Re: [PATCH v1 1/4] lib: introduce copy_struct_from_user() helper
Hi Aleksa, Thank you for the patch! Yet something to improve: [auto build test ERROR on linus/master] [cannot apply to v5.3 next-20190924] [if your patch is applied to the wrong git tree, please drop us a note to help improve the system. BTW, we also suggest to use '--base' option to specify the base tree in git format-patch, please see https://stackoverflow.com/a/37406982] url: https://github.com/0day-ci/linux/commits/Aleksa-Sarai/lib-introduce-copy_struct_from_user-helper/20190926-010527 config: sh-rsk7269_defconfig (attached as .config) compiler: sh4-linux-gcc (GCC) 7.4.0 reproduce: wget https://raw.githubusercontent.com/intel/lkp-tests/master/sbin/make.cross -O ~/bin/make.cross chmod +x ~/bin/make.cross # save the attached .config to linux build tree GCC_VERSION=7.4.0 make.cross ARCH=sh If you fix the issue, kindly add following tag Reported-by: kbuild test robot All errors (new ones prefixed by >>): lib/strnlen_user.o: In function `is_zeroed_user': >> strnlen_user.c:(.text+0x160): undefined reference to `__get_user_unknown' --- 0-DAY kernel test infrastructureOpen Source Technology Center https://lists.01.org/pipermail/kbuild-all Intel Corporation .config.gz Description: application/gzip
Re: [PATCH v1 1/4] lib: introduce copy_struct_from_user() helper
On Wed, Sep 25, 2019 at 12:43 PM Al Viro wrote: > > FWIW, I would probably add a kernel-space analogue of that thing at the > same time - check that an area is all-zeroes is not all that rare. Hmm. Maybe. > Another thing is that for s390 we almost certainly want something better > than word-by-word. IIRC, word-sized userland accesses really hurt there. > It's nowhere near as critical as with strncpy_from_user(), but with the > same underlying issue. Well, s390 does have those magic "area" instructions, but part of why it's expensive on s390 is that they haven't implemented the "user_access_begin()/end()' stuff. I think s390 could use that to at least minimize some of the costs. With the common case presumably being just a couple of words, it migth not be worth it doing anything more than that even on s390. Interestingly (or perhaps not) if I read the internal s390 implementation correctly, they kind of _have_ that concept and they use it internally. It's just that they call it "enable_sacf_uaccess()" and "disable_sacf_uaccess()" instead. Linus
Re: [PATCH v1 1/4] lib: introduce copy_struct_from_user() helper
On Wed, Sep 25, 2019 at 11:13:18AM -0700, Linus Torvalds wrote:
> On Wed, Sep 25, 2019 at 11:04 AM Al Viro wrote:
> >
> > IMO it's better to lift reading the first word out of the loop, like this:
> > ...
> > Do you see any problems with that variant?
>
> That looks fine to me too.
>
> It's a bit harder for humans to read because of how it reads the word
> from user space one iteration early, but from a code generation
> standpoint it probably is better.
>
> So slightly subtler source code, and imho harder to read, but it looks
> correct to me.
FWIW, I would probably add a kernel-space analogue of that thing at the
same time - check that an area is all-zeroes is not all that rare.
In fact, most of the callers of memchr_inv() are exactly that:
arch/x86/kernel/fpu/xstate.c:492: if (memchr_inv(hdr->reserved, 0,
sizeof(hdr->reserved)))
crypto/async_tx/async_xor.c:225:return !memchr_inv(page_address(p) +
offset, 0, len);
crypto/dh_helper.c:109: if (memchr_inv(params->p, 0, params->p_size) == NULL)
crypto/testmgr.c:446: memchr_inv(&divs[i], 0, (count - i) *
sizeof(divs[0])) == NULL;
crypto/testmgr.c:470: if (memchr_inv(cfg->dst_divs, 0,
sizeof(cfg->dst_divs)))
crypto/testmgr.c:3802: if (memchr_inv(outbuf_dec, 0, out_len - m_size) ||
drivers/block/rbd.c:3138: if (memchr_inv(page_address(bv.bv_page)
+ bv.bv_offset, 0,
drivers/gpu/drm/drm_dp_mst_topology.c:1808: if (memchr_inv(guid, 0, 16))
drivers/gpu/drm/drm_edid.c:1359:if (memchr_inv(in_edid, 0, length))
drivers/gpu/drm/i915/gem/selftests/i915_gem_dmabuf.c:238: if
(memchr_inv(ptr, 0, dmabuf->size)) {
drivers/gpu/drm/i915/gem/selftests/i915_gem_dmabuf.c:284: if
(memchr_inv(ptr, 0, PAGE_SIZE)) {
drivers/infiniband/core/uverbs_cmd.c:2741: if
(memchr_inv(kern_spec_filter +
drivers/infiniband/core/uverbs_ioctl.c:143: return !memchr_inv((const void
*)&uattr->data + len,
drivers/infiniband/hw/efa/efa_verbs.c:127: !memchr_inv(reserved, 0,
sizeof(reserved))
drivers/infiniband/hw/mlx4/main.c:1330: memchr_inv((void *)&filter.field +\
drivers/infiniband/hw/mlx4/qp.c:728:if (memchr_inv(ucmd.reserved, 0,
sizeof(ucmd.reserved)))
drivers/infiniband/hw/mlx5/main.c:2516:
!(memchr_inv(MLX5_ADDR_OF(fte_match_param, match_criteria, headers), \
drivers/infiniband/hw/mlx5/main.c:2621: memchr_inv((void *)&filter.field +\
drivers/infiniband/hw/mlx5/qp.c:3921: memchr_inv(&ucmd.reserved,
0, sizeof(ucmd.reserved)) ||
drivers/infiniband/hw/mlx5/qp.c:3922:
memchr_inv(&ucmd.burst_info.reserved, 0,
drivers/md/dm-integrity.c:3844: if (memchr_inv(ic->sb, 0,
SB_SECTORS << SECTOR_SHIFT)) {
drivers/mtd/nand/raw/qcom_nandc.c:1548: if (memchr_inv(data_buf, 0xff,
data_len)) {
drivers/net/ethernet/mellanox/mlx5/core/en/tc_tun_geneve.c:179: if
(memchr_inv(&enc_opts.mask->data, 0, sizeof(enc_opts.mask->data)) &&
drivers/net/ethernet/mellanox/mlx5/core/en/tc_tun_geneve.c:240: if
(!memchr_inv(option_key->opt_data, 0, option_key->length * 4)) {
drivers/net/ethernet/mellanox/mlx5/core/eswitch_offloads.c:113: if
(memchr_inv(misc, 0, MLX5_ST_SZ_BYTES(fte_match_set_misc)))
drivers/net/ethernet/mellanox/mlxsw/core_acl_flex_keys.c:374: if
(!memchr_inv(mask_value, 0, len)) /* If mask is zero */
drivers/net/geneve.c:1243: memchr_inv(&info->key.u, 0,
sizeof(info->key.u)));
drivers/nvme/host/core.c:1619: memchr_inv(ids->nguid, 0,
sizeof(ids->nguid)) ||
drivers/nvme/host/core.c:1620: memchr_inv(ids->eui64, 0,
sizeof(ids->eui64));
drivers/nvme/host/core.c:2884: if (memchr_inv(ids->nguid, 0,
sizeof(ids->nguid)))
drivers/nvme/host/core.c:2887: if (memchr_inv(ids->eui64, 0,
sizeof(ids->eui64)))
drivers/nvme/host/core.c:2962: !memchr_inv(ids->nguid, 0,
sizeof(ids->nguid)))
drivers/nvme/host/core.c:2966: if (!memchr_inv(ids->nguid, 0,
sizeof(ids->nguid)))
drivers/nvme/host/core.c:2970: if (!memchr_inv(ids->eui64, 0,
sizeof(ids->eui64)))
drivers/nvme/target/admin-cmd.c:544:if (memchr_inv(&ns->uuid, 0,
sizeof(ns->uuid))) {
drivers/nvme/target/admin-cmd.c:551:if (memchr_inv(ns->nguid, 0,
sizeof(ns->nguid))) {
drivers/target/target_core_iblock.c:425:not_zero = memchr_inv(buf,
0x00, cmd->data_length);
fs/btrfs/ioctl.c:4586: if (memchr_inv(loi->reserved, 0,
sizeof(loi->reserved))) {
fs/cramfs/inode.c:351: return memchr_inv(tail_data, 0, PAGE_SIZE - partial) ?
true : false;
fs/ext4/ioctl.c:649:if (memchr_inv(head.fmh_reserved, 0,
sizeof(head.fmh_reserved)) ||
fs/ext4/ioctl.c:650:memchr_inv(head.fmh_keys[0].fmr_reserved, 0,
fs/ext4/ioctl.c:652:memchr_inv(head.fmh_keys[1].fmr_reserved, 0,
fs/gfs2/lock_dlm.c:485: return !memchr_inv(lvb + JID_BITMAP_OFFSET, 0,
fs/ubifs/auth.c:565:return !memchr_inv(hmac, 0, c->hmac_desc_len);
fs/xfs/scrub/agheader.c:276:if (memchr_inv(&sb->sb_features_compat,
0,
fs/xfs/scrub/aghe
Re: [PATCH v1 1/4] lib: introduce copy_struct_from_user() helper
Hi Aleksa, Thank you for the patch! Yet something to improve: [auto build test ERROR on linus/master] [cannot apply to v5.3 next-20190924] [if your patch is applied to the wrong git tree, please drop us a note to help improve the system. BTW, we also suggest to use '--base' option to specify the base tree in git format-patch, please see https://stackoverflow.com/a/37406982] url: https://github.com/0day-ci/linux/commits/Aleksa-Sarai/lib-introduce-copy_struct_from_user-helper/20190926-010527 config: arm-at91_dt_defconfig (attached as .config) compiler: arm-linux-gnueabi-gcc (GCC) 7.4.0 reproduce: wget https://raw.githubusercontent.com/intel/lkp-tests/master/sbin/make.cross -O ~/bin/make.cross chmod +x ~/bin/make.cross # save the attached .config to linux build tree GCC_VERSION=7.4.0 make.cross ARCH=arm If you fix the issue, kindly add following tag Reported-by: kbuild test robot All errors (new ones prefixed by >>): lib/strnlen_user.o: In function `is_zeroed_user': >> strnlen_user.c:(.text+0x1a8): undefined reference to `__get_user_bad' strnlen_user.c:(.text+0x204): undefined reference to `__get_user_bad' --- 0-DAY kernel test infrastructureOpen Source Technology Center https://lists.01.org/pipermail/kbuild-all Intel Corporation .config.gz Description: application/gzip
Re: [PATCH v1 1/4] lib: introduce copy_struct_from_user() helper
On Wed, Sep 25, 2019 at 11:04 AM Al Viro wrote: > > IMO it's better to lift reading the first word out of the loop, like this: > ... > Do you see any problems with that variant? That looks fine to me too. It's a bit harder for humans to read because of how it reads the word from user space one iteration early, but from a code generation standpoint it probably is better. So slightly subtler source code, and imho harder to read, but it looks correct to me. Linus
Re: [PATCH v1 1/4] lib: introduce copy_struct_from_user() helper
On Wed, Sep 25, 2019 at 10:48:31AM -0700, Linus Torvalds wrote:
> On Wed, Sep 25, 2019 at 10:21 AM Aleksa Sarai wrote:
> >
> > Just to make sure I understand, the following diff would this solve the
> > problem? If so, I'll apply it, and re-send in a few hours.
>
> Actually, looking at it more, it's still buggy.
>
> That final "size smaller than unsigned long" doesn't correctly handle
> the case of (say) a single byte in the middle of a 8-byte word.
>
> So you need to do something like this:
>
> int is_zeroed_user(const void __user *from, size_t size)
> {
> unsigned long val, mask, align;
>
> if (unlikely(!size))
> return true;
>
> if (!user_access_begin(from, size))
> return -EFAULT;
>
> align = (uintptr_t) from % sizeof(unsigned long);
> from -= align;
> size += align;
>
> mask = ~aligned_byte_mask(align);
>
> while (size >= sizeof(unsigned long)) {
> unsafe_get_user(val, (unsigned long __user *) from,
> err_fault);
> val &= mask;
> if (unlikely(val))
> goto done;
> mask = ~0ul;
> from += sizeof(unsigned long);
> size -= sizeof(unsigned long);
> }
>
> if (size) {
> /* (@from + @size) is unaligned. */
> unsafe_get_user(val, (unsigned long __user *) from,
> err_fault);
> mask &= aligned_byte_mask(size);
> val &= mask;
> }
IMO it's better to lift reading the first word out of the loop, like this:
align = (uintptr_t) from % sizeof(unsigned long);
from -= align;
unsafe_get_user(val, (unsigned long __user *) from, err_fault);
if (align) {
size += align;
val &= ~aligned_byte_mask(align);
}
while (size > sizeof(unsigned long)) {
if (unlikely(val))
goto done;
from += sizeof(unsigned long);
size -= sizeof(unsigned long);
unsafe_get_user(val, (unsigned long __user *) from, err_fault);
}
if (size != size(unsigned long))
val &= aligned_byte_mask(size);
done:
Do you see any problems with that variant?
Re: [PATCH v1 1/4] lib: introduce copy_struct_from_user() helper
On Wed, Sep 25, 2019 at 10:21 AM Aleksa Sarai wrote:
>
> Just to make sure I understand, the following diff would this solve the
> problem? If so, I'll apply it, and re-send in a few hours.
Actually, looking at it more, it's still buggy.
That final "size smaller than unsigned long" doesn't correctly handle
the case of (say) a single byte in the middle of a 8-byte word.
So you need to do something like this:
int is_zeroed_user(const void __user *from, size_t size)
{
unsigned long val, mask, align;
if (unlikely(!size))
return true;
if (!user_access_begin(from, size))
return -EFAULT;
align = (uintptr_t) from % sizeof(unsigned long);
from -= align;
size += align;
mask = ~aligned_byte_mask(align);
while (size >= sizeof(unsigned long)) {
unsafe_get_user(val, (unsigned long __user *) from, err_fault);
val &= mask;
if (unlikely(val))
goto done;
mask = ~0ul;
from += sizeof(unsigned long);
size -= sizeof(unsigned long);
}
if (size) {
/* (@from + @size) is unaligned. */
unsafe_get_user(val, (unsigned long __user *) from, err_fault);
mask &= aligned_byte_mask(size);
val &= mask;
}
done:
user_access_end();
return (val == 0);
err_fault:
user_access_end();
return -EFAULT;
}
note how "mask" carries around from the very beginning all the way to
the end, and "align" itself is no longer used after mask has been
calculated.
That's required because of say a 2-byte read at offset 5. You end up
with "align=5, size=7" at the beginning, and mask needs to be
0x0000 (on little-endian) for that final access.
Anyway, I checked, and the above seems to generate ok code quality
too. Sadly "unsafe_get_user()" cannot use "asm goto" because of a gcc
limitation (no asm goto with outputs), so it's not _perfect_, but
that's literally a compiler limitation.
But I didn't actually _test_ the end result. You should probably
verify that it gets the right behavior exactly for those interesting
cases where we mask both the beginning and the end.
Linus
Re: [PATCH v1 1/4] lib: introduce copy_struct_from_user() helper
On 2019-09-25, Linus Torvalds wrote:
> On Wed, Sep 25, 2019 at 10:00 AM Aleksa Sarai wrote:
> >
> > +int is_zeroed_user(const void __user *from, size_t size)
>
> I like how you've done this, but it's buggy and only works on 64-bit.
>
> All the "u64" and "8" cases need to be "unsigned long" and
> "sizeof(unsigned long)".
>
> Part of that requirement is:
>
> > + unsafe_get_user(val, (u64 __user *) from, err_fault);
>
> This part works fine - although 64-bit accesses migth be much more
> expensive and the win of unrolling might not be sufficient - but:
>
> > + if (align) {
> > + /* @from is unaligned. */
> > + val &= ~aligned_byte_mask(align);
> > + align = 0;
> > + }
>
> This part fundamentally only works on 'unsigned long'.
Just to make sure I understand, the following diff would this solve the
problem? If so, I'll apply it, and re-send in a few hours.
--8<--
int is_zeroed_user(const void __user *from, size_t size)
{
- u64 val;
- uintptr_t align = (uintptr_t) from % 8;
+ unsigned long val;
+ uintptr_t align = (uintptr_t) from % sizeof(unsigned long);
if (unlikely(!size))
return true;
@@ -150,8 +150,8 @@ int is_zeroed_user(const void __user *from, size_t size)
if (!user_access_begin(from, size))
return -EFAULT;
- while (size >= 8) {
- unsafe_get_user(val, (u64 __user *) from, err_fault);
+ while (size >= sizeof(unsigned long)) {
+ unsafe_get_user(val, (unsigned long __user *) from, err_fault);
if (align) {
/* @from is unaligned. */
val &= ~aligned_byte_mask(align);
@@ -159,12 +159,12 @@ int is_zeroed_user(const void __user *from, size_t size)
}
if (val)
goto done;
- from += 8;
- size -= 8;
+ from += sizeof(unsigned long);
+ size -= sizeof(unsigned long);
}
if (size) {
/* (@from + @size) is unaligned. */
- unsafe_get_user(val, (u64 __user *) from, err_fault);
+ unsafe_get_user(val, (unsigned long __user *) from, err_fault);
val &= aligned_byte_mask(size);
}
--
Aleksa Sarai
Senior Software Engineer (Containers)
SUSE Linux GmbH
Re: [PATCH v1 1/4] lib: introduce copy_struct_from_user() helper
On Wed, Sep 25, 2019 at 07:18:11PM +0200, Christian Brauner wrote:
> On Wed, Sep 25, 2019 at 06:59:12PM +0200, Aleksa Sarai wrote:
> > A common pattern for syscall extensions is increasing the size of a
> > struct passed from userspace, such that the zero-value of the new fields
> > result in the old kernel behaviour (allowing for a mix of userspace and
> > kernel vintages to operate on one another in most cases).
> >
> > While this interface exists for communication in both directions, only
> > one interface is straightforward to have reasonable semantics for
> > (userspace passing a struct to the kernel). For kernel returns to
> > userspace, what the correct semantics are (whether there should be an
> > error if userspace is unaware of a new extension) is very
> > syscall-dependent and thus probably cannot be unified between syscalls
> > (a good example of this problem is [1]).
> >
> > Previously there was no common lib/ function that implemented
> > the necessary extension-checking semantics (and different syscalls
> > implemented them slightly differently or incompletely[2]). Future
> > patches replace common uses of this pattern to make use of
> > copy_struct_from_user().
> >
> > [1]: commit 1251201c0d34 ("sched/core: Fix uclamp ABI bug, clean up and
> > robustify sched_read_attr() ABI logic and code")
> >
> > [2]: For instance {sched_setattr,perf_event_open,clone3}(2) all do do
> > similar checks to copy_struct_from_user() while rt_sigprocmask(2)
> > always rejects differently-sized struct arguments.
> >
> > Suggested-by: Rasmus Villemoes
> > Signed-off-by: Aleksa Sarai
> > ---
> > include/linux/uaccess.h | 4 +++
> > lib/Makefile| 2 +-
> > lib/strnlen_user.c | 52 +
> > lib/struct_user.c | 73 +
> > 4 files changed, 130 insertions(+), 1 deletion(-)
> > create mode 100644 lib/struct_user.c
>
> Hm, why the new file?
> Couldn't this just live in usercopy.c?
>
> >
> > diff --git a/include/linux/uaccess.h b/include/linux/uaccess.h
> > index 34a038563d97..824569e309e4 100644
> > --- a/include/linux/uaccess.h
> > +++ b/include/linux/uaccess.h
> > @@ -230,6 +230,10 @@ static inline unsigned long
> > __copy_from_user_inatomic_nocache(void *to,
> >
> > #endif /* ARCH_HAS_NOCACHE_UACCESS */
> >
> > +extern int is_zeroed_user(const void __user *from, size_t count);
> > +extern int copy_struct_from_user(void *dst, size_t ksize,
> > +const void __user *src, size_t usize);
> > +
> > /*
> > * probe_kernel_read(): safely attempt to read from a location
> > * @dst: pointer to the buffer that shall take the data
> > diff --git a/lib/Makefile b/lib/Makefile
> > index 29c02a924973..d86c71feaf0a 100644
> > --- a/lib/Makefile
> > +++ b/lib/Makefile
> > @@ -28,7 +28,7 @@ endif
> > CFLAGS_string.o := $(call cc-option, -fno-stack-protector)
> > endif
> >
> > -lib-y := ctype.o string.o vsprintf.o cmdline.o \
> > +lib-y := ctype.o string.o struct_user.o vsprintf.o cmdline.o \
> > rbtree.o radix-tree.o timerqueue.o xarray.o \
> > idr.o extable.o \
> > sha1.o chacha.o irq_regs.o argv_split.o \
> > diff --git a/lib/strnlen_user.c b/lib/strnlen_user.c
> > index 7f2db3fe311f..7eb665732954 100644
> > --- a/lib/strnlen_user.c
> > +++ b/lib/strnlen_user.c
> > @@ -123,3 +123,55 @@ long strnlen_user(const char __user *str, long count)
> > return 0;
> > }
> > EXPORT_SYMBOL(strnlen_user);
> > +
> > +/**
> > + * is_zeroed_user: check if a userspace buffer is full of zeros
> > + * @from: Source address, in userspace.
> > + * @size: Size of buffer.
> > + *
> > + * This is effectively shorthand for "memchr_inv(from, 0, size) == NULL"
> > for
> > + * userspace addresses. If there are non-zero bytes present then false is
> > + * returned, otherwise true is returned.
> > + *
> > + * Returns:
> > + * * -EFAULT: access to userspace failed.
> > + */
> > +int is_zeroed_user(const void __user *from, size_t size)
>
> *sigh*, I'm probably going to get yelled at because of this but: does
> this really provide any _performance_ benefits over the dumb get_user()
> loop that we currently have that we care about right now? My point
> being, that the loop - imho - is much easier to understand than what is
> going on here with all the masking, and aligning etc. that we have here.
> But I'm not going to fight it.
>
> > +{
> > + u64 val;
> > + uintptr_t align = (uintptr_t) from % 8;
> > +
> > + if (unlikely(!size))
> > + return true;
>
> Nit: I'd prefer int variables be checked with if (size != 0) :)
>
> > +
> > + from -= align;
> > + size += align;
> > +
> > + if (!user_access_begin(from, size))
> > + return -EFAULT;
> > +
> > + while (size >= 8) {
> > + unsafe_get_user(val, (u64 __user *) from, err_fault);
> > + if (align) {
> > + /* @from is unaligned. */
> > + val &= ~
Re: [PATCH v1 1/4] lib: introduce copy_struct_from_user() helper
On Wed, Sep 25, 2019 at 06:59:12PM +0200, Aleksa Sarai wrote:
> A common pattern for syscall extensions is increasing the size of a
> struct passed from userspace, such that the zero-value of the new fields
> result in the old kernel behaviour (allowing for a mix of userspace and
> kernel vintages to operate on one another in most cases).
>
> While this interface exists for communication in both directions, only
> one interface is straightforward to have reasonable semantics for
> (userspace passing a struct to the kernel). For kernel returns to
> userspace, what the correct semantics are (whether there should be an
> error if userspace is unaware of a new extension) is very
> syscall-dependent and thus probably cannot be unified between syscalls
> (a good example of this problem is [1]).
>
> Previously there was no common lib/ function that implemented
> the necessary extension-checking semantics (and different syscalls
> implemented them slightly differently or incompletely[2]). Future
> patches replace common uses of this pattern to make use of
> copy_struct_from_user().
>
> [1]: commit 1251201c0d34 ("sched/core: Fix uclamp ABI bug, clean up and
> robustify sched_read_attr() ABI logic and code")
>
> [2]: For instance {sched_setattr,perf_event_open,clone3}(2) all do do
> similar checks to copy_struct_from_user() while rt_sigprocmask(2)
> always rejects differently-sized struct arguments.
>
> Suggested-by: Rasmus Villemoes
> Signed-off-by: Aleksa Sarai
> ---
> include/linux/uaccess.h | 4 +++
> lib/Makefile| 2 +-
> lib/strnlen_user.c | 52 +
> lib/struct_user.c | 73 +
> 4 files changed, 130 insertions(+), 1 deletion(-)
> create mode 100644 lib/struct_user.c
Hm, why the new file?
Couldn't this just live in usercopy.c?
>
> diff --git a/include/linux/uaccess.h b/include/linux/uaccess.h
> index 34a038563d97..824569e309e4 100644
> --- a/include/linux/uaccess.h
> +++ b/include/linux/uaccess.h
> @@ -230,6 +230,10 @@ static inline unsigned long
> __copy_from_user_inatomic_nocache(void *to,
>
> #endif /* ARCH_HAS_NOCACHE_UACCESS */
>
> +extern int is_zeroed_user(const void __user *from, size_t count);
> +extern int copy_struct_from_user(void *dst, size_t ksize,
> + const void __user *src, size_t usize);
> +
> /*
> * probe_kernel_read(): safely attempt to read from a location
> * @dst: pointer to the buffer that shall take the data
> diff --git a/lib/Makefile b/lib/Makefile
> index 29c02a924973..d86c71feaf0a 100644
> --- a/lib/Makefile
> +++ b/lib/Makefile
> @@ -28,7 +28,7 @@ endif
> CFLAGS_string.o := $(call cc-option, -fno-stack-protector)
> endif
>
> -lib-y := ctype.o string.o vsprintf.o cmdline.o \
> +lib-y := ctype.o string.o struct_user.o vsprintf.o cmdline.o \
>rbtree.o radix-tree.o timerqueue.o xarray.o \
>idr.o extable.o \
>sha1.o chacha.o irq_regs.o argv_split.o \
> diff --git a/lib/strnlen_user.c b/lib/strnlen_user.c
> index 7f2db3fe311f..7eb665732954 100644
> --- a/lib/strnlen_user.c
> +++ b/lib/strnlen_user.c
> @@ -123,3 +123,55 @@ long strnlen_user(const char __user *str, long count)
> return 0;
> }
> EXPORT_SYMBOL(strnlen_user);
> +
> +/**
> + * is_zeroed_user: check if a userspace buffer is full of zeros
> + * @from: Source address, in userspace.
> + * @size: Size of buffer.
> + *
> + * This is effectively shorthand for "memchr_inv(from, 0, size) == NULL" for
> + * userspace addresses. If there are non-zero bytes present then false is
> + * returned, otherwise true is returned.
> + *
> + * Returns:
> + * * -EFAULT: access to userspace failed.
> + */
> +int is_zeroed_user(const void __user *from, size_t size)
*sigh*, I'm probably going to get yelled at because of this but: does
this really provide any _performance_ benefits over the dumb get_user()
loop that we currently have that we care about right now? My point
being, that the loop - imho - is much easier to understand than what is
going on here with all the masking, and aligning etc. that we have here.
But I'm not going to fight it.
> +{
> + u64 val;
> + uintptr_t align = (uintptr_t) from % 8;
> +
> + if (unlikely(!size))
> + return true;
Nit: I'd prefer int variables be checked with if (size != 0) :)
> +
> + from -= align;
> + size += align;
> +
> + if (!user_access_begin(from, size))
> + return -EFAULT;
> +
> + while (size >= 8) {
> + unsafe_get_user(val, (u64 __user *) from, err_fault);
> + if (align) {
> + /* @from is unaligned. */
> + val &= ~aligned_byte_mask(align);
> + align = 0;
> + }
> + if (val)
> + goto done;
> + from += 8;
> + size -= 8;
> + }
> + if (size) {
> + /* (@from + @size) is unaligned. */
>
Re: [PATCH v1 1/4] lib: introduce copy_struct_from_user() helper
On Wed, Sep 25, 2019 at 10:00 AM Aleksa Sarai wrote:
>
> +int is_zeroed_user(const void __user *from, size_t size)
I like how you've done this, but it's buggy and only works on 64-bit.
All the "u64" and "8" cases need to be "unsigned long" and
"sizeof(unsigned long)".
Part of that requirement is:
> + unsafe_get_user(val, (u64 __user *) from, err_fault);
This part works fine - although 64-bit accesses migth be much more
expensive and the win of unrolling might not be sufficient - but:
> + if (align) {
> + /* @from is unaligned. */
> + val &= ~aligned_byte_mask(align);
> + align = 0;
> + }
This part fundamentally only works on 'unsigned long'.
Linus
[PATCH v1 1/4] lib: introduce copy_struct_from_user() helper
A common pattern for syscall extensions is increasing the size of a
struct passed from userspace, such that the zero-value of the new fields
result in the old kernel behaviour (allowing for a mix of userspace and
kernel vintages to operate on one another in most cases).
While this interface exists for communication in both directions, only
one interface is straightforward to have reasonable semantics for
(userspace passing a struct to the kernel). For kernel returns to
userspace, what the correct semantics are (whether there should be an
error if userspace is unaware of a new extension) is very
syscall-dependent and thus probably cannot be unified between syscalls
(a good example of this problem is [1]).
Previously there was no common lib/ function that implemented
the necessary extension-checking semantics (and different syscalls
implemented them slightly differently or incompletely[2]). Future
patches replace common uses of this pattern to make use of
copy_struct_from_user().
[1]: commit 1251201c0d34 ("sched/core: Fix uclamp ABI bug, clean up and
robustify sched_read_attr() ABI logic and code")
[2]: For instance {sched_setattr,perf_event_open,clone3}(2) all do do
similar checks to copy_struct_from_user() while rt_sigprocmask(2)
always rejects differently-sized struct arguments.
Suggested-by: Rasmus Villemoes
Signed-off-by: Aleksa Sarai
---
include/linux/uaccess.h | 4 +++
lib/Makefile| 2 +-
lib/strnlen_user.c | 52 +
lib/struct_user.c | 73 +
4 files changed, 130 insertions(+), 1 deletion(-)
create mode 100644 lib/struct_user.c
diff --git a/include/linux/uaccess.h b/include/linux/uaccess.h
index 34a038563d97..824569e309e4 100644
--- a/include/linux/uaccess.h
+++ b/include/linux/uaccess.h
@@ -230,6 +230,10 @@ static inline unsigned long
__copy_from_user_inatomic_nocache(void *to,
#endif /* ARCH_HAS_NOCACHE_UACCESS */
+extern int is_zeroed_user(const void __user *from, size_t count);
+extern int copy_struct_from_user(void *dst, size_t ksize,
+const void __user *src, size_t usize);
+
/*
* probe_kernel_read(): safely attempt to read from a location
* @dst: pointer to the buffer that shall take the data
diff --git a/lib/Makefile b/lib/Makefile
index 29c02a924973..d86c71feaf0a 100644
--- a/lib/Makefile
+++ b/lib/Makefile
@@ -28,7 +28,7 @@ endif
CFLAGS_string.o := $(call cc-option, -fno-stack-protector)
endif
-lib-y := ctype.o string.o vsprintf.o cmdline.o \
+lib-y := ctype.o string.o struct_user.o vsprintf.o cmdline.o \
rbtree.o radix-tree.o timerqueue.o xarray.o \
idr.o extable.o \
sha1.o chacha.o irq_regs.o argv_split.o \
diff --git a/lib/strnlen_user.c b/lib/strnlen_user.c
index 7f2db3fe311f..7eb665732954 100644
--- a/lib/strnlen_user.c
+++ b/lib/strnlen_user.c
@@ -123,3 +123,55 @@ long strnlen_user(const char __user *str, long count)
return 0;
}
EXPORT_SYMBOL(strnlen_user);
+
+/**
+ * is_zeroed_user: check if a userspace buffer is full of zeros
+ * @from: Source address, in userspace.
+ * @size: Size of buffer.
+ *
+ * This is effectively shorthand for "memchr_inv(from, 0, size) == NULL" for
+ * userspace addresses. If there are non-zero bytes present then false is
+ * returned, otherwise true is returned.
+ *
+ * Returns:
+ * * -EFAULT: access to userspace failed.
+ */
+int is_zeroed_user(const void __user *from, size_t size)
+{
+ u64 val;
+ uintptr_t align = (uintptr_t) from % 8;
+
+ if (unlikely(!size))
+ return true;
+
+ from -= align;
+ size += align;
+
+ if (!user_access_begin(from, size))
+ return -EFAULT;
+
+ while (size >= 8) {
+ unsafe_get_user(val, (u64 __user *) from, err_fault);
+ if (align) {
+ /* @from is unaligned. */
+ val &= ~aligned_byte_mask(align);
+ align = 0;
+ }
+ if (val)
+ goto done;
+ from += 8;
+ size -= 8;
+ }
+ if (size) {
+ /* (@from + @size) is unaligned. */
+ unsafe_get_user(val, (u64 __user *) from, err_fault);
+ val &= aligned_byte_mask(size);
+ }
+
+done:
+ user_access_end();
+ return (val == 0);
+err_fault:
+ user_access_end();
+ return -EFAULT;
+}
diff --git a/lib/struct_user.c b/lib/struct_user.c
new file mode 100644
index ..57d79eb53bfa
--- /dev/null
+++ b/lib/struct_user.c
@@ -0,0 +1,73 @@
+// SPDX-License-Identifier: GPL-2.0-or-later
+/*
+ * Copyright (C) 2019 SUSE LLC
+ * Copyright (C) 2019 Aleksa Sarai
+ */
+
+#include
+#include
+#include
+#include
+#include
+
+/**
+ * copy_struct_from_user: copy a struct from userspace
+ * @dst: Destination address, in kernel space. This buffer must be @ksize
+ * bytes lon
