Operating Systems 2018F Lecture 17: Difference between revisions
Created page with "==To Do== * Demo kernel oops in newgetpid * increase storage in remember" |
R.samanfar (talk | contribs) |
||
(One intermediate revision by one other user not shown) | |||
Line 1: | Line 1: | ||
== | ==Video== | ||
* | The video from the lecture on November 9, 2018 [https://homeostasis.scs.carleton.ca/~soma/os-2018f/lectures/comp3000-2018f-lec17-20181109.m4a is now available]. | ||
==Notes== | |||
alloc pages -> frames -> page | |||
''''FUSE'''' - filesystem in userspace | |||
interface for userspace programs to export a filesystem to the LINUX kernel | |||
'''What is a filesystem from a kernel?''' | |||
* A device file (character device, block device), depends on file operations. | |||
* A filesystem is just this scaled up, it should not only include these file operations, but it should also include operations for directories, and mounting, unmounting. | |||
If you can define those, you can make a filesystem | |||
FUSE makes a "special" filesystem that allows you to create other filesystems | |||
Filesystems are abstractions of file operations | |||
Filesystems talk to rest of the kernel through a virtual filesystem interface | |||
'''Are all userspace filesystem in memory?''' | |||
Standard ntfs implementation on most linux distributions, is written in userspace | |||
'''Microkernals vs monolithic kernal''' | |||
microkernel: everything turns into a process | |||
monolithic kernel: it's faster | |||
'''memoryll.py''' | |||
mounting dev/fuse | |||
any file operation is going to this python program | |||
'''remember.c''' | |||
To increase storage, update: | |||
saved_data_max | |||
saved_data_order | |||
Debugging kernel code is difficult, but kernel can be watched while running | |||
===In Class=== | |||
<pre> | |||
Lecture 17 | |||
---------- | |||
Filesystems | |||
What is it? | |||
Well, what's a device file? | |||
- struct file_operations which gives functions for file operations | |||
(read, write, etc) | |||
filesystems just take this idea further | |||
When you access a file | |||
- process does a file-related system call | |||
- kernel calls generic system call function for requested operation | |||
- kernel calls filesystem-specific function for requested operation | |||
e.g. | |||
- process does a write system call on file on an ext4 filesystem | |||
- kernel runs generic write syscall code | |||
- kernel runs ext4 write code | |||
filesystems are abstractions of file and directory-related operations | |||
- allow many instantiations that we call filesystems | |||
filesystems talk to rest of the kernel through a virtual filesystem interface | |||
(VFS) | |||
Sidebar: | |||
Microkernels vs monolithic kernels | |||
Why microkernels | |||
- better security: if a service crashes, it is just a process | |||
- easier development: servers are just a process | |||
Why monolithic kernels | |||
- faster | |||
- you depend on it anyway, so security benefits are illusory | |||
</pre> |
Latest revision as of 08:47, 23 November 2018
Video
The video from the lecture on November 9, 2018 is now available.
Notes
alloc pages -> frames -> page
'FUSE' - filesystem in userspace interface for userspace programs to export a filesystem to the LINUX kernel
What is a filesystem from a kernel?
- A device file (character device, block device), depends on file operations.
- A filesystem is just this scaled up, it should not only include these file operations, but it should also include operations for directories, and mounting, unmounting.
If you can define those, you can make a filesystem
FUSE makes a "special" filesystem that allows you to create other filesystems
Filesystems are abstractions of file operations Filesystems talk to rest of the kernel through a virtual filesystem interface
Are all userspace filesystem in memory? Standard ntfs implementation on most linux distributions, is written in userspace
Microkernals vs monolithic kernal microkernel: everything turns into a process monolithic kernel: it's faster
memoryll.py
mounting dev/fuse
any file operation is going to this python program
remember.c To increase storage, update: saved_data_max saved_data_order
Debugging kernel code is difficult, but kernel can be watched while running
In Class
Lecture 17 ---------- Filesystems What is it? Well, what's a device file? - struct file_operations which gives functions for file operations (read, write, etc) filesystems just take this idea further When you access a file - process does a file-related system call - kernel calls generic system call function for requested operation - kernel calls filesystem-specific function for requested operation e.g. - process does a write system call on file on an ext4 filesystem - kernel runs generic write syscall code - kernel runs ext4 write code filesystems are abstractions of file and directory-related operations - allow many instantiations that we call filesystems filesystems talk to rest of the kernel through a virtual filesystem interface (VFS) Sidebar: Microkernels vs monolithic kernels Why microkernels - better security: if a service crashes, it is just a process - easier development: servers are just a process Why monolithic kernels - faster - you depend on it anyway, so security benefits are illusory