mkswap /dev/block/mmcblk0p3
But I don't think it is usually necessary to remake the swap after the initial creation (could be wrong on that though and I guess it can't really hurt)

This thread should be listed in the "Informative Links" sticky

simliarly, how would you go about removing the entire ext3 partition, so reclaim that 500mb back into the fat32? i have a htc magic and no longer need apps2sd, so having a ext partition isnt neccesary in my case. i rather use that extra 500mb for media and storage.
is there a special command to do this? i dont want to reformat the entire card since i have a lot of stuff on my fat32. i was hoping there was an easy way. thanks for any advice!

NguyenHuu said:
simliarly, how would you go about removing the entire ext3 partition, so reclaim that 500mb back into the fat32? i have a htc magic and no longer need apps2sd, so having a ext partition isnt neccesary in my case. i rather use that extra 500mb for media and storage.
is there a special command to do this? i dont want to reformat the entire card since i have a lot of stuff on my fat32. i was hoping there was an easy way. thanks for any advice!
Click to expand...
Click to collapse
Not really.
To do this you have to modify the partition table. Personally I would just back up the data, use fdisk (available in recovery console) to delete old and create new partitions, use windows to format the new fat32 partition created with fdisk, and then restore the data.
However since you don't want to reformat, you could try using gparted(or even parted for that matter) to remove the ext3 partition and resize the fat32. Still probably want to backup first though. There are a few tutorials already that walk you through adding an ext3 partition without losing your fat32. It shouldn't be to hard to follow one of them and tweak the commands to remove ext3 and expand the fat32.

thanks for all the advice, i have removed the ext partitions and resize my fat32 successful without data lost. i backed up just in case too. thanks.

[DEV] Someone ever tried an "init.d/rc0.d" script on Android ?

While messing around with ext2 and data2ext i noticed that android doesn't umount the sd-ext partition (Vold process unmounts the fat partition just fine as it can be seen with logcat). Also cyanogen states the same in this posting (Im wasn't sure on this but i don't doubt him EDIT: i doubt him now ... ).
I wonder now why the recommended solution for file corruption issues is using ext3/4 (wich slows things down due to the journal and reduces the lifetime of flashdrives due to limited r/w cycles) instead of just make a shutdown script wich unmounts the ext partition? (About lifetime: I think 10.000-100.000 r/w cycles is a rather high value and lifetime is only a theoretical and not a practical issue - but im not sure on this)
Since i'm not really expirenced in scripting on either linux or android i wanted to ask if a script in "init.d/rc0.d" (+"init.d/rc6.d" for reboot) would be possible? I could write one but i don't know if this would be executed and i actually don't know how i could check if it works (i think logcat would be shutdown before this is done).
Edit: At least if rc0.d scripts are run BEFORE logcat stops it doesn't work. Logcat log is empty if i "grep test" an "echo "test"" script.
Need help. Someone ...

melethron said:
While messing around with ext2 and data2ext i noticed that android doesn't umount the sd-ext partition (Vold process unmounts the fat partition just fine as it can be seen with logcat). Also cyanogen states the same in this posting (Im wasn't sure on this but i don't doubt him ).
I wonder now why the recommended solution for file corruption issues is using ext3/4 (wich slows things down due to the journal and reduces the lifetime of flashdrives due to limited r/w cycles) instead of just make a shutdown script wich unmounts the ext partition? (About lifetime: I think 10.000-100.000 r/w cycles is a rather high value and lifetime is only a theoretical and not a practical issue - but im not sure on this)
Since i'm not really expirenced in scripting on either linux or android i wanted to ask if a script in "init.d/rc0.d" (+"init.d/rc6.d" for reboot) would be possible? I could write one but i don't know if this would be executed and i actually don't know how i could check if it works (i think logcat would be shutdown before this is done).
Click to expand...
Click to collapse
Does /system/bin/shutdown get called to turn off the phone?
Logcat looks like its the last thing to stop

coutts99 said:
Does /system/bin/shutdown get called to turn off the phone?
Logcat looks like its the last thing to stop
Click to expand...
Click to collapse
Code:
adb logcat | grep /system/bin/shutdown > ~/shutdowntest.log
Empty file
Although the file is present in ROM's so maybe it IS called after logcat stops. But i don't know if it's added to Custom ROMs and if it's also present in RUU's.
If it's not called do you see any possibility to use "Vold" process? Cause it does the unmounting:
Code:
I/Vold ( 145): /mnt/secure/staging/.android_secure sucessfully unmounted
I/Vold ( 145): /mnt/secure/asec sucessfully unmounted
I/Vold ( 145): /mnt/secure/staging sucessfully unmounted
I/Vold ( 145): /mnt/sdcard unmounted sucessfully
D/Vold ( 145): Volume sdcard state changing 5 (Unmounting) -> 1 (Idle-Unmounted)

melethron said:
Code:
adb logcat | grep /system/bin/shutdown > ~/shutdowntest.log
Empty file
Although the file is present in ROM's so maybe it IS called after logcat stops. But i don't know if it's added to Custom ROMs and if it's also present in RUU's.
If it's not called do you see any possibility to use "Vold" process? Cause it does the unmounting:
Code:
I/Vold ( 145): /mnt/secure/staging/.android_secure sucessfully unmounted
I/Vold ( 145): /mnt/secure/asec sucessfully unmounted
I/Vold ( 145): /mnt/secure/staging sucessfully unmounted
I/Vold ( 145): /mnt/sdcard unmounted sucessfully
D/Vold ( 145): Volume sdcard state changing 5 (Unmounting) -> 1 (Idle-Unmounted)
Click to expand...
Click to collapse
hmm, interesting. Maybe have to poke through the source and see what is called to shutdown the phone.
Sent from my HTC Desire using XDA App

coutts99 said:
hmm, interesting. Maybe have to poke through the source and see what is called to shutdown the phone.
Sent from my HTC Desire using XDA App
Click to expand...
Click to collapse
I think this would be a major improvement. Since journaling is actually bad for sd's. Im not sure on this, but the journal is in some specific area on the filesystem. If this is the case I'm not sure if algorithms used to distribute r/w on flash drives will help if the journal gets overwritten (as i understand it so far the r/w is distributet over FREE space but not if something gets overwriten). If this is the case the part of the SD with the journal will be stressed heavy.

I've been playing around with data2sd with ext2 for a few weeks now.
I might be wrong but I think that the system is unmounting cleanly /data (and not /dev/block/mtdblock5) because I always have a clean e2fsck log after rebooting the phone.
So basically if you mount something under the /data mountpoint, it won't be cleanly unmounted and you'll end-up with corrupted FS. That what happens for A2SD because /data is still /dev/block/mtdblock5. But if you mount straight over /data, like I do with DATA2SD, the system unmounts /data on shutdown, witch is now your ext partition (whether your ext partition or a file on ext partition mounted via a loop device) and your FS is not corrupted. In that case you don't need journaling and can stick with ext2.

sibere said:
I've been playing around with data2sd with ext2 for a few weeks now.
I might be wrong but I think that the system in unmounting cleanly /data (and not /dev/block/mtdblock5) because I always have a clean e2fsck log after rebooting the phone.
So basically if you mount something under the /data mountpoint, it won't be cleanly unmounted and you'll end-up with corrupted FS. That what happens for A2SD because /data is still /dev/block/mtdblock5. But if you mount straight over /data, like I do with DATA2SD, the system unmounts /data on shutdown, witch is now your ext partition (whether your ext partition or a file on ext partition mounted via a loop device) and your FS is not corrupted. In that case you don't need journaling and can stick with ext2.
Click to expand...
Click to collapse
Im pretty sure for 3 reasons that it doesnt unmount data:
1. Android unmounts "/sdcard" as it can be seen in logcat while other stuff isn't unmounted. It may be still that the other unmount stuff is not not by the "Vold" process. Sdcard needs to be able to get unmounted when the phone is plugged to usb.
2. I also did quite a lot testing with sd-ext directly mounted to /data and i did get quite a lot corruption. But it is situation depended. If my phone was "idle" i didn't get corruption (1 time in 20 reboots) with no corruption. Then to make absolutly sure that it works i shut it down right after it boots up. There is many stuff running then wich means there is a much higher chance that something is still written while it shutdowns. Doing this i had corruption every 3rd reboot.
3. And this is the most important reason why you didn't get corruption but i did : You use sync as a mount option so you there is no need to unmount it ^^ !!!!
Sync is not an option for me for some reason:
-No cache will slowdown write on small files heavily
-Without cache every other process has to wait if some writing is done (lags).
-Much more wear for the sd since small files aren't cached
Using ext2 with sync isn't an option for me. Either i get ext2 with a proper unmount to work or i'll use Ext4 with an "ordered" journal then (or maybe i give a s*** about data2ext then).

melethron said:
Im pretty sure for 3 reasons that it doesnt unmount data:
1. Android unmounts "/sdcard" as it can be seen in logcat while other stuff isn't unmounted. It may be still that the other unmount stuff is not not by the "Vold" process. Sdcard needs to be able to get unmounted when the phone is plugged to usb.
2. I also did quite a lot testing with sd-ext directly mounted to /data and i did get quite a lot corruption. But it is situation depended. If my phone was "idle" i didn't get corruption (1 time in 20 reboots) with no corruption. Then to make absolutly sure that it works i shut it down right after it boots up. There is many stuff running then wich means there is a much higher chance that something is still written while it shutdowns. Doing this i had corruption every 3rd reboot.
3. And this is the most important reason why you didn't get corruption but i did : You use sync as a mount option so you there is no need to unmount it ^^ !!!!
Sync is not an option for me for some reason:
-No cache will slowdown write on small files heavily
-Without cache every other process has to wait if some writing is done (lags).
-Much more wear for the sd since small files aren't cached
Using ext2 with sync isn't an option for me. Either i get ext2 with a proper unmount to work or i'll use Ext4 with an "ordered" journal then (or maybe i give a s*** about data2ext then).
Click to expand...
Click to collapse
To me, the main purpose with data2Sd was to get more storage with at least the same amount of performances than the original NAND. I found that, on top of the extra space I have now, I also have a very good performance, roughly the same as NAND on small files and a much better one on bigger files. And I'm not thinking about quadrant score here
Talking about the unmount, if I run an e2fsck -n on the ext2 file while it's in use by the system, I have a report of a few errors (like deleted inode with nodtime, or wrong count of free blocks) But if I shut down the phone and turn in on again, the e2fsck I run at boot reports no errors at all. I might be wrong but my guess is that there is a clean umount done at some stage...
What I am going to do is an other test. You know that when an ext2 partition is not umounted properly, a flag is activated and this ext2 partition must be checked, regardless of the number of mounts done before running a check. This flag can be check by running a tune2fs -l on the partition before mounting it. It is reported as well by the system that the partition hasn't been unmounted properly when you try to mount it and that it should be checked .
So If i turn off my phone, restart in recovery mode, run a tune2fs -l on the ext2 file or mount it manually, I should be able to see if the partition has been unmounted by the system or not before the actual reboot.
I'll report the results and let you know if I'm wrong

OK
I've just turned off my phone. rebooted in recovery, mounted the 2nd FAT32 partition and run a tune2fs -l ext2 (that's the name of my ext2 file mounted via a loop device over /data)
Here is the result:
Code:
tune2fs 1.40.8 (13-Mar-2008)
Filesystem volume name: userdata
Last mounted on: /data
Filesystem UUID: 296f98d0-21e8-48b6-82ac-6c94c4edf28d
Filesystem magic number: 0xEF53
Filesystem revision #: 1 (dynamic)
Filesystem features: ext_attr dir_index filetype sparse_super
Filesystem flags: unsigned_directory_hash
Default mount options: (none)
Filesystem state: clean
Errors behavior: Continue
Filesystem OS type: Linux
Inode count: 60800
Block count: 243056
Reserved block count: 0
Free blocks: 168874
Free inodes: 57516
First block: 0
Block size: 4096
Fragment size: 4096
Blocks per group: 32768
Fragments per group: 32768
Inodes per group: 7600
Inode blocks per group: 475
Filesystem created: Tue Dec 7 15:00:51 2010
Last mount time: Thu Dec 9 17:40:16 2010
Last write time: Thu Dec 9 21:16:43 2010
Mount count: 1
Maximum mount count: 21
Last checked: Thu Dec 9 17:40:10 2010
Check interval: 15552000 (6 months)
Next check after: Tue Jun 7 17:40:10 2011
Reserved blocks uid: 0 (user root)
Reserved blocks gid: 0 (group root)
First inode: 11
Inode size: 256
Default directory hash: half_md4
Directory Hash Seed: 8ff9238f-aef7-48a7-85e1-7d2d45b9e809
You can see that it's reporting that the system is clean. If the ext2 partition wasn't cleanly unmounted by the system at shutdown, it would never report as being clean.
Obviously, if I check the state now that, after booting to android, the FS is mounted over /data, it is reported as being unclean.
It's reporting 1 as mount count because I'm forcing an fsck on boot anyway.
This doesn't mean that I will never get errors tho. I even have some errors from time to time when I reboot my linux home server

melethron said:
-No cache will slowdown write on small files heavily
-Without cache every other process has to wait if some writing is done (lags).
-Much more wear for the sd since small files aren't cached
Using ext2 with sync isn't an option for me. Either i get ext2 with a proper unmount to work or i'll use Ext4 with an "ordered" journal then (or maybe i give a s*** about data2ext then).
Click to expand...
Click to collapse
Remember that there is no writeback cache (AFAIK) with YAFFS2
But I should try with sync off and see if the system is still "clean" after a shutdown. If the systems unmounts ext2 properly, then the cache should be flushed by the system before unmounting the filesystem.

sibere said:
OK
I've just turned off my phone. rebooted in recovery, mounted the 2nd FAT32 partition and run a tune2fs -l ext2 (that's the name of my ext2 file mounted via a loop device over /data)
You can see that it's reporting that the system is clean. If the ext2 partition wasn't cleanly unmounted by the system at shutdown, it would never report as being clean.
Obviously, if I check the state now that, after booting to android, the FS is mounted over /data, it is reported as being unclean.
It's reporting 1 as mount count because I'm forcing an fsck on boot anyway.
This doesn't mean that I will never get errors tho. I even have some errors from time to time when I reboot my linux home server
Click to expand...
Click to collapse
You are right about the "clean" flag being set while unmounted. But you forgot one thing. The "clean" flag must be cleared when the partition is mounted. You should check if it as also reported while it IS mounted .... if so mount doesn't write proberly to the superblock. I would do this myself but im on ext4 and afaik it isn't cleared on mount then. For me this far it shows also "clean" when it is mounted.
My guess that mount doesn't write the superblock proberly is also supported by the fact that your and my mount count both shows 1. This could be a coincidince (chance for this is 0,147928994083 %) but could also support my assumption.
So check again while you mounted your system.
EDIT: I completly agree with you that data2ext is about space. Considering speed i also expirienced that reading larger amounts of data from data/data is faster with data2ext. But i also expirienced some slowdown sometimes on ext 4. It was better on ext2 and with that im really fine.
Remount done: Mount count still 1. I bet you will also get clean when you mount your ext.

melethron said:
Since i'm not really expirenced in scripting on either linux or android i wanted to ask if a script in "init.d/rc0.d" (+"init.d/rc6.d" for reboot) would be possible? I could write one but i don't know if this would be executed and i actually don't know how i could check if it works (i think logcat would be shutdown before this is done).
Edit: At least if rc0.d scripts are run BEFORE logcat stops it doesn't work. Logcat log is empty if i "grep test" an "echo "test"" script.
Need help. Someone ...
Click to expand...
Click to collapse
i think that won't work with rc.0 as init.d script execution is initiated in
init.rc only for "before booting", its not as a "normal" linux init system
# Execute files in /etc/init.d before booting
service sysinit /system/bin/logwrapper /system/xbin/busybox run-parts /system/etc/init.d
disabled
oneshot
i wonder if a shutdown.rc could be populated, there is a shutdown.bravo.rc but its empty
for init-handling, in the kernel source its in system/core/init
but im not fit with c/c++ to look into
maybe the init binary would have to be modified by a kernel dev to enable shutdown.rc / also possible that htc has motified it to not to be used.

woti23 said:
i think that won't work with rc.0 as init.d script execution is initiated in
init.rc only for "before booting", its not as a "normal" linux init system
# Execute files in /etc/init.d before booting
service sysinit /system/bin/logwrapper /system/xbin/busybox run-parts /system/etc/init.d
disabled
oneshot
i wonder if a shutdown.rc could be populated, there is a shutdown.bravo.rc but its empty
for init-handling, in the kernel source its in system/core/init
but im not fit with c/c++ to look into
maybe the init binary would have to be modified by a kernel dev to enable shutdown.rc / also possible that htc has motified it to not to be used.
Click to expand...
Click to collapse
This would have been to easy ....
Well maybe some Kernel Dev will look into this.... coutts, sibere, anyone ... can you help ?
Btw: i still wonder why /system/bin/shutdown is present in ROMs if it's not used ....

melethron said:
This would have been to easy ....
Well maybe some Kernel Dev will look into this.... coutts, sibere, anyone ... can you help ?
Btw: i still wonder why /system/bin/shutdown is present in ROMs if it's not used ....
Click to expand...
Click to collapse
this works for the script that should be executed part
/system/bin/logwrapper /system/xbin/busybox run-parts /system/etc/shutdown.d
runs every script which is in /system/etc/shutdown.d
if you create /system/etc/shutdown.d before
and put one or more executable script/s that do something in

melethron said:
You are right about the "clean" flag being set while unmounted. But you forgot one thing. The "clean" flag must be cleared when the partition is mounted. You should check if it as also reported while it IS mounted .... if so mount doesn't write proberly to the superblock. I would do this myself but im on ext4 and afaik it isn't cleared on mount then. For me this far it shows also "clean" when it is mounted.
My guess that mount doesn't write the superblock proberly is also supported by the fact that your and my mount count both shows 1. This could be a coincidince (chance for this is 0,147928994083 %) but could also support my assumption.
So check again while you mounted your system.
EDIT: I completly agree with you that data2ext is about space. Considering speed i also expirienced that reading larger amounts of data from data/data is faster with data2ext. But i also expirienced some slowdown sometimes on ext 4. It was better on ext2 and with that im really fine.
Remount done: Mount count still 1. I bet you will also get clean when you mount your ext.
Click to expand...
Click to collapse
Sorry mate, but, like I wrote in my previous post, it shows unclean when mounted.
Sent from my HTC Desire using XDA App

sibere said:
Sorry mate, but, like I wrote in my previous post, it shows unclean when mounted.
Sent from my HTC Desire using XDA App
Click to expand...
Click to collapse
ROFL! I should learn to read proberly:
Obviously, if I check the state now that, after booting to android, the FS is mounted over /data, it is reported as being unclean.
Click to expand...
Click to collapse
Well thats actually good news thanks for helping me reading. ^^
Well so this isn't a problem. Now its time for ext4 (no loop) without journal mounted to data. About the loopdevice: you might wanna read what ownhere has to say about loop.

melethron said:
If it's not called do you see any possibility to use "Vold" process? Cause it does the unmounting:
Code:
I/Vold ( 145): /mnt/secure/staging/.android_secure sucessfully unmounted
I/Vold ( 145): /mnt/secure/asec sucessfully unmounted
I/Vold ( 145): /mnt/secure/staging sucessfully unmounted
I/Vold ( 145): /mnt/sdcard unmounted sucessfully
D/Vold ( 145): Volume sdcard state changing 5 (Unmounting) -> 1 (Idle-Unmounted)
Click to expand...
Click to collapse
http://android.git.kernel.org/?p=pl...2af4c72427ad069245d0aa572;hb=refs/heads/froyo
Some interesting stuff in this code here

coutts99 said:
http://android.git.kernel.org/?p=pl...2af4c72427ad069245d0aa572;hb=refs/heads/froyo
Some interesting stuff in this code here
Click to expand...
Click to collapse
i'm running this now on btrfs partition
/dev/block/mmcblk0p6** on /system/sd type btrfs (rw,relatime,ssd,noacl)
/dev/block/mmcblk0p6 5.2G 1.3G 3.9G 26% /system/sd
tmpfs 85.8M 24.0M 61.8M 28% /cache
a2sd btrfs *
d2sd btrfs *
dalvik-cache in /data mtd
cache in ram 90M
* without loop mounted img filesystem
** logical partition on extented partition

woti23 said:
i'm running this now on btrfs partition
/dev/block/mmcblk0p6** on /system/sd type btrfs (rw,relatime,ssd,noacl)
/dev/block/mmcblk0p6 5.2G 1.3G 3.9G 26% /system/sd
tmpfs 85.8M 24.0M 61.8M 28% /cache
a2sd btrfs *
d2sd btrfs *
dalvik-cache in /data mtd
cache in ram 90M
* without loop mounted img filesystem
** logical partition on extented partition
Click to expand...
Click to collapse
Running what now?

conventional (not froyo app2sd) a2sd + data2sd both on btrfs without loopmounted filesystem files
/data/dalvik-cache left in nand
/cache tmpfs in RAM
btrfs instead of ext4

[Q] ext4 optimizations

Now, i know there are a *few* places in the dev section that have ext4 optimizations. But these "optimizations" are for the conversion from ext3 to ext4 without having the reformat and everything.
I've been trying (unsuccessfully) to change some of the ext4 mount flags for better speed. My problem is, everything i do in making a CWM flashable with a ext4 optimize sh script errors out with a Status 0. it doesn't even run.
But this time, i though i'd try a few things using adb shell. I got some tune2fs stuff working, so i thought i'd run an fsck on it. There isn't any fsck located on the phone. So i copied over a fsck.ext4, ran it but then got this error (and it also left my rom unuseable, but luckily i had a nandroid):
Code:
***file system was modified*** invalid argument passed to ext2 library while rewriting block inode bitmaps
.
And when doing an ls of /sbin, there isn't anything in there related to ext4, let alone ext3. So i'm wondering, why is our phone supports ext4 partitions, but doesn't actually have anything to support them?
EDIT: running just the tune2fs seems to have worked, just need to test out any differences found first.

[IDEA] Need community help! A safer solution for memory woes on Marshmallow.

When I installed Android 5.0 and later 6.0 ROM on our Xperia L experience was much better than stock. These ROM's are simply fantastic! However, we have memory issues. Yeah we can use apps like Link2SD or foldermount, but that's too much of a hassle. Thet's not really a long-term solution.
The main problem is that /data partition is too small to handle modern app file sizes. It's only 1GB.
And /sdcard partition also exists (which is seperate from data) and represents internal memory - that's 4GB.
On top of that our phone has externalSD slot - /extSD option.
When you install an app it always directly installs to /data. Yeah you can move it. But still 1GB is not enough, somehow I keep getting not enough memory notification.
So the first solution by @Kahana82 was to repartition. That way we could remove 2GB from /sdcard and give it to /data. But the problem is that when we change partitions our phone gets bricked when stock ROM is flashed again.
But I have come across a new idea that could work. Just need community help for that as I am not that experienced.
So instead of repartitioning we could simply mount /SDCard as /data (just rename it!) and mount /data as /SDCard. That way we could increase /data from 1GB to 4GB and I think it is enough for basic app needs. And therefore flashing stock ROM should work as it is just change in kernel or system. Only 1GB internal memory wouldn't be a problem since Android 6.0 allows us to mount externalSD as internal memory.
Now how to do it. I have taken a look at this thread for 2011 Xperia Arc. Looks like ramdisk could be modified to do so as I see that mountpoint is there.
EDIT: OK, I've managed to unpack boot.img and ramdisk. Here's what I get:
fstab.qcom file
Code:
# Android fstab file.
# The filesystem that contains the filesystem checker binary (typically /system) cannot
# specify MF_CHECK, and must come before any filesystems that do specify MF_CHECK
#TODO: Add 'check' as fs_mgr_flags with data partition.
# Currently we dont have e2fsck compiled. So fs check would failed.
#<src> <mnt_point> <type> <mnt_flags and options> <fs_mgr_flags>
/dev/block/platform/msm_sdcc.1/by-name/boot /boot emmc defaults recoveryonly
/dev/block/platform/msm_sdcc.1/by-name/system /system ext4 ro,barrier=1 wait
/dev/block/platform/msm_sdcc.1/by-name/userdata /data ext4 noatime,nosuid,nodev,barrier=1,data=ordered,noauto_da_alloc,journal_async_commit,errors=panic wait,check,encryptable=/persist/footer
/dev/block/platform/msm_sdcc.1/by-name/userdata /data f2fs rw,nosuid,nodev,noatime,nodiratime,inline_xattr wait,formattable,check,encryptable=/persist/footer
/dev/block/platform/msm_sdcc.1/by-name/cache /cache ext4 noatime,nosuid,nodev,barrier=1,data=ordered,noauto_da_alloc,journal_async_commit,errors=panic wait,check
/dev/block/platform/msm_sdcc.1/by-name/cache /cache f2fs rw,nosuid,nodev,noatime,nodiratime,inline_xattr wait,check
/dev/block/platform/msm_sdcc.1/by-name/FOTAKernel /recovery emmc defaults defaults
/devices/platform/msm_sdcc.1/mmc_host/mmc0* auto auto defaults voldmanaged=sdcard0:32,nonremovable,noemulatedsd
/devices/platform/msm_sdcc.3/mmc_host* auto auto nosuid,nodev voldmanaged=sdcard1:auto,encryptable=userdata

I propose another solution.
Format "SD 0" as EXT4.
Then mount during the boot of Android, "SD 0" as DATA partition.
Simply replace "date" with "SD 0" at the time of installation of partitions.
This does not alter the structure of partitions and would give the 4 GB device for apps installation.
I've done tests. It's easy to format using the TWRP "SD 0" as EXT4.
The same process is easy to reverse.
It is safe because it does not change the structure of the partitions. If there is any problem just restore the Stock ROM.
Soon just need help to know which Android configuration files need to change mount "SD 0 in EXT4" like the DATA unit.
The Xperia L pass from 1,57GB to 4 GB !!!
Help, please !!!!

All the MM ROMs for Xperia L should have Adoptable Storage. You could format a partition of your SD Card as Adoptable Storage (using Apps2SD). And you can then move most of the apps to that partition in its entirety. Apps downloaded from Play Store would automatically go to the Adopted Storage.
Only certain apps like WhatsApp and Maps would refuse to move. You could use Link2SD to force move such apps to Internal SD of your device. It will move a big part of it.
On older ROMs, you should be able to setup a partition of your SD as linkable storage, allowing Link2SD or Apps2SD to move the whole app to the partition.

DragonClawsAreSharp said:
All the MM ROMs for Xperia L should have Adoptable Storage. You could format a partition of your SD Card as Adoptable Storage (using Apps2SD). And you can then move most of the apps to that partition in its entirety. Apps downloaded from Play Store would automatically go to the Adopted Storage.
Only certain apps like WhatsApp and Maps would refuse to move. You could use Link2SD to force move such apps to Internal SD of your device. It will move a big part of it.
On older ROMs, you should be able to setup a partition of your SD as linkable storage, allowing Link2SD or Apps2SD to move the whole app to the partition.
Click to expand...
Click to collapse
There are advantages to swap partitions instead of using the SD Card:
- The new partition date would be within the device's NAND memory. So there are the risks involved in using a sdcard always fragile ...
- How the partition is within the NAND memory beyond the security there is the highest performance.
The only downside is that the restriction would increase from 1,57GB to 4GB.
In my case I not use the smartphone for gaming. All of my APPs would fit into 2.5GB.
If I still needed more space, could use "Adoptable Storage".
- Could unmount and remove the SD card when wanted, after all was being used only for personal files and not apps.
Also it would be a great experimental situation of low risk to learn how to configure mount partitions on Android.

ConceptBR said:
There are advantages to swap partitions instead of using the SD Card:
- The new partition date would be within the device's NAND memory. So there are the risks involved in using a sdcard always fragile ...
- How the partition is within the NAND memory beyond the security there is the highest performance.
The only downside is that the restriction would increase from 1,57GB to 4GB.
In my case I not use the smartphone for gaming. All of my APPs would fit into 2.5GB.
If I still needed more space, could use "Adoptable Storage".
- Could unmount and remove the SD card when wanted, after all was being used only for personal files and not apps.
Also it would be a great experimental situation of low risk to learn how to configure mount partitions on Android.
Click to expand...
Click to collapse
Guys, devs are apparently working on unifying partitions. That's the best solution.
It is already working on Xperia T. So it is possible on Xperia L as well.
Let's just wait and see for now.
IMO This is actually top priority for Xperia L lol other that this the device is working just fine.

ConceptBR said:
I propose another solution.
Format "SD 0" as EXT4.
Then mount during the boot of Android, "SD 0" as DATA partition.
Simply replace "date" with "SD 0" at the time of installation of partitions.
This does not alter the structure of partitions and would give the 4 GB device for apps installation.
I've done tests. It's easy to format using the TWRP "SD 0" as EXT4.
The same process is easy to reverse.
It is safe because it does not change the structure of the partitions. If there is any problem just restore the Stock ROM.
Soon just need help to know which Android configuration files need to change mount "SD 0 in EXT4" like the DATA unit.
The Xperia L pass from 1,57GB to 4 GB !!!
Help, please !!!!
Click to expand...
Click to collapse
That's the same thing I was thinking. But yeah, thanks for the info that SDCard can be formatted to ext4. Then I think this can happen:
Code:
# Android fstab file.
# The filesystem that contains the filesystem checker binary (typically /system) cannot
# specify MF_CHECK, and must come before any filesystems that do specify MF_CHECK
#TODO: Add 'check' as fs_mgr_flags with data partition.
# Currently we dont have e2fsck compiled. So fs check would failed.
#<src> <mnt_point> <type> <mnt_flags and options> <fs_mgr_flags>
/dev/block/platform/msm_sdcc.1/by-name/boot /boot emmc defaults recoveryonly
/dev/block/platform/msm_sdcc.1/by-name/system /system ext4 ro,barrier=1 wait
/devices/platform/msm_sdcc.1/mmc_host/mmc0 /data ext4 noatime,nosuid,nodev,barrier=1,data=ordered,noauto_da_alloc,journal_async_commit,errors=panic wait,check,encryptable=/persist/footer
/dev/block/platform/msm_sdcc.1/by-name/cache /cache ext4 noatime,nosuid,nodev,barrier=1,data=ordered,noauto_da_alloc,journal_async_commit,errors=panic wait,check
/dev/block/platform/msm_sdcc.1/by-name/cache /cache f2fs rw,nosuid,nodev,noatime,nodiratime,inline_xattr wait,check
/dev/block/platform/msm_sdcc.1/by-name/FOTAKernel /recovery emmc defaults defaults
/devices/platform/msm_sdcc.3/mmc_host* auto auto nosuid,nodev voldmanaged=sdcard1:auto,encryptable=userdata
Ehh, will try to make boot.img soon. You guys think this code would succeed? Then /sdcard wouldn't exist but let's just see if data would.

I've tried it and it f..ked my phone. I had to reflash stock and go to mm from there ...

cojocar.andrei said:
I've tried it and it f..ked my phone. I had to reflash stock and go to mm from there ...
Click to expand...
Click to collapse
Sorry to hear that...

In my case I need an even simpler solution ...
Using TWRP formatted "SD 0" as EXT4.
I managed to UnMount "/DATA" and "/sdcard".
I would like to reverse the mounts.
Partitioning modifies the structure of partitions.
Mounting is by software changes and safe.
The solution will allow any time to install the ROM Stock since the structure of the partitions will not be changed.
As in TWRP Terminal can ride correct partitions?
if I'm wrong then correct me, please ...
Device Configuration Partitons
sdcard -> /dev/block/mmcblk0p32
Userdata-> /dev/block/mmcblk0p31
CHANGE
sdcard -> /dev/block/mmcblk0p31
Userdata-> /dev/block/mmcblk0p32
------------------------------------------------------------
Is it correct? ADB Shell or Terminal?
unmount /sdcard/
unmount /Userdata/
mount -o rw /dev/block/mmcblk0p32/ /userdata/
mount -o rw /dev/block/mmcblk0p31/ /sdcard/
And reboot smartphone?
/system/init/init.rc?
------------------------------------------------------------
Sorry my bad english...

ConceptBR said:
In my case I need an even simpler solution ...
Using TWRP formatted "SD 0" as EXT4.
I managed to UnMount "/DATA" and "/sdcard".
I would like to reverse the mounts.
Partitioning modifies the structure of partitions.
Mounting is by software changes and safe.
The solution will allow any time to install the ROM Stock since the structure of the partitions will not be changed.
As in TWRP Terminal can ride correct partitions?
if I'm wrong then correct me, please ...
Device Configuration Partitons
sdcard -> /dev/block/mmcblk0p32
Userdata-> /dev/block/mmcblk0p31
CHANGE
sdcard -> /dev/block/mmcblk0p31
Userdata-> /dev/block/mmcblk0p32
------------------------------------------------------------
Is it correct? ADB Shell or Terminal?
unmount /sdcard/
unmount /Userdata/
mount -o rw /dev/block/mmcblk0p32/ /userdata/
mount -o rw /dev/block/mmcblk0p31/ /sdcard/
And reboot smartphone?
/system/init/init.rc?
------------------------------------------------------------
Sorry my bad english...
Click to expand...
Click to collapse
In theory it should work. I've tried it but it didn't come up right for me. I suspect the line in fstab.qcom for the user data that I've replaced (/devices/platform/msm_sdcc.1/mmc_host/mmc0).
I think we should replace
/dev/block/platform/msm_sdcc.1/by-name/userdata
with
/dev/block/platform/msm_sdcc.1/by-name/sdcard
and of course format sdcard as ext4.
I see no reason why it shouldn't work. Probably I'll try it again in a few weeks, when I'll have some spare time ...

So for this to work the fstab.qcom needs to look like this:
#TODO: Add 'check' as fs_mgr_flags with data partition.
# Currently we dont have e2fsck compiled. So fs check would failed.
#<src> <mnt_point> <type> <mnt_flags and options> <fs_mgr_flags>
/dev/block/platform/msm_sdcc.1/by-name/boot /boot emmc defaults recoveryonly
/dev/block/platform/msm_sdcc.1/by-name/system /system ext4 ro,barrier=1 wait
/dev/block/platform/msm_sdcc.1/by-name/sdcard /data ext4 noatime,nosuid,nodev,barrier=1,data=ordered,noauto_da_alloc,journal_async_commit,errors=panic wait,check,encryptable=/persist/footer
/dev/block/platform/msm_sdcc.1/by-name/sdcard /data f2fs rw,nosuid,nodev,noatime,nodiratime,inline_xattr wait,formattable,check,encryptable=/persist/footer
/dev/block/platform/msm_sdcc.1/by-name/cache /cache ext4 noatime,nosuid,nodev,barrier=1,data=ordered,noauto_da_alloc,journal_async_commit,errors=panic wait,check
/dev/block/platform/msm_sdcc.1/by-name/cache /cache f2fs rw,nosuid,nodev,noatime,nodiratime,inline_xattr wait,check
/dev/block/platform/msm_sdcc.1/by-name/FOTAKernel /recovery emmc defaults defaults
/devices/platform/msm_sdcc.1/mmc_host/mmc0* auto auto defaults voldmanaged=sdcard0:31,nonremovable,noemulatedsd
/devices/platform/msm_sdcc.3/mmc_host* auto auto nosuid,nodev voldmanaged=sdcard1:auto,encryptable=userdata

Great job! It's usable?

Option58 said:
Guys, devs are apparently working on unifying partitions. That's the best solution.
It is already working on Xperia T. So it is possible on Xperia L as well.
Let's just wait and see for now.
IMO This is actually top priority for Xperia L lol other that this the device is working just fine.
Click to expand...
Click to collapse
That's great, but what are people who can't use SD cards are supposed to do? Where will they store their data?

stuckbootloader said:
That's great, but what are people who can't use SD cards are supposed to do? Where will they store their data?
Click to expand...
Click to collapse
Want to hear my advice? Install ROM that gives you most data space, I don't know which ROM that is. And use your device as-is until you buy new one. You could also replace your motherboard.
I know this is not xda-spirit like, but I have to say give up messing with your existent motherboard before you hard-brick your device, I destroyed mine. Had to buy new device. Xperia L can be hard-bricked if you mess with partitions.

stuckbootloader said:
That's great, but what are people who can't use SD cards are supposed to do? Where will they store their data?
Click to expand...
Click to collapse
In Xperia T, i think what @Adrian DC use a sdcard emuled for this...
Option58 said:
Want to hear my advice? Install ROM that gives you most data space, I don't know which ROM that is. And use your device as-is until you buy new one. You could also replace your motherboard.
I know this is not xda-spirit like, but I have to say give up messing with your existent motherboard before you hard-brick your device, I destroyed mine. Had to buy new device. Xperia L can be hard-bricked if you mess with partitions.
Click to expand...
Click to collapse
Other Roms? In Galaxy S2 for example, with PIT partitions(other method), all ROMS working today with pits (official cyanogenmod inclusive) . I'm sorry, but, this is the price of testing

Thread closed at OPs request