| Message ID | 20200325111347.32553-1-will@kernel.org (mailing list archive) |
|---|---|
| State | Mainlined |
| Commit | b2a84de2a2deb76a6a51609845341f508c518c03 |
| Headers | show |
| Series | mm/mremap: Add comment explaining the untagging behaviour of mremap() | expand |
On Wed, Mar 25, 2020 at 11:13:46AM +0000, Will Deacon wrote: > Commit dcde237319e6 ("mm: Avoid creating virtual address aliases in > brk()/mmap()/mremap()") changed mremap() so that only the 'old' address > is untagged, leaving the 'new' address in the form it was passed from > userspace. This prevents the unexpected creation of aliasing virtual > mappings in userspace, but looks a bit odd when you read the code. > > Add a comment justifying the untagging behaviour in mremap(). > > Cc: Catalin Marinas <catalin.marinas@arm.com> > Reported-by: Linus Torvalds <torvalds@linux-foundation.org> > Signed-off-by: Will Deacon <will@kernel.org> Reviewed-by: Catalin Marinas <catalin.marinas@arm.com>
On Wed, Mar 25, 2020 at 4:13 AM Will Deacon <will@kernel.org> wrote: > > Add a comment justifying the untagging behaviour in mremap(). Ack. This makes that odd "do it on one address, not the other" thing clear. Otherwise I bet somebody would come along in a few years and go "Oh, people missed that other case". Linus
On Wed, Mar 25, 2020 at 11:13:46AM +0000, Will Deacon wrote: > Commit dcde237319e6 ("mm: Avoid creating virtual address aliases in > brk()/mmap()/mremap()") changed mremap() so that only the 'old' address > is untagged, leaving the 'new' address in the form it was passed from > userspace. This prevents the unexpected creation of aliasing virtual > mappings in userspace, but looks a bit odd when you read the code. > > Add a comment justifying the untagging behaviour in mremap(). > > Cc: Catalin Marinas <catalin.marinas@arm.com> > Reported-by: Linus Torvalds <torvalds@linux-foundation.org> > Signed-off-by: Will Deacon <will@kernel.org> I queued this patch via the arm64 tree (for 5.7-rc1). Thanks.
diff --git a/mm/mremap.c b/mm/mremap.c index af363063ea23..d28f08a36b96 100644 --- a/mm/mremap.c +++ b/mm/mremap.c @@ -606,6 +606,16 @@ SYSCALL_DEFINE5(mremap, unsigned long, addr, unsigned long, old_len, LIST_HEAD(uf_unmap_early); LIST_HEAD(uf_unmap); + /* + * There is a deliberate asymmetry here: we strip the pointer tag + * from the old address but leave the new address alone. This is + * for consistency with mmap(), where we prevent the creation of + * aliasing mappings in userspace by leaving the tag bits of the + * mapping address intact. A non-zero tag will cause the subsequent + * range checks to reject the address as invalid. + * + * See Documentation/arm64/tagged-address-abi.rst for more information. + */ addr = untagged_addr(addr); if (flags & ~(MREMAP_FIXED | MREMAP_MAYMOVE))
Commit dcde237319e6 ("mm: Avoid creating virtual address aliases in brk()/mmap()/mremap()") changed mremap() so that only the 'old' address is untagged, leaving the 'new' address in the form it was passed from userspace. This prevents the unexpected creation of aliasing virtual mappings in userspace, but looks a bit odd when you read the code. Add a comment justifying the untagging behaviour in mremap(). Cc: Catalin Marinas <catalin.marinas@arm.com> Reported-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Will Deacon <will@kernel.org> --- mm/mremap.c | 10 ++++++++++ 1 file changed, 10 insertions(+)