hpr_transcripts/hpr1232.txt

Episode: 1232
Title: HPR1232: LiTS 028: extended attributes
Source: https://hub.hackerpublicradio.org/ccdn.php?filename=/eps/hpr1232/hpr1232.mp3
Transcribed: 2025-10-17 22:01:20

---

Welcome to Linux and the Shell Episodes 28, Extended Attributes and the LSATTR and
CHATTR commands.
My name is Dan Washko, I'll be your host today, and as always I'd like to thank Hacker
Public Radio for hosting the website and the audio files, Hacker Public Radio, fantastic
resource for the community, consider contributing to Hacker Public Radio by doing your own
episode on anything that matters to your geeky little heart, or head on over there to listen
some really great content put out every day by geeks across the world.
Let's talk about Extended Attributes and the way we can manage Extended Attributes.
Back in episode 7 we talked about file permissions, their attributes, basic attributes are read
right and execute, and they could be applied by either the owner to the owner to the
owning group or everybody else, it gives you three attributes at three different levels.
While there are some Extended Attributes that do some other stuff, not necessarily permissions
related, but some other information about the file and how to handle stuff with regards
to the file.
So what I'm going to do is I'm going to talk about those attributes and then talk about
how you can see them and manipulate them.
Now not all of them are able to be manipulated by these commands, and I'll cover that along
the way.
This is going to have some pretty detailed information.
So I suggest that if you have not already gone over to the website, you do so after listening
to this to get a fuller understanding and see the links to some more information because
some of these attributes are pretty complicated.
Alright, let's start off.
I believe I have them in alphabetical order, maybe I'll mess up.
First attribute, lower case A, stands for append only.
What that does is sets the file so that it cannot be overwritten or deleted.
You can only append information to it.
Now this can only be set by a super user account, and what is going to end up happening with
this is it's probably going to break the standard way that you might manipulate or interact
with this file.
For instance, you can redirect text to the file so long as you are appending and not overwriting,
but you're not going to be able to open it up in text editor or something like that.
You're going to get a permissions denied.
So that attribute append only does exactly what it says.
It only allows you to append to the file, not write or overwrite it.
Now why doesn't it open in some applications like a text editor?
Because when a text editor goes to write it, it doesn't append to the file, it actually
just overwrites it and it won't let you do that because it is this attribute being set.
Next attribute, capital A, that stands for do not update the access time or A time.
So what that is is if you use a command like stat or LS-U, you should be able to see the
access time on the file, that means the last time the file was accessed.
Now you got to be careful with that because if you access the file like doing LS on the
file, it hits the file and the file system becomes back, you're going to see that the access
time gets updated.
But if you do that again, you're not going to see the access time gets updated more than
likely because chances are that file is cached in memory and it's just going to the memory
and set it to the file system to get that information.
So be aware of that.
Now again, what this does is exactly what it says.
It will not update the access time.
They also have to realize that another reason why you might not see the access time updated
is depending upon how the file system that it's on is mounted.
Now you can have it mounted with the option no way time or real A time or it's relational
A time, not real.
And the other one, I can't remember what is off the bat, but anyway, what ends up happening
like with relay time, relational A time.
The way that that behaves is if you access the file, it doesn't update the time stamp.
If the time's access time stamp is greater than the last modification time stamp, it doesn't
update it, but if it's older than the access time, then it will update the access time.
But until you manipulate it and then read it again, the access time might not be updated.
So there's another reason why you might not see access time updated.
So it gets a little confusing, but this attribute sets no access time.
It does not update the access time on this file.
The C or compressed attribute sets this file to be compressed when written to on the disk.
So if you write to the file, it compresses the data and writes it to the file.
If you read to it from the file, it uncompresses the file as it's reading to the data and displays
it.
It just does compression.
Now there's overhead associated with that.
Just be aware of that.
It can cause some overhead.
Chances are a lot of these attributes you're probably never going to need.
So just understand what those mean.
So you can compress on read and it'll compress and write.
The capital C means no copy on write.
What this means is the copy on write.
What that does is a way that a lot of modern unic systems work, that if a file is accessed,
instead of creating multiple open copies of the file, instead of creating multiple copies
of the information for each task, a pointer to the shared resources used.
So instead of, when multiple times that file is open, it's all pointing to the same
thing as a bunch of pointers.
So should the task, should that file be written to, it kind of breaks the pointer and it
actually writes it to a private copy.
So it breaks it off, writes it to a private copy.
And then, so you're not manipulating the shared resource.
Now you might see that, something similar to that happen if you're in VI and you're editing
a file and you do something that a file changes the file and it'll come back and say the
file has changed, you want to reload, just be aware of that.
So you have shared pointers to it.
It'll break off a private copy if you need to manipulate it and then put it back when
you actually write it to the disk.
Now what this does is it prevents that from happening, it's no copy on right.
So it writes directly instead of creating a shared copy or a private copy so that private
copy is never created, it just goes right to the resource.
There is a lowercase D, which stands for no dump.
That means the file is not to be a candidate for backups when the dump command is used.
Okay, pretty basic.
Capital D, synchronous directory updates.
Okay, so when a change is made to file in a directory, those changes are written synchronously
to the disk.
What that means is that any change to files are immediately written to the disk and available.
That's not necessarily how the behavior is in a network file system where it'll write
the changes but you might not see them on the network file system immediately.
When a file is created in a share, the directory is immediately updated in this case so that
it's visible to the systems accessing the share.
It's primarily for a network based file system like NFS or if you're mounted with NFS
and NFS and you make a change to it with the capital D set on the directory, it immediately
updates those files in there so that the update is pushed out to the other systems that
might have that file system mounted.
Lowercasee, extents.
This attribute indicates the file is using extents from mapping the blocks on the disk.
Now you're going to see this in ec2 or 3.
This is more of ec4, more modern file systems.
It replaces the traditional block mapping schemes and what it does is defines a continuous
range of physical blocks for storing data and what that means is it defines those blocks.
It means that file, when it has an e or an extents attribute to it, it means that it's
using extents.
Now you cannot set this attribute but you can only view it.
This is set by the file system itself that's using it.
There's a capital e attribute, now this is experimental and it's set by a compression
program to indicate that a compressed file has a compression error.
Again this is not something that you can set.
This is set by the program that does the compression and it means that that file has
a compression error so that you can only view that, not actually set it.
Lowercasee, another attribute that you cannot set, its presence indicates that the file
is storing of blocks and units of the file system block size as opposed to units of sectors.
This is primarily used when the file is larger than two terabytes or it was one time larger
than two terabytes so chances are unless you're working with huge data systems, you're not
going to see this attribute set and you can't set it anyway, it's set by the file system.
The lowercasee or immutable, this renders a file impervious to change, no matter what
the permissions are on it, you cannot change the file, it cannot be written to, it cannot
be appended to or it cannot be deleted, it cannot be renamed or it cannot have a link
created to it.
Now this attribute can only be set by a super user account and if you want to delete
the file or change it or whatever you have to remove the attribute, again it can only
be set by the super user account.
Capital i means directory is being indexed.
Again this is not something you can set but it is set when the directory is being indexed
by hash trees except by another application, you can see it but you can't set it.
J, data journaling, this is the lowercase J.
Now if the file system is mounted with the data equals ordered or data equals right back
options, what this will do is force the data to be written to like an EST3 journal before
it being written to the file.
This is what data journaling does, instead of the file going to the journal first and
then being a later written to the file, it goes directly, no wait I got that backed
up, instead of writing to the file directly it writes to the journal first and then now
to the file, if the data is ordered, if it is ordered or right back what this attribute
does is force it to be, instead of written to the journal it writes it to the data first
and then to the journal.
My correct on that, let me double check.
Yes, I've got that, in a situation where data is ordered or right back, it's written
to the file then to the journal, it can be written to the file then to the journal, this
forces it to be written to the journal first and then out to the file.
The shares amounted with data equals journal, then this attribute does no effect at all
on this file because that's the way data equals journal behaves.
Secure deletion, lowercase S, when a file is deleted the blocks of the file are zeroed
out, so if you set secure deletion it zeros out the file instead of just removing the
pointers to the data, it zeros out the data that used to be there.
Capital S is synchronous updates, with synchronous update attributes set, any change to a file
are immediately written to the disk, otherwise it changes to the file or cash and then updated
it later time.
So in a situation where it writes out to the cash, changes to the file system and then
later updates the file on the disk, that's why you don't want to just power off the system
because there could be stuff waiting to be written to the disk.
But this attribute does, if there's ever a change it immediately writes it to the
disk, it doesn't push it out there to the cash.
T, lowercase T means no tail merging.
Tail merging, what is also known as block sub allocation, is when a single block is used
to hold the tail end of a number of files of data.
Now in traditional file system, block allocation, when you have like most, I think modern Linux
systems are still using four kilobyts blocks.
When data gets written out, when a file gets written out to the file system, it writes it
out in four kilobyts blocks.
Now the last bit of data may or may not take up that entire four kilobytes.
And if it doesn't, you have wasted space at the end.
It could be one byte or it could be 300 or 350 bytes and you still have extra spaces
not being used.
Tail merging does, is on file systems that support tail merging, it will take some of
those files that have x that don't fill up the last block and it will crunch them together
so that you have three or four files sharing whose tails share one single block.
Now I think it's EXT 2 and I don't think EXT 3, I'm not sure maybe net won't support
it.
I think EXT 4 does, but what no tail merging does is it prevents that file from having
that happen to it or that directory, if you put it on directory, prevents those files
in there from doing the block sub allocation or tail merging.
Capital T. Top of directory hierarchy, okay.
This works on what's called Warlov's block allocator algorithm.
And the idea is that when you store files in related directories, store them closer
together so that there will be faster disk access.
For instance, in your home directory, the way that you ideally want stuff to be stored
in your home directory on the file is if a person A's files would be closer to person
A's home directory on the disk.
And then in another disk location person, B's files would be closer together to their
home directory on the disk.
And without using this algorithm, you would end up having like at the top level, you'd
have everybody's home directory and then all the files would get written and they would
be not allocated closer together on the disk.
What the dash capital T does is it would take something at a directory and kind of force
the or loss of the top directory hierarchy so that it would start grouping the files in
the sub directories closer to that directory and into those sub directories on the disk
to theoretically reduce disk access time instead of just having them placed as they are
on there at Willie Nilly.
It says, it says, take this directory and understand that anything underneath it needs
to follow Warlove's block allocator algorithm as closely as possible.
So it will force sub directories to be unrelated and should be spread apart.
So then you would have the directories under there, they'll be unrelated but the contents
of those directories would be related so that they would be spread apart and you would
have faster access.
Hope that makes sense.
Here's underscore, lowercase u, undeliedable.
This sets the attributes of a file to be recoverable, unrecoverable should it be deleted.
The contents of a file are actually saved.
Well, if you delete it, it would be recoverable.
Now this option was back in EXT1 so the first extended file system, you're not going to
really see it on EXT2 and EXT3, it's a holdover from an older version but basically what
just says if you delete it and you have the undeliedable attribute set, you'd be able
to recover it.
But that's no longer available, it's deprecated, you're not really going to find it.
There are a few other experimental attributes right now, you're not going to be able to
really use them to the extent you can see them in the man page.
But those are the basic attributes, extended attributes right there and what they do.
Now you can see attributes by using the LSATTR command or LSATribute command.
It's like an LS command in general listing but it shows the files in a current directory.
If you just type LSATTR, it'll show you all the files and directories in then prior,
right before it where you'd see normal permissions, you're going to see the attributes instead.
So that's the way LSATTR works, it's just doing your home directory right now, you're
going to see a list of all your files and more than likely the only attribute that you
are going to see in there might be lowercasee, extents, that's what I see because I'm using
an EXT4 file system.
If I wasn't using AXT4, probably I wouldn't see any attributes whatsoever.
So just be aware of that.
Now there are a couple of flags that you could pass to LSATTR, one is capital dash or
dash capital R or uppercase R and that's for recursively listing attributes and directories
and sub-directories.
So if you do LS dash capital R, you're going to see all the attributes of the file in the
current directory and as it hits each directory is going to recurse into the sub-directories,
show the files and keep recursing until it's done everything.
Dash A, lowercase A is going to list all the files and directory including hidden files.
So if you add the two together, dash capital R and dash capital R and lowercase A, you're
going to see a recursive listing of all the attributes and of all the files including
the dot hidden files, the hidden files.
Dash D suppresses, and this is lowercase, these suppresses directory, contents of directory
and just list directories like other files.
Each capital V displays the version of LSATTR and dash lowercase V displays the file version
or generation number.
Now the file version or generation number, that's something that's handled by the file system.
This is a number that's handy, particularly in a network file system like the NFS, where
the version of the number, version number can be checked to see if a file has been changed
or deleted by another user when multiple users are accessing the file or the file system.
So that is usually typically set by the file system.
The other command that you can use with extended attributes is the change attribute or CHATTR.
Now this allows you to manipulate some of the attributes, not all of them because we talked
about some of them like extents are not able to be set by a change attribute.
Some of those like append-only or immutable require you to be super user privileges to
be able to change attributes and just like regular file permissions with the CHON or CHMOD,
not CHON, CHMOD, you set the permissions with a plus or a minus or you set the equal.
Now if you do plus, it sets the attribute.
If you do minus, it removes the attribute, both in those cases if they're apt, if the
attribute is already set, it doesn't change anything.
If the attribute wasn't available, it doesn't change anything.
The equals, on the other hand, sets it to exactly what you say.
So if there are any other attributes that are applied to that file that you can change
and use use equal and those files, you don't specify those attributes, they will be removed
and the attributes you specify to be applied.
Very simple.
So then you just do a plus minus equals and whatever label is that you want to set.
Like if you want to set a append-only compression, it would be chatter plus AC test.txt.
Now you have to be rude to do that or a super user, you have to pseudo chatter plus AC test.txt.
So that's basically how to manipulate the attributes, the extended attributes of the
file.
There are a few flags that chatter takes.
One is the capital V, which tells chatter to be verbose on what it's doing and it also
prints a program version.
There's the dash F switch, which will suppress any error messages.
And then there's the dash V switch, which will allow you to set the version number of a
file.
I do not recommend doing that or messing with that unless you absolutely know what you're
doing because I wouldn't understand that.
If you look at the version number, I wouldn't be confident in myself trying to set a version
number because the version numbers are fairly long numbers.
And I don't know what happens if you set it the one.
I don't know what that does.
So you should probably leave that alone.
So that is extended attributes in a nutshell right there.
There's a lot of complicated stuff and understanding some heavy duty file system and stuff like
that that I just pretty much glossed over through it out there.
It's complicated, but I have a lot of links to where you can get more information on some
of this stuff over there on the website at linuxandtheschel.org.
Check it out.
Also go to read up the right up and the video of showing some examples.
Probably not going to be that great of a video because what are you going to really show?
How to use LSATTR and chatter.
My name is Dan Washco.
This is Linux and the Shell, extended attributes.
Thank you for listening, and I'll see you in a couple of weeks.
You have been listening to Hacker Public Radio or Hacker Public Radio does our.
We are a community podcast network that releases shows every weekday Monday through Friday.
Today's show, like all our shows, was contributed by a HPR listener by yourself.
If you ever consider recording a podcast, then visit our website to find out how easy
it really is.
Hacker Public Radio was founded by the digital dog pound and the infonomicum computer
cloud.
HPR is funded by the binary revolution at binref.com, all binref projects are crowd-sponsored
by linear pages.
For shared hosting to custom private clouds, go to lunarpages.com for all your hosting
needs.
Unless otherwise stasis, today's show is released under a creative commons, attribution, share
a line, 3.0 license.