]> git.proxmox.com Git - mirror_ubuntu-eoan-kernel.git/commitdiff
mm/mincore.c: make mincore() more conservative
authorJiri Kosina <jkosina@suse.cz>
Tue, 14 May 2019 22:41:38 +0000 (15:41 -0700)
committerLinus Torvalds <torvalds@linux-foundation.org>
Wed, 15 May 2019 02:52:48 +0000 (19:52 -0700)
The semantics of what mincore() considers to be resident is not
completely clear, but Linux has always (since 2.3.52, which is when
mincore() was initially done) treated it as "page is available in page
cache".

That's potentially a problem, as that [in]directly exposes
meta-information about pagecache / memory mapping state even about
memory not strictly belonging to the process executing the syscall,
opening possibilities for sidechannel attacks.

Change the semantics of mincore() so that it only reveals pagecache
information for non-anonymous mappings that belog to files that the
calling process could (if it tried to) successfully open for writing;
otherwise we'd be including shared non-exclusive mappings, which

 - is the sidechannel

 - is not the usecase for mincore(), as that's primarily used for data,
   not (shared) text

[jkosina@suse.cz: v2]
Link: http://lkml.kernel.org/r/20190312141708.6652-2-vbabka@suse.cz
[mhocko@suse.com: restructure can_do_mincore() conditions]
Link: http://lkml.kernel.org/r/nycvar.YFH.7.76.1903062342020.19912@cbobk.fhfr.pm
Signed-off-by: Jiri Kosina <jkosina@suse.cz>
Signed-off-by: Vlastimil Babka <vbabka@suse.cz>
Acked-by: Josh Snyder <joshs@netflix.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Originally-by: Linus Torvalds <torvalds@linux-foundation.org>
Originally-by: Dominique Martinet <asmadeus@codewreck.org>
Cc: Andy Lutomirski <luto@amacapital.net>
Cc: Dave Chinner <david@fromorbit.com>
Cc: Kevin Easton <kevin@guarana.org>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Cyril Hrubis <chrubis@suse.cz>
Cc: Tejun Heo <tj@kernel.org>
Cc: Kirill A. Shutemov <kirill@shutemov.name>
Cc: Daniel Gruss <daniel@gruss.cc>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
mm/mincore.c

index 218099b5ed31d1e971d8d64b0951a31b8e310c01..c3f058bd0faf3e90857a762dd0828b621139ea5b 100644 (file)
@@ -169,6 +169,22 @@ out:
        return 0;
 }
 
+static inline bool can_do_mincore(struct vm_area_struct *vma)
+{
+       if (vma_is_anonymous(vma))
+               return true;
+       if (!vma->vm_file)
+               return false;
+       /*
+        * Reveal pagecache information only for non-anonymous mappings that
+        * correspond to the files the calling process could (if tried) open
+        * for writing; otherwise we'd be including shared non-exclusive
+        * mappings, which opens a side channel.
+        */
+       return inode_owner_or_capable(file_inode(vma->vm_file)) ||
+               inode_permission(file_inode(vma->vm_file), MAY_WRITE) == 0;
+}
+
 /*
  * Do a chunk of "sys_mincore()". We've already checked
  * all the arguments, we hold the mmap semaphore: we should
@@ -189,8 +205,13 @@ static long do_mincore(unsigned long addr, unsigned long pages, unsigned char *v
        vma = find_vma(current->mm, addr);
        if (!vma || addr < vma->vm_start)
                return -ENOMEM;
-       mincore_walk.mm = vma->vm_mm;
        end = min(vma->vm_end, addr + (pages << PAGE_SHIFT));
+       if (!can_do_mincore(vma)) {
+               unsigned long pages = DIV_ROUND_UP(end - addr, PAGE_SIZE);
+               memset(vec, 1, pages);
+               return pages;
+       }
+       mincore_walk.mm = vma->vm_mm;
        err = walk_page_range(addr, end, &mincore_walk);
        if (err < 0)
                return err;