Bitbake-type build question

[Steve Grubb]

On Wednesday, March 11, 2020 6:01:54 PM EDT Aaron Biver wrote:
> I'm attempting to cross-compile the 2.8 maintenance branch with petalinux,
> which uses bitbake yocto under the hood to build it, and I run it on a
> system V target which is a 64 bit arm architecture.  I'm seeing some errors
> when I attempt to load rules for system some calls (like open, creat,
> rename, link, etc) with auditctl, and I'm not sure how critical these
> errors are.  This is my first encounter with linux auditing, and I'll
> apologize in advance because I'm sure this is something easy, but I could
> not find a way to search the mailing list archives.
> 
> These are some of my errors from auditctl's output:
> > Syscall name unknown: creat
> > There was an error in line 14 of /etc/audit/audit.rules
> > Syscall name unknown: link
> > There was an error in line 15 of /etc/audit/audit.rules
> > Syscall name unknown: open
> > There was an error in line 16 of /etc/audit/audit.rules
> > Syscall name unknown: open
> > There was an error in line 17 of /etc/audit/audit.rules
> > Syscall name unknown: rename
> > There was an error in line 19 of /etc/audit/audit.rules
> > Syscall name unknown: rename
> > There was an error in line 20 of /etc/audit/audit.rules
> 
> Those lines are below in an excerpt from my rules file (with line numbers
> prepended for easy reading):
>  13 -w /etc/hostname -p wa -k system-locale
>  14 -a always,exit -F arch=b64 -S
> creat,link,mknod,mkdir,symlink,mknodat,linkat,symlinkat -F exit=-EACCES -F
> key=creation
>  15 -a always,exit -F arch=b64 -S link,mkdir,symlink,mkdirat -F exit=-EPERM
> -F key=creation
>  16 -a always,exit -F arch=b64 -S open,openat,open_by_handle_at -F
> exit=-EACCES -F key=open
>  17 -a always,exit -F arch=b64 -S open,openat,open_by_handle_at -F
> exit=-EPERM -F key=open
>  18 -a always,exit -F arch=b64 -S close -F exit=-EIO -F key=close
>  19 -a always,exit -F arch=b64 -S rename -S renameat -S truncate -S chmod
> -S setxattr -S lsetxattr -S removexattr -S lremovexattr -F exit=-EACCES -F
> key=mods
>  20 -a always,exit -F arch=b64 -S rename -S renameat -S truncate -S chmod
> -S setxattr -S lsetxattr -S removexattr -S lremovexattr -F exit=-EPERM -F
> key=mods
> 
> The startup script and rules I'm repurposing had rules for open, create,
> link, unlink, rename, and others, and it seems like those would be part of
> a reasonable auditing security plan, but these system calls do not appear
> to be auditable on my system.

If I were to venture a guess, I'd say that the ABI was created after glibc 
started using openat, renameat, linkat, etc. So, there was never a need to 
implement the older "insecure" syscalls.


> My target platform is a 64 bit arm architecture.   I have a bitbake recipe
> which uses  the --with-aarch64.
> I build the kernel with auditing support (CONFIG_AUDIT and others like it),
> and I pass the audit=1 argument on the kernel boot command line.
> 
> Are there other incantations I am missing?  Any help would be greatly
> appreciated.

You can always strace a program to see what syscalls its using. Also, 
ausyscall is handy for telling you if a syscall exists. I think you can drop 
the missing syscalls from your rules and you'll be fine.

-Steve