danog/gllvm
1
0
Fork 0
mirror of https://github.com/danog/gllvm.git synced 2024-11-30 06:39:06 +01:00
Code Issues Packages Projects Releases Wiki Activity

Updated Readme to include the bitcode build

Browse Source
This commit is contained in:
paul-naert 2018-05-22 17:07:41 -07:00 committed by GitHub
parent d2c65b6564
commit 95e123dc93
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
1 changed files with 24 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

25
examples/linux-kernel/README.md
View File

@ -20,7 +20,7 @@ Vagrant.configure("2") do |config|
vb.memory = "4096"
vb.customize ["modifyvm", :id, "--ioapic", "on"]
vb.customize ["modifyvm", :id, "--memory", "4096"]
vb.customize ["modifyvm", :id, "--cpus", "4"]
vb.customize ["modifyvm", :id, "--cpus", "2"]
end
end
@ -43,6 +43,8 @@ sudo apt-get install -y flex bison bc libncurses5-dev
sudo apt-get install -y libelf-dev libssl-dev
echo ". /vagrant/bash_profile" >> /home/vagrant/.bashrc
bash /vagrant/init_script.sh
```
## Shell Settings
@ -147,3 +149,24 @@ You can also build from a [git clone using gllvm,](https://github.com/SRI-CSL/gl
or build from a [git clone using wllvm.](https://github.com/SRI-CSL/gllvm/blob/master/examples/linux-kernel/build_linux_wllvm_git.sh)
Though using a tarball is faster, and seemingly more reliable.
# Building a bootable kernel from bitcode
The init_script.sh script will build a bootable kernel from mostly bitcode (drivers and ext4 file system are currently not translated).
## Building the kernel with gclang
The init script will build the required folder architecture for the build and call build_linux_gllvm, only this time with a default configuration instead of tinyconfig.
## Building the kernel from bitcode
The copy.sh script will then extract the bitcode from the archives in the linux build folder, and copy them along with necessary object files (the files compiled straight from assembly will not emit a bitcode file).
It will then call the link command on those files and generate a vmlinux executable containing the kernel.
## Booting on the generated kernel
The gclang build of the kernel adds llvm_bc headers to most files, and those mess with the generation of a compressed bootable kernel.
We need to have a separate folder built form clang or gcc on which to finish the kernel build and install.
That is the purpose of the bootable-kernel script.
Finally, calling the install-kernel script will copy the new kernel into the clang generated folder and finish the build and install. Rebooting will be on the bitcode kernel.
NB: The generated kernel boots properly but it may be buggy. Most notably, I have experienced issues when shutting down and booting the machine a second time.