Re: [PATCH 1/2] module: allow loading module from fd
[Posted October 9, 2012 by mkerrisk]
| From: |
| Rusty Russell <rusty-AT-rustcorp.com.au> |
| To: |
| Kees Cook <keescook-AT-chromium.org>, linux-kernel-AT-vger.kernel.org,
"Linus Torvalds" <torvalds-AT-linux-foundation.org> |
| Subject: |
| Re: [PATCH 1/2] module: allow loading module from fd |
| Date: |
| Thu, 06 Sep 2012 10:44:08 +0930 |
| Message-ID: |
| <87y5kohs0v.fsf@rustcorp.com.au> |
| Cc: |
| Serge Hallyn <serge.hallyn-AT-canonical.com>,
James Morris <james.l.morris-AT-oracle.com>,
Al Viro <viro-AT-zeniv.linux.org.uk>,
Eric Paris <eparis-AT-redhat.com>,
Kees Cook <keescook-AT-chromium.org>,
Jiri Kosina <jkosina-AT-suse.cz>,
linux-security-module-AT-vger.kernel.org |
| Archive-link: |
| Article, Thread
|
Kees Cook <keescook@chromium.org> writes:
> This changes the init_module syscall so that when the first argument
> (blob address) is NULL, the second argument is used as a file descriptor
> to the module (instead of length). The third argument (module arguments)
> remains unchanged.
Do you know why Linus hates ioctls? It's because they encourage
casual introduction of new ABIs.
I see you managed the same feat with a different system call. It avoids
the agony of debate a new system call would entail. Very clever.
Now do it properly.
Rusty.
(
Log in to post comments)
