Skip to contentSkip to conversation

Add interface xdm_run_domain

PATCH In review Sugar, David 3 participants 7 messages Apr 4, 2019 – Apr 8, 2019

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 Add interface xdm_run_domain

This interface is needed when using the various script options in lightdm (others?) configuration file. I'm using with 'display-setup-script' and 'greeter-setup-script', but I suspect the same would be useful for 'display-stopped-script', 'session-cleanup-script' and others. In these cases lightdm will be executing a configured script to perform some action during the specified phase of the greeter/session. But without this interface there is no way to transition into a new domain to execute the action.

---
 policy/modules/services/xserver.if | 31 ++++++++++++++++++++++++++++++
 1 file changed, 31 insertions(+)
policy/modules/services/xserver.if +31 -0
index 854f3b90..2faef71d 100644
@@ -734,6 +734,37 @@ interface(`xserver_relabel_console_pipes',`
 	allow $1 xconsole_device_t:fifo_file { getattr relabelfrom relabelto };
 ')
 
+########################################
+## <summary>
+##	Allow display manager to execute the specified program in
+##	the specified domain.
+## </summary>
+## <desc>
+##	<p>
+##	This is a interface to support the lightdm *-setup-script
+##	options.  This will allow the command run by lightdm to
+##	be run in a domain other than xdm_t.
+##	</p>
+## </desc>
+## <param name="domain">
+##	<summary>
+##	Domain to execute in.
+##	</summary>
+## </param>
+## <param name="entry_file">
+##	<summary>
+##	Domain entry point file.
+##	</summary>
+## </param>
+#
+interface(`xdm_run_domain',`
+	gen_require(`
+		type xdm_t;
+	')
+
+	domtrans_pattern(xdm_t,$2,$1)
+')
+
 ########################################
 ## <summary>
 ##	Use file descriptors for xdm.
Signature · 1 line
2.20.1
#1 lore raw

PATCH 1/1 Resolve some denials related to sending journal messages

type=AVC msg=audit(1554324562.840:159): avc:  denied  { sendto } for  pid=7277 comm="systemd-backlig" path="/run/systemd/journal/socket" scontext=system_u:system_r:systemd_backlight_t:s0 tcontext=system_u:system_r:kernel_t:s0 tclass=unix_dgram_socket permissive=1
type=AVC msg=audit(1554324271.863:245): avc:  denied  { sendto } for  pid=7421 comm="systemd-user-se" path="/run/systemd/journal/socket" scontext=system_u:system_r:systemd_sessions_t:s0 tcontext=system_u:system_r:kernel_t:s0 tclass=unix_dgram_socket permissive=0
type=AVC msg=audit(1554324635.844:313): avc:  denied  { sendto } for  pid=7744 comm="systemd-cryptse" path="/run/systemd/journal/socket" scontext=system_u:system_r:lvm_t:s0 tcontext=system_u:system_r:kernel_t:s0 tclass=unix_dgram_socket permissive=1
---
 policy/modules/system/lvm.te     | 1 +
 policy/modules/system/systemd.te | 4 ++++
 2 files changed, 5 insertions(+)
policy/modules/system/lvm.te +1 -0
index ec3b4a18..aea199d4 100644
@@ -222,6 +222,7 @@ filetrans_pattern(lvm_t, lvm_etc_t, lvm_metadata_t, file)
 files_etc_filetrans(lvm_t, lvm_metadata_t, file)
 files_search_mnt(lvm_t)
 
+kernel_dgram_send(lvm_t)
 kernel_request_load_module(lvm_t)
 kernel_get_sysvipc_info(lvm_t)
 kernel_read_system_state(lvm_t)
policy/modules/system/systemd.te +4 -0
index f6455f6f..541117a9 100644
@@ -228,6 +228,8 @@ dev_rw_sysfs(systemd_backlight_t)
 # for udev.conf
 files_read_etc_files(systemd_backlight_t)
 
+kernel_dgram_send(systemd_backlight_t)
+
 # for /run/udev/data/+backlight*
 udev_read_pid_files(systemd_backlight_t)
 
@@ -981,6 +983,8 @@ allow systemd_sessions_t self:process setfscreate;
 allow systemd_sessions_t systemd_sessions_var_run_t:file manage_file_perms;
 files_pid_filetrans(systemd_sessions_t, systemd_sessions_var_run_t, file)
 
+kernel_dgram_send(systemd_sessions_t)
+
 selinux_get_enforce_mode(systemd_sessions_t)
 selinux_get_fs_mount(systemd_sessions_t)
 
Signature · 1 line
2.20.1
#2 lore raw

Re: [PATCH] Resolve some denials related to sending journal messages

Quoted 3 lines · David
type=AVC msg=audit(1554324562.840:159): avc:  denied  { sendto } for  pid=7277 comm="systemd-backlig" path="/run/systemd/journal/socket" scontext=system_u:system_r:systemd_backlight_t:s0 tcontext=system_u:system_r:kernel_t:s0 tclass=unix_dgram_socket permissive=1
type=AVC msg=audit(1554324271.863:245): avc:  denied  { sendto } for  pid=7421 comm="systemd-user-se" path="/run/systemd/journal/socket" scontext=system_u:system_r:systemd_sessions_t:s0 tcontext=system_u:system_r:kernel_t:s0 tclass=unix_dgram_socket permissive=0
type=AVC msg=audit(1554324635.844:313): avc:  denied  { sendto } for  pid=7744 comm="systemd-cryptse" path="/run/systemd/journal/socket" scontext=system_u:system_r:lvm_t:s0 tcontext=system_u:system_r:kernel_t:s0 tclass=unix_dgram_socket permissive=1

I think we've reached the limit of the explicit kernel_dgram_send() and it's time to reevaluate putting this in logging_send_syslog_msg().

Quoted 41 lines
---
  policy/modules/system/lvm.te     | 1 +
  policy/modules/system/systemd.te | 4 ++++
  2 files changed, 5 insertions(+)
policy/modules/system/lvm.te +1 -0
index ec3b4a18..aea199d4 100644
@@ -222,6 +222,7 @@ filetrans_pattern(lvm_t, lvm_etc_t, lvm_metadata_t, file)
  files_etc_filetrans(lvm_t, lvm_metadata_t, file)
  files_search_mnt(lvm_t)
 
+kernel_dgram_send(lvm_t)
  kernel_request_load_module(lvm_t)
  kernel_get_sysvipc_info(lvm_t)
  kernel_read_system_state(lvm_t)
policy/modules/system/systemd.te +4 -0
index f6455f6f..541117a9 100644
@@ -228,6 +228,8 @@ dev_rw_sysfs(systemd_backlight_t)
  # for udev.conf
  files_read_etc_files(systemd_backlight_t)
 
+kernel_dgram_send(systemd_backlight_t)
+
  # for /run/udev/data/+backlight*
  udev_read_pid_files(systemd_backlight_t)
 
@@ -981,6 +983,8 @@ allow systemd_sessions_t self:process setfscreate;
  allow systemd_sessions_t systemd_sessions_var_run_t:file manage_file_perms;
  files_pid_filetrans(systemd_sessions_t, systemd_sessions_var_run_t, file)
 
+kernel_dgram_send(systemd_sessions_t)
+
  selinux_get_enforce_mode(systemd_sessions_t)
  selinux_get_fs_mount(systemd_sessions_t)
 
Signature · 1 line
Chris PeBenito
#3 lore raw

Re: [PATCH] Resolve some denials related to sending journal messages

Quoted 17 lines · Chris PeBenito
Quoted 7 lines · David

type=AVC msg=audit(1554324562.840:159): avc:  denied  { sendto } for pid=7277 comm="systemd-backlig" path="/run/systemd/journal/socket" scontext=system_u:system_r:systemd_backlight_t:s0 tcontext=system_u:system_r:kernel_t:s0 tclass=unix_dgram_socket permissive=1
type=AVC msg=audit(1554324271.863:245): avc:  denied  { sendto } for pid=7421 comm="systemd-user-se" path="/run/systemd/journal/socket"

Quoted 8 lines

scontext=system_u:system_r:systemd_sessions_t:s0 tcontext=system_u:system_r:kernel_t:s0 tclass=unix_dgram_socket permissive=0
type=AVC msg=audit(1554324635.844:313): avc:  denied  { sendto } for pid=7744 comm="systemd-cryptse" path="/run/systemd/journal/socket" scontext=system_u:system_r:lvm_t:s0 tcontext=system_u:system_r:kernel_t:s0 tclass=unix_dgram_socket permissive=1

Quoted 3 lines

I think we've reached the limit of the explicit kernel_dgram_send() and it's time to reevaluate putting this in logging_send_syslog_msg().

That makes sense.  I will submit a patch after I test a bit.  But it looks like everywhere (except the systemd module) that uses kernel_dgram_send() already has logging_send_syslog_msg() so once kernel_dgram_send() is added to logging_send_syslog_msg(), kernel_dgram_send() can removed from those domains.

The exceptions are:

1) In the systemd module there isn't much use of logging_send_syslog_msg().  For this case I will add logging_send_syslog_msg() as required.

2) In the interface init_daemon_domain(), it uses kernel_dgram_send() directly, there it probably doesn't need to use logging_send_syslog_msg(). This may be able to go away as modules might already use logging now, I will try removing and see what breaks.

Quoted 43 lines
Quoted 41 lines

---
  policy/modules/system/lvm.te     | 1 +
  policy/modules/system/systemd.te | 4 ++++

  2 files changed, 5 insertions(+)
policy/modules/system/lvm.te +0 -0
index ec3b4a18..aea199d4 100644
@@ -222,6 +222,7 @@ filetrans_pattern(lvm_t, lvm_etc_t, 
 lvm_metadata_t, file)
   files_etc_filetrans(lvm_t, lvm_metadata_t, file)
   files_search_mnt(lvm_t)
   +kernel_dgram_send(lvm_t)
   kernel_request_load_module(lvm_t)
   kernel_get_sysvipc_info(lvm_t)

  kernel_read_system_state(lvm_t)

policy/modules/system/systemd.te +1 -0
index f6455f6f..541117a9 100644
@@ -228,6 +228,8 @@ dev_rw_sysfs(systemd_backlight_t)
   # for udev.conf
   files_read_etc_files(systemd_backlight_t)
   +kernel_dgram_send(systemd_backlight_t)
+
   # for /run/udev/data/+backlight*
   udev_read_pid_files(systemd_backlight_t)
   @@ -981,6 +983,8 @@ allow systemd_sessions_t self:process setfscreate;

  allow systemd_sessions_t systemd_sessions_var_run_t:file manage_file_perms;
  files_pid_filetrans(systemd_sessions_t, systemd_sessions_var_run_t, file)
  +kernel_dgram_send(systemd_sessions_t)
+
  selinux_get_enforce_mode(systemd_sessions_t) selinux_get_fs_mount(systemd_sessions_t)

#4 lore raw

Re: [PATCH] Resolve some denials related to sending journal messages

Quoted 27 lines · David
Quoted 17 lines · Chris PeBenito
Quoted 7 lines · David

type=AVC msg=audit(1554324562.840:159): avc:  denied  { sendto } for pid=7277 comm="systemd-backlig" path="/run/systemd/journal/socket" scontext=system_u:system_r:systemd_backlight_t:s0 tcontext=system_u:system_r:kernel_t:s0 tclass=unix_dgram_socket permissive=1
type=AVC msg=audit(1554324271.863:245): avc:  denied  { sendto } for pid=7421 comm="systemd-user-se" path="/run/systemd/journal/socket"

Quoted 8 lines

scontext=system_u:system_r:systemd_sessions_t:s0 tcontext=system_u:system_r:kernel_t:s0 tclass=unix_dgram_socket permissive=0
type=AVC msg=audit(1554324635.844:313): avc:  denied  { sendto } for pid=7744 comm="systemd-cryptse" path="/run/systemd/journal/socket" scontext=system_u:system_r:lvm_t:s0 tcontext=system_u:system_r:kernel_t:s0 tclass=unix_dgram_socket permissive=1

Quoted 3 lines

I think we've reached the limit of the explicit kernel_dgram_send() and it's time to reevaluate putting this in logging_send_syslog_msg().

That makes sense.  I will submit a patch after I test a bit.  But it looks like everywhere (except the systemd module) that uses kernel_dgram_send() already has logging_send_syslog_msg() so once kernel_dgram_send() is added to logging_send_syslog_msg(), kernel_dgram_send() can removed from those domains.

Quoted 5 lines

The exceptions are:

1) In the systemd module there isn't much use of logging_send_syslog_msg().  For this case I will add logging_send_syslog_msg() as required.

see:

https://github.com/SELinuxProject/refpolicy/blob/master/policy/modules/system/systemd.te#L410

So "systemd_log_parse_environment()" implies "logging_send_syslog_msg()"

Quoted 49 lines

2) In the interface init_daemon_domain(), it uses kernel_dgram_send() directly, there it probably doesn't need to use logging_send_syslog_msg(). This may be able to go away as modules might already use logging now, I will try removing and see what breaks.

Quoted 43 lines
Quoted 41 lines

---
  policy/modules/system/lvm.te     | 1 +
  policy/modules/system/systemd.te | 4 ++++

  2 files changed, 5 insertions(+)
policy/modules/system/lvm.te +0 -0
index ec3b4a18..aea199d4 100644
@@ -222,6 +222,7 @@ filetrans_pattern(lvm_t, lvm_etc_t, 
 lvm_metadata_t, file)
   files_etc_filetrans(lvm_t, lvm_metadata_t, file)
   files_search_mnt(lvm_t)
   +kernel_dgram_send(lvm_t)
   kernel_request_load_module(lvm_t)
   kernel_get_sysvipc_info(lvm_t)

  kernel_read_system_state(lvm_t)

policy/modules/system/systemd.te +1 -0
index f6455f6f..541117a9 100644
@@ -228,6 +228,8 @@ dev_rw_sysfs(systemd_backlight_t)
   # for udev.conf
   files_read_etc_files(systemd_backlight_t)
   +kernel_dgram_send(systemd_backlight_t)
+
   # for /run/udev/data/+backlight*
   udev_read_pid_files(systemd_backlight_t)
   @@ -981,6 +983,8 @@ allow systemd_sessions_t self:process setfscreate;

  allow systemd_sessions_t systemd_sessions_var_run_t:file manage_file_perms;
  files_pid_filetrans(systemd_sessions_t, systemd_sessions_var_run_t, file)
  +kernel_dgram_send(systemd_sessions_t)
+
  selinux_get_enforce_mode(systemd_sessions_t) selinux_get_fs_mount(systemd_sessions_t)

Signature · 3 lines
Key fingerprint = 5F4D 3CDB D3F8 3652 FBD8 02D5 3B6C 5F1D 2C7B 6B02
https://sks-keyservers.net/pks/lookup?op=get&search=0x3B6C5F1D2C7B6B02
Dominick Grift
#5 lore raw

Re: [PATCH] Resolve some denials related to sending journal messages

Quoted 36 lines · Dominick Grift
Quoted 27 lines · David
Quoted 17 lines · Chris PeBenito
Quoted 7 lines · David

type=AVC msg=audit(1554324562.840:159): avc:  denied  { sendto } for pid=7277 comm="systemd-backlig" path="/run/systemd/journal/socket" scontext=system_u:system_r:systemd_backlight_t:s0 tcontext=system_u:system_r:kernel_t:s0 tclass=unix_dgram_socket permissive=1
type=AVC msg=audit(1554324271.863:245): avc:  denied  { sendto } for pid=7421 comm="systemd-user-se" path="/run/systemd/journal/socket"

Quoted 8 lines

scontext=system_u:system_r:systemd_sessions_t:s0 tcontext=system_u:system_r:kernel_t:s0 tclass=unix_dgram_socket permissive=0
type=AVC msg=audit(1554324635.844:313): avc:  denied  { sendto } for pid=7744 comm="systemd-cryptse" path="/run/systemd/journal/socket" scontext=system_u:system_r:lvm_t:s0 tcontext=system_u:system_r:kernel_t:s0 tclass=unix_dgram_socket permissive=1

Quoted 3 lines

I think we've reached the limit of the explicit kernel_dgram_send() and it's time to reevaluate putting this in logging_send_syslog_msg().

That makes sense.  I will submit a patch after I test a bit.  But it looks like everywhere (except the systemd module) that uses kernel_dgram_send() already has logging_send_syslog_msg() so once kernel_dgram_send() is added to logging_send_syslog_msg(), kernel_dgram_send() can removed from those domains.

Quoted 5 lines

The exceptions are:

1) In the systemd module there isn't much use of logging_send_syslog_msg().  For this case I will add logging_send_syslog_msg() as required.

Quoted 6 lines

see:

https://github.com/SELinuxProject/refpolicy/blob/master/policy/modules/system/systemd.te#L410

So "systemd_log_parse_environment()" implies "logging_send_syslog_msg()"

Thanks for that, I hadn't noticed.

Quoted 6 lines
Quoted 6 lines

2) In the interface init_daemon_domain(), it uses kernel_dgram_send() directly, there it probably doesn't need to use logging_send_syslog_msg(). This may be able to go away as modules might already use logging now, I will try removing and see what breaks.

Initial testing is showing this can be removed with no negative side effect (at least for the domains I'm using). I will continue testing before submitting a patch.

Quoted 44 lines
Quoted 43 lines
Quoted 43 lines
Quoted 41 lines
---
   policy/modules/system/lvm.te     | 1 +
   policy/modules/system/systemd.te | 4 ++++
   2 files changed, 5 insertions(+)
policy/modules/system/lvm.te +0 -0
index ec3b4a18..aea199d4 100644
@@ -222,6 +222,7 @@ filetrans_pattern(lvm_t, lvm_etc_t,
 lvm_metadata_t, file)
   files_etc_filetrans(lvm_t, lvm_metadata_t, file)
   files_search_mnt(lvm_t)
   +kernel_dgram_send(lvm_t)
   kernel_request_load_module(lvm_t)
   kernel_get_sysvipc_info(lvm_t)
   kernel_read_system_state(lvm_t)
policy/modules/system/systemd.te +1 -0
index f6455f6f..541117a9 100644
@@ -228,6 +228,8 @@ dev_rw_sysfs(systemd_backlight_t)
   # for udev.conf
   files_read_etc_files(systemd_backlight_t)
   +kernel_dgram_send(systemd_backlight_t)
+
   # for /run/udev/data/+backlight*
   udev_read_pid_files(systemd_backlight_t)
   @@ -981,6 +983,8 @@ allow systemd_sessions_t self:process setfscreate;
   allow systemd_sessions_t systemd_sessions_var_run_t:file

manage_file_perms;
  files_pid_filetrans(systemd_sessions_t, systemd_sessions_var_run_t,
file)
  +kernel_dgram_send(systemd_sessions_t)
+
  selinux_get_enforce_mode(systemd_sessions_t) selinux_get_fs_mount(systemd_sessions_t)

#6 lore raw
In reply to Sugar, David
Quoted 7 lines · David

This interface is needed when using the various script options in lightdm (others?) configuration file. I'm using with 'display-setup-script' and 'greeter-setup-script', but I suspect the same would be useful for 'display-stopped-script', 'session-cleanup-script' and others. In these cases lightdm will be executing a configured script to perform some action during the specified phase of the greeter/session. But without this interface there is no way to transition into a new domain to execute the action.

I'm reluctant to add this "reversed" interface. In other cases, the domain (init, cron, inetd, etc.) is legitimately transitioning to many domains. I don't see that here.

Quoted 48 lines
---
  policy/modules/services/xserver.if | 31 ++++++++++++++++++++++++++++++
  1 file changed, 31 insertions(+)
policy/modules/services/xserver.if +31 -0
index 854f3b90..2faef71d 100644
@@ -734,6 +734,37 @@ interface(`xserver_relabel_console_pipes',`
  	allow $1 xconsole_device_t:fifo_file { getattr relabelfrom relabelto };
  ')
 
+########################################
+## <summary>
+##	Allow display manager to execute the specified program in
+##	the specified domain.
+## </summary>
+## <desc>
+##	<p>
+##	This is a interface to support the lightdm *-setup-script
+##	options.  This will allow the command run by lightdm to
+##	be run in a domain other than xdm_t.
+##	</p>
+## </desc>
+## <param name="domain">
+##	<summary>
+##	Domain to execute in.
+##	</summary>
+## </param>
+## <param name="entry_file">
+##	<summary>
+##	Domain entry point file.
+##	</summary>
+## </param>
+#
+interface(`xdm_run_domain',`
+	gen_require(`
+		type xdm_t;
+	')
+
+	domtrans_pattern(xdm_t,$2,$1)
+')
+
  ########################################
  ## <summary>
  ##	Use file descriptors for xdm.
Signature · 1 line
Chris PeBenito