* [PATCH] mm: fix use-after-free in sys_remap_file_pages @ 2013-12-12 22:07 Kees Cook 2013-12-13 3:41 ` Rik van Riel 0 siblings, 1 reply; 5+ messages in thread From: Kees Cook @ 2013-12-12 22:07 UTC (permalink / raw) To: linux-kernel Cc: Andrew Morton, Michel Lespinasse, Rik van Riel, Cyrill Gorcunov, Hugh Dickins, linux-mm, PaX Team, Dmitry Vyukov From: PaX Team <pageexec@freemail.hu> http://lkml.org/lkml/2013/9/17/30 SyS_remap_file_pages() calls mmap_region(), which calls remove_vma_list(), which calls remove_vma(), which frees the vma. Later (after out label) SyS_remap_file_pages() accesses the freed vma in vm_flags = vma->vm_flags. Reported-by: Dmitry Vyukov <dvyukov@google.com> Signed-off-by: PaX Team <pageexec@freemail.hu> Signed-off-by: Kees Cook <keescook@chromium.org> Cc: stable@vger.kernel.org --- mm/fremap.c | 2 ++ 1 file changed, 2 insertions(+) diff --git a/mm/fremap.c b/mm/fremap.c index 5bff08147768..afad07b85ef2 100644 --- a/mm/fremap.c +++ b/mm/fremap.c @@ -218,6 +218,8 @@ get_write_lock: BUG_ON(addr != start); err = 0; } + vm_flags = vma->vm_flags; + vma = NULL; goto out; } mutex_lock(&mapping->i_mmap_mutex); -- 1.7.9.5 -- Kees Cook Chrome OS Security -- To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to majordomo@kvack.org. For more info on Linux MM, see: http://www.linux-mm.org/ . Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a> ^ permalink raw reply related [flat|nested] 5+ messages in thread
* Re: [PATCH] mm: fix use-after-free in sys_remap_file_pages 2013-12-12 22:07 [PATCH] mm: fix use-after-free in sys_remap_file_pages Kees Cook @ 2013-12-13 3:41 ` Rik van Riel 2013-12-13 6:24 ` Cyrill Gorcunov 2013-12-13 11:10 ` PaX Team 0 siblings, 2 replies; 5+ messages in thread From: Rik van Riel @ 2013-12-13 3:41 UTC (permalink / raw) To: Kees Cook Cc: linux-kernel, Andrew Morton, Michel Lespinasse, Cyrill Gorcunov, Hugh Dickins, linux-mm, PaX Team, Dmitry Vyukov On Thu, 12 Dec 2013 14:07:57 -0800 Kees Cook <keescook@chromium.org> wrote: > From: PaX Team <pageexec@freemail.hu> > > http://lkml.org/lkml/2013/9/17/30 > > SyS_remap_file_pages() calls mmap_region(), which calls remove_vma_list(), > which calls remove_vma(), which frees the vma. Later (after out label) > SyS_remap_file_pages() accesses the freed vma in vm_flags = vma->vm_flags. > index 5bff08147768..afad07b85ef2 100644 > --- a/mm/fremap.c > +++ b/mm/fremap.c > @@ -218,6 +218,8 @@ get_write_lock: > BUG_ON(addr != start); > err = 0; > } > + vm_flags = vma->vm_flags; > + vma = NULL; > goto out; If the vma has been freed by the time the code jumps to the out label (because it was freed by a function called from mmap_region), surely it will also already have been freed by the time this patch dereferences it? Also, setting vma = NULL to avoid the if (vma) branch at the out: label is unnecessarily obfuscated. Lets make things clear by documenting what is going on, and having a label after that dereference. Maybe something like this patch: ---8<--- Subject: [PATCH] mm: fix use-after-free in sys_remap_file_pages remap_file_pages calls mmap_region, which may merge the VMA with other existing VMAs, and free "vma". This can lead to a use-after-free bug. Avoid the bug by remembering vm_flags before calling mmap_region, and not trying to dereference vma later. Signed-off-by: Rik van Riel <riel@redhat.com> Reported-by: Dmitry Vyukov <dvyukov@google.com> Cc: stable@vger.kernel.org --- mm/fremap.c | 5 ++++- 1 file changed, 4 insertions(+), 1 deletion(-) diff --git a/mm/fremap.c b/mm/fremap.c index 87da359..d28ede5 100644 --- a/mm/fremap.c +++ b/mm/fremap.c @@ -203,6 +203,7 @@ SYSCALL_DEFINE5(remap_file_pages, unsigned long, start, unsigned long, size, if (mapping_cap_account_dirty(mapping)) { unsigned long addr; struct file *file = get_file(vma->vm_file); + vm_flags = vma->vm_flags; addr = mmap_region(file, start, size, vma->vm_flags, pgoff); @@ -213,7 +214,8 @@ SYSCALL_DEFINE5(remap_file_pages, unsigned long, start, unsigned long, size, BUG_ON(addr != start); err = 0; } - goto out; + /* mmap_region may have freed vma */ + goto out_freed; } mutex_lock(&mapping->i_mmap_mutex); flush_dcache_mmap_lock(mapping); @@ -248,6 +250,7 @@ SYSCALL_DEFINE5(remap_file_pages, unsigned long, start, unsigned long, size, out: if (vma) vm_flags = vma->vm_flags; +out_freed: if (likely(!has_write_lock)) up_read(&mm->mmap_sem); else -- To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to majordomo@kvack.org. For more info on Linux MM, see: http://www.linux-mm.org/ . Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a> ^ permalink raw reply related [flat|nested] 5+ messages in thread
* Re: [PATCH] mm: fix use-after-free in sys_remap_file_pages 2013-12-13 3:41 ` Rik van Riel @ 2013-12-13 6:24 ` Cyrill Gorcunov 2013-12-13 11:10 ` PaX Team 1 sibling, 0 replies; 5+ messages in thread From: Cyrill Gorcunov @ 2013-12-13 6:24 UTC (permalink / raw) To: Rik van Riel Cc: Kees Cook, linux-kernel, Andrew Morton, Michel Lespinasse, Hugh Dickins, linux-mm, PaX Team, Dmitry Vyukov On Thu, Dec 12, 2013 at 10:41:18PM -0500, Rik van Riel wrote: > > If the vma has been freed by the time the code jumps to the > out label (because it was freed by a function called from > mmap_region), surely it will also already have been freed > by the time this patch dereferences it? > > Also, setting vma = NULL to avoid the if (vma) branch at > the out: label is unnecessarily obfuscated. Lets make things > clear by documenting what is going on, and having a label > after that dereference. This patch is a bit easier to read, at least for me. And if I understand the code flow right, the issue is due to remap_file_pages -> mmap_region -> find_vma_links -> do_munmap. -- To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to majordomo@kvack.org. For more info on Linux MM, see: http://www.linux-mm.org/ . Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a> ^ permalink raw reply [flat|nested] 5+ messages in thread
* Re: [PATCH] mm: fix use-after-free in sys_remap_file_pages 2013-12-13 3:41 ` Rik van Riel 2013-12-13 6:24 ` Cyrill Gorcunov @ 2013-12-13 11:10 ` PaX Team 2013-12-13 14:43 ` [PATCH -v2] " Rik van Riel 1 sibling, 1 reply; 5+ messages in thread From: PaX Team @ 2013-12-13 11:10 UTC (permalink / raw) To: Kees Cook, Rik van Riel Cc: linux-kernel, Andrew Morton, Michel Lespinasse, Cyrill Gorcunov, Hugh Dickins, linux-mm, Dmitry Vyukov On 12 Dec 2013 at 22:41, Rik van Riel wrote: > If the vma has been freed by the time the code jumps to the > out label (because it was freed by a function called from > mmap_region), surely it will also already have been freed > by the time this patch dereferences it? oops, yes, i meant to save the flags away before mmap_region, no idea how i ended up with this ;). > Also, setting vma = NULL to avoid the if (vma) branch at > the out: label is unnecessarily obfuscated. Lets make things > clear by documenting what is going on, and having a label > after that dereference. on that note, how about this as well: > --- a/mm/fremap.c > +++ b/mm/fremap.c > @@ -203,6 +203,7 @@ SYSCALL_DEFINE5(remap_file_pages, unsigned long, start, unsigned long, size, > if (mapping_cap_account_dirty(mapping)) { > unsigned long addr; > struct file *file = get_file(vma->vm_file); > + vm_flags = vma->vm_flags; > > addr = mmap_region(file, start, size, > vma->vm_flags, pgoff); ^^^^^^^^^^^^^ pass in vm_flags instead of vma->vm_flags just to prevent someone from 'optimizing' away the read in the future? > @@ -213,7 +214,8 @@ SYSCALL_DEFINE5(remap_file_pages, unsigned long, start, unsigned long, size, > BUG_ON(addr != start); > err = 0; > } > - goto out; > + /* mmap_region may have freed vma */ > + goto out_freed; perhaps {copy,move} this comment above the previous hunk since that's where the relevant action is? cheers, PaX Team -- To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to majordomo@kvack.org. For more info on Linux MM, see: http://www.linux-mm.org/ . Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a> ^ permalink raw reply [flat|nested] 5+ messages in thread
* [PATCH -v2] mm: fix use-after-free in sys_remap_file_pages 2013-12-13 11:10 ` PaX Team @ 2013-12-13 14:43 ` Rik van Riel 0 siblings, 0 replies; 5+ messages in thread From: Rik van Riel @ 2013-12-13 14:43 UTC (permalink / raw) To: pageexec Cc: Kees Cook, linux-kernel, Andrew Morton, Michel Lespinasse, Cyrill Gorcunov, Hugh Dickins, linux-mm, Dmitry Vyukov On Fri, 13 Dec 2013 12:10:17 +0100 "PaX Team" <pageexec@freemail.hu> wrote: > pass in vm_flags instead of vma->vm_flags just to prevent someone > from 'optimizing' away the read in the future? In that case, we should probably also use ACCESS_ONCE, if only to be explicit about it. > perhaps {copy,move} this comment above the previous hunk since that's > where the relevant action is? See the new version below: ---8<--- Subject: mm: fix use-after-free in sys_remap_file_pages remap_file_pages calls mmap_region, which may merge the VMA with other existing VMAs, and free "vma". This can lead to a use-after-free bug. Avoid the bug by remembering vm_flags before calling mmap_region, and not trying to dereference vma later. Signed-off-by: Rik van Riel <riel@redhat.com> Reported-by: Dmitry Vyukov <dvyukov@google.com> Cc: stable@vger.kernel.org --- mm/fremap.c | 8 +++++--- 1 file changed, 5 insertions(+), 3 deletions(-) diff --git a/mm/fremap.c b/mm/fremap.c index 87da359..c85e2ec 100644 --- a/mm/fremap.c +++ b/mm/fremap.c @@ -203,9 +203,10 @@ SYSCALL_DEFINE5(remap_file_pages, unsigned long, start, unsigned long, size, if (mapping_cap_account_dirty(mapping)) { unsigned long addr; struct file *file = get_file(vma->vm_file); + /* mmap_region may free vma; grab the info now */ + vm_flags = ACCESS_ONCE(vma->vm_flags); - addr = mmap_region(file, start, size, - vma->vm_flags, pgoff); + addr = mmap_region(file, start, size, vm_flags, pgoff); fput(file); if (IS_ERR_VALUE(addr)) { err = addr; @@ -213,7 +214,7 @@ SYSCALL_DEFINE5(remap_file_pages, unsigned long, start, unsigned long, size, BUG_ON(addr != start); err = 0; } - goto out; + goto out_freed; } mutex_lock(&mapping->i_mmap_mutex); flush_dcache_mmap_lock(mapping); @@ -248,6 +249,7 @@ SYSCALL_DEFINE5(remap_file_pages, unsigned long, start, unsigned long, size, out: if (vma) vm_flags = vma->vm_flags; +out_freed: if (likely(!has_write_lock)) up_read(&mm->mmap_sem); else -- To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to majordomo@kvack.org. For more info on Linux MM, see: http://www.linux-mm.org/ . Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a> ^ permalink raw reply related [flat|nested] 5+ messages in thread
end of thread, other threads:[~2013-12-13 15:00 UTC | newest] Thread overview: 5+ messages (download: mbox.gz follow: Atom feed -- links below jump to the message on this page -- 2013-12-12 22:07 [PATCH] mm: fix use-after-free in sys_remap_file_pages Kees Cook 2013-12-13 3:41 ` Rik van Riel 2013-12-13 6:24 ` Cyrill Gorcunov 2013-12-13 11:10 ` PaX Team 2013-12-13 14:43 ` [PATCH -v2] " Rik van Riel
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).