The bitmap_scatter() mistakenly refers to itself for detailed explanation
about the relationships of two. Instead of simply fixing this, align text
in both making a cross-reference.
Fixes: de5f843389 ("lib/bitmap: Introduce bitmap_scatter() and bitmap_gather() helpers")
Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Signed-off-by: Yury Norov <yury.norov@gmail.com>
Pull bitmap updates from Yury Norov:
- switch all bitmamp APIs from inline to __always_inline (Brian Norris)
The __always_inline series improves on code generation, and now with
the latest compiler versions is required to avoid compilation
warnings. It spent enough in my backlog, and I'm thankful to Brian
Norris for taking over and moving it forward.
- introduce GENMASK_U128() macro (Anshuman Khandual)
GENMASK_U128() is a prerequisite needed for arm64 development
* tag 'bitmap-for-6.12' of https://github.com/norov/linux:
lib/test_bits.c: Add tests for GENMASK_U128()
uapi: Define GENMASK_U128
nodemask: Switch from inline to __always_inline
cpumask: Switch from inline to __always_inline
bitmap: Switch from inline to __always_inline
find: Switch from inline to __always_inline
'inline' keyword is only a recommendation for compiler. If it decides to
not inline bitmap functions, the whole small_const_nbits() machinery
doesn't work.
This is how a standard GCC 11.3.0 does for my x86_64 build now. This patch
replaces 'inline' directive with unconditional '__always_inline' to make
sure that there's always a chance for compile-time optimization. It doesn't
change size of kernel image, according to bloat-o-meter.
[[ Brian: split out from:
Subject: [PATCH 1/3] bitmap: switch from inline to __always_inline
https://lore.kernel.org/all/20221027043810.350460-2-yury.norov@gmail.com/
But rewritten, as there were too many conflicts. ]]
Co-developed-by: Brian Norris <briannorris@chromium.org>
Signed-off-by: Brian Norris <briannorris@chromium.org>
Reviewed-by: Kees Cook <kees@kernel.org>
Reviewed-by: Nathan Chancellor <nathan@kernel.org>
Signed-off-by: Yury Norov <yury.norov@gmail.com>
copy_fd_bitmaps(new, old, count) is expected to copy the first
count/BITS_PER_LONG bits from old->full_fds_bits[] and fill
the rest with zeroes. What it does is copying enough words
(BITS_TO_LONGS(count/BITS_PER_LONG)), then memsets the rest.
That works fine, *if* all bits past the cutoff point are
clear. Otherwise we are risking garbage from the last word
we'd copied.
For most of the callers that is true - expand_fdtable() has
count equal to old->max_fds, so there's no open descriptors
past count, let alone fully occupied words in ->open_fds[],
which is what bits in ->full_fds_bits[] correspond to.
The other caller (dup_fd()) passes sane_fdtable_size(old_fdt, max_fds),
which is the smallest multiple of BITS_PER_LONG that covers all
opened descriptors below max_fds. In the common case (copying on
fork()) max_fds is ~0U, so all opened descriptors will be below
it and we are fine, by the same reasons why the call in expand_fdtable()
is safe.
Unfortunately, there is a case where max_fds is less than that
and where we might, indeed, end up with junk in ->full_fds_bits[] -
close_range(from, to, CLOSE_RANGE_UNSHARE) with
* descriptor table being currently shared
* 'to' being above the current capacity of descriptor table
* 'from' being just under some chunk of opened descriptors.
In that case we end up with observably wrong behaviour - e.g. spawn
a child with CLONE_FILES, get all descriptors in range 0..127 open,
then close_range(64, ~0U, CLOSE_RANGE_UNSHARE) and watch dup(0) ending
up with descriptor #128, despite #64 being observably not open.
The minimally invasive fix would be to deal with that in dup_fd().
If this proves to add measurable overhead, we can go that way, but
let's try to fix copy_fd_bitmaps() first.
* new helper: bitmap_copy_and_expand(to, from, bits_to_copy, size).
* make copy_fd_bitmaps() take the bitmap size in words, rather than
bits; it's 'count' argument is always a multiple of BITS_PER_LONG,
so we are not losing any information, and that way we can use the
same helper for all three bitmaps - compiler will see that count
is a multiple of BITS_PER_LONG for the large ones, so it'll generate
plain memcpy()+memset().
Reproducer added to tools/testing/selftests/core/close_range_test.c
Cc: stable@vger.kernel.org
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Now that we have generic bitmap_read() and bitmap_write(), which are
inline and try to take care of non-bound-crossing and aligned cases
to keep them optimized, collapse bitmap_{get,set}_value8() into
simple wrappers around the former ones.
bloat-o-meter shows no difference in vmlinux and -2 bytes for
gpio-pca953x.ko, which says the optimization didn't suffer due to
that change. The converted helpers have the value width embedded
and always compile-time constant and that helps a lot.
Suggested-by: Yury Norov <yury.norov@gmail.com>
Signed-off-by: Yury Norov <yury.norov@gmail.com>
Signed-off-by: Alexander Lobakin <aleksander.lobakin@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The number of times yet another open coded
`BITS_TO_LONGS(nbits) * sizeof(long)` can be spotted is huge.
Some generic helper is long overdue.
Add one, bitmap_size(), but with one detail.
BITS_TO_LONGS() uses DIV_ROUND_UP(). The latter works well when both
divident and divisor are compile-time constants or when the divisor
is not a pow-of-2. When it is however, the compilers sometimes tend
to generate suboptimal code (GCC 13):
48 83 c0 3f add $0x3f,%rax
48 c1 e8 06 shr $0x6,%rax
48 8d 14 c5 00 00 00 00 lea 0x0(,%rax,8),%rdx
%BITS_PER_LONG is always a pow-2 (either 32 or 64), but GCC still does
full division of `nbits + 63` by it and then multiplication by 8.
Instead of BITS_TO_LONGS(), use ALIGN() and then divide by 8. GCC:
8d 50 3f lea 0x3f(%rax),%edx
c1 ea 03 shr $0x3,%edx
81 e2 f8 ff ff 1f and $0x1ffffff8,%edx
Now it shifts `nbits + 63` by 3 positions (IOW performs fast division
by 8) and then masks bits[2:0]. bloat-o-meter:
add/remove: 0/0 grow/shrink: 20/133 up/down: 156/-773 (-617)
Clang does it better and generates the same code before/after starting
from -O1, except that with the ALIGN() approach it uses %edx and thus
still saves some bytes:
add/remove: 0/0 grow/shrink: 9/133 up/down: 18/-538 (-520)
Note that we can't expand DIV_ROUND_UP() by adding a check and using
this approach there, as it's used in array declarations where
expressions are not allowed.
Add this helper to tools/ as well.
Reviewed-by: Przemek Kitszel <przemyslaw.kitszel@intel.com>
Acked-by: Yury Norov <yury.norov@gmail.com>
Signed-off-by: Alexander Lobakin <aleksander.lobakin@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Pull networking updates from Jakub Kicinski:
"Core & protocols:
- Large effort by Eric to lower rtnl_lock pressure and remove locks:
- Make commonly used parts of rtnetlink (address, route dumps
etc) lockless, protected by RCU instead of rtnl_lock.
- Add a netns exit callback which already holds rtnl_lock,
allowing netns exit to take rtnl_lock once in the core instead
of once for each driver / callback.
- Remove locks / serialization in the socket diag interface.
- Remove 6 calls to synchronize_rcu() while holding rtnl_lock.
- Remove the dev_base_lock, depend on RCU where necessary.
- Support busy polling on a per-epoll context basis. Poll length and
budget parameters can be set independently of system defaults.
- Introduce struct net_hotdata, to make sure read-mostly global
config variables fit in as few cache lines as possible.
- Add optional per-nexthop statistics to ease monitoring / debug of
ECMP imbalance problems.
- Support TCP_NOTSENT_LOWAT in MPTCP.
- Ensure that IPv6 temporary addresses' preferred lifetimes are long
enough, compared to other configured lifetimes, and at least 2 sec.
- Support forwarding of ICMP Error messages in IPSec, per RFC 4301.
- Add support for the independent control state machine for bonding
per IEEE 802.1AX-2008 5.4.15 in addition to the existing coupled
control state machine.
- Add "network ID" to MCTP socket APIs to support hosts with multiple
disjoint MCTP networks.
- Re-use the mono_delivery_time skbuff bit for packets which user
space wants to be sent at a specified time. Maintain the timing
information while traversing veth links, bridge etc.
- Take advantage of MSG_SPLICE_PAGES for RxRPC DATA and ACK packets.
- Simplify many places iterating over netdevs by using an xarray
instead of a hash table walk (hash table remains in place, for use
on fastpaths).
- Speed up scanning for expired routes by keeping a dedicated list.
- Speed up "generic" XDP by trying harder to avoid large allocations.
- Support attaching arbitrary metadata to netconsole messages.
Things we sprinkled into general kernel code:
- Enforce VM_IOREMAP flag and range in ioremap_page_range and
introduce VM_SPARSE kind and vm_area_[un]map_pages (used by
bpf_arena).
- Rework selftest harness to enable the use of the full range of ksft
exit code (pass, fail, skip, xfail, xpass).
Netfilter:
- Allow userspace to define a table that is exclusively owned by a
daemon (via netlink socket aliveness) without auto-removing this
table when the userspace program exits. Such table gets marked as
orphaned and a restarting management daemon can re-attach/regain
ownership.
- Speed up element insertions to nftables' concatenated-ranges set
type. Compact a few related data structures.
BPF:
- Add BPF token support for delegating a subset of BPF subsystem
functionality from privileged system-wide daemons such as systemd
through special mount options for userns-bound BPF fs to a trusted
& unprivileged application.
- Introduce bpf_arena which is sparse shared memory region between
BPF program and user space where structures inside the arena can
have pointers to other areas of the arena, and pointers work
seamlessly for both user-space programs and BPF programs.
- Introduce may_goto instruction that is a contract between the
verifier and the program. The verifier allows the program to loop
assuming it's behaving well, but reserves the right to terminate
it.
- Extend the BPF verifier to enable static subprog calls in spin lock
critical sections.
- Support registration of struct_ops types from modules which helps
projects like fuse-bpf that seeks to implement a new struct_ops
type.
- Add support for retrieval of cookies for perf/kprobe multi links.
- Support arbitrary TCP SYN cookie generation / validation in the TC
layer with BPF to allow creating SYN flood handling in BPF
firewalls.
- Add code generation to inline the bpf_kptr_xchg() helper which
improves performance when stashing/popping the allocated BPF
objects.
Wireless:
- Add SPP (signaling and payload protected) AMSDU support.
- Support wider bandwidth OFDMA, as required for EHT operation.
Driver API:
- Major overhaul of the Energy Efficient Ethernet internals to
support new link modes (2.5GE, 5GE), share more code between
drivers (especially those using phylib), and encourage more
uniform behavior. Convert and clean up drivers.
- Define an API for querying per netdev queue statistics from
drivers.
- IPSec: account in global stats for fully offloaded sessions.
- Create a concept of Ethernet PHY Packages at the Device Tree level,
to allow parameterizing the existing PHY package code.
- Enable Rx hashing (RSS) on GTP protocol fields.
Misc:
- Improvements and refactoring all over networking selftests.
- Create uniform module aliases for TC classifiers, actions, and
packet schedulers to simplify creating modprobe policies.
- Address all missing MODULE_DESCRIPTION() warnings in networking.
- Extend the Netlink descriptions in YAML to cover message
encapsulation or "Netlink polymorphism", where interpretation of
nested attributes depends on link type, classifier type or some
other "class type".
Drivers:
- Ethernet high-speed NICs:
- Add a new driver for Marvell's Octeon PCI Endpoint NIC VF.
- Intel (100G, ice, idpf):
- support E825-C devices
- nVidia/Mellanox:
- support devices with one port and multiple PCIe links
- Broadcom (bnxt):
- support n-tuple filters
- support configuring the RSS key
- Wangxun (ngbe/txgbe):
- implement irq_domain for TXGBE's sub-interrupts
- Pensando/AMD:
- support XDP
- optimize queue submission and wakeup handling (+17% bps)
- optimize struct layout, saving 28% of memory on queues
- Ethernet NICs embedded and virtual:
- Google cloud vNIC:
- refactor driver to perform memory allocations for new queue
config before stopping and freeing the old queue memory
- Synopsys (stmmac):
- obey queueMaxSDU and implement counters required by 802.1Qbv
- Renesas (ravb):
- support packet checksum offload
- suspend to RAM and runtime PM support
- Ethernet switches:
- nVidia/Mellanox:
- support for nexthop group statistics
- Microchip:
- ksz8: implement PHY loopback
- add support for KSZ8567, a 7-port 10/100Mbps switch
- PTP:
- New driver for RENESAS FemtoClock3 Wireless clock generator.
- Support OCP PTP cards designed and built by Adva.
- CAN:
- Support recvmsg() flags for own, local and remote traffic on CAN
BCM sockets.
- Support for esd GmbH PCIe/402 CAN device family.
- m_can:
- Rx/Tx submission coalescing
- wake on frame Rx
- WiFi:
- Intel (iwlwifi):
- enable signaling and payload protected A-MSDUs
- support wider-bandwidth OFDMA
- support for new devices
- bump FW API to 89 for AX devices; 90 for BZ/SC devices
- MediaTek (mt76):
- mt7915: newer ADIE version support
- mt7925: radio temperature sensor support
- Qualcomm (ath11k):
- support 6 GHz station power modes: Low Power Indoor (LPI),
Standard Power) SP and Very Low Power (VLP)
- QCA6390 & WCN6855: support 2 concurrent station interfaces
- QCA2066 support
- Qualcomm (ath12k):
- refactoring in preparation for Multi-Link Operation (MLO)
support
- 1024 Block Ack window size support
- firmware-2.bin support
- support having multiple identical PCI devices (firmware needs
to have ATH12K_FW_FEATURE_MULTI_QRTR_ID)
- QCN9274: support split-PHY devices
- WCN7850: enable Power Save Mode in station mode
- WCN7850: P2P support
- RealTek:
- rtw88: support for more rtw8811cu and rtw8821cu devices
- rtw89: support SCAN_RANDOM_SN and SET_SCAN_DWELL
- rtlwifi: speed up USB firmware initialization
- rtwl8xxxu:
- RTL8188F: concurrent interface support
- Channel Switch Announcement (CSA) support in AP mode
- Broadcom (brcmfmac):
- per-vendor feature support
- per-vendor SAE password setup
- DMI nvram filename quirk for ACEPC W5 Pro"
* tag 'net-next-6.9' of git://git.kernel.org/pub/scm/linux/kernel/git/netdev/net-next: (2255 commits)
nexthop: Fix splat with CONFIG_DEBUG_PREEMPT=y
nexthop: Fix out-of-bounds access during attribute validation
nexthop: Only parse NHA_OP_FLAGS for dump messages that require it
nexthop: Only parse NHA_OP_FLAGS for get messages that require it
bpf: move sleepable flag from bpf_prog_aux to bpf_prog
bpf: hardcode BPF_PROG_PACK_SIZE to 2MB * num_possible_nodes()
selftests/bpf: Add kprobe multi triggering benchmarks
ptp: Move from simple ida to xarray
vxlan: Remove generic .ndo_get_stats64
vxlan: Do not alloc tstats manually
devlink: Add comments to use netlink gen tool
nfp: flower: handle acti_netdevs allocation failure
net/packet: Add getsockopt support for PACKET_COPY_THRESH
net/netlink: Add getsockopt support for NETLINK_LISTEN_ALL_NSID
selftests/bpf: Add bpf_arena_htab test.
selftests/bpf: Add bpf_arena_list test.
selftests/bpf: Add unit tests for bpf_arena_alloc/free_pages
bpf: Add helper macro bpf_addr_space_cast()
libbpf: Recognize __arena global variables.
bpftool: Recognize arena map type
...
These helpers scatters or gathers a bitmap with the help of the mask
position bits parameter.
bitmap_scatter() does the following:
src: 0000000001011010
||||||
+------+|||||
| +----+||||
| |+----+|||
| || +-+||
| || | ||
mask: ...v..vv...v..vv
...0..11...0..10
dst: 0000001100000010
and bitmap_gather() performs this one:
mask: ...v..vv...v..vv
src: 0000001100000010
^ ^^ ^ 0
| || | 10
| || > 010
| |+--> 1010
| +--> 11010
+----> 011010
dst: 0000000000011010
bitmap_gather() can the seen as the reverse bitmap_scatter() operation.
Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Link: https://lore.kernel.org/lkml/20230926052007.3917389-3-andriy.shevchenko@linux.intel.com/
Co-developed-by: Herve Codina <herve.codina@bootlin.com>
Signed-off-by: Herve Codina <herve.codina@bootlin.com>
Acked-by: Yury Norov <yury.norov@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Similarly to cpumask_weight_and(), cpumask_weight_andnot() is a handy
helper that may help to avoid creating an intermediate mask just to
calculate number of bits that set in a 1st given mask, and clear in 2nd
one.
Signed-off-by: Yury Norov <yury.norov@gmail.com>
Reviewed-by: Jacob Keller <jacob.e.keller@intel.com>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
lib/bitmap.c and corresponding include/linux/bitmap.h are intended to
hold functions related to operations on bitmaps, like bitmap_shift or
bitmap_set. Historically, some string-related operations like
bitmap_parse are also reside in lib/bitmap.c.
Now that the subsystem evolves, string-related bitmap operations became a
significant part of the file. Because they are quite different from the
other bitmap functions by nature, it's worth to split them to a separate
source/header files.
CC: Andrew Morton <akpm@linux-foundation.org>
CC: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
CC: Rasmus Villemoes <linux@rasmusvillemoes.dk>
Signed-off-by: Yury Norov <yury.norov@gmail.com>
Besides the fact it's not used anywhere it should be implemented
differently, i.e. via helpers from linux/byteorder/generic.h.
Yet the helpers themselves need to be introduced first.
Also note, the function lacks of the test cases, they must be provided.
Hence, drop the current dead code for good.
Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Signed-off-by: Yury Norov <yury.norov@gmail.com>
bitmap_{from,to}_arr64() optimization is overly optimistic on 32-bit LE
architectures when it's wired to bitmap_copy_clear_tail().
bitmap_copy_clear_tail() takes care of unused bits in the bitmap up to
the next word boundary. But on 32-bit machines when copying bits from
bitmap to array of 64-bit words, it's expected that the unused part of
a recipient array must be cleared up to 64-bit boundary, so the last 4
bytes may stay untouched when nbits % 64 <= 32.
While the copying part of the optimization works correct, that clear-tail
trick makes corresponding tests reasonably fail:
test_bitmap: bitmap_to_arr64(nbits == 1): tail is not safely cleared: 0xa5a5a5a500000001 (must be 0x0000000000000001)
Fix it by removing bitmap_{from,to}_arr64() optimization for 32-bit LE
arches.
Reported-by: Guenter Roeck <linux@roeck-us.net>
Link: https://lore.kernel.org/lkml/20230225184702.GA3587246@roeck-us.net/
Fixes: 0a97953fd2 ("lib: add bitmap_{from,to}_arr64")
Signed-off-by: Yury Norov <yury.norov@gmail.com>
Tested-by: Guenter Roeck <linux@roeck-us.net>
Reviewed-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Reviewed-by: Alexander Lobakin <aleksander.lobakin@intel.com>
The function calculates Hamming weight of (bitmap1 & bitmap2). Now we
have to do like this:
tmp = bitmap_alloc(nbits);
bitmap_and(tmp, map1, map2, nbits);
weight = bitmap_weight(tmp, nbits);
bitmap_free(tmp);
This requires additional memory, adds pressure on alloc subsystem, and
way less cache-friendly than just:
weight = bitmap_weight_and(map1, map2, nbits);
The following patches apply it for cpumask functions.
Signed-off-by: Yury Norov <yury.norov@gmail.com>
Pull bitmap updates from Yury Norov:
- fix the duplicated comments on bitmap_to_arr64() (Qu Wenruo)
- optimize out non-atomic bitops on compile-time constants (Alexander
Lobakin)
- cleanup bitmap-related headers (Yury Norov)
- x86/olpc: fix 'logical not is only applied to the left hand side'
(Alexander Lobakin)
- lib/nodemask: inline wrappers around bitmap (Yury Norov)
* tag 'bitmap-6.0-rc1' of https://github.com/norov/linux: (26 commits)
lib/nodemask: inline next_node_in() and node_random()
powerpc: drop dependency on <asm/machdep.h> in archrandom.h
x86/olpc: fix 'logical not is only applied to the left hand side'
lib/cpumask: move some one-line wrappers to header file
headers/deps: mm: align MANITAINERS and Docs with new gfp.h structure
headers/deps: mm: Split <linux/gfp_types.h> out of <linux/gfp.h>
headers/deps: mm: Optimize <linux/gfp.h> header dependencies
lib/cpumask: move trivial wrappers around find_bit to the header
lib/cpumask: change return types to unsigned where appropriate
cpumask: change return types to bool where appropriate
lib/bitmap: change type of bitmap_weight to unsigned long
lib/bitmap: change return types to bool where appropriate
arm: align find_bit declarations with generic kernel
iommu/vt-d: avoid invalid memory access via node_online(NUMA_NO_NODE)
lib/test_bitmap: test the tail after bitmap_to_arr64()
lib/bitmap: fix off-by-one in bitmap_to_arr64()
lib: test_bitmap: add compile-time optimization/evaluations assertions
bitmap: don't assume compiler evaluates small mem*() builtins calls
net/ice: fix initializing the bitmap in the switch code
bitops: let optimize out non-atomic bitops on compile-time constants
...
bitmap_weight() doesn't return negative values, so change it's type
to unsigned long. It may help compiler to generate better code and
catch bugs.
Signed-off-by: Yury Norov <yury.norov@gmail.com>
Some bitmap functions return boolean results in int variables. Fix it
by changing return types to bool.
Signed-off-by: Yury Norov <yury.norov@gmail.com>
Intel kernel bot triggered the build bug on ARC architecture that
in fact is as follows:
DECLARE_BITMAP(bitmap, BITS_PER_LONG);
bitmap_clear(bitmap, 0, BITS_PER_LONG);
BUILD_BUG_ON(!__builtin_constant_p(*bitmap));
which can be expanded to:
unsigned long bitmap[1];
memset(bitmap, 0, sizeof(*bitmap));
BUILD_BUG_ON(!__builtin_constant_p(*bitmap));
In most cases, a compiler is able to expand small/simple mem*()
calls to simple assignments or bitops, in this case that would mean:
unsigned long bitmap[1] = { 0 };
BUILD_BUG_ON(!__builtin_constant_p(*bitmap));
and on most architectures this works, but not on ARC, despite having
-O3 for every build.
So, to make this work, in case when the last bit to modify is still
within the first long (small_const_nbits()), just use plain
assignments for the rest of bitmap_*() functions which still use
mem*(), but didn't receive such compile-time optimizations yet.
This doesn't have the same coverage as compilers provide, but at
least something to start:
text: add/remove: 3/7 grow/shrink: 43/78 up/down: 1848/-3370 (-1546)
data: add/remove: 1/11 grow/shrink: 0/8 up/down: 4/-356 (-352)
notably cpumask_*() family when NR_CPUS <= BITS_PER_LONG:
netif_get_num_default_rss_queues 38 4 -34
cpumask_copy 90 - -90
cpumask_clear 146 - -146
and the abovementioned assertion started passing.
Signed-off-by: Alexander Lobakin <alexandr.lobakin@intel.com>
Signed-off-by: Yury Norov <yury.norov@gmail.com>
Thanks to the recent commit 0a97953fd2 ("lib: add
bitmap_{from,to}_arr64") now we can directly convert a U64 value into a
bitmap and vice verse.
However when checking the header there is duplicated helper for
bitmap_to_arr64(), but no bitmap_from_arr64().
Just fix the copy-n-paste error.
Signed-off-by: Qu Wenruo <wqu@suse.com>
Signed-off-by: Yury Norov <yury.norov@gmail.com>
Andy Shevchenko
Geert Uytterhoeven
Linus Torvalds
Yury Norov
Al Viro
Alexander Lobakin
Syed Nayyar Waris
Herve Codina
Bartosz Golaszewski
Alexander Lobakin
Qu Wenruo
Kees Cook