The XDP_REDIRECT map devmap can avoid using ndo_xdp_flush, by instead
instructing ndo_xdp_xmit to flush via XDP_XMIT_FLUSH flag in
appropriate places.
Notice after this patch it is possible to remove ndo_xdp_flush
completely, as this is the last user of ndo_xdp_flush. This is left
for later patches, to keep driver changes separate.
Signed-off-by: Jesper Dangaard Brouer <brouer@redhat.com>
Acked-by: Song Liu <songliubraving@fb.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
This patch only change the API and reject any use of flags. This is an
intermediate step that allows us to implement the flush flag operation
later, for each individual driver in a separate patch.
The plan is to implement flush operation via XDP_XMIT_FLUSH flag
and then remove XDP_XMIT_FLAGS_NONE when done.
Signed-off-by: Jesper Dangaard Brouer <brouer@redhat.com>
Acked-by: Song Liu <songliubraving@fb.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
The assignment dev = dev is redundant and should be removed.
Detected by CoverityScan, CID#1469486 ("Evaluation order violation")
Signed-off-by: Colin Ian King <colin.king@canonical.com>
Acked-by: Song Liu <songliubraving@fb.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Extending tracepoint xdp:xdp_devmap_xmit in devmap with an err code
allow people to easier identify the reason behind the ndo_xdp_xmit
call to a given driver is failing.
Signed-off-by: Jesper Dangaard Brouer <brouer@redhat.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
This patch change the API for ndo_xdp_xmit to support bulking
xdp_frames.
When kernel is compiled with CONFIG_RETPOLINE, XDP sees a huge slowdown.
Most of the slowdown is caused by DMA API indirect function calls, but
also the net_device->ndo_xdp_xmit() call.
Benchmarked patch with CONFIG_RETPOLINE, using xdp_redirect_map with
single flow/core test (CPU E5-1650 v4 @ 3.60GHz), showed
performance improved:
for driver ixgbe: 6,042,682 pps -> 6,853,768 pps = +811,086 pps
for driver i40e : 6,187,169 pps -> 6,724,519 pps = +537,350 pps
With frames avail as a bulk inside the driver ndo_xdp_xmit call,
further optimizations are possible, like bulk DMA-mapping for TX.
Testing without CONFIG_RETPOLINE show the same performance for
physical NIC drivers.
The virtual NIC driver tun sees a huge performance boost, as it can
avoid doing per frame producer locking, but instead amortize the
locking cost over the bulk.
V2: Fix compile errors reported by kbuild test robot <lkp@intel.com>
V4: Isolated ndo, driver changes and callers.
Signed-off-by: Jesper Dangaard Brouer <brouer@redhat.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
When sending an xdp_frame through xdp_do_redirect call, then error
cases can happen where the xdp_frame needs to be dropped, and
returning an -errno code isn't sufficient/possible any-longer
(e.g. for cpumap case). This is already fully supported, by simply
calling xdp_return_frame.
This patch is an optimization, which provides xdp_return_frame_rx_napi,
which is a faster variant for these error cases. It take advantage of
the protection provided by XDP RX running under NAPI protection.
This change is mostly relevant for drivers using the page_pool
allocator as it can take advantage of this. (Tested with mlx5).
Signed-off-by: Jesper Dangaard Brouer <brouer@redhat.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Notice how this allow us get XDP statistic without affecting the XDP
performance, as tracepoint is no-longer activated on a per packet basis.
V5: Spotted by John Fastabend.
Fix 'sent' also counted 'drops' in this patch, a later patch corrected
this, but it was a mistake in this intermediate step.
Signed-off-by: Jesper Dangaard Brouer <brouer@redhat.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Like cpumap create queue for xdp frames that will be bulked. For now,
this patch simply invoke ndo_xdp_xmit foreach frame. This happens,
either when the map flush operation is envoked, or when the limit
DEV_MAP_BULK_SIZE is reached.
V5: Avoid memleak on error path in dev_map_update_elem()
Signed-off-by: Jesper Dangaard Brouer <brouer@redhat.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Functionality is the same, but the ndo_xdp_xmit call is now
simply invoked from inside the devmap.c code.
V2: Fix compile issue reported by kbuild test robot <lkp@intel.com>
V5: Cleanups requested by Daniel
- Newlines before func definition
- Use BUILD_BUG_ON checks
- Remove unnecessary use return value store in dev_map_enqueue
Signed-off-by: Jesper Dangaard Brouer <brouer@redhat.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
All map types reimplement the field-by-field copy of union bpf_attr
members into struct bpf_map. Add a helper to perform this operation.
Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Reviewed-by: Quentin Monnet <quentin.monnet@netronome.com>
Acked-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
There were quite a few overlapping sets of changes here.
Daniel's bug fix for off-by-ones in the new BPF branch instructions,
along with the added allowances for "data_end > ptr + x" forms
collided with the metadata additions.
Along with those three changes came veritifer test cases, which in
their final form I tried to group together properly. If I had just
trimmed GIT's conflict tags as-is, this would have split up the
meta tests unnecessarily.
In the socketmap code, a set of preemption disabling changes
overlapped with the rename of bpf_compute_data_end() to
bpf_compute_data_pointers().
Changes were made to the mv88e6060.c driver set addr method
which got removed in net-next.
The hyperv transport socket layer had a locking change in 'net'
which overlapped with a change of socket state macro usage
in 'net-next'.
Signed-off-by: David S. Miller <davem@davemloft.net>
An integer overflow is possible in dev_map_bitmap_size() when
calculating the BITS_TO_LONG logic which becomes, after macro
replacement,
(((n) + (d) - 1)/ (d))
where 'n' is a __u32 and 'd' is (8 * sizeof(long)). To avoid
overflow cast to u64 before arithmetic.
Reported-by: Richard Weinberger <richard@nod.at>
Acked-by: Daniel Borkmann <daniel@iogearbox.net>
Signed-off-by: John Fastabend <john.fastabend@gmail.com>
Acked-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Introduce the map read/write flags to the eBPF syscalls that returns the
map fd. The flags is used to set up the file mode when construct a new
file descriptor for bpf maps. To not break the backward capability, the
f_flags is set to O_RDWR if the flag passed by syscall is 0. Otherwise
it should be O_RDONLY or O_WRONLY. When the userspace want to modify or
read the map content, it will check the file mode to see if it is
allowed to make the change.
Signed-off-by: Chenbo Feng <fengc@google.com>
Acked-by: Alexei Starovoitov <ast@kernel.org>
Acked-by: Daniel Borkmann <daniel@iogearbox.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
Devmap is used with XDP which requires CAP_NET_ADMIN so lets also
make CAP_NET_ADMIN required to use the map.
Signed-off-by: John Fastabend <john.fastabend@gmail.com>
Acked-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
It was reported that syzkaller was able to trigger a splat on
devmap percpu allocation due to illegal/unsupported allocation
request size passed to __alloc_percpu():
[ 70.094249] illegal size (32776) or align (8) for percpu allocation
[ 70.094256] ------------[ cut here ]------------
[ 70.094259] WARNING: CPU: 3 PID: 3451 at mm/percpu.c:1365 pcpu_alloc+0x96/0x630
[...]
[ 70.094325] Call Trace:
[ 70.094328] __alloc_percpu_gfp+0x12/0x20
[ 70.094330] dev_map_alloc+0x134/0x1e0
[ 70.094331] SyS_bpf+0x9bc/0x1610
[ 70.094333] ? selinux_task_setrlimit+0x5a/0x60
[ 70.094334] ? security_task_setrlimit+0x43/0x60
[ 70.094336] entry_SYSCALL_64_fastpath+0x1a/0xa5
This was due to too large max_entries for the map such that we
surpassed the upper limit of PCPU_MIN_UNIT_SIZE. It's fine to
fail naturally here, so switch to __alloc_percpu_gfp() and pass
__GFP_NOWARN instead.
Fixes: 11393cc9b9 ("xdp: Add batching support to redirect map")
Reported-by: Mark Rutland <mark.rutland@arm.com>
Reported-by: Shankara Pailoor <sp3485@columbia.edu>
Reported-by: Richard Weinberger <richard@nod.at>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Cc: John Fastabend <john.fastabend@gmail.com>
Acked-by: Alexei Starovoitov <ast@kernel.org>
Acked-by: John Fastabend <john.fastabend@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
If bpf_map_precharge_memlock in dev_map_alloc, -ENOMEM is returned
regardless of the actual error produced by bpf_map_precharge_memlock.
Fix it by passing on the error returned by bpf_map_precharge_memlock.
Also return -EINVAL instead of -ENOMEM if the page count overflow check
fails.
This makes dev_map_alloc match the behavior of other bpf maps' alloc
functions wrt. return values.
Signed-off-by: Tobias Klauser <tklauser@distanz.ch>
Acked-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Be a bit more friendly about waiting for flush bits to complete.
Replace the cpu_relax() with a cond_resched().
Suggested-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Daniel Borkmann <daniel@iogearbox.net>
Signed-off-by: John Fastabend <john.fastabend@gmail.com>
Acked-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
No need to test for it in fast-path, every dev in bpf_dtab_netdev
is guaranteed to be non-NULL, otherwise dev_map_update_elem() will
fail in the first place.
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Alexei Starovoitov <ast@kernel.org>
Acked-by: John Fastabend <john.fastabend@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Some minor code cleanups, while going over it I also noticed that
we're accounting the bitmap only for one CPU currently, so fix that
up as well.
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
In the current code, dev_map_free() can still race with dev_map_notification().
In dev_map_free(), we remove dtab from the list of dtabs after we purged
all entries from it. However, we don't do xchg() with NULL or the like,
so the entry at that point is still pointing to the device. If a unregister
notification comes in at the same time, we therefore risk a double-free,
since the pointer is still present in the map, and then pushed again to
__dev_map_entry_free().
All this is completely unnecessary. Just remove the dtab from the list
right before the synchronize_rcu(), so all outstanding readers from the
notifier list have finished by then, thus we don't need to deal with this
corner case anymore and also wouldn't need to nullify dev entires. This is
fine because we iterate over the map releasing all entries and therefore
dev references anyway.
Fixes: 4cc7b9544b ("bpf: devmap fix mutex in rcu critical section")
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
The current map creation API does not allow to provide the numa-node
preference. The memory usually comes from where the map-creation-process
is running. The performance is not ideal if the bpf_prog is known to
always run in a numa node different from the map-creation-process.
One of the use case is sharding on CPU to different LRU maps (i.e.
an array of LRU maps). Here is the test result of map_perf_test on
the INNER_LRU_HASH_PREALLOC test if we force the lru map used by
CPU0 to be allocated from a remote numa node:
[ The machine has 20 cores. CPU0-9 at node 0. CPU10-19 at node 1 ]
># taskset -c 10 ./map_perf_test 512 8 1260000 8000000
5:inner_lru_hash_map_perf pre-alloc 1628380 events per sec
4:inner_lru_hash_map_perf pre-alloc 1626396 events per sec
3:inner_lru_hash_map_perf pre-alloc 1626144 events per sec
6:inner_lru_hash_map_perf pre-alloc 1621657 events per sec
2:inner_lru_hash_map_perf pre-alloc 1621534 events per sec
1:inner_lru_hash_map_perf pre-alloc 1620292 events per sec
7:inner_lru_hash_map_perf pre-alloc 1613305 events per sec
0:inner_lru_hash_map_perf pre-alloc 1239150 events per sec #<<<
After specifying numa node:
># taskset -c 10 ./map_perf_test 512 8 1260000 8000000
5:inner_lru_hash_map_perf pre-alloc 1629627 events per sec
3:inner_lru_hash_map_perf pre-alloc 1628057 events per sec
1:inner_lru_hash_map_perf pre-alloc 1623054 events per sec
6:inner_lru_hash_map_perf pre-alloc 1616033 events per sec
2:inner_lru_hash_map_perf pre-alloc 1614630 events per sec
4:inner_lru_hash_map_perf pre-alloc 1612651 events per sec
7:inner_lru_hash_map_perf pre-alloc 1609337 events per sec
0:inner_lru_hash_map_perf pre-alloc 1619340 events per sec #<<<
This patch adds one field, numa_node, to the bpf_attr. Since numa node 0
is a valid node, a new flag BPF_F_NUMA_NODE is also added. The numa_node
field is honored if and only if the BPF_F_NUMA_NODE flag is set.
Numa node selection is not supported for percpu map.
This patch does not change all the kmalloc. F.e.
'htab = kzalloc()' is not changed since the object
is small enough to stay in the cache.
Signed-off-by: Martin KaFai Lau <kafai@fb.com>
Acked-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Alexei Starovoitov <ast@fb.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
In the devmap alloc map logic we check to ensure that the sizeof the
values are not greater than KMALLOC_MAX_SIZE. But, in the dev map case
we ensure the value size is 4bytes earlier in the function because all
values should be netdev ifindex values.
The second check is harmless but is not needed so remove it.
Signed-off-by: John Fastabend <john.fastabend@gmail.com>
Acked-by: Daniel Borkmann <daniel@iogearbox.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
Originally we used a mutex to protect concurrent devmap update
and delete operations from racing with netdev unregister notifier
callbacks.
The notifier hook is needed because we increment the netdev ref
count when a dev is added to the devmap. This ensures the netdev
reference is valid in the datapath. However, we don't want to block
unregister events, hence the initial mutex and notifier handler.
The concern was in the notifier hook we search the map for dev
entries that hold a refcnt on the net device being torn down. But,
in order to do this we require two steps,
(i) dereference the netdev: dev = rcu_dereference(map[i])
(ii) test ifindex: dev->ifindex == removing_ifindex
and then finally we can swap in the NULL dev in the map via an
xchg operation,
xchg(map[i], NULL)
The danger here is a concurrent update could run a different
xchg op concurrently leading us to replace the new dev with a
NULL dev incorrectly.
CPU 1 CPU 2
notifier hook bpf devmap update
dev = rcu_dereference(map[i])
dev = rcu_dereference(map[i])
xchg(map[i]), new_dev);
rcu_call(dev,...)
xchg(map[i], NULL)
The above flow would create the incorrect state with the dev
reference in the update path being lost. To resolve this the
original code used a mutex around the above block. However,
updates, deletes, and lookups occur inside rcu critical sections
so we can't use a mutex in this context safely.
Fortunately, by writing slightly better code we can avoid the
mutex altogether. If CPU 1 in the above example uses a cmpxchg
and _only_ replaces the dev reference in the map when it is in
fact the expected dev the race is removed completely. The two
cases being illustrated here, first the race condition,
CPU 1 CPU 2
notifier hook bpf devmap update
dev = rcu_dereference(map[i])
dev = rcu_dereference(map[i])
xchg(map[i]), new_dev);
rcu_call(dev,...)
odev = cmpxchg(map[i], dev, NULL)
Now we can test the cmpxchg return value, detect odev != dev and
abort. Or in the good case,
CPU 1 CPU 2
notifier hook bpf devmap update
dev = rcu_dereference(map[i])
odev = cmpxchg(map[i], dev, NULL)
[...]
Now 'odev == dev' and we can do proper cleanup.
And viola the original race we tried to solve with a mutex is
corrected and the trace noted by Sasha below is resolved due
to removal of the mutex.
Note: When walking the devmap and removing dev references as needed
we depend on the core to fail any calls to dev_get_by_index() using
the ifindex of the device being removed. This way we do not race with
the user while searching the devmap.
Additionally, the mutex was also protecting list add/del/read on
the list of maps in-use. This patch converts this to an RCU list
and spinlock implementation. This protects the list from concurrent
alloc/free operations. The notifier hook walks this list so it uses
RCU read semantics.
BUG: sleeping function called from invalid context at kernel/locking/mutex.c:747
in_atomic(): 1, irqs_disabled(): 0, pid: 16315, name: syz-executor1
1 lock held by syz-executor1/16315:
#0: (rcu_read_lock){......}, at: [<ffffffff8c363bc2>] map_delete_elem kernel/bpf/syscall.c:577 [inline]
#0: (rcu_read_lock){......}, at: [<ffffffff8c363bc2>] SYSC_bpf kernel/bpf/syscall.c:1427 [inline]
#0: (rcu_read_lock){......}, at: [<ffffffff8c363bc2>] SyS_bpf+0x1d32/0x4ba0 kernel/bpf/syscall.c:1388
Fixes: 2ddf71e23c ("net: add notifier hooks for devmap bpf map")
Reported-by: Sasha Levin <alexander.levin@verizon.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Signed-off-by: John Fastabend <john.fastabend@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
We forgot to set the error code on two error paths which means that we
return ERR_PTR(0) which is NULL. The caller, find_and_alloc_map(), is
not expecting that and will have a NULL dereference.
Fixes: 546ac1ffb7 ("bpf: add devmap, a map for storing net device references")
Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Acked-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: John Fastabend <john.fastabend@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The BPF map devmap holds a refcnt on the net_device structure when
it is in the map. We need to do this to ensure on driver unload we
don't lose a dev reference.
However, its not very convenient to have to manually unload the map
when destroying a net device so add notifier handlers to do the cleanup
automatically. But this creates a race between update/destroy BPF
syscall and programs and the unregister netdev hook.
Unfortunately, the best I could come up with is either to live with
requiring manual removal of net devices from the map before removing
the net device OR to add a mutex in devmap to ensure the map is not
modified while we are removing a device. The fallout also requires
that BPF programs no longer update/delete the map from the BPF program
side because the mutex may sleep and this can not be done from inside
an rcu critical section. This is not a real problem though because I
have not come up with any use cases where this is actually useful in
practice. If/when we come up with a compelling user for this we may
need to revisit this.
Signed-off-by: John Fastabend <john.fastabend@gmail.com>
Acked-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Jesper Dangaard Brouer <brouer@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
For performance reasons we want to avoid updating the tail pointer in
the driver tx ring as much as possible. To accomplish this we add
batching support to the redirect path in XDP.
This adds another ndo op "xdp_flush" that is used to inform the driver
that it should bump the tail pointer on the TX ring.
Signed-off-by: John Fastabend <john.fastabend@gmail.com>
Signed-off-by: Jesper Dangaard Brouer <brouer@redhat.com>
Acked-by: Daniel Borkmann <daniel@iogearbox.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
BPF programs can use the devmap with a bpf_redirect_map() helper
routine to forward packets to netdevice in map.
Signed-off-by: John Fastabend <john.fastabend@gmail.com>
Signed-off-by: Jesper Dangaard Brouer <brouer@redhat.com>
Acked-by: Daniel Borkmann <daniel@iogearbox.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
Device map (devmap) is a BPF map, primarily useful for networking
applications, that uses a key to lookup a reference to a netdevice.
The map provides a clean way for BPF programs to build virtual port
to physical port maps. Additionally, it provides a scoping function
for the redirect action itself allowing multiple optimizations. Future
patches will leverage the map to provide batching at the XDP layer.
Another optimization/feature, that is not yet implemented, would be
to support multiple netdevices per key to support efficient multicast
and broadcast support.
Signed-off-by: John Fastabend <john.fastabend@gmail.com>
Acked-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Jesper Dangaard Brouer <brouer@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Jesper Dangaard Brouer