On Fri, Mar 5, 2021 at 10:23 AM David Hildenbrand david@redhat.com wrote:
On 05.03.21 19:08, Suren Baghdasaryan wrote:
On Fri, Mar 5, 2021 at 9:52 AM David Hildenbrand david@redhat.com wrote:
On 05.03.21 18:45, Shakeel Butt wrote:
On Fri, Mar 5, 2021 at 9:37 AM David Hildenbrand david@redhat.com wrote:
On 04.03.21 01:03, Shakeel Butt wrote:
On Wed, Mar 3, 2021 at 3:34 PM Suren Baghdasaryan surenb@google.com wrote: > > On Wed, Mar 3, 2021 at 3:17 PM Shakeel Butt shakeelb@google.com wrote: >> >> On Wed, Mar 3, 2021 at 10:58 AM Suren Baghdasaryan surenb@google.com wrote: >>> >>> process_madvise currently requires ptrace attach capability. >>> PTRACE_MODE_ATTACH gives one process complete control over another >>> process. It effectively removes the security boundary between the >>> two processes (in one direction). Granting ptrace attach capability >>> even to a system process is considered dangerous since it creates an >>> attack surface. This severely limits the usage of this API. >>> The operations process_madvise can perform do not affect the correctness >>> of the operation of the target process; they only affect where the data >>> is physically located (and therefore, how fast it can be accessed). >>> What we want is the ability for one process to influence another process >>> in order to optimize performance across the entire system while leaving >>> the security boundary intact. >>> Replace PTRACE_MODE_ATTACH with a combination of PTRACE_MODE_READ >>> and CAP_SYS_NICE. PTRACE_MODE_READ to prevent leaking ASLR metadata >>> and CAP_SYS_NICE for influencing process performance. >>> >>> Cc: stable@vger.kernel.org # 5.10+ >>> Signed-off-by: Suren Baghdasaryan surenb@google.com >>> Reviewed-by: Kees Cook keescook@chromium.org >>> Acked-by: Minchan Kim minchan@kernel.org >>> Acked-by: David Rientjes rientjes@google.com >>> --- >>> changes in v3 >>> - Added Reviewed-by: Kees Cook keescook@chromium.org >>> - Created man page for process_madvise per Andrew's request: https://git.kernel.org/pub/scm/docs/man-pages/man-pages.git/commit/?id=a144f... >>> - cc'ed stable@vger.kernel.org # 5.10+ per Andrew's request >>> - cc'ed linux-security-module@vger.kernel.org per James Morris's request >>> >>> mm/madvise.c | 13 ++++++++++++- >>> 1 file changed, 12 insertions(+), 1 deletion(-) >>> >>> diff --git a/mm/madvise.c b/mm/madvise.c >>> index df692d2e35d4..01fef79ac761 100644 >>> --- a/mm/madvise.c >>> +++ b/mm/madvise.c >>> @@ -1198,12 +1198,22 @@ SYSCALL_DEFINE5(process_madvise, int, pidfd, const struct iovec __user *, vec, >>> goto release_task; >>> } >>> >>> - mm = mm_access(task, PTRACE_MODE_ATTACH_FSCREDS); >>> + /* Require PTRACE_MODE_READ to avoid leaking ASLR metadata. */ >>> + mm = mm_access(task, PTRACE_MODE_READ_FSCREDS); >>> if (IS_ERR_OR_NULL(mm)) { >>> ret = IS_ERR(mm) ? PTR_ERR(mm) : -ESRCH; >>> goto release_task; >>> } >>> >>> + /* >>> + * Require CAP_SYS_NICE for influencing process performance. Note that >>> + * only non-destructive hints are currently supported. >> >> How is non-destructive defined? Is MADV_DONTNEED non-destructive? > > Non-destructive in this context means the data is not lost and can be > recovered. I follow the logic described in > https://lwn.net/Articles/794704/ where Minchan was introducing > MADV_COLD and MADV_PAGEOUT as non-destructive versions of MADV_FREE > and MADV_DONTNEED. Following that logic, MADV_FREE and MADV_DONTNEED > would be considered destructive hints. > Note that process_madvise_behavior_valid() allows only MADV_COLD and > MADV_PAGEOUT at the moment, which are both non-destructive. >
There is a plan to support MADV_DONTNEED for this syscall. Do we need to change these access checks again with that support?
Eh, I absolutely don't think letting another process discard memory in another process' address space is a good idea. The target process can observe that easily and might even run into real issues.
What's the use case?
Userspace oom reaper. Please look at https://lore.kernel.org/linux-api/20201014183943.GA1489464@google.com/T/
Thanks, somehow I missed that (not that it really changed my opinion on the approach while skimming over the discussion :) will have a more detailed look)
The latest version of that patchset is: https://lore.kernel.org/patchwork/patch/1344419/ Yeah, memory reaping is a special case when we are operating on a dying process to speed up the release of its memory. I don't know if for that particular case we need to make the checks stricter. It's a dying process anyway and the data is being destroyed. Allowing to speed up that process probably can still use CAP_SYS_NICE.
I know, unrelated discussion (sorry, I don't have above thread in my archive anymore due to automatic cleanups ...) , but introducing MADV_DONTEED on a remote processes, having to tweak range logic because we always want to apply it to the whole MM, just to speed up memory reaping sounds like completely abusing madvise()/process_madvise() to me.
You want different semantics than MADV_DONTNEED. You want different semantics than madvise.
Simple example: mlock()ed pages in the target process. MADV_DONTNEED would choke on that. For the use case of reaping, you certainly don't care.
I am not sure if process_madvise() is the right interface to enforce discarding of all target memory.
Not to mention that MADV_FREE doesn't make any sense IMHO for memory reaping ... no to mention exposing this via process_madvise().
Yeah, that was the last comment from Christoph Hellwig on https://lore.kernel.org/patchwork/patch/1344418/ I'll be rethinking the whole approach. Previously I proposed couple different approaches that would make reaping a part of the kill by adding a new flag for pidfd_send_signal: https://lore.kernel.org/patchwork/patch/1338196/ https://lore.kernel.org/patchwork/patch/1060407/ but maybe a separate syscall for reaping is indeed the right way to go...
-- Thanks,
David / dhildenb