Fix SELinux denials against target driver
Subsystems (2)
v2 cover Fix SELinux denials against target driver
Steps to reproduce:
1) install the ibacm, rdma-core and targetcli
2) service ibacm start (ignore the errors)
3) Look at the dmesg, you will see an error message like "db_root: cannot open: /etc/target"
4) Execute $ sudo ausearch -m AVC,USER_AVC -ts recent
type=AVC msg=audit(1707990698.893:610): avc: denied { read } for pid=26447 comm="systemd-modules" name="target" dev="dm-0" ino=973050 scontext=system_u:system_r:systemd_modules_load_t:s0 tcontext=system_u:object_r:targetd_etc_rw_t:s0 tclass=dir permissive=0
Fix inspired by commit 581dd69830341d299b0c097fc366097ab497d679
V2: fix a memory leak in the error path, add a patch to set
a freed pointer to NULL to avoid possible double frees
Maurizio Lombardi (2):
target: fix selinux error when systemd-modules loads the target module
target: set the xcopy_wq pointer to NULL after free.
drivers/target/target_core_configfs.c | 12 ++++++++++++ drivers/target/target_core_xcopy.c | 4 +++- 2 files changed, 15 insertions(+), 1 deletion(-)
Signature · 1 line
2.39.3
PATCH v2 1/2 target: fix selinux error when systemd-modules loads the target module
If the systemd-modules service loads the target module, the credentials of that userspace process will be used to validate the access to the target db directory.
selinux will prevent it, reporting an error like the following:
kernel: audit: type=1400 audit(1676301082.205:4): avc: denied { read } for pid=1020 comm="systemd-modules" name="target" dev="dm-3" ino=4657583 scontext=system_u:system_r:systemd_modules_load_t:s0 tcontext=system_u:object_r:targetd_etc_rw_t:s0 tclass=dir permissive=0
Fix the error by using the kernel credentials to access the db directory
---
v2: call target_xcopy_release_pt() to destroy the xcopy_wq in case of failure
drivers/target/target_core_configfs.c | 12 ++++++++++++ 1 file changed, 12 insertions(+)
drivers/target/target_core_configfs.c +12 -0
@@ -3656,6 +3656,8 @@ static int __init target_core_init_configfs(void) {struct configfs_subsystem *subsys = &target_core_fabrics;struct t10_alua_lu_gp *lu_gp; +struct cred *kern_cred; +const struct cred *old_cred;int ret; pr_debug("TARGET_CORE[0]: Loading Generic Kernel Storage"
@@ -3732,11 +3734,21 @@ static int __init target_core_init_configfs(void)if (ret <0 ) goto out; +/* We use the kernel credentials to access the target directory */ + kern_cred =prepare_kernel_cred (&init_task); +if (!kern_cred) { + ret = -ENOMEM ; + goto out; + } + old_cred =override_creds (kern_cred);target_init_dbroot (); +revert_creds (old_cred); +put_cred (kern_cred);return 0 ; out: +target_xcopy_release_pt ();configfs_unregister_subsystem (subsys);core_dev_release_virtual_lun0 ();rd_module_exit ();
Signature · 1 line
2.39.3
Re: [PATCH V2 1/2] target: fix selinux error when systemd-modules loads the target module
Quoted 5 lines · Maurizio Lombardi
If the systemd-modules service loads the target module, the credentials of that userspace process will be used to validate the access to the target db directory.
selinux will prevent it, reporting an error like the following:
Quoted 7 lines
kernel: audit: type=1400 audit(1676301082.205:4): avc: denied { read } for pid=1020 comm="systemd-modules" name="target" dev="dm-3" ino=4657583 scontext=system_u:system_r:systemd_modules_load_t:s0 tcontext=system_u:object_r:targetd_etc_rw_t:s0 tclass=dir permissive=0
Fix the error by using the kernel credentials to access the db directory
Do you need something similar for the pr related dirs/files or how does that work but not this?
Re: [PATCH V2 1/2] target: fix selinux error when systemd-modules loads the target module
čt 15. 2. 2024 v 17:44 odesílatel <[email protected]> napsal:
Quoted 3 lines
Do you need something similar for the pr related dirs/files or how does that work but not this?
I think that in that case it won't be necessary because the pr code is executed by a kernel thread that calls the execute_cmd() callback, not by a user process in kernel context,
but I will try and eventually I will report back the findings
Maurizio
Mike Christie
Reviewed-byRe: [PATCH V2 1/2] target: fix selinux error when systemd-modules loads the target module
Quoted 10 lines · Maurizio Lombardi
čt 15. 2. 2024 v 17:44 odesílatel <[email protected]> napsal:
Quoted 3 lines
Do you need something similar for the pr related dirs/files or how does that work but not this?
I think that in that case it won't be necessary because the pr code is executed by a kernel thread that calls the
execute_cmd()callback, not by a user process in kernel context,
but I will try and eventually I will report back the findings
Ignore my comment. I was thinking of something completely different and figured out I was wrong when I looked into it.
The target parts look ok:
I have no idea about the creds part. If the patch:
commit 581dd69830341d299b0c097fc366097ab497d679
Author: Thiébaud Weksteen <[email protected]> Date: Mon May 2 10:49:52 2022 +1000
firmware_loader: use kernel credentials when reading firmware
you referenced is the correct way to do it, then it looks ok.
PATCH v2 2/2 target: set the xcopy_wq pointer to NULL after free.
In reply to Maurizio LombardiDo not leave a dangling pointer after free.
--- drivers/target/target_core_xcopy.c | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-)
drivers/target/target_core_xcopy.c +3 -1
@@ -491,8 +491,10 @@ int target_xcopy_setup_pt(void)void target_xcopy_release_pt (void ) { -if (xcopy_wq) +if (xcopy_wq) {destroy_workqueue (xcopy_wq); + xcopy_wq =NULL ; + } }/*
Signature · 1 line
2.39.3
Re: [PATCH V2 2/2] target: set the xcopy_wq pointer to NULL after free.
Quoted 12 lines · Maurizio Lombardi
Do not leave a dangling pointer after free.
Signed-off-by Maurizio Lombardi <[email protected]>--- drivers/target/target_core_xcopy.c | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-)drivers/target/target_core_xcopy.c +0
-0
drivers/target/target_core_xcopy.c int target_xcopy_setup_pt(void) · line 499
Quoted 10 lines
drivers/target/target_core_xcopy.c +3
-1@@ -491,8 +491,10 @@ int target_xcopy_setup_pt(void)void target_xcopy_release_pt (void ) { -if (xcopy_wq) +if (xcopy_wq) {destroy_workqueue (xcopy_wq); + xcopy_wq =NULL ; + } }
Why do you need this? Isn't this only called when the module is unloaded?
We don't normally do this for that type of case in general. In the target code alone we have lots of places we don't do this.
Re: [PATCH V2 2/2] target: set the xcopy_wq pointer to NULL after free.
čt 15. 2. 2024 v 17:42 odesílatel <[email protected]> napsal:
Quoted 5 lines
Why do you need this? Isn't this only called when the module is unloaded?
We don't normally do this for that type of case in general. In the target code alone we have lots of places we don't do this.
I don't really need this, it's just a matter of personal preference, we can drop this one.
Maurizio
Martin K. Petersen
AppliedRe: (subset) [PATCH V2 0/2] Fix SELinux denials against target driver
In reply to Maurizio LombardiQuoted 8 lines · Maurizio Lombardi
Steps to reproduce:
1) install the ibacm, rdma-core and targetcli
2) service ibacm start (ignore the errors)
3) Look at the dmesg, you will see an error message like "db_root: cannot open:/etc/target"[...]
Applied to 6.9/scsi-fixes, thanks!
[1/2] target: fix selinux error when systemd-modules loads the target module
https://git.kernel.org/mkp/scsi/c/97a54ef596c3
Signature · 1 line
Martin K. Petersen Oracle Linux Engineering