hsiehkaofeng / linux Goto Github PK

Linux kernel
============

There are several guides for kernel developers and users. These guides can
be rendered in a number of formats, like HTML and PDF. Please read
Documentation/admin-guide/README.rst first.

In order to build the documentation, use ``make htmldocs`` or
``make pdfdocs``.  The formatted documentation can also be read online at:

    https://www.kernel.org/doc/html/latest/

There are various text files in the Documentation/ subdirectory,
several of them using the Restructured Text markup notation.

Please read the Documentation/process/changes.rst file, as it contains the
requirements for building and running the kernel, and information about
the problems which may result by upgrading your kernel.


============
Editor : Kao-Feng Hsieh

This project is based on linux 5.9.10.
In this project, in order to reduce TLB overhead, I aim to make different 
processes share their ASID. I mainly focus on containered processes, 
which are run by Docker.

In the current X86 system, the kernel only utilizes 6 ASIDs in TLB. Each 
ASID has a corresponding ctx_id, which is a unique identifier for an address 
space(mm_struct). Everytime when a process is asking for an ASID, the kernel 
will first iterate through the current ASIDs in the TLB and see if there is 
an ASID’s corresponding ctx_id that matches the process’s ctx_id. If there 
it is, return the ASID to the process. Otherwise, return an ASID and flush 
the TLB. However, different containered processes run the same application may 
use different address space. That means they utilize different ASIDs although 
they run the same application. In this scenario, it could cause TLB overhead 
and invalidate TLB.

**tlbstate**
|asid|ctx_id|
| --- | --- |
|1|0|
|2|3|
|3|5|
|4|7|
|5|8|
|6|10|

We can create another field in an address space, let’s call it “group_ctx_id” for 
now. Everytime when a process is being created, we assign a group_ctx_id to it. 

**tlbstate**
|asid|ctx_id|group_ctx_id|
| --- | --- | --- |
|1|0|1|
|2|3|2|
|3|5|5|
|4|7|0|
|5|8|0|
|6|10|0|

If we want to make those processes run the same application to share their ASID, we 
assign them with the same group_ctx_id. Next, we add an additional step to the kernel,
it needs to examine the group_ctx_id in TLB before the kernel assigns a new ASID to a 
process. Although there is no matched ctx_id for the process which is asking for an 
ASID, the kernel still needs to check if there is a matched group_ctx_id in TLB. In a 
word, it makes different processes can share their ASID even though they use different 
address spaces.


We leave group_ctx_id 0 for those normal processes, instead of containered processes. 

modified:
tlb_context of tlbstate in tlbflush.h(arch/x86/include/asm/tlbflush.h)
mm_struct in mm_types.h(include/linux/mm_types.h)
choose_new_asid() in tlb.c(arch/x86/mm/tlb.c)
fork.c (kernel/fork.c)
we also need to make shared processes have the same page table layout.
begin_new_exec() in exec.c(fs/exec.c)