My solutions to Eudyptula challenges.
The task descriptions were taken from Agelastic's repository.
Write a Linux kernel module, and stand-alone Makefile, that when loaded prints to the kernel debug log level, "Hello World!" Be sure to make the module be able to be unloaded as well.
The Makefile should build the kernel module against the source for the currently running kernel, or, use an environment variable to specify what kernel tree to build it against.
Please show proof of this module being built, and running, in your kernel. What this proof is is up to you, I'm sure you can come up with something. Also be sure to send the kernel module you wrote, along with the Makefile you created to build the module.
Remember to use your ID assigned to you in the Subject: line when responding to this task, so that I can figure out who to attribute it to. You can just respond to the task with the answers and all should be fine.
If you forgot, your id is "". But of course you have not forgotten that yet, you are better than that.
Now that you have written your first kernel module, it's time to take off the training wheels and move on to building a custom kernel. No more distro kernels for you, for this task you must run your own kernel. And use git! Exciting isn't it! No, oh, ok...
The tasks for this round is:
- download Linus's latest git tree from git.kernel.org (you have to figure out which one is his, it's not that hard, just remember what his last name is and you should be fine.)
- build it, install it, and boot it. You can use whatever kernel configuration options you wish to use, but you must enable CONFIG_LOCALVERSION_AUTO=y.
- show proof of booting this kernel. Bonus points for you if you do it on a "real" machine, and not a virtual machine (virtual machines are acceptable, but come on, real kernel developers don't mess around with virtual machines, they are too slow. Oh yeah, we aren't real kernel developers just yet. Well, I'm not anyway, I'm just a script...) Again, proof of running this kernel is up to you, I'm sure you can do well.
Hint, you should look into the 'make localmodconfig' option, and base your kernel configuration on a working distro kernel configuration. Don't sit there and answer all 1625 different kernel configuration options by hand, even I, a foolish script, know better than to do that!
After doing this, don't throw away that kernel and git tree and configuration file. You'll be using it for later tasks, a working kernel configuration file is a precious thing, all kernel developers have one they have grown and tended to over the years. This is the start of a long journey with yours, don't discard it like was a broken umbrella, it deserves better than that.
Remember to use your ID assigned to you in the Subject: line when responding to this task, so that I can figure out who to attribute it to.
If you forgot, your id is "". But why do I repeat myself? Of course you know your id, you made it through the first task just fine with it.
Now that you have your custom kernel up and running, it's time to modify it!
The tasks for this round is:
- take the kernel git tree from Task 02 and modify the Makefile to and modify the EXTRAVERSION field. Do this in a way that the running kernel (after modifying the Makefile, rebuilding, and rebooting) has the characters "-eudyptula" in the version string.
- show proof of booting this kernel. Extra cookies for you by providing creative examples, especially if done in intrepretive dance at your local pub.
- Send a patch that shows the Makefile modified. Do this in a manner that would be acceptable for merging in the kernel source tree. (Hint, read the file Documentation/SubmittingPatches and follow the steps there.)
Remember to use your ID assigned to you in the Subject: line when responding to this task, so that I can figure out who to attribute it to.
If you forgot, your id is "". Surely I don't need to keep saying this right? I know, you wouldn't forget, but someone else, of course they would, so I'll just leave it here for those "others".
Wonderful job in making it this far, I hope you have been having fun. Oh, you're getting bored, just booting and installing kernels? Well, time for some pedantic things to make you feel that those kernel builds are actually fun!
Part of the job of being a kernel developer is recognizing the proper Linux kernel coding style. The full description of this coding style can be found in the kernel itself, in the Documentation/CodingStyle file. I'd recommend going and reading that right now, it's pretty simple stuff, and something that you are going to need to know and understand. There is also a tool in the kernel source tree in the scripts/ directory called checkpatch.pl that can be used to test for adhering to the coding style rules, as kernel programmers are lazy and prefer to let scripts do their work for them...
And why a coding standard at all? That's because of your brain (yes, yours, not mine, remember, I'm just some dumb shell scripts). Once your brain learns the patterns, the information contained really starts to sink in better. So it's important that everyone follow the same standard so that the patterns become consistent. In other words, you want to make it really easy for other people to find the bugs in your code, and not be confused and distracted by the fact that you happen to prefer 5 spaces instead of tabs for indentation. Of course you would never prefer such a thing, I'd never accuse you of that, it was just an example, please forgive my impertinence!
Anyway, the tasks for this round all deal with the Linux kernel coding style. Attached to this message are two kernel modules that do not follow the proper Linux kernel coding style rules. Please fix both of them up, and send it back to me in such a way that does follow the rules.
What, you recognize one of these modules? Imagine that, perhaps I was right to accuse you of the using a "wrong" coding style :)
Yes, the logic in the second module is crazy, and probably wrong, but don't focus on that, just look at the patterns here, and fix up the coding style, do not remove lines of code.
As always, please remember to use your ID assigned to you in the Subject: line when responding to this task, so that I can figure out who to attribute it to. And if you forgot (which of course you have not, we've been through all of this before), your id is "".
Yeah, you survived the coding style mess! Now, on to some "real" things, as I know you are getting bored by these so far.
So, simple task this time around:
- take the kernel module you wrote for task 01, and modify it so that when a USB keyboard is plugged in, the module will be automatically loaded by the correct userspace hotplug tools (which are implemented by depmod / kmod / udev / mdev / systemd, depending on what distro you are using.)
Yes, so simple, and yet, it's a bit tricky. As a hint, go read chapter 14 of the book, "Linux Device Drivers, 3rd edition." Don't worry, it's free, and online, no need to go buy anything.
As always, please remember to use your ID assigned to you, yadda yadda yadda... It's "" of course.
Nice job with the module loading macros, those are tricky, but a very valuable skill to know about, especially when running across them in real kernel code.
Speaking of real kernel code, let's write some!
The task this time is this:
- Take the kernel module you wrote for task 01, and modify it to be a misc char device driver. The misc interface is a very simple way to be able to create a character device, without having to worry about all of the sysfs and character device registration mess. And what a mess it is, so stick to the simple interfaces wherever possible.
- The misc device should be created with a dynamic minor number, no need running off and trying to reserve a real minor number for your test module, that would be crazy.
- The misc device should implement the read and write functions.
- The misc device node should show up in /dev/eudyptula.
- When the character device node is read from, your assigned id is returned to the caller.
- When the character device node is written to, the data sent to the kernel needs to be checked. If it matches your assigned id, then return a correct write return value. If the value does not match your assigned id, return the "invalid value" error value.
- The misc device should be registered when your module is loaded, and unregistered when it is unloaded.
- Provide some "proof" this all works properly.
As you will be putting your id into the kernel module, of course you haven't forgotten it, but just to be safe, it's "".
Great work with that misc device driver. Isn't that a nice and simple way to write a character driver?
Just when you think this challenge is all about writing kernel code, this task is a throwback to your second one. Yes, that's right, building kernels. Turns out that's what most developers end up doing, tons and tons of rebuilds, not writing new code. Sad, but it is a good skill to know.
The tasks this round are:
- Download the linux-next kernel for today. Or tomorrow, just use the latest one. It changes every day so there is no specific one you need to pick. Build it. Boot it. Provide proof that you built and booted it.
What is the linux-next kernel? Ah, that's part of the challenge.
For a hint, you should read the excellent documentation about how the Linux kernel is developed in Documentation/development-process/ in the kernel source itself. It's a great read, and should tell you all you never wanted to know about what Linux kernel developers do and how they do it.
As always, please respond to this challenge with your id. I know you know what it is. I'll not even include it this time, I trust you. Don't make me feel that is a mistake.
We will come back to the linux-next kernel in a later exercise, so don't go and delete that directory, you'll want it around. But enough of building kernels, let's write more code!
This task is much like the 06 task with the misc device, but this time we are going to focus on another user/kernel interface, debugfs. It is rumored that the creator of debugfs said that there is only one rule for debugfs use, "There are no rules when using debugfs." So let's take them up on that offer and see how to use it.
debugfs should be mounted by your distro in /sys/kernel/debug/, if it isn't, then you can mount it with the line: mount -t debugfs none /sys/kernel/debug/
Make sure it is enabled in your kernel, with the CONFIG_DEBUG_FS option, you will need it for this task.
The task, in specifics is:
- Take the kernel module you wrote for task 01, and modify it to be create a debugfs subdirectory called "eudyptula". In that directory, create 3 virtual files called "id", "jiffies", and "foo".
- The file "id" operates just like it did for example 06, use the same logic there, the file must be readable and writable by any user.
- The file "jiffies" is to be read only by any user, and when read, should return the current value of the jiffies kernel timer.
- The file "foo" needs to be writable only by root, but readable by anyone. When writing to it, the value must be stored, up to one page of data. When read, which can be done by any user, the value must be returned that is stored it it. Properly handle the fact that someone could be reading from the file while someone else is writing to it (oh, a locking hint!)
- When the module is unloaded, all of the debugfs files are cleaned up, and any memory allocated is freed.
- Provide some "proof" this all works.
Again, you are using your id in the code, so you know what it is by now, no need to repeat it again.
Nice job with debugfs, that is a handy thing to remember when wanting to print out some debugging information. Never use /proc/ that is only for processes, use debugfs instead.
Along with debugfs, sysfs is a common place to put information that needs to move from the user to the kernel. So let us focus on sysfs for this task.
The task this time:
- Take the code you wrote in task 08, and move it to sysfs. Put the "eudyptula" directory under the /sys/kernel/ location in sysfs.
- Provide some "proof" this works.
That's it! Simple, right? No, you are right, it's more complex, sysfs is complicated. I'd recommend reading Documentation/kobject.txt as a primer on how to use kobjects and sysfs.
Feel free to ask for hints and help with this one if you have questions by sending in code to review if you get stuck, many people have dropped out in the challenge at this point in time, so don't feel bad about asking, we don't want to see you go away just because sysfs is so damn complicated.
Yeah, you conquered the sysfs monster, great job!
Now you know to never want to mess with a kobject again, right? Trust me, there are easier ways to create sysfs files, and we will get into that for a future task, but for now, let's make it a bit more simple after all of that coding.
For this task, go back to the linux-next tree you used for task 07. Update it, and then do the following:
- Create a patch that fixes one coding style problem in any of the files in drivers/staging/
- Make sure the patch is correct by running it through scripts/checkpatch.pl
- Submit the code to the maintainer of the driver/subsystem, finding the proper name and mailing lists to send it to by running the tool, scripts/get_maintainer.pl on your patch.
- Send a web link back to me of your patch on the public mailing list archive (don't cc: me on the patch, that will only confuse me and everyone in the kernel development community.)
- If you want to mention the Eudyptula Challenge as the reason for writing the patch, feel free to do so in the body of the patch, but it's not necessary at all.
Hopefully this patch will be accepted into the kernel tree, and all of a sudden, you are an "official" kernel developer!
Don't worry, there's plenty more tasks coming, but a little breather every now and again for something simple is always nice to have.
You made a successful patch to the kernel source tree, that's a great step!
But, let's not rest, time to get back to code.
Remember that mess of kobject and sysfs code back in task 09? Let's move one level up the tree and start to mess with devices and not raw kobjects.
For this task:
- Write a patch against any driver that you are currently using on your machine. So first you have to figure out which drivers you are using, and where the source code in the kernel tree is for that driver.
- In that driver, add a sysfs file to show up in the /sys/devices/ tree for the device that is called "id". As you might expect, this file follows the same rules as task 09 as for what you can read and write to it.
- The file is to show up only for devices that are controlled by a single driver, not for all devices of a single type (like all USB devices. But all USB maibox LEDs would be acceptable, if you happen to have the device that that driver controls.)
Submit both the patch, in proper kernel commit form, and "proof" of it working properly on your machine.
And as always, please use your id in the subject line. If you happened to forget it, as it has been a while since I reminded you of it, it is "".
Nice job with the driver patch. That took some work in finding the proper place to modify, and demonstrates a great skill in tracking down issues when you can't get a specific piece of hardware working.
Now let's step back from drivers (they are boring things), and focus on the kernel core. To do that, we need to go way back to the basics -- stuff you would learn in your "intro to data structures" class, if you happened to take one.
Yes, I'm talking about linked lists.
The kernel has a unique way of creating and handling linked lists, that is quite different than the "textbook" way of doing so. But, it turns out to be faster, and simpler, than a "textbook" would describe, so that's a good thing.
For this task, write a kernel module, based on your cleaned up one from task 04, that does the following:
-
You have a structure that has 3 fields: char name[20]; int id; bool busy; name this structure "identity".
-
Your module has a static variable that points to a list of these "identity" structures.
-
Write a function that looks like: int identity_create(char *name, int id) that creates the structure "identity", copies in the 'name' and 'id' fields and sets 'busy' to false. Proper error checking for out of memory issues is required. Return 0 if everything went ok; an error value if something went wrong.
-
Write a function that looks like: struct identity *identity_find(int id); that takes a given id, iterates over the list of all ids, and returns the proper 'struct identity' associated with it. If the identity is not found, return NULL.
-
Write a function that looks like: void identity_destroy(int id); that given an id, finds the proper 'struct identity' and removes it from the system.
-
Your module_init() function will look much like the following:
struct identity *temp; identity_create("Alice", 1); identity_create("Bob", 2); identity_create("Dave", 3); identity_create("Gena", 10); temp = identity_find(3); pr_debug("id 3 = %s\n", temp->name); temp = identity_find(42); if (temp == NULL) pr_debug("id 42 not found\n"); identity_destroy(2); identity_destroy(1); identity_destroy(10); identity_destroy(42); identity_destroy(3);
Bonus points for properly checking return values of the above functions.
As always, please send the full module (following the proper kernel coding style rules), and "proof" of it working properly. And remember to use your id of "" in the Subject line properly.
Weren't those lists fun to play with? You should get used to them, they are used all over the kernel in lots of different places.
Now that we are allocating a structure that we want to use a lot of, we might want to start caring about the speed of the allocation, and not have to worry about the creation of those objects from the "general" memory pools of the kernel.
This task is to take the code written in task 12, and cause all memory allocated from the 'struct identity' to come from a private slab cache just for the fun of it.
Instead of using kmalloc() and kfree() in the module, use kmem_cache_alloc() and kmem_cache_free() instead. Of course this means you will have to initialize your memory cache properly when the module starts up. Don't forget to do that. You are free to name your memory cache whatever you wish, but it should show up in the /proc/slabinfo file.
Don't send the full module for this task, only a patch with the diff from task 12 showing the lines changed. This means you will have to keep a copy of your 12 task results somewhere to make sure you don't overwrite them.
Also show the output of /proc/slabinfo with your module loaded.
Now that you have the basics of lists, and we glossed over the custom allocators (the first cut at that task was much harder, you got off easy), it's time to move on to something a bit more old-school: tasks.
For this task:
- Add a new field to the core kernel task structure called, wait for it, "id".
- When the task is created, set the id to your id. Imaginative, I know. You try writing these tasks.
- Add a new proc file for every task called, "id", located in the /proc/${PID}/ directory for that task.
- When the proc file is read from, have it print out the value of your id, and then increment it by one, allowing different tasks to have different values for the "id" file over time as they are read from.
- Provide some "proof" it all works properly.
As you are touching files all over the kernel tree, a patch is the required result to be sent in here. Please specify which kernel version you make this patch against, to give my virtual machines a chance to figure out how to apply it.
Also provide some kind of proof that you tested the patch.
And, in case you happened to forget it, your id is "".
That process task turned out to not be all that complex, but digging through the core kernel was a tough task, nice work with that.
Speaking of "core kernel" tasks, let's do another one. It's one of the most common undergraduate tasks there is: create a new syscall! Yeah, loads of fun, but it's good to know the basics of how to do this, and, how to call it from userspace.
For this task:
- Add a new syscall to the kernel called "sys_eudyptula", so this is all going to be changes to the kernel tree itself, no stand-alone module needed for this task (unless you want to do it that way without hacking around the syscall table, if so, bonus points for you...)
- The syscall number needs to be the next syscall number for the architecture you test it on (some of you all are doing this on ARM systems, showoffs, and syscall numbers are not the same across all architectures). Document the arch you are using and why you picked this number in the module.
- The syscall should take two parameters: int high_id, int low_id.
- The syscall will take the two values, mush them together into one 64bit value (low_id being the lower 32bits of the id, high_id being the upper 32bits of the id).
- If the id value matches your id (which of course you know as "", then the syscall returns success. Otherwise it returns a return code signifying an invalid value was passed to it.
- Write a userspace C program that calls the syscall and properly exercises it (valid and invalid calls need to be made).
- "Proof" of running the code needs to be provided.
So you need to send in a .c userspace program, a kernel patch, and "proof" that it all works, as a response.
Good job with the new syscall. The odds of you ever having to write a new syscall is pretty rare, but now you know where to look them up, and what the code involved in creating one looks like, which is useful when you try to debug things.
Let's take a breather after all of that code, and go back to doing a bit of reading, and learn some more about some common kernel tools.
Go install the tool 'sparse'. It was started by Linus as a static-analysis tool that acts much like a compiler. The kernel build system is set up to have it run if you ask it to, and it will report a bunch of issues in C code that are really specific to the kernel.
When you build the kernel, pass the "C=1" option to the build, to havesparse run on the .c file before gcc is run. Depending on the file, nothing might be printed out, or something might. Here's an example of it being run on the ext4 code:
$ make C=1 M=fs/ext4 CHECK fs/ext4/balloc.c CC fs/ext4/balloc.o CHECK fs/ext4/bitmap.c CC fs/ext4/bitmap.o CHECK fs/ext4/dir.c CC fs/ext4/dir.o CHECK fs/ext4/file.c CC fs/ext4/file.o CHECK fs/ext4/fsync.c CC fs/ext4/fsync.o CHECK fs/ext4/ialloc.c CC fs/ext4/ialloc.o CHECK fs/ext4/inode.c CC fs/ext4/inode.o CHECK fs/ext4/page-io.c CC fs/ext4/page-io.o CHECK fs/ext4/ioctl.c CC fs/ext4/ioctl.o CHECK fs/ext4/namei.c CC fs/ext4/namei.o CHECK fs/ext4/super.c CC fs/ext4/super.o CHECK fs/ext4/symlink.c CC fs/ext4/symlink.o CHECK fs/ext4/hash.c CC fs/ext4/hash.o CHECK fs/ext4/resize.c CC fs/ext4/resize.o CHECK fs/ext4/extents.c CC fs/ext4/extents.o CHECK fs/ext4/ext4_jbd2.c CC fs/ext4/ext4_jbd2.o CHECK fs/ext4/migrate.c CC fs/ext4/migrate.o CHECK fs/ext4/mballoc.c fs/ext4/mballoc.c:5018:9: warning: context imbalance in 'ext4_trim_extent' - unexpected unlock CC fs/ext4/mballoc.o CHECK fs/ext4/block_validity.c CC fs/ext4/block_validity.o CHECK fs/ext4/move_extent.c CC fs/ext4/move_extent.o CHECK fs/ext4/mmp.c CC fs/ext4/mmp.o CHECK fs/ext4/indirect.c CC fs/ext4/indirect.o CHECK fs/ext4/extents_status.c CC fs/ext4/extents_status.o CHECK fs/ext4/xattr.c CC fs/ext4/xattr.o CHECK fs/ext4/xattr_user.c CC fs/ext4/xattr_user.o CHECK fs/ext4/xattr_trusted.c CC fs/ext4/xattr_trusted.o CHECK fs/ext4/inline.c CC fs/ext4/inline.o CHECK fs/ext4/acl.c CC fs/ext4/acl.o CHECK fs/ext4/xattr_security.c CC fs/ext4/xattr_security.o LD fs/ext4/ext4.o LD fs/ext4/built-in.o Building modules, stage 2. MODPOST 0 modules
As you can see, only one warning was found here, and odds are, it is a false-positive, as I'm sure those ext4 developers know what they are doing with their locking functions, right?
Anyway the task this time is:
- Run sparse on the drivers/staging/ directory, yes, that horrible code again, sorry.
- Find one warning that looks interesting.
- Write a patch that resolves the issue.
- Make sure the patch is correct by running it through scripts/checkpatch.pl
- Submit the code to the maintainer of the driver/subsystem, finding the proper name and mailing lists to send it to by running the tool, scripts/get_maintainer.pl on your patch.
- Send a web link back to me of your patch in the public mailing list archive (don't cc: me on the patch, that will only confuse me and everyone in the kernel development community.)
- If you want to mention the Eudyptula Challenge as the reason for writing the patch, feel free to do so in the body of the patch, but it's not necessary at all.
That's it, much like task 10 was, but this time you are fixing logical issues, not just pesky coding style issues. You are a real developer now, fixing real bugs!
Another patch made and sent in. See, that wasn't so hard. Keep sending in kernel patches outside of this challenge, those lazy kernel developers need all the help they can get in cleaning up their code.
It is time to start putting the different pieces of what we have done in the past together, into a much larger module, doing more complex things. Much more like what a "real" kernel module has to do.
Go dig up your code from task 06, the misc char device driver, and make the following changes:
- Delete the read function. You don't need that anymore, so make it a write-only misc device and be sure to set the mode of the device to be write-only, by anyone. If you do this right, udev will set up the node automatically with the correct permissions.
- Create a wait queue, name it "wee_wait".
- In your module init function, create a kernel thread, named of course "eudyptula".
- The thread's main function should not do anything at this point in time, except make sure to shutdown if asked to, and wait on the "wee_wait" waitqueue.
- In your module exit function, shut down the kernel thread you started up.
Load and unload the module and "prove" that it works properly (that the thread is created, it can be found in the process list, and that the device node is created with the correct permission value.) Send in the proof and the .c file for the module.
Be sure to keep this code around, as we will be doing more with it next time.
Nice job with the kernel thread. It really doesn't take much code at all to create a new thread. So now let us get some data into the module to give the thread something to do.
Base all of this work on your task 17 codebase.
Go back and dig up task 12's source code, the one with the list handling. Copy the structure into this module, and the identity_create(), identity_find(), and identity_destroy() functions into this module as well.
Write a new function, identity_get(), that looks like: struct identity identity_get(void); and returns the next "identity" structure that is on the list, and removes it from the list. If nothing is on the list, return NULL.
Then, hook up the misc char device "write" function to do the following:
- If a write is larger than 19 characters, truncate it at 19.
- Take the write data and pass it to identity_create() as the string, and use an incrementing counter as the "id" value.
- Wake up the "wee_wait" queue.
In the kernel thread function:
- If the "wee_wait" queue wakes us up, get the next identity in the system with a call to identity_get().
- Sleep (in an interruptable state, don't go increasing the system load in a bad way) for 5 seconds.
- Write out the identity name, and id to the debug kernel log and then free the memory.
When the module exits, clean up the whole list by using the functions given, no fair mucking around with the list variables directly.
Yes, it's a bit clunky, but it shows the basics of taking work from userspace, and then quickly returning to the user, and then going off and doing something else with the data and cleaning everything up. It's a common pattern for a kernel, as it's really all that a kernel ends up doing most of the time (get a disk block, write a disk block, handle a mouse event, etc.)
Load and unload the module and "prove" that it works properly by writing and looking at the debug log, and that everything cleans up properly when the module is unloaded.
Send in the proof and the .c file for the module.
A good test script would be the following: echo "Alice" > /dev/eudyptula echo "Bob" > /dev/eudyptula sleep 15 echo "Dave" > /dev/eudyptula echo "Gena" > /dev/eudyptula rmmod task18
Removing the module while there is pending work is always a good stress test.
Handling delayed work is easy now, right? So, time to move on to something totally different. How about networking? We have been ignoring that part of the kernel, so let us now focus on the network side of the kernel, as that is a huge reason for why Linux has taken over the world.
For this task, write a netfilter kernel module that does the following:
- monitors all IPv4 network traffic that is coming into the machine
- prints the id to the kernel debug log if the network traffic stream contains your id.
- properly unregisters you from the netfilter core when the module unloads.
Test this by sending yourself an email with your id in the subject, much like the email you need to send back to me.
Send in the proof and the .c file for the module.
Networking filters turned out to be not all that complex in the end, great work with that.
So, here's the final task.
There might be other tasks that get created and sent out later on, but the original challenge had 20 tasks, so after finishing this one, you can consider yourself done!
Let's try something a bit harder. Something that might cause some data loss on a filesystem, always a fun thing to play with, if for no other reason than to not be afraid of things like that in the future.
This task requires you to work on the fat filesystem code:
- Add an ioctl to modify the volume label of a mounted fat filesystem. Be sure to handle both 16 and 32 bit fat filesystems {hint!}
- Provide a userspace .c program to test this new ioctl.
That's it! Seems so simple, right? I wonder why that option isn't in the kernel tree already...
Anyway, provide a patch to the kernel, and the .c file used to test the new ioctl, as well as "proof" of this working. Make sure you don't run into 32/64bit kernel issues with the ioctl, if you do things correctly, you shouldn't have any problems.
I recommend doing this work on either a loop-back fat filesystem on your "normal" filesystem, or on a USB stick. Either will work just as well, and make things easier to debug and test.
Watch out for locking issues, as well as dirty filesystem state problems.
Best of luck!