Re: Re: Re: Re: Re: [PATCH 1/3] switch_creds: Syscall to switch creds for file server ops

[Jim Lieb <jlieb@panasas.com>]

On Friday, November 01, 2013 22:24:12 Tetsuo Handa wrote:
> Jim Lieb wrote:
> > Subsequent uses look like:
> > 	use_creds(cached fd);
> > 
> > followed by
> > 
> > 	open/creat/mknod/write
> > 
> > followed by
> > 
> > 	use_creds(-1);
> 
> Are you aware that calling commit_creds() is prohibitted between
> override_creds() and revert_creds() ?
> 
> If the caller does some operation that calls commit_creds() (like
> example below), the kernel triggers BUG().

Yes, I do.  I caught this in an early pass.  I only use override_creds() and 
revert_creds().  
> 
> ---------- example module start ----------
> #include <linux/module.h>
> #include <linux/cred.h>
> #include <linux/fs.h>
> #include <linux/file.h>
> 
> static int __init test_init(void)
> {
>         { /* switch_creds() syscall */
>                 struct fd f = fdget(0);
>                 if (!f.file)
>                         return -EBADF;
>                 put_cred(override_creds(f.file->f_cred));
>                 fdput(f);
>         }
>         { /* something that calls commit_creds() */
>                 struct cred *cred = prepare_creds();
>                 if (cred)
>                         commit_creds(cred);
>         }
>         return 0;
> }
> 
> static void test_exit(void)
> {
> }
> 
> module_init(test_init);
> module_exit(test_exit);
> MODULE_LICENSE("GPL");
> ---------- example module end ----------
> 
> Since nobody can guarantee that the caller of switch_creds() never does
> some operation that calls commit_creds(), I don't think switch_creds()
> based on override_creds() will work.

-- 
Jim Lieb
Linux Systems Engineer
Panasas Inc.

"If ease of use was the only requirement, we would all be riding tricycles"
- Douglas Engelbart 1925–2013