From: Masami Hiramatsu <mhiramat@kernel.org>
To: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Steven Rostedt <rostedt@goodmis.org>,
Andy Lutomirski <luto@amacapital.net>,
Linux List Kernel Mailing <linux-kernel@vger.kernel.org>,
Ingo Molnar <mingo@kernel.org>,
Andrew Morton <akpm@linux-foundation.org>,
stable <stable@vger.kernel.org>,
Changbin Du <changbin.du@gmail.com>, Jann Horn <jannh@google.com>,
Kees Cook <keescook@chromium.org>,
Andy Lutomirski <luto@kernel.org>
Subject: Re: [PATCH 1/2 v2] kprobe: Do not use uaccess functions to access kernel memory that can fault
Date: Sat, 23 Feb 2019 12:47:46 +0900 [thread overview]
Message-ID: <20190223124746.d021973004c7c892c3b3fde1@kernel.org> (raw)
In-Reply-To: <CAHk-=whNf_n1WXWW+ugAVeL5ZK0GcEP3cTYocju1nS85VtMjjQ@mail.gmail.com>
On Fri, 22 Feb 2019 09:43:14 -0800
Linus Torvalds <torvalds@linux-foundation.org> wrote:
> On Fri, Feb 22, 2019 at 12:35 AM Masami Hiramatsu <mhiramat@kernel.org> wrote:
> >
> > Or, can we do this?
> >
> > long __probe_user_read(void *dst, const void *src, size_t size)
> > {
>
> Add a
>
> if (!access_ok(src, size))
> ret = -EFAULT;
> else {
> .. do the pagefault_disable() etc ..
> }
Since kprobes handler runs in IRQ context, we can not use access_ok() in it.
(only on x86 + CONFIG_DEBUG_ATOMIC_SLEEP=y)
In arch/x86/include/asm/uaccess.h:
#define access_ok(addr, size) \
({ \
WARN_ON_IN_IRQ(); \
likely(!__range_not_ok(addr, size, user_addr_max())); \
})
Do we need acccess_ok_inatomic()?
BTW, it seems a bit strange that this WARN_ON_IN_IRQ() is only in x86
access_ok() implementation, since CONFIG_DEBUG_ATOMIC_SLEEP(which defines
WARN_ON_IN_IRQ) doesn't depend on x86, and access_ok() is widely used
in kernel. I think it would be better that each arch provides __access_ok()
and include/linux/uaccess.h provides access_ok() with WARN_ON_IN_IRQ().
> to after the "set_fs()", and it looks good to me. Make it clear that
> yes, this works _only_ for user reads.
>
> Adn that makes all the games with "kernel_uaccess_faults_ok"
> pointless, so you can just remove them.
OK.
>
> (note that the "access_ok()" has to come after we've done "set_fs()",
> because it takes the address limit from that).
>
> Also, since normally we'd expect that we already have USER_DS, it
> might be worthwhile to do this with a wrapper, something along the
> lines of
>
> mm_segment_t old_fs = get_fs();
>
> if (segment_eq(old_fs, USER_DS))
> return __normal_probe_user_read();
> set_fs(USER_DS);
> ret = __normal_probe_user_read();
> set_fs(old_fs);
> return ret;
>
> and have that __normal_probe_user_read() just do
>
> if (!access_ok(src, size))
> return -EFAULT;
> pagefault_disable();
> ret = __copy_from_user_inatomic(dst, ...);
> pagefault_enable();
> return ret ? -EFAULT : 0;
>
> which looks more obvious.
OK.
>
> Also, I would suggest that you just make the argument type be "const
> void __user *", since the whole point is that this takes a user
> pointer, and nothing else.
Ah, right.
> Then we should still probably fix up "__probe_kernel_read()" to not
> allow user accesses. The easiest way to do that is actually likely to
> use the "unsafe_get_user()" functions *without* doing a
> uaccess_begin(), which will mean that modern CPU's will simply fault
> on a kernel access to user space.
Or, use __chk_user_ptr(ptr) to check it?
Thank you,
>
> The nice thing about that is that usually developers will have access
> to exactly those modern boxes, so the people who notice that it
> doesn't work are the right people.
>
> Alternatively, we should just make it be architecture-specific, so
> that architectures can decide "this address cannot be a kernel
> address" and refuse to do it.
>
> Linus
--
Masami Hiramatsu <mhiramat@kernel.org>
next prev parent reply other threads:[~2019-02-23 3:47 UTC|newest]
Thread overview: 85+ messages / expand[flat|nested] mbox.gz Atom feed top
2019-02-15 17:47 [PATCH 0/2 v2] [GIT PULL (take two)] tracing: Two more fixes Steven Rostedt
2019-02-15 17:47 ` [PATCH 1/2 v2] kprobe: Do not use uaccess functions to access kernel memory that can fault Steven Rostedt
2019-02-15 17:55 ` Linus Torvalds
2019-02-15 22:15 ` Steven Rostedt
2019-02-15 23:49 ` Andy Lutomirski
2019-02-16 0:19 ` Steven Rostedt
2019-02-16 1:32 ` Andy Lutomirski
2019-02-16 2:08 ` Steven Rostedt
2019-02-16 2:14 ` Andy Lutomirski
2019-02-16 2:21 ` Steven Rostedt
2019-02-18 17:58 ` Linus Torvalds
2019-02-18 18:23 ` Linus Torvalds
2019-02-19 16:18 ` Steven Rostedt
2019-02-19 18:43 ` Linus Torvalds
2019-02-19 19:03 ` Steven Rostedt
2019-02-20 8:10 ` Masami Hiramatsu
2019-02-20 13:57 ` Jann Horn
2019-02-20 14:47 ` Steven Rostedt
2019-02-20 15:08 ` Masami Hiramatsu
2019-02-20 14:49 ` Steven Rostedt
2019-02-20 16:04 ` Masami Hiramatsu
2019-02-20 16:42 ` Steven Rostedt
2019-02-21 7:37 ` Masami Hiramatsu
2019-02-22 8:27 ` Masami Hiramatsu
2019-02-22 8:35 ` Masami Hiramatsu
2019-02-22 17:43 ` Linus Torvalds
2019-02-22 17:48 ` Andy Lutomirski
2019-02-22 18:28 ` Linus Torvalds
2019-02-22 19:52 ` Andy Lutomirski
2019-02-22 19:27 ` Alexei Starovoitov
2019-02-22 19:30 ` Steven Rostedt
2019-02-22 19:34 ` Alexei Starovoitov
2019-02-22 19:39 ` Steven Rostedt
2019-02-22 19:55 ` Andy Lutomirski
2019-02-22 21:43 ` Jann Horn
2019-02-22 22:08 ` Nadav Amit
2019-02-22 22:17 ` Jann Horn
2019-02-22 22:21 ` Nadav Amit
2019-02-22 22:39 ` Nadav Amit
2019-02-22 23:02 ` Jann Horn
2019-02-22 23:22 ` Nadav Amit
2019-02-22 23:59 ` Andy Lutomirski
2019-02-23 0:03 ` Alexei Starovoitov
2019-02-23 0:15 ` Nadav Amit
2019-02-24 19:35 ` Andy Lutomirski
2019-02-25 13:36 ` Masami Hiramatsu
2019-02-22 21:20 ` Linus Torvalds
2019-02-22 21:38 ` David Miller
2019-02-22 21:59 ` Linus Torvalds
2019-02-22 22:51 ` Alexei Starovoitov
2019-02-22 23:11 ` Jann Horn
2019-02-22 23:16 ` David Miller
2019-02-22 23:16 ` Linus Torvalds
2019-02-22 23:56 ` Alexei Starovoitov
2019-02-23 0:08 ` Linus Torvalds
2019-02-23 2:28 ` Alexei Starovoitov
2019-02-23 2:32 ` Linus Torvalds
2019-02-23 3:02 ` Steven Rostedt
2019-02-23 4:51 ` Masami Hiramatsu
2019-02-26 3:57 ` Christoph Hellwig
2019-02-26 15:24 ` Joel Fernandes
2019-02-28 12:29 ` Masami Hiramatsu
2019-02-28 15:18 ` Joel Fernandes
2019-02-23 3:47 ` Masami Hiramatsu [this message]
2019-02-24 0:44 ` Steven Rostedt
2019-02-24 4:38 ` Andy Lutomirski
2019-02-24 15:17 ` Masami Hiramatsu
2019-02-24 17:26 ` Linus Torvalds
2019-02-25 2:40 ` Masami Hiramatsu
2019-02-25 4:49 ` Andy Lutomirski
2019-02-25 8:09 ` Masami Hiramatsu
2019-02-25 16:40 ` Steven Rostedt
2019-02-26 1:35 ` Masami Hiramatsu
2019-02-25 8:33 ` Peter Zijlstra
2019-02-25 14:52 ` Peter Zijlstra
2019-02-25 16:48 ` Kees Cook
2019-02-25 16:58 ` Andy Lutomirski
2019-02-25 17:07 ` Kees Cook
2019-02-21 7:52 ` Masami Hiramatsu
2019-02-21 14:36 ` Steven Rostedt
2019-02-21 15:58 ` Masami Hiramatsu
2019-02-21 16:16 ` Masami Hiramatsu
2019-02-21 16:32 ` Steven Rostedt
2019-02-23 14:48 ` Masami Hiramatsu
2019-02-15 17:47 ` [PATCH 2/2 v2] tracing: Fix number of entries in trace header Steven Rostedt
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20190223124746.d021973004c7c892c3b3fde1@kernel.org \
--to=mhiramat@kernel.org \
--cc=akpm@linux-foundation.org \
--cc=changbin.du@gmail.com \
--cc=jannh@google.com \
--cc=keescook@chromium.org \
--cc=linux-kernel@vger.kernel.org \
--cc=luto@amacapital.net \
--cc=luto@kernel.org \
--cc=mingo@kernel.org \
--cc=rostedt@goodmis.org \
--cc=stable@vger.kernel.org \
--cc=torvalds@linux-foundation.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).