Similar to the earlier commit that changed sk_setsockopt() to
use sockopt_{lock,release}_sock() such that it can avoid taking
lock when called from bpf. This patch also changes do_ip_getsockopt()
to use sockopt_{lock,release}_sock() such that a latter patch can
make bpf_getsockopt(SOL_IP) to reuse do_ip_getsockopt().
Signed-off-by: Martin KaFai Lau <martin.lau@kernel.org>
Link: https://lore.kernel.org/r/20220902002834.2891514-1-kafai@fb.com
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Similar to the earlier patch that changes sk_getsockopt() to
take the sockptr_t argument. This patch also changes
do_ip_getsockopt() to take the sockptr_t argument such that
a latter patch can make bpf_getsockopt(SOL_IP) to reuse
do_ip_getsockopt().
Note on the change in ip_mc_gsfget(). This function is to
return an array of sockaddr_storage in optval. This function
is shared between ip_get_mcast_msfilter() and
compat_ip_get_mcast_msfilter(). However, the sockaddr_storage
is stored at different offset of the optval because of
the difference between group_filter and compat_group_filter.
Thus, a new 'ss_offset' argument is added to ip_mc_gsfget().
Signed-off-by: Martin KaFai Lau <martin.lau@kernel.org>
Link: https://lore.kernel.org/r/20220902002828.2890585-1-kafai@fb.com
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Similar to the earlier commit that changed sk_setsockopt() to
use sockopt_{lock,release}_sock() such that it can avoid taking
lock when called from bpf. This patch also changes do_tcp_getsockopt()
to use sockopt_{lock,release}_sock() such that a latter patch can
make bpf_getsockopt(SOL_TCP) to reuse do_tcp_getsockopt().
Signed-off-by: Martin KaFai Lau <martin.lau@kernel.org>
Link: https://lore.kernel.org/r/20220902002821.2889765-1-kafai@fb.com
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Similar to the earlier patch that changes sk_getsockopt() to
take the sockptr_t argument . This patch also changes
do_tcp_getsockopt() to take the sockptr_t argument such that
a latter patch can make bpf_getsockopt(SOL_TCP) to reuse
do_tcp_getsockopt().
Signed-off-by: Martin KaFai Lau <martin.lau@kernel.org>
Link: https://lore.kernel.org/r/20220902002815.2889332-1-kafai@fb.com
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Similar to the earlier commit that changed sk_setsockopt() to
use sockopt_{lock,release}_sock() such that it can avoid taking
lock when called from bpf. This patch also changes sk_getsockopt()
to use sockopt_{lock,release}_sock() such that a latter patch can
make bpf_getsockopt(SOL_SOCKET) to reuse sk_getsockopt().
Only sk_get_filter() requires this change and it is used by
the optname SO_GET_FILTER.
The '.getname' implementations in sock->ops->getname() is not
changed also since bpf does not always have the sk->sk_socket
pointer and cannot support SO_PEERNAME.
Signed-off-by: Martin KaFai Lau <martin.lau@kernel.org>
Link: https://lore.kernel.org/r/20220902002809.2888981-1-kafai@fb.com
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
This patch changes sk_getsockopt() to take the sockptr_t argument
such that it can be used by bpf_getsockopt(SOL_SOCKET) in a
latter patch.
security_socket_getpeersec_stream() is not changed. It stays
with the __user ptr (optval.user and optlen.user) to avoid changes
to other security hooks. bpf_getsockopt(SOL_SOCKET) also does not
support SO_PEERSEC.
Signed-off-by: Martin KaFai Lau <martin.lau@kernel.org>
Link: https://lore.kernel.org/r/20220902002802.2888419-1-kafai@fb.com
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
A latter patch refactors bpf_getsockopt(SOL_SOCKET) with the
sock_getsockopt() to avoid code duplication and code
drift between the two duplicates.
The current sock_getsockopt() takes sock ptr as the argument.
The very first thing of this function is to get back the sk ptr
by 'sk = sock->sk'.
bpf_getsockopt() could be called when the sk does not have
the sock ptr created. Meaning sk->sk_socket is NULL. For example,
when a passive tcp connection has just been established but has yet
been accept()-ed. Thus, it cannot use the sock_getsockopt(sk->sk_socket)
or else it will pass a NULL ptr.
This patch moves all sock_getsockopt implementation to the newly
added sk_getsockopt(). The new sk_getsockopt() takes a sk ptr
and immediately gets the sock ptr by 'sock = sk->sk_socket'
The existing sock_getsockopt(sock) is changed to call
sk_getsockopt(sock->sk). All existing callers have both sock->sk
and sk->sk_socket pointer.
The latter patch will make bpf_getsockopt(SOL_SOCKET) call
sk_getsockopt(sk) directly. The bpf_getsockopt(SOL_SOCKET) does
not use the optnames that require sk->sk_socket, so it will
be safe.
Signed-off-by: Martin KaFai Lau <martin.lau@kernel.org>
Link: https://lore.kernel.org/r/20220902002756.2887884-1-kafai@fb.com
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
The put lowers the reference count to 0 and frees ctx, reading it
afterwards is invalid. Move the put after the uses and determine the
last use by the reference count being 1.
Fixes: 39e940d4ab ("selftests/xsk: Destroy BPF resources only when ctx refcount drops to 0")
Signed-off-by: Ian Rogers <irogers@google.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Magnus Karlsson <magnus.karlsson@intel.com>
Link: https://lore.kernel.org/bpf/20220901202645.1463552-1-irogers@google.com
BPF object files are, in a way, the final artifact produced as part of
the ahead-of-time compilation process. That makes them somewhat special
compared to "regular" object files, which are a intermediate build
artifacts that can typically be removed safely. As such, it can make
sense to name them differently to make it easier to spot this difference
at a glance.
Among others, libbpf-bootstrap [0] has established the extension .bpf.o
for BPF object files. It seems reasonable to follow this example and
establish the same denomination for selftest build artifacts. To that
end, this change adjusts the corresponding part of the build system and
the test programs loading BPF object files to work with .bpf.o files.
[0] https://github.com/libbpf/libbpf-bootstrap
Suggested-by: Andrii Nakryiko <andrii@kernel.org>
Signed-off-by: Daniel Müller <deso@posteo.net>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Link: https://lore.kernel.org/bpf/20220901222253.1199242-1-deso@posteo.net
Introduce new mode to xdpxceiver responsible for testing AF_XDP zero
copy support of driver that serves underlying physical device. When
setting up test suite, determine whether driver has ZC support or not by
trying to bind XSK ZC socket to the interface. If it succeeded,
interpret it as ZC support being in place and do softirq and busy poll
tests for zero copy mode.
Note that Rx dropped tests are skipped since ZC path is not touching
rx_dropped stat at all.
Signed-off-by: Maciej Fijalkowski <maciej.fijalkowski@intel.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Magnus Karlsson <magnus.karlsson@intel.com>
Link: https://lore.kernel.org/bpf/20220901114813.16275-7-maciej.fijalkowski@intel.com
For single threaded poll tests call pthread_kill() from main thread so
that we are sure worker thread has finished its job and it is possible
to proceed with next test types from test suite. It was observed that on
some platforms it takes a bit longer for worker thread to exit and next
test case sees device as busy in this case.
Signed-off-by: Maciej Fijalkowski <maciej.fijalkowski@intel.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Magnus Karlsson <magnus.karlsson@intel.com>
Link: https://lore.kernel.org/bpf/20220901114813.16275-6-maciej.fijalkowski@intel.com
Currently, architecture of xdpxceiver is designed strictly for
conducting veth based tests. Veth pair is created together with a
network namespace and one of the veth interfaces is moved to the
mentioned netns. Then, separate threads for Tx and Rx are spawned which
will utilize described setup.
Infrastructure described in the paragraph above can not be used for
testing AF_XDP support on physical devices. That testing will be
conducted on a single network interface and same queue. Xskxceiver
needs to be extended to distinguish between veth tests and physical
interface tests.
Since same iface/queue id pair will be used by both Tx/Rx threads for
physical device testing, Tx thread, which happen to run after the Rx
thread, is going to create XSK socket with shared umem flag. In order to
track this setting throughout the lifetime of spawned threads, introduce
'shared_umem' boolean variable to struct ifobject and set it to true
when xdpxceiver is run against physical device. In such case, UMEM size
needs to be doubled, so half of it will be used by Rx thread and other
half by Tx thread. For two step based test types, value of XSKMAP
element under key 0 has to be updated as there is now another socket for
the second step. Also, to avoid race conditions when destroying XSK
resources, move this activity to the main thread after spawned Rx and Tx
threads have finished its job. This way it is possible to gracefully
remove shared umem without introducing synchronization mechanisms.
To run xsk selftests suite on physical device, append "-i $IFACE" when
invoking test_xsk.sh. For veth based tests, simply skip it. When "-i
$IFACE" is in place, under the hood test_xsk.sh will use $IFACE for both
interfaces supplied to xdpxceiver, which in turn will interpret that
this execution of test suite is for a physical device.
Note that currently this makes it possible only to test SKB and DRV mode
(in case underlying device has native XDP support). ZC testing support
is added in a later patch.
Signed-off-by: Maciej Fijalkowski <maciej.fijalkowski@intel.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Magnus Karlsson <magnus.karlsson@intel.com>
Link: https://lore.kernel.org/bpf/20220901114813.16275-5-maciej.fijalkowski@intel.com
So that "enp240s0f0" or such name can be used against xskxceiver.
While at it, also extend character count for netns name.
Signed-off-by: Maciej Fijalkowski <maciej.fijalkowski@intel.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Magnus Karlsson <magnus.karlsson@intel.com>
Link: https://lore.kernel.org/bpf/20220901114813.16275-4-maciej.fijalkowski@intel.com
In order to prepare xdpxceiver for physical device testing, let us
introduce default Rx pkt stream. Reason for doing it is that physical
device testing will use a UMEM with a doubled size where half of it will
be used by Tx and other half by Rx. This means that pkt addresses will
differ for Tx and Rx streams. Rx thread will initialize the
xsk_umem_info::base_addr that is added here so that pkt_set(), when
working on Rx UMEM will add this offset and second half of UMEM space
will be used. Note that currently base_addr is 0 on both sides. Future
commit will do the mentioned initialization.
Previously, veth based testing worked on separate UMEMs, so single
default stream was fine.
Signed-off-by: Maciej Fijalkowski <maciej.fijalkowski@intel.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Magnus Karlsson <magnus.karlsson@intel.com>
Link: https://lore.kernel.org/bpf/20220901114813.16275-3-maciej.fijalkowski@intel.com
Currently, xdpxceiver assumes that underlying device supports XDP in
native mode - it is fine by now since tests can run only on a veth pair.
Future commit is going to allow running test suite against physical
devices, so let us query the device if it is capable of running XDP
programs in native mode. This way xdpxceiver will not try to run
TEST_MODE_DRV if device being tested is not supporting it.
Signed-off-by: Maciej Fijalkowski <maciej.fijalkowski@intel.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Magnus Karlsson <magnus.karlsson@intel.com>
Link: https://lore.kernel.org/bpf/20220901114813.16275-2-maciej.fijalkowski@intel.com
Get the tunnel flags in {ipv6}vxlan_get_tunnel_src and ensure they are
aligned with tunnel params set at {ipv6}vxlan_set_tunnel_dst.
Signed-off-by: Shmulik Ladkani <shmulik.ladkani@gmail.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Link: https://lore.kernel.org/bpf/20220831144010.174110-2-shmulik.ladkani@gmail.com
Existing 'bpf_skb_get_tunnel_key' extracts various tunnel parameters
(id, ttl, tos, local and remote) but does not expose ip_tunnel_info's
tun_flags to the BPF program.
It makes sense to expose tun_flags to the BPF program.
Assume for example multiple GRE tunnels maintained on a single GRE
interface in collect_md mode. The program expects origins to initiate
over GRE, however different origins use different GRE characteristics
(e.g. some prefer to use GRE checksum, some do not; some pass a GRE key,
some do not, etc..).
A BPF program getting tun_flags can therefore remember the relevant
flags (e.g. TUNNEL_CSUM, TUNNEL_SEQ...) for each initiating remote. In
the reply path, the program can use 'bpf_skb_set_tunnel_key' in order
to correctly reply to the remote, using similar characteristics, based
on the stored tunnel flags.
Introduce BPF_F_TUNINFO_FLAGS flag for bpf_skb_get_tunnel_key. If
specified, 'bpf_tunnel_key->tunnel_flags' is set with the tun_flags.
Decided to use the existing unused 'tunnel_ext' as the storage for the
'tunnel_flags' in order to avoid changing bpf_tunnel_key's layout.
Also, the following has been considered during the design:
1. Convert the "interesting" internal TUNNEL_xxx flags back to BPF_F_yyy
and place into the new 'tunnel_flags' field. This has 2 drawbacks:
- The BPF_F_yyy flags are from *set_tunnel_key* enumeration space,
e.g. BPF_F_ZERO_CSUM_TX. It is awkward that it is "returned" into
tunnel_flags from a *get_tunnel_key* call.
- Not all "interesting" TUNNEL_xxx flags can be mapped to existing
BPF_F_yyy flags, and it doesn't make sense to create new BPF_F_yyy
flags just for purposes of the returned tunnel_flags.
2. Place key.tun_flags into 'tunnel_flags' but mask them, keeping only
"interesting" flags. That's ok, but the drawback is that what's
"interesting" for my usecase might be limiting for other usecases.
Therefore I decided to expose what's in key.tun_flags *as is*, which seems
most flexible. The BPF user can just choose to ignore bits he's not
interested in. The TUNNEL_xxx are also UAPI, so no harm exposing them
back in the get_tunnel_key call.
Signed-off-by: Shmulik Ladkani <shmulik.ladkani@gmail.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Link: https://lore.kernel.org/bpf/20220831144010.174110-1-shmulik.ladkani@gmail.com
When CONFIG_SECURITY_NETWORK is disabled, there will be build warnings
from resolve_btfids:
WARN: resolve_btfids: unresolved symbol bpf_lsm_socket_socketpair
......
WARN: resolve_btfids: unresolved symbol bpf_lsm_inet_conn_established
Fixing it by wrapping these BTF ID definitions by CONFIG_SECURITY_NETWORK.
Fixes: 69fd337a97 ("bpf: per-cgroup lsm flavor")
Fixes: 9113d7e48e ("bpf: expose bpf_{g,s}etsockopt to lsm cgroup")
Signed-off-by: Hou Tao <houtao1@huawei.com>
Link: https://lore.kernel.org/r/20220901065126.3856297-1-houtao@huaweicloud.com
Signed-off-by: Martin KaFai Lau <martin.lau@kernel.org>
The assignment of the else and else if branches is the same, so the else
if here is redundant, so we remove it and add a comment to make the code
here readable.
./kernel/bpf/cgroup_iter.c:81:6-8: WARNING: possible condition with no effect (if == else).
Link: https://bugzilla.openanolis.cn/show_bug.cgi?id=2016
Reported-by: Abaci Robot <abaci@linux.alibaba.com>
Signed-off-by: Jiapeng Chong <jiapeng.chong@linux.alibaba.com>
Link: https://lore.kernel.org/r/20220831021618.86770-1-jiapeng.chong@linux.alibaba.com
Signed-off-by: Martin KaFai Lau <martin.lau@kernel.org>
Hou Tao says:
====================
From: Hou Tao <houtao1@huawei.com>
Hi,
The patchset aims to make the update of per-cpu prog->active and per-cpu
bpf_task_storage_busy being preemption-safe. The problem is on same
architectures (e.g. arm64), __this_cpu_{inc|dec|inc_return} are neither
preemption-safe nor IRQ-safe, so under fully preemptible kernel the
concurrent updates on these per-cpu variables may be interleaved and the
final values of these variables may be not zero.
Patch 1 & 2 use the preemption-safe per-cpu helpers to manipulate
prog->active and bpf_task_storage_busy. Patch 3 & 4 add a test case in
map_tests to show the concurrent updates on the per-cpu
bpf_task_storage_busy by using __this_cpu_{inc|dec} are not atomic.
Comments are always welcome.
Regards,
Tao
Change Log:
v2:
* Patch 1: update commit message to indicate the problem is only
possible for fully preemptible kernel
* Patch 2: a new patch which fixes the problem for prog->active
* Patch 3 & 4: move it to test_maps and make it depend on CONFIG_PREEMPT
v1: https://lore.kernel.org/bpf/20220829142752.330094-1-houtao@huaweicloud.com/
====================
Signed-off-by: Martin KaFai Lau <martin.lau@kernel.org>
Under full preemptible kernel, task local storage lookup operations on
the same CPU may update per-cpu bpf_task_storage_busy concurrently. If
the update of bpf_task_storage_busy is not preemption safe, the final
value of bpf_task_storage_busy may become not-zero forever and
bpf_task_storage_trylock() will always fail. So add a test case to
ensure the update of bpf_task_storage_busy is preemption safe.
Will skip the test case when CONFIG_PREEMPT is disabled, and it can only
reproduce the problem probabilistically. By increasing
TASK_STORAGE_MAP_NR_LOOP and running it under ARM64 VM with 4-cpus, it
takes about four rounds to reproduce:
> test_maps is modified to only run test_task_storage_map_stress_lookup()
$ export TASK_STORAGE_MAP_NR_THREAD=256
$ export TASK_STORAGE_MAP_NR_LOOP=81920
$ export TASK_STORAGE_MAP_PIN_CPU=1
$ time ./test_maps
test_task_storage_map_stress_lookup(135):FAIL:bad bpf_task_storage_busy got -2
real 0m24.743s
user 0m6.772s
sys 0m17.966s
Signed-off-by: Hou Tao <houtao1@huawei.com>
Acked-by: Alexei Starovoitov <ast@kernel.org>
Link: https://lore.kernel.org/r/20220901061938.3789460-5-houtao@huaweicloud.com
Signed-off-by: Martin KaFai Lau <martin.lau@kernel.org>
sys_pidfd_open() is defined twice in both test_bprm_opts.c and
test_local_storage.c, so move it to a common header file. And it will be
used in map_tests as well.
Signed-off-by: Hou Tao <houtao1@huawei.com>
Acked-by: Alexei Starovoitov <ast@kernel.org>
Link: https://lore.kernel.org/r/20220901061938.3789460-4-houtao@huaweicloud.com
Signed-off-by: Martin KaFai Lau <martin.lau@kernel.org>
Both __this_cpu_inc_return() and __this_cpu_dec() are not preemption
safe and now migrate_disable() doesn't disable preemption, so the update
of prog-active is not atomic and in theory under fully preemptible kernel
recurisve prevention may do not work.
Fixing by using the preemption-safe and IRQ-safe variants.
Fixes: ca06f55b90 ("bpf: Add per-program recursion prevention mechanism")
Signed-off-by: Hou Tao <houtao1@huawei.com>
Acked-by: Alexei Starovoitov <ast@kernel.org>
Link: https://lore.kernel.org/r/20220901061938.3789460-3-houtao@huaweicloud.com
Signed-off-by: Martin KaFai Lau <martin.lau@kernel.org>
Now migrate_disable() does not disable preemption and under some
architectures (e.g. arm64) __this_cpu_{inc|dec|inc_return} are neither
preemption-safe nor IRQ-safe, so for fully preemptible kernel concurrent
lookups or updates on the same task local storage and on the same CPU
may make bpf_task_storage_busy be imbalanced, and
bpf_task_storage_trylock() on the specific cpu will always fail.
Fixing it by using this_cpu_{inc|dec|inc_return} when manipulating
bpf_task_storage_busy.
Fixes: bc235cdb42 ("bpf: Prevent deadlock from recursive bpf_task_storage_[get|delete]")
Signed-off-by: Hou Tao <houtao1@huawei.com>
Acked-by: Alexei Starovoitov <ast@kernel.org>
Link: https://lore.kernel.org/r/20220901061938.3789460-2-houtao@huaweicloud.com
Signed-off-by: Martin KaFai Lau <martin.lau@kernel.org>
Hou Tao says:
====================
From: Hou Tao <houtao1@huawei.com>
Hi,
The patchset aims to fix the issues found during investigating the
syzkaller problem reported in [0]. It seems that the concurrent updates
to the same hash-table bucket may fail as shown in patch 1.
Patch 1 uses preempt_disable() to fix the problem for
htab_use_raw_lock() case. For !htab_use_raw_lock() case, the problem is
left to "BPF specific memory allocator" patchset [1] in which
!htab_use_raw_lock() will be removed.
Patch 2 fixes the out-of-bound memory read problem reported in [0]. The
problem has the root cause as patch 1 and it is fixed by handling -EBUSY
from htab_lock_bucket() correctly.
Patch 3 add two cases for hash-table update: one for the reentrancy of
bpf_map_update_elem(), and another one for concurrent updates of the
same hash-table bucket.
Comments are always welcome.
Regards,
Tao
[0]: https://lore.kernel.org/bpf/CACkBjsbuxaR6cv0kXJoVnBfL9ZJXjjoUcMpw_Ogc313jSrg14A@mail.gmail.com/
[1]: https://lore.kernel.org/bpf/20220819214232.18784-1-alexei.starovoitov@gmail.com/
Change Log:
v4:
* rebased on bpf-next
* add htab_update to DENYLIST.s390x
v3: https://lore.kernel.org/bpf/20220829023709.1958204-1-houtao@huaweicloud.com/
* patch 1: update commit message and add Fixes tag
* patch 2: add Fixes tag
* patch 3: elaborate the description of test cases
v2: https://lore.kernel.org/bpf/bd60ef93-1c6a-2db2-557d-b09b92ad22bd@huaweicloud.com/
* Note the fix is for CONFIG_PREEMPT case in commit message and add
Reviewed-by tag for patch 1
* Drop patch "bpf: Allow normally concurrent map updates for !htab_use_raw_lock() case"
v1: https://lore.kernel.org/bpf/20220821033223.2598791-1-houtao@huaweicloud.com/
====================
Signed-off-by: Martin KaFai Lau <martin.lau@kernel.org>
One test demonstrates the reentrancy of hash map update on the same
bucket should fail, and another one shows concureently updates of
the same hash map bucket should succeed and not fail due to
the reentrancy checking for bucket lock.
There is no trampoline support on s390x, so move htab_update to
denylist.
Signed-off-by: Hou Tao <houtao1@huawei.com>
Link: https://lore.kernel.org/r/20220831042629.130006-4-houtao@huaweicloud.com
Signed-off-by: Martin KaFai Lau <martin.lau@kernel.org>
In __htab_map_lookup_and_delete_batch() if htab_lock_bucket() returns
-EBUSY, it will go to next bucket. Going to next bucket may not only
skip the elements in current bucket silently, but also incur
out-of-bound memory access or expose kernel memory to userspace if
current bucket_cnt is greater than bucket_size or zero.
Fixing it by stopping batch operation and returning -EBUSY when
htab_lock_bucket() fails, and the application can retry or skip the busy
batch as needed.
Fixes: 20b6cc34ea ("bpf: Avoid hashtab deadlock with map_locked")
Reported-by: Hao Sun <sunhao.th@gmail.com>
Signed-off-by: Hou Tao <houtao1@huawei.com>
Link: https://lore.kernel.org/r/20220831042629.130006-3-houtao@huaweicloud.com
Signed-off-by: Martin KaFai Lau <martin.lau@kernel.org>
Per-cpu htab->map_locked is used to prohibit the concurrent accesses
from both NMI and non-NMI contexts. But since commit 74d862b682
("sched: Make migrate_disable/enable() independent of RT"),
migrate_disable() is also preemptible under CONFIG_PREEMPT case, so now
map_locked also disallows concurrent updates from normal contexts
(e.g. userspace processes) unexpectedly as shown below:
process A process B
htab_map_update_elem()
htab_lock_bucket()
migrate_disable()
/* return 1 */
__this_cpu_inc_return()
/* preempted by B */
htab_map_update_elem()
/* the same bucket as A */
htab_lock_bucket()
migrate_disable()
/* return 2, so lock fails */
__this_cpu_inc_return()
return -EBUSY
A fix that seems feasible is using in_nmi() in htab_lock_bucket() and
only checking the value of map_locked for nmi context. But it will
re-introduce dead-lock on bucket lock if htab_lock_bucket() is re-entered
through non-tracing program (e.g. fentry program).
One cannot use preempt_disable() to fix this issue as htab_use_raw_lock
being false causes the bucket lock to be a spin lock which can sleep and
does not work with preempt_disable().
Therefore, use migrate_disable() when using the spinlock instead of
preempt_disable() and defer fixing concurrent updates to when the kernel
has its own BPF memory allocator.
Fixes: 74d862b682 ("sched: Make migrate_disable/enable() independent of RT")
Reviewed-by: Hao Luo <haoluo@google.com>
Signed-off-by: Hou Tao <houtao1@huawei.com>
Link: https://lore.kernel.org/r/20220831042629.130006-2-houtao@huaweicloud.com
Signed-off-by: Martin KaFai Lau <martin.lau@kernel.org>
This patch adds a test to ensure bpf_setsockopt(TCP_CONGESTION, "not_exist")
will not trigger the kernel module autoload.
Before the fix:
[ 40.535829] BUG: sleeping function called from invalid context at include/linux/sched/mm.h:274
[...]
[ 40.552134] tcp_ca_find_autoload.constprop.0+0xcb/0x200
[ 40.552689] tcp_set_congestion_control+0x99/0x7b0
[ 40.553203] do_tcp_setsockopt+0x3ed/0x2240
[...]
[ 40.556041] __bpf_setsockopt+0x124/0x640
Signed-off-by: Martin KaFai Lau <martin.lau@linux.dev>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Link: https://lore.kernel.org/bpf/20220830231953.792412-1-martin.lau@linux.dev
When bpf prog changes tcp-cc by calling bpf_setsockopt(TCP_CONGESTION),
it should not try to load module which may be a blocking operation.
This details was correct in the v1 [0] but missed by mistake in the
later revision in commit cb388e7ee3 ("bpf: net: Change do_tcp_setsockopt()
to use the sockopt's lock_sock() and capable()"). This patch fixes it by
checking the has_current_bpf_ctx().
[0] https://lore.kernel.org/bpf/20220727060921.2373314-1-kafai@fb.com/
Fixes: cb388e7ee3 ("bpf: net: Change do_tcp_setsockopt() to use the sockopt's lock_sock() and capable()")
Signed-off-by: Martin KaFai Lau <martin.lau@linux.dev>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Link: https://lore.kernel.org/bpf/20220830231946.791504-1-martin.lau@linux.dev
The bpf_tail_call_static function is currently not defined unless
using clang >= 8.
To support bpf_tail_call_static on GCC we can check if __clang__ is
not defined to enable bpf_tail_call_static.
We need to use GCC assembly syntax when the compiler does not define
__clang__ as LLVM inline assembly is not fully compatible with GCC.
Signed-off-by: James Hilliard <james.hilliard1@gmail.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Link: https://lore.kernel.org/bpf/20220829210546.755377-1-james.hilliard1@gmail.com
Support dumping info of a cgroup_iter link. This includes
showing the cgroup's id and the order for walking the cgroup
hierarchy. Example output is as follows:
> bpftool link show
1: iter prog 2 target_name bpf_map
2: iter prog 3 target_name bpf_prog
3: iter prog 12 target_name cgroup cgroup_id 72 order self_only
> bpftool -p link show
[{
"id": 1,
"type": "iter",
"prog_id": 2,
"target_name": "bpf_map"
},{
"id": 2,
"type": "iter",
"prog_id": 3,
"target_name": "bpf_prog"
},{
"id": 3,
"type": "iter",
"prog_id": 12,
"target_name": "cgroup",
"cgroup_id": 72,
"order": "self_only"
}
]
Signed-off-by: Hao Luo <haoluo@google.com>
Reviewed-by: Quentin Monnet <quentin@isovalent.com>
Acked-by: Yonghong Song <yhs@fb.com>
Link: https://lore.kernel.org/r/20220829231828.1016835-1-haoluo@google.com
Signed-off-by: Martin KaFai Lau <martin.lau@linux.dev>
There is a potential for us to hit a type conflict when including
netinet/tcp.h and sys/socket.h, we can replace both of these includes
with linux/tcp.h and bpf_tcp_helpers.h to avoid this conflict.
Fixes errors like the below when compiling with gcc BPF backend:
In file included from /usr/include/netinet/tcp.h:91,
from progs/connect4_prog.c:11:
/home/buildroot/opt/cross/lib/gcc/bpf/13.0.0/include/stdint.h:34:23: error: conflicting types for 'int8_t'; have 'char'
34 | typedef __INT8_TYPE__ int8_t;
| ^~~~~~
In file included from /usr/include/x86_64-linux-gnu/sys/types.h:155,
from /usr/include/x86_64-linux-gnu/bits/socket.h:29,
from /usr/include/x86_64-linux-gnu/sys/socket.h:33,
from progs/connect4_prog.c:10:
/usr/include/x86_64-linux-gnu/bits/stdint-intn.h:24:18: note: previous declaration of 'int8_t' with type 'int8_t' {aka 'signed char'}
24 | typedef __int8_t int8_t;
| ^~~~~~
/home/buildroot/opt/cross/lib/gcc/bpf/13.0.0/include/stdint.h:43:24: error: conflicting types for 'int64_t'; have 'long int'
43 | typedef __INT64_TYPE__ int64_t;
| ^~~~~~~
/usr/include/x86_64-linux-gnu/bits/stdint-intn.h:27:19: note: previous declaration of 'int64_t' with type 'int64_t' {aka 'long long int'}
27 | typedef __int64_t int64_t;
| ^~~~~~~
Signed-off-by: James Hilliard <james.hilliard1@gmail.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Link: https://lore.kernel.org/bpf/20220829154710.3870139-1-james.hilliard1@gmail.com
There is a potential for us to hit a type conflict when including
netinet/tcp.h with sys/socket.h, we can remove these as they are not
actually needed.
Fixes errors like the below when compiling with gcc BPF backend:
In file included from /usr/include/netinet/tcp.h:91,
from progs/bind4_prog.c:10:
/home/buildroot/opt/cross/lib/gcc/bpf/13.0.0/include/stdint.h:34:23: error: conflicting types for 'int8_t'; have 'char'
34 | typedef __INT8_TYPE__ int8_t;
| ^~~~~~
In file included from /usr/include/x86_64-linux-gnu/sys/types.h:155,
from /usr/include/x86_64-linux-gnu/bits/socket.h:29,
from /usr/include/x86_64-linux-gnu/sys/socket.h:33,
from progs/bind4_prog.c:9:
/usr/include/x86_64-linux-gnu/bits/stdint-intn.h:24:18: note: previous declaration of 'int8_t' with type 'int8_t' {aka 'signed char'}
24 | typedef __int8_t int8_t;
| ^~~~~~
/home/buildroot/opt/cross/lib/gcc/bpf/13.0.0/include/stdint.h:43:24: error: conflicting types for 'int64_t'; have 'long int'
43 | typedef __INT64_TYPE__ int64_t;
| ^~~~~~~
/usr/include/x86_64-linux-gnu/bits/stdint-intn.h:27:19: note: previous declaration of 'int64_t' with type 'int64_t' {aka 'long long int'}
27 | typedef __int64_t int64_t;
| ^~~~~~~
make: *** [Makefile:537: /home/buildroot/bpf-next/tools/testing/selftests/bpf/bpf_gcc/bind4_prog.o] Error 1
Signed-off-by: James Hilliard <james.hilliard1@gmail.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Link: https://lore.kernel.org/bpf/20220826052925.980431-1-james.hilliard1@gmail.com
MAX_TAIL_CALL_CNT is 33, so min(MAX_TAIL_CALL_CNT, 0xffff) is always
MAX_TAIL_CALL_CNT, it is better to use MAX_TAIL_CALL_CNT directly.
At the same time, add BUILD_BUG_ON(MAX_TAIL_CALL_CNT > 0xffff) with a
comment on why the assertion is there.
Suggested-by: Daniel Borkmann <daniel@iogearbox.net>
Suggested-by: Johan Almbladh <johan.almbladh@anyfinetworks.com>
Signed-off-by: Tiezhu Yang <yangtiezhu@loongson.cn>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Link: https://lore.kernel.org/bpf/1661742309-2320-1-git-send-email-yangtiezhu@loongson.cn
Due to bpf_map_lookup_elem being declared static we need to also
declare subprog_noise as static.
Fixes the following error:
progs/tailcall_bpf2bpf4.c:26:9: error: 'bpf_map_lookup_elem' is static but used in inline function 'subprog_noise' which is not static [-Werror]
26 | bpf_map_lookup_elem(&nop_table, &key);
| ^~~~~~~~~~~~~~~~~~~
Signed-off-by: James Hilliard <james.hilliard1@gmail.com>
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Acked-by: Song Liu <song@kernel.org>
Link: https://lore.kernel.org/bpf/20220826035141.737919-1-james.hilliard1@gmail.com
The sys/socket.h header isn't required to build test_tc_dtime and may
cause a type conflict.
Fixes the following error:
In file included from /usr/include/x86_64-linux-gnu/sys/types.h:155,
from /usr/include/x86_64-linux-gnu/bits/socket.h:29,
from /usr/include/x86_64-linux-gnu/sys/socket.h:33,
from progs/test_tc_dtime.c:18:
/usr/include/x86_64-linux-gnu/bits/stdint-intn.h:24:18: error: conflicting types for 'int8_t'; have '__int8_t' {aka 'signed char'}
24 | typedef __int8_t int8_t;
| ^~~~~~
In file included from progs/test_tc_dtime.c:5:
/home/buildroot/opt/cross/lib/gcc/bpf/13.0.0/include/stdint.h:34:23: note: previous declaration of 'int8_t' with type 'int8_t' {aka 'char'}
34 | typedef __INT8_TYPE__ int8_t;
| ^~~~~~
/usr/include/x86_64-linux-gnu/bits/stdint-intn.h:27:19: error: conflicting types for 'int64_t'; have '__int64_t' {aka 'long long int'}
27 | typedef __int64_t int64_t;
| ^~~~~~~
/home/buildroot/opt/cross/lib/gcc/bpf/13.0.0/include/stdint.h:43:24: note: previous declaration of 'int64_t' with type 'int64_t' {aka 'long int'}
43 | typedef __INT64_TYPE__ int64_t;
| ^~~~~~~
make: *** [Makefile:537: /home/buildroot/bpf-next/tools/testing/selftests/bpf/bpf_gcc/test_tc_dtime.o] Error 1
Signed-off-by: James Hilliard <james.hilliard1@gmail.com>
Link: https://lore.kernel.org/r/20220826050703.869571-1-james.hilliard1@gmail.com
Signed-off-by: Martin KaFai Lau <kafai@fb.com>
This allows to have a better control over maps from the kernel when
preloading eBPF programs.
Acked-by: Yonghong Song <yhs@fb.com>
Signed-off-by: Benjamin Tissoires <benjamin.tissoires@redhat.com>
Link: https://lore.kernel.org/r/20220824134055.1328882-8-benjamin.tissoires@redhat.com
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Add BPF_MAP_GET_FD_BY_ID and BPF_MAP_DELETE_PROG.
Only BPF_MAP_GET_FD_BY_ID needs to be amended to be able
to access the bpf pointer either from the userspace or the kernel.
Acked-by: Yonghong Song <yhs@fb.com>
Signed-off-by: Benjamin Tissoires <benjamin.tissoires@redhat.com>
Link: https://lore.kernel.org/r/20220824134055.1328882-7-benjamin.tissoires@redhat.com
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
bpf_cgroup_iter_order is globally visible but the entries do not have
CGROUP prefix. As requested by Andrii, put a CGROUP in the names
in bpf_cgroup_iter_order.
This patch fixes two previous commits: one introduced the API and
the other uses the API in bpf selftest (that is, the selftest
cgroup_hierarchical_stats).
I tested this patch via the following command:
test_progs -t cgroup,iter,btf_dump
Fixes: d4ccaf58a8 ("bpf: Introduce cgroup iter")
Fixes: 88886309d2 ("selftests/bpf: add a selftest for cgroup hierarchical stats collection")
Suggested-by: Andrii Nakryiko <andrii@kernel.org>
Acked-by: Andrii Nakryiko <andrii@kernel.org>
Signed-off-by: Hao Luo <haoluo@google.com>
Link: https://lore.kernel.org/r/20220825223936.1865810-1-haoluo@google.com
Signed-off-by: Martin KaFai Lau <kafai@fb.com>
The helper value is ABI as defined by enum bpf_func_id.
As bpf_helper_defs.h is used for the userpace part, it must be consistent
with this enum.
Before this change the comments order was used by the bpf_doc script in
order to set the helper values defined in the helpers file.
When adding new helpers it is very puzzling when the userspace application
breaks in weird places if the comment is inserted instead of appended -
because the generated helper ABI is incorrect and shifted.
This commit sets the helper value to the enum value.
In addition it is currently the practice to have the comments appended
and kept in the same order as the enum. As such, add an assertion
validating the comment order is consistent with enum value.
In case a different comments ordering is desired, this assertion can
be lifted.
Signed-off-by: Eyal Birger <eyal.birger@gmail.com>
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Reviewed-by: Quentin Monnet <quentin@isovalent.com>
Link: https://lore.kernel.org/bpf/20220824181043.1601429-1-eyal.birger@gmail.com
When `data` points to a boolean value, casting it to `int *` is problematic
and could lead to a wrong value being passed to `jsonw_bool`. Change the
cast to `bool *` instead.
Fixes: b12d6ec097 ("bpf: btf: add btf print functionality")
Signed-off-by: Lam Thai <lamthai@arista.com>
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Reviewed-by: Quentin Monnet <quentin@isovalent.com>
Acked-by: John Fastabend <john.fastabend@gmail.com>
Link: https://lore.kernel.org/bpf/20220824225859.9038-1-lamthai@arista.com
Hao Luo says:
====================
This patch series allows for using bpf to collect hierarchical cgroup
stats efficiently by integrating with the rstat framework. The rstat
framework provides an efficient way to collect cgroup stats percpu and
propagate them through the cgroup hierarchy.
The stats are exposed to userspace in textual form by reading files in
bpffs, similar to cgroupfs stats by using a cgroup_iter program.
cgroup_iter is a type of bpf_iter. It walks over cgroups in four modes:
- walking a cgroup's descendants in pre-order.
- walking a cgroup's descendants in post-order.
- walking a cgroup's ancestors.
- process only a single object.
When attaching cgroup_iter, one needs to set a cgroup to the iter_link
created from attaching. This cgroup can be passed either as a file
descriptor or a cgroup id. That cgroup serves as the starting point of
the walk.
One can also terminate the walk early by returning 1 from the iter
program.
Note that because walking cgroup hierarchy holds cgroup_mutex, the iter
program is called with cgroup_mutex held.
** Background on rstat for stats collection **
(I am using a subscriber analogy that is not commonly used)
The rstat framework maintains a tree of cgroups that have updates and
which cpus have updates. A subscriber to the rstat framework maintains
their own stats. The framework is used to tell the subscriber when
and what to flush, for the most efficient stats propagation. The
workflow is as follows:
- When a subscriber updates a cgroup on a cpu, it informs the rstat
framework by calling cgroup_rstat_updated(cgrp, cpu).
- When a subscriber wants to read some stats for a cgroup, it asks
the rstat framework to initiate a stats flush (propagation) by calling
cgroup_rstat_flush(cgrp).
- When the rstat framework initiates a flush, it makes callbacks to
subscribers to aggregate stats on cpus that have updates, and
propagate updates to their parent.
Currently, the main subscribers to the rstat framework are cgroup
subsystems (e.g. memory, block). This patch series allow bpf programs to
become subscribers as well.
Patches in this series are organized as follows:
* Patches 1-2 introduce cgroup_iter prog, and a selftest.
* Patches 3-5 allow bpf programs to integrate with rstat by adding the
necessary hook points and kfunc. A comprehensive selftest that
demonstrates the entire workflow for using bpf and rstat to
efficiently collect and output cgroup stats is added.
---
Changelog:
v8 -> v9:
- Make UNSPEC (an invalid option) as the default order for cgroup_iter.
- Use enum for specifying cgroup_iter order, instead of u32.
- Add BPF_ITER_RESHCED to cgroup_iter.
- Add cgroup_hierarchical_stats to s390x denylist.
v7 -> v8:
- Removed the confusing BPF_ITER_DEFAULT (Andrii)
- s/SELF/SELF_ONLY/g
- Fixed typo (e.g. outputing) (Andrii)
- Use "descendants_pre", "descendants_post" etc. instead of "pre",
"post" (Andrii)
v6 -> v7:
- Updated commit/comments in cgroup_iter for read() behavior (Yonghong)
- Extracted BPF_ITER_SELF and other options out of cgroup_iter, so
that they can be used in other iters. Also renamed them. (Andrii)
- Supports both cgroup_fd and cgroup_id when specifying target cgroup.
(Andrii)
- Avoided using macro for formatting expected output in cgroup_iter
selftest. (Andrii)
- Applied 'static' on all vars and functions in cgroup_iter selftest.
(Andrii)
- Fixed broken buf reading in cgroup_iter selftest. (Andrii)
- Switched to use bpf_link__destroy() unconditionally. (Andrii)
- Removed 'volatile' for non-const global vars in selftests. (Andrii)
- Started using bpf_core_enum_value() to get memory_cgrp_id. (Andrii)
v5 -> v6:
- Rebased on bpf-next
- Tidy up cgroup_hierarchical_stats test (Andrii)
* 'static' and 'inline'
* avoid using libbpf_get_error()
* string literals of cgroup paths.
- Rename patch 8/8 to 'selftests/bpf' (Yonghong)
- Fix cgroup_iter comments (e.g. PAGE_SIZE and uapi) (Yonghong)
- Make sure further read() returns OK after previous read() finished
properly (Yonghong)
- Release cgroup_mutex before the last call of show() (Kumar)
v4 -> v5:
- Rebased on top of new kfunc flags infrastructure, updated patch 1 and
patch 6 accordingly.
- Added docs for sleepable kfuncs.
v3 -> v4:
- cgroup_iter:
* reorder fields in bpf_link_info to avoid break uapi (Yonghong)
* comment the behavior when cgroup_fd=0 (Yonghong)
* comment on the limit of number of cgroups supported by cgroup_iter.
(Yonghong)
- cgroup_hierarchical_stats selftest:
* Do not return -1 if stats are not found (causes overflow in userspace).
* Check if child process failed to join cgroup.
* Make buf and path arrays in get_cgroup_vmscan_delay() static.
* Increase the test map sizes to accomodate cgroups that are not
created by the test.
v2 -> v3:
- cgroup_iter:
* Added conditional compilation of cgroup_iter.c in kernel/bpf/Makefile
(kernel test) and dropped the !CONFIG_CGROUP patch.
* Added validation of traversal_order when attaching (Yonghong).
* Fixed previous wording "two modes" to "three modes" (Yonghong).
* Fixed the btf_dump selftest broken by this patch (Yonghong).
* Fixed ctx_arg_info[0] to use "PTR_TO_BTF_ID_OR_NULL" instead of
"PTR_TO_BTF_ID", because the "cgroup" pointer passed to iter prog can
be null.
- Use __diag_push to eliminate __weak noinline warning in
bpf_rstat_flush().
- cgroup_hierarchical_stats selftest:
* Added write_cgroup_file_parent() helper.
* Added error handling for failed map updates.
* Added null check for cgroup in vmscan_flush.
* Fixed the signature of vmscan_[start/end].
* Correctly return error code when attaching trace programs fail.
* Make sure all links are destroyed correctly and not leaking in
cgroup_hierarchical_stats selftest.
* Use memory.reclaim instead of memory.high as a more reliable way to
invoke reclaim.
* Eliminated sleeps, the test now runs faster.
v1 -> v2:
- Redesign of cgroup_iter from v1, based on Alexei's idea [1]:
* supports walking cgroup subtree.
* supports walking ancestors of a cgroup. (Andrii)
* supports terminating the walk early.
* uses fd instead of cgroup_id as parameter for iter_link. Using fd is
a convention in bpf.
* gets cgroup's ref at attach time and deref at detach.
* brought back cgroup1 support for cgroup_iter.
- Squashed the patches adding the rstat flush hook points and kfuncs
(Tejun).
- Added a comment explaining why bpf_rstat_flush() needs to be weak
(Tejun).
- Updated the final selftest with the new cgroup_iter design.
- Changed CHECKs in the selftest with ASSERTs (Yonghong, Andrii).
- Removed empty line at the end of the selftest (Yonghong).
- Renamed test files to cgroup_hierarchical_stats.c.
- Reordered CGROUP_PATH params order to match struct declaration
in the selftest (Michal).
- Removed memory_subsys_enabled() and made sure memcg controller
enablement checks make sense and are documented (Michal).
RFC v2 -> v1:
- Instead of introducing a new program type for rstat flushing, add an
empty hook point, bpf_rstat_flush(), and use fentry bpf programs to
attach to it and flush bpf stats.
- Instead of using helpers, use kfuncs for rstat functions.
- These changes simplify the patchset greatly, with minimal changes to
uapi.
RFC v1 -> RFC v2:
- Instead of rstat flush programs attach to subsystems, they now attach
to rstat (global flushers, not per-subsystem), based on discussions
with Tejun. The first patch is entirely rewritten.
- Pass cgroup pointers to rstat flushers instead of cgroup ids. This is
much more flexibility and less likely to need a uapi update later.
- rstat helpers are now only defined if CGROUP_CONFIG.
- Most of the code is now only defined if CGROUP_CONFIG and
CONFIG_BPF_SYSCALL.
- Move rstat helper protos from bpf_base_func_proto() to
tracing_prog_func_proto().
- rstat helpers argument (cgroup pointer) is now ARG_PTR_TO_BTF_ID, not
ARG_ANYTHING.
- Rewrote the selftest to use the cgroup helpers.
- Dropped bpf_map_lookup_percpu_elem (already added by Feng).
- Dropped patch to support cgroup v1 for cgroup_iter.
- Dropped patch to define some cgroup_put() when !CONFIG_CGROUP. The
code that calls it is no longer compiled when !CONFIG_CGROUP.
cgroup_iter was originally introduced in a different patch series[2].
Hao and I agreed that it fits better as part of this series.
RFC v1 of this patch series had the following changes from [2]:
- Getting the cgroup's reference at the time at attaching, instead of
at the time when iterating. (Yonghong)
- Remove .init_seq_private and .fini_seq_private callbacks for
cgroup_iter. They are not needed now. (Yonghong)
[1] https://lore.kernel.org/bpf/20220520221919.jnqgv52k4ajlgzcl@MBP-98dd607d3435.dhcp.thefacebook.com/
[2] https://lore.kernel.org/lkml/20220225234339.2386398-9-haoluo@google.com/
Hao Luo (2):
bpf: Introduce cgroup iter
selftests/bpf: Test cgroup_iter.
====================
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Add a selftest that tests the whole workflow for collecting,
aggregating (flushing), and displaying cgroup hierarchical stats.
TL;DR:
- Userspace program creates a cgroup hierarchy and induces memcg reclaim
in parts of it.
- Whenever reclaim happens, vmscan_start and vmscan_end update
per-cgroup percpu readings, and tell rstat which (cgroup, cpu) pairs
have updates.
- When userspace tries to read the stats, vmscan_dump calls rstat to flush
the stats, and outputs the stats in text format to userspace (similar
to cgroupfs stats).
- rstat calls vmscan_flush once for every (cgroup, cpu) pair that has
updates, vmscan_flush aggregates cpu readings and propagates updates
to parents.
- Userspace program makes sure the stats are aggregated and read
correctly.
Detailed explanation:
- The test loads tracing bpf programs, vmscan_start and vmscan_end, to
measure the latency of cgroup reclaim. Per-cgroup readings are stored in
percpu maps for efficiency. When a cgroup reading is updated on a cpu,
cgroup_rstat_updated(cgroup, cpu) is called to add the cgroup to the
rstat updated tree on that cpu.
- A cgroup_iter program, vmscan_dump, is loaded and pinned to a file, for
each cgroup. Reading this file invokes the program, which calls
cgroup_rstat_flush(cgroup) to ask rstat to propagate the updates for all
cpus and cgroups that have updates in this cgroup's subtree. Afterwards,
the stats are exposed to the user. vmscan_dump returns 1 to terminate
iteration early, so that we only expose stats for one cgroup per read.
- An ftrace program, vmscan_flush, is also loaded and attached to
bpf_rstat_flush. When rstat flushing is ongoing, vmscan_flush is invoked
once for each (cgroup, cpu) pair that has updates. cgroups are popped
from the rstat tree in a bottom-up fashion, so calls will always be
made for cgroups that have updates before their parents. The program
aggregates percpu readings to a total per-cgroup reading, and also
propagates them to the parent cgroup. After rstat flushing is over, all
cgroups will have correct updated hierarchical readings (including all
cpus and all their descendants).
- Finally, the test creates a cgroup hierarchy and induces memcg reclaim
in parts of it, and makes sure that the stats collection, aggregation,
and reading workflow works as expected.
Signed-off-by: Yosry Ahmed <yosryahmed@google.com>
Signed-off-by: Hao Luo <haoluo@google.com>
Link: https://lore.kernel.org/r/20220824233117.1312810-6-haoluo@google.com
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
This patch extends bpf selft cgroup_helpers [ID] n various ways:
- Add enable_controllers() that allows tests to enable all or a
subset of controllers for a specific cgroup.
- Add join_cgroup_parent(). The cgroup workdir is based on the pid,
therefore a spawned child cannot join the same cgroup hierarchy of the
test through join_cgroup(). join_cgroup_parent() is used in child
processes to join a cgroup under the parent's workdir.
- Add write_cgroup_file() and write_cgroup_file_parent() (similar to
join_cgroup_parent() above).
- Add get_root_cgroup() for tests that need to do checks on root cgroup.
- Distinguish relative and absolute cgroup paths in function arguments.
Now relative paths are called relative_path, and absolute paths are
called cgroup_path.
Signed-off-by: Yosry Ahmed <yosryahmed@google.com>
Signed-off-by: Hao Luo <haoluo@google.com>
Link: https://lore.kernel.org/r/20220824233117.1312810-5-haoluo@google.com
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Enable bpf programs to make use of rstat to collect cgroup hierarchical
stats efficiently:
- Add cgroup_rstat_updated() kfunc, for bpf progs that collect stats.
- Add cgroup_rstat_flush() sleepable kfunc, for bpf progs that read stats.
- Add an empty bpf_rstat_flush() hook that is called during rstat
flushing, for bpf progs that flush stats to attach to. Attaching a bpf
prog to this hook effectively registers it as a flush callback.
Signed-off-by: Yosry Ahmed <yosryahmed@google.com>
Acked-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Hao Luo <haoluo@google.com>
Link: https://lore.kernel.org/r/20220824233117.1312810-4-haoluo@google.com
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Add a selftest for cgroup_iter. The selftest creates a mini cgroup tree
of the following structure:
ROOT (working cgroup)
|
PARENT
/ \
CHILD1 CHILD2
and tests the following scenarios:
- invalid cgroup fd.
- pre-order walk over descendants from PARENT.
- post-order walk over descendants from PARENT.
- walk of ancestors from PARENT.
- process only a single object (i.e. PARENT).
- early termination.
Acked-by: Yonghong Song <yhs@fb.com>
Acked-by: Andrii Nakryiko <andrii@kernel.org>
Signed-off-by: Hao Luo <haoluo@google.com>
Link: https://lore.kernel.org/r/20220824233117.1312810-3-haoluo@google.com
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Cgroup_iter is a type of bpf_iter. It walks over cgroups in four modes:
- walking a cgroup's descendants in pre-order.
- walking a cgroup's descendants in post-order.
- walking a cgroup's ancestors.
- process only the given cgroup.
When attaching cgroup_iter, one can set a cgroup to the iter_link
created from attaching. This cgroup is passed as a file descriptor
or cgroup id and serves as the starting point of the walk. If no
cgroup is specified, the starting point will be the root cgroup v2.
For walking descendants, one can specify the order: either pre-order or
post-order. For walking ancestors, the walk starts at the specified
cgroup and ends at the root.
One can also terminate the walk early by returning 1 from the iter
program.
Note that because walking cgroup hierarchy holds cgroup_mutex, the iter
program is called with cgroup_mutex held.
Currently only one session is supported, which means, depending on the
volume of data bpf program intends to send to user space, the number
of cgroups that can be walked is limited. For example, given the current
buffer size is 8 * PAGE_SIZE, if the program sends 64B data for each
cgroup, assuming PAGE_SIZE is 4kb, the total number of cgroups that can
be walked is 512. This is a limitation of cgroup_iter. If the output
data is larger than the kernel buffer size, after all data in the
kernel buffer is consumed by user space, the subsequent read() syscall
will signal EOPNOTSUPP. In order to work around, the user may have to
update their program to reduce the volume of data sent to output. For
example, skip some uninteresting cgroups. In future, we may extend
bpf_iter flags to allow customizing buffer size.
Acked-by: Yonghong Song <yhs@fb.com>
Acked-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Hao Luo <haoluo@google.com>
Link: https://lore.kernel.org/r/20220824233117.1312810-2-haoluo@google.com
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
