Chroot environment locks up?

[AeroFengBlade]

Originally Posted by knc1

Well, if you find you have to use a partition (a device) rather than a file...

You can also make a new node inside /dev/ called swap for easier access:

mknod /dev/swap b 7 8 #b=block device, 7=use loop driver, 8=loop device number

Make a separate directory in /mnt/us/ called swap:

mkdir /mnt/us/swap
cd /mnt/us/swap

Make a new swap file:

dd if=/dev/zero of=swapfile bs=1M count=128
mkswap swapfile

Turn the file into a block device:

losetup /dev/swap /mnt/us/swap/swapfile

I made some scripts in /mnt/us/swap/ for easy swapping on/off:
swapon.sh

Spoiler:

mknod /dev/swap b 7 8
losetup /dev/swap /mnt/us/swap/swapfile
swapon /dev/swap

swapoff.sh

Spoiler:

swappoff /dev/swap #You don't really need a script for this, but I'm lazy

Swap on...swap off Sounds like wax on, wax off from the original Karate Kid, if you know what I'm talking about

[qlob]

I liked the old Karate Kid movie better than the new one...

[aditya3098]

No host found for xkid.biz

[twobob]

So yeah. it's down... so I dug around...

dd if=/dev/zero of=debian.ext3 bs=1M count=500 oflag=append conv=notrunc e2fsck -f debian.ext3 resize2fs debian.ext3

is the only fragment I could find from that host.

Which looks a lot like

https://www.mobileread.com/forums/showthread.php?t=96048

here's more of the same:

http://tomashg.com/?p=330
hope that helps


enjoy

[AeroFengBlade]

Luckily I downloaded the image a few days before it went down (like a few months ago). I don't have the original, but if you want my current copy of it, I'm currently uploading it.

[AeroFengBlade]

Ok guys, here it is:

https://dl.dropbox.com/s/mft5c4vgjk8...t3.tar.gz?dl=1

It's on Dropbox since it's too large to attach, and because I don't have a site to host it on.

[twobob]

Originally Posted by AeroFengBlade

Ok guys, here it is:

https://dl.dropbox.com/s/mft5c4vgjk8...t3.tar.gz?dl=1

It's on Dropbox since it's too large to attach, and because I don't have a site to host it on.

http://bit.ly/k5debian
for a ge.tt mirror of that mate.
HTH

[EternalCyclist]

Originally Posted by geekmaster

But I thought we were supposed to avoid using /mnt/base-us/.

Using /mnt/base-us seems to work unless you are accessing userstore from host PC at the same time.

The lock ups might not be caused by loop mounting in general, but by fsp. /etc/fstab includes:

 fsp#/mnt/base-us  /mnt/us       fuse   
rw,max_write=65536,max_readahead=65536,noatime,exec,nosuid,nodev,nonempty

Quick look at fsp:

strings /sbin/fsp

says that it is a "filesystem proxy".

ps  -eo %mem,rss,args | grep fsp

shows

 0.4  1116 fsp /mnt/base-us /mnt/us -o rw,nosuid,nodev,noatime,max_write=65536,max_readahead=65536,nonempty

fsp uses 0.4% of physical memory, 1116 kB. Not too bad. But maybe it changes dynamically.

[twobob]

I use winscp (on linux) to push stuff onto the mounted /mnt/debain/stuff/here directly from my linux mint box.

never had a complaint, mounted or otherwise. Since that works flawlessly seemingly I am sticking with that. HTH (it's mounted via /mnt/base-us IIRC at mount time)

[knc1]

Originally Posted by twobob

I use winscp (on linux) to push stuff onto the mounted /mnt/debain/stuff/here directly from my linux mint box.

never had a complaint, mounted or otherwise. Since that works flawlessly seemingly I am sticking with that. HTH (it's mounted via /mnt/base-us IIRC at mount time)

It has to be.
At least it does if you want the device nodes to work within the chroot environment.

Note the "nodev" attribute on /mnt/us.

[twobob]

Originally Posted by knc1

It has to be.
At least it does if you want the device nodes to work within the chroot environment.

Note the "nodev" attribute on /mnt/us.

ah yeah. I do, now you mention it

[aditya3098]

That clears my cross compiling nightmares!

[twobob]

Originally Posted by AeroFengBlade

You can also make a new node inside /dev/ called swap for easier access:

mknod /dev/swap b 7 8 #b=block device, 7=use loop driver, 8=loop device number

Make a separate directory in /mnt/us/ called swap:

mkdir /mnt/us/swap
cd /mnt/us/swap

Make a new swap file:

dd if=/dev/zero of=swapfile bs=1M count=128
mkswap swapfile

Turn the file into a block device:

losetup /dev/swap /mnt/us/swap/swapfile

I made some scripts in /mnt/us/swap/ for easy swapping on/off:
swapon.sh

Spoiler:

mknod /dev/swap b 7 8
losetup /dev/swap /mnt/us/swap/swapfile
swapon /dev/swap

swapoff.sh

Spoiler:

swappoff /dev/swap #You don't really need a script for this, but I'm lazy

Swap on...swap off Sounds like wax on, wax off from the original Karate Kid, if you know what I'm talking about

I do exactly this save for one thing

losetup /dev/swap /mnt/base-us/swap/swapfile

I run mine via base-us.

Unsure if this has any impact but the other way occasionally locked up.
(possibly the whole Fuse file proxy thing again)

Thought I would share...

[knc1]

Yup - you can't run swap over the anything fuse - it is a deadlock path in the kernel.
I think the same restriction applies to NFS (or, at least, it used to apply).

And there is no reason to turn the swap file into a device - 'swap' can use a file as well as a device.
So just "use it" as a file after you run mkswap on it.

[NullNix]

Originally Posted by knc1

I think the same restriction applies to NFS (or, at least, it used to apply).

Recent kernels can swap over NFS, but it was a very hard problem: a huge patchset brewing for literally years. The solution involves an emergency memory pool that the network subsystem can dip into for packet buffers, so it won't make it any more possible to swap to FUSE. *That* is risky as hell even if you mlockall() your entire FUSE daemon into memory, because stack allocations can still happen...