init: in redhat derived distributions the kernel (systemd) dynamically transitions to init
PATCH 3/5 init: in redhat derived distributions the kernel (systemd) dynamically transitions to init
I used the kernel_dyntrans_to() to stay consistent with the kernel_domtrans_to. I am not sure why this was done this way
init: systemd wants to block suspend in distro_redhat
--- policy/modules/kernel/kernel.if | 53 +++++++++++++++++++++++++++++++++++++++++ policy/modules/system/init.fc | 2 ++ policy/modules/system/init.te | 23 ++++++++++++++++++ 3 files changed, 78 insertions(+)
policy/modules/kernel/kernel.if +53 -0
@@ -32,6 +32,25 @@ interface(`kernel_domtrans_to',` ######################################## ## <summary> +## Start userland processes by dynamically +## transitioning to the specified domain. +## </summary> +## <param name="domain"> +## <summary> +## The process type entered by kernel. +## </summary> +## </param> +# +interface(`kernel_dyntrans_to',` + gen_require(` + type kernel_t; + ') + + dyntrans_pattern(kernel_t, $1) +') + +######################################## +## <summary> ## Allows to start userland processes ## by transitioning to the specified domain, ## with a range transition.
@@ -71,6 +90,40 @@ interface(`kernel_ranged_domtrans_to',` ######################################## ## <summary> +## Start userland processes by dynamically +## transitioning to the specified domain, +## with a range transition. +## </summary> +## <param name="domain"> +## <summary> +## The process type entered by kernel. +## </summary> +## </param> +## <param name="range"> +## <summary> +## Range for the domain. +## </summary> +## </param> +# +interface(`kernel_ranged_dyntrans_to',` + gen_require(` + type kernel_t; + ') + + kernel_dyntrans_to($1) + + ifdef(`enable_mcs',` + range_transition kernel_t $1:process $2; + ') + + ifdef(`enable_mls',` + range_transition kernel_t $1:process $2; + mls_rangetrans_target($1) + ') +') + +######################################## +## <summary> ## Allows the kernel to mount filesystems on ## the specified directory type. ## </summary>
policy/modules/system/init.fc +2 -0
@@ -42,6 +42,8 @@ ifdef(`distro_gentoo', ` # /usr/bin/sepg_ctl -- gen_context(system_u:object_r:initrc_exec_t,s0) +/usr/lib/systemd/systemd -- gen_context(system_u:object_r:init_exec_t,s0) + /usr/libexec/dcc/start-.* -- gen_context(system_u:object_r:initrc_exec_t,s0) /usr/libexec/dcc/stop-.* -- gen_context(system_u:object_r:initrc_exec_t,s0)
policy/modules/system/init.te +23 -0
@@ -16,6 +16,13 @@ gen_require(` ## </desc> gen_tunable(init_upstart, false) +## <desc> +## <p> +## Enable support for systemd as the init program. +## </p> +## </desc> +gen_tunable(init_systemd, false) + # used for direct running of init scripts # by admin domains attribute direct_run_init;
@@ -88,10 +95,22 @@ ifdef(`distro_gentoo',` domain_entry_file(initrc_t, rc_exec_t) ') +ifdef(`distro_redhat',` + tunable_policy(`init_systemd',` + kernel_dyntrans_to(init_t) + ') +') + ifdef(`enable_mls',` kernel_ranged_domtrans_to(init_t, init_exec_t, s0 - mls_systemhigh) ') +ifdef(`enable_mls && distro_redhat',` + tunable_policy(`init_systemd',` + kernel_ranged_dyntrans_to(init_t, s0 - mls_systemhigh) + ') +') + ######################################## # # Init local policy
@@ -189,6 +208,10 @@ ifdef(`distro_redhat',` fs_read_tmpfs_symlinks(init_t) fs_rw_tmpfs_chr_files(init_t) fs_tmpfs_filetrans(init_t, initctl_t, fifo_file) + + tunable_policy(`init_systemd',` + allow init_t self:capability2 block_suspend; + ') ') tunable_policy(`init_upstart',`
Signature · 1 line
1.8.4.2
PATCH 3/5 init: in redhat derived distributions the kernel (systemd) dynamically transitions to init
Quoted 6 lines · Dominick Grift
I used the
kernel_dyntrans_to()to stay consistent with thekernel_domtrans_to. I am not sure why this was done this wayinit: systemd wants to block suspend in
distro_redhatSigned-off-by Dominick Grift <[email protected]>
Whoops these patches were sent from the wrong email address. git-send-email or dependency is broken in fedora so i resorted to mutt which was configured with this email address
Anyhow:
This might be a first step to supporting systemd.
Basically run the main component in init_t, and add the systemd specific policy in a boolean init_systemd
These changes also only apply to rh derived distros (at least for now)
debian is not using systemd currently but is using some systemd components
however these components are started by dbus system bus, and not by systemd like they are in fedora for example
So that needs special attention
The main message of this patch i guess is that i suggest we take advantage of Fedoras experience with systemd and follow suit as much as possible
Earlier on i was hoping we could do more but now i am of the opinion that this is probably best