Hello,
I am trying to understand how smack works (looking at
smack_inode_permission hook). I can see that that smack security
information is taken from cred->security and inode->i_security but how
did they get in there? Also, when does it happen? (for a task and for a
file)
--
Regards, Denis Obrezkov
On 12/31/2021 5:08 AM, Denis Obrezkov wrote:
Hello,
I am trying to understand how smack works (looking at
smack_inode_permission hook). I can see that that smack security
information is taken from cred->security and inode->i_security but how
did they get in there?
The LSM infrastructure (security/security.c) allocates cred and inode
security blobs. This allows multiple security modules to use them.
Also, when does it happen? (for a task and for a
file)
security_cred_alloc() and security_inode_alloc().
I am the Smack maintainer. Feel free to ask anything you'd
like here and to me directly.
The LSM infrastructure (security/security.c) allocates cred and inode
security blobs. This allows multiple security modules to use them.
quoted
Also, when does it happen? (for a task and for a
file)
security_cred_alloc() and security_inode_alloc().
I mean how is information from SMACK64EXEC and SMACK64 getting into
those blobs? Do I understand the sequence right:
First, both labels (SMACK64EXEC and SMACK64) are installed in
smack_inode_post_setxattr. Second, when we launch a program, there is a
hook smack_bprm_creds_for_exec that installs a security label from the
program file inode into the corresponding smack task structure. Third,
when the program tries to access a file, it is caught in the
smack_inode_permission.
I am also not sure what is happening in security_inode_alloc. Does it
just copy a pointer to a security structure of a current task?
I also can't find where security_cred_alloc is used. I found
security_cred_alloc_blank but it is called only from cred_alloc_blank
from cred.c (and I can't find from where the latter is called).
--
Regards, Denis Obrezkov
On 1/1/2022 1:34 PM, Denis Obrezkov wrote:
quoted
The LSM infrastructure (security/security.c) allocates cred and inode
security blobs. This allows multiple security modules to use them.
quoted
Also, when does it happen? (for a task and for a
file)
security_cred_alloc() and security_inode_alloc().
I mean how is information from SMACK64EXEC and SMACK64 getting into
those blobs? Do I understand the sequence right:
First, both labels (SMACK64EXEC and SMACK64) are installed in
smack_inode_post_setxattr. Second, when we launch a program, there is a
hook smack_bprm_creds_for_exec that installs a security label from the
program file inode into the corresponding smack task structure. Third,
when the program tries to access a file, it is caught in the
smack_inode_permission.
Seems right. Note that few programs use SMACK64EXEC, while
all files will have SMACK64.
I am also not sure what is happening in security_inode_alloc. Does it
just copy a pointer to a security structure of a current task?
Smack labels are stored on a list in the kernel. Once a label
is introduced (smk_import_entry()) it never gets forgotten. The
inode contains a pointer into this list.
I also can't find where security_cred_alloc is used. I found
security_cred_alloc_blank but it is called only from cred_alloc_blank
from cred.c (and I can't find from where the latter is called).