Mapping the delay slot emulation page as both writeable & executable
presents a security risk, in that if an exploit can write to & jump into
the page then it can be used as an easy way to execute arbitrary code.
Prevent this by mapping the page read-only for userland, and using
access_process_vm() with the FOLL_FORCE flag to write to it from
mips_dsemul().
This will likely be less efficient due to copy_to_user_page() performing
cache maintenance on a whole page, rather than a single line as in the
previous use of flush_cache_sigtramp(). However this delay slot
emulation code ought not to be running in any performance critical paths
anyway so this isn't really a problem, and we can probably do better in
copy_to_user_page() anyway in future.
A major advantage of this approach is that the fix is small & simple to
backport to stable kernels.
Reported-by: Andy Lutomirski <luto@kernel.org>
Signed-off-by: Paul Burton <paul.burton@mips.com>
Fixes: 432c6bacbd ("MIPS: Use per-mm page to execute branch delay slot instructions")
Cc: stable@vger.kernel.org # v4.8+
Cc: linux-mips@vger.kernel.org
Cc: linux-kernel@vger.kernel.org
Cc: Rich Felker <dalias@libc.org>
Cc: David Daney <david.daney@cavium.com>
When using the legacy mmap layout, for example triggered using ulimit -s
unlimited, get_unmapped_area() fills memory from bottom to top starting
from a fairly low address near TASK_UNMAPPED_BASE.
This placement is suboptimal if the user application wishes to allocate
large amounts of heap memory using the brk syscall. With the VDSO being
located low in the user's virtual address space, the amount of space
available for access using brk is limited much more than it was prior to
the introduction of the VDSO.
For example:
# ulimit -s unlimited; cat /proc/self/maps
00400000-004ec000 r-xp 00000000 08:00 71436 /usr/bin/coreutils
004fc000-004fd000 rwxp 000ec000 08:00 71436 /usr/bin/coreutils
004fd000-0050f000 rwxp 00000000 00:00 0
00cc3000-00ce4000 rwxp 00000000 00:00 0 [heap]
2ab96000-2ab98000 r--p 00000000 00:00 0 [vvar]
2ab98000-2ab99000 r-xp 00000000 00:00 0 [vdso]
2ab99000-2ab9d000 rwxp 00000000 00:00 0
...
Resolve this by adjusting STACK_TOP to reserve space for the VDSO &
providing an address hint to get_unmapped_area() causing it to use this
space even when using the legacy mmap layout.
We reserve enough space for the VDSO, plus 1MB or 256MB for 32 bit & 64
bit systems respectively within which we randomize the VDSO base
address. Previously this randomization was taken care of by the mmap
base address randomization performed by arch_mmap_rnd(). The 1MB & 256MB
sizes are somewhat arbitrary but chosen such that we have some
randomization without taking up too much of the user's virtual address
space, which is often in short supply for 32 bit systems.
With this the VDSO is always mapped at a high address, leaving lots of
space for statically linked programs to make use of brk:
# ulimit -s unlimited; cat /proc/self/maps
00400000-004ec000 r-xp 00000000 08:00 71436 /usr/bin/coreutils
004fc000-004fd000 rwxp 000ec000 08:00 71436 /usr/bin/coreutils
004fd000-0050f000 rwxp 00000000 00:00 0
00c28000-00c49000 rwxp 00000000 00:00 0 [heap]
...
7f67c000-7f69d000 rwxp 00000000 00:00 0 [stack]
7f7fc000-7f7fd000 rwxp 00000000 00:00 0
7fcf1000-7fcf3000 r--p 00000000 00:00 0 [vvar]
7fcf3000-7fcf4000 r-xp 00000000 00:00 0 [vdso]
Signed-off-by: Paul Burton <paul.burton@mips.com>
Reported-by: Huacai Chen <chenhc@lemote.com>
Fixes: ebb5e78cc6 ("MIPS: Initial implementation of a VDSO")
Cc: Huacai Chen <chenhc@lemote.com>
Cc: linux-mips@linux-mips.org
Cc: stable@vger.kernel.org # v4.4+
When a system suffers from dcache aliasing a user program may observe
stale VDSO data from an aliased cache line. Notably this can break the
expectation that clock_gettime(CLOCK_MONOTONIC, ...) is, as its name
suggests, monotonic.
In order to ensure that users observe updates to the VDSO data page as
intended, align the user mappings of the VDSO data page such that their
cache colouring matches that of the virtual address range which the
kernel will use to update the data page - typically its unmapped address
within kseg0.
This ensures that we don't introduce aliasing cache lines for the VDSO
data page, and therefore that userland will observe updates without
requiring cache invalidation.
Signed-off-by: Paul Burton <paul.burton@mips.com>
Reported-by: Hauke Mehrtens <hauke@hauke-m.de>
Reported-by: Rene Nielsen <rene.nielsen@microsemi.com>
Reported-by: Alexandre Belloni <alexandre.belloni@bootlin.com>
Fixes: ebb5e78cc6 ("MIPS: Initial implementation of a VDSO")
Patchwork: https://patchwork.linux-mips.org/patch/20344/
Tested-by: Alexandre Belloni <alexandre.belloni@bootlin.com>
Tested-by: Hauke Mehrtens <hauke@hauke-m.de>
Cc: James Hogan <jhogan@kernel.org>
Cc: linux-mips@linux-mips.org
Cc: stable@vger.kernel.org # v4.4+
We don't really need gic_get_usm_range() to abstract discovery of the
address of the GIC user-visible section now that we have access to its
base address globally.
Switch to calculating it ourselves, which will allow us to stop
requiring the irqchip driver to care about a counter exposed to userland
for use via the VDSO.
Signed-off-by: Paul Burton <paul.burton@imgtec.com>
Cc: Jason Cooper <jason@lakedaemon.net>
Cc: Marc Zyngier <marc.zyngier@arm.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: linux-mips@linux-mips.org
Patchwork: https://patchwork.linux-mips.org/patch/17040/
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
When a non-cooperative userfaultfd monitor copies pages in the
background, it may encounter regions that were already unmapped.
Addition of UFFD_EVENT_UNMAP allows the uffd monitor to track precisely
changes in the virtual memory layout.
Since there might be different uffd contexts for the affected VMAs, we
first should create a temporary representation for the unmap event for
each uffd context and then notify them one by one to the appropriate
userfault file descriptors.
The event notification occurs after the mmap_sem has been released.
[arnd@arndb.de: fix nommu build]
Link: http://lkml.kernel.org/r/20170203165141.3665284-1-arnd@arndb.de
[mhocko@suse.com: fix nommu build]
Link: http://lkml.kernel.org/r/20170202091503.GA22823@dhcp22.suse.cz
Link: http://lkml.kernel.org/r/1485542673-24387-3-git-send-email-rppt@linux.vnet.ibm.com
Signed-off-by: Mike Rapoport <rppt@linux.vnet.ibm.com>
Signed-off-by: Michal Hocko <mhocko@suse.com>
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Acked-by: Hillf Danton <hillf.zj@alibaba-inc.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: "Dr. David Alan Gilbert" <dgilbert@redhat.com>
Cc: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Pavel Emelyanov <xemul@virtuozzo.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
The page structures associated with the vDSO pages in the kernel image
are calculated using virt_to_page(), which uses __pa() under the hood to
find the pfn associated with the virtual address. The vDSO data pointers
however point to kernel symbols, so __pa_symbol() should really be used
instead.
Since there is no equivalent to virt_to_page() which uses __pa_symbol(),
fix init_vdso_image() to work directly with pfns, calculated with
__phys_to_pfn(__pa_symbol(...)).
This issue broke the Malta Enhanced Virtual Addressing (EVA)
configuration which has a non-default implementation of __pa_symbol().
This is because it uses a physical alias so that the kernel executes
from KSeg0 (VA 0x80000000 -> PA 0x00000000), while RAM is provided to
the kernel in the KUSeg range (VA 0x00000000 -> PA 0x80000000) which
uses the same underlying RAM.
Since there are no page structures associated with the low physical
address region, some arbitrary kernel memory would be interpreted as a
page structure for the vDSO pages and badness ensues.
Fixes: ebb5e78cc6 ("MIPS: Initial implementation of a VDSO")
Signed-off-by: James Hogan <james.hogan@imgtec.com>
Cc: Leonid Yegoshin <leonid.yegoshin@imgtec.com>
Cc: linux-mips@linux-mips.org
Cc: <stable@vger.kernel.org> # 4.4.x-
Patchwork: https://patchwork.linux-mips.org/patch/14229/
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
In some cases the kernel needs to execute an instruction from the delay
slot of an emulated branch instruction. These cases include:
- Emulated floating point branch instructions (bc1[ft]l?) for systems
which don't include an FPU, or upon which the kernel is run with the
"nofpu" parameter.
- MIPSr6 systems running binaries targeting older revisions of the
architecture, which may include branch instructions whose encodings
are no longer valid in MIPSr6.
Executing instructions from such delay slots is done by writing the
instruction to memory followed by a trap, as part of an "emuframe", and
executing it. This avoids the requirement of an emulator for the entire
MIPS instruction set. Prior to this patch such emuframes are written to
the user stack and executed from there.
This patch moves FP branch delay emuframes off of the user stack and
into a per-mm page. Allocating a page per-mm leaves userland with access
to only what it had access to previously, and compared to other
solutions is relatively simple.
When a thread requires a delay slot emulation, it is allocated a frame.
A thread may only have one frame allocated at any one time, since it may
only ever be executing one instruction at any one time. In order to
ensure that we can free up allocated frame later, its index is recorded
in struct thread_struct. In the typical case, after executing the delay
slot instruction we'll execute a break instruction with the BRK_MEMU
code. This traps back to the kernel & leads to a call to do_dsemulret
which frees the allocated frame & moves the user PC back to the
instruction that would have executed following the emulated branch.
In some cases the delay slot instruction may be invalid, such as a
branch, or may trigger an exception. In these cases the BRK_MEMU break
instruction will not be hit. In order to ensure that frames are freed
this patch introduces dsemul_thread_cleanup() and calls it to free any
allocated frame upon thread exit. If the instruction generated an
exception & leads to a signal being delivered to the thread, or indeed
if a signal simply happens to be delivered to the thread whilst it is
executing from the struct emuframe, then we need to take care to exit
the frame appropriately. This is done by either rolling back the user PC
to the branch or advancing it to the continuation PC prior to signal
delivery, using dsemul_thread_rollback(). If this were not done then a
sigreturn would return to the struct emuframe, and if that frame had
meanwhile been used in response to an emulated branch instruction within
the signal handler then we would execute the wrong user code.
Whilst a user could theoretically place something like a compact branch
to self in a delay slot and cause their thread to become stuck in an
infinite loop with the frame never being deallocated, this would:
- Only affect the users single process.
- Be architecturally invalid since there would be a branch in the
delay slot, which is forbidden.
- Be extremely unlikely to happen by mistake, and provide a program
with no more ability to harm the system than a simple infinite loop
would.
If a thread requires a delay slot emulation & no frame is available to
it (ie. the process has enough other threads that all frames are
currently in use) then the thread joins a waitqueue. It will sleep until
a frame is freed by another thread in the process.
Since we now know whether a thread has an allocated frame due to our
tracking of its index, the cookie field of struct emuframe is removed as
we can be more certain whether we have a valid frame. Since a thread may
only ever have a single frame at any given time, the epc field of struct
emuframe is also removed & the PC to continue from is instead stored in
struct thread_struct. Together these changes simplify & shrink struct
emuframe somewhat, allowing twice as many frames to fit into the page
allocated for them.
The primary benefit of this patch is that we are now free to mark the
user stack non-executable where that is possible.
Signed-off-by: Paul Burton <paul.burton@imgtec.com>
Cc: Leonid Yegoshin <leonid.yegoshin@imgtec.com>
Cc: Maciej Rozycki <maciej.rozycki@imgtec.com>
Cc: Faraz Shahbazker <faraz.shahbazker@imgtec.com>
Cc: Raghu Gandham <raghu.gandham@imgtec.com>
Cc: Matthew Fortune <matthew.fortune@imgtec.com>
Cc: linux-mips@linux-mips.org
Patchwork: https://patchwork.linux-mips.org/patch/13764/
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
most architectures are relying on mmap_sem for write in their
arch_setup_additional_pages. If the waiting task gets killed by the oom
killer it would block oom_reaper from asynchronous address space reclaim
and reduce the chances of timely OOM resolving. Wait for the lock in
the killable mode and return with EINTR if the task got killed while
waiting.
Signed-off-by: Michal Hocko <mhocko@suse.com>
Acked-by: Andy Lutomirski <luto@amacapital.net> [x86 vdso]
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Add user-mode implementations of gettimeofday() and clock_gettime() to
the VDSO. This is currently usable with 2 clocksources: the CP0 count
register, which is accessible to user-mode via RDHWR on R2 and later
cores, or the MIPS Global Interrupt Controller (GIC) timer, which
provides a "user-mode visible" section containing a mirror of its
counter registers. This section must be mapped into user memory, which
is done below the VDSO data page.
When a supported clocksource is not in use, the VDSO functions will
return -ENOSYS, which causes libc to fall back on the standard syscall
path.
When support for neither of these clocksources is compiled into the
kernel at all, the VDSO still provides clock_gettime(), as the coarse
realtime/monotonic clocks can still be implemented. However,
gettimeofday() is not provided in this case as nothing can be done
without a suitable clocksource. This causes the symbol lookup to fail
in libc and it will then always use the standard syscall path.
This patch includes a workaround for a bug in QEMU which results in
RDHWR on the CP0 count register always returning a constant (incorrect)
value. A fix for this has been submitted, and the workaround can be
removed after the fix has been in stable releases for a reasonable
amount of time.
A simple performance test which calls gettimeofday() 1000 times in a
loop and calculates the average execution time gives the following
results on a Malta + I6400 (running at 20MHz):
- Syscall: ~31000 ns
- VDSO (GIC): ~15000 ns
- VDSO (CP0): ~9500 ns
[markos.chandras@imgtec.com:
- Minor code re-arrangements in order for mappings to be made
in the order they appear to the process' address space.
- Move do_{monotonic, realtime} outside of the MIPS_CLOCK_VSYSCALL ifdef
- Use gic_get_usm_range so we can do the GIC mapping in the
arch/mips/kernel/vdso instead of the GIC irqchip driver]
Signed-off-by: Alex Smith <alex.smith@imgtec.com>
Signed-off-by: Markos Chandras <markos.chandras@imgtec.com>
Cc: linux-kernel@vger.kernel.org
Cc: linux-mips@linux-mips.org
Patchwork: https://patchwork.linux-mips.org/patch/11338/
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
Add an initial implementation of a proper (i.e. an ELF shared library)
VDSO. With this commit it does not export any symbols, it only replaces
the current signal return trampoline page. A later commit will add user
implementations of gettimeofday()/clock_gettime().
To support both new toolchains and old ones which don't generate ABI
flags section, we define its content manually and then use a tool
(genvdso) to patch up the section to have the correct name and type.
genvdso also extracts symbol offsets ({,rt_}sigreturn) needed by the
kernel, and generates a C file containing a "struct mips_vdso_image"
containing both the VDSO data and these offsets. This C file is
compiled into the kernel.
On 64-bit kernels we require a different VDSO for each supported ABI,
so we may build up to 3 different VDSOs. The VDSO to use is selected by
the mips_abi structure.
A kernel/user shared data page is created and mapped below the VDSO
image. This is currently empty, but will be used by the user time
function implementations which are added later.
[markos.chandras@imgtec.com:
- Add more comments
- Move abi detection in genvdso.h since it's the get_symbol function
that needs it.
- Add an R6 specific way to calculate the base address of VDSO in order
to avoid the branch instruction which affects performance.
- Do not patch .gnu.attributes since it's not needed for dynamic linking.
- Simplify Makefile a little bit.
- checkpatch fixes
- Restrict VDSO support for binutils < 2.25 for pre-R6
- Include atomic64.h for O32 variant on MIPS64]
Signed-off-by: Alex Smith <alex.smith@imgtec.com>
Signed-off-by: Markos Chandras <markos.chandras@imgtec.com>
Cc: Matthew Fortune <matthew.fortune@imgtec.com>
Cc: linux-mips@linux-mips.org
Patchwork: https://patchwork.linux-mips.org/patch/11337/
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
The motivation for this patchset was that I was looking at a way for a
qemu-kvm process, to exclude the guest memory from its core dump, which
can be quite large. There are already a number of filter flags in
/proc/<pid>/coredump_filter, however, these allow one to specify 'types'
of kernel memory, not specific address ranges (which is needed in this
case).
Since there are no more vma flags available, the first patch eliminates
the need for the 'VM_ALWAYSDUMP' flag. The flag is used internally by
the kernel to mark vdso and vsyscall pages. However, it is simple
enough to check if a vma covers a vdso or vsyscall page without the need
for this flag.
The second patch then replaces the 'VM_ALWAYSDUMP' flag with a new
'VM_NODUMP' flag, which can be set by userspace using new madvise flags:
'MADV_DONTDUMP', and unset via 'MADV_DODUMP'. The core dump filters
continue to work the same as before unless 'MADV_DONTDUMP' is set on the
region.
The qemu code which implements this features is at:
http://people.redhat.com/~jbaron/qemu-dump/qemu-dump.patch
In my testing the qemu core dump shrunk from 383MB -> 13MB with this
patch.
I also believe that the 'MADV_DONTDUMP' flag might be useful for
security sensitive apps, which might want to select which areas are
dumped.
This patch:
The VM_ALWAYSDUMP flag is currently used by the coredump code to
indicate that a vma is part of a vsyscall or vdso section. However, we
can determine if a vma is in one these sections by checking it against
the gate_vma and checking for a non-NULL return value from
arch_vma_name(). Thus, freeing a valuable vma bit.
Signed-off-by: Jason Baron <jbaron@redhat.com>
Acked-by: Roland McGrath <roland@hack.frob.com>
Cc: Chris Metcalf <cmetcalf@tilera.com>
Cc: Avi Kivity <avi@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Quoting from Jiri Slaby's patch of a similar nature for x86:
When initrd is in use and a driver does request_module() in its
module_init (i.e. __initcall or device_initcall), a modprobe
process is created with VDSO mapping. But VDSO is inited even in
__initcall, i.e. on the same level (at the same time), so it may
not be inited yet (link order matters).
Move init_vdso up to subsys_initcall to avoid the issue.
Signed-off-by: David Daney <ddaney@caviumnetworks.com>
To: linux-mips@linux-mips.org
Cc: David Daney <ddaney@caviumnetworks.com>
Cc: Jiri Slaby <jslaby@suse.cz>
Patchwork: http://patchwork.linux-mips.org/patch/1386/
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
In addition to being useless, it was mis-spelled.
Signed-off-by: David Daney <ddaney@caviumnetworks.com>
To: linux-mips@linux-mips.org
Cc: David Daney <ddaney@caviumnetworks.com>
Patchwork: http://patchwork.linux-mips.org/patch/1385/
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
This is a preliminary patch to add a vdso to all user processes. Still
missing are ELF headers and .eh_frame information. But it is enough to
allow us to move signal trampolines off of the stack. Note that emulation
of branch delay slots in the FPU emulator still requires the stack.
We allocate a single page (the vdso) and write all possible signal
trampolines into it. The stack is moved down by one page and the vdso is
mapped into this space.
Signed-off-by: David Daney <ddaney@caviumnetworks.com>
To: linux-mips@linux-mips.org
Patchwork: http://patchwork.linux-mips.org/patch/975/
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
