Date:   Fri, 20 Aug 2021 09:36:12 -0500
From:   ebiederm@...ssion.com (Eric W. Biederman)
To:     David Hildenbrand <david@...hat.com>
Cc:     Linus Torvalds <torvalds@...ux-foundation.org>,
        Linux Kernel Mailing List <linux-kernel@...r.kernel.org>,
        Andrew Morton <akpm@...ux-foundation.org>,
        Thomas Gleixner <tglx@...utronix.de>,
        Ingo Molnar <mingo@...hat.com>, Borislav Petkov <bp@...en8.de>,
        "H. Peter Anvin" <hpa@...or.com>,
        Alexander Viro <viro@...iv.linux.org.uk>,
        Alexey Dobriyan <adobriyan@...il.com>,
        Steven Rostedt <rostedt@...dmis.org>,
        Peter Zijlstra <peterz@...radead.org>,
        Arnaldo Carvalho de Melo <acme@...nel.org>,
        Mark Rutland <mark.rutland@....com>,
        Alexander Shishkin <alexander.shishkin@...ux.intel.com>,
        Jiri Olsa <jolsa@...hat.com>,
        Namhyung Kim <namhyung@...nel.org>,
        Petr Mladek <pmladek@...e.com>,
        Sergey Senozhatsky <sergey.senozhatsky@...il.com>,
        Andy Shevchenko <andriy.shevchenko@...ux.intel.com>,
        Rasmus Villemoes <linux@...musvillemoes.dk>,
        Kees Cook <keescook@...omium.org>,
        Greg Ungerer <gerg@...ux-m68k.org>,
        Geert Uytterhoeven <geert@...ux-m68k.org>,
        Mike Rapoport <rppt@...nel.org>,
        Vlastimil Babka <vbabka@...e.cz>,
        Vincenzo Frascino <vincenzo.frascino@....com>,
        Chinwen Chang <chinwen.chang@...iatek.com>,
        Catalin Marinas <catalin.marinas@....com>,
        "Matthew Wilcox \(Oracle\)" <willy@...radead.org>,
        Huang Ying <ying.huang@...el.com>,
        Jann Horn <jannh@...gle.com>, Feng Tang <feng.tang@...el.com>,
        Kevin Brodsky <Kevin.Brodsky@....com>,
        Michael Ellerman <mpe@...erman.id.au>,
        Shawn Anastasio <shawn@...stas.io>,
        Steven Price <steven.price@....com>,
        Nicholas Piggin <npiggin@...il.com>,
        Christian Brauner <christian.brauner@...ntu.com>,
        Jens Axboe <axboe@...nel.dk>,
        Gabriel Krisman Bertazi <krisman@...labora.com>,
        Peter Xu <peterx@...hat.com>,
        Suren Baghdasaryan <surenb@...gle.com>,
        Shakeel Butt <shakeelb@...gle.com>,
        Marco Elver <elver@...gle.com>,
        Daniel Jordan <daniel.m.jordan@...cle.com>,
        Nicolas Viennot <Nicolas.Viennot@...sigma.com>,
        Thomas Cedeno <thomascedeno@...gle.com>,
        Michal Hocko <mhocko@...e.com>,
        Miklos Szeredi <miklos@...redi.hu>,
        Chengguang Xu <cgxu519@...ernel.net>,
        Christian König 
        <ckoenig.leichtzumerken@...il.com>,
        Florian Weimer <fweimer@...hat.com>,
        David Laight <David.Laight@...lab.com>,
        linux-unionfs@...r.kernel.org,
        Linux API <linux-api@...r.kernel.org>,
        the arch/x86 maintainers <x86@...nel.org>,
        linux-fsdevel <linux-fsdevel@...r.kernel.org>,
        Linux-MM <linux-mm@...ck.org>
Subject: Re: [PATCH v2 2/7] kernel/fork: factor out replacing the current MM exe_file

David Hildenbrand <david@...hat.com> writes:

> On 19.08.21 22:51, Linus Torvalds wrote:
>> So I like this series.
>>
>> However, logically, I think this part in replace_mm_exe_file() no
>> longer makes sense:
>>
>> On Mon, Aug 16, 2021 at 12:50 PM David Hildenbrand <david@...hat.com> wrote:
>>>
>>> +       /* Forbid mm->exe_file change if old file still mapped. */
>>> +       old_exe_file = get_mm_exe_file(mm);
>>> +       if (old_exe_file) {
>>> +               mmap_read_lock(mm);
>>> +               for (vma = mm->mmap; vma && !ret; vma = vma->vm_next) {
>>> +                       if (!vma->vm_file)
>>> +                               continue;
>>> +                       if (path_equal(&vma->vm_file->f_path,
>>> +                                      &old_exe_file->f_path))
>>> +                               ret = -EBUSY;
>>> +               }
>>> +               mmap_read_unlock(mm);
>>> +               fput(old_exe_file);
>>> +               if (ret)
>>> +                       return ret;
>>> +       }
>>
>> and should just be removed.
>>
>> NOTE! I think it makes sense within the context of this patch (where
>> you just move code around), but that it should then be removed in the
>> next patch that does that "always deny write access to current MM
>> exe_file" thing.
>>
>> I just quoted it in the context of this patch, since the next patch
>> doesn't actually show this code any more.
>>
>> In the *old* model - where the ETXTBUSY was about the mmap() of the
>> file - the above tests make sense.
>>
>> But in the new model, walking the mappings just doesn't seem to be a
>> sensible operation any more. The mappings simply aren't what ETXTBUSY
>> is about in the new world order, and so doing that mapping walk seems
>> nonsensical.
>>
>> Hmm?
>
> I think this is somewhat another kind of "stop user space trying
> to do stupid things" thingy, not necessarily glued to ETXTBUSY:
> don't allow replacing exe_file if that very file is still mapped
> and consequently eventually still in use by the application.
>
> I don't think it necessarily has many things to do with ETXTBUSY:
> we only check if there is a VMA mapping that file, not that it's
> a VM_DENYWRITE mapping.
>
> That code originates from
>
> commit 4229fb1dc6843c49a14bb098719f8a696cdc44f8
> Author: Konstantin Khlebnikov <khlebnikov@...nvz.org>
> Date:   Wed Jul 11 14:02:11 2012 -0700
>
>     c/r: prctl: less paranoid prctl_set_mm_exe_file()
>
>     "no other files mapped" requirement from my previous patch (c/r: prctl:
>     update prctl_set_mm_exe_file() after mm->num_exe_file_vmas removal) is too
>     paranoid, it forbids operation even if there mapped one shared-anon vma.
>       Let's check that current mm->exe_file already unmapped, in this case
>     exe_file symlink already outdated and its changing is reasonable.
>
>
> The statement "exe_file symlink already outdated and its
> changing is reasonable" somewhat makes sense.
>
>
> Long story short, I think this check somehow makes a bit of sense, but
> we wouldn't lose too much if we drop it -- just another sanity check.
>
> Your call :)

There has been quite a bit of conversation of the years about how bad is
it to allow changing /proc/self/exe as some userspace depends on it.

I think this check is there to keep from changing /proc/self/exe
arbitrarily.

Maybe it is all completely silly and we should not care about the code
that thinks /proc/self/exe is a reliable measure of anything, but short
of that I think we should either keep the code or put in some careful
thought as to which restrictions make sense when changing
/proc/self/exe.

Eric

Powered by blists - more mailing lists